4 ~~~~~~AND fiuF 1 28989 S ' ,, .e t s . -pl 2004 ,, li1t 1 5 t T %. . ii~~~~~~~~~~~~~~~~~~~~~~A -~I 4t_'t ''.'' 4 V~~~~~~~4 II Vl~~.8 -S. r 3 IC~~~~~~~~~~~~~~~~~~~~~~~~~~~~W zj~~~~~~~~~~~~~~K ~~~~~~~~~~~~~~4 r~ ~ ~~~~~~~~~dh -Jr~~~~~~~~~~~~~~~~~~ 44kb ~ ~ ~ ~ ~ ~ 7 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES INEQUALITY IN LATIN AMERICA Breaking with History? David de Ferranti Guillermo E. Perry Francisco H. G. Ferreira Michael Walton THE WORLD BANK Washington, D.C. C) 2004 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www.worldbank.org E-mail feedback@worldbank.org All rights reserved. 0405 0607 4 3 2 1 The findings, interpretations, and conclusions expressed herein are those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Library of Congress Cataloging-in-Publication Data Inequality in Latin America: breaking with history? / David de Ferranti ... [et al.l. p. cm.-(World Bank Latin American and Caribbean studies. Viewpoints) Includes bibliographical references and index. ISBN 0-8213-5665-8 1. Equality-Latin America. 2. Equality-Caribbean Area. 3. Latin America-Social conditions-1982- 4. Caribbean Area-Social conditions-1945- 5. Latin America-Economic conditions-1982- 6. Caribbean Area-Economic conditions-1945- 1. De Ferranti, David M. II. Series. HN110.5.Z9S646 2004 305'.09729-dc22 2003063203 Contents Acknowledgments ................................................................ xi Summary ................................................................ 1 Chapter 1: Introduction: Motivation and Conceptual Framework ........ ............................... 17 1.1 A conceptual framework . ................................................................ 18 1.2 The consequences of high inequality . ........................................................ 24 Part l: The Nature of Inequality in Latin America ............ ..................................... 33 Chapter 2: Different Lives: Inequality in Latin America .................... ......................... 35 2.1 Some conceptual issues .35 2.2 Income inequality and beyond .37 2.3 Measurement-related issues and data limitations .48 2.4 Inequality in Latin America in perspective .53 2.5 Looking inside household income .57 2.6 Inequality beyond income .66 2.7 Conclusions 71 Chapter 3: Group-Based Inequalities: The Roles of Race, Ethnicity, and Gender .......... .. ................ 77 3.1 Who are the people of Latin America? . ...................................................... 78 3.2 Inequality among individuals during the lifecycle ............................................... 84 3.3 Would income inequality decline if returns to human capital were more equal? ......................... 96 3.4 Conclusions . ................................................................... 104 Part II: The Determinants of Inequaity in Latin America .......... ................................. 107 Chapter 4: Historical Roots of Inequality in Latin America .......................................... 109 4.1 Factor endowments, inequality, and institutions .109 4.2 The persistence of inequality: The colonial period .112 4.3 The persistence of inequality: Post-independence .112 4.4 The 20th century .120 4.5 The 21St century and beyond ..122 Chapter 5: State-Society Interactions as Sources of Persistence and Change in Inequality ........ ... ........ 123 5.1 Political and social structures as forces for the reproduction of inequality ............................. 125 5.2 The potential for equalizing political and social change .......................................... 134 5.3 Conclusions ................................................................. 145 Chapter 6: Economic Mechanisms for the Persistence of High Inequality in Latin America ........ ... ........ 149 6.1 Asset distributions: Education and land .151 6.2 Job match quality .157 6.3 Remuneration in the labor markets .161 6.4 Household formation .165 6.5 Conclusions ..173 CONTENTS Part III: Policies for Lower Inequality .................................................. .. 175 Chapter 7: Policies on Assets and Services .................................................... 177 7.1 Education .................................................... 177 7.2 Property rights, land, and housing ......................................................... 189 7.3 Infrastructure services and the distributional impact of privatization ................................ 204 7.4 Conclusions .......................................................... 215 Chapter 8: Policies on Markets and Institutions ................................................. 217 8.1 Markets and inequality .............. ...................................... 217 8.2 Labor market policies and inequality ....................................................... 224 8.3 Inequality and macroeconomic crises ....................................................... 227 8.4 C onclusions .......................................................... 243 Chapter 9: Taxation, Public Expenditures, and Transfers ........................................... 247 9.1 Taxes and distribution . .......................................................... 248 9.2 Public social spending and distribution .................................................... 257 9.3 Cash transfers and distribution . .................................................... 268 9.4 Conclusions .......................................................... 280 Statistical Appendix .................................................... 285 Bibliography .................................................... 351 Boxes Chapter 1 1.1 Measuring inequality of opportunities in Brazil ................................................. 20 Chapter 2 2.1 Social class .46 2.2 Mobility in Latin America: What little is known? .51 2.3 Some simple decompositions .63 2.4 Social capital and trust .66 2.5 Disability and distribution .70 Chapter 3 3.1 Distribution of the population of the Americas: An historical evolution .79 3.2 Todos Contamos: National census and social inclusion .80 3.3 The challenge of racial, ethnic, and gender identification and measurement .82 3.4 What if we do hold "all else constant?" .85 3.5 Women's other job: Housework .90 3.6 Oaxaca-Blinder decomposition methodology .94 3.7 Econometric methodology .98 Chapter 5 5.1 Racial inequality and social spending: Evidence from the United States and Europe .128 5.2 Clientelism and the underprovision of public services in the Dominican Republic .131 5.3 Are there political and social reasons for the contrast between Latin America and East Asia? .133 5.4 Political agency and the potential for redistributive strategy in rich countries: Lessons from the OECD .135 5.5 Increased equity through taxation and social spending in a democratic Chile .136 5.6 Failed redistributive efforts in a fragmented democracy: Social security reforms in Brazil .138 5.7 Neopopulism and policies on social funds in Peru .139 5.8 Transition at the sectoral level? Mexico's targeted antipoverty programs 141 5.9 Popular budgeting in Porto Alegre: Explaining a transition to a new political equilibrium 142 5.10 Local contexts and the transition from clientelism: Ibague versus Pasto ............................... 144 Chapter 6 6.1 Schematic representation of household income determination .150 vi CONTENTS Chapter 7 7.1 Business elites and the fear of education: Is there a low-road trap? Evidence from northeastern Brazil and the southern United States ..................................................... 186 7.2 Bilingual education: Preserving cultural heritage without being left behind ........................... 188 7.3 The poverty reduction potential of land in rural Mexico .......................................... 193 Chapter 8 8.1 Alternative views on the potential impact of market-oriented reforms and inequality .................... 219 8.2 The evolution of small-scale firms in the "Third Italy" ............................................. 225 8.3 Unions, firms, and the expansion of high-value export crops in Petrolina-Juazeiro in northeastern Brazil ...... 228 8.4 Institutions, macroeconomic policy, and volatility .............................................. 230 Chapter 9 9.1 Do tax incidence numbers mean much? ..................................................... 253 9.2 Issues in the valuation of public services ..................................................... 260 9.3 Summarizing benefit incidence using Gini income elasticities (GIE) ................................. 265 Figures Summary 1. Difference in average years of education between top and bottom quintiles for people 31-40 and 51-60, in 2000 . 3 2. Trends in income inequality in Latin America between the early 1990s and early 2000s ................... 4 3. The elasticity of poverty reduction with respect to growth for different levels of inequality .................. 4 4. Wages of various groups in relation to the wages of white males ..................................... 5 5. Marital sorting and income inequality ........................................................ 6 6. Distribution of mathematics and reading test scores for 15-year-olds in Brazil, Mexico, and OECD countries ..... 8 7. Tax revenue in relation to GDP per capita ..................................................... 12 8. Average and marginal incidence of access to various health services .................................. 13 9. Targeting to the poor and reach among the poor for different types of programs in Brazil .................. 14 Chapter 1 (Box 1.1) Inequality of opportunities as a share of total inequality, 1996 .20 1.1 A simple conceptual framework .22 1.2 Perceptions of fairness regarding the income distribution in Latin America .25 1.3 Inequality and poverty with a constant mean .26 1.4 Growth elasticities become less negative with inequality .26 Chapter 2 2.1 Income shares by deciles (household per capita income) ........................................... 39 2.2 Household per capita income distribution, excluding the richest 1 percent, Mexico, 2000 .................. 40 2.3 Equivalized household income ............................................................. 41 2.4 Indices of aggregate welfare for Argentina under alternative assumptions of a society's aversion to inequality ... . 47 2.5 Brazil's consumption inequality in international perspective ........................................ 49 2.6 Distribution of household per capita income, per region, 1990s ..................................... 54 2.7 Difference in average years of education between top and bottom quintiles, per country ................... 58 2.8 Hourly wage gap between skilled and unskilled workers, men aged 25 to 55 ........................... 60 2.9 Hourly wage gap between semi-skilled and unskilled workers, men aged 25 to 55 ....................... 61 2.10 Increase in hourly wages for college-educated workers, men aged 25 to 55 ............................. 61 2.11 Labor market outcomes by gender .62 2.12 Ratio of number of children under age 12 years of age per household between bottom and top quintiles .65 2.13 School enrollment rates by age and country .68 2.14 Gaps in access to services, by country .72 Chapter 3 (Box 3.1) The changing ethnic composition of the Americas .79 3.1 Completed primary education or less, by race, ethnicity, and gender .86 3.2 Proportion of workers in service or managerial occupations, by group and country .89 3.3 Wage gaps by race/ethnicity and gender .92 vii CONTENTS 3.4 Urban gender wage gap-ratio of women's hourly wages to men's hourly wages ......................... 93 (Box 3.7) a. Nonindigenous male vs. American Indian male (simulating earnings regression coefficients) .... ........ 99 (Box 3.7) b. Nonindigenous male vs. American Indian male (simulating only education) ...................... 100 (Box 3.7) c. Nonindigenous male vs. American Indian male (simulating both a and b) ........................ 100 3.5 Actual and simulated earnings distributions for white and indigenous men in Bolivia .................... 102 3.6 Actual and simulated earnings distributions for Indo-Guyanese men and women ....................... 103 Chapter 5 5.1 The syndrome of weak and unequal states: correlations between the rule of law and development outcomes .... 124 5.2 Political financing and its impact, by type ................................................... 128 5.3 Impact of bribes in shaping policy, according to a survey of firms by a recipient group, Peru, 2001 .... ...... 129 5.4 The rise and fall of the labor share in Argentina ............................................... 132 Chapter 6 6.1 Income and education inequalities across countries .152 6.2 Lorenz curves of years of schooling, selected countries .153 6.3 Difference in mean incomes per hundredth of the mean-normalized distribution: U.S.-Brazil and U.S.-Brazil (with U.S. conditional distribution of education) .156 6.4 Income and land inequalities across countries .157 6.5 Labor force participation and inequality across countries .158 6.6 Unemployment and inequality across countries .159 6.7 Informality and inequality across countries .160 6.8 Returns on schooling and inequality across countries .162 6.9 Distribution of and returns on education explain some of the differences between income distributions .163 6.10 Marital sorting and inequality across countries ................................................ 165 6.11 Youth dependency and inequality across countries .............................................. 166 6.12 Old-age dependency and inequality across countries ............................................ 167 6.13 The role of reproductive behavior in accounting for differences in income distribution between Brazil and the United States ............................................................. 168 6.14 Total public spending and income inequality ................................................. 168 6.15 Public expenditure on primary education and income inequality ................................... 169 6.16 Public income transfers to households and income inequality ...................................... 169 6.17 U.S.-Brazil differences, actual and simulated including nonlabor incomes and reweighting .... ............ 170 6.18 Retirement income as a share of total household income, Brazil and the United States .................... 171 6.19 Nonlabor incomes account for some of the differences between Italian and Chilean income distributions ...... 172 Chapter 7 7.1 The long chain of interactions from unequal schooling to more equal outcomes .178 7.2 Education dynamics in selected countries; years of schooling of men and women by age group .179 7.3 Years of schooling by cohort for blacks and whites in Brazil and South Africa .181 7.4 Spending by education level in Chile, Mexico, and Brazil .185 7.5 Polarization and property rights across countries .190 (Box 7.3) The marginal welfare impact of increased land ............................................. 193 (Box 7.3) a. The welfare impact of land assets and interactions with education .............................. 194 (Box 7.3) b. The welfare impact of land assets and interactions with indigeneity ............................ 194 (Box 7.3) c. The welfare impact of land assets and interactions with road access ............................. 194 7.6 Land inequality and growth, 1970-99 ............ .......................................... 195 7.7 Property taxation as a share of GDP in Latin America and other countries ............................ 203 7.8 Public and private investment in selected infrastructure sectors in Latin America, 1980-98 .... ........... 208 7.9 Increases in access to infrastructure, by income decile in Argentina and Nicaragua ...................... 209 7.10 Views on privatization in Latin America at different points in time ................................. 210 Chapter 8 8.1 The extent of policy reforms in Latin America, with country and policy variations .218 8.2 Correlation between changes in reform and changes in income inequality .220 8.3 Crony bias versus democratic voice and accountability .223 8.4 Tax compliance versus crony bias .224 8.5 Labor market rigidity and informality .227 viii CONTENTS 8.6 The association between macroeconomic volatility and measures of institutional conditions .... ............ 229 8.7 The evolution of the labor share in Chile, Mexico, and Peru ....... ............................... 235 8.8 Cumulative percentage change in time deposits (by residence and size) in Argentina ..................... 238 8.9 The evolution of foreign asset holdings in Argentina ............................................ 238 Chapter 9 9.1 Highest personal and enterprise income tax rates in Latin America .................................. 250 9.2 Value-added tax rate (in Latin America) ..................................................... 251 9.3 Unweighted average of tax revenue and taxes per category in Latin America ........................... 252 9.4 A simple cross-country regression of total tax revenues against log GDP per capita ...................... 253 9.5 The relationship between personal income taxes and income ...................................... 257 9.6 Public social spending in Latin America: expenditures per capita, as share of public spending and GDP, 1990s . 258 9.7 Average and marginal access incidence for various health services ................................... 262 9.8 Average and marginal benefit incidence for access to water in Latin America ........................... 262 9.9 Concentration coefficients for public social spending, Mexico 2000 ................................. 264 9.10 Targeting, coverage of the poor, and spending on various programs, Brazil 1997 .. . ...................... 266 9.11 Ratio of income of the 1st to 5th income quintiles, with and without public transfers, Chile, 1990-1998 ..... 267 9.12 Total social security expenditure .270 Tables Summary 1. Indicators of inequality for selected Latin American countries, the United States, and Italy. 2 Chapter 1 1.1 Theoretical elasticities of poverty with respect to aggregate income growth .27 Chapter 2 2.1 Gini coefficients of the distribution of household per capita income: Common sample countries, decadal averages by region .55 2.2 Distribution of household per capita income, 1992 (income shares by deciles and vintiles) .56 2.3 Distribution of household per capita income, 1990s (quintiles) .57 Chapter 3 3.1 Distribution of Afro-descended, indigenous, and female populations in the Americas and the Caribbean, various years (percent) .78 3.2 Interviewer perception versus self-perception of race-Brazil, 1997 (percent) .83 3.3 Interviewer versus self-perception of ethnicity-Peru, 2000 (percent) .83 (Box 3.4) Labor market outcomes for different groups .85 3.4 Ratio of female to male educational attainment and literacy .86 3.5 Difference in the probability of being in each sector, by race and by gender (percent) .87 3.6 Share of sector that is female (percent) .88 3.7 Occupational distribution, by gender .88 (Box 3.5) Average weekly hours spent on housework, by household member .90 3.8 Asset ownership, by race, ethnicity, and gender (percent) .91 3.9 Access to services by race, ethnicity, and gender (percent) .91 3.10 Decompositions of group mean log wage differentials .94 3.11 Within-group inequality share (for distributions of labor income) .95 3.12 Theil index decompositions (for hourly wages) .96 3.13 Individual earnings and inequality measures, based on actual and simulated earnings within racial and gender groups .97 Chapter 4 4.1 The distribution and composition of the population in New World economies (percent) .... .............. 111 4.2 Laws governing the franchise and the extent of voting in selected countries, 1840-1940 ............................................................... 114 4.3 Literacy rates in the Americas, 1850-1950 ................................................... 116 4.4 Land ownership rates in rural regions of Mexico, the United States, Canada, and Argentina in the early 1900s (percent) ................................................. 119 ix CONTENTS Chapter 6 6.1 Variation in student performance in PISA 2000 examinations ..................................... 155 6.2 Wage, work, and self-employment in selected countries, 1995 ..................................... 161 Chapter 7 7.1 Years of education by income quintile for Brazil and Jamaica, men and women aged 21-30 .... ........... 180 7.2 Mathematics test results for Brazilian students who identify themselves as black or of mixed blood, compared with white ....................................................... 182 7.3 Schools account for only a small part of variance in student learning outcomes ......................... 183 7.4 Gini coefficients for the distribution of operational holdings of agricultural land across regions .... ......... 191 7.5 Gini coefficients for the distribution of operational holdings of agricultural land in Latin American countries ............................................................. 191 7.6 Selected land reforms in Latin America in the 20,h century ....................................... 195 7.7 Distribution of housing wealth Brazil, Chile, and Peru .......................................... 201 7.8 Access to selected services for top and bottom quintiles, selected countries (percent) ..................... 204 7.9 Enterprise profitability and access to infrastructure in Guatemala (net income of owner in quetzales per-worker hour) ........................................... 205 7.10 Price changes after privatization . ........................................................... 211 Chapter 8 8.1 Effects of policy reforms on inequality: Estimates from three studies ................................. 221 8.2 Changes in real per capita income across the income distribution in Mexico, 1992-2000 .... .............. 232 8.3 Financial income during the 1994-96 crisis in Mexico .......................................... 233 8.4 Total fiscal costs of selected banking crises ................................................... 236 Chapter 9 9.1 Income inequality before and after taxes and transfers in selected industrialized countries .... ............. 248 9.2 Typology of Latin American countries by fiscal position and social spending, 1998 ...................... 248 9.3 Tax structure in the 1990s, Latin America and developed countries (in percent of GDP, consolidated central government) ......................................... 252 9.4 Tax incidence studies for Latin American countries, 1975-98 ...................................... 255 9.5 Increase in social spending with and without social security, ECLAC and IMF databases ..... ............. 259 9.6 Synthesis of case studies using estimates of the Gini income elasticity (GIE) ........................... 264 9.7 Social security expenditures around the world, as percentage of GDP ................................ 269 9.8 Social security expenditures in Latin America and the Caribbean, as percentage of GDP ..... ............. 271 9.9 Distribution of government expenditures on pensions and unemployment insurance in Brazil by quintile (percent) . .......................................................... 271 9.10 Conditional cash transfers: comparison of program sizes .......................................... 273 9.11 Benefit structure of Progresa, July-December 1999 (pesos/month) .................................. 275 9.12 Cumulative distribution of beneficiaries across national consumption deciles, in percent ..... ............. 277 x Acknowledgments I NEQUALITY IN LATIN AMERICA: BREAKING WITH HISTORY? IS THE PRODUCT OF A COLLABORATIVE EFFORT that brought together a diverse team of professionals from both within and outside the World Bank. The report was prepared under the guidance of David de Ferranti and Guillermo Perry by a team led by Francisco H. G. Ferreira and Michael Walton. Team members included David Coady (Institute for Food Policy Research), Wendy Cunningham (World Bank), Leonardo Gasparini (Universidad Nacional de la Plata, Argentina), Joyce Jacobsen (Wesleyan University), Yasuhiko Matsuda (World Bank), James Robinson (University of California, Berkeley), Kenneth Sokoloff (University of California, Los Angeles) and Quentin Wodon (World Bank). The preparation of the report benefited greatly from two meetings with the authors and a group of advisors, in Rio de Janeiro in August 2002 and in Washington DC in April 2003. The group of advisors comprised Barry Ames, Richard Bird, Francois Bourguignon, John DiNardo, Patrick Heller, James Mahoney, Lant Pritchett, Sergio Schmuckler, Judith Tendler, and Andre Urani. The report was also signifi- cantly influenced by commentary of the three principal reviewers: Martin Ravallion, Ana Revenga, and Michael Woolcock. Initial findings of the work were presented at two meet- Ocampo, Ricardo Paes de Barros, and Miguel Szekely, as well ings of the Network on Inequality and Poverty, a group that as by meeting participants. operates under the auspices of the Latin American and While the writing of this report has been a collective Caribbean Economics Association (LACEA), with the insti- effort, the principal authors for the various chapters are as tutional involvement of the Inter-American Development follows: Bank and the World Bank. These meetings, both held in June 2003, were organized, respectively, by Maximo Rossi * Summary: David de Ferranti, Francisco Ferreira, in Costa Rica, on behalf of the whole network, and by Guillermo Perry, and Michael Walton Sebastian Galiani and Leonardo Gasparini in Argentina, on * Chapter 1: Francisco Ferreira, Guillermo Perry, and behalf of the Argentine Chapter of the Network. A pre- Michael Walton liminary conference edition of the report was released to the * Chapter 2: Leonardo Gasparini public in Mexico City in October 2003 and then discussed * Chapter 3: Wendy Cunningham and Joyce Jacobsen at a combined meeting of the Network on Inequality and * Chapter 4: Kenneth Sokoloff and James Robinson Poverty and the World Bank's Annual Conference on Devel- * Chapter 5: Yasuhiko Matsuda and Michael Walton opment, in Puebla, Mexico, as part of the annual LACEA * Chapter 6: Francisco Ferreira meeting. Valuable commentary was provided byJose Pablo * Chapters 7 and 8: Michael Walton Arellano, Raquel Fernandez, Sebastian Galiani, Santiago * Chapter 9: David Coady, Francisco Ferreira, Guillermo Levy, Ricardo L6pez-Murphy, Nora Lustig, Jose Antonio Perry, and Quentin Wodon xi ACKNOWLEDGMENTS This report would never have been completed without the Estache, Marianne Fay, Ariel Fiszbein, Vivien Foster, excellent assistance of a number of individuals: Anna Crespo Daniel Kaufmann, Peter Lanjouw, Daniel Lederman, for chapters 1 and 3; Phillippe Leite for chapters 1 and 6; William Maloney, Andrew Mason, Ernesto May, Harry Jorge Balat, Matias Busso, Cecilia Calder6n, Martin Patrinos, John Redwood, Luis Serven, and Mauricio Cicowiez, Nicolas Epele, Federico Gutierrez, Marcela Santamaria. Massini, Augusto Mercadier, Alejandro Tamola, and Julieta Finally, a number of background papers and notes were Trias for chapter 2; Lucas Siga for chapter 3; Rodrigo Suescun commissioned especially for this study. These papers are by and Mariana Fajardo for chapter 9; Mayuresh Kshetramade Angela Albernaz, Barry Ames, Richard Bird, Wendy for chapter 6; Mary Fran Malone for chapter 5; Marcela Cunningham and Maurico Santamarfa, Ariel Fiszbein and Rubio Sanchez for chapters 7 and 8; and Kihoon Lee for assis- Sebastian Galiani, Marina Halac and Sergio Schmuckler, tance in provision of data sets. Patrick Heller and James Mahoney, Alberto Maldonado, The authors would also like to thank the many dedicated Valeria Pero, and Judith Tendler. A preliminary version of professionals within the World Bank who provided specific this report and background papers can be downloaded at contributions or commentary. These included, in particular, http://www.worldbank.org/laceconomist. Ana-Maria Arriagada, Jose Maria Caballero, Antonio xii Summary I NEQUALITY IN LATIN AMERICA IS EXTENSIVE: THE COUNTRY IN THE REGION WITH THE LEAST income inequality is still more unequal than any Organization for Economic Cooperation and Development (OECD) or Eastern European country. Latin American inequality is also pervasive, characterizing every aspect of life, including access to education, health and public services; access to land and other assets; the functioning of credit and formal labor markets; and attainment of political voice and influence. Inequality is also resilient; in its modern form, high inequality is rooted in exclusionary institutions that have been perpetuated ever since colonial times and that have survived different political and economic regimes, from interventionist and import substitution strategies to more market-oriented policies. Significant racial and ethnic differences also continue to persist even today. High inequality has major costs. It increases poverty and reduces the impact of economic development on poverty reduction. It is probably bad for aggregate economic growth, especially when associated with unequal access to credit and education, and with social tensions. A large majority of Latin Americans judge current levels of income inequality to be unfair. Inequality of opportunities is especially unacceptable. For all these reasons, Latin American countries must make an effort to break with their long history of inequality. Can this be done? This report suggests that the answer Third, the state needs to strengthen its capacity to redis- is yes, if there is decisive action to tackle the range of mech- tribute. For most countries in the region this will imply anisms that reproduce inequality. First and foremost, there increasing their (low) tax efforts, and, over the longer term, is a need to reduce inequality in access to productive assets. making taxes more progressive through effective collection Equalizing access to quality education is central because of of personal income and property taxes. Increased taxes only its influence on economic opportunities, social status, and make sense if they are used effectively. Despite some progress political influence. However, more equal education will take and the increased progressivity of social spending in the decades to subsequently transform other inequalities. Also 1990s, much public expenditure remains highly regressive important is the achievement of more equal access to land, (for example subsidies to tertiary education and the "trun- property rights, and other assets such as infrastructure. cated" welfare state, especially with respect to pension pay- Second, there is a need to make market institutions work ments). Redistributive transfers have a role to play precisely better for everyone through deeper financial and product because asset-based strategies take time to implement. A par- market, and more inclusive labor institutions that balance ticular area of promise is conditional cash transfers, which can flexibility with protection for workers. In the macroeconomic have a significant impact on income redistribution today, domain, a concern with distribution reinforces the case for while at the same time extending social protection against sound management, since crises tend to be highly regres- shocks to the poor and encouraging investment in the human sive. Attaining this will require building institutions and capital of the poor. These and other instruments can provide rules to reduce the risk of crisis, and to make the distribu- the basis for a truly progressive social protection system in tion of losses less unequal when crises do occur. Latin America. I INEQUALITY IN LATIN AMERICA Given the deep historical, institutional roots of high According to household surveys, the richest 10 percent inequality, progress on all these fronts will require decisive of individuals receive between 40 and 47 percent of total social action and political leadership. This implies making income in most Latin American societies, while the poor- progress toward more inclusive political institutions, since est 20 percent receive only 2-4 percent (see table 1). These inequalities in influence lie behind many of the mechanisms differences are substantially higher than in OECD coun- that reproduce overall inequality. tries, Eastern Europe, and most of Asia. Moreover, the most This is no easy task. But it may be more possible to accom- distinctive attribute of Latin American income inequality plish today than in the past, given increased social demands is the unusually large concentration of income at the very for deeper democracies, more equal sharing of political influ- top of the distribution. (Only some countries in Africa and ence, broader access to education and health, and greater the successor states of the former Soviet Union have com- recognition for Afro-descendents and indigenous groups. parable inequalities.) By way of comparison, the richest These demands are palpable across the region, and are partly 10 percent in the United States receive 31 percent of total a consequence of the globalization of information, economic income, and in Italy they receive 27 percent. Even the opportunities, and human rights. There are promising exam- most equal countries in Latin America (Costa Rica and ples of change underway, notably at subnational levels, with Uruguay) have significantly higher levels of income new alliances between progressive elites, public officials, inequality. Inequality in consumption-where it can be middle classes, and the poor currently spurring the estab- adequately measured-also appears to be higher in Latin lishment of more inclusive and efficient institutions. American countries, although differences with other These are the main messages of this report, on which the regions seem not to be as sharp as in the case of inequal- rest of this summary section elaborates. Readers interested ity of incomes.2 in the evidence and analysis that underpin these conclusions Inequalities with respect to education, health, water, san- will have to immerse themselves in the entire, extensive text. itation, electricity, and telephony are also typically large and We hope they will find it worth the effort. correlated with differences in income. For example, differ- ences in average years of education between the top and Latin America suffers from extreme inequality- bottom income quintiles ranged between 5 and 9 years for on many dimensions 31-40 and 51-60 year-olds across the region (figure 1). Inequality is a pervasive feature of Latin American' societies Standard surveys do not provide comparable material on in terms of differences in income, access to services, power inequalities of power or influence within a society, but a and influence and, in many countries, treatment by police wealth of political, historical, and sociological information and justice systems. attests to both their salience and association with wealth. TABLE 1 Indicators of inequality for selected Latin American countries, the United States, and Italy Share of top Share of bottom Gini 10 percent in total 20 percent in total Ratio of incomes of coefficient income (percent) income (percent) 1oth to 15t decile Brazil (2001) 59.0 47.2 2.6 54.4 Guatemala (2000) 58.3 46.8 2.4 63.3 Colombia (1999) 57.6 46.5 2.7 57.8 Chile (2000) 57.1 47.0 3.4 40.6 Mexico (2000) 54.6 43.1 3.1 45.0 Argentina (2000) 52.2 38.9 3.1 39.1 Jamaica (1999) 52.0 40.1 3.4 36.5 Dominican Republic (1997) 49.7 38.6 4.0 28.4 Costa Rica (2000) 46.5 34.8 4.2 25.1 Uruguay (2000) 44.6 33.5 4.8 18.9 United States (1997) 40.8 30.5 5.2 16.9 Italy (1998) 36.0 27.4 6.0 14.4 Source: Statistical appendix and SDStats database. 2 SJ M MARY FIGURE 1 Difference in average years of education between top and bottom quintiles, for people 31-40 and 51-60, in 2000 Years 10 _ 0 Age 51-60 M Age 31-40 9 8- 7- 6- 4 Source: Statistical appendix table A.23. A further divide that cuts across all these dimensions is one key area, that of access to college education, the domi- that of race and ethnicity, that affects all societies with a sig- nant pattern was of widening differences between the rich nificant indigenous or Afro-descended population, as dis- and the poor as attendance rose faster for members of house- cussed further below. holds at the top of the distribution. This is important for Is inequality improving? During the last decade, diverse future income dynamics, in light of the high and rising pre- patterns have emerged with respect to income differences, mium placed on tertiary education in the labor market. The with more countries experiencing a worsening than an wave of democratization in the 1980s brought, at a basic improving trend. On balance, relatively equal countries expe- level, some equalization in citizenship, and potentially in dif- rienced some worsening, with a dramatic deterioration for ferences in power. However, inequalities in influence and in Argentina both before and during its economic crisis. At the the application of the rule of law remained large in most of other end of the scale, Brazil, historically the most unequal the region, even under democratic auspices. country in the region, experienced a modest but significant decline in income inequality (see figure 2). In previous High inequali't hurts both povert reduction decades, there was a trend toward reduced inequality in the and development 1970s and a more pronounced trend toward rising inequal- Most people would agree that some degree of inequality in ity during the crisis-ridden 1980s. However, the more strik- a society is necessary to provide incentives for work and ing fact for the long term is the resilience of high inequality investment. However, the levels of inequality prevailing in in the face of diverse economic and political regimes. Latin America are clearly costly for well-being. There are With respect to services and human investment, the gen- three broad reasons for concern. First, higher inequality, eral tendency in the past decade was more positive. Most of whether in income or other dimensions of well-being, the region experienced at least some equalization in terms means more poverty at any given point in time. High of access to services and basic education status. However, in inequality also implies a lower dynamic impact on poverty 3 INEQUALITY IN LATIN AMERICA FIGURE 2 Trends in income inequaliy in Latin America between the early 1990s and early 2000s a. South America b. Central America and the Caribbean Gini coefficients for equivalized Gini coefficients for equivalized household income per capita household income per capita 60 - 60 - '55- - - 50- 50 ~-0 45 0 45 ----_-- - - - - 40 1 , i 40- Early-1990s Mid-1990s Early-2000s Early-1990s Mid-1990s Early-2000s i Argentina -*- Bolivia - Brazil Chile |U- Costa Rica . El Salvador - Honduras Colombia c- Peru -_- Uruguay 4 RB de Venezuela | Jamaica _ Mexico + Nicaragua -0 Panama Source: See figure A.1, statistical appendix. in ten years with 3 percent growth and an improvement of FIGURE 3 5 percent in the Gini coefficient (the most common mea- The elasticity of poverty reduction with respect to growth for sure of income inequality); it would take the country 30 different levels of inequaliy years to achieve the same objective with 3 percent growth Gini coefficient g and no improvements in income distribution. Percent 0.3 0.4 0.5 0.6 Second, inequality can slow the overall development 0.0 - -o.s ; - - process. In contrast to some earlier strands of development thought, most economists (and other social scientists) now 1.0 L, 1S+>. 4 see high inequality as a potential drag on development for -.5- a variety of reasons: unequal access to credit implies miss- -2.0- L -2.5 - ming highly profitable investment opportunities for the econ- -3.0 - 1 . ' omy as a whole; unequal educational opportunities limit the potential contribution to society of some of the most talented -3.5 - l . individuals; distributional conflicts are heightened, especially -4.0 in the context of managing adverse shocks; crime and vio- -4.5 lence increase; and, under some conditions, the institutional Note: The elasticity is the percentage change in poverty : underpinnings for growth become weaker, for example with corresponding to a one percent change in mean income. respect to property rights. Source: Authors' calculations based on a log-normality Third, according to surveys from Latinobar6metro, the assumption, from a 33 percent initial rate of poverty incidence. public opinion survey organization, high inequality is widely - ~ < * ^ disliked: in almost all countries surveyed, some 80-90 percent from development, unless significant redistribution takes of citizens consider prevailing levels of income inequality to place (figure 3 provides a numerical illustration of the be unfair or very unfair. relation between economic growth and income poverty Inequality of opportunities is especially unacceptable in reduction). For example, Brazil could reduce poverty to half ethical terms-implying that individuals at birth face totally 4 SUMMARY different life options-and is particularly harmful for the FIGURE 4 overall growth potential of societies. Although it is difficult W of u Wages of various groups in relation to the wages of wfhite to disentangle the effect of inequality of opportunities from males the effect of other factors (such as differences in preferences and efforts) on the inequality of outcomes, some attempts Ratio indicate that they do indeed account for a significant fraction 1 * 0.9 of inequality of income in countries such as Brazil (see 08 chapter 1). 0.7 - 0.6- Latin America's inequality has deep historical 0.5 - roots and pervades contemporary institutions 0.4- The genesis of current structures of inequality lies in Latin 0.3 - America's colonial past, and in particular in the interactions 0.2 - between European colonists and subordinate populations. 0.1 - During the early colonial period, influences on inequality Brazil Guyana uatemala Bolivia were reinforced by economic conditions and power differ- ences. European settlement was initially concentrated in areas a Nonwhite male M White female I Nonwhite female where natural resources-especially mining and sugar production-could be exploited with the use of unskilled Source: Authors' calculations based on microdata from labor. Labor was provided either by subjugated indigenous household surveys; see chapter 3. populations or imported African slaves. The colonists developed institutions-notably those related to labor man- the United States and Canada, but was ahead of countries agement (including slavery), land use, and political control- such as Brazil, Peru, and Mexico. that consolidated and perpetuated their influence and wealth. Despite the immense political, social, and economic In the post-independence period, domestic elites continued changes of the past century, these historically-formed sources to shape institutions and policies to maintain their privileged of inequality persisted until the present time, albeit in shift- positions, for example in the areas of restricted suffrage, ing institutional forms. In countries with indigenous or Afro- access to education, and land policy. descended populations, differences remain large: incomes of In those parts of the Americas where there was no eco- these groups are half that of their "white" counterparts in nomic potential for combining subordinate labor with valu- Bolivia, Brazil, and Guatemala. (See figure 4; Guyana is an able natural resources, the evolution of inequality was exception, with relatively small differences between Afro- and different. This is most clearly the case in North America, Indian-descended groups.) These disparities in turn reflect dif- where immigrant smallholders (who were able to pursue ferences in education and other assets of disadvantaged ethnic settlement with lower mortality rates) successfully resisted and racial groups. At the other end of the scale, elites have attempts to impose authoritarian forms of government. perpetuated their position through a variety of mechanisms, (Nonetheless, the use of slavery in the American South was including in the realm of social relations. One example is the a story similar to that found in Latin American sugar- high propensity ofpeople to choose marital partners with the producing countries.) Similarly, in Costa Rica and the South- same education (figure 5). ern Cone, the depth of social divisions was less than that Of particular importance for the design of public action is found in the areas where indigenous and slave populations the historical heritage in the capacities of the state. At the end were concentrated. However, in the latter countries, a range of the 20"h century, most Latin American states still conformed of forces, including abundant land and high concentrations most closely to a model of patronage and clientelism that was of power among elites, led to these societies pursuing paths embedded within a broader pattern of unequal societal rela- that also fostered relatively high levels of inequality. The tions, albeit with islands of high levels of technical compe- granting of suffrage and extension of education in Argentina tency. This pattern is associated with high degrees of inequality and Chile, for example, substantially lagged behind that of of influence, with disproportionate influence over the state by 5 INEQUALITY IN LATIN AMERICA FIGURE 5 Marital sorting and income inequality Gini coefficient 65 -pi 0.68* 0 Consumption Gini * Other countries, income Gini 0 Latin America. income Gini P 0.63* - Income Income or consumption ~~~~~~~~~~~~~~~~~~~~~~~Brazil 60 - Income - Income or consumption Paraguay 0 Bolivia Colombia Panama0 OChile 0 M 55 - 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Ecuador 0 Argentina 50~~~~~~~~~~~~~~~~~ 45 - O Uruguay 40- 35 40 30~~~~~~~~~~~~~~~~~~~~~ 25- . 0 O S * * O 20 - 15 - 10 l l l l l 30 35 40 45 50 55 60 65 70 75 80 Marital sorting Data Sources: Chapter 2 for Latin American income Ginis, World Development Indicators Database for other income Ginis, and Fernandez et. al. (2001) for marital sorting Pearson correlation coefficients. Note: *Significant at 5 percent level; **Significant at 10 percent level. Source: Authors' calculations based on microdata from household surveys; see chapter 6. wealthy individuals or corporations while poorer groups typ- is also likely to be slow for overall income inequality. However, ically interact with the state through vertical relations of change can be relatively fast in terms of patterns of service pro- patronage, or are excluded. Few states have effected the tran- vision, for specific parts of the income distribution (includ- sition to programmatic parties and autonomous bureaucracies ing among the poorest), and, under some conditions, in that was a feature of institutional change in much of Europe reducing inequalities of power. In addition, precisely because and North America.3 This is further reflected in the relatively change takes time, concerted, early action is of great impor- weak capacity of Latin American states to deliver key public tance. A range of economic policy levers that are key to goods (especially macroeconomic stability, safe property rights, making a difference are outlined below, followed by consid- and citizenship) and a broad-based provisioning of services eration of the underlying political and social changes that will financed by taxes. Such failures in the provisioning of public be required to support and sustain shifts in economic policy. goods and services are almost always disequalizing. ... Through a range of economic policy levers But a "break with history" is possible ... Guidance for economic policy choice is provided by an analy- While this report emphasizes the influence of history and its sis of the correlates of high inequality in the region, drawing contemporary reflection in institutional structures, it is not both on international comparisons and diagnoses within deterministic. Economic, political, and social analyses indi- countries. This analysis indicates that there is no single eco- cate the potential for change, and of political agency as a source nomic reason for excess inequality in Latin America. In eco- of change, whether with regard to experiences within Latin nomic terms, high levels of inequality are reproduced via America or comparisons with other societies. The speed of interactions among several influences: moderate levels of potential change varies across dimensions of inequality. It is inequalities in education (exacerbated by low and unequal intrinsically slow for education and, in part as a consequence, educational quality); high market returns on education, 6 SUMMARY especially college education; high levels of concentration of a significant jump in both coverage and quality across the land and other productive assets; unequal access to critical public school system at both primary and, in particular, markets (particularly financial and labor markets); and weak secondary levels. redistribution through the state.4 An economic strategy to While there is no specific institutional blueprint for this reduce inequality needs to reflect these interactions. This in strategy, successful actions are likely to involve mechanisms particular implies complementary actions in the following three to increase the accountability of teachers and schools. These areas: asset inequalities; the institutions that shape market may take the form of incentives for results (as is the case access and returns; and use of the redistributive potential of with Chile's school contests and vouchers), special funds to the state. In some areas, policies to reduce inequality will supplement school budgets (as with Brazil's FUNDEF pro- support higher aggregate growth and even bring gains to gram), or greater participation by local communities (as with all (or almost all). In other areas there will be tradeoffs, if El Salvador's EDUCO approach). Such supply-side measures not in aggregate growth then in reducing the privileges of can be complemented by demand-side incentives that those who currently benefit from inequality. encourage kids to stay in school, as with Oportunidades (pre- viously Progresa) in Mexico and Bolsa Escola in Brazil. Broadening asset ownership: Redistribution (These latter measures can also serve the purpose of broad- with growth ening social protection systems to poorer groups, as discussed further below.) All these experiments will have to be accom- Democratizing education panied by ongoing evaluation of which approaches have the Education is the most important productive asset most highest impact on student learning, in order to guarantee people will ever own. Apart from its economic effects, edu- an efficient use of education spending. cation is intimately linked to sociocultural and political For tertiary education, the challenge is to increase the inequalities. More equal education has potentially multiple access of bright children from poor backgrounds. Part of the influences on more equal outcomes and practices. In addi- problem in this regard is low quality and lack of coverage tion, it has two important advantages as a strategy: its dis- at lower levels. However, there is also potential for tackling tribution can be improved without the need to redistribute the credit and informational constraints that limit entry. it away from someone else, and improvements in its distri- Measures such as sharply expanding the availability of edu- bution (which go hand-in-hand with increases in overall cational credit, targeted scholarships for students from poor- mean levels of education) are good for efficiency and growth. est families, state exams, accreditation programs, and labor Although the importance of education is now largely observatories would greatly facilitate the required expansion accepted by governments, implementation has not always of tertiary education. In this context, higher cost recovery been easy. In particular, the recent "massification" of basic in public universities and performance-based budget sup- education appears to be associated with new forms of inequal- port would be important complements. ity, in particular those associated with high variance in qual- In addition to such general measures, there is a need for ity (see figure 6), and with the fact that elites have the action to deal with the exclusionary structures faced by his- financial means to opt out of public systems. torically subordinate groups. A case can be made for invest- Despite widespread experimentation, the magic bullet ing in the curricula and teachers to provide multicultural for raising educational quality has yet to be found. Fur- and multiracial education, including bilingual education in thermore, the key divide in terms of market returns is now areas with a strong presence of indigenous people. At the the attainment of a college education. The increased value tertiary level, there is a prima facie case for use of affirma- of higher education is driven by the skill-biased nature of tive action programs, especially for such excluded groups technological change. But higher education remains largely such as Afro-descendents in Brazil. However, such programs the preserve of the children of the rich, with, as noted can lead to backlash and stigmatization and need to be care- above, increases in inequality of enrollment in the past fully discussed, designed, and monitored. decade. In this context, an important element of a deeper Although a push toward equalizing access to quality edu- education strategy could be the introduction of a civic drive, cation is critical for the distribution of income in the long involving both the public and the business sector, to achieve run, the effects of education will take time. Since children 7 INEQUALITY IN LATIN AMERICA FIGUPE 6 Distribution of mathematics and reading test scores for 15-year-olds in Brazil, Mexico, and OECD countries Distribution of Distribution of mathematics test score reading test score 0.5 -o.s - OECD average 50 percent above 500 50 percent above 500 _ | O- _ _ ~~~~ ~ ~~~~~~~~I O-I 367 500 625 366 500 623 0.5 - 0.5- Brazil 4.4 percent above 500 percnt above 500 212 334 464 288 396 507 0.5 -0.5 - Mexico << Mexico / 1.8 percent above 500 8.6 percent above 500 ~~-OI I O-I 281 387 496 311 422 535 Note: Figures are based on results from the OECD's Programme for International Student Assessment; distributions are approximated on the basis of the mean and standard deviation reported in the original source. Source: OECD 2001. who benefit from better schooling now will enter the labor a comparison between Chile and Italy found that differences market in the future, it will take decades to transform the in the structure of income were related more to differences educational composition of the entire labor force. During the in returns than to differences in the distribution of years of transition, tertiary expansion could even lead to increased education. While these comparisons are only statistical exer- inequality, if the high returns to tertiary level graduates per- cises, they suggest that many factors influence the effect of sist. There is also uncertainty over what will happen to the education on income inequality. pattern of returns on education over the long term. A com- parison between Brazil and the United States (see chapter 6 The distribution of land and the security of and table 6.3) found that differences in the pattern of edu- property rights cation in the labor force "explained" about a third of the dif- Unequal distribution of land was historically a major source ference in income inequality, with another third explained by of both high levels of economic inequality and unequal polit- differences in the pattern of returns on education. However, ical influence. With the enormous rise in urbanization and, 8 SUMMARY in most countries, the disappearance of landed elites as a major measures to make housing markets work better, provide source of political influence, rural landholdings are less impor- local infrastructure, and ensure that local governments are tant than in the past. However, rural land distribution still responsive to poor urban dwellers. matters for an important, and relatively poor, minority in In both rural and urban areas, strengthening property middle income countries, as well as for larger groups in the taxes is an important area for public action. Latin America's poorer countries in the region. property taxes are low by international standards, even after Latin America can be categorized as a region of extensive controlling for incomes. Increased property taxation can both but incomplete land reform. Some past efforts succeeded in be a valuable source of revenue for local governments and breaking up haciendas, but almost all failed to transform the provide incentives for more efficient land use. position of poor peasants, because of shortcomings in design and the political context. When property rights became more Broadening the provision of infrastructure insecure, landed elites often used other means to preserve Unequal access to public infrastructure is another impor- their economic position, such as evicting tenants and secur- tant determinant of inequality dynamics. Recent cross-coun- ing subsidies on capital or outputs. In addition, newly try work indicates that the more infrastructure a country reformed areas were not supported by the infrastructure and has in place (including electric power, telecommunications, other economic services necessary to generate a vibrant small- roads, and railways), the less unequal it is. Evidence sug- scale sector. gests that this relationship is causal, and that the effects can There is scope for rural land reform, but its nature and be large over the long term (Calder6n and Chong 2004). degree depend on the country. Where property rights for The sharp reduction in investments in infrastructure during land are generally recognized and legitimate (as in most of the 1980s and 1990s in many countries in Latin America the Southern Cone, for example), coercive action would be (see Calder6n, Easterly and Serven 2003) is thus cause for undesirable and counterproductive. However, there are many concern from the perspective of both growth and equity. countries that have regions of land settlement where rights This trend has in turn led to a large and growing gap in are ill defined or illegitimate. In those cases, redistribution terms of infrastructure between Latin America and both can potentially go hand-in-hand with a strengthening of OECD and successful East Asian countries. The decline was property rights if the process is well managed. This is likely driven by large declines in public spending on infrastruc- to involve a range of techniques, from repossession and redis- ture, first during the debt crisis of the 1980s and then tribution of land owned by druglords to the changed use of within a context of bullish expectations related to the poten- land that is still in public hands. In Brazil and Colombia, tial of privatization in the 1990s. Although private provi- processes of state purchase for redistribution and experiments sioning did increase, it only fully compensated for public in community and market-based land reform are ongoing. declines in the case of telecommunications. In terms of post-reform designs, indigenous groups some- Tackling inequalities in the future will require a judicious times prefer to maintain traditional collective forms of land mix of public provisioning and privatization. The privati- management, with the issue being how to provide adequate zation process has become highly politically charged. In access to credit and services in the context of these property many areas, pure private provisioning is unlikely, especially and governance structures. In all areas, it is essential to com- in such areas as rural road networks and water supply. In plement small-scale production with infrastructure and other these subsectors, public and private investments in infra- support services, to integrate new schemes within territor- structure are complements, not substitutes. The issues here ial strategies for development, and to provide the legal are analogous to the challenges of making any public ser- underpinnings for rental markets. vice responsive to the needs of all citizens. In many areas, Issues of urban land and housing have increased in impor- however, a case for privatization can still be made, includ- tance with urbanization. An important divide is that of the ing on grounds of equity. security of property rights, since poorer groups are more Many public utilities have been part of clientelistic struc- likely to suffer from lack of legal title. Experience from Peru tures and have been both inequitable and inefficient. Priva- indicates that it is possible to run large-scale titling pro- tization efforts can break through these barriers, but will only grams, an approach that needs to be complemented by succeed if they are well designed and subject to strong, 9 INEQUALITY IN LATIN AMERICA autonomous regulation. Evidence of past privatizations of with a recent leveling off in wage inequalities in the case of utilities suggests that most tend to lead to more equal access both Chile and Mexico, two of the most internationally inte- and higher quality services, but price effects have varied from grated economies in the region. case to case depending on initial conditions and the quality The best response to such inequality-augmenting forces of regulation. In addition, the process is key: in most cases, is not to avoid the use of markets, but to tackle the under- privatization processes have not been used to distribute own- lying sources of inequality. Doing so involves broadening and ership more broadly and, where transparency and competi- equalizing the asset base-notably in education and tion have been weak, regressive wealth transfers and infrastructure-and reforming social protection systems to corruption have occurred. This is particularly the case in non- make them more inclusive. It is also important to extend competitive sectors, in which the establishment of sound reg- and deepen the reach of markets. Part of the problem is that ulatory structures alongside privatization is essential to market access is very different for poorer households and achieving both efficient and equitable outcomes. informal sector workers than for the rich and established companies. Financial markets are of particular importance Developing deeper markets and more in this regard, both for improving equality of opportunities equitable institutions in productive investment and facilitating more equal access to assets such as land, housing, and education. Transparency, Deeper markets disclosure, good corporate governance, credit bureaus, and Promarket reforms are often accused of favoring the rich at strong property rights for creditors and minority shareholders the expense of the poor. While this tends to ignore the fact have proven to be important determinants of financial deep- that most nonmarket institutions in Latin America were ening, and thus of greater access to financial services. There historically shaped to favor the rich, it is important to are also a host of specific institutions, products, and collat- understand the distributional impacts of market reforms in eral regulations that may increase access to credit by micro- light of their importance for the growth agenda. A survey and small-scale enterprises. of the literature finds no strong general pattern associated with the impact of structural reforms on inequality. On bal- Inclusive and efficient labor markets ance, market-based reforms have been associated with The labor market is the most important domain for income greater income inequality more often than not, but the determination for most households. Labor market institu- effects are not large and often not statistically robust. The tions can have a significant influence in this area, particu- big picture is that inequality has been notably resilient to larly with regard to the quality of work. Current labor a range of policy experiments, from import-substituting market and social security institutions in Latin America (such industrialization to populist policies and market-oriented as unions, minimum wage rules, and employment benefits) reforms. protect the interests of formal sector workers and largely More can be said when specific reforms are considered exclude informal workers and most of the unemployed. In individually. Whenever fiscal reforms have led to price sta- contrast with conditions in OECD countries, formal sector bilization, the result has been a reduction of inequality. Fur- workers in Latin America seldom make up a majority of the thermore, the effects of privatization on distribution have working class, and are usually located in the middle or upper often been positive (see discussion above). However, many parts of the income distribution. In almost all countries, a countries in the region experienced rising inequality in the majority of poor workers are employed (or self-employed) wages of individual workers in the past decade or so,5 which in the informal sector, where they are not protected by formal appears to have been mediated by an opening up of trade benefits, minimum wage rules, or unions. Organized work- and foreign direct investment (De Ferranti and others 2003). ers sometimes effectively practice a form of "opportunity Trade liberalization, in addition to the unequal distribution hoarding," or protection of their privileges, that can limit of education, has been associated with increased wage dis- the scope for equitable reforms. Examples include efforts to parity because it has led to faster adoption of new techniques protect favored public sector pensions in Brazil and the resis- and production processes that are intensive in skilled labor. tance of some teachers' unions to quality-enhancing reforms However, there is some evidence that this is a once-off effect, in basic education that could benefit all children. 10 SUMMARY There is a need to move toward a labor and social secu- markets, improving debt management, and increasing tax rity framework that is both more inclusive and less distort- ratios. Since the latter objectives will take time to achieve, ing. Such a framework should seek, on the one hand, to it would also be desirable to generate cyclically adjusted extend basic labor rights and security protection to all work- public sector surpluses for a period of time in countries with ers while, on the other hand, increasing labor flexibility, espe- high debt burdens, in order to reduce vulnerability to shocks cially in areas where groups have significant, entrenched and the likelihood of crisis. privileges. This approach is consistent with an active role A key element is the establishment of rules and institu- for unions within competitive environments. tions that would allow for the operation of countercyclical fiscal policies, by mitigating the political economy and informa- Avoiding crises and regressive workouts from tional problems that lay behind procyclical fiscal policies. In macroeconomic shocks turn, this would overcome the procyclicality of social expen- A notable finding of this report is the regressivity of finan- ditures and avoid their reduction when they are most needed. cial crises. Crises are regressive because the wealthy and Pursuit of fiscal prudence on equity grounds may seem better informed move money out of their countries first, surprising to those who have long equated propoor gover- sometimes enjoying capital gains in terms of purchasing nance with budget deficits. This view is consistent with the power at home. Crises are also regressive because of the fiscal need for substantial amounts of public action-and public costs of the resolution mechanisms that have been adopted spending-to reduce poverty, inequality, and various sources to deal with recent banking problems in Latin America. of inefficiency across Latin America. However, financing for During the 1980s and 1990s, public transfers to large these expenditures should not be generated through larger depositors and bank shareholders, undertaken as crisis- fiscal deficits because doing so would increase the likelihood resolution mechanisms (that is, bailouts), were in the range of inflation and balance-of-payment and banking crises, both of 13-5 5 percent of gross domestic product (GDP). These of which have highly regressive resolutions. Instead, financ- transfers dwarfed poverty-targeted programs and were ing should come from other sources, including higher taxes financed by a combination of higher taxes and lower public and the redirection of existing spending, some of which is benefits and services. Since the incidence of overall tax sys- almost certainly wasteful and much of which is regressive. tems is generally proportional (or possibly mildly progres- Finally, since crises can never be avoided entirely, it is sive) and the marginal incidence of service expansion in the critical to have in place ex ante efficient and equitable crisis 1990s was typically progressive, these transfers were resolution mechanisms (especially with respect to potential strongly regressive. It is unlikely that such large regressive financial problems) in order to avoid the wasteful and highly transfers to large debtors and depositors-and even share- regressive transfers that have characterized the management holders on occasion-were actually needed to avoid the of most banking crises in the past. These mechanisms should collapse of banking systems. be complemented by safety nets that automatically kick in Policies to reduce the likelihood and severity of crises are to protect the poor when crises do hit (De Ferranti and thus a necessary part of an agenda to reduce inequality. Latin others 2000). American economies have remained excessively vulnerable to reversals of capital flows due to a combination of moder- Using the redistributive power of the state ate to high public debt levels, excessive reliance on foreign and dollar-linked debt, low export and tax ratios, weak pru- Raising taxes for an effective state dential regulation and supervision, and procyclical fiscal poli- A healthy tax system is a vital ingredient for an effective cies. The best course for macroeconomic policy in the region, state. Currently, most countries in Latin America have low from the point of view of reducing poverty and inequality, tax-to-GDP ratios given their level of development (see would most likely be one of increasing public sector savings figure 7). Some, notably Uruguay and to a lesser extent Chile and adopting "super-prudent" regulatory and supervisory and Costa Rica, have moderately high taxes and significant practices in the financial system during good economic times, and somewhat redistributive public provisioning. Brazil while at the same time increasing the level of openness of stands out as a case in which a relatively high tax effort coin- the economies, developing long-term domestic capital cides with highly unequal public provisioning. While 11 INEQUALITY IN LATIN AMERICA FIGURE 7 Tax revenue in relation to GDP per capita Tax revenue as a share of GDP Percent 45- * S 40 - Italy * * r France 30- 30 - *~~~~~~~~~~~~~~~~~~~~~~~~ Spain r 25 - Nicaragua * Uruguay i 20 - S 20 Costa *ICa * = *0 United States i * _ i ~~~~~~~~~~~~~~~Costa Ric 15- Peru Meco 0 *Dominicarkjepublic * Bolivia 5 10- Colombia 0@ Argentina Guatemala 5 - 0O I I I I I I I 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 Log GDP per capita (in constant 1995 U.S. dollars) Average 1990-2000 Source: Authors' calculations based on microdata from household surveys; see chapter 9. methodological problems bedevil incidence analysis and little rate for most other products, and a higher rate for luxury can be stated with confidence, a reasonable working items. Excise taxes may be imposed on goods that carry high assumption is that, given very low collections of personal negative externalities, such as tobacco, alcohol, and motor income and property taxes, most tax regimes in the region vehicles. However, some excise taxes (such as those placed are proportional or only mildly progressive or regressive. This on kerosene and tobacco) are typically quite regressive, situation implies that the poor, middle class, and rich pay implying the need to balance equity with other objectives. roughly the same proportion of their incomes in taxes. There is also an important long-term role for personal What kinds of changes in tax policy are desirable income and property taxes to play, since tax collections are will depend on the initial position of a country. In poorer very low in most Latin American countries in comparison to countries, often the greatest priority lies in developing a the OECD. Income tax structures need not have many rates reasonably well-functioning tax administration in order to to be progressive. Efforts to tax the rich a little more should expand the resources available to the state. Where this has focus largely on improving enforcement and closing loopholes not yet been done, there is a case to be made for moving indi- and exemptions in order to reduce tax avoidance and evasion. rect tax systems in the direction of applying broad-based This is also true for property taxes, which exist in many coun- value-added taxes (VATs) that can combine strong revenue- tries but generally raise almost no revenue, and which are crit- generating capacities with reasonable levels of efficiency. In ically important to financing local public services. Reform order to achieve modest levels of progressivity, a new VAT efforts should not focus on raising the rates of these taxes, system should have a simple rate structure, possibly con- but on eliminating exemptions and devising disclosure mech- sisting of a few product exemptions (largely for food), a basic anisms to improve collection and enforcement. 12 SUMMARY FIGURE 8 Average and marginal incidence of access to various health services a. Ratio of average incidence, quintile 1 to 4 b. Ratio of marginal incidence, quintile 1 to 4 Ratio Ratio 3.5 3.5 Antenatal. 3.0 - 3.0 - care: 2+ visits 2.5 - 2.5 - 2.0- 2.0- 1.5- 1.5- attendance: * Treated for Antenatal pubic :1.0 diarrhea care: 2+ visits i.o * *~~~~~- * * _$ ~~~~~~~~~~~~~~*Antenatal O.S - 0 * O.S - * Treated care: doctor for diarrhea 0.0- I I I 0.0- I I I 15 35 55 75 95 15 35 55 75 95 Access rate (percent) Access rate (percent) Source: Ajwad and Wodon 2003. Providing services for all average and marginal incidence of health programs, and Given the limits on redistribution from taxes, especially in indicates how marginal incidence is more progressive for the short term, the spending side of the equation is of fun- those programs that have initially high access rates. In infra- damental importance. Evidence confirms what historical and structure sectors with low coverage (such as telecommuni- political analyses often suggest: most public services reach cations in most of the region and electricity in the poorest the rich before they reach the poor. The average incidence countries), expansions have also tended to favor richer of a large number of services remains regressive in most groups first. countries in Latin America. The good news, however, is that Latin American countries also have significant experience during the 1990s, the expansion of access to basic educa- in the use of targeted delivery mechanisms. Some programs tion and health was quite progressive. To a large extent, this are highly successful in targeting the poor, but they tend to is simply because expansions automatically brought in remain small. Indeed, there appears to be a tradeoff between poorer people, since the rich already had access to such ser- effective targeting and the degree of coverage of programs vices. In many cases, however, this shift also resulted from among the poor (see figure 9 for examples from Brazil). Such deliberate, well-meaning, effective public action. a tradeoff may be partly due to administrative difficulties, Given the weight of these basic services, the overall but is also likely to be an outcome of interactions with increase in social spending that took place in almost every political and economic institutions. Progressive policies country in the region in the 1990s (due to both increases generally need to be based on broad coalitions of support. in the ratio of social expenditures to GDP and growth) was Thus, targeted programs probably should be linked or "pack- in general progressive, in spite of the fact that an impor- aged" together with universal access programs that also tant chunk went to pension subsidies, which are usually benefit the middle class. The use of transfers to the poor that highly regressive. Nonetheless, problems remain. There are are conditional on children participating in general school- still very large differences across programs with regard to ing and health systems is a good example (see below). their distributional impact. As noted above, the subsidized Finally, the historical pattern of social spending has been expansion of tertiary education enrollments has been dise- far from optimal in relation to the business cycle. Social qualizing. This also seems to be true for any service for spending per poor person is in fact highly procyclical almost which the initial access rate is low. Figure 8 shows the everywhere in Latin America, which means that resources 13 INEQUALITY IN LATIN AMERICA FIGURE 9 Targeting to the poor and reach among the poor for different ypes of programs in Brazil Share of the bottom 20% of the population who have access to the program per year 1.0 - Public health Primary education 0.8 - Water 0.6* School feeding 0.4 - Kindergarten - - Social assistance 0 * Sewage t ~~~~0.2 - Pensions Unemployment Secondary education LOAS/BPCa insurance Adult education Creche __ N 0.2 0.4 0.6 W 0.8 1.0W Share of program spending accruing to the bottom 20% of the population Note: Circle size is proportional to program budgets. a. Lei Organmca da Assistencaa Social/Benefico de Prestaczo Continuada Source: World Bank (2000e); see chapter 9. are scarce exactly when they are most needed. This is not a ownership is likely to take time. Yet Latin America's his- pattern unique to this region, and it is easy to understand torical evolution has led to the formation of what might be the political processes that may lead to this outcome. Yet described as "truncated" welfare states that provide benefits the fact remains that it has high costs for overall economic only to formal workers. Such states tend to be highly activity, and especially for the poor. Hence the importance unequal; for example, Mexican social security systems for of adopting fiscal rules and institutions that would help over- public and private workers result in some of the most come the procyclicality of fiscal policies in general, and of unequal programs in the country. propoor social expenditures in particular. Many countries have been reforming their welfare states to make them more efficient and less burdensome. The Reforming and extending the truncated welfare state concern here is to design instruments that make states more Transfers have an important role to play in all societies in inclusive, while simultaneously minimizing efficiency costs. helping households manage risks-for example, those This is likely to involve a range of policies, depending on related to health, unemployment, and the failure of crops- the area of vulnerability. For example, reaching poor senior and in redistributing resources to the poor. The case for redis- citizens is best achieved through basic pensions financed by tributive transfers is heightened by the fact that the general revenues, as illustrated by Brazil's pension system transition to more equal societies through broadening asset for farmers. Dealing with unemployment risks is likely to 14 SUMMARY involve a range of employment-related transfers-including provisioning of broad-based services. Democratization was unemployment insurance systems that already exist-and probably necessary for transformation, but has not been suf- workfare programs targeted to poor, informal sector work- ficient. Clientelistic relationships tend to continue under ers, such as Trabajar in Argentina (which was redesigned as democratic auspices, especially when associated with unequal Jefas y Jefes de Hogar during the recent economic crisis), A cultures across groups. This pattern is by no means unique Trabajar in Peru, and Manos a la Obra in Colombia. to Latin America: the political machines of major U.S. cities Particularly promising are cash transfer programs that are also persisted for a significant part of the 20'h century. targeted toward certain groups on the basis of need and Both international and Latin American experiences delivered only when recipients have taken a set of actions, underscore the potential for change. This is most likely to which usually consist of investments in their own human occur when there are both shifts in the opportunities for capital. Evidence underscores the capacity of some of these political influence (including via electoral competition) and programs (such as Oportunidades in Mexico, Bolsa Escola strengthening of the capacities of poorer groups to articu- in Brazil, and the Red de Protecci6n Social in Nicaragua) late their interests and organize for effective action. These to reach the poor in an efficient manner and to have posi- processes may take various forms. In the long run, frequently tive effects on their accumulation of human capital. In addi- maligned political parties have a crucial role to play, pro- tion to effecting redistributive transfers, such programs also vided they are formed along programmatic or policy lines have potential as risk management instruments, and could rather than on a vertical, patronage-oriented basis. There are become an effective part of a comprehensive and equitable also important complementary relationships between pro- social protection system. As mentioned above, coupling con- grammatic demands and the development of autonomous ditional cash transfers to the poor with overall expansions bureaucracies. Chile is the best example in the region of a of basic education and health services makes eminent sense case in which the combination of programmatic parties and both from the point of view of efficiency (since otherwise the bureaucratic competency helped support important shifts to poor may not demand these services) and the political sus- a more effective and redistributive state following the tran- tainability of targeted programs. sition to democracy-with the result that progressive con- sequences flowed from the improved provisioning of key Policy reforn must be underpinned by political public goods and both broad-based and targeted services. and social change Other national cases, including Brazil over the past decade, The economic policies described above provide a potentially appear to be moving along this path. potent set of instruments to make Latin American societies Where programmatic parties are absent, change is often substantially more equal, especially over the long run. But effected by outsiders, generally in the form of political entre- neither the design of policy nor its implementation is a preneurs or movements that can be catalysts for breaking purely technocratic affair. Analysis of history and current with history. However, such outsider-driven changes are only social and political conditions shows that redistributive and likely to be sustainable if they are embedded in formal insti- efficient economic policies are only likely to be chosen and tutions, especially those that create the capacity of lower and effectively implemented if underpinned by supportive middle groups to articulate and organize for an effective, political and social conditions. To varying degrees most Latin redistributive state. Important complements to-but not American societies do not have such conditions today. Public substitutes for-change from within or outside the pre- action needs to foster shifts to a new political equilibrium. vailing system include measures to increase transparency, promote public debate, and strengthen public administra- From cilentelistic and weak to effective tive capacities. These aspects both sustain societal pressure and redistributlve states and increase the capacity for developing and maintaining a History bequeathed Latin America with unequal social rela- responsive government. tions and weak states, in degrees that vary across the region. The subnational state has become an increasingly impor- This has been a fundamental source of the characteristic mix tant arena of action for political change and economic policy, of underprovisioning of the public goods of macro stability, owing to the combination of political decentralization (includ- property rights, and citizenship, as well as low levels of ing the introduction of elections for mayors) and devolution 15 INEQUALITY IN LATIN AMERICA of responsibilities. In this context, a number of cases of major with inclusive political and social structures, a process that changes in performance of the state have occurred in just a may first require a shift in the political equilibrium. That this few years, notably in Bogota, Colombia, and Porto Alegre, can happen at all is illustrated by developments both within Brazil. These cases show signs of a real shift toward a new and outside the region. The forging of alliances between poorer political equilibrium, with large, redistributive improvements and middle groups, and with progressive elements of elites, of services, reductions in patronage and corruption, and will be essential. A key consequence of such alliances will be increases in tax efforts. the equalizing of economic and political opportunity, includ- ing moving toward societies with more open elite structures. Confronting ethnic and racial divisions History suggests that this will be good for economic and History has also left some countries in Latin America with societal dynamism, and certainly for equity. major divides between ethnic and racial groups, in particu- lar indigenous and Afro-descended groups. (Although gender Notes differences are still present, tangible progress appears to 1. General statements about Latin America throughout this report have been made in this area.) While this report could not refer to continental Latin America, excluding Belize, French Guiana, deal with these issues in depth, tackling group-based divi- and Suriname. Unfortunately, the authors are unable to make general sions is clearly an important part of the overall agenda of statements regarding the Caribbean, since household surveys were reducing inequality in the region. As with the transition to available for only 3 of the 15 Caribbean countries: the Dominican democracy, most countries have made formal moves in this Republic, Jamaica, and Trinidad and Tobago. However, results are direction, for example by recognizing multiculturalism in reported for these three countries, and they are included in the regional national constitutions. averages presented. Nevertheless, there would be no statistical justi- The real issues now lie in changing the social dynamics fication for making inferences for the Caribbean subregion as a whole, that perpetuate such group-based inequalities. One area for on the basis of a primary data analysis for those three nations. action, noted above, is bilingual and multicultural educa- 2. Although higher volatility in income and consumption may tion. There are also rising demands in a range of other areas, overstate the degree of excess inequality that exists in Latin America including land rights and judicial and health practices. In (in comparison to other regions), the underreporting of capital income the recent past, these demands have been forcefully articu- in household surveys, particularly in top deciles, may tend to under- lated by indigenous movements, especially in the Andes, state it. For a detailed discussion of such measurement problems, see Mexico, and Central America. However, in many countries chapter 2. there is a related but distinct agenda around Afro-descended 3. There is, of course, wide variation within the region, with Chile groups, as evidenced by growing public recognition of the probably closest to such a model. issue in Brazil. In many cases movements representing 4. Unequal social relations can exacerbate these influences-as in indigenous and Afro-descended groups have an "outsider" the example of marital sorting, which can cause household inequal- quality. A challenge for the coming decade and beyond will ity to magnify individual inequality-and can also be sources of eco- be their inclusion within core political and social processes. nomic differences, for example by playing a key role in the processes that determine educational levels and market returns. Aillances, open-minded elites, and economic 5. Nonetheless, rising wage inequality did not necessarily trans- dynamism late into rising household income inequality, for many reasons. For Latin American societies have the economic policy instru- example, the increased participation of women in the labor marker at ments to make, over time, major changes in their embedded low wages can increase wage inequality without increasing household patterns of inequality. Sound choices and the effective execu- income inequality. tion of these policies will only occur when they are associated 16 CHAPTER 1 Introduction: Motivation and Conceptual Framework F O OR AS LONG AS DATA ON LIVING STANDARDS HAVE BEEN AVAILABLE, LATIN AMERICA HAS BEEN one of the regions of the world with the greatest inequality.' With the possible exception of Sub-Saharan Africa, this is true with regard to almost every conceivable indicator, from income and consumption expenditures to measures of political influence and voice, and including most aspects of health and education. Whereas the richest tenth of the population in the region earn 48 percent of total income, the poorest tenth earn only 1.6 percent. By contrast, in developed countries the top tenth receive 29 percent of total income, compared to 2.5 percent for the bottom tenth.2 Gini coefficients tell a similar story: whereas they averaged 0.52 in Latin America in the 1990s, the averages for the OECD, Eastern Europe, and Asia during the same period were much lower: 0.34, 0.33 and 0.41, respectively.3 This trend implies that very high ratios of income accrue to the wealthiest segments of the population relative to the poorest. In Guatemala, the ratio of the top to the bottom tenth of the population was 58.6 in 2000. In Panama, it declined from 71.6 in 1991 to 53.5 in 2000. Even the lowest 10/1 ratio in the region in 2000-15.8 in Uruguay-is higher than most figures found in Europe. (For example, the closest comparison is a ratio of 11.2 in Italy.) Such enormous differences in the incomes of citizens of the attainment are equally striking to those in income. In same country clearly imply correspondingly different degrees Mexico, the average person in the poorest fifth of the pop- of access to the goods and services that people consume in ulation has 3.5 years of schooling, as compared with 11.6 order to satisfy their needs and wants. However, disparities years for the average person in the richest fifth. These num- extend much beyond private consumption. Following the ter- bers are likely to underestimate actual educational differences minology of Amartya Sen, there are profound differences in because of marked differences in the quality of education. the freedom, or capability, of different individuals and groups In many countries, educational attainment also differs among to follow lives of their choosing-to do things that they have gender, ethnic, and racial groups. In most places, differences cause to value (Sen 1985a, 1992, 1999). Private resources and between men and women are becoming less marked with patterns of public provisioning affect such capabilities, while time and have even reversed for young cohorts, but severe social and political arrangements affect the capacity to par- disparities still exist among older people. In Bolivia, for ticipate meaningfully in society, influence decision-making, instance, average years of schooling for persons age 61 or or live without shame. older are 4.1 for men and 2.4 for women.4 With respect to education, even though public systems Health outcomes also vary dramatically along with exist in most countries in Latin America, the disparities of income distribution, resulting in enormous impacts on life Note: All tables and figures referred to in the text with the designation "A" can be found in the statistical appendix, p. 285. 17 INEQUALITY IN LATIN AMERICA opportunities and quality. In Brazil, children born to house- authors look at the levels of and trends in personal distrib- holds in the poorest fifth of the population are three times utions (of income or other indicators), as well as at group- as likely to die before they reach the age of five as are chil- based differences, whether by race, ethnicity, or gender. In dren born to households in the richest fifth. In Bolivia, this addition to private incomes, the authors consider goods and figure is more than four times as high among the poor, with services provided publicly and the taxes levied to finance children in the bottom fifth of the population experiencing their provision. under-five mortality rates of 146.5 per 1,000, which is as Second, the authors investigate the causes of Latin America's high as the South Asian average. extreme inequality by considering the historical roots and cur- In fact, it is not an exaggeration to say that every aspect rent processes that result in the reproduction of the prob- of life is affected by pervasive inequality. A Guatemalan lem. This takes us to questions of the social, cultural, and family whose income places it in the bottom fifth of the political sources of inequality that have interacted system- income distribution has on average three children, whereas atically with economic mechanisms. its counterpart in the top fifth has 1.9 children. In the former The third and final objective of this report is to consider household, 4.5 people live in each room, compared to 1.6 some of the options available to policymakers in the region for break- in the latter. The former household has a 57 percent chance ing with the long history of inequality that has character- of being connected to water mains and a 49 percent proba- ized the countries studied. In so doing, the authors suggest bility of having access to electricity. The corresponding prob- policies and policy directions that can help reform economies abilities for the latter household are 92 and 93 percent, and societies in such a way as to make them more equitable, respectively. without detracting from economic efficiency. In every conceivable way, the lives of these two families To prepare readers for the journey contained in the rest have very little in common. The very meaning of being a of this report, the remainder of this chapter is divided into citizen of a country is almost certainly substantively differ- two parts. The first poses the question "inequality of what?" ent for the families. A poor Guatemalan household has at and defines the realm of concern and conceptual framework. worst experienced violence and repression, and at best "low- The second section considers the question "why should we density" citizenship and the "unrule" of law in recent care?" decades.5 Poor Guatemalan families are predominantly indigenous and have experienced centuries of exploitation 1.1. A conceptual framework and exclusion, with weak influence over local and national The idea of inequality generally refers to a measure of dis- decision-making. The richest 20 percent of the population persion in a distribution.6 Most economic analysis is con- is more likely to be white and to have enjoyed at least some cerned with inequality in the distribution of some measure measure of normal citizenship. Although Guatemala is at the of individual well-being, with household income (or con- upper end of the inequality league in Latin America, and has sumption expenditure) per capita being the most commonly particularly unequal and weak social and political institu- used proxy. In light of the growing recognition that well- tions, the picture in most of the region is qualitatively sim- being has many dimensions besides income, inequalities are ilar. Worryingly, even in countries that were distinctly more discussed with regard to other variables, such as education, egalitarian by Latin American standards (such as Argentina, health, safety, and access to services.7 In addition, the report Uruguay, and Venezuela), the recent trend has been one of is not concerned only with economic welfare. Political power, growing inequality, at least in terms of income. or influence within a society, is also unequally distributed, In this report, the World Bank seeks to do three things. and these "inequalities of agency" are powerfully intertwined First, facts related to inequality will be presented, to the extent with economic inequality. allowed by the data. The authors build on an array of unit- One of the themes considered in this report is that dif- level household survey data sets in order to construct an ferences in voice, influence, and power are both driven by up-to-date picture of the distributions of income and other economic differences and are a key element in ensuring the indicators of living standards across 20 countries in Latin resilience and adaptability of such differences. While much America. For three-quarters of these countries, the pattern of the focus will be on differences across the whole distrib- of change during the 1990s is examined. In so doing, the ution, there is often particular interest in "horizontal" 18 INTRODUCTION MOTIVATION AND CONCEPTUAL FRAMEWORK inequalities in all these dimensions between groups, for outcome that reduces the number of people living below a example between races, indigenous and nonindigenous minimum income level (that is, below a socially accepted groups, and men and women (discussed in particular in income poverty line) and that guarantees everyone's access chapter 4). to, say, basic education and health services and social secu- In addition to looking at the distribution of outcomes rity of a minimum quality. Governments around the world (such as incomes, health indicators, or safety from crime), have already taken a strong stance on these issues by com- it is also possible to look at the distribution of assets and mitting to the Millennium Development Goals as a guide opportunities. Opportunities are crucial determinants of out- to their own actions and the actions of international orga- comes. In fact, a long-standing view among students of the- nizations. In many countries in the region, broad consulta- ories of social justice is that "equity" and "fairness" are more tions have led to strong endorsements of these goals by properly defined in terms of opportunities than outcomes, political parties, nongovernmental organizations, the private since the latter depend also on a range of variable human sector, and other social actors. characteristics, including age, gender, talent, physical abil- It is probably harder to construct a similar consensus ity, social background, and preferences. about valuing equality of outcomes across the entire distrib- Sen (1992) argues that it is of great importance to dis- ution. However, it may be easier to instead form broad coali- tinguish between "achievement" and "the freedom to tions in favor of the concept of equality of opportunities, even achieve" in assessing both the extent and normative signif- if such a concept means different things to different people icance of inequalities.' Among achievements, Sen emphasizes and is more difficult to operationalize (see below). In many the doings and beings (or "functionings") that constitute chapters in this report (notably chapter 5), the authors stress well-being rather than the means with which to achieve well- the need to build effective coalitions that bring together the being, such as incomes and resources. Such achievements can poor, the middle classes, and enlightened elites to support range from quite basic functionings, including "being well- policies and programs oriented toward reductions of nourished, avoiding escapable morbidity and premature mor- inequality and, in particular, of poverty. It might be easier tality, etc. to more sophisticated ones, such as having to achieve such coalitions if objectives are defined in terms self-respect, being able to participate in the life of the com- of two key outcomes, income poverty reduction and mini- munity and so on." The "capability set" represents the range mum access to basic services, as well as greater equality of of potential functionings that an individual can achieve, or opportunities for all. the "overall freedom a person enjoys to pursue her well- In this report, the authors take the position that the dis- being." tributions of both outcomes and opportunities matter and A core theme of this report is that the profound differ- therefore provide information on both aspects, leaving the ences in capabilities among individuals and groups in Latin reader to make his or her own social judgments. In addition America are based on interactions between economic assets, to the existence of different views on valuation, there is also economic opportunities, political forces, and sociocultural a practical reason for adopting this approach. While an processes. One aspect of this line of thought concerns the approach based on freedom to achieve, or capabilities, has moral assessment of which differences are fair. It is some- important conceptual and ethical attractions, it also poses sig- times argued that differences in outcomes that may be nificant measurement challenges. Sen (1992) argues that using morally offensive or "socially unjust" if they are caused by a capability-based approach to assess freedom to achieve has disparities in opportunities or life chances beyond the con- to rely heavily on measures of actual achievements. trol of the individual may actually be much less objection- The measurement challenge is illustrated by the applica- able if they are due instead to choices made in the level of tion of a specific approach to the question of opportunities. effort-for example, studying or working hard rather than This example follows Roemer (1998) in defining opportu- enjoying more leisure-or to differences in needs.9 nities as the set of circumstances that affect people's outcomes, In this context, it may be useful to draw a distinction in but that do not depend on their own efforts or decisions and the way societies judge outcomes between the poor and other are instead determined beyond their control. Even this appar- segments of the income distribution. It is probably possi- ently simple conceptual definition, however, turns out to be ble to gain widespread consensus about valuing strongly any rather difficult to operationalize. Box 1.1 presents the results 19 INEQUALITY IN LATIN AMERICA BOX I I Mesudng inequalft of opponle hI Bail The Peaquwja Nacional por Amostra de Domirilioj (PNAD) is race, parental schooling (mean and difference between par- Brazil's main household survey instrument. In 1996, it con- enrs), the father's occupation, region where the individual tained a set of supplemental questions on the parents was born, years of schooling (linear and squared), and a of respondents, including their level of educational dummy if the individual had migrated at some point in his achievement and occupation when the respondents were or her lifetime. These regressions are estimated separarely 15 years old. Using this data, Bourguignon. Ferreira, and fbr men and women. In both cases, the sample is restricted Mendndez (2003) seek to decompose total income inequal- to workers aged 26-60 living in urban areas of Brazil. ity into both a component based on inequaliry in observed The authors then classify the last two variables (edu- opportuniries and a residual. cation and migration) as efforts, on [he assumption thar Roemer (1998) defines opportunities as the set of cir- they were at least in part under the individual's own con- curnstances that lie beyond an individual's own control. The trol. The remaining variables are treated as circumstances. key quesiion is delineating what is and is not beyond an However, education and migration may also be influenced individual's control. To provide the basis for this determi- by circumstances (for example, parents affect the educa- nation, Bourgusgnon, Ferreira, and Men6ndez (2003) esri- tion of their children). To partly account for this, the mate an augmented Mfincer earnings equation, regressing authors regress education and migration on the vector ot an individual's laboreramings on several variables: a constant, circumsrances as well and use a Monte Carlo simulation Inequaliy of opportunities as a shame of total inequality, 1996 Theil-T *0.9 os| Residual inequality i Inequality of observed opportunities 0.803 0.805 0.8 075 0.779 0.767 0.753 0.2 0.44 0.511 0.489 0.499 0.750 0.753 .5 0.7 - 0.684 '0.53 0.6 - O- _ , l l l l l _ , an 2~~~~~~~~~~~02 0.5- 0.4- 0.3- 0.511 0490.499 0.489 0.459 0.455 02 0.44 0.1 b1936-40 b1941-45 b1946-50 b1951-55 b1956-60 b1961-65 b1966-70 Cohorts Source Bourgugnon. Ferrelra and Menendez 2003 20 INTRODUCTION: MOTIVATION AND CONCEPTUAL FRAMEWORK procedure Co correct for the attendant endogeneity bias. The authors test a variety of specificarions and assump- This effectively treats education and migration as being tions about the nature of the residual rerm z' They pre- partly due to circumstance and partly to effort. They end sent results f'or earnings and household incomes per capita up with a svstem that looks as follows: under various assumptions. The figure below presents the results of the preferred specification for the seven cohorts In u = Ca + El+tue EE3 -C +,most active in the Brazilian labor market in 1996. The I ' + height of rhe bar is overall observed income inequality in In this equation, u denotes earnings, C is a vector of household incomes per capita, as measured by the Theil- observed circumstance variables, E is a vector of observed T index. The bottom part is residual inequaliry after ef'ort variables, and v and z are unobserved dererminants observed circumstances are equalized. The upper part thus that might in principle be either circumstance or efforts. corresponds to one estimate of inequality of observed The method used to estimate inequality of observed oppor- opportunities in Brazil in 1996. tunities in this setup is to simulate the disrribution of pre- Even though a number of important circumstances dicted earnings when each and every circumstance variable (such as family wealth and contacts, cultural history. or is assumed ro be equal to a constant (for example, an aver- rhe quality of the school where a child is firsr placed) are age i) n both equations of the system. The resulting not explicitly controlled for, inequality of opportunities inequality is clearly not due to any variance in observed still appears to be high in Brazil. According to this mea- circumstances. It must instead be due to effort made and sure, ir accounts for 36-45 percent of total inequaliry for to any unobserved circumstances, such as family wealth, the older cohorts and 3 3-39 percent for the younger ones. which might be included in u and r. Inequaliry of observed There is a slight but perceptible downward trend in the opportuniries-denoted by 0-would, under the assump- share of inequalities accounted for by opportunities, iions, correspond to the difference between observed which seems closely connected to a reducrion in the income inequality and the simulated inequality when degree of intergenerational transmission of educational circumstances are equal for all. inequality. e = ICy) - A(VIC =*- of a 1996 study in Brazil, one attempt to empirically iden- expenditures; individual earnings; hourly wages; hours tify the share of inequality that is due to opportunities. The worked; access to various public services; years of schooling; study highlights both the methodological difficulties and the and land owned and farmed. potential insights inherent in focusing on the inequality of The authors do not attempt to classify, for example, edu- opportunities, which account for a considerable share of cational attainment as either an asset or an outcome because Brazil's high level of income inequality. it clearly is both things. Education, like health, affects worker More generally in this report, the authors note that oppor- productivity and remuneration, and is thus a human capi- tunities are highly correlated with the set of assets on which tal asset. But, as Sen (2000) and others have forcefully argued, people can draw, as well as on the set of markets they can health and education are also of value in their own right, both access and the institutions that surround them. For this as functionings and as determinants of capabilities. This is reason, the report presents information on the distributions equally true of power and influence: political rights are valu- of individual assets and access to services and markets, and able as such, and can thus be thought of as outcomes with discusses the operation of a range of formal and informal intrinsic value. However, they are equally important as institutions likely to affect people's opportunities. The capabilities that influence the set of opportunities available authors list inequality measures for a wide range of outcomes to agents. and assets, including, among others, distributions of house- Opportunities, assets, and outcomes are also related in hold per capita incomes; household per capita consumption causal ways. This report clearly cannot attempt to provide 21 INEQUALITY IN LATIN AMERICA viduals match to form households and on decisions related FIGURE I 1 to family size and composition. A simple conceptual framework Finally, secondary incomes are also affected by taxes and Assets & transfers, grouped together here under the category of public opportunities redistribution. There are obviously feedback loops in all these \1~ processes-notably from market processes-and this linear Markets Markets presentation should be treated as a way of organizing think- Household formation ing, and not as a unidirectional view of causation. This part / ' of the framework is obviously a highly stylized way of seeing Public redistribution economic influences on household per capita incomes. How- Governance institutions ever, it is also one that is useful for the authors' attempts to understand just where along the chain of assets, markets, Power -~-----Otoe Power <& inWi§ Outcomes households, and governments Latin American countries gen- erate so much more inequality than others. These economic processes will be further examined in chapter 6. an exhaustive treatment of the determinants of inequality. These economic processes do not take place in a vacuum. After all, as just stated, even "simple" income inequality At every step of the way, they are mediated by social and refers to dispersion in the distribution of all incomes. This political institutions that are very broadly understood to distribution is determined within the general equilibrium encompass the rules and norms of behavior in society, as well of complex modern economies in which market imperfec- as more specific institutions that formally or informally tions, incomplete information, strategic interactions, and regulate markets and affect governments."0 This approach political processes simultaneously occur. Going beyond static includes both the "macroinstitutional" arrangements in soci- economics, the institutions that mediate a number of these eties and the sociocultural processes that are products of economic processes have long historical roots and reflect cul- interactions between different groups in a society, notably tural patterns that are often deeply ingrained. those that occur between dominant and subordinate groups. Pretense to an exhaustive discussion of the causes of As sociologists such as Charles Tilly and Pierre Bourdieu inequality is thus a shortcut to failure. Instead, a very simple, have emphasized, the latter are essentially relational, and are general framework of the circular interaction between the deeply intertwined with both the organization of economic distributions of assets and opportunities, incomes and other production and the structure of power. Social arrangements outcomes, and power and influence in society is schemati- and sociocultural processes are the sources of large "inequal- cally presented in figure 1.1. The (joint) distribution of assets ities in agency" of different groups, or differences in their and opportunities is depicted at the top. These assets include capacity to shape and influence the conditions in which they not only physical and financial wealth such as land or stocks, live." These political and social forces are discussed in but also human capital endowments such as education. Indi- chapter 5. viduals may inherit some of these, but others are produced In turn, formal and informal institutions are neither during a person's lifetime. Education is of particular inter- immutable nor indifferent to economics. Income and est since it is, for most people, the main productive asset and wealth-or "economic power"-is closely linked to politi- is primarily acquired early in life. cal power, influence, and voice. This report represents these People then make choices about the use of their assets in latter aspects through the link between income and power the specific markets in which they are remunerated. Finan- and considers them to be mediated by political institutions cial savings generate interest or dividends in bonds or stock and sociocultural processes. In figure 1. 1, both the economic markets; land generates returns through rents or profits; and processes on the hypotenuse of the triangle and the social human capital is remunerated in labor markets. The com- and political processes on the other sides of the triangle bination of incomes arising from the remuneration of assets succinctly summarize broad traditions of economics, social in these different markets combines to form individual science, and political economy in an extremely reduced form. incomes. Primary household incomes depend on how indi- The concepts underlying this figure are considered in more 22 INTRODUCTION MOTIVATION AND CONCEPTUAL FRAMEWORK detail in chapters 5 and 6, while salient aspects of group- in a farmers' market to microenterprise credit or financial based differences (that is, those related to race, ethnicity, and derivatives-the enforcement of property rights and the need gender) are discussed in chapter 3. to correct for informational and other market failures also In addition, power can be used to shape economic reali- grow. Financial regulators, inspectors of working conditions ties in a number of different ways. Both historical and con- in labor markets, and antitrust investigators are examples temporary experiences vividly illustrate the linkages between of governance institutions through which political systems power, asset accumulation, market functioning, and fiscal provide a framework for the activities of markets, which action. Despite the magic of the "invisible hand," markets in turn shapes economic processes and outcomes. Of par- themselves never really operate independently of regulatory ticular importance to Latin America are the interactions institutions or the purposeful activity of different groups between policies and macroeconomic conditions. Some of the holding greater or lesser degrees of political and market policy mechanisms available to increase equity through mar- power-a theme that Adam Smith himself forcefully kets and other institutions are considered in chapter 8. emphasized."2 Third, governments inevitably affect the distribution of At the most basic level, the enforcement of law and order disposable household incomes through the need to finance is necessary to secure the property rights of those who sell, their operations. Taxes serve to raise the revenue that sup- as well as of those who buy. At a more sophisticated level, ports the regulatory roles discussed above, but may also a number of markets (in particular some financial and finance a number of redistributive services and transfers, "knowledge-intensive" markets) would be highly incom- whether in kind (such as through the provision of free edu- plete, or missing altogether, in the absence of regulatory and cation or health care) or in cash (such as in the form of unem- enforcement institutions that guarantee the rights of cred- ployment insurance or various cash subsidies). Chapter 9 itors, investors, or innovators. The use of power, or politi- considers policy options in this realm. cal agency, to effect change will therefore typically work By treating policies as emanating from a given social, cul- through specific policy actions. A significant part of the tural, and political context-which itself reflects a certain report is concerned with the ways in which policies can influ- distribution of power and influence and underlying distri- ence inequality. In terms of economic policies, this can work butions of income and wealth-the authors acknowledge through a number of the points summarized in the that policies are not designed in a vacuum. The usefulness hypotenuse of the triangle in figure 1.1. of this contribution to the policy debate will be enhanced First, policy can affect the distribution of assets, either by the recognition that the context of a society places impor- through direct redistribution (such as land reform or priva- tant constraints on what can be done and helps shape pref- tization processes), through taxation and subsidies (as in the erences about desirable outcomes. As figure 1.1 indicates, direct provision of education and health services) or through the policies arising from social and political institutions affect regulations that facilitate access to information, credit, and both the economic opportunities available to people and the insurance. The overall economic incentives with which market processes in which they engage. agents are faced are influenced by political forces through Such "circular causality" between wealth, income, and mechanisms ranging from the design of taxes and subsidies power mediated through institutions evolves throughout to the enforcement of property rights and the general invest- time and history. The position of a country at a given point ment climate. In turn, economic incentives shape differen- is powerfully shaped by the initial distributions of assets, as tial rates of accumulation, which then influence how asset well as by the past history of institutions that exist today. distributions evolve. Some policy options referring to asset A brief treatment of the historical roots of Latin American distributions are discussed in chapter 7. inequality is provided in chapter 4. Second, policy can influence how markets work. Even in While acknowledging the importance of history and of the most minimalist view of governments (that is, in the existing social and political institutions is crucial in order tradition of Hobbes, Locke, and Nozick), room was always to avoid policy-related mistakes, a fatalistic view of the world made for a "night-watchman" state. As commodities them- is both wrong and counterproductive. To propose policies selves and the markets in which they are traded become without an understanding of history and the specific con- increasingly complex-moving, for example, from apples texts in which they were developed often leads to failure. 23 INEQUALITY IN LATIN AMERICA However, this is not equivalent to the view that no policies economy's ability to respond effectively to macroeco- should be suggested at all because they will emerge endoge- nomic shocks. nously from historically predetermined institutions. Such a view fails to recognize the major role played by purposeful Each of these three issues is discussed in turn below. social and political action in achieving significant policy and institutional changes, and would ultimately result in fatal- Inequality is a bad thing in itself ism and inaction. Both in economics and in political philosophy, there is a A core theme of this report is that the circular causality common presumption that a unit increment in the income that underpins the resilience of inequality can be altered by of a poor person should be valued more highly than the same both economic forces and the social agency of different increment accruing to a richer person. This normative view groups (see, in particular, chapter 5). The authors believe that was initially associated with utilitarian thinkers dating back the possibility exists for domestic policymakers to break with to Jeremy Bentham, who attributed it to the process whereby Latin America's long history of persistently high inequality utilities increase with income or consumption, but at a by advancing equity-enhancing policies in the economic, decreasing rate. This property of "declining marginal utility" social, and political realms. They also believe that chances is consistent with a preference for equity, but it is not its of success are higher if policymakers are equipped with a sole source. deep historical understanding of the social and political Beyond this consideration, there is also a generalized norms and institutions that shape societies. agreement among economists around social welfare functions that value greater increases of utility (and hence of incomes) 1.2. The consequences of high inequality among the poor than among the rich.13 Other thinkers, such Before embarking on an entire volume on inequality, it may as John Rawls, have derived grounds for even stronger be worth pausing to ask a central question: "Why should we degrees of inequality aversion from different philosophical care?" The mission of the World Bank is to help countries and ethical sources. However it is ethically motivated, the eliminate poverty. However, poverty and inequality, although fact is that aversion to inequality or, conversely, a preference related, are very different phenomena. It is therefore legiti- for equity, has been a dominant (although not consensual) mate to ask why the World Bank should be concerned with view in political philosophy and the theories of social justice. inequality. There are three key reasons, namely: Recent evidence suggests that this view is not restricted to rarefied intellectual debate. Opinion surveys support the * The peoples and governments in World Bank client view that most Latin Americans are unhappy with the extent countries dislike inequality per se, both in terms of of inequality that exists in the region, which is consistent with outcomes and opportunities. what economists call "concave" social welfare functions. The * For a given level of mean income, greater inequality gen- results of a region-wide opinion survey conducted in 2001 erally means greater poverty. Perhaps worse, for a given by Latinobar6metro, a public opinion project based in Chile, rate of growth in mean incomes, greater inequality usu- indicate that on average, 89 percent of Latin Americans regard ally implies a slower rate of poverty reduction. the income distribution in their countries to be unfair or very * Evidence suggests that-in addition to reducing unfair. With the exception of Venezuela, no country had fewer poverty more slowly for each percentage point in the than 80 percent of respondents answering affirmatively in growth rate of the economy-high inequality of oppor- those categories. Figure 1.2 reports results for the 17 coun- tunities and outcomes reduces the rate of growth itself. tries for which they are available. In cases of extreme inequality and significant negative Such high levels of disapproval of income distribution are influences of inequality on growth, the combination of a cause for concern. They imply that the vast majority of the these two effects may imply that high inequality coun- population believes that the way the national income is tries find it difficult or even impossible to escape divided up is not just. Nor is it necessarily the case that such absolute poverty (see Ravallion 1997). There is also disapproval will go away with sustained growth. While evidence that inequality is associated with a greater relative differences in income often change little with prevalence of conflict and violence and may impair an growth, absolute income differences systematically increase. 24 INTRODUCTION MOTIVATION AND CONCEPTUAL FRAMEWORK FIGURE 1.2 Perceptions of fairness regarding the income distuibution in Latin America Average Argentina Bolivia Brazil Chile Colombia Costa Rica Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Pe ru Uruguay RB de Venezuela 0 20 40 60 80 100 Percent of responses E|C Very fair M Fair Cl Unfair M Very unfair| Source: Latinobar6metro, 2001. Responses to the question: "Do you think that the income distribution in your country is...?" There is evidence that many people think about inequality associated with high poverty levels. This is true in two in such absolute, rather then relative, terms.'4 different, although related, senses. Moreover, it would not be surprising if this belief also First, poverty will rise if the poverty line in a given soci- implied a general lack of ownership of or belief in the entire ety is below the mean of the income distribution, and if set of institutions that are responsible for income distribu- inequality increases in that distribution in a way that affects tion. Indeed, evidence from other opinion surveys indicates the poor without affecting the mean.'" Figure 1.3 illustrates that the level of trust for institutions in Latin America is this point: if society moves from the less unequal distribu- unusually low. The first reason the World Bank should care tion A to the more unequal distribution B, and both dis- about inequality is thus a rather simple one: the Bank's tributions have the same mean, then the incidence ofpoverty clients do. Inequality makes people unhappy and reduces will rise from area C to area C + D. their faith in national institutions. This argument is based on theory. It is consistent with empirical observations of historical experiences of the High inequality makes poverty reduction harder relationships between distributional and poverty changes. The second reason governments and international institu- On the basis of a sample of 114 episodes for 50 countries, tions may care about inequality is that high levels are Bourguignon (2002) confirms previous findings that higher 25 INEQUALITY IN LATIN AMERICA growth elasticity of poverty reduction) becomes smaller in FIGURE I 3 absolute value as initial inequality rises. Inequality and poverty with a constant mean neait aIn other words, more unequal countries convert a per- centage point of growth in mean household income into a smaller reduction in the incidence of poverty than do more .1 / \ Distribution A § egalitarian countries. Conversely, in order to obtain a 1 per- cent reduction in the number of people living in poverty, more unequal countries need to grow faster than do more egalitarian countries. Bourguignon (2002) analytically es- Distribution tablishes this negative impact of inequality on the rate of poverty reduction for the case of lognormal distributions, _c I j which are the most common functional form approximations for empirical income distributions. At the same time, he also finds strong empirical support (under various alternative model specifications) for the negative relationship within his rates of economic growth are unambiguously associated with sample. Drawing on the same data set, figure 1.4 illus- higher rates of poverty reduction."6 However, he also finds trates the basic result: after controlling for initial income that the impact of growth on poverty (that is, the negative levels, the growth elasticity of poverty reduction falls in FIGuRE I 4 Growth elasticities become less negatve with inequality Adjusted growth elasticity of poverty reduction 8 - Ei = -2.2 + 0.05Gini + e* p-value Gini: 0.09 6 - 4 - 2 - -2 * * *-. *** -4 * * *~~~~~ -6 -8- * - I I I I I 25 30 35 40 45 50 55 60 65 Gini coefficient Note: This is a country scatterplot in which adjusted elasticity (El) is the residual from a simple regression of the empirical elasticity on initial mean household survey income. Source: Bourguignon 2002, with sample truncated at empirical elasticity values of -8, +8. 26 INTRODUCTION MOTIVATION AND CONCEPTUAI FRAME'WORK absolute terms (that is, it becomes less negative, so that Inequality impairs the development process itself more growth is needed to reduce poverty) along with the The third and final reason development practitioners should Gini coefficient. The result is statistically significant at the care about inequality is that there is evidence that it may 10 percent level. have negative consequences for the overall process of Bourguignon (2002, p. 14) concludes that .. . income development, including slowing down the rate of economic redistribution plays essentially two roles in poverty reduc- growth. In fact, it likely has an effect on a number of other tion. A permanent redistribution of incomes reduces development goals and processes-such as the capacity to poverty instantaneously [through the effect described in resolve conflict without resorting to violence and the abil- figure 1.2]. But, in addition, it also contributes to a per- ity to manage aggregate shocks effectively. Effects on growth manent increase in the elasticity of poverty reduction with must therefore be considered along with other aspects con- respect to growth and therefore to an acceleration of poverty tained in the broader category of the impact of inequality reduction for a given rate of economic growth." (Emphasis in on the development process. the original; comment added.) The first aspect to consider is economic growth. After all, In a similar exercise, L6pez (2003) corroborates the as Robert Lucas famously wrote (Lucas 1988, p. 5), "Once finding that the absolute value of the theoretical growth one starts to think about [economic growth], it is hard to elasticity of poverty reduction (again under a lognormal think about anything else." Whereas during the 1960s, assumption) declines with inequality and with the ratio of 1970s, and 1980s, most of the thinking about economic the poverty line to mean income. These results are reported growth ignored distributional considerations, in the last in the first panel of table 1.1 below. The second panel of 10 years or so the distribution of income and wealth has the table reveals that the magnitude of the direct (that is, returned to center stage of the concerns related to growth instantaneous) effect of inequality on poverty also falls (Atkinson 1997). While there is as yet no consensus through- with initial inequality. In other words, the effect depicted out the economics profession on this question, it is proba- in figure 1.2 of an inequality reduction on poverty is concave. bly fair to say that the balance of academic opinion leans toward the view that high levels of inequality in incomes or in assets are causally related to lower rates of growth in TABLE 1.1 mean incomes. Theoretical elasticities of poverty with respect to aggregate There are two main conceptual reasons why this might income growth be the case. First, if credit or insurance markets are imper- fect, people may depend on their initial wealth to undertake Growth important investment decisions. If this indeed happens, a Gini situation may in turn arise in which the poorest people in PL' 0.3 0.4 0.5 0.6 society are unable to invest in socially efficient (that is, prof- itable) projects, while richer individuals receive lower returns 0.33 -3.9 -2.1 -1.3 -0.8 on the marginal dollar of their wealth. Credit market imper- 0.50 -2.8 -1.6 -1.0 -0.7 0.67 -2.0 -1.2 -0.8 -0.5 fections refer precisely to the reasons why intermediation 1.00 -1.2 -0.8 -0.5 -0.4 might fail, preventing the rich from lending to the poor and Inequality enabling both groups to gain from higher returns. Econo- Gini mists refer to such a situation-in which both parties might be made better off without anyone losing out-as a Pareto PLI 0.3 0.4 0.5 0.6 inefficient allocation. 0.33 5.2 3.3 2.4 2.0 Why would credit markets fail in this way? Largely 0.50 2.5 1.7 1.3 1.2 because lenders do not have all the information they might 0067 0.2 0.2 0.3 0.4 like to have about borrowers and because contracts may not be fully enforceable. Because enforcing credit contracts may aPoverty line as share of per capita GDR be costly, interest rates may be higher for borrowers than for Source: L6pez 2003. lenders. This difference may mean that poorer agents borrow 27 INEQUALITY IN LATIN AMERICA too little. If some projects (such as a child's education) require effective public action. The private costs of education, for minimum "lumpy" investments, the result may be aggre- instance, can be considerably reduced (though never elim- gate underinvestment and an inefficient overall equilibrium inated) through a variety of government actions, such as the (Galor and Zeira 1993). provision of free public schooling or of good roads and Credit market imperfections may also manifest themselves public transportation by which students can reach schools. through collateral requirements. If loan amounts increase However, unequal societies in which political power is inter- with the collateral available to an agent, the poor will again twined with wealth may be less likely to choose policies that be at a disadvantage. Some people may not be able to start reduce those inefficiencies than to allocate scarce resources projects with positive expected social returns-including to alternative uses. These can include private consumption investments in human and physical capital, land, and of the rich (through lower taxes) or public spending on housing, -and the result will be inefficiency. Indeed, if there alternative programs that do not reach the poor (Benabou are "too many" such people, they may drive down the aggre- 2000). gate wage rate, with additional implications for the long- One study, inspired by a Latin American brand of polit- term steady-state of the economy (Banerjee and Newman, ical economy, characterizes the links between wealth 1993). A third variant on this story is that those agents who inequality, income inequality, and political inequality as fol- need to borrow the most (that is, the poorest) are faced with lows (Ferreira 2001). Wealth inequality means that some such a large "debt overhang" that they have very little incen- children attend good private schools, while others have no tive to work on their projects. Lenders ration excessively large choice but to enroll in lower-quality public schools. Those credit burdens because they may have a "moral hazard" effect who attend public schools enter the labor market with lower on effort supply (Aghion and Bolton 1997). Similar argu- levels of human capital and thus command lower incomes. ments explain how the poor may be rationed out of insur- If decisions about public spending are influenced by ance and other financial markets. incomes, the inequality in incomes that arises from large edu- In sum, economists have modeled a variety of ways in cational inequalities in turn leads to political outcomes in which the combination of imperfect financial markets and which a richer, better educated minority blocks the votes for unequal wealth distributions lead to outcomes which are not more funding for public schools. efficient in the "first best" sense. The specifics of financial The authors shows that there are multiple equilibria at markets may vary, but the essential issue is that inequality play in this system, with societies where initial wealth generates allocations through which the poor do not have inequality is lower ending up being more egalitarian than the means to undertake projects that, if undertaken, would those with higher initial levels of inequality. Furthermore, have high expected social returns. At the same time, the rich a change in political regime leading to less influence of are left receiving lower marginal returns on their wealth. Fur- income on political decisions (that is, a process of real democ- ther, income and wealth inequality, coupled with imperfect ratization such as that discussed in chapter 5) could lead to financial markets, will also limit the capacity of the poor to equilibrium-switching in this model. The result would be acquire assets such as human capital, land, and housing, a shift from a high-inequality, lower-welfare equilibrium to which will in turn limit their future opportunities and the one that is dominant both in inequality and welfare terms. possibility of smoothing consumption in the presence of large The mechanisms whereby elites act in ways that do not shocks. Such indirect effects will also reduce overall growth serve the interests of all segments of the population may be and welfare. more pervasive than can be captured in these economic The second conceptual reason that inequality may lead models. A strand of thought in the political science and soci- to lower growth involves questions of political economy. In ological literature that is relevant to parts of this report is societies with high degrees of concentration of power and that elites help form and perpetuate institutional structures wealth, elites may have more leeway to choose strategies that that are characterized by governance, limited accountabil- benefit themselves rather than middle- and lower-income ity, and high levels of corruption. Such "weak" institutional groups. Within the economics literature, this has often been arrangements, rather than being symptomatic of immature developed in terms of the links between the credit market institutional evolution, may be functional for the rich (Heller failures (discussed above) that might be remedied through and Mahoney 2003). In addition, where institutions are 28 INTRODUCTION MOTIVATION AND CONCEPTUAL FRAMEWORK weak, it is the poor who suffer most, since the rich can use work by Ravallion and Chen (1997) and Dollar and Kraay their political power and financial influence to further their (2002a)-but rather statistically significant evidence that interests (Glaeser, Scheinkman, and Schleifer 2002). Such initial inequality reduces growth. processes would have powerful adverse impacts on overall In addition, it seems that the harmful effects of high growth and other aspects of the development process. inequality on economic development are not restricted to Do the data support the hypothesis that high inequality economic efficiency and growth, important though these is bad for growth and efficiency? Most studies do find a neg- aspects are. A broader view of development includes more ative coefficient of initial inequality when this aspect is than output per capita, considering, for instance, a country's included as an explanatory variable in empirical growth institutional ability to cope with economic and other aggre- models. In regressing the average growth rate during gate shocks, such as a change in terms of trade or a key for- 1960-1985 on Gini coefficients for income and land around eign interest rate. Rodrik (1999) suggests that countries 1960, Alesina and Rodrik (1994) find statistically significant suffering from more pervasive social divisions-whether coefficients for both. Persson and Tabellini (1994) employ an ethnic and racial in nature or income- and class-based-seem alternative measure: using the share of income accruing to not to adjust to large shocks as well as other, more egali- the middle fifth of the income distribution as a proxy for tarian and cohesive societies. The implicit mechanism is that equality. The coefficient for that variable is statistically sig- the institutions responsible for sharing the burdens of adjust- nificant and positive, which is consistent with the Alesina and ment work less well in economies in which distributions are Rodrik results. Although Perotti (1996) uses a larger set of more unequal. countries and tests additional specifications, the results remain In order to understand the process of adjustment to the significant. first oil-price shock of 1973, Rodrik (1999) regresses the dif- More recently, this budding consensus has been chal- ference between the per capita growth rate during lenged by Forbes (2000) based on analysis of a new data set 1975-1989 and the same rate during 1960-1975. He finds known as the Deininger and Squire "high-quality" data a significant negative coefficient on the dummy variable for set."7 Forbes found a positive relationship between lagged Latin America, suggesting that the region suffered more from inequality and growth. There are, however, a number of rea- the shock (in terms of reduced ex-post growth) than did sons why these findings, although important, remain a developed or East Asian countries. Interestingly, this effect minority view in the economics profession. First, as the disappears (that is, the coefficient becomes insignificant) author herself points out, this study differs from the others when three specific aspects are controlled for: income in considering the effect of time-varying lagged inequality inequality in the 1970s; land inequality; and the murder rate rather than some fixed, initial level of inequality. The inter- (the latter serves as a general proxy for the prevalence of pretation thus has a more short-run effect. Second, ques- violence or weakness in conflict resolution in a society).18 tions have been raised both about the data set and the This result is interpreted as indicating that the prevalence econometric technique employed (see Aghion, Caroli, and of high inequality in Latin America (when compared to, for Garcia-Pefialosa 1999). example, East Asia) was partly to blame for the region's poorer Finally, other studies have continued to find significant performance in adjusting to the severe terms-of-trade shocks negative coefficients for measures of wealth and asset inequal- of the 1970s. Institutions were not able to achieve closure ity in growth regressions. For example, Birdsall and Londofio on the incidence of the costs of adjustments. In the mean- (1997) use a subset of the Deninger-Squire data set and con- time, fiscal and current account deficits grew, planting the clude (p. 35) that "Initial inequalities in the distribution of seeds for the ensuing debt crises of the 1980s. Chapter 8 land and human capital have a clear negative effect on eco- discusses the continuing importance of crisis management nomic growth, and the effects are almost twice as great for and the links with inequality in Latin America. the poor as for the population as a whole." L6pez (2003) uses The link between high inequality and weak institutions, a similar econometric approach as Forbes and the Deininger- which may appear tenuous at this stage and which is treated Squire data. Taking into account the simultaneous nature of in a much reduced form in Rodrik (1999), is further explored the determination of growth and inequality dynamics, he finds in chapters 4 and 5. In those chapters, the theme emerges no sign of an effect of growth on inequality-in keeping with that Latin America is, in a range of areas, characterized by 29 INEQUALITY IN LATIN AMERICA institutions that are weak and controlled by unequal power These are the main reasons why a better understanding of structures. Both these aspects tend to lead to sub-optimal inequality is important. Most people and governments in the levels of development and to perpetuate unequal gains from region regard Latin American levels of inequality as too high. development that does occur. They would like to reduce them in terms of both outcomes Returning for a moment to the significance of the murder and opportunities. In addition, such a reduction would likely rate variable in Rodrik's specifications, it is important to note help policymakers become more effective in fighting poverty that the fact that it mimicked the effect of the omitted and promoting broader economic development. inequality variables in the growth regression is not a coin- cidence. Evidence across countries suggests that high An overview of the report inequality is significantly and positively associated with the The remainder of the report is divided into three parts, one prevalence of violence.'9 Fajnzylber, Lederman, and Loayza for each objective outlined above. Part I establishes basic (2000) regress homicide rates (from the World Health facts about inequality in Latin America, to the extent allowed Organization database) on income inequality (measured by by the available data, in terms of both evidence on differ- the Gini coefficient) and a number of control variables in a ences across individuals and households (chapter 2) and panel data set of about 40 countries and 140-190 observa- between groups, across ethnic, racial, and gender lines tions (depending on the specification). Results show that (chapter 3). Part II turns to the causes and determinants of income inequality is always positive and strongly significant, high and persistent inequality in Latin America, from its and remains so even when educational inequality and income historical origins (chapter 4), contemporary social and polit- polarization are added as controls. Both social conflict and ical influences (chapter 5), and economic forces for the repro- personal violence are associated with higher levels of inequal- duction of inequality (chapter 6). Part III considers what ity, especially when institutions of conflict management and can be done in terms of specific policies in the domains of the rule of law are weak. assets (chapter 7), market institutions (chapter 8), and the Finally, it is worth noting that interaction occurs redistributive role of the state through taxation, services, and between possible effects of inequality on the growth rate and transfers (chapter 9). the adverse effects of inequality on poverty reduction. If inequality is sufficiently high, countries that would have Notes good growth prospects with strong poverty-reducing poten- 1. General statements about Latin America throughout this report tial at low levels of inequality will experience little or no refer to continental Latin America, excluding Belize, French Guiana, growth or poverty reduction. Ravallion (1997) calculates that and Suriname. Unfortunately, the authors are unable to make general about one-fifth of observations in a data set for developing statements regarding the Caribbean, since household surveys are avail- countries could fall into a category in which there would be able for only three Caribbean countries: the Dominican Republic, no absolute poverty reduction. Jamaica, and Trinidad and Tobago. Nonetheless, results for these three The balance of the evidence is such that high levels of countries are reported and included in the regional averages presented. income and wealth inequality observed in Latin America However, there would be no statistical justification for making infer- have the following features. They: ences for the Caribbean subregion as a whole on the basis of the pri- mary data analysis for those three countries. * Are considered unfair by large majorities of the 2. This refers to unweighted averages for the distribution of house- continent's population. hold per capita income in 1992, with estimates based on Bourguignon * Slow down the pace of poverty reduction in the region and Morrison 2002. See table A. 18 in the statistical appendix. by lowering the growth elasticity of poverty reduction. 3. The Gini coefficient is a standard measure of inequality in a dis- * Slow down economic growth and development itself. tribution. It ranges from zero to one, and increases with inequality. * Possibly hamper the region's ability to manage eco- A value of zero corresponds to perfect equality and a value of one cor- nomic volatility and worsen the quality of its macro- responds to a distribution in which a single unit receives all of the economic responses to shocks, which are unfortunately income and the other units receive nothing. all too frequent. 4. These figures refer to the national 1999 survey. See table A.22 * Make violent crime more pervasive. in the statistical appendix. 30 INTRODUCTION MOTIVATION AND CONCEPTUAL FRAMEWORK 5. See O'Donnell 1999b for discussion of low-density citizenship abuse of unequal influence and the role of the purposive agency of and Mendez and others 1999 for the unequal reach of the rule of law different groups in shaping economic and social outcomes. in much of Latin America. The most recent and important symbolic 13. In technical jargon, this refers to the "concavity" of social wel- change in Guatemala was the signing of the peace accords in 1996. fare functions. However, major institutional weaknesses and patterns of social exclu- 14. In experiments on attitudes toward inequality, Amiel and sion remain; see World Bank 2002b. Cowell (1997) found that 40 percent of participants thought of inequal- 6. Most measures frequently used in the economics literature eval- ity in terms of absolute, rather than relative, differences. uate dispersion in ways that are consistent with certain desirable attrib- 15. An increase in overall inequality could occur solely through utes, known as axioms of inequality measurement. For a discussion of increases in differences above the poverty line, for example shifts from measurement issues, see Cowell 1995 or 2000. the moderately rich to the very rich. This type of increase would not 7. This report does not focus on full multivariate welfare analy- affect poverty, although it would be judged as a worsening in social sis. For more on this subject, see Bourguignon and Chakravarty 2003. welfare under most welfare functions. 8. Quotes in this paragraph are from Sen 1992, pp. 4-5 and p.150. 16. That result was established for the same data set by Ravallion 9. The literature on the theories of social justice is vast. See Sen and Chen 1997. 2000 for a recent survey. 17. Forbes also used a different econometric technique, applying 10. SeeNorthl990foraclassictreatmentofinstitutionsasnorms the Arellano and Bond (1991) Generalized Method of Moments and "rules of the game." (GMM). 11. See Heller and Mahoney 2003 for a review, Tilly 1999 and 18. There are, of course, other factors that influence the effects of Bourdieu 1990 for key examples, and Rao and Walton 2004 for a dis- shocks on countries, such as the overall degree of openness, which tends cussion of "inequality of agency" from a cultural perspective. A related to be associated with lower long-term impacts of adverse external strand of the economics literature on measurement of inequality con- shocks. cerns the development of the theory and measurement of polarization. 19. However, there is some evidence that the Caribbean is an excep- This starts from the conceptual insight that individuals' evaluation of tion to this pattern. Inequality levels are often lower in that subre- the salience of inequalities is linked to differences between the groups gion than they are on the continent, although crime rates can be rather with which they identify relative to the groups from which they feel high in particular countries due largely to the drug trade. alienated(see Esteban and Ray 1994). 12. See Rothschild 2001 for a rich account of how Adam Smith and other enlightenment thinkers were greatly concerned with the 31 PART I The Nature of Inequality in Latin America i I CHAPTER 2 Different Lives: Inequality in Latin America T HIS CHAPTER SUMMARIZES INFORMATION FOR LATIN AMERICA ON MANY DIMENSIONS OF inequality, a topic that by itself deserves a large book. The chapter highlights some of the main features of inequality in the region in order to help readers understand the level and structure of and trends related to this phenomenon and to set the stage for the following chapters of this report, in which determinants of inequality and redistributive public policies are analyzed. a The empirical counterpart of the concept of inequality is far from trivial. The most important issue is to identify the variables for which the measurement of inequality is informative with regard to social unfairness. This process involves both theoretical and empirical problems, which are discussed briefly in section 2.1. Most of the statistics in this chapter are drawn from a sample of household surveys for 20 countries at three points during the period 1989-2001. These statistics and other results of the current study are periodically updated with new information on the Web site . Section 2.2 introduces the sample of household surveys, presents a large set of inequality measures for the distribution of household income (adjusted for demographics), and reports results for other dimensions of the income distribution (that is, aggregate welfare, poverty, and polarization). Section 2.3 is devoted to identifying problems generated by the measurement errors typically encountered in household surveys and to assessing the impact of such errors on the ability to measure, and hence understand, inequality. Section 2.4 places the results of section 2.2 in an inter- attendance and educational mobility, although statistics on national perspective by comparing inequality in Latin inequality in health, political representation, crime victim- America to other regions in the world. This analysis draws ization, and some basic social services (for example the pro- on recent studies that have pulled together large data sets vision of clean water) are also presented and discussed. with inequality information from many countries. Finally, section 2.7 provides concluding remarks. Household per capita income is derived from three main components: remuneration of assets (including labor and 2.1. Some conceptual issues human capital), financial transfers, and household demo- As discussed in chapter 1, it is safe to state that most people graphics. Section 2.5 considers inequality statistics on these have preferences for social fairness and associate the concept variables that have been computed from the sample of house- of unfairness to some sort of inequality. Differences arise hold surveys and drawn from recent studies by other authors. at the stage of defining the variable(s) considered most impor- Section 2.6 goes beyond the distribution of income to tant to equalize among individuals in order to attain a fairer present information on the distribution of some goods and society. A first choice to be made is between outcomes and services with which people are especially concerned. Most opportunities. Should we try to reduce disparities in outcomes of the section focuses on evidence of inequality in school (for example, income or consumption levels), or instead 35 INEQUALITY IN LATIN AMERICA guarantee equality of opportunities to achieve those outcomes? other variables, it largely follows that tradition. The implicit Many authors have argued in favor of the second alternative.' assumption is that current household income, as measured According to this view, inequality should not be a societal con- in household surveys, is highly correlated to individual oppor- cern if it arises between people subject to the same constraints; tunities and intertemporal living standards. Although these is the consequence of individual choices regarding the effort correlations are surely positive and probably high, it is dif- invested in improving outcomes; or is related to other vari- ficult to know how distorted the picture drawn from house- ables for which people should be accountable. Unfortunately, hold survey income data is in comparison to reality. the concept of opportunity is difficult to define and measure, It is important to be aware that, in focusing on household and hence in practice it is usually abandoned in favor of the income inequality, it is possible to implicitly consider as unfair analysis of inequality in outcome variables (see box 1.1 in chap- situations that are not (for example, two people may have dif- ter 1 for an example of an attempt to measure it.) ferent incomes due only to the different levels of individual Probably the most relevant outcome variable to compare effort or because they have different income profiles associ- among individuals is the intertemporal living standard, that ated with the same mean income in intertemporal terms). It is, the "average" well-being of people over their entire life- is also possible to assess as fair situations that are not (for exam- times. However, conceptual and, in particular, data-related ple, an unskilled prime-age worker who will remain relatively limitations restrict the comparison of living standards to poor the rest of her life and a college student who will be rich time periods much shorter than a lifetime. As a result, sur- in the future may currently have the same income). veys usually capture dimensions of well-being for periods no In addition to the analysis of household income or con- longer than one year. sumption inequality, the literature on measuring inequal- Within the group of variables usually measured in a house- ity increasingly reflects concern over other variables (for hold survey, consumption is generally considered to be the example, schooling, health status and services, and politi- preferred overall measure with which to approximate living cal representation). Two arguments are behind this concern. standards.2 It has three main advantages over its main com- The first states that individual well-being depends on fac- petitor, household income. First, if people can borrow and tors beyond consumption of goods and services, including lend (as can most people at least in small amounts, for short health status, security from crime and violence, degree of periods of time, and in informal markets), consumption is freedom, and respect for human rights. One possible ana- more closely associated with individual well-being than is lytical strategy is to try to value such factors in monetary income.3 Second, underreporting is usually a less severe prob- terms, add them to income or consumption, and measure lem in relation to consumption than income. Third, most inequality only in the aggregate. A less ambitious approach surveys report gross income rather than after-tax income, even is to measure inequality in different variables without any though the latter is a more relevant indicator of welfare and attempt to aggregate the results. usually more reflective of consumption levels. A second argument states that societies have a particu- Consumption can be estimated from household surveys lar concern with the distribution of certain variables, such conducted in many countries. In particular, the Living as consumption of basic education and health services. Even Standards Measurement Surveys (LSMS) project of the World if these aspects were included in the computation of total Bank makes use of questionnaires designed to measure consumption, there would be normative arguments for consumption, or at least expenditures. Unfortunately, con- assessing inequality separately for each variable. To inform sumption surveys are not common in Latin America. The consideration of these arguments, section 2.6 presents great majority of countries in the region conduct surveys statistics on the distribution of school enrollment, health without questions related to consumption or expenditure. status and services, political representation, safety, and In our sample of 20 countries, in only five were a minimum coverage of some basic social services. of two expenditure surveys conducted in the last decade. There is great interest in measuring inequality among For this reason, the measurement of social unfairness individuals. However, individuals usually live in households in Latin America has been mainly associated with the and share a common budget. This fact implies that an indi- measurement of inequality in the distribution of household vidual's well-being depends on the resources available in, income. Although the current study includes information on the size and structure of, and norms of sharing within the 36 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA household. The most commonly used indicator of individ- surveyed in 20 countries: Argentina, Bolivia, Brazil, Chile, ual well-being is household per capita income, that is, total Colombia, Costa Rica, the Dominican Republic, Ecuador, El household income divided by the number of persons in the Salvador, Guatemala, Honduras, Jamaica, Mexico, Nicaragua, household. Although widely used, this variable ignores three Panama, Paraguay, Peru, Trinidad and Tobago, Uruguay, and relevant factors: (1) consumption-related economies of scale Republica Bolivariana de Venezuela. The sample is fully within the household, which might, for example, allow a representative of most of Latin America but only partially of couple to live with less than double the budget of a person the Caribbean countries, which do not regularly conduct or living alone; (2) differences in needs among individuals that publish household surveys.5 are a function of age and gender (such differences are behind For most countries, the sample contains three observations the adjustments often made for adult equivalents); and corresponding to the early 1990s, mid-1990s, and the late (3) unequal allocation of resources within the household.4 1990s, 2000, or 2001. In each period, the sample represents Following the tradition in Latin America, this chapter more than 92 percent of the region's total population. All considers inequality measures for the distribution of house- household surveys included in the sample are nationally rep- hold per capita income among individuals, although statis- resentative. The exceptions are Argentina and Uruguay, where tics that take points (1) and (2) above into consideration are surveys cover only urban residents but nonetheless represent also computed. There is also considerable interest in more than 85 percent of the total population in both coun- inequalities between groups or categories of individuals, for tries.6 All surveys record a basic set of demographic, educa- example, inequalities drawn along ethnic, racial, and tion, labor, and income variables at the household and gender lines. This subject is presented briefly in the following individual levels. Although differences exist among countries, pages, but is treated in more depth in chapter 3. the surveys are roughly comparable in terms of the type of Summing up, the ideal objective would be to measure the questionnaire and sampling techniques used. degree of social unfairness in Latin American countries. Due Table A. 1 presents the main characteristics of each house- to conceptual and data-related limitations, we have measured hold survey. The table shows the names of the surveys, their inequality in the distribution of household income, adjusted coverage (urban or national), and the sample size (in number for demographics, and complemented these statistics with of individuals). For reference, the population estimates of indicators of inequality in the distribution of other dimen- each country are presented in column (v). Household income sions of well-being. Although the limitations of this is reported in all surveys. Those surveys that also cover expen- approach are evident, the authors believe that the statistics ditures are indicated in column (vi). All surveys have specific provided in this chapter are useful to help characterize and questions related to labor income and nearly all also cover understand social unfairness in the region. nonlabor income (that is, capital income, property income, profits, and transfers), although they differ in terms of the 2.2. Income inequality and beyond detail of the questions and the possibility of separating out Despite the many caveats associated with household surveys, different sources of nonlabor income. Surveys that include they remain the most reliable and appropriate source for dis- questions related to nonmonetary income and the implicit tributional analysis. A data set of household surveys for most rent of own housing are also marked in the table. Latin American countries since 1989 was prepared for this Most surveys were obtained through the Program for the report. The sample is introduced first, followed by some basic Improvement of Surveys and the Measurement of Living inequality statistics and discussion of the limitations of the Conditions in Latin America and the Caribbean (MECOVI), data. A set of tables provided in the statistical appendix from a joint effort of the Inter-American Development Bank the household surveys and other sources forms the basis for (IDB), the World Bank, and the United Nations Economic the following discussion. Commission for Latin America and the Caribbean (ECLAC). This program promotes improvements in the collection, The data organization, and analysis of household surveys in the region. The authors were able to assemble a data set containing Other surveys used in this chapter are part of the LSMS 52 household surveys covering the period 1989-2001. The program; these usually yield richer data because they include sample comprises approximately 3.6 million individuals questions on social services and expenditures. 37 INEQUALITY IN LATIN AMERICA This study is not the first one to analyze inequality in Latin household per capita disposable income among all individ- America based on a set of household surveys. Altimir (1994) uals in the population. Population weights are used in the and Morley (2001) at ECLAC, Wodon and others (2000) at calculations and missing and zero-income observations are the World Bank, and, in particular, Szekely and his co-authors discarded. Following the practice of national statistical (2000 and 2001) at the IDB, have gathered information from offices, the authors use a broad definition of the household household surveys to analyze income distribution in the but exclude servants and renters and their families. Both region.7 Compared to these studies, the current sample covers monetary and nonmonetary incomes are considered when more countries, provides more information on some countries that information is available. Although most income (mainly Argentina), and includes surveys for 2000 and 2001. sources are included (that is, labor, capital, profits, property This study also presents a larger set of statistics across coun- rents, and transfers), some potentially relevant items are tries and over time that are related to several dimensions of ignored (for example, the implicit rent from own-housing, inequality (that is, in addition to household income) and to in-kind gifts, and government in-kind transfers). Estimates inequality in the distribution of relevant demographic and of some of these variables are available in only a few surveys socioeconomic variables. and are of dubious quality. The authors made all possible efforts to make statistics The relevant concept for welfare analysis is net income comparable across countries and over time by using similar rather than gross income. In household surveys in the region, definitions of variables for each country and year and by some income sources are generally reported after labor and applying consistent methods of processing data. However, income taxes (for example, earnings of salaried workers), while perfect comparability is not assured, since the coverage of some others are typically not (for example, earnings of non- and questionnaires used in household surveys differ among salaried professionals or capital income). In addition, cash countries and frequently also within countries over time.8 transfers are reported in surveys but the value of government Three ways of alleviating comparability problems are adhered in-kind transfers (such as the provision of education and to throughout the chapter. First, when major changes in health services) is ignored. This differential treatment calls methodology or coverage occur, ways of assessing the impact for a detailed analysis of the distributional incidence of taxes on inequality statistics are provided. For instance, in Bolivia and public spending. Unfortunately, this analysis is rarely the household survey was urban in 1992 and nationally rep- done on a regular basis since it poses numerous theoretical resentative in 1996. Two sets of statistics are therefore pre- challenges and demands information not typically included sented for Bolivia in 1996: one for the whole sample and one in household surveys. This chapter follows the usual practice for those urban areas also surveyed in 1992. of computing statistics over the distribution of income Second, in addition to presenting statistics for more gen- reported in the surveys. (See chapter 9 for a survey of stud- eral variables (for example, household income from all ies of the incidence of taxes and government spending.) sources), more specific variables with fewer problems of com- Table A.2 shows one of the most tangible measures of parability (for example, wages from primary jobs for male inequality: the income shares of different income strata.9 prime-age workers) are also considered. Third, the tables People are sorted according to their household per capita document the particularities of each survey that may blur income and divided into ten groups of equal size (called comparisons with other countries and years. (Most of this deciles). In all the Latin American countries, the share of total information is available on the Web site of this study. Read- income held by the poorest 10 percent of the population has ers interested in technical details are advised to visit the site.) been always less than 2 percent, while the share of the richest Although every effort has been made to clean the data and 10 percent has been always higher than 30 percent. The first present consistent statistics, the reader interested in a spe- panel in figure 2.1 shows the income shares by decile in the cific country is advised to consult country-specific literature. three largest economies of the region: Argentina, Brazil, and Mexico. In these three countries, income shares slowly Income Inequality in the 1990s increase at a rate of less than two percentage points between This section uses the data set to study income inequality consecutive deciles along most of the distribution. Differ- across countries and over time. It first takes a look at the most ences between deciles are greater in the upper quarter of the analyzed distribution in Latin America: distribution of distribution, especially between deciles 9 and 10. This 38 DIFFERENT LIVES: INEQUALITY IN LATIN AMERICA FIGURE 2.1 Income shares by deciles (household per capita income) a. Argentina, Mexico, Brazil b. Costa Rica, El Salvador, Nicaragua Income shares Income shares 50 50 ED Argentina m Mexico 1 Brazil | Costa Rica M El Salvador Nicaragua 40 -40- 30 -30- 20 -20- 10 10 , I 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Deciles Deciles Note: Data are from Argentina (2001), Mexico (2000), and Brazil (2001). Note: Data are from Costa Rica (2000), El Salvador (2000), and Nicaragua (1998). Source: Authors' calculations based on microdata from household surveys. income gap is more than 20 percentage points in Argentina between a person located at the 95th percentile and one and more than 30 in Brazil. located at the 80th percentile. Inequality is lower in Argentina, since the income share To illustrate the long "upper tail" of the distributions, for each of deciles 1-9 is greater than in the other two figure 2.2 shows a histogram of the household per capita economies and, consequently, the top decile share is smaller. income distribution in Mexico for 2000, ignoring the richest The comparison between Brazil and Mexico is also clear: the 1 percent of the population. Most people are concentrated income share of each of the eight poorest deciles is higher in the first quarter of the income line. Including the richest in Mexico, implying lower inequality. The second panel of 1 percent would make the graph illegible, as most of the figure 2.1 replicates the analysis for three Central American population would be concentrated in a small segment very countries. Nicaragua is the most unequal economy of the close to the origin. three, while Costa Rica appears to be substantially more In the academic literature, more sophisticated measures equal than the rest."0 of inequality are preferred to simple statistics on income Column (xi) in table A. 2 reports the income ratio between shares and ratios. Table A. 3 presents a set of commonly used the average individual in the top decile and a typical person indices: the Gini coefficient, the Theil index, the coefficient in the bottom decile. This ratio ranges from 16 in Uruguay of variation, the Atkinson index, and the generalized entropy (in 1989) to values above 60 in several countries. In column index with different parameters.12 All indices are designed (xii), individuals at the outer limits of these deciles are com- to increase as the distribution becomes more unequal. By far, pared, that is, the poorest in the top decile are compared with the most used one is the Gini coefficient, which in the sample the richest in the bottom decile. The income ratios are much ranges from 42.2 in Uruguay for 1989 to 61.2 in Brazil for smaller than in the previous column, a fact driven by the 1990.13 presence of a few individuals with extremely large household Although widely used, household per capita income is incomes compared to even the incomes of most people in probably not the best available measure of individual well- the top decile.11 It has been argued that Latin American dis- being based on household income, as it ignores household tributions are characterized by large differences between the economies of scale and differential needs by age. An indi- rich and the middle class. To show these differences more vidual's equivalized household income is defined as total closely, column (xiii) in table A.2 provides the income ratio household income divided by (A+a,;K+a2;K2)j where A 39 INEQUALITY IN LATIN AMERICA FIGURE 2.2 Household per capita income distribution, excluding the richest 1 percent, Mexico, 2000 Density 0.12 0.10 0.08- 0.06 0.04- 0.02 0 2,000 4,000 6,000 8,000 Monthly household per capita income (in pesos) Source: Authors' calculations based on microdata from the ENIGH, 2000. is the number of adults, K1 the number of children under experienced significant changes in inequality. Brazil is the 5 years old, and K, the numnber of children between 6 and only South American economy where there has been a clear 14. Parameters a allow for different weights for adults and reduction in inequality in the 1990S, although this change children, while 6 regulates the degree of household was small enough to not change Brazil's position as the most economies of scale. Following Deaton and Zaidi (2002), unequal country in the region relative to all other countries intermediate values of the ats (a,= 0.5 and a2= 0.75) and for which information from the 1990s is available. a rather high value of 6 (0.9) as the benchmark case are (Guatemala was more unequal by some measures than Brazil applied.'" Statistics for the distribution of equivalized at the end of the decade, but there is no information for household income constructed in this way are presented in earlier periods.) Most of these results are in accordance with tables AA4 and AS5. Table A.6, which reproduces the Gini those found in other studies for the period 1990-1999 coefficient of that distribution for all the countries in the (Morley 200 1, Sz6kely 200 1, and Wodon and others 2000). sample, is the basic input for figures A.2, A.3, and A.4 These studies, however, overlooked the two most relevant Inequality has risen in most South American economies distributional changes in the region: the large increase in during the last decade (see first panel of figure A. 1). inequality in Argentina and the distributional improvement Argentina experienced by far the biggest jump (7.7 Gini in Brazil. points between 1992 and 2001).5 Venezuela follows with In Central America and the Caribbean, changes have been an increase of nearly 4 Gini points.'" The income distribu- milder (see second panel of figure A.l). The income distri- tion has also become more unequal in Bolivia, Chile, bution has remained remarkably stable in Jamaica, Ecuador, Peru, Uruguay, and possibly Paraguay (see below Nicaragua, and Panama and has become more equal in for a discussion of the Paraguay). Colombia has not Honduras and somewhat more unequal in Costa Rica and 40 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA FIGURE 2.3 Equivalized household income a. Early 1990s b. Late 1990s to early 2000s Gini coefficient Gini coefficient 60 -60- 55 -55- 50 [1 50 40 -,J=Lr[LT 1~T1 11 40 - <,ob~~~~~~~~~< PvP \0*,o9 Source: Authors' calculations based on microdata from household surveys. El Salvador."7 Inequality indices went down in Mexico by Although Latin America has traditionally been very homo- 2000, although not enough to confirm that the fall is really geneous compared to other regions of the world, it has become significant in statistical terms (see below for a discussion on even more so in distributional terms in the last ten years. this point). Initial analysis by the Government of Mexico of Figure 2.3 illustrates this fact. In the early 1990s, coun- the 2002 survey finds a larger further decline in inequality, tries fell into three groups: (1) those with low inequality (rel- but this was not available in time to include in the analysis ative to regional standards), comprising Argentina, Costa for this report. Rica, possibly Peru, Uruguay, and Republica Bolivariana de The assessment of inequality patterns is quite robust with Venezuela; (2) those with high inequality; and (3) Brazil, regard to the choice of inequality measures. It is interesting which stood out as significantly more unequal than the rest. to note, however, that the share of the poorest deciles has Ten years later, the differences among groups are not so clear. increased significantly only in Brazil and Panama, while it A sort of convergence of inequality levels seems to have taken has shrunk in most of the region's economies. Consequently, place. Figure A.3 shows this distributional convergence for for instance, the Atkinson index with inequality-aversion the three largest economies of the region. Although still sig- parameter 2 (see column (vi) in tables A.3 and A.5), which nificantly different, the income distributions of Argentina, gives more weight to changes affecting the poorest individ- Mexico, and Brazil have rapidly become more alike during uals than the Gini coefficient, generates a somewhat more the last decade. pessimistic picture of distributional changes in the region. A decade of differential changes has had some impact on It should be noted that less unequal countries have the inequality ranking of countries in the region. Argentina performed worse on average than more unequal countries. moved up the inequality ladder, closer to the mid- and high- While the distribution has become more unequal in inequality group; Bolivia and Chile also rose in the rank- Argentina, Uruguay, and Rep6blica Bolivariana de Venezuela ings; and Colombia, Costa Rica, and Honduras all moved (three low-inequality economies), it either has not changed down. In spite of having the largest reduction in inequality or has become more equal in Brazil, Colombia, Mexico, and in the last decade, Brazil remains at the top of the list of these Panama (four mid- to high-inequality economies). The stan- countries. dard deviation for the distribution of Gini coefficients in the As a result of the reported changes, the average Gini coef- region fell substantially in the last decade, from 6.1 to 4.6. ficient across countries increased almost 1 point in the period 41 INEQUALITY IN LATIN AMERICA (from 50.5 to 51.4). The population-weighted average, how- inequality in the region. Unfortunately, the view of this issue ever, shows a small decrease (from 51.9 to 51.5) because of becomes increasingly blurred as one goes back in time. As the positive performance of Brazil and Mexico and the sta- recently as the 1970s, many countries did not conduct bility of Colombia, the three most populated countries in national surveys or even any household surveys. In fact, it the region. was only after World War II that countries around the world Table A.7 reports the Gini coefficient for the distribution started to conduct household surveys and to compute of household income divided by alternative equivalent scales. inequality statistics in a systematic way. Mexico and some In columns (vii) and (viii), distributions in urban and rural Caribbean countries (including Barbados, Guyana, Jamaica, populations are separated wherever possible. In some coun- and Trinidad and Tobago) were the first in the region to tries inequality is higher in cities, while in others it is higher participate in that trend in the 1950s. in rural areas. However, in most countries inequality dif- Only Mexico has continued with a systematic program of ferences between urban and rural areas appear to be minor. surveying household incomes and expenditures. The available Household surveys do not usually properly capture nonla- statistics for that country show a mild increase in income bor and nonmonetary income. Columns (ix) to (xii) report inequality in the 1950s and the first half of the 1960s (Felix the Gini coefficient for the distribution of household per 1982, Fields 1989, and Altimir 1996b). There is some evidence capita income, including as income sources only labor that inequality also increased in some of the countries in the income, monetary income, labor monetary income, and labor region where distributional statistics became available in the monetary income in urban regions. These are the most 1960s (such as Brazil, Costa Rica, Chile, and Uruguay)."8 homogeneous household income variables to compare across Most countries either consolidated or introduced countries. household surveys in the 1970s. The picture of income According to some, inequality only has normative sig- inequality from that decade on is therefore clearer. Some nificance for the distribution of variables that are beyond the international organizations (including ECLAC, IDB, and the control of individuals. This idea may indicate, for instance, World Bank) shed additional light on the issue by generat- concern about the distribution of total household income and ing periodic reports depicting the level, structure, and trends not household income adjusted for demographics, since fer- of income inequality in the region. Table A.8 indicates signs tility decisions are (mostly) under the control of individuals. of changes in inequality trends in most Latin American coun- Without judging the validity of this argument, column (xiii) tries in the last three decades. During the 1970s, inequality shows the Gini coefficient for the distribution of total house- increased significantly only in the Southern Cone countries hold income. Finally, table A.7 reports the Gini coefficient (that is, Argentina, Chile, and Uruguay). for the distribution of equivalized household income for In contrast, several countries in Latin America (the people in certain age ranges in order to control for factors Bahamas, Colombia, Mexico, Panama, Peru, and RepCiblica related to the lifecycle. Bolivariana de Venezuela) experienced equalizing changes, Most of the qualitative results related to inequality trends while others showed stable distributions. The 1980s was a and cross-country comparisons do not significantly vary when "lost decade" in distributional terms, as well as in terms of any of the distributions in table A.7 are considered instead growth. Most countries suffered a significant increase in the of the household per capita or equivalized income distribu- level of income inequality. In about half, inequality contin- tions given in tables A.3 and A.5. There are a few ranking ued to increase in the 1990s, although in most cases at a reversals and changes in trends as different income variables slower pace. As a result of the patterns described above, most are considered, but the main results remain quite robust countries in the region now have more unequal income dis- despite these methodological changes. tributions than they did around 1970, and very likely than they did at the end of World War II. There are some excep- A brief history of income Inequality In Latin tions, but for the majority of Latin American countries the America since the 1950s economic changes of the last half-century have been mainly This section combines information from the data set prepared disequalizing. for this report with evidence from other sources for previ- The previous evidence refers to countries in the region ous decades to draw a general picture of trends in income considered separately. Londofio and Szekely (2000) compute 42 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA inequality indicators for the region as a whole by calculat- Due to the limited number and changing nature of house- ing a Lorenz curve using the percentiles of each country. They hold surveys, distributional information for Bolivia and conclude that inequality fell in the 1970s, increased in the Paraguay before the mid-1990s is scarce. Inequality appeared 1980s, and increased a bit more in the first half of the 1990s. to increase slightly in Bolivia during the 1990s, a result that The average income ratio of top to bottom quintiles went is confirmed by other studies (Morley 2001, Szekely 2001). from 22.9 in 1970 to 18.0 in 1982, back to 22.9 in 1991, Paraguay did not have reliable national household surveys and then to 24.4 in 1995. Londofio and Szekely (2000) also until the mid-1990s. In order to gain some insight on the conclude that both the level and the change of overall evolution of inequality, the authors of the current study com- inequality are mainly due to differences within rather than puted the Gini coefficient for two years, 1990 and 1995, across countries. In fact, in the last 20 years a slow conver- using only data from the metropolitan area of Asunci6n; the gence in per capita income has occurred among countries in finding was a sizable increase in inequality.21 Inequality the region. Consequently, the increase in regional inequal- seems to have decreased during the second half of the 1990s ity is due exclusively to disequalizing changes in the income in Paraguay, although possibly not enough to compensate distributions within countries. for the increase that occurred during the first half.22 Having described the main changes in the region as a Income distribution in Colombia and Republica Bolivari- whole, the rest of this section is devoted to presenting a broad ana de Venezuela became more equal in the 1970s and more picture of income inequality patterns by country. unequal in the 1980s. In the 1990s, there was no recovery Inequality has dramatically increased in Argentina during from the distributional losses of the 1980s; inequality con- the last three decades.'9 The Gini coefficient for the house- tinued to increase in Repuiblica Bolivariana de Venezuela and hold per capita income distribution in the Greater Buenos the pattern for Colombia was stable (see also Ocampo and Aires area increased from 34.5 in 1974 to 53.8 in 2002 others 1998, Szekely 2001). In Peru, although there is no (CEDLAS 2003). Even if observations for the recent years clear evidence that income distribution became more of economic crisis are ignored, the trend toward an increase unequal in the 1970s and 1980s, data for the 1990s sug- is noticeable. None of the other Latin American countries gests a significant movement toward a greater concentration has experienced such deep distributional changes as of income. Studies that use expenditure data have had similar Argentina.20 results. The distribution seems to have also become some- Inequality also increased in neighboring Uruguay during what more unequal in neighboring Ecuador, at least in the the 1990s, although the increase was smaller. Moreover, there second half of the 1990s. were no significant distributional changes in Uruguay in the The income distribution in Mexico has changed in differ- 1970s and 1980s. As a consequence of these divergent pat- ent directions in the last three decades. After improving in terns, the distributions of Argentina and Uruguay, once the 1970s, the distribution became substantially more almost identical, now are significantly different. The third unequal in the 1980s. In the 1990s, despite important eco- country in the Southern Cone, Chile, has always had higher nomic changes and shocks, income distribution has remained inequality indicators, but that country's income distribution remarkably stable (however, as noted above, preliminary became more unequal during the 1970s and 1980s. results for 2002 indicate a reduction in inequality). The tables Although this "storm" seemed to end in the 1990s (Ferreira in this section illustrate this fact, which has also been high- and Litchfield 1999), there are no signs of distributional lighted by other authors (Morley 2001, Szekely 2001).23 recovery; in fact, inequality measures increased slightly The inequality pattern for Panama is similar. In Costa during the last decade (see Contreras and others 2001). Rica, the distribution remained stable for decades at low Brazil has traditionally been the most unequal economy levels of inequality (Londofio and Szekely 2000). Despite a in the region. The Brazilian economy experienced a signif- small increase in inequality during the 1990s, Costa Rica icant increase in income inequality during the 1980s remains one of the most equal countries in the region (see (Ferreira and Litchfield 1996), but since then inequality has also Trejos 1999). Inequality is much higher in the other stabilized and even started to decline (Neri and Camargo Central American countries. The evidence suggests no 1999). As mentioned above, a drop in income inequality in significant inequality changes in Nicaragua, a drop in Brazil during the last decade has been identified. Honduras, and a small increase in El Salvador during the 43 INEQUALITY IN LATIN AMERICA 1990s. Only one household survey is available for Guatemala, by Latinobar6metro,24 nearly 90 percent of the population on the basis of which the country emerges as one of the most considers the current income distribution in their countries unequal countries of the region. to be "unfair" or "very unfair" (see figure 1.3 and table A.9). Once more widespread among Caribbean countries, The correlation between the level of income inequality as household surveys are now scarce in the region. This makes measured by the Gini coefficient for the equivalized house- monitoring the social situation and designing poverty reduc- hold income distribution and the proportion of "very unfair" tion policies in the subregion much harder. For the hand- answers (or the sum of "unfair" plus "very unfair") is posi- ful of countries where data is available, however, some trends tive but only marginally significant. The unconditional can be discerned. During the 1970s and 1980s, inequality relationship between the change in the Gini coefficient increased in the Bahamas and decreased in Trinidad and during the 1990s and the perception of justice in the income Tobago, according to Fields (1989) and the World Institute distribution seems to be nonsignificant. for Development Economics Research (WIDER) (2000). Some studies report mild inequality increases in the Consumption inequality Dominican Republic (Hausman and Rigob6n 1993). Income As discussed in section 2.1, household consumption is a distribution has remained quite stable inJamaica in the last better measure of well-being than income. Unfortunately, decade, as shown in this and other studies (Chen, Datt, and only a few countries in the region conduct expenditure sur- Ravallion 1995, World Bank 1999). Inequality in the veys on a regular basis, with most of them participants in Caribbean seems to have always been significantly lower than the World Bank's Living Standards Measurement Study pro- in the rest of Latin America. ject. Only Ecuador, Jamaica, Nicaragua, Mexico, and Peru It is always tempting to account for inequality patterns have conducted more than one expenditure survey in the last with a simple explanation, for instance by referring to a few decade.25 The inequality patterns that can be traced with that macro variables. Inequality decreased in the 1970s during information do not differ significantly from the one times of relative economic prosperity and increased during depicted in this paper using income data (WIDER 2000). the "lost decade" of the 1980s. According to this simple view, As expected, inequality levels are much lower when com- the recovery of the 1990s should have brought significant puted over the distribution of expenditures, as people tend distributional improvements. However, there is no evidence to smooth their consumption from more volatile income pro- that this happened. Of course, many changes that occurred files. However, the changes over time are similar: inequal- in the 1990s can be blamed for the lack of distributional ity increased in Peru, probably also in Ecuador, and stayed improvement, but their consideration would result in more roughly unchanged in Jamaica, Mexico, and Nicaragua.26 complex explanations. A sign of this complexity is the mul- tiplicity of distributional patterns across relatively homo- Other dimensions of the distribution geneous countries that result from the evidence provided in Inequality is just one dimension of income distribution. This this chapter. (Discussion of the inequality determinants in section briefly considers three other relevant dimensions: the region is deferred to part II of this report.) polarization, aggregate welfare, and poverty. Are the levels of inequality in the region's income dis- tributions high? The answer seems to be "yes" without the Polarization need for much qualification of the question. That current The notion of polarization refers to relatively homogeneous inequality levels in the region are high is illustrated in three clusters of households or individuals that identify with each ways: (1) compared to previous decades, (2) compared to other and are antagonistic toward or alienated from other countries in other regions of the world, and (3) according clusters. A case of maximum polarization in terms of income to the perceptions of people throughout Latin America. This would be one in which half the population is penniless and section has shown evidence for point (1), while the next the other half shares total income equally among its mem- section is devoted to consideration of point (2). bers. A central conjecture that motivates research on polar- Regarding point (3), evidence is of course more elusive. ization is that differences between homogeneous groups can According to a recent survey in various Latin American coun- cause social tension. Some indices have recently been devel- tries on perceptions about various economic and social issues oped to measure income polarization.27 These measures 44 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA depend on three factors: (1) the number of groups and their box 2.1, while differences among ethnic, racial, and gender relative sizes, (2) the degree of equality within each group groups are the subject of chapter 3. ("identification"), and (3) the degree of income differences among groups ("alienation"). Intuitively, higher levels of Aggregate welfare identification and alienation would increase polarization. It To assess the aggregate welfare of an economy, both the mean is worth noting that polarization can increase when inequal- and the inequality level of the income distribution should be ity decreases (and vice versa). For instance, some transfers taken into account. It could be the case that inequality from the middle class to the poor and the rich can lead to increases but everybody's incomes go up. In that case, most both lower inequality and higher polarization (see Esteban people would agree that aggregate welfare in an economy has and Ray 1994). Thus, the analysis of income polarization is increased despite the growth in inequality. Since the perfor- complementary to that of income inequality. mance of an economy should not be assessed only by consid- From the sample of household surveys, the authors com- ering inequality statistics, the opposite mistake of just looking puted two bipolarization indices for each country and year: at average statistics-which is very common in the economics the Wolfson Index, which divides the distribution at the field-should be avoided as well. Average income may rise median income, and the Esteban, Gradfn, and Ray 1999 but inequality may also increase in such a way that some (EGR) Index, which finds the optimal income cut-off point. people suffer reductions in their real incomes, which may then Table A. 10 shows the results for these bipolarization mea- translate into a negative assessment of the overall performance sures, along with the Gini coefficient, for both the distrib- of the economy, according to some value judgments. ution of household per capita income and the distribution Table A. 14 presents welfare measures for all the countries of equivalized household income. in the sample that have more than one observation. Each As with inequality measures, polarization increased in column shows the value of a given aggregate welfare func- several South American countries and remained stable in tion for a given country and year. Values are rescaled so as Central America, the Dominican Republic, and Jamaica. to make the first observation for each country equal to 100. Argentina, Bolivia, Uruguay, and Venezuela experienced the Four social welfare functions are considered. The first uses largest increases in polarization. Among the economies with the average income of the population; according to this value falling bipolarization measures, there are cases in which judgment, inequality is irrelevant. In columns (ii) to (iv) and inequality increased, for example Chile. (Note that in (vi) to (viii), three widely used functions that take inequal- tables A.2 and A.4, the share of the top decile increased ity into account are considered.29 In the first panel, real per significantly in Chile in the last decade, driving inequality capita gross domestic product (GDP) from national accounts measures up.) is taken as the average income measure and combined with Among the main losers of the distributional changes of the inequality indices shown in table A. 3, in order to obtain the 1990s were people in deciles 7 to 9, that is, people who rough estimates of the value of aggregate welfare according are considered according to bipolarization measures as to different value judgments.30 For various reasons, per capita belonging to the same "class" as the winners in the top decile. income from household surveys differs from estimates in the This fact weakens measured "identification" within the high- national accounts, so in the second panel of table A.14, the income group, and in turn drives bipolarization measures exercise is repeated using information only from household down. Between 1995 and 1999, Paraguay exhibited an oppo- surveys.31 site pattern: the share of the top decile went down, while Most Latin American economies managed to grow during the share of deciles 7-9 increased significantly, implying a the 1990s, but in many the income distribution became decline in inequality but an increase in bipolarization that more unequal. This combination led to ambiguous results was driven by a tighter identification within the high-income for aggregate welfare in some cases. In all ten economies por- group.28 trayed in figure A.3, real per capita GDP increased during Theories of polarization in economics emphasize the the 1990s. However, in Peru and Venezuela, according to importance of group-based differences. There is, of course, value judgments that attach more weight to the poorest indi- a much stronger tradition ofanalysis ofgroups in other social viduals (Atk(2) in figure 2.4), welfare declined, while in science disciplines. Issues of social class are discussed in others, such as El Salvador and Uruguay, the welfare increase 45 INEQUALITY IN LATIN AMERICA BOX 2 I Soca clss The concept of ,ocial Clai is an analytical rool widely used income for each class relative to thar ot' pey entrepreneurs. in sociology and other social sciences, but mostly ignored In Argentina, capitalists. professionals, and executives in modern economics. The term refers to "discrete and gained significantly in relative terms compared to petty durable categories of the population characterized by dif- enrrepreneurs and, in particular, ro informal workers. In ferential access to power-conferring resources and related addition, the income gap berween formal and irLormal life chances' (Portes and Hoffman 2003). In practice, social workers widened during the 1990s in Argentina. classesareusuallydefined in termsoftincomesources. Portes In contrast, relative incomes across classes were prertt and Hoffman (2003) recently presented evidence on class stable in Brazil. The same conclusion applies when con- structure and trends in Latin America, based on informa- sidering equivalized household income instead of mdi- rion from household surveys gathered by ECLAC (2000). vidual income (as indicared in the third panel). The fourth In that study, fivegroupsareconsidered: capitalists, pro- panel of table A.12 shows that the dominant' classes fessionals and execurives, perty enrrepreneurs (or the petty receive around 20 percent of total income. These estimates. bourgeoisie'. formal workers, and informal workers. The however. are very likely affected by the difficulties in first two groups comprise the "dominant" class. Empirical including capitalists and landlords in household surveys estimates are rough. Capitalists are operationally defined and by the problem of underreporting of income. Finally, as owners of firms with more rhan five workers, while pro- the fifth panel of the Table shows Gmin coefficients within fessionals and execurives work in the public secror and in classes. Inequaliries within groups are lower in Argentina firmsemploving fiveor moreworkers. The petty bourgeoisie than in Brazil, although the gaps are narrowing includes owners of small firms, own account professionals, There is a sizable degree of' income overlap among andtechnicians. Formalworkersaredefinedasthoseinthe social classes. Figure A.5 shows the relarive income of each public sector or in firms with five or more workers. percentile of rhe within-class distriburions in Argentina Based on information from ECLAC (2000), the first and Brazil. Both graphs show substantial income panel of table A. ll shows the relative occupational income overlapping across classes, especially among capitalists. of each social class for each country and year. Except for a professionals, and petty entrepreneurs on the ont hand. few countrLes, there are no clear signs that the "dominant" and between formal and informal workers on the other. class, defined in this very narrow way, has become richer Table A. 13 shows cross-tabulations of the deciles, based in relative terms during the 1990s. In addition, in most on the distribution of individual income and the class countries the shareof low-paid informal workers in the pop- structure. Around 75 percent ot capitalists declare ulation has not increased significantly since the 1990s, incomes that place them in the top rwo deciles of the indi- according to the second panel of rable A.l 1. vidual income distribution. However, it should be noted LUsing the definitions in Portes and Hoffman (2003), rhat those deciles also include around 80 percent of all rhe authors of rhe current study have applied the data set professionals and executives, more than 60 percent ot to compute the class structure in Argentina and Brazil, the petty entrepreneurs, more than 15 percent ot tormal work- two paradigmaric cases of disrriburional changes in Latin ers, and around 10 percent of informal workers. Of course. America used in this analysis. The main results are shown informal workers are more concentrated in the bottom in table A.12. Compared to Argentina, Brazil has a sig- deciles of the income distribution. It is interesting to note nificantly higher share of informal workers and a lower that informal workers are Increasingly moving toward share of perry entrepreneurs and formal workers. The the bottom deciles in Argentina. in contrast to a more second panel of table A.12 shows average individual stable situation in Brazil. 46 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA of inequality would end up reducing the number of people FIGURE 2 4 whose incomes fall below an absolute poverty line. In fact, Indices of aggregate welfare for Argentina under alternative the significant fall in poverty in Latin America during the assumptions of a society's aversion to inequality 1970s was fed by growing economies with either stable or Social welfare index more equal distributions. The story in the 1980s was exactly 130 - the opposite: falling incomes and more unequal distributions 120- combined to generate a tangible increase in poverty statis- 110 - tics. The 1990s again showed a different combination: despite 100 - - - -'' 0000t some disequalizing changes in income distribution, the strong 90- recovery of several Latin American economies generated a 80- reduction in the poverty indicators both for the region as a _-- Bentham -0- Sen whole and in most countries (see Wodon and others 2000 and 70 - b Atk(2) -o- Atk(1) 2001, Szekely 2001, and Sala-i-Martin 2002). Overall, in the 1992 1996 1998 2001 last three decades the region has experienced a substantial decline in the incidence of poverty.33 Note: Welfare indices normalized to 100 m 1992. Source: Authors' calculations based on microdata from Table A. 15 summarizes part of the large body of litera- household surveys; see table A.14 in the appendix. ture on poverty indicators in the region.34 It should be emphasized that poverty measures are very sensitive to the explicit and implicit assumptions made by researchers was significantly smaller than growth in GDP. In Argentina, (Szekely and others 2000). Thus, caution is required when the contrast is more dramatic: despite an 11 percent increase comparing results obtained by different methodologies. The in per capita GDP measured by national accounts between general trend during the 1990s was toward a reduction in 1992 and 2001, aggregate welfare decreased according to all income poverty. In the mid-1990s, various authors found the value judgments implicit in the calculations that do not that there were 1.5-2 percent fewer poor people in the region neglect distributional issues. The increase in inequality was than at the beginning of the decade (Wodon and others large enough to offset the growth in mean income. In con- 2001, Londofio and Szekely 2000). This reduction is mainly trast, aggregate welfare unambiguously increased in Costa explained by the performance of Brazil, where the headcount Rica and Chile despite disequalizing distributional changes. ratio fell around 7 percentage points. In Brazil and Panama, growth in aggregate welfare was Chile was the other country that had an extraordinary per- fueled by both a growing per capita income and a more equal formance in terms of poverty reduction (around 16 percent- income distribution.32 age points). Colombia (until the late 1990s), Costa Rica, and The scope of these exercises is limited, as it is assumed Panama also managed to reduce poverty thanks to stable that aggregate welfare is a function only of household income distributions and growth. On the other hand, in income. Other factors-including freedom, security, polit- Mexico the percentage of poor people increased about 5 per- ical power, access to basic services, and health status-should centage points during the first half of the 1990s. Much of also be considered as arguments for individual well-being. the literature fails to report the increase in poverty in A comprehensive welfare study including these aspects is Argentina, even though the official headcount ratio in beyond the scope of this chapter. However, section 2.5 below Greater Buenos Aires went from 17.8 in 1992 to 25.9 in provides statistics on the distribution of some of the vari- 1998 and to 35.4 in 2001. That poverty increase, also notice- ables that, arguably, influence individual utility. able in the rest of the country, took place during a period of growing per capita income. Venezuela is the other case Poverty where, despite a growing economy, the increase in inequal- Although inequality and poverty are different concepts, they ity pulled poverty rates up significantly. are closely related. Changes in income-related poverty can be Figure A.4, based on data from Szekely (2001), shows the thought of as the result of changes in average income and poverty headcount ratio of most countries in the region inequality. For instance, a growing economy with stable levels on the basis of a poverty line of US$2 a day (adjusted for 47 INEQUALITY IN LATIN AMERICA differences in purchasing power across countries) in the late consumption will thus be smaller than that of the 1990s. High-income countries with relatively low inequal- distribution of income at any point in time. ity, such as Uruguay, Republica Bolivariana de Venezuela, * It is generally argued that income is recalled and mea- and Argentina, have relatively low poverty levels. Chile has sured with a greater level of error than is consump- recently joined this group due to its growth performance, tion (see Deaton 1997), adding to variance (and despite being a high inequality country. On the other hand, therefore inequality) in the distribution of current low-income countries with high inequality, such as Bolivia, incomes. In particular, underreporting can be the result El Salvador, Guatemala, Honduras, Nicaragua, and Paraguay, of several factors. These include a deliberate decision have very high poverty levels. by the respondent to misreport; the absence of ques- tions that could capture some income sources (for 2.3. Measurement-related issues example, the implicit rent from housing for home- and data limitations owners); or the difficulty of recalling or estimating Although household surveys are the most appropriate source income from certain sources (for example, earnings of information for distributional analysis, they have many from informal activities, in-kind payments, home pro- limitations. It is important to make these limitations explicit duction, and capital income). This problem probably in order to ensure that statistics are interpreted with cau- implies a downward bias on the measured living stan- tion and to identify areas for future improvements. While dards of both poor people, who rely on a combination most limitations are present everywhere, some are particu- of informal activities or production for their own con- larly important in Latin America.35 Probably the main dif- sumption, and of rich people, who derive a larger pro- ference between household surveys in the region and in other portion of income from nonlabor sources and are less developed parts of the world is the lack of expenditure- probably more prone to underreport. based questionnaires in most Latin American countries. As noted in section 2.1, expenditures are a better measure of Whatever the sources of difference between income- and living standards than income, especially in less developed expenditure-based information, what can be said about how areas. The extension of consumption-based questionnaires in Latin America would compare with regions in which the region is certainly one of the main areas in need of atten- inequality is only measured in terms of consumption? tion by national and international agencies. A new study by Elbers and others (2003) sheds some ini- tial light on this question. The study takes advantage of a Does use of Income rather than expenditures 1996 pilot household survey conducted in Brazil's north- overstate Latin America's Inequality relative eastern and southeastern regions. Known as the Pesquisa sobre to other regions? Padr6es de Vida (PPV), that survey is modeled on the World It may be that Latin America appears to be more unequal Bank's LSMS. The PPV is a multimodule integrated survey than other regions that only report information on expen- that collects data on both household consumption and ditures because of systematic differences between income and income. Fairly detailed information on consumption expen- expenditure inequality. There are two reasons why, in any ditures is also collected, making it possible to impute values given country, incomes are generally more unequally dis- of consumption streams from items such as housing and food tributed than consumption expenditures financed by those products made at home. While generally viewed as a high- same incomes: quality survey, the PPV has a small sample size and limited geographic coverage, and is therefore not widely used to Households dislike sudden variation in their con- study inequality at the national level. sumption patterns, and thus tend to smooth con- Elbers and others (2003) employ a recently developed sumption patterns over time. For a given individual, methodology to impute consumption from the PPV survey consumption flows therefore behave almost like the into Brazil's traditional (income-based) PNAD survey.36 averages of income streams over time (see Friedman This approach allows the authors to estimate consumption- 1956). The law of large numbers implies that the rather than income-related inequality in Brazil, based on cross-sectional dispersion of the distribution of the same underlying data from the PNAD household survey. 48 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA FIGURE 2.5 Brazil's consumption inequality in international perspective Household per capita consumption Gini coefficient 0.65 - 0.60 - 0.55- 0.50- 0.45 - 0.40 - 0.35 - 0.30 - 0.25 - 0.20 - 0.15 - 0.10 -* Three different estimates of Brazil's consumption inequality 0.05 * Consumption inequality in other countries 0 .0 0 I I I I I I I I I I I I I I I I I I IC, e; ,~ <0 Source: Elbers and others 2003. Figure 2.5 indicates that when inequality is measured in Brazil is less obviously an outlier. To the extent that simi- the PNAD on the basis of the Gini coefficient of imputed lar concerns can be raised about the quality of income data consumption, Brazil's inequality is less extreme. This new in other Latin American countries, inequality in the whole Gini coefficient in Brazil as a whole is measured at around region, relative to that in other regions, may be lower than 0.45, compared with around 0.6 for household per capita has traditionally been assumed. income. If attention is given only to the northeastern and southeastern regions, inequality is a bit higher, at 0.46. Income volatility and mobility Inequality in these two areas, measured directly with the A related concern arises from the fact that there is con- PPV household survey data, is a bit higher still, at 0.49. siderable income volatility within Latin America. When All three of these measures are, however, significantly incomes are volatile from month to month, measured below those observed in a number of other countries. It is inequality tends to reflect an overestimation of underlying interesting to note that, with the exception of Russia and inequality. De Ferranti and others (2000) showed that Nicaragua, most of the countries with high levels of con- aggregate volatility of real private consumption in Latin sumption inequality are in Africa, which is also a high- America was almost three times as high as in the developed inequality region (and possibly also one with weaker economies between 1960 and 1999. During the same statistical traditions and so less accurately measured con- period, it was slightly higher than in most countries in Asia, sumption). While still unequal by international standards, but lower than in Sub-Saharan Africa. This aggregate 49 INEQUALITY IN LATIN AMERICA volatility translates into high levels of risk and uncertainty behind the large difference between income inequal- at a more disaggregated level. In Brazil, for instance, the ity and consumption inequality measures in Brazil dis- mean income of the poorest fifth of the population fell by cussed above. Consumption is a better indicator of more than 30 percent during the 1982-1983 recession, rose permanent income. by more than 30 percent during the 1984-86 boom, fell * To the extent that people are risk-averse, volatility is by some 12 percent in 1990-1991, and then rose again by costly in terms of social welfare. On the one hand, 15 percent in 1994-1995. Volatility in both gross national long-run interpersonal inequalities may be overesti- disposable income and real private consumption is also mated by focusing on current incomes. On the other very high-indeed higher than in the rest of Latin hand, some of this measurement reflects inequality in America and in the Caribbean countries (World Bank, the incomes of the same people over time, which is 2003e). Under the standard assumption that consumers are detrimental to individual (and thus social) welfare for averse to intertemporal volatility in their consumption path reasons having to do with preferences for lower risk. (or averse to risk), this sort of roller coaster clearly reduces Although this is less of a problem to the extent that social welfare. people can smooth out patterns of consumptions, it is Income variations over time are caused by a number of only cost-free if capital markets (including, impor- factors. As discussed above, a large self-employed sector is tantly, insurance markets) are perfect. usually characterized by more volatility than is the formal * Some of what appears to be intragenerational short- sector, since profits adjust to changes in demand and cost term mobility is simply volatility by a nicer name. conditions directly but wages are usually more protected. Some of the evidence on mobility in Latin America is The rate at which firms close tends to be much higher discussed in box 2.2, but the point here is that evi- among small firms in the informal sector than among large dence of large, short-term disruptions in wage or firms in the formal sector (see Levenson and Maloney 1998). income distributions is likely to reflect high risk and However, in Latin America real wages have often oscillated volatility. quite dramatically as a result of high and often unstable inflation rates combined with imperfect indexation mech- The absence of multiyear panel data in the region means anisms. In addition to wages, income may also be affected that very little is actually known about intragenerational by changes in employment status. Unemployment risk is mobility in Latin America (however, see Graham and Pet- of increasing concern in Latin America, particularly for tinato 2002 for a discussion of results of a study in Peru). young people and women, as reviewed in de Ferranti and If and when more becomes known, it is likely that analysts others (2000). will face trade-offs between welfare gains (arising from In rural areas, climatic events represent an additional lower inequality during multiple periods when later source of risk, beyond those which arise from terms-of-trade incomes are less predictable on the basis of earlier and other macroeconomic shocks. A study of household incomes) and welfare losses (arising from aversion to income strategies for coping with the downturn in agricultural pro- fluctuation over time). These trade-offs will be the same duction in El Salvador in 1997 (Conning, Olinto, and as those that result from mobility and volatility trends Trigueros 2000) found that landless workers, who tended to observed in more developed countries. Gottschalk and rely on agricultural employment prior to the crisis, were hit Spolaore (2002, p. 193), for instance, find that "When aver- the hardest. sion to income fluctuations and (beyond-the-veil) risk are What does the evidence of large income volatility in Latin introduced, those larger costs offset the benefits stemming America (reviewed in more detail in de Ferranti and others from reduced multi-period inequality. Consequently, Ger- 2000) imply in the context of a study of inequality? There mans and Americans end up obtaining similar net bene- are three basic implications: fits from mobility, although for very different reasons." Recently, this sort of careful thinking about mobility has * As the volatility of current income increases, so does challenged earlier facile assertions that higher inequality the gap between inequality in current income and in the United States than in Europe was not problematic, inequality in permanent income. This is one factor since mobility was also higher there. 50 BOX 22 MobUt In Latlh Amern What IM Is know.? Since the idea of intragenerational short-term mobility is In keeping wirh points presented above, members of the strongly relared ro volatility-or at leasr rhe two concepts younger generation should be able to achieve better positions are not easily separable-rhe literature usually focuses more in rhe labor market in comparison to their parents In other on intergenerarional long-term mobility. The main vari- words, an increase in occupational mobiliry could be expected ables used in this kind of analysis are generally assets or as each generation becomes more educated. Although the choices related to education and occupation, which can be data suggest this partem-since more than half the sons in good proxies for permanent income. With these two vari- Brazil are in different occuparional classes from those ot'Eheir ables, Val6ria Pero (20031 studied the Brazilian case and &athers-uhe occupational mobility rate in that country seems showed that the country does not have significant educa- to be very low compared with other countries. most of which tional or occupational inrergenerarional mobility. are developed (Scalon and Ribeiro 2001). In fact. Brazil has one of the lowest levels of intergen- In some sense, the occupational mobility rate can be erational educational mobility in the world. According ro considered a good indicaror of development. According to Behrman, Gaviria, and Szekely (2001), Larin American Pero (2003), mobility can resulr from "economic growth countries have lower intergenerational educational mobil- and its impacts on job creation and on sectoral and occu- iry than developed ones. Furthermore, comparing Brazil pational composition and demographic aspects' and may and the other Latin American countries, there is evidence be a consequence of "distribution of opportunitles, upon that Brazil has an even lower level of mobiliry (Mienezes- which the society builds mobiliry channels to match people Filho 2001). The main reason for this phenomenon is the efficiently in the social structure in a fair system. fact rhat educational performance in Brazil is associated This perspective is corroborated by analvses of different more with family background-in particular the school- Brazilian states, since there is greater mobiliry in more devel- ing of parents-than it is in other countries ' oped areas. The lirerarure concemed with educational and The educational levels of individuals and their parents occupational mobility in Brazil indicates that during the are highly correlated. in the sense that sons of parents 1970s,mobilirywastheresultofindustrializationandurban- with little education also have little education. Accord- ization processes. On the other hand, thisdoes not appear to ing to Paes de Barros and others (2001), the schooling berrue with regard to rhedistributionofopporrunities. Silva of parents appears to be the most important variable in and Roditi (1988) -using a log-linear model ro test the explaining educational performance in Brazil. Moreover, hypothesis of constant circular or relative mobility patterns the schooling of mothers seems ro have a stronger effect over time-argue that the disrribution ofopporturuties have than that of fathers on educational performance. Based not changed signuficantly. However, Scalon and Ribeiro on the lirerature, Pero (2003) goes further, concluding (2001) assert rhat there is a trend toward a more even dis- that 'The influence of parenrs' schooling on educational tribution of the achievement ofsocial positions, even though performance is stronger for men than women, for black Brazilian sociery is rigid in comparison to other countries. than nonblack, and for residents in the Northeast than In conclusion, Pero (2003) finds that occupational in the Southeast." mobility increased in Brazil between 197 3 and 1996. Pero Nevertheless, Bourguignon, Ferreira, and Men6ndez also suggests that other channels of mobiliry were more (2003), as well as others, point to an increase in educa- important in derermining position in the social structure tional mobiliry over time. According to this conclusion, than occupation or class inheritance The Brazilian daca also an additional year of schooling for parents means a larger indicares thar mobility currently tends ro be more relared increase in schooling for older cohorts than for younger to positional changes than to the creation of new positions. ones. A complementary analysis (Pero 2003) indicates that it is possible to find evidence that educational mobility is a Contrary to common senst. variabIes such as the iruning of teachers and more noticeable among less educated individuals. Thus, Intrastrucrure iinr exampit. books. computers. and night courts are not that indications that educational mobiliti is rising for young important in explaining educational pertormance The key point is that although thtse variables hase a poiitise effect. [he} ariounr for just a malil cohorts could be related to the fact thar educarional policy part cit'ihe leel no education as a whole is closing the gaps related to farnily background in Brazil. 51 INEQUALITY IN LATIN AMERICA Issues of coverage and measurement of Incomes changes. Szekely and Hilgert (1999) find that some inequal- A further issue concerns how comprehensive and accurate ity rankings among Latin American countries vary as a wide are questions related to income. All household surveys in range of adjustments are performed to deal with the infor- Latin America include questions on monetary income from mational issues sketched here. However, even when there salaried work, but some use very simple (and therefore less might be some changes, the general picture remains reliable) questions and many do not include estimates of non- robust: low-inequality and high-inequality countries remain salary and nonmonetary payments. Among those countries in their groups regardless of the methodology used for the that focus on this issue, efforts to obtain correct estimates analysis. vary significantly. All countries make some effort to capture A common observation among users of household surveys income from self-employment and capital income. However, is that they do not typically include "very rich" individu- the intensity of these efforts varies among countries and als; millionaires, wealthy landlords, powerful entrepreneurs, sometimes also within countries over time. Differential mis- and capitalists do not usually show up in the surveys. The reporting behavior among respondents and varying efforts highest individual incomes in surveys conducted in Latin in the survey design can distort inequality comparisons America correspond primarily to urban professionals. This among countries. If these behaviors and efforts change over fact may be the natural consequence of random sampling time, they can also distort the view of inequality trends. (that is, there are so few millionaires that it is unlikely that Researchers apply three kinds of strategies to alleviate they would be chosen by a random sample-selection proce- these problems. The first is to restrict the analysis to more dure to answer the survey), nonresponse, or extensive under- homogeneous variables that are subject to fewer problems reporting. The fact is that rich people in the surveys are of misreporting. Typically, people look at the distribution "highly educated professionals obtaining labor incomes, of labor income or, in an even more restricted approach, at rather than capitalist owners living on profits" (Szekely and the distribution of monetary wages from salaried work in Hilgert 1999). The omission of this group surely implies urban areas (see section 2.4). Of course, the cost of doing this underestimation of inequality of a size that is difficult to is ignoring a sometimes sizable part of the overall income predict. distribution. The second strategy is to apply some grossing- Real rather than nominal incomes should be used in any up procedures. Income from a given source in the household distributional analysis. If prices faced by all households were survey is adjusted to match the corresponding value in the same, the distinction would be irrelevant. However, national accounts. This adjustment usually leads to inflat- prices usually differ by location. If two households located ing capital income relatively more than the other income in different regions have the same nominal income but face sources, and hence generates higher inequality estimates. different prices, they will have different living standards. Finally, a third strategy is to estimate some incomes based Despite the fact that many authors have highlighted the on other pieces of information in a survey. For instance, the importance of considering spatial variations of prices in a dis- implicit rent from own-housing can be estimated using tributional study (for example, Deaton 1997, Ravallion and hedonic price regressions if a survey records some housing Chen 1997), price adjustments are rarely performed in coun- characteristics and the amount paid by renters in the rental tries that do not routinely collect information on local prices sector (see Fay and others 2002). In addition, multivariate as part of household surveys. regressions can be run to estimate wages for workers who Unfortunately, most Latin American countries fall into do not report or clearly misreport wages, but do report indi- this category. Some countries have regional price informa- vidual characteristics (for example, education and age). tion, which is useful but does not solve the problem since Since surveys differ with regard to the severity of these price dispersion may be high within a single region (espe- problems, adjustments should be made on a case-by-case cially between urban and rural areas). However, the inequal- basis. This is a task that goes beyond the scope of this ity results in this study appear to be quite robust when chapter. Researchers who have made different types of adjust- adjustments for regional prices are made. For instance, in ments have generally found that most results for inequality Argentina the Gini coefficient for the distribution of house- trends within countries are robust.37 Results across countries hold per capita income for 2001 slightly decreases from 52.2 are somewhat less robust in the face of methodological to 51.9 as regional prices are taken into consideration. For 52 DIFFERENT LIVES: INEQUALITY IN LATIN AMERICA Chile in 2000, the Gini coefficient increases from 57.1 to dispersion measures for inequality indices in a simple and 57.3 when this is done. efficient way.38 Table A. 16 shows the estimated Gini coefficient for the Is measured Inequality over- or understated? distribution of household per capita income for each coun- Is "real" inequality lower or higher than the estimates derived try and year, its bootstrapped standard error, the coefficient here from household surveys? Unfortunately, the answer is of variation, and the corresponding confidence interval for not clear. Some factors lead to an underestimation of inequal- a 95 percent level of significance.39 Given the large size of ity (for example, misreporting of capital incomes or the samples in most household surveys, the Gini coefficients are absence of very rich people in the surveys), while others result estimated with high precision. This is reflected in the low in an overestimation (for example, using monthly income values of the standard errors. Column (iii) shows that the instead of permanent income or consumption). More work standard error is almost always smaller than 1 percent of the is definitely needed in this area. However, it should be noted estimated coefficient. However, in many cases this is enough that the key concern here is not to know the exact level of to cast doubt over the statistical significance of the inequal- inequality of a country in a given year, but instead to make ity changes. For instance, although the recorded Gini time and cross-country comparisons. The authors implicitly increased in Mexico between 1996 and 2000, the two con- assume in this analysis that factors that bias the measure- fidence intervals overlap and make the change in the Gini ment of inequality remain stable among countries and over coefficient insignificant, that is, consistent with when dif- time. ferent samples are taken from a population with a stable It is necessary to avoid either of the two extreme posi- income distribution. tions taken toward household surveys, that is, either to dis- card them or to use them without qualifications. Even with 2.4. Inequalit in Latin America in perspective all their limitations in mind, household surveys provide valu- This brief section is devoted to placing the evidence from able information for a distributional analysis and are the best the previous section in geographical perspective. How available source of data with which to generate representa- unequal are Latin American economies compared to the rest tive distributional statistics of the population. However, it of the world? is important to be aware of their drawbacks. Despite impor- In the last 10 years, several studies have surveyed and tant steps taken by Latin American governments and inter- computed inequality measures across countries and over national organizations in the last decade (for example, the time. Deininger and Squire (1996) put together a large data MECOVI program), there is still a long way to go before a set of quintile shares and Gini coefficients for most coun- more reliable, richer, and more homogeneous set of national tries since World War II. This panel data set-which greatly household surveys is available. In this regard, Latin America stimulated the empirical study of the links between lags behind some other less developed regions. inequality and other economic and political variables-was updated and extended in the World Income Inequality Sample variability and confldence intervals Database, a joint project of the United Nations University, Measures of the different dimensions of a distribution are WIDER, and the United Nations Development Pro- subject to sample variability problems because they are gramme (WIDER 2000).4 Using these and other secondary derived from surveys, not census data. If a sample consisted sources, Milanovic (2002), Bourguignon and Morrison of only two individuals, indicators would surely vary widely (2002), and Sala-i-Martin (2002) recently computed income over time even when the population remains completely distributions for the world and its regions.4" Other authors unchanged, since two different individuals would be ran- have used microdata to compare distributions from differ- domly selected each year. This problem can be alleviated ent regions of the world. Bourguignon, Ferreira, and Leite but not completely eliminated through the use of larger (2003) and Szekely and Hilgert (2001) have compared Latin samples. This point is illustrated here by assessing the American countries with some developed countries at the robustness of some inequality comparisons with the help microdata level. of confidence intervals estimated using bootstrapping This empirical literature unambiguously suggests that techniques. This method provides interval estimations and Latin America is the region with the highest levels of 53 INEQUALITY IN LATIN AMERICA inequality in the world, and that this has been true for as question has been table 5 of Deininger and Squire (1996), long as statistics have been kept.42 Each bar in figure 2.6 which is reproduced here in table A.17. The table, which indicates the value of the Gini coefficient for the distribu- shows nonweighted averages of Gini coefficients by region, tion of household per capita income in countries located in indicates that at least since the 1960s, inequality in Latin four "regions" of the world.43 Inequality in Latin America American countries has been higher than in any other region is higher than in Asia, Eastern Europe, and the developed of the world. With the exception of countries in Sub-Saharan countries.44 Income inequality in the least unequal Latin Africa, the differences in Gini points between Latin America American country (Uruguay) is higher than in the most and other regions are large. This gap narrowed in the 1970s unequal country in Eastern Europe and the industrialized and became wider again in the 1980s. There was no clear pat- countries, and not too much different from that in the most tern in the 1990s, when Latin America performed better than unequal country in Asia. The nine most unequal countries some regions in distributional terms (for example, Eastern in the sample are in Latin America. In addition, the 14 Europe) and worse than others (for example, South Asia). economies included in the graph are all among the 20 most It is sometimes argued that inequality is related to the unequal countries in the sample. state of development in a country and comparisons should Figure 2.6 refers to income inequality. When inequality therefore be made that control for this factor. Londonio and is measured with regard to the distribution of household Szekely (2000) use regression analysis to compute the dif- expenditures, the conclusions are similar. Using information ference between actual inequality and expected inequality from the same source (WIDER 2000), the average Gini coef- given the level of development of many countries in the ficient in the seven Latin American countries for which world. The difference for Latin American countries is posi- expenditure data from the 1990s are available (44.0) is far tive; that is, the region suffers from "excess inequality," which above the average Gini coefficient in Asia (36.6) and East- Londonio and Szekely (2000) find has fluctuated around 13 ern Europe (30.4) and slightly higher than in Africa (43.3).45 Gini points over time. Has Latin America always been more unequal than the rest Although widely cited, Deininger and Squire's table of the world? The most widely cited source to answer this should be interpreted very cautiously. Among its problems FIGURE 2.6 Distuibution of household per capita income, per region, l990s Gini coefficient 60 - 50- 40- 30- 20- 10 0,1 - _ 1 __ llllllllll l 0U'U I Latin American countries Asia Developed countries Eastern Europe Source: Authors' calculations based on WIDER 2000. 54 DIFFI!RLNT LIVES INEQIfALITYV IN LATIN AMERICA TABLE 2.1 Gini coefficients of the distribution of household per capita income: Common sample countries, decadal averages by region Overall Region 1970s 1980s 1990s average Levels Latin America and the Caribbean 48.4 50.8 52.2 50.5 Asia 40.2 40.4 41.2 40.6 OECD 32.3 32.5 34.2 33.0 Eastern Europe 28.3 29.3 32.8 30.1 Changes 70s-80s 80s-9Os 70s-90s Latin America and the Caribbean 2.4 1.3 3.7 Asia 0.2 0.8 1.1 OECD 0.2 1.7 1.9 Eastern Europe 1.0 3.5 4.5 Difference in Gini points: LAC vs. Asia 8.3 10.4 10.9 9.9 OECD 16.1 18.3 18.0 17.5 Eastern Europe 20.2 21.6 19.4 20.4 Source: Authors' calculations based on WIDER 2000, Smeeding and Grodner 2000, Szekely 2001, and estimates for Latin America. are four issues: (1) the sample of countries used to compute the last 50 years. The story at the level of individual coun- the regional statistics is unbalanced, (2) income inequality tries is similar: the five Latin American countries/subregions statistics are mixed with expenditure inequality measures, considered in the study have been among the eight most (3) some of the figures that are averaged come from studies unequal countries/subregions in the world since 1950. that use different methodologies, and (4) data from the 1990s As mentioned before, there are no consistent statistics for are very scarce.46 Some of these problems are corrected in inequality before the 1950s. To calculate world distribution table 2.1, in which Gini coefficients are computed from a between 1820 and 1950, Bourguignon and Morrison common sample of countries, come from a small set of stud- (2002) assume no changes in distribution in Latin America, ies and hence are methodologically more consistent, and use arguing that "the absence of strong evidence suggesting the income as the living standard variable. The general picture distribution was much less unequal during the 19th cen- is not very different from table A.17. Inequality in Latin tury." They also refer to Malthus's suggestion that in 1820 America has been significantly higher than in Asia, OECD inequality was much larger in Mexico than in England. In countries, and Eastern Europe in the last three decades. There chapter 4 of this report, it is argued that extreme inequal- are no signs that this gap is narrowing.47 ity emerged soon after the Europeans began to colonize the In a recent study, Bourguignon and Morrison (2002) com- Americas. puted world income inequality statistics since 1820, includ- In another recent study, Milanovic (2002) computes ing information at the regional level. Table A. 18 shows three regional distributions in which all individuals are treated inequality measures computed for five regions since 1950. The equally as inhabitants of a given region (see table A. 19).5° figures reflect nonweighted and population-weighted aver- Latin America again appears as a region of high inequality. ages across countries or groups of countries.48 Although for However, Asia is currently the region with the highest Gini some regions the statistics go back to 1820, inequality mea- coefficient. This is mainly due to the great disparities in sures for Latin America start in 1950.9 Once again, the region national incomes across countries in that region (for exam- emerges as very unequal compared to the rest of the world. ple, Japan, Hong Kong (China), and the Republic of Korea The last panel of the table shows the difference in inequal- in comparison with India or Bangladesh). In the decompo- ity between Latin America and the rest of the regions in terms sitions performed by Milanovic and Yitzhaki (2002) with of Gini points. There is no evidence that the gap between the same data set, only 7 percent of overall inequality in Latin America and the rest of the regions has narrowed in Latin America is due to between-country group inequality. 55 INEQUALITY IN LATIN AMERICA By contrast, the contribution of the between-country The last five decades have witnessed economic booms and Gini coefficient is 39 percent in Africa, 72 percent in Asia, crude recessions, inward growth models and export-led 39 percent in Eastern Europe and the former Soviet Union, growth strategies, widespread public sector interventions and 18 percent in Western Europe, North America, and and extensive promarket reforms, dictatorships and democ- Oceania. Compared to the rest of the world, Latin America racies, yet in none of the region's countries did any of these is a region comprising relatively similar countries, within scenarios change the income distribution to make it sig- which high inequality prevails. nificantly more similar to distributions in other parts of the Summing up, although there are many methodological world. drawbacks related to the available evidence, differences in This observation suggests an important point: Latin magnitude are sufficiently large to indicate that inequality America seems to be more unequal than the rest of the world in Latin America has been greater than in the rest of the for reasons beyond the economic cycle or particular economic world since at least World War II, with the possible excep- policies. These factors have undoubtedly played an impor- tion of Sub-Saharan Africa. Moreover, there are no signs that tant role in shaping income distribution, a role that should this gap has narrowed over time. Changes in inequality have be studied and better understood. However, there seem to been more or less similar on average to those found in the be underlying factors that are stronger determinants of rest of the world in the last half century. The widespread drop inequality levels in the region. Several chapters in this report in inequality in the 1970s was probably more pronounced elaborate this point (see part II). in Latin America, but this relative gain was lost in the 1980s Which parts of the Latin American income distributions when inequality in the region increased more rapidly than differ most from the rest of the world? Inequality, for in the rest of the world. During the 1990s, inequality instance, may be higher in Latin America because of greater increased in Latin America at about the global average rate. income concentration in the middle class and lower con- It is interesting to note that the position of most Latin centration in the bottom strata, compared to other regions. American countries in the ranking of global inequality has Tables 2.2 and 2.3 suggest that this is not the case. Latin been nearly the same during the last several decades, despite American distributions are mainly characterized by a changes in the economic, social, and political environment. higher income share among the rich relative to countries in TABLE 2.2 Distribution of household per capita income, 1992 (income shares by deciles and vintiles) Latin Eastern Developed World Difference America Africa Asia Europe countries without LA (i)H(vi) Regions (i) (ii) (iii) (iv) (v) (vi) (vii) Deciles 1 1.6 2.1 2.6 2.2 2.5 2.4 -0.8 2 2.4 3.0 3.5 3.8 3.4 3.4 - 1.0 3 3.0 3.7 4.8 5.1 5.3 4.8 -1.8 4 3.4 4.6 5.8 5.7 6.3 5.7 -2.2 5 5.0 5.5 6.5 7.5 7.3 6.7 -1.8 6 6.0 6.5 7.5 8.2 8.6 7.8 -1.7 7 7.6 8.6 9.0 9.4 10.5 9.5 -1.9 8 9.0 10.5 10.5 10.8 12.2 11.1 -2.2 9 14.0 13.3 12.4 12.8 14.8 13.5 0.5 10 48.0 42.2 37.4 34.7 29.1 35.1 12.9 Total 100.0 100.0 100.0 100.0 100.0 100.0 Vintiles 19 10.8 11.3 9.7 8.9 10.7 10.3 0.5 20 37.2 30.8 27.7 25.8 18.5 24.8 12.4 Source: Authors' estimates based on Bourguignon and Morrison 2002. 56 DIFFERENT LIVES INEQIJALITY IN LATIN AMERICA TABLE 2.3 Distribution of household per capita income, 1990s (quintiles) Sub-Saharan East Asia & Eastern Middle East & Developed LAC Africa Pacific South Asia Europe North Africa countries (i) (ii) (iii) (iv) (v) (vi) (vii) 1 and 2 13.2 14.1 18.1 21.7 22.2 17.8 18.4 3 and 4 33.8 33.5 37.5 38.4 40.0 36.8 41.8 5 52.9 52.4 44.3 39.9 37.8 45.4 39.8 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: Authors' estimates based on Deininger and Squire 1996. other regions.5' Who are the losers from this "excess share?" assumed to be fixed, this leaves three sources of differences: The tables suggest that the eight bottom deciles in Latin labor income, nonlabor income, and family size and struc- America have lower income shares than they do in the rest ture. These three components are discussed below. of the world. If anything, the "losses" seem to be larger for middle-income groups. If a typical Latin American distri- Labor income bution had to mimic a typical income distribution in the Labor is the main income source for most individuals. This rest of the world, the income share of the top vintile (the role is magnified in household surveys, since most nonlabor richest 5 percent of the population) would have to be reduced sources are usually not well represented. Column (i) in and the proceeds distributed more or less evenly across the table A.21 shows that the share of labor sources in total poorest 80 percent of the population. reported income is more than 80 percent in most Latin Despite its high inequality, Latin America performs American countries-substantially greater than what is better in terms of poverty levels (when measured by stan- recorded in national accounts. The table also shows the Gini dardized international poverty lines) than do some other less coefficient for the distribution of individual labor income in developed regions of the world, primarily because of the columns (vii) to (ix). Most of the conclusions in section 2.2 region's higher per capita income. Poverty is lower in the on inequality rankings and trends remain valid when the region than in Africa and Asia (see table A.20). World analysis is restricted to individual labor income.52 Wodon poverty has been significantly reduced in the last decade. and others (2001) perform source decompositions of the Gini The Latin American record is better than that of Africa, coefficient, and conclude that about three-quarters of the Eastern Europe, and South Asia, but not nearly as good as Gini coefficient for the distribution of per capita income can East Asia's. be attributed to the contribution of inequality in the labor income distribution, since that source represents a very large 2.5. Looking inside household income share of total income in household surveys. Section 2.2 considered the distribution of household equiv- alized real income among individuals. This variable can be Education written as, (YL+ YNL)/(AEO.P), where yL stands for house- Individuals earn labor income from the use of their endow- hold total labor income, yNL represents household total non- ment of productive "labor assets," such as physical capabil- labor income, AE are the number of equivalent adults in the ity, human capital, and connections. This section deals household, 0 is a parameter for consumption economies of mostly with one of these assets, formal education. Although scale, and P is the price index for the bundle consumed by education is certainly very important as a determinant of the household. Differences in well-being among individu- income, its central position in the literature is also based on als, which are approximated by differences in equivalized its observability in surveys and censuses. This contrasts with household income, depend on differences in each of the fac- the difficulty of obtaining statistics for other relevant income tors in this equation. Since differential prices are not stud- determinants, such as natural ability, labor market connec- ied in this paper due to a lack of information and 0 is tions, on-the-job training, and work ethics. 57 INEQUALITY IN LATIN AMERICA Table A.22 shows the average number of years of formal increase in years of education in the bottom quintile of the education for adults aged 25-65 by income quintile and by income distribution. age and gender group for each country and year.53 There are Another remarkable fact evident in table A.22 is the significant differences across countries in the average years reversal of the gap in years of education between men and of education. Although in Argentina, Chile, and Panama that women. In all Latin American countries, men older than 50 average is around ten years, in Guatemala, Honduras, and have more years of education than women of the same age, Nicaragua the corresponding figure is less than six years. but the difference is in favor of women in the 10-30 year These differences hold for all income quintiles, although age bracket."5 For the working age population (aged 25-65), the gap between the Southern Cone and the rest of Latin years of education are slightly greater for women in some America is wider for the poorest quintile. countries (including Argentina, Brazil, Jamaica, Panama, One remarkable phenomenon in table A.22 is the sub- Trinidad and Tobago, Uruguay, and Republica Bolivariana stantial increase in the average years of education in all Latin de Venezuela) and somewhat higher for men in the rest of American countries during the 1990s, which continued a the region's countries. process initiated decades ago. In most countries, education In table A.23, people are divided according to age and increased in all income groups, except in Argentina and Peru, household income quintile. For most countries, the gap in where years of education among adults from the poorest years of education between the top and bottom quintiles is households decreased. Figure 2.7 shows that the gap in years wider for young adults than for older people, which suggests of education between the bottom quintile and the top increasing educational inequality in the last few decades. For quintile increased not only in these two countries, but also instance, in Bolivia in 1999 the gap was 7.8 years for people in more than half of the countries in the sample."4 The edu- aged 51-60, 8.5 years for people aged 41-50, and 9.1 for cation gap has widened in Brazil and Mexico from already individuals in their 30s. This trend is a sign of an unbalanced large values by Latin American standards, despite a sizable increase in education in Bolivia that is also present in most FIGURE 2.7 Difference in average years of education between top and bottom quiffhles, per country Years of schooling 8- E| Around 1990 M Around 2000 7- 6- 4- 3 4'J +0 9t9 Source: Authors' calculations based on microdata from household surveys. 58 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA Latin American countries in the sample. Figure 2.7 shows have achieved nearly 100 percent literacy. However, the that only in Chile and Mexico is the educational gap between percentage of illiterate people is still very significant the poor and the rich substantially lower in the cohort aged among the poor, reaching more than 30 percent in several 3 1-40 than in the cohort aged 51-60. Table A.23 also shows countries. that during the last decade, the gap in years of education In the analysis of the labor market that follows, the adult between top and bottom quintiles for youth aged 21-30 has population is classified into three educational groups not shrunk in almost all of the Latin American countries. This according to years of education: low (less than 8 years of is also the case for children aged 10-20 in most countries, schooling), medium (9-13 years of schooling), and high with the exception of Brazil, Chile, Ecuador, Mexico, and (more than 14 years of schooling). This roughly corresponds Panama. to unskilled, semiskilled, and skilled workers.7 Table A.28 Recently, effort has been made to gather educational infor- shows the shares of each educational group in the adult pop- mation from most countries in the world. Table A.24 sum- ulation. The share of adults with high levels of education marizes data from an updated version of Barro and Lee (2000). has significantly increased in all the Latin American coun- All figures in the table correspond to adults over 25. Results tries in the sample, especially among women. for the 1990s are in general consistent with our estimates. The Latin American average for years of schooling is almost Hourly wages: Returns to labor assets at the level of the world mean and has also increased in the Investing in education usually pays in the labor market, espe- last four decades at rates similar to the world mean. cially in terms of higher hourly wages. Looking at the dis- Thomas, Wang, and Fan (2002) calculated Gini coeffi- tribution of hourly wages is important for two reasons. On cients over the distribution of years of education for 140 the one hand, most of the differences in average earnings countries for the period 1960-2000. Data for Latin among educational groups are due to differences in hourly American countries and regional averages are reproduced in wages and much less to differences in hours of work or unem- table A.25. Educational Gini coefficients for the region are ployment rates. The second reason the distribution of hourly close to the world mean, higher than in the developed coun- wages is relevant is because of normative issues. If people tries and Eastern Europe, slightly lower than in Asia, and were completely free to choose hours of work and family size, significantly lower than in Africa. This ranking has not sub- the distribution of hourly wages (and some inherited assets) stantially varied in the last four decades. Table A.28 shows would become the primary concern in an equality-of- the results of computing educational Gini coefficients from opportunity scenario of social fairness. the current sample of household surveys. The Southern Cone Table A.29 shows the Gini coefficients for the distrib- countries (that is, Argentina, Chile, and Uruguay), Jamaica, ution of hourly wages for different groups of workers. Most Panama, and Trinidad and Tobago have the lowest inequal- of the conclusions drawn from previous tables hold here. ity levels. The distribution of hourly wages has become more unequal Educational Gini coefficients fell for most of the Latin in Argentina, Peru, Uruguay, and Republica Bolivariana American countries during the 1990s. This result is not de Venezuela, but also in countries like Colombia and inconsistent with the widening gaps illustrated in figure 2.6. Nicaragua where household income inequality did not sig- In the case of Brazil, as indicated in table A.22, between nificantly change. In contrast, hourly earnings inequality 1990 and 2001 years of education increased from 1.9 to 3.0 decreased in Brazil. For the rest of the countries, changes in the bottom quintile and from 8.9 to 10.4 in the top have been small and the levels usually depend on the group quintile. The absolute difference in years of education for which inequality is computed. between the rich and the poor has increased (as is shown in The literature has stressed the relevance of studying wage figure 2.4), but the ratio has decreased. This latter effect is gaps among educational groups as main determinants of captured by the Gini coefficient, a measure of relative rather inequality in hourly earnings. The "wage premium" earned than absolute differences among individuals6.5 by skilled workers has always been a major topic of interest Table A.27 shows an alternative rough measure of edu- for labor economists and other social scientists. Table A.30 cation: the self-reported literacy rate by income quintile. shows wage gaps among three educational groups. For Most countries have made substantial progress and some instance, in Brazil in 2001, the hourly wage in the primary 59 INEQUALITY IN LATIN AMERICA job of a skilled worker was on average 6.5 times more than secondary school implied a wage increase of 40 percent over that of an unskilled worker. All numbers in the table the earnings of a worker with only a primary school educa- are higher than 1, meaning that more educated workers on tion; in other words, the marginal return of completing average have higher hourly earnings. The wage premium for secondary school versus only primary school is 40 percent. skilled workers increased in most countries during the 1990s, The wage premium for a college education is an additional even in economies in which household income inequality did 76 percent. In Argentina, returns on primary and secondary not significantly change or even decreased, such as Brazil (see school education have not significantly changed in the last also figure 2.8). In contrast, the wage gap between semi- decade. In contrast, there was a large jump in returns on col- skilled and unskilled workers (column (iii) in table A.30) lege education (from 54 to 76 percent). That jump was also did not significantly increase in most countries and noticeable for working women and for both male and female decreased in some (see figure 2.9). urban salaried workers. Although not of the magnitude of The figures in table A.30 are unconditional means. In the changes seen in Argentina, the marginal returns to col- order to investigate further the relationship between edu- lege education increased in several other Latin American cation and hourly wages, the authors ran regressions of the countries during the 1990s, particularly for urban salaried logarithm of hourly wage in the primary job on educational workers, confirming the results from the unconditional dummies and other control variables (that is, age, age means (see figure 2.10).59 squared, regional dummies, and an urban/rural dummy) for Mincer equations are also informative with regard to two men and for women.58 Table A. 31 shows the results of these interesting factors: the role of unobservable variables and the Mincer equations. For instance, in Argentina in 2001, a male gender wage gap. The error term in the Mincer regression worker between 25 and 55 years old with a primary edu- is usually interpreted as capturing the effect on hourly wages cation degree earned on average nearly 22 percent more of factors that aren't considered in household surveys, such than a similar worker without that degree. Completion of as natural ability, connections, and work ethics. An increase FIGURE 2.8 Hourly wage gap between skilled and unskilled workers, men aged 25 to 55 Difference in log wages 7.0 EC Around 1990 M Around 2000 6.0 - 3.0 - 4.0 -~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2.0- 1.0 1 0.0 Source: Authors' calculations based on microdata from household surveys. 60 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA FIGURE 2.9 Hourly wage gap between semi-skilled and unskilled workers, men aged 25 to 55 Difference in log wages 3.0 CM Around 1990 M Around 2000 2.5- 2.0- 1.5- 1.0 II 0.5 - 0.0 Source: Authors' calculations based on microdata from household surveys. FIGURE 2.10 Increase in hourly wages for college-educated workers, men aged 25 to 55 Change in log wages 0.9 | E Around 1990 E Around 2000 0.8- 0.7- 0.6 -t .0.5- 0.4 0.3 Source: Authors' calculations based on microdata from household surveys. 61 INEQUALITY IN LATIN AMERICA in the dispersion of this error term may reflect an increase The coefficients in the Mincer regressions are different in the returns on such unobservable factors in terms of hourly for men and women, indicating that they are paid differ- wages (Juhn, Murphy, and Pierce 1993). Table A.30 shows ently even when they have the same observable character- the standard deviation of the error term of each Mincer equa- istics (that is, education, age, and location). To further tion. The returns on unobservable factors have clearly investigate this point, the authors simulated the counter- increased in Argentina and Repiublica Bolivariana de factual wage that men would earn if they were paid like Venezuela, while there were either no clear changes or reduc- women. The last column in table A.32 reports the ratio tions in the rest of the Latin American countries. between the average of this simulated wage and the actual Another way of investigating the influence of factors other average wage for men. In all cases this ratio is less than one, than education on inequality is to compute the distribution reflecting the fact that women earn less than men even when of hourly wages within each educational group. Table A.33 controlling for observable characteristics.60 This result has does this for prime-age males. Again, Argentina and two possible interpretations: it is either the consequence of Republica Bolivariana de Venezuela stand out from the rest discrimination against women or the result of men having of the countries due to significant increases in inequality more valuable unobservable factors than do women (for within all educational groups. In most countries, inequal- example, that they are more attached to their work). It ity increased among skilled workers and did not increase (or seems that the gender wage gap has shrunk in all countries even decreased) for others. Chile provides one of the clear- during the last decade (see also figure 2.11). Brazil has the est examples of this trend, since the distribution of hourly widest gap between men and women, while in Argentina, wages has become significantly more equal for workers with Colombia, Mexico, and Peru the gap is narrower than in low and medium levels of education and more unequal the rest of the region. Chapter 3 in this report focuses more within the skilled group. on this aspect. FIGURE 2.11 Labor market outcomes by gender Hourly wage ratio, women to men 1.0 |Z3 Around 1990 M Around 2000 0.9 0.8- 1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 0.7 - 0.6 I-- Note: The ratios are derived from regressions that control for worker characteristics. Source: Authors' calculations based on microdata from household surveys. 62 DIFFERENT LIVES. INEQUALITY IN LATIN AMERICA The use of labor assets: Hours worked, unskilled and the skilled with regard to unemployment has participation, and unemployment become larger over the last decade (see figure A.7), imply- Are the differences in hourly wages reinforced by differences ing a disequalizing effect on the income distribution. Labor in hours worked? Table A.34 suggests that the opposite is patterns differ among countries. While unemployment true. Correlations between hours worked and hourly wages increased in South America, it either decreased or remained are negative and significant in all countries. Also, the second low in the rest of Latin America. Finally, the last panel of panel shows that in most countries, workers with low levels table A.36 reports shorter but increasing spells of unem- of education tend to work longer hours than workers with ployment for the unskilled. more years of formal education. However, this gap is nar- rowing in most Latin American countries, with hours worked Nonlabor income decreasing for the unskilled and increasing for the skilled Income from nonlabor sources, once a primary object of inter- during the last decade (see figure A.6). See box 2.3 for the est for economists, is not at the center of inequality studies description of a set of decompositions that sheds additional today. This is due in part to the increasing relevance of labor light on the structure of inequality in hourly earnings. as a main income source, and in part because of the diffi- So far, this discussion has focused on the group of workers culties of getting reliable information on nonlabor income in the labor market. However, some able-bodied people may sources. decide not to work or may not find a job even after actively Column (ii) in table A.21 shows the share of nonlabor looking for one. Table A.36 contains basic statistics on employ- income in Latin American household surveys, while column ment, unemployment, and duration of unemployment by edu- (x) presents the Gini coefficient for the distribution of that cation and gender. People with a college education participate variable. Nonlabor income comprises capital income, prof- in the labor market more than other groups. In addition, the its, rents, and various types of transfers (that is, both pri- employment rate for men is much greater than for women. vate and public, including pensions). The coverage of There does not seem to be a clear pattern of changes in the nonlabor income sources varies greatly among surveys. employment gap between skilled and unskilled workers. Dif- While, for instance, some countries include detailed ques- ferences in employment rates among educational groups have tions on income from capital and rents, some include only increased in some countries and decreased in others. a general, vague question and others don't broach the sub- The unemployment rate for the unskilled is higher than ject at all. for the skilled in more than half of the countries in the Capital income, land rents, and profits are highly con- sample. Also, in most countries the difference between the centrated in the richest stratum of the income distribution BOXe - A rough but illustrative way of investigating what is men and women are negligible, but instead that these behind a distribution is to perform simple decompositions. differences are very small compared to differences in wages A popularion is divided into groups according to a given within each of the two gender groups. variable, and toEal inequaliry is expressed as a combina- The role of location (urban or rural) varies across coun- tion of inequality between and within groups. Table A 15 tries. For instance, the wage gap between urban and rural shows the results of performing decompositions of the areas accounts for I1 percent of overall wage inequality in Theil inequality index compured over the distributLon of Bolivia and just 2 percent in Repuiblica Bolivariana de wages tor working adults aged 25-55 The table suggests Venezuela In many Latin American countries, the rele- that even rhough formal education is an important factor vance of urban-rural dif'erences has decreased during the in accounting for differences in wages in all Latin American last decade. countries, the roles of gender and age are small. This does See also Wodon and others 2000 for similar decompo- not imply that, for instance, differences in wages between sitions with a smaller sample. 63 INEQUALITY IN LATIN AMERICA (see table A.37). An increase in the share of these income transfers relative to public income support programs.64 This sources (for example, an increase in the rate of return on pattern, however, seems to be changing, since the share of capital) may imply a disequalizing change in the income the bottom quintiles in total cash transfers has been increas- distribution. Inequality trends and differences across coun- ing during the last decade. tries can then be accounted for by differences in the share In some countries, the main item in nonlabor income is of these nonlabor sources. Unfortunately, capital income pensions. In Argentina, Brazil, Panama, and Uruguay, pen- and land rents and profits are seriously underestimated in sions account for more than half the value of nonlabor income household surveys. In nearly all countries, the share of all (see column (v) in table A.2 1). Pensions are also concentrated of these income sources is 2-4 percent (see column (iii) in in the upper-income strata, although to a smaller degree than table A.21).61 Given this minor role, neither the reported is capital income (see table A.37). level nor the changes in household income inequality dis- Land is one of the most important assets in agrarian cussed in this chapter are driven by capital income. As men- economies such as those in Latin America. Table A.38 repro- tioned in section 2.2, the inequality that can be measured duces data on the distribution of operational holdings of agri- on the basis of household surveys essentially stems from cultural land assembled by the United Nations Development differences in labor income and demographic factors. Programme (UNDP) (1993) and Deininger and Olinto The previous discussion suggests that if good estimates (2000) (who in turn mainly used the decennial United of individual capital income and profits were available, the Nations Food and Agriculture Organization World Census measured level of inequality would be a good deal higher. of Agriculture). The data do not include adjustments for soil What about trends in inequality? According to the United quality, land improvements, or communal tenure arrange- Nations National Account Statistics, the Latin American ments and refer to operational rather than ownership dis- countries' average share of nonlabor sources in the GDP did tribution. Deininger and Olinto (2000) highlight the fact not significantly change during the 1990s. Harrison (2002) that the distribution of land is more concentrated than the shows that during the period of 1960-1997, the average distribution of income, and also that the variation across labor share in Latin American countries was almost constant countries is higher than that of income. Again, Latin (or slightly decreasing in some cases). America emerges as a very unequal region compared to other In a recent paper, Gollin (2002) finds that the labor share, regions in the world.65 Table A.39 reproduces the Gini coef- when appropriately measured to include those workers who ficients for operational holdings of agricultural land in Latin are self-employed or employed outside the corporate sector, American countries. does not vary much across countries or time periods.62 On Housing is probably the main asset that most people own. the other hand, there does seem to be some short-run volatil- Several household surveys in Latin America report whether ity in shares that are possibly associated with crises (see chap- a house is owned by the family who lives in it, although very ter 8). The evidence of no significant changes in the share few report the property or rental value of the dwelling. Table of labor and nonlabor sources increases confidence in the A.40, derived from the sample of household surveys used in reported inequality changes obtained from household sur- this report, presents for each income quintile the share of veys that mostly ignore capital income. However, it is clear families owning a house (that is, both the building and the that more effort and resources should be devoted to improve lot). Housing ownership is widespread along the income dis- the measurement of capital income, rents, and profits in Latin tribution. In fact, in several countries the share of poor people American household surveys.63 who own a dwelling is higher than the corresponding share Transfers are an important component of nonlabor for the rich. However, figure A.8 shows that in most coun- income. People receive private and public transfers, the latter tries in the last decade, housing ownership among the rich in the form of cash subsidies and in-kind programs such as has grown relative to the poor. Poor families live in houses free education and health. The third panel of table A.37 that are smaller (in terms of number of rooms) than do rich shows the distribution of cash transfers, excluding pensions. families. Since poor families also tend to be larger in size, Perhaps surprisingly, the distribution is pro-rich, meaning the number of persons per room is significantly greater. Dif- that higher shares of total transfers go to high-income strata, ferences across income quintiles have not significantly varied probably as a consequence of the greater relevance of private over time in most Latin American countries. 64 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA Fay, Yepes, and Foster (2002) find that the distribution across the income strata. The exception is Argentina, where of housing values is more unequally distributed than income the number of children under 12 in the bottom quintile in Chile and Peru. They find that in the last decade, hous- increased between 1992 and 2001. In most Latin American ing markets have increasingly excluded the poor, a conclu- countries, the ratio of the number of children in the bottom sion that with different intensities is also evident in and the top quintiles has increased in the last decade, thereby Argentina, Bolivia, Brazil, Colombia, Honduras, Jamaica, contributing to higher income inequality (see figure 2.12).66 Mexico, and Uruguay. Table A.42 shows household size by equivalized income quintiles and by education of the household head. The results Family size and structure are similar to those in the previous table. Countries differ The resources available to each person depends on the with regard to average family size and the gap between poor number of people with whom an individual has to share total and rich families. Most show a similar pattern of falling household resources. Size and composition of the household number of persons per household in each part of the income are therefore key determinants of an individual's economic distribution, with the exception of Argentina and Uruguay, well-being. Table A.41 shows the number of children under where poorer families have become larger. 12 years of age by parental income quintiles and by educa- Inequality is reinforced if marriages take place between tion of the household head. The table reveals significant persons with similar income potential. Table A.43 presents differences in means, with the Southern Cone nations, some simple linear correlations that suggest the existence of Jamaica, and Panama being the areas where families are of assortative mating in all Latin American countries (see also smaller size. All nations have experienced substantial chapter 6 and Fernandez, Gunerm, and Knowles 2001). Men reductions in the number of children per household during with more years of formal education tend to marry women with the last decade. In most, reductions have been generalized a similar educational background, as shown in column (i). FIGURE 2.12 Ratio of number of children under 12 years of age per household between bottom and top quintiles Ratio 2.0 - |3 Around 1990 - Around 2000 1.8- 1.6- 1.0 1. Source: Authors' calculations based on microdata from household surveys. 65 INEQUALITY IN LATIN AMERI( A BOX 24 Social capital and tust Social capital has been broadly defined as the set of infor- results for the period 1996-2001, they found that inter- mal rules embedded in social relations and society's insti- personal and institutional trust increase with age and sub- tutional arrangements that enable members to achieve jective income and decrease with education and percept ion their individual and community objectives (Coleman of income inequality (see table A.46).c 1990). Social interactions, in particular repeated interac- tions, produce obligations and expectations among indi- This box was primarily written by Matfas Busso. viduals that generate trust among people, in turn easing a. The literature has found that higher levels of social capital are associated and promoting cooperation and participation. This idea with higher economic growth (Knack and Keefer 1997), increased judicial efficiency, lower government corruption (La Porta and others 1997), and has spurred a growing body of research that tries to assess improved local government efficiency (Putman 1993). the influence of social capital on a wide range of economic b. The question used to construct the measure of ineterpersonal trut is: "Can and political outcomes.a you trust most people?" If the person answers affirmatively, then trust is given Measuring social capital is not an easy task. Most of the a value of 1; otherwise it is given a value of zero. The s.ri.,l-.l in-ritii,.r-r/ Mrut was constructed as a simple average of eight (qut.ic.n' LUt.nubar.r.-etrr literature at the country level has relied on the World asks"Howmuchcanyoutrustthek.IlI-',int -.rtfiniz,cti,n' (,C,trrrrertr. Values Surveys, which include microdata for 30 mostly congress, the judiciary, the church, the military, police, political parties, and developed countries in response to questions about inter- TV." To construct the index of institutional trust, the aurhors .irb'rrilui assign a value of 1 if respondents say "a lot," 0.66 if they say "something," personal trust and civic engagement. Using Latino- 0.33 if the answer is "a little," and 0 if it is "not at all." bar6metro surveys, the authors of this report constructed c. The results of multivariate regression analysis, controlling for country- two measures of social capital: a measure of interpersonal fixed effects, confirm the unconditional results shown in table A.46. trust and a measure of institutional trust.b Averaging out This is one of the factors that contribute to a positive corre- Finally, some rough estimates of the distribution of social lation of hourly wages within couples, as shown in column (ii). capital-and of the related concepts of interpersonal and There is no evidence of changes in the degree of assortative institutional trust-are reported in table A.46 and discussed mating in the last decade, according to these simple statis- in box 2.4. tics. Finally, columns (iii) and (iv) show positive, though small, correlations in hours of work, both when considering and 2.6. Inequality beyond income excluding people who do not work. Most of the empirical studies that aim to measure the fair- Concern over child labor has increased recently. Table ness of social arrangements focus on the distribution of indi- A.44 shows the proportion of working children between 10 vidual welfare. However, in the real world people seem to and 14 years of age. Unfortunately, many surveys do not care also-and probably especially-about the distribution report labor statistics for younger children. While child labor of particular goods and services. It is likely that more people is negligible in some countries, it is a strong phenomenon would support programs that seek to guarantee equality in in others. There are significant differences in child labor basic education and health care than programs aimed at across income strata. reducing income or total consumption inequality. This spe- In some countries, household surveys report the race cific egalitarianism is further supported by normative argu- or ethnicity of individuals. Based on that information, ments based on the idea of equality of opportunity (see Tobin table A.45 presents the ratio between nonwhites and whites 1970 and Roemer 1996, among others). This section pro- with regard to wages and household per capita income. In vides statistics on inequality distribution for some of the vari- all countries for which information is available, nonwhites ables for which people show particular concern: school earn less and are poorer than whites. This is discussed in more enrollment, basic health status and services, political repre- detail in chapter 3. sentation, safety from crime, and some basic social services 66 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA (for example, the provision of clean water). Having a basic of inequities in access to education. The authors estimated level of these variables is often seen as a right, and hence these conditional probabilities from logit models of the inequality is viewed as particularly disturbing to the extent attendance decision, using parental equivalized income, age, that it impinges on the realization of this right. gender, location, and parental education as independent vari- ables.69 The Gini coefficients for the distribution of these School enrollment conditional probabilities for different age groups are Guaranteeing equality of access to formal education is one reported in table A.48 for each country and year. The higher of the goals of most societies. The authors have used the the Gini coefficient, the higher the differences in the prob- sample of household surveys to calculate school enrollment ability of attending school among children of the same age, statistics by income strata in order to compute inequality after controlling for the independent variables. These dif- measures of school attendance and to investigate the issue ferences can be due to parental income, but also to parental of educational mobility that links the education of children education, location, and gender, all of which are implicitly and youth to that of their parents. considered here to be unacceptable sources of differences in Table A.47 shows school enrollment rates by equivalized access to education. income quintiles. The table indicates, for example, that Inequality in the probability of being enrolled in school among Brazilian children aged 6-12 in the bottom quin- is low for kids aged 9-12 (see column (i) of table A.48). tile, 70 percent attended school in 1990 and 93 percent in Countries that have nearly achieved full enrollment natu- 2001. For the top quintile, 96 and 99 percent attended rally have a Gini coefficient close to 0. Chile, and especially school in 1990 and 2001, respectively. These numbers reflect Brazil, have experienced large decreases in this measure. three important phenomena: (1) attendance rates increase Column (ii) shows the Gini coefficient for the distribution along with household income, (2) enrollment rates have of conditional probabilities of attending secondary school for increased over time for all quintiles, and (3) the gap in atten- youth aged 15-17 who finished primary school. Gini coef- dance rates between poor and rich children has significantly ficients have been falling over the last decade in many coun- narrowed during the last decade. tries. One notable exception is Brazil, where good results in Are these results applicable to other age brackets and primary school are not replicated at the high school level. countries? The first result is quite general: in all countries, Table A.48 also shows substantial differences in the Gini schooling rates increase along with income. The differences coefficients across countries, from a negligible 4.5 in between poor and rich are smaller for children of primary Argentina to 26.2 in Brazil. Column (iii) shows similar sta- school age and larger for youth of college age. Differences tistics for high school graduates aged 19-21. are also very large for children under five years of age.67 The Inequities are in general higher in college than in high second result is also quite general: enrollment rates have school, even when the analysis is restricted to those youth increased over time among all income distribution points who completed the previous educational level. In two-thirds in nearly all countries. On average, increases have been of the countries, dispersion in the distribution of probabil- greater in pre-primary school, followed by high school, col- ities of attending college for those who finished high school lege, and finally primary school, where several countries are increased during the period considered. The last column in close to achieving nearly universal schooling. table A.48 (also shown in figure A.9) summarizes inequal- The third result-the shrinking gap in enrollment rates ities for all educational levels. The Gini coefficient for the between the poor and the rich-is quite general for children distribution of conditional probabilities of attending college under 12 (see figure 2.13) 6R However, the gap has widened for all youth aged 19-21 significantly fell in half of the Latin in some countries for youth aged 13-17 and has become American countries in the survey, and increased or did not larger in most countries for youth aged 18-23. Differences significantly change in the rest. in college attendance between the poor and the rich have increased throughout the region over the last decade. Educational mobility The level of inequality in the distribution of conditional The analysis of schooling decisions is closely related to the probabilities of attending school can be viewed as a measure topic of educational mobility. The concept is simple: if family 67 INEQUALITY IN LATIN AMERICA FIGURE 2.13 School enrollment rates by age and country a. 3 to 5 years old b. 6 to 12 years old Difference between top and bottom quintiles Difference between top and bottom quintiles 06- 0.6 - 0.5 -0.5 0.4- 0.4- 0c3 - 0.3 - 0 2 0.2 0.1 0.1 '05 0.0 °$ 4e3' '' ., ,CS i'& . '., c. 3 c tol7ayearsnold d. o8to 23 years old Difference between top and bottom quintiles Difference between top and bottom quintiles 0.6 -0.6- 0.5 0.5- 0.4 -0.4- 0.3 -0.3- f0.2 0.2 ' .01 0.1 f 0.0 0.0 4 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 6 m Around 1990 M Around 2000 Source: Authors' calculations based on microdata from household surveys. background explains a child's opportunities, then social that is, the degree to which parental education and income mobility is low. As we have seen in previous sections, during determine a child's education. The dependent variable is the the 1990s many Latin American countries experienced an schooling gap, defined as the difference between (1) years of increase in income inequality. High inequality is often seen education that children would have completed had they as less worrisome when it is combined with high social mobil- entered school at a normal age and advanced one grade each ity than when it occurs in a context of low social mobility. year, and (2) actual years of education. In other words, the Ideally, mobility for opportunities or for living standards schooling gap measures years of missing education. The would be computed. However, surveys in Latin American Educational Mobility Index (EMI) is defined as 1 minus countries do not have long panels to allow this step. This the proportion of the variance of the school gap that is section therefore follows the methodology developed in explained by family background. In an economy with low Andersen (2001) to provide estimates of educational mobility, mobility, family background would be important and 68 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA thus the index would be small (for technical details see children. As expected, inequalities in health exist to the dis- Andersen 2001). advantage of the poor. Latin America stands out as a region Table A.49 shows the EMI for teenagers (aged 13-19) of relatively good average health status measures, but also and young adults (aged 20-25) in all Latin American coun- as a region of high inequality (see also Wagstaff 2001). The tries in the sample. Educational mobility is relatively high concentration index for the region is, in absolute terms, in the Southern Cone and the Caribbean. There have not larger than the world mean for both under-five mortality and been substantial improvements in mobility in the region underweight children. Some countries, such as Bolivia, as a whole (see figure A.1O). In many countries, the EMI Brazil, the Dominican Republic, and Peru have very high has not significantly changed over the last decade, while levels of inequality. in others it has decreased for both age groups. Only in Under-five mortality among Peruvians in the top quin- Brazil and Panama are there unambiguous signs of higher tile is lower than in the countries in East Asia included in mobility. the DHS program (that is, Indonesia, the Philippines, and Vietnam). In contrast, under-five mortality in the bottom Health quintile is higher in Peru. Health inequality measures for Inequality in health is causing growing concern in both the prevalence of diarrhea (see panel B) are also relatively high public policy arena and the academic literature. This con- in Latin America, with Peru and Brazil again standing out cern has translated into a better understanding and mea- as particularly unequal countries. In contrast to the inequal- surement of disparities in health status and services ity statistics on health status, inequality in services such as indicators. The Demographic and Health Surveys (DHS) immunization, basic antenatal care, and attended delivery program is the main initiative in gathering information on in Latin America do not seem higher than in other devel- a large number of health variables, as well as data on oping regions of the world. At any rate, the disparities are respondents' demographic, social, and economic charac- worrisome. In Peru, while a medically trained person attends teristics (see . nearly all deliveries of babies in the top quintile, that Table A.50, developed with information from that pro- proportion is only 14 percent for babies in the bottom gram, presents statistics for different health status measures quintile. and health service indicators for each Latin American coun- Wagstaff and Watanabe (2000) computed measures of try in the sample and for the average of other regions in inequality for stunting, underweight, and wasting, working the developing world. These are presented by quintile with a sample of 20 countries and ranking individuals by of socioeconomic status, defined in terms of the ownership equivalized consumption.70 Latin American countries in the of assets in the household rather than as income or sample systematically appeared at the top of the inequality consumption. rankings. The most negative concentration index for stunt- Along with the statistics for each quintile, each panel of ing and underweight was in Peru and for wasting in the table shows the concentration index (CI), a measure of Nicaragua. Using consumption as a welfare indicator, the extent to which a particular variable is distributed Wagstaff (2000) reports that inequality in under-five mor- unequally across the income strata (see Lambert 1993). tality is particularly high in Brazil compared to other coun- "Bads," like child mortality, are usually more common tries in the world. among poor households. In this case, the CI is negative. The A related topic that is also attracting increased attention higher the CI in absolute value, the more concentrated the in public debate and from policymakers across Latin America "bads" are in the poor households. In contrast, "goods," such is the extent and distribution of physical and mental dis- as immunization, are more frequent in the richer percentiles. abilities. Because of conceptual issues, sampling frames, and In this case, the CI is positive. The higher the CI, the more reporting difficulties, the true extent of disability is seldom concentrated the "goods" are in the most affluent households. easy to infer from standard household surveys. Nonetheless, In sum, if a region has values of CI close to zero, health progress has recently been made with regard to both the inequalities are assessed to be relatively low. measurement and understanding of disabilities in Latin Panel A of table A.50 provides statistics on two measures America-and of the relationship of these aspects with dis- of health status: under-five mortality and underweight tribution and poverty (see box 2.5). 69 INEQUALITY IN LATIN AMERICA BOX 2 i DbabIt and dhIIbutIn The linkage between poverry and disabiliry is strong, with disability must incorporate the physical and cultural con- evidence suggesting that causaliry goes in both directions. text in which they live. and so also includes a list of envi- Poverty can cause disability through malnutrition, poor ronmental factors. health care, and dangerous living conditions. Disability can In 2001, following the ICF methodology, the Brazilian cause poverry by preventing the full participation of dis- census broadened the scope of questioning, including ques- abled people in rhe economic and social ac[ivities in their tions on degrees of impairmenr (including mental, phys- communities. especially if the proper support systems and ical or motor, sight, and hearing aspects). Some preliminary accommodations are not available. As many as 15-20 per- findings include: cent of the poor in developing countries have disabilities. One problem in studying the link between income. * More women than men are disabled. mainly because poverty, and disability is rhe difficulty in obtaining high women live longer and disabiliry is more common quality data, especially data thar can be used ro make com- among the elderly. Fifty-four percent of disabled parisons across countries. The rate of disabilitv found in people in Brazil are women. household surveys and censuses varies dramatically. This * Disability rates vary across racial groups. More than variation resulrs from differing measures of disability, dif- 17.5 percent of blacks have a disability, compared ferent data collection rechniques, and different reactions to 13.8 percent of whites. to survey questions by respondents. For example, the rate * Disability is linked to poverty. The poorest areas in of disability in Paraguay was measured as I percent, but Brazil present higher shares of disabled people: in in Uruguay and Brazil It was measured as 16 percent (see the ten stares with the highest rates of disability, table A.51). It is unlikely that a common definition and eight are located in the Northeas. one in rhe Center- comparable quality data would have yielded such disparate West and one in the North. Moreover, 40.1 percent rates of disabiliry across reasonably similar, neighboring of disabled people earn less than minimum countries In fact, a 1991 census in Bra2il measured the wage, compared ro 29.6 percent of nondisabled rate of disability at 0.9 percent. The higher rate measured people. in 2001 was not the result of an explosion of disability in * Disability is negatively correlated with education. Brazil, but of the use of better measurement techniques. and thus with permanent incomes. The incidence What is disability? Disability refers to a long orshort- of disability falls from 22 percent among those term reduction in a person's activity resulting from an with no formal education, to 20 percent among those acute or chronic condition. Table A.52 shows the break- with 1-3 years of schooling, to 16 percent among down of the disabled population by type of disability, in those with 4-10 years of schooling, and to 15 per- countries with available data. But disability goes beyond cenr among those with more than 10 years of the description of a particular medical condition. schooling. Disability is multi-faceted, complex, and difficult to define. In fact, it is best understood as the interaction A deeper analysis of data and broader data collection between a condirion and a host of personal, social, and methods across the region are necessary in order to explore environmental factors. The International Classification of further the relationship between disabiliry and the distri- Functioning, Disabiliry, and Health (ICF) developed by butions of education and income. Grearer understanding the WX'orld Health Organizarion (WHO) explicirly recog- should pave rhe way to betrer policies that can help break nizes that any evaluation of a person's functioning and the links between poverty and disability. 70 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA Political representation Di Tella, Galiani, and Schargrodsky (2002) use a survey Disparities in income and wealth interact with disparities specially designed for victimization analysis and conclude in representation and political influence. More influential that in Buenos Aires, "although high-income households groups tend to have higher economic rents, while wealth- used to suffer a significantly higher home victimization rate ier individuals tend to have more political influence. Sta- than low-income households, the difference has now turned tistics on inequality in political influence are naturally very nonsignificant. For street robberies, both groups show sim- hard to obtain. In recent papers, Samuels and Snyder empha- ilar augments in victimization." Fiszbein, Giovagnoli, and size legislative malapportionment as a measure of inequal- Ad6riz (2002) use a national household survey in Argentina ity in representation (see for instance Samuels and Snyder and find that the rate of households that reported being a 2001). Malapportionment is a discrepancy between the share victim of crime or violence in the previous six months is not of legislative seats and the share of population within elec- significantly different along the income distribution. Crime toral districts, and implies a failure to uphold the "one is also a large problem in some countries in the Caribbean, person, one vote" rule. such as Haiti and Jamaica (see, for instance, World Bank, Table A.53 shows measures of malapportionment in 2003f). both chambers of congress or parliament in several Latin American countries. A score of x percent means that x per- Basic services cent of seats are allocated to districts that would not receive Tables A.46 and A.48 report statistics by income strata on those seats in cases of perfect apportionment. The table access to some basic services, including water, hygienic suggests that malapportionment is significantly higher in restrooms, and electricity.73 The size of the gaps between Latin America than in the rest of the world.71 Note, how- the poor and the rich in terms of these services varies widely ever, that although malapportionment is linked to inequal- across Latin America. Figure 2.14 is informative with ity in political representation, it does not necessarily imply regard to these gaps, which tend to be larger for hygienic a bias against the poor. There is clearly a need for more restrooms than for electricity and water, where coverage is empirical work in this field. Issues of power and political more widespread. At any rate, in several countries the dif- representation are discussed in more depth in chapters 4 ference in the fraction of households with direct access to and 5. water or with electricity in the house is more than 30 per- centage points between the top and bottom quintiles. Safety from crime Figure 2.14 shows that the gaps in the access to water, Safety from crime is one of the top concerns of the Latin hygienic restrooms, and especially electricity have American population. The available evidence suggests that significantly narrowed in most countries in the last decade. the region has the highest rates of homicide and crime vic- Information on access to a telephone is also included, timization in the world (see for instance Shrader 2001). The although arguably it is not a basic social service.74 Access probabilities of being victim of a crime are not uniform along to a telephone has dramatically increased in the region, the income distribution. Although rich people are a more especially for the wealthiest households. valuable target to criminals, they also have better means with which to protect themselves against crime. 2.7. Conclusions Recent studies have tried to assess whether crime has a This chapter has presented and analyzed statistics on different higher impact on poor people than on rich people. Since dimensions of inequality in Latin America, setting the stage national victimization surveys are rarely available, researchers for the following chapters of this report. As discussed above, use a variety of other sources. Gaviria and Pages (1999) use a data set of household surveys from 20 Latin American coun- Latinobar6metro surveys from 1996 to 1998 to compute tries was assembled and used to compute statistics on the crime victimization across quintiles of a socioeconomic index distribution of a wide set of variables. Results drawn from based on the ownership of durable goods and household char- various authors complete the picture of inequality presented acteristics. Table A.54 reproduces these results, generally in the preceding pages. indicating that victimization moderately increases with The authors emphasize the need for improvements in wealth in all countries.72 data collection. The measurement of living standards in the 71 INEQUALITY IN LATIN AMERICA FIGURE 2.14 Gaps in access to services, by country a. Water b. Electricity Difference between top and bottom quintiles Difference between top and bottom quintiles 0.80 0.80 0.70 0.70 - 0.60 - 060 -. 0.50 0.50 0.40 0 0.40- 0.30 0.30 0.20 0.20 0.10 -0.10 0.00 0.00 4' ~~~~~~~~~~~~4' ~ ~ ~ ~ ~ ~ ~~~O c. Hygienic restrooms d. Telephone Difference between top and bottom quintiles Difference between top and bottom quintiles 0.80 0.80 0.70 -0.70- 0.60 -0.60- 0.50 -0.50- 0.40 -0.40- 0.30 -0.30 020 0.20 0.1 0 0.10 0.00 0.00 t*6Otw t t~~~~~~b d4 N2,4 *@o+ /4C /i |3 Around 2000 m Around 1990 | Note: Water refers to the availability of a source of water in the house or lot. The variable restroom is equal to 1 when the household has a restroom with a toilet connected to the sewage system or to a septic tank. The variable electricity includes all sources of electricity, while telephone includes fixed and cellular phones. Source: Authors' calculations based on microdata from household surveys. region has some shortcomings that blur the picture of The existing data, however, are useful to provide at least inequality and hamper the possibilities of conducting preliminary answers to relevant questions on inequality comparative analysis. Governments and international levels, trends, and structure. The tables provided in the sta- organizations should move toward standardization of ques- tistical appendix provide useful information on the distri- tionnaires across countries (without losing valuable idio- bution of household income, as well as on labor, educational, syncratic components), improvement in capturing nonlabor and demographic variables by country and year. and nonmonetary income, generalization of consumption Income inequality has increased in Latin America since surveys, and introduction of long panel surveys. Data avail- World War II. During the 1990s, the trends were not uni- ability is currently particularly poor in the Caribbean, where form across countries. On average, inequality has increased recent household surveys were completed in only 3 out of in South America and remained stable in Central America, 15 countries. the Dominican Republic, and Jamaica. Two paradigmatic 72 DIFFERENT LIVES INEQUALITY IN LATIN AMERICA cases of changes in inequality are the neighbors Argentina Notes and Brazil. Argentina, once a very low-inequality country 1. See Le Grand 1991 and Roemer 1996 and 1998 for overviews by Latin American standards, has experienced dramatic dis- of this debate. equalizing changes in recent decades. In contrast, inequal- 2. See Deaton 1997 and Deaton and Zaidi 2002 for arguments ity has fallen significantly in Brazil, the most unequal supporting the use of consumption as the best welfare indicator. country in the region. The evidence presented in this chapter 3. In fact, according to the life-cycle theory, with full access to suggests a movement toward convergence in the country- capital markets current consumption should be closely associated to by-country inequality levels for the whole region. Latin interremporal living standards. American economies, already quite homogeneous in terms 4. See Buhmann, Rainwater, Schmaus, and Smeeding 1988 and of inequality, are becoming even more uniform. Deaton 1997 for discussions on these points. The available evidence suggests that Latin America is 5. Only the Dominican Republic, Jamaica, and Trinidad and and has been the most unequal region in the world. Dif- Tobago are included in the sample. The quality and frequency of house- ferences in inequality with respect to the rest of the world hold surveys in the Caribbean are significantly lower than in the main- have not significantly changed in the last 50 years. The land Latin Amenrican countries. For instance, although household surveys assessment of Latin America as a high-inequality region in Guyana and St. Lucia are available, the authors were not able to is not restricted to income. The same conclusion arises compute consistent household income statistics for these countries. when considering the distribution of consumption, land 6. For reference purposes, the authors have worked with some sur- holdings, health status, and some measures of political veys that cover only urban areas in Bolivia, Colombia, and Paraguay representation. in the early 1990s. This chapter has also analyzed other dimensions of dis- 7. See IDB 1998, Londoiio and Szekely 2000, Szekely and Hilgert tribution beyond inequality. Fueled by growth in GDP, 1999 and 2001, and Szekely 2001. poverty measures fell and aggregate welfare increased in most 8. Whenever a trade-off arises, the author has generally decided Latin American countries during the 1 990s. However, to preserve comparability within a country over time rather than across increases in inequality have reduced the positive effects of countries. growth: in several Latin American countries, the assessment 9. For some countries, income definitions have varied over time. of the performance of the economy is less optimistic when Although statistics for alternative definitions have not been computed distributional issues are considered. for the sake of breviry, and in most cases the tables provide a single line Although there are many contributing factors, household for each country and year. For instance, although the survey in El Salvador income inequality computed from household surveys is in for 2000 includes nonmonetary income, in the tables statistics are given particular the result of an unequal distribution of education, without those incomes in order to preserve comparability with previous combined with high wage premiums for skilled workers and surveys in that country. Alternative results for El Salvador, as well as for decreasing fertility rates due to improvements in income and the Dominican Republic, Honduras, Paraguay, Peru, Uruguay, and education. The intensity of these three factors seems to have Republica Bolivariana de Venezuela are available upon request. The main increased in most Latin American countries during the results reported in the paperdo not vary in any of these cases because alter- 1990s. native income definitions are considered. During 2001, Argentina was Finally, this chapter examined the information on non- in a deep recession. For reference, tables A.2 and A.3 also include staris- income measures of inequality, including health status, tics for 1998, when the economy was still growing. The surveys excluded crime victimization, political influence, and access to basic some areas of Bolivia in 1992, Colombia in 1992, and Peru in 1991. For services. In some cases, notably in health status, the region reference, all the statistics for Bolivia in 1996, Colombia in 1996, and is highly unequal by international standards. This may also Peru in 1994 were computed using, alternatively, the whole national be true of political influence, but cross-country information survey or only the observations from rhe areas covered in the early 1990s. is weaker in this area. Basic services are typically unequally The label regional in Peru refers to all regions covered in the 1991 survey distributed, but less so than with regard to income, while (which excluded Costa Rural, Selva Urbana, and Selva Rural). the weak information on crime victimization that exists 10. DataforNicaraguaincludenonmonetaryincome,whiledatafor suggests that this aspect may be at least relatively equally Costa Rica and El Salvador do nor. However, ignoring nonmonerary distributed, payments from the Nicaraguan survey does not significantly alter the 73 INEQUALITY IN LATIN AMERICA results of the inequality comparison among these three countries (see significant changes between 1995 and 1998 and report an increase also table A.7). between 1998 and 1999. On the other hand, Gonzalez (2001) finds 11. The richest individual in the household survey of Mexico in a drop of 1 point in the Gini coefficient between 1998 and 1999. 2000 has an income 18 times greater than the median individual in 23. The Gini coefficient actually fell around 1 point, which is just the top decile. That gap (18 times) separates the median individual within the limit to be considered a nonsignificant change from a in the top decile from a person in the poorest second decile of the over- statistical point of view (at a 95 percent confidence level). all income distribution. This is an example of the long "upper tail" 24. Latinobar6metro is an annual survey of public opinion that of the distributions. started in 1995. Data is gathered in 17 Latin American countries. 12. See Lambert 1993 and Cowell 1995 and 2000 for details on 25. OneLSMSsurveyisavailableforpartsofBrazilfor 1996-1997, these inequality indices. Guatemala for 2000, Guyana 1992-1993, and Panama for 1997. 13. The Gini coefficient ranges from 0 (complete equality) to 100 26. See Chen and others 1995, World Bank Development Indica- (all national income concentrated in the hands of one individual). It tors Database (1998-2000), and various World Bank Poverty Reports is also common to present that coefficient in terms of the [0,11 inter- on these countries. val, instead of the [0,100) interval. 27. Readers interested in technical details can consult Esteban and 14. There is room for debate as to the appropriate value of these Ray 1994, Wolfson 1994, Esteban, Gradfn, and Ray 1999, and a note measures. For a case for a lower value of 8, see Lanjouw and Ravallion by Marfas Busso on the Web site of this study. (1995). The best practice is to test the robustness of any conclusions 28. The next step in the research agenda would be to consider mea- against different assumptions, but space did not allow presentation of sures of polarization with more than two groups. the results. 29. These include one proposed by Sen (equal to the mean times 15. Even ignoring the most recent economic crisis, the inequality 1 minus the Gini coefficient) and two proposed by Atkinson (CES increase is very large (around 5 Gini points between 1992 and 1998). functions with two alternative parameters of inequality aversion). See 16. The survey in Venezuela for 1989 is not strictly comparable Lambert (1993) for technical details. with 1995 and 1998, since it does not include nonlabor income and 30. The sources for GDP figures are World Bank 2000b and World nonmonetary payments. However, ignoring these incomes in 1995 and Development Indicators Database. 1998 does not significantly modify the results. For instance, the Gini 31. See Gasparini and Sosa Escudero 2001 for a more complete coefficient for the distribution of household per capita income in 1995 justification of this kind of study. goes from 46.9 with all income sources to 46.7 with only labor mon- 32. Note that in Panama the share of the bottom deciles increased, etary income. leading to a fall in inequality indices with greater weights in that part 17. The survey frame changed significantly in the Dominican of the distribution (for example, the Atkinson Index with a parameter Republic between 1995 and 1997, making the results of the com- equal to 2). parisons difficult to interpret. 33. The estimates range from 37 percent (Londofio and Szekely 18. See Fields 1989 for Brazil, Gonzales-Vega and Cespedes 1993 2000) to 53 percent (Sala-i-Martin 2002). for Costa Rica, and Altimir 1994 and 1996a for the remaining countries. 34. This study does not attempt to compute poverty statistics with 19. See Altimir 1986 and Gasparini, Marchionni, and Sosa Escud- our sample of household surveys because the World Bank has recently ero 2001, among others, for documentation of similar inequality trends issued a report on poverty in Latin America and the Caribbean (Wodon in Argentina. and others 2001) and frequently produces Poverty Reports in several 20. This pattern is hardly attributable to informational problems, countries in the region. such as the urban coverage of the household survey. More than 85 per- 35. For reviews on the usual limitations of household surveys for cent of Argentineans live in cities and significant migratory movements distributional analysis, see Deaton 1997, Gottschalk and Smeeding have not occurred in the last three decades. 2000, and Atkinson, Brandolini, and Smeeding 2002. 21. This result is in line with those reported by ECLAC 1996a, 36. See chapter 2 and table A.l in the statistical appendix for more Morley and Vos 1997, and Robles 1999. details on the PNAD survey. 22. A decrease of nearly 3 points in the Gini coefficient between 37. For example, Gasparini and Sosa Escudero (2001) found that 1995 and 1999 was identified, even after eliminating an extreme the measured increasing trend in income inequality in Argentina is outlier in the 1995 survey. Szekely and Hilgert (1999) do not find robust when the three types of adjustments mentioned above are made. 74 DIFFERENT LIVES: INEQUALITY IN LATIN AMERICA 38. The implementation of the bootstrap method here follows Sosa increased in the 1990s, while the Gini coefficient for household income Escudero and Gasparini 2000. For more theoretical references on the remained unchanged. subject, see Biewen 2002, Davidson and Duclos 2000, and Mills and 53. Educational systems differ across countries and sometimes also Zandvakili 1997. over time within countries. See the Web site of this study for details on 39. Results for other variables and indices are available from the the construction of educational variables. The variable years of education author upon request. are recorded in most surveys. For those in which it is not, we have esti- 40. Tabatabai (1996) at the International Labor Organization also mated it from the maximum educational degree attained by people and made an independent effort to put together distributional statistics their age. Years of education are truncated at a maximum of 17. for many countries in the world. 54. This fact would have been even more noticeable if years of 41. Secondary data sets have some problems that were recently education had not been truncated at 17. reviewed by Atkinson and Brandolini (2001) and Atkinson, Brandolini, 55. Bolivia and Guatemala are the two clearest exceptions to this and Smeeding (2002). pattern. 42. Deininger and Squire (1996), for instance, highlight the 56. The Gini coefficient as most of the inequality indices, is scale- "familiar fact that inequality in Latin America is considerably higher invariant (see Lambert 1993). than in the rest of the world." 57. In addition, this roughly corresponds to (1) completion ofpri- 43. The Gini coefficients are taken from the UNU/WIDER-UNDP mary education or less, (2) completion or noncompletion of secondary World Income Inequality Database. All countries with at least one education, and (3) at least some higher education. observation in the period 1991-1999 with the quality rating "reli- 58. See Wodon and others 2000 and Duryea and Pages 2002 for able data" are included. When several observations are available for a estimates of returns on years of education in several Latin American given country, the most recent data point has been used. countries. 44. Africa is not included in this graph, since there are not enough 59. Nicaragua and Peru experienced changes in the returns on observations on income inequality in that region. skilled labor similar to those in Argentina. 45. The data set includes observations for Bolivia, Ecuador, Guyana, 60. The only exception is Mexico in 1996; nonetheless, the coeffi- Jamaica, Mexico, Nicaragua, and Peru. This set of countries does not cient is not significantly different from 1. have an average Gini coefficient for the household per capita income 61. Only Chile has a share higher than 10 percent, which may be distribution that is significantly different from the overall mean for the consequence of a better survey design. This higher share does not Latin America (that is, it is just 0.4 Gini points higher). seem to account for the high inequality level in Chile, which is still 46. For instance, in the sample Brazil has 15 observations from one of the highest in Latin America when the analysis is restricted to 1960 but observations end in 1989. Being a country with very high the distribution of labor income (see table A.7). inequality, omitting Brazil in the 1990s reduces the regional value for 62. Bernanke and Gurkaynak (2002) replicate and update Gollin's that decade and biases the results for the decadal changes. (2002) calculations for a larger sample of countries. They find that labor 47. The exception is the gap with Eastern Europe, a region that shares in Latin America are higher than those computed directly in suffered strong distributional transformations in the 1990s. the UN National Account Statistics. Moreover, the heterogeneity 48. Forinstance,forthecaseofLatinAmerica,BourguignonandMor- among Latin American countries seems to be lower. rison (2002) consider five "countries" (Brazil, Mexico, Argentina/Chile, 63. The need for more reliable information on capital income has Colombia/Republica Bolivariana de Venezuela/Peru, and a group of 37 been emphasized again as recent studies have cited the distribution smaller countries). of nonlabor assets as a key determinant in income distribution, income 49. Most statistics are obtained from Maddison 1992, Deininger mobility, and growth. See Birdsall and Londonlo 1997, Deininger and and Squire 1996, and Altimir 1996b. Squire 1998, and Deininger and Olinto 2002. 50. Notice that this is not international inequality within a region 64. The only exception to this pattern is Chile. obtained by averaging (with or without weights) national levels of 65. Cardoso and Helwege (1992) report that the largest 7 percent inequality. of land holdings in Latin America account for 77 percent of the land. 51. This fact is also highlighted in IDB 1998. See also Thiesenhusen 1995 for more evidence. 52. Two exceptions are Colombia and Nicaragua, where the Gini 66. The absolute difference in the number of children has also coefficient for the distribution of individual labor income significantly increased in many countries (see Table 5.20). 75 INEQUALITY IN LATIN AMERICA 67. In some countries, statistics refer to schooling only for children 72. The question in Latinobar6metro used for the study is: "Have 5 years of age or older, since no information is recorded for younger you or any member of your family been assaulted, robbed, or victim- children. See the Web site of this study for details. ized in any way during the past 12 months?" 68. Of course, it could be the case that the gap in attendance rates 73. Water refers to the availability of a source of water in the house narrows down, but the gap in the quality of education becomes larger. or on the building lot. The variable restroom is equal to I when the 69. This analysis follows Gasparini 2002. household has a restroom with a toilet connected to a sewage system 70. "Stunting" is used to describe a condition in which children or septic tank. The variable electricity includes all sources of electricity. fail to gain sufficient height given their age. The term "wasting" refers Some definitions and classifications differ among surveys, so compar- to a situation in which children have failed to achieve sufficient weight isons should be made carefully. See the Web site of this study for more given their height. details on the definitions for specific countries. 71. Samuels and Snyder 2001 show that this tesult holds when con- 74. The variable telephone includes fixed and cellular phones. trolling for institutional variables. 76 CHAPTER 3 Group-Based Inequalities: The Roles of Race, Ethnicity, and Gender L ~~ATIN AMERICA IS A RACIALLY ANT) ETHNICALLY DIVERSE REGION, WITH MEN AND WOMEN, AFRO- AND Euro-descendents, and indigenous and nonindigenous people actively pursuing higher living standards for themselves and their families. However, levels of well-being are not equal across races or ethnic and gender groups. The extent of inequality in well-being and the reasons for such differences have been the focus of various regional and country studies. The most notable study on indigenous groups, by Psacharopoulos and Patrinos (1994), concludes that indigenous people are systematically poorer than nonindigenous people and that this pattern can be largely traced to fewer "endowments"-education, job experience, family structure, and occupation-as well as to other causes, such as differential preferences, institutions, social relations, cultural norms, and discrimination. In terms of gender analysis in the Latin American region, Psacharopoulos and Tzannatos (1992), find that across the region, women's earnings and labor force participation are consistently lower than men's and can also be traced to "endowment" and other effects. Regional studies on race are scarce, with country-level studies of Brazil making up the bulk of the research. This chapter investigates the extent to which inequality The following chapter has four sections in addition to the within and between racial, ethnic, and gender groups con- current introduction. Section 3.1 describes circumstances related tributes to national inequality. Standard measures of inequal- to gender, race, and ethnicity in Latin America and considers ity are applied to racial, ethnic, and gender subgroups and some conceptual issues encountered in analyzing these demo- these results are aggregated up to the national level. This graphic groups. Section 3.2 uses descriptive statistics to out- chapter will focus primarily on wage inequalities and inequal- line the extent of inequality within and between the different ity in endowments that impact wages. It draws on sociolog- demographic groups (defined as white and Afro-descendents, ical and anthropological work, but the bulk of the information Indo- and Afro-Guyanese, and indigenous and nonindigenous is based on economics and derived using economic tools. depending on the countty), within which the gender variable Although race, ethnicity, and gender are all examined in this cuts across racial and ethnic group lines. Inequality through- chapter, this approach does not suggest that the underlying out the lifecycle is examined in terms of education, sector of factors that lead to group-based inequalities in any one of employment and occupation, household formation, wealth and these dimensions necessarily are responsible for the observed access to services, and labor income. Section 3.3 uses econometric group-based inequality with regard to the other dimensions. tools to clarify how differential returns on characteristics and Instead, the patterns and underlying factors that lead to dif- differential levels of education, fertility, and sector of work con- ferences according to race, ethnicity, and gender need to be tribute to the observed inequality between individuals' labor considered and understood separately for each category. incomes and to the contribution of race, ethnicity, and gender 77 INEQUALITY IN LATIN AMERICA to national inequality. Finally, section 3.4 provides concluding TABLE 3.1 remarks. Dtibution of Afro-descended, indigenous, and female populaions in the Americas and the Caribbean, various year 3.1. Who are the people of Latin Amienca? (pent) Table 3.1 shows that an estimated 17-30 percent of the Country Afro-descendents' Indigenous Femaleb Latin American population identify themselves as Afro- descendents, with Brazil having the largest concentration of Latin America such citizens. Some 44.7 percent of Brazilians identify them- Argentina - 1.0' 51 Bolivia 2.0 71.0c 50 selves as being black or of mixed race (preto or pardo), which Brazil 44.7' 0.4e 51 is the equivalent of nearly 80 million people. The next largest Chile - 8.0' 50 Colombia 25.09 i.8f 51 Afro-descendent populations in Latin America are in Costa Rica 2.0 0.8a 49 Colombia (25 percent), Nicaragua (13 percent), and Ecuador 10.0 38.0' 50 Republica Bolivariana de Venezuela (10 percent), while the El Salvador - 7.0' 51 Guatemala - 66.0' 50 majority of the populations of most Caribbean countries Honduras 5.0 15.0' 50 identify themselves as Afro-Creole. More than 80 percent of Mexico 0.5 14.0e 51 the citizens of Haiti, Barbados, Grenada, Guadeloupe, Nicaragua 13.0 5.0' 50 Panama 73.5 1o.Oe 50 Jamaica, St. Vincent and the Grenadines, and St. Lucia claim Paraguay 3.5 1 5e 50 African descent. Peru 9.7 47.0' 50 Uruguay 5.9 0.4e 51 Although indigeneity rates have been historically much RB de Venezuela 10.0 0.9' 50 higher, only 10 percent of the population of Latin America Caribbean identify themselves as indigenous today (see box 3.1 and Barbuda and Barbuda ~ 97.9 -52 table 3.1). These rates vary strongly across countries: the Bahamas 85.0 - 51 majority of Bolivians and Guatemalans are identified as Barbados 95.8 - 51 Belize 57.0 19.0d 49 indigenous, comprising 71 and 66 percent of the national Bermuda 61.3 - 52 populations, respectively. Peru (47 percent) and Ecuador Cuba 62.0 - 50 Dominican Republic 84.0 - 49 (38 percent) also have sizable indigenous populations. French Guiana 42.6 40d 49 Indigenous populations in the Caribbean countries are rarely Grenada 84.0 - 50 quantified, but when they are they appear as a small pro- Guadeloupe 87.0 - 51 Guyana 42.6 6 0d 52 portion of the population (table 3.1). Haiti 100.0 - 51 With the exception of Belize, Costa Rica, French Jamaica 91.4 2 0d 51 Guiana, and the Dominican Republic, one half or more of St. Viuncent and 90.3 51 the population of Latin America is female (table 3.1), with Grenadines 95.0 - 50 women making up the majority in nearly one-half of the Suriname 41.0 6 od 50 Trinidad and Tobago 43.0 -50 countries. It is not unusual for women to make up a higher Other proportion of a population than men, since women tend to Canada 2.21 1. d 50 United States 12.3 o ge 51 have higher longevity and are less prone to die prematurely United States from violence (Jacobsen 2002). - not reported. Note: Table includes data from 2000 for Latin America and the United States, except: Afro-descendents from 1992 for Bolivia, The evolution of race, ethnicity, and gender Guatemala, Paraguay, Peru, and Venezuela; 1990 for Ecuador as concepts and Panama; 1995 for Nicaragua; and 2001 for Canada. Indigenous from 1999 for Canada. Data are for 2001 for Ethnicity and race Colombia, 1999 for Honduras, and 1997 for Uruguay. For the Caribbean, all Afro-descendent data are for 1992 and all The unique colonial history of Latin America forms the indigenous data are for 1999. The gender data are for 2000. basis for the conceptualization of race and ethnicity and Sources: aFlorez, Medina, and Urrea 2001 for all figures except: bUNDP 2000, 'Mahoney and Vom Hau 2002, dDeruyttere 2001. related policies today (see also chapter 4). Sociologists eNational census data, fxNorld Bank 2003a, gVice-presidencia de argue that in Latin America, the population of European la Republica de Colombia 2002. 78 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER BOX 3.1 Dkbtbom of Om popul of tUr Amecas: A historical ewkihon The racial and ethnic composirion of the Americas changed United States and Canada were considered nonindigenous. substanrially t'ollowing the arrival of the Europeans. The The share of African and indigenous descendencs dropped graph below shows that in 15'0, nearly 100 percent of by 1935 as Euro-descendent popularions in the Americas the Americas was populated by people identified as continued to grow rapidly and the assimilarion of indige- *indigenous," but by 1825, nearly 40 percent of Spanish nous people into European-style institutions continued. America, 80 percent of Brazil, and 98 percent of the The changing ethnic composition of the Americas Percent 100 80 60 40- 20- 0 F 1570 1650 1825 1935 1570 1650 1825 1935 1570 1650 1825 1935 Spanish America Brazil U.S. & Canada m Indigenous m Afro-descendent E3 White] Source: Engerman and Sokoloff 1997. descendents strove for ethnic unity in order to build Given the history of the development of racial and ethnic national unity that could counter the newly arriving identity in Latin America, racial and ethnic differences have colonists and help to control the nonwhite population often been denied in favor of the claim that countries in the through assimilation with the primarily indigenous (rather region are "post-racial," that is, a single mixed community than Afro-descendent) populations (Agier 1995). Privately, (Graham 1990). Racial inequality that is recognized has been however, individuality has been emphasized and mobility treated either as the result of an incomplete process of has been seen as a result of "self-whitening" and wealth national integration or as an expression of class-based rather than movement of a group as a whole; race and class inequality. With this mindset, differences between racial or status were thus linked. ethnic groups have been attributed to socioeconomic 79 INEQUALITY IN LATIN AMERICA influences that happened to be correlated with, but are not of power relations that create an uneven playing field by set- caused by, power structures (Torres-Saillant 1998). Discus- ting up a situation in which endowments, opportunities, and sions on race and ethnicity have been combined with those expectations differ by demographic group (Baiocchi 2001). on poverty and class, thereby ignoring many specific aspects These theories are becoming salient as ethnic and racial groups that contribute to intergroup differentials. It has also been increasingly reshape their cultures and raise their voices for claimed that Latin America was more racially tolerant than equal rights and a greater share of opportunities in their coun- the United States, for example, with its history of legalized tries (Arocha 1998). This shift has also led to increased segregation, suppressed ethnic and racial identity, and rights demands and efforts to include ethnic and racial groups in movements. This claim has formed the basis for an ideol- national statistical exercises (see box 3.2), and to address the ogy of racially and ethnically blind national policies focused difficult identification and measurement issues that arise in on poverty reduction. those contexts (see box 3.3). This conceptualization is, however, changing as racial and Due to the nature of colonization in the Caribbean, racial ethnic disparities in well-being are increasingly acknowl- identity evolved differently than the way it did in the rest edged in academic and policy circles and reflected in social of Latin America. Racial lines were more starkly drawn under movements. In the mid-1970s, scholars began to argue that, colonial rule, a factor that-partly due to the absence of despite ideologies of integration, differences in well-being plantation owners-facilitated the maintenance of the between demographic groups are partly based on racial and white population largely separate from the ethnic and Afro ethnic discrimination. More recently, sociologists have the- populations. This physical distance in turn fostered a social orized that racial or ethnic inequalities are a result of a history distance between whites and blacks. Planters identified with BOX '.2 Todos Contamos: Na_i cens ai soea h mslon 1 have a problem with this census questionnaire, I am not A follow-up workshop, Todos CojJtamnoi II: National included', said Jos6 Chala after listening to a presentation Crniuj andSoca/l Incluiton. was held in Lima, Peru in Octo- on Ecuador's 2000 census Mr. Chala was referring Io ber 2002. I[ brought togerher more than 100 representa- absence of data on Ecuadorans of African descent. His tives from 18 countries throughout Latin America tO take status in this census is hardly unique. Not all countries stock of progress made. discuss lessons, explore method- in Latin America and the Caribbean include a caregory to ologies, and share examples of how census questions can account for those of African descent, which limits a gov- be further improved. At the end of the conference, each ernmen's ability to address inequaliry and to inform poli- participating counrry revealed a plan of action that cies and programs oriented toward poverty reduction and included six key aspects: 1 I Incorporating racial and erhnic social inclusion. questions inro national censuses and refining existing ques- The World Bank and other multilateral organizations tions; (2) integraring in a systematic manner the question have been working with governments in Latin America to of race and ethnic self-idenrification in national household rectify rhis shortcoming. In November 2000, rhe gov- surveys and poverty analyses; (3) tostering greater par- ernment of Colombia, wirh rhe sponsorship of interna- ticipation by ethnic groups in data collection, (4) dis- tional organizations, held rhe first Todos Contamo; seminating information that is already available: ("Everyone Counts") conference to raise awareness arnong (5) implementing capacity-building programs to help demographers, government census personnel, and indige- translate data into social inclusion-oriented policies: and nous and Afro-descended individuals and organizations in (6) promoting national Tudor Coniarnos conferences and sup- Latin America, about rhe need to include questions on race porting country-to-country learning exchanges Progress and ethnicity in national censuses. By 2002, 10 countries has been made as Latin America slowly moves toward a in the region included a racial or ethnic self-identification better understanding of who really makes up the region's question in their censuses. population. 80 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER England much more than hacienderos (large estate owners) did Despite these gains, gender disparities in various mea- with Spain, with the result that planters in the Caribbean sures of well-being still exist and new challenges continu- did not develop a sense of national identity. Instead, a ally face societies. Women still lag behind men in most stronger sense of subordinate racial identities developed over indicators of well-being-including wage gaps, control over time. Once the fight for independence did ensue, the black resources, political voice, and health care-but meanwhile political elite became stronger and continued to reinforce there are increasingly notable trends of disadvantage among black racial identity in many Caribbean countries (Safa 1998, men, including shorter longevity, greater alcoholism, and Torres-Saillant 1998). higher suicide rates than women (Jacobsen 2002). In Brazil As the racial composition of Latin America continues to and the Caribbean countries, male educational attainment change, another racial and ethnic group has recently estab- rates are lagging and difficulty in obtaining and maintain- lished a presence in Caribbean countries. Migrants from India ing employment is increasing, in turn placing male-oriented have formed sizable Indo-Guyanese and Indo-Trinidadian gender issues firmly on the agenda. populations. These groups continue to grow, integrate into society, and increasingly compete with the Afro-Guyanese The Roots of Difference and Afro-Trinidadians for economic, social, and political Observed differences between subordinate and dominant positions in their adopted countries (Safa 1998). groups are commonly attributed to discrimination (broadly construed), but the process of creating these differences Gender involves a complex interaction between individual choice, While racial and ethnic variables are newly recognized as opportunities, and the institutions with which individuals important factors in explaining differences in well-being interact throughout their lifetimes. Two insights come from among demographic groups, gender has a longer history of sociological perspectives on the formation and persistence of acceptance as a salient variable that affects the likelihood of group-based differences. (See Heller and Mahoney 2003 for poverty and inequality. Many Latin American countries have a discussion of concepts and processes related to such dif- had strong women's movements that began in the 1970s and ferences.) First, they are fundamentally relational, that is, 1980s, resulting in an ever-increasing presence of women in interactions that occur between groups shape the behaviors, the economy, political arena, and society as a whole. All gov- dispositions and attitudes of both subordinate and dominant ernments in the region now have Women's Offices or groups. In the language of economists, such interactions Women's Ministries. determine "preferences," which should be seen as products Over time, women's well-being has improved relative to of social relations. Second, an important reason for the per- men's, whether because of better opportunities for women, sistence of what Tilly (1999) refers to as durable inequali- changing needs of society that have altered gender roles, or ties, is that they serve organizational purposes, allowing the worsening of men's situations. Maternal mortality has people in workplaces, schools, and other areas of activity to decreased, women's educational attainment is approaching use group-based differences to influence how they interact or surpassing that of men (particularly in the English- with others. As Tilly (1999, p.7) states: speaking Caribbean countries), women's labor force partic- Durable inequality among categories arises because ipation continues to increase, and the wage gap is decreasing over time. In addition, women's involvement in political, pole whoscontrlraccesstovaluproducing resou solve pressing organizational problems by means of cultural, and intellectual realms continues to grow, with par- tiiato rae in seio goenmn inmn.onre categorical distinctions. Inadvertently or otherwise, those people set up systems of social closure, exclusion, exceeding those in the United States and some other Organ- and control. isation for Economic Cooperation and Development (OECD) countries. These successes are not equally experienced by all Generally, the outcomes that can be observed today-such women, however, as those living in urban areas who are not as wages, labor supply, housework time, and access to indigenous or of African descent are making gains at a much services-are the result of a lifelong process of accumulation faster rate than are women in rural areas or those of indige- of experiences, human capital, preferences, and constraints. nous or African descent (World Bank 2000f, 2002c). At each stage of life, decisions are made and events occur that 81 INEQUALITY IN LATIN AMERICA BOX ' , The chalene of racia, ethl, ad gener eaidati ad mea_ement Race, ethniciry, and gender are difficult concepts to mea- for example, counts 23 echnolinguistic indigenous groups: sure due to Ehe social underpinnings of these demographic poverry, inequality, access to services, education. demo- groupings. Individuals are commonly classified into race graphic, and labor force statistics differ greatly amongst these based on phenotype; into ethnicity by language, customs, groups, even though all are classified as Lndigenous (World and dress; and into gender by biology. However, inherent Bank 2002b, Edwards and Winkler 2002). in these definitions are power relationships, behaviors, and The absence of a precise method for ethnic identification expectations that are not easily observed or classified is highlighted inthecontrast berween individual and inter- (Baiocchi 2001) Despite the difficulty in defining race, viewer perceptions of ethnicity This contrast differed more ethnicity, and gender, many countries in Latin America in the classification of indigenous groups than in the iden- have made great strides in identify'ing the demographic tificacion of racial groups. Individuals in a Peruvian survey composition of their populations. All census data now tend to rneitjzar-st (racially mix themselves), rather than identify the sex of respondents, although the identifica- ''whiren themselves. One-half of chose who were identified ion of race or ethnicity is less common.a by (heintervieweras"white self-identifiedasnksf-o(mixed Assignment ot a person to a racial category is difficult. race), while rwo-thirds of those identified by the interviewer Although "race' is based on observable physical charac- as indigenous" declared themselves as metzvzf (see table i.31. terisiics that do not change over a lifetime Ifor example, The idenrification of a person's gender is less contro- skin color, hair type, and facial features), racial mixing over versial since gender and sex are highly correlated variables. the years ha-s led to a continuum of physical characteris- However, they are different concepis. While "sex' is iden- tics in all Latin American countries. For example, table 3.2 tified by those basic biological attribures that different,- illustrates the high degree of disagreement on racial clas- ate males and the females, "gender" is taken to refer to rhe sification among Brazilian individuals and census inter- full ensemble of norms, values, customs and practices that viewers. The racial category self-reporred by individuals are correlated with a parricular sex but contribute to much tends to be whiter' than that identified by the interviewer. wider social differences between women and men (Kabeer Among those individuals identified as "black" by the inrer- 199'4 and Kabeer and Subrahmaniam 2000). When viewer, only 57.9 percent also identified themselves as women take on tradirional "male' roles-such as unmar- black, whereas 39.8 percent replied that they were of ried women who are the sole breadwinners and have no muxed race and the remaining 2.2 percenc identified them- childcare responsibilities-they behave more similarly to selves as whire. In addition, race seems to be strongly tied unmarried men than to women with more traditional toeconomic status,suggestingthatthemeaningof"race" female roles (Cunningham 2000, Bosch and Maloney is i part socially. rather than compleiely biologically, con- 003). The social connotation of'gender" is therefore dif- structed; wealthierclasses tend to be "whiter." This is sup- ficult to capture through a single variable, alihough the ported by classifications of race based on both self-reporting high correlation among sex and gender roles allows tor and interviewer perceptions, where the discrepancy the sex variable to serve as a proxy. btiween white and nonwhire increases with economic These measures of racial, ethnic, and gender classifica- status (Florez, NIedina, and Urrea 2001). tion by statisticians are clearly far from perfect. Thus, any The difficuiry of identification is exacerbated in the case analysis using these measures (including the statistical of eihnicilY, because linguistic or cultural practices, which analysis in this chapter) must take into account measure- may change over a lifetime, are used for classification, as ment errors when using the variables ot .race, 'language opposed to the relatively fixed biological features used for spoken," and "sex' ro caprure the complex biological-social racial identification Even ifa person is identified as indige- concepts of'race, ethnicity, and gender. nous, this is a crude measurement since many distinct indige- Cen sain 1- ci the p5 Urin Ameriixm .ounrr- do nor aik abour rat nous groups are present in Latin American countries. The or ,chnic origin and tcn lever labor force or houaeh,zld sur"(v& ask relared Guatemalan National Living Standards survey (ENCOV, qileions i Flor0z. Niedini. and Lirrea 200l 1l. 82 GROUP-BASED INEQUALITIES: THE ROLES OF RACE, ETHNICITY, AND GENDER TABLE 3.2 Although some roles are based on biology (such as child- Interviewer perception versus self-perception of race-Brazil, birth for women), many are based on assumptions of differ- 1997 (percent) ences between groups (for example, an assertion that white men are more capable than women or nonwhite individu- White Mulatto Black All als), obsolete societal arrangements (women care for the Interviewer classification for all household, indigenous people work in the fields), or an All 56.9 31.7 13.4 assignment of roles used to create or sustain a social hierar- Interviewer versus self-classification chy. This process of molding individual motivations, White 88.6 20.2 2.2 56 preferences, and behaviors is driven by cultural and social Self- Mulatto 11.0 71.0 39.8 33 contexts (Baron and Hannan 1994), thus potentially trans- classification Black 0.4 8.8 57.9 11 lating biological differences into social differences. Total 100.0 100.0 100.0 100 Social differences influence individual choices through- out the lifecycle. In response to unequal social histories and Source: Telles and Lim 1999. life changes, subordinate groups often develop patterns of behavior, practices, and dispositions that influence lower create opportunities for or constraints on the next stage of aspirations, or acquire "constraining preferences." For exam- life. Thus, the reason for observed outcomes today can be ple, the behavior of children as a "girl" or "boy" may be partly attributed to past events, sometimes all the way back partly biological, but it is also largely based on the prefer- to childhood, and can in turn be formed by historical processes ences that a parent or society teaches children based on their of group formation and intergroup interactions. These expe- (biological) sex. This behavior in turn shapes choices of how riences and decisions are affected by how individuals and to behave in school, both of which influence the selection groups view themselves and how dominant groups behave of career, and all three of which affect the choice of family toward them. Such processes occur within the context of the formation and employment. These patterns subsequently formal and, especially, informal institutions that present lead to the contemporaneous differences in such aspects as opportunities or barriers at each stage of formation. wages, labor force participation, assets, and family roles. A From birth, individuals are socialized to conform to their similar argument can be made for children of different ethnic biological, demographic, and social group. Males and or racial groups. females, indigenous and nonindigenous, rich and poor, develop collective expectations and acquire preferences that Discrimination: The costs of engaging with different comply with the rules of behavior, emotions, and expecta- groups tions that are assigned to individuals based on the group to Discrimination is a key factor that influences the accumu- which they belong. Collective identities are constructed as lation of skills and experiences of individuals and groups products of social histories. There is, of course, individual during their lifetimes. It is also part of the relational, his- variation within groups, but that is formed in relation to torically formed interactions between groups. In the eco- group-based characteristics. nomic approach, discrimination is defined as a process of TABLE 3.3 Interviewer versus seH-perception of ethniciy-Peru, 2000 (percent) Interviewer classification White Mestizo Indigenous Other Total White 55 10 8 3 13 Mestizo 45 87 74 55 79 Self-classification Indigenous 0 2 18 1 6 Other 0 1 1 41 2 Total 100 100 100 100 100 Source: GRADE 2001, as cited in Gacit6a-Mari6 and Wodon (forthcoming). 83 INEQUALITY IN LATIN AMERICA unequal treatment of individuals due to characteristics that A key conclusion of this line of thinking is that it is are unrelated to the decision at hand and which lead to dif- impossible to precisely identify the degree to which ferentially observed outcomes. Discrimination is classified employer and worker preferences, employer and worker con- into two broad categories. straints, and institutions lead to observed differential out- comes among individuals from various demographic groups. * Pure discrimination is a differential treatment of indi- However, it is possible to analyze how the outcomes would viduals because they possess characteristics that are differ if some of these cross-group differences were controlled "distasteful" to the discriminator. For example, if an for (see box 3.4 for an example). Although gender, race, and employer is not comfortable working with Afro- ethnic differences are explained through the same model, descendents, the employer will compensate for this dis- they should be understood differently due to the relative comfort by penalizing Afro-descendents at a rate that importance of different factors for each group. is equal to the value of the cost of the unhappiness that This chapter primarily focuses on income and educational the employer feels. This is realized through not hiring inequality and the processes that take place earlier in life that Afro-descendents, hiring them at lower wages, or may be responsible for the observed income differences giving them fewer opportunities than those given to among groups. Thus, the chapter uses data to measure the otherwise observationally equivalent white counter- differences between men and women, indigenous and non- parts (Becker 1971). indigenous people, Indo- and Afro-descendents, and white * Statistical discrimination is differential treatment of and Afro-descended residents in Latin America at each stage individuals based on observable characteristics that are of the lifecycle with regard to educational attainment (child- correlated with certain undesired behaviors, but hood), occupational and sector of employment, family for- which may or may not accrue to the person in ques- mation (youth/early adulthood), and observed contemporary tion. For example, women are more likely than men wages (adulthood). to leave the labor market due to childbearing, so Although the empirical work presented in this chapter employers may offer less training to female than to is representative of the type of approach generally taken to male employees since there is a higher probability that analyze inequality, the approach does not definitively solve they will not be with the firm for as long as otherwise the puzzle of how to attribute the observed differentials to equivalent men. All women in the firm would then the different potential causes. Nevertheless, it does delin- receive less training, even though the original moti- eate the extent of differences between groups and provides vation for lower pay may not apply to certain indi- some clues about how institutions, choice, and discrimina- vidual women. tion may be the source of the observed differences. It is often difficult to disentangle which observed differences 3.2. Inequality among individuals during are due to the behavior of the group adversely affected and the lifecycle which are due directly to discrimination. Of course, no one This section presents differences in levels of education, par- would choose lower wages over higher wages, all else being ticipation in occupation and employment sectors, family equal, but choices may have been made earlier in life that structure, wealth, access to services, and income among dif- are unobservable to the researcher-such as the choice to ferent racial, ethnic, and gender groups. Where possible, select a lower-paying career that is more compatible with gender is considered jointly with race or ethnicity. The objec- family responsibilities rather than choosing a higher- tive of this section is not to compare the distribution of these paying career that allows little personal time-but that affect variables along ethnic or racial lines across countries, as observed wages. Thus, it is not possible to measure the underlying histories and resulting institutions make such degree to which discrimination is responsible for differen- cross-country comparisons meaningless. Instead, the fol- tial outcomes (Becker 1971, Altonji and Blank 1999). How- lowing discussion presents various examples of inequality ever, as noted above, such "choices" by subordinate groups, related to race and ethnicity across Latin America, thus can themselves be products of histories of exploitation and revealing the great range of differences that exist through- unequal conditions. out the region. 84 GROUP-BASED INEQUALITIES. THE ROLES OF RACE, ETHNICITY, AND GENDER BOXi Wat E we do hold Meluse const Men and women make different choices throughour their capital have nor diverged as much by gender as rhat of liferimes that lead to different labor market outcomes. older individuals-with rhe same level of education, the Thus, comparisons by sex are simplistic since they assume labor force participation gap falls to 10 percenr, while the thar men and women respond identically to incenrives and wage gap does nor show a clear trend. When the sample are treated the same in the labor market. The rable below is further limited to only those individuals who are nor shows that by successively holding factors constant married and do not have children (two key variables thar berween men and women, the gap berween men's and particularly affect women's labor force participation and women's labor force participation and wages decreases. performance), the labor force participation gap disappears Women in Argentina, Brazil, and Costa Rica are 30-40 and the wage gap shrinks. In other words, some of the percentage points less likely to work than are men and difference in labor market outcomes are due to personal they earn lower wages. However, when the sample tesred choices made throughout life that are nor directly relared is limited to young individuals-whose human and social to the labor market. Labor mdrket outeomes for dEest gpmps Argentina Brazil Costa Rica Difference in labor force participation rates All men and women 31.5 30.4 39.6 Age 25-35, university education 9.8 8.27 7.9 Age 25-35, university education, -2.34 0 0 no spouse or children Ratio of female to male wages All wage earners 0.98 0.65 0.9 Wage earners, age 25-35, university education 0.88 0.68 0.9 Wage earners, age 25-35, university education, 0.98 0.75 0.94 no spouse or children Source: Authors' calculations. Educational attainment is unequal by race and areas, or it may demonstrate that gender-and the prefer- ethnicity, but not by gender ences and discrimination associated with it-has a greater Men and women have increasingly similar levels of educa- effect on parents' decisions to send their sons or daughters tion in Latin America. This is particularly true in urban areas to school in indigenous households (figure 3.1). By contrast, (see table 3.4), where women's educational attainment women of Afro descent in Brazil and Guyana have a small exceeds men's accumulated education in several countries, advantage compared with men in the same group. especially among younger age groups. While more women Within each country, the white/nonindigenous/Indo than men are illiterate in the age group 55-70 in all coun- populations have higher levels of education than the tries, this same pattern holds true only for youth in the indigenous/Afro-descended groups, but no single education Dominican Republic, El Salvador, and Peru.' level is strictly segregated by race or ethnicity. Indigenous However, when examining the sample according to racial and Afro-descended people are clearly over-represented and ethnic characteristics, gender equality is not universal. among the least educated (figure 3.1) and under-represented The education level of indigenous women in Bolivia and among the most educated. For example, 63 percent of indige- Guatemala lags behind that of men from the same ethnic nous Bolivian women have primary school education or less, group. This difference may reflect the education gap in rural compared to 38 percent of nonindigenous Bolivian women. 85 INEQUALITY IN LATIN AMERICA TABLE 3.4 This illustrates both the education gap by ethnicity and that Ratio of female to male educational attainment and literacy a large proportion of nonindigenous Bolivian women also Female/male illiteracy, have this low level of education. Thus, even though the edu- by age cation system does not strictly leave behind men and women Female/male Year education 20-24 55-70 of only one ethnic or racial group, certain groups are more left behind than others. The single group that is left behind Argentina 1997 - 0.33 2.80 the most is indigenous women, who lag behind white! Bolivia 1999 - 0.99 4.09 Brazil 1999 1.03 0.55 1.24 nonindigenous/Indo-Guyanese women due to racial and Chile 1998 1.00 0.87 1.22 ethnic gaps and behind indigenous men due to a gender Colombia 1998 0.97 0.96 1.62 Costa Rica 1998 0.97 - - factor that does not appear to be as strong among white! Dominican 1996 0.88 1.42 1.45 nonindigenous groups or Afro-descended groups. Republic The explanation for higher educational attainment El Salvador 1997 0.92 1.22 1.75 Honduras 1998 1.00 0.71 1.42 among white, nonindigenous men and women has been Mexico 1999 0.96 - - attributed by various researchers to the following factors: Nicaragua 1998 0.99 0.63 1.70 Paraguay 1999 0.95 0.64 3.00 Peru 1997 0.96 1.13 5.09 The rural-urban divide A collection of studies for Latin RB de Venezuela 1996 1.04 0.45 2.02 America finds that any difference in school attendance and per- formance between rural and urban schools nearly disappears - data not available. . Source: Author's calculations based on microdata from in Ecuador, Peru, and Bolivia when controlling for the loca- household surveys. tion of the school (Garcfa Aracil and Winkler 2002 for Ecuador, FIGURE 3.1 Completed primary education or less, by race, ethnicity, and gender Population share 100 - 80- 60 - 40- 20- 0 --I- Guyana Guatemala Bolivia Brazil Cl Male white/nonindigenous/lndo-Guyanese m Female white/nonindigenous/lndo-Guyanese E Male indigenous/Afro-descended m Female indigenous/Afro-descended Note: Figures are for Guatemala and Bolivia, indigenous and nonindigenous; for Brazil, white and Afro-descended; for Guyana, lndo-Guyanese and Afro-descended. Source: Authors' calculations based on microdata from household surveys. 86 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY. AND GENDER Saavedra and Cardenas 2002 for Peru, and McEwan and Poverty and child labor Poverty is cited as a reason behind Jimenez 2002 for Bolivia). However, Saavedra and Cardenas school dropouts (Gacitua and Wodon, forthcoming, World (2002) caution that there is a great deal of heterogeneity among Bank 2001a), whether due to the direct cost of attending schools within rural or urban areas as well. school (that is, for uniforms, books, transportation, and sup- plies) or to the opportunity cost of school attendance (that Discrimination in schools Is racial discrimination a factor in is, forgone labor market earnings). Since the poor are over- schooling? Oliveira Barbosa (2002) suggest that in Brazil, represented among indigenous and Afro-Latino groups, they racial discrimination is not a factor in teacher evaluations of are more likely to be in this situation. Furthermore, these students, but that social position of the student is. Since children are more likely to hold jobs (Psacharopoulos and Afro-Brazilian students are over-represented among the poor Patrinos 1994). (Henriques 2001), this suggests that racial discrimination is conflated with socioeconomic discrimination. Conversely, Quality of education and available markets Education- girls (not controlling for race) experience positive discrim- particularly beyond secondary school-is pursued largely ination in teacher evaluations. On the other hand, Albernaz, in preparation for the job market. Girls, however, have Ferreira, and Franco (2002) find evidence of Afro-Brazilian another market that they may enter: the marriage market. children doing less well after controlling for social status and Preparation for the marriage market does not occur in school quality. schools, but in the home (Knaul 2001). Thus, girls who live in remote areas where job opportunities are limited, such as Intergenerational transfer of low levels of education Many stud- rural indigenous communities, may find that dropping out ies have found that parents (especially fathers) pass on their of school and learning skills in the home will return a higher low levels of education to their children (Arias, Yamada, and investment than education. Tejerina 2002 for Brazil, Sosa and Marchionni 1999 for Argentina, Garcia Aracil and Winkler 2002 for Ecuador, and Employment sectors and occupations are highly McEwan and Jimenez 2002 for Bolivia). The causes of the segregated strong intergenerational transfer of low levels of education Another factor that affects individual welfare is occupational are unknown, but it may be due to the underestimation by and sectoral choice. Observed variations by race, ethnicity, parents of the value of education, the lower ability of racial and gender in occupations and sectors of work are correlated (and ethnic) minorities to convert education into earnings with differences in earning potential and well-being. (Valle Silva 2001), or a lack of funding for low-educated par- Sectoral allocation is noticeably gender-specific across ents to send children to school. countries. In all three countries considered in table 3.5, a TABLE 3.5 Difference in the probability of being in each sector, by race and by gender (percent) Difference in probabilities, by Difference by race/ethnicity for men gender, for white/nonindigenous Bolivia Guatemala Brazil Bolivia Guatemala Brazil Formal sector employee -3.4 -8.2 -4.5 -3.3 -9.9 -7.5 Informal sector employee -8.1 4.1 5.1 -16.7 -16.0 -1.9 Public sector employee -1.3 2.6 -0.6 -1.3 -2.0 0.6 Self-employed 23.9 7.7 -2.4 -4.7 -9.2 -11.3 Nonincome earner -11.2 -1.0 3.5 26.0 37.1 20.2 Note: A negative value indicates that the probability is greater that a white/nonindigenous male is in that sector relative to an Afro-indigenous male (left columns) or relative to a female of the same race/ethnicity (right columns). Formal sector employee is defined as an individual whose employer pays social security benefits for the worker (Bolivia and Guatemala) or a Brazilian worker with a carteira assinada (signed work contract). Source: Author's calculations based on microdata from household surveys. 87 INEQUALITY IN LATIN AMERICA TABLE 3.6 TABLE 3.7 Share of sector that is female (percent) Occupational disWibution, by gender Labor Formal Self- Informal Domestic worker force wage employed wage (percent) Year Duncan indexb Female Male Argentinae 39 36 36 47 Brazila 43 38 38 56 Argentinaa 1999 29 15 1.0 Chile 33 28 40 - Brazil 1999 41 22 1.0 Costa Rican 40 38 35 50 Chile 1998 31 15 0.2 El Salvador 40 32 54 - Colombia 1998 31 8 0.2 Nicaragua 34 32 29 - Clmi 983 . Paraguay 38 37 38 Costa Rica 1998 41 17 0.2 El Salvador 1997 - 9 0.3 Honduras 1998 36 10 0.6 - not reported. Mexico 31 - - a. Cunningham (2001), data from 1995 for Costa Rica and Nicaragua 1993 - 14 0.3 Brazil, 1997 for Argentina. Formal wage employee is identified Paraguay 1999 - 21 0.8 for Argentina and Costa Rica as an individual whose employer Peru 36 - - pays social security benefits and in Brazil as a person with a carteira assinada (signed work contract). Source: Authors' calculations, 1998, for all except as indicated -not reported. in note a ~~~~~~~~~~~a. World Bank 2001b. b. Percentage of the men and women who would have to trade occupations to create gender equality in all occupations. Source: Authors' calculations based on microdata from much higher proportion of white and nonindigenous household surveys. women than white or nonindigenous men are nonincome earners, meaning that they are not in the labor force or are sent women's share of the labor force. For women and men unpaid workers.2 Compared to the subset of women who do to be equally represented in the labor force, approximately work, men comprise a disproportionate share of the urban 29-41 percent of men and women would have to exchange formal sector, while the other sectors are more mixed. occupations (table 3.7).3 Table 3.6 shows women's share of the labor force and of A higher proportion of women than men are in low- the formal, self-employed, and informal wage sectors; only paying service occupations regardless of racial or ethnic back- in Paraguay does women's participation in the sectors reflect ground. Women also dominate the high-paying managerial their participation in the labor force as a whole. Conversely, field, but here there is more segregation along racial and in El Salvador, 40 percent of workers are women but only ethnic lines. Figure 3.2 shows that white and nonindigenous 32 percent of the workers in the formal sector are women. (Indo-Guyanese) women are particularly engaged in service Even in the modern economies of Chile and Argentina, sector work, and nonwhite women are not far behind. In women are underrepresented in the formal sector by 5 and Bolivia, for example, 28 percent of nonindigenous women 3 percentage points, respectively. On the other hand, women and 15 percent of indigenous women with jobs are are overrepresented in the lower paying informal wage sector, employed in the service sector. This compares to less than comprising more than half of the informal wage workers but 10 percent of men of either ethnic category. less than half of the labor force (Cunningham 2001). One explanation for women's high participation in the service sector is their overrepresentation as domestic servants. Occupation In Argentina, for example, domestic service is the occupa- Men and women are not equally represented across occupa- tional category with the largest proportion of the female tions. Men are particularly overrepresented in the manufac- labor force, at 15 percent (World Bank 2001b). More than turing sector-in which more than 80 percent of jobs are 20 percent of the female labor force is employed in domes- held by men in most countries-while women hold approx- tic service in Brazil and Paraguay, while fewer than 1 per- imately 60 percent of service sector jobs even though they cent of men are employed as domestic workers (table 3.7). comprise only 40 percent of the labor force in most coun- Domestic work is often pointed to as the most vulnera- tries. Other sectors are more mixed and more closely repre- ble of jobs due to its isolated nature, the full dependence of 88 GROUP-BASED INEQUALITIES: THE ROLES OF RACE, ETHNICITY, AND GENDER FIGURE 3.2 Proportion of workers in service or managerial occupations, by group and country a. Service occupations b. Managerial occupations Percent Percent 50 - 20 - 40- 2 0 1 0 1 |, W ] g ~ ~~~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~~~ 5 LL 30- 10 20- 10 0 0 Guyana Guatemala Bolivia Brazil Guyana Guatemala Bolivia Brazil | 1 Male white/nonindigenous/lndo-Guyanese m Female white/nonindigenous/lndo-Guyanese |J Male indigenous/Afro-descended M Female indigenous/Afro-descended Source: Authors' calculations based on microdata from household surveys. the worker on the employer, limits on career advancement, Family formation and the excessive work hours for domestic workers who live The choice of family formation, in terms of spouse, children, with their employers and thus are essentially at their work- and extended family, is an important factor that determines place 24 hours a day, 6 days a week. However, interviews future access to earning opportunities. Perhaps more impor- with domestic workers reveal that there are positive factors tant, however, is the fact that the gender roles associated with associated with this type of work. A sample of Argentine being a wife or a husband, or being a parent or not, influ- domestic workers, for example, report leaving low-paying ence one's work and nonwork life (see box 3.5). formal sector jobs to return to the domestic service sector, Most individuals in Latin America are married with chil- in which employers are more flexible in terms of work hours, dren. More than 50 percent of households are headed by a salary payments, and bonuses than are formal sector couple that is married or in a consensual union with chil- employers (World Bank 2001b). dren present in the household, while another 20 percent are A different pattern emerges in the high-paying man- headed by a couple without any children. Single mothers are agerial sector, in which differences in participation are based household heads in 5-14 percent of Latin American house- more on race or ethnicity than on gender. In Guatemala, holds, while less than 1 percent of households are headed by Bolivia, and Brazil, nonindigenous or white women are most single fathers. Unmarried heads of households account for represented in the sector, followed by nonindigenous or the remaining 15 percent of the sample. white men. The trend deviates in Guyana, where women Women are twice as likely as men to be unmarried with- are more in the managerial sector than men; this is due to out children, partly due to their longevity when husbands men's high participation in agricultural and manufacturing have died and children have moved out of the house, a factor occupations. that in turn results in a large proportion of elderly women 89 INEQUALITY IN LATIN AMERICA BOX * ' men, but Guatemalan households headed by men have Wonml's other job houwwd roughly 4.5 people, regardless of the ethnicity of the house- hold head. Mlarried women's contribution to caring for their own Extended families are more prevalent in households homes and tamilies is equivalent co a full-time job. headed by women, but this household arrangement is not Mtexican women who do not hold jobs in the labor particular to specific ethnic or racial groups. More than two- market spend an average of 49 hours each week taking thirds of male-headed households do not include family care oF their homes and families, while those in the labor members beyond a spouse and children, a proportion that market still spend a substantial amount of time engaged is similar among households headed by indigenous men. In in home care-33 hours weekly. This compares to the contrast, approximately half of households headed by 5.5 hours contributed by husbands each week, with an women include family members besides a spouse and chil- extra 30 minutes of housework added to this figure if dren. This is particularly true of indigenous female-headed wives are employed. Thus, women's work in the home households in Bolivia and Guatemala and in Afro-Guyanese is not simply a specialization of labor," since even when female-headed households. However, female-headed house- women work in the labor market they still undertake holds make up fewer than 30 percent of households in the majority of housework. Guyana and 20 percent of households in Bolivia and Even before arriving in (he labor market, housework Guatemala, making nuclear households the dominant is assigned according to gender. Daughters contribute model in society. 14 hours weekly helping cheir mothers, while sons spend the same time as their fathers (that is, 5-6 hours weekly). Households headed by Indigenous and Afro-descended people have less wealth Averwgeee hs set on h"oso by housed Assets and access to services are proxies for wealth and well- member* being. Assets are an intertemporal measure of wealth since they are accumulated over time and may be converted into If wife employed If wife not employed income at any point. Public services also add value to a Wife 32.8 48.9 household by improving quality of life and decreasing Husband 6.1 5.S demands on household time. Teenage daughter 14.3 14.0 Households headed by white and nonindigenous people Teenage son 6.2 5.5 generally own more assets than those headed by Afro- or *Only households with a married or consensual union indigenous people. Table 3.8 shows the average dwelling size household head, spouse present. per capita (measured by number of rooms) and rates of own- Source: Encuesta Nacional de Empleo Urbano, Mexico, ership of an automobile, refrigerator, and home. Dwelling 1999. area per capita is larger in female-headed households, which partly reflects the smaller average number of people living who care for themselves. A broad breakdown of this struc- in such households. In addition, houses inhabited by white ture shows that household structures generally do not differ or nonindigenous families are larger than those inhabited by across racial and ethnic group lines. Differences are greater Afro or indigenous families. However, Afro-Brazilians are according to the gender of the head of household: male- marginally more likely to own their homes than are white headed households are approximately one person larger than Brazilians. In terms of ownership of durable goods, house- female-headed households. It turns out that this is not due holds headed by nonindigenous men and women are more to differences in the numbers of children, but rather reflects likely to own both automobiles and refrigerators, and male- the missing male partner in female-headed households. headed households are more likely to own automobiles. Across the four-country sample, no single race or ethnicity Inequality by race rather than gender is most notable in has larger families. Afro-Brazilian and indigenous terms of access to services. In the three countries considered Guatemalan men have households that are a half-person in table 3.9, households headed by white or nonindigenous larger, on average, than those of white and nonindigenous people have greater access to water, sewage systems, electricity, 90 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER TABLE 3.8 Asset ownership, by race, ethnicity, and gender (percent) Rooms per capita Own home Auto Refrigerator Guatemala Nonindigenous male 0.58 - 20.4 38.5 Indigenous male 0.37 - 5.7 9.8 Nonindigenous female 0.82 - 10.6 46.7 Indigenous female 0.56 - 2.9 9.7 Brazil White male 2.0 73.1 - 84.2 Afro-male 1.8 74.7 - 70.4 White female 2.9 72.8 - 85.2 Afro-female 2.4 74.8 - 73.6 - missing data. Source: Author's calculations based on microdata from household surveys. TABLE 3.9 Access to services by race, ethnicity, and gender (percent) Water Sewage systems Electricity Garbage collection Guatemala Nonindigenous male 75.0 44.1 78.3 30.7 Indigenous male 64.6 23.0 59.0 8.7 Nonindigenous female 81.6 58.2 85.6 37.4 Indigenous female 71.5 31.2 67.2 12.9 Bolivia Nonindigenous male 41.5 52.2 81.2 Indigenous male 16.2 29.7 55.8 Nonindigenous female 38.3 59.1 84.3 Indigenous female 18.2 36.9 64.3 Brazil White male 99.9 47.4 95.4 78.8 Nonwhite male 99.8 40.7 89.0 65.4 White female 99.9 50.0 96.4 81.7 Nonwhite female 99.8 43.1 90.7 69.5 Source: Authors' calculations based on microdata from household surveys. and garbage collection than do Afro- or indigenous-headed rural indigenous men have access to water, 85 percent of households. For example, in Bolivia more than 50 percent of households headed by urban indigenous men have access to households headed by nonindigenous men or women have that service. Similarly, 68 percent of households headed by access to sewage systems compared to 30 percent and 37 per- nonindigenous rural women and 55 percent of households cent of households headed by indigenous men and women, headed by indigenous rural women have access to electric- respectively. Within ethnic or racial groups, households headed ity, while 83 percent of households headed by indigenous by women seem to have particularly high levels of access to urban women have electricity. These differences highlight public services. the general pattern of urban-rural differentials being much Access to services is greater for all urban dwellers greater than ethnic differentials with regard to access to ser- compared to any group of rural dwellers. For example, vices. However, since indigenous people are disproportion- in Guatemala, although 56 percent of households headed ately represented in rural and marginal urban areas, they by rural white men and 56 percent of households headed by experience more of the burden of limited access in such areas. 91 INEQUALITY IN LATIN AMERICA FIGURE 3.3 Wage gaps by race/ethnicity and gender Ratio of group mean wage to white male wages 1.0 -E Male Afro/indigenous 0.9s- m Female white/nonindigenous M Female Afro/indigenous 0.8- 0.7- 0.6 - 0.5 - 0.4 [- . 0.3 - 1: 0.2- 0.1 0- Brazil Guyana Guatemala Bolivia Chile Mexico Peru Note: Statistics are not gender-disaggregated for Chile, Mexico, or Peru; the comparison is all nonindigenous to all indigenous, using monthly wages. Figures for Brazil and Guyana refer to Afro-descendents and white (for Brazil) and lndo-Guyanese for Guyana. Source: Wodon 2003 for Chile, Mexico, and Peru, authors' calculations based on primary data for other countries. White/nonindigenous men have the highest white/nonindigenous men and women have higher average average Income, but all groups have high earnings than do Afro- or indigenous men or women.6 within-group inequality In wages Among the countries in figure 3.3, indigenous men earn Within racial and ethnic groups, women have lower earn- 40-65 percent less than nonindigenous men. A similar pat- ings than men, although the gap narrows in urban areas. tern exists between white Brazilian and Afro-Brazilian men, National figures show that women earn 60-80 percent as with the latter group earning 48 percent of the wages earned much as do men of the same race and ethnic group (fig- by the former. In addition, a similar gap exists between the ure 3.3). This compares to a national gender wage ratio of wages of women of different races. This pattern differs in 0.78 in the United States in 2002.4 In urban labor markets- Guyana, however, where Afro- and Indo-Guyanese men (and disregarding racial and ethnic variables-the gender wage women) earn very similar average wages (Indo-Guyanese have gap is not very large. Figure 3.4 shows that women in urban a slight edge). areas earn lower wages than do men, but that the wage gap When putting together race, ethnicity, and gender, hovers around a ratio of 0.85; the average hourly earnings two lessons emerge. First, indigenous/Afro-descended for women in Argentina and Costa Rica actually exceed the women earn the lowest wages among the four groups while hourly average wages of men.5 white/nonindigenous/Indo men earn the highest. Among White and nonindigenous people earn the highest wages the countries in figure 3.3 that allow disaggregation of sta- in Latin America. Data from across the region show that tistics according to race, ethnicity, and gender, nonwhite 92 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER FIGURE 3.4 Urban gender wage gap-ratio of women's hourly wages to men's hourly wages Argentina-1 997 Brazil-1998 Chile-1998 Colombia-1998 Costa Rica-1998 = Dominican Republic-1997 El Salvador-1997 Honduras-1998 Mexico-1999 Nicaragua-1998 Paraguay-1995 , Peru-1997 0 0.2 0.4 0.6 0.8 1.0 1.2 Ratio Median Me Source: Gender Stats database. women earn 35 percent (Bolivia), 42 percent (Brazil), and level and racial-ethnic disparities in urban-rural residence- 64 percent (Guyana) of white men's wages. Second, race and and reflect the heterogeneity in characteristics described ear- ethnicity play a greater role in wage gaps than does gender. lier in this chapter. In terms of the contribution of the return In three of the four countries in figure 3.3, white/non- on various characteristics that drive the observed wage gap, indigenous women have higher earnings than indigenous/ the high wage premium placed on education enjoyed by Afro-Brazilian men. Only in Guyana do Afro-descended men white/nonindigenous workers is not experienced by the have wages that exceed those of the Indo-Guyanese women. Afro/indigenous workers. In the case of Guyana, neither dif- ferences in characteristics nor returns on characteristics Sources of wage Inequality explain the wage gap between racial groups; however, there A standard Oaxaca-Blinder decomposition (table 3.10) shows is little difference in the observed wages that can be that wage differentials between racial and ethnic groups are explained. driven more by differences in productivity-related charac- For all four countries, differences in returns on teristics than by differences in the returns to those charac- productivity-related characteristics (coefficients) are the teristics.7 Guyana is the exception (for a discussion of the driving force behind the gender-based wage differentials methodology used, see box 3.6).8 Two characteristics are most group. Indeed, the similarity in women's and men's responsible for the wage gap-inequality in the education characteristics serves to mitigate rather than increase the 93 INEQUALITY IN LATIN AMERICA TABLE 3.10 Decompositions of group mean log wage differentials Attributed to differences Attributed to differences Differential in characteristics in coefficients Decompositions of white/nonwhite men wage differentials Bolivia 0.94 0.57 (61%) 0.37 (39%) Brazil 0.62 0.47 (76%) 0.15 (24%) Guatemala 0.72 0.44 (61%) 0.28 (39%) Guyana 0.01 -0.04 (-400%) 0.05 (500%) Decompositions of white men and women wage differentials Bolivia 0.29 -0.07 (-24%) 0.36 (124%) Brazil 0.41 -0.12 (-29%) 0.53 (129%) Guatemala 0.92 0.14 (15%) 0.78 (85%) Guyana 0.56 0.01 (2%) 0.55 (98%) Source: Authors' calculations based on microdata from household surveys. BOX qe6 OaxcaUdeadcmpam massdol The standard merhodology for identifying the factors rearranging the equarion gives equation (3) behind the difference in wages between two groups is the (3) Le')' )-Ln( Y,) =(X - X ,4q +I (',( -13*.)X, Oaxaca-Blinder methodology. The merhod quantifies the 3 ' - ' = _ - I portion of the wage differential that can be artributed ro Estimated (J3)and observed mean values (X) are plugged a difference in characteristics between two groups and the into equation (3) to give the difference in the wages for each proportion rhat is due to the difference in payoffs from group. The first term on the right-hand side of the equa- those characteristics (that is, rhe differences in coefficients tion quantifies the amount of rhe wage differeniial that can associated with those characteristics In a regression frame- be attributed to the difference in observed productivity work). To generate the estimate, each member of the characteristics and the second term on the right-hand side sample (i) is assigned to one of two groups, defined as m gives the amount that can be artributed to the difference andf: The first step is to esrimate the wages for each group in returns to productivity characteristics. X-bars are vec- separately, using a standard Mincerian wage equation: tors, where each component is the average level of that productivity characteristic in group m orf. The ,B and (3, (1) Ln(Y,^) = aw + ,3X,q +8E,m, are taken from equations (1) and (21. Commonly, the i2) Ln(Y,,) = a, + p3iX,, + c,, contriburion of each term to the overall distribution is stated as a proportion. The contribution of each produc- Where V, , is earnings of individual i in group m, X,is rivirv characteristic in X can also be calculated using a vector of productivity-related characteristics of individ- ual i. e i. s the error term, and a. and 13 are the coeffi- equation (3) cient estimates for group m; the same pattern is followed So Oaxaa 19P3 for group f. Subtracting equation (2) from equation tI) and gender-based wage differential in Bolivia and Brazil. While The male-female difference in returns on characteristics also the difference in returns (coefficients) is largely driven by captures the effects of those variables that were not controlled returns on education, returns related to all characteristics for in the vector of endowments, including actual job expe- prove important. Although the gap in returns is commonly rience, preferences, pre-market effects, and mismeasured interpreted as "discrimination," this is only part of the story. variables.9 94 GROUP-BASED INEQUALITIES: THE ROLES OF RACE, ETHNICITY, AND GENDER TABLE 3.11 Wuilin-group inequality share (for distributions of labor income) Afro-descended White/Ando-Guyanese Afro-descended White/lndo-Guyanese indigenous nonindigenous indigenous All nonindigenous men men women women Theil (1) Bolivia 0.60 0.47 0.53 0.56 0.69 Brazil 0.65 0.60 0.51 0.59 0.50 Guatemala 0.78 0.73 0.55 0.67 0.69 Guyana 0.32 0.32 0.29 0.41 0.26 Theil (0) Bolivia 0.73 0.52 0.69 0.61 0.82 Brazil 0.58 0.56 0.45 0.53 0.44 Guatemala 0.86 0.72 0.65 0.85 0.85 Guyana 0.29 0.27 0.27 0.34 0.26 Gini Bolivia 0.56 0.51 0.53 0.54 0.60 Brazil 0.57 0.56 0.51 0.54 0.49 Guatemala 0.61 0.59 0.54 0.58 0.61 Guyana 0.39 0.37 0.39 0.43 0.37 90th percentile wages/ 10th percentile wages Bolivia 35.4 14.5 32.6 20.4 39.0 Brazil 10.4 13.7 8.8 10.0 9.0 Guatemala 36.8 23.2 26.2 43.2 42.8 Guyana 5.5 4.3 5.0 4.9 5.2 50th percentile wages/ 1othpercentile wages Bolivia 9.8 4.2 10.7 5.9 9.6 Brazil 2.6 3.4 2.5 2.5 2.8 Guatemala 10.3 6.3 7.5 12.8 8.0 Guyana 2.6 2.3 2.0 2.4 2.6 Source: Authors' calculations based on microdata from household surveys. In addition to inequality between groups, there is great The factors behind national Inequality: inequality within gender, racial, and ethnic groups, especially Within-group inequality among women (table 3.11). Using the Theil coefficient as a In statistical terms, national wage inequality is mostly measure of income inequality, white and nonindigenous men explained by within-group, rather than between-group, dif- appear to have the most unequal wages in Brazil and ferences."2 Table 3.12 shows the contribution to national Guatemala. However, indigenous women's incomes in inequality of the within- and between-group inequalities for Bolivia are the most unequal and Indo-Guyanese women the two Theil indices of the four countries in the study, as have the highest Theil coefficient among the Guyanese well as for all persons, among men by race and ethnicity and groups.'" Considering the ratio of the 90th and the 50rh per- among white/nonindigenous/Indo persons by gender. centile wages to the 10th percentile wages suggests that wage In each case, within-group inequality accounts for the inequality is greater among women than men, since the greatest proportion of overall inequality. In no case does women's ratios tend to exceed those of men in Guatemala between-group inequality account for more than 14 percent and Guyana and somewhat in Bolivia. This high variance of overall inequality. Similarly, inequality between women may reflect the greater heterogeneity among the female than and men accounts for less than 3 percent of national inequal- the male labor force in terms of actual experience, training, ity. Thus, measures of national inequality between individ- and work choice." uals is largely due to heterogeneity within racial, ethnic, and 95 INEQUALITY IN LATIN AMERICA TABLE 3.12 Theil index decompositions (for hourly wages) Theil (1) Theil (0) Within-group inequality Between-group inequality Four-way partition Bolivia 0.60 0.53 (88%) 0.07 (12%) 0.73 0.66 (90%) 0.07 (10%) Brazil 0.65 0.57 (88%) 0.08 (12%) 0.58 0.50 (86%) 0.08 (14%) Guatemala 0.78 0.69 (88%) 0.09 (12%) 0.86 0.75 (87%) 0.11 (13%) Guyana 0.33 0.31 (94%) 0.02 (6%) 0.29 0.28 (97%) 0.01 (3%) Inequality among men by race/ethnicity Bolivia 0.55 0.50 (91%) 0.05 (9%) 0.67 0.62 (93%) 0.05 (7%) Brazil 0.63 0.57 (90%) 0.06 (10%) 0.57 0.51 (89%) 0.06 (11%) Guatemala 0.76 0.69 (91%) 0.07 (9%) 0.77 0.69 (90%) 0.08 (10%) Guyana 0.31 0.31 (100%) 0.00 (0%) 0.27 0.27 (100%) 0.00 (0%) Inequality among whites/nonindigenous/Indo by gender Bolivia 0.51 0.50 (98%) 0.01 (2%) 0.57 0.55 (96%) 0.02 (4%) Brazil 0.62 0.60 (97%) 0.02 (3%) 0.57 0.55 (96%) 0.02 (4%) Guatemala 0.74 0.72 (97%) 0.02 (3%) 0.79 0.77 (97%) 0.02 (3%) Guyana 0.35 0.34 (97%) 0.01 (3%) 0.30 0.28 (93%) 0.01 (7%) Source: Authors' calculations based on microdata from household surveys. gender groups, and not to the differences between demo- were treated alike, would overall measures of income inequal- graphic groups. This is not to say that between-group dif- ity go down? Or is inequality so high-both overall and ferences are not important. As shown earlier in this chapter, within gender, racial, and ethnic groups, including the they are in fact substantial and facilitate understanding of advantaged groups in society-that such changes would have poverty. Moreover, group-based differences may be of par- little effect on the overall level of inequality? ticular societal concern (Kanbur 2000), or may be particu- This question is examined here in two steps. First, the larly durable because of the role they play in maintaining returns on characteristics of the white/nonindigenous/Indo overall differences (Tilly 1999 and chapter 4). and/or male group are applied to the Afro/indigenous and/or female group in order to simulate wages, which are then used 3.3. Would income inequality decline if returns to generate inequality measures. This allows us to understand to human capital were more equal? how national inequality would change if everyone were The discussion in the preceding pages shows that, starting rewarded the same as the relative "wage-dominant" group.'3 in childhood, people from different racial, ethnic, and gender Second, both the returns on characteristics and the condi- groups face different circumstances and accumulate differ- tional distributions of some of the personal characteristics ent levels of human capital characteristics. This partly that affect wages-education level, sector of employment, explains the varying wage distributions between groups, but and children (that is, fertility)-of the white/nonindigenous/ the question emerges as to whether differences in treatment Indo and/or male group are applied to the Afro/indigenous according to race, ethnicity, and gender have measurable and/or female group in order to simulate wages, which are effects on overall inequality. In other words, if all persons then used to generate national inequality measures. This step 96 GROUP-BASED INEQUALITIES: THE ROLES OF RACE, ETHNICITY, AND GENDER facilitates understanding of how equalizing the playing field and Guyana). Hence, minority versus majority status and its in terms of both returns on characteristics and wage-related relation to wage-disadvantage can be considered separately. characteristics would affect national inequality. In addition, the definition of wage-disadvantaged group in The simulations are based on data from four recent house- some cases follows racial lines (Brazil and Guyana) and in hold surveys from four countries: Brazil (1996), Bolivia other cases is more obviously drawn along ethnic lines (1999), Guatemala (2000), and Guyana (1999). These coun- (Bolivia and Guatemala). Also evident are different levels of tries were chosen because their surveys asked questions female participation in the paid labor force and different edu- regarding racial and ethnic status as well as a number of other cational distributions. 14 variables of interest, including occupational sector, educa- tional attainment, household composition, and earnings. Equalizing returns on endowments Together, these surveys provide a range of information, The inequality measures from observed and simulated wages including cases in which the wage-disadvantaged group is in are shown in table 3.13. Wage inequality that is calculated the majority (Bolivia), the wage-disadvantaged group is in from observed wages is shown in column (a) and inequality the minority (Guatemala), and the wage-disadvantaged and measures from simulated wages is shown in columns (b)-(d), wage-advantaged groups are of relatively similar size (Brazil where the simulated wages are generated by equalizing the TABLE 3.13 Individual earnings and inequality measures, based on actual and simulated earnings within racial and gender groups Equalize returns and the distribution of Equalize returns only some characteristics All segments All segments treated as treated as Equal treatment Equal treatment white/ Equal treatment Equal treatment white/ by race/ethnicity by gender within nonindigenous/ by race/ethnicity by gender within nonindigenous/ Observed within gender race/ethnicity Indo men within gender race/ethnicity Indo men (a) (b) (c) (d) (e) (f) (g) Theil(1) Bolivia 0.60 0.53 0.56 0.51 0.57 0.58 0.55 Brazil 0.65 0.61 0.64 0.63 0.64 0.66 0.65 Guatemala 0.78 0.75 0.79 0.76 0.75 0.76 0.79 Guyana 0.32 0.35 0.32 0.34 0.36 0.31 0.33 Theil(0) Bolivia 0.73 0.63 0.68 0.59 0.66 0.72 0.65 Brazil 0.58 0.57 0.58 0.57 0.58 0.59 0.59 Guatemala 0.86 0.85 0.89 0.84 0.81 0.77 0.86 Guyana 0.29 0.31 0.27 0.29 0.31 0.28 0.30 Gini Bolivia 0.56 0.54 0.55 0.53 0.55 0.56 0.55 Brazil 0.57 0.56 0.57 0.56 0.56 0.59 0.57 Guatemala 0.61 0.61 0.62 0.61 0.59 0.60 0.63 Guyana 0.39 0.41 0.38 0.40 0.40 0.39 0.40 goth percentile wages /10th percentile wages Bolivia 35.4 24.3 31.3 22.2 26.9 34.0 26.6 Brazil 10.4 11.8 12.2 12.1 14.2 15.5 16.7 Guatemala 36.8 32.5 27.2 23.9 30.3 28.3 21.4 Guyana 5.5 5.4 4.7 4.9 5.3 4.7 4.7 50th percentile wages /10th percentile wages Bolivia 9.8 7.5 9.7 7.1 7.6 9.9 7.7 Brazil 2.6 2.9 2.9 2.9 3.5 3.3 4.2 Guatemala 10.3 9.3 7.9 7.2 8.7 7.5 5.8 Guyana 2.6 2.4 2.2 2.3 2.5 2.2 2.3 Source: Authors' calculations based on microdata from household surveys. 97 INEQUALITY IN LATIN AMERICA returns on characteristics across groups. Within the "equal- (that is, by holding gender constant) are greater than those ize returns only" section, three columns are presented. The in the simulated Theils by gender (that is, by holding race first simulation (column (b)) considers what would happen constant). The increased inequality in Guyana is due to the to the overall earnings distribution if white/nonindige- greater wage inequality among Indo-Guyanese, compared to nous/Indo and Afro/indigenous receive the same returns on Afro-Guyanese of both genders. The equalization of returns characteristics within gender groups, while the second sim- by gender has more to do with the decrease in the differ- ulation considers what would happen if women were treated ences between certain parts of the whole distribution, but like men within each racial and ethnic group (column (c)). is not strong enough to drive the overall inequality in the The third simulation considers what would happen if both distribution. women and Afro/indigenous people received the same return on characteristics as did white/nonindigenous/Indo men (column (d)). Columns (e)-(g) are considered in the discus- Equalization of returns on the distribution sion in the next section. of characteristics If everyone were rewarded the same for their character- As argued earlier in this chapter, characteristics accumulated istics as are white/nonindigenous/Indo- men (that is, com- throughout the lifecycle that affect current wages vary greatly paring columns (a) and (d)), inequality would not uniformly by gender, race, and ethnicity. In addition, returns on these increase or decrease across the sample. It would stay the characteristics (in other words, endowments) leads to dif- same or decrease in Bolivia and Guatemala (depending on ferent observed wages. To consider what would happen if dif- the inequality measure used) and increase in Guyana. A ferent groups both had similar characteristics and received the decrease in inequality in Bolivia is expected, since non- same compensation for various factors affecting their earn- indigenous men there have the lowest within-group inequal- ings, simulations were run of earnings distributions for wage- ity of the four groups. disadvantaged groups in those cases in which they would Equalized returns among racial or ethnic groups (column accumulate endowments (including education, sector of (b)) explain the great proportion of the observed changes in employment, and, for women, number of children) and be the overall Theil index (column (d)). In Bolivia, Guatemala, compensated by the same earnings mechanism as the advan- and Guyana, changes in the simulated Theil indices by race taged groups (box 3.7). BOX A ' Ecomtc -I In order to perform a multistage simulation-in which household income distribution across countries, this human capital-related characterisrics as well as returns approach is readily modifiable to considering the same on human capital can be considered-the authors uti- factor-this cime in the form of individual income and lized the microeconometric simulation method devel- earnings distributions-across demographic groups oped in recent work by Bourguignon, Ferreira, and Leite within a country. (2002) The merhod expands rhe standard earnings In order to walk through this process, imagine two decomposition method (the Oaxaca-Blinder method, groups, X and Y. Here earnings for group X are simulated described in box 3.6) by allowing some of the variables under the assumprion that they are treated like group Y. in [he earnings equarion to be determined by earlier The first srep in chis process istoesrimareeducarion level processes that are also estimated separately by group. for members of group Y as a function of age, mother's level Thus, a distribution of characteristics is simulated and of schooling (when available), and region. Then educacion substituted rather than using the mean values of char- is simulated for members of group X by using their values acteristics, as is done in rhe standard Oaxaca-Blinder for age and mother's level of schooling and the estimated met hod. While Bourguignon, Ferreira, and Leite (2002) coefficients from the education equation for group Y. A ran- apply their method to considering differences in domly generated error term for each X-person is then drawn 98 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER from a normal distribution that is standardized to reflect the The fburth step is to estimate month]) earnings fbr group X-group's error term variance. (See Bourguignon, Ferreira, Y as a function of age, morher's level of school ing, education, and Lustig 1998 fur detaElL on how this methodology works.) occupational sector, and region Then earnings are simulated The second step, for women in group X only, is to esti- for members of group X by using their values for age, mate the number of children for group Y as a function of mother's level of schooling. (simulated) education, and age, mother's level of schooling, region, and education. The (simulated) sector. A randomly generated error term for each number of children is simulated for members of group X X-person is then drawn from a normal distriburion stan- by substituting their values for age, mother's level of dardized to reflect the Y-group's error term variance. schooling. and (simulated) education in the fert iliry equa- Steps 1-3 utilize multinomial logic as the estimation tion for group Y. A randomly generated error term for each technique, since people fall into distinct groups, while step X-person is then drawn from a normal distribution stan- 4 utilizes Ordinary Least Squares (OLS) as the estimation dardized to reflect the X-group's error term variance. technique to deal with the continuous earnings disiribu- The third step is to estimate the sector of employment tion with a build up at the zero earnings point. forgroup Y as a function of age, mother's level of school- The figures below show graphically thetypeofoutput ing, education, household composition, and (for women that is generated from these successive levels of simula- only) number of children in the household. The sector is tion, so that ir is possible to see how each additional under- simulared for members of group X by using their values for lying stage brings the simulated earnings distribution for age, mother's level of schooling, household composition. the wage-disadvantaged group closer to that of rhe wage- (simulated) education, and (for women only the simulated advanraged group. In this case, actual and simulated number of children in the sectoral choice equation for earnings are presented for white and indigenous men in group Y A randomly generated error term for each Guatemala. The largest effect comes from simulating the X-person is then drawn from a normal distribution earnings distribution for indigenous men. using the coef- standardized to reflect the X-group's error term variance. ficient estimates from their Mi ncer wage equation: a. Nonindigenous male vs. American Indian male (simulating earnings regression coefficients) Density 0.45- - Nonindigenous male - American Indian male o~~~~~~~~~~~~~~~~~~~~~~~~ --American Indian male simulated I~~~~~~~~~~~~~~~~~~~~ aR?.X 0 -I~~~~~~~~~~~~~~~~~ Log earnings 1 Source: Authors' calculations based on microdata from household surveys. (Continue ) 99 INEQUALITY IN LATIN AMERICA BOX - (continued) There is litrle difference between the simulated densities on the above graph and the following one, which presents the disrribution while simulating education: b Nonind.genous male vs Amer-can IndCan male (simulating only education) Density o 45 N 5on-r-dgenoui male - American Indan male 1 ~~~~~~~~~~~~~Log earnings Source- Autt--cs calculaI.onS tased ons microdata from householdj surve~s Fi'nally, the last graph shows rhat simulaEed occupational sector and education are nor major f:actors in causing the disEributions to move closer togerher: c. Non.naigenou. male vs Amer.ran Indian male (s5r,ulallng both a and b) DensltV 0 45- - Non,rndigenoLus male - Amer,can Ind.an male ^ A --American Indian male simulated 0 -~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~ I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Log earnings 1 Source: Authors' calculations based on microdata from household surveys. 100 GROUP-BASED INEQUALITIES THE ROLES OF RACE ETHNICITY, AND GENDER The multistage procedure simulates endowments accord- single demographic group and the contribution of this spread ing to the same process through which the male and/or to national inequality. white/nonindigenous/Indo group accumulated these char- acteristics. The simulation procedure also allows for the Inequality Increases when endowments effects of possible discrimination in access to or variation in are equalized the choice of educational attainment and employment sector Equalizing both returns and the distribution of some (as well as the feed-through to fertility outcomes in the case endowments-so that education, occupation, and fertility of women) to be considered directly. In addition, it then levels of the nondominant group are modified to be close to becomes possible to consider the effects of potential dis- that of the dominant group-has the effect of returning the crimination in terms of earnings, once an individual has a inequality measures back to their original levels. Returning particular educational level, employment sector, and family to table 3.13, the set of columns entitled "equalize returns structure. and the distribution of some characteristics" replicate the set entitled "equalize returns only," but also allow for certain Equal treatment by race, gender, and ethnicity characteristics of the simulated group to be distributed in explains some, but not most, of the differences the same way as those from the white/nonindigenous/Indo in wage distributions or male group. Thus, columns (e)-(g) show the level of The result of these simulations for each country can be visu- inequality if the effect of those earlier events and decisions alized in figures 3.5 and 3.6. Figure 3.5 shows a simula- (that is, education, occupation choice, and fertility) were tion for men in Bolivia, in which indigenous men are given determined in the same way in each group. the earnings specification of nonindigenous men.15 Note that While inequality decreased overall (but increased for the spread of both actual earnings distributions is signifi- Guyana) when only the returns on characteristics were cant (as shown earlier in table 3.4 regarding the various equalized, it increased again when some of those character- inequality measures). The simulated distribution for indige- istics were also equalized. A comparison of observed nous men shifts somewhat closer to the distribution for (column (a)) and completely simulated (column (g)) overall nonindigenous men, but continues to display a significant inequality shows almost identical Gini inequality measures. spread, as was observed for Guatemala in box 3.7. This The Gini coefficient remains unchanged in Bolivia and actu- remaining gap is due, at least in part, to the difference in ally increases in Brazil under the simulations that address parental education levels across the two ethnic groups, which gender differences (apparently because treating women like reflects a historical heritage of ethnic differences across men in the Brazilian system increases the overall spread of generations. earnings). Figure 3.6 performs a similar exercise as that in figure 3.5, In the case of the Theil indices, there is also little notice- this time for Guyana by comparing Indo-Guyanese men to able change and little similarity in the direction of change Indo-Guyanese women, in which the latter group is given across countries, particularly for the Theil (1) measures. the earnings specification of the former. Note again the con- There are more noticeable changes in the 90/10 and 50/10 tinued spread in the simulated distribution for white women, ratios, with general reductions in these ratios relative to the even as the distribution conforms more closely to the dis- base case in Bolivia, Guatemala, and Guyana and increases tribution for white men. in the case of Brazil. However, significant spread remains in These results, which are consistent with results for the all four countries' earnings distributions under any of these other two comparisons (indigenous men to indigenous scenarios, particularly in Bolivia and Guatemala. women and nonindigenous women to indigenous women) The last result suggests that the way in which endow- for Bolivia and Guatemala, and also with similar graphs ments are generated for white/nonindigenous/Indo or male generated for Brazil and Guyana, point to results from a individuals creates greater within-group inequality than do formal calculation of inequality measures (presented in returns on these endowments. This can be rigorously tested tables 3.13 and discussed below). Racial and gender differ- by decomposing the Theil indices generated by the full sim- ences in inequality contribute little to national inequality, ulation into within- and between-group inequalities and compared to the large wage inequality observed within any comparing this result to the within- and between-group 101 INEQUALITY IN LATIN AMERICA FIGURE 3.5 Actual and simulated earnings distibulions for white and Indigenous men in Bolivia a. Nonindigenous male vs. American Indian male (simulating earnings regression coefficients) Density 0.5 - Nonindigenous male - American Indian male A - American Indian male simulated 0.1 1 . Log earnings b. Nonindigenous male vs. American Indian male (simulating only education) Density 0.5 -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. 0.1 s0. g ~~~~~~~~~~~~~~~~Log earnings c. Nonindigenous male vs. American Indian male (simulating both a and b) Density 0.5 -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. 0 01 10.1 Log earnings sSource: Authors' calculations based on microdata from household surveys. 102 GROUP-BASED INEQUALITIES: THE ROLES OF RACE, ETHNICITY, AND GENDER FIGURE 3 6 Actual and simulated earnings distribooUs for Indo-Guyanew men and wovmen a. East Indian male vs. East Indian female (simulating earnings regression coefficients) Density 0.75 - _ - East Indian male - East Indian female --East Indian female simulated 10 2 14.5t 1 2 Log earnings 14.5 b. East Indian male vs. East Indian female (simulating only education) Density 0.75- I~~~~ I~~~~~ 0 2 14.5 Log earnings c. East Indian male vs. East Indian female (simulating both a and b) Density 0.75 - 0 t 2 . 14.5 Log earnings Source: Authors' calculations based on microdata from household surveys. 103 INEQUALITY IN LATIN AMERICA measures for both the observed wages and partly simulated sions related to power other than relative earnings contri- (that is, with returns on endowments) wages. Although butions (or potential for earnings). inequality did not change when using simulated wages from Second, there may be some important determinants of the wage-dominant group's return on characteristics and income missing that are correlated with race, ethnicity, and their distribution, the nature of the inequality did change. gender. The estimated equations presented here do not In other words, nearly all inequality can be attributed to explain fully variations in education, employment sector, within-group inequality. fertility, or earnings.16 Third, the regional controls used in the current study may National inequality does not uniformly decrease capture racial and ethnic (and, though less likely, gender) when leveling the playing field differences to some degree. To the extent that there are dif- Section 3.2 described a range of notable differences between ferential distributions of racial and ethnic groups between racial and ethnic groups and between men and women, many rural and urban areas and between regions of a country, the of which appear to be of a type that would affect the abil- controls used here may downplay the linkage between ity of members of these groups to earn a reasonable income location and difference. and achieve a reasonable standard of living within their soci- Fourth, other important dimensions of income inequal- eties. Such differences are related to educational attainment, ity may not have been captured due to the focus on labor employment sector participation and occupation, household income. An additional step, that of considering what would structure, and dependency ratios. happen if unearned income (for example, pensions, interest, Given the differences in the way race, ethnicity, and rent, and returns on equity) were distributed equally, would gender apparently interact with advantage and majority provide an important addition to the research reported status across the four countries studied, it is striking to notice herein. It is more difficult to obtain this type of information, the relatively similar levels of inequality and results from although other studies (notably Bourguignon, Ferreira, and the simulations reported above. Indeed, it appears that these Leite 2002) have also considered unearned income distri- differences, while creating different average levels of bution before calculating household income distributions. income, actually contribute little to the spread of earnings, Finally, these results do not relate inequality measures to that is, to earnings inequality. the relative normative importance of between- and within- These results may run counter to what might have been group differences. As already noted, there are many reasons expected given the widespread nature of beliefs among both why societies may be particularly concerned with within- social scientists focusing on Latin America and the broader group differences on grounds of social justice and because public concerned with the effect of demographic factors on of the particular historical and organizational role such dif- income distribution. The difference in results may be partly ferences can play. due to limited comprehension that within-group variation is so much larger than between-group differences, when mea- 3.4. Conclusions sured in terms of inequality indices that are based on dif- The evidence presented in this chapter confirms that racial, ferences between all individuals or households. ethnic, and gender inequalities exist across Latin America, At the same time, there are a number of other reasons that along a number of dimensions and for a variety of different the results presented here might differ from those generated measures. White/nonindigenous men and their households from other sources. These require further exploration before have more-in terms of labor income, formal sector jobs, the current results can be asserted as the standard against assets, and education-than does any other ethnic or racial which others should be measured. First, some of the social group, including in comparison to women. However, not all science evidence relates to differences in dimensions other white/nonindigenous men are wealthy; instead men and than income, and thus is not necessarily relevant to under- women, whites and Afro and Indo groups, and indigenous standing income differences. For instance-to the degree that and nonindigenous people are represented in all advantaged much of the sociological literature is concentrated on describ- and disadvantaged sectors of society. ing and understanding power relationships within societies On average, women earn less than men, and nonwhite and households-it is not possible to characterize dimen- workers have much lower earnings than white workers. A 104 GROUP-BASED INEQUALITIES THE ROLES OF RACE, ETHNICITY, AND GENDER closer look at the wage distribution shows that these aver- spread in incomes, not the difference in mean incomes by age tendencies do not persist across the distribution. Instead, group) cannot be assumed to be greatly affected or even while the poorest part of the population is relatively evenly decreased by a leveling of the playing field along racial, divided among white/nonindigenous women and Afro or ethnic, and gender lines. indigenous men and women, white/nonindigenous women Using various measures of inequality and simulated wages have a larger presence than Afro or indigenous individuals from four countries, national wage inequality may increase higher up the distribution. The exception is Guyana, where or decrease if the processes that determine wages for gender is the more important variable. white/nonindigenous men also determine the wages of Guyana is an important special case, since its inclusion women and Afro/indigenous men. Naturally, however, a in the sample shows that being of Afro descent does not nec- decrease in wage and income inequality that is based on essarily allocate a person to a position of inferiority. Although racial, ethnic, and gender factors is of great value in itself, Afro-Brazilians have situations similar to those patterns since it would reduce one of the major horizontal inequities observed for the indigenous in countries with high indige- in the region-one that has deep historical roots and great nous populations, Afro-Guyanese fare nearly as well as Indo- durability. Guyanese. Instead, the gender divides are much greater in the Guyanese economy. This suggests that the power struc- Notes ture along racial lines is a greater determinant than simply 1. Throughout the rest of this chapter, unless otherwise noted, the phenotypical characteristics. In Guyana, this is a rapidly statistics for Bolivia, Brazil, Guatemala, and Guyana are generated by evolving structure. the authors using: Encuesta Continua de Hogares, 1999 for Bolivia; Finally, the mix of being female with being indigenous Pesqvisa Nacional por Amostra de Domicilio, 1996 for Brazil; Encuesta or of Afro descent is highly disadvantageous-indigenous Nacional sobre Condiciones de Vida, 2000 for Guatemala; and Survey of and Afro-descended women are at the bottom of any dis- Living Conditions, 1999 for Guyana. tribution, even in Guyana. Thus, the "double strike" con- 2. Cunningham (2001) shows that nearly all unpaid Mexican cept seems to apply in that inequalities arising from being workers are married women, with 10 percent of married working female and from the nondominant racial or ethnic group women citing unpaid work as their sector of employment. The major- combine into a particularly difficult position. ity of these women are married to men who declare themselves "self- The within-group inequality measures and graphs pre- employed." Thus, these appear to be family businesses to which women sented in this chapter showed a high degree of inequality contribute and from which they are likely to share in the fruits of their within demographic groups and much overlap between labor to a limited degree. groups. When distributions were plotted separately for dif- 3. One-digit occupational categories give crude estimates of the ferent groups, the mode of the Afro/indigenous distribution Duncan Index, due to the heterogeneity of jobs within occupations. fell to the left of that of the white/nonindigenous/Indo group For example, in Peru, both physicians (0.21 percent of men and 0.08 (both for men and women). However, examination of the percent of women) and midwives (0.01 percent of women and 0 per- wage distributions for each group showed a high degree of cent of men) are classified as "professional occupations." The gender overlap among them, suggesting that within-group varia- segregation within these occupations is lost using the one-digit occu- tions are larger than between-group differences. Comparisons pational categories, while it clearly emerges when physician and mid- of inequality measures across countries also showed that no wife are treated as distinct occupations. The interpretations of the labor one ethnic, racial, or gender group consistently experiences market greatly change when more disaggregated occupational codes greater inequality. are used. The Brazilian data show that the Duncan Index is 0.41 when The simulations presented in this chapter showed that using one-digit occupational codes for 7 occupations but is 0.62 when even when the playing field is leveled, within-group inequal- using 35 occupations, indicating that another 20 percent of men and ities persist. Equalizing those factors that determine women would have to switch occupations for the distribution of occu- employment and wages-education, sector of employment, pations to be equal by gender. and fertility-decreases the spread in the wage distribution 4. This figure refers to median weekly earnings for year-round, in some countries but increases it in others. Hence, measured full-time workers; see frp://ftp.bls.gov/pub/news.release/Hisrory/ overall income inequality (in terms of the amount of the wkyeng.01 172003.news. 105 INEQUALITY IN LATIN AMERICA 5. The statistics presented in Figure 3.4 are based on hourly wages. variation in women's monthly work hours. The Guyana data does not Since women tend to work fewer average hours per month, as com- permit using a similar control, since a variable for the number of hours pared to men, the use of hourly wages to construct the gender wage worked monthly is not reported. gap is more accurate than the use of monthly wages, which are com- 12. The relative contribution of inequality within and between monly reported by statistical agencies. demographic groups to national inequality can be estimated by decom- 6. Monthly wages are used for Chile, Mexico, and Peru. Hourly posing the inequality measure into these two components. While it wages are a more appropriate comparison variable than monthly wages is not possible to do this with the Gini index or with the wage ratios, since the number of hours worked by each group differs, the Theil index is decomposable in this fashion. 7. This has been found for other studies using Latin American 13. This is necessarily a partial equilibrium analysis, since the data. These studies include Maclsaac and Patrinos (1995), which uses assumption is made that increasing everyone's rewards to match those Peruvian data and finds that 50-70 percent of the log earnings gap of the wage-dominant group will not result in an overall downward between indigenous and nonindigenous men is due to differences in pressure on returns, as may occur in a general equilibrium model. characteristics. Patrinos, Velez, and Psacharopoulos (1994) find that 14. The authors attempted to estimate a relatively similar specifi- 78 percent of the difference between the Spanish and Guarani speak- cation for each of the four countries. While this did not affect the abil- ers is due to differences in characteristics. In addition, Patrinos (1997) ity to create counterfactuals (since they are all within-country), it does points out variation in the proportion attributable to characteristics make the results more comparable across countries in terms of under- among the various ethnic groups within Guatemala. standing that the results capture the same effects in each country. 8. Using a standard set of characteristics, including age, educa- Hence, although the particular levels of education vary for each coun- tional attainment, occupational sector, and rural-urban location for a try, for instance, the educational variable enters in a similar way into four-country sample, the decomposition for two of the possible group each country's specification. Full numerical results from these simu- comparisons-white/nonindigenous versus Afro/indigenous men and lations are reported in the companion working paper for this chapter, white/nonindigenous men versus women-are calculated. (Details of Cunningham and Jacobsen (2003). the estimation are in the companion working paper Cunningham and 15. Note that other counterfactual states could have been used in Jacobsen 2003.) the simulation process, for instance by attributing to white men the 9. For example, age is a proxy for labor force experience and is a earnings specification of nonwhite men or utilizing weighted inter- good measure for both men and women if they have similar levels of mediate values of the white and nonwhite male coefficients for the education (which they do) and they both work full-time after com- simulation coefficients. However, this particular version of the coun- pleting school. However, a smaller proportion of women than men are terfactual has interpretive appeal as it addresses directly the "what if " in the labor force and women enter and exit the labor force with greater scenario of members of the disadvantaged group being treated as if frequency than men. As a result, "actual experience" is impossible to they were members of the advantaged group, under current labor measure for women, leading to a poor estimate of the returns due to market conditions. age for women and greater deviation from men's better measured 16. Omitted variables would have to be correlated with these demo- returns. The claim that the "other" is an upper bound for discrimi- graphic factors in order to be important explanatory variables for demo- nation is also not entirely correct, since discrimination may even be graphic differences in earnings. Another aspect of "goodness of fit" is greater than the value of the "other" proportion; however, some of the that nonsystematic factors perhaps matter more in the societies stud- omitted effects decrease the observed discriminatory effects. ied here than in other societies; that is, these earnings functions are 10. For the formal definitions of the Gini, Theil (1), and Theil (0) potentially inherently more stochastic. However, this stochastic factor index, as well as a discussion of index decomposition, see Litchfield would still have to differ systematically by demographic group, in that (1999). certain groups are more likely to have a higher variance in earnings 11. Hourly wage is used to calculate the statistics for Bolivia, Brazil, due to stochastic factors. and Guatemala in tables 3.4 and 3.5. This controls for the higher 106 PART II The Determinants of Inequality in Latin America - CHAPTER 4 Historical Roots of Inequality in Latin America JI T IS WELL KNOWN THAT NO OTHER REGION HAS AS EXTREME INEQUALITY IN WEALTH OR INCOME AS Latin America. This chapter addresses some basic questions about the roots of this inequality. Has Latin America always been more unequal than other regions? What does the historical record reveal about comparative inequality? Following Engerman and Sokoloff (1997, 2000, 2002) and Acemoglu, Johnson, and Robinson (2001, 2002), the authors of this chapter argue that the contemporary situation cannot be understood without recognizing that extreme inequality emerged soon after the Europeans began to colonize the Americas half a millennium ago, and has been reflected in the institutions they put in place. Both this initial inequality and institutions were shaped largely by the factor endowments that the Europeans found in Central and South America, rather than the nature of the colonial powers themselves. Although these colonies ultimately gained independence and the development of technology and the world econ- omy brought about important changes, extreme inequality persisted into the 19th and 20th centuries because the evolution of political and economic institutions tended to reproduce and reinforce highly unequal distributions of wealth, human capital, and political influence. The difficulties of overcoming gross disparities in cir- effects are filtered through economic, social, and political cumstances became all the more formidable during the late institutions. Economic and political institutions define and 19th century, when major increases occurred in the relative enforce property rights and provide the legal framework returns on land-an asset with extraordinarily concentrated within which markets and other voluntary agreements ownership-and on education, in which Latin American soci- between private agents are organized. Thus, they have a fun- eties were slow to invest. A legacy of political institutions bor- damental impact upon the distributions of income, capital, dering on authoritarianism also spurred inequality by and political standing. There is no doubt that the colonies discouraging the evolution of truly consolidated democracies. established by the Europeans in South and Central America Finally, the "burden of history" must be considered. If the cur- were characterized by extreme inequality from the begin- rently very high levels of inequality in Latin America are the ning, but institutions ultimately account for its persistence. result of long historical processes, does this mean that they The basis for the initial inequality is in some ways easy to are immutable? What can be done to turn this pattern around? understand. There were very few Europeans among much larger populations of indigenous peoples and slaves, but they 4.1. Factor endowments, inequality, were greatly advantaged by their higher levels of human cap- and institutions ital (that is, knowledge of technology and the sorts of legal Although factor endowments, technology, and relative and economic institutions established in the colonies), scarcity have important implications for inequality, their wealth, and legal standing. However, it is institutions that 109 INEQUALITY IN LATIN AMERICA must be considered to gain understanding of why this landholders the means and the motive to operate on a large inequality persisted for nearly 500 years and generated pat- scale. terns that can still be seen today. The small group of elites that enjoyed large rents had the Brazil, like many other European colonies with soils and political power to protect their interests, and protect them climates suitable for cultivating sugar-the most valuable they did. Severe restrictions on trade routes and on further commodity in world trade from the late 16th through the immigration from Europe bolstered their position and 18th centuries-rather quickly developed a vibrant agricul- hindered the ability of regions of New Spain without sub- tural sector oriented toward exploiting the country's com- stantial native populations (such as Argentina and Uruguay) parative advantage. With the "technology" of gang labor, to attract resources or otherwise compete with colonial sugar could be produced at a very low cost on large slave centers in Bolivia, Mexico, and Peru until late in the 18th plantations. Brazilian planters tapped the readily available century. Despite their small share of the population-less supply made possible by a world market in slaves. In the than 20 percent as late as 1825-people of European descent country's first 250 years, roughly 70 percent of the immi- living in Spanish American benefited from greater wealth, grants to this Portuguese colony arrived in chains. Europeans human capital, and political influence than did the rest of could do very well in Brazil's booming economy, which the population, and were therefore able to shape policies and attracted more than 25 percent of those people who chose institutions to their advantage. These circumstances in turn to come to the Americas through 1760. Although the frac- perpetuated extreme inequality, and conditions changed little tion of the population that was white and privileged rose after the colonies gained independence. The elites maintained over time (from 7.4 percent in 1650 to 23.4 percent in effective control in the new republics and the fundamental 1825), the structure of Brazilian society did not change, and basis for economic inequality remained the same. remained dominated by a small but entrenched elite. It is illustrative to consider what happened to the Spanish America also came to be characterized by extreme European colonies in the northern part of North America, inequality early in its history. The most powerful country where initial conditions were quite different and a much in Europe during the 16th century, Spain was the leader in greater degree of equality prevailed. This relative equality, colonizing the New World. It chose to focus its efforts on however, was not due to the original intentions of the colo- areas such as Mexico and Peru, establishing colonies that nial powers or of the early colonizers. When the English set were characterized by both a substantial native population up their colony at Jamestown, Virginia, in 1607, for exam- surviving contact with the European colonizers and by dis- ple, the original plan was inspired by the experience of the tribution among a privileged few of claims to enormous blocs Spaniards in Latin America. The English looked for gold and of land, mineral resources, and native labor. Among the many tried to capture the local Algonquian chief Powhatan, in the examples of institutions that aided members of the elite and hope that this would empower them in the same way that offered little to the bulk of the population were the encomienda the capture of Atahulapa had for Pizarro in Peru in 1532. (which gave Spanish conquistadors the right to Amerindian This strategy proved to be an utter failure, however, and a labor), the mita (a system of forced labor used in the mines), new strategy had to be developed when the colonists starved and the repartimiento (the forced sale of goods to Indians, typ- and the Virginia Company went bankrupt (see Morgan ically at highly inflated prices). 1975). The resulting large-scale estates and mines, established In addition, the English, French, and Dutch learned that early in the histories of these colonies, were to some degree there was no gold and few indigenous people to exploit, and based on pre-conquest social organizations in which Native that the soils and climates of North America would not sup- American elites extracted tribute from the general popula- port the production of highly profitable crops such as sug- tion. These arrangements endured even when principal pro- arcane (and thus large slave plantations). The reality was that duction activities were lacking in economies of scale. land was cheap and labor was costly. Such circumstances were Although small-scale production was typical of the grain radically different from those found elsewhere in the hemi- agriculture sector during this era, the essentially nontrad- sphere, and fostered a remarkable degree of equality. Virginia able property rights to tribute (in the form of labor and and other colonies had to compete with each other to attract other resources) from rather sedentary groups of natives gave migrants by providing favorable working conditions to even 110 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA unskilled laborers coming to North America as indentured economic growth. The populations of these colonies grew pri- servants. Attempts to establish tenant agriculture and the marily due to a healthy excess of births over deaths until early hierarchical social structures of England-such as in Mary- industrialization got under way, and consequently European land under Lord Baltimore and in Pennsylvania under perceptions of the opportunities available in the region William Penn-were repeatedly frustrated by the reality of changed. From these conditions emerged a society in which factor endowments and initial circumstances. Labor scarcity the population was relatively homogeneous and there was rel- made for high wages, but also low returns on land, and thus ative (compared to most of the hemisphere) equality in limited prospects for attaining gentlemanly status. In the wealth, human capital, and ultimately political influence. words of David Galenson, the noted economic historian of Of course, Native Americans and African Americans did colonial America (Galenson 1996, p. 144): not enjoy the same conditions as did whites. However, these Although the establishment of large estates to be groups accounted for a much smaller proportion of the pop- worked by tulation than in colonies established elsewhere, and the pat- worked by tenants and landless laborers was the mil- tern of institutions in the United States was established tsually mo sedel , whih gathesecroprie ery cloniesered before the arrival of large numbers of slaves (see table 4.1). usually based, the greater economic power conferred What was key to the formation of colonial institutions on settlers by the New World's labor scarcity prevented was not the identity of the colonizing power, but rather the these English tenures and practices from tively initial conditions in the colonies (see Engerman and taksimply seing hold,handeproprietors wee oftet fosrc a Sokoloff 1997 and Acemoglu, Johnson, and Robinson 2001). by simply selling their land outright to settlers. Thus, British colonies such as Guyana, British Honduras, Relatively few Europeans chose to migrate to New and Jamaica developed degrees of inequality, economic struc- England and Canada (net migration to these areas was actu- tures, and institutions that looked very much like those in ally negative during the early period), but those who did came neighboring colonies established by other European powers. to enjoy greater equality, protection of the law, and institu- Similarly, Spanish colonies that had small populations of tions that were in the long run more conducive to widespread Native Americans, such as Argentina, Costa Rica, and participation in the commercial economy, investment, and Uruguay, seemed to follow significantly different paths of TABLE 4.1 The distribution and composition of the population in New World economies (percent) Composition of population Native Share in New World Colonial region and year White Black American population Spanish America 1570 1.3 2.5 96.3 83.5 1650 6.3 9.3 84.4 84.3 1825 18.0 22.5 59.5 55.2 1935 35.5 13.3 50.4 30.3 Brazil 1570 2.4 3.5 94.1 7.6 1650 7.4 13.7 78.9 7.7 1825 23.4 55.6 21.0 11.6 1935 41.0 35.5 23.0 17.1 United States and Canada 1570 0.2 0.2 99.6 8.9 1650 12.0 2.2 85.8 8.1 1825 79.6 16.7 3.7 33.2 1935 89.4 8.9 1.4 52.6 Source: Engerman and Sokoloff 1997. 111 INEQUALITY IN LATIN AMERICA institutional development than did those with factor indigenous into urban centers in order to be better able to endowments that disposed them toward extreme inequal- mobilize them to work in the silver mines of Potosi. Even ity. The sharp contrasts in many institutional dimensions the Bourbon reforms in the second half of the 18'h century between Costa Rica-which had relatively few Native were meant only to rationalize and increase the efficiency of Americans-and Guatemala is well-known. the system (from the point of view of the Spanish state), not Another example, however, concerns a broader pattern to fundamentally alter it. influencing the capabilities of local governments. The lowest The situation was similar in the United States, despite the level of Spanish administration was the cabildo, a council that fact that very different sorts of institutions persisted there. at least nominally had the power to levy taxes and held For example, even though the British Crown was at times able responsibility for administering municipal services such as to effectively intervene and restrain or overrule colonial leg- schools. The cabildo was generally appointed by higher offi- islatures (such as when Charles II abolished the practice of uni- cials, who were themselves appointed by the Spanish Crown. versal male suffrage in Virginia), they remained a vital force. Yet there was the possibility for a cabildo abierto, or open council, where all the vecinos, or citizens, could meet to 4.3. The persistence of inequality: express their opinions and attempt to influence policy. This Post-independence institution was the closest thing to a truly democratic entity Like the United States, most of Latin America had fought in Spanish Latin America. Noting that these meetings were for, and won, independence from their colonial masters by seldom (if ever) held in many areas, such as colonial Peru, the early 19,h century. Despite the rhetoric of the revolu- Lang (1975, p. 28) has suggested that: tionary movement-and the fact that nearly all of the new countries were nominal democracies-the political break Relative insignificance was often the best guarantee cutiswr oia eorce-h oiia ra Relative iownsignif tainc was oft ith breest guaratee from Spain appears to have yielded little or no reduction in that a town would retain all of its privileges unre- the extent of inequality throughout the continent as a whole. stricted by subsequent clashes between the cabi/do and Indeed, in most Latin American societies, inequality prob- various bureaucratic bodies. Santiago and Buenos al nrae uigtefloigdcds Aires . . . considered to be marginal and undesirable On the face of it, this appears to be much harder to places, convened the cabi/do abierto numerous times account for than persistence of inequality during the colo- during the Hapsburg era. nial period. The fundamental reason for the persistence of Thus, the political variation within the Spanish colonies extreme inequality appears to have been the continued pat- strongly supports the notion that it was initial conditions tern of institutional change that favored the interests of the that mattered most for the development of institutions. elite and provided the bulk of the population with only lim- ited access to economic opportunities. Given the disparities 4.2. The persistence of inequality: in resources that were a legacy of the colonial period, it The colonial period should be no surprise that the Creole elite were quickly able Patterns of inequality persisted once they were established to regain effective control of the independent countries and during the early colonial period. This occurred because the determine the general structure of institutions. initial nexus of institutions survived, as did the rationale for Indeed, in several cases, most notably Mexico, indepen- these institutions. Even though the century after the con- dence was favored precisely because it enabled domestic elites quest of the continent saw a huge collapse of the indigenous to avoid liberalizing currents that were emerging from Spain. population of Latin America (the conventional estimate This was particularly true after the writing of the Cadiz Con- being that the 1492 population declined by about 90 per- stitution in 1812. Peter Bakewell (Bakewell 1997, p. 377) cent during the following century), exploiting Indian labor sums up the consensus on this issue when he notes that and resources remained the most attractive strategy whenever "Mexican independence in 1821 has long been viewed as a it was feasible. Colonial institutions often responded to the reaction to the liberalism of the revived Spanish Cortes, a population collapse simply by allowing for more efficient rejection by influential conservative Creoles." Thus, inde- exploitation of the remaining indigenous peoples. In Peru, pendence did not fundamentally alter the structures of polit- the reforms of Viceroy Toledo in the 1 570s concentrated the ical power in Latin America. 112 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA An important study of institutional persistence in 19,h and early 20th centuries. The estimates reveal that although century Latin America is that by Woodward (1966) on it was common in all countries to give the right to vote only Guatemala. He shows that the economic life of Guatemala to adult males until the 20th century, the United States and during the colonial period was controlled by a merchant guild Canada were the clear leaders in doing away with restrictions called the Consulado de Comercio, the members of which based on wealth and literacy, and a much higher proportion continued to control the economy after independence. As of the population voted in these countries than anywhere else Woodward (1966, p. 34) notes, "The merchants continued in the Americas. Not only did the United States and Canada to hold their exclusive court privilege until 1871. Little or adopt the secret ballot and extend the franchise to even the no change was made in the basic concept of justice before that poor and illiterate much earlier-although restrictions were year. As in the 1790s, in the 1860s the wealthy, aristocratic reintroduced in the United States in the 1890s at the expense merchants continued to rely on their own court to interpret of African Americans-but the proportion of the population justice in the best interests of their own class, and in main- that voted was at least a half-century ahead of even the most taining their leadership over the commerce of the land." progressive and most equal countries in South America Woodward also points out (1966, p. xiv) that "This seem- (namely, Argentina, Costa Rica, and Uruguay). ingly paradoxical existence of a royal institution under a It is a remarkable testament to the extent of political republican government is explained by the nature of the gov- inequality in Latin America and the Caribbean that as late ernment which ruled Guatemala from 1839 to 1871. Repub- as 1900 none of the countries in the region had a secret lican in form ... the monopolistic position of the Consulado ballot system or more than a minuscule fraction of the pop- was consistent with the policies and objectives of the regime." ulation casting votes. The Australian ballot (a system of As noted above, truly representative institutions never state-provided ballots with provisions to ensure secrecy) was really developed during the colonial period in Spanish not introduced until 1958 in Chile and 1988 in Colombia. America, while the relatively representative legislative assem- Before this time, political parties directly issued their own blies were established quite early in most of the British ballot papers. Baland and Robinson (2003) show that these colonies in North America. This contrast in political inequal- circumstances had dramatic effects on voting patterns in ity, along with corresponding variation in economic inequal- Chile, effectively allowing conservative parties to control ity, endured into the 19th century and beyond. The elections in rural areas since landowners could use the threat persistence of these radical disparities in the distribution of of job dismissal to regulate the voting behavior of workers. political power can be seen by examining both the laws gov- The great majority of European nations, as well as the erning who was entitled to vote, as well as the fraction of United States and Canada, achieved secrecy in balloting and the population that actually participated in elections. universal adult male suffrage long before other countries in Since most of the societies in the Americas were nomi- the western hemisphere, and the proportion of the popula- nally republican democracies by the mid-19th century, this tions that voted was always higher (often by as much as 4-5 sort of information had a direct bearing on (although it likely times). Although many factors may have contributed to the understates) the extent to which elites-whose position was low levels of participation in Latin America, wealth and lit- based largely on wealth, human capital, race, and gender- eracy requirements-in addition to the absence of secrecy- were able to wield disproportionate influence in the forma- were serious binding constraints. Some countries, such as tion and implementation of government policies. The Barbados, maintained wealth-based suffrage restrictions until understatement arises not only from the advantages elites the mid-20th century, but most European societies joined the had in making effective use of informal channels of politi- United States and Canada in moving away from economic cal influence (including an approach based on the force of requirements over the course of the 19t5 century. However, arms), but also from the fact that many elections were accom- whereas the states in the United States frequently adopted panied by fraud and, in the absence of an effective secret explicit racial limitations on voting, when they abandoned ballot, explicit or implicit coercion (see Graham 1990 on the economic requirements (that is, when the constitutional case of Brazil). amendments following the Civil War ended the practice), Table 4.2 summarizes information about how the right to Latin American countries typically chose to screen accord- vote was restricted across New World societies in the late 19th ing to literacy. 113 INEQUALITY IN LATIN AMERICA TABLE 4.2 Laws governing the franchise and the extent of voting in seected countrhies, 1840-1940 Lack of secrecy Wealth Literacy Percent of the Period and country Year in balloting requirement requirement population voting 1840-80 - Chile 1869 No Yes Yes 1.6 1878 No No Noa Costa Rica 1890 Yes Yes Yes Ecuador 1848 Yes Yes Yes 0.0 1856 Yes Yes Yes 0.1 Mexico 1840 Yes Yes Yes Peru 1875 Yes Yes Yes Uruguay 1840 Yes Yes Yes 1880 Yes Yes Yes Venezuela 1840 Yes Yes Yes 1880 Yes Yes Yes Canada 1867 Yes Yes No 7.7 1878 No Yes No 12.9 United States 1850 No No No 12.9 1880 No No No 18.3 1881-1920 Argentina 1896 Yes Yes Yes 1.8b 1916 No No No 9.0 Brazil 1894 Yes Yes Yes 2.2 1914 Yes Yes Yes 2.4 Chile 1881 No No No 3.1 1920 No No Yes 4.4 Colombia 1918c No No No 6.9 Costa Rica 1912 Yes Yes Yes 1919 Yes No No 10.6 Ecuador 1888 No Yes Yes 2.8 1894 No No Yes 3.3 Mexico 1920 No No No 8.6 Peru 1920 Yes Yes Yes Uruguay 1900 Yes Yes Yes 1920 No No No 13.8 RB de Venezuela 1920 Yes Yes Yes Canada 1911 No No No 18.1 1917 No No No 20.5 United States 1900 No No yesd 18.4 1920 No No Yes 25.1 1921-40 Argentina 1928 No No No 12.8 1937 No No No 15.0 Bolivia 1951 - Yes Yes 4.1 Brazil 1930 Yes Yes Yes 5.7 Colombia 1930 No No No 11.1 1936 No No No 5.9 Chile 1920 No No Yes 4.4 1931 No No Yes 6.5 1938 No No Yes 9.4 (continued) 114 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA TABLE 4.2 (continued) Lack of secrecy Wealth Literacy Percent of the Period and country Year in balloting requirement requirement population voting Costa Rica 1940 No No No 17.6 Ecuador 1940 No No Yes 3.3 Mexico 1940 No No No 11.8 Peru 1940 No No Yes - Uruguay 1940 No No No 19.7 RB de Venezuela 1940 No Yes Yes Canada 1940 No No No 41.1 United States 1940 No No Yes 37.8 a. After having eliminated wealth and education requirements in 1878, Chile instituted a literacy requirement in 1885 that seems to have contributed to a sharp decline in the proportion of the population that was registered to vote. b. This figure is for the city of Buenos Aires, and likely overstates the proportion that voted at the national level. c. The information on restrictions refers to national laws. The 1863 Constitution empowered provincial state governments to regulate electoral affairs. Afterward, elections became restricted (in terms of the franchise for adult males) and indirect in some states. It was not until 1948 that a national law established universal adult male suffrage throughout the country. This pattern was followed in other Latin American countries, as it was in the United States and Canada to a lesser extent. d. There were 18 states-7 in the South and 11 elsewhere-that introduced literacy requirements between 1890 and 1926. Source: Engerman, Haber, and Sokoloff 2000. The contrast between the United States and Canada, on taxes, and access to land, as well as other policies that were the one hand, and the Latin American countries, on the other, expected to be attractive to potential migrants. was not so evident at the outset. Despite the sentiments pop- Labor scarcity was a crucial element in determining the ularly attributed to the Founding Fathers, voting in the initial level of inequality across New World colonies, and United States was until early in the 1 9'h century largely a continued to have an important effect on the level of polit- privilege reserved for white men with significant property ical inequality even within the United States. It is strik- holdings. By 1815, only four states had adopted universal ing that the pioneers in extending suffrage (such as new white male suffrage, but as the movement to do away with states being formed in Argentina, Uruguay, and the United political inequality gained strength, the rest of the country States) did so during periods in which they were striving followed suit. Virtually all new entrants to the Union to attract migrants. The right to suffrage was often one of extended voting rights to white men (with explicit racial a set of policy measures that were thought to be attractive restrictions generally introduced in the same state consti- to those contemplating relocation. When elites-such as tutions that did away with economic requirements), and holders of land or other assets-want common men to locate older states revised their laws in the wake of protracted polit- in the polity, they may choose to extend access to privileges ical debates. and opportunities even in the absence of the threat of civil The politically and economically key states of New York disorder. In fact, a polity (or one set of elites) may find itself and Massachusetts made the break with wealth-based restric- competing with another to attract labor or whatever else is tions in the 1820s and the shift to full white adult male suf- needed. (See Acemoglu and Robinson 2000 for cases in frage was largely complete by the late 1850s (with North which the franchise was extended under threat of civic Carolina, Rhode Island, and Virginia being the laggards). disorder.) Alternative explanations, such as the importance The relatively more egalitarian populations of the western of national heritage, are not very useful in identifying states were the clear leaders in the movement. The rapid why Argentina, Costa Rica, and Uruguay pulled so far ahead extension of access to the franchise in these areas not coin- of their Latin American neighbors, or why other British cidentally paralleled liberal policies toward public schools, colonies in the New World lagged behind Canada with 115 INEQUALITY IN LATIN AMERICA regard to the pace at which access to suffrage was education, including the ability to read and write. It was extended. common for schools to be organized and funded at the vil- Differences in the distribution of political power had a lage or town level, especially in New England. feedback effect on both access to economic opportunities and The United States probably had the most literate popu- investment in public goods in such a way as to in turn influ- lation in the world by the beginning of the 1 9'h century, but ence the long-term direction of institutional development and the common school movement-which got under way in the inequality. Educational institutions provide a highly relevant 1820s following the movement to extend the franchise- example because increases in a society's levels of schooling and put the country on an accelerated path of investment in edu- literacy have been related to many socioeconomic changes cational institutions. Between 1825 and 1850, nearly every conducive to growth, including higher labor productivity, northern state that had not already done so enacted a law more rapid technological change, and higher rates of com- strongly encouraging or requiring localities to establish free mercial and political participation. Moreover, in addition to schools that were open to all children and supported by gen- promoting growth, public investment in education also has eral taxes (Cubberley 1920). Although the education move- a major influence on the distribution of the benefits of growth. ment made slower progress in the South, schooling had Although most New World societies were so prosperous spread sufficiently by the mid-19h century that more than by the early 1 9'h century that they had the material resources 40 percent of the school-age population was enrolled and needed to support the establishment of a widespread net- nearly 90 percent of white adults were literate (see table 4.3). work of primary schools, before the 20th century only a few Schools were also widespread in early 19th century Canada. made such investments on a scale sufficient to serve the gen- However, the northernmost English colony lagged behind eral population. The United States and Canada were excep- the United States by several decades with regard to estab- tions in terms of leadership in investing in institutions of lishing tax-supported schools with universal access, even primary education. Virtually from the time of early settle- though its literacy rates were nearly as high. ment, the North Americans seemed to have generally been The rest of the hemisphere trailed far behind the United convinced of the value of providing children with a basic States and Canada in primary schooling and the attainment TABLE 4.3 Literacy rates in the Americas, 1850-1950 Country Year Age group Rate' (percent) Argentina 1869 6 and above 23.8 1895 6 and above 45.6 1900 10 and above 52.0 1925 10 and above 73.0 Barbados 1946 10 and above 92.7 Bolivia 1900 10 and above 17.0 Brazil 1872 7 and above 15.8 1890 7 and above 14.8 1900 7 and above 25.6 1920 10 and above 30.0 1939 10 and above 57.0 British Honduras 1911 10 and above 59.6 (Belize) 1931 10 and above 71.8 Chile 1865 7 and above 18.0 1875 7 and above 25.7 1885 7 and above 30.3 1900 10 and above 43.0 1925 10 and above 66.0 1945 10 and above 76.0 (continued) 116 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA TABLE 4.3 (continued) Country Year Age group Rate' (percent) Colombia 1918 15 and above 32.0 1938 15 and above 56.0 1951 15 and above 62.0 Costa Rica 1892 7 and above 23.6 1900 10 and above 33.0 1925 10 and above 64.0 Cuba 1861 7 and above 23.8 (38.5, 5.3) 1899 10 and above 40.5 1925 10 and above 67.0 1946 10 and above 77.9 Guatemala 1893 7 and above 11.3 1925 10 and above 15.0 1945 10 and above 20.0 Honduras 1887 7 and above 15.2 1925 10 and above 29.0 Jamaica 1871 5 and above 16.3 1891 5 and above 32.0 1911 5 and above 47.2 1943 5 and above 67.9 1943 10 and above 76.1 Mexico 1900 10 and above 22.2 1925 10 and above 36.0 1946 10 and above 48.4 Paraguay 1886 7 and above 19.3 1900 10 and above 30.0 Peru 1925 10 and above 38.0 Puerto Rico 1860 7 and above 11.8 (19.8, 3.1) Uruguay 1900 10 and above 54.0 1925 10 and above 70.0 RB de Venezuela 1925 10 and above 34.0 Canada 1861 All 82.5 English-majority counties 1861 All 93.0 French-majority counties 1861 All 81.2 United States North whites 1860 10 and above 96.9 South whites 1860 10 and above 91.5 Total population 1870 10 and above 80.0 (88.5, 21.1) 1890 10 and above 86.7 (92.3, 43.2) 1910 10 and above 92.3 (95.0, 69.5) a. In some cases, the figures for whites and nonwhites, respectively, are reported in parentheses. Source: Engerman, Haber, and Sokoloff 2000. 117 INEQUALITY IN LATIN AMERICA of literacy. Even the most progressive Latin American coun- inequality is land policy. Virtually all the economies in the tries, such as Argentina and Uruguay, were more than 75 Americas had ample supplies of public land well into the years behind the United States and Canada. These societies 19th century and beyond. Since the respective governments began to boost their investments in public schooling at of each colony, province, or nation were regarded as owners roughly the same time that they were experiencing an eco- of this resource, they were able to influence the distribution nomic boom and intensifying their efforts to attract of wealth-as well as the pace of settlement for effective migrants from Europe, which was well before they imple- production-by implementing policies to control the avail- mented a general liberalization of the franchise. ability of land, set prices, establish minimum or maximum While this association might be interpreted as provid- acreages, provide credit for such purposes, and design tax ing for the socialization of foreign immigrants, it also sug- systems. Because agriculture was the dominant sector every- gests that elites may have been inclined to extend access to where, questions of how best to employ this public resource opportunities as part of an effort to attract the scarce labor for the national interest and how to make the land available for which they were directly or indirectly competing. The for private use were the subject of policy debates, if not polit- latter perspective is supported by the observation that major ical struggles, in many countries. Land policy also became investments in primary schooling did not generally occur an instrument with which to shape the labor force, because in any Latin American country until the national govern- it either encouraged immigration by making land readily ments provided the funds; in contrast to the pattern in North available or influenced the regional distribution of labor (or America, local and state governments in Latin America were supply of wage labor) by limiting access to and raising the not willing or able to take on this responsibility on their price of land. own. Most of these societies therefore did not achieve high There were never major obstacles to acquiring land in the levels of literacy until well into the 20th century. Fairly gen- United States, the terms of which became easier over the erous support was made available, however, for universities course of the 19,h century (Gates 1968). The well-known and other institutions of higher learning that were more Homestead Act of 1862-which essentially made land free geared toward the children of the elite. in plots suitable for family farms to all those who settled and Two mechanisms help explain why extreme levels of worked the land for a specified period-was perhaps the cul- inequality depressed investments in schooling. First, in set- mination of a policy of promoting widespread access to land. tings where private schooling predominated or parents paid Canada pursued similar policies: the Dominion Lands Act user fees for their children, greater wealth or income inequal- of 1872 closely resembled the Homestead Act in both spirit ity generally reduced the proportion of the school-age pop- and substance. Argentina and Brazil instituted similar ulation enrolled, which in turn held per capita income changes in the second half of the 19,h century as a way to constant. Second, greater inequality likely exacerbated the encourage immigration, but these efforts were much less collective-action problems associated with the establishment directed toward providing land to smallholders, and were and funding of universal public schools, either because the thus less successful, than the programs in the United States distribution of benefits across the population was quite dif- and Canada. ferent from the incidence of taxes and other costs or simply In Argentina, for example, a number of factors explain because the heterogeneity of the population made it more the contrast in outcomes. First, the elites of Buenos Aires- difficult for communities to reach consensus on such public whose interests favored keeping scarce labor in the province projects. Where the wealthy enjoyed disproportionate rather than in the capital city-were much more effective political power, they were able to procure schooling services at weakening or blocking programs than were their urban for their own children and to resist being taxed to under- counterparts in North America. Second, even those policies write or subsidize services for others. Although the children nominally intended to broaden access to land tended to of the elite may have been well educated, not everyone was involve large grants to land developers (with the logic that so fortunate. Furthermore, it is a fact that no society has ever allocative efficiency could best be achieved through realized high levels of literacy without public schools. exchanges among private agents) or transfers to occupants Another example of how the evolution of institutions in who were already using the land (including those who were Latin America contributed to the persistence of extreme grazing livestock). Policies in Argentina thus generally 118 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA conveyed public lands to private owners in much larger and TABLE 4.4 concentrated holdings than did policies in the United States Land ownership rates in rural regions of Mexico, the United and Canada. Third, the processes by which large landhold- States, Canada, and Argentina in the early 1900s (percent) ings might have broken up in the absence of economies of scale may have operated very slowly in Argentina. Once land Proportion of household heads was in private hands, its potential value for grazing may have Country, year, and region who own land, set a floor on land prices that was too high for immigrants and other ordinary would-be farmers to manage, especially Mexico, 1910 5.6 given the general underdevelopment of mortgage and finan- North 3.4 cial institutions. Central 2.0 Gulf 2.1 Argentina, Canada, and the United States all had an extra- South Pacific 1.5 ordinary abundance of virtually uninhabited land to trans- Total rural Mexico 2.4 fer to private hands in order to bring this public resource United States, 190079.2 into production and serve other general interests. In soci- South Atlantic 55.8 eties such as Mexico, however, the issues at stake with regard North Central 72.1 South Central 51.4 to land policy were very different. Good land was relatively West 83.4 scarce and labor was relatively abundant. Here the lands in Alaska/Hawaii 42.1 Total United States 74.5 question had long been controlled by indigenous peoples, Canada, 1901 but without the existence of individual private property British Columbia 87.1 rights. Alberta 95.8 Saskatchewan 96.2 Mexico was not unique in pursuing policies, especially Manitoba 88.9 in the final decades of the 19th and the first decade of the Ontario 80.2 20th century, that had the effect ofconferring ownership of Maritimeb 95.0 much of this land to large nonindigenous landholders. The Total Canada 87.1 1856 Ley Lerdo and the 1857 Constitution had established Argentina, 1895 Chaco 27.8 methods of privatizing public lands in a manner that could Formosa 18.5 originally have been intended to help indigenous farmers Misiones 26.7 enter a national land market and commercial economy. La Pampa 9.7 Neuqu6n 12.3 Under the regime of Porfirio Dfaz, however, these laws Rio Negro 15.4 became the basis for a series of new statutes and policies. Chubut 35.2 Santa Cruz 20.2 Between 1878 and 1908, regulations resulted in a massive Tierra del Fuego 6.6 transfer of public lands (more than 10.7 percent of the national territory) to large holders, such as survey and land a. Land ownership is defined as follows: In Mexico, household development companies, either in the form of outright grants heads who own land; in the United States, farms that are owner-operated; in Canada, occupiers of farmland who own for services rendered by the companies or for prices set by that land; and in Argentina, the ratio of landowners to the decree. number of males between the ages of 18 and 50. Table 4.4 presents estimates of the fractions of household b. The Maritime region includes Nova Scotia, New Brunswick, and Prince Edward island. heads (or a near equivalent) that owned land in agricultural Source: Engerman and Sokoloff 2002. areas in the late 19,h and early 20"h centuries in Argentina, Canada, Mexico, and the United States. The figures indicate enormous differences across the four countries regarding measures taken in Mexico at the end of the 19"h century may the prevalence of land ownership among the adult male have succeeded in privatizing most of the public lands, but population in rural areas. On the eve of the Mexican Revo- they left the vast majority of the rural population without lution, the figures from the 1910 census suggest that only any land at all. The evidence obviously conforms well with 2.4 percent of household heads in rural Mexico owned land. the idea that in societies that began with extreme inequal- This figure is astoundingly low. The dramatic land policy ity (such as Mexico), institutions evolved in such a way as 119 INEQUALITY IN LATIN AMERICA to greatly advantage the elite in terms of access to economic markets was likely another factor that reinforced inequality. opportunities, and thus contributed to the persistence of In his assessment ofthe caseofArgentina, Scobie (1964, p5.) extreme inequality over time. certainly thought this to be the case: In contrast, the proportion of adult males who owned land ' ~~~~~~~~~~~~Those whose forebears had been able to acquire and in rural areas was quite high in the United States, at just keep enormous land grants or who now secured estates below 75 percent in 1900. Although the prevalence of land enjoyed a gilded existence. Lands whose only worth had ownership was markedly lower in the South, where African been in their herds of wild cattle, land which could be Americans were disproportionately concentrated, the over- reached by horseback or oxcart, land occupied largely all picture is one of land policies (such as the Homestead Act) by hostile Indians underwent a total transformation. that provided broad access to this fundamental type of eco- British capital had built railroads. Pastoral techniques nomic opportunity. Canada had an even better record, with ' ~~had been improved and the resources of the pampas nearly 90 percent of household heads owning the agricul- tural lands they occupied in 1901. The estimates of land- werined mro ughly. Imrants, newly arrived from European poverty, were available . . . as holding in these two countries support the notion that land .. . . ~~~~~~~~~~~sharecroppers, tenant farmers, or peons to raise policies made a difference, especially when compared to corn, wheat, or alfalfa. Under such conditions, land Argentina. The rural regions of Argentina constitute a set provided an annual return of from 12 to 15 percent to of frontier provinces, where one would expect higher rates ' ~~~~~~~~~~the owner and land values often rose 1000 percent in of ownership than in Buenos Aires. The numbers, however, ' ' ~~a decade. Those who already had land, power, or money suggest a much lower prevalence of land ownership than in monopolized the newly developed wealth of the the two North American economies. Nevertheless, all of these countries were far more effec- pampas. tive than was Mexico in making land ownership available By creating new market opportunities in Latin America to the general population. The contrast between the United and increasing the value of land, the integration of world States and Canada, with their practices of offering easy access markets may have stimulated efforts by elites to influence to small units of land, and the rest of the Americas (as well the sorts of land policies discussed above, with the aim of as the contrast between Argentina and Mexico) is consistent capturing a greater share of potential gains for themselves. with the argument that the initial extent of inequality influ- This flexibility in altering institutions to serve the purposes enced the way in which institutions evolved, and in turn of the politically and economically powerful in response to fostered persistence in the degree of inequality over time. changing circumstances is also evident in many other areas Land policies in Mexico and Argentina reflect another of economic policy, such as the regulation of financial insti- general pattern in the way that institutions evolved. As tech- tutions and foreign trade. What was good for the elites was nology and market conditions changed, institutions adjusted likely not good for reducing inequality or spurring economic because societies-as well as the segments of the populations growth. In terms of these aspects, the Latin American that had real influence-sought to take advantage of the new economies fell far behind their neighbors to the north over circumstances. The late 19th century was a period of falling the course of the 19,h century. transportation costs and extraordinary growth in world trade. Among the greatest beneficiaries of this development were 4.4. The 20th century relatively land-abundant countries in the Americas, such as This chapter has suggested that Latin America entered the Argentina and Uruguay, which were now in a better position 20,h century with very high levels of inequality, which vir- to exploit their strong comparative advantage in agriculture. tually all of the evidence discussed elsewhere in this volume Increased access to vast international markets in chilled beef, indicates persisted during the rest of the century. For exam- grains, and other agricultural products dampened ratios ple, the estimates of Calvo, Torre, and Szwarcberg (2002) for of wage rates to land rents (see the estimates of Williamson Argentina suggest that the extent of inequality changed little 1999, O'Rourke and Williamson 2002, and Bertola and during the 20,h century. This is consistent with the evidence Porcile 2002). Given the already high concentration of land in Diaz-Alejandro (1970) and Randall (1978). A study of ownership, the effect of this improved integration of world inequality in Colombia since the 1930s by Londoiio (1995) 120 HISTORICAL ROOTS OF INEQUALITY IN LATIN AMERICA likewise argues that inequality in the 1990s was at about appear to be uncorrelated with economic growth, which is the same level as it was in 1938. Similarly, a study of income consistent with the idea that educational expansion can take distribution in Mexico by Reynolds (1970) found that little place without radically transforming inequality. The most change occurred between the 1940s and the late 1960s. plausible explanations for Pritchett's findings are either that Explaining the persistence of inequality is again prob- educational quality is often so low that there is no impact lematic. During the course of the 20" century, Latin Amer- on development, or that the social payoff from education ican societies underwent significant social, economic, and depends on other institutions in society. This latter idea may political changes. Urbanization increased considerably and also be important for explaining the disappointing impact the role of agriculture declined. Thus, the highly inegali- of educational expansion on inequality in Latin America. tarian pattern of landownership that characterized Latin Despite the increase in education, there are still deeply America ought to have become less and less significant as a ingrained patterns of social exclusion and discrimination determinant of income inequality. In contrast, educational in Latin America that are hard to break down and which attainment still lagged behind that of North America, and prevent poor people from becoming upwardly mobile. This there was a large amount of catching up to do, as the num- may be why even apparently radical changes, such as the bers in table 4.3 indicate. Bolivian revolution, end up having little effect on inequal- In addition, modernization moved nearly all countries in ity (see Kelley and Klein 1981). the direction of more open and democratic politics. Even if Another possible explanation for the failure of educational democracy was perhaps relatively shallow and unconsoli- reform to reduce inequality is the type of development strat- dated, it seems to have undergone a significant change egy that evolved in Latin America starting in the 1930s. This compared to the 19'h century. Moreover, several countries, approach, which was based on looking inward, import sub- including Bolivia, Mexico, and Nicaragua (and possibly stitution, and isolation from world markets, may have Colombia, depending on the interpretation of La Violencia, reduced incentives for investments in education and skill the period of violence in the 1950s) underwent dramatic building. It is significant that countries that adopted export- social revolutions, while others, such as Peru after 1968, oriented development strategies in the 20,h centuries (for implemented radical redistributive policies. example, Mauritius and Singapore) often also experienced This pattern in Latin America also contrasts with what declining inequality. we know about Europe, where economic inequality appears These effects may have been reinforced by the factor to have fallen dramatically during most of the 20"h century endowments of Latin America. Even in the 20"h century, (Morrison 2000). It is also different from the pattern seen many economies were dominated by the export of primary in the United States, where measures of inequality indicate commodities, such as oil in Venezuela; coffee in Brazil, a marked decline from the 1930s through the 1960s, Colombia, and Central America; bananas in Central before a recent upturn began (see, for example, Goldin and America; sugar and bananas in the Caribbean; and copper Katz 1999). and nitrates in Chile. It has been argued (for example, by What accounts for the radically different courses of coun- Leamer and others 1999) that commodity-intensive devel- tries throughout the Americas? Why did inequality appar- opment reduces the returns on education, which may have ently remain unchanged in the face of the types of social and been another factor in reducing the impact of educational economic transformations that led to falling inequality else- expansion on inequality. where? Unfortunately, there is little research on this topic, Finally, it is also possible that in the presence of imper- and given the severe limitations of the evidence it is possible fect financial markets, changes in the structure of the only to offer some intelligent, but nevertheless speculative, economy from agriculture to industry have little effect on conjectures. (These are explored in somewhat more depth in inequality. Although there are no comparative data on this the next two chapters of this report.) connection, it seems likely that the ownership of industry First, it is surely the case that mass education has been in Latin America is much more concentrated, and financial of low quality in Latin America, especially when compared markets much less developed, than in North America. Haber to North America. Pritchett (2001) has pointed out that mass (2002a) has shown how banking was very tightly controlled expansions of educational enrollment in the last 40 years in Mexico compared to the United States. For example, in 121 INEQUALITY IN LATIN AMERICA 1910 the United States had roughly 25,000 banks and a historical processes tell us about the future evolution of highly competitive market structure. In contrast, Mexico had inequality in Latin America? Certainly the recognition of the 42 banks, 2 of which controlled 60 percent of total bank- powerful factors that shaped and allowed extreme inequal- ing assets and virtually none of which actually competed ity to persist for centuries in Latin America does not inspire with another bank. This example, and others like it, sug- optimism, but a historical perspective reminds us that gest that financial markets in Latin America were also very changes that take a generation or two to be realized should thin and monopolized (see the essays in Haggard, Lee, and perhaps not be considered to be so slow. Although it is Maxfield 1993). Without access to financial markets, poor impossible to conceive of a quick-fix solution, there is abun- people stayed poor. dant and impressive historical and cross-country evidence Although the focus of this chapter is economic inequal- that democracy (or political equality), equality in the dis- ity, it is hard to separate this issue from political inequal- tribution of human capital, and economic equality tend to ity. Political inequality both causes and is caused by develop together. economic inequality. Just as in the 19th century, one of the It is generally difficult to attack economic inequality central reasons for the persistence of inequality in the 20,h directly (for example through redistribution measures) with- century may be the prevalence of authoritarianism in Latin out harming some of the incentive structures thought crit- America. Although democratization has taken place ical to ensure processes of economic growth. For that and throughout the region, this process has usually been uncon- other reasons, more indirect policies that would broaden solidated and limited in what it can achieve. effective access to economic opportunities-not just for Moreover, those popular democratic parties that have been the current population, but, more importantly, for future able to obtain power have also often been driven by narrow generations-would be preferable. Such policies would political considerations and opted for policies that were dra- include not only major investments in extending access to matic in the short run but conveyed little in the way of last- high-quality education, but also efforts to provide more equal ing benefit. For example, in Argentina, despite the fact that access to political influence and to the protection of laws policy under the first administration of Per6n in the late more generally. The historical experiences of Europe and the 1940s and early 1950s significantly reduced inequality, these United States suggest that the most reliable ways to induce effects quickly unraveled when the military took over in falling inequality are to promote sustained educational 1955 and did not lead to lasting improvements in terms of investment and asset accumulation by the poor and to ensure inequality. Chapter 5 discusses these issues further, avenues for social and economic mobility. In general, this may entail solving market and other institutional failures 4.5. The 21t century and beyond that discourage the poor and their children from making the If the roots of Latin American inequality lie deep in history, sorts of investments that would help them succeed in an is there anything that can be done about it today? What do increasingly global economy. 122 CHAPTER 5 State-Society Interactions as Sources of Persistence and Change in Inequality HE PREVIOUS CHAPTER EXAMINED HOW INITIAL FACTOR ENDOWMENTS AND EARLY INSTITUTIONAL development played a central role in the generation of egregious levels of inequality in Latin America. The next two chapters consider the nature of the political and economic processes that continue to shape inequality in the region today. This chapter discusses both political and social factors and seeks to fill in the interactions between social and political institutions, power, and governance that were summarized in two sides of the triangle used to portray the conceptual framework presented in chapter 1 (figure 1.1). In assessing both current problems and possibilities for change, the current chapter's primary focus is on the perfor- mance of the state, which is viewed as a product of historically-formed state-society relations. The state is examined in terms of a syndrome, or malaise, that is characterized by unequal social relations and "weak" governing capacities. "Weak" has a specific meaning: doing a poor job of providing public goods such as macroeconomic stability, property rights and citizenship, and basic services such as education, health, water, sanitation, roads, electricity, and social protection. In all of these areas, failures in state action are typically countries, as indicated in table A.27). At the same time, regressive, hurting the poor and weak disproportionately. while most countries in the region have made substantial They are also, in most areas, bad for overall development. progress in expanding basic education coverage, a major chal- Where states fail in providing publicgoods, wealthy members lenge related to school quality remains unsolved for most of society typically can either exert private influence over the of the public school systems in the region (de Ferranti and state for selective delivery, or can opt out of the system. For others 2003). Wealthy groups again have greater options for example, elites can shift savings abroad and develop private either exerting influence on public service provisioning for means of protecting property rights and their personal secu- themselves or opting out for private provision. rity. Poor and middle groups live with the consequences. Similar patterns can be identified with regard to other ser- With respect to services, some Latin American states his- vices. Although most Latin American states have developed torically did a reasonable job of providing services for elites social welfare systems since the early 20th century, these are (for example through the early formation of universities in often best characterized as "truncated" systems that only many countries). However, most did a poor job of broad- reach formal sector workers, leaving out the neediest seg- based provisioning; by 1925, only Argentina and Uruguay ments of the populations (see chapter 9). These issues are had achieved literacy rates of 70 percent or higher (see tightly linked to the tax effort, which reflects the underly- table 4.3). In 2000, literacy was still only 65 percent in ing implicit social contract that conditions the willingness Guatemala, 77 percent in Nicaragua, and 87 percent in of all groups-but especially elites and middle groups-to Brazil (and considerably lower for poorer groups in all pay taxes in return for effective state action. 123 INEQUALITY IN LATIN AMERICA Why has this pattern of weak and truncated state actions the issues of state-society relations in Latin America, especially persisted? Addressing this question involves taking some ten- in the fields of political science and sociology. This chapter does tative steps into a complex terrain. There is considerable het- not seek to present new theories, but rather has the more erogeneity in conditions across the region, and it would be modest objective of drawing on major strands in the existing impossible to do justice to the issues in a short chapter. Yet, literature-on both Latin America and elsewhere-to help as seen in chapter 4 on the historical formation of high inequal- inform understanding of the ways in which state-society rela- ity in Latin America, the links among economic, political, and tions can be a force for both the persistence and transforma- social conditions are fundamental. Attempting to understand tion of economic, social, and political inequality. these links is central to any enterprise that seeks to draw con- To motivate the discussion, figure 5.1 shows bivariate cor- clusions on policy options. There is an immense literature on relations between selected indicators of social difference, state FIGURE 5.1 The syndrome of weak and unequal states: correlaions between the rule of law and development outcomes a. Governance b. Social cleavage Government effectiveness index Ethnolinguistic fractionalization index 3.0 1.0 0.9 y = -0.1064x + 0.4403 2.0 R2 =0.1697 .BH RS.... SRgOL LZ. %~~~~UB ~ ~ L .MEX . P &;ER.)Y A .SUR EA. BA -1.5 -4. 0H . BLZO.5 1.0 1.5 2.0 2.5 *BH :, ... :3XDMoA Rule of law * .vc *TW s PYY 0 * DMA sA"GRD.UR'I?, -2.0 y 0.9299x + 0.0068 *ARD * - LCl . BRB R 0.8607 .HTI * .* -3.0 - 0.0- I -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Rule of law c. Volatility d. Fiscal effort Standard deviation of growth Total tax revenue (as a share of GDP) 16 45%- 14 - y = -1.3077x + 4.9985 40% - R2 0.2328 . . * 12 - 10~~~~~~~~~~~~~~0 NIC . . LCA . 2.0 /0 ^~~~~~.-B -BO ; loi_. ARG la~ a' - *10% *2 ; y 0.0609x + 0.1835 , I I O I I I - 0% -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Rule of law Rule of law Sources: Governance Indicators Database for rule of law, Alesina and others 2002 for ethnic fractionalization, and World Development Indicators Database for other variables. 124 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY capacity, and economic performance, with the "rule of law" between historically-formed, unequal social and political used as an anchor for the graphs. An ethnic fractionalization structures and relatively weak and regressive states (as evi- index is used as a (highly imperfect) measure of unequal social denced in the areas of the provision of public goods and basic relations, though similar results are obtained with a Gini coef- services, noted in the introduction to this chapter). In some ficient of income inequality. The rule of law can be thought cases-for example the Dominican Republic, El Salvador, of as a rough proxy for a state's performance in providing the Guatemala, Honduras, Nicaragua, or Paraguay for much of public goods of protection of property and citizenship rights, the past century-this has typically taken the form of tra- government effectiveness is a proxy for bureaucratic capacity ditional authoritarian or oligarchic regimes characterized by to provide broad-based public services, and macroeconomic profound exclusion of the bulk of the population, as well as instability and tax effort are two of the outcome measures. high levels of state capture by small groups of elites.' These are only correlations, but they may capture the two-way In Argentina, Brazil, and Mexico, where industrialization patterns of causality between inequality and state performance. advanced to a relatively greater extent, significant processes of The remainder of this chapter is divided into two main sec- incorporation of the working classes into the political struc- tions. The first discusses the origins of weak and unequal state ture occurred. However, this trend was both partial (that is, pri- performance in state-society interactions, including a discus- marily of only the formal working class) and effected in a way sion of why the process of democratization has been halting that tended to perpetuate relations based on patronage, which and lengthy, as well as insufficient to effect a transformation typically created vested interests among middle-class groups in political and social inequalities. The second section turns as well as elites (Cohen 1989, Middlebrook 1995, O'Donnell to the question of how to effect change in order to achieve more 1973). In countries such as Bolivia, Ecuador, or Peru, where effective and equal states, a condition that is characterized here deep and complex social and ethnic divisions have existed, the as the move toward a new "political equilibrium." processes of incorporation have tended to be even less complete, even for the formal working class, and to be marked with con- 5.1. Political and social structures as forces flict. Traditional oligarchic politics was often replaced with for the reproduction of inequality "populist" outsider movements (for example, Bolivia's 1952 Chapter 4 on the historical roots of inequality in Latin America revolution, Ecuador's Velasquismo, and Peru's Aprismo), which posed a puzzling question: why has inequality persisted in the raised hopes among the underprivileged without building an region, in spite of significant social, economic, and political effective state to cater to their general demands.2 changes? This section addresses this question and discusses While formally operating in electoral democracies, elite- how enduring features of political institutions and social struc- dominated political parties in Colombia and Venezuela in tures have limited the scope for equalizing public actions, and the second half of the 20,h century grew increasingly dis- thus contributed to reproduction of inequality in the region connected from their popular support bases and pursued long after its colonial origins faded. Has inequality in the patronage-driven policies.3 The challenge of political incor- region persisted simply because of the lateness of changes (espe- poration was exacerbated by weak democratic consolidation, cially the arrival of universal suffrage), or are there other forces as well as periodic unconstitutional transitions, especially that are preventing democracies in the region from becom- during the Cold War period. Chile, Costa Rica, and Uruguay ing more potent drivers of equalizing changes? It is argued to varying degrees developed states that were, when under that while the late arrival of competitive democracy is a rel- democratic auspices, more effective in providing public evant fact, the evolution of the institutional structure in the goods and redistributive services, although both Chile and region in response to, and in defense of, the predominant elite Uruguay suffered periods of repressive authoritarian rule.4 interests has been a fundamental cause of the persistent inequality, rather than the type of regime in place. Unequalsocial relations and weak states: twin Institutional evolution has taken a variety of forms within facets of negative institutional legacies the region. However, the underlying characteristics of state relationships have been strikingly similar across different pat- Unequal relations and clientelism terns of political and social development in the region. The The persistence of unequal social relations between elites main commonality is the theme of a two-way causal relation and poorer groups is classically manifested in the form of 125 INEQUALITY IN LATIN AMERICA clientelist relations, which are characterized by the unequal, Republic) and in poorer sub-regions such as northeastern although reciprocal, exchange of favors, benefits, or rents for Brazil or the southern states of Mexico. However, clientelism support or acquiescence.5 In a clientelist system, citizens do is not by any means peculiar to Latin America or developing not base their support for candidates and parties according countries. It has been an important aspect of the political to their political preferences in particular issue areas; instead, history of many (perhaps most) Organisation for Economic their support is contingent upon particularistic material Cooperation and Development (OECD) countries, for exam- incentives. Rather than pursuing public policies that ple in the political machines of U.S. cities and at both a address the shared needs of a broad segment of society (as local and national level under Christian Democrats in Italy might a disciplined programmatic party with broad geo- (Shefter 1994). Where Latin America is distinctive is in the graphic coverage), clientelist politicians generally draw sup- depth and resilience of clientelist structures. port from relatively narrow, geographically-defined An additional, particular feature of the relationship constituencies and concentrate on providing tangible mate- between states and subordinate groups in many Latin rial goods (that is, support for patronage) that is narrowly American countries in the mid 20'h century was a more formal targeted to a specific group of supporters in exchange for "corporate" form of inclusion, in which governments brought their support.6 Clientelist practices enable the poor to make working class or peasant movements into national, institu- isolated efforts to attain local goods, but these rarely form tionalized, patronage-based systems. These processes were the basis of broad-based policy initiatives that address the typically marked by co-option. Such conditions were exem- underlying causes of poverty and inequality. plified by the highly "successful" incorporation of labor, The other side of clientelistic, or patronage-based, sys- peasants, and other functional groups in society in Mexico, tems is the absence of broad, organized alliances of social which was fully institutionalized by Lazaro Cardenas into a groups pressing for sustained improvements in their col- remarkably stable form of state-party organization under the lective welfare and public services (for example, general Partido Revolucionario Institucional (PRI). In Argentina, Juan improvements in mass education). This condition has Per6n successfully co-opted a previously independent union reduced incentives for political elites to strengthen the capac- movement into the corporate wing of Peronism. Co-option ities of the state to provide such services. Strengthening the of groups was also reflected in the less successful attempts by state in this sense would have required considerable invest- the Bolivian Movimiento Nacional Revolucionario to emulate the ment to improve its capacities to raise revenues through tax- PRI-style form of governance and state formation after the ation and to organize its administrative apparatus. Given that 1952 revolution, which succeeded in first bringing the peas- the elites already had selective access to public goods, such ant movement under the patronage of the national movement, as security of property, and to basic services (that is, either but failed to provide the economic and services support that through the market itself or through appropriation of the might have made land reform a success (see chapter 7 on land limited state functions), it was often not rational to make reform). these investments in state building. In cases of the partial In most cases, however, these incorporation processes incorporation of poorer groups into national systems of ser- simply transferred unequal social structures to the formal vice provision, especially in the middle-income countries, political arena by reproducing unequal vertical relations at new vested interests were created, for example to sustain the the national political level. Although corporatism is increas- "truncated" welfare state. ingly a phenomenon of the past, its heritage persists in many The importance of clienrelism in both rural and urban areas that are relevant to current policy, including widespread settings has been documented throughout the region, for vested interests that have accrued to the formal working example in Brazil (Leal 1948, Carvalho 1997), Mexico classes and public sector servants. This phenomenon is evi- (Cornelius 1977, Grindle 1977), Colombia (Schmidt 1977, denced in dualistic labor structures (see chapter 8) and the Archer 1990), Peru (Stokes 1995), and Venezuela (Powell truncated welfare state (see chapter 9). 1977). More traditional forms of clientelism, often linked to Clientelism and corporatism-and more generally the powerful local personalities, are probably of greater importance vertical forms of inclusion of subordinate groups-have also in the relatively less developed parts of the region, that is, left a damaging legacy in terms of political organization. It in less developed countries (see box 5.2 on the Dominican worked against the development of programmatic parties 126 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY (either those that represented working classes or more services are provided by the government. (See box 5.1 for conservative forces), which was an important part of the evidence from the United States and Europe.) political evolution of most European countries and the United States (Shefter 1994). Chile is one of the exceptions Weak state capacity to this pattern in Latin America, in the sense of having a The implementation of political actions to reduce inequal- long history of genuine programmatic parties, especially ity, either through the overt redistribution of assets or the those that represent both the middle- and working class. As efficient production of public goods, requires adequate discussed below, political scientists attribute Chile's relative institutional capacities of the state apparatus. Most Latin success in developing an effective and redistributive state American countries have relatively "weak" states relative to after its transition to democracy in part to the continuity of OECD comparators and other developing countries of sim- party performance originating in the country's earlier his- ilar income levels, and therefore typically score low on a tory (Valenzuela 1977). range of governance indices, for example the rule of law, gov- In some cases, corporatism allowed for more blatant forms ernment effectiveness, and corruption (Kaufmann, Kraay, and of political subordination, as well as state capture by some Mastruzzi 2003). As noted above, this means, in particular, of the more privileged interest groups. Political scientist relative weaknesses in providing the public with goods such James Malloy (1979) describes how the corporatist arrange- as the rule of law, citizenship, and protection of property ments in Brazil's social security systems-initially developed rights, and with basic services such as quality primary and by the authoritarian government of Getulio Vargas as a way secondary education.' Such weaknesses are reflected in the to preempt radicalization of the working class-were even- difficulties a number of the states in the region face with tually captured by the organized labor in the formal sector regard to raising revenues. as the movement gained a series of policy concessions from The potential for the state to enact pro-equity changes the state, as well as privileged access to state rents at the cost is also constrained by informal political realities, whereby of workers outside formal organizations.7 the excessive influence of elites and some well-organized Although unequal, vertical social relations have been a middle-class groups (such as unions) on public decisions general phenomenon in Latin America, in many countries often results in the effective "capture" of the state. For public they have been overlaid and reinforced by ethnic and racial actions to be effective against entrenched interests, the state divisions. This is most apparent in Brazil, much of the itself needs to be more or less autonomous from such influ- Caribbean, the Central Andes, parts of Central America, ence. Many Latin American states have historically faced Colombia, and Mexico. As discussed in chapter 4, these difficulties with state capture in its most blatant form, as in divisions reach back to profoundly unequal conditions that the oligarchic domination by the so-called 14 families in emerged in the colonial period. Such relational inequalities El Salvador, the dynastic rule of the Somozas in Nicaragua, have had major effects on processes of socialization, collec- the long personalist rule of Stroessner in Paraguay, the tive expectations, and daily interactions, in turn affecting "partidocracia" in Venezuela, and the recent case of behaviors and conditions related to school, work, and social Vladimiro Montesinos in Peru during the 1990s. In addi- interactions. tion, more subtle, everyday manifestations of the uneven For example, in an ethnographic study of poor black application of the rule of law, which discriminates against women in a sugarcane-producing area of northeastern Brazil, the poor, also occur. Scheper-Hughes (1992) documents the harsh dilemmas faced International evidence of the salience of state capture by the poor in seeking to manage their daily lives, with many comes from international surveys of executives by the hoping to find the "good patron" who will solve their prob- World Economic Forum. Representatives of firms were lems. Such stories are repeated in sociological and ethno- asked questions on the extent in their countries of exces- graphic studies in the region (see Calder6n and Szmukler sive or illegal influence of powerful conglomerates, firms, 2004). Such group-based differences continue to affect both or individuals on state policymaking (figure 5.2). Such the design and implementation of policy and, as in other unequal influence (or "crony bias") is correlated with a more parts of the world, magnify the challenges of forming the negative assessment of the fairness and impartiality of social and political base through which public goods and courts and the enforceability of court decisions, less secure 127 INEQUALITY IN LATIN AMERICA BOX . I Raci leqmaly mud sodal spending: evidence fot th Urtd States and Europe The impact of grnup-based differences is not limited to which a histonr of racial stigma can continue to have Larin America and the Caribbean: it has had a profound salience in contemporary behavior and policy-making. effect on social policy in the United States as well. Alesina, Baqir, and Easterly I 1999) find lower spending Although the Linited States and Europe enjoy compara- on publicly provided services in less racially homogeneous ble levels of economic prosperity, the United States invests areas. -iince universal programs would benefir rhe popu- considerably less of its resources into social programs and lation as a whole, including the "undesering" poor, aver- public goods The Linited States spends approximately age Americans tend to vote for more limited government 16 percent of its gross domestic product (GDP,) on social expenditures on transters or services targeted co the poor. programs. in contrast to the European average of 25 per- In contrast. racial differences were historically much less cent. Not surprisingly, the United States also registers important to oxerall sucial interactlons and policy formula- levels of income inequalirv that are substantially higher tion in European narionsduring the 20"bcentury. Europeans than its European counterparts. A quantitative compari- were more likely roview thepooras unfortunareand did not son of the LUnited States and Europe reveals the influence stigmatize them as members ofa socially undesirable group. ot a history of racial inequaliry on these phenomena (see This historical absence of racial inequalicy helps explain the Alesina, Glaeser, and Sacerdore 2001). differing levels of social spending in the ULnited States and Since African Americans are overrepresented in the Europe. forerecent racialandethnic tensionsin Europeand, lowest income brackers. the whire majority tends to view in some countries, the related rise of far-right. anti-immi- social spending and red ist ribut ve policies as dispropor- grant parties, further illusrrates the point rhar the central issue tionatel y benefiring minorities, effectively undercutting ot' concern is the interact ions bewmeen group-based difterences support for policies that would reduce levels of inequal- and policy choice. and not an intrinsically 'European' pref- it)." Loury (2002) has explored the often subtle ways in erence tor redistributive policies FIGURE 5.2 Political financing and its impact, by type % of survey respondents 100 Pervasive 55- Not a problem t 10 °0 sttKN A,,t O 9 # 1: Illegal political financing M Legal political financing influences policy M Firms' capture of laws/policy/regulation Source: World Economic Forum 2003. 128 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY property rights, lower levels of tax compliance, and higher spheres by the rich and the powerful, including in some cases levels of bribery. by organized segments of the middle classes. This trend has occurred at the cost of poorer and subordinate groups, who Views of firms on unequal influence over policies, have lacked the power or wealth to exert such influence. in international perspective Property rights are of particular interest in light of their The importance of state capture in the region is supported potent influence on investment and growth. In some by careful documentation of key episodes in the history of countries where general protection of property rights has lawmaking in Mexico by Haber (2002a) and by country- been absent, alternative means of protecting the elites' rights specific surveys. For example, in surveys of firms and public emerged. officials in Peru in 2001, some 80 percent of respondents In a major study of Mexico, Haber, Maurer, and Razo judged that there was a "high impact" of bribes given to (2003) use historical, political, and economic analysis to show parliamentarians to shape laws, the judiciary to influence deci- how property rights were selectively enforced, especially in sions, and public officials to shape ministerial decrees. A the turbulent period of 1910-1929. Governments made slightly lower proportion of respondents reported significant selective, credible commitments to subsets of asset holders, levels of "state capture" of affected regulatory agencies, and guaranteed by third-party actors who shared in the rents. much less with regard to the Central Bank (see figure 5.3). The third-party actors varied with the asset and period, A similar pattern was found in a survey of firms in Colom- including local military factions (for example, agricultural bia, but with "only" about 60 percent reporting influence on products during the revolution), the United States (for exam- the judiciary. ple, for oil), and the union movement (in the industrial The weakness of the rule of law and property rights has sector). These arrangements were institutionalized in the cor- in turn led to the partial "privatization" of these public porate state created under the PRI and supported investment FIGURE 5.3 Impact of bribes in shaping policy, according to a survey of firns by a recipient group, Peru, 2001 Groups taking bribes Parliamentarians (to influence laws) Judiciary (influencing major g E court decisions) High officials (influencing presidential decrees/decisions) Regulatory agencies El Firms Public officials Central Bank decisions 40 70 100 Percentage of survey respondents Source: Kaufmann, Recanatini, and Biletsky 2002. 129 INEQUALITY IN LATIN AMERICA and rapid growth that lasted till the 1980s debt crisis. How- with a national reach, individual politicians, including the ever, this economic improvement came at the cost of a pat- president, cannot credibly offer a major public policy ini- tern of development that was both inefficient and tiative (for example, educational reform) and expect to receive inequitable, since selective protection of rents and rent- electoral support. Cognizant of the governing elites' inabil- sharing were central to the institutional basis for growth."0 ity to commit to a major policy reform, the electorate would A more extreme case of elite influence and state capture prefer to vote for locally prominent candidates who promise is Guatemala. Starting in the late 1800s, landed elites and to deliver tangible public works or public employment the military have exercised strong influence over the programs. Guatemalan state (Heller and Mahoney 2003). The 1870s were marked by the rise of export agriculture, controlled pri- Institutional legacies of inequality and the limits marily by landed plantation coffee elites. The state encour- of democratization aged the growth of the coffee industry by catering to the This section focuses on the process of democratization in needs of coffee elites, providing infrastructure to service Latin America. On the one hand, an authoritarian legacy (see export agriculture and implementing coercive labor policies chapter 4) and the dynamics induced by fragmented social over the indigenous population to secure plantation labor. structures and weak states led to the delayed and halting con- The economic base of the elites has evolved, but high levels solidation of democracy, with periodic bouts of unconstitu- of state capture and exclusion have persisted to the present tional threats and authoritarianism (especially during the day. In the second half of the 20th century, the exclusionary Cold War period). On the other, when democracy did return politics unfolded in violent conflict and extreme repression to the region in the 1980s, it proved insufficient to trans- by the state, especially of indigenous groups. These events form the region to one with genuinely equal citizenship and finally culminated in the 1996 Peace Accords, but even today effective, inclusive states (O'Donnell 1998b). the Guatemalan state's capacity for public goods provision These processes can be understood in terms of the broader is quite limited. Guatemala collects only 8 percent of gross question of how poorer and subordinate groups are incor- domestic product (GDP) in taxes and spends 6 percent of porated into the polity. The ineffectiveness of politically GDP on social sectors. legitimate representation mechanisms, especially broad- Costa Rica provides a contrasting example: since the based programmatic political parties, in including the poor 1870s, the state has maintained substantial autonomy from has often created tensions and dynamics that were inimical dominant class actors, including the coffee elite. (See Paige to sustained redistributive change (Collier and Collier 1991, 1997 and Mahoney 2001 for comparative discussions.) This Rueschmeyer, Stephens, and Stephens 1992). These same pattern was coupled with increasing political openness, the dynamics have had significant impacts not only on frequent mobilization of class interests through political parties, and regime changes in the mid-20"h century throughout Latin the channeling of demands through democratic institu- America, but also the prospects for democratic consolidation tions. In turn, the societal demand for services rose, to in more recent years (Collier and Collier 1991). which the state responded with programs of broad-based When social and economic modernization began to put provisioning, notably in basic education. Costa Rica raises strains on the traditional exclusionary politics in the early 18 percent of GDP and spends 17 percent of GDP from 20"h century, political elites and society responded in a vari- taxes and other sources on social sectors (see table 9.2 and ety of ways to this new dynamic. As discussed above, in more figure 9.3). economically advanced countries (including Argentina, The case of the Dominican Republic further illustrates Brazil, and Mexico), elites attempted to incorporate "pop- how an extreme form of clientelism can combine with other ular" sectors (primarily organized urban labor) through state- institutional characteristics to crystallize in a political equi- dominated corporatist arrangements. In some cases (notably librium of a small state with limited service provision (see Argentina and Brazil, but also Chile and Uruguay), these box 5.2). In sum, clientelism induces politicians to provide incorporation processes led to the radicalization of segments targeted goods (such as small-scale public works) in highly of the popular sector, which in turn triggered authoritarian visible ways to specific constituencies (such as local com- reactions from the military, backed by the countries' munities). In the absence of a programmatic political party economic elites. 130 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY BOX 2 2 CUentmusm and lMe munderpmrslon of public sevkes In the Domican Rpbik Politics in the Dominican Republic has been characterized taken relatively little public spending and concentrated by a strong clienrelist tradition, whereby parries rarely it in the presidency. For example, in the Dominican compete for votes on the basis of national policy issues, Republic, almost 30 percent of total public investments and by strong presidentialism, whereby presidents dom- in 2000 were executed by the Ministry of the Presidency. mnare the narional political scene but rely on the support Presidents could push for improved public sector finan- of local political bosses to obtain votes. A combined effect cial management to better monitor spending by line min- of these institutional characteristics is significant empha- iscries, in principle increasing presidential willingness to sis on government provision of targered goods (in other allow more delegated spending and greater accountability. words, public works), low spending on public goods (for However, they are reluctant to do this since (unlike Alberto example, education), and low overall spending. Fujimori in Peru) they have only a limnied ability [o counter The clientelist tradition in the Dominican Republic is efforts by rivals to use resources as they see fit, even if a the legacy of 30 years of the repressive, personalist dicta- sophisticated system of public sector financial management torship of Trujillo from 1930 to 1961, which stifled the makes those uses completely rransparent The inability to development of credible political parties with clear and tightly control expenditure execution, however, gives the credible stances on public policy issues Absent disciplined president a strong incentive to keep overall expenditure parties with clear policy reputations. presidents face greater levels low. In the end, rhe government provides highly challenges from political rivals even within their own party targeted parronage goods and under-provides public goods and cabinet. These rivals are always interested in laying such as high-quality public primary education thar could their hands on portions of public expenditures in order to have made a dent in inequality in the country. provide patronage goods to their own personalized con- srituencies. Responding to this, presidents have under- Sor. Keefer 2002 In countries where initial incorporation was more limited periodically emerged with promises to the masses that were (such as Ecuador and Peru), political exclusion offered a fer- not mediated by institutional structures. However, populist tile ground for populism, as exemplified by the leftist mili- leaders were in most cases ineffective precisely because of tary regime of Velasco and the Aprista administration of their charismatic, personalist status, and were incapable Garcia in Peru or by new parties and "outsider" movements. of effecting change in overall patterns of privilege and In earlier periods, this trend was evidenced by periodic indige- institutional processes. In the case of Velasco in Ecuador, nous insurrections (for example, in the Yucatan, Mexico) or Conaghan (1995, p. 446) comments: "Velasco's early success in millenarian movements of withdrawal, such as the Canudos was the product of his close ties to traditional elites. Unlike movement in northeastern Brazil. In the 20'h century, polit- classic populism in Latin America (for example, Peronism), ical exclusion and populism were manifested in violent civil Velasco never produced real material and political advances conflict, notably in parts of Central America, Bolivia, and for lower-class supporters . . . He did, however, school an Peru. Especially in the Cold War, post-Cuban Revolution entire generation of politicians in how to win political power atmosphere, social tensions tended to spur repressive, without building organizations or refining party platforms. authoritarian reactions, as in Guatemala (1956), Chile (1973), His career seemed to underscore the unimportance of party and El Salvador (1980s), to name a few of the more violent and the dominance of personality . . ." In other countries, examples. reforms (such as the Peruvian land reform) were implemented Populism did have redistributive aims-or at least incompletely, failed to resolve distributional struggles, and rhetoric-manifested in both authoritarian and democratic spilled over into macroeconomic instability (notably in regimes, but the movement was also a symptom of unequal, Argentina, Bolivia, and Peru). The result was that previous vertical institutions through which charismatic leaders "gains" in redistribution were reversed. 131 INEQUALITY IN LATIN AMERICA The contemporary history of Argentina vividly illustrates above rather than the result of working-class demands." Fol- how a nation's inability to successfully deal with the ques- lowing the ouster of Per6n in a coup, democratic and mil- tion of political inclusion has stymied the development of an itary regimes alternated until a more extensive period of effective state, with periodic unresolved distributional strug- democracy set in during the 1980s. The second period of gles and a steady worsening of inequality (see chapter 2). In rising labor share, 1967-1974, was one of relatively good Argentina, the incorporation of the poorer groups (especially economic performance and strong labor union pressure, both urban workers) started in the 1940s under the populist under dictatorship and the elected Peronist government. regime ofJuan Per6n, who mobilized industrial workers into However, this culminated in economic crisis and the 1976 state-sanctioned (and controlled) labor unions and introduced coup, with a sharp reduction in labor share. This pattern a series of policy measures aimed at addressing their needs exemplifies a broader international pattern of crises often (including large increases in worker wages and benefits and "resolving" underlying distributional struggles, with losses the creation of a social security system). There was a large being borne by formal labor (see chapter 8 and Diwan 2001). increase in labor share under Per6n that was then sharply If Argentina represents a case of extreme regime insta- reversed when he was out of power (see figure 5.4, which bility masking the underlying persistence of unequal social illustrates a long time series of labor shares in Argentina). relations and a weak state, Venezuela is a case in which formal The government's relationships with organized labor was democracy has failed to resolve distributive conflicts highly dependent on Per6n's personal relations and charis- throughout the 20t century. Venezuela has therefore not matic appeal rather than institutionalized mechanisms of been evolved into a redistributive state capable of fostering representation. In the words of Snow and Wynia (1990, inclusive development. Although the country maintained a p. 138), "The material benefits obtained by the workers healthy economic performance and democratic stability during the Per6n administration were essentially gifts from during the 1960s and 1970s, once its oil-dependent FIGURE 5.4 The rise and fall of the labor share in Argenbna Labor incomes as a share of total 55.0 - 50.0- 45.0- 40.0- 35.0- 30.0- -4- Ministry of Economy Conade-Cepal 25.0 - -A- BCRA, without employers contributions to the pension system (De Pablo 1979) -0- BCRA, with employers contributions to the pension system (De Pablo 1979) -0- Beccaria (1991), without employers contributions to the pension system 1935 1939 1943 1947 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1990 Source: Fiszbein and Galiani 2003b. 132 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQllALITY economy ran into trouble in the 1980s, successive govern- As in the case of Argentina, the country's inability to resolve ments failed to adjust economic policy and, in turn, faced distributive conflict in a peaceful and sustainable manner is an escalating crisis in governance. vividly illustrated by the current stand-off between a pop- For a long time, oil revenues have provided Venezuela ulist president and ardent opponents. with a relatively easy source of fiscal revenues with which Democratization constituted an important shift in the to finance the expansion of public services, but have also political opportunity structure, but it was not a sufficient con- sharply limited the incentive of political elites to invest in dition for change. This was precisely because the process of strengthening the state's tax capacity. When the role of oil democratization interacted with the institutional structures revenues role as a social lubricant weakened in the 1980s and of unequal social relations and weak states and, in particular, 1990s, the original elite group that established 20th century the weak agency of subordinate and middle-class groups rel- democracy increasingly came to serve to protect the privi- ative to the influence of elites. The result was weak effective leges of new elites-those associated with established demand for change on the part of states. In addition, unequal parties-and to ossify the country's democratic institutions. structures tended to persist under democratic auspices. BOX 5.3 Are there political and social reasons for the contast between Latn Amrica and Ead Asia? East Asia is often characterized as a region ofgrowth with on that group for its primary social and political base equity par rxc/llence. Reality is more complex, and some (under the rule of Miahathir)and also recognized the need East Asian countries have experienced significant increases to protect the properry rights ot other domestic and for- in inequality in the past two decades (Ahuja and others eign investors. Unequal social relations therefore had less 199"). However. it is clear thar a number of countries in salience in East Asia. the region-from Japan to ( most recently) Vietnam-have Second, successful cases of simultaneous growth and experienced long phases of high growth with moderate equalitv evolved a form of economic management that levels of inequality. Mloreover, this pattern has occurred in combined a degree of bureaucraric autonomy with close many countries when they had authoritarian governments. linkages to malor interest groups. Such "embedded auEon- Based on an extensive literature on the underpinnings omy" 4 Evans 1995 was a central element of the success- of East Asian success, two strands are of particular rele- ful development of states that emerged in the region. This vance to the current discussion. First, successes almost was by no means a clean affair, and often coexisted with always involve broad alliances between elites and subor- cronyism and corruption However, for major development dinare groups, especially peasants during the early phases choices-from export drives in Korea to dealing with of transirion This pattern reflects many facrors, includ- macroeconomic crises in Tndonesia prior to the deep eco- ing long traditions of smallholder agriculture; land reforms nomic and political crisis of 1998-embedded autonomy during periods of major political transition (rhar is. in iacilitated rhe pursuit of pro-development paths rather China, Republic ot Korea, and Taiwan, China). the need than protection of vested inrerests. States were effective in to deliver incomes and services to peasants in the face ot key areas of providing public goods and services. external threats ro national security (Korea and Taiwan, The key concrast with Latin America should not be seen China); and rhrears of peasant-based revolutionary in terms of a spectrum ot authoritarian and democratic movements (Indonesia). systems, but in terms of how East Asia's political and These alliances did not necessarily involve grassroors social histories-which evolved under different external mobilization, but it was in the interests of elites ro eco- environments-underpinned more inclusive and effective nomically include poorer groups. In Mlalaysia, this action insticutional structures. took the particular form of improving the posirion of the historically poorer ethnic group of //wmiputera3 ( Malays), but Soutes: World Bank 1993, Teranishi 1997, Evans 1995, Heller 2001. was effected by a governmenr that both directly depended 133 INEQUALITY IN LATIN AMERICA Clientelism can function under democracy, as well as comparison between Spain and the United Kingdom in the under authoritarianism (see Shefter 1994 and Chubb 1983 1980s, recent work (Boix 1998) shows that the different pref- for OECD countries and Hagopian 1996 for the persistence erences of political parties with respect to inequality were of clientelist politics in Brazil). Group-based social inequal- acted upon when governments were elected, especially in the ities also show great persistence across formal regime domain of structural policies (see box 5.4). changes. A comparison between Latin America and East Asia Programmatic parties, conceived of as "organized coali- is useful. Although there has been plenty of cronyism and tions of interests around ideas" (Boix 1998, p. 14) are less corruption in East Asia, other factors in the social and polit- well-developed in Latin America, but the principle that ical structures of most countries in that region have led to coalitions of interests and ideas can lead to different policy the construction of states that have been both more effec- choices that influence distribution is directly relevant. Under rive and redistributive during both authoritarian and demo- what conditions are there effective coalitions for change? cratic periods. How do these coalitions interact with the different initial institutional conditions that prevail across Latin America? 5.2. The potential for equalizing political and Even though the region (with the exception of Cuba) has not social change seen major change in its distributional patterns based on fun- The first section of this chapter sought to explain how insti- damental political and social transformations, some experi- tutional forces cause inequalities to persist. With this in ences are already promising. The section below focuses on mind, how can societies break with the history of unequal change both at the level of countries and at the sub-national social relations and weak states? Success will depend upon level. developing institutions that are resilient to the predations of the rich and powerful and can tackle vested interests. For The potential for more effective and redistributive most countries in the region, success also involves strength- states at the national level ening the implicit social contract that underpins an adequate tax effort with which to finance government spending. Patronage vs. programmatic parties in a The authors argue here that there is potential for change. democratizing context Some features of the regional context are favorable. Although Despite being an insufficient force to foster a more effective democratization alone has not so far led to a swift transfor- and redistributive state, democratization did help bring into mation of social relations and political structures, it remains the political arena generalized demands for state action in an important potential catalyst for change in the opportunity the social sphere throughout Latin America in the 1990s. structure. While the authors do not seek to address inter- These demands were undoubtedly heightened in the wake national influences in this chapter, it is worth noting that of the period of economic crisis and worsening income with the passing of the Cold War and the globalization of inequality during the 1980s. Most governments did respond information flows, awareness of human rights, and connec- to these problems in one way or another. Social spending tivity of social movements, the pressure and potential for grew significantly in most countries, with part of the growth inclusive political and social change may be higher than it (except for large increases in pension outlays) being pro- was in the past. (See Castells 1997 for a discussion of the gressive in impact because of the expansion of basic services relationships between identity formation, globalization, and such as education, health, water, and sanitation to parts of information.) the population that had previously not been served (see In addressing these issues, the authors take the view that chapter 9). However, there were large variations in the political agency matters. History and institutions are also effectiveness of such actions in terms of outcomes, and an important, but are not determinist sources of action. This expansion in public spending as such did not reflect any is based on a reading of country and local experiences in Latin change in the underlying political equilibria. America that are discussed in the following sections. This Both in Latin America and throughout the history of position echoes debates on whether governments in OECD Europe and the United States, changes to political equilib- countries can affect conditions of inequality. Drawing on an ria that underpin more effective and redistributive states analysis of experiences across the OECD and a structured have occurred both from within and outside of existing state 134 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHAN(CE IN INEQUIALITY BOX S.4 Pdical agmncy and the pobntial for redsibutlve s btegy hI rih counbies lsons fom the OECD A comparison of the different paths taken by governments housing polic and encouragement of broad-share owner- in the OECD finds that insrirucional dererminism does not ship through privatization That parry successfully reduced hold true, at least in certain domains of public polic) the power of vested interests, notably in the union move- Although both social democrartic and conservative parties ment. In contrast, the base of the Spanish Socialists was support growth, very different strategies have been pur- initially among the working and lower-middle classes. sued ro this end. This is particularly the case with regard Although there was some erosion of support from the ro distributional choices related to the trade-off between former group due to its need to pursue a tight macroeco- employment and equaliry on the one hand and efficiency nomic policy, this was compensated by a significant expan- on the other. During the 1970s and 1980s, social demo- sion of support due topolicies that expanded social benefits. cratic governments chose strategies based on significantly The point of the comparison here is not to evaluate the higher public capital f'ormation, human investment, and impact of policy choices; this would require both a care- tax efforts, and maintained a larger public business role ful assessment of instrumental effects of policies and nor- in the economy. In contrast, conservative governments mative views on inequalit) and unemployment. It is, chose lower investment in physical and human capital, however, noteworthy that Britain experienced avery large lower taxes. and privatization. increase in inequality in the 1980s, by the standards of rich The contrast in policies among types of governments can countries. Spain experienced a small decline in its (already be clearly seen in a comparison berween rwo programmatic lower) level of inequaliry. Moreover, some observers argue parties in the 1980s: the British Conservatives under that the substantial expansion in social welfare was a key Margaret Thatcher and the Spanish Socialists under complement to the extensive industrial restructuring asso- Felipe Gonzalez These parties were based initially on very ciated with Spain's opening up and integration into different electoral coalitions The British Conservatives Europe. appealed to the middle and upper classes and sought to exprand their base in rhe former, for example through Sornce: Boix 1998. structures. The key question is whether new or reformed par- traditions and fragmented societies, and is a case illustrated ties or groups, when they attain power, adopt the potent tools by Peru. The general weakness of programmatic parties in of patronage in order to shore up support or pursue collec- the region-a product of historical patterns of political evo- tive goals.1" In the following discussion, different cases of lution-makes the challenge of improving the state harder, transition within Latin America are considered, with boxes and in turn makes the role of other actors in society (includ- used to illustrate areas of policy action. ing business associations, unions, grassroots movements, and Chile is a good example of how the pre-existence of pro- nongovernmental organizations) even more important. grammatic parties combined with relative bureaucratic Chile has been relatively successful in consolidating an autonomy to facilitate a transition toward political equilib- effective and redistributive state, including a significant rium with a more effective and redistributive state. Where degree of redistribution of services; however, the country has these pre-conditions are weak, change has occurred either not been able to reduce its high level of income inequality. because of the internal mobilization of new parties or "out- Unusually for Latin America, Chile benefited from a long sider" movements. Brazil illustrates both in the form of the history of party formation along class-based rather than ver- Cardoso and Lula administrations. Latin America also has a tical corporatist lines, as well as a tradition of a relatively variant of the outsider movement, that is, of a populist leader autonomous and competent public administration. The first who appeals strongly and directly to a (disaffected) populace, democratic administration pursued a policy of "growth with but is weakly embedded in organizational structures or local equity."12 This approach was implemented by balancing the mobilization. This is itself a product of the weakness of party fruits of market-oriented economic policies (introduced by 135 INEQUALITY IN LATIN AMERICA BOX % i Increased equity dUsmo twmax and social spending In a democratic Chile The Aylwin administration's (1990-1994) equity- Conrreras, and Milian 2002). Mtany vrulnerable segments of enhancing policies included an increase in taxes to finance society, such as the elderly and the poorl) educated. experi- expanded social spending. The tax reform was moderately enced large gains in well-being, and the benefits of economic progressive (ar least in intention), raising the share of gov- growth were distributed more or less equally across regions. emnment revenues generated through direct taxes from 18 The reform package. which was the first piece of maior percent to 24 percent and increasing the share of corpo- legislation by the Coinertacidn coalition. was politically rate and upper-income personal taxes within the direct tax negotiated in order to assuage rhe conservative economic category. Tax retorm was accompanied by an arrangement elites of the government commitment to fiscal responsi- to earmark a specific share of revenues for social spending. bl iry and moderate economic policies without radical Between 1990 and 1994, social spending increased by redistributive efforts This negotiated settlement reflected almost 50 percent. This had little impact on the high level nor only the relative political strengths of the conserva- of measured income inequality during the same period, tive elites and the Aylwin government (which lacked a which was largely driven by the effect of market forces on majority in Congress), but also what the parties compris- labor incomes Howe%er, there were significant reductions ing Con;e-eacz6n had learned from experiences in the 19'0s. in other dimensions of inequality, reflecting the redistribu- when radicalism eventually culminated in a brutal coup tivecharacterofsocialspending: if the imputed -aluecofsocia] and nearly rwo decades of authoritarianism under Pinocher. benefits is included, the distribution of "income" improved significantly in the 1990s (see chapter 9 and Bravo, Sormne Buoiand 19%6.We%1and 1)9'. the former military government) with the objective of have resulted in unsustainable economic populism. Why was achieving social equity and justice by systematically Chile in the 1990s able to avoid a similar fate, and instead strengthening social policies (see box 5.5). succeed in introducing a progressive policy reform agenda As a result, key public goods were delivered, inflation was without jeopardizing business confidence and macroeco- brought under control, and fiscal prudence was maintained. nomic stability? There were no major challenges to property rights, and a sig- One plausible explanation is found in the encompassing nificant expansion of citizenship rights (relative to the nature of key political institutional actors and the relations authoritarian era) was complemented by a strong increase among them. The institutional backbone of Chile's "growth in broad-based, targeted services. With macroeconomic and with equity" approach was the broad center-left alliance of social stability apparently assured, an investment boom the Concertacidn, which comprised 17 political parties around occurred that in turn sustained rapid growth and poverty the core of the Socialists and the Christian Democrats. reduction. Democratic administrations also invested in mod- Deprived of access to state patronage during the struggle ernizing the public administration, building on already solid against the military dictatorship in the 1980s, these parties (by regional standards) administrative capacities. Today, evolved throughout the decade into moderate, pragmatic Chile is known for its effectiveness and the probity of its political entities with strong ties to social organizations public administration, as well as for its achievements in the (particularly labor unions) and a solid commitment to plu- realm of economic and social policies."3 ralist, representative social democracy. The Chilean mix of prudent macroeconomic policies with As the negotiated democratic transition allowed conserv- gradual social sector reform might appear to be a sensible ative forces to retain significant institutional advantages, the yet predictable outcome of a good government. However, Concertacio'n deliberately chose a consultative form of gover- in so many other countries (including several in Latin nance-a so-called "politics of agreements"-whereby the America), similarly strong social desires for redistribution government discussed every important piece of legislation not 136 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY only with coalition partners, but also with opposition parties Brazil's political institutions have traditionally fostered and social organizations.14 Such negotiated settlements were political deadlock.'7 The institutional fragmentation of the possible because both the political parties and key societal Brazilian political system-in which a large number of inter- organizations involved (in particular labor and business con- est groups and individual politicians, unconstrained by cohe- federations) had broad, encompassing bases of representation sive party discipline, engage in the exchange of short-term and support.'" In the face of strong demand for social equity political favors-is not conducive to redistributive change. and expectations for prudent economic management, actors Instead, powerful and well-organized groups successfully managed to achieve sensible compromises on a broad national lobby to press for or protect their own interests and privi- agenda instead of pursuing parochial, sectoral interests.16 leges, which in turn largely reflect the underlying inequal- If Chile has, by Latin American standards, formed a rel- ity of political power in Brazilian society. Such weak atively effective (and redistributive) state, why does it remain governance is clearly illustrated by the dynamics of the one of the most unequal societies in the region in terms of process of forming a constitution in the late 1980s, as well income? (See chapter 2.) Three factors are relevant. First, as by difficulties that the country has since faced in reform- Chile's historical formation (especially during the 19th and ing its regressive social security system (see box 5.6). first half of the 20th century) has always involved a mixture Since the mid-1990s, the Cardoso and the Lula admin- of relatively high degrees of constitutional rule and of con- istrations have consistently pursued a policy mix of a strong centration of wealth, which is in part associated with the commitment to macroeconomic stability, social develop- importance of mining in the economy. Second, most ment, and building state capacities for revenue collection and observers agree that there were significant increases in service delivery.'8 Actions by both governments have no inequality in the 1970s (with the transition to the Pinochet doubt been constrained by the institutional fragmentation regime) and the 1980s, which is evident both in the income of the Brazilian political system.'9 Nonetheless, the achieve- data (see chapter 2 for a brief discussion and references) and ments made and the consistency with which these govern- also in the consolidation of the position of major conglom- ments have pursued more equitable policies are notable.20 erates (Schamis 2002). Third, shifting income distributions Pro-equity policies have been pursued by these two is always a slower process than, for example, increasing access reformist governments, each of which has different partisan to services, due to the centrality of market forces in deter- origins. Cardoso's party, the PSDB, was created as an inter- mining wage distributions. nally mobilized group by defectors of the PMDB-which As discussed in a recent World Bank report (de Ferranti was itself the outgrowth of the "opposition" party created and others 2003), international integration was a catalyst for by the military and traditional politicians during the author- skill-biased technical change in countries such as Chile, ini- itarian period-as an attempt to overcome the PMDB's clien- tially leading to a sharp increase in skill premiums and wage telist orientation. Lula's PT, on the other hand, was an inequality. However, the large expansion in schooling (espe- outsider movement during the authoritarian period (and thus cially at the tertiary level) appeared by the end of the 1990s lacked access to state patronage for mobilization) and devel- to be offsetting such powerful demand-side forces. oped as a genuine horizontal alliance between unions and Like Chile, Brazil went through a negotiated transition other mainly lower-class groups, including poorer households from an authoritarian military government. Throughout the dependent on the informal sector.2' dictatorship and even during the democratic transition, the Despite these differences, both parties found opportuni- continued importance of unequal social relations, the lack ties to govern the country in a gradually maturing democ- of programmatic parties, and the strength of special inter- ratic environment. These developments suggest that est groups characterized Brazil's political economy. Although multiple possibilities for positive political agency exist even Brazil did inherit a relatively effective state in terms of in a difficult governance environment, which may also indi- bureaucratic capacities in certain state functions, such as cate that Brazil is in a time of transition to a political equi- industrial promotion (Evans 1979, Schneider 1991), the librium that is more favorable for the reduction of effective provision of key public goods-notably macro sta- inequality. bility, property rights, and citizenship-was not part of the Peru under Fujimori illustrates how an outsider leader can legacy of the country's military rulers. be effective in breaking through established institutional 137 INEQUALITY IN LATIN AMERICA BOX se. Faibe mdbbtutlv sleor i a ragmed d.meea socil ...t refnor hi Brad Following a democratic transition in 1985, Brazil spent virtually all government employees-regardless of whether nearly rwo years debating and drafting a new constitution. or nor they had been recruted through a competitive qual- The National Constitutional Assembly (ANC) operated ification process or had ever made contributions to the pen- within a highly fragmented political system, in which lose sion system-would be "reclassified" as tenured civil coalitions of weakly organized parties and numerous spe- servants eligible for generous pension benefits. cial interest groups engaged in targeted polirical bargain- In the mid-1990s, rhe Cardoso administration ing.9 As a result, rhe 1988 Brazilian Constitution is an attempted a major overhaul of the social security system. incoherent and complex document that simply pulls Although progress was made, some of the most far-reach- together a very broad range of special inrerests expressed ing reform proposals, especially those affecring the through the ANC.' Although the Constitution includes a entrenched inierests of public sector unions, were strong social component-including the universal nature defeated.d The recently elected Lula adminisrration is now of basic services such as education and health-it also embarking on another major reform attempt to gradually cemenrs a number of special privileges of powerful inter- bring the very generous public sector pension regime more est groups, such as granting public servants excessively gen- in line with the private sector regime. Although the reform erous and fiscally unsustainable pension benefits.' has nor gained final approval, the govemment has already The incoherence of the Brazilian Constitution has neces- had to make several important concessions vis-a-vis its sitated endless attempts by subsequent governments to original proposal, again reflecting the power of corporate amend the document in order ro introduce major reforms, interests in the areas of public service, the military, and most notably (but not exclusivelv) reforms of the social the judiciary. security sysrem. Already in the constiturional debate, pro- posals by rechnocrats in the Ministry of Social Security a The Nationoi ConstItutionalssermbly ANC) was the same national con- aimed at making the social security system distribution- giess that had been elected in 1985, and thuis tendedi to reflect the parochial ally fairer-for example calling for universal social secu- and short-term political interests of legislators. as %ell as their vulnerabil- itV to spec al intercst group pressure rity coverage, improvements in the pension benefits of poor b By one count. at least 383 separate groups representing dmierse occupa- rural workers, the diversification of social securiry taxes, tional and functiona] interests pressed fior inclusion of their agendas In the and an age-based threshold for receipt of the time-of- Consrirution-vividly llustralingl he collecive ACtion poblem caused by the fragmented. vertical, corporatist structure of start -socetry relations in service pension-were defeated by various interesr groups. Brazil (Vianna 19981 Business interests opposed the diversification of social c. see Reich 1998 for an analysis ol the connsitutiionmaking process security taxes, while urban unions and pensioners chal- d. This case is analyzed In decail in Nfelo 1998 lenged rhe proposed restrictions on time-of-service bene- Sorcti Weyland 1996b. Mianna 1998. Meio 1998 firs. Public sector unions achieved a rnajor concession when structures in a time of crisis. A decade after the democratic ending the conflict with Sendero and supporting a major transition occurred, Fujimori inherited almost the exact expansion in basic services, in particular to poor, rural opposite attributes of an effective and redistributive state: populations. hyperinflation; a violent conflict with the Sendero Luminoso How was this done? Instead of reforming the state to (Shining Path) guerrilla movement; uncertain property upgrade its overall capacity (as both Chile and Brazil have rights; weakened citizenship rights; and a collapse of the attempted to do), "islands" of effective state action were cre- state's capacity to raise revenues and deliver services, espe- ated, notably in customs and taxes (in order to provide cially in poorer areas. In the early 1990s, Fujimori sub- revenues to support spending), in macro-management, and stanrially improved the provision of public goods-especially in the concentration of a set of service delivery programs in with regard to macroeconomic and social stability-by the presidential administration. These actions involved 138 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY BOX 5.7 NeopopUIISm ad poalkes on sochil uds in Pem The case of Peru highlights both the possibility (at least marginal voters. After a disappoinring result in a national in the short run) and the difficulry of sustaining equality- referendum held in 1993. the Fujimori administration enhancing policies through outsider, neopopulisr" redirected resources in an attempt to buy back the vote leaders.a Tnder President Fujimori, traditional sources ot in turncoat provinces. Since Fujimori targeted poorer authority and intermediation between citizens and the constituents, FONCODES protects tended to Favor the state were weakened because he relied upon direct, per- poorest provinces, with the result that the program had sonal appeals to poorer citizens. Traditional political par- a redistributive function. ties all but disappeared, and government became much Although evaluations of the pro-poor impacr ot more centralized FONCODES on poverty have been positive, the program An important example is the centralization of has often been criticized for its politicization and for a FONCODES and other social programs under immedi- design that bypassed local governments. Consistent with ate control of the Presidency. This shift both enhanced Fujimori's other policy and political actions, FONCODES service delivery and helped Fujimori garner electoral did little tocontnbute to the institutional development and support outside party structures. One study finds that empowerment of local governments. FONCODES expenditures increased significanrly in the three months prior to national elections, and that projects i See Roberts 1)95 on the aoncept nf neopopulism and ins applictnion to were directed to provinces in which their political effect the Fujimori Tegime. was likely to be largest. Berween 1991 and 1993, Sosmei 9hadv24.0).WotidBank 2001.: FONCODES expenditures favored core supporters and breaking through existing clientelist institutional structures of an institutional base of representation to link the gov- (for example, through closing parliament) or circumventing erning elites to the poor. In the case of Peru, costs were mas- them (for example, by avoiding traditional bureaucratic sively higher than could have been imagined. The lack of channels). Political support was sustained by initially wide- institutional constraints on the authoritarian and personal- spread approval for the improvements relative to the pre- ized presidential behavior-exacerbated by the well- ceding crisis; the delivery of services to the poor, albeit in a documented abuse of power by Montesinos, Fujimori's close form of a national clientelist project (see box 5.7); and, associate-led to a steady and systematic undermining of the increasingly, connections with the rich. However, this situ- rule of law with respect to property rights and citizenship, ation had high costs in terms of undermining democratic and the eventual loss of legitimacy and collapse of the institutions, including the capture of the judiciary and parts regime.22 There have been important gains in the subsequent of the media. transition to democracy, but the road from unequal social In terms of political structure, the Fujimori experience relations and a weak state is long, and one that is made more has parallels with the model of the populist leader who difficult with a socially fractured population, high expecta- reaches across institutional structures and promises redis- tions, and the absence of programmatic parties. tribution, but who has the resolve and capacity to challenge vested interests and maintain macro-level stability, and uses Other problematic or incomplete transitions authoritarian methods when needed. At the beginning of Other countries in Latin America have had failed or prob- Fujimori's rule, the depth of the economic crisis and a dis- lematic transitions to democracy (albeit with less dramatic credited political elite facilitated change. forward and backward shifts than in Peru). Guatemala, a clas- However, the charismatic outsider leader model is inher- sic case of unequal social relations and a weak state, signed ently risky, and ultimately unsustainable because of the lack Peace Accords in 1996 and achieved some gains in taxes and 139 INEQUALITY IN LATIN AMERICA service provision from a very low base level, including some poverty programs Programa Nacional de Solidaridad (National symbolic provisioning of bilingual education. However, at Solidarity Program, or PRONASOL), a strongly clientelis- the national level there were few signs of major transition tic policy instrument,23 and its successor PROGRESA (cur- from a polity captured by old elites and new drug- and crime- rently called Oportunidades, or "Opportunities"), better related interests. Also symbolic was the proposal for a mul- conceived technically as a policy lever for long-term ticultural constitution (now increasingly common in the poverty reduction (see box 5.8 and chapter 9). region), which was rejected in a referendum. As discussed in box 5.2, the Dominican Republic, despite a relatively good The role of democratic decentralization growth performance, also remains mired in a low-level equi- Can political and administrative decentralization lead to local librium with high levels of patronage and a weak tax effort. political and social transformations? (See de Ferranti and Bolivia is a more complex case. Starting from a situation others 2000 for a general discussion.) The process of decen- of deeply clientelist structures and a weak state, the coun- tralization has in its very nature bred a large variety of try pursued a number of measures to strengthen governance experiences, since new opportunities for greater political and accountability. It restored macroeconomic stability in autonomy have interacted in diverse local political, social, the mid 1980s, and sought wide-ranging reforms in public and economic contexts. Absent proactive national compen- administration, but it is not clear if these significantly trans- satory policies, decentralization would generally increase formed clientelist relations. Finally, Bolivia took innovative inter-jurisdictional inequalities, for example through the measures to support democratic decentralization (see below), greater capacity of richer jurisdictions to raise taxes that which were intended to increase the influence of poorer, could then be applied to the provision of better economic including indigenous, groups. and social services. At the same time, Bolivia has been swept by waves of In some cases, local democratization, symbolized by the anger that coalesced in the formation of groups in a largely election of mayors, has actually changed the opportunity indigenous-based protest movement, which forced the res- structure for local action for aspiring leaders and local move- ignation of President Sanchez de Lozada in late 2003, rem- ments. This has been referred to as a "silent revolution" iniscent of the earlier ouster of President Mauhad in Ecuador. (Campbell 2003). Yet outcomes are contingent upon pre- (However, Ecuador had registered a less clear commitment cisely the types of issues raised in the preceding discussion to policy and structural reforms than had Bolivia.) In both of national conditions and transitions. How embedded are cases, the protest movement fits the mold of a classic out- unequal social relations and how strong or weak is the local sider movement of poor, indigenous groups allied with vested state, in the sense of its capacity to provide (local) public interests, such as teachers unions. The movement's strength goods and services? Decentralization will be disequalizing flowed from a combination of severe adverse shocks to the if it strengthens the hands of local elites, but equalizing if economy (including the emblematic destruction of coca pro- it is associated with local democratic processes that favor duction in Bolivia) and a long history of social and politi- poorer groups and lead to more effective state action.24 cal exclusion and disillusionment with the political class and That good local governance can make a positive differ- capacity of government. ence to state performance is illustrated by the cases of Porto Alegre (Brazil) and Bogota (Colombia), where major Transition at the sectoral level changes in patterns of action on the part of governments and Even when the realm of national-level politics remains inhos- citizen groups have fostered cultures of citizenship, which pitable to equity-enhancing change, transition away from have in turn led to a successful break with the social and cul- clientelistic modes of governance has been possible at the tural bases of inequality that have defied national-level public sectoral level in a number of countries. Well-known exam- intervention. Clearly, local contexts matter with regard to ples in the region include EDUCO, an effective community- the distributive outcomes of decentralization, as illustrated based school program in El Salvador, and the Family Health by the Bolivian experience with decentralization, in which Program in the State of Ceara, Brazil (for details, see Tendler outcomes of the explicitly pro-community decentralization 1997). Political dynamics of transition at the sectoral level process have depended on the political and social contours is well illustrated by the case of Mexico's targeted anti- of newly created municipalities. 140 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY BOX SH Trandsin at th setl vel? Mexico's targb ed anfipovet programs Although much of the funding for Pengrama Nacoinal de poor and to be less vulnerable to use as an instrument of Soldari,ad(National Solidarity Program, or PRONASOL) patronage. PROGRESA illustrates that even in a clientelist was apparently directed toward the poor,a several studies political environment, technical design ot a policy can have found thar political variables heavily influenced the make a difference in terms of both development impacts allocation of rhe program's resources.' Aside from its politi- and rhe scope for polirical manipulation. cized management, PRONASOLs programmatic focus on infrastrucrure investments-the largest area of funding was a With regard to health carte gcseinrnent programs benebted more than basic services, such as water, sewage treatment, and elec- 4 miLlion people in 1985 An additional i 4 million people were aided b) visible and effective for the purpose of gen- PRONASOUs hospitals in areas with large indigenous populations, and tricity-was visi 1 6 miLlion families bentfed from new clinks With regard to education. erating electoral support. PRONASOL was effeccively PRONASOLs EMelak Dz,m program benehied 4 million students In replaced by PROGRESA (currently called Oportunadadej. 1989-199(. 5 million people benefited froir a drinkable water and sewage or Opportunities'). which focused primarily on invesc- program, and another five million benefted from rural electrificatlon pro- eects By 1992. PRONASOL spending as a share ol oserall public spend- ment in human development within more technocratic Igo oli errhdrahd1 4preleuigli4 tog on social sectors had re-ached 12 3-1 percent I Lustig li94li decision-making processes.c Impact evaluations of b One stud) Niolinar Horcasitas and Weldon 1994 -lound thai the tlederal PROGRESA have been more fiavorable than evaluations government spent more in ares where concurrent local and federal elections of PRONASOL.d An analysis of the political determinants scheduled and that PRONASJL Fands were targeted notr to states in which Ehe PR] tradirionuli) dominated, but to marginal stains that were not of PROGRESA allocarions found thar although elecroral neressarily the poorest in Mexico but where the opposition parry Pmrdo Ret- considerations appear to be somewhat related to funding olhaceiar,o DemnrSraa ithe Democratic Revolutionar) Parry, or PRDt was patterns, poverry level has by far been the best predictor strong. Anotherstudyiliaz-Ca)eros,Magaloni.andVeingast02l2li,ound of PROGRESA expenditures (lilenocal 2001). thar the PRI targeted clectoraJly marginal municipalities and withheld finds oiom PRD-dominated municipalities, but sipent little on municipalities What accounrs for this episode of depoliticization of where it had a big margin ofvi,von W'here the PRI won b ad small marein. antipoverry prograrns in Mexico? The shift coincided with it spent the mno the democratization, or at least pluralizarion, of Mlexican c. At the end of 1999. PROGRESA cosered approximately 2 6 million Lamn slies. or about -tU peLtentI ol all rural families and one-ninih of all families politics. The change in this program can be seen in the in Mexicoi At that time, rhe program o,perated in almostfAl 51i00 localities context of President Zedillo's response to the economic in more than 2!000 municipalities and 31 states crisis of the period and general dissatisfaction with the pol- d The Iniemational Food PohLcy Research Insiltute LfFPRII concluded that itics surrounding spending by PRONASOL, which was a PROGRESA had significantl; increased ennoIlment ofboys and girls in sec ondars school, reduced Illness anong benefciaries by 12 percent for chil flagship program under his predecessor President Salinas. dren and 1) percent nin terms of ob-related sick days) for adults. reduced With each report of corruption of social funds under Sali- srunting caused by nutritional dtfciencies. and increased calcrie consump- nas, pressure on Zedillo to act in an "above-board' manner tion and dietary diversirt For a surnmarn if IEPRI resulIs, iee increased. His response took various forms, including a hrrp poverty wcrldbank org librar) iew.596I e By decentralizing expenditures. primarily to PR] go.ernors. and then politicall) driven process of decentralizarion. The case of lening rhose politicians decide who got whar. Zedillc' mra in effect ha%e PROGRESA is one of a redesigned i.elt rally managed pro- been allowing lotr the politi-al use of these hinds but a ihe saime time keep gram, shaped to both have a righter link to the extreme ing himself at arms length from criricism Two cases of a shift to more effective dynamic interactions between local groups, parry, and and redistributive local states government-through which the PT has repeatedly been Deepdemocraticdecentralization Oneofthebest-documented reelected-have allowed the government to learn from its cases of local-level transition from clientelism is that of Porto experience and democratization to take root (see box 5.9). Alegre. This process was facilitated by the introduction of When opportunity was created by national political and local elections and then by the election of the Partido dos governmental decentralization, local capacities were propitious Trabalhadores (Workers Party, or the PT). Subsequent because of the presence of an externally mobilized party and 141 INEQUALITY IN LATIN AMERICA BOX i.) Polar budget In Porte Ahgr: expiat a tamite to a pego" equbdum Mlunicipal budgets have long been the domain of rent- The close and dynamic relarionship between the PT as seeking elites in Brazil. Politicians, bureaucrats, business a formal organization and its social movement partners has interests, and local powerbrokers have typically colluded created a number of synergies. Mtost obviously, rhe PT in controlling the allocation of city resources. The parnic- has provided a crirical brokerage function by creating the ipatorv budgeting (PB) system represents an imporrant formal spaces and political opportunities in which a range change in the traditional configuration of power interests, of social movements and civic organizations have been since as a botrom-up process it has significanrly shifted able to articulate their demands, exchange ideas, and decision-making powers from traditional elites to ordinary propose reforms. However. if the political opening and the citizens. The role of the PT-an ideologically cohesive and opportunity structure for PB was created from above, it programmatically left-of-cenrer parry-as the catalyst of is civil society that provided the critical inf'ormational change has been critical, since ir created the 'opening from resources for the PB and fostered its mobilization. In above" and provided the ongoing support that has made opting for the introduction of PB. the PT piggybacked on PB possible. demands first presented to an earlier administration by the In contrast to the more hierarchical, rigid, and orga- Union of Neighborhood Associations of Porto Alegre nizationally insulated character of most Latin American (UAMIPA), which had called for popular, communiry-based socialist parties, the PT is deeply embedded both organi- 'control ofthe definition of the city budget ' A wide range zationally and ideologically in civil sociery. From rhe time of civic organizations (including neighborhood and of its establishment in 1980, the PT has been a social church-based groups, women's organizations, nongovern- movement party, having been formed as an alliance of pro- mental organizations, and unions) then provided the mobi- gressi'e elements within the church, unions, peasant asso- lizing nerworks, innovative Ideas, and participatory ciations, human rights groups and an array of revolutionary practices that came to define the PB. With every year, organizations. Created as an instrument of struggle neighborhood delegates to the budget council became against an authoritarian regime and associated corporatist increasingly better at linking local demands to city-wide structures, during its first decade of existence the PT, in needs and objectives (Abers 1996, 2000b, Biaocchi, 2001) the words of Rebecca Abers, "constantly sought to main- tain ties to a multitude of grassroors social movements and a Abers 1996, p.;' developed highly decentralized internal srructures." the local traditions of associational activity. It was, however, redistributive element. A rising tax effort is evidence of a the development of the institution of participatory budget- broader willingness to support the project, as has been the ing (PB) that facilitated the split from clientelist structures. reelection of the PT. Although Porto Alegre is a good illus- The design of the PB process-by which delegates are allo- tration of a shift to a new political equilibrium, caution should cated to the budget council in proportion to the level of par- be exercised in drawing specific institutional lessons, espe- ticipation by associations in open forums-has stimulated cially on the process of PB. By Brazilian and regional stan- associational activity. PB has also had the effect of promoting dards, the city enjoys a high degree of social inclusion. PB horizontal coalition building; this contrasts with the past when has been exercised in a little more than 100 municipalities neighborhoods processed demands vertically through politi- in Brazil (out of a total of more than 5,500). Assessing the cal patrons, with the effect of severely compromising their asso- actual determinants and impact of PB will require examin- ciational autonomy and weakening civil society. ing some of the 100 or so other cases of the process in Brazil. State action in Porto Alegre has been successful in key public arenas, especially in service provision, a trend that has Entry of outsider political entrepreneurs and mobilization of middle- both improved conditions across the city and had a strong income group support Bogota provides a second case of a 142 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY dramatic shift in the performance of local government in a per resident increased from 39,000 pesos in 1988 to 86,000 large city. Bogota's political transformation has unfolded pesos in 1999-while at the same time the city government against the backdrop of a national political scene character- implemented a series of administrative modernization mea- ized by power-alternating and power-sharing agreements sures and kept its operational expenses (such as personnel between the two established political parties, based on the costs) more or less constant, thus allowing increased revenues post-violencia accord among the elites. This context helped to finance the expansion of public services (Maldonado 2003). embed, at both the national and local levels, a particularly The experience of Bogota shows how change is possible well-entrenched form of clientelism, through which the tra- when electoral competition attracts talented outsiders to the dition of exchanging favors for political support persisted political arena, who can then use their position to effect in a wide range of areas (for example, jobs and local public changes in political cultures and state effectiveness. Why did works). This pattern has often been stronger in poor areas this occur in Bogota and not in some other cities? The in both urban and rural settings. authors do not have a complete answer to this question, but As in Brazil, opportunity has been created in Bogota by suggest that it is driven by an interaction between the degree the advent of mayoral elections and electoral competition. to which clientelist structures are entrenched with the capac- However, in this case the transforming event appears to have ity of individuals and groups to make use of new alliances been electoral victories by outsider political entrepreneurs to effect change. Both international and Latin American who were not part of the patronage-based party systems, experience suggests that institutionalizing such changes will rather than (as in Porto Alegre) a programmatic political be of great importance, in terms of both building con- party with a strong grassroots base. These entrepreneurs have stituencies for autonomous bureaucracies and societal chosen to pursue changes in governmental practice and state- demands for performance in the provision of collective society interactions that shifted the domain of political com- goods.25 petition from the capacity to deliver patronage to the capacity to deliver public goods (that is, a cleaner city, How local institutional context affects the potential reduced violence) and broad-based services, notably in edu- for shifts cation and transport. Such competition was reflected in the Throughout this chapter, the authors have emphasized the alternation between two mayors, Mockus and Pefialosa. importance of the institutional context within which a par- Mockus, in particular, placed great emphasis on building ticular political equilibrium of unequal social relations and new coalitions with a range of social groups and on changes effective state may evolve into different patterns of gover- in the culture of citizenship. nance. The critical role of local context is visible in quasi- The effects of these institutional changes are evident in experimental settings of nationwide decentralization large improvements in city performance, both in terms of processes, whereby political opportunity structures open up the level of provisioning of services and (although more anec- more or less equally across a given country, and yet each local dotally) in the quality of service delivery. The moderniza- jurisdiction develops its own path of institutional evolution. tion of Bogota's public transport system resulted in a A crucial question is how local political agency can capitalize dramatic decrease in accident deaths from around 1,300 per on the new opening in the local political arena. As the cases year in the early 1990s to less than 800 by 2001. Govern- of Porto Alegre and Bogota demonstrate, the strength of local ment efforts to improve public safety resulted in the reduc- associational activities and the quality of competitive elec- tion of homicides per 1,000 inhabitants from 80 in 1993 to tions are important elements of these local contexts.26 When 31 in 2001. Net enrollment in primary and the secondary these factors interact positively with each other, political schools improved from 85 percent in 1993 to 91 percent in incentives and capacities for building an effective, redis- 2001, with the secondary level showing the most dramatic tributive local government are enhanced. When one or both increase. of these factors is missing, the potential of democratic decen- Similarly positive tendencies were registered with regard tralization is likely to be underexploited. to other indicators of social service delivery, such as public One of the most radical cases of decentralization in the health, water, and sanitation provision. Trends are also, again 1990s is Bolivia's Participacidn Popular (popular participa- emblematically, reflected in a rise in tax effort-tax burden tion), which aimed to strengthen citizens' ability to oversee 143 INEQUALITY IN LATIN AMERICA the actions of municipal governments and hold them effectively. This may be because the committees were co-opted accountable. This process introduced institutional mecha- by the traditional elites or because both municipal govern- nisms of social control, including vigilance committees and ments and the committees became arenas for open conflict. territorial base organizations (known as OTBs) that were Instead of being used as an ultimate means of holding mayors charged with overseeing municipal government performance electorally accountable for their performance, the construc- and the "constructive vote of censure," a formal procedure tive vote of censure was often abused as a tool of partisan con- to remove mayors on account of poor performance. Decen- flict. In a sample of 38 of the 311 municipalities surveyed in tralization in Bolivia brought about significant shifts in the a recent World Bank study (World Bank 2000c), 92 mayors allocations of public investments from a range of sectors- came and went between 1996 and 2000, or 2.4 mayors per including transport, hydrocarbons, energy, education, urban municipality in less than four years. It is hard to imagine that development, and water and sanitation-and the apparent effective governance could emerge in such a context. equalization of their geographic distribution. (For a more The influence of varying local institutional context in the detailed analysis see Faguet 2000.) This evidence indicates same country is also illustrated by a comparison of medium- that decentralization in the aggregate has brought about pos- size cities in Colombia that underwent the same nationwide itive distributive consequences by allowing local demands process of political decentralization (see box 5.10). to drive public resource allocations. These city-by-city cases illustrate the potential for ben- However, a closer look at Bolivia's experience, at least in eficial cycles of change that bring about complementary or the initial phase of popular participation, shows varying expe- joint shifts in different areas of local governance, citizenship, riences across municipalities in terms of improvements in local and public services. Pro-poor, participatory development pat- governance. Oversight and participatory instruments-in the terns have in some cases gone hand-in-hand with pro- form of vigilance committees, the OTBs, and the construc- business and local economic development initiatives, better tive vote of censure-sometimes became attractive targets local education services, reduced corruption, resource mobi- for political parties and interest groups, thereby weakening lization, and changes in the work culture of public sector accountability and transparency. In those localities where tra- labor. Dynamic mayors have also been important in creat- ditional parties remained strong (and civic associational activ- ing positive change, but they work within a system and the ities weak), vigilance committees tended to function less incentive to be a mayor depends on the context. Holding BOX E i'-1 Local contexts and te tra ee khom clentelhm Ib.au vsns Pa_e Colombia went through a major political and functional mired in old structures-at least until the election of a decentralization, but effects varied across municipalities businesswoman in the late 1990s-and appears to have depending on initial conditions. A clientelist political cul- made limited progress in tackling corruption and rure developed into an instirutionalized arr form in Ibague, improving services. In Pasto, in contrast, an apparently which was essentially run as a well-controlled political transformative change took place under rhe mayoral lead- machine. In Pasto, one study on the 19'0s asserts that "the ership of Antonio Navarro (one of the ex-leaders of rhe totality of the municipal budget Ls devoured by the par- recently legalized M-19 movement), who brought about asitic bureaucracy created by the clientelist machinery."' radical changes in ciry governance, promoted openness and The key facilitating changes in the political environment transparency, and reduced corruption. This shift also led at the city level, which altered the opporrunity structure, to new alliances, especially with local business associations included rhe introduction of democratic mayoral elections and cirizen groups in 1988 and ihe subsequent radical decentralization of responsibilities and resources in 1991 However, the results a. V'elAsquez. 1982. as quored in Angell, Lowden and Thorp (2'001 p SI of rhese events varied sharply, with the biggest difference SEmnr Angell, Lowden. and Thorp 2001 seen between lbague and Pasro. lbague remained politically 144 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY that office is more attractive for talented and ambitious Important complements, but not substitutes, for changes individuals in Colombia than, for example, in Chile, where from within or outside the system include measures to political decentralization has advanced much more slowly. increase transparency and public debate, to strengthen public However, when basic political and social institutional struc- administrative capacities, to sustain societal pressure, and to tures are achieved in the "right" way or radically improved, increase the capacity for effective government. These mea- there is potential for shifting to better and more equal sures could take the form of a variety of transparency ini- institutions, with subsequent gains in multiple areas. tiatives that are already underway in various parts of the world, such as citizens' scorecards to independently assess 5.3. Conclusions the performance of government agencies in serving citizens History has bequeathed Latin America with unequal social (for example, in Bangalore, India, and the Philippines); relations and weak states, in degrees that vary across the efforts, promoted by both governments themselves and some region. Clientelist and patronage-driven relationships, nongovernmental organizations, to improve transparency in backed by varied political and social cultures across groups, public budgets (such as Argentina's "Lupa Fiscal"); and the tend to perpetuate inequalities. Weak states-in the sense careful use of information technology to modernize infor- of low capacities to deliver the public goods of macroeco- mation management systems (for example, Brazil's SIAFI, nomic stability, ensure property and citizenship rights, and which the media uses to report on the federal government's deliver basic services-are pernicious contexts for inequal- budget management). ity, as they are particularly vulnerable to the predations of Various countries in the region, such as Bolivia, Brazil, and the rich and organized middle-income groups (such as public Peru, have begun to institutionalize various forms of public sector unions). consultations and participatory governance for government Nonetheless, both international and Latin American expe- planning purposes. Although not all of these initiatives may riences illustrate the potential for change. This is most likely be resounding successes, since their impact inevitably to occur when there are shifts in the opportunity structure depends on local conditions, they represent a promising (including via electoral competition) and strengthening of approach to enhancing citizen voice and accountability. Tra- the capacities of poorer groups to articulate their interests ditional measures to strengthen state bureaucracies, such as and organize, often in alliance with middle-income groups modernization of administrative procedures, training of per- and progressive elites. These processes may take various sonnel, and organizational restructuring, will continue to be forms. In the long run, the (much maligned) political par- necessary. However, the implementation record in this area is ties have a crucial role to play, provided they are formed along mixed at best, and supply-side interventions to strengthen programmatic or policy lines, rather than the more common administrative capacities are unlikely to be effective unless they tradition of vertical, patronage-oriented parties. are appropriately tailored to local conditions and backed by Where programmatic parties are absent, change is often clear political demands emerging from a well-functioning effected by political outsiders in the form of political entre- political accountability arrangement (Burki and Perry 1998, preneurs or movements. This can be an important catalyst World Bank 2000e, World Bank 2003a). for breaks with history. However, such outsider-driven In closing, the question of whether history is fate can be changes are only likely to be sustainable if they become revisited. This report places considerable emphasis on the embedded in formal institutions, especially those that create centrality of history in explaining current conditions. the capacity of lower and middle groups to articulate their Institutions-in the sense of the formal and informal "rules goals and interests and organize for an effective and redis- of the game" of social interaction-can be seen as the crys- tributive state. There have been cases of significant shifts in tallization of historically-formed interactions between power government action within a few years, primarily at the local and wealth. However, as the case discussions above show, the level (as noted in the cases of Bogota and Porto Alegre). interpretation of how change takes place is not institution- However, more often a shift to a new political equilibrium ally deterministic. Rather, the authors see institutional involves a series of advances and difficulties, as appears to change as occurring in response to the political and social have occurred in Bolivia, or a series of steps that come to be agency of different groups, as well as to external forces that seen as a strategic transition, as may be occurring in Brazil. include economic factors. 145 INEQUALITY IN LATIN AMERICA How this change takes place will be a function of the action harnessed by elites to perpetuate their positions, at times dynamic evolution of the capacities of different groups to through the use of violence. make effective demands on the state, and of the pattern of 9. While racial attitudes are the most powerful predictors of social political opportunities, of which the capacity of the state to spending, other categorical inequalities play a role as well. Histori- respond is a key element. The bottom line is: change is pos- cally, ethnicity has divided the U.S. working class and hampered class- sible, even though it does not come about easily or fre- based demand formation, as immigration introduced ethnic divides quently. Under certain conditions under which opportunities in the working class. Workers in the United States have not devel- arise, it is the voluntary actions of political and social actors oped class solidarity to the degree that they have in Europe, U.S. work- operating within specific institutional constraints that deter- ers have frequently identified according to ethnicity rather than class. mine the courses of change. In the early part of the 20,h century, for example, newer immigrant groups (mainly of Catholic background) were not viewed as potential class partners by the "older" Protestant working class. Indeed, many Notes of these newer Italian and Irish immigrants pressed their demands 1. See Paige (1997) and Mahoney (2001) on the Central American through clientelistic ties to local level political machines in many urban countries, Hartlyn (1998) on the Dominican Republic, and Abente centers. These clientelistic networks impeded broader class ties and (1995) on Paraguay. the subsequent articulation of class-based demands. 2. For analyses of these dynamics, see Malloy and Gamarra (1988) 10. See Fernando Henrique Cardoso's classic study on "associated- on Bolivia, Cotler (1978) on Peru, and Martz (1983) and Quintero dependent development" (Cardoso 1973) for an argument linking the (1980) on Ecuador. specific industrialization strategy pursued by Brazil in the mid 20,h 3. See Archer (1990) on Colombia and Levine (1973), Coppedge century, the country's political development, and inequality. (1994), Karl (1995), and Rey (1990) on Republica Bolivariana de 11. Externally mobilized parties, typified by European labor or Venezuela. socialist parties, characteristically sustained the strong programmatic 4. See Valenzuela (1977) and Garret6n (1989) on Chile and traditions that evolved in their formation once they were in office, Gonzalez (1991) on Uruguay. whether at local or national levels. For internally mobilized parties 5. Originally understood as relationships between "patrons" (for or political groupings, behavior depended on history and context and, example, landlords and local strongmen) and their "clients" (for exam- in particular, on whether a tradition of bureaucratic autonomy and ple, peasants) in traditional agrarian societies, clientelism has shown independence had already been established prior to the process of party remarkable resilience and "survived" industrialization and urbaniza- formation. If such a tradition existed, the ruling party's capacity to tion (see Powell 1970). deploy the tools of patronage were sharply diminished and the incen- 6. Recent empirical work that tries to model the effects of clien- tive to sustain support through a programmatic orientation grew telism on policy outcomes under different assumptions all point to the correspondingly stronger. Where bureaucratic autonomy was weak, same conclusion: clientelism results in less than optimal levels of public patronage tended to persist, if with new masters (Shefter 1994). goods production. Although not all public goods would directly impact 12. This policy was based on three pillars: (1) change in governance inequality reduction positively, it is safe to assume that inequality is style to ensure inclusive, participatory processes appropriate to a unlikely to decline in the absence of more public goods production. Rep- renewed democracy; (2) maintenance of economic stability and growth; resentative work includes Keefer (2002) and Robinson and Verdier and (3) commitment to social justice. (2002), as summarized by Ames 2003. 13. Chileconstantlyrankshighamongdevelopingcountriesinvarious 7. By contrast, societal corporatist arrangements that emerged in indicators of good governance and state capacity, such as Transparency several West European countries in the post World War 11 period International's corruption index. See Marcel (1999) fora detailed account served as mechanisms for building national-level class compromises. of Chile's public sector modernization efforts during the 1990s. For representative work in this area, see Schmltter and Streeck (1985) 14. For example, the 1990 labor reform was based on a "national and Katzenstein (1984) and (1985). agreement" between the government, the Central Unitaria de 8. This does not mean Latin American states were weak in all Trabajadores (Central Workers Union), and the Confederaci'n de la dimensions. Indeed, there were many instances of strong and effective Producci6n y el Comercio (Confederation of Production and Trade) and 146 STATE-SOCIETY INTERACTIONS AS SOURCES OF PERSISTENCE AND CHANGE IN INEQUALITY resulted in a 22 percent increase in the real minimum wage between programmatic nor grassroots based, and to negotiate with governors. 1990 and 1992. See Foxley and Sapelli (1999). This ensures engagement in the traditional forms of political 15. The adoption of negotiated approaches was also facilitated by exchanges of votes in congress, with the subsequent release of public the fact that the conservative elites had to some extent been discred- expenditures and other aspects. ited by their association with the Pinochet regime. 20. The Cardoso administration was more successful in these areas 16. Equallyimportantwastheimprovedqualityofthestateinterms of positive action that did not threaten the deep vested interests in of both policymaking and administration. A series of policy and insti- Brazilian society, compared to major structural reforms that implied tutional changes implemented during and since the dictatorship had some kind of redistribution. While it is too early to assess the Lula reduced or even eliminated the scope for patronage politics (for exam- administration's performance, that government has shown willingness ple, the uniform tariff that limits the scope of private sector lobbying; not only to build on the successful achievements of the previous gov- privatization of the social security system that limits politicians' incen- ernment, but also to tackle some of the entrenched privileges through tives to increase pension benefits by increasing public spending; the inde- structural reforms, most notably the reform of the public sector pension pendence of the central bank that prohibits it from financing fiscal regime. deficits; and the strict, executive-centered budget process that limits the 21. For the PT, the union movement was central, but it was not scope of congressional logrolling and financing of pork barrel projects structurally incorporated into the state in the way the main Mexican by inflating the budget). For more details, see Foxley and Sapelli (1999). union confederation was (see Heller and Mahoney 2003). 17. There isa growing literature on the institutional roots of Brazil's 22. Property rights were threatened, for example through targeted governability problems. Some representative work is found in Ames tax raids on political opponents, while citizenship rights were under- (2001), Mainwaring (2000), and Stepan (2000). mined through excess military and police actions in the name of 18. The Cardoso administration launched deep reforms of the public counter terrorism. This latter factor resulted in many accusations of administration that involved constitutional changes (Bresser Pereira human rights abuses as well as the manipulation and extortion of the 1998). Although some achievements were registered, the outcomes fell regime's opponents by Montesinos. short of the fundamental objectives of the reform due to resistance from 23. PRONASOL was inspired and created directly by President the public sector unions, sometimes supported by the Judiciary (Rinne Carlos Salinas de Gortari (1988-94), who concluded in his own doc- 1999 and Melo 1998). During the two Cardoso administrations, the toral thesis that federal spending in rural communities did not pro- federal government's capacity for revenue collection was enhanced sig- duce sufficient popular support for the regime and that the poor nificantly. The Lula administration is now trying to bring to the relationships between government agencies and beneficiaries were to national level the principles of transparency and participation that blame (Salinas 1978). served PT administrations well in various municipalities. In the social 24. Between and within jurisdictional effects can interact in a way arena, the Cardoso administration introduced several policy innova- that adversely affects inequality, particularly if poorer jurisdictions are tions. These include FUNDEF, a formula-based distribution of a por- systematically more unequal and weak in institutional terms. tion of the education budget to either states or municipalities based 25. In this context, it is too soon to interpret the implications of on the number of primary-level students enrolled in schools in their the recent election of a left-leaning candidate in Bogota. It is hope- jurisdiction that has had a positive incentive effect on coverage expan- fully a sign not of a rejection of the institutional shifts facilitated by sion, and Bolsa Escola and other conditional cash transfer programs that previous mayors, but rather of a shift toward a more mature pattern have recently been combined into the larger Bolsa Fanflia program, of electoral competition across programmatic lines that is sustained the Family Health program pioneered in the State of Ceara that uses by both social demands and new bureaucratic traditions. family health agents to promote preventive health at the community 26. Even in a nondecentralizing, patronage-laden context, such as level. The Lula administration has placed a new focus on social efforts, Mexico under Salinas, electoral competitiveness at the local level such as the flagship Zero Hunger and Youth Employment programs, appears to have significant impact on government performance. Thus, and is also attempting to scale up some successful innovations like con- there is evidence that even PRONASOL, the centrally controlled anti- dirional cash transfer programs. poverty program with a well-documented politicized management, 19. For example, both the Cardoso and Lula administrations have produced better outcomes in those municipalities characterized by a had to share cabinet posts with coalition partners that are neither competitive electoral environment (Hiskey 1999). 147 CHAPTER 6 Economic Mechanisms for the Persistence of High Inequality in Latin America T HIS CHAPTER TURNS TO ECONOMIC MECHANISMS AND THE DISTRIBUTION OF HOUSEHOLD PER capita income. This focus complements the discussion of political and social mechanisms that were reviewed in chapter 5; in a sense, the current chapter constitutes a different prism through which the same, interrelated, set of phenomena can be viewed. Despite the various conceptual and data quality caveats highlighted in chapters 1 and 2, it remains the case that household per capita income is probably the variable most commonly used to construct a distribution, against which inequality is then measured. In addition, as indicated in chapter 1, income helps shape personal freedoms and affects political power and patterns of participation, thus feeding back into the broader set of attributes with which this book is concerned. Household income distributions are not simple constructs, however. They are the result of a complex process, in which initial opportunities available to individuals interact with their educational, professional, and personal choices. Both opportunities and choices are in turn shaped by the institutional environment in which people live, including such aspects as family values; the quality of schools; labor market institu- tions (for example, hiring and firing rules, the size of the informal sector, the size and role of unions, and the prevalence of minimum wages); and the nature of state taxes and transfers. For given individual trajec- tories, the household income distribution also depends on the pattern of household formation, from the choice of partner to reproductive decisions. Even though it is essential to acknowledge the complexity particular emphasis on the state taxes and transfers that lead of the processes that lie behind an income distribution, from the primary to the secondary income distribution. progress in understanding these processes can only be made Box 6.1 lays out this logical sequence schematically. It can by tackling such complexity analytically. This chapter takes be seen as a slightly more detailed version of the link between the view that income distributions are determined by the the "assets and opportunities" and "outcomes and incomes" following logical sequence. The first aspect considered is pat- presented in figure 1.1 in chapter 1. terns of individual asset accumulation, which determine the This representation can be thought of as a sequence of distributions of the assets that people later use to generate functionals of distributions, as follows. Think of I (Z, w) as income. The second concern is how, given these assets, people the joint distribution, over the population, of all relevant choose where to work, that is, which sectors of employment, innate characteristics (Z) and inherited family wealth (w). jobs, and types of contracts are chosen. Third, the chapter Then let X denote the set of acquired human capital char- looks at how remuneration rates in these jobs are determined. acteristics, such as health status and educational attainment. Fourth, how individual earnings combine into household P (X, Z, w) may then represent the joint distribution of incomes is examined from the perspective of household for- family wealth and both innate and acquired human capital mation. The fifth issue considered is nonlabor income, with characteristics throughout the population. The process 149 INEQUALITY IN LATIN AMERICA BOX 6 I Schenmic reprusettoion of houueud hbe dietwaulad I(Z, w) Investment in human capital P(X, Z, w) V(J) "Oe,l The marching function D(IE(X, Z, J), X, Z, J, w) Remuneration in the labor market G(w, w) Household formation F(y) Redistribution H(y + t) through which P is generated from I is enormously complex addition, if signaling is at play or discrimination of some in practice, but can be thought of as a human capital for- sort exists, elements of X and Z may affect the wage rate w mation functional. In reasonably abstract terms, this through channels other than productivity. In the simplified depends on Z and w and is mediated by a number of family scheme presented here, all of these processes are subsumed and educational institutions. under the remuneration functional, which leads from the dis- Once we have a distribution P of the relevant character- tribution of match attributes among matches (D) to an indi- istics of potential workers and a distribution V of job attrib- vidual earnings distribution G, which is here written jointly utes (J) across all potential vacancies, then the process by with wealth. which labor market institutions allocate workers to vacan- Once a joint distribution of individual wage rates and cies (or to unemployment) is often called a "matching func- wealth (G) has been determined, the distribution of house- tion" (see, among others, Pissarides 1990). These matches hold incomes is obtained through two processes: (1) the com- are characterized by sets of attributes of workers and jobs, bination of individuals into households, including the as well as by the productivity of the match ir (X, Z, J). outcome of their reproductive decisions, and (2) the return The labor market does not only match workers and on nonhuman wealth, designated as nonlabor income. These vacancies-important though that task is. It also generates processes are subsumed here under the household formation wage rates. These rates determine how firms remunerate functional. Finally, this primary income distribution is con- workers, which is generally a function of the productivity verted into the secondary income distribution after allow- of the match. However, if the labor market is segmented, ing for the state's redistribution through taxes and transfers. remuneration may also be a function of elements of J. In When pension incomes are mediated through the state 150 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA (except when they are kept in individual accounts), this sort 6.1. Asset distributions: Education and land of income will also be included, given the extent of inter- It is possible that one reason why income inequality personal redistribution that generally exists. remains so high in Latin America is that the ownership This scheme is perforce synthetic. It does not pretend to of assets-which generates incomes-is itself fairly con- do justice to the full complexity of the processes represented. centrated in the first place. As was shown in chapter 4, Having focused on labor incomes, this framework was par- concentration in the ownership of land and other natural ticularly reduced-form with regard to portfolio decisions for resources played a central role in the birth of inequality physical and financial wealth and the remuneration functions in colonial Latin America. Today, for the vast majority of for those assets in capital markets. However, as with more the population in the region, total wealth is held pre- developed models, the essential purpose of analytical tools dominantly in the form of two assets: education and is often to abstract from details on some fronts in order to housing. For residents in rural areas, the distribution of shed light on other aspects of reality. This framework is used agricultural land is also critical. Housing values are diffi- to facilitate the investigation of some of the mechanisms cult to measure and information about their distribution through which income distributions in Latin America are is very hard to obtain. (What little is known about the dis- currently reproduced. tribution of housing assets in Latin America is summarized A comparative approach is taken throughout the chapter, in chapter 7.) Most of the following section focuses on the drawing both on cross-country correlations and on detailed relationship between the distributions of education and microeconometric pair-wise comparisons, in order to shed (rural) land on the one hand, and that of income on the light on the following question: What factors account for other. Latin America's excess income inequality vis-a-vis the rest of It seems natural to start by looking at how inequality the world? For each step in the above conceptual framework measures for education and income are correlated across for income determination (that is, asset accumulation, labor countries. This can be done using the international set of market matching and remuneration, household formation, Gini coefficients for years of schooling compiled by Thomas, and government redistribution), the authors investigate how Wang, and Fan (2002), based on the Barro and Lee (2000) Latin American countries compare with other countries. data on educational indicators. The "income" Gini coeffi- There are two reasons why cross-country regressions were cients are obtained from two sources: table A.3 in the not run in order to explain inequality levels. First, the model statistical appendix of this report for the Latin American used here to determine income inequality (which was dis- countries and the World Development Report 2003 database for cussed in chapter 1) emphasizes circular causality flows all other countries.2 Figure 6.1 plots the sample of 68 coun- between incomes, political power, the distribution of assets, tries for which information on both dimensions is available. and the nature of institutions. These variables are jointly Latin American countries are indicated by their country determined and it would be incorrect to specify a single- acronyms. equation model. Second, cross-country data scarcity would One problem with scatter plots such as figure 6.1 is that, not allow for a meaningful estimation of a single-equation whereas all Gini coefficients in the Latin American database model even if it were appropriate (which it is not).i Instead, are based on income distributions, those in the World the authors present bivariate scatter plots and report the Development Report database are based on both income dis- associated correlation coefficients. These diagrams offer tributions and distributions of consumption expenditure, insight into the position of Latin American nations within depending on the country. Since these are obviously not the set of observations. They are not to be interpreted as strictly comparable, the indicator on which a country's Gini being suggestive of causality. This information is comple- coefficient is based is indicated by denoting income Gini mented by the results of two pair-wise comparisons of countries with full circles and expenditure Gini countries income distributions: Brazil and the United States and Chile with empty circles. Correlation coefficients are also reported and Italy. Naturally, given the enormous diversity within for the sample that includes all countries (p,c) and for the Latin America, and the even greater variations outside the sample that includes only countries with Gini coefficients region, these comparisons are meant to be illustrative rather that refer to income distribution (p,). This latter group is than comprehensive. more comparable to the Latin American countries included 151 INEQUALITY IN LATIN AMERICA FIGuRE 6 1 Income and education inequalires across counties Income Gini 70 p = 0.76* O Consumption Gini pi = 0.40* f 65 - 0 Latin America, income Gini * Other countries, income Gini o o - Income BRa 60 - I income or consumption P BRA * / UA PAR __CHL GUA PANG,*O ECUNI 55 - MEX ARGGo JAM o ° oELS 50 - O TRN ( DOM ° ° 40~~~~~~~~~~~~~~ 45- U R F / @ *~~~~~~~~~~0 00 0 30 ~~~~~~~~~~~~00 0 ~ ~ ~ ~ ~~~~00 30 - 0 o 25 . 20- 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Education Gini Note: *Significant at 5 percent level; **Significant at 10 percent level. Sources: Table A3 for Latin American income Ginis, World Development Indicators Database for income and expenditure Ginis elsewhere, and Thomas and others 2002 for education Ginis. in the analysis.3 This convention is followed in a number of higher than both the figure of 0.27 found by Castell6 and figures throughout this chapter. Domenech (2002), who used their own education-related Another problem with scatter plots is that of interpreta- Gini coefficients, and the income-related coefficients from tion. Figure 6. 1, like all other cross-country scatter plots pre- Deininger and Squire (1996), although this latter figure was sented in this chapter, shows covariance patterns between also significantly positive. income inequality (on the vertical axis) and some other Figure 6.1 also shows that Latin American countries do variable (on the horizontal axis). As indicated above, the not have particularly high levels of educational inequality authors believe that most of these variables are jointly deter- by world standards. Instead, they are concentrated toward mined. These diagrams should therefore not be interpreted the middle range of the horizontal axis, with educational as suggesting a direction ofcausation. Simple regression lines Gini coefficients ranging between 0.29 (Argentina) and (of the income-related Gini coefficient on the x-variable) are 0.60 (Guatemala). According to Castell6 and Domenech drawn exclusively for purposes of illustration. (2002), the average Latin American Gini coefficient for The correlation across countries between educational and education is lower than that of every other developing income inequality is clearly positive and significant. The region, except for the transitional economies. Figure 6.2 Pearson correlation coefficient between the Gini indices is further illustrates this by plotting the Lorenz curves of years 0.76 for the income-only sample and 0.40 for the joint of schooling for two Latin American countries (Chile and income and education sample. These numbers are significant Nicaragua) that are close to opposite ends of the regional at the 1 percent level in both cases. They are also somewhat spectrum of educational inequality and alongside the curve 152 ECONOMIIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMfERICA FIGURE 6.2 Lorenz curves of years of schooling, seectd countries Cumulative proportion of schooling (%) 100 100 -1- India 1990 _ Chile 1990 - Nicaragua 1993 80~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~9. 60- /<633 20- 5.9~~~~~~~~~~~~~~~~~~~4 J 60 - 0 20 40 60 80 100 Cumulative proportion of population (percent) Sources: Thomas and others 2002 for India, chapter 2 for Chile and Nicaragua. for India. Although income inequality in both Chile and as discussed further below. However, before jumping to the Nicaragua is higher than India's expenditure inequality, conclusion that educational disparities are definitely not the inequality in years of schooling is unambiguously higher reason for high income inequality in Latin America, it should in India. be noted that years of schooling is a very imperfect measure Since they do have high income inequality levels, Latin of the value of the human capital stock embodied in a person. American countries tend to also have some of the highest In particular, this indicator does nor convey the quality of levels of income inequality conditional on educational dis- the education achieved during a given period. It is therefore persion in the world. All Latin American countries plotted possible that the ratio of income to education inequality in figure 6.1 lie above the regression line of the joint sample, in Latin America simply reflects the fact that disparities in and most also lie above the regression line of the income- human capital accumulation in this region occur to a greater related sample. In other words, in light of the average extent (that is, compared to other parts of the world) because cross-country relationship, Latin American countries appear of differences in the quality of education among various to have "too much" income inequality, given their levels of schools, rather than because of differences in the number of inequality in years of schooling. years of schooling among individuals. This finding suggests that other factors may play a greater It is difficult to test this hypothesis, however, because the role in accounting for the region's egregious inequality levels, quality of education is very hard to measure in a comparable 153~~~~~~3. INEQUALITY IN LATIN AMERICA manner. It is put forth simply as a caveat against concluding cally understated. On the basis of a comparison between a that education differentials are unimportant in explaining single Latin American country (Brazil) and the OECD Latin America's high income inequality, since true human cap- sample, however, it is possible that similar (or even worse) ital inequality may be understated as a result of inadequately quality differentials exist in other developing countries.4 In capturing quality differentials. Two pieces of evidence sug- particular, there are no data on quality dispersion for coun- gest that this possibility is worth exploring. tries in Africa and South Asia, which tend to lie to the right The first comes from the most recent attempt to compare of those in Latin America in figure 6.1. student achievements internationally through the Organisa- The second piece of evidence on the importance of dif- tion for Economic Co-operation and Development (OECD) ferences in educational quality comes from a study of "edu- Program for International Student Assessment (PISA) (2000). cational production functions" in Brazil, which uses test score This exercise was undertaken for 31 countries, only 4 of which data from the 1999 wave of the national primary education do not belong to the OECD. One of these was Brazil. Since examinations known as the Sistema de Acompanhamento do Mexico belongs to the OECD, there are two Latin American Ensino Basico (SAEB). Using hierarchical linear models countries in this sample, which otherwise includes rich coun- (which permit an identification of the sources of inter- and tries, some transition economies, and the Republic of Korea. intra-school variation in achievement scores), the study found Table 6.1 reports the means, the coefficients of variation, and that 28 percent of the total variation in the sample occurred the 90,h to 10'h percentile ratios for test scores with regard to across schools. Most of this variation was accounted for by the three dimensions for which they are reported in OECD differences in the mean socioeconomic levels of students in 2001: literacy/reading, mathematics, and sciences. these schools, indicating that sorting plays a large role in the The results are striking. In terms of absolute levels, Mexico determination of educational outcomes. However, variables and Brazil rank at the bottom of the table on every scale, with related to the quality of school infrastructure (such as whether the lowest mean scores of all 31 countries. Even the internal classrooms are systematically stuffy or noisy) and to the edu- ranking is consistent, with Mexico always scoring above cational attainment of teachers were also significant. This find- Brazil. In terms of the two measures of dispersion in test scores ing indicates that disparities in the quality of educational (or "quality inequality") used here, results are more mixed services being provided throughout Brazil contributed to dis- for Mexico, where the 90th to 1 0'h percentile ratios are the parities in the ultimate value of those services to students.5 13th, 6h, and 215' highest in the literacy, mathematics, and This brief foray into issues pertaining to the measure- scientific scales, respectively. However, dispersion results are ment of the quality of education serves mainly as a caveat still rather stark for Brazil, for which the 90th to 1 0'h per- to the apparent simplicity of the visual message conveyed centile ratios rank 5th highest in literacy but 15 highest in in figure 6.1. The issue of quality is a reminder of how both mathematics and science. The coefficients of variation imperfectly information on years of schooling captures present a similar picture, as can be easily seen in table 6.1. human capital accumulation for the purposes of making These findings imply, first and foremost, that any com- comparisons or aggregation within countries; this is even parison of educational distributions across countries that more evident with regard to international comparisons. relies on years of schooling as a measure should be treated Analyzing quality also raises the possibility that educational with considerable circumspection. If test scores are any indi- inequality in Latin America might be understated with cation, what students learn in any given year varies consid- respect to countries in other regions. If that is the case, erably across countries. At the same time, looking beyond income inequality conditional on educational inequality country means, this finding also suggests that variations in might not be so high for this region. However, this possi- the quality of education within countries, while present bility is by no means established, given the severity of the everywhere, appear to be more pronounced in some Latin data limitations. The balance of the analysis so far must still American countries than in the OECD. be that-since Latin America has very large income-related It would be tempting to read into the latter finding that differentials but not very high education-related ones- the educational inequality of Latin American countries, mea- educational inequality cannot be the sole source of very high sured along the horizontal axis of figure 6. 1, was systemati- income inequality in the region. 154 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA TABLE 6.1. Variation in student perfonnance In PISA 2000 examinations Reading/literacy scale Mathematical scale Scientific scale Coeff. 90t/10th Coeff. 90th/10th Coeff. 90th/10th Country Mean score of var. percentile Mean score of var. percentile Mean score of var. percentile Australia 528 0.19 1.66 533 0.17 1.55 528 0.18 1.61 Austria 507 0.18 1.62 515 0.18 1.61 519 0.18 1.59 Belgium 507 0.21 1.79 520 0.20 1.76 496 0.22 1.82 Canada 534 0.18 1.59 533 0.16 1.51 529 0.17 1.56 Czech Republic 492 0.20 1.66 498 0.19 1.67 511 0.18 1.62 Denmark 497 0.20 1.68 514 0.17 1.55 481 0.21 1.77 Finland 546 0.16 1.52 536 0.15 1.47 538 0.16 1.52 France 505 0.18 1.62 517 0.17 1.58 500 0.20 1.74 Germany 484 0.23 1.85 490 0.21 1.77 487 0.21 1.77 Greece 474 0.20 1.74 447 0.24 1.93 461 0.21 1.75 Hungary 480 0.20 1.69 488 0.20 1.71 496 0.21 1.74 Iceland 507 0.18 1.62 514 0.17 1.53 496 0.18 1.59 Ireland 527 0.18 1.60 503 0.17 1.54 513 0.18 1.60 Italy 487 0.19 1.63 457 0.20 1.69 478 0.21 1.72 Japan 522 0.16 1.54 557 0.16 1.50 550 0.16 1.53 Korea 525 0.13 1.40 547 0.15 1.48 552 0.15 1.48 Luxembourg 441 0.23 1.81 446 0.21 1.70 443 0.22 1.76 Mexico 422 0.20 1.72 387 0.21 1.77 422 0.18 1.62 New Zealand 529 0.20 1.73 537 0.18 1.63 528 0.19 1.67 Norway 505 0.21 1.73 499 0.18 1.62 500 0.19 1.64 Poland 479 0.21 1.76 470 0.22 1.79 483 0.20 1.70 Portugal 470 0.21 1.76 454 0.20 1.72 459 0.19 1.68 Spain 493 0.17 1.58 476 0.19 1.65 491 0.19 1.67 Sweden 516 0.18 1.61 510 0.18 1.62 512 0.18 1.62 Switzerland 494 0.21 1.75 529 0.19 1.64 496 0.20 1.71 United Kingdom 523 0.19 1.66 529 0.17 1.57 532 0.18 1.64 United States 504 0.21 1.75 493 0.20 1.72 499 0.20 1.71 OECD Total 499 0.20 1.71 498 0.21 1.75 502 0.20 1.71 OECD Average 500 0.20 1.70 500 0.20 1.70 500 0.20 1.70 Brazil 396 0.22 1.76 334 0.29 2.19 375 0.24 1.88 Latvia 458 0.22 1.82 463 0.22 1.81 460 0.21 1.75 Liechetenstein 483 0.20 1.72 514 0.19 1.67 476 0.20 1.67 Russian Federation 462 0.20 1.70 478 0.22 1.79 460 0.22 1.77 Mean 493.82 0.19 1.68 493.52 0.19 1.67 493.12 0.19 1.68 Source: OECD 2001. A similar message arises from a rather different type of 0.505), which corresponds to just over half of the total Gini analysis that compares the microdata on the income distri- gap between the two countries. butions of two countries in much greater detail. One recent Other inequality measures had even more impressive study (Bourguignon, Ferreira, and Leite 2002) compared the declines in this analysis. For example, the Theil index fell household income distributions for Brazil and the United from 0.644 to 0.460, or more than 60 percent of the Brazil- States by simulating what the Brazilian distribution might United States Theil gap. Figure 6.3 illustrates this exercise look like if certain aspects of U.S. economic behavior were in a more disaggregated manner through plots of the dif- "imported" into Brazil. It found that replacing the Brazilian ferences in logarithms between the mean incomes of each conditional distribution of education with that of the United 100"h of the Brazilian and U.S. distributions (normalized States-but changing nothing else-would reduce the to have the same mean) in the solid line, as well as the dif- Brazilian Gini coefficient by 6.4 points (from 0.569 to ference between the counterfactual "Brazil with the U.S. 155 INEQUALITY IN LATIN AMERICA FIGURE 6.3 Difference in mean incomes per hundredth of the mean-nornalized distribution: U.S-Brazil and US.-Brazil (with U.S. conditional distribution of education) Log difference 0.8- - Actual U.S.-Brazil differences - Educational endowment effects 0.6 - 0.4- 0.2 - 9- 0. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -0.2 0 10 20 30 40 50 60 70 80 90 100 Percentiles Source: PNAD/IBGE 1999, CPS/ADS 2000 surveys; authors' calculations. conditional distribution of years of schooling" and the U.S. products in which the region developed an early compara- distribution. Like the comparisons across countries, this more tive advantage (such as sugar and cocoa) were most efficiently disaggregated exercise suggests that educational disparities produced on large slave plantations. This fact, along with account for an important share of Latin America's high large power differences between groups, led to the devel- income-related inequality, but are not the only explanatory opment of societies that were polarized between slaves and factor. slave owners, or between large landowners and indigenous A contrasting message comes from countries in Latin workers or small landholders. Is inequality in land owner- America with greater average educational attainment (and ship still abnormally high in Latin America? Could this less educational inequality). A similar microeconometric factor still be driving income inequality even in today's comparison between Chile and Italy found that "importing" mostly urban societies? the parameters of the Italian conditional distribution of edu- To investigate this possibility, figure 6.4 plots the income cation into Chile accounted for only 2 of the 20 Gini point and expenditure Gini coefficients used in figure 6.1 against difference between the two countries. Chile's Gini stood at land-related Gini coefficients from the Deininger and Olinto 0.5 57, Italy's at 0.357, and Chile with an Italian distribu- (2000) data set. Data are available for both Gini coefficients tion of education moved to 0.537. Factors other than the for 75 countries, all of which are plotted below. (Those in structure of education lie behind the sources of differences Latin America are once again denoted by their country in income inequality between these two countries. abbreviations.) The search for other possible sources for Latin America's Evidence of correlation across countries is a little more excess inequality" therefore continues, in part by consid- mixed with regard to the association between land and ering another asset of great importance for poor people. In income inequality. The simple correlation coefficient for the chapter 4, the historical process that led to high inequality joint sample is 0.22 and is only significantly different from in Latin America was based on the fact that some of the zero at the 10 percent level. However, for the income-only 156 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA FIGURE 6 4 Income and land inequalities across countuies Income Gini 70 0 Consumption Gini 0 Latin America, income Gini * Other countries, income Gini pi = 0.48* 65 - Income - Income or consumption pc =0.22** 0 0 60 - °e BRA ea BOL 55 0 OMEX OECU O PAN 50~~~~~~~~ 55 - a MEX ° HON 0ELS 3 0 ARG 25 * * 0* 0 OVEN~~~~~~~~~~~ 20- I I I I I I I 20 30 40 50 60 70 80 90 100 Land Gini Note: *Significant at 5 percent level; **Significant at 10 percent level. Soulrces: Table A3 for income Ginis in Latin America, World Development Indicators Database for expenditure and income Ginis elsewhere, ' Deininger and Olinto 2000 for land Ginis. sample, it is 0.48 and significant at the 1 percent level. All 6.2. Job match quality in all, there does appear to be a positive association between According to the schematic representation of household land and income inequalities across countries, although it income determination presented in box 6.1, once individu- is weaker than the one that exists between education and als are endowed with a basic allotment of human and other income inequalities. assets, they decide whether or not to participate in the labor In regional terms, however, Latin America's inequality market and are matched up with a job vacancy. Their earn- ranks seem to be closer for land and income. The cluster of ings will depend to a large extent on the characteristics of the region's countries has moved from the upper-middle of this match. It follows that the distribution of earnings across figure 6.1 to the upper-right quadrant in figure 6.4. Latin the population might well depend on the nature of labor America is overrepresented among the highest Gini coeffi- force participation, unemployment, and the formal or infor- cients in the world, with regard to both income and land. mal nature of the labor market. It is still the case, however, that Latin American income It turns out that total labor force participation, as reported inequality conditional on land inequality is higher than the by the International Labour Organization (ILO) for 116 world average, suggesting that-as is the case for educational countries, is essentially uncorrelated with inequality in the disparities-considering land dispersion on its own can joint sample. (The correlation coefficient is -0.04, with a result in an underestimation of income inequality in Latin p-value of 0.66). In the income sample only, however, the America. The search for the culprits of inequality must con- correlation is negative (-0.42) and significant. It is also the tinue beyond the realm of asset accumulation and into the case that this latter result is driven to a rather large extent functioning of the labor market, by relatively low rates of female labor force participation in 157 INEQUALITY IN LATIN AMERICA FiGURE 6 5 Labor force participation and inequaity across countries Gini coefficient 70 0 Consumption Gini 0 Latin America, income Gini 0 Other countries, income Gini p - 0.42 65- - Income - Income or consumption | =-0.04 0 ~~~~~~~~~~0 60 - COL o GUA 0 °OBRA 0 OCHL 0 o ~~~~~0 OBOL 0 C) CHL 00 0 0 PAR0O * ECU O NIC PA 0 OM@ HON SS - PAN O e MEXO * O OEC 0 0 O ARG ELS 0 0 * JAM O 50 - * OM 0 *PER 35 30 0 0~ ~ ~~~~~~~0 as 0 0 ~~~~~~0 0 450- 0 0 | . | | | |0 55 60 65 70 75 0 85 90 95 Labor force participation (percent) Note: *Significant at S percent level; **Significant at 10 percent level. Sources: World Development Indicators Database for income Ginis. chapter 2 for income Ginis and labor force participation rates in Latin America, Key Indicators of the Labour Market database for labor force participation rates elsewhere. Latin America compared to more developed regions, which unemployment rates, once again drawn from the ILO data- report low income Gini coefficients and have higher overall base. The correlation coefficient is insignificant in the pooled labor market participation rates. The scatter plot illustrat- sample and -0.34 and significant at the 5 percent level in ing these patterns is presented in figure 6.5. the income-only sample. The latter result is likely to be Overall, there does not seem to be evidence of a significant spuriously driven by a positive correlation between unem- covariance pattern between labor force participation and ployment and gross domestic product (GDP) per capita, as inequality across countries. An exception to this occurs when well as by the negative correlation between inequality and only income-reporting countries are considered, which GDP per capita in the sample. Once again, all Latin American excludes much of Africa and Asia. Nevertheless, what does countries lie above both regression lines. This time, however, remain true is that Latin American countries lie overwhelm- they are more spread out along the horizontal axis--along ingly above both regression lines. In this data set, the coeffi- which, in this case, the unemployment rate is measured- cient of a Latin American dummy included in a cross-country than was the case in previous graphs, suggesting that Latin regression of income inequality on total labor force participa- American countries have less in common in terms of their rion would be significantly positive, as it would have been in unemployment rates than they did, for instance, with respect similar regressions using education and land Gini coeffficients. to patterns of land or education distributions.6 Similarly, no clear pattern of correlation emerges from The situation is somewhat different when we consider the the joint cross-country distribution of inequality and extent of duality in the labor market, as measured by the unemployment rates. Figure 6.6 presents the plot for total share of informal sector employment in total employment. 158 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN ANIERICA FIGURE 6 6 Unemployment and inequalit acmss countries Gini coefficient 70- pi 7 -0.34* 0 Consumption Gini e Latin America, income Gini * Other countries, income Gini p -0.16 65- - Income - Income or consumption 0 60 - 0 BOL 0 A BRA GUA OCOL GUAECU0 *PAR PASOI 55 - MEXO O HON 0 ELS I JAM ARG e 50 - S o PER * O TRN § t V~~~~~~~EN O 45 - URUO 40 - * 0_ ~~ 40~ 0 35- 30 0 00~~CP00 0 :0 * 30 - 0 ~~~~~~000 125 *0 0 0 6 20 1 0 2 4 6 8 10 12 14 16 18 20 22 Unemployment rate (percent) Note: *Significant at 5 percent level; **Significant at 10 percent level. Sources: World Development Indicators Database for income Ginis, chapter 2 for income Ginis and unemployment rates in Latin America, Key Indicators of the Labour Market database for unemployment rates elsewhere. The scatter plot of this share and the income Gini coefficients A comparison of mean earnings across sectors reinforces is shown in figure 6.7. The correlation coefficient between the claim of heterogeneity (Cunningham and Maloney 2003). the two is 0.35 and significant at the 2 percent level in the Clearly, unpaid workers are at the bottom of the earnings pooled sample.' Latin American countries lie along the distribution.8 Employees in small firms tend also to earn middle ranges of informality, between African countries to lower average wages than do formal sector employees. How- the right and more developed countries to the left, and ever, average earnings for self-employed workers are very remain above the regression line. similar to the wages of formal sector employees, although The positive association between a large informal sector the variance of wages is higher for the former group.9 and income inequality across countries may reflect the fact Many factors help explain the higher inequality among that the informal sector is quite heterogeneous. It includes, the self-employed. First, self-employment is a risky venture, among others, unpaid family workers, voluntary owners of so the self-employed may require higher wages to compen- small family businesses, street vendors who cannot find work sate for the extra insecurity that they absorb by owning their elsewhere, employees in small firms who receive training in own businesses. Second, the selection process for survival in their first jobs, young mothers earning pocket money, and the self-employment sector typically leads to a broader dis- well-educated owners of small firms that are just being estab- tribution of earnings for any given level of human capital, lished. The heterogeneity of the informal sector contributes compared to what would be obtained if workers were all to the difficulty in understanding its nature and may explain salaried with nonstochastic, smoothly increasing wages rel- why it tends to be more unequal than the formal sector. ative to human capital. Since the self-employed have full 159 INEQUALITY IN LATIN AMERICA FIGURE 6 7 Infonuality and inequality across countries Gini coefficient 70- 70 - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~p, = 0.37 65 - pic = 0.35* a 60 0 CHL O BRA O GUA OCOL ECU BO 55 - PAN MEX Q.:.rj PAR a JAM 3CP I 50 - G O PER 45 - 40~~~~~~~~~~~ 40 - ° O 0 ~~~~~00 35- 0 o 0 o~~~~~~~~~~~~~~ 30 - 0 0 8 25 - 0 Slovak Republic 20- 15 - 10 0 Consumption Gini 0 Latin America, income Gini 0 Other countries, income Gini - Income - Income or consumption 0 -I I I I I , I I I I I I I I I I I 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Share of the informal sector in total employment (percent) Note: *Significant at 5 percent level; **Significant at 10 percent level. Also see chapter 6, note 7. Sources: World Development Indicators Database for income Ginis, chapter 2 for informality rates in Latin America, Key Indicators of the Labour Market database for informality rates elsewhere. information about their abilities and do not pay efficiency of this inequality is within groups, while that between wages, their returns best approximate marginal productiv- groups is very small in all countries except Chile.11 ity, unlike employees who are hired based on the few These findings suggest that a greater level of labor market characteristics that employers can observe.10 informality contributes to higher inequality through a com- Third, inequality among the self-employed may simply position effect-that is, greater weight in a sector with capture measurement error, since regular employees report higher within-group inequality-rather than through large wages but the self-employed may not accurately report prof- differences in means between the formal and the informal its. Due to the abstract and difficult task of estimating sector, as a more traditional approach might have predicted. forgone earnings from capital investments, the self-employed It may also be the case that increases in the share of the infor- are likely to overestimate their earnings, thus leading to mal sector during the 1990s might have exerted some unequal earnings between two particular observations upward pressure on overall inequality in countries such as within a sample that actually have the same net earnings. Colombia, Costa Rica, and Mexico (see Cunningham and In Latin America, inequality is greater among the self- Santamaria 2003). employed than among salaried workers. The informal sector Overall, the picture that arises from this analysis of the qual- comprises 30-70 percent of the labor force and is made up ity of job matches and income inequality is murkier than the mostly of self-employed individuals. As shown in table 6.2, picture from asset distributions. The correlation coefficients evidence from six Latin American countries indicates that between education and land inequalities on the one hand and earnings inequality in the self-employment sector is double income inequality on the other are convincingly positive. How- the degree of inequality that exists in the wage sector. Most ever complex the process of joint determination might be, the 160 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA TABLE 6.2 Wage work and self-employment in selected countries, 1995 Argentina Bolivia Chile Colombia Uruguay RB de Venezuela Share of the self-employed in the labor force 26% 56% 29% 33% 26% 37% Theil index, all workers 0.362 0.642 0.735 0.667 0.398 0.340 Theil index, self-employed 0.484 0.819 0.867 0.972 0.499 0.470 Theil index, wage workers 0.295 0.430 0.411 0.433 0.350 0.264 Inequality decomposition Within-group 0.355 0.642 0.639 0.653 0.395 0.340 Between-group 0.007 0.001 0.096 0.013 0.004 0.000 Source: Wodon and Maloney 2000. covariance between asset inequality and income inequality does dard way in which economists view the determination of seem to be borne out, both by the cross-country data and by earnings is through the human capital model, originally the microeconometric comparisons between Brazil and the developed by Gary Becker and Jacob Mincer. If education United States. In contrast, the microeconometric comparison and experience increase a worker's productivity, then earn- between Chile and Italy found only small effects of educational ings ought to vary positively with each of these factors, even assets on the distribution of income. when controlling for the other. A standard Mincer equation On the other hand, no association was identified between is therefore as follows:"2 income inequality and either labor force participation or unem- ployment. Once again, this is consistent with findings from + b (age-s -6)2 + ) a microeconometric comparison of Brazil and the United States 3 ' 1 by Bourguignon, Ferreira, and Leite (2002). Importing the Where s, captures the years of schooling of individual i, occupational structure from the United States onto Brazil and e, measures that person's earnings or wage rate. It follows (which is conditional on observed worker characteristics) had that, for a given distribution of education (the dispersion of very little effect on inequality. A similar result is found when which may be measured by the Gini coefficient, as shown Italy's occupational structure is imported onto Chile. in figure 6.1), earnings inequality should rise with the coef- The microeconometric studies confirm the view that occu- ficient b1 in equation (1) above. In figure 6.8, the familiar pational structure variables play a smaller role in accounting income Gini coefficients are plotted against the Mincer coef- for Latin America's excess inequality than do either educational ficients (b1) for the 33 Latin American and OECD countries, or land endowments. The one exception might be the extent on which Fernandez, Gunerm, and Knowles (2001) esti- of labor market informality due to the prevalence of greater mated equation (1) above on household survey microdata. income disparities within the informal sector, as discussed The result is the highest correlation coefficients reported in above. This tentative conclusion suggests that rigid labor this chapter: 0.70 in the pooled sample and 0.81 in the market institutions, such as high hiring and firing costs, may income-only sample, with p-values of 0.000 in both cases. contribute to more rather than less inequality (see Heckman In this (unfortunately small) sample of countries, it seems and Pages-Serra 2000). Other labor market institutions also that returns on education in the labor market are closely asso- play a role in the context of the determination of labor market ciated with income inequality. Latin American countries lie remuneration or wage formation, as discussed below. in the upper quadrant of the diagram, with high estimated Mincer coefficients and high income inequality. 6.3. Remuneration in the labor markets The microeconometric comparison of Brazil and the United In addition to its allocative role, through which workers and States provides further support for the conclusion that higher vacancies are matched, the labor market affects income dis- returns to education in Latin America are an important factor tribution directly through the determination of wage rates in accounting for the region's high levels of inequality. Replac- paid to different workers holding different jobs. One stan- ing Brazil's structure of returns with that of the United States 161 INEQUALITY IN LATIN AMERICA FIGURE 6.8 Returns on schooling and inequality across counties Gini cefficient 65 0Cnupinpi =0.81* O Cosumpion ini 0 Latin America, income Gini 0 Other countries noeGn .0 - Income - Income or consumption 60 40~~~~~~~~~~~~C 30~~~~~~~~~~~~~ 25 -. . 40 - 435 0 1 2 3 1 5 1 7 1 0 0 0~~~~~~~~~icr ofiiet(kllpeim 2001 or sill remim, a meauredin Mncerequatons.int(kl peim led to a reduction of 4 points in the Gini coefficient, or about education and its actual distribution in the population one-third of the total gap. Even more revealing is the fact that accounts for an important part of the difference between the jointly replacing Brazil's conditional distribution of years of income distributions of Brazil and the United States. schooling and returns structure with those of the United States An analogous picture for the Chile-Italy comparison is led to a reduction of 7.5 Gini points, from 0.569 to 0.494. shown in figure 6.9(b). Here the actual differences between This change represents a full 60 percent of the difference in the mean-normalized incomes per percentile between Italy Gini coefficients between the two countries, and Chile are represented by the thick, uppermost line. The The disaggregated impact of the simulation on the entire dotted line represents price effects, that is, the difference distribution can be gauged from the information provided between Chile "as is" and Chile with the Italian structure of in figure 6.9(a), in which the thick, solid line shows the labor market returns. just above the dotted line, the thin solid actual differences between incomes per percentile in the line depicts the differences between Chile with the Italian United States and Brazil. The dotted line represents the dif- returns structure and conditional educational distribution. ferences between the Brazilian distribution and the simu- There is very little difference between these lines, indicating lated distribution for Brazil with the U.S. returns and that almost all the difference related to education flows from occupational structures, while the thin solid line in between differences in returns on education. A substantial share of the those two lines represents the differences between the distributional difference between the two countries is Brazilian distribution and the simulated distribution for accounted for. In fact, the Gini coefficient for that counter- Brazil with U.S. returns and occupational and educational factual distribution (0.445) lies approximately midway structures. It is clear that the combination of returns on between the coefficients for Chile (0. 5 57) and Italy (0. 3 57).1 162 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA FIGURE 6 9 Distribution of and returns on education explain some of the differences between income distributions a. United States and Brazil Difference in log incomes 0.8 _ - Actual U.S.-Brazil differences - Combined price and occupational structure effects 0.6- - - Combined price, occupational structure, and educational endowment effects 0.4- 0.2 -- - …,- -.-- ----- -0.4- 0 10 20 30 40 50 60 70 80 90 100 Percentiles b. Italy and Chile Difference in log incomes - Actual Italy-Chile differences \ -~~~~~~~~~~~~Combined price and occupational structure effects O.S - -Combined price, occupational structure, and educational endowment effects 0.5 -0.5 / -1.5 1 10 19 28 37 46 55 64 73 82 91 100 Percentiles Source: PNAD/IBGE 1999, CPS/ADS 2000, CASEN 2000 surveys; Survey of Italian Households' Income and Wealth, 2000; and authors' calculations. 163 INEQUALITY IN LATIN AMERICA The evidence does provide support for two conjectures: seem to arise from an historical pattern of underinvestment (1) Latin America is characterized by higher than average in education combined with an intermediate position in the returns to human capital, particularly education, and (2) this world trading system, which implies that most sectors in the is an important part of the reason for the region's "excess region involving low-skilled, intensive labor are "price-takers." inequality." Why returns to education are so high in Latin The way in which these differentials between wages and skill America remains something of an open question. To some levels have evolved recently appears to be driven predomi- extent, the obvious but not terribly helpful answer is that nantly by process-related, managerial, and technical innova- the ratio of demand for highly skilled workers to their supply tions, which in many cases have been mediated through is too high, while the ratio of demand for low-skilled work- international trade and foreign direct investment. ers and their supply is too low. This conclusion in turn begs Before concluding this section, it is important to say a word the question of why this is the case. about labor market institutions in Latin America. Even though It seems inevitable that part of the reason lies in the wage rates, like all prices, ultimately depend on demand and changing pattern of comparative advantage for middle- supply conditions, these interact through institutions. In income countries. As Wood (1997, p. 49) puts it: "The eco- addition, since labor markets are-due to heterogeneity and nomic world of the 1960s and 1970s consisted effectively information asymmetries-particularly complex, related only of developed and middle-income countries, and thus institutions are particularly important.16 This chapter has the middle-income countries had a comparative advantage already argued that hiring and firing costs, as well as other in goods of low skill intensity. In the 1980s, when low- regulatory features, may affect workers' decisions about income Asia started to realize its own comparative advan- whether to enter the formal or informal sectors, which in turn tage in goods of low skill intensity, the comparative has implications for the overall distribution of incomes. advantage of middle-income countries shifted to goods of Wage-setting is also affected by other institutions, such intermediate skill intensity." The popular analogy is that the as labor unions or the prevalence of minimum wages. Whereas wages of poor Latin Americans are set in Beijing, but those in most OECD countries, unions succeed in compressing of highly educated Latin Americans are set in New York. wage scales and reducing overall wage inequality in covered It is logical to recognize that this intermediate position sectors (and often throughout the economy, through collec- between countries with an abundance of skilled labor and tive bargaining spillovers), this effect is not robust in Latin those with an abundance of unskilled labor must have some America. In some countries, such as Brazil, unions actually bearing on return structures in Latin America. However, such appear to have the opposite effect: unionized workers appear understanding should not rule out consideration of the to have greater wage disparities than nonunionized workers, following views: and unionization appears to contribute to greater wage dispersion in the economy as a whole. (See Cunningham and * The political economy of high agency inequality Santamarfa 2003 for a brief survey of the evidence.) throughout the history of Latin America has had an Arbache (2002) argues that this trend is largely due to impact by limiting the supply of schooling, which has the fact that Brazilian unions, unlike those in most other in turn resulted in a lower ratio of skilled to unskilled countries, are organized according to professional categories. labor supply than would otherwise have been the case.14 Employers thus bargain separately with various unions that * Most of the recent increases in wage differentials by represent different grades of employees and are concerned skill in Latin America (as elsewhere) appear to be only with their own salaries. In contrast to more integrated driven by skill-biased technical change, rather than by labor movements, there is little pressure to have compressed static trade effects of the type usually associated with wage scales. More generally across Latin America, the lim- the Stolper-Samuelson theorem in Hecksher-Ohlin ited impact of unions appears to largely be a result of both trade theory. The authors' views on that debate do in low levels of union participation and the fact that the fact generally favor this interpretation, as compre- spillover effects from unionized to nonunionized workers hensively set out in de Ferranti and others 2003.'5 (the so-called bargaining coverage rates) are small (see Cunningham and Santamarfa 2003). In fact, the authors agree with both of these views. The higher Minimum wages, on the other hand, do appear to have the than usual rates of return on human capital in Latin America potential to generate equalizing effects on wage distributions 164 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA in Latin America. In Colombia and Brazil, for example, a levels: dispersion in asset distributions (education and land); 1 percent increase in the minimum wage results in increases indicators of labor market matching (participation, unem- along the formal sector wage distribution, with effects great- ployment, and informality rates); and a key indicator of returns est below the minimum wage and diminishing across the to human capital (estimated Mincer coefficients). Moving fur- wage distribution (Maloney and Nuinez 2000, Neri, ther along the schematic determination of household incomes Gonzaga, and Camargo 2001). Although the spillover (or depicted in box 6.1, the authors suggest that the process of "lighthouse") effects of minimum wage rises can be felt quite household formation affects how earnings distributions are far up the wage distribution in these two countries, they are transformed into household income distributions. much larger among low-wage workers. The overall effect of In particular, how men and women sort themselves into this pattern is clearly to reduce inequality. couples matters a great deal. Consider two societies with iden- It should be noted that the scope for the minimum wage tical earnings distributions. Household income inequalities to be used as a policy variable that can reduce inequality is would clearly be different if in one of them the highest- obviously limited by the fact that at some point employment earning woman married the lowest-earning man (and so on), effects may become too great (see Angel-Urdinola 2003). In while in the other the highest-earning woman married the addition, if the minimum wage is used to index public sector highest-earning man (and so on). Income inequality would liabilities (such as pension outlays, in the case of Brazil), the be much higher in the latter society than in the former. More opportunity costs of public funds in terms of other forgone generally, this example simply suggests that when shifting equitable expenditures may outweigh any equity gains in from earnings distributions to distributions of household the wage distribution. income per capita, marital sorting may be an important factor. Figure 6.10 plots Gini coefficients related to both income 6.4. Household formation and marital sorting for all 33 countries (once again only in Latin America's position in the world has now been consid- Latin America and the OECD), which were computed by ered in terms of a number of covariates of income inequality Fernandez, Gunerm, and Knowles (2001). The marital FIGURE 6.10 Marital sorting and inequalit across countries Gini coefficient pi = 0.68* 65 - |0 Consumption Gini 0 Latin America, income Gini * Other countries, income Gini p, 0.63* 60 - Income - Income or consumption PAR 0 BRA OL O ARG - 50 - VEN O PER PAO o CSR cc 50~~~~~~~~~~~~~~~~ 45 - O URU 40 - 35- *0 30- O * O 25- _ S o ** 20 - 15 10 ------- 30 35 40 45 50 55 60 65 70 75 80 Marital sorting Note: *Significant at 5 percent level; **Significant at 10 percent level. Sources: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere, Fernandez and others 2001 for marital sorting correlation coefficients. 165 INEQUALITY IN LATIN AMERICA sorting coefficients are defined as Pearson correlation coef- ages 0-15 to the number of persons ages 16-64) and the ficients for years of schooling between husbands and wives old-age dependency ratio (defined as the ratio of persons among couples within a country. The correlation coefficients ages 65 or over to the number of persons ages 16-64). between marital sorting and income inequality in this sample These ratios are very imperfect measures, since they rep- are high: 0.63 in the pooled sample and 0.68 in the income- resent simple population shares that do not take into account only sample (withp-values of 0.000 in both cases). As before, the correlation between total household size and total most plausible models of household formation and income household income. The latter factor clearly matters for how determination would suggest the existence of considerable dependency ratios affect the distribution of household per simultaneity in this relationship: in more unequal societies, capita income. Be that as it may, the correlation coefficients men may be likelier to marry women from the same social considered here are still significant in both cases: 0.50 in the stratum and with similar education levels. At the same time, pooled sample and 0.84 in the income-only sample for youth if education increases labor market participation and earn- dependency, and -0.56 in the pooled sample and -0.83 in ings, this will likely contribute to the persistence of income the income-only sample for old-age dependency."7 In this inequality in the future. sample of 121 countries, a larger share of youth in a popu- Although whom a person lives with is important, it does lation is associated with higher inequality-primarily not fully determine household composition. Income per capita because of a negative correlation between the number of chil- will also depend on how many children a person has and on dren in the household and household per capita income- the age structure in the household. Figures 6.11 and 6.12 while a larger share of the elderly in a population is associated capture these two dimensions, albeit imperfectly, through two with lower inequality. commonly used demographic variables: the youth depen- The detailed Brazil-United States comparison by dency ratio (defined as the ratio of the number of persons Bourguignon, Ferreira, and Leite (2002) also found that |FIGURE 6.11l j Youth dependency and inequality across counties Gini coefficient 70 - O Consumption Gini 0 Latin America, income Gini * Other countries, income Gini pi = 0.84* 65 - Income - Income or consumption P = 0.50* 60~~~~~~~~~~~~~~~~~~~~ 50~~~~~~~~~~~~~~~~~~~~ 60 * BRA PAN COL O GUA° 50 - * TRN 40~~~~~~~~~~~~~~~~~ 30 2- 0 20 I I I I I 20 30 40 50 60 70 80 90 100 Youth dependency ratio (percent) Note: *Significant at 5 percent level; **Significant at 10 percent level. i~~~~~~~~~~~~~~ S ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ t ource5: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere and dependency ratios. Dependency ratios of youth are defined as 0- to 15S-year-old population as a proportion of total workforce, that is, total population = 16- to 64-year-olds. 30 - --- - -0 -- " 0 ~ ~ ~~~~~~~6 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA FIGURE 6.12 Old-age dependency and inequality acros countnes Gini coefficient 75- - -056* °n0 Consumption Gini 0 Latin America, income Gini * Other countries, income Gini 70 p~~~ = ~~0.56* 0 ~- Income -income or consumption 65 - 00 60 - 0Ba 0RA GU0NC COL CH 55 - R C8PAN 40~~~~~~~ L 50 - 6> 6D V~TR O JAM ARG O 50 - ~ ~ 0 0S o URU 45 - o URU 40 - ° O (C) 0 ° ° 325- * 20 - l l l l l l l l l l l 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Old-age dependency ratio (percent) Note: *Significant at 5 percent level; **Significant at 10 percent level. Sources: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere and dependency ratios. Dependency ratios of elderly are defined as 65+ year-old population as a proportion of total workforce, that is, total population = 16- to 64-year-olds. replacing Brazil's larger family sizes with those prevalent in intervention and inequality, it would be ideal to examine the United States helped reduce inequality, although not by more disaggregated categories of spending. Since interna- very much. When combined with the changes reported in tionally comparable data on the incidence of public programs figure 6.9, importing the parameters from a multinomial logit are scarce, however, it is difficult to disaggregate much. Here model for the number of children in U.S. households into the only one category of public spending is considered, which Brazilian model led to a further decline of 1 point in the Gini due to its commonly universal coverage is not usually regres- coefficient. Figure 6.13 once again plots the differences sive: public expenditure on primary education. Figure 6.15 between incomes in Brazilian and U.S. households, with the plots the income-based Gini coefficients against the ratio of thick solid line referring to actual mean-normalized this expenditure per student to per capita GDP. As one differences. The dotted line refers to the counterfactual dis- would expect, the correlation coefficients are even lower tribution for Brazil with U.S. educational, occupational, and (-0.51 in the pooled sample and -0.67 in the income-only returns structures. The solid line between these two lines sample) and remain significant. A similar result is obtained corresponds to the simulation that incorporates demographic if, instead of considering public expenditure on primary edu- effects. It can be seen that the effect is to further reduce Brazil's cation, the ratio between government transfers to nonprofit inequality and to bring the simulated distribution closer to institutions and households and GDP is used (International that of the United States. In this specific case, it is also Monetary Fund Government Finance Statistics database). apparent that the effect of demographic behavior is not as This correlation is shown in figure 6.16. quantitatively important as are the effects related to both the Figures 6.14-6.16 suggest that states do in fact play an distribution of and returns to education. active role in affecting the distribution of disposable house- Because the incidence of this aggregate spending is likely hold income through their basic taxation and public expen- to matter a great deal for the relationship between public diture choices (see chapter 9). This is not to suggest that the 167 INEQUALITY IN LATIN AMERICA FIGURE 6.13 The role of reproduictive behavior in accounting for differences in income distnb.aton between Brazil and the United States Difference in log incomes 0.8 _____________________________________ - Actual U.S.-Brazil differences - Combined price, occupational structure, and educational endowment effects --Combined price, occupation, education, and fertility effects 0.6 -____________________________________ 0.4- … 0.2 -- 0 -0.2- -0.4 0 10 20 30 40 50 60 70 80 90 100 Percentiles Source: PNAD/IBGE 1999, CPS/ADS 2000 surveys; authors' calculations. FIGURE 6.14 Total public spending and income inequality Gini coefficient 70 ,= - .8 0 Consumption Gini 0 Latin America, income Gini 0 Other countries, income Gini p~ .8 65 - Income - Income or consumption p, 0.32* 0 0 60 -*e BR 0 S PAR O PAN ~ I O ECU CHL 0 I 50~~~~~~~R 30~~~~~~ 425 -0 UR 20 - I 0 30 10 15 20 230 354 5 05 Ratio of government spending to GOP (percent) Note: 'Significant at 5 percent level; "*Significant at 10 percent level. Sources: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere and government spending-to-GDP ratios. 168 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA FIGURE 6.15 Public expenditure on primary education and income inequalit Gini coefficient 70 - |0 Consumption Gini * Latin America, income Gini * Other countries, income Gini p, = -0.67* - Income - Income or consumption |pic=-0.51* 60 - * GUA * BRA 55 - MEX PAR * NIC ° ° 50 - AR JAM 50 TRN~~~~~~~~~~~PER0 45 - 40- 303 ° 0 0 0~ ~ ~ ~~~~~ k 25- * ** o- * o 20 I I I I I 0 4 8 12 16 20 24 28 32 Ratio of per student expenditure on primary education to per capita GDP (percent) Note: *Significant at 5 percent level; **Significant at 10 percent level. Sources: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere and per student expenditure on primary education-to-per capita GDP ratios. i...... J__a _s_ fi_ __________ _______I FIGURE 6.16 5 Public income transfers to households and income inequality Gini coefficient 70 p- 0.6 o Consumption Gini 0 Latin America, income Gini * Other countries, income Gini pi = -0.46* 5 65- - Income - Income or consumption p,,=-0.43' 0 60 ~ GUA C COL OBRABOL O5- 0PAN * PAR OBNOC l O ° @ ARG 50 - cp * PE * TRN| 4 0 - * *°°**° --= _i 35 - 0 0 0 02 4 6 8 10 12 14 16 j ; ~~~~~~~~~~~~~~Public income transfers to GDP ratio } | Note: *Significant at 5 percent level; **Significant at 10 percent level.| Sources: Chapter 2 for income Ginis in Latin America, World Development Indicators Database for income Ginis elsewhere and dependency ratios, and Government Finance Statistics database for transfers (nonprofit institutions and households) to GDP ratios. 169 INEQUALITY IN LATIN AMERICA influence of the state on distributional outcomes is limited The cross-country nature of these comparisons means that to taxation and spending levels. There clearly are a number they are limited to aggregate indicators of public spending. of other important channels through which state institutions The incidence patterns of the aggregate amounts spent are affect the distribution of power and income (such as the not taken into account, even though the impact of the state degree to which decisionmaking processes are democratic and on distribution can clearly vary enormously among countries participatory). Nevertheless, the primary impact of the state with the same expenditure-to-GDP ratio. Distributional as an economic actor is indeed felt through the raising and impacts depend on who receives the benefits of such spend- spending of revenue. In addition, although the direction of ing. Latin America has often been singled out as a region causation is once again impossible to ascertain from the fig- where the richer and more powerful segments of society ures provided here, it is clear that at least some categories of appropriate large shares of the benefits of public programs public expenditure (such as on primary education and trans- for themselves (see, for example, International Development fers) are negatively correlated with income inequality. Bank 1998, United Nations Economic Commission for Latin Interestingly, whereas Latin American countries are rea- America and the Caribbean 2001, and chapter 9 of this sonably spread out along the horizontal axis in figure 6.14- report). This pattern would tend to increase the importance indicating a large variation across the region in ratios of total of the public expenditure component in accounting for the public expenditure to GDP-they are rather concentrated region's high levels of inequality beyond what is suggested toward the left in figures 6.15 and 6.16. This suggests that by figures 6.14-6.16. the region tends to lie toward the low end of the interna- This conclusion is confirmed by what Bourguignon, tional range of public spending per student in primary Ferreira, and Leite (2002) found in terms of the comparison education and transfers, relative to GDP. between Brazil and the United States discussed above. In that FIGURE 6.17 U.S.-Brazil differences, actual and simubled, including nonbabor incomes and reweighting Difference in log incomes 0.8 - - Actual U.S.-Brazil differences -- Combined price, occupation, education, and fertility effects * 0.6 -*----.Full simulation, with nonlabor incomes, without reweighting - Full simulation, with nonlabor incomes and reweighting 0.4 - =_ 0.2 - ; _t - - __---_ _ _0-2 - ,,, \ , $ -0.4 0 1 10 20 30 40 50 60 70 so 90 1i00 Percentiles Source: PNAD/IBGE 1999, CPS/ADS 2000 surveys; authors'calculations. 170 -0_.4 I ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA study, when the impact of importing the parameters of the U.S. parameters with those for nonlabor incomes. This curve U.S. conditional distribution of nonlabor incomes to Brazil comes quite close to the actual differences between mean- is simulated, the Gini coefficient is reduced by more than normalized Brazil and the United States. Most of this effect 3.5 points. When combined with the other effects previously is due to replacing the conditional distribution of retirement discussed (that is, the distribution of education, the struc- pensions in Brazil (which are primarily publicly funded, at ture of returns, and occupational and demographic struc- least in part) with that of the United States. tures), the distribution of nonlabor incomes practically closes Whereas retirement pensions as a share of total income the inequality gap: the simulated Brazilian Gini coefficient decline with household income per capita in most countries (incorporating all the changes) now comes within 1.7 points (including the United States), they make up a rising share of the United States. Other measures of inequality are also of the total in Brazil, as indicated in figure 6.18. comparably near their U.S. "target levels." A similar (although somewhat weaker) equalizing effect The interesting thing with this result is that the bulk of of nonlabor incomes is observed in the comparison between the change is due to pensions, which account for 83 percent Chile and Italy. As counterfactual income distributions for of total reported nonlabor income in Brazil. Figure 6.17 Chile are simulated by importing various elements of the reproduces the main curves shown in figure 6.13, that is, the Italian distribution, the inequality gap between the two actual Brazil-United States differentials and the intermedi- countries progressively narrows. Importing just the condi- ate simulated distribution corresponding to Brazil with tional distribution of nonlabor income (which includes all imported U.S. parameters for educational endowments and public transfers) contributes two Gini points, or one-tenth returns and occupational and demographic structures. of the gap between the two countries. When this simula- Figure 6.17 adds the curve that combines all of the imported tion is combined with all other simulated parameters (that FIGURE 6.18 Retirement income as a share of total household income, Brazil and the United States Retirement income as a share of total household income 0.30- lU.S. Brazil 0.25- 0.20- 0.15- 0.10 0.05 3 0.00 0 10 20 30 40 s0 60 70 80 9010 Percentiles of the distribution of household per capita income Source: PNAD/IBGE 1999. CPS/ADS 2000 surveys; authors' calculations. 171 INEQUALITY IN LATIN AMERICA FIGURE 6 19 Nonlabor incomes account for some of the differences between Italian and Chilean income distributions Difference in log incomes 1.5 - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - Actual Italy-Chile differences i - Combined price, occupation, education, and fertility effects 1 Full simulation, with nonlabor incomes 0.5- .............. o~~~~ - - -- -- -- --------- - 0' -05 4 _5 - I -1.5- 1 1 0 20 30 40 50 60 70 80 90 100 Percentiles Source: CASEN 2000 survey and Survey of Italian Households' Income and Wealth 2000 (Banca d' Italia). is, for the structure of returns, the distribution of education, Gini for Chile was not large, it had a significant effect on occupational behavior, and reproductive choices), it adds the lowest relative incomes in that country. Whereas the pre- almost three Gini points. This complete simulation changes vious stages of the simulation succeeded in making the top the Chilean Gini coefficient from 0.557 to 0.391, which is 80 percent of the Chilean income distribution look more not too far from the Italian "target" of 0.357. similar to that of Italy, they failed to raise the incomes of Figure 6.19, which is analogous to figure 6.18, expands people at the bottom of the distribution. At this stage, the the Chile-Italy comparison represented in figure 6.9(b). The Italian structure of nonlabor incomes had a substantial thick solid line at the top represents, once again, differences positive impact on the bottom 20 percent of the Chilean in the logarithm of incomes accruing to corresponding per- distribution. centiles of the Italian and Chilean distributions after means Since most income from self-employment among the were equalized. The dashed line at the bottom represents the poor is usually reported as labor income, this is unlikely same differences between Italy and a counterfactual distri- to represent the effect of capital incomes. It is more likely bution for Chile after the Italian returns and occupational, to represent the effect of larger and better targeted public educational, and reproductive structures were imported. transfers and benefits. As in the case of the Brazil-United Finally, the dotted line between these two lines represents States comparison, this microeconometric evidence sup- the same differences when the conditional Italian distribu- ports the view-suggested by the pattern of correlations tion of nonlabor incomes (including public transfers) is also across countries-that Latin American states are relatively incorporated into the simulated Chilean distribution. unsuccessful at transferring resources to their poorest It is interesting to note that although the impact of the citizens, and that this is one factor behind the region's Italian conditional distribution of nonlabor incomes on the excessive inequality. 172 ECONOMIC MECHANISMS FOR THE PERSISTENCE OF HIGH INEQUALITY IN LATIN AMERICA 6.5. Conclusions and household income per capita; and (4) the role of high Drawing on a schematic representation of the determinants but badly targeted public spending. In addition to these four of household incomes, this chapter considered Latin key factors, it should also be noted that some of Latin America's position in the world in terms of six broad groups America's excess inequality may be illusionary and due to of factors: (1) the distribution of the assets that underlie the predominance of income rather than expenditure surveys income generation; (2) the structure of occupation and (see chapter 2.) employment across the labor force and the labor market Strikingly, there is apparently no single culprit that can institutions that influence them; (3) the structure of remu- be blamed for Latin America's inequality. The interaction neration in the labor market and, in particular, returns to among all the four factors described above reinforces each education; (4) patterns of household formation and com- one's individual role and generates the final outcome. This position; and (5) the level and incidence of taxes and public is particularly clear in the case of the distributions of years spending. of schooling, which are not extremely unequal by develop- The preceding discussion was framed in terms of the cor- ing country standards. It is only when these distributions relation between indicators pertaining to each of the first five interact with unusually high returns on education (partic- areas and income-based Gini coefficients, and complemented ularly post-secondary education) and with higher than aver- by two specific country comparisons of disaggregated income age correlation coefficients between spouses that they lead distributions: Brazil and the United States and Chile and to high inequalities in earnings, and therefore in income. Italy. These two comparisons are clearly not representative Finally, both the cross-country correlations and the Brazil- of Latin America as a whole, but interestingly the informa- United States comparison also pointed to the role of the tion that they provide resonated well with the patterns regressive incidence of public finance in failing to reduce- emerging from correlations across countries. and indeed in some cases in exacerbating-inequality in the Cross-country diagrams provided illustrations of joint dis- secondary distribution of income. (The magnitude of this tributions, or patterns of correlation. These cannot be inter- phenomenon and the areas in which it is most pronounced preted in causal terms, since in general both variables are are discussed in chapter 9.) The general picture that arises determined jointly through economic and political processes. is therefore consistent with a view of the world in which An additional caveat is that the correlations presented were wealth and educational inequalities are self-perpetuating, all pair-wise, and hence did not control for other attributes. often through decisions taken within political systems. The Multivariate regressions on the cross-section were run, but- pattern of taxation and public expenditure is not exogenously in addition to the usual problems of data comparability and determined in a social and political vacuum. It is also likely interpretation often ascribed to them-this analysis suffered to be influenced by the distributions of education, income, from the fact that two of the variables that were most closely and political power, as discussed in other parts of this report. correlated with income inequality (namely the Mincer and the marital sorting coefficients) were only available for a Notes nonrandom sample of 33 countries. The results of the 1. Since data for Mincer coefficients and marital sorting coefficients regressions were therefore so problematic that they were left are only available in comparable form for 33 countries, and some of unreported. these have no reliable Gini coefficients, the best cross-country regres- The balance of the evidence presented-whether from the sion model would have 19 degrees of freedom. The authors did run international correlations or from the more detailed com- such regressions, and found almost no significant partial correlations. parison between Brazil and the United States-suggests that Given the variety of omitted variable, simultaneity, and attenuation four factors are jointly responsible for Latin America's high biases from which the specification might have suffered, the authors income inequality. These are: (1) a moderately unequal dis- were unable to conclude much from those results, and rherefore pre- tribution of educational endowments; (2) the prevalence of sent the bivariate correlations merely as a descriptive tool. high rates of return on education in the labor market, which 2. In all cases, the Gini coefficient used was for the latest avail- may operate through specific institutions; (3) household for- able year in the respective database. mation patterns with high levels of marital sorting and a 3. This is clearly a relative statement. Income variables collected large, negative correlation between the number of children from different survey instruments across countries are not strictly 173 INEQUALITY IN LATIN AMERICA comparable. However, they are generally held to be more closely com- formal sector workers will receive, the value of "being one's own boss," parable among themselves than in relation to consumption expendi- or the value of low job security. ture variables, given that sources of measurement errors are quite 10. This difference also implies that standard wage regressions will different across the two concepts and agents tend to choose con- have less explanatory power in the self-employment sector, as found by sumption streams that are inter-temporally smoother than income Rees and Shah (1986) and Borjas and Bronars (1989). In fact, Rees and streams. Shah found that in the United Kingdom there is no significant rela- 4. This is the case particularly because quality dispersion in Mexico tionship between self-employed earnings and human capital variables. (the other Latin American country in the sample) was rather average 11. If the level of self-employment in Latin American countries was in relation to the OECD sample. similar to that encountered in most OECD countries (that is, about 5. The study also found that, even when controlling for socio- 10 percent versus an average of more than 30 percent in the countries economic status, black students performed worse than did nonblack included in table 6.2), the within-group component of the inequal- students (see chapter 3). In addition, even after controlling for a ity indices would be much lower because within-group inequality is number of attributes related to family and education, students in pri- lower among salaried workers. As a result, the inequality indices for vate schools significantly outperformed those in the public school Latin America would also be lower. system. See Albernaz, Ferreira, and Franco (2002). 12. This equation can, of course, be augmented to allow for other 6. It should be noted, however, that both this greater dispersion determinants of earnings in segmented markets (such as sector or region in Latin American countries and the low correlation coefficients might of activity) or in cases where there is discrimination (such as that related be the consequence of a large measurement error in the unemployment to race and gender). variable, which in turn could result from the fact that statistical agen- 13. Note that the impact of adding the educational effects to the cies and labor ministries in different countries define and collect data price effects is much larger in the Brazil-United States case than in on unemployment rates in widely disparate ways. the Chile-Italy case. This arises from the fact that the Chilean distri- 7. The pooled sample is the only relevant one in figure 6.7, since burion of years of schooling is much closer to that of Italy than the there is only one country in the income-only sample that is not in Latin Brazilian distribution is to that of the United States. America: the Slovak Republic. 14. See Bourguignon and Verdier (2000) for a model in which high 8. The category of "unpaid" workers may simply reflect a mis- inequality slows down educational expansion through a political reporting of income, particularly if the worker is a relative of the channel. "employer." Most unpaid workers are the spouses of a firm owner. In 15. As one would expect, this is not clear for every country. The Mexico, for example, 10 percent of married women are unpaid work- pattern of wage differentials in Brazil during 1988-1995, for instance, ers, compared to the members of almost no other family group is quite consistent with a standard Stolper-Samuelson interpretation. (Cunningham 2001). With this in mind, unpaid workers may actu- See Gonzaga and others (2002). ally receive their income "in-kind" if the firm owner shares the income 16. See Blau and Kahn (1999) and DiNardo, Fortin, and Lemieux throughout a household. (1996) for discussions on labor market institutions in developed countries. 9. The comparison of these two groups is difficult since it is nearly 17. The p-values were 0.000 in all four cases. impossible to quantify the value of the social security benefits that 174 PART III Policies for Lower Inequality CHAPTER 7 Policies on Assets and Services T H HE THIRD PART OF THIS REPORT TURNS TO THE QUESTION OF WHAT CAN BE DONE IN SPECIFIC policy-related arenas to reduce inequality in Latin America. How can public action lead to greater equity with a minimal cost in terms of efficiency and, in turn, result in faster and more effective poverty reduction? In the next three chapters, the authors combine a discus- sion of practical choices for public action with recognition of the centrality of the broader social and political processes that were discussed in chapter 5. There are multiple, complex areas for action, since inequalities are linked to almost all aspects of economic and social policy. There is no way that a report of this length can do justice to the full spec- trum of issues. The authors therefore seek to cover a broad range of concerns, in the spirit of providing an overview of some of the major policy questions. Although this pursuit is not comprehensive-and it certainly cannot go into the depth that any particular area warrants-the aim here is to draw out key themes and contribute both to debate and more in-depth analysis. Possible areas for public action are organized into three categories. This chapter focuses on influences on the distribution of assets (an issue closely linked to service provision), with emphasis on education; property rights, land, and housing; and the provision of infrastructure. Chapter 8 considers policies that affect how economic distribution of well-being in a society and the dynamics of institutions function, moving from the micro to the macro change. Unequal assets and unequal institutions lower the level. Evidence on the overall effect of market-oriented poli- well-being of poorer groups and are associated with politi- cies and on the particular role of labor markets is considered, cal structures and sociocultural processes that lead to the followed by macro-level policies and shocks. Chapter 9 then perpetuation of inequality. This chapter focuses on how a looks at factors that influence the secondary distribution of more equal distribution of assets can be achieved in Latin income (that is, taxes and transfers), with a particular focus on America in three areas: education; property rights, land, and measures that can help manage risk and, specifically, condi- housing; and infrastructure. Although by no means exhaus- tional cash transfer instruments. An important consideration tive, this chapter provides enough material to highlight some when examining policy options is the assessment oftradeoffs, of the major themes and potential areas for public action. particularly between distributional and efficiency goals. Whenever feasible, this question is discussed in general 7.1. Education terms, although country-specific analysis would be required Education lies at the center of the perpetuation of inequal- to develop a more definitive assessment. ity, both reflecting and influencing unequal economic In part II of this report, it was argued that the interac- conditions, power, and social status. Education also has the tion between the distribution of assets and institutions makes potential to reduce inequalities. This perspective is central the most fundamental difference with regard to both the to thinking in many disciplines, from economists who view 177 INEQUALITY IN LATIN AMERICA more equal education as a means of achieving more equal labor income to social scientists who see new forms of edu- FIGURE 7 1 cation as essential to changing the aspirations, outlook, and The long chain of interactions from unequal schooling to more equal outcomnes behaviors of subordinate groups.' The interactions between education and the various i i ~~~~~~~~unequall dimensions of inequality are complex. The discussion in E UnequaIpoltkal * schooig accss i sKol-utural L j* economic opportunities chapter 6 indicated that a relatively equal distribution of historiles * aprations r education across the population is no guarantee of equal income. By international standards, Latin America has on /lt r average middling levels of education inequality but high epano bas) levels of income inequality. Some societies, such as Chile, *chovi, have relatively extensive and equal education by Latin . or equal American standards (that is, an average of ten years of eI nmic& Expansionein education and a Gini coefficient of 24 for 25- to 65-year- i olds in 2000), but also high income inequality. Even broad-based and high quality education systems, such as I _ those in France, can deeply reflect and perpetuate patterns Compoto tskaill. / Sh tudehior of social difference, as documented in the classic work of dizensNp o aspiratho of studenparens French sociologist Pierre Bourdieu (1984). The embedded and complex relationship between edu- cation and society should inspire caution when confronted with the simple view that more education can solve high benefit from sizeable subsidies. Democratizing tertiary levels of inequality with regard to income, power, and social education, and increasing access to bright but poorer kids, status. However, this is not to say that education does not is important for reasons related to both efficiency and equity. matter. It most certainly does; education has been central The following pages examine the potential for public to every case of a successful and equal development process, action to effect a large and equalizing expansion of in both economic and political terms. This fact applies to education that can also support greater equality in other the U.S. experience in the 1 9th and 20th centuries, the dimensions. Most notable among such aspects is income, for European experience (most notably during Scandinavian which there is much more information on quantitative transitions), the Japanese process of development after the relationships. Meiji restoration, and in almost all the East Asian success Figure 7.1 represents some of the complexities of these stories, from the Republic of Korea to Vietnam. processes. It is useful to draw a distinction between first-stage Moreover, the importance of education appears to be reforms that are primarily about expanding access to basic increasing as global integration intensifies and the transi- education via increased schools and teachers, and second-stage tion to urban, industrial, and service-based societies takes reforms that focus on both equitable access throughout hold. Education is quickly becoming the most important the system and the achievement of high levels of quality. The economic asset for the majority of the population, a factor latter are typically more complex politically and institu- that only reinforces the need to expand access to quality edu- tionally, thus raising questions about uncertainties in the cation. Most countries have made significant progress on chain of changes. Most Latin American countries have made basic education, placing priority on expanding secondary the push toward universal basic education, although some schooling, democratizing college education, and strength- have not completed that process. All are engaged to varying ening quality throughout the system (de Ferranti and others degrees in second-stage reforms. 2003). As noted in chapter 2, many countries have experi- The primary focus here is on service delivery and use as enced increases in the premium placed on a college educa- the major levers that states and social groups have to effect tion. Yet most of Latin America's college systems are still change in educational attainment. The first issue considered primarily the domain of the children of elites, who often is the recent dynamics of educational expansion. Then the 178 POLICIES ON ASSETS AND SERVICES potential impact of a future large-scale expansion on income tion in the region (de Ferranti and others 2003) emphasized inequality is examined through structured comparisons that on average education expanded more slowly in the between Latin America and more developed regions. With region than in other regions of the world, especially East Asia this foundation in mind, the central challenge of what can and the European periphery. be done to improve educational equality is discussed. A picture of past expansions can be obtained by using the data assembled for this report to look at how educational Recent dynamics of educational expansion attainment varies across cohorts.2 The current education of How fast have societies in Latin America pursued an edu- the 51- to 60-year-old cohort reflects educational efforts cational transition? Education dynamics of the recent past launched 40-50 years ago. Low levels of average education suggest that Latin American countries show some signs of in a country, or within a cohort, are typically associated with breaking with a political and social history of education- high inequality of education. (Actual types of inequalities based mechanisms for exclusion. There is a great variety of are discussed below). experiences, and all countries in Latin America have The patterns in Latin America can be roughly classified launched significant educational expansions in the past into four typical past processes, as illustrated for four selected decades. However, last year's World Bank report on educa- countries in figure 7.2. In the first group are countries with FIGURE 7 2 Education dynamics in selected counbies; years of schooling of men and women by age group a. Argentina, 2001 b. Brazil, 2001 R Years of education Years of education 12 -12 - 10 10 8 8 6 6 4 4- 4 -i 2 2 - 0- II I I I 0- I I 21-30 31-40 41-50 51-60 60+ 21-30 31-40 41-50 51-60 60+ Age group Age group c. Guatemala, 2000 d. Ecuador, 1998 Years of education Years of education 12- 12- 10 10 8 -8 6 -6 - 4 4 2 2 A 0 06I 21-30 31-40 41-50 51-60 60+ 21-30 31-40 41-50 51-60 60+ Age group Age group | + Female-Ml|l Source: Authors' calculations from surveys covered in statistical appendix. 179 INEQUALITY IN LATIN AMERICA relatively high levels of educational provisioning in the attained by 10- to 20-year-olds (many of whom are still in middle of the 20,h century, followed by subsequent expan- school), which barely changed (from 3.7 to 3.6 years) sion, especially the Southern Cone countries. This group between 1990 and 2000. ended up with average years of schooling comparable to In contrast, Jamaica has continued to have a remarkably Spain and Malaysia in 2000.3 Second, there is a group that equal educational profile: 21- to 30-year-olds in all income started with a lower level of education, but had a major quintiles had more than nine years of education in both 1990 expansion in recent decades. This includes some countries and 2000, with a top-to-bottom quintile difference of only with a high concentration of indigenous people, including one year. (This is partly because of unusually low tertiary Mexico, Peru, Ecuador, and Bolivia. It is noteworthy that education enrollments in Jamaica at all income levels.) In at least in Bolivia, Mexico, and Peru, the 20th century was terms of incomes, Jamaica is more equal than Brazil but is in part characterized by episodes of major social mobiliza- still unequal by international standards, with a Gini coeffi- tion or revolution with socially inclusive aims, such as strong cient of 49 in 2000. The patterns and sources of income societal pressures to expand education. Third, Brazil stands difference must therefore work in domains other than acqui- out as a country with a dismal historical legacy but a much sition of years of education. more recent acceleration in provisioning. Fourth, there is a set of countries in which educational provisioning has Group-based differences are still large among racial continued to be very low despite recent progress; this is but not between gender groups especially the case in the poor Central American societies of In terms of categorical, or group-based, variations, chapter 3 Guatemala, Honduras, and Nicaragua (as well as, undoubt- showed that there continue to be large differences in edu- edly, Haiti). cational attainment between ethnic and racial categories. In The evolution of educational inequality can be illustrated Brazil, Bolivia, Guatemala, and Peru, the wages of prime- by comparing educational attainment for different age, nonwhite men are on average half those of whites, and categories of the same cohort (see table A.23 ). Doing so the difference in mean per capita incomes is even lower. confirms continued large inequalities, especially in those (See table A.45 in the statistical appendix). Differences in countries where overall progress has been relatively weak education always account for the larger share of the differ- or late. Table 7.1 compares Brazil and Jamaica. There has ences that are revealed in the labor market. As figure 7.3 been important progress in Brazil in the recent past, but shows, progress in reducing black-white differentials in the differences between the bottom and top income quin- schooling has been much slower in Brazil than in South tiles remain very large: 7.1 years in 1990 and 6.9 years in Africa. Years of schooling for South African blacks are lower 2000 for 21- to 30-year-olds. Of equal concern is the fact than for Brazilian whites among older age cohorts, but higher that there is also a large gap in the years of education among younger ones. In contrast, with respect to gender women have made sig- TABLE 7.1 nificant advances relative to men. Among younger cohorts Years of education by income quintile for Brazil and Jamaica, in most countries, women are at an educational advantage, men and women aged 21-30 at least with respect to years of education attained. This also applies to gender differences among blacks in Brazil, but does Quintiles First Second Third Fourth Fifth Average not apply to indigenous groups in Bolivia and Guatemala, where girls continue to have lower attainment and Brazil enrollment. 1990 3.0 4.4 5.7 7.3 10.1 6.6 2000 4.3 5.7 7.2 8.8 11.2 7.6 Spatial differences have remained large during educational Jamaica expansions, at least in some countries. This is markedly so 1990 9.7 9.0 10.0 9.7 10.8 10.0 1999 9.3 9.5 10.0 10.2 10.5 10.0 In the case of northeastern Brazil and the southern states of Mexico. In the latter case, which includes Chiapas, Source: Calculations from microdata; see table A. 23 in Guerrero, and Oaxaca, there is also a strong overlap with statistical appendix. differences associated with indigeneity. 180 POLICIES ON ASSETS AND SERVICES FIGURE 7.3 Years of schooling by cohort for blacks and whites in Brazil and South Afiica Years of schooling 13 - 12- 11 10 9 7 -…- …-- -- - --- - 6------ ...-----.. ---------....... --------........... -----................. ...................... 5 -. 4- 3 - 2 South Africa white --Brazil white 1 - South Africa black - Brazil black] 0 20 25 30 35 40 45 50 Age Source: Lam 2001. Tertiary education is increasingly the most However, in the 1990s differences between the top and salient divide bottom quintiles in the enrollment of 18- to 23-year-olds One of the key correlates of advantage in the past 10-15 widened substantially in all countries except Brazil, El years has been achievement of tertiary education. In 19 out Salvador, and Honduras, all three of which had the highest of the 20 countries for which data exist for the end of the levels of inequality in tertiary enrollments in the region (see 1980s and 1990s (Guatemala is the exception), marginal figure 2.13 in chapter 2). There are high and often rising returns to tertiary education are substantially above the inequalities related to tertiary education, which gives the returns to secondary education (see table A.31).4 Further- greatest gains and (as discussed below) receives the highest more, for 11 of the 17 countries for which time series have subsidies. Moreover, the privileged access to tertiary educa- been produced, marginal returns to tertiary education rose tion is in many countries associated with either private during the 1990s. secondary schooling-which is affordable only to the These differences in tertiary education make some of the better-off--or with passage through higher-quality public trends in expansion of enrollment particularly worrisome schools. from the perspective of future inequality dynamics. As noted in chapter 2, a compression of differences in enrollment at The importance of educational quality the primary school level has occurred between income cat- Educational levels are only part of the story. Of at least equal egories, in turn a consequence of the crowding-in of excluded importance is the quality of education. As discussed in last groups. Experiences have been varied at the secondary level, year's World Bank report (de Ferranti and others 2003) and with neither clearly equalizing nor unequalizing patterns in chapter 6, issues of low quality appear to be pervasive seen across countries. throughout Latin America. Those countries that have 181 INEQUALITY IN LATIN AMERICA participated in international tests have scored substantially 20 percent of differences in test results can be explained by below the Organisation for Economic Cooperation and differences in school quality and that overall, about half of Development (OECD) average or scores for high perform- the variation in student test scores occurs within schools, ers elsewhere (especially some East Asian and Eastern with the other half explained by differences between schools European countries). This is shown in the results for Chilean (see table 7.3). However, with regard to differences between and Colombian students in the Third International Mathe- schools, more than half of the difference is due to measured matics and Science Study (TIMSS), test and for Brazil and gaps in the socioeconomic status of students. Mexico in a standardized test undertaken by the OECD. These results reflect not only a lower average but a wide The possible transformation of education dispersion in Latin America. Only 4.4 percent of Brazilian In the future students and 8.6 percent of Mexicans scored above the OECD What would Latin American countries look like with more average. equal educational conditions? The answer depends on three Differences in educational quality are strongly associated factors: the distribution of both the quantity and quality of with other dimensions of inequality, especially the socio- education; the pattern of remuneration in the labor market; economic status of households. As one illustration of this and the mechanisms that translate such remuneration into link, table 7.2 shows the differences in test results among household income distributions. Indirect effects of expanded Brazilian students by self-identified racial category. This education on fertility and labor force participation would also information shows large deficits for nonwhites, which are play a role. The relative increase in the supply of more edu- higher for blacks than for people with mixed blood. Between cated workers would tend to reduce the wage premia paid one-third and one-half of the deficit is associated with dif- to these workers, but what actually happens is a product of ferences in the socioeconomic status of individuals or with both interactions with demand-related side effects and the the condition of schools. A slightly higher proportion of the influence of formal and informal institutions. deficit is attributable to both socioeconomics and school con- In light of the incomplete understanding of the deter- ditions taken together. Thus, poorer, nonwhite students get minants of wage structures, this report does not seek to both worse results and worse schools, and there remains an model or predict changes. However, as reported in chapter 6, additional effect associated with race. This problem is two paired comparisons were undertaken, between Brazil and typically interpreted as the product of some combination of the United States and Chile and Italy. Brazil and Chile are race-based differences in family or social experiences and both unequal societies even by Latin American standards, expectations or factors in the schooling experience that are yet have very different educational structures; Chileans have not captured in data on schooling conditions (for example, higher and more equally distributed education levels. The the social distance between teachers and students, poorer United States is a rich country with high levels of educa- teaching skills, and cultural aspects of the curricula). tion, and has relatively flexible labor markets by OECD The importance of socioeconomic status in determining standards. Italy also has a highly educated workforce, but is test results is a common finding. An analysis of Brazilian and relatively equal and has more policy and institutional inter- Mexican schools by the OECD (2001) found that only ventions in place in the labor market. TABLE 7.2 Mathematics test results for Brazilian students who identify themselves as black or of mixed blood, compared with white Difference adjusted Difference adjusted Observed Difference adjusted for schooling for both SES difference for SES conditions and school Mixed blood (pardo) -16 -9 -8 -7 Black (preto) -24 -18 -14 -13 Source: SAEB 1999 INEP, 8a serie Matemitica. Note: The standard deviation of math tests is 50; all differences shown are statistically significant. SES = socioeconomic status. 182 POLICIES ON ASSETS AND SERVICES TABLE 7.3 Schools account for only a small part of variance in student learning outcomes Share of total variance in test results for all students (percent) 4. Share of total variation in student test 3. Fraction of performance that is total variation school-specific and not 1. Due to differences attributable to school attributable to student in student 2. Due to student background background differences performance differences and differences across schools within schools across schools across schools (column 2 minus 3) Brazil 55 45 25 20 Russian Federation 63 37 17 20 Czech Republic 48 52 4 18 Korea, Republic of 62 38 14 24 Mexico 46 54 32 22 Developed country average 66 34 20 14 Source: OECD 2001, Annex Bi, table 2.4. The two paired comparisons led to the contrasting results In addition to these comparisons across countries and that were presented in chapter 6.5 In the case of the Brazil- possible long-run changes, an educational expansion has United States comparison, differences in the distribution of potentially significant transitional effects. When there are educational endowments and returns to education explained high returns to tertiary education (and convex returns about two-thirds of the difference in overall income overall), as well as relatively small groups of tertiary grad- inequality. This suggests that benefits of an equalizing edu- uates, there can be a temporary rise in overall inequalities. cation expansion exist for Brazil, which could in turn induce This is because initial expansion of tertiary enrollments will movements toward the more equal returns observed in the tend to increase the share of the highly paid. Only when this United States. However, for the Chile-Italy comparison, the group becomes sufficiently large will overall inequality difference in inequality in educational endowments hardly decline. explained any of the difference in income inequality. Much These simulations should be considered as accounting more was due to the pattern of returns to education and the exercises that provide some quantitative information on what greater equality associated with unobserved factors in Italy. might occur if the educational (and other) characteristics of This would suggest that the sources of the differences in Latin American countries were to shift substantially. Simu- income inequality between the two countries stem from fac- lations are certainly not predictions. However, they suggest tors other than the structure of education. that although equalizing education expansion is undoubt- A related exercise for the poor Brazilian northeastern state edly of fundamental importance, it would be unwise to of Ceara yielded results that were also consistent with a cau- assume that education alone would lead to a transformation tious assessment (Ferreira and Leite 2004 and World Bank, of income inequality. 2003d). That study simulated a large educational expansion and, instead of making a comparison with another country, Policy choices for educational transformation explored the implications of a plausible range of changes in What are the political, policy-related, and institutional the pattern of returns on skills. Even with a strong equal- conditions that would lead to an equalizing expansion of izing shift in the pattern of returns, the overall influence on education, in terms of both quantity and quality? Both policy income inequality was small, in part because expanded basic formulation and implementation matter. Experiences in Latin education induced entry of relatively low-skilled women into America and elsewhere suggest that there is no magic bullet the labor market. In contrast, the effect on poverty was for attaining higher quality and more equitable education substantial. systems. There are many promising experiments in 183 INEQUALITY IN LATIN AMERICA decentralization, demand-side measures, transparency, and in total resource allocations explain only part of the differences testing, but none alone is a panacea. The way in which sys- in educational outcomes. According to international standards, tems operate depends on the context, that is, on the patterns most Latin American countries have already allocated signif- of interactions emphasized in this report, in which economic icant resources to education. The share of gross domestic prod- factors are intertwined with political, social, and cultural uct (GDP) spent on education at all levels in the region fell influences. General work on service provision (World Bank from 3.7 percent to 2.9 percent between 1980 and 1990, but 2003a) has emphasized the varieties of influence and account- then rose to 4.8 percent by 1995. This trend compares with ability that determine the design and execution of policy. 2.6 percent on average for East Asian developing economies These relationships occur both via indirect paths of influence (de Ferranti and others 2003). on governmental decisionmaking and the direct interactions The real issue at hand is the allocation and effective use between communities and households on the one hand of resources. In this regard, there are large differences among and service delivery organizations and frontline workers on countries. Figure 7.4 illustrates three very different patterns the other. of spending across primary, secondary, and tertiary levels. Chile has seen a large, equalizing convergence across these The shape of an expanding education system levels; Mexico has experienced steady growth at all levels, As emphasized by de Ferranti and others (2003), the overall thereby maintaining unequal patterns; and Brazil has a large way in which an education system expands makes a signif- bias toward tertiary education, which receives seven times icant difference with regard to its equity properties. A major more funding than does secondary education. theme developed in that report was the need for extensive Issues also exist that are related to the distribution of expansion from the bottom up, in keeping with the path of resources across inputs. Especially in the wake of macro- East Asian and Scandinavian countries. For most countries economic crises, there has been a tendency in Latin America in Latin America, this approach would imply a particular to cut nonpersonnel-related expenses disproportionately, focus on a large expansion of secondary education. However, resulting in a situation in which a high proportion of spend- this type of expansion alone is not enough. ing is on teachers. In contrast, all "production function" First, of equal importance is the need to address the analyses find that the marginal effect of nonsalary spending potentially adverse effects of the "massification" of educa- is many times higher than spending on teachers. This reflects tion, especially through measures that both ensure high qual- the priority that both governments and teachers' unions place ity and prevent the large-scale opting-out of education by on the relative protection of personnel. members of the middle class, whose political support for Finally, there are questions to be asked about the depth quality education is crucial. This makes the issues around of support for education spending. Are citizens, businesses, accountability and teacher performance fundamental. Second, and policymakers eager to improve education and enthusi- there is a strong need to improve access to tertiary astic about tax-based support, especially in an increasingly education-which is primarily the domain of elites-both globalized world? This is not always clearly the case. At least to ensure that societies make effective use of the pool of for some poorer regions, there appears to be some resistance potential talent in the population and to create incentives to educational expansion on the part of modernizing busi- and opportunities for all groups taking part in secondary- ness elites, who often prefer lower-cost labor and training, level schooling. subsidies, and tax advantages from governments rather than concerted efforts to expand secondary education (see box 7.1). Societal commitment and resources It is particularly interesting that this echoes the 20th century Discourse in Latin America is strongly proeducation, experience of the southern states in the United States, which typically emphasizing the right of everyone to at least a basic also pursued a relatively low-tax, low-education, cheap-labor education. This trend is often linked to the rising demands course as part of efforts to attract private investment. for higher skills as a result of globalization. However, actual This pattern underscores a more general message: the need behavior and conditions result in a more mixed picture. for civic efforts to facilitate high-quality, broad-based edu- Resource allocations are indicative of the level of commit- cation, including access to tertiary education. Economic ment of national or local governments. However, differences incentives do matter, but are insufficient, and may not always 184 POLICIES ON ASSETS AND SERVICES FIGURE 7.4 Spending by education lvel in Chile, Mexico, and Brazil a. Chile: public spending per student, 1980-2000 Real 2000 pesos per student 1800 _ 1600 -+- Basic ed. expends/stud -U- Sec gen expends/stud 1400 | * Sec technical expend/stud 4-O Higher ed expends/stud 1200 . .. 1000 . 800 - 600- 400- 200- 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 b. Mexico: federal spending per student, 1987-1998 Real 1994 pesos per student 12000 4-| Primary -_- Bachillerato -Higher ed. 10000 - 8000 - 6000- 4000- 2000 - 0- I I I + 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 c. Brazil: federal expenditures per student, secondary and higher education, 1980-1995 Real 1998 reais per student 7000 - -- Sec expends/student -_- Tertiary expends/student 6000- 5000 - 4000 - 3000- 2000 1000 . . * . . 9 9 * 9 9 . 9 9 9 9 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 Source: Carnoy 2002. 185 INEQUALITY IN LATIN AMERICA BOX - I Business sites ad ft far of eduat s flie a bw-ro tap? ESeuce from northeas e r.ail and the souhe, Unid Stats In light of the current combination of democratization and public discourse of the business secror on the value of edu- allegedly surging demands for developing skills in a world cation. However, it is consistent with a survey of Brazilian of economic reform and globalization, it would seem that elites by Elisa Reis ( 999), who found that business elites fear of, or resistance to, education would be a part of past were unwilling to support increased spending or taxes on history. In conducting background work for this reporr, education, despite the fact that rhey often named educa- Judith Tendler (2003) explored a specific, surprising tion as an important factor in poverty reduction. dimension of this issue in the context of attitudes among The low-education preferences of business elites are in business elites, who constitute an important, somewhat practice supported by specific policies of northeastern neglected source of demand for education. Brazilian stare goverrnments. In particular, as in other poor Drawing on fieldwork in northeastern Brazil, other sur- regions (notably the southern United Srates), state gov- veys of the business world's atiltudes in Brazil, and the his- ernments have emphasized low wages and tax breaks as torical experience of the southern LJnited SEares, Tendler sources of competitive advantage. This tendency has been found that "modernizing" business elites (including those complemented by an array of specific training and infra- who have relocated from southeastern Brazil) prefer work- structure subsidies that help firms upgrade their own labor ers with only basic education This is generally considered forces but take away resources from the greater collecrive sufficient for workers ro rapidly catch up to productivity need of generalized educational upgrading, especially at levels thar prevail in higher-wage areas, especially when secondary levels. These patterns seem to make up a "low combined with on-the-job training. (Comparable results road" to globalization that keeps northeastern Brazil (and were found in surveys referred to in Tendler 2003 in orher similar poor areas and poor countries) in a low education/ parts of Brazil). Minimum education levels are also seen high inequality equilibrium relarive to the altemarive of as key to maintaining a more docile and less mobile work- broad-based, tax-financed educational upgrading. force with weak aspirations to advance to less grueling ser- vice secror jobs. This preference is in contrast to the typical Soure Tendler 2003 be supportive of educational upgrades (as described in box 7.1 thus also to the distribution of schooling quality. In many with regard to the business world's preference for only basic countries, the teaching profession has been greatly affected education). This is consistent with historical experiences of by the patterns of political and institutional development education expansion. As noted in de Ferranti and others 2003, discussed in chapter 5. Teaching appointments are often part the "Great Transformation" of education in the United States of clientelistic forms of patronage (see Angell and others during 1910-40 was due to a range of economic and soci- 2001 on the case of Colombia). Teachers' unions are usually etal factors, including high rates of return on secondary important forces, whether as elements of historical corpo- education, campaigns by school administrators, and egali- ratist structures (in Mexico, for example) or as sites for polit- tarian features of the schooling system. Societal pressures that ical mobilization (as in Bolivia). This fact provides a context served to change the collective expectations of all participants for reform efforts. mattered, as they did during earlier periods of education Considerable experimentation with education reforms has expansion in the United States (Goldin 1999 and 2001). taken place in Latin America, including decentralization both at the school and local government level (see Di Decentralization, accountability, and teacher Gropello 2003 for a review). However, there has been little performance rigorous evaluation of the impact of decentralization on the The skills and performance of teachers, as well as their inter- quality of service delivery and educational outcomes, and actions with pupils, are key to the quality of schooling, and even less on distributional effects.6 With respect to increased 186 POLICIES ON ASSETS AND SERVICES school autonomy, EDUCO in El Salvador has been frequently winners scored 0.2-0.3 standard deviations higher on stan- cited as a major experiment in greater school autonomy that dardized tests, after controlling for other characteristics of allowed educational expansion for poorer groups without a students and their families (see Angrist and others 2002. loss in quality. Most studies of this and similar reforms find Note that the program was discontinued in 1997.) A a significant impact either on educational achievement or on tentative conclusion is that vouchers targeted toward poor intermediate variables, such as teacher management and children in urban areas-where there is significant school learning, with a positive but weaker link occurring between choice-may play a useful role. However, vouchers do not autonomous decisionmaking in pedagogical processes and appear to be as important an instrument as the conditional learning. Schools that are more active in tracking and cash transfers referred to below, which also involve transfers monitoring teacher activity are likely to be more successful for education but with different design characteristics. in increasing student achievement. By contrast, little evidence exists on the impact of Proactive support for excluded groups decentralization of local or intermediate governments on Although broad-based expansion of education systems educational achievement. Actual performance is likely to be helps to expand coverage to poorer groups, there is also a context-specific. For example, Angell and others (2001) strong case to be made for special action to include those document significant differences in the extent to which local students that have historically suffered educational deficits education reform was effectively pursued across municipal- due to location, socioeconomic class, or culturally based ities in Colombia. Pasto stood out for the vigor of its local inequalities. Ongoing experiments therefore suggest the need reform efforts, which were part of an overall shift to more to work on both the demand and supply sides of the effective local governance and facilitated in part by the elec- education issue. tion of a dynamic outsider mayor (see chapter 5). This is an With respect to the demand side, there is evidence that area in which further experimentation would be highly scholarships conditional on school attendance have positive useful, along with ongoing monitoring and evaluation. effects on deprived groups. These include Oportunidades A more contentious issue is that of vouchers. Advocates (previously known as PROGRESA) in Mexico and Bolsa see vouchers as having the potential to increase school quality Familia (previously Bolsa Escola) in Brazil, among others. by allowing households to exercise choice among different (These are discussed in more detail in chapter 9, since an schools, which would in effect apply competitive pressures important element of the design is redistributive transfers on educational systems to improve performance. It is also targeted toward poor families.) In the case of Oportunidades, argued that vouchers can benefit the underprivileged by geographic targeting has led to the relatively high partici- allowing poor kids to choose better schools. On the other pation of indigenous people (although those in the most hand, critics fear that vouchers will heighten the polariza- remote areas without schools are not reached) and specific tion among schools and between students that do well and grants for girls, who in this group suffer higher dropout rates those who don't, and undercut the morale of teachers more than boys. Such programs have become one of the core than they spur better teaching. International evidence on instruments for providing incentives for excluded groups to these points is ambiguous (Gauri and Vawda 2003). enter school. The effects of vouchers are not clearcut, judging from the On the supply side, bilingual education is of particular results of the two main experiments with the system in Latin relevance to indigenous people. Indigeneity is not coter- America: a large-scale expansion in Chile and a five-year minous with speaking indigenous languages, especially in experiment in Colombia that was targeted toward low- the context of the major migrations of indigenous or mixed income neighborhoods. Some studies of the Chile case find blood groups to urban areas. However, for many of the more modestly higher test scores, but this is not a robust result; rural and poorer indigenous groups, language is an impor- others argue that neither overall quality nor the achievement tant aspect of both their learning experience and cultural gap between subsidized and nonsubsidized elite schools has identity. Ethnographic and other case study evidence doc- changed (see, for example, Hsieh and Urquiola 2003). In uments the role of social distance in schooling, which can Colombia, vouchers were allocated through a lottery, which be amplified by language factors. This is not solely an issue allowed for a quasi-experimental evaluation that found that for traditional indigenous groups; work by Portes and 187 INEQUALITY IN LATIN AMERICA BOX - 2 BlUugal educado Prsmwl cdtrd hltags ae bh* hAtbehdi Research shows rhat for students whose "home" language immediately puts them ar a linguistic disadvanrage that is not the dominanr language in a society, those in buin- cannot be overcome later in the education process. This gual education schools have higher reading comprehension debate is partly serrled by empirical evidence, which shows levels than rhose who learn in only the dominant language. that the bilingual educarion approach provides equal or Bilingual education schools rypically reach skills in the higher reading competencies ro children across the world, child's native language and then transfer these skills to the in such diverse countries as Canada, Guatemala, Haiti, use of the nationally dominant language. Nigeria, the Philippines, and the United States. Prelim- Those in favor of this educational model argue that inary work in Guatemala further suggests that bilingual simultaneously irnmersing a srudent in a new language and educarion is cost-effective in terms of learning outcomes. a new skill does not allow the student to learn either very well, thus accounring for low performance on reading rests. Sonv Parnnosand Vdciz(1996). addirionail informaiInn provided by Harry Conversely, those who are againsr this model claim rhar Parnnos by teaching a child in his or her native language first Rumbaut (2001a) on migrants (especially Latino and Asian) and others (2003) a number of measures will need to be to the United States finds analogous problems, which can taken, including overcoming credit and information con- become more acute in the case of the children of migrants straints for potential applicants, increasing cost recovery for who are torn between two cultures. Bilingual education can public universities (in light of high private returns), and link- play a role both in terms of enhancing education, since ing public resources to performance. A particular issue is initial learning in a mother tongue seems to help students, whether to introduce affirmative action programs based on and managing cultural transitions by reducing social race, ethnicity, or class. This is currently a controversial issue distance and making learning more relevant to children (see in Brazil with respect to Afro-descended groups. The evi- box 7.2). One issue in this context that has not been dence is clear that such groups are both greatly under- addressed in core educational work is the position of represented in tertiary education and suffer a series of Afro-descended children. In this case, the issue is not so disadvantages in schooling and socialization processes. much language as more subtle cultural aspects of social dif- Achieving a more equal society in Brazil would require ference that an education system can accentuate or active greater representation. The goal of affirmative action pro- design can help combat. This is an area for further research grams in that country would be to build an education system and efforts. that allows talented Afro-Brazilians (as well as the indige- nous and poor) to advance through the educational system. Toward more democratic tertiary systems The experience with affirmative action in the United States Tertiary education systems are predominantly the domain provides some lessons, including with regard to the com- of elites, and are typically highly subsidized (though much plexity of the issue. In July 2003, the U.S. Supreme Court more so in Brazil than in, for example, Chile). This trend is ruled on a case that supported the goal of including racial both inefficient and inequitable. The talent pool is reduced diversity in the objectives of recruitment policies, but ruled and the intergenerational transmission of wealth and elite that mechanistic means of doing so (as in across-the-board status is facilitated. Reforming tertiary education is often adjustment in test scores) are nor constitutional. Although contentious, with (understandable) resistance to subsidies for the authors of this report do not have a clear basis for being less advantaged groups. There are, however, experiments for or against affirmative action in college recruitment in the underway that seek to make attending a university more Brazilian case, there is no question that action of some kind affordable to the poor and to make the quality of different is needed to increase the chances that poor and talented universities more transparent. As discussed by de Ferranti Afro-Brazilians will gain access to tertiary education. 188 POLICIES ON ASSETS AND SERVICES Educational reform "against the odds?" property rights were featured in the original "Washington First-phase reforms involving access to and expansion of basic Consensus" as one of the ingredients for growth.7 Links education often command broad support from various stake- between property rights and inequality may also be a factor holders, including politicians, teachers' unions, and parents. in the bivariate association between measures of polarization However, second-phase reforms that are needed to deal with and of the credibility and predictability of property and problems of quality-including for poorer groups-typically contractual rights (figure 7.5). On the one hand, greater involve measures that affect the conditions of work for teach- polarization may be a source of challenges to property rights, ers and increase the power of parents and communities. At if manifested in distributive struggles (as argued by Keefer the tertiary level, reform can involve dealing with the vested and Knack 2000). Such challenges can be in the form of out- interests of universities and students who are part of richer right expropriation or the more subtle form of changes in groups in society. Is making education more equal therefore policies, such as taxes, that affect returns to property. a classic political economy case of reform "against the odds," On the other hand, unequal political structures can cause in which benefits are diffuse but costs are associated with inequalities in both the ownership and security of property concentrated and organized groups? rights. Unequal land ownership has been a central feature of Actual experiences show that reforms do occur, with the Latin American history (see chapter 4) and remains prevalent role of different groups varying across cases. This is seen in in many parts of the region today. However, inequalities in a five-country study of Bolivia, Brazil (in the state of Minas security are also prevalent, as made clear by typically weaker Gerais), Ecuador, Mexico, and Nicaragua by Grindle (2002; land rights in poorer urban areas (de Soto 1989 and 2000). see also discussion in Ames 2003). Efforts in all five countries Furthermore, in politically unstable societies, one solution involved decentralization to the state, municipality, or school, has been to generate institutional mechanisms for selective and to differing degrees involved providing training and commitments on property rights that typically involve tight career advancement opportunities to teachers and promoting alliances between political and economic elites, as discussed student testing, parental involvement, and curricula reform. in chapter 5 (see Haber, Maurer, and Razo 2003 for a detailed Teachers' unions were often found to be sources of resistance. discussion on the case of Mexico). However, in every case except Ecuador, significant reforms More secure property rights for all will thus generally be were undertaken that involved leadership from the execu- good for equity, especially if they influence patterns of invest- tive branch (that is, by presidents or ministers of education) ment by households and property owners across the income and persistent action by reform teams. distribution. However, there are potential tradeoffs. Where Do unions have to be the villains in reform processes? property ownership (as opposed to the security of property Maceira and Murillo (2001) argue that it is necessary to take rights) is very unequally distributed, a case can be made for into account the market position, beliefs, and alliances of redistribution. Nonetheless, threats of expropriation for teachers unions, but that engagement and negotiation with redistributive purposes will lower the investments of prop- worker representatives can be a source of both sensible design erty owners, in turn reducing growth (and possibly employ- and, importantly, greater buy-in by stakeholders. Indeed, in ment). These issues are discussed below, first in the context at least one of Grindle's (forthcoming) cases (the Brazilian of rural land and then with regard to urban areas, which also state of Minas Gerais), unions appear to have been advocates includes the consideration of housing. for reform and to have long called for greater democracy in education. Similarly, the Colombia teachers' union has sig- Rural land reform nificant progressive elements, especially in the pedagogic Inequalities in land ownership have been at the center of the movement, that favor decentralization, greater school historical formation of social, income, and political inequal- autonomy, and teacher appointments by open competition ities in Latin America (see chapter 4). The typical pattern rather than clientelism (Angell, Lowden, and Thorp 2001). in the region has been the creation and maintenance of large landholdings for groups of colonial origin, followed in most 7.2. Property rights, land, and housing countries by the extensive appropriation of land by elites after Property rights matter for both growth and inequality. independence in both the 19,h century, and, in some coun- Security of property rights is essential to investment, and tries in the 20th century. A wide range of mechanisms were 189 INEQUALITY IN LATIN AMERICA FIGURE 7 5 Poladzation and property rights across countries ICRG composite index 95- y = -20.218x + 78.092 j 0 R2= 0.2097 00 850- , as * 0 * 75 - O*TTO 65 - o o o e ~0 URU @ < 1 @@ G U@ O PRY * OVEN 55 -00 S S NIC O HTI * ARG 0 45~~~ - 0~~~~~~~~~~~ 45- 0 0 o * 35 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Ethnic fragmentation index ICRG composite index 95.0 - y = -0.4487x + 87.589 R2= 0.2134 0~~~~ O %. 85.0- 00 o a 0 0~~~~~~~~~~~~~~~ 65.0~ ~ ~ ~ ~ ~~~~~~~~~~. H 75.0 - 0~ ~ ~ I $TTo * H ° ~~ * C e O VEN O S PRY 55.0- 0 0 ~ ~ ~ * * ARGO° NIC e e 45.0 - 55.0 -~~~~~~~~~~~~~~ 35.0- I I I 15.0 25.0 35.0 45.0 55.0 65.0 75.0 Gini coefficient Source: Authors' calculations, based on the ethnic fragmentation index in Alesina and others 2002, income Gini coefficients calculated for this report, and the composite index from the International Country Risk Guide (ICRG) that is a proxy of credibility and predictability of property and contractual rights. (See Knack and Keefer 1997 for a discussion.) 190 POLICIES ON ASSETS AND SERVICES TABLE 7.4 TABLE 7.5. Gini coefficients for the distribution of operational holdings of Gin! coefficients for the dstriibution of operational holdings of agricultural land across regions agricultural land in Latin American countries Deininger and Olinto UNDP Deininger and Olinto UNDP Region (I) (ii) 1950-1979 1980-1994 1981 (i) (ii) (ii) Latin America 0.81 0.74 Middle East and 0.67 0.56 Antigua 0.74 North Africa Argentina 0.86 0.85 North America 0.64 Bahamas 0.90 0.87 Sub-Saharan Africa 0.61 0.51 Barbados 0.90 0.93 Western Europe 0.57 Belize 0.72 0.71 East and South Asia 0.56 0.52 Bolivia 0.77 Brazil 0.83 0.85 0.86 Colombia 0.85 0.77 0.70 Note: Column (i) shows averages for the 1950-1994 period. Costa Rica 0.81 Column (ii) shows values around 1981. Chile 0.64 Source: Deininger and Olinto 2000 and UNDP 1993. Dominican Rep. 0.80 0.70 Ecuador 0.86 0.69 El Salvador 0.83 used to effect this process, including expropriation of Grenada 0.78 0.74 0.69 Guatemala 0.86 indigenous lands during both the colonial period and the Guyana 0.68 19th century (for example in El Salvador, Guatemala, and Honduras 0.75 0.64 Mexico), control of labor through slavery to make land pro- JMexaicoa 0.59 0.81 ductive (notably in sugar production in Brazil), forced labor Nicaragua 0.80 requirements, debt peonage, imported indentured labor, and Panama 0.71 0.87 0.84 Paraguay 0.86 0.78 0.94 vagrancy laws. These mechanisms were complemented by Peru 0.94 0.61 interventions related to taxes, inputs, and other market- Puerto Rico 0.73 0.77 related aspects. Trinidad 0.68 Comparative information on the inequality of land Uruguay 0.82 0.80 0.84 ownership is unreliable, since standard measures don't adjust RB de Venezuela 0.92 for land quality and use. With this word of caution in mind, tables 7.4 and 7.5 show the commonly quoted land-related Note: The values for each country correspond in column (i) to the first value in the 1950-1979 period and in column (ii) to Gini coefficients. For Latin America as a whole, the Gini the most recent observation during the 1980-1994 period. coefficient was 0.81 in the second half of the 20,h century, Source: Authors' calculations based on Deininger and Olinto compared with 0.56 in Asia.8 As discussed in chapter 4, the 2000 and UNDP 1993. prevailing pattern of landholding and production became intimately linked with forms of political domination and the weak influence of subordinate groups, trends that were government, these reforms were only partly reversed, with supported by the neglect of education and other social a significant component of distribution to small and services. medium farmers.) Costa Rica-for reasons related to socio- Although the frontier settler countries of Argentina, economic structure and political transition in the late 19th Chile, and Uruguay had small indigenous or Afro-descended century-developed a system of political influence that gave populations, they also developed highly unequal patterns of substantial influence to small and medium farmers over the land ownership. Argentina and Uruguay had, if anything, shaping of policy, along with an associated early broadening slightly higher than average levels of inequality, in large part of suffrage and the provision of basic education. This process because of the extensive pattern of livestock production. was associated with relatively equal land distribution in the Chile's relatively modest land inequality, represented by a Valle Central, the most populated and important area of the Gini coefficient of 0.64 around 1981, was influenced by land country. Outside this area, land was unequally distributed, reforms of the 1960s and early 1970s. (Under the Pinochet which explains the high overall Gini coefficient. 191 INEQUALITY IN LATIN AMERICA Redistributive land reform has long been advocated of marked agricultural dualism-which are typical in Latin both as a source of greater short-run equity and efficiency American countries-the small farm sector usually remains and as a measure to support the transition to a more rapid disadvantaged with respect to access to financial and tech- and equal development path.9 However, it is important to nical services. recognize the complexity of the relationship between land In addition, small, poor farms are generally located on size and productivity. Current thinking on the land size- marginal land, often in mountainous areas, that offer poor efficiency relationship in Latin America recognizes this conditions for commercial farming. The advantage of large complexity.10 farms has frequently resulted from an array of distortions in For most crops and under normal conditions of avail- policy induced by the political influence of landed elites. ability of mechanical services, production is neutral with Such distortions range from output, input, and credit sub- regard to scale for a wide range of farm sizes, starting with sidies to the public provisioning of infrastructure to meet fairly small areas of one or two hectares. For grains and the needs of large farmers (see Binswanger, Deininger, and other field crops that require a small labor input and little Feder 1995). management, the relevant area can be large (that is, up to This situation suggests that small farms can be efficient several hundred hectares). For farms growing labor- alternatives for production, but that this depends on con- intensive crops (mostly fruit and horticulture) or when ditions specific to particular crops and associated factors such management requirements are substantial (because of such as marketing and credit. The importance of context in turn factors as varying soil conditions, erratic weather conditions, implies that attention should be given to the likely hetero- frequent pest incidence, and mixed farming), returns on geneity of conditions. In one review of empirical work in six production begin to decline rather quickly. This is because countries, L6pez and Valdes (2000a, 2000b) found that the under similar technology and market conditions, family contribution of land to per capita income is low (even though farms are generally more efficient than wage-labor farms returns on production are typically either constant or decreas- with regard to managing labor and ensuring quality man- ing), especially when compared with other factors of pro- agement practices. duction such as human capital. For crops that require immediate processing (like coffee, However, recent work on Mexico by Finan, Sadoulet, and sugar, or African palm), small farm size is not a problem at deJanvry (2002) (summarized in box 7.3) finds a more com- the farm level (that is, in terms of production), but can be plex relationship between land size and welfare, with the a disadvantage at the industry level if coordination between potential for making large, poverty-reducing gains from farmers and mills is inadequate. For coffee, farm size is not levels of land ownership as low as one or two hectares. These a significant problem because processing-in which authors also emphasize the powerful influence of other house- timeliness is essential-is carried out on the farm with hold characteristics, including education and indigeneity, simple means. However, processing can make a difference and of the economic context, with roads having a strong for sugar, palm oil, and banana production; for these crops, influence on productivity and household welfare. both wage-labor plantation systems and family farm systems There is also some evidence of the dynamic advantages have proved to be competitive. Industrial crops that do not of lower land inequality. The negative influence of land require immediate processing (such as cotton, a variety of inequality on subsequent growth has been documented oil seeds, and to some extent cocoa) tend to be grown com- in econometric work, as illustrated by the bivariate petitively on small and medium family farms. relationship shown in figure 7.6. Such cross-country results Large farms often have advantages in terms of access to are only suggestive, as has been emphasized throughout this input and output markets, financing, and contracting of tech- report. Earlier chapters have discussed a range of potential nical assistance. This can sometimes offset diseconomies of mechanisms through which land inequalities may be asso- scale associated with management and labor supervision. The ciated with lower growth. These include the impact on advantages of large farms could potentially be replicated by policy distortions and, perhaps more profoundly, links with small and medium farms if they coordinated input and the creation of weak and unequal institutions that dampen output needs through cooperatives or similar associations, growth prospects. In contrast, work that interprets the as is done in many parts of Europe. However, in situations sources of rapid East Asian growth has often emphasized 192 POLICIES ON ASSETS AND SERVICES BOX 7.3 lhT povety reductin potena of land in rural Mexico A major question that arises when developing land policy is the effect of increased land on the welfare of the poor. The marginal welfare impact of increased land Although almost all studies find a positive impact both on production and income, empirical work on several Latin American countries by L6pez and Valdes (2000a and 1500 - 2000b) concludes that the effect of land on the size of income is low for small farmers. This implies that only 1000 - large increases in land would lift poor rural farmers out t*. of poverty. Work on rural households in Mexico by Finan, 500 - Sadoulet, and de Janvry (2002), based on a 1 99' survey, casts important light on this issue. These authors use a o I t I t methodological approach (specifically. semiparametric 0 1 2 3 4 5 regression techniques) that allows for highly nonlinear rela- Land size (hectares) tionships with land size. They also explore the impact of Note: The upper and lower lines are the 95 percent confidence complementary factors and use a broad welfare measure intervals. Source: Finan, Sadoulet, and de Janvry 2002. that captures the multiple dimensions of well-being, as well as income alone. There are two main results of this approach. First, access to even a small plot of land can raise household welfare members had the nonland assers and the same returns as significantly: for small landholders with at most one the members ofgroup B. The third line WA2) adds in the hectare of land (which means 30 percent of farmers in the higher returns received by group B Finally the top line survey), an additional hectare increases welfare by an shows the relarionship between land size and nonland average of 1.3 times the earnings of an agricultural worker. assets. retrurns on education, and recurns to land for group (This is illustrated in the figure below, which shows the B. The combinatioin of differences in nonland assets and peso value of marginal increases in land.) This pattern is differences in returns on these assets explains a high pro- partly due to the fact that land can facilitate the more portion of the gap effective deployment of a household's labor resources in the The second figure (b) undertakes a similar exercise context of imperfect labor markets and local unemploy- between indigenous and nonindigenous households. ment. As the size of land increases, its marginal value tall s Results show that nornind igenous households not only pos- to a much lower level. sess higher nonland assets, but also get higher returns on Second, there are important complementarities between these assets, there is an unexplai ned premium of 5 5 percent land and other influences. FX nan, Sadoulet, and de Janv r) ofthe welfare differenrial If indigenous households had the (2002) explore this by comparing farming households with same ret urn on cheir assets as do nonindigenous households, different characteristics and decomposing the differences they would only require access to lIs than 3 hectares toreach in asset holdings and differences in these assets across the poverty line. compared with the 15 hectares needed households. This is shown in the figures (a-c) below. The given the rerurns they actually receive Finally, the third first figure (a) explores the impact of nonland assets for figure (ci shows the impact of road access; the infrasrruc- households with low levels of education (group A) and high ture context has a powerful impact on returns and on the education (group B). The bottom line (WA) shows the efficacyofincreased land in improving welfare.These rtsulrs estimated relationship between welfare and land size for underscore the importance of developing an inregrared households with low education. The second line (WAI) strategy for rural development, including the need tounder- estimates the relationship for group A, assuming that its stand the special conditions faced by indigenous groups. 193 INEQUALITY IN LATIN AMERICA BOX - i (Continued) a. The welfare impact of land assets and interactions with b. The welfare impact of land assets and interactions with education indigeneiHy Wellare index VVellare .rdex I S - US ~~~~~~~~~high educat,on 1 2 - 1r~ A~. ii~i ed~0Ig I 2 Non ndigenous 0 8 0 6- O S-= Pover line 0 4 - bne 05 ~~~~~~~~~~~~~~~~~~~~~0.2 4 5 6 71 -0 4 I L_ ~[ovv educat.on -0 6 - 08 Indigenous - I S Label needed -- 1 Label needed c. The welfare impact of land assets and interactions with road access velliare rndex 04 - oo 42 - ~~~~~~ZZ g~~ Poverty line 0- ~ ~ ,~ z X; Lnarea (ha) -0O2 / 4 5 6 7 8 9 10 - 0 4 - itho road access - 0 6 Label needed Source Finan, Sadouler. and Jan.rs 2002 how relatively equal land distributions, combined with the El Salvador, Nicaragua, and Peru (see table 7.6). These were political imperative to provide benefits to peasants, have sometimes motivated by peasant mobilizations, but also- resulted in policy choices that favor the broad-based especially during the Cold War era and in the wake of the provisioning ofeconomic and social services.' This pattern Cuban revolution-by pressure from above and outside, has fostered a relatively dynamic rural base and the including from the United States under the Alliance for upgrading of the labor force that contributed to the indus- Progress. These forms of pressure had the joint objectives trial takeoff. of increasing the well-being of peasants and reducing the probability of a rural communist insurgency. The history of "incomplete" land reforms Throughout the 20"h century, the possibility of redis- in Latin America tributive land reform was also politically facilitated by rising Despite the resistance and influence of landed elites, it is urbanization and the growing importance of urban relative notable how much redistributive land reform actually to rural elites. In some cases (Chile for example), alliances occurred in Latin America during the 20,h century. The between the peasantry and urban-based, social democratic Mexican revolution led the way, and was followed over movements formed the political foundation for land reform. the decades by large-scale reforms in Bolivia, Chile, Cuba, Yet the results were largely disappointing. By the late 1980s, 194 FiGURE 7 6 1970-99 Land inequalfY and grom**l7O' 1970-99 0~~~~~~~~~~~- -0255X 3~a42 3 Ave,,ge g,ovwth rateS n6PpfP8-O05s -3 7 6 *DOM 4 o BOA ~~~~~~~~~~~~~~~~~~~~~~~~0 *E C U La 35 45 7nd Gini coeffCicent 25~~~~~~~~~~~~~~~~~~~~~~~~~~~~9 o eIc.ulatori ased on 13nd ineqU a TABLE 7.6 in Latn hMeliCa ill the 20CeNW,9 27ne 1hou 2.ehl Bolivia~~~~~~~~~~~~~~~~~~~~~~~geeiir hUeOd E l s a l v a d o rS ;e O u a NicaragUa 1973 ous decade, for which data a not availablre.s of arable ( Total9lani d l n e de Janv and Sdoulet 1939) (treousantd n hethle 6lost game of Latin A ericaoad reom bc3eo telc otiutdt h akof4 intres MedCium nd 1a9g3fa 7er onte ttea 5Cbf9tl theas opn9n thMol ak 90W rdDeeomn Repor ESavdimiised19h7pedO innc offrt uhriain5n1W rl7ak 90 de ln eorm toi be t god fre Nicaragua 1973 e,~~~~~~~~19 INEQUALITY IN LATIN AMERICA poverty reduction in principle, but to rarely be feasible The "incompleteness" of land reform processes also played except under the unusual circumstances of colonial action out in a variety of settings. In both Mexico and Bolivia, a or revolution. genuine large-scale transfer of land from haciendas to peas- What happened with the major land reforms of the 20,h ants occurred in the wake of revolutions. However, in both century? They clearly did not usher in the hoped-for trans- cases long-term productivity has been low and stagnant and formation of social and economic inequalities in Latin rural poverty has been deep and persistent. In Mexico, there America. Nor, for the most part, did they generate the kind was initially strong state support for the newly created ejido of vibrant smallholder sectors that were so important in sector-a system of communal lands based on expropriated furthering East Asia's dynamic development path. However, private landholdings-with regard to efforts related to irri- the reason was not because land reform was in principle ill- gation, credit, and technical assistance. The poor performance conceived. Rather, major land reform efforts were often of the sector may in fact partly reflect weaknesses in longer- poorly designed-notably where ill-fated production coop- term support, but an important element was an underlying eratives were emphasized-and, more fundamentally, were problem in the design of land management. "incomplete" (de Janvry and Sadoulet 2002b). Restrictions on tenancy and the subdivision and sale of This incompleteness existed both with respect to the gen- land undermined incentives for production and structural eration of competitiveness in the post-reform sector and to change (World Bank 2001c). The 1992 reforms were a giving power to peasants, whether in terms of direct or response to these problems (as discussed below). The ejido implicit influence over decisionmaking (the latter being a sector was also incorporated into vertical, clientelistic feature of East Asian successes even under authoritarianism). patterns of political relationships, backed by food-related Although the communist transition (particularly in China programs and the provisioning of local works that helped and Vietnam) and communist threat (in much of the rest of sustain rural support, in particular for the Partido Revolu- East Asia) made governments respond to the interests of the cionario Institucional (Institutional Revolutionary Party, or peasantry, the threat of Cuban-style revolutions in Latin PRI) (Fox 1993). There is evidence of the evolution of America spawned halfhearted land reforms and heightened unequal power within ejidos that provided local counterparts repression, at least until the democratization movements of of statewide and national forms of clientelism (World the 1980s ensued. Bank 2001c). The political agency of landed elites was one of the fac- In Bolivia, the failure to generate a competitive and influ- tors behind the weaknesses in reforms. A specific element ential peasant sector was much starker. The land reform was the economic response to the threat of land redistribu- occurred mainly in the Altiplano and valleys, with little tion, or the heightened insecurity of property rights. This action taking place in the then sparsely populated Oriente. in particular induced both low levels of tenancy and the Relatively egalitarian land structures were formed in the eviction of tenants, followed by shifts into more capital- or former areas, but there was again a failure to support peas- land-intensive forms of production, such as livestock pro- ant-based production. This failure had political roots. The duction (Binswanger, Deininger, and Feder 1995, deJanvry horizontal alliances that were a force behind the 1953 and Sadoulet 1989). In some cases, resistance took the form revolution were quickly reoriented toward vertical incorpo- of the suppression of rural movements, tenant evictions, or ration of these groups into the new national structures, large-scale violence between peasants and governments or notably within the leading political agent of the revolution, other forces associated with landed elites. This occurred, for the Movimiento Nacional Revolucionario (National Revolu- example, in Colombia, El Salvador, and Guatemala during tionary Movement, or MNR). the 20th century. Even in countries where major violence did With this change, the peasantry entered a new phase and not occur, the effect on the whole society of such "prema- took on a new form of neglect (albeit without haciendas, or ture ejection" of labor from productive land was strong, and estates, controlling them), except for their continuing impor- increased inequalities by placing greater pressure on mar- tant role in providing votes when elections were held. The ginal lands, rural, nonfarm production, and urban unskilled restructuring of unequal vertical relationships overlapped labor markets (see Binswanger, Deininger, and Feder 1995, with the continuation of identity-based differences despite, Conning and Robinson 2002). or perhaps because of, the much greater predominance of 196 POLICIES ON ASSETS AND SERVICES indigenous groups in Bolivia than in Mexico.12 Meanwhile, became a source of radical resentment. This in turn provided in the Bolivian Oriente, where soils and agroecological con- fertile ground for the emergence of the violent and mil- ditions were better, a concentration of land occurred behind lenarian Sendero Luminoso (Shining Path) movement, which the backs of the legislature, often in a corrupt fashion. Only is much closer in approach to the Khmer Rouge in during the 1990s did elements of a more broadly inclusive Cambodia than to traditional guerrilla movements in Latin national project reemerge under a "pluricultural" constitu- America. The main victims of both Sendero and the mili- tion. This effort did build on some of the social changes of tary effort to respond to the movement were poor rural the 1950s, notably the granting of legal status to and the peasants."3 Political and economic order was finally restored potential influence of "base groups" that were typically a under the semi-authoritarian regime of Alberto Fujimori syncretic mix of indigenous structures and rural unions. How- (1990-2000), who reversed some of the earlier land ever, progress in land reform remained slow, and reemerged reforms (especially those affecting the more fertile coastal as a central issues in new waves of protest in 2001-2003. areas) and undertook substantial expansion of social and Both Peru and Chile introduced collective land reforms. infrastructural services to peasants through a kind of Peru's case is both instructive and tragic, especially with national-level clientelistic project.14 However, little was done respect to the dynamics of change in the poorer Sierra (the to strengthen the competitiveness of peasant-based agri- strongly indigenous Andean region). The reforms were part cultural production. of an ideology of giving power to the peasants. As President Other land reform efforts also remained largely mired in General Velasco Alvarado said in 1969 after seizing power, problems related to failure to achieve competitiveness, "Peasants, the landlords will no longer eat your poverty!" notably under cooperative structures (Cuba and Nicaragua (Seligmann 1995, p. 1). However, during the course of under the Sandinistas) or in sites of conflict (El Salvador). implementation, the Peruvian land reforms had features With the shift to a heightened emphasis on markets starting of "high modernist" projects (Scott 1998) that favored the in the mid-1980s, issues of redistributive land reform took larger-scale cooperative production of expropriated hacien- a much lower place on the agenda. (On the Washington das over support to smallholder agriculture. Consensus list, the important issue of property rights was The reforms also failed to provide complementary services typically interpreted as a message of increasing those of exist- and favored an integrationist cultural agenda rather than the ing incumbents, thus implying that the redistribution of empowerment of indigenous aspirations. Despite (or partly property rights was to be avoided.) because of) attempts to shift the highly unequal balance of Major 20,h century efforts for direct land reform were power in land-related court cases to peasants, land disputes disappointing. Did the past 15-20 years of market liberal- and local conflict were rife. With worsening aggregate eco- ization yield better results by providing an incentive regime nomic conditions in the 1970s and 1980s, service provision for labor-intensive agricultural development? There were to poor rural areas remained dismal and poverty worsened, positive developments in terms of the reduction of inefficient even as deep, old resentments were stirred and new aspects distortions in product, input, and credit markets that had of expectation, mobilization, and conflict emerged. The dis- been created under the influence of landed elites. In some mantling of agrarian reform cooperatives took place in the cases, this helped bring down land prices (World Bank early 1980s, with the parceling of land occurring as an 2003b). The reduced protection of import-substituting implosive phenomenon rather than as a counter reform. industries (through tariffs or the overvaluation of real Credit and extension services were cut back and the new exchange rates) shifted incentives to tradable production and parceleros (parcel owners)-many of whom were agricultural helped some parts of the agricultural sector-though not workers with little entrepreneurial experience-were left on producers of subsistence or other nontraded products, which their own, with disastrous results. are often important among the rural poor (see Schiff and Even important educational advances had mixed effects. Valdes 1992 for a review). The first major cohorts of indigenous groups with an edu- The last few decades also sparked high hopes that labor- cation emerged with hopes of attaining modern jobs, but intensive, export-oriented agriculture would help solve the in reality faced miserable economic opportunities. Local employment and rural poverty problem and, at least implic- teaching was one outlet, but low-paid school teachers itly, render the politically difficult challenge of land reform 197 INEQUALITY IN LATIN AMERICA unnecessary. Chile appeared to be a model for this process. ship involve transfers from medium to small producers. This Under the rule of Augusto Pinochet (1973-1990), about pattern is attributable to four factors: the high levels of labor one-third of expropriated lands were restituted to their large- interactivity required in the production process (where small- scale owners and most of the remainder was allocated as holders have an advantage); an initially highly fragmented private parcels averaging around 20 hectares. When growth land ownership structure in this part of Guatemala; took off (following the economic crisis of the early 1980s), contractual linkages with processors that also facilitated export-oriented agricultural expansion was a central com- working capital; and the ability of farmers to pursue self- ponent. During the 1980s, there was an enormous expan- insurance strategies by mixing exports with food crops. sion in nontraditional agro-exports. In Paraguay, both the pattern of initially adopting soya Chile was not alone. Other cases-involving in large part and wheat and the induced structural changes were exclu- the successful association of small farmers and processing and sionary. This was because of a prevailing mixture of techni- exporting companies-include the development of nontra- cal factors (some crops required less labor interactivity), ditional exports, for example, decorative plants, palm kernels, economic institutions (smallholders lacked the means to and tropical fruits, on the Atlantic coast of Costa Rica; access working capital), and initial land allocation processes vegetable production in Guatemala; successful exports to the (the frontier region had relatively large land allocations and United States of fruits and vegetables by irrigation farmers a land market that facilitated unequal agrarian change). in northern Mexico; production of tropical fruits and veg- There is no guarantee that small farmers or rural work- etables for international markets in northeastern Brazil; soya ers will be major beneficiaries of agro-export booms. Agro- and wheat production in Paraguay; and the recent expan- exports can favor poor farmers or workers, but the degree sion of palm oil exports from Colombia. Yet while advocates to which this actually happens depends on both technical saw such developments as a source of growth and jobs, critics factors and the institutional context. Although smallhold- emphasized the adverse effect that they had on the rural poor, ers have often been favored by labor interactivity in some primarily in the form of reduced access to land, uncertain of the agro-export crops and initially fragmented land hold- work, and rising food prices. ings, most competitive biases work against them. This trend A comparison between export-oriented agriculture in is due to the labor- and skill-intensive character of most pro- Chile, Guatemala, and Paraguay by Carter, Barham, and duction processes, price-quality measurement issues, the per- Mesbah (1996) provides insight into the question of who ishability of many crops, the extended gestation periods of benefited from such agro-export booms. The study found some investments, and the lack of complementary insurance that outcomes were contingent on initial conditions, types markets. Similarly, as the case of Petrolina-Juazeiro in north- of crops, patterns of support, and the induced processes of eastern Brazil illustrates (see chapter 8), export production structural change. In Chile, for example, the agro-export on large farms can be consistent with decent work for agri- boom was dominated by medium to large farmers, in part cultural laborers, but this again depends on the institutional due to the information, packaging, and marketing require- context. In this case, the critical issue was the nature of ments of fruit production."5 With new pressure on the interactions between the union movement and modern traditional crops grown in the smallholder sector, there was farms. a substantial amount of selling out to large farmers, with almost 60 percent of the parceleros who had received land New opportunities and options under the Pinochet land reform selling their land by the late Land reform had moved to the bottom of the policy agenda 1980s. This exclusionary pattern (in terms of landownership) by the end of the 20'h century, reflecting shifts in the tides was partly offset by the rapid growth in employment on the of opinion and disappointments over earlier efforts. Yet the large farms, but the new jobs were mostly seasonal and paid issues that spurred reform have not gone away. Rural poverty stagnant or declining wages. remains pervasive and is almost always deeper than urban In contrast, in the Guatemalan highlands smallholders poverty. Agro-exports can help alleviate the situation, but have been the main actors in the boom in winter vegetable will not be an automatic solution. The context of the early crops. These crops are 50-300 percent more labor-intensive 2lIs century provides potential opportunities for intensified than traditional crops. In this case, changes in land owner- action. There is heightened concern today over poverty and 198 POLICIES ON ASSETS AND SERVICES social exclusion and continued demands from established considerations. In Colombia, recent legislation allows for the movements (such as the Movimento Sem Terra in Brazil and confiscation of the lands of druglords. The situation is indigenous groups in many countries). Perhaps most impor- different in countries where the land rights system has tant, more recent experience suggests that land reforms can legitimacy, such as in the Southern Cone countries or where be designed to be effective and to reflect lessons learned from there are no lands to redistribute (for example, in the Bolivian the mistakes of the past. highlands, Mexico, and Peru). An approach involving multiple paths and activities is out- Five approaches to land reform are outlined below. It lined below. There is no single "high modernist" or other should be noted that in all of these examples, infrastructural, route to rural development and agrarian change. It is increas- technical, and institutional support tailored to local condi- ingly understood that there are multiple exit paths out of rural tions is essential to ensure competitiveness. poverty, including migration, intensified smallholder pro- duction, and unskilled and (increasingly) semiskilled work Strengthening tenancy markets As already noted, the tenancy on farms and in rural nonfarm activities and small towns. market in Latin America is severely underdeveloped relative Transfers for redistribution and risk management can also play to other parts of the world, especially when compared with a complementary and transitional role (see chapter 9). developed countries. The primary reasons for this are weak This complexity implies the need to avoid standardized property rights and a lack of conflict resolution mechanisms, solutions. One example involves plot size. "Unviable" plots sometimes combined with prohibitions on renting (for exam- for agricultural production can in fact play an important ple, pre-1992 in the Mexican ejido sector). Yet rentals can role in income generation and diversification, even when be a major mechanism for the redistribution of land access, the bottom 20 percent of rural households typically gets with significant efficiency gains. Renting is also often part only 30-50 percent of its income from agriculture.'6 As of the land acquisition "ladder" for small producers. There discussed in box 7.3, increasing land to the 1-2 hectare is a strong case to be made for tackling existing disincen- range in Mexico can have a significant impact on poverty. tives to rent (see below for an example from Mexico More broadly, achieving competitiveness is a factor of post-1992). There are also a variety of experiments that seek overall spatial strategies, local participation, and linkages to support tenancy and deserve careful tracking and evalu- between agricultural and nonagricultural sectors, as empha- ation; examples include the Landless Workers' Consortium sized in "la nueva ruralidad." (Food and Agriculture Orga- and Sharecropper and Rural Leasing Exchange in Brazil and nization and the World Bank 2003). In addition, as noted group rentals in Honduras. above, support for smallholder production also involves tackling remaining distortions that favor large-scale Land titling Insecure property rights can reduce produc- production and developing infrastructural and institutional tion growth because of reduced incentives to make invest- services to support smallholder production. Finally, agrarian ments and the limited ability to use land as collateral for change is often intimately connected with the identity- credit. According to surveys conducted in the early- to mid- based inequalities and subsequent needs for and policies on 1990s, about 63 percent of farmers in Chile, Colombia, social incorporation that were major themes in chapters 3 Honduras, and Paraguay lacked legal title to their land and 7. (L6pez and Valdes 2000a). Empirical work on Honduras and Both farmers and countries across Latin America-as well Paraguay has identified significant impacts of this trend, as the extent to which land reform is on the political and for example an increase in per capita income by 5 percent economic agenda throughout the region-vary greatly. It is in Honduras. However, most benefits were captured by important to distinguish between countries (and among areas medium and large farms. Work on Asia suggests that land within countries) with regard to where existing land rights titling alone is insufficient. To reap significant benefits, are and are not contested. The former applies to parts of titling should be complemented by a fair and effective legal Bolivia, Brazil, and Colombia (for example in parts of the system, cadastral surveys, and implementation of enforce- Bolivian Oriente). In these cases, agrarian reforms with ment mechanisms, in addition to actions that enhance some expropriatory elements may have a role to play, with competitiveness in terms of infrastructure and other eco- their design depending on both political and technical nomic services. 199 INEQUALITY IN LATIN AMERICA Community- andmarket-basedlandreform In the early 1990s, with mechanisms for regulating the overuse of common the World Bank and other agencies became associated with resources. (See chapter 1 in World Bank 2003b for a review a new approach to land reform. This involved decentralized of such methods.) Indigenous groups in Ecuador and Bolivia and voluntary land sales backed by both credit and grants have succeeded in obtaining collective rights that involve for land purchase (in addition to complementary actions to restitution of ancient claims. When traditional mechanisms enhance competitiveness). Such assistance is needed because work well, such approaches can be both equitable and effi- land prices typically exceed in-use values due to other sources cient solutions, but there is also a risk of maladaptation to of land value (for example, as an investment and source of dynamic processes of social and economic change. With rapid security and status), and even after distortionary policies are change, problems of insecurity, conflict, and overuse of removed. There are some interesting experiments under way, common property become more important. In addition, notably in the C6dula da Terra project in Brazil, but rigor- indigenous sociocultural systems can perpetuate traditional ous evaluations on them are not yet available. There are also patterns of inequality. Institutional design needs to be shaped disadvantages to this approach, especially those associated in a way that simultaneously allows for change and integrates with fiscal cost. In addition to upfront fiscal subsidies, a con- transparent, local democratic processes so that decisions can tingent fiscal liability applies if farmers default on loans used be made as local land policy evolves. for land purchase. These fiscal costs may limit the potential An example of the transition from collective toward more for scaling up community- or market-based land reform. individual landholdings is the ejido (communal land) sector Nevertheless, this approach may be a valuable instrument in Mexico. The 1992 land reforms, which involved changes for some groups of farmers, especially under conditions in to the 1917 constitution, were the product of a political com- which there is mutual acceptance of the legitimacy of land promise between those resistant to change in the ejido sector ownership by both the seller and the purchaser. and advocates for a full-blown shift to property rights. The resulting hybrid reform may have been a judicious choice. Negotiated recuperation of lands There are substantial but The self-governance of ejidos was strengthened through the underutilized opportunities for recuperation of lands for use effective recognition of their legal standing and autonomous by poor people at a relatively low cost. This includes the decisionmaking power over internal matters, and in partic- potential distribution of public lands. In Bolivia, the gov- ular by allowing members to decide on the form of land own- ernment has identified 900,000 hectares of land suitable for ership within their entity (including full individual property settlement; in the Dominican Republic there is a backlog titles). Land rental was fully supported and sales were legal- of already expropriated land. This approach may also involve ized but restricted to others within the ejido, except when a recuperation of illegal settlements, given the potential community decision was made to shift to dominio pleno. or welfare dimensions of this process. There are also opportu- fully individualized titles. In addition, an independent nities for negotiated deals in areas where land title is system of agrarian justice was set up to administer a wide ambiguous, that is, where partial release of land with uncer- range of land disputes. tain title is conducted in return for titling of the smaller Some 50 million hectares and 3 million households have part. The mayor of Brasilia has an ongoing scheme of this had land regularized under the ejido reform, including 1 mil- kind, and a comparable scheme is under way in the Domini- lion households that previously had no land rights at all.17 can Republic. Many Latin American land reforms have No major sell-off or concentration of landholdings has incomplete titling and present new opportunities. Releas- occurred, as feared by some. Panel data for 1994-1997 show ing land in return for greater security of the remaining plot that those who participated in ejidos also experienced rising can be an attractive deal for landowners. incomes, primarily because of an expansion in off-farm work. An extensive program of resolution of conflicts over land was Land reform and indigenous, communalprocesses One aspect of also established and there is some evidence of greater trans- land reform that is of particular relevance to indigenous parency and deepening in governance within ejidos, which groups involves collective rights to land. Indigenous cultures have been subject to both external party influence and sub- often have effective ways of managing property that balance stantial internal inequalities in power. The agenda remains the need to provide security and incentives for investment unfinished, however, especially in areas where land disputes 200 POLICIES ON ASSETS AND SERVICES and other conflicts are severe; this is particularly the case in TABLE 7.7 the southern states, which are poorer and have a higher Distibuton of housing wealth in Brazil, Chile, and Peru concentration of indigenous people than other areas. Nonetheless, the ejido process still represents a complex Peru Chile Brazil reform that may facilitate a process of modernization with- 1991 1999 1992 2000 1992 1999 out the sudden removal of existing mechanisms that provide Housing wealth Gini 0.66 0.72 0.69 0.59 0.73 0.66 security (World Bank 2001c). Income Gini 0.45 0.44 0.56 0.57 0.56 0.57 Share of housing value Urban land and housing held (percent): With regard to urban areas, the focus is placed here on the Income quintile 1 14 4 17 11 7 8 Income quintile 2 17 6 11 13 9 9 housing issue, which has important links with issues of land Income quintile 3 18 9 13 17 13 13 tenure."8 For most people in urban areas, housing is the Income quintile 4 25 18 17 22 20 20 Income quintile 5 25 62 42 38 52 50 single most valuable item they will ever own. For the poor Income_quintile ___25 _62 _42 _38_52___ and much of the middle class, home ownership and the accu- Note: The housing wealth Gini coefficient shows that the mulation of consumer durables are typical asset-building "poor" are nonowners with zero housing wealth. This strategies. Expenditures on housing account for an average measurement is therefore not strictly comparable with the share of housing value that is held by the various income of 25 percent of household consumption. What happens to quintiles. The Brazil data are from Reis and others (2000) and the distribution of housing is likely to have a strong impact the other data are from Fay and others (2002). on asset inequality in any given country. Homeownership itself is not highly correlated with fixed sum for rent every month. This can be important in income (see table A.8). Homeownership often, but not times of crisis. For example, in Uruguay 10 percent of people always, increases along with an individual's income. In who did not own the house in which they live had to Argentina and Ecuador, for example, it is higher in the move following the economic crisis in order to cut down on poorest quintile than in the second and third quintiles. In housing costs. Housing services can also be monetized quite addition, homeownership even among the very poor is quite easily by taking in renters or additional household mem- high (above 60 percent for most countries), with the excep- bers.19 However, being a homeowner may also have some tion of Ecuador and Colombia. A large proportion of poor downsides, for example by tying the poor to locations that homeowners have informal tenure; in Mexico, for example, are disadvantageous for work or unsafe due to crime or nat- a survey of the country's poor urban neighborhoods shows ural disasters. It is unclear how liquid housing markets are that only about half the homeowners had a formal title to in poor neighborhoods. their land. Even in Argentina, where the housing market is Most governments in the world have pursued active quite mature, a full 18 percent of all homeowners lack a full homeownership policies. Throughout Latin America, these set of titles. have typically taken two forms: mortgage subsidies that ben- Despite relatively equal access to homeownership, in the efit the middle and upper classes and the provision of "low- countries for which data are available the overall distribution income" housing of a size and value that is typically beyond of housing wealth is extremely unequal, even more so than the means of the low-income population. In Mexico, about inequality in incomes (see table 7.7). Chile in 2000 is a partial one percent of GDP is spent on housing subsidies that largely exception, with housing wealth inequality similar to income benefit the middle and upper classes, substantially more than inequality. This may be the result of a widely praised home- is spent on Oportunidades, the country's largest targeted social ownership program. The distribution of housing wealth program.20 Developed countries, on the other hand, have appears to be less stable than income distribution, but it is been more successful with their homeownership policies. unclear whether this is due to greater measurement errors or Two good examples are the United States and the United real trends (Fay, Yepes, and Foster 2002). Kingdom, where a secular trend of decreasing wealth Is housing a good asset to hold? Being a homeowner inequality has been associated with the spread of popular brings advantages of a constant flow of services and frees assets such as housing and consumer durables (Davies and liquidity-constrained households from having to generate a Shorrocks 2000). 201 INEQUALITY IN LATIN AMERICA Most low-income families in Latin America acquire hous- ment agencies. Today, public policies aim for a more inte- ing through informal markets, the main characteristic of which grated approach and are adopted to support demand through is the gradual acquisition and upgrading of housing. It is esti- property rights, mortgage financing, and the rationalization mated that informal housing accounts for about one-quarter of demand subsidies. This approach in turn helps to promote of all urban homes in Latin America (Angel 2000), varying institutional development and to organize supply by pro- from a low of 10 percent in Buenos Aires to a high of 59 per- viding infrastructure for residential land development, reg- cent in Bogota (ECLAC 2000). Low-income families are ulating land and housing development, and coordinating the concentrated in informal housing for several reasons. The building industry. avoidance of cumbersome regulation and excessive standards High supply costs are often due, at least in part, to can lead to substantially cheaper houses. (In Buenos Aires, for regulatory or market failures. Many national or local example, the cheapest formal sector house or apartment costs governments impose minimum land use and housing stan- 2.7 times the median income; similar housing in an informal dards that place housing out of reach for large segments of settlement costs 0.8 times the median income.) The informal the population. Moreover, cities in many countries suffer sector also offers the opportunity to live in houses that are built from a scarcity of serviced land, usually because of the cum- gradually as financing resources become available. In the infor- bersome process required to develop land. In Ecuador, the mal sector, the gradually built house consists completely of laws governing residential development have created a equity, and can be lived in, sold, rented, or passed on as family cumbersome approval process for residential subdivisions patrimony, however modest and incomplete. that averages 16 months and imposes costs estimated at While informal housing is the most common choice for 30 percent of the value of new homes. low-income families, it has disadvantages from a public policy Public policies to ease supply-side constraints focus on point of view. The informal sector is poor at performing lowering costs, increasing the supply of serviced land, and the collective action role of ensuring that settlements have engaging residents in the provision of housing. This requires adequately defined rights of way, properly titled properties, administrative reforms to: and access to urban services such as water and sanitation. Fail- ure to plan properly for these aspects means greater capital * Make land tenure regularization and land transactions outlays in the future, for example to resettle families away easier. Peru offers an excellent example of successful from high-risk areas (such as floodplains), replot rights of way reforms of this type. In just over five years, the COFO- so that emergency vehicles and collective transport can access PRI project regularized 1.6 million lots and registered communities, and untangle legal claims on property that can more than 1.2 million titles by streamlining admin- take years or decades to resolve. istrative and legal processes and adopting a large-scale An effective and equitable public policy should be based approach to regularizing vast tracks of illegal housing. on a good understanding of the informal sector, and seek to * Make buildings and land subdivision regulations more work on both the demand and supply sides of the housing flexible and allow for reduced norms and standards. and land markets. To a large extent, policies have evolved. Experience in Brazil and elsewhere has shown that In the 1960s and 1970s, the public sector sought to directly with flexible standards, a house can be built at about provide low-income housing, typically linking it to clear- half the price of a formal housing unit. ing out slums and relocating populations in new settlements, * Allow for and encourage progressive provision of infra- which were often in more distant, cheaper locations that structure, whereby public investment focuses on major made transportation and other aspects of daily life costly and trunk infrastructure and leaves to the residents (in inconvenient. As a result, the poor would often sell their new partnership with developers) the process of installing houses and move into informal settlements closer to town. connections to their plots. Gradually, projects began to focus more on upgrading slums, * Develop innovations in materials and reduce system but high costs and the absence of cost-recovery policies installation and material costs. resulted in government agencies running out of resources. The 1980s saw a greater emphasis on reforming housing On the demand side, the main objective is to ease the finan- financing and the financial performance of related govern- cial constraints of the poor. This can be done through small 202 POLICIES ON ASSETS AND SERVICES savings schemes; by pooling group savings into community programs to upgrade slums. One of the better known is the trusts and linking the trusts to national housing finance pro- case of El Mezquital, a low-income informal settlement with grams; by providing access to staged and progressive loans a population of close to 40,000 people on the outskirts of and technical assistance; by encouraging community-based Guatemala City. In this community, an average public invest- production and employment; and through small targeted sub- ment of US$1,200 per household has stimulated substantial sidies in the form of up-front capital grants for people who increases in private investment and has resulted in signifi- cannot afford a down payment. Developing financing strate- cant improvements in the quality of life (for example, infant gies to meet the demand for low-income housing also helps mortality rates fell from 80 to 20 per 1,000 live births in stimulate the interest of private sector entrepreneurs in devel- four years) and an increase in land values by a factor of 11. opment and construction. A number of countries have In El Mezquital, community participation extends beyond successfully experienced such demand-side interventions. Per- the simple contribution of labor to involvement in planning, haps the most successful example of a direct subsidy is Chile's implementation, and evaluation of all activities. Management Unified Subsidy and Basic Housing Program, which enrolls capacity is promoted by working through the community. the lowest income groups in a savings program that eventu- Both project funding and contracting are carried out directly ally allows them to acquire a house through a combination by a community association. Community-based organizations of savings, a direct subsidy (which varies according to income ensure that the costs of investments are recovered and that but cannot exceed about US$4,000), and an optional mort- families pay for the consumption of urban services. gage credit. Again, these programs are flexible in that they allow participants to purchase an old or a new housing unit Land and property taxation or to build a house for those that already own a plot. The issues discussed in this section are linked with the tax In recent years, these principles-along with a greater issues discussed in chapter 9. Land and property taxes are emphasis on community participation and cost recovery- substantially underdeveloped in Latin America relative to have been applied to a new generation of very successful OECD comparators (as shown in figure 7.7). Even though FIGURE 7 7 Property taxation as a share of GDP in Latin America and other countries Property taxes as a share of GDP (percent) 3.5 3.0 - 2.5 - 2.0 - 1.5 - 0 Bolivia :1Uruguay 1.0 * C ~~~~~~~~~~~~~ -Italy E Fra%~e 0.5 - * * * ~ _n l wadpVenezuela 3 * t O - j * * Dorninican Costa Ric,* Argentina Spain United States I Republic Brazil I 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 -0.5j Log GDP per capita (constant 1995, U.S. dollars) Average 1990-2000 Source: Authors' estimates. 203 INEQUALITY IN LATIN AMERICA these are not elastic taxes, they could play a much larger role Why Infrastructure mafters and yield benefits in terms of efficiency, equity, and local gov- Infrastructure can have both direct and indirect effects on well- ernment functioning. Technical issues related to this shift being. Access to and use of electricity, roads, telephony, clean exist, but are not insuperable. In the context of rising decen- water, and improved sanitation has intrinsic value: in many tralization, the current emphasis on social inclusion, and the societies, access to at least some of these categories of basic need for more accountable local government, the time has infrastructure is considered a right that should be extended come to have a major push for a substantially greater use of to all citizens. There are also important indirect effects on these instruments in both rural and urban areas. A new tax- incomes, via the influence of roads on access to markets and related strategy could be politically linked to local reform jobs, and of electricity and telephony on the productivity of and development, whether through agrarian change or urban small and large enterprises. Finally, there are indirect influ- land development. ences on health, education, and security, for example through access to schools and clinics. As discussed in chapter 2 and 7.3. Infrastructure services and the illustrated in table 7.8, access to infrastructure remains distributional impact of privatization unequal, especially in poorer countries such as Bolivia and This section turns to the relationship between infrastructure Honduras. In addition, even where access is reasonably equal provision and inequalities. Infrastructure services can have there are large differences in consumption that are correlated large effects on the level and distribution of well-being. Yet with income. Often differences in access are magnified by dif- infrastructure provision is also embedded in the political ferences in quality, with lower quality in peri-urban and rural economy of Latin American societies. The amount, efficiency, areas. Infrastructure expansion has the potential to equalize and equity of infrastructure provision has been compromised income and other dimensions of well-being if it increasingly in particular by clientelistic structures and the persistently integrates poorer groups into regional and national systems low tax effort typical in much of the region. Since the 1980s, of production, commerce, and service provision. there have been major attempts to reform infrastructure pro- Evidence of the influence of infrastructure on incomes is vision, both through privatization and reforms within the provided in recent country level analyses (Calder6n and public sector. Some of these have brought substantial ben- Chong 2004, Calder6n and Serven 2003). This work explores efits for both efficiency and equity. However, these changes the cross-country relationship between infrastructure, are incomplete, with problems remaining in regard to the growth, and inequality by using an international data set process and regulation of privatization (which partly of both developing and developed countries. It is not explains its extreme unpopularity) and, in most countries, surprising that infrastructure is correlated with growth: a strong need for both more spending and deepening of higher-income countries can afford the investment in infra- reforms within the public sector. structure. What is of particular interest in these analyses is TABLE 7.8 Access to selected services for top and bottom quintiles, selected countries (percent) Water Sewerage Electricity Telephones Country Top Bottom Top Bottom Top Bottom Top Bottom (ranked by income) quintile quintile quintile quintile quintile quintile quintile quintile Uruguay (1989) 98 80 82 28 100 89 n/a n/a Brazil (1992) n/a n/a 87 17 99 60 61a 2a Mexico (1992) 96 63 91 28 99 75 54 3 Colombia (1996) 96 60 92 35 98 81 71 12 Bolivia (1996) 86 26 49 4 93 21 59b lb Honduras (1990) 89 49 55 4 80 15 n/a n/a Notes: a: 1995, b: 1999. Source: National household surveys analyzed for this report. 204 POLICIES ON ASSETS AND SERVICES their attempt to disentangle the influence of infrastructure TABLE 7.9 on both growth and inequality. The analyses use data on the Eaterpuse profitability and access to infrastructure in stock of telecommunications, electricity, and land trans- Guatemala (net income of owner in quetzales per wouker hour) portation (roads and railways) and examine the relationship between expansions in indices of infrastructure stocks and Basic service subsequent changes in growth and inequality. (Technically, Coverage Yes No this is done by using panel data methods such as the Gen- Electricity 8.0 3.4 eralized Method of Moments to minimize endogeneity prob- Piped water 7.9 4.7 lems, as well as to control for various other influences.) Fixed phone 15.9 4.6 These studies find that more and better quality infra- Cellular phone 20.2 5.2 structure both enhances growth and reduces income inequal- Notes: Refers only to enterprises that operate in dwelling. ity. With respect to growth effects, it is quantity, not quality, The null hypothesis of equality of profits between enterprises that is significant in the econometric analysis. After con- covered and not covered is rejected in all cases at either 99 or trolling for a range of other influences on growth, Calder6n 95 percent levels of significance. an3find that about a quarter of the variation Source: World Bank calculations using the ENCOVI 2000 survey and Serven (2003) find that about a qurter of the vanatlon (Instituto Nacional de Estadistica, Guatemala), World Bank in Latin American growth in 1996-2000 relative to 2002b. 1981-85 can be explained by differences in infrastructure stocks. With respect to income inequality, both the quan- with disproportionate gains accruing to poorer groups tity and quality of infrastructure has a significant negative (World Bank, forthcoming). In an analysis of the effect of influence: more and higher quality infrastructure is associ- rural road rehabilitation in Peru, Escobal and Ponce (2002) ated with lower inequality, after controlling for other influ- find a significant impact of road quality on income, espe- ences (such as initial income, education, financial assets, and cially in terms of wage employment. (This study seeks to health). Calder6n and Chong (2004) estimate that an increase control for reverse causation, that is from incomes to rural in infrastructure equivalent to moving up one quartile of the roads, by matching households in areas close to rehabilitated global distribution of infrastructure stocks is linked with a roads with those in other areas that did not benefit from the reduction in the Gini coefficient of 2.2 points in a five-year intervention.) period and 12 points in the subsequent 35 years. Such a This selective review of macroeconomic and microeco- long-run effect is equivalent to a shift from the level of nomic evidence supports the premise that infrastructure inequality prevailing in Brazil to that in Costa Rica. matters with regard to both the level and distribution of Even the most careful cross-country analysis has to be well-being. Other work, provides evidence on additional treated with some caution given the multiple, interacting dimensions, notably education and health (see Fay and others influences at play. It is therefore important to complement 2003). This finding complements the role of other assets this type of evidence with micro analyses. When this is done, (such as education and land), as illustrated, for example, by there is evidence of the positive influence of infrastructure. the discussion of farming in Mexico in box 7.3. The rest of For example, in Guatemala (which has low levels of infra- this section turns to issues in the provision of infrastructure. structure provision), infrastructure is strongly correlated with the profitability of microenterprises (table 7.9). Latin American problems In infrastructure Econometric analysis that controls for other influences on delivery: Flnance and institutions profitability and seeks to deal with the endogeneity of infra- The 1980s debt crisis resulted in the culmination of long- structure2' further supports the view that access to electricity, standing problems in infrastructure provision in Latin water, and telephony have large, significant influences on America. Although conditions varied across the region, the incomes: expansion of these services to poor, underserved primary symptoms of insufficient infrastructure were low populations can lead to equalizing income increases. Simi- levels and unequal provision of infrastructure services, dismal larly, in El Salvador, panel data on incomes in rural areas quality, and lack of financing. This malaise can be seen as finds an association between expanded infrastructure and a product of deeper institutional conditions, as discussed proximate influences on incomes (such as time to market), in chapter 5. Throughout the region, infrastructure was 205 INEQUALITY IN LATIN AMERICA traditionally provided by public utilities or government of the region in the middle of the 20'h century, nationaliza- departments. It is therefore useful to place infrastructure con- tion formed part of the broader process of incorporation of cerns in the context of differing intellectual perspectives on various domestic groups into the state (see chapter 5). the role of public or private enterprise, and to distinguish Nationalization generally involved the expropriation of for- three views (La Porta and L6pez-de-Silanes 1999): eign actors, while beneficiaries included domestic businesses (as both suppliers and consumers of goods and services at * The "social welfare" view, in which public enterprises low prices), formal workers, and urban consumers (Murillo can pursue goals set by governments that reflect soci- 2002). This resulted in specific dynamics of behavior on the ety's broader interests (including those related to part of these enterprises for clientelistic purposes (Foster equity), as opposed to the "narrower" profit motive of 2002, World Bank 2003a). private firms. Politicians maintained control over the public sector by * The "agency" view, in which economic incentives are appointing and dismissing managers and providing subsi- crucial to behavior (especially that of managers) and dies to support unsustainable enterprises. In return for this public provisioning raises special challenges in pro- patronage, utilities were obliged to do political favors in the viding sufficiently "high-powered" incentives. form of providing jobs, keeping tariffs down, and allocat- * The "political economy" view, in which the owners and ing new investment and public works contracts on the basis managers of both public and private enterprises are of political criteria. This pattern undermined incentives embedded in political structures and reflect the groups for efficient performance and, in turn, resulted in high costs, and distributional alliances that they are a part of. excess employment, and a lack of internally generated resources to finance badly needed expansions in service cov- The first perspective, on social welfare, formed the overt erage. With the fiscal crises of the 1980s, money from the ideological basis for the nationalization of public utilities national budget also dried up. Customers had little if any and, in many countries, large parts of the tradable industrial power to hold providers directly accountable for services and sector. Sometimes this process has been more closely linked national elections were much too broad a mechanism through to nationalist ideologies (as in Mexico) or "populist" projects which to express discontent with specific utilities. with vaguely socialist or national socialist leanings (as in the Most countries responded to this crisis in service provi- case of Peronism in Argentina). However, the principle in sion with significant institutional reforms. Privatization all its various forms is essentially that public ownership could dominated the discourse of the late 1980s and 1990s, even better protect the national patrimony, support the expansion though it formed only one element of the reforms pursued. of employment, and provide goods and services for all. These also included decentralization of water provision By the 1970s and early 1980s, public enterprise sectors (mainly to municipal levels) and attempts to corporatize throughout Latin America were increasingly tattered, char- national utilities that remained in the public sector. acterized by low productivity, bloated payrolls, and the rising However, privatization is of particular interest in light of its drain on government budgets. There were also high levels centrality in debates over the relationship between reforms of inequality in supply in the utility sector (which still holds and inequality. For advocates, privatization seemed to offer to some degree, as discussed in chapter 2), with significant a straightforward solution to the twin problems of a lack of segments of the population at the bottom of the income dis- finance for infrastructure expansion and severe inefficiencies: triburion rationed out of the public provision of electricity, using the institutional transition to facilitate a genuine break water, sanitation, and telephony. The middle and upper with the past. Fiscal proceeds from sell-offs only added to classes benefited from low prices but suffered from low ser- the attraction. Although distribution was emphasized less, vice quality, especially with respect to obtaining new con- there were hopes that solving the financing problem would nections. These conditions best fit a combination of the both allow expansion to underserved populations who had agency and political economy views. been rationed out under old models, and yield benefits by The public provision of infrastructure was typically releasing resources for social spending. embedded within the broader clientelistic patterns of rela- Privatization was pursued across the region under diverse tionships that long defined state-society interactions. In most political regimes. These included, in the words of Murillo 206 POLICIES ON ASSETS AND SERVICES (2002, p. 462), "not only right-wing regimes such as of the 1980s to 2 percent in the second half of the 1990s. Pinochet's in Chile, Chamorro's in Nicaragua, and Calder6n This shift helped release resources for the expansion in social Sol's in El Salvador, but also ... old populist parties like spending (see chapter 9), but at the cost of a further reduc- the MNR in Bolivia, the Peronists in Argentina, and the PRI tion in infrastructure provision relative to comparators. In in Mexico [which] led the push to privatize . . . even though relation to the East Asian "tigers,"22 the infrastructure gap they had previously been the champions of nationalization." rose substantially. For example, with regard to electric power Executive power and technocratic designs undoubtedly generating capacity, the median infrastructure stock in East played important roles in reform efforts. However, a richer Asia grew from 90 percent of Latin America's in 1980 to account of policy design reveals preferences for change as 230 percent in 2000; with regard to paved roads, the ratio products both of political constituencies and of the updat- rose from 780 percent to 1560 percent (Calder6n and Serven ing of beliefs in light of new information (Murillo 2002). 2003). Moreover, it is a mistake to see the post-reform world as one Second, expansions in service access did occur due to pri- of diffuse benefits, free of the rent-seeking and unequal vate and public investment (figure 7.9). Impacts of this influence that were features of the pre-reform, clientelist, investment on inequalities in access varied depending on ini- protectionist model. tial conditions. Where the middle class and elites were Under privatization, new rents and gains were afforded already covered, the expansion that did occur typically went to particular groups-notably domestic conglomerates and to the poor, as exemplified by water and electricity service wealthy individuals-that politicians could make use of to in Argentina. However, where access was initially low, the support old or create new constituencies. Beliefs, fiscal crises, beneficiaries of expansion were typically the nonpoor, espe- external pressures, and a common culture among decision- cially those in the middle of the distribution. This pattern making elites in the region facilitated convergence. However, is consistent with the general dynamics of service expansion, the way in which beliefs were updated was influenced by the which typically function along the income distribution (see views and socialization experiences of political elites, leading chapters 2 and 9 with regard to social service provision). to varying political biases in the design of privatization. Third, privatization became highly unpopular. This Chile under Pinochet pursued an approach that was friendly became evident in the streets of Arequipa, Peru, in June to foreign capital, distrustful of state control, and had no 2002, when days of rioting against the planned privatiza- overarching regulatory agency. Mexico, which was more tion of water and electricity services occurred. The privati- nationalist and statist in tradition, placed more restrictions zation plan was viewed as a betrayal, a selling off of national on foreign participation (93 percent of privatizations went assets that would only benefit the rich and, most likely, new to domestic businesses, according to Schamis 2002) and had foreign owners. Such sentiments have echoed across the more regulatory structures. Argentina, which was less nation- region, from the dominantly negative view on electricity pri- alist and had a populist-statist tradition, was open to foreign vatization expressed in the Mexican Congress to the intense capital but sought strong, independent regulation (with feelings and violent action against the botched water mixed effects across agencies). These institutional choices privatization process in Cochabamba, Bolivia. A negative have affected the subsequent evolution of privatization. view of privatization is also reflected in the sharp fall in What were the overall patterns of change? First, there was assessments of privatization in public opinion surveys a substantial expansion of private investment, especially in conducted by the organization Latinobar6metro (see fig- telecoms and electricity and to a lesser extent in water, roads, ure 7.10). Across the countries surveyed, on average only and railways. As a result, the private sector in the region 22 percent of people thought that privatization had bene- now serves some 85 percent of domestic telephone con- fited their country, which marked a sharp decline since 1998. sumers, 50 percent of domestic electricity consumers, and Firms surveyed held a similarly negative (and declining) view 15 percent of domestic water consumers (Foster 2003). How- of privatization. ever, with the exception of telecoms, the growth of private The remainder of this section first looks at the distribu- investment was insufficient to offset a drastic decline in tional effects of privatization and then at the policy issues public investment (figure 7.8), such that overall investment related to equitable expansion of infrastructure in both the declined sharply from 3.5 percent of GDP in the first half public and private sector. 207 INEQUALITY IN LATIN AMERICA FIGURE 7.8 Public and private invesbnent in seleted infrastructure sectors in Latin America, 1980-98 Percent of GDP 4.5 - Total --Public - Private 4.0- 3.5- 3.0 - ___ \E 2.5- 2.0- 1.5- 1.0 0.5 - 0.0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Note: GDP-weighted average for Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, and RB de Venezuela. Source: Authors' calculations using World Bank infrastructure investment database; see also Calder6n and Serven (2003). The effects of infrastructure privatization similar and are illustrated here by a cross-country study of A range of work has examined the impact of the privatiza- Argentina, Bolivia, Mexico, and Nicaragua that applied tion of utilities and other companies on profitability, effi- common techniques to assessing the impacts of privatiza- ciency, a variety of distributional effects, and financing. tion on inequality in the electricity, telecommunications, and Most analyses find significant gains in profitability and water sectors (McKenzie and Mookherjee 2003). All analy- efficiency following privatization. This can be illustrated ses come with an empirical warning: it is difficult to with partial measures of productivity; for example, sales undertake a clean comparison of the experience of privati- per worker in Chile, Mexico, and Peru, rose by 88, 92, and zation with that of continued public ownership under 112 percent, respectively. In Argentina, output per worker changed economic or managerial conditions, since there are rose by 46 percent, while in Brazil the ratio of costs to sales never precisely similar companies or conditions to facilitate implied a 45 percent increase in efficiency.2" More compre- such an evaluation. Thus, most analyses are of a before- hensive estimates of efficiency changes in Argentina-which and-after character with attempts to obtain approximate account for joint effects on improvements across inputs and comparisons when feasible. outputs-also indicate annual efficiency gains ranging from 1 percent for electricity distribution to 6 percent for water Consumers (Estache 2002). How were these gains distributed among Impacts of privatization on consumers are due to effects in consumers, workers, firms, and the government? three areas: increased access, prices, and quality. Such effects With respect to economic analyses of privatization and tend to reach consumers who are on average poorer than those inequality, there have been three recent research projects with who already have access; as noted above, the extent of poverty a Latin American orientation.24 The qualitative results are orientation depends on the initial distribution. 208 POLICIES ON ASSETS AND SERVICES FIGURE 7.9 Increases in access to infrastructure, by income decile in Argentina and Nicaragua a. Argentina: water and electricity Percentage of households with access 100- 90- 80- 70- 60- 50 - 40- 30- 20- 10 -4- 1985-86 1996-97 1 2 3 4 5 6 7 8 9 10 Deciles (household expenditure per capita) b. Argentina: telephone Percentage of households with access 100- 90 80- 70- 60- 50- 40- 30- 20 - 10- 1985-861996-97 0-- 1 2 3 4 5 6 7 8 9 10 Deciles (household expenditure per capita) c. Nicaragua: electricity Percentage of households with access 100- 90 - 80- 70- 60- 50- 40 - 30- 20- 10- -- + _1993 1 2 3 4 5 6 7 8 9 10 Deciles (household expenditure per capita) Source: Authors' estimates from household survey information. 209 INEQUALITY IN LATIN AMERICA FIGURE 7 10 Views on privatization in Latin America at different points in time Latin America :______._|___2003 Ei 2002 m 1999-2000 - 1998 Costa Rica Panama Argentina El Salvador . Guatemala Uruguay Bolivia Ecuadori a u~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Nicaragua _ _ _ _ _ Peru Paraguay Colombia U Honduras RB de Venezuela Brazil = = S 0 10 20 30 40 50 60 70 80 90 X The proportion of respondents who said they thought privatization had benefited their country Source: Latinobar6metro, 2003. Critics often argue that privatization leads to higher variety of effects were noted, with prices falling in five out prices, which in turn finance higher profits. In fact, price of the ten cases studied and rising in the other five changes vary greatly depending on initial conditions and pat- (table 7. 10). Even within Bolivia, there were sharp contrasts terns of change, especially with regard to the balance between between real price declines in water in the La Paz/El Alto gains in efficiency, the need to compensate for historically region and large increases related to the infamous subsidized prices, and the regulatory and institutional frame- Cochabamba water privatization project. Also noteworthy work that determines profits. A case can also typically be is the large increase in prices for telephony in Mexico, which made for tariff rebalancing so that tariffs reflect costs (for may have been associated with the granting of a monopoly example, between local and long distance charges for tele- for several years to the group that won the privatization con- phony, across sectors for water and electricity, and between tract. Finally, with respect to quality, the available evidence connection and consumption charges in all cases). In the four- is of significant improvements following privatization, in country study (McKenzie and Mookherjee 2003), a wide terms of indicators such as reliability and waiting times for 210 POLICIES ON ASSETS AND SERVICES TABLE 7.10 Price changes after privatization Argentina Bolivia Mexico Nicaragua Before After Before After Before After Before After Telephones 100 84 100 92 100 148 Electricity 100 67 100 126 100 124 Water 100 84 100 109 La Paz/El Alto 100 90 Cochabamba 100 143 Note: Table shows real price indices, relative to the Consumer Price Index, with preprivatization prices set at 100. Source: McKenzie and Mookherjee 2003. connection. (Again, privatization may not have been neces- Workers sary to secure such improvements, as the experience of the Extensive job layoffs in privatized enterprises are probably Peruvian public water company in Lima illustrates.) the most publicized and politicized aspect of apparent cost, An analysis of the overall impact of consumer effects on a pattern that follows naturally from the analysis of the poverty and inequality suggests generally small changes, public company as a clientelist provider of jobs. Large layoffs largely because households devote only small shares of their of regular workers have indeed occurred in Latin America. budget to infrastructure services.25 In all cases, the effects A cross-country study by the International Development on inequality were minor. In all but one, effects on poverty Bank (Chong and L6pez-de-Silanes 2003) estimates total were positive but not large. For example, in Argentina access losses in employment in privatized companies from about and the price effects of electricity and telephony were asso- 23-25 percent in Argentina, Chile, and Peru to around ciated with minor effects on income inequality and a 1-1.5 40 percent in Colombia and more than 50 percent in Mexico. percent reduction in poverty. In Bolivia, privatization and Firm-specific effects can be larger. Ennis and Pinto (forth- water in the La Paz/El Alto region was also associated with coming) estimate losses of 75 percent in privatized utilities little change in inequality and a 1-1.5 percent reduction in in Argentina (that is, electricity, natural gas, water, telecom- poverty. In contrast, the large price increases following the munications, airlines, and railways), and indicate that if those Cochabamba water privatization project did increase poverty workers who were laid off had received zero income, poverty by some 2 percent. In Nicaragua, large increases in the would have risen from 29.5 to 32.0 percent. welfare of those who gained access to electricity (in the order However, the net effects or layoffs depend on the labor of 12-16 percent in the bottom three deciles of income market and macroeconomic situation. All studies find that distribution) were roughly offset by welfare declines, which net employment effects are less adverse than gross losses, and were in turn due to price increases for those who already had are sometimes positive. In the case of Mexico, L6pez-Calva access. and Rosell6n (2002) used panel surveys to follow the expe- A significant study by Galiani, Gertler, and Schargrod- riences of individual workers. They found that 45-50 per- sky (2002) on Argentina estimates the impact of water cent of laid-off workers found jobs in the same sector within privatization on child mortality, a key nonmonetary measure a year, without any loss of social security or health benefits. of welfare, by exploiting the variation over time and space Only 0.5 percent of workers were unemployed for a full year. between municipalities with and without privatized water. This pattern of generally positive gains is common for the Child mortality fell an additional 5-7 percent in areas with medium term. In the Peruvian telecommunications sector, privatized water services and an estimated 24 percent in the employment rose from 13,000 to 34,000 jobs between 1993 poorest municipalities, with the decline in mortality asso- and 1998. About 145,000 new jobs were created in the same ciated with reductions in deaths from water-borne infectious sector in Brazil during that period. Even in Argentina, which and parasitic diseases. had large losses due to a slack labor market, the share of 211 INEQUALITY IN LATIN AMERICA employment in privatized utilities had recovered to almost year, yet this had no fiscal implications for or effects on social its pre-privatization levels (around 7 percent) by 1996, sug- spending. Given the poor standing of Nicaragua in terms gesting that the long-term trajectory of employment was of corruption, it is highly likely that the money that disap- similar to that for overall employment. peared went to a few rich individuals or groups. The more Effects on wages have been varied. In Argentina, economy- recent privatization of electricity raised 5 percent of wide wages rose following privatization, although the impact GDP, 80 percent of which actually accrued to government of privatization itself may have been mildly depressing reserves and supported future, if not present, public spend- because public sector workers commanded a 10 percent ing (Freije and Rivas 2002). premium. However, the impact is unlikely to have been sig- nificant, and would have been offset by the increased hours Domestic business sector effects of the workers who shifted to the private sector (Ennis and Did privatization lead to excess profits for domestic and for- Pinto forthcoming). In Mexico, La Porta and L6pez-de-Silanes eign capitalists? In most cases profits did go up, but often (1999) found that wages increased in the privatized sectors; from a starting point of unsustainable, loss-making positions. the pace of the increase was fastest for blue-collar workers, In addition, a proximate source of these gains was the large implying that part of the large efficiency increases were passed layoff of surplus workers. However, as just discussed, lay- on to workers. One of the countries in which wage effects offs were typically associated with net expansions in may have been significant and negative is Nicaragua, where employment at the sectoral level. Reasonable profits are a reallocation involved some 5-7 percent of the workforce and necessary part of private production, and the search for prof- the public sector wage premium was significant (20 percent its is key to the realization of productivity gains under the in urban and 29 percent in rural areas in 1993). 'agency" interpretation of enterprise performance. The more relevant question is whether profits were exces- Fiscal effects sive. A study of all privatizations in Mexico found that prof- Fiscal gains were one of the drivers of privatization. There its in the noncompetitive sector were not, on average, were often large fiscal gains from sales, followed by significantly above the competitive sector (La Porta and increased tax receipts from profitable private companies. Cap- L6pez-de-Silanes 1999). A more recent study focusing on util- ital receipts ran as high as US$25 billion in Argentina and ities finds suggestive evidence of higher profits in noncom- US$33 billion in Mexico. The distributional effects of these petitive than competitive sectors following privatization in receipts depend on how they were applied by governments, Chile, Mexico, and Peru, but not in Colombia (Chong and most of which used part of the gains to retire debt. In addi- L6pez-de-Silanes 2003). However, the increase in profits fol- tion to positive stabilization effects, doing so allowed for lowing privatization was not larger in noncompetitive than reductions in debt service and created fiscal room for the in competitive sectors. To the extent excess profits existed, expansion in social spending that characterized most they likely reflected regulatory failures, preferences of gov- economies in the 1990s. As discussed in chapter 9, this ernments for higher profits to support their tax base, or expansion was generally mildly or significantly equalizing. specific features of the design of privatization. In some cases, However, it is noteworthy that some governments ended the explicit or implicit profit guarantees were made that were put decade with higher indebtedness despite gaining large cap- in place to attract investors or to maximize fiscal benefits from ital receipts from privatization, as was markedly the case in higher sales prices. Typical cases involved the Mexican and Argentina and Brazil. This pattern is consistent with the use Peruvian telecommunications sectors, where a specified period of fiscal expansions by governments to postpone distribu- of monopoly was guaranteed. As noted earlier, prices tional struggles to the future (see chapter 8), which makes increased in the former case. In 2002, the charge structure the challenge of achieving growth with greater equity all the pursued by Peruvian telecoms became a source of public con- more difficult in the present. troversy since it was higher than Latin American standards, Finally, privatizations have often been undertaken with even though this had been legally agreed to as part of the orig- less than full transparency, which was dramatically the case inal sale. Finally, much less is known about illegal gains, in Nicaragua. During the period of active privatization in although accusations are common (as previously noted with the 1990s, receipts were running at 2.5 percent of GDP every regard to the missing millions of dollars in Nicaragua). 212 POLICIES ON ASSETS AND SERVICES Who gained control of the privatized assets? There is Consumers also now expect decent service from their utili- some evidence of the consolidation of gains by domestic ties, and no longer use the dismal service levels of the 1980s conglomerates, along with a relatively small number of inter- as a benchmark (Estache 2003). national companies. Schamis (2002) has documented how domestic conglomerates and large-firm business elites in Policies for equitable Infrastructure expansion Argentina, Chile, and Mexico were major protagonists and Privatization alone did not solve the financing and insti- beneficiaries of the privatization process. In the case of Chile tutional malaise that afflicted the infrastructure sector in under Pinochet, this situation was associated with a revolv- the 1980s. In light of the importance of infrastructure to ing door of individuals moving among major conglomerates, both growth and equity, there is a need for both policies the state, and newly privatized companies. In Mexico, such to restore public sector provisioning and a strengthened conditions were part of a long tradition of close forward and policy and regulatory framework for private provisioning. backward linkages between business and the state (see Haber A case can also be made for specific subsidy policies on 2002b), but also marked a change in the patterns of alliance. distributional grounds, for both the public and private While the historical influence of business associations rep- sectors. resenting medium-sized firms (which were judged to be backward and protectionist) was reduced, connections with Restoring public spending and extending public large-scale conglomerates increased (Schamis 2002). These sector reforms processes involved new forms of distributional alliances in The case for restoring public spending on public infra- pursuit of rents. structure is vividly illustrated by figure 7.8, as well as the previous analysis of Latin America's widening infrastructure An overall assessment deficit with respect to other regions. Measures to close the The preceding analysis of experiences in various countries gap are good for both growth and inequality. Some sectors- suggests that utility privatization brought substantial ben- sanitation and rural water and roads, for example-are likely efits, but that room for criticism remains. There is wide- to remain in the public sector in the long run. Moreover, spread evidence that greater efficiency has resulted, although the private sector may also need a stronger public presence higher profits and productivity were often achieved in part if demand is insufficient. However, achieving this will by laying off excess workers. There is also evidence that many require both facing issues of finance and tackling persistent privatizations led to increased access by poorer groups, that institutional challenges in public provisioning. price increases declined as often as they rose, and that net The financing issue is intimately connected to Latin employment effects were often positive within a sector. More- America's relatively low tax effort and relatively high debt over, severance pay at least partly shielded the blow of worker burden. As discussed in chapter 9, raising taxes in ways that layoffs, with aggregate effects on income inequality typically minimize both efficiency costs and regressive effects is cen- being very small. tral to a more equitable development path, provided that Poverty impacts are almost always beneficial and can be resources are used effectively. User charges that reflect costs significant when large increases in access occur, with appar- can make an essential contribution, although this has to be ently few exceptions (Cochabamba in Bolivia being one). On balanced with distributional objectives (see below). the other hand, there is also evidence that privatization has Effective use of resources in the infrastructure sector been associated with increased power on the part of con- depends equally on institutional problems. Resolving glomerates and their foreign partners, as well as with agency-related hurdles and avoiding the clientelistic use of relatively high profits in noncompetitive sectors. Finally, utilities-two factors that were important motivations for accusations of favors being given or corruption taking place privatization-remain major challenges for the public during the privatization process have often been made. Such sector. Obras (public works) are still a classic tool of patron- concentrated gains by a few actors, whether made legally or age. Institutional reforms can bring gains in terms of effi- illegally, combined with losses by workers, help to explain ciency and equity under public auspices, especially if actions the apparent paradox that privatization is deeply unpopu- are taken to separate policymaker from provider under a lar even though it has often yielded significant benefits. strong agreement, increase transparency, and strengthen the 213 INEQUALITY IN LATIN AMERICA direct voice of all consumers, especially those who have pre- teristics, but can also be affected by design. It is now rec- viously been excluded. ognized that the process of electricity privatization in Chile Indirect evidence for this process is provided again by the (as was the case in Great Britain) created firms with Bolivian case, in which the privatized La Paz/El Alto water substantial monopoly power, and thereby missed an oppor- utility was compared with the cooperative water utility in tunity to create a more competitive market. It is a fact that Santa Cruz. It performed better, but the gains were signif- once a system is privatized, it can be harder to change the icantly smaller than those seen in a before-and-after com- structure. This not only increases economic (and perhaps parison. Another example from the same sector is the Lima political) influence, but can make the task of any regulator water company known as SEDAPAL. Even though the com- more difficult. pany has so far remained under public ownership, it has Complementary to design specifics are questions of trans- undertaken a series of reforms (including with regard to cor- parency and participation, in both the privatization process poratization and outreach) and has achieved substantial (in order to support measures to minimize corruption) and improvements in both efficiency and access. On the other in production. Such questions can involve participatory hand, a recent review of experience in the regulation of engagement by groups with weak influence. For example, municipalized water companies concluded that many adhere the case of water privatization through Aguas Argentinas to a clientelistic model, although at local levels (Foster 2003). involved a redesign of the approach to access after complaints That review also argues for the need to improve and expand were heard on the local level (Estache, Foster, and Wodon civil society, strengthen regulation (as well as civil society's 2002). Factors such as openness, the presence of watchdogs, knowledge of regulatory decisions and processes), and align and independent analysis as a means of increasing public central finance with the realization of regulatory goals. information and debate are also important. Participation can reduce incentives for collusion between providers and policy- Strengthening privatization processes and regulating makers (potentially in both public and private models). In private production France, the indirect expression of voice-via the potential While assessment of the effects of privatization on efficiency to vote out mayors who fail to design effective agreements and equity provided here is substantially more positive than with local water companies-provides an example. Pub- that provided by opinion surveys in Latin America, there are lishing and explaining regulatory decisions, establishing nonetheless key areas where improvements can be made. appeals processes, and involving consumers can help increase Privatization in principle facilitates the separation between the independence of regulators, despite the fact that policymaker and provider, but remains susceptible to infor- consumer involvement has generally been low in Latin mal and "revolving door" ties and to excess power on the America (World Bank 2003a). part of the provider. Independent regulators have to be part of the solution, at least in noncompetitive sectors, but they User charges and subsidies largely do not yet perform the central role of protecting the User charges that cover costs are essential to the financial public interest, especially when governments and operators viability of utilities in both the public and private sectors. have a common interest in high profits and associated high However, the extent to which specific charges reflect costs taxes (Estache 2003). However, classic problems of regula- needs to be balanced against the distributional objectives of tory capture or relative weakness remain, especially in the assuring basic levels of supply to all groups. Should such dis- context of large private firms with highly qualified, well-paid tributional goals be handled by the general tax and benefit staff. Ensuring the legal, budgetary, and staffing indepen- system? There are in fact good arguments for explicitly dence of regulators can help. In Argentina, for example, a including equity considerations within the pricing policies stronger legal framework for electricity and gas regulators of utilities: most societies view at least some infrastructure was associated with better regulatory performance than that services as basic needs, and so are concerned with actual con- seen in other sectors (Estache 2002, Foster 2003). sumption of utility services; general welfare systems are far Also important is the process of privatization itself. The from perfect; and paying attention to social concerns may post-privatization market structure is influenced by the be essential to the social acceptance of private production extent to which an activity has natural monopoly charac- (G6mez-Lobo and Contreras 2000). 214 POLICIES ON ASSETS AND SERVICES As noted above, during the 1980s, many utilities had 7.4. Conclusions pricing policies that were bad on both efficiency and equity This chapter has focused on assets in three major areas: edu- grounds. Prices did not cover costs, but subsidies rarely cation; property rights, land, and housing; and infrastruc- reached the poor, who typically did not have access to ser- ture. Discussions on each area have emphasized the intimate vices. Regressive subsidy policies have persisted in some links with institutions (in the broad sense in which the word cases, especially in some public utilities. In Mexico, elec- is used in this report). These three areas are undoubtedly sig- tricity subsidies are large (substantially greater than the flag- nificant in any quest for greater equity, and there are strong ship poverty-focused Oportunidades program) and regressive, complementarities between them. However, this is not a if somewhat less so than in the past. comprehensive list of aspects that should be explored; other Subsidies to ensure access to and basic consumption levels areas related to asset formation are also important, and health by the poor are desirable, but pose challenges with regard is one important omission here. Although this chapter could to targeting that are common to the social sectors (see chap- only sketch issues and options in these three central areas, ter 9). Although discussion of design details is beyond the it has nonetheless supported the view that a major, serious, scope of this chapter, some general points can be made here. redistributive effort to promote asset distribution is indeed It is generally desirable to have low connection fees in feasible, and substantial scope exists for asset creation strate- order to encourage uptake, as well as subsidies to encourage gies that are good for both growth and equity. Making sure basic levels of consumption. If the fiscal position of a country this happens will both require and support critical, com- allows, it is also desirable to include subsidies in the public plementary changes related to equalizing political action and budget, so that utilities don't have to cross-subsidize by social agency, and potentially help shift the cycles of inequal- "taxing" other consumers (as is the case in Chile, for exam- ity from vicious to virtuous. ple). However, doing so will not always be feasible, especially in cases in which the fiscal position is tight. In assessing the Notes pre-crisis performance of privatized utilities in Argentina, 1. Paulo Freire (1970) is a classic example. Estache (2002) argues that efficiency gains were insufficiently 2. Some biases are inherent in these patterns, since poorer, less edu- passed on to consumers, and that cross-subsidies for poorer cared people on average lead shorter lives. groups are likely to be necessary in any practical solution. 3. See table 4.2 in de Ferranti and others (2003). These interre- The Colombian water utilities finance subsidies through gional comparisons are made on the basis of the Barro and Lee (2001) higher tariffs for richer consumers, for example. data set. All numbers presented for intraregional comparisons are taken It is therefore clear that choices exist between subsidiz- from the microdata analyzed for this report; note that there are dis- ing categories of consumption (as in the case of lifeline tariffs, crepancies between these sources. that is, basic tariffs at low rates) or categories of consumers, 4. These are the marginal returns on finishing an educational level which would include an assessment of the capacity of dif- obtained from Mincer equations, as indicated in table A.31 in the ferent consumers to pay for services. For example, Chile statistical appendix. makes use of a national proxy-means test (known as theficha 5. For a brief discussion of the methodology used, see chapter 6; CAS) to determine water subsidies, reaping substantial sav- for details, see Bourguignon, Ferreira, and Leire (2002). ings from spreading the administrative burden of means 6. Recent studies on Latin American schooling include Glewwe assessment across many programs. Colombia has a geo- and others (1995), King and Ozler (1998), Paes de Barros and graphic targeting system, based on the classification of a Mendonca (1998), Jimenez and Sawada (1998), and Eskeland and geographic area within a municipality into one of several Filmer (2002). See also Wossmann (2000) for a cross-country study. socioeconomic strata. Despite this apparently less precise 7. The "Washington Consensus" refers to the set of market- method, the Chilean and Colombian schemes have similar oriented policies that were commonly adopted in Latin America in the targeting properties: both are progressive, but both make sig- late 1980s and 1990s. See chapter 8 for a discussion. nificant errors of exclusion and inclusion (G6mez-Lobo and 8. A recent analysis of Colombia makes an adjustment for land Contreras 2000). Subsidy design should therefore be shaped characteristics (see World Bank 2003b). However, this leads to only around the characteristics of a country and a particular ser- a modest reduction in the Gini coefficient for land distribution from vice, and be subject to regular monitoring and evaluation. 0.93 to 0.85. 215 INEQUALITY IN LATIN AMERICA 9. There is a large body of literature on this point; (see chapter 1 18. This subsection is based on inputs from Marianne Fay and Anna of World Bank 2003a) for a review. Wellenstein. 10. This synthesis is based on comments from Jose Marfa 19. In Uruguay, for example, 6.8 percent of the households in the Caballero. For a review, see Chavas (2001). bottom quintile of the population merged with other households in 11. For one of many overall syntheses, (see World Bank 1993). For order to reduce housing expenses. an approach that emphasizes the centrality of political support of the 20. This includes spending by the federal government and peasantry, see Teranishi (1997). implicit finance subsidies provided by the quasi-public pension fund 12. See Klein (1992) for a general history and Gray-Molina (2002) and housing lenders INFONAVIT and FOVISSSTE, which are funded on children of the 1953 revolution. through loan repayments and compulsory savings (5 percent of pay- 13. See Seligmann (1995) for an ethnographic account of agrarian rolls) from private and public sector employees, respectively. reform and its aftermath in a poor area of Cusco province in the Andes. 21. Endogeneity can arise since causation may also flow from greater McClintock (1998) discusses Sendero Luminoso in a comparative analy- profitability to the use of modern utilities. The analysis used com- sis with El Salvador. munity-level access as an instrument to deal with this issue, at least 14. Just as PRONASOL was in part an instrument for gathering partially. rural votes for President Salinas in Mexico in the 1990s, so at times 22. Hong Kong (China), Indonesia, Republic of Korea, Malaysia, was the social fund FONCODES in Peru. As with many social funds, Singapore, Taiwan (China) and Thailand. FONCODES combined genuine success in service provision and a 23. Estimates from a synthesis by the IDB (2002). See also the strong propoor orientation with the goal of maintaining political sup- extensive analysis of the privatization process for Mexico in La Porta port. See Schady (2000) and Paxton and Schady (2002) for more on and L6pez-de-Silanes (1999), which identifies large efficiency gains in this and related points. both competitive and noncompetitive sectors. 15. lt is also important to note that the foundation of major suc- 24. These were conducted under the auspices of the World Insti- cesses was partly established through government projects that were tute for Development Economics Research (WIDER), the IDB (2002), launched during the Frei and Allende administrations. This was rel- Chong and L6pez-de-Silanes (2003), and the Latin American and evant in the cases of temperate fruits, relocation of dairy from the Caribbean Economic Association (LACEA)/World Bank/IDB Network Central Valley to the south, and the development of timber for cellulose on Inequality and Poverty (summarized in McKenzie and Mookherjee production. 2003). For earlier work on this subject, see Galal and Shirley (1994) 16. See World Bank (2001a) for an example of multiple paths and Estache, Foster, and Wodon (2002). toward rural development in Brazil; figure 4 in deJanvry and Sadoulet 25. To quantify access effects, this study estimated the "virtual" (2002b) on shares of off-farm incomes in six Latin American countries; prices associated with rationing of access, by estimating the price level and Lanjouw and Ravallion (1999). that would cause zero consumption. 17. These avecindados and posesionarios occupied land within ejidos, but were not formally members and therefore lacked traditional rights of land use. 216 CHAPTER 8 Policies on Markets and Institutlons T H HIS CHAPTER DISCUSSES THE RELATIONSHIP BETWEEN MARKETS AND INEQUALITY. IN TERMS OF the conceptual framework presented in chapters 1 and 6 (see figure 1.1), such an exploration involves moving from policies that affect asset distributions (the subject of chapter 7) to policies that influence market returns to these assets. In turn, the role of one set of institu- tions, those that are directly linked to market outcomes, must be considered. The subject of this chapter is only part of the broader question of how and to what extent institutions influence inequality (the central role of political and social institutions was discussed in chapter 5). Links among all factors will be considered and referred to at various points throughout this chapter. The question of the impact of market-oriented reforms on inequality is one of the more controversial areas of debate in Latin America. The issue is often framed in black and white terms: have market-oriented (and macroeconomic) policy reforms been a curse or a cure for Latin America's ills of slow growth and high inequality? This report takes a more nuanced view. As the structure of the third, current part of the report indicates, institutions affecting markets are only one domain of economic policy that is relevant to distribu- tion (and growth, of course.) It is also important to look separately at different market-related institutions, and to recognize that the consequences of markets will depend on the structure of asset ownership and social and political structures. This chapter first looks at overall evidence on the rela- opened their economies to greater market influence. Structural tionship between structural market-based reforms and changes have included the substantial liberalization of trade inequality. It then briefly surveys issues in the labor market, and domestic financial policy, opening up of capital accounts, recognizing that this area in particular warrants a much fuller privatization of state-owned companies, and tax reforms that treatment than there is room for here. Particular focus is entail more uniform tax treatments and greater reliance on placed on the relationship among macroeconomic crises, value added taxes. The following section reviews these over- policy responses, and inequality, which is a major concern all issues (see chapter 7 for a more detailed treatment of the in Latin America. Both the review of evidence and new work privatization of infrastructure). undertaken for this report support the view that financial The initial driver for policy reforms in Latin America was crises will typically be regressive, a fact that underscores the economic crisis and the imperative to restore growth. There importance of strengthening policies to avert crises and are numerous accounts of both the reform process and its effects developing both policies and institutions that ensure more on growth. (For a recent example, see Loayza, Fajnzylber, and equitable resolutions to crises when they do occur. Calder6n 2002 and Burki and Perry 1998.) Two important conclusions can be drawn from these analyses. First, there is 8.1. Markets and inequality consensus that many reforms were indeed undertaken. To illus- Over the past two decades or so, most countries in Latin trate, figure 8.1 presents reform indices constructed by Sam America have undertaken major policy reforms that have Morley, then of the Economic Commission for Latin America 217 INEQUALITY IN LATIN AMERICA FIGURE 8 1 The extent of policy refonns in Latin America, writh country and polky variations a. Average reform index for selected countries Average reform index 0.8 0.6 0.4- A 0.2 - Argentina - Brazil * Chile | Costa Rica Uruguay -_ RB de Venezuela b. Reform indices by policy area Specific reform indices 0.9 - 0.8- 0.7- 0.6 - 0.5 0.4 - 0.3 - 0.2- -4- Commercial index -_- Financial reform index - Capital account liberalization index 0.1 - - Privatization index Tax reform index O - Note: The average reform index is the simple average of the index of reform in the five policy areas for each country; the regional index for a policy area is the simple average of all countries. Source: Morley 2001. and the Caribbean (ECLAC), that are based on earlier work ers on many dimensions, but Argentina and Costa Rica had by Eduardo Lora of the Inter-American Development Bank caught up with them by the mid-1990s. Brazil has under- (IDB). Although such indices pose difficulties,' they nonethe- taken fewer reforms on average, while the Rep6blica less provide support for the view that there has been a sig- Bolivariana de Venezuela has been a laggard. In terms of the nificant move in the direction of more liberal, market-oriented type of reform that has taken place, a steady and large wave policies throughout the region. of reform in trade liberalization has occurred since 1987, fol- Significant variation has occurred both across countries and lowed by a burst of reform in the financial sector around 1990, reform categories. Among a few countries selected as illus- and significant capital account opening in the 1990s. Tax trations, Chile and Uruguay were early and strong reform- reform has been slower. Privatization started at a relatively 218 POLICIES ON MARKETS AND INSTITUTIONS high level-since many countries did not have large public The achievement of growth has complex institutional sectors-and then underwent less change, although impor- underpinnings. There is aggregate evidence that the qual- tant and high profile privatization processes did take place ity of institutions is a powerful determinant of sustained in some countries. long-term growth, although debates continue over the Second, there is a somewhat weaker consensus that reforms relative importance and independence of "institutions" and had a positive impact on growth, but that this effect was more market-oriented policies that foster greater economic inte- muted than originally hoped for. Reforms helped, but didn't gration. Some authors argue at one extreme that "institu- lift Latin America up onto East Asian-style growth paths. tions rule over policies," while others believe that the role A recent analysis that includes reform indices as well as other of institutions is overstated in some of the recent literature structural factors such as education and governance (2003) on economic growth.2 A fair statement might be that both finds that the reform response of countries in Latin Amer- perspectives matter and that there are causal interrelations ica has been in line with results based on global experience. in both directions. However, achieving rapid growth requires far more than a The issue of concern here is the potential impact of market- specific set of market-oriented reforms. As Rodrik (2003, oriented reforms on inequality, a subject about which there p. 2) has argued in a review of growth strategies, "First-order are diverse views. Box 8.1 outlines two opposing views, each economic principles-protection of property rights, market- of which is intellectually coherent. Those reform advocates who based competition, appropriate incentives, sound money and have focused on questions of inequality anticipated strong, so on-do not map into unique policy packages. Reformers positive effects. Others predicted disequalizing consequences. have substantial room for creatively packaging these prin- The approach of this report-which has emphasized the strong ciples into institutional designs that are sensitive to local interactions between markets, institutions, and sociopolitical opportunities and constraints." (For a general discussion on factors-would suggest that theory is ambiguous, and will this topic, see Lindauer and Pritchett 2002.) often be context-specific. What does empirical work suggest? BOX 8 1 AlUbnate vws on te pteild hpact of mist-ede refornbu ad qualibt Theorecical expectations of the relationship berween reform substitutes, and capital-intensive production choices and inequaliry are diverse. Ar rhe risk of putring forth a rather than labor-intensive exports. slightly caricatured perspective, rwo of the currenEs of (3) Access to markers. Deepening markets would have sig- thought prevalent in debates in Latin America (as well as nificant equalizing effects, since most often the poor elsewhere) are outlined here. are precisely the ones with the weakest market access. In the blue corn-er. Miarket-liberalizing reforms are equal- In ihx red Lorner. Neoliberal policies are disequalizing izing tor three central reasons: because they dismantle systems of support and protection (1) Renr reduction. A panoply of state controls has only that nurture employment growth. provide institutional served to create "rents," that is. opportunities for gains support For agriculture, and facilirate widespread social based on protected privileges that were sysrematically provisioning. The chill winds of international competition appropriated by the rich and powerful These rents also hurt the weak and benefit the strong. This is especially true entailed efficiency costs. Market liberalization reduces for those companies and individuals with an initial basis rents and opportunities for corruption, allowing for economic strength in terms ofeconomic capital. skills, people to receive income that is in line with their access to financial systems, and international and domes- economic worth, not their influence. tic connections, and which are therefore best posirioned (2) Reduction of distortions Trade protection, domestic to benefit from market opporrunities. Liberalization also regulation, and financial repression maintain a sys- reduces rhe power of labor, both by making capital more tematic bias against labor, including unskilled labor, accessible and by withdrawing stare-mandared protections which encourages indusrry over agriculture, import for workers. 219 INEQUALITY IN LATIN AMERICA The big picture suggests that there Is no simple story FiGURE 8 2 Chapter 2 presented patterns of change in income inequal- | CorTelation between changes in reforn and changes in eincomie inequality ity over the past decade and reviewed the literature on evi- dence from previous decades. Some patterns emerged from Change in income Gini (early to late 1990s, percentage points) that analysis, with a number of countries showing mild dis- i 10 - tributional improvements in the 1970s, many experiencing 8 *ARG a worsening in the 1980s, and a subset of South American 6- countries showing slight worsening in the past decade. How- ever, the dominant pattern is of resilient high inequality 4 * VEN PER under a wide array of policy regimes. 2 - CHI * C C Moreover, every pattern has its exception. In the past decade, 0o- BOL JAM* ELS Brazil and Mexico both implemented major reforms and have, 2 * MEX -2 ~ ~~~~~*BRA DOM respectively, experienced a small reduction in inequality and I HON * PAR no clear trend. In the 1970s and 1980s, Chile experienced 0.1 0.2 0.3 0.4 0.5 0.6 growing income inequality. Eagerness for an overarching expla- Change in general reform index, 1980-95 nation might attribute this trend to the vigorous shift toward J Source: Authors' calculations based on Morley (2001) and market-oriented policies under the Pinochet regime, starting inequality measures reported in the Appendix. in the 1970s. However, there was a lot more going on in Chile - than market liberalization. The military junta explicitly sought cases have helped reduce inequalities in well-being and to reverse the redistributive measures of the previous govern- opportunities.) Several studies on market-oriented policy ment and destroy the power of organized labor. The variety of reforms have been conducted, especially those associated with country experiences suggests a multiplicity of distributional ECLAC (see in particular Morley 2001). Also notable are a stories across relatively homogenous countries. cross-country review in Berry (1998) and a cross-country analy- The lack of any dramatic or obvious pattern is further sis of effects on wage inequality by Behrman, Gaviria, and illustrated by looking at the simple bivariate relationships Szekely (2001). An additional important study is by Ganuza, between reform effort and changes in income inequality. In Taylor, and Morley (1998), although it focused on the effects figure 8.2, the change in the average reform index between of reforms on poverty rather than on inequality. 1980 and 1995 is compared by country, with the change in These studies generally compare changes in measures of income inequality measured by the Gini coefficient at the income inequality in relation to changes in policy, using beginning and end of the 1990s. (This is a rough form of reform indices such as those put forth by Lora (2001) (as calculation that allows for possible lags between reforms and illustrated in figure 8.1) or other similar work. Cross-country effects. A similar lack of a bivariate relationship was found analysis in particular exploits variations in the timing of between the level of reform and the change in inequality.) As different reforms to look for statistical associations with indicated, no clear pattern emerges. changes in inequality. Table 8.1 summarizes results from this approach, along with recent work undertaken at the World Attempts to disentangle effects suggest Bank (as background for this report) on a global database modest and mixed effects (see L6pez 2003). The latter was conducted within the same With the words of caution given above, attempts to disentangle overall framework as the growth study referred to above the effects of market-oriented policy reforms on the income (Loayza and others 2002). dimension of inequality can be examined. The lack of an Most studies find modest negative effects on household obvious bivariate pattern, as illustrated in figure 8.2, may hide or individual income inequality in at least some areas of more complex causal effects. (As frequently emphasized, this policy change. However, the effects are not always statisti- is only one dimension ofinequality that is ofinterest. Chapters cally robust and different studies find different effects for 2, 7, and 9 discuss some important gains in the distribution different categories of policy reform. For example, Morley of spending and access to associated services, which in many (2001) finds slightly negative effects of trade liberalization 220 POLICIES ON MARKETS AND INSTITUTIONS TABLE 8.1 tertiary level that typically outweighs increases in the rela- Effects of policy reforms on inequaity: Estimates from three tive supply of these workers. The extent of increase in this studies premium varies across countries, but has been particularly strong during various periods in Argentina, Chile, Colombia, Behrman, Birdsall, and Mexico, and present but weaker in Brazil. By contrast, Study Morley/ECLAC and Szekely L6pez there has not been a significant rise in the premium placed Variable Gini index Wage differentials Gini index on workers with a secondary education (primarily because Trade liberalization + 0 + the substantial expansion in the relative supply of workers Financial 0 + + with secondary schooling largely offset increases in demand). liberalization Capital account - + There has been debate in the economic literature as to liberalization whether these observed changes are due to the classical effects Tax reform + + of increased trade, in which those countries with relatively Privatization O 0 All 0 + abundant unskilled labor would (in simple models) have experienced rising demand for such labor and subsequent Note: + means inequality increasing; - indicates inequality increases in the wages of the unskilled.4 According to this reducing, 0 means no robust effect. account the fall in relative demand for unskilled labor in Source: Morley (2001) and Behrman and others (2001) address a Latin America; L6pez (2003) uses a global database. Latin America would have been caused by the entry into world markets of such labor-intensive giants as Bangladesh, on income distribution, while Behrman, Birdsall, and China, India, and Indonesia. Szekely (2001) find positive influences of trade liberalization In fact, economic analysis identifies a different pattern of but negative effects of financial liberalization on wage generalized biases against the unskilled and in favor of the inequality. skilled across both more and less labor-intensive industries. It appears that market-based policy reforms have been This provides support for the view that the dominant source neither a curse nor a cure for income inequality, as the oppos- of the observed changes is what economists refer to as skill- ing views presented in box 8.1 would have suggested. On biased technical change (SBTC). It should be noted that such balance, the evidence suggests mild disequalizing effects. skill-biased changes can come from many sources, including Modest and complex effects could well have been lost in a technological advances in the layman's sense (for example, sea of other variables that bring about changes in inequal- owing to information technology), as well as changes in the ity, both in terms of economic assets and the broader set of occupational and sectoral composition of work organization sociocultural and political variables that can be forces for that favor tertiary or secondary school graduates. sustaining or changing income differences. Although significant changes in labor market policies have Although general cross-country studies do not lead to firm not occurred, there is some evidence that some shifts have conclusions, one category of work is more persuasive. There sharpened trends toward rising wage differentials, owing in is an emerging consensus that economic integration is a force particular to declines in unionization and a more binding behind the transmission of skill-biased technical change in minimum wage (particularly in the case of Colombia). More the labor markets of many countries. This is based on a care- generally, it is possible that the combination of easier access ful analysis of labor market trends conducted through a to capital and periodic crises have weakened the bargaining prism of interactions between the relative supply and rela- power of labor (see section 8.3 for more on this point). tive demand of different skill categories. A number of coun- Finally, although the classical effects of trade on wage try studies and the new work conducted for last year's World inequality have not occurred, economic integration and Bank report (de Ferranti and others 2003) have applied such liberalization-including the expansion of trade, financial a supply-and-demand framework.3 Since this approach was liberalization, and Foreign Direct Investment (FDI) detailed in that report, only a brief summary of the results inflows-appear to have a mediating influence that has is provided here. spurred technological, sectoral, and work organization It is unambiguous that many countries have experienced changes, which have in turn led to rapid growth in the rising demands for workers who are educated through the relative demand for skills. 221 INEQUALITY IN LATIN AMERICA This is probably the strongest evidence to support the view illustrated in the case of China, where dynamic non- that polices that spur economic integration often lead to state activities (for example, town and village enter- increases in wage inequality. This process reflects a tradeoff prises) and private firms expanded significantly in the between growth and distribution, at least during economic 1980s and 1990s, whilst large numbers of unpro- transition. A growing demand for skills has been associated ductive jobs in state firms were preserved. with processes of innovation and economic change that are This was also the case in Japan during its high- central to Latin America getting on a faster growth path. growth era. Internationally competitive export indus- Moreover, there is some evidence that such change is once- tries grew while inefficient domestic sectors (for off. For example, both Chile and Mexico-the two countries example, agriculture and construction) were preserved that have gone furthest in terms of international economic through explicit public policies (which were in turn integration-appear to have experienced a leveling off in the facilitated by political pressures and underpinned by past few years (see discussion in de Ferranti and others 2002 substantial electoral malapportionment). However, and 2003). It should also be noted that increases in wage the Japan example also illustrates the risk of such an inequality have not been systematically associated with rising approach: it is difficult to stop protecting inefficient income inequality in terms of household-level per capita sectors. In the 1980s, when the Japanese economy incomes, as discussed in chapters 2 and 6. looked strong, observers declared victory for the Japanese model of capitalism (that is, a system with What can be done so that market-based reforms strong state intervention via industrial policy with have more equalizing effects? strategic foresight). However, once the Japanese econ- The effects of both opening up markets and economic inte- omy began to perform poorly in the 1990s, the gration are complex, and some of them may be once-off or weight of opinion shifted to favor the U.S. model, in temporary. However, the existence of likely disequalizing which the market presumably operates to clear inef- economic pressures in already highly unequal societies is an ficiency out of the system in at least some areas. The important issue. What can be done to offset such disequal- U.S. agricultural sector is an important counterex- izing effects? Potential policies can be organized into four ample that again illustrates the political difficulties categories, as follows. of ending protection of a politically influential sector. Policy change is an important area that is influ- (1) Complementary measures to provide more equal assets across enced by both economic design and political econ- households and groups. The assets that individuals, omy. However, this report has neither conducted households, and groups possess have a powerful influ- original work on this issue, nor found a clear synthesis ence over their capacity to respond to new market from existing work. Nonetheless, it can be suggested opportunities. This is illustrated by differing returns that these factors may be quite significant in some on educational levels, but also applies to other eco- cases. For example, opening up an economy at a high nomic assets such as land and infrastructure. Policies and fixed exchange rate (with significant declines in should facilitate a rapid expansion of educational the price of capital in Argentina being one result) attainment in response to increased demand for and may have contributed to large job losses among rel- premiums placed on skills, and support broader access atively unskilled workers. Sorting out these issues will to land and infrastructure (as discussed in chapter 7). clearly require much more careful work in the future. (2) Sequencing of policy changes. Economic development (3) Measures to provide safety nets or "ropes. " The process of involves a process of creating and destroying jobs, job destruction can lead to temporary or permanent firms, and economic activities. Part of the dynamic income losses, with adverse effects on various aspects that empirical trends indicate is a more rapid destruc- of both social and psychological well-being. A cen- tion than creation of activities for the unskilled. Some tral part of a well-functioning market economy is the countries have managed to create new productive provision of safety nets to catch those who are hurt, activities while simultaneously shielding old, less or of safety "ropes" to help limit their fall. To the productive ones. This is perhaps most dramatically extent that these exist in Latin America, they have 222 POLICIES ON MARKETS AND INSTITUTIONS tended to be unequal in scope; options for reform are discussed in chapter 9. FiGURE 8 3 (4) Deeper markets and a more equal investment environment. Crony bias versus democratic voice and accountabilty Markets are by no means uniform with regard to both Crony bias index their reach and effects. An important part of the High 1.0- objective of achieving more equal effects through r -0.41 * market-oriented policy reform involves the deepen- * +# 4* ing of markets to provide more equal opportunities for all potential market participants. Doing so will 0.5s- require increased attention to the economic, politi- cal, and sociocultural factors that contribute to the unequal effects of markets. Lo 00- -2.0 0.0 2.0 It should be remembered that the investment environ- Democratic voice and accountability index ment is central to the dynamics of economic change and job Low High growth (Stern, forthcoming). The determinants of an Source: Hellman and Kaufmann 2003, World Economic Forum 2003 (Executive Summary) for 80 countries. investment environment are complex, including market opportunities, access to inputs, infrastructural quality, the extent to which transactions are associated with corruption, Hellman and Kaufmann (2003) construct an index of and the role of government agents in facilitating or obstruct- "crony bias," which is "perceived by the firm as the differ- ing the wide variety of measures needed for investment and ence between the firm's characterization of the influence of production by firms. individuals or firms with close, personal ties to political lead- Important for the present focus is the fact that, although ers and the influence of its own business or trade association the investment "environment" connotes a public good, in on recently enacted laws, rules, and regulations affecting most cases conditions stray significantly from that quality, their business" (Hellman and Kaufmann 2003, p. 9). In A public good implies that its benefits would be the same extreme cases, state capture, or the use of illegal means by for everyone and the actions of some agents would not affect private groups to obtain favorable government policies, the position of others. In most Latin America societies, it is result. This phenomenon has been well documented in the more realistic to talk of a series of investment environments case of Russia and other successor states of the former Soviet for different groups of firms, which are influenced by market Union, but there is also evidence from global analyses and structure, property rights, a firm's characteristics, political recent business surveys that attests to at least the percep- influence, and social connectivity (as discussed in chapter 5). tion of its importance in many other places. (See Kaufmann, In other words, investment environments are embedded in Kraay, and Mastruzzi 2003 for a summary of the results of unequal sociocultural and political structures.5 surveys.) It will be necessary to pay more attention to factors that A global survey finds an inverted U-shaped relationship influence both inequalities in the investment environment between the crony bias index and an index of voice and and market access and inequalities among individuals and democratic accountability, as indicated in figure 8.3 and households. Doing so will require careful empirical analy- using data from the World Bank governance database. The sis that combines economic and social science approaches. highest level of crony bias occurs in societies with interme- A recent analysis by Hellman and Kaufmann (2003) on the diate levels of democratization, which is typical of most of relationship between "inequalities of influence" in the busi- Latin America. Apart from providing advantageous influence ness sector and the quality of the institutional environment to the large and powerful, crony bias is also associated with is suggestive of the importance of these issues. This issue is weakness in public institutions, as discussed in chapter 5. also closely related to one of the general themes of this report: In societies with high levels of crony bias, firms trust and that there is a circular relationship between weak institu- use the courts less, pay fewer taxes, and have less secure tions and inequalities of power, wealth, and position. property rights; bribery is also more prevalent. Figure 8.4 223 INEQUALITY IN LATIN AMERICA their comparative analysis of experiences in different FIGURE 8.4 medium-sized Colombian cities. Tax compliance versus crony bias Tax compliance VerSUS C~ONY biaS The policy agenda related to crony bias involves build- ing the institutional basis for increased transparency, less Unofficial economy index High 60.0 corruption, better corporate governance, and more effective s r = 0.64 + regulation. This is likely to involve alliances between new * * * * forms of governance (as discussed in chapter 5) and business * *++ ** + >associations, especially those that represent small and 30.0 -* j * medium firms. Although reducing unequal influence at the "top" and , - + * ; $ + strengthening the effectiveness of institutions will provide Low 0.0 general support for more equal treatment across firms, there 0.0 0.5 1.0 are also special issues related to the informal sector to con- Crony bias index sider. Participants are highly heterogeneous, including the Low < 0 High self-employed, micro and small firms, and medium and large Source: Hellman and Kaufmann 2003, World Economic Forum 2003 firms with informal practices (especially with regard to (Executive Survey) for 80 countries, employment). Extensive literature on conditions and policy exists, and two key points can be made in this context. First, there is an agenda of proactive measures to improve the busi- illustrates this problem in relation to the extent of unoffi- ness and investment environment for small and medium cial business activities. Hellman and Kaufmann (2003) firms that is rooted in some of the fundamental questions suggest that these patterns can lead to a self-reinforcing of market failure and weak influence that were discussed in dynamic, with lower support for public institutions per- chapter 1. This agenda is likely to involve a range of areas petuating the weakness of institutions and, in turn, a greater of action, including reducing excess regulation, strength- likelihood of capture by more influential groups. ening property rights, providing infrastructure, streamlining As always, one should be cautious about overinterpret- bureaucracy, and deepening financial markets (see de Soto ing cross-country results, and new in-depth work will be 1989 and 2000 and Tendler 2002, among others). Second, needed to explore these mechanisms further. However, it is this agenda is not institution-free, but is very much about worth noting that the themes considered here are consistent the forms of political and social interaction that exist in the with a current focus in the political economy literature: broader societal context. As one example, the spectacular the new distributional coalitions that emerged post- success of the "Third Italy" small-scale sector was not a liberalization and could take advantage of new means of product of government favors, but of vibrant associational securing better deals. (See Schamis 2002 for a discussion on activity that provided specific services to firms in order to this point in the context of privatization.) facilitate the processes of formalization and recognition of A complementary issue concerns the role of business asso- workers rights (see box 8.2). ciations, especially those that represent small and medium firms. As noted above, the crony bias index is itself largely 8.2. Labor market policies and inequality6 a product of the perceptions of these groups. Related cross- Can well-designed labor policy and institutions lead to country work finds a positive relationship between the greater equality without the occurrence of major losses in importance and activity of such associations and measures efficiency? As with overall structural policies, there are of institutional performance. This is consistent with case sharply divided opinions on this question. One school of study evidence of the potential role of such associations in thought argues that inflexibilities in the labor market and transitions away from clientelist and patronage-based high levels of protection and privilege for formal workers relationships between government and business. There is are sources of higher inequality between "insiders" and "out- literature on this issue from developed countries, and Angell, siders" with respect to formal labor policies and institutions, Lowden, and Thorp (2001) document related patterns in as well as sources of lower growth in "good" jobs. Another 224 POLICIES ON MARKETS AND INSTITUTIONS BOX 8.2 The evolution of snail-scale firns In the T'rhird ItaWly The expansion of the small firm sector in central and north- sector in terms of interactions with governments on policy eastern Italy is one of the great economic successes in the and supporting small firms in eft'ecting change in organi- second half of the 2 r0 century. This change contributed zation and practices to the absolute and relative economic and social transfor- Second, the type ot policies and practices adopted mation of the region. Modena, for example, went from strongly tavored what would now be termed formalization. being a high unemployment region in the early 1950s, Ideologically committed to worker rights, and pragmat- ranking j)0"h amongst Italy's 95 provinces in terms of per ically committed to produc[ivity growth, the CNA was capita income, to the top of the list in 1980 and one of an important advocate for good working condciions and the wealthiest areas in Europe efficient organization. The major rype ot instrument it used Part of this story provides lessons on the dynamics of was what Criscuolo terms development burden-relieving,' small firms in developing countries today As Criscuolo or the provision of a range of both adminis[rative services (2002) discusses in a case study of firms in Bologna and around accounting, payroll management, and fiscal Mlodena, dynamic growth emerged neither in the form counseling) and production-targeted support (around pro- of special government favors nor generalized burden- ducers'consortia. industrial sites, and innovation centers). relieving or deregulation In the developing country con- These activities helped small firms both comply with the text. Tendler (20021 has argued that such approaches risk formal regulations that the CNA advocated and sustain confining the small firm sector to social policN, and ren- high levels of productivity growth. dering it vulnerable to the exchange of favors or relief for A remarkably dynamic small-scale sector emerged as a political support. product of mobilization and the central role of deliberate, Two elements of Criscuolo's (2002) diagnosis are worth proactive. and persistent support of association. This highlighting here. First, central to the dynamics of the helped artisan firms pursue the "high road ro small sector and its interactions wirh the state was vigorous business development While institutional context is associational activity The Coni/dna~zone Nazionale intrinsically specific to geography and history, there are &P//'.Arttza1)zto (National Confederation of Artisans, or potentially valuable lessons to be learned for deepening CNA) emerged from anti-fascist resistance in the 1930s markets in Latin America and I 9-0s, and had close connections with the commu- nist movement. It became the key acror for the small s(ale S/r,e Cr[licsr,lc. 't)0O school of thought argues that state policies on labor are essen- holds true in the context of partial coverage. In some cases, tial to ensure decent working conditions, reduce the risk of such coverage is associated with differential privileges, which exploitation of the vulnerable, and provide reasonable levels impose implicit costs on outsiders; social security provisions of income security against job loss, sickness, and old age. in many countries are an important example. There are also Sorting through these issues is a major (and important) areas in which labor institutions lead to inflexible and undertaking (see World Bank 1995 for an overall discussion). inequitable outcomes. The design of labor policy is complex, and the literature on Teachers unions are an example of a labor institution that the issue is expansive. Specifics need to be carefully worked has often (though not, of course, always) been an impediment out in particular contexts. As with many issues covered in to reforms that would lead to higher quality and more equi- this report, the current discussion is restricted to a few gen- table schooling, which is an area of fundamental importance eral policy themes. to the expansion of more equal capabilities. (Such cases reflect Current labor policy and institutions are an unholy com- "opportunity hoarding," in the language of sociologists.) promise in much of Latin America. To the extent that they More generally, while unions often have helped improve the make a difference to labor market outcomes, this assessment working conditions of their members, this rarely leads to 225 INEQUALITY IN LATIN AMERICA greater equality. Estimated effects are usually negligible. In with their deeply reified blood ties, rituals of membership, the case of Brazil, unions seem to lead to greater inequality and high existence costs are only the extreme case" (Heller of wages. This is because, in contrast to Organisation of and Mahoney 2003, p. 41). Economic Co-operation and Development (OECD) countries, A number of directions that could be taken to achieve a unionized workers tend to be in the more skilled and better- more equitable labor policy are put forth here, in the spirit paid part of the overall wage distribution. of seeking to both extend worker rights to all and increase The other side of this situation is that large segments of the flexibility in those sectors of the labor market that are the workforce are not covered by union action, protected in relatively protected. These policies can generally be framed terms of health standards, discrimination, or abuse, or pro- in terms of increasing incentives for formalization by both vided with even basic forms of income security. Many Latin reducing costs and increasing benefits. American countries are reasonably flexible in overall labor First, policies that support the informal sector are needed market adjustments (though there are important differences (as noted in the preceding section). In particular, Latin between, for example, relatively flexible Mexico and American countries should seek innovative ways to increase Argentina), but flexibility is the flip side of having large infor- the productivity of firms in this sector and facilitate their mal sectors that are weakly covered by formal working participation in formal institutions. The informal sector conditions. Minimum wages generally have modest and should not be viewed exclusively as a pool of workers left mildly equalizing effect on the structure of wages, but this out of formal employment due to excessive regulations, but is offset by reductions in employment, especially for young, rather as a group of entrepreneurs, and their employees, that low-skilled, and female workers (World Bank 2004). Where need support and amenable policies in order to grow and significantly binding, as in Colombia, minimum wages prob- succeed. Such policies include the reduction of transaction ably contribute to lower levels of formal employment. costs, the availability of credit and formal mechanisms Large informal sectors are symptomatic of both weak through which to operate, and, most importantly, labor formal employment growth and a high ratio of the costs to laws that permit firms to adapt to changing economic the benefits of formalization, whether for employers or conditions. workers. However, this does not imply that workers in the Second, labor market laws and minimum wage-setting informal sector are necessarily worse off, or that they are procedures should be assessed carefully in each country. Dual queuing up for better work in the formal sector. The key objectives should be to remove excessive rigidity and to word is heterogeneity. Many informal sector participants are reduce the impact of minimum wages on unemployment. self-employed entrepreneurs who are in that sector out of More specifically, removing excessive rigidity means giving choice because of higher returns or preferred working con- employers and employees alike efficient mechanisms with ditions. (See Maloney 2003 for a survey of issues related to which to adapt to the business cycle, which would then be informal sector employment.) In some cases, households may reflected in higher formal employment rates in the long run. adopt diversified employment strategies, with one worker Regarding minimum wages, it is important to ensure that maintaining a formal job in order to secure social security the mechanism used to set them be based not only on and health benefits and others pursuing informal activities. purchasing power, but also on considerations related to However, informal work can also be more susceptible to productivity and employment. socially based inequalities. As Heller and Mahoney (2003) If managed effectively, these policies can be consistent argue, because the informal sector is largely outside the reach with greater support for workers. This can be done through of the state and transactions within it are largely extra-legal, both reduced informality and the spinoff effects of a generally success is often dependent on networks and on noninstitu- more supportive approach to labor in terms of institutional tionalized mechanisms of reducing transaction costs. In other functioning. Figure 8.5 illustrates the tradeoff between more words, "Access to capital, control of space and markets, rigid protection and informality. However, note that Costa enforcement of transactions, and maintaining buyer and Rica has both more rigidity (as proxied by an index of the seller networks thus all involve controlling access to a costs of firing workers) and lower levels of informality than resource or a network by erecting strong boundaries of does the Dominican Republic. In an in-depth comparison ethnicity, gender, community, and race. Criminal networks, of the labor markets in these two countries, Itzigsohn (2000) 226 POLICIES ON MARKETS AND INSTITUTIONS FIGURE 8.5 Labor market rigidity and informality Informal sector share of total employment 55 50 - * HON PER BOL 45 - *DOM 40- 35- PAN * ~~~~~ECU 30- 25 - + *UR ~ ~ ~~U *ARG 20 - *CSR 15- 10 - 0 0.5 1 1.5 2 25 3 3.5 4 4.5 S Labor market rigidity index Source: Heckman and Pages-Serra 2000 and LABORSTA database. argues that this pattern is the product of an overall strategy interests of broader groups of poor workers and households. in Costa Rica that was more supportive of worker rights (This is not, however, to say that they will not still seek to and working conditions than efforts in the Dominican vigorously defend the interests of their members.) Republic. The effects of this approach were therefore as much Finally, there is no inherent inconsistency between unions, a result of the impact on the functioning of institutions as state regulation, and integration into the global economy via of the details of policies. a high productivity route. This is illustrated by a case study Third, reforms in the area of social security should con- of union involvement in export-oriented, high-value crops tinue to be made with the intention of achieving a more in northeastern Brazil (box 8.3). Unions have been an impor- equitable form of risk management for all workers and house- tant element of a shift toward a path to integration that has holds (see chapter 9). benefited both workers and firms, primarily at a local level Fourth, unions have a central role to play in both the in a specific industry. Cases in which productivity has workplace and society, but the context in which they func- increased cooperation at a national level are probably harder tion influences the distributional effects of their activities. to find in Latin America, in contrast, for example, to the role Where unions have become means for protecting privileges of the Swedish union movement in that country's develop- that impose costs on others-especially those who are less ment (de Ferranti and others 2002). fortunate-then more flexibility will be of greater value. There is a contrast between those union movements that were 8.3. Inequality and macroeconomic crises effectively incorporated into vertical state structures (Mexico, Macroeconomic instability has been the bane of long-term eco- for example) and those with stronger horizontal affiliations nomic performance in Latin America. Periodic crises have been (as in Brazil). The latter will be more likely to act in the sources of pervasive and deep social scars. In recent decades, 227 INEQUALITY IN LATIN AMERICA HOX K . Unions, finns, and the expansion of hig6-ale export cmp in petnlluaia h iohetastn razg A study on the evolution of worker-firm relationships in timing for quality in the agricultural process (tor example rhe high-value export crop secror in Petrolina-Juazeiro in harvesring grapes) gave the workers potential levcrage. underscores the potential positive role of unions in increas- The rural union had to significantly change its strategy ing both productiviry and the well-being of workers. In from focusing on its rradirional base among small farmers a seemingly unpromising environment of landless in order to represent the interest of landless workers. In agriculiural laborers, union activiry was an important addition, the farm owners were primarily large-scale firms ingredient in the shift ro work practices that both offer from southeastern Brazil with experience in collective bar- berter working conditions and achieve high productivity- gaining, in contrast ro the more traditional large-scale as well as, importantly, high qualiry-in a successful area sugar farmers of rhe northeast who were accustomed to ot export agriculture. This outcome was facilitated by a more repressive and conflicti%e labor relations. combination of market conditions and the strategy and history of institutional actors. The critical importance of Sowne: Dnnilani 2003 the 1980s debt crisis brought to a halt an extended period of of macroeconomic instability and the unequal resolution of growth in much of the region, albeit growth that occurred crises? on highly unequal terms. By the beginning of the 1990s, most The usual suspects for crisis causation have traditionally countries appeared to have gained a sounder macroeconomic been faulty macro-level policies, especially those associated policy footing, and market-based debt reductions had helped with excessive public spending, overvalued exchange rates, reduce the debt overhang. Yet crises continued to hit the and loose monetary management. There are well-documented region in the subsequent decade. The 1994-1995 Mexican links between such factors and volatility, high inflation, and and the 2001-2003 Argentine crises will be remembered currency and banking crises. These have correctly received as much as the steady gains in health and education that significant attention-not least of all by international finan- occurred throughout the region during this time. cial institutions-with the objective of achieving better The causes of crises-as well as of optimal macroeco- macroeconomic policies to prevent crises. In the wake of the nomic, financial, and debt management policies to prevent 1980s debt crisis, realistic exchange rates and prudent fiscal and manage crises-is an immense area of study. The and sound monetary policies were central to the policy literature and debates are not summarized here, but instead platforms being advocated for, and eventually adopted by, crises are examined through the prism of the concerns and countries in Latin America. There were important gains that approaches contained in this report. Are there any causal have, in particular, been associated with large reductions in links between weak and unequal institutions and crises? inflation. Most studies find that lower inflation tends to have What are the distributional consequences of crises? Does the positive effects on income distribution. perspective of this report cast light on feasible policy options Substantial gains in these policy areas seem to have ush- that are better for reducing inequality and managing crises? ered in a period of stability. However, this changed with the Each of these questions is considered in turn. 1994-1995 Mexican crisis, the ripple effects of the East Asian and Russian crises in 1997 and 1998, and macroeconomic What causes macroeconomic crises? Proximate difficulties in countries such as Argentina, Bolivia, and and deep factors Venezuela starting in 2000. All problems had clearly not been Earlier parts of this report explored how weak and unequal solved. In the economics field, these and other crises (includ- institutions can be self-sustaining with regard to long-term ing those that occurred in Europe during the 1990s) helped development processes and underlying policy choices. Some spawn a range of new thinking. Recent perspectives have institutional weaknesses may actually be functional for the emphasized asymmetric information problems in financial powerful. Can weak and unequal institutions also be a source markets, the possibility of multiple equilibria with high and 228 POLICIES ON MARKETS AND INSTITUTIONS FIGURE 8.6 The association between macroeconomic volatiliy and measures of institutional conditions Standard deviation of GDP growth 25- Correlation coefficient: -0.3638 20- 15 - 10- S - * + + C l'Y ECU AR E TRO CHL VEN ~~~~~~~PRY 0 I HTI I HON GME c 0 1 2 3 4 5 6 7 8 Constraints on the executive index Source: Authors' calculations, using World Bank statistics and the Polity IV database for constraints on the executive. low external capital flows, currency balance sheet mismatches by the bivariate relationship between a measure of the extent in government and the corporate and financial sectors, and to which the executive is constrained by national institutions weak financial sector regulation and performance.' and the standard deviation of growth in gross domestic Even when fiscal and monetary policies are reasonably product (GDP) (figure 8.6). This association says nothing sound (as they were in much of East Asia before crisis about causation, but some studies do provide an empirical ensued), surges in capital inflows reinforced by premature basis for the inference of some causal processes on the or poorly designed capital account liberalization, as well as relationship. weak prudential regulation and supervision, have led to the An example of empirical work on crises is a recent study excessive accumulation of short-term external debt and credit by Acemoglu and others (2002) that explores the relation- booms. This pattern rapidly reversed when the booms ended, ships among macroeconomic policies, institutions, and and led to sharp domestic asset price declines and large macroeconomic volatility. It argues that the strength of currency devaluations. These problems in turn resulted in institutions is a principal long-term causal force behind widespread corporate bankruptcy and deep recessions. volatility, which can be manifested through macroeconomic One theme of relevance to this report can be emphasized policies or other factors (box 8.4). here. As in many areas discussed in other chapters, economic This analysis is useful, but does not explore the specific processes interact with institutional conditions. Weak mechanisms of crises. Two are likely to be of importance institutions (which are related to inequality in ways discussed and to lie at the intersection of concerns over inequality: in chapters 4 and 5) may have a causal influence on the occur- management of distributional conflict and the relationship rence of crises. There is a suggestive association between between unequal influence and the functioning of the weak institutions and macroeconomic volatility, as illustrated financial system. 229 INEQUALITY IN LATIN AMERICA BOX b 4 Ins_aailonl mcro.cmlc poky, and uMUYm Recent work by Acemoglu and others (2002) explores the based on the substantive view that extractive instilucions relationships among institutions, macroeconomic policy formed in colonial periods can have a persistent negarive choices, and economic volatility. Countries that have had influenceon contemporary institutional conditions(see chap- tauLry macroeconomic policies (such as excessively expan- rer 4.) Such extractive institutions are more likely to evolve sionary fiscal policy, loose monetary policy, and overval- in situations where there is relatively modest settlement. ued exchange rates) have also experienced higher whichallowsforthe useofan indexofsettlermortalirycduring macroeconomic volatiliry, as well as lower growth, in the the colonial period. This measure ofthe hospitabiliry ofthe past few decades. The question is whether the end of the environment iscorrelated wirh recent institutional measures, story is that governments make the "mistake" of choos- but there is no reason why it should have any direct influ- ing bad policies that [hen cause volatility. enceon current macroeconomic policies. An alternative view is chat both the choice of policy and When the analysis of influences on volarility is under- economic volatility reflect underlying institutional weak- taken using settler mortality as an instrument to identify nesses. By institutions, the authors are concerned with a the exogenous, historically determined component of clu.er olf sLnalarrangementj that include constitutionali and recent institutional conditions, rhe relationship between social limits on poliricians and elites' power, the ruJe of macroeconomic policies and volatility largely disappears, law, provisions for mediating social cleavages, strong prop- while measures of instirutional strength remain robustly erty rights enforcement, a minimum of equal opportunity, significant. This trend applies both when macroeconomic and broad-based access to education, etc."a (Also see the policies are treated as exogenous (which would generally discussion in chapter S.) The key argument in this regard bias the results ro favor the influence of such policies, if is that insiltutional weak-nesses will tend to lower invest- these variables are actually endogenous), or if lagged values ment and increase volarility and policies, for example of macroeconomic policies are used as instruments. There through the inability to manage distributional struggles. is weak; support for rhe role of exchange rate overvalua- Cross-country evidence in support of rhis argument is tion, but not for the role of inflation or measures of public provided by explonng the empirical relarionship between spending and deficits. The authors suggest that where macroeconomic policies-an indicator of institutional underlying institutional weaknesses exist, they can spill strengrh using an index ofconstraints on the executive from over into crises in various ways, sometimes through macro- rhe Polity [V database-and measures of economic volatil- economic policy mechanisms and sometimes via micro- iry, including the standard deviation of growth and largest economic routes. This is further illustrated by case study declines in output. The empirical challenge is that both analyses of Argentina and Ghana. macroeconomic policies and institutions are likely to be endogenous, or to reflecr the influence of omitted variables. Sokr,,. Acemoglu and others 2002 To deal with this, the authors use an idenrification strategy a Acrmoglu and oEhers 21012, p 4 With respect to the management of distributional conflict, struggles on the one hand, and institutions for conflict man- there is a tradition of work that focuses on how fights between agement on the other. The study looks in particular at groups can have potent influences on the conduct of macro- differences in growth during 1960-1975 and 1975-1989, economic policy. (See Bates 1981 for a seminal piece on this periods when most of Latin America suffered a significant subject.) The interaction between distributional struggle and decline in economic growth while East Asia actually experi- institutional weakness is illustrated in cross-country work by enced accelerated growth; this difference occurred despite the Rodrik (1999). That analysis seeks to explain the varying fact that both regions experienced shocks. impacts of the shocks of the 1970s on growth in terms of the Countries that experienced the largest declines in growth interaction between latent social conflict or distributional were not necessarily those that suffered the greatest shocks, 230 POLICIES ON MARKETS AND INSTITUTIONS but rather those with a combination of more divided societies The studies referred to here are by no means the last word (as measured by indices of inequality or social difference in empirical investigation of interactions among macro-level between groups) and weak institutions for conflict man- policies, institutions, and volatility with regard to both cross- agement (as proxied by indicators of the quality of govern- country analysis and, perhaps more importantly, in-depth mental institutions, the rule of law, democratic rights, and case studies. As noted earlier, an extensive economics liter- social safety nets). In Rodrik's (1999) analysis, more con- ature on the sources of crisis already exists. Questions related ventional influences on economic vulnerability-including to institutional weakness and links to inequality intersect economic openness, the share of government consumption, with only some of the mechanisms discussed in that litera- and levels of indebtedness-become insignificant predictors ture. Macro-level policy choices of course have an effect at of growth collapse once the measures of latent conflict and least in the short- to medium-term, and the functioning of institutions for conflict management are introduced. The the financial system will be central to whether crises occur discussions in this report (especially in chapters 4 and 5) and how severe they are. The mechanisms whereby institu- support the view that Latin America is generally a region tional weaknesses interact with distributional struggles to of high social cleavage and weak institutions, including weak form a source of volatility require considerably more explo- constraints on the power of elites, when compared with East ration. Nonetheless, there is a need to bring these issues back Asian and OECD countries.8 to the center of an analysis of crises. Completing this type of story would require careful case study research. A vivid contemporary example is Venezuela, The distributional consequences of where social-political conflict and the associated stoppage by macroeconomic crises the oil industry directly led to approximately a 26 percent Macroeconomic crises affect society in multiple ways. Four drop in GDP from March 2002 to March 2003. Less dra- economic channels of influence are emphasized in most matic episodes also occurred earlier, with the government's research: the impact of reduced demand on jobs, wages, and ability to engineer "social pacts" increasingly exhausted from enterprise income; changes in prices; public spending cut- the mid-1980s on, a trend that in turn limited society's backs; and changes in the value of assets.9 As households and ability to resolve distributional conflict through semicor- communities respond to these effects, secondary effects can poratist settlements. This is an interesting example precisely be felt in terms of household decisions on schooling or health, because Venezuelan democracy was founded on the basis of the increased work of women and other secondary workers, an explicit pact between the government and the elites, as depletion of financial assets, and reliance on social networks well as on the deliberate use of distributional-but not redis- for support. tributional-public policies stemming from the abundant Many observers also emphasize the costs of macroeco- public resources generated by oil revenues. This scheme nomic crises to the social fabric, as individuals are forced into worked for a while: Venezuela succeeded in maintaining illegal activities or as young men and women adapt to remarkable macroeconomic stability and quite respectable adverse opportunities by shifting into cultures of drugs, levels of growth during the 1960s and 1970s. violence, or prostitution rather than skills acquisition and The second area in which institutional weaknesses may productive work. Adverse effects can be persistent (a phe- be associated with crises is the links between unequal nomenon known as "hysteresis" in the jargon of economists), influence of the corporate sector and the functioning of the for example through losses in organizational, human, or financial system. As noted in section 8.1, "crony bias" is asso- social capital that can take considerable time to rebuild. As ciated with a range of weaknesses in institutions. To the one example, it is striking that personal violence in Latin extent that this phenomenon weakens the lending practices America increased in the 1980s and then did not go down and prudential behavior of the financial sector, it can have even when economies recovered in the 1990s. (For more on a pernicious influence both on crisis causation (via excessive this subject, see Fajnzylber, Lederman, and Loayza 2000.) credit expansion) and crisis dynamics (through poor port- This section looks at three categories of evidence related folios). This is of particular importance to interpretations to the effects of macroeconomic crises. First, the results of of some of the more recent crises that were not based on fiscal studies on income distribution, as measured by household sur- laxity. veys and via spending, are summarized. Second, information 2 31 INEQUALITY IN LATIN AMERICA on the distribution between labor and capital is discussed. TABLE 8.2 Third, distributional aspects of the resolution of crises Changes in real per capita income across the income through financial sector channels are considered. While these distribution in Mexico, 1992-2000 two latter sources are relatively new and represent early research efforts, they have a particular bearing on the issues Increase in real per capita income (in percent) discussed throughout this report. Deciles 1992-94 1994-96 1996-98 1998-2000 1 21.3 -23.4 0.1 21.9 Effects on income distribution 2 10.0 -19.9 4.9 14.1 Are the costs of crises unequally distributed? It is often 3 11.9 -21.6 4.6 21.7 asserted that the poor suffer most when a crisis hits. Past 5 615 -22.1 9.6 17.3 work actually reveals a mixed picture in this regard. With 6 7.6 -20.0 9.0 11.3 respect to influences on incomes, evidence from household 7 6.0 -20.9 13.9 19.6 8 8.6 -22.6 11.9 16.4 surveys typically finds losses across all strata that are par- 9 8.7 -21.5 11.5 12.5 tially disequalizing in some cases but not in others. For 10 14.5 -24.1 20.4 5.6 Top 5% 17.5 -24.7 22.2 4.7 example, Lustig (2000), in reviewing 20 Latin American Top 1% 18.4 -23.9 37.0 -3.5 crises, found small increases in the Gini coefficient in 15 cases that compared post- with pre-crisis periods, but small Source: Authors' calculations from Mexican household income declines in others. DeJanvry and Sadoulet (2000) reviewed and expenditure surveys. 48 instances in 12 countries in Latin America between 1970 and 1994 and found asymmetric effects between recessions L6pez-Acevedo and Salinas (2000) found relatively high that tended to increase income inequality and periods of losses in the top decile of the income distribution, for which positive growth, which in effect led to a ratcheting up of the share of total income declined from 42.3 percent in 1994 inequality over this period. to 40.7 percent in 1996. This was the main force behind the Morley (2001) found that during the 1980s, recessions fall in the Gini coefficient from 0.534 to 0.519 during this typically led to increases in inequality, while recoveries were period. Table 8.2 shows the pattern of income decline that associated with lower inequality. However, Morley also found occurred between 1994 and 1996. no clear pattern of response to macroeconomic crises during Why did richer groups in Mexico lose out in the crisis? It the 1990s. On the basis of worldwide evidence, Ravallion might have been expected that the rich would be particularly and Chen (1997) found no overall pattern of change in the hard hit by losses in financial income in the wake of the asset income distribution during episodes of macroeconomic price declines associated with the currency and financial crisis. decline in most regions, with increases and declines occur- In fact, the opposite appears to have occurred, at least on aver- ring in roughly equal proportion. The major exception in age. The decline in the share of the rich was instead driven the 1990s was the successor states of the former Soviet by sharp declines in labor earnings in the financial and other Union, which experienced significant increases in inequal- high-paying service sectors. By contrast, the contribution of ity during episodes of large-scale income decline. (These were financial income to total income increased (table 8.3), with clearly associated with the large-scale institutional changes growth in real financial income per capita at the top of the taking place during the transition from socialism (World distribution rising from 6 percent for the top decile to 14 per- Bank 2000d). cent for the top 1 percent of the income distribution. Although there is some evidence that recession increases Evidence on financial income from household surveys is inequality, some of the time, most of these results suggest generally weak, covering only a fraction of total financial that there is no general pattern of the influence of adverse trends in a country. Although these results should be treated shocks on income distribution (or at least not on that part with caution given the usually weak coverage of capital of the distribution captured by surveys; see below). This is income, the apparent capacity of the rich to increase their further borne out by case study evidence, for example with financial income during the crisis is striking."0 It is possi- regard to the contrast between Mexico and Argentina in ble that this occurred because of earnings on money that was recent years. With respect to the Mexican Tequila crisis, successfully shifted out of the country just before the crisis, 232 POLICIES (N MARKETS AND INSTITUTIONS TABLE 8.3 Financial income during the 1994-96 crisis in Mexico Real increase in Mean financial Share of financial income income per capita financial income in total per capita (in 2000 pesos) income (percent) (percent) Deciles 1994 1996 1994 1996 1994-1996 1 20.5 16.9 1.8 2.0 -17.5 2 26.3 26.1 1.6 2.0 -0.8 3 56.8 33.5 2.7 2.0 -41.0 4 47.1 38.8 1.9 2.0 -17.5 5 65.3 67.5 2.1 2.8 3.3 6 107.4 90.4 2.9 3.0 -15.8 7 141.0 134.9 3.2 3.8 -4.3 8 161.8 147.2 2.8 3.3 -9.0 9 257.3 264.7 3.2 4.2 2.9 10 1,444.1 1,535.0 7.1 10.0 6.3 Top 5% 2,524.1 2,833.4 8.8 13.1 12.2 Top 1% 7,416.5 8,448.1 13.1 19.4 13.9 Source: Authors' calculations from Mexican household income and expenditure surveys. and that consequently experienced capital gains when the Mexico has long had a pattern of adjustment at the exchange rate collapsed. This is only a partial picture, and bottom of the labor market in which firms and workers large asset price movements occurred that imposed capital adjust through lower real wages-whether based on shifts losses on those holding Mexican assets, especially assets that into lower-paying work (for example, in the informal sector) were denominated in pesos. (The role of the financial sector or declines in real wages that are facilitated by inflation. in redistribution is discussed further below.) During the Tequila crisis, unemployment spiked, but then The finding that labor earnings became slightly more declined rapidly to the 3 percent range a couple of years later. equal during Mexico's crisis is consistent with findings by At the top of the labor earnings distribution, by contrast, Sanchez-Paramo and Schady (2002) in a study of six major the pre-crisis period had seen large increases. This was espe- Latin American countries, which indicated that the premium cially true in leading sectors such as high-paying services and placed on tertiary education tends to fall during crises. This finance, which were hit hard by crisis. premise is also consistent with an international analysis of In contrast, Argentina shifted to a high unemployment influences on income distribution by L6pez (2003) that finds labor market even during the relatively good times of the banking crises to have equalizing effects on measured income 1990s. This change was influenced by high layoffs due to after controlling for a range of other factors. Again, for most restructuring, the relatively low relative price flexibility countries, measured household income is dominated by labor associated with a fixed exchange rate and low inflation, income, and capital income is severely underreported. and, possibly, the influence of labor market policies.'2 The In Argentina, the recent crisis has so far led to a sub- Argentine crisis further increased unemployment, which rose stantial worsening in inequality, with an increase in the Gini to 21.5 percent in May 2002 before declining to 17.8 per- coefficient from 0.49 in 1999 to 0.55 in 2002. (In contrast cent in October of that year. The spike in inflation after the to Mexico, reported financial income in Argentina is so small exchange rate devaluation was a source of large real wage as to be implausible and is therefore not analyzed.") As a declines and a sharp rise in poverty. result, the ratio of income between the top and bottom A comparison of Argentina and Mexico suggests a general deciles rose from 15 times in 1999 to 28 times in 2002 lesson: the way in which crises affect household distribution (compared to 11 times in 1990). The differences between is conditional on the structure of the labor market, the sec- the two cases are undoubtedly complex, but one of the more toral incidence of adverse shocks, and the way that the labor striking contrasts lies in labor market adjustment. market adjusts. 233 INEQUALITY IN LATIN AMIERICA The above discussion refers to the evolution of income dif- debt and rising risk premiums. During the economic slow- ferences as measured in household surveys, which capture downofthelate 1990s,twoexceptionstothisrulewereChile, three channels of influence: labor market changes, relative which had developed both the credibility and the budgetary prices, and returns to nonhuman capital assets. Although no processes to run a countercyclical policy, and Bolivia, a coun- overall pattern of distribution effects is evident, it is impor- try dependent on concessional finance and therefore spared the tant to emphasize that impacts on poverty are unambiguously vicissitudes ofprivate market sentiments. In the case of Bolivia, adverse, since declines in income push some nonpoor people the deficit rose from 3.5-3.7 percent ofGDP in 1999 and 2000 into poverty and deepen the poverty of those who are already to 7 percent in 2001 under International Monetary Fund (IMF) poor before a crisis. Even an equiproportionate decline in programs. incomes across the income distribution is likely to hurt those Second, when crises induce spending declines, there is a at the bottom most; a 10 percent decline in income for those tendency to protect those categories of spending that are prac- in poverty is more costly in terms of well-being than it is for tically, legally, or politically more difficult to adjust. This those in the middle or upper parts of the distribution. includes debt service (except when crises are very deep and This result applies with even greater force to households countries stop paying), permanent public employment, living in extreme poverty, which generally have insufficient and social security entitlements (notably pensions for retired resources with which to purchase adequate food or pay public sector workers). Maintaining teachers and nurses on for medical expenses, especially during deep or protracted the public payroll can help keep basic social services going, recessions. (See de Ferranti and others, 2000.) In addition, par- although there can be discontinuities when crises are deep and ticular individuals and households can experience much larger workers don't get paid, or when declines in real wages become drops in income, a phenomenon that is hidden in data based very costly in terms of morale and complementary inputs. on averages. Large income declines for individuals and house- Moreover, there is evidence that cuts disproportionately holds are concerns that are distinct from chronic poverty on hurt programs (especially discretionary ones) that are rela- both welfare and political grounds. (See Pritchett 2001 for tively progressive. Just as many programs tend to function an analysis of Indonesia.) This fact has implications for the along the income distribution as they expand, thereby crowd- design of safety net instruments, as discussed in chapter 9. ing in poorer groups (see chapter 9), spending cuts work in The other main channel that has been studied in past the opposite fashion (see Ravallion 2000 and 2002). Of par- research concerns the distribution of public spending and ticular concern are cutbacks in targeted transfer programs, the consequential distribution of services and transfers. Two the resources of which decline or fail to expand as needs rise general processes are important in this regard. First, in the alongside an increase in poverty. aggregate, public spending in most Latin American coun- tries tends to be procyclical, expanding in good times and The distribution between capital and labor contracting in bad (see chapter 9). This trend is driven by Household surveys are relatively good at capturing labor political pressures to increase spending when the going is income and some transfers, but are particularly weak at cap- good, the strongly procyclical nature of private market sen- turing capital income. Yet the distribution between labor timents, and the proclivity for "sudden stops" in market and capital income is an important part of the story. This flows when the going is bad (see de Ferranti and others 2000 issue can be examined using information from national and Calvo and Reinhart 2000 for more on this point). Sharp accounts. Doing so involves considering the functional dis- reversals in capital flows are also associated with the lack of tribution of income between workers and owners of capital credibility given to countercyclical fiscal policies because of rather than the personal distribution across households. asymmetric information problems: it is difficult for markets There is evidence that the share of income going to labor to distinguish between a prudent countercyclical fiscal policy decreases during crises and that there is some persistence in and fiscal laxity, especially when governments have not this redistribution.'3 proven able to produce surpluses in good times. The underlying sources of information for labor and This trend in turn affects the capacity to finance deficits and capital shares in national accounts are different from those increases the burden of debt service because of exchange rate contained in household surveys. While household surveys ask depreciations for foreign and domestic dollar-denominated household members where they work, how much they earn, 234 POLICIES ON MARKETS AND INSTITUTIONS and what they spend money on, national accounts are based the interpretation of what happens during crisis episodes. primarily on information from production sources (that is, In order for patterns of change in the distribution of income surveys of firms, estimates of agricultural production, and in the national accounts to reflect real changes, any biases sales statistics). Detailed information on the composition or inaccuracies have to be the same during the entire cycle. of value added is typically only available periodically-for This may not be the case during economic downturns, for example from agricultural and industrial censuses and example if informal labor incomes are underestimated and special surveys of service sectors-so that estimates of the the relative share of this source of income rises. functional distribution of income are likely to contain sub- With these now familiar caveats on the quality of the data stantial inaccuracies. This is especially the case when the in mind, the resulting patterns can be examined. Using a informal sector is significant. global database, Diwan (2001 and 2002) examines the rela- The data on labor shares provided here cover primarily tionship between crises (in currency, banking, or both) and formal work. Because of all the limitations discussed in chap- labor shares. These studies find that labor shares systemat- ter 2, household surveys are the preferred instrument for ically fall during crises and don't fully recover afterward, a obtaining direct measures of the well-being of households cross-country result that is illustrated for four countries in and of changes over time. However, national account sources Latin America in figure 8.7. The cross-country pattern is may be superior in capturing what is happening to all sources highlighted by the case of Mexico, where the labor share fell of income, including capital income. The fact that capital sharply in 1982, steadily recovered during the following income and the incomes of the rich appear to be relatively decade, weathered the more moderate macroeconomic poorly captured by the surveys adds another dimension to setback of 1989, and then fell significantly following the FIGURE 8.7 The evoluton of the labor share in Chile, Mexico, and Peru a. Chile b. Mexico Labor income as a Labor income as a share of total income share of total income 60- 60- 55 - 55 -1 50 :50 -I 40 -40 -~ 35 -35 - 25 25 20 20 c. Peru - Crisis year| Labor income as a share of total income 60 - 55 -f 50 - 45 - 40j 35 A. 25- 20 + Note: Crisis years are defined as years in which at least two out of three of the following occur: a 25 percent nominal devaluation, negative growth, and 50 percent inflation. Source: Authors' calculations, based on national accounts data. I 235 INEQUALITY IN LATIN AMERICA 1994-1995 crisis. The flip side of this pattern, of course, is labor share around the time of crises, which at least in some that the shares of corporate and financial sector capital cases appear to be associated with the resolution of politi- income rise relative to wages. There are also significant inter- cal fights between formal labor and other interests at the actions with structural variables. In particular, closed trade, expense of labor. This is perhaps most clearly the case in capital controls, and fiscal deficits are associated with higher the episode of high inflation and stabilization under the labor shares in normal times, but also with large falls in labor Argentine military government in the 1970s. shares when crises occur. What are the mechanisms behind these trends? There are The impact of financial sector "adjustments" both proximate and deep explanations. In terms of proximate on income distribution factors, labor market and prices channels tend during crises While the evidence surveyed here provides an important part to lead to a disproportionate drop in the portion of value of the picture, it still leaves out some of the larger issues at added that workers obtain through a diverse combination play. In particular, currency and financial crises are both of wage and employment declines. Worker and household driven by, and cause, large movements within the financial adjustments-whether occurring through shifts into system, which in turn have significant fiscal consequences lower-paying informal work or the increased labor force par- for the future. According to a survey of 40 banking crises, ticipation of secondary workers-do not compensate for the the average fiscal costs were 14.7 percent of GDP. These costs relatively adverse effects on initial employment conditions. were incurred because of a variety of resolution tools, includ- (If they do, it is not captured in the data.) With regard to ing guarantees for depositors and creditors, liquidity sup- capital, most Latin American countries have relied signifi- port, repeated capitalizations, and public debt relief. While cantly on external savings to finance investment (leaving them the costs are distributed over many years, these are clearly with high debt burdens and vulnerable to capital outflows) nontrivial amounts. In addition, they have often been larger, and on large movements of asset prices when confidence falls. as table 8.4 illustrates, for Latin America and countries hit Interest rates are typically increased to reduce capital flight, by the East Asian crisis. Mexico's Tequila crisis cost 19.3 per- thereby also protecting domestic financial income. cent and Chile's 1981-83 crisis cost 41 percent of GDP. At a deeper level, the interpretation of causes depends on The bailout of Banco Intercontinental in the Dominican the underlying view of how labor market outcomes are deter- Republic in 2003 (which is still too recent to be included mined. In a world in which wages are determined by mar- in table 8.4) is estimated to have cost between 14 and 17 per- ginal productivity-and are therefore subject to the multiple cent of GDP (see World Bank 2003g). The bill for the influences of institutional or policy-related effects on wages, current Argentina crisis is not yet in. By contrast, the Savings from minimum levels to severance pay-the explanation and Loans crisis in the United States during the 1980s cost should be sought in the disproportionately high decline in just 3.2 percent of GDP. labor productivity, relative to returns on capital, that occurs during crises. It is not clear why this pattern ensues. Total fiscal costs of selected banking crises Alternatively, if bargaining or, more broadly, power rela- tionships are an important influence on private and public Fiscal cost wages, the explanation lies in a decline in the relative power Country and episode (in percent of GDP) of workers during a crisis. Under this premise, crises can be Argentina, 1980-82 55.1 seen as mechanisms for the resolution of distributional con- Brazil, 1994-96 13.2 Chile, 1981-83 41.2 flicts that are not tackled during good economic times. The Ecuador, 1996- 13.0 latter is the preferred interpretation of Diwan (2001 and Mexico, 1994- 19.3 2002), who sees crises as mechanisms that societies use to Venezuela, RB de, 1994-97 250. "digest" social losses across groups. Labor is relatively immo- Korea, 1997- 26.5 bile and so typically bears a higher proportion of the cost.14 United States, 1981-91 3.2 Further exploration of this issue will require in-depth Source: Honohan and Klingebiel 2000. research on case studies. Fizsbein and Galiani (2003), in Note: Costs refer to both fiscal and quasi-fiscal outlays and are initial work on the Argentine case, find large changes in the the accumulated total. 236 POLICIES ON MARKETS AND INSTITUTIONS What are the distributional dimensions of financial crises? The bulk of borrowers were from the formal and, in particular, Are richer financial sector participants bailed out at the the large-scale corporate sector. Owners of this sector are typ- expense of poorer groups outside the financial system? Do ically wealthy and workers come from the middle of the richer participants within the financial system gain (relatively income distribution. Most of the beneficiaries of the financial speaking), or receive higher levels of protection for their transfers were therefore well-off by Mexican standards. It is losses, when crises occur? In background work for this report, possible that there were secondary benefits in employment Halac and Schmuckler (2003) found evidence of adverse stemming from financial support measures, but international distributional effects of crises through both effects. Large, evidence on bailouts does not find any association between the foreign depositors enjoyed the greatest level of compensa- size of a resolution and output. tion, and sometimes capital gains, in crises in Argentina, Transfers that took place were financed by higher bud- Ecuador, and Uruguay, while small depositors suffered capital getary surpluses than would have otherwise been needed in losses. In addition, there is evidence that large borrowers the absence of financial crisis. The distribution between higher with close connections with banks particularly benefited taxes and lower spending is hard to judge in terms of its from crises in Chile, Ecuador, and Mexico. This is illustrated relation to the counterfactual without a financial transfer. here with case study evidence from Mexico and Argentina. However, the main tax instrument, the value added tax (VAT), There is no direct, survey-based evidence of the house- is essentially a flat tax in terms of income distribution; in other hold impacts of financial sector changes. An assessment words, the poor and the rich pay in equal proportion. As therefore has to be formed on the basis of indirect evidence. discussed in chapter 9 and noted above, spending expansions The total size of the Mexican bailout is estimated at $112 typically crowd in poorer groups. Assuming that this would billion (Honohan and Klingebiel 2000), with a large addi- have occurred in Mexico in the late 1990s, poorer groups tional amount spent trying to prevent crises from occurring disproportionately lost out from the lower spending than with regard to liquidity support, sovereign bond swaps, and would otherwise be the case. This indirect evidence suggests the financing of large investors who withdraw money from that the winners were predominantly in the top part of the projects. Halac and Schmukler (2003) use the $23 billion distribution (including some of the very wealthy), while the decline in the Central Bank's reserves between February and losers were spread throughout the distribution, with a December 1994 as a proxy, thereby calculating a total fiscal probable bias toward poorer groups. and quasi-fiscal cost of the crisis of $135 billion. This amount The ongoing Argentine crisis provides a second case. The represents about one-quarter of Mexico's GDP in 2000 and effects of the crisis are still being worked through, and the some four times the $33 billion received in capital receipts design and final fiscal costs of any bailout are still unknown. from privatization during the 1990s. Nonetheless, it is possible to look at changes within the Winners and losers in the large financial transfers in financial sector that occurred in the context of the very large Mexico can be roughly assessed by examining the charac- reductions in well-being (and increases in inequality) asso- teristics of different social groups. Those whose losses were ciated with rising unemployment, falling wages, and partially or wholly compensated by the workouts were share- spending cuts. A central element of financial management holders, depositors, and borrowers. Shareholders in Mexico was the asymmetric "pesification" of loans and deposits. are primarily wealthy individuals and families. With respect Debtors enjoyed a conversion of their pre-crisis debts at a to depositors, evidence of the distribution of deposits for rate of one peso to the dollar, while deposits were converted Mexico City in 2000 found that only 14 percent of the pop- at 1.4 pesos to the dollar; the market price of the dollar at ulation has a savings/debit account and a much smaller this time was 1.7 pesos. proportion a checking account or time deposit. Among those Depositors also faced restrictions on cash withdrawals on who have any type of account, 84 percent of deposits are bank accounts and experienced forced reprogramming of held by individuals in the top half of the distribution, with pesified time deposits. Since the peso depreciated to almost 36 percent of individuals from the top 10 percent of the 4 per dollar in 2002 (before stabilizing at closer to 3 in the distribution. first quarter of 2003), depositors suffered further capital Deposit holdings are clearly highly skewed even within losses compared to the counterfactual of an earlier conver- Mexico City, which is a relatively well-off part of the country. sion to dollars. In addition, banks suffered a large capital loss 237 INEQUALITY IN LATIN AMERICA FIGURE 8.8 Cumulative percentage change in tme deposits (by residence and sizel in Argentina December 2000-March 2001 December 2000-November 2001 21.7 *101 5.6 64 1.3 3.2 16.9 156641 I ~-5. -0. sKI I 6 I 21-0 -143 -221 152 r 230 I I 1 59 -0,9 l_o.5 1 1 ~-11.3 -21 9V -11.7 -52 r -223 -113 W2 6 - *~~~~~~~~~~~~~~~~~~~~~~~~62.2 -56.8 -945 Upto 5 5-20 20-100 100-500 Over 500 Foreign Upto 5 5-20 20-100 100-500 Over 500 Foreign residents residents In thousands of U.S. dollars In thousands of U.S. dollars m Peso deposits M Dollar deposits Source: Central Bank of Argentina and Halac and Schmuckler 2003. FIGURE 8.9 1 The evoluton of forign asset holdings in Argenbna a. Private foreign asset holdings, in billions of U.S. dollars b. Net private foreign assets, in percent of GDP 5 apital flight =12.9 billion dollars 0. 1. 93.9 102.2 97.2 94.4 9. 20.0 22.4 21.0 24.4 28.6 Dec. 2000 Mar. 01 Jun. 01 Sep.01 Dec. 01 Mar. 02 Dec. 2000 Mar. 01 Jun. 01 Sep. 01 Dec. 01 Mar.02 Source: Ministry of the Economy, Argentina. from the asymmetric pesification, but the government has (and before the currency crash and financial sector adjustments sought to finance this with a "compensation bond" that is that caused the capital losses occurred). Although withdrawals currently estimated at 14.6 billion pesos (roughly 12 per- became more widespread during 2001, small depositors con- cent of Argentina's current GDP). The story is far from over, tinued to keep money in the country in dollar deposits. however, and other mechanisms are likely to emerge, the cost The scale of capital flight amounted to about US$12.9 of which will once again be paid for in the form of either billion in the year ending in December 2001 (figure 8.9). higher taxes or lower spending. Those who got their money out experienced large capital Depositors experienced large losses in Argentina; of par- gains in terms of purchasing power in Argentina, a conse- ticular interest is the question of which ones were most quence of the steep real exchange rate depreciation. Private affected. There is striking evidence of the pattern and timing net foreign assets in the first quarter of 2002 were equiva- of changes in both peso and dollar deposits (figure 8.8). Large lent to 94 percent of GDP, up from between 20-30 percent deposit holders were moving money out of both peso and in preceding years. As was speculated above, it is possible dollar deposits between December 2000 and March 2001, well that a similar phenomenon may have been behind increases before the final unraveling of the economy in December 2001 in the financial income of the top decile during Mexico's 238 POLICIES ON MARKETS AND INSTITUTIONS Tequila crisis, as reflected in the (undoubtedly incomplete) steps have often had positive effects, the analysis here suggests data provided in household surveys.15 that such measures will not be panaceas in situations that An assessment of distributional dimensions of the finan- involve complex underlying problems and weak institutions. cial effects of crises has to rely on indirect evidence of the A striking example of such situations is Argentina, which type illustrated by these cases. A tentative conclusion is was a poster child for the international financial community that financial sector crises can lead to large, disequalizing during the 1990s. This role had a strong foundation, since redistributions. Some large financial sector participants were Argentina seemed to have broken out of its chronic insta- able to avoid losses-or even to achieve gains-by antici- bility and poor performance through a series of measures that pating the crisis, while crisis resolution mechanisms largely moved the country along a more prudent path. However, protected richer groups. problems spilled over into other directions in an example of the "seesaw" effect. In other words, when underlying What can be done? conflicts and institutional weaknesses are not resolved, the Countries in Latin America have relatively high levels of domains of action are pushed into new areas (Acemoglu, macroeconomic volatility and crisis-related risk, and there Johnson, and Robinson 2002). is evidence to support the view that fundamental reasons for Manifestations of this trend in Argentina were weak fiscal this stem from weak and unequal institutions. Many coun- control over the provinces, continued debilitating corrup- tries are on a path of lower growth and higher inequality, tion, and strained relations between the executive and the with a greater proclivity for disruptive attempts at redis- judiciary. These problems interacted with policy mistakes- tribution. Others, notably Chile, have developed strong including pegging to the wrong currency and applying mechanisms for economic management and encompassing expansionary fiscal policies during the economic boom, deep coalitions that effectively manage potential distributional structural problems (such as a de facto "dollarized" finan- conflicts (see chapter 5) and allow for relatively rapid and cial system), and a weaker international environment-to steady growth with continued high inequality. produce the deep economic and social crisis of the recent past. What does this pattern imply for policy choice? Macro- (See Perry and Serven 2002 for a review of the economic economic management clearly is not a policy with which to causes of the crisis.) facilitate the exercise of immediate "client power" over ser- Where crises are products of underlying social and polit- vice delivery. However, there should be a role for public ical structures, the long-run solution must involve changes debate that indirectly influences both policymakers and the in these domains or, as Robinson (2002) has termed it, the relationship between policymaking and actors or organiza- move toward a new "political equilibrium." Doing so would tions that manage policy directly (such as ministries of involve effective, inclusive, and accountable political and finance, central banks, and financial regulators). The dis- other types of institutions, a goal that is closely linked with cussion below organizes policy implications into four areas, tackling issues related to inequality in influence, recogni- with emphasis first placed on the role of fundamental social tion, and wealth, constraining clientelistic practices, and and political aspects, and suggestions then given for three abuses of power (as discussed in chapter 5). areas that may help bridge the apparent gulf between prox- In a good political equilibrium, everything tends to go imate and deep solutions. right. Choices of policies and specific institutional arrange- ments are of great importance, but as instruments of the Social and political fundamentals polity rather than as a means of protecting society from The diagnosis that institutions matter with regard to crises the vagaries of politicians, or "politician-proofing" (as might naturally lead to fixing institutions, in keeping with phrased in Robinson 2002). This is completely consistent the view that the world is easily amenable to such solutions. with recent work on economic performance that emphasizes Some of the measures of the past 15 years have been taken in the absence of specific, generalized policy and institutional this vein. Examples are the granting of autonomy to central solutions to economic challenges; instead, successes occur banks, policies on fiscal rules at national or local levels, greater when societies shape solutions to fit specific contexts. (See transparency in fiscal and financial accounts, and strengthen- Rodrik 2003 for a discussion on this segue.) Such solutions ing of the institutional basis of financial systems. While these are, of course, constrained by the rules of arithmetic and 239 INEQUALITY IN LATIN AMERICA fundamental economic conditions; for example, budgets have overall social and political arrangements in a society. As to be financed and prices do make a difference to behavior. O'Donnell (1999a) has argued, while broad provisioning for Chile provides an illustration of this perspective. As dis- risk was central to the process of inclusion of the working cussed in chapter 5, following its transition to democracy, classes in Europe, this same process generally didn't occur the country has been characterized by a relatively encom- in Latin America. Germany's Bismarckian welfare state was passing coalition of political and social actors and extensive a response to the distributional threat from the Marxist consensus on the strategic directions of social and economic Social Democrats. The Beveridge welfare state in the United policy. These circumstances have entailed maintaining sound Kingdom was a response to a history of mobilization by macroeconomic policy and supporting property rights and lower class groups, the unifying experience of the Second open trade in the context of inclusive and equitable policies, World War, and the shadow of the depression in the 1930s. especially in the area of public spending. Chile has been one By contrast, the weak and vertical forms of incorporation of the few countries in Latin America that has managed to of subordinate groups in Latin America (as discussed in pursue countercyclical policies during good times, and chapter 5) led to truncated forms of social risk manage- recently adopted a fiscal rule that institutionalizes this ment that divided the formal working class from outsider approach. These processes are products of a reasonably effec- groups. tive political equilibrium. Such political economy considerations imply the need for more comprehensive risk management institutions that Distributionally inclusive risk-management appeal to middle class groups as well as to the poor and that institutions provide "safety ropes" for a broad segment of the population, The importance of safety nets in mitigating the costs of crises not just a highly targeted net for the very poor. (Gelbach is a longstanding focus for policy advisors, especially in the and Pritchett 1997, Sumarto, Suryahadi, and Pritchett wake of the 1980s debt crisis (see World Bank 1990) and 2003). The mix of programs will depend on conditions in the East Asian crisis. Support for safety nets has become stan- specific countries, but will also typically involve politically dard in programs supported by the World Bank and the IME acceptable, comprehensive transfers for the extreme poor Since creating new institutions and programs during a crisis (notably those linked to human capital formation), various is a bad idea, it has been argued that such safety nets should forms of labor-based measures (from public works to unem- be developed independently of a crisis, and designed to kick ployment insurance), and measures for the young and old in automatically when crises hit. (For more on this, see de that can be expanded during times of recession. (These Ferranti and others 2000, Ferreira, Prennushi, and Ravallion aspects are discussed further in chapter 9.) 1999, and World Bank 2000b.) Putting broad-based risk management at the center of The problem is that this rarely occurs and safety nets are policymaking has an important corollary in macroeconomic usually fiscally constrained (especially during crises) and policy. The design and performance of safety nets has a inadequate relative to the problem. There are also risks that powerful influence on macroeconomic policy options such targeted programs are captured and exploited politi- (Bourguignon 2000). Designing broad-based risk manage- cally. The Trabajar (work) program, established under the ment institutions to manage distributional losses and Menem administration in Argentina, was correctly renowned conflicts will effectively free up macroeconomic policy to for its highly effective targeting of the poor (see Jalan and more efficiently respond to shocks.'6 Ravallion 1999 and 2003). However, the program also became known for local political capture, and lost political Policymaking when not in crisis: Exercising support with the change in government in 1999. This led "superprudence" in fiscal, financial, and to significant delays in building a more modified and exten- debt policies sive program, known as Jefas y Jefes de Hogares (head of The material surveyed here suggests a pessimistic conclu- household) some time after the new crisis occurred. sion on the capacity of societies in Latin America to effec- Safety nets for the poor are often treated as policy tively manage crises in an equalizing fashion. When shocks add-ons. In this regard, the issue can be seen as part of occur, the cards are stacked against discretionary choices and the incompleteness of inclusion of subordinate groups in behaviors that favor poorer groups. Rich, powerful groups 240 POLICIES ON MARKETS AND INSTITUTIONS have both more information and more influence; when displaying a countercyclical pattern) that form a healthy base all groups strive to protect their wealth and incomes, this for long-term development. influence is likely to be exercised. This problem is seen in However, history suggests that when the next upswing the initial results of patterns of adjustment and distribution occurs, private lenders will be back and borrowers (especially of workouts within the financial sector; in the relatively lim- in the private corporate sector) will be happy to receive their ited coverage of safety nets; in poor track records related to loans. There is some evidence that a structural break occurred the distributional dimensions of spending cuts; and in cross- after the Russian crisis, and that flows are unlikely to return country evidence of mechanisms to resolve distributional to the levels that prevailed in 1996-1997 for some time; in struggles. Although it is important to keep seeking mech- the long run, this is likely to be a good thing. The interac- anisms through which to make distributional resolutions tions among external debt, crisis, and distribution suggest more equitable, doing so underscores the importance of that most countries in Latin America should be more debt- developing institutions during noncrisis periods that can averse than in the past. automatically foster better responses during crises. Indeed, Latin American economies have remained exces- This statement has a possibly surprising implication. In sively vulnerable to the reversal of capital flows due to a com- public debates, adopting a less stringent macroeconomic bination of moderate to high public debt levels, excessive stance is often portrayed as a distributionally progressive reliance on foreign and dollar-linked domestic debt, low approach, whether in good times or bad. While there will export and tax ratios, weak prudential regulation and super- always be specific judgments made about the distributional vision, and procyclical fiscal policies. The implication is that impacts of a range of fiscal and monetary policy options, the the best course for macroeconomic policy in the region to analysis here suggests that taking a "superprudent" position take, from the point of view of reducing poverty and inequal- over the course of the cycle provides greater hope of sup- ity, would be twofold: increasing public sector savings and porting a more equal development pattern. On one level, this adopting superprudent regulatory and supervisory practices sharply reinforces the common prescription to break pro- in the financial system during good economic times, while cyclical policy positions. Macroeconomic restraint in good at the same time increasing the level of openness of times will facilitate automatic stabilizers and a sensible easing economies, developing long-term domestic capital markets, of policies to be applied in a disciplined fashion when adverse improving debt management, and increasing tax ratios. Since shocks occur. Building fiscal rules and institutions that help it will take time to achieve the latter objectives, the authors overcome both the political economy-related pressure to recommend the generation of cyclically-adjusted public deplete potential surpluses in good times and informational sector surpluses for a period of time for countries with high asymmetry problems-and hence improve credibility in debt burdens, in order to reduce vulnerability to shocks and countercyclical fiscal policies during downturns-therefore the likelihood of crisis. become a major priority (Perry 2003). This strategy would This recommendation implies the importance of paying in particular provide the macroeconomic foundation for attention to private sector indebtedness, balance sheet cur- broad-based, self-expanding safety nets. rency mismatches, and patterns of capital inflows. There is Countercyclical fiscal policy, while eminently desirable, also a prima facie case to be made for managing short-term has been elusive in Latin America, domestically for politi- capital inflows, although the verdict is out on whether poli- cal economy reasons and externally because of the herd-like cies such as Chile's to tax such inflows have had significant, swings of private sector sentiment (de Ferranti and others lasting effects. Technical issues on the feasibility and design 2000). In 2002 and early 2003, the region was in the down- of measures to manage capital flows are left for another swing of a private lending cycle. This sharpened macro-level report. Developing long-term capital markets in domestic pressures, forced defensive monetary strategies, and increased currency is also critical, not just for governments but for the domestic costs. Yet the retreat of private loan resources from corporate sector as well, in order to overcome the dilemma Latin America has had the positive side effect of leaving the between undertaking major currency or rollover liquidity region's external nontrade inflows dominated by foreign risks. direct investment and remittances, both of which are much A superprudent stance implies confronting distributional less volatile sources of financing (with remittances often conflicts in the tax and spending decisions that are made 241 INEQUALITY IN LATIN AMERICA during good times. The buildup of debt in the 1970s than blanket protection, whether or not the latter is explicit and 1990s in many Latin American countries-despite the or implicit. substantial privatization receipts of the latter period-helped avoid distributional struggles, as the cases of both Brazil and Distributional resolutions within crisis Argentina illustrate. Debt also allowed for expansions in It is now standard advice to protect propoor spending and social spending without confronting unaffordable entitle- increase safety net spending when crises hit. While the past ments to the nonpoor (such as public sector social security record is not good (chapter 9), moving in the direction of commitments in Brazil), and actually facilitated new broad-based risk management institutions and superpru- expansions in spending (especially at the provincial and state dence during good times will increase the probability that levels in both countries). However, this trend only shifted it will occur. On the fiscal front, more attention could be the distributional conflicts to the future. given to increased, temporary taxation on luxuries, such as In cases in which fiscal policy is limited, at least in the cars and other consumer durables. Even if the quantitative short run, the financial sector policy becomes all the more impact is not large, this action would serve as a political important. This is especially true because the evidence pre- signal of a government's efforts to share the burden of crises sented above suggests that a lot of distributional action takes more equitably, and may also facilitate negotiations in other place between the financial sector and the rest of the econ- spheres. The risk is that these measures will not be removed omy, as well as within the financial sector itself. Taking action after the crisis passes. It would therefore be wise to focus on a in good times will generally be more efficient and less costly very few, high profile items for a limited time, for example than any rescue package offered in the aftermath of a crisis, by adopting "sunset" clauses that require formal legislative and may be more politically feasible than restrictive fiscal renewal after a certain period. policy when the going is good. The analysis of the financial sector workouts in Mexico Governments can encourage market discipline and and Argentina underscores the importance of policies that improve regulation and supervision to strengthen the finan- limit adverse redistributions through pro-rich resolution cial system. This includes, among other things, a sound con- processes. This approach can be considered in terms of dis- tractual and regulatory environment, rigorous monitoring tributions both between the financial sector and the rest of that fosters risk awareness, risk-based capital requirements the economy and within the financial system. With regard (including risks arising from currency mismatches), and to the former, the major question to be considered is the costs countercyclical provisioning requirements. Banks on aver- of not bailing out. Cross-country evidence indicates that res- age create too few provisions in good times and are then olution policies increase fiscal costs without reducing output forced to increase them during cyclical downturns, thus mag- declines (Honohan and Klingebiel 2000). This implies finan- nifying losses (Laeven and Majnoni 2002). cial redistributions in favor of the relatively rich, without A good example of countercyclical provisioning require- the often-asserted gains in protecting employment. ments is the system established in late 1999 by the Bank of However, discontinuity is likely to arise if a financial Spain (Banco de Espania 1999). The Spanish regulation seeks system were to collapse, with major consequences in terms to deal with the movement of solvency risks throughout the of economic conditions.Assuming that some action is nec- economic cycle by requiring "statistical loan loss provisions," essary to avoid financial collapse, what designs make sense? in addition to both general and specific provisions for loans The literature on banking crises emphasizes the importance classified as doubtful. In other words, provisions must be set of a prompt resolution to reduce adverse effects and limit aside in order to cover transactions that have not yet been overall costs."7 A rapid and strict resolution reduces the like- specifically identified as doubtful assets, but for which expe- lihood that fragile institutions will gamble for resurrection rience shows that a risk may exist. These provisions are built and stop the flow of funds to loss-incurring borrowers.18 It up in good times and then shifted into specific areas of pro- makes sense to grant liquidity support to financial institu- vision in bad times (without passing through the income tions that do not present solvency problems, provided that statement) as the loan portfolio decays. Also important is the liquidity and solvency problems can be disentangled. Some design of deposit insurance and the desirability of having guarantees for depositors are desirable, but it is advisable that explicit, limited, and funded deposit insurance systems rather these not be generalized and that associated conditions be 242 POLICIES ON MARKETS AND INSTITUTIONS determined ex ante. (In Argentina, the blanket bailout from more liquid private banks. The legitimacy of the offered by pesification was unnecessary, even considering that pesification-and thus the ultimate effects of the crisis- the authorities would assume part of the balance sheet effect were subject to the influence over policymaking of differ- resulting from the devaluation.) Limiting a bailout reduces ent powers, such as Congress, the governors, the judiciary, both resolution costs and adverse signaling effects; unre- and large debtors.20 Although the specific impact of stricted aid can create problems of moral hazard for the political economy factors on policy measures and financial future. transfers is difficult to predict, those parties with greater With respect to transfers within the financial sector, it is power and capacities to organize are often likely to exert again important to take measures before a crisis to increase greater influence. transparency and minimize information asymmetry prob- lems, for example through good accounting and information 8.4. Conclusions disclosure standards. Once a crisis is underway, the issue To the extent that the causes and distributional consequences becomes how to intervene. Freezing or restructuring mea- of macroeconomic crises are functions of weak and unequal sures could discriminate across deposits by type and size institutions, the shift to a better equilibrium will not occur (De la Torre, Levy Yeyati, and Schmukler 2003). For exam- overnight. As a final comment, there are two corollaries of ple, during the recent crisis in Uruguay, the authorities the policy conclusions sketched above. decided to concentrate central bank reserves on fully backed First, one strand of thinking in the past two decades has demand deposits at troubled banks in order to preserve focused on the need to give greater autonomy to macro-level the functioning of the payment system. By contrast, time and financial sector management in order to shield these deposits at troubled banks were restructured by decree. policy domains from the fickle, myopic, or ill-informed pres- Such a policy approach could be complemented by sim- sures of public opinion and politicians. This is half correct. ilar measures aimed at protecting small depositors. Since There is a large domain of technical analysis in macro and large, foreign investors tend to run first, ex ante stop-loss financial management. There is also a strong case to be made clauses on the use of liquidity could be imposed to ensure for creating institutions that provide incentives for behav- that something is left to share among small depositors. The ior independent of electoral cycles, as occurs with indepen- government could also force the reprogramming of time dent central bankers with nonoverlapping terms. deposits only for deposits larger than a certain threshold However, there is also a case to be made for both greater amount, thereby giving some form of priority of claim over transparency and public debate over the consequences of available liquidity to small deposits. actions in the macro sphere, especially in the financial sector. Finally, the financing of restricted assistance to the finan- In light of the technical and indirect nature of the rela- cial sector is a question that deserves attention. Since tionships involved, there is a particular role for independent, increases in standard taxes, spending cuts, and inflation cause technical analyses of fiscal and financial policy to play in the losses to individuals outside the financial system, govern- domain of fiscal management (as illustrated by the work of ments could consider alternative financing schemes (such as the Institute for Democracy in South Africa or the Institute a capital gains tax) that could potentially shift some of of Fiscal Studies in the United Kingdom). the burden to the financial system and its high-income Second, there are implications for the rules-versus- participants. discretion issue in macroeconomic policy, as already noted Although policy tools with which to minimize financial above. Support is increasing for rules as a means of ensuring transfers can be proposed, political economy factors also play greater prudence and credibility and in order to self-discipline an important role in regulatory and resolution-related poli- the unruly child of policy management of "immature" states. cies. In the case of Mexico, for example, bank borrowers This is illustrated by the range of fiscal rules (at national and organized themselves and forced the government to reassign local levels); by the Argentine currency choice of the 1990s part of fiscal resources to support debt restructuring (see De (as well as the somewhat surreal layers of self-disciplining Luna-Martinez 2000). '9 In Argentina, the late 2001 deposit rules proposed in a desperate attempt to prevent crisis in the freeze was driven in part by the liquidity problems of state final year of the de la Rua administration); and in proposed banks, which helped the authorities to recirculate liquidity rules to protect categories of social spending. 243 INEQUALITY IN LATIN AMERICA Rules can have a positive impact when they are well- Well-designed rules have an important role to play, but are designed and the surrounding incentives, including politi- a complement rather than a substitute for the development cal incentives, are strong.21 Italy's vigorous fiscal adjustment of sound fiscal and monetary institutions and political to join the European Union's Euro club is a case in point; this behaviors that facilitate both greater prudence and fairer occurred in the context of strong domestic support for being distributional resolutions. at the core of the European project. Brazil's fiscal responsi- bility law has probably been a useful general instrument in Notes the attempt by the political center to bring state finances 1. For example, in the early 1990s the Financial Reform Index under control. Chile's Copper Stabilization Fund and the did not weight regulation and supervision sufficiently and then over- recent structural surplus rule are good examples of how to stated the extent of reform. incorporate the countercyclical elements emphasized here. 2. Thiscross-countryanalysishasusedavarietyofmeasuresofinsti- Badly designed rules, on the contrary, will have little tutional performance and has paid particular attention to exploring the value. Fiscal responsibility laws in Argentina and Peru that exogenous, or independent, influences of institutions, since it would be set rigid deficit and debt goals that were not cyclically expected that the quality of institutions would improve with general adjusted became rapidly unviable in the recessionary envi- economic development. See Acemoglu, Johnson, and Robinson (2001 ronment of 2000-2001, and were therefore abandoned. Even and 2002); Easterly and Levine (2002); Rodrik, Subramanian, and more dramatic was the experience with the Argentine Trebbi (2002); Dollar and Kraay (2002b), among other contributions Currency Board: it worked well as long as the external envi- to this debate. Kaufmann, Kraay, and Mastruzzi (2003) have argued that ronment was favorable, but once Argentina suffered major cross-country analysis indicates that although better institutions lead adverse external shocks after the Russian crisis and the to better long-term economic performance, there is no evidence that Brazilian devaluation, it led to a significant real exchange growth is an independent source of institutional improvement. overvaluation and a protracted recession, under which the 3. See Sanchez-Paramo and Schady (2002) and de Ferranti and required fiscal adjustment (that should have been made well others (2003) for multiple country analyses and syntheses. There is a before, during good times) proved to be not only difficult range of country-specific work in the region. Some examples are Cragg but even counterproductive. In the end, the rule became and Epelbaum (1996) for Mexico, Galiani and Sanguinetti (2000) for unsustainable but the costs of abandoning it were huge, as Argentina, Robbins (1994) for Chile, Robbins and Gindling (1999) well as amplified by poor crisis management. This example for Costa Rica, Santamarfa (2000) for Colombia, and Sanguinerti, shows that there are no short cuts to credibility, even less so Arim, and Pantano (2001) for Uruguay. when this is attempted through rigid rules that do not fit a 4. These arguments are based on models in the Hechsker-Ohlin given country's specific economic structure. A Currency tradition and, in particular, on the Stolper-Samuelson theorem. See Board pegged to the dollar can work well in a country that Wood (1997) for arguments in this tradition. carries out most of its trade in dollars and is closely integrated 5. It should be noted that Evans (1995) used the term "embedded" in the U.S. economy, but it implies huge risks in a closed in a different but related way, also in the context of industrial develop- economy that conducts most of its trade with the European ment. He argued that East Asian governments managed a combination Union and neighboring Brazil (Perry and Serven 2002). of some autonomy of action with embeddedness in societal structures, Further, rules work only in a context. Indonesia under including connections with the business community. The line between Suharto formally had a balanced budget rule that was often embedded autonomy and state capture is possibly a fine one. cited as a source of its relative macroeconomic prudence from 6. This section draws on the background paper by Cunningham the 1970s to the mid-1990s. This situation was, in fact, an and Santamaria (2003). illusion, since the government's domestic borrowing and 7. For a brief survey of the different generations of models of crises surpluses were hidden in the financial system. The budget see Krugman (1999) and for selected references see Aghion, Bacchetta, rule was instead endogenous to the fear of hyperinflation and and Baneriee (2001), Chang and Velasco (2001), Krugman (1979), instability among the ruling elite in the wake of the macro- Obstfeld (1996), and Velasco (1996). economic and social conflagration of the mid-1960s. The 8. While the measures used in Rodrik's analysis go some way in political powers were willing to turn over policy to the explaining Latin America's worst growth collapse, there remains an technocrats when macro-level problems became a threat. additional, negative "Latin American effect." This suggests that the 244 POLICIES ON MARKETS AND INSTITUTIONS indicators can only partially serve as proxies for the patterns of weak 17. See, for example, Sheng (1996), Dziobec and Pazarbasioglu and unequal institutions that this report has argued are so central to (1997), De Luna-Martinez (2000), and Honohan and Klingebiel the region's performance. (2000). 9. Seearangeofauthors,includingBaldacci,deMelo,andlnchauste 18. Honohan and Klingebiel (2000) provide empirical evidence (2002), Ferreira, Prennushi, and Ravallion (1999), and Manuelyan Atinc that accommodating policies such as unlimited deposit guaran- and Walton (1998). tees, open-ended liquidity support, repeated recapitalization, debtor 10. The surveys are repeated cross-sections; although representa- bailouts, and regulatory forbearance increases overall fiscal costs by tive of earnings in different parts of the distribution, they do not track a significant amount. Providing open-ended, extensive liquidity changes of the same households, as is done in panel surveys. support, for example, adds to fiscal costs 6.3 percentage points of 11. This statement is based on official government statistics. GDP. 12. Inflation and devaluation are often mechanisms for effecting real 19. Debtors also impeded reforms to the bankruptcy law that would wage adjustments; downward nominal wage flexibility is typically have made it easier for banks to take possession of guarantees and for limited or sluggish. the government to rapidly sell the acquired impaired assets. 13. This section draws in particular on Diwan (2001 and 2002). 20. As part of a long dispute, the Supreme Court sponsored a 14. See also Harrison (2002) and Ortega and Rodriguez (2002) for resolution that declared pesification unconstitutional and mandated work on determinants of labor shares over the long term, within a the redollarization of most pesified bank liabilities, significantly chang- somewhat similar overall framework as that presented here. ing the allocation of the costs of the crisis. The government responded 15. Household survey data in Argentina do not appear to be cred- with a failed attempt to impeach all of its judges. ible sources of financial income information because they show large 21. See Perry (2003) for a review of the issues and the case for rules pre-crisis declines in such income in the 1990s, which is inconsistent that have a counter-cyclical element, with the objective of reducing with the evidence provided by national accounts. political economy pressures for excessive fiscal expansion in good 16. This is consistent with the cross-section finding that greater times. social insurance is associated with lower propensity for crisis. For more on this topic, see Rodrik (1999). 245 CHAPTER 9 Taxation, Public Expenditures, and Transfers A T THE BOTTOM OF THE SCHEMATIC REPRESENTATION OF HOW HOUSEHOLD INCOMES ARE determined (introduced in figure 1.1 in chapter 1 and further developed in box 6.1 in chapter 6), there is a clear link between the primary distribution of income, which is determined by market rewards to the private assets and efforts of individuals, and the secondary distribution, which refers to the income available for household consumption and savings. That link, which is described simply as "redistribution," refers to the fact that in virtually all modern societies, the state takes some income away from households through various types of taxes and gives some income back to households through various types of transfers. After considering policy options in the realms of institutional reform and asset redistribution in chapters 7 and 8, chapter 9 now focuses on inequality-reducing initiatives that are suitable for Latin America, in the areas of taxation, general public expenditures, and, in particular, income transfers. States levy taxes in order to finance various responsibilities, of which-following Musgrave's (1959) classic taxonomy-there are three basic types: stabilization, allocation, and distribution. The first refers to the countercyclical role that government can play with respect to the business cycle, which leads to dampened fluctuation in economic activity and consumption. If households are risk-averse, this result improves welfare. The second refers to the provision of public goods and the may or may not be efficient, but it is generally not distribu- possible internalization of externalities. The third is derived rionally neutral. It is now widely accepted that even the uni- from the long-held notion that societies need not necessar- versal public provision of goods and services in kind can have ily accept the primary distribution of income as final and a progressive incidence, if quality is rationed and the richer may, through a collective decisionmaking process, seek to members of society opt our of the public system and purchase redistribute income in order to reach a more desirable dis- higher-quality substitutes privately (Besley and Coate 1991). tribution of welfare.' To provide a benchmark of the potential for redistribu- For a variety of reasons, tax revenues are used to finance a tive action by the state over the long run, table 9.1 provides wide range of public actions across countries. Such reasons estimates of income inequality before and after taxes and range from the historical to the political and include differ- transfers for a set of industrialized societies. There are likely ent levels of administrative capacity, the nature of public to be large standard errors around estimates such as these goods, and the quality of information about individual (see the discussion of tax incidence in section 9.1), and the incomes available to the government. Basic roles, such as capacity of the state in rich countries to pursue redistribu- national defense and the enforcement of law and order, are tion is much greater than in Latin America. Yet these dif- financed by public revenues virtually everywhere. However, ferences are nonetheless significant, and show that very equal a number of private goods, such as education and health ser- societies (for example in Scandinavia) are much less equal vices, are also publicly provided in kind in most nations. This pretax than after tax. 247 INEQUALITY IN LATIN AMERICA TABLE 9.1 expenditures, and on conditional cash transfers-a new and Income inequality before and after taxes and transfers in promising variety-in particular. selected industrialized countries 9.1. Taxes and distribution Gini coefficient for Gini coefficient for In terms of the size of government (or aggregate fiscal posi- income inequality income inequality before taxes and after taxes and tion), as is the case with many other dimensions, there is con- transfers transfers siderable variation across Latin America. In table 9.2, Australia 46.3 30.6 countries are organized into a simplified typology along two Belgium 52.7 27.2 dimensions, the aggregate engagement of the state in the Denmark 42.0 21.7 economy (where spending shares equal to government rev- Finland 39.2 23.1 Germany 43.6 28.2 enues plus government deficit serve as a proxy) and the share Italy 51.0 34.5 of total public spending devoted to social sectors (includ- Japan 34.0 26.5 Netherlands 42.1 25.3 ing social security). Sweden 48.7 23.0 Large differences in social spending among countries are United States 45.5 34.4 apparent. Argentina, Brazil, and Uruguay have high levels of social spending as a share of gross national product (GNP) Source: OECD as reported in Burniaux and others 1998. because they have high spending levels overall and devote large shares to social programs. As a result, social spending What is the evidence of the impact of taxation and social accounts for more than 20 percent of national income. With spending on inequality in Latin America? Has inequality, somewhat different configurations of overall spending and in terms of access to services, changed for the better? Are social shares, Chile, Costa Rica, and Panama also have levels there specific programs and financing efforts that hold sig- of social spending of 16 percent or more. So does Bolivia nificant promise for redistribution, while at the same time despite its low income, which is a notable result of both rad- enhancing growth opportunities or at least having low effi- ical shifts in spending priorities during the 1990s and sub- ciency costs? This chapter attempts to answer these ques- stantial donor support. At the other extreme are the tions, using the information that is available. It is organized Dominican Republic, El Salvador, and Peru, which have both as follows. Section 9.1 looks at the level and composition of low aggregate spending levels and low shares of social spend- taxation; section 9.2 considers the structure and incidence ing, at less than 7 percent of national income. Guatemala, of the social components of public spending; and section 9.3 Honduras, and Paraguay also have notably low social spend- focuses more narrowly on cash transfers as an element of those ing because of varying mixes of relatively low aggregate TABLE 9.2 Typolgy of Latin American countries by fiscal position and social spending, 1998 Fiscal priority: Share of total public spending going to social sectors Expenditures as share of GNP Less than 40 percent Between 40 and 60 percent More than 60 percent More than 30 percent Nicaragua (12.7) Costa Rica (16.8) Argentina (20.5) Colombia (15.0) Brazil (21.0) Panama (19.4) Uruguay (22.8) Between 20 and 30 percent Honduras (7.4) Bolivia (16.1) Chile (16.0) Venezuela (8.6) Less than 20 percent El Salvador (4.3) Guatemala (6.2) Dominican Republic (6.6) Mexico (9.1) Peru (6.8) Paraguay (7.4) Note: The numbers in parentheses correspond to the percentage share of social spending in a country's gross national product (GNP). Source: ECLAC 2001. 248 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS spending and low shares of social spending. Although these wealth. This view is the basis for the current preference for patterns are influenced by the overall income level of the a broad-based consumption tax, with few exemptions and economy, this is only part of the story. For example, Mexico selected higher tax rates for items that either carry negative is an upper-middle-income country with unusually low taxes externalities or are considered "bads" by society (such as gaso- and spending, while Nicaragua is a low-income society with line, tobacco, and alcohol). It is also now believed that a high aggregate tax-spending position. income tax bases should be as broad as possible and treat This variance in terms of both overall taxing and spend- all incomes, no matter from what source, as uniformly as ing and of the composition of spending reflects different possible. sociopolitical histories, tax collection capacities, and reactions Moreover, the decline of taxes on international trade due to changes in thinking about tax and development policy to unilateral liberalization and the multilateral rounds at the across the continent. Ideas about the role of the state in the World Trade Organization (WTO), as well as increased com- economy have changed considerably, even during the last half petition for foreign investment, have moved international century, both in Latin America and elsewhere. concerns from the bottom to the top of the tax policy action In the 1950s and 1960s, tax policy discussions in Latin list in many countries. This in turn has led to reduced tax rates America reflected the "developmentalist" views common in related to trade (which strengthens the trend toward the use the postwar era and the emerging reformist approach of the of value added taxes, or VATs) and has put downward pressure largely U.S.-trained advisers and consultants financed by the on tax rates for internationally mobile factors of production, Alliance for Progress and the United States Agency for Inter- notably capital and highly skilled labor. At the same time, in national Development (USAID). The two main aims of tax- many countries the fiscal implications of decentralization have ation were to raise revenue-and lots of it-in order to become a major issue (IDB 1998, World Bank 2000b). finance the state as the "engine of development" and to redis- Reflecting this "new view," income tax rates on both tribute income and wealth. Both goals could, it was gener- people and corporations were cut sharply and are now almost ally thought, be achieved largely by imposing high effective universally in the 20-30 percent range throughout Latin tax rates on income. The depressing effects of taxes on invest- America (down from around 50 and 40 percent for personal ment and saving were judged to be small. and corporate income tax, respectively, in the mid-1980s), Indeed, it was sometimes argued that the extra bonus of as is the case elsewhere in the world (Shome 1999). Figure 9.1 high tax rates is that they make it easier to lead private shows the decline in the highest personal and corporate investors by the very visible hand of well-designed fiscal income tax rates between the 1980s and the present. This incentives into those channels most needed for development decline was accompanied by an increase in the personal purposes. Most analysts took it for granted that a highly pro- exemption rate and reductions in the income bracket to which gressive personal income tax (with marginal rates ranging top rates applied. Trade tariffs were also drastically cut, from as high as 60 percent) buttressed by a substantial corporate an average of about 40 percent to around 11 percent between income tax (often at 40-50 percent) constituted something the mid-1980s and the mid-1990s, and export taxes were close to an "ideal" tax system. Consumption taxes-which eliminated altogether.2 On the other hand, the VAT is now at the time consisted mainly of excises, custom duties, and seen as the mainstay of the revenue system in the region, cascading manufacturer sales taxes-were grudgingly reflecting-or even anticipating-global trends (Ebrill and accepted as necessary for revenue purposes. No one talked others 2001). Figure 9.2 details the pattern of VAT rates, about local taxes, since almost all the action took place at which rose on average from 10 percent when first introduced the central government level. In addition, no one seems to (in various years) to about 15 percent in 2001. have worried much about the international context, and tax This combination of falling income tax rates and trade policy was almost invariably considered to be a largely tariffs and rising indirect domestic taxes yielded modest domestic affair. increases in overall tax revenues across the region. Between Views on the appropriate role and structure of taxation the 1970s and 2000, the unweighted average of aggregate began to change in the 1970s and 1980s. Today, most tax effort rose by about 2 percentage points of gross domes- economists and policymakers consider that high tax rates not tic product (GDP) to 14 percent of national income (fig- only significantly discourage and distort economic activity ure 9.3). Taxes on goods and services (mostly VATs) were but are largely ineffective in redistributing income and the primary source of this increase, more than offsetting 249 INEQUALITY IN LATIN AMERICA FIGURE 9.1 Highest personal and enterprise income tax rates in Latin Amnerca a. Highest personal income tax rates Percent of taxable income Percent of taxable income 70 60- l 60- 50 - 50 I 4 40 I 3~~~~~~~~~~~~~~~~~~0- o0~~~~~~~~~~~~~~~~3 - 30 20 ~~~~~~~~~~~~~~~~~~~~20- w 1992 data instead of 1986.. b. Highest enterprise income tax rates Percent of taxable income Percent of taxable income 70 4 60 - 140 t.50 30 ' 20i 151 o - ~~~~~~~~~~~~~~~~~~~0 'g .< e,#4 o .................................. ,3 Unweighted regional average el' \O~ eg O bA?qe 1992 data i d o 61986 2001or2002 Source: Stotsky and WoldeMariam 2002. declines in taxes on trade. Income tax collections rose is the much higher take from social security taxes in richer slightly, in spite of the reduction in rates. countries, which is associated with much more developed Despite this increase in the overall level of taxation, Latin welfare states. Developed societies also tax trade less and get America still taxes a considerably smaller share of total more from VAT and sales taxes. Property taxes are low in output than do richer countries. Table 9.3 compares the total developed countries and even lower in Latin America, revenue and its composition in Latin America at two points although these results are probably particularly biased by the in time, with the relevant averages for developed countries, exclusion of local government revenues. The most striking feature of the comparison with richer The differences in the levels of taxation between Latin societies is the much lower take from income taxes. This is American and richer countries are in part due to what econ- especially due to low levels of personal (as opposed to cor- omists call Wagner's Law: the observation that government porate) incomes taxes, which account for less than 1 percent expenditures as a share of GDP tend to rise with per capita of GDP in Latin America compared with more than 7 per- incomes. However, as figure 9.4 illustrates, Latin American cent in developed countries. The other significant difference tax efforts tend to lie below the average even in relation to 250 _~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~3 - TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS FIGURE 9.2 Value-added tax rate (in Latin America) Percent Percent 25 '16 14- 20 12 *~ ~ ~ a.& ..(; i ,2 'B, Unweighted regional average 0.70 - 0.40 - ow,L~ ~~~'~ ,,,,,,,,[,,,,,,a 0, 0 - 0~~ ,,vx~ ~~~~~d jo> Nb r g rA Unweighted regional average E3 At introducton m20 VAT revenue productivity Percent Percent 0.70 -0.45- 0.60 0.4 0.35- 0.50 -0.30- 0.40 -0.25- 0.30 0.20- 0.20 ,ilfT {l-FilfI i.__ __ 0.10- 0.05 0.00* ___ ___ _ io leve, wh te e ion of xP Nag and ba cang) a Unweighted regional average 0s0 IZZ VAT revenue productivity (latest year) Source: Stotsky and WoldeMariam 2002. income levels, with the exception of Nicaragua and Uruguay change) and in Brazil (in the early 1990s). There is no pattern (Brazil is also close to the norm). This is particularly true among moderate-tax countries, and only modest increases with regard to personal income and property taxes collec- in tax ratios in the low-tax countries. tion, as well as social security contributions (the latter reflect- Before turning to the possible policy alternatives ing the low coverage of social security). which might be considered in the area of tax policy, it is In terms of comparisons across Latin America, there is important to understand the incidence of taxation in Latin considerable diversity in tax efforts and tax composition, America. but also a relatively high degree of consistency over time. Countries that had relatively high tax ratios at the end of Who pays the taxes? the 1970s were still above the regional average in the 1990s. The question of who pays taxes is surprisingly hard to answer, Among relatively high tax countries, the major changes despite decades of economic work on taxation. This is occurred in Nicaragua (in the early 1980s, following a regime essentially because there can be major gaps between the 251 INEQUALITY IN LATIN AMERICA FIGURE 9 3 Unweighted average of tax revenue and taxes per category in Latn America Percent of GDP 16- | O 1970s EM 1980s M 1990s m 2000 O Total 14- 12 - 10 8- 6- 4- 2 -Tj_ 0 - I Tax/GDP Tax on Income/GDP Tax on goods and Tax on trade/GDP Other taxes/GDP Social security services/GDP contributions/GDP Source: World Development Indicators Database. intention of tax systems embodied in statutory tax structures and actual payments. Firms and individuals respond to taxes, TABLE 9.3 TAxL structure in the 1990s, Latin Amenca and developedadjusting wages and prices when they can, to reduce taxes Tax trutur in he 990, Lain merca ad dvelpedon themselves and shift the burden on to others. Most tax countries (in percent of GDP, consolidated central government) incidence studies resort to highly simplified assumptions Developed about such responses (see box 9.1), so their results must be Latin America and the Caribbean countries taken with due caution. Tax category 1990-94 1995-99 1991-2000 A recent review of quantitative studies of tax incidence in developing countries (Chu, Davoodi, and Gupta 2000) finds Income taxes 3.6 3.4 9.7 that taxes generally have little redistributive effect in Latin Individual 0.5 0.9 7.1 Corporate 1.9 1.7 2.3 America, largely because of heavy dependence on indirect Social security 2.5 2.9 7.8 taxes in most countries.3 The authors of the study find that Taxes on goods 5.6 7.4 9.5 and services tax structure is the most important factor in determining this VAT and sales 3.2 4.8 6.5 outcome, since personal income taxes are assumed to be basi- Excises 2.1 2.3 3.0 cally progressive and consumption taxes are not. Most stud- Trade taxes 2.2 1.8 0.3 Imports 1.9 1.8 0.3 ies assume that personal income taxes and specific taxes on Exports 0.1 0.0 0.0 'luxury" items are progressive, property taxes marginally pro- Property taxes 0.4 0.3 0.8 All taxes 14.2 16.1 28.7 gressive, and the rest usually regressive (although exemptions of basic foodstuffs and other items that have high weights in Source: Stotsky and WoldeMariam 2002 and Government the consumption baskets of the poor may render sales or VAT Finance Statistics database. taxes close to neutral or even mildly progressive). 252 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS FIGuRE 9 4 A simple cross-country regression of total tax revenues against log GDP per capita Tax revenue as a share of GDP 0.45 0.40 - Italy ' France 0.35 - 0.30 * Spain 0.25 - Nicaragua *Uruay 0.20 - * * Costa Rica " United States 0.15 - eru xico * * * *tIominicarRepublic 0.10 * Colombia 0 * Argentina Guatemala E 0.05 0.00* 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 Log GDP per capita (in constant 1995 U.S. dollars) Average 1990-2000 Source: Authors' calculations from World Development Indicators Database. BOX 9.1 Do tax lncidenee numbens much? Ir is not clear that the evidence provided even by such attempt to pass on (shifr) raxes to someone else whenever careful incidence studies as that by Engel, Galetovic, and they can, including to consumers rhrough higher product Raddatz (1998) is of much value. Starements such as "most prices, to workers through lower wages, and to owners of the burden of the tax system falls on the bottom rwo of land through lower land prices. MIost tax incidence income quitiles are based on a series of assumptions analyses apply a parial equilibrium approach based on about the incidence of different taxes rhat are invariably assumptions of demand and supply elasticiEies (actual subject to considerable dispute (Whalley 1984).' In addi- measures of such elasticities are rare in tax incidence rion, in practice-though not in most analyses-these analysis). statements depend upon many facrors specific to a partic- Depending on these assumptions. taxes are thus ular country at a particular time, for instance wirh respect assumed to be shifted predominantly forward (if demand to market srructures and macroeconomic conditions.5 is assumed to be less elastic than supply) or backward. In What really matters is of course the economic and not some cases, assumptions are even cruder: personal income the legal incidence of taxation. As a general rule, the eco- taxes are not thought to be shifted ar all (hence most nomic incidenceofa tax(at least in the long run)isdeter- studies estimate a progressive effect)and sales rax and VAT mined by market conditions and nor by whether the tax are assumed ro be shifted wholly to consumers. As a result, is legally imposed on the buyer or rhe seller. Businesses most studies estimate a slightly regressive effect chat is 253 INEQUALITY IN LATIN AMERICA normally tempered by the exemption of basic foodstuffs economic incidence (Shah and Whalley 1990), the specific and other consumprion items (such as clothing, trans- characteristics of each country need to be taken into portation, and housing) that weigh heavily in the con- account when evaluating the economic incidence of each sumprion baskets of the poor. tax. Thus, one might attribute all taxes on corporations Some studies artempt to go beyond this by using com- to owners in the highest income deciles if low capiral putable general equilibrium models (CGEs) It would be mobiliry is assumed, or to unskilled workers if perfect expected that relative prices, output composition and level, international capital mobility is assumed and technological change would be altered wich a differ- This argument may seem unduly nihilisric. The key ent tax system. One result is that most of rhe rime the exer- point abour such incidence srudies is simply that they do cise of comparing the incidence ofa real tax structure with nor matter much. Policy questions in raxation are almost an imaginar) counterfacrual is fundamentally flawed, since never related to the question "who pays how much?" under doing so holds constanr what simply cannot be held con- the present system; rather, rhe relevant policy question is stant (Meerman 1980). However, tax incidence analysis almosr always "who will pay more (or less) if this or that with CGEs is still plagued by a variety of assumptions. particular change is made?" If the change is small enough, (These include the model "closure" and the choice of coun- it will be possible to assume that relative prices will not terfactual, both of which are normally crirical to the end change and to carry out a simple partial-equilibrium analy- result, most exercises use static CGE models, in which sis. Usually, however, some form of general equilibrium effects on investment and technological change are either analysis. with all the atrendant sensitivity to parameter dismissed or exrremely simplified, and many parameters, choices, clearing mechanisms, and the like will be including elasticities, are not measured but assumed.) necessary. (For an excellenr recent example, see the study To illustrate the difficulties inherent in incidence analy- of Colombia by Rutherford, Lighr, and Barrera 2002.) sis, consider the incidence of corporate income taxes (which Whichever method is used, ir is always and everywhere make up rhe bulk of income raxes in most Latin both more meaningful and more policy-relevant to analyze American counrries) Incidence will depend on such fac- rhe incidence of taxation "at the margin." rather rhan "on tors as the openness of the overall economy; inflows and the average." This is essentially the same point as made by outflows of capital investment and the extent to which for example, Ahmad and Stem (1991) with respect to the capital moves berween rhe corporate and unincorporared grearer ease and meaningfulness of analyzing 'optimal rax sectors; the relative capital-intensity of corporations; the reform" rather rhan "optimal tax policy" per se. The elasticity of demand for goods produced by corporarions assumptions needed ro do so are less heroic, more testable, and other businesses; and the elasticiry of supply of inputs and more believable, and rhe results may help answer the (most importantly, labor). None of these factors is easy to crirical question of the distributional impacr of proposed measure, and thus neither is the incidence of the corpo- policy reforms. Despite all the problems of incidence rate income tax. Furthermore, since differing economic analysis, it remains important to do the best one can, and conditions in different countries mean that taxes that have at the same time not to overstate what is possible to the same legal incidence may have a quite different accomplish. Given the low importance of personal income taxes last three decades. A comparative recent study by Gemmel and property taxes in Latin American countries, the direct and Morrissey (2002) also found that tax systems in the redistributive leverage of the tax system in most countries region ranged from slightly progressive to slightly regres- should be expected to be very small or even negative. It is sive. Their review of recent individual studies shows simi- therefore not surprising that the Chu, Davoodi, and Gupta lar conclusions (see table 9.4). (2000) study, like the much earlier work of Bird and De Wulf In one of the more careful studies conducted, Engel, (1973), found little evidence of much fiscal redistribution Galerovic, and Raddatz (1998) find Chile's taxes to be slightly through taxes in Latin America, despite the sometimes regressive (with a pretax Gini coefficient of 0.488 rising to wrenching tax changes undergone by many countries in the 0.496 after taxes). This was the case despite the fact that 254 TAXATION, PUBLIC, EXPENDITURES, AND TRANSFERS TABLE 9.4 Tax incidence studies for Latin American countries, 1975-98 Households, individuals Taxes Measures of Author's Author/s Country Period Income Concept covered Included progressivity conclusions Foxley (1979) Chile 1969 Household income All Direct & indirect Not indicated Semi-proportional and expenditure Engel, Galetovic, Chile 1996 Household income All Direct & indirect Gini coeff. Slightly regressive Raddatz (1998) and expenditure Gillis, McLure (1978) Colombia 1974 National income High inc. Direct & indirect Not indicated Progressive Berry, Soligo (1980) Colombia 1970 Household income All Direct & indirect Not indicated Substantial and expenditure progressivity Santana, Rathe Dominican 1989 Household income All Direct & indirect Not indicated A degree of (1993) Rep. progressivity Galper, Ramos (1993) Guatemala 1992 Household income All Overall tax system Not indicated Progressive Bahl, Martinez- Guatemala 1994 Household income All Overall tax system Not indicated Progressive Vazquez,Wallace (est.) (1996) Hicks,Lee (1997) Guatemala 1992 Household income All Direct & indirect Not indicated Mildly progressive and expenditure Gil-Diaz (1984) Mexico 1972, Permanent All Direct, indirect, Not indicated Progressive, but 1980 household income and inflation uneven in 1972, smoother in 1990 Escobal etaL. (1993) Peru 1985-90 Household All Fueltaxes Not indicated Gasoline tax is expenditure progressive; kerosene tax is regressive Source: Chu, Davoodi, and Gupta 2000. Chile's tax system is the most effective in Latin America,6 The empirical difficulties that plague average tax inci- collects the most from the personal income tax, and has the dence studies (highlighted in box 9.1) may seem somewhat highest marginal rates (up to 45 percent, compared to 25-30 dispiriting. However, there is also a somewhat more opti- percent in most countries). The authors of that study also find mistic message contained in those arguments: in designing that if the income tax were made much more progressive by tax reforms, there is less concern with levels than with eliminating all tax allowances and all (estimated) underre- changes, since it is possible to be more confident about the porting of income, the Gini coefficient would fall to only marginal effects of such changes. Although the dominant 0.484 and the ratio of incomes between the top and bottom influence of taxes on distribution is probably their contri- deciles would remain virtually unchanged at 13.4. Engel, bution to the overall redistributive capacities of a government, Galetovic, and Raddatz (1998) also argue that the more there is almost certainly some potential to make tax systems unequal the pretax distribution, the greater the distortion somewhat more progressive, although the extent to which this costs and the less the redistributive effect of progressive tax- is possible will depend on issues of overall political and social ation. They conclude that the most important factor affect- consensus as much as on the details of tax instruments. ing the distributional impact of the tax system is how much revenue it generates for potential redistribution by the public Implications for tax policy sector. Estimates from developed countries find diverse So how can taxes be made somewhat more progressive in results. While the figures given in table 9.1 were not broken Latin America, without excessively compromising the down between taxes and transfers, in individual country stud- gains in economic efficiency that have been attained over the ies the estimates of this division range from income taxes last 30 years through the reduction in tax rates, broadening accounting for the bulk of redistribution in Denmark to very of tax bases, and other reforms? It should first be recognized little in the United Kingdom; Canada falls in between, with that tax design is a policy realm in which efficiency-equity a third of redistribution being attributable to taxes.7 tradeoffs are unlikely to disappear. In this regard, there really 255 INEQUALITY IN LATIN AMERICA is no such thing as a free lunch, and it is likely that efforts obviously regressive excise taxes (as well as import to make taxes more equitable will, in most cases, have some duties), for example on food items or kerosene, is efficiency costs. The principles outlined below are therefore another obvious measure. It should also be remem- intended only as suggestions of aspects of tax policy through bered that the incidence of taxes on alcohol and tobacco which it may be possible to make taxes marginally more is very variable, and in some instances a case can be progressive at a relatively low cost. They are only general made on equity grounds for moderating efficiency- principles; in every case a concrete recommendation can only driven rate increases. To make up for lost revenue from be made after the country-specific context has been prop- any of these measures, there are often good reasons for erly evaluated. Bearing these caveats in mind, the authors considering higher levels of taxation on private trans- propose six broad principles for equity-enhancing tax reforms port. Export taxes, on the other hand, are generally best in Latin America, as follows. avoided both on efficiency and on equity grounds. (4) There is scope for raising personal income tax collections. (1) Tax bases should be as broad as possible. A broad-based Collections frompersonal income taxes are low in Latin consumption tax, for example, will still discourage America, even when compared with countries with the work effort, but choices between tradable and non- same level of income (figure 9.5). This suggests that tradable goods and services will not be altered if all there is scope for increasing such revenues, ofcoursepro- are taxed.8 A few items (such as gasoline, tobacco vided that increased revenues finance socially useful, and products, and alcohol) may be chosen for differentially hopefully progressive, expenditures. The experience of higher taxation rates for regulatory reasons or because Chile shows that this is indeed possible when there is the demand for these products is relatively unre- political will and good administration. In light ofpoint sponsive to taxation. As a result, at any given tax rate, (2) above, higher revenues should be sought first by efficiency costs will be relatively low and revenues rel- closing loopholes and enforcing greater compliance, and atively high. Income tax bases should also be as broad only later through higher marginal rates. Income taxes as possible and treat all incomes, no matter from what need to apply both to persons and to corporations. To source, as uniformly as possible. keep tax avoidance in check, it would be wise to keep (2) Tax rates should be as low as possible, provided they raise the top marginal rate of the personal income tax fairly sufficient revenue to finance the appropriate expendi- close to the rate of the corporate income tax, which tures of government. Of course, the broader the base, means that it is not likely to be all that high. the lower the rate needed to generate a given revenue (5) Property taxes are currently underutilized and should be made level. However, lower rates are worthwhile in their own to generate more revenue. Property taxes at present account right because the efficiency cost of taxes arises from for only about 0.3 percent of GDP for the region as a their effect on relative prices, and the size of this effect whole (Stotsky and WoldeMariam 2002). Collections is directly related to the tax rate. In fact, the general from property taxes are low in Latin America even rule is that the distorting effect of taxes increases pro- when compared with countries with the same level of portionally to the square of the tax rate, so that dou- income. Coverage of property taxes is not compre- bling the rate of a tax implies a four-fold increase in hensive and both assessments and collection rates are its efficiency costs. From an efficiency perspective, it often low. Although nominal rates are also low, gov- is therefore always better to raise revenue by impos- ernments usually find rate increases in this very visi- ing a single rate on a broad base of taxpayers, rather ble tax difficult to sell politically. Simply raising the than dividing that base into segments and imposing legal tax rate would in any case usually burden only differential rates on each one. In practice, of course, this the few from whom taxes are actually collected. consideration needs to be balanced against the equity Increased nominal rates are likely to be acceptable only argument for imposing graduated rate schedules. along with improvements in tax administration, such (3) Indirect taxes need not be regressive. In particular, VATs- as more comprehensive coverage, better and more fre- which are generally preferable to excise or import quently evaluated assessments, and enforced penalties taxes-can often be made less regressive with a few for late payment. (For a recent detailed discussion of key exemptions.9 Similarly, reducing or removing land and property tax policy, see Bird and Slack 2002.) 256 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS FIGUIRE 9 5 The relationship between personal income taxes and income Revenue as a share of GDP 0.14 . 0.12 Italy e 0.10 0 * United States 0.08 * 0.06 - 0.04 - 0.02 Dom ePuv vC, Peru Costa Ric Uugua y 0 - ~~~~~~~~~~~~~Colonnbia Partna o Brazil * Argentina 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 -0.02 - Log GDP per capita (in constant 1995 U.S. dollars) Average 1990-2000 i 0.04 - Source: Authors' calculations on data from World Development Indicators Database and IMF Government Finance Statistics database. (6) In many instances, the most important taxes affecting less and to tax the middle class both more fairly and more the poor are not formal ones levied in the government efficiently. budget, but rather implicit taxes, which include bribes and exactions of various kinds (see Prud'homme 1990), 9.2. Public social spending and distribution as well as the inflation tax. As explored at length in The point of taxing, as indicated above, is to raise resources Bird (1991), implicit taxes also include many instances that can be spent well. This chapter now turns to the evo- of "regulation as taxation" such as quasi-taxes imposed lution, composition, and incidence of public spending in through controls on trade, prices, credit, foreign Latin America, focusing on what is known as "social spend- exchange, or capital markets. ing." Although the exact definitions of this spending cat- egory vary from country to country, the term generally The authors recognize, of course, that raising income and includes expenditures on social security, education, health, property tax revenues, as well as rebalancing the composi- social protection, social assistance, and active labor market tion of indirect taxes so that they shift away from goods con- programs. In some cases, housing subsidies and expenditures sumed relatively more intensively by the poor and toward, on basic urban services and infrastructure (for example, say, private transport, are highly charged political issues. water and sanitation) are also included. First considered are Ultimately, just how much the rich are taxed remains a overall trends in social spending, followed by the distrib- domestic political choice. The opening of capital markets, utional aspects of access to various programs. The section combined with weak international tax enforcement capa- concludes with a brief examination of the effects of economic bilities, does make it harder to tax the rich heavily. How- cycles. ever, there is no doubt that if they wanted to, most Latin American countries could do more in this respect, as Chile's Changes in public social spending In the 1990s experience with personal income taxes shows. Similarly, According to an analysis of 17 Latin American countries most countries could make a greater effort to tax the poor conducted by the United Nations Economic Commission for 257 FIGURE 9.6 Public social spending in Latin America: expenditures per capita, as shame of public spending and GNP, 1990s a. Social public expenditures b. Social public expenditures as a c. Social public expenditures per capita (in US$ 1997) share of total public spending as a share of GNP Argentina11 1687 Uruguay 62 25Uruguay 18 2. 1539 66.8 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~21 Uruguay8813 Chile 0. Brazil 1. 888 60.8~~~~~______________________2. Br.:i ~ ~ 111Argentina6. Argentina 1. 827 ~~~~~~~~~~~~~~60.4 __________________19.4 Chile 87Brazil 49Panama 1. Panama 47642 Meio58.5 Costa Rica 16.8 M ic 40.8 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1.7 622 56.5 1. Costa Rica 46Bol ivia 49. Boivi 16.1 402 46.2 165 Colombia 1831Guatemala 29.9 462Colombia RB de 39 aRa43.1 1. Venezuela 37 Cot ia3.9 Nicaragua Peru 68 192 ~~~~~~~~~ ~ ~~~~Domincn 39.791 cc Peru ~~~~~~~~~~~~~~~~~~eubi 8.4 eic Blva 168 38.6 RB de . Bolivia ~~~~~~~~~~~ Panama ______ 4.0 ~~~~~~~~~~~~~~~~~~~Venezuela9 Dominican 135 Peru 31138.3 Paraguay 3. Republic 6 132 ~~~~~~~~~~~~~~RB de 373 odua Paraguay 563. Hnua 56 ~~~~~~~~~~Venezuela 349 Guatemala 17Nicaragu 35.4 82 Clm35.5 Dominican 6.6 El Salvador 6Combiai88 euli . 574362 Honduras H60du6.5 Guatemala6. 57 El Salvador 27 El Salvador Nicaragua 48 21.3 Latin America 540 Latin America ~~~~~~~~47.8 Latin America Z 11 540 Latin America~~~~~~~ 1 2 3 41.8 6 0 00 5 10 i 0 2 I I I I~~N I I; I d 1 ~~~~~~~~~~~~ ~~~~~~~~~~~Percent Percent U.S. dollars 101998-99m1901 Source: Authors' calculations based on microdata from household surveys. TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS TABLE 9.5 Increase in social spending with and without social secuity, from ECLAC and IMF databases Social spending per capita Social spending as share of Social spending as (US$1995) total spending (%) share of GDP (%) With social security ECLAC 1990-91 360 41.8 10.4 ECLAC 1998-99 540 47.8 13.1 IMF GFS 1991 369 48.7 8.5 IMF GFS 1998 549 54.8 11.0 Without social security ECLAC 1990-91 212 30.5 6.5 ECLAC 1998-99 304 33.1 8.0 IMF GFS 1991 218 27.4 4.7 IMF GFS 1998 289 28.7 5.7 Source: ECLAC 2001 and authors' own estimates from the IMF Government Finance Statistics database for eight countries for which data are available. Latin America and the Caribbean (ECLAC) (2001), annual (table 9.5). According to the ECLAC (2001) study, when social public spending per capita increased on average by social security is excluded from social spending, the increase 50 percent between 1990-1991 and 1998-1999, from the as a share of GDP is only 1.5 percentage points, versus 2.7 equivalent of US$360 to US$540 per capita at 1997 prices percentage points when social security is included. Estimates (figure 9.6).1 Although the exact magnitude of the increase made for this report, based on the International Monetary in public social spending differed among the 17 countries, Fund's (IMF) Government Finance Statistics (GFS) data- the increase was fairly general. It was greatest in the coun- base, provide results which are fairly close to those in the tries with initially low and medium levels of per capita social ECLAC study, with an even smaller gain in social spending expenditure, notably Colombia, the Dominican Republic, during the 1990s when social security (and welfare spend- Guatemala, Paraguay, and Peru, where spending per capita ing according to the IMF classification) is excluded from the more than doubled. In countries with relatively high initial analysis (see http://wwwl.worldbank.org/publicsector/pe/ levels of spending, specifically Argentina, Brazil, Costa Rica, crosscountrydata.htm). and Panama, the increase was approximately 30 percent over Despite a common trend toward higher public spending the course of the decade. on social sectors over time, there are still large differences Three factors explain the increase in social public spend- in spending levels among countries. Countries with high ing per capita during the 1990s: economic growth in GDP levels of GDP per capita tend to have higher levels of social per capita (which accounted for some 40 percent of the gain); spending as a share of GDP (in part due to higher social secu- increased total spending as a share of GDP; and shifts in the rity outlays). In addition, in many countries the increased composition of spending to benefit the social sector (the share share of social spending has taken place at the expense of of which grew from 42 to 48 percent). While the growth public infrastructure investment, with the likelihood of effect was particularly important in the first half of the adverse effects on long-term growth and inequality (see decade, the impact of shifts in spending toward the social chapter 7). While in some countries this shift was a delib- sector mattered more during the second half. This change erate consequence of higher private sector involvement in allowed most countries to increase social spending per capita, infrastructure, in others a "crowding out" effect of infra- even though economies were not doing well. structure investment by higher pensions and other social Social spending as a percent of GDP grew from 10.4 in expenditures occurred during the 1990s (see various chap- 1991-1992 to 13.1 in 1998-1999. However, a dispropor- ters in Easterly and Serven 2003). tionate portion of the increase in the ratio of social expen- Why didn't large increases in social spending translate ditures to GDP and in the share of social spending in total into falling income inequality? There are two reasons. First, expenditures was due to rapidly increasing pension outlays social spending is not necessarily progressive (and can be 259 INEQUALITY IN LATIN AMERICA quite regressive in the case of social security spending). Most of the reduction in inequality measures adjusted for Second, much social spending is in-kind, so that even if it in-kind transfers is due to spending on education and health. affects the distribution of services (and indeed capabilities), With regard to education, Wodon (2003) estimated levels of it doesn't have a short-term impact on incomes. usage of various education levels by income quintile in 11 Latin American countries: Bolivia (1999), Brazil (1998), Who uses public services? Chile (1998), Colombia (1999), El Salvador (1997), While much social spending-for example, that related to Guatemala (1998-1999), Honduras (1999), Mexico (1996), better education and child health-affects incomes only with Nicaragua (1998), Peru (1997), and Venezuela (1998). Net long lags, it may be of both direct value for well-being and and gross enrollment rates were calculated by income quin- can expand the capabilities of different groups. To properly tile and by schooling cycle (that is, preschool, primary school, assess the incidence of public expenditures, it is necessary and secondary school)."' first to establish who benefits from access to public services Although these estimates were conducted nationally, in and then to value this access in equivalent income terms. urban and rural areas, and for both the male and female pop- Conceptually, it is critical to distinguish between average and ulations, only the national results are discussed here. In addi- marginal benefit incidence. Average benefit incidence refers tion, because for many surveys information on publicly versus to access to (and valuation of) services at a given time. Mar- privately provided education was not available, the analysis ginal benefit incidence refers to who benefits from increases does not make that differentiation; as a result, the progres- in access that are related to increases in expenditures at the sivity of public expenditure benefits may often be under- margin or over a period of time. There are difficulties stated (since the rich most likely opt for private services). involved in both stages (see box 9.2). Finally, the results obtained for the various countries were BOX ) 2 lss bI vhe ahuMw of pubikc swlc Strong assumptions are used to value the distribution of schools attended by the poor receive less funding. On the the benefits of public spending. In theory it would be best other hand, if the rich go to private school and this factor ro base valuations on estimares of underlying demand func- is not taken into account in the analysis, then some of the (ions for public senrices. However, in practice benefit inci- inequality in the disrribution of public educarion bene- dence analysis typically proceeds with valuations based on fits may be overesrimated. government unir costs, which are implicitly assumed to Demery (2003) highlights rwo other problems involved be equivalent ro the benefits of the spending. This assump- in benefit incidence analysis. First (as observed in the tion begets entirely the problems of efficiency in public analysis of who benefits from health care services), request- spending and differences in qualiry of services (see below). ing care is conditional on an illness or injury occurring Other problems are relared to limits in the available in the household. In surveys, illnesses and injuries are often household survey data to march the unir costs to house- self-reported only when those who declare an illness or holds. With regard to educarion, for example, ir is common injury are asked questions on the use of health faciliries. practice to multiply enrollment rates by level (thar is, pri- If some among the poor fail to report illnesses or injuries mary, secondary, or rerriary) among various income because they consider such events to be a common occur- groups, or within geographic areas, by an average estimate rence, a downward bias in the estimate of the benefits of of the cost ro rhe state of providing educarion services by rhe health care system for rhe poor may result. Second, level. However, if schools in rural areas-where a higher household surveys, while representarive at the national proportion of the population as poor-tend to be less well level, may not be representative of selected population sub- sraffed and equipped than urban schools, inequaliry in the groups, such as those attending a universiry Small sample distriburion of the benefits of public spending on educa- sizes may rhen also lead ro a bias in rhe estimare of benefit tion will be underestimated. The same will be true if incidence. 260 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS aggregated into overall measures of net and gross enrollment (DHS), combined with a measure of assets used to rank for a typical country in Latin America (each country was households into income quintiles. As for education, most given an equal weight in the analysis.) average benefits tend to accrue to the rich at a higher rate The analysis found that even if present average access to than to the poor. For example, immunization rates among all levels of education is generally regressive (that is, it children ages 12-23 months increase along with assets.12 On benefits the rich more), marginal increases in enrollment of average, about 50.6 percent of all households obtain all children ages 6-11 have been strictly progressive, benefit- vaccines for their children, while 5.6 percent do not get any ing the poorest quintiles more than those above them. Mar- immunizations. However, the rate of immunization for the ginal increases in enrollment up to 17 years of age were also poorest households is only 78 percent of the mean for all vac- progressive, although households in quintiles 2 and 3 ben- cines (a value of 0.8 in figure 4.6), while it is 112 percent efited more than the poorest (which nonetheless benefited for the richest households. more than quintiles 4 and 5). Marginal enrollments were not Antenatal care and delivery services are also used dis- progressive, however, for ages 18-24 or under 5 (see also proportionately by better-off households. For example, discussion in chapter 2). households in the first (lowest) income quintile are 50 per- The story line is clear enough. When overall enrollment cent less likely than the average to have a doctor preside over rates are low (as they are for preschool and tertiary educa- a birth, while households in quintile five are 45 percent more tion), increased access benefits the rich more than the poor. likely. As expected, the use of private facilities is even more When overall enrollment rates are already high, on the con- strongly associated with household assets, with the poorest trary, increased access benefits the poor more, since children households capturing about 14 percent and the richest house- from rich families are already overwhelmingly enrolled in holds capturing 254 percent of the overall mean. While there school. is less inequality in the use of public facilities, the benefits The implication of this cross-country analysis is that the are still higher for better-off households. progressivity of an expansion in education services varies This trend notwithstanding, expansions in the use of significantly among countries, depending on current levels health services are much more likely to be progressive, as in of enrollment. In most cases, increased access to primary the case of basic education. Figure 9.7 illustrates this through school is highly progressive, since average enrollment rates scatter plots of the ratio of average and marginal benefit inci- are already high. Increased access to secondary school is dence of 15 health services for the first and fourth quintiles, highly progressive in countries with high average enrollment in terms of the distribution of assets of the 9 Latin American rates, such as Argentina, Chile, and Uruguay. In most other countries (as an unweighted average) for which the analysis countries, access to secondary school is just slightly pro- has been conducted. A ratio of 1 means that the first quin- gressive or regressive. However, once average enrollment tile has the same access indicator as the fourth quintile. rates are pushed up (as they should be), further increases in In the figure on the left, apart from a few indicators such coverage will become progressive in most countries after a as treatment for diarrhea, most observations are well below few years. Increased access to tertiary education is, in con- 1, meaning that the access to health services for the first quin- trast, highly regressive and will remain so for many years in tile is well below that for the fourth quintile. In the figure most countries. Together with evidence of much higher on the right, for marginal benefit incidence, the ratios tend private returns on tertiary than on secondary or primary edu- to be much larger than 1, suggesting that at the margin the cation (de Ferranti and others 2003), these results indicate first quintile benefits more from increases in access than does that present patterns of allocation of public resources in most the fourth quintile. The figure also shows that the higher countries in Latin America-which is much higher for stu- the level of access to the service (the horizontal axis in both dents in tertiary than in primary or secondary education- figures), the more progressive the marginal benefit incidence. may well lead to overall regressive outcomes with regard to The overall pattern for most publicly provided services public spending in education. (including education, health, water, and infrastructure) is for Similar results were obtained with regard to health, nutri- access to be expanded from the top of the distribution down. tion, and population services. This conclusion is based on an Where service coverage is initially high, marginal incidence analysis of data from Demographic and Health Surveys is often quite progressive. Figure 9.8 illustrates this in terms 261 INEQUALITY IN LATIN AMERICA FIGURE 9.7 Average and marginal access incidence for various health services a. Ratio of average incidence, quintile 1 to 4 b. Ratio of marginal incidence, quintile 1 to 4 Ratio Ratio 3.5 - 3.5 - Antenatal5 3.0 3.0 - care: 2+ visits 2.5 2.5- 2.0 2.0- r ~~~~~~~~~~~~~~~~~~~~~~~Delivery 15 -Antenatal 1.5 attendance: * |15 Treated for Atnal15-public/l diarrhea + visits Yt * ~~~~~~~~~~ * * * *A~~~~~~~~~~~~~ntenatal 0.5 - 0.5 - care: doctor for diarrhea 0.0- 0.0 I I I I 0.0- 15 35 55 75 95 15 35 55 75 95 Access rate (%) Access rate (%) Source: Ajwad and Wodon 2003. FIGURE 9.8 Average and marginal benefit incidence for access to water in Latin America a. Water (urban) b. Water (rural) Share of average coverage Share of average coverage 2.0- 2.0- 1.6 - ~~~~~~~~~~~~~~~~~~~1.6- ~~~~~~~~~~~~~~~4 144< 6 1.2 - ~~~~~~~~~~~~~~~1.2 - 0.8 X 048 -1 0.8 - 0444 0.0 0.0 4 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Deciles Deciles --Actual (marginal) - Estimated (marginal) - Average benefit incidence Source: Ajwad and Wodon 2003. of access to water, and shows that although average incidence already enjoy high access rates can be quite significant, as is regressive for both rural and urban access to water, mar- shown by the progressive impact of the expansion of basic ginal incidence is quite progressive in both cases. The redis- education and health services in Latin America during the tributional effect of increasing the coverage of services that 1990s. This is encouraging, but it must also be recognized 262 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS that at best it brings poorer households up to service cover- according to their incomes. Unlike the Gini coefficient, this age levels routinely enjoyed by better-off households; in other can be negative, when poorer people receive absolutely more words, this trend should not be interpreted as evidence that benefits than richer people. The concentration coefficients policy is sufficiently propoor. Moreover, since coverage of for the various categories of spending in Mexico are presented many services (for example, secondary and tertiary education, in figure 9.9. Their values range from 0.66 to 0.73, indi- sewerage, and telecommunications) is not yet high in most cating widely different redistributive effects. Programs such Latin American countries, their initial expansion may not as Oportunidades, health services for the uninsured provided benefit the poor and may in fact have a regressive effect, unless by the Ministry of Health, basic education (primary and deliberate, targeted efforts are made to improve access levels lower secondary levels), school breakfasts, and Procampo all at the bottom of the income distribution. have negative concentration coefficients that indicate high to mild absolute progressive incidence. Most other programs Valuing the distribution of public spending show a positive concentration coefficient that implies that Following the analysis of access to services, it is natural to public outlays are regressive. Values between zero and the ask how much such access is worth to receiving households. value of the Gini for income inequality are still regressive This leads us to the difficult question of valuing the bene- in absolute terms, but are less unequally distributed than fits of public spending across the income distribution. The income. Public expenditures on higher secondary education valuation of public goods, and indeed also of publicly pro- and electricity subsidies, for example, are regressive, but not vided private goods, is still in its infancy. Box 9.2 summa- as badly distributed as income itself. Expenditures on ter- rizes the main difficulties faced when conducting studies on tiary education and pensions are almost as badly distributed this topic. Although it is beyond the scope of this chapter as income, and thus highly regressive, or worse, as is the case to cover this complex issue in depth, and despite the seri- of pensions from ISSSTE (for public sector workers). ous methodological caveats, there is enough evidence from Although incidence characteristics vary across countries- existing case studies to make three essential points. in addition to across programs within a country-some The first point is that the (average) incidence of the ben- common patterns do emerge (see table 9.6). Basic education, efits of public social expenditures varies enormously across for instance, tends to be progressive everywhere, with a programs. This is well-illustrated by a Mexican example, higher share of total spending going to the lower part of the drawing on the work by Scott (2003). Using household income distribution. Gini income elasticities (GIEs) (defined survey data from 2000 and unit-level administrative cost in box 9.3) are negative with regard to total spending for figures, Scott considers public spending on education (pri- public education in Argentina (-0.28), Chile (-0.44), mary, lower secondary, higher secondary, tertiary, and post- Mexico (-0.06), and Uruguay (-0.30). As expected, spend- graduate levels); health (using data on social security ing on primary education is highly redistributive, with GIEs institutions serving formal sector and state workers, as well ranging from -0.91 to -0.34, while spending for tertiary as data on the services provided by the Ministry of Health education is regressive, with GIEs ranging from 0.54 to 1.25 to the uninsured population); social security pensions and (the highest value is observed for Brazil). On average, the other nonhealth benefits; and other subsidies and transfers, benefits from spending on secondary education tend to be including programs targeted toward the poor (such as distributed in a neutral way on average, with GIEs ranging Oportunidades, previously known as Progresa, a grant from -0.33 to 0.32, depending on the initial rate of cover- program that is conditional on participation in education, age; this distribution is more progressive in countries with nutrition, and health programs) and farmers (such as higher levels of coverage. Procampo, a capacity-building project). Subsidies for Data on health are weaker than data on education, but residential electricity were also included, and measured by there are indications that total health spending is distrib- comparing the cost for households to the production costs uted in a neutral way overall, with large variations between published by the government. countries. The GIE is negative (that is, propoor) for programs An index of the inequality in incidence of benefits, the in Chile and Colombia, zero in Brazil and Uruguay, and concentration coefficient, can be calculated using the same positive in Mexico. However, intercountry comparisons formula as the Gini coefficient, but with individuals ranked should be treated with caution. In general, spending by 263 INEQUALITY IN LATIN AMERICA FIGURE 9 9 Concentration coeffcients for public social spending, Mexico 2000 Pensioners (ISSSTE) GINI Pensioners (IMSS) Active workers (ISSSTE) Health (ISSSTE) Tertiary education Electricity subsidy 7 777772. i2 I Health (IMSS) Active workers (IMSS) Upper secondary education ALL (excluding pensions) Milk subsidy (LICONSA) PROCAMO| Lower secondary education Tortilla (FIDELIST) Primary education Health (SSA) School breakfasts (DIF) Oportunidades school Oportunidades nutrition I I ~ ~ ~ ~ ~~~I I I -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 Concentration coefficients Source: Scott 2003. TABLE 9.6 Synthesis of case studies using estimates of the Gini income elsticiy (GIEI GIE (Gini income elasticities) Argentina Brazil Chile Colombia Mexico Uruguay Education -0.28 0.39 -0.44 -0.05 -0.06 -0.3 Primary -0.52 -0.34 - -0.58 -0.5 -0.91 Secondary -0.19 0.32 - -0.29 -0.005 -0.33 Tertiary 0.54 1.25 - 0.67 0.81 1.13 Health -0.25 0.06 -1.05 -0.5 0.25 0.00 Pensions - 0.7 0.91/ 0.58* - 1.0 1.15 Child care (or child development services) - -0.6 -0.6 -0.44 - -1.79 Family allowances - -1.03 - - -1.17 Cash transfers - - -0.78 - -1.12 ** -0.1 Unemployment benefits (insurance or assistance) - 0.17 - - - -0.02 Nutrition -1.06 -0.47 - - - -1.56 Housing programs -0.08 0.59 -0.56 - - -0.29 Infrastructure and transport - 0.35 - - - - Subsidies - - -0.35 - 0.49 Water -0.35 - - Electricity - - 0.53 *Noncontributory pensions **PROGRESA food transfers (-0.99) and PROGRESA scholarships (- 1. 25) Source: Wodon and others 2000. 264 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS BOX9 9 SuumarWng benelit incidene usn Igni G icomne e bsde IGIE) The GIE is a parameter that indicates the impact on equally. A GIE berween zero and one means that the dis- inequality of a marginal percentage increase In spending tribution of rhe increase in spending is correlated wirh for anv given income source or public transfer (in cash or income, and therefore is regressive in absolute terms but the monetary value of an in-kind transfer). A GIE less than progressive relative to income. A GIE of one means thar zero means that income and spending are negatively cor- public expenditures are distribured in the same way as the related, implving thar people in the lower part of the dis- distribution of total per capita income, and a GIE greater triburion benefit more than those in the higher part from than one implies an even more unequal distribution than the increase in spending. At zero. everybody benefits that associated with income. Ministries-which tend to cater to the uninsured-are often old of consumption, below which water or electricity is sold redistributive, while spending on the health plans linked to at a discount. However this "lifeline" is typically set at a rel- social security tend to favor better-off households. atively high level. Evidence on unemployment benefits (via Pensions are closely correlated with income, with bene- insurance or assistance) is also mixed. Some programs do fits typically as unequally distributed as other sources of reach the poor, but with low coverage rates (unemployment income. Related GIEs vary from 0.70 in Brazil to 1.15 in assistance in Chile, for example), while other programs are Uruguay. However, noncontributory pension schemes can limited to formal sector workers with benefits that are reach the poor. In the case of Chile, the GIE for pensions positively correlated with income. Households that benefit (known as PASIS) is -0.58. In Mexico City, a new program from job training tend to be in the middle of the income of transfers to the elderly was recently implemented, which distribution. also had a negative GIE (MNPTSG 2002). Overall, this comparison reinforces the enormous vari- Existing programs that provide childcare or early child ability in the redistributive impact of different social pro- development services tend to be redistributive, with nega- grams. This is true both for individual programs within the tive GIEs ranging from -1.79 in Uruguay to -0.16 in same country and for similar programs across countries. Colombia. Family allowances also tend to be redistributive Several factors explain such disparities. Programs that are (at least when they are means tested, as is the case in Chile intended to eventually have universal coverage are redistrib- and Uruguay). Targeted nutrition programs also tend to be utive (and marginal increases are especially progressive) when highly redistributive, for example in Argentina, Chile, the present access rate is already high, but are regressive Mexico, and Uruguay. However, school breakfast and lunch (even marginally so) when access rates are low. Thus, the programs, as well as food subsidies, appear to be much less expansion of programs with relatively low access rates should effective in reaching the poor and contributing to a reduc- be accompanied by targeting mechanisms to ensure that the tion in inequality. poor share in the benefits of expansion. In some other cases, there are no common patterns across This leads to the second essential point of this section: countries. The impact of housing programs on inequality, in many countries, a tradeoff seems to exist between cover- for instance, varies greatly. Uruguay's rural housing program age of the poor and targeting to the poor. In other words, those is highly redistributive, with a GIE of -1.61, and both programs that effectively approach universal coverage Chile's programs and Brazil's favela upgrading programs have explain most of the current redistributive effects of social negative GIEs. However, benefits from several other pro- public expenditures (that is, basic education in most coun- grams are positively correlated with income and are there- tries and basic health in many). Meanwhile, several well- fore regressive. Utility subsidies for water and electricity are targeted programs that are highly cost-efficient in in most cases positively correlated with income. These redistributing income or other benefits are generally small subsidies tend to be allocated according to a "lifeline" thresh- and cover only a small share of the poor. This apparent 265 INEQUALITY IN LATIN AMERICA FIGURE 9.10 Targeting, coverage of the poor, and spending on various programs, Brazil 1997 Share of the bottom 20% of the population who have access to the program per year 1.0 , Public health ;. , _, Primary education 0.8 L. _ Water 0.6 0 School feeding 0.4- 0.4~~~~~~~~~L AM Kindergarten ^ 0r Social assistance 0.2 - S Sewage O 0.2 - Pensions Unemployment -f Secondary education LOAS/BPC' insurance Adult education '911 0.2 0.4 0.6 0.8 Share of program spending accruing to the bottom 20% of the population Note: Circle size is proportional to program budgets. a. Lei Org3nica da Assistencia Social/Beneficio de PrestasAo Continuada Source: World Bank 2000e; see chapter 9. dilemma between targeting and coverage probably reflects In figure 9.10, each bubble is proportional to the size of both administrative and political economy constraints, which an expenditure category or program. The horizontal axis must be taken into account in the design and "packaging" measures the share of benefits accruing to the bottom quin- of programs. tile, while the vertical axis measures coverage of the program An excellent example of this tradeoff is the case of Brazil, within this group (that is, the share of households in the as documented by von Amsberg, Lanjouw, and Nead (2003). bottom quintile benefiting from the program). Programs in These authors consider public social spending for education the lower left corner (pensions, unemployment insurance, and health services, urban services, infrastructure (clean sewage system provision, and secondary education) are poorly water, sanitation, transportation, and housing), and social targeted and do not reach many of the poor. Programs in the protection (pensions and unemployment insurance). For each bottom right corner (for example, social assistance programs) type of spending or service, three aspects were calculated: are well targeted but have low coverage among the poor. Pro- (1) the amount of government spending; (2) the share of grams near the top left (water, public health, public prepri- spending accruing to the urban poor (that is, the targeting mary and primary education) tend to be universal. There are or benefit incidence for the poor); and (3) coverage among no programs near the top right, that is, programs that are the urban poor for the service. The authors also present a both well-targeted and reach most of the poor. These pat- ranking of programs according to effectiveness in transfer- terns clearly illustrate the tradeoff between targeting and ring resources to the poor. coverage in Brazil. The higher the coverage among the poor, 266 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS the more difficult it is to avoid leakage to the nonpoor; in contrast, tightly targeted programs are small and tend to FIGURE 9 11 have low coverage even among the poor. Ratio of income of the 5th to 1st inmome quintiles, with and without public transfers, Chile, 1990-98 The third main point of this section is that, despite large variance in the incidence of the benefits of public social Ratio expenditures across both program types and countries, the 25-j evidence suggests that social public expenditure tends, on 20 - balance, to reduce inequality. This is generally true, as dis- 15 - * cussed above, with regard to childcare, primary education, lo -_ and basic health care. In many cases, it is also true of sec- I 5 - U l U U ondary education and well-targeted housing programs. 0 - More generally, however, the overall impact of social 1990 1994 1996 1998 expenditures tends to be less unequally distributed than are FM Without public transfers M With public transfers incomes (as indicated in figure 9.9 in the case of Mexico), and would therefore contribute to a reduction of overall Source Bravo and others 2001. inequality, if only values for all public programs could be imputed into household incomes. However, such an exercise is fraught with difficulties (as partly summarized in box 9.2). transfers (17,000 pesos per month per person) was almost Nevertheless, in the few cases where it has been (bravely) as large as the monetary income ofhouseholds (19,000 pesos attempted, the results do suggest that the overall effect of per month per person). In-kind subsidies for education were public expenditures on social sectors is inequality-reducing. the main contributors to the reduction in inequality (59.9 Such is the case, for example, of Bravo, Contreras, and percent of total transfers), followed by health (25.5 percent), Millan (2002), who estimated the value of government trans- monetary transfers (11.1 percent), and housing (6.5 percent). fers in cash and in kind in Chile and compared the distrib- Although a reduction of 6 percentage points in the Gini ution of these transfers to the distribution of monetary is not revolutionary, it is quite substantial. There is no evi- income. Public cash transfers include family allowances, non- dence that this order of magnitude is representative of other contributory pensions, utility and other subsidies, and unem- countries in the continent, since Chile has one of the most ployment benefits. In-kind transfers arise from public progressive incidence patterns for social spending in Latin expenditures on education, health, and housing. To estimate America. In the case of Colombia, for instance, Velez and the value of in-kind benefits provided by government expen- Foster (2003) found a concentration coefficient13 for total ditures, Bravo, Contreras, and Millan (2002) conducted a public social spending of-0.06, implying that in absolute detailed analysis of the various subcomponents of each type terms the subsidies were distributed in a marginally pro- of expenditure program and developed valuation criteria. gressive way across quintiles and amounted to a substantial For example, 17 health categories were identified (such as redistribution relative to the unequal pattern of income dis- surgery, dental services, laboratory tests, preventive check- tribution. Given the valuation assumptions made, social ups, X-rays, emergency services, and hospital expenses) and, spending contributed to a reduction in the Gini index of for each of these categories, average monthly values were close to 3 percentage points. estimated. More than 25 subcomponents were identified for Once again, there were large differences in the extent to education and 6 for housing. which programs were redistributive. The patterns by pro- The main conclusion of this analysis is that rising, pro- gram type conform to those identified above. Some nursery gressive public social expenditures more than offset the schools, primary education, and subsidized health insurance observed increase in income inequality during the 1990s were the most redistributive programs. Housing subsidies, (figure 9.11). The authors argue that including in-cash and tertiary education, and family subsidies were the least in-kind transfers in the welfare aggregate reduces the Gini redistributive. inequality index by 6 percentage points, from 0.56 to 0.50. Finally, it is important to note that the incidence patterns In the first income quintile, for example, the value of the considered in this section-whether average or marginal and 267 INEQUALITY IN LATIN AMERICA related to access or value-have all referred, in some sense, economic expansions, social spending as a share of total to a temporal average. They have been discussed without ref- spending increases, but it is better to save at least some erence to whether they were observed in good times or in public funds during expansions in order to better protect the bad times, or whether the economy was in a boom or in a poor during recessions. Although governments try to be recession. However, macroeconomic volatility is an unfor- propoor (or at least prosocial) whenever they can, they often tunate but ubiquitous fact of Latin American life, and the do so in a shortsighted way that does not support good safety manner in which social spending behaves throughout the nets during crises. economic cycle is a key attribute to consider. Does it pro- In a deeper sense, procyclicality in public expenditures vide some measure of insurance against increased risks in is the combined result of three factors (Perry 2003). First is downturns? Or does it aggravate the volatility of the cycle? the procyclicality of tax revenues: tax-to-GDP ratios are higher during economic booms because of high dependency The lack of countercyclical targeted spending on consumption taxes and commodity exports. Second are during downturns political economy factors that lead to the spending of any Macroeconomic shocks have plagued Latin American coun- potential surpluses in good times (that is, to procyclical tries since the 1980s. The poor are particularly vulnerable spending during economic booms). Third are procyclicality to these shocks, as they are typically more exposed to risk and informational asymmetry problems in financial markets, and have little access to appropriate risk management instru- in particular with regard to the "true" nature of fiscal policy. ments."4 Social protection systems, including safety nets, In other words, markets can not distinguish between should help the poor to deal with the risks brought about responsible counter-cyclical policies and relaxation of the by economic shocks, as well as with idiosyncratic risks. intertemporal fiscal stance, especially when the track record According to the risk management framework spelled out in booms has been bad. in de Ferranti and others (2000), social public expenditures These three factors combine to force major expenditure designed to manage risks should go up during economic restraints during recessions due to a lack of financing. In crises. countries with no access to private financial markets, pro- Yet the opposite appears to be true. It is difficult to pro- cyclicality might be partially attributed to deficiencies in offi- tect the poor during recessions because several forces combine cial adjustment programs. Hence, overcoming the problems to put downward pressures on the amount of public transfers associated with the procyclicality of social expenditures that can be allocated per poor person. First, due to the highly requires building appropriate institutions (including fiscal procyclical nature of fiscal policy in most Latin American rules) to overcome structural problems (see discussion in countries,'5the share of GDP devoted to overall public spend- chapter 8). Equally important is creating some automatic- ing tends to decrease in bad times. Second, GDP itself declines ity in the expansion of propoor programs in bad times, for in a crisis, so that even if the share of GDP devoted to social example through the earmarking of funds, especially in light public spending remains constant, there will still be fewer of the substantial pressures to spend regressively that are resources available. These two factors tend to make aggregate associated with financial crises (see chapter 8 and de Ferranti targeted public spending for the poor procyclical rather than and others 2000). countercyclical. Third, poverty increases during a crisis, so that the available aggregate resources targeted toward the poor 9.3. Cash transfers and distribution have to be distributed among a larger pool of applicants, Following the discussion of the trends and incidence pat- resulting in lower spending per poor person. terns of social public spending as a whole, this section turns On average, in Latin America during the 1990s, a decrease to a specific subcomponent of public expenditures, namely of GDP by 1 percentage point led to a reduction of at least direct cash transfers to households. Whereas governments 2 percentage points in targeted public spending per poor might choose to spend money on education, labor market person (Wodon and others 2003). About half of the impact programs, or even housing purely for efficiency reasons was due to the reduction in per capita GDP and half to the (having to do with internalizing externalities, for instance), increase in the number of poor people; the share of the GDP cash transfers appear, at first glance at least, to be more likely in targeted spending remained relatively stable. During driven by equity concerns. 268 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS TABLE 9.7 Social security expenditures around the wouid, as percentage of GDP Total social security Pensions' Health care' Others3 expenditure4 Africa 1.4 1.7 1.2 4.3 Asia 3.0 2.7 0.7 6.4 Europe 12.1 6.3 6.4 24.8 Latin America and Caribbean 2.1 2.8 3.9 8.8 North America 7.1 7.5 2.0 16.6 Oceania 4.9 5.6 5.6 16.1 Notes: 'Pension expenditures cover old age, disability, and survivors' pensions. 2Health care expenditures cover health care services. 3Other expenditures cover employment injury, sickness, family, housing, and social assistance benefits in cash and in kind, including administrative expenditures. 4Total social security expenditures cover pensions, health care, and other aspects. Source: Table originally published in ILO 2000, Statistical Annex, table 14. Regional averages are weighted by gross national product evaluated at Pricing and Purchasing Parities (PPP) prices. See http://www.ilo.org/public/english/protection/socsec/publ/ css/table 1 4.htm. Indeed, in most developed countries the bulk of cash as social assistance to the poor and are formally intended to transfers are part of what is known as "social security sys- be redistributive. Examples include the Working Families tems" which have historically been associated with distrib- Tax Credit (WFTC) in the United Kingdom or the Tem- utional concerns. As they developed over the course of the porary Assistance for Needy Families (TANF) in the United latter half of the last century, it became evident that social States. The key point is that all public transfers of cash to security systems were fundamentally more about risk (that households are to some extent a mix of social insurance and is, inequality across states of nature) than about inequality social assistance, since all such transfers can help mitigate per se (that is, a cross-section of inequality). Three princi- the effects of negative shocks on the consumption stream and pal risks seem to draw most public resources: the risk of virtually all incorporate some element of progressive redis- having undersaved for and, in particular, falling into poverty tribution. in old age; the risk of disability; the risk of unemployment; With this in mind, how do cash transfers-and indeed and the risk of income loss through poor health. In devel- social security-in Latin America compare with those oped countries, all social security payments typically in other countries? Are they effective risk-management account for almost half of total public spending, or roughly systems? Do they also enhance equity? Table 9.7 compares one-fifth of GDP. Old-age, disability, and survivor pensions official social security expenditures (as a percentage of GDP) account for just under half of those shares in Organisation across regions of the world. Social security expenditures are for Economic Co-operation and Development (OECD) broken down into pensions, health care, and other (mostly countries. social assistance) according to the official definitions of the However, it would clearly be incorrect, or at least incom- International Labour Organisation (ILO). The regional aver- plete, to characterize these social security transfers simply ages indicate that rich countries (in Europe, North America, as "insurance." Whereas the incidence of unemployment, dis- and Oceania) allocate a much greater output share to over- ability, and illness do indeed have random components all social expenditures than do developing countries. (though different income groups may be faced by different At 8.8 percent, the share of Latin America's GDP that is probabilities), old age is reasonably (if not perfectly) fore- accounted for by social security is only one-third of the seeable. In addition, transfer levels are not identical among European share and roughly half that of North America and people faced with identical concerns, since benefits are gen- Oceania, but it is still substantially higher than in Africa or erally means tested in order to achieve progressive incidence. Asia. Indeed, overall social security expenditures in many In other words, transfers depend not only on the realization Latin American countries are around what would be of the random variable of interest, but also on private expected for their level of income per capita (See figure 9.12.) incomes. Some benefit types are indeed explicitly designed In addition, a large share of the total, corresponding to 3.9% 269 INEQUALITY IN LATIN AMERICA FIGURE 9.12 Total social security expenditure Revenue as a share of GDP 0.40 1 0.35 0.30 -France 0.25 - 0.20 - Uruguay 1 0.15 Chi * UnitedStates l ~~~~~~~~ ~ ~ ~ ~~~~~~~Paraguay gnia f * _Y ~~~~~~~~~~~~~~~~~~~~~~Brazil o 0.10 - f~~ ~ ~~~~~~~~~~~~~~~ _*Pnama 0.05 - Nicaragua Colombia l * * e Fz S~~0 ?Atbr * Mexico 0.00 4 6 7 8 9 10 11. -O.OS Log GDP per capita (in constant 1995 U.S. dollars) Average for the 1990s -0.10 Source: Authors' calculations based on microdata from household surveys. World Development Indicators Database and ILO 2000. of GDP, is allocated to "other" expenditures, which include cent in Uruguay, with equally impressive heterogeneity social assistance. If used well, these resources would hold real among areas within individual countries. As is the case with potential for redistribution in Latin America. other public expenditures, the effectiveness and incidence of Table 9.8 presents some summary information on social social security expenditures vary considerably both across and security expenditures for 24 countries in Latin America. This within countries. Cash transfers under the general heading information should be treated with some circumspection, as of "social security" include very disparate programs, with it is generally harder to aggregate across social security and widely different incidence patterns. In the case of Mexico, social protection programs in developing countries than it for instance, cash transfers include both some of the most is in industrial nations. The reason for this is that a wide progressive of all public programs (such as Oportunidades), variety of programs are seldom integrated and not all pro- and some of the most regressive (such as the pensions paid grams are well documented. Even basic data on program size by the ISSSTE, and the generalized electricity subsidies) (see and eligibility criteria are not always available. In addition, figure 9.9). The same is true in Brazil (see chapter 6 for a cash transfers appear to have traditionally been used less discussion on the regressivity of the subsidies to civil ser- extensively in less developed countries than in developed vant retirement pensions), but where other components, such countries; instead, a number of safety net programs in poor as rural pensions and the payments made under the countries distribute food or other commodities. Although BPC/LOAS scheme, are actually quite progressive. food stamps, for example, are also common in developed Unfortunately, the evidence suggests that in most countries-indeed, they are prominent in the United countries, the regressive components of the "welfare state" States-these countries tend to rely more on cash transfers in present-day Latin America quantitatively dominate pro- than do developing countries. gressive elements. Table 9.9, drawn from Paes de Barros and Social security as a share of total public expenditures in Foguel (2000), illustrates this point through figures on retire- Latin America ranges from 14 percent in Belize to 68 per- ment pensions and unemployment insurance in Brazil. 270 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS TABLE 9.8 Social security expenditures in Latin America and the Caribbean, as percentage of GDP Total social security expenditure Percentage Pementage of total public Pensions' Health careb Others' of GDPd expenditure Argentina 4.1 4.3 4.0 12.4 41.2 Bahamas - 2.5 Barbados 4.1 4.4 1.5 9.0 - Belize - 2.1 - 3.5 14.2 Bolivia - 2.3 - 7.0 29.3 Brazil 2.4 2.1 7.7 12.2 36.7 Chile 5.9 2.3 3.1 11.3 45.6 Colombia 0.9 5.1 0.1 6.1 - Costa Rica - 6.8 - 13.0 42.6 Cuba - - - - - Dominica 1.4 0.4 3.0 4.8 - Dominican Republic - 1.8 - 2.5 15.7 Ecuador 1.2 0.3 0.5 2.0 - El Salvador 1.3 1.3 1.0 3.6 Grenada - 2.8 Guatemala - 1.7 Guyana 0.9 4.3 0.6 5.8 Jamaica 0.3 2.5 1.7 4.5 - Mexico 0.4 2.8 0.5 3.7 22.6 Nicaragua 1.4 4.3 3.4 9.1 28.1 Panama 4.3 5.6 1.4 11.3 41.3 Peru - 2.2 Trinidad and Tobago 0.6 2.5 3.5 6.6 22.7 Uruguay 8.7 2.0 11.7 22.4 67.8 Simple average 2.5 2.9 2.9 8.0 34.0 - Not available. Notes: a. Pension expenditures cover old-age, disability, and survivors' pensions. b. Health care expenditures cover health care services. c. Other expenditures cover employment injury, sickness, family, housing, and social assistance benefits in cash and in kind, including administrative expenditures. d. Total social security expenditures cover pensions, health care, and other aspects. Source: Table originally published in ILO 2000, Statistical Annex, table 14. See http://www.ilo.org/public/english/protection/ socsec/publ/css/tablel4.htm. TABLE 9.9 sion payouts. In the case of old-age pensions, this low share Distribution of govemnment expenditures on pensions and in part reflects different past levels of contributions during unemployment insurance in Brazil by quintile (percent) the working years. However, as indicated in chapter 6, those differences are actually magnified, rather than dampened, by Income quintiles Pensions Unemployment insurance a pension system that rewards well-off civil servants at the First 2.4 3.0 expense of poorer taxpayers. Second 6.4 21.0 The case of unemployment insurance is even more sur- Third 9.7 20.2.. Fourth 16.5 36.3 prising, at least at first glance. One might have expected that Fifth 65.1 19.5 the unemployed would be overrepresented in the bottom fifth of the income distribution, and it is thus disappoint- Source: Paes de Barros and Foguel 2000. ing to see that this group receives by far the lowest share of total disbursements and more than six times less than the The bottom 20 percent of the Brazilian income distrib- richest quintile. Upon further analysis, however, the reason ution receives a mere 3 percent of total disbursements in is clear: only workers in the formal sector qualify for unem- unemployment insurance, and an even smaller share of pen- ployment insurance, and the poorest of the poor in Brazil 271 INEQUALITY IN LATIN AMERICA do not work in the formal sector. More than half of the labor the continued receipt of transfers is conditional on households force has no signed labor documentation and are overrepre- investing in the education and health status of their mem- sented among the poorest 20 percent of the population. bers, in particular children. With some exceptions (discussed Table 9.9 serves as a reminder of the truncated and elit- below), these programs are typically designed and imple- ist nature of the social security systems that developed in mented by the federal or central government. Participating Latin America over the course of the 20th century. Never- communities and households are selected by program officials theless, the existence of some social assistance and social situated in the central government and transfers go directly protection programs (such as Oportunidades in Mexico, to eligible households without passing through state budgets. Familias en Acci6n in Colombia, Trabajar in Argentina, and It has been hoped that this arrangement would eliminate Bolsa Familia in Brazil) which make cash transfers that actu- the unnecessary bureaucracy inherent in many existing pro- ally reach some of the poorest citizens in their respective grams, as well as the potential for corruption. The programs countries, and thereby attain highly progressive incidence use a range of targeting methods (for example, geographic measures, holds out the possibility of serious reform of the and community targeting or household proxy-means) to historically truncated "welfare state." Also noteworthy are ensure that program benefits reach the poorest households. initiatives in some countries that consolidate eligible ben- It is often argued that such "fine" targeting can result in a eficiaries for social assistance programs into one system. loss of support from the excluded middle classes, and thus Examples are the ficha CAS in Chile and SISBEN (Sistema adversely affect the political sustainability of programs. How- de Selecci6n de Beneficiarios para Programas Sociales) in ever, in many cases the argument that programs reach only Colombia. Such efforts lead to significant savings on admin- those who are very needy, with little leakage to the nonpoor, istrative costs and are associated with a set of highly pro- is used to generate political support domestically, as well as gressive programs. external support from international development institutions Some of these programs (such as Trabajar and its succes- and donors. sorJefes y Jefas de Hogar in Argentina) are essentially public Despite an increased supply of education and health ser- employment schemes. Jefes y Jefas entails additional con- vices (see section 9.2 above), empirical evidence clearly shows ditions and complexities, but the essence of the program that poor households have substantially lower utilization remains self selection by the urban poor into a program rates, a fact that reflects the high opportunity cost of access that provides a transfer corresponding to a low wage, in (for example, forgone earnings and the cost of travel, uni- exchange for community or other work. Although these forms, and educational materials). Supply-side interventions programs run the risk of being used for clientelistic purposes, in isolation are therefore not very effective at enticing poor Trabajar has been evaluated as being highly effective in households to access these services. Conditioning targeted targeting the poorest citizens (Jalan and Ravallion 1999, cash transfers on access to services and facilities-which 2003). Although they are no panacea, and do suffer from effectively transforms pure transfers into human capital some leakage to the nonpoor and incomplete coverage of the subsidies-is an attractive policy response to this problem.17 poor, this family of programs has a long history and a number Yet access, although a necessary condition for capital accu- of features that deserve praise. (See Galasso and Ravallion mulation, is in and of itself insufficient. Human capital 2003 for a discussion of the impact of Jefes y Jefas during accumulation requires that such access be complemented Argentina's 2002 crisis.) Following is a brief discussion on with quality services. For this reason, many transfer programs a related family of programs known as conditional cash include a supply-side component in order to reinforce the transfers (CCTs). impact of demand-side subsidies. One of the main attractions of demand-side transfers is Smart transfers: A new breed of social that they can help generate a sustained decrease in poverty assistance programs independent of the transfers themselves, in addition to the In recent years, many developing countries, particularly in obvious immediate decrease in poverty. This characteristic Latin America, have taken a relatively innovative approach of program design can also be used to generate political sup- to social safety nets, now known generically as conditional cash port for larger transfers, since it can be argued that the trans- transfers.16 One of the key defining features of CCTs is that fers constitute investments in needy households (that is, 272 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS a "hand up" as opposed to a handout) and that they are a of a country's GDP, total public expenditure (TPE), and total required complement or component of universal coverage government expenditure on education, respectively. programs (for example, in basic education and health) that benefit the whole population. The requirement that house- Program implementation and selection holds undertake such investments simultaneously recognizes of beneficiaries the right of individuals to a basic living standard and their As indicated above, a distinctive element of many of the responsibility to achieve this goal. transfer programs is the direct relationship between the fed- Another innovative feature of the new CCT programs is eral or central government and the beneficiaries. In the case the fact that in many cases, special emphasis has been placed of Progresa, for example, once the program identifies which on the need for a built-in, comprehensive evaluation of the communities and households are eligible to receive program program from its early stages. Some of the features identi- benefits, a general assembly is held in the community to fied above are discussed below in more detail with regard inform households of their eligibility, to explain the work- to a subset of these programs, that is, those for which suf- ings of the program with regard to household entitlements ficient information on their design, implementation, and and responsibilities, and to answer any questions on program impact is currently available. For this purpose, we focus in operations or objectives. detail on the programs that have been implemented in Brazil Although the Progresa, PRAF, and RPS programs are (Programa Nacional de Bolsa Escola, henceforth denoted as heavily centralized in terms of both design and implemen- BE, and Programa Bolsa Alimentacao, henceforth denoted tation, their structures also allow for direct community as BA);18 in Honduras (Programa de Asignacion Familiar, involvement. For example, at Progresa's general assemblies, henceforth denoted as PRAF); Nicaragua (Red de Protecci6n participants select a woman beneficiary to act as the "com- Social, henceforth denoted as RPS); and Mexico (Programa munity promoter." This person plays the role of a liaison de Educaci6n, Salud, y Alimentaci6n, henceforth denoted as officer, arranging regular community meetings with bene- Progresa and now known as Oportunidades).19 ficiaries, informing them of their rights and responsibilities Table 9.10 summarizes information on the size of these under the program, and communicating beneficiary concerns transfer programs, including Chile's Subsidio Unico Famil- to program officials. It is this person who informs benefi- iar (SUF) and Bangladesh's Foodfor Education (FFE) for com- ciaries when and where to collect their transfers and answers parison purposes, but excluding BA, for which data were not any questions regarding the amounts they receive. Com- available. Column 1 presents information on the coverage of munity promoters are volunteers who do not receive pay- the programs, while column 2 describes the size of their bud- ment for their services, except for a small amount to cover gets. Columns 3, 4, and 5 show the annual budget as a share expenses (for example, for travel to promoter meetings). TABLE 9.10 CondiUonal cash transfers: comparison of program sizes Budget as Budget as percentage of percentage of Total annual budget Budget as total public public expenditure Program Coverage (U.S. dollars) % GDP expenditure on education Progresa (Mexico, 2.6 million 998 million in 2000 0.2 1.6 4.1 introduced August 1997) households in 1999 PRAF (Honduras, 47.8 thousand 12.5 million in 2001 0.2 2.0 5.0 introduced late 2000) households RPS (Nicaragua, pilot 10 thousand 10 million in 0.2 introduced October 2000) households in 2001 2001-2002 FFE (Bangladesh, 2.1 million students 77 million in 1999 0.2 4.2 7.9 introduced 1993) in 2000 SUF (Chile 1998) 954,000 students 70 million 0.1 0.9 3.5 BE (Brazil 2001) 5 million families 680 million 0.15 0.7 2.5 Source: Morley and Coady 2003. 273 INEQUALITY IN LATIN AMERICA Once Progresa beneficiaries have been informed of pro- used to identify the most marginal communities to be gram operations and their rights and responsibilities, they included in the program. In addition, communities must have must register at schools and clinics. Proof of registration access to education and health facilities within a certain radius must be presented in order to receive the initial cash pay- and to have 50-2,500 inhabitants. ment. Registration forms are then logged into electronic In PRAF, information from a 1999 height-for-age school databases, which are then transmitted from Progresa's state census of first graders is used to identify the 70 rural munic- office to the organization's central office. Attendance at school ipalities (out of a total of 297) in which malnourishment is and health clinics is monitored and recorded by service greatest. Geographic targeting in the RPS pilot program was providers, and this information is transmitted every two based on poverty rates and on access to social infrastructure months to the central office. Household transfer levels are and organizational capacity. Using these criteria, two depart- calculated based on this information and communicated to ments were initially chosen. Within these two departments, households at local distribution points via the national 6 (out of 20) municipalities were chosen using similar poverty telegraphic system. and operational criteria. All of the 59 rural comarcas (locali- One of the issues that arose during the program imple- ties) within these municipalities were eligible for the program. mentation and evaluation process was the importance of Once the participating localities are identified, each ensuring that the effectiveness and integrity of the moni- program undertakes further targeting using proxy-means, toring process was continuously evaluated. The validity of categorical, or community-targeting methods. In Progresa, a beneficiary compliance cannot be confirmed based solely on community census is conducted in all localities. The resulting impact evaluations, since these rely on reporting by house- socioeconomic data on households is then used to calculate a holds, which obviously have an incentive to report atten- proxy-means score for each household (using discriminant dance. Similarly, service providers may have incentives to analysis, with income as the left-hand-side variable) and falsely report attendance, for example because of community then to classify households as "poor" (that is, eligible for the pressure, the consequences of withdrawing benefits from poor program) and "nonpoor" (that is, not eligible).20 households, and the need to avoid congestion at schools and In the 70 participating municipalities selected by PRAF, clinics. only households with pregnant or lactating women or with Although a preliminary evaluation of the monitoring children under the age of 12 are eligible for the program. process indicated that incentives for truthful reporting also For both the BE and BA programs in Brazil, only households existed (for example, monitoring by beneficiaries who do classified as poor based on a monetary income cutoff level comply with the rules, a commitment by service providers of 90 reais per month (equivalent to one half the official min- to improve human capital status, and the fact that there is imum wage in 2001) are eligible. a provision for noncompliance for health-related reasons), this The process of household selection is a bit more complex process needs to be looked at systematically. in the RPS. To determine which households are eligible for All of the programs considered here have paid particular the program, a marginality index is calculated for all of the attention to the need to ensure that benefits actually reach 59 comarcas in the six participating municipalities. The their intended populations. This has been achieved by using index was based on the following variables from the 1995 a combination of targeting methods to select beneficiaries. National Population and Housing Census: family size, access Most programs use a two-step process to determine the eli- to potable water, access to latrines, and illiteracy rates. These gibility of households, combining some form of geographic 59 localities were then divided into four groups on the basis targeting with proxy-means, categorical, or community- of this index. In principle, all the households in the "poorest" targeting methods. two groups, which contained the 42 most marginal locali- In Progresa, the first stage of geographic targeting involves ties, were eligible. However, during implementation it was choosing the poorest or most marginal communities. Infor- decided to exclude around 2 percent of the households that mation from the national census on the demographic, hous- were deemed to be nonpoor because of ownership of a vehi- ing, infrastructure, occupation, and education characteristics cle or large landholdings. In the remaining localities that of communities is used to construct a "marginality index" (or were considered to be least marginal, 20 percent of house- score) for each community in the country, which is in turn holds were excluded based on predicted consumption levels. 274 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS Benefit structure TABLE 9.11 A defining feature of CCT programs is that in order for ben- Benefit sbcture of Progmsa, Jly-December 1999 eficiaries to receive transfers, they must undertake a series (pesos/month) of actions that are intended to enhance the human capital status of household members. In some programs, partici- Education scholarships Boys Girls pating households receive two sets of demand-side transfers. Primary One is the education subsidy, which is conditioned on reg- Grade 3 80 ular attendance at school. The other set-often referred to Grade 4 95 Grade 5 125 as the food or nutrition transfer-is conditioned on regular Grade 6 165 attendance by household members (particularly children) at Materials (annual) 160 health and information clinics at which the importance of Secondary Grade 7 240 250 practicing specific nutrition and hygiene behaviors in the Grade 8 250 285 home is discussed. In other programs, such as Brazil's Bolsa Grade 9 265 305 . . ~~~~~Materials (annual) 205 205 Famflia, a single transfer is conditioned on both sets of Food transfer 125 per family Benefit cap 750 per family In Progresa, only children aged 7 or older are eligible for Benefit_cap_750_per_family education transfers, starting when they are in grade 3. This Note: The cap on the total benefits a household can receive is guideline is partly motivated by a desire to prevent incen- applied only to the sum of the education scholarships and tives for higher fertility, as well as by the fact that school food transfer (that is, excluding transfers for materials). ' ~~~~~~~~~Source: Skoufias 2001 and Morley and Coady 2003. enrollment rates for the first two grades are very high even in the targeted "poor" populations. The benefit structure has a number of important features. Transfers increase by grade All children in this age group must attain an 85 percent (primarily because of the higher opportunity costs for older attendance rate. Under BE, households with children ages children) and are higher for girls in middle school (grades 6-15 receive a monthly transfer of 15 reais per child, up to a 7-9). Higher subsidies for girls in middle school reflect a maximum of three such transfers and conditional on 85 per- desire to reduce gender bias in education, as well as to inter- cent attendance. nalize the widely documented social externalities that accrue With regard to food transfers (that is, health subsidies), from female education. Each eligible child also receives in Progresa this was fixed for each family at 250 pesos per a stipend for school materials, fixed at 160 pesos annually month in the second half of 1999 and was conditioned on for primary school children and at 205 pesos annually for households making regular trips to health clinics for a range secondary school. This information is summarized in of preventive health checks and on attendance at monthly table 9.11. nutrition and hygiene information sessions. In addition to Education subsidies in the other programs have a sim- the cash transfer, beneficiary households with children under pler structure. In the RPS, only households with children the age of 3 receive a monthly nutritional supplement (usu- ages 7-13 who have not yet completed grade 4 of primary ally a box of 30 sachets per month) that contains essential school are eligible to receive transfers, which in 2000 were micronutrients. The monthly nutrition and hygiene infor- fixed at a monthly rate of 120 cordobas per household and mation sessions focus on such issues as the importance of con- an annual transfer of 275 cordobas to finance additional suming nutritional foods, preparing food properly, and education expenses. All such children in the family are following simple hygiene and preventive health care prac- required to maintain an 85 percent school attendance rate. tices at home. Food transfers were similarly conditioned in Unlike in Progresa, transfers through RPS are not inflation- the other programs. indexed. In all of the programs considered here, a capping system In PRAF, all households with children ages 6-12 who is in place that puts a ceiling on the total amount of trans- have not yet completed grade 4 are eligible for transfers. Each fers a household can receive. This is usually motivated by a participating child receives a monthly transfer of US$58/12 desire to not create dependency on the part of recipients, to reais, with a maximum of three such transfers per household. not erode the incentive for self-help, and to avoid incentives 275 INEQUALITY IN LATIN AMERICA for higher fertility. In Progresa, this ceiling is set at 750 pesos depends on the size of the population covered and range from for education and health transfers combined. On average, the US$3,318 to US$15,000, with a ceiling of US$15,000 and transfer to beneficiary households is substantial, constitut- an average of US$6,020. As with the education supply side- ing around 20 percent of total household consumption. In and possibly to an even greater degree-the nutrition and PRAF, households can receive at most three education trans- health program has encountered many unanticipated prob- fers and two health transfers. On average, total transfers are lems in implementing supply-side measures. less than 5 percent of total household expenditures for beneficiaries. Do conditional cash transfers work? In RPS, the transfer is fixed at the household level and One of the most innovative aspects of many CCT programs is not dependent on the number of eligible children. On is the emphasis placed from the beginning on having a built- average, the total transfers received by beneficiaries are just in, credible evaluation process. The evaluation of Progresa over 20 percent of total household expenditures. In both BE was perhaps the most extensive such evaluation. Although and BA, households can receive transfers for at most three the decision to evaluate the program was "homegrown," the children, equivalent to a transfer ceiling of 45 reais per Inter-American Development Bank (IDB) also played a key family. With regard to BE, it has been estimated that the role in facilitating and encouraging discussions on design, total transfers received by beneficiary households make up implementation, and evaluation issues. less than 5 percent of total income for households in the Focus is placed here on the evaluation of Progresa, a dis- bottom two national income deciles. cussion that is occasionally complemented with results for One of the key differences among the programs is how RPS. For both these programs, experimental evaluation they coordinate with the supply-side provision of education designs were built into their structure from the start and the and health services. Although Progresa, BE, and BA do not results of evaluations are now publicly available. Some esti- have an explicit supply-side component in their education mates for BE that were obtained using an alternative or health interventions, program personnel coordinate with methodology are also reported. national education and health ministries to plan for In Progresa, 506 communities were identified by the geo- increased supply-side investments in areas where the pro- graphic targeting method as being the most marginal and grams operate, with demand for services therefore expected thus eligible for the program (Behrman and Todd 1999b, to expand. Skoufias 2001). Out of these, 320 were randomly chosen as However, both RPS and PRAF have explicit supply-side treatment communities (that is, those that would receive the components. In RPS, each beneficiary student receives a program's services), and 186 as control communities for com- monthly transfer of 5 cordobas, which is handed over to the parison purposes. The justification for withholding program school at the time of enrollment registration. It is expected benefits from the control group was that there were bud- that this transfer will be split evenly between teachers' getary and operational constraints that would be relaxed over salaries and school resources. On the health side, informa- time and that a phased introduction of the program was tion clinics and health services are provided by nongovern- therefore necessary. mental organizations (NGOs). With this in mind, the sequential expansion of Progresa In PRAF, schools directly receive transfers linked to the due to such constraints presented the opportunity for a rig- number of students and teachers, ranging from US$1,600 to orous evaluation of the impact of the program by compar- US$23,000 annually, with an average of US$4,000. These ing the outcomes in participating communities with those funds are also channeled through school coordinating bodies, of control communities that were to be incorporated into the which include NGOs and parents and are intended to have program at a later stage. The control group was eventually discretion over how monies are spent. However, in practice incorporated in late November 1999, two years after the pro- it appears that for legal reasons, schools have not received the gram was first launched. The decision to maintain a control funds directly but have had to place orders for materials group predictably generated heated public debate, but pro- through conventional school channels. On the nutrition and gram officials were determined to defend this decision since health side, resources are distributed to both clinics and com- it was seen to be crucial to maintaining the integrity of the munity health promotion groups. The size of these transfers program evaluation process. 276 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS Another defining feature of the Progresa evaluation was poor receive between 1.55 and 2.22 times their population its extensiveness in terms of the use of both quantitative and shares. qualitative evaluation techniques and the wide range of issues addressed. At the core of the quantitative evaluation was the Human capital impacts collection of both "baseline" and "follow-up" household sur- The edacation impacts of the CCTs that have been evaluated veys during the period November 1997-November 1999. are very impressive. Prior to the Progresa program, primary These surveys collected detailed socioeconomic information enrollments in the evaluated communities were about from all 24,000 households in the 506 evaluation commu- 93 percent in all primary grades. Enrollment rates dipped nities. Altogether, five rounds of data were collected, one in substantially after the final primary school level (grade 6), 1997 and two each in 1998 and 1999. These data provided with enrollments rates in middle school (that is, grades 7-9) the opportunity to undertake a scientific evaluation that falling to 67 percent for girls and 73 percent for boys. could provide a clean estimate of program impacts, since the Enrollment rates in high school showed another dip, with presence of a control group enabled evaluators to "wash out" only 58 percent of those who successfully completed middle confounding factors, for example inherent time trends and school enrolling in high school. This is the main motiva- exogenous shocks. As it turned out, having such a control tion to increase grants by around 50 percent for children in group was crucial because the program area went through middle school, compared to the level available to primary an economic shock in the form of a drought-which caused school students. much distress and adversely affected many of the outcomes As expected, enrollment impacts in primary school were considered in the evaluation-that was exogenous to the pro- small, with the program increasing enrollment rates by gram itself. As a final note, the design of the evaluation of 0.74-1.07 percentage points for boys and 0.96-1.45 per- RPS was very similar to that of Progresa (IFPRI 2002). centage points for girls (Schultz 2000a-d, Behrman, Sengupta, and Todd 2000a-b). This relatively small change Targeting outcomes reflects high initial enrollment rates and captures a range of The targeting performance of the CCT programs has been impacts, including improvements in starting dates, progres- quite impressive. Table 9.12 presents the distribution of sion rates, and primary completion rates. In countries with program beneficiaries across consumption deciles for four lower primary enrollment, larger impacts would be expected. programs. The percentage of program beneficiaries in the The largest impact was on enrollments in middle bottom 40 percent of the national consumption distribution school, with enrollment rates for girls increasing by varies from more than 62 percent for Progresa to almost 7.2-9.3 percentage points (that is, from 67 percent to around 89 percent for SUE In other words, households classified as 75 percent) and for boys by 3.5-5.8 percentage points (that TABLE 9.12 Cumulative distribution of beneficiaries across national consumption deciks, in percent Progresa Decile PRAF (Honduras) RPS (Nicaragua) (Mexico) SUF (Chile) 1 22.1 32.6 22 2 42.5 55 39.5 67 3 66.9 70.2 51.9 4 79.5 80.9 62.4 88.8 5 88.6 89.6 70.9 6 93.5 94.3 80.5 97.2 7 97 97.1 87.8 8 97.3 99.1 93 99.8 9 97.7 99.8 98 10 100 100 100 100 Sources: For Chile, MIDEPLAN 1998; for Honduras, Nicaragua, and Mexico, MNPTSG 2002. 277 INEQUALITY IN LATIN AMERICA is, from 73 percent to around 78 percent). Most of this With regard to RPS, Maluccio (2003) reports a decrease increase came during the transition year from primary to in the incidence ofchild labor by 8.8 percentage points (from middle school (that is, from grade 6 to 7). Among those stu- 27 percent) for 10- to 13- year-olds. For those children who dents who successfully completed primary school, Progresa work, the number of hours worked also decreased by 9 hours increased enrollment rates in the first year of middle school "in the previous week" (from an initial level of around 24 by 14.8 percentage points for girls and 6.5 percentage points hours). The levels of children who only attend school increased for boys. There is also evidence that much of this overall substantially, from 59 percent to 84 percent, with this impact increase in education rates is due to an increase in continu- attributed primarily to those children who were neither in ation rates (that is, continued enrollment by those already school nor working prior to the onset of the RPS program. enrolled), as opposed to encouraging those who had dropped Comparable results are not available for BE in Brazil due out prior to the program to return to school. to the nonexperimental design of the program. There was The accumulated effect of increased schooling in grades no random element in the assignment of treatment to com- 1-9 is consistent with an average increase in educational munities. An alternative approach, adopted by Bourguignon, attainment of 0.66 years by grade 9, or from an average of Ferreira, and Leite (2003), is to simulate the impact of the 6.2 to 6.9 years of schooling. The impact on girls is higher, conditional transfer on occupational decisions by the fami- at 0.72 extra years of schooling compared to 0.64 years for lies of the children who would qualify for the program. Using boys. It has been estimated that these educational gains are a model estimated econometrically on household survey data consistent with an increase of 8 percent in the earnings of from Brazil in 1999 and then simulating the program these children when they reach adulthood. The use of con- parameters, those authors find that BE had as substantial an ditional transfers also seems to be a much more cost-effective impact on enrollment as did Progresa and RPS. Some instrument for generating educational improvements than 8.9 percent of children ages 10-15 in households classified the increase in the building of schools that accompanied the as poor were not enrolled in school in 1999. Simulation Progresa program (Coady and Parker 2002). However, one results suggest that this number would be reduced by almost study explores whether the program could be more effective 60 percent to 3.7 percent under the transfer rules and with respect to the objective of getting kids into school, and amounts stipulated in BE. finds scope for substantial improvements (de Janvry and Results for child labor are somewhat less clear in the BE Sadoulet 2003). This is because many transfers go to chil- case, however. The share of children enrolled in school and dren in very poor households who would have attended not working would rise by only 4 percentage points, from school in any case, especially at primary levels, while some 68 percent to 72 percent. The greatest incidence of child labor children in nonpoor households who do not attend school would remain in the category of children who both study and are excluded from the program. Whether programs such as work, which would account for approximately one-quarter these should be designed to maximize impacts on changes of all children ages 10-15 who live in poor households. in school attendance or on reducing extreme poverty is an Impacts on health and nutrition have also been substantial. assessment that has to be made in a country-specific context. Prior to the Progresa program, stunting levels (that is, low Both objectives matter-and in this report significant weight height-for-age) among children ages 12-36 months were very is given to the second objective of extending the current high, at 44 percent. The program had a substantial effect on social protection system to the poor-but there will be trade- reducing the probability of stunting, increasing the annual offs between them. mean growth rate by 16 percent (or 1 centimeter per year) Evaluations of both Progresa and RPS also find that the for children receiving treatment in this critical age range. programs have reduced child labor. The results for Progresa The potential for increases in the health and nutrition indicate that the program led to a substantial reduction in status of households as a whole can also be seen from the the incidence of child labor among boys and girls in both fact that median food expenditures in 1999 were 11 per- salaried and nonsalaried activities. For example, labor force cent higher than in 1998. Since this increase is mainly due participation rates for boys decreased by 15-25 percent due to higher expenditures on fruits, vegetables, and meat and to the program. Nonetheless, a substantial number of chil- other animal products, it is consistent with substantial dren continue to combine work with school. improvements in the availability of crucial micronutrients. 278 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS By November 1999, the median calorific value of household For example, there is evidence that the percentage of diets had also risen by 7.1 percent, accompanied by a clear households receiving transfers from other government pro- sense among households that they were eating better. Even grams decreased due to Progresa, as did the income received controlling for the amount of the program transfer, there is by children ages 8-17 (Skoufias 2001). Simulations that evidence that the acquisition of calories from fruits, ignore such responses suggest that, in these communities, vegetables, and animal products increased consistently with the program decreases the poverty gap index by 30 percent a change in consumption behavior, which is one of the and the severity of the poverty index by 45 percent (Skoufias, objectives of the nutrition information meetings. There is Davis, and de la Vega 2001). Regression estimates using also evidence that this dimension of the program has had actual outcomes (which incorporate responses) find that the positive spillover effects on nonbeneficiaries in the same com- poverty gap and poverty indices decline by 36 percent and munity. Finally, there is no evidence to support the view that 46 percent, respectively, suggesting that adverse indirect nutritional supplementation would "crowd out" calorie con- income effects are negligible (Handa and others 2000, sumption from other foods. Skoufias 2001). This interpretation is supported by results Progresa has also substantially increased preventive health that find that the labor force participation rates of men and care visits, including an increase of 8 percent in visits by women have not decreased as a result of participation in Pro- women in their first trimester of pregnancy. This has in turn gresa (Skoufias and Parker 2001). In addition, one analysis led to substantial improvements in the health of babies and did not find any evidence that program transfers crowd out pregnant women. There has also been a substantial decrease previously existing public transfers (Teruel and Davis 2000). in the incidence of illness, ranging from a 25 percent decrease Similar results were found for the RPS. That program among newborns, a 19 percent decrease for children ages 0-2 increased household per capita expenditures by 19 percent, years, and a 22 percent decrease among children ages 3-5 which is almost identical to the average magnitude of the years. Child height has increased by between 1 and 4 per- other transfers (IFPRI 2001). These analyses of the impact cent and children are on average 3.5 percent heavier. The of the program on consumption also help to highlight the prevalence of anemia among children 24-48 months has importance of including a control group in studies. For decreased by 19 percent. Health status appears to have example, over the course of the year in which the analysis improved among adults as well, with this group reporting was conducted, per capita consumption in RPS control on average 19 percent fewer days with difficulties under- households decreased by 16 percent due to the ongoing crisis taking daily activities due to illness, 17 percent fewer days in the coffee industry, while per capita consumption of being incapacitated due to illness, 18 percent fewer days increased by around 4 percent among the households par- in bed due to illness, and being able to walk 7.5 percent fur- ticipating in the program. Thus, the consumption impact ther without getting tired. of the program arose from its ability to protect household incomes in the presence of an adverse economic shock. Impacts on current poverty levels As in the case of child labor, simulated results for the BE In both Progresa and RPS, transfers represent a substantial program suggest a less pronounced impact on current increase in the total incomes of beneficiary households, cor- poverty levels. Bourguignon, Ferreira, and Leite (2003) find responding on average to around 20 percent. However, ben- that, under the assumptions in their model, the incidence eficiaries of these programs are required to forgo any existing of poverty in Brazil would fall by 1.3 percentage points (from transfers from other programs. Due to the potential for 30.1 percent) and that the FGT(2) measure of the severity household responses that reduce income and consumption of poverty would fall from 7.9 to 6.8 percent due to the BE (for example, reduced income from child labor, changes in program. Since targeting is assumed perfect in the simula- the adult labor supply, or reductions in private transfers), this tion, these modest impacts are evidently due to the small means that the effect of the programs on consumption and size of the transfers. Larger poverty declines were obtained welfare may in fact be somewhat lower. (For more on child for larger (simulated) program budgets. labor and labor supply responses, see Skoufias and Parker As a final word, one should also consider the costs of con- 2001; for more on private transfers, see Teruel and Davis ditional cash transfer programs. The available evidence sug- 2000.) gests that despite their various successes, these programs are 279 INEQUALITY IN LATIN AMERICA not particularly expensive. Coady, Perez, and Vera-Llamas tax effort and Mexico for its low tax receipts; among lower- (2002) found that the administrative costs of Progresa income countries, Nicaragua is a case of a high and Guatemala accounted for less than 9 percent of total program costs, of a low tax effort. implying that it cost less than a peso for every ten pesos Can taxes be raised in a more progressive fashion? The transferred to households. This number is relatively low com- authors have been very cautious about average incidence stud- pared to those calculated for other programs such the ies, while also recognizing that it is important to pay atten- LICONSA (subsidized milk) and TORTIVALES (subsidized tion to the distributional dimensions of tax reform. The tortilla) programs, which cost 40 and 14 pesos per 100 trans- analysis in this report suggests that in most countries there ferred to beneficiaries, respectively. In addition, since is scope to both reduce the tax burden on the poorest mem- Progresa has large up-front, establishment costs, this bers of society and raise more of the required additional rev- number should decrease over time. For example, if the costs enue from higher income groups. This process is likely to associated with the geographic and household proxy-means involve eliminating or reducing informal quasi-taxes, such targeting methods are removed, it costs less than 6 pesos to as the inflation tax and local corruption. It involves revisit- transfer 100 pesos to beneficiaries. ing the detail of indirect taxes in order to reduce rates on particularly regressive excises and to keep or consider exemp- 9.4. Conclusions tions of basic food items from VATs. The process also involves As discussed in chapters 7 and 8, deep reforms in economic enforcing the collection of property and income taxes and institutions and policies and a broader, more egalitarian pat- eliminating most exemptions and loopholes, without nec- tern of asset ownership are two fundamental components in essarily increasing the marginal rates of the latter. the process of breaking with a long history of persistent On the spending side, there is wide variation across coun- inequalities in Latin America. This chapter has considered tries and even greater variation across programs. Much a third key element: a reform in the redistributive role of spending is regressive, but most is less regressive than the state itself through both taxes and public expenditures. income patterns. Pension systems stand out for their regres- In all areas, this reform process should be complemented by sivity, as do tertiary education and health programs associ- changes in political institutions, as discussed in chapter 5. ated with social security systems. Overall, however, social Throughout this report, the authors' analysis has been spending tends to dampen inequality. The biggest redis- guided by an understanding that existing taxation and tributive impact generally comes from large-scale, near- expenditure patterns reflect complex political equilibria, universal programs, including in basic education and which in turn are the result of a long social and economic health. The highest level of targeting toward the poor comes history. From a normative viewpoint, however, the justifi- from an impressively large range of social assistance and cation for taxation in any country is that taxes need to be focused propoor programs. However, these contributions raised in order either to pay for goods or services which, tend to be relatively small in the overall spending picture. because of market failure, require subsidies or are better pro- During the 1990s, there was a large increase in social duced by the government, or to make transfers to other spending per capita in most Latin American countries due to households because of equity-related motives. the combined effect of economic growth and an increase in From this normative point of view, it is probably true that both the ratio of total expenditures to GDP and the share of most countries in the region raise inadequate levels of taxes. social spending in total expenditures. Although a significant There have been modest gains in tax effort over the past few part of these trends was related to large increases in pension decades, including as recently as the 1990s. However, many outlays-which in Latin America are on average as badly dis- countries in Latin America still have levels of taxes signifi- tributed as incomes-there were also major increases in pro- cantly below what is needed to support rapid social and grams such as basic education and health and targeted social economic development and efficient redistribution, and less assistance. Indeed, significant increases in the largest social than prevailing levels in countries in other regions with sim- programs (that is, basic education and health) were highly ilar levels of income per capita. There is also substantial vari- redistributive, since access by high- and medium-income ation among the countries. For example, among the families was already significant and most of the increase in upper-middle-income countries, Brazil stands out for its high access benefited the bottom two quintiles in particular. 280 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS Latin America has experimented with a variety of targeted ventions are expensive and require considerable adminis- programs and mechanisms. Some of these have been highly trative capacity. In addition, existing CCTs are relatively successful in reaching the poor, while in others the leakage small (in comparison to national populations) and categor- of subsidies to the nonpoor is quite large. Unfortunately, the ical in selection. Because participation is usually linked to better targeted programs tend to be small, and thus have lim- school year cycles and other structures, beneficiary selection ited coverage even among the poor (with some encouraging is not usually very flexible, which may impair their exceptions such as the Oportunidades program in Mexico). effectiveness as social insurance (or risk management) There thus seems to be a tradeoff between the degree of effec- programs. Nevertheless, given their various attractive fea- tive targeting and the degree of coverage of the poor. This tures, effort is being made to expand the coverage of and may be partially explained by administrative limitations, but refine CCTs so that they provide better coverage against risk, mostly by political economy considerations: although whether aggregate or idiosyncratic.21 better, more progressive programs have been proposed, the CCT programs have been proven to reach the poorest beneficiaries of the older, larger, regressive expenditures are parts of the population in a way that other programs could well-organized and resist reductions in their budgets. Under not. They complement efforts aimed at improving the under- tight overall budget constraints, the tradeoff between large lying distribution of assets because they enhance the supply programs and well-targeted ones persists. of health and education services to poor people. CCTs also It is difficult to achieve support to enlarge highly targeted present fewer incentive problems than do traditional, uncon- programs, precisely because they do not benefit middle- ditional transfers. If they can be expanded to provide uni- income groups. Obviously (as discussed in chapter 5), such versal social insurance coverage-a process that is currently limitations can be overcome through political leadership and being tested in Brazil-they may well deserve a place as one the adequate "bundling" of programs to include both those of the pillars of a new, more inclusive, social security system with universal coverage and those that are highly targeted. in Latin America. Such a system would, together with open In particular, the expansion of programs that attempt uni- and more democratic institutions, deeper markets, and more versal coverage can be successfully coupled with conditional, egalitarian asset distributions, represent a real break with the targeted transfers (such as those in Oportunidades in Mexico region's history of inequality. or Bolsa Familia in Brazil) that provide cash transfers to the poor with the stipulation that families keep kids in Notes school and attend maternal, child, and other health-related 1. The phrase "need not necessarily" deliberately glosses over cen- programs. turies of debate, in both political philosophy and economics, over what The third part of this chapter focused on CCT programs. constitutes social justice and what is the optimal role of the state in A review of the evidence suggests that these programs rank redistribution. Opinions have ranged from the view that the state among the most progressive in Latin America, and indeed should redistribute only in order to correct some past injustice or coer- in the developing world as a whole. They also tend to have cive expropriation (Nozick 1974), to the view that equality is a goal a substantial impact on current income levels among ben- in its own right that should be given weight independently of the level eficiaries, thus contributing to measurable reductions in involved. In between, a variety of utilitarian views indicate that social current poverty levels. In addition, by their very nature welfare depends on the distribution of outcome levels, with positive they contribute to greater investments by the poor in their value placed on individual levels but at a nonincreasing rate. Within own human capital. The evidence suggests that these this school, the two extremes are Jeremy Bentham's view that social impacts are real in terms of both educational attainment utility is simply the unweighted sum of individual utilities and John and health outcome indicators, particularly in Mexico and Rawls's argument that the entire weight should be placed on the wel- Nicaragua. fare of the worst-off individual. These arguments are revised in greater Conditional cash transfers (CCTs) are not a new panacea. detail in a number of places, including Sen 2000. While this is not They are likely to be most successful when combined with the place to discuss them further, it should be acknowledged that supply-side interventions that improve the availability and readers' degree of sympathy with these different schools of thought is quality of the public services in which beneficiaries have to likely to affect the resonance of some of the arguments presented in participate. Such combined supply- and demand-side inter- this chapter, which is written largely from a utilitarian viewpoint. 281 INEQUALITY IN LATIN AMERICA 2. There has, however, been some resurgence in export taxes in vaccine, three doses of Diptheria, Pertussis, and Tetanus (DPT), or all cases of recent crises (for example, in Argentina and the Dominican vaccines (that is, Bacillus Calmette-Guerin (BCG), three doses each Republic). of DPT and oral polio, and measles). 3. For other recent reviews, see Martfnez-Vizquez (2001) and 13. The concentration coefficient is an index of inequality along Gemmell and Morrissey (2002). For an earlier, detailed examination some dimension-in this case public spending-that is calculated in of tax incidence in Latin America, see Bird and De Wulf (1973). the same way as the Gini coefficient, but with individuals or house- 4. To illustrate, Harberger (2003) suggests that Engel, Galetovic, hold ranked according to their income. and Raddatz (1998) underestimate the regressivity of the Chilean 14. For more on this point, see de Ferranti and others (2000), system because they overestimate the progressivity of the corporate Glewwe and Hall (1998), Lustig (1995 and 1999), and Ganuza, Taylor, income tax in an open economy. and Morley (1998). 5. Engel, Galetovic, and Raddatz (1998) recognize this argument, 15. For more on this point, see Perry (2003), Gavin and Perotti but suggest that it is not relevant in the case of Chile, which is viewed (1997), and Talvi and Vegh (2000). See also World Bank (2003e) on essentially more like a European (that is, developed) than a Latin the nature and extent of the procyclicality of fiscal policy in Caribbean American country. nations. 6. As one small illustration of this effectiveness, note that the VAT 16. For reviews of these types of programs, see Patrinos (2002); evasion rate in Chile is estimated to be about 20 percent, compared Coady, Grosh, and Hoddinott (2003); Rawlings and Rubio (2003); and to the range of 30 percent (Uruguay) to 68 percent (Peru) that is esti- Morley and Coady (2003). In this chapter, focus is placed solely on mated for other large Latin American countries (Sour 2003). Similarly, programs in Latin America. Lledo, Schneider, and Moore (2003) note that the shadow economy in 17. Although the transfers are paid in cash, the fact that they are Chile is 20 percent of GDP, compared to the Latin American average conditional on the "consumption" of certain public services (such as of 41 percent. attending public schools or visiting health care facilities) turns them 7. Results for Denmark come from a Danish study (as reported into subsidies. Since those services are usually already provided free in translation in a comment on a draft of this report by Andreas Blom), of charge, what is being subsidized are the indirect (for example, while studies for the United Kingdom and Canada are reported in travel) costs involved or the opportunity costs of consumption. Key Bird (2003). among the latter is the opportunity cost of a child's school attendance 8. In theory, in order to minimize efficiency losses, different tax time, in terms of forgone earnings from child labor or the forgone rates should be imposed on each commodity, with higher rates imposed value of domestic work. Since the transfers are subsidies, there must on those goods and services for which changes in behavior are the small- be a market failure that makes them preferable to unconditional cash est. To do so, however, requires much more information about how transfers; otherwise, families could be trusted to make the optimal taxes alter behavior than is available in most countries. Moreover, this decisions for themselves with the additional resources. Three reasons approach does not take administrative and equity concerns into are thought to justify the paternalism implicit in these conditions: account. For these reasons, in practice it seems generally advisable to (1) externalities are generated from keeping kids healthy and in school; impose a uniform tax rate to the greatest extent possible. (2) child health and education decisions are the outcome of an 9. To illustrate, as Bird and Miller (1989) show in the case of intrafamily bargain, and thus not necessarily optimal from the social Jamaica, exempting just five specific items from VAT reduces the viewpoint; and (3) discount rates among very poor families may be burden imposed on the poorest 40 percent of the population by half. myopically high. 10. The analysis by ECLAC is especially interesting because in sev- 18. Bolsa Escola and Bolsa Alimentac,ao-together with two other, eral of the larger, more decentralized countries (including Argentina, smaller programs-were integrated into a single initiative, known as Brazil, and Colombia, but not Mexico), the authors were able to include Bolsa Famflia, in late 2003. The discussion here is based on data from spending by local governments in the estimates of total social spending. when the individual programs were still run separately. 11. Net enrollment rates are the ratio of children attending school 19. Although similar programs exist in other countries in Latin at a particular level and at the age appropriate for the level in rela- America (for example, Subsidio Unico Familiar in Chile, Familias en tion to all children of that age; gross enrollment rates include chil- Acci6n in Colombia, Bono Solidario in Ecuador, and the Program for dren of all ages attending school at a particular level. Advancement through Health and Education in Jamaica), much less 12. The immunization variables considered here are the percent- detail is available on their design, implementation, and impact. For age of children in the appropriate age bracket who receive the measles more details on the programs detailed in this chapter, see Skoufias 282 TAXATION, PUBLIC EXPENDITURES, AND TRANSFERS (2001) and Coady (2000) on Progresa; Maluccio (2003) and IFPRI which in effect disproportionately excluded very poor elderly couples (2002) on RPS; IFPRI (2001) on PRAF; Bourguignon, Ferreira, and and households without children. It was thus decided that the eligi- Leite (2003) on BE; and IFPRI (2003) on BA. bility rate for Progresa would be increased to around 80 percent, with 20. Initially, around 50 percent of households concentrated in the the result that many previously excluded households were now most marginal communities were deemed eligible. However, after included. discussions with community leaders, it was decided that the propor- 21. See, for example, Camargo and Ferreira (2001), which partly tion of households being classified as poor was too low. This reflected inspired the consolidation of Brazil's cash transfer programs under Bolsa the crudeness of the statistical procedures employed and the fact that Familia in 2003. having children was weighted too heavily in the classification outcome, 283 Statistical Appendix These statistics and other results of the current study are periodically updated with new information. See http://www.depeco.econo.unlp.edu.ar/cedlas/wb/. 285 INEQUALITY IN LATIN AMERICA TABLE A.1 Household surveys in Latin America: coverage and characteuiscs Does the survey report Name of Sample size Population Nonlabor Nonmonetary Implicit rent Year survey Coverage individuals (in millions) Expenditures? income? income? own housing? Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) Argentina 1992 EPH Urban 67,776 33.4 No Yes No No 1996 EPH Urban 63,387 35.2 No Yes No No 2001 EPH Urban 48,048 37.5 No Yes No No Bolivia 1992 EIH Urban 28,502 6.9 Yes Yes No No 1996 ENE National 35,648 7.6 No Yes No No 1999 ECH National 13,031 8.1 Yes Yes Yes Yes Brazil 1990 PNAD National 306,493 144.7 No Yes No No 1995 PNAD National 334,106 155.8 No Yes No No 2001 PNAD National 378,837 172.6 No Yes No No Chile 1990 CASEN National 105,189 13.1 No Yes Yes Yes 1996 CASEN National 134,262 14.4 No Yes Yes Yes 2000 CASEN National 252,748 15.2 No Yes Yes Yes Colombia 1992 ENH-FT Urban 13,936 36.4 No Yes Yes No 1996 ENH-FT National 137,423 39.3 No Yes Yes No 1999 ENH-FT National 152,298 41.6 No Yes Yes No Costa Rica 1990 EHPM National 36,272 3.0 No Yes No No 1995 EHPM National 40,613 3.3 No Yes No No 2000 EHPM National 40,509 3.6 No Yes No No Dominican Republic 1995 ENFT National 23,730 7.7 No No No No 1997 ENFT National 15,842 8.0 No Yes Yes No Ecuador 1994 ECV National 20,873 11.2 Yes Yes Yes Yes 1998 ECV National 26,129 12.2 Yes Yes Yes Yes El Salvador 1991 EHPM National 90,624 5.4 No Yes No Yes 1995 EHPM National 40,004 5.7 No Yes No Yes 2000 EHPM National 71,665 6.3 No Yes Yes No Guatemala 2000 ENCOVI National 37,771 11.4 Yes Yes Yes Yes Honduras 1990 EPHPM National 47,056 4.8 No No No No 1995 EPHPM National 29,804 5.6 No No No No 1999 EPHPM National 33,772 6.4 No Yes Yes No Jamaica 1990 JSLC/LFS National 8,269 2.4 Yes Yes No No 1996 JSLC/LFS National 8,280 2.5 Yes Yes No No 1999 JSLC/LFS National 8,921 2.6 Yes Yes No No Mexico 1992 ENIGH National 50,862 86.4 Yes Yes Yes Yes 1996 ENIGH National 64,916 92.7 Yes Yes Yes Yes 2000 ENIGH National 42,535 98.0 Yes Yes Yes Yes Nicaragua 1993 EMNV National 25,162 4.3 Yes Yes Yes No 1998 EMNV National 22,423 4.8 Yes Yes Yes No Panama 1991 EH-MO National 38,000 2.4 No Yes No No 1995 EH-MO National 40,320 2.6 No Yes No No 2000 EH-MO National 39,562 2.9 No Yes No No Paraguay 1990 EH-MO Urban 4,795 4.2 No Yes Yes No 1995 EH-MO National 21,910 4.8 No Yes Yes No 1999 EPH National 24,193 5.4 No Yes Yes No Peru 1991 ENNIV National 11,845 22.0 Yes Yes Yes Yes 1994 ENNIV National 18,662 23.1 Yes Yes Yes Yes 2000 ENNIV National 19,961 25.7 Yes Yes Yes Yes Trinidad & Tobago 1992 ECV National 6,220 1.2 Yes No No No Uruguay 1989 ECH Urban 31,766 3.1 No Yes Yes Yes 1995 ECH Urban 64,930 3.2 No Yes Yes Yes 2000 ECH Urban 57,984 3.3 No Yes Yes Yes RB de Venezuela 1989 EHM National 224,172 18.9 No No No No 1995 EHM National 92,450 21.8 No Yes Yes Yes 1998 EHM National 80,311 23.4 No Yes Yes Yes Note: EPH = Encuesta Permanente de Hogares; EIH = Encuesta Integrada de Hogares; ENE = Encuesta Nacional de Empleo; ECH = Encuesta Continua de Hogares; PNAD = Pesquisa Nacional por Amostra de Domicilios; CASEN = Encuesta de Caracterizaci6n Socioecon6mica Nacional; ENH-FT = Encuesta Nacional de Hogares-Fuerza de Trabajo; EHPM = Encuesta de Hogares de Propositos Multiples; ENCOVI = Encuesta Nacional sobre Condiciones de Vida; ENFT = Encuesta Nacional de Fuerza de Trabajo; ECV = Encuesta de Condiciones de Vida; EPHPM = Encuesta Permanente de Hogares de Propositos Multiples; JSLC = Jamaica Survey of Living Conditions; LFS = Labor Force Survey; ENIGH = Encuesta Nacional de Ingresos y Gastos de los Hogares; EMNV and ENNIV = Encuesta Nacional de Hogares Sobre Medicion de Niveles de Vida; EH-MO = Encuesta de Hogares-Mano de Obra; EHM = Encuesta de Hogares por Muestreo. 286 TABLE A.2 Distribution of household per capita incomie: share of deciles and income rafos Share of deciles Income ratios 1 2 3 4 S 6 7 8 9 10 10/1 90/10 9S/80 Country (i (li (Dii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xi!) (xiii) Argentina 1992 1.8 3.1 4.1 5.2 6.3 7.7 9.4 12.1 16.6 33.8 18.5 7.8 2.0 1996 1.4 2.7 3.7 4.7 5.9 7.3 9.2 11.9 16.8 36.4 25.4 9.6 2.2 1998 1.3 2.6 3.6 4.6 5.7 7.0 9.0 11.8 16.6 37.8 29.0 10.3 2.4 2001 1.0 2.1 3.1 4.1 5.4 6.9 9.0 12.0 17.5 38.9 39.1 13.8 2.4 Bolivia Urban 1992 1.7 2.8 3.7 4.6 5.6 6.8 8.5 11.1 15.6 39.5 23.1 8.4 2.3 1996 1.7 2.7 3.5 4.4 5.4 6.6 8.2 10.9 15.9 40.8 24.5 9.3 2.4 National 1996 O.S 1.5 2.6 3.7 S.0 6.4 8.4 1 1.1 16.4 44.4 81.2 20.8 2.5 1999 0.3 1.0 2.3 3.6 5.1 6.8 8.9 11.9 17.8 42.3 143.5 38.6 2.4 Brazil 1990 0.8 1.5 2.2 3.0 4.1 5.4 7.3 10.4 16.5 48.7 63.2 19.2 3.0 1995 0.8 1.6 2.4 3.3 4.3 S.S 7.4 10.3 16.3 48.1 58.0 17.6 3.0 2001 0.9 1.7 2.5 3.4 4.5 5.8 7.5 10.4 16.1 47.2 54.4 16.1 2.9 Chile 1990 1.3 2.3 3.0 3.8 4.8 6.0 7.6 10.1 15.4 45.8 36.2 11.1 2.9 1996 1.2 2.2 3.0 3.8 4.7 5.9 7.6 10.3 15.7 45.5 36.4 11.5 2.7 2000 1.2 2.2 2.9 3.7 4.7 5.8 7.4 10.0 15.2 47.0 40.6 11.4 2.9 Colombia Bogota 1992 1.0 2.2 3.0 3.9 4.9 6.3 8.1 11.0 16.8 42.7 42.0 12.9 2.8 1996 1.7 2.6 3.4 4.2 5.2 6.4 7.8 10.2 16.1 42.5 25.6 10.1 2.7 National 1996 0.9 2.1 3.0 3.9 S.0 6.2 7.9 10.4 15.1 45.4 50.3 12.3 2.7 1999 0.8 1.9 2.8 3.7 4.8 6.1 7.7 10.3 15.4 46.5 57.8 14.5 2.8 Costa Rica 1990 1.3 2.9 4.1 5.1 6.3 7.8 9.7 12.3 16.4 34.0 25.5 9.1 2.1 1995 1.4 2.9 4.0 5.1 6.3 7.7 9.6 12.2 16.5 34.2 24.1 9.0 2.0 2000 1.4 2.8 3.9 S.0 6.1 7.6 9.5 12.2 16.7 34.8 25.1 9.5 2.2 Dominican Rep. 1995 1.5 2.6 3.5 4.4 5.4 6.7 8.3 10.9 15.6 41.2 26.8 9.4 2.5 1997 1.4 2.6 3.6 4.6 5.8 7.1 8.9 11.5 15.8 38.6 28.4 9.5 2.3 Ecuador 1994 0.9 2.2 3.1 4.1 5.2 6.6 8.2 10.6 15.5 43.7 51.2 12.9 2.6 1998 0.7 1.9 2.9 3.9 S.0 6.4 8.3 10.8 15.9 44.2 63.6 15.2 2.6 El Salvador 1991 1.1 2.3 3.3 4.3 5.4 6.7 8.5 11.1 15.7 41.5 37.4 10.8 2.4 1995 1.0 2.4 3.4 4.5 5.7 7.1 8.9 11.4 16.1 39.6 38.3 11.1 2.3 2000 0.9 2.0 3.1 4.2 S.S 6.9 8.8 11.4 16.5 40.6 47.4 14.1 2.3 Guatemala 2000 0.7 1.7 2.6 3.6 4.7 6.1 7.8 10.4 15.6 46.8 63.3 16.6 2.9 Honduras 1990 0.9 1.8 2.7 3.6 4.7 6.0 7.9 10.5 15.7 46.1 52.6 14.7 2.6 1995 1.0 2.0 2.8 3.8 4.9 6.3 8.0 10.8 16.2 44.2 44.9 13.4 2.5 1999 0.9 1.9 2.8 3.9 5.1 6.7 8.5 11.4 16.7 42.2 49.1 15.1 2.5 Jamaica 1990 1.2 2.3 3.3 4.2 5.4 7.1 9.1 11.5 15.7 40.1 32.7 10.8 2.3 1996 0.9 2.1 3.1 4.0 5.3 6.7 8.2 10.9 16.2 42.7 46.9 13.7 2.6 1999 1.1 2.3 3.3 4.3 S.S 7.0 8.9 11.5 16.1 40.1 35.5 11.2 2.3 Mexico 1992 1.0 2.1 3.0 4.0 S.0 6.3 7.9 10.4 15.6 44.8 47.1 13.2 2.5 1996 1.0 2.2 3.2 4.1 5.2 6.5 8.2 10.8 15.6 43.3 41.9 11.7 2.6 2000 1.0 2.1 3.1 4.1 5.2 6.5 8.2 10.7 16.0 43.1 45.0 12.9 2.5 Nicaragua 1993 0.8 1.8 2.8 3.8 4.9 6.5 8.6 11.4 16.5 43.0 55.3 15.5 2.4 1998 0.8 1.9 2.9 4.0 5.2 6.6 8.3 11.0 15.6 43.9 56.2 14.6 2.3 Panama 1991 O.S 1.5 2.7 3.8 S.0 6.4 8.6 11.9 17.7 42.0 80.5 22.7 2.4 1995 0.6 1.7 2.7 3.8 S.0 6.5 8.5 11.6 17.0 42.5 69.4 17.7 2.5 2000 0.7 1.7 2.7 3.8 4.9 6.3 8.3 1 1.3 17.0 43.3 62.3 18.2 2.6 Paraguay 1995 0.7 1.5 2.4 3.4 4.5 6.0 7.8 10.5 15.6 47.5 67.9 18.0 2.8 1999 0.6 1.6 2.7 3.8 S.0 6.5 8.4 11.2 16.5 43.8 70.4 19.0 2.4 Peru Regions 1991 1.1 2.5 3.8 5.1 6.5 8.1 9.9 12.6 16.8 33.7 30.9 11.3 2.0 1994 1.0 2.5 3.7 4.9 6.1 7.8 9.7 12.2 16.4 35.6 34.1 11.0 2.0 National 1994 1.0 2.4 3.6 4.9 6.1 7.6 9.7 12.2 16.6 35.9 36.5 1 1.3 2.0 2000 0.8 2.3 3.6 4.8 6.3 7.8 9.5 12.0 16.0 36.9 46.2 12.2 2.1 Trinidad & Tobago 1992 0.9 2.3 3.6 4.9 6.0 7.4 9.2 11.9 17.2 36.6 40.6 12.7 2.1 Uruguay 1989 2.0 3.4 4.5 5.6 6.8 8.0 9.7 11.9 15.7 32.4 16.0 6.5 1.9 1995 1.8 3.2 4.4 S.S 6.7 8.0 9.7 12.2 16.4 32.1 17.6 7.6 2.0 2000 1.8 3.0 4.1 5.2 6.4 7.8 9.5 12.1 16.6 33.5 18.9 8.1 2.1 RB de Venezuela 1989 1.7 3.1 4.2 5.3 6.5 7.9 9.7 12.2 16.3 33.2 19.5 7.9 2.0 1995 1.5 2.8 3.8 4.9 6.2 7.6 9.4 11.9 16.3 35.6 23.6 8.7 2.1 1998 1.3 2.7 3.7 4.9 6.1 7.6 9.4 12.0 16.7 35.6 28.2 9.5 2.2 Notes: Column (xi) = income ratio between deciles 10 and 1; column (xii) = income ratio between percentiles 90 and 10; column (xiii) = income ratio between percentiles 95 and 80. Data for Dominican Republic (1995) and for Honduras, Trinidad and Tobago, and RB de Venezuela (1989) include only monetary income from labor sources. Source: Authors' calculations based on microdata from household surveys. 287 TABLE A.3 Distribution of household per capita income: inequality indices Gin! Theil CV A(.5) A(l) A(2) E(0) E(2) Country (i) (ii) ONi) (iv) (v) NOi (vii) (viii) Argentina 1992 44.7 0.362 1.074 0.162 0.295 0.503 0.349 0.576 1996 48.2 0.435 1.249 0.191 0.343 0.588 0.419 0.780 1998 49.5 0.451 1.208 0.200 0.359 0.599 0.444 0.730 2001 52.2 0.497 1.276 0.223 0.405 0.677 0.517 0.814 Bolivia Urban 1992 49.5 0.490 1.408 0.203 0.346 0.5411 0.425 0.991 1996 51.1 0.532 1.539 0.216 0.363 0.551 0.450 1.185 National 1996 57.6 0.675 1.846 0.282 0.493 0.790 0.679 1.704 1999 57.8 0.633 1.643 0.287 0.532 0.851 0.760 1.349 Brazil 1990 61.2 0.767 2.062 0.310 0.511 0.739 0.716 2.125 1995 60.0 0.735 1.875 0.299 0.494 0.722 0.681 1.759 2001 59.0 0.713 1.866 0.289 0.481 0.714 0.655 1.740 Chile 1990 55.9 0.668 1.944 0.262 0.430 0.655 0.562 1.889 1996 56.1 0.652 1.803 0.261 0.431 0.651 0.564 1.626 2000 57.1 0.703 2.022 0.274 0.447 0.674 0.592 2.043 Colombia Bogota 1992 54.6 0.569 1.440 0.246 0.434 0.778 0.569 1.037 1996 52.4 0.540 1.422 0.224 0.374 0.551 0.469 1.010 National 1996 56.1 0.707 2.811 0.270 0.447 0.701 0.593 3.951 1999 57.6 0.721 2.191 0.282 0.469 0.728 0.633 2.399 Costa Rica 1990 45.6 0.381 1.11 1 0.173 0.321 0.581 0.387 0.617 1995 45.7 0.383 1.111 0.173 0.319 0.573 0.384 0.617 2000 46.5 0.389 1.083 0.177 0.326 0.581 0.396 0.586 Dominican Rep. 1995 51.5 0.542 1.578 0.221 0.371 0.563 0.462 1.244 1997 49.7 0.498 1.520 0.207 0.359 0.580 0.444 1.155 Ecuador 1994 54.8 0.627 1.758 0.255 0.436 0.706 0.573 1.546 1998 56.2 0.658 1.866 0.269 0.463 0.755 0.623 1.741 El Salvador 1991 52.7 0.585 1.867 0.236 0.402 0.655 0.514 1.742 1995 51.3 0.526 1.511 0.223 0.393 0.669 0.499 1.141 2000 53.2 0.582 1.914 0.241 0.422 0.699 0.548 1.831 Guatemala 2000 58.3 0.697 1.823 0.285 0.482 0.739 0.656 1.662 Honduras 1 990 57.8 0.733 2.295 0.283 0.466 0.696 0.627 2.633 1995 56.1 0.653 1.793 0.264 0.444 0.678 0.586 1.608 1999 SS.0 0.586 1.525 0.251 0.440 0.705 0.580 1.163 Jamaica 1990 51.7 0.520 1.406 0.222 0.388 0.637 0.491 0.988 1996 54.4 0.583 1.535 0.247 0.427 0.685 0.558 1.178 1999 52.0 0.585 1.954 0.232 0.394 0.627 0.501 1.909 Mexico 1992 55.9 0.667 1.935 0.264 0.441 0.685 0.582 1.872 1996 54.4 0.616 1.864 0.249 0.424 0.683 0.551 1.738 2000 54.6 0.609 1.692 0.251 0.429 0.693 0.561 1.431 Nicaragua 1993 55.9 0.629 1.711 0.263 0.454 0.719 0.605 1.463 1998 55.9 0.693 2.202 0.270 0.455 0.719 0.606 2.424 Panama 1991 56.4 0.603 1.518 0.267 0.483 0.784 0.659 1.153 1995 55.9 0.593 1.465 0.261 0.469 0.771 0.632 1.073 2000 56.4 0.613 1.531 0.265 0.466 0.748 0.626 1.172 Paraguay 1995 59.5 0.728 1.830 0.297 0.497 0.742 0.688 1.675 1999 56.8 0.690 2.370 0.277 0.477 0.760 0.649 2.808 Peru Regions 1991 46.5 0.394 1.131 0.182 0.344 0.642 0.422 0.640 1994 47.9 0.444 1.338 0.196 0.362 0.666 0.449 0.895 National 1994 48.6 0.453 1.344 0.200 0.371 0.676 0.463 0.903 2000 49.4 0.477 1.358 0.211 0.392 0.721 0.497 0.923 Trinidad & Tobago 1992 49.5 0.472 1.480 0.208 0.383 0.687 0.482 1.095 Uruguay 1989 42.2 0.364 1.383 0.151 0.268 0.457 0.311 0.956 1995 42.7 0.326 0.982 0.149 0.275 0.487 0.321 0.482 2000 44.6 0.357 1.040 0.161 0.293 0.497 0.347 0.541 RB de Venezuela 1989 44.2 0.360 1.087 0.161 0.294 0.521 0.348 0.591 1995 46.9 0.418 1.230 0.183 0.327 0.571 0.398 0.757 1998 47.6 0.420 1.216 0.188 0.345 0.626 0.424 0.740 Notes: CV = coefficient of variation. A(6) refers to the Atkinson index with a CES function with parameter 6.E(6) refers to the generalized entropy index with parameter E. Ell) = Theil. Data for Dominican Republic (1 995) and Honduras, Trinidad and Tobago, and Venezuela (1 989) include only monetary income from labor sources. Source: Authors' calculations based on microdata from household surveys. 288 TABLE A.4 Distribution of equivalized household income: shar-e of dec!Ws and income ratios Share of dediles Income ratios 1 2 3 4 S 6 7 8 9 10 10/1 90/10 95/80 Country (i) (ii) (jii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii) Argentinal 1992 2.0 3.4 4.4 5.4 6.5 7.9 9.5 12.0 16.5 32.4 15.9 7.0 2.0l 1996 1.7 3.0 4.0 S.0 6.2 7.6 9.3 11.9 16.6 34.7 20.9 8.4 2.1l 2001 1.1 2.4 3.4 4.4 5.6 7.1 9.0 1 1.9 17.2 37.8 32.9 1 1.8 2.4l Bolivial Urbanl 1992 1.9 3.0 3.9 4.8 S.9 7.0 8.6 11.1 15.4 38.2 20.5 7.5 2.3l 1996 1.8 2.9 3.8 4.6 5.6 6.8 8.4 10.9 15.7 39.6 21.6 8.1 2.3l National 1996 0.6 1.6 2.8 4.0 5.3 6.7 8.7 11.3 16.3 42.8 72.9 18.6 2.5 1999 0.3 1.1 2.5 3.9 S.S 7.2 9.1 12.1 17.6 40.7 126.5 33.8 2.4l Brazil 1990 0.9 1.7 2.4 3.2 4.3 5.7 7.5 10.5 16.6 47.3 53.4 17.1 2.9 1995 1.0 1.8 2.6 3.5 4.5 5.7 7.6 10.4 16.3 46.6 48.1 15.3 2.9 2001 1.0 1.9 2.8 3.7 4.7 5.9 7.7 10.5 16.1 45.7 45.4 14.0 2.8 Chile 1990 1.4 2.5 3.2 4.0 4.9 6.0 7.6 10.1 15.3 45.0 32.5 10.1 2.9 1996 1.4 2.4 3.1 4.0 4.9 6.0 7.7 10.3 15.5 44.7 32.7 10.5 2.7 2000 1.3 2.3 3.1 3.9 4.8 5.9 7.4 10.0 15.0 46.3 37.0 10.4 2.8 Colombia Bogota 1992 1.1 2.4 3.3 4.1 5.2 6.7 8.4 11.2 17.0 40.5 36.1 11.7 2.7 1996 1.8 2.8 3.6 4.5 5.4 6.5 7.9 10.2 15.9 41.4 23.1 9.1 2.7 National 1996 1.0 2.3 3.3 4.2 5.3 6.5 8.1 10.5 14.9 43.9 44.1 11.1 2.7 1999 0.9 2.1 3.1 4.0 S.0 6.3 7.9 10.4 15.3 45.0 51.3 13.0 2.8 Costa Rica 1990 1.4 3.1 4.4 S.S 6.6 8.0 9.8 12.2 16.2 32.9 22.9 8.3 2.1 1995 1.5 3.1 4.3 5.4 6.6 7.9 9.7 12.1 16.2 33.1 21.5 8.0 2.0 2000 1.5 3.0 4.2 5.3 6.4 7.8 9.7 12.2 16.4 33.5 22.4 8.7 2.2 Dominican Rep. 1995 1.6 2.7 3.6 4.6 S.S 6.8 8.5 10.8 15.6 40.2 24.5 8.9 2.4 1997 1.5 2.8 3.8 4.9 6.0 7.3 9.1 11.6 15.8 37.3 25.3 8.9 2.2 Ecuador 1994 0.9 2.3 3.4 4.3 S.S 6.8 8.4 10.7 15.2 42.3 45.4 11.2 2.6 1998 0.7 2.1 3.2 4.2 5.3 6.7 8.5 10.9 15.8 42.6 57.5 13.7 2.5 El Salvador 1991 1.2 2.6 3.6 4.7 5.7 7.0 8.7 11.2 15.5 39.7 32.6 9.4 2.3 1995 1.1 2.6 3.7 4.8 6.0 7.3 9.0 11.4 15.9 38.1 33.7 9.9 2.3 2000 0.9 2.2 3.3 4.4 5.7 7.2 9.0 11.5 16.2 39.6 43.0 12.8 2.2 Guatemala 2000 0.8 1.9 2.9 4.0 5.1 6.4 8.2 10.6 15.6 44.6 54.2 14.6 2.8 Honduras 1990 1.0 2.0 3.0 3.9 S.0 6.4 8.2 10.9 15.9 43.9 45.7 13.2 2.5 1995 1.1 2.1 3.1 4.1 5.2 6.6 8.4 11.0 16.0 42.5 39.3 11.9 2.4 1999 1.0 2.0 3.0 4.2 5.4 7.0 8.8 11.6 16.7 40.3 42.4 13.5 2.4 Jamaica 1990 1.4 2.6 3.5 4.5 5.7 7.4 9.3 11.6 15.5 38.4 28.3 8.9 2.2 1996 1.1 2.3 3.3 4.4 5.7 7.2 8.9 11.0 16.0 40.1 37.5 11.7 2.5 1999 1.3 2.5 3.6 4.7 5.9 7.4 9.4 11.9 16.1 37.1 28.4 10.1 2.0 Mexico 1992 1.1 2.3 3.3 4.2 5.3 6.5 8.1 10.5 15.5 43.2 40.7 11.7 2.5 1996 1.2 2.4 3.4 4.3 5.4 6.7 8.3 10.8 15.6 41.8 36.1 10.6 2.6 2000 1.1 2.3 3.3 4.3 5.4 6.7 8.5 10.9 15.9 41.5 39.0 11.8 2.6 Nicaragua 1993 0.9 2.0 3.0 4.0 5.2 6.8 8.8 11.6 16.4 41.4 48.1 14.9 2.4 1998 0.8 2.0 3.1 4.2 S.S 6.8 8.6 11.1 15.5 42.3 S0.0 13.5 2.2 Panama 1991 0.6 1.6 2.9 4.0 5.3 6.8 8.8 11.9 17.5 40.5 71.6 21.1 2.4 1995 0.7 1.9 3.0 4.1 5.3 6.8 8.7 1 1.7 16.8 41.0 60.2 16.3 2.5 2000 0.8 1.9 3.0 4.1 5.2 6.6 8.5 11.4 16.9 41.5 53.5 15.7 2.5 Paraguay 1995 0.8 1.7 2.6 3.6 4.8 6.3 8.1 10.6 15.4 46.1 58.6 15.7 2.7 1999 0.7 1.7 2.9 4.2 5.3 6.8 8.6 11.3 16.4 42.1 60.3 17.0 2.4 Peru Regions 1991 1.1 2.6 4.0 5.3 6.7 8.2 10.1 12.6 16.6 32.7 28.6 10.5 1.9 1994 1.2 2.7 4.0 5.3 6.4 8.0 9.9 12.3 16.3 33.9 29.4 9.6 1.9 National 1994 1.1 2.6 4.0 5.2 6.4 7.9 9.8 12.3 16.5 34.2 31.4 10.0 1.9 2000 0.9 2.4 3.8 5.1 6.5 8.1 9.7 12.0 15.9 35.6 41.0 11.0 2.0 Trinidad & Tobago 1992 1.0 2.6 3.9 5.2 6.3 7.7 9.4 12.1 17.0 34.8 35.5 11.6 2.1 Uruguay 1989 2.3 3.7 4.7 5.7 6.8 8.1 9.7 11.8 15.5 31.7 14.0 5.8 1.9 1995 2.1 3.5 4.7 5.7 6.8 8.1 9.8 12.2 16.2 31.0 15.0 6.7 2.0 2000 2.0 3.4 4.4 S.S 6.6 7.9 9.6 11.9 16.4 32.3 15.8 7.1 2.0 RB de Venezuela 1989 1.9 3.4 4.5 5.6 6.8 8.2 9.9 12.3 16.0 31.4 16.6 6.9 1.9 1995 1.7 3.1 4.2 5.3 6.5 7.9 9.6 12.0 16.0 33.8 20.2 7.6 2.0 1998 1.4 2.9 4.0 5.2 6.4 7.9 9.6 12.1 16.5 34.0 24.6 8.4 2.1 Notes: Column (xi) = Income ratio between deciles 10 and 1; column (xii) = income ratio between percentiles 90 and 10, and column (xiii) =income ratio between percentiles 95 and 80. Data for Dominican Republic (199S) and Honduras, Trinidad and Tobago, and Venezuela (1989) include only monetary income from labor sources. Source: Authors' calculations based on microdata from household surveys. 289 TABLE A.5 DistribuUion of equivalized household income: inequaliy indices Gini Theil CV A(.5) A(l) A(2) E(O) E(2) Country (i) (Oi) (iii) (iv) (v) (vi) (vii) (viii) Argentina 1992 42.6 0.327 0.996 0.147 0.269 0.463 0.313 0.496 1996 45.8 0.391 1.155 0.173 0.311 0.541 0.373 0.667 2001 50.4 0.461 1.210 0.207 0.376 0.639 0.471 0.732 Bolivia Urban 1992 47.7 0.457 1.349 0.190 0.324 0.515 0.392 0.910 1996 49.3 0.496 1.473 0.202 0.340 0.522 0.416 1.085 National 1996 55.8 0.630 1.727 0.267 0.473 0.777 0.640 1.492 1999 55.9 0.587 1.487 0.271 0.511 0.836 0.715 1.105 Brazil 1990 59.5 0.717 1.916 0.293 0.486 0.710 0.666 1.836 1995 58.3 0.684 1.752 0.280 0.466 0.690 0.628 1.535 2001 57.2 0.665 1.780 0.271 0.453 0.681 0.603 1.584 Chile 1990 54.7 0.635 1.837 0.251 0.412 0.632 0.531 1.687 1996 54.9 0.624 1.722 0.251 0.414 0.631 0.535 1.483 2000 56.1 0.681 1.955 0.265 0.432 0.655 0.565 1.911 Colombia Bogota 1992 52.4 0.510 1.295 0.226 0.407 0.761 0.522 0.838 1996 50.8 0.508 1.374 0.211 0.354 0.525 0.437 0.944 National 1996 54.3 0.662 2.735 0.253 0.423 0.677 0.551 3.740 1999 55.8 0.676 2.102 0.266 0.446 0.706 0.591 2.209 Costa Rica 1990 43.9 0.350 1.039 0.160 0.300 0.559 0.357 0.540 1995 44.0 0.352 1.043 0.160 0.298 0.545 0.353 0.544 2000 44.6 0.354 1.008 0.163 0.304 0.553 0.363 0.508 Dominican Rep. 1995 50.2 0.516 1.524 0.210 0.354 0.543 0.437 1.161 1997 48.1 0.462 1.438 0.194 0.339 0.557 0.413 1.033 Ecuador 1994 53.0 0.585 1.670 0.240 0.415 0.688 0.535 1.395 1998 54.3 0.606 1.709 0.252 0.441 0.740 0.581 1.460 El Salvador 1991 50.5 0.536 1.726 0.218 0.375 0.626 0.470 1.490 1995 49.4 0.490 1.462 0.207 0.368 0.641 0.459 1.068 2000 51.8 0.558 1.839 0.230 0.404 0.681 0.518 1.690 Guatemala 2000 56.0 0.632 1.659 0.263 0.450 0.713 0.599 1.376 Honduras 1990 55.6 0.664 2.083 0.262 0.438 0.672 0.577 2.169 1995 54.1 0.598 1.657 0.245 0.417 0.653 0.540 1.373 1999 53.0 0.537 1.414 0.234 0.415 0.680 0.535 1.000 Jamaica 1990 49.6 0.484 1.362 0.206 0.362 0.611 0.449 0.927 1996 51.5 0.518 1.406 0.222 0.391 0.648 0.496 0.989 1999 49.0 0.513 1.756 0.206 0.357 0.584 0.441 1.541 Mexico 1992 53.9 0.612 1.773 0.246 0.415 0.659 0.536 1.571 1996 52.5 0.571 1.770 0.233 0.398 0.655 0.508 1.566 2000 52.7 0.558 1.568 0.233 0.404 0.665 0.518 1.229 Nicaragua 1993 54.2 0.583 1.602 0.246 0.431 0.697 0.564 1.283 1998 54.1 0.639 2.008 0.253 0.432 0.699 0.565 2.017 Panama 1991 54.7 0.561 1.423 0.252 0.460 0.766 0.616 1.012 1995 54.0 0.549 1.378 0.244 0.443 0.749 0.585 0.949 2000 54.4 0.565 1.440 0.246 0.437 0.721 0.575 1.037 Paraguay 1995 57.8 0.693 1.808 0.282 0.473 0.717 0.640 1.635 1999 54.9 0.640 2.314 0.259 0.451 0.735 0.599 2.677 Peru Regions 1991 45,2 0.369 1.061 0.172 0.329 0.625 0.399 0.563 1994 45.9 0.399 1.207 0.179 0.335 0.635 0.409 0.729 National 1994 46.4 0.406 1.213 0.183 0.344 0.647 0.421 0.735 2000 47.7 0.443 1.284 0.198 0.371 0.700 0.464 0.825 Trinidad & Tobago 1992 47.2 0.415 1.277 0.188 0.355 0.661 0.438 0.815 Uruguay 1989 40.8 0.344 1.359 0.142 0.250 0.425 0.287 0.923 1995 40.9 0.297 0.923 0.136 0.252 0.448 0.290 0.426 2000 42.5 0.324 0.980 0.146 0.266 0.454 0.309 0.480 RB de Venezuela 1989 41.7 0.317 0.989 0.144 0.266 0.484 0.309 0.489 1995 44.5 0.374 1.138 0.165 0.300 0.539 0.356 0.647 1998 45.5 0.382 1.133 0.173 0.321 0.598 0.387 0.642 Notes: CV = coefficient of variation. A(E) refers to the Atkinson index with a CES function with parameter e. E(e) refers to the generalized entropy index with parameter e. E(1) = Theil. Data for Dominican Republic (1995) and Honduras, Trinidad and Tobago, and Venezuela (1989) include only monetary income from labor sources. Source: Author's calculations based on microdata from household surveys. 290 STATISTICAL APPENDIX TABLE A.6 Distribution of equivalized household inconie: Gini coefficients Early 90s Mid-9Os Early OOs Change Countries (i) (ii) (iii) (iv) Argentina 42.6 45.8 50.4 7.7 Bolivia 54.3 55.8 55.9 1.6 Brazil 59.5 58.3 57.2 -2.3 Chile 54.7 54.9 56.1 1.4 Colombia 55.9 54.3 55.8 -0.1 Costa Rica 43.9 44.0 44.6 0.8 El Salvador 50.5 49.4 51.8 1.3 Honduras 55.6 54.1 53.0 -2.6 Jamaica 49.6 51.5 49.0 -0.6 Mexico 53.9 52.5 52.7 -1.2 Nicaragua 54.2 54.1 -0.1 Panama 54.7 54.0 54.4 -0.3 Peru 45.7 46.4 47.7 2.0 Uruguay 40.8 40.9 42.5 1.7 RB de Venezuela 41.7 44.5 45.5 3.8 Average (nonweighted) 50.5 50.7 51.4 0.9 Average (weighted) 51.9 51.2 51.5 -0.4 Dominican Rep. 50.2 48.1 Ecuador 53.0 54.3 Guatemala 56.0 Paraguay 57.8 54.9 Trinidad & Tobago 47.2 Notes: The Gini coefficients for Bolivia and Colombia for the early 1990s were estimated by extrapolating the changes in the Gini for urban areas (see table 3.5). A similar procedure was applied for Peru using the areas covered in 1991. To com- pute the Latin American average for the mid-1990s, a Gini of 54.1 was assumed for Nicaragua. Source: Authors' calculations based on microdata from household surveys. 291 TABLE A.7 Distribution of different concepts of houmhold income among indvduals, Gini coefficients Per capita Per capita Per capita Equivalized Equivalized Equivalized Equivalized Equivalized Per capita Per capita Per capita income income income Total Equivalized Equivalized Equivalized Equivalized Per capita income income income income income income income income Only Only labor Urban labor household income A income A income A income A income A 6 C D E Only Urban Only rural Only labor monetary monetary monetary income Age 0-10 Age 20-30 Age 40-50 Age 60-70 Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii) (xiv) (xv) (xvi) (xvii) Argentina 1992 44.7 42.6 41.9 41.9 41.3 43.0 44.7 44.5 44.7 44.5 44.5 44.2 42.9 40.0 44.1 41.8 1996 48.2 45.8 44.8 45.0 44.1 46.3 48.2 47.9 48.2 47.9 47.9 45.3 45.0 43.8 48.0 42.7 2001 52.2 50.4 49.4 49.6 48.8 50.8 52.2 53.3 52.2 53.3 53.3 47.9 51.9 46.5 50.7 47.2 Bolivia Urban 1992 49.5 47.7 47.2 47.1 46.6 47.9 49.5 49.3 49.5 49.3 49.3 48.7 47.1 46.6 49.7 48.9 1996 51.1 49.3 48.7 48.5 48.1 49.3 51.0 51.3 51.1 51.3 51.3 51.0 49.2 47.8 48.7 49.5 National 1996 57.6 55.8 55.4 55.1 54.8 55.8 50.7 59.1 57.4 57.6 57.4 51.1 58.0 54.9 52.4 57.5 61.7 1999 57.8 55.9 55.3 55.2 54.8 56.1 48.2 63.0 58.3 57.7 58.4 48.9 56.8 57.0 50.3 55.6 60.9 Brazil 1990 61.2 59.5 58.7 58.9 58.2 59.9 58.6 53.9 61.0 61.2 61.0 58.5 58.3 59.6 56.0 59.9 82.0 1995 60.0 58.3 57.5 57.6 56.9 58.6 58.0 54.2 60.6 60.0 60.5 58.2 56.9 57.8 55.7 58.8 57.4 2001 59.0 57.2 66.4 56.5 55.8 57.5 57.7 53.1 59.9 59.0 60.0 58.3 55.8 56.7 54.2 56.9 56.4 Chile 1990 55.9 54.7 54.1 54.3 53.8 54.9 54.9 58.2 56.8 57.1 55.5 55.4 55.7 52.7 54.4 55.1 1996 56.1 54.9 54.4 54.6 54.1 55.2 55.2 49.9 58.0 57.6 56.4 55.3 56.1 52.9 54.9 53.7 2000 57.1 56.1 55.5 55.9 55.3 56.4 56.5 52.4 58.5 57.8 56.9 55.5 57.9 52.6 59.4 52.7 Colombia Bogota 1992 54.6 52.4 51.6 52.2 51.3 52.9 54.6 55.4 56.0 55.5 55.5 51.7 52.5 50.7 52.2 55.0 1996 52.4 50.8 50.3 50.1 49.8 51.0 52.4 52.6 53.1 52.9 52.9 51.8 49.7 48.0 53.0 49.5 National 1996 56.1 54.3 53.5 53.4 52.9 54.5 53.6 50.2 55.0 56.9 55.3 52.9 54.4 53.1 51.2 55.4 57.7 1999 57.6 55.8 55.0 55.2 54.4 56.3 55.1 55.0 57.2 58.6 57.7 55.9 55.1 55.1 52.0 58.9 58.2 Costa Rica 1990 45.6 43.9 43.3 43.2 42.7 44.0 42.9 43.2 45.5 45.6 45.5 43.2 45.0 43.8 41.2 42.9 46.8 1995 45.7 44.0 43.3 43.3 42.8 44.1 42.9 43.7 45.5 45.7 45.5 43.1 44.9 44.8 40.1 45.1 45.1 2000 46.5 44.6 44.1 43.8 43.4 44.8 44.2 44.0 46.4 46.5 46.4 44.7 46.4 44.0 42.7 43.4 48.6 Dominican Rep. 1995 51.5 50.2 49.6 49.7 49.2 50.5 53.5 44.4 51.5 51.5 51.4 53.4 50.3 49.0 48.0 53.2 52.4 1997 49.7 48.1 47.5 47.3 46.9 48.3 48.0 47.5 48.8 50.0 49.0 47.5 49.7 46.5 45.1 47.3 52.2 Ecuador 1994 54.8 53.0 52.2 52.3 51.6 53.3 51.8 51.5 54.0 55.3 55.3 52.2 52.8 53.8 50.4 55.3 53.7 1998 56.2 54.3 53.5 53.6 52.9 54.8 52.2 54.1 55.1 57.3 56.6 52.1 54.4 52.6 50.8 55.1 60.8 El Salvador 1991 52.7 50.5 49.7 49.5 48.9 50.7 49.5 47.8 54.9 52.7 54.9 50.0 50.8 49.4 48.6 51.1 54.3 1995 51.3 49.4 48.6 48.5 47.9 49.6 47.4 44.1 51.5 51.3 51.5 47.4 49.7 49.6 46.6 50.2 50.8 2000 53.2 51.8 51.0 51.0 50.4 52.2 50.3 46.8 52.8 53.2 52.8 50.8 51.3 48.2 49.0 50.6 59.2 Guatemala 2000 58.3 56.0 55.3 54.8 54.3 56.2 54.2 50.7 57.5 59.0 58.0 54.0 55.8 54.1 53.7 56.3 61.6 Honduras 1990 57.8 55.6 54.8 54.8 54.1 55.9 55.3 49.4 57.8 57.8 57.8 55.3 55.2 52.6 54.9 57.6 64.1 1995 56.1 54.1 53.3 53.2 52.6 54.4 52.2 55.5 56.1 56.1 56.1 52.2 53.6 51.9 50.4 57.3 54.3 1999 55.0 53.0 52.3 52.2 51.6 53.3 50.2 53.2 55.0 55.0 55.0 50.2 52.9 52.9 50.3 53.5 54.8 Jamaica 1990 51.7 49.6 48.7 48.8 48.1 49.9 49.8 52.4 52.6 51.7 52.6 50.3 48.8 49.5 48.9 49.1 45.9 1996 54.4 51.5 50.3 50.4 49.4 51.7 59.1 48.0 58.0 54.4 58.0 61.5 50.4 45.3 50.4 52.6 52.4 1999 52.0 49.0 47.4 47.9 46.5 49.7 54.9 46.8 55.4 52.0 55.4 57.3 48.2 46.5 46.8 54.8 49.1 Mexico 1992 55.9 53.9 53.0 53.1 52.3 54.2 52.7 52.4 56.3 55.2 55.5 52.4 52.8 52.6 52.5 56.8 59.5 1996 54.4 52.5 51.5 51.8 50.9 52.9 51.7 50.8 55.7 53.8 55.0 51.8 50.3 51.8 50.9 53.4 52.4 2000 54.6 52.7 51.8 52.0 51.3 53.0 50.9 52.1 55.0 54.6 54.9 50.8 51.3 51.2 49.9 53.6 57.6 Nicaragua 1993 55.9 54.2 53.3 53.6 52.9 54.6 52.5 53.9 56.3 57.7 57.6 54.4 53.4 54.1 52.2 53.4 53.1 1998 55.9 54.0 53.5 53.4 52.9 54.2 53.0 54.3 55.5 58.4 57.8 54.8 55.4 52.4 49.3 62.4 53.9 Panama 1991 56.4 54.7 53.8 54.0 53.3 55.1 54.4 56.4 54.4 54.0 54.9 51.0 55.5 57.0 1995 55.9 54.0 53.2 53.3 52.6 54.4 51.5 52.7 55.1 55.9 55.1 52.5 53.5 54.8 49.9 53.3 55.5 2000 56.4 54.4 53.6 53.5 52.9 54.7 52.2 54.2 56.9 56.4 56.9 53.2 54.4 53.6 50.6 52.5 59.0 Paraguay 1995 59.5 57.8 57.3 57.0 56.6 57.8 53.3 56.7 59.9 50.7 48.7 56.7 57.2 53.5 58.4 58.2 1999 56.8 54.9 54.3 53.9 53.5 54.8 50.3 59.9 57.2 65.7 65.2 54.3 53.2 51.4 56.8 52.9 Peru Regions 1991 46.5 45.2 44.9 44.8 44.5 45.4 42.8 48.7 46.5 47.1 47.0 43.4 47.5 46.5 42.8 46.0 51.2 1994 47.9 45.9 45.2 44.9 44.5 46.0 44.0 49.8 48.4 48.3 48.7 45.3 46.9 45.2 43.1 46.7 46.2 National 1994 48.6 46.4 45.8 45.5 45.0 46.6 44.2 49.3 48.8 48.9 49.2 45.4 47.2 45.9 43.7 47.4 46.2 2000 49.4 47.7 47.4 46.9 46.7 47.8 43.8 45.7 49.5 49.5 49.6 44.4 49.9 48.5 45.8 46.6 49.2 Trinidad & Tobago 1992 49.5 47.2 45.7 46.5 45.2 47.8 51.4 47.8 49.5 49.5 49.5 51.4 43.7 46.4 44.6 44.2 81.8 Uruguay 1989 42.2 40.8 40.2 40.3 39.9 41.0 42.2 45.0 43.3 46.1 46.1 43.2 41.5 37.6 40.2 41.8 1995 42.7 40.9 40.2 40.3 39.8 41.3 42.7 46.2 44.0 47.8 47.8 42.3 42.0 38.7 40.7 39.7 2000 44.6 42.5 41.5 41.8 41.0 43.1 44.6 47.9 45.8 49.5 49.5 42.0 43.0 40.9 41.6 40.6 RB de Venezuela 1989 44.2 41.7 40.6 40.7 39.7 42.1 42.3 42.6 44.2 44.2 44.2 42.3 40.7 40.7 39.0 43.5 44.2 1995 46.9 44.5 43.7 43.4 42.8 44.7 45.4 45.3 46.7 46.9 46.7 45.1 45.1 42.7 42.4 43.8 46.2 1998 47.6 45.5 44.7 44.7 44.0 45.9 46.4 45.4 47.9 47.6 47.9 46.9 46.0 45.1 43.5 44.9 47.4 Notes: Equivalized income A: 0= 0.9, oa1 = 0.5 and a2 0.75; B: 0 = 0.75, ota = 0.5 and 2= 0.75; C: 0 =0.9, x, =0.3 and (X2 = 0.5; D: 0 = 0.75, a1 = 0.3 and °X2 = 0.5; E: Amsterdam scale. Adult equivalent equal to 0.98 for men aged 14-17, 0.9 for women over 14, 0.52 for children under 14, and 1 for all others. Source: Author's calculations based on microdata from household surveys. INEQUALITY IN LATIN AMERICA TABLE A.8 Distrbution of household per capita income: signs of change in Gin! coefficients 1970s 1980s 1990s Londofio & Morley/ Londoho & Morley/ Sz6kely Altimir Szekely Altimir Szekely This study Country (i) (ii) (iii) (iv) (v) (vi) Argentina + + + Bahamas + = Bolivia + + Brazil -or = + + Chile + + or = + + = + Colombia - or = + or- - Costa Rica - or = = = or + Dominican Rep. = + + Ecuador + El Salvador + = or + Guatemala = + Honduras + Jamaica Mexico - + + = or- Nicaragua + Panama - + + Paraguay + Peru - + or = = + + Uruguay + or = + RB de Venezuela - + + + Note: Signs refer to changes in the Gini coefficient of the distribution of household per capita income. Source: Altimir 1994 and 1996b, Morley 2001, Londono and Szekely 2000, and Szekely 2001. 294 STATISTICAL APPENDIX TABLE A.9 Perceptions about unfairness of the income distribution: Latinobarometro survey for 2002 Do you think the income distribution is ..? Gini coefficient Very Level Change Very Fair Fair Unfair unfair Total around 2000 in the 90s (i) (ii) (iii) (iv) (v) (vi) (vii) Argentina 1.0 2.3 42.4 54.3 100.0 50.4 7.7 Bolivia 1.6 6.3 57.3 34.8 100.0 55.9 1.6 Brazil 0.8 12.8 50.7 35.7 100.0 57.2 -2.3 Chile 1.7 9.4 52.3 36.5 100.0 56.1 1.4 Colombia 0.6 3.5 41.4 54.5 100.0 55.8 -0.1 Costa Rica 1.3 14.2 51.9 32.7 100.0 44.6 0.8 Ecuador 3.2 6.6 57.9 32.3 100.0 54.3 El Salvador 1.8 13.5 49.1 35.7 100.0 51.8 1.3 Guatemala 0.7 13.1 57.9 28.2 100.0 56.0 Honduras 1.5 6.6 61.8 30.1 100.0 53.0 -2.6 Mexico 1.3 14.4 51.2 33.1 100.0 52.7 -1.2 Nicaragua 2.1 13.3 54.7 29.9 100.0 54.1 -0.1 Panama 0.4 4.7 40.8 54.0 100.0 54.4 -0.3 Paraguay 0.8 4.1 40.0 55.1 100.0 54.9 Peru 1.3 4.0 57.7 36.9 100.0 47.7 2.0 Uruguay 1.1 6.8 60.5 31.6 100.0 42.5 1.7 RB de Venezuela 6.3 21.1 48.1 24.5 100.0 45.5 3.8 Average 1.7 9.4 51.7 37.3 100.0 Source: Latinobar6metro (2002) and table 3.5. 295 TABLE A.10 Indices of bipolarization and inequaliy (Gini coefficient) Household per capita income Equivalized income EGR Wolfson Gini EGR Wolfson Gini Country (i) (ii) (iii) (iv) (v) (vi) Argentina 1992 0.149 0.427 44.7 0.140 0.397 42.6 1996 0.158 0.454 48.2 0.147 0.419 45.8 2001 0.175 0.519 52.2 0.164 0.472 50.4 Bolivia Urban 1992 0.171 0.442 49.5 0.162 0.413 47.7 1996 0,179 0.462 51.1 0.171 0.429 49.3 National 1996 0.192 0.543 57.6 0.178 0.502 55.8 1999 0.200 0.630 57.8 0.188 0.587 55.9 Brazil 1990 0.245 0.634 61.2 0.236 0.611 59.5 1995 0.239 0.629 60.0 0.230 0.593 58.3 2001 0.229 0.460 59.0 0.221 0.429 57.2 Chile 1990 0.208 0.478 55.9 0.202 0.452 54.7 1996 0.195 0.482 56.1 0.189 0.454 54.9 2000 0.190 0.458 57.1 0.183 0.431 56.1 Colombia Bogota 1992 0.200 0.504 54.6 0.189 0.467 52.4 1996 0.190 0.425 52.4 0.183 0.393 50.8 National 1996 0.183 0.422 56.1 0.172 0.391 54.3 1999 0.195 0.462 57.6 0.185 0.434 55.8 Costa Rica 1990 0.146 0.430 45.6 0.137 0.396 43.9 1995 0.147 0.414 45.7 0.137 0.384 44.0 2000 0.155 0.442 46.5 0.145 0.412 44.6 Dominican Rep. 1995 0.187 0.475 51.5 0.181 0.459 50.2 1997 0.163 0.446 49.7 0.156 0.428 48.1 Ecuador 1994 0.196 0.496 54.8 0.185 0.466 53.0 1998 0.197 0.547 56.2 0.185 0.509 54.3 El Salvador 1991 0.175 0.455 52.7 0.162 0.422 50.5 1995 0.163 0.446 51.3 0.151 0.414 49.4 2000 0.166 0.485 53.2 0.156 0.460 51.8 Guatemala 2000 0.209 0.567 58.3 0.193 0.520 56.0 Honduras 1990 0.209 0.519 57.8 0.196 0.483 55.6 1995 0.190 0.483 56.1 0.178 0.446 54.1 1999 0.189 0.496 55.0 0.177 0.461 53.0 Jamaica 1990 0.168 0.492 51.7 0.155 0.447 49.6 1996 0.185 0.448 54.4 0.163 0.432 51.5 1999 0.171 0.473 52.0 0.155 0.440 49.0 Mexico 1992 0.190 0.489 55.9 0.178 0.464 53.9 1996 0.185 0.490 54.4 0.176 0.460 52.5 2000 0.178 0.480 54.6 0.170 0.457 52.7 Nicaragua 1993 0.204 0.569 55.9 0.194 0.537 54.2 1998 0.194 0.532 55.9 0.183 0.497 54.1 Panama 1991 0.207 0.587 56.4 0.195 0.545 54.7 1995 0.199 0.537 55.9 0.186 0.499 54.0 2000 0.202 0.566 56.4 0.189 0.528 54.4 Paraguay 1995 0.213 O.590 59.5 0,201 0.544 57.8 1999 0.225 0.609 56.8 0.214 0.564 54.9 Peru Regions 1991 0.144 0.409 46.5 0.137 0.398 45.2 1994 0.151 0.461 47.9 0.143 0.430 45.9 National 1994 0.158 0.471 48.6 0.148 0.436 46.4 2000 0.146 0.423 49.4 0.136 0.400 47.7 Trinidad & Tobago 1992 0.161 0.445 49.5 0.151 0.414 47.2 Uruguay 1989 0.130 0.352 42.2 0.126 0.343 40.8 1995 0.136 0.385 42.7 0.130 0.365 40.9 2000 0.151 0.426 44.6 0.144 0.398 42.5 RS de Venezuela 1989 0.135 0.392 44.2 0.125 0.364 41.7 1995 0.144 0.418 46.9 0.132 0.381 44.5 1998 0.149 0.440 47.6 0.139 0.404 45.5 Note: EGR = Esteban, Gradin, and Ray 1999. Source: Authors' calculations based on microdata from household surveys. 296 STATISTICAL APPENDIX TABLE A.11 Class stncture in Latin America: rebtive occupational income and share in working population Relative occupational incomes Shares in working population Professional Petty Formal Informal Professional Petty Formal Informal Employers executives entrepreneurs workers workers Capitalists executives entrepreneurs workers workers (i (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) Argentina (Greater Buenos Aires) 1980 280 226 267 96 71 2.1 3.3 4.3 44.2 46.2 1990 322 147 288 70 91 1.6 6.9 6.4 44.8 40.3 1997 336 321 1.4 6.9 5.6 46.1 40.1 Bolivia 1989 386 183 281 86 77 1.1 4.3 3.9 31.4 59.1 1994 294 209 231 77 57 1.4 6.8 7.8 28.3 55.8 1997 281 244 197 89 59 2.0 6.7 6.9 24.8 59.5 Brazil 1979 389 168 296 86 72 1.5 7.5 3.8 49.7 37.5 1990 343 174 240 81 62 2.2 4.6 3.3 45.9 43.1 1996 382 214 280 78 63 2.5 4.9 3.9 44.6 44.1 Chile 1990 528 157 404 74 83 1.6 12.9 2.7 45.7 37.0 1994 544 155 290 65 79 1.5 15.4 6.2 44.9 32.1 1998 457 158 331 58 86 1.6 17.0 7.2 43.4 30.8 Colombia 1980 428 208 55 64 1994 345 208 68 55 1997 287 182 71 53 Costa Rica 1981 198 173 195 73 82 1.5 2.7 3.4 60.1 32.2 1994 208 162 177 85 69 1.1 6.1 6.5 54.5 31.7 1997 150 161 132 86 58 1.6 8.8 8.9 49.9 30.8 Ecuador 1990 171 214 143 104 67 1.4 4.5 4.2 39.3 50.3 1994 228 179 210 90 63 1.4 5.6 8.5 35.5 49.0 1997 220 190 217 97 63 1.9 6.0 7.5 34.0 50.6 Mexico 1984 308 183 277 92 33 0.2 6.2 3.3 1994 416 216 314 68 36 0.5 9.0 4.4 1998 444 168 285 76 39 0.9 6.6 5.8 Panama 1979 116 243 89 35 2.1 4.6 69.9 23.4 1991 236 188 154 82 45 0.8 7.4 3.6 53.6 34.6 1997 275 179 207 73 53 1.0 10.8 3.6 53.4 31.2 Paraguay (Asunci6n) 1986 290 223 245 84 61 1.7 6.1 7.8 35.3 49.1 1990 303 138 241 76 86 2.1 5.5 8.3 36.8 47.3 1996 294 181 200 86 69 1.4 4.8 8.5 33.0 52.2 Uruguay 1981 347 147 293 60 42 1.6 3.9 4.6 55.8 34.0 1990 279 177 207 86 51 1.9 5.1 5.0 51.9 36.2 1997 235 200 200 94 47 1.7 6.5 5.9 48.3 37.7 RB de Venezuela 1981 153 196 145 91 69 1.5 5.2 6.4 43.5 44.3 1990 264 147 211 80 79 2.6 5.8 6.0 51.4 34.2 1997 311 161 261 67 84 1.9 6.1 6.1 45.2 40.6 Note: In panel 1, income is relative to the mean occupational income. Source: Authors' calculations based on ECLAC 2000. 297 INEQUALITY IN LATIN AMERICA TABLE A. 12 Class structure in Argentina and Brazil: percentage of working population and relafve income Argentina Brazil 1992 2001 1995 2001 (i) ODi (iii) (iv) 1. Share in working population Capitalists 1.6 1.1 1.2 1.3 Professionals/executives 4.4 5.8 4.1 4.5 Petty entrepreneurs 4.9 5.4 3.7 3.7 Formal workers 46.5 45.9 33.8 31.7 Informal workers 42.6 41.8 57.1 58.8 2. Relative individual income Capitalists 144 173 145 149 Professionals/executives 108 127 105 110 Petty entrepreneurs 100 100 100 100 Formal workers 41 46 26 29 Informal workers 38 28 17 19 3. Relative equivalized household income Capitalists 136 140 127 129 Professionals/executives 113 140 122 120 Petty entrepreneurs 100 100 100 100 Formal workers 51 50 32 33 Informal workers 48 35 21 22 4. Income share Capitalists 4.8 4.1 6.2 6.1 Professionals/executives 10.0 15.5 15.1 16.2 Petty entrepreneurs 10.5 11.3 13.0 12.1 Formal workers 40.4 44.2 31.1 29.6 Informal workers 34.2 24.9 34.6 36.0 5. Within Gini coefficient for individual income Capitalists 37.1 53.1 54.9 57.1 Professionals/executives 34.1 34.2 45.7 45.7 Petty entrepreneurs 40.9 40.9 48.7 49.4 Formal workers 33.1 36.3 45.0 41.8 Informal workers 39.9 43.1 51.8 49.6 Source: Authors' calculations based on microdata from Encuesta Permanente de Hogares (EPH) and Pesquisa Nacional por Amostra de Domicilios (PNAD). 298 STATISTICAL APPENDIX TABLE A.13 Individual income and cbss structure in Argentina (1992, 2001) and Brazil (1995, 20011: docile distribuiion Argentina, 1992 Argentina, 2001 Classes Classes Professional Petty Formal Informal Professional Petty Formal Informal Deciles Capitalists executives entrepreneurs workers workers Deciles Capitalists executives entrepreneurs workers workers 1 0.3 1.0 5.3 5.8 16.9 1 0.3 0.7 2.1 4.5 19.0 2 1.2 0.5 1.1 11.0 11.1 2 0.0 0.8 4.1 5.7 17.2 3 0.7 0.6 3.2 10.8 11.1 3 5.7 1.4 2.4 8.1 14.5 4 1.3 1.2 3.0 11.8 9.8 4 0.1 1.0 3.4 11.9 10.1 5 0.2 4.0 1.7 11.1 10.5 5 0.3 2.1 6.3 10.9 10.7 6 5.7 3.9 8.1 10.6 10.2 6 3.2 3.1 6.9 12.8 8.3 7 2.2 6.6 8.0 12.0 8.3 7 6.7 7.0 6.5 13.4 7.0 8 7.1 11.2 8.9 11.2 8.6 8 9.6 6.1 12.3 14.3 5.2 9 27.0 22.3 20.6 9.6 7.5 9 20.7 21.4 18.0 11.6 5.3 10 54.4 48.7 40.0 6.0 6.0 10 53.4 56.6 38.0 7.0 2.7 Total 100.0 100.0 100.0 100.0 100.0 Total 100.0 100.0 100.0 100.0 100.0 Brazil, 1995 Brazil, 2001 Classes Classes Professional Petty Formal Informal Professional Petty Formal Informal Deciles Capitalists executives entrepreneurs workers workers Deciles Capitalists executives entrepreneurs workers workers 1 1.0 0.1 0.4 3.2 16.8 1 1.6 0.1 0.6 5.0 17.3 2 1.1 0.4 0.6 6.1 14.6 2 2.2 0.4 0.8 7.9 14.3 3 1.9 0.5 0.7 8.4 12.9 3 1.5 0.6 1.1 9.7 12.5 4 2.8 0.7 1.3 10.2 11.5 4 2.5 0.8 1.8 11.2 10.9 5 3.6 1.5 3.1 10.9 10.8 5 3.5 1.7 2.8 11.9 9.9 6 3.0 2.8 2.7 13.7 8.7 6 3.0 2.2 3.7 12.8 8.9 7 5.9 5.2 6.8 14.2 7.7 7 5.8 5.2 8.6 12.1 8.7 8 7.5 10.4 12.3 13.4 7.4 8 6.1 10.1 9.8 12.8 7.3 9 14.1 23.6 21.1 12.7 5.9 9 17.0 24.5 22.7 10.6 6.3 10 59.1 54.8 50.9 7.0 3.8 10 56.9 54.6 48.1 5.8 4.0 Total 100.0 100.0 100.0 100.0 100.0 Total 100.0 100.0 100.0 100.0 100.0 Source: Authors' calculations based on microdata from Encuesta Permanente de Hogares (EPH) and Pesquisa Nacional por Amostra de Domicilios (PNAD). 299 INEQUALITY IN LATIN AMERICA TABLE A.14 Aggregate welfare: mean income Mean income from national accounts Mean income from household surveys Mean income Sen Atk(1) Atk(2) Mean income Sen Atk(1) Atk(2) Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) Argentina 1992 100 100 100 100 100 100 100 100 1996 109 102 102 90 91 86 85 76 1998 119 109 108 96 101 92 92 82 2001 111 96 94 72 89 77 75 58 Bolivia 1996 100 100 100 100 100 100 100 100 1999 104 104 96 74 122 122 112 87 Brazil 1990 100 100 100 100 100 100 100 100 1995 108 111 112 115 137 141 142 145 2001 113 120 120 124 133 140 141 145 Chile 1990 100 100 100 100 100 100 100 100 1996 148 147 148 150 142 142 142 144 2000 161 157 157 153 153 149 148 144 Costa Rica 1990 100 100 100 100 100 100 100 100 1995 116 116 116 118 119 118 119 121 2000 123 121 122 123 125 123 124 125 Dominican Rep. 1995 100 100 100 100 100 100 100 100 1997 112 116 114 108 119 123 121 114 Ecuador 1994 100 100 100 100 100 100 100 100 1998 100 97 95 83 96 93 91 80 El Salvador 1991 100 100 100 100 1995 120 123 122 115 2000 126 125 122 110 Honduras 1990 100 100 100 100 1995 103 107 107 109 1999 101 108 106 98 Mexico 1992 100 100 100 100 100 100 100 100 1996 98 101 101 99 78 80 80 78 2000 116 120 119 114 104 107 107 102 Nicaragua 1993 100 100 100 100 100 100 100 100 1998 108 108 108 108 111 111 111 111 Panama 1991 100 100 100 100 100 100 100 100 1995 111 112 114 118 128 129 131 135 2000 123 123 127 143 130 130 134 151 Paraguay 1995 100 100 100 100 100 100 100 100 1999 92 99 96 86 90 96 94 84 Peru 1991 100 100 100 100 100 100 100 100 1994 112 107 107 101 2000 124 118 115 97 114 108 106 89 Uruguay 1989 100 100 100 100 100 100 100 100 1995 116 115 115 110 100 99 99 94 2000 123 118 118 114 112 107 108 104 RB de Venezuela 1989 100 100 100 100 100 100 100 100 1995 109 104 104 98 90 86 86 81 1998 109 102 101 85 94 88 87 73 Notes: See Lambert 1993 for details on the aggregate welfare functions. Atk(e) refers to the function proposed by Atkinson 1970, that is, a CES function with parameter equal to e. First observation for each country = 100. Source: Authors' calculations based on microdata from household surveys and per capita GDP from World Bank 2000b and World Development Indicators Database. 300 STATISTICAL APPENDIX TABLE A.15 Poverty in Latin American counties: headcount ratio (share of population below the poverty linel Szekely Szekely Wodon Sala-i-Martin Szekely Szekely Wodon Sala-i-Martin (1997) (2001) (2001) (2002) (1997) (2001) (2001) (2002) Country (i) (ii) (iii) (iv) Country (i) (ii) (iii) (iv) Argentina Honduras Early 90s 14.6 Early 90s 67.8 77.2 60.1 36.7 Mid-90s 18.4 15.2 Mid 90s 65.6 76.3 63.0 Late 90s 17.9 Late 90s 75.3 47.2 Bolivia Jamaica Early 90s 63.4 73.4 26.5 Early 90s 27.4 8.1 Mid 90s 62.1 64.5 Mid 90s 25.1 Late 90s 61.4 40.1 Late 90s 6.1 Brazil Mexico Early 90s 46.3 48.3 46.7 16.5 Early 90s 19.3 16.2 26.9 4.5 Mid 90s 43.5 44.7 36.7 Mid 90s 19.7 21.2 32.1 Late 90s 41.3 12.9 Late 90s 21.2 1.8 Chile Nicaragua Early 90s 31.0 32.4 46.1 10.1 Early 90s 70.7 28.8 Mid 90s 18.3 28.0 Late 90s 72.7 47.3 Late 90s 16.1 2.2 Panama Colombia Early 90s 48.4 47.8 17.9 Early 90s 23.8 42.4 53.6 17.2 Mid 90s 47.8 Mid 90s 38.8 52.2 Late 90s 36.6 10.0 Late 90s 37.8 17.0 Paraguay Costa Rica Early 90s 39.7 22.8 Early 90s 24.7 34.2 11.2 Mid 90s 52.1 39.5 Mid 90s 22.1 28.7 Late 90s 61.1 22.3 Late 90s 30.5 9.2 Peru Dominican Rep. Early 90s 41.9 15.3 Early 90s 49.5 22.6 Mid 90s 35.0 44.0 Mid 90s 38.1 47.4 Late 90s 42.4 12.5 Late 90s 34.5 8.4 Trinidad & Tobago Ecuador Early 90s 0.4 Early 90s 49.2 13.9 Late 90s 0.1 Mid 90s 49.5 55.2 Uruguay Late 90s 48.0 13.6 Early 90s 23.2 18.0 0.9 El Salvador Mid 90s 16.6 19.4 Early 90s 21.6 Late 90s 13.6 0.1 Mid 90s 58.6 RB de Venezuela Late 90s 64.0 16.8 Early 90s 14.3 12.6 30.2 4.2 Guatemala Mid 90s 13.4 15.2 41.2 Early 90s 28.5 Late 90s 20.6 4.8 Late 90s 25.1 Guyana Early 90s 27.2 Late 90s 12.8 Notes: The poverty lines in Londono and Szekely 1997, Szekely 2001, and Sala-i-Martin 2002 is US$2 per day. In Wodon and others 2001, poverty line equals twice the food poverty lines, which is based on the cost of country-specific food baskets providing 2,200 kcal per day per person. Source: Sala-i-Martin 2002, Londono and Szekely 1997, Szekely 2001, Wodon and others 2001. 301 INEQUALITY IN LATIN AMERICA TABLE A.16 DisIbution of household per capita Income: Ginl coefficient (value, standard error, coefficient of variation, and 95% confidence interval) Standard Coefficient of Standard Coefficient of Value error variation 95% interval Value error variation 95% interval Country (i) (ii) (iii) (iv) Country (i) (ii) (iii) (iv) Argentina Guatemala 1992 44.7 0.326 0.7 (44.1, 45.4) 2000 58.3 0.437 0.7 (57.4, 59.1) 1996 48.2 0.431 0.9 (47.8, 48.8) Honduras 1998 49.5 0.354 0.7 (49.0, 50.2) 1990 57.8 0.487 0.8 (56.8, 58.8) 2001 52.2 0.304 0.6 (51.7, 52.9) 1995 56.1 0.548 1.0 (55.1, 57.2) Bolivia 1999 55.0 0.448 0.8 (54.1, 55.9) Urban Jamaica 1992 49.5 0.406 0.8 (48.8, 50.4) 1990 51.7 0.691 1.3 (50.3, 52.9) 1996 51.1 0.529 1.0 (50.3, 52.1) 1996 54.4 1.216 2.2 (52.0, 56.8) National 1999 52.0 1.245 2.4 (49.6, 54.5) 1996 57.6 0.360 0.6 (56.9, 58.4) Mexico 1999 57.8 0.573 1.0 (56.8, 59.1) 1992 55.9 0.493 0.9 (55.0, 57.1) Brazil 1996 54.4 0.383 0.7 (53.6, 55.0) 1990 61.2 0.134 0.2 (60.9, 61.5) 2000 54.6 0.508 0.9 (53.7, 55.6) 1995 60.0 0.119 0.2 (59.8, 60.3) Nicaragua 2001 59.0 0.113 0.2 (58.8, 59.2) 1993 55.9 0.4 0.8 (54.9, 56.6) Chile 1998 55.9 0.9 1.5 (54.4, 57.6) 1990 55.9 0.281 0.5 (55.3, 56.4) Panama 1996 56.1 0.360 0.6 (55.4, 56.7) 1991 56.4 0.294 0.5 (55.7, 56.9) 2000 57.1 0.435 0.8 (56.4, 58.1) 1995 55.9 0.293 0.5 (55.3, 56.5) Colombia 2000 56.4 0.283 0.5 (55.9, 56.9) Bogota Paraguay 1992 54.6 0.451 0.8 (53.7, 55.4) 1995 59.5 0.401 0.7 (58.8, 60.3) 1996 52.4 0.496 0.9 (51.4, 53.3) 1999 56.8 0.516 0.9 (55.9, 58.0) National Peru 1996 56.1 0.377 0.7 (55.3, 56.8) 1991 46.5 0.416 0.9 (45.7, 47.2) 1999 57.6 0.375 0.7 (56.9, 58.3) 1994 48.6 0.475 1.0 (47.7, 49.2) Costa Rica 2000 49.4 0.387 0.8 (48.5, 50.1) 1990 45.6 0.297 0.7 (45.0, 46.2) Trinidad & Tobago 1995 45.7 0.250 0.5 (45.3, 46.2) 1992 49.5 0.950 1.9 (47.6, 51.4) 2000 46.5 0.234 0.5 (46.1, 47.0) Uruguay Dominican Rep. 1989 42.2 0.391 0.9 (41.5, 43.0) 1995 51.5 0.407 0.8 (50.6, 52.2) 1995 42.7 0.137 0.3 (42.5, 43.0) 1997 49.7 0.497 1.0 (48.7, 50.6) 2000 44.6 0.147 0.3 (44.3, 44.9) Ecuador RB de Venezuela 1994 54.8 0.545 1.0 (53.8, 55.9) 1989 44.2 0.137 0.3 (43.8, 44.4) 1998 56.2 0.494 0.9 (55.2, 57.2) 1995 46.9 0.237 0.5 (46.4, 47.4) El Salvador 1998 47.6 0.213 0.4 (47.2, 48.0) 1991 52.7 0.272 0.5 (52.1, 53.1) 1995 51.3 0.331 0.6 (50.7, 51.9) 2000 53.2 0.664 1.2 (52.1, 54.7) Note: Estimation made using bootstrap technique with 200 replications. Source: Authors' calculations based on microdata from household surveys. 302 STATISTICAL APPENDIX TABLE A.17 Distribution of household income or expenditures, averages by decade, by region: Gini coefficients Overall Region 1960s 1970s 1980s 1990s average Gini coefficients Latin America and the Caribbean 53.2 49.1 49.8 49.3 49.8 Sub-Saharan Africa 49.9 48.2 43.5 47.0 46.1 Middle East and North Africa 41.4 41.9 40.5 38.0 40.5 East Asia and Pacific 37.4 39.9 38.7 38.1 38.8 South Asia 36.2 34.0 35.0 31.9 35.1 Industrial countries and high-income developing countries 35.0 34.8 33.2 33.8 34.3 Eastern Europe 25.1 24.6 25.0 28.9 26.6 Difference in Gini points: LAC vs. Sub-Saharan Africa 3.3 0.9 6.3 2.4 3.7 Middle East and North Africa 11.9 7.1 9.3 11.3 9.3 EastAsia and Pacific 15.8 9.2 11.1 11.2 11.0 South Asia 17.0 15.1 14.7 17.4 14.7 Industrial countries and high-income developing countries 18.2 14.3 16.5 15.6 15.5 Eastern Europe 28.2 24.4 24.7 20.4 23.2 Source: Deininger and Squire 1996. TABLE A.18 Distribution of household per capita income, selected years by region: Gini coefficients Nonweighted statistics Population-weighted statistics 1950 1960 1970 1980 1992 1950 1960 1970 1980 1992 Share bottom quintile Africa 5.4 5.4 5.3 5.3 5.1 6.8 6.8 6.6 6.6 6.5 Asia 6.0 6.1 6.1 6.0 6.1 6.3 6.5 6.4 6.3 6.2 Latin America 4.3 4.2 4.2 4.0 4.0 4.2 4.1 4.1 4.0 4.0 Eastern Europe 6.8 6.8 6.8 6.8 6.8 7.0 6.9 6.9 6.8 6.7 Developed countries 5.0 5.0 5.2 5.3 5.4 4.7 4.7 4.8 4.9 4.8 Share top quintile Africa 54.2 54.2 55.0 55.3 55.5 49.1 49.2 50.1 50.5 50.7 Asia 50.0 48.8 49.0 49.7 49.8 48.7 48.0 48.1 50.5 51.0 Latin America 59.2 59.6 61.2 62.0 62.0 59.7 60.1 62.1 62.6 62.6 Eastern Europe 45.7 45.7 45.7 45.7 45.7 44.4 44.6 44.7 44.9 45.2 Developed countries 45.9 45.7 44.6 44.3 44.5 45.8 45.8 44.8 44.5 45.4 Gini coefficient Africa 46.5 46.5 47.3 47.6 48.0 40.8 41.0 41.8 42.2 42.5 Asia 41.7 40.7 40.9 41.5 41.6 40.3 39.5 39.7 41.7 42.3 Latin America 51.9 52.3 53.4 54.2 54.2 52.4 52.8 54.2 54.7 54.8 Eastern Europe 37.2 37.2 37.2 37.2 37.2 35.8 36.1 36.3 36.5 36.9 Developed countries 39.5 39.3 38.2 37.8 37.8 39.8 39.8 38.9 38.6 39.2 Difference in Gini points: Latin America vs. Africa 5.4 5.8 6.1 6.6 6.3 11.6 11.8 12.4 12.5 12.3 Asia 10.1 11.7 12.5 12.8 12.7 12.1 13.3 14.4 13.0 12.4 Eastern Europe 14.7 15.1 16.2 17.1 17.1 16.6 16.8 17.9 18.2 17.9 Developed countries 12.4 13.0 15.1 16.5 16.5 12.6 13.0 15.2 16.1 15.5 LAC Latin America and the Caribbean Source: Authors' calculations based on Bourguignon and Morrison 2002. 303 INEQUALITY IN LATIN AMERICA TABLE A.19 Distribution of $PPP income/expenditures per capita, 1988 and 1993: regional Gini coefficients Regions 1988 1993 Gini coefficients Africa 42.7 48.7 Asia 55.9 61.8 Latin America and the Caribbean 57.1 55.6 Eastern Europe, FSU 25.6 46.4 Western Europe, North America, Oceania 37.1 36.6 Difference in Gini points: LAC vs. Africa 14.4 6.9 Asia 1.2 -6.2 Eastern Europe, FSU 31.5 9.2 Western Europe, North America, Oceania 20.0 19.0 LAC Latin America and the Caribbean Note: Coefficients calculated on distributions of individuals for countries that were in the sample in both years. Source: Milanovic 2002. TABLE A.20 Poverty in different regions of the world: headcount rato According to Chen and Ravallion (2001) 1987 1998 Change 1987-98 East Asia 67.0 48.7 -18.3 Eastern Europe and Central Asia 3.6 20.7 17.1 Latin America and the Caribbean 35.5 31.7 -3.8 Middle East and North Africa 30.0 20.9 -9.2 South Asia 86.3 83.9 -2.4 Sub-Saharan Africa 76.5 78.0 1.4 Total 61.0 57.9 -3.1 According to Sala-i-Martin (2002) 1970 1980 1990 1998 Change 1970-98 Asia 60.3 48.3 29.7 15.6 -44.7 Latin America 22.2 10.5 14.0 10.5 -11.7 Africa 53.0 55.2 57.9 63.6 10.6 Total 41.0 34.6 25.8 18.6 -22.4 Source: Chen and Ravallion 2001 and Sala-i-Martin 2002. 304 TABLE A.21 Share of income sources in percent and Gini coeflicients of indivMdual income Share in total individual income Gini coefficient Capital & Individual Labor income Labor income Nonlabor Labor Nonlabor profits Transfers Pensions income Labor income monetary monetary & urban income Country (i) (ii) (Oil) (iv) (v) (vi) (vii) (viii) (ix) (x) Argentina 1992 82.2 17.8 2.3 15.5 13.4 43.1 39.2 39.2 39.2 42.4 1996 78.0 22.0 2.7 19.4 16.5 44.1 41.6 41.6 41.6 44.3 2001 77.8 22.2 3.0 19.3 14.9 46.6 45.7 45.7 45.7 44.1 Bolivia Urban 1992 89.3 10.7 10.7 4.9 50.8 50.2 50.2 50.2 56.7 1996 89.6 10.4 6.6 2.6 53.5 52.8 52.8 52.8 60.7 National 1996 89.6 10.4 4.0 6.4 2.3 55.7 55.7 55.7 52.5 65.7 1999 87.9 12.1 2.0 10.1 4.1 56.4 56.3 56.7 51.0 66.0 Brazil 1990 85.4 14.6 3.9 10.7 10.7 62.0 60.2 60.2 59.1 85.7 1995 82.8 17.2 3.1 14.1 13.6 59.2 58.6 58.5 57.5 60.5 2001 78.7 21.3 2.8 18.5 17.8 57.1 56.4 56.4 55.6 58.2 Chile 1990 78.4 21.6 11.2 10.3 9.5 57.7 53.0 53.0 51.9 61.5 1996 82.6 17.4 9.5 7.9 7.0 60.9 54.5 53.7 53.2 68.8 2000 80.0 20.0 11.5 8.5 7.5 61.8 55.0 53.8 53.4 70.3 Colombia Bogota 1992 86.9 13.1 51.9 49.7 50.2 50.2 55.0 1996 83.7 16.3 51.7 50.1 50.3 50.3 59.3 National 1996 82.8 17.2 0.9 16.3 52.5 50.0 50.3 48.4 64.2 1999 82.9 17.1 5.4 11.7 6.7 53.9 52.1 52.7 50.4 58.7 Costa Rica 1990 91.1 8.9 8.9 44.4 41.4 41.4 41.2 56.9 1995 89.8 10.2 10.2 5.7 46.1 43.0 43.0 42.6 55.9 2000 89.6 10.4 10.4 5.6 45.8 41.8 41.8 41.6 56.5 Dominican Rep. 1995 100.0 48.9 48.9 52.2 1997 85.5 14.5 1.8 12.7 2.9 48.8 43.9 44.0 44.5 62.5 Ecuador 1994 96.0 4.0 1.4 2.6 55.6 53.8 55.9 53.4 71.9 1998 92.8 7.2 3.3 3.9 58.7 54.5 56.8 53.7 78.0 El Salvador 1991 79.3 20.7 9.8 10.9 2.0 51.3 48.6 48.6 46.6 63.2 1995 84.8 15.2 49.5 47.4 47.4 46.4 56.3 2000 85.0 15.0 3.3 11.7 5.4 52.2 49.1 49.1 49.0 58.7 Guatemala 2000 86.9 13.1 2.7 10.4 3.2 57.3 55.8 56.5 54.9 71.2 Honduras 1990 100.0 55.0 55.0 55.0 54.1 1995 100.0 53.3 53.3 53.3 50.7 1999 100.0 53.9 53.9 53.9 49.9 Jamaica 1990 98.6 1.4 42.0 42.1 42.1 40.6 39.8 1996 97.0 3.0 44.9 45.2 45.2 49.5 61.4 1999 98.3 1.7 41.9 42.4 42.4 44.6 42.4 Mexico 1992 93.9 6.1 1.6 4.5 2.8 54.3 54.3 53.3 51.3 52.6 1996 91.4 8.6 1.6 7.0 2.9 53.4 53.7 52.5 50.1 53.9 2000 89.9 10.1 1.6 8.5 5.2 54.5 52.7 52.2 49.5 65.5 Nicaragua 1993 96.2 3.8 0.5 3.3 1.3 51.9 51.3 52.9 50.3 59.4 1998 89.3 10.7 2.5 8.3 0.9 55.4 53.8 56.4 55.4 68.4 Panama 1991 78.9 21.1 1.9 19.2 13.2 54.1 47.9 47.9 63.3 1995 78.7 21.3 1.6 19.7 9.4 55.2 49.5 49.5 48.3 65.8 2000 78.7 21.3 1.6 19.6 13.7 53.7 47.9 47.9 46.5 62.3 Paraguay 1995 89.2 10.8 2.7 8.1 4.0 56.6 55.5 43.9 43.1 64.3 1999 87.0 13.0 2.4 10.6 5.4 51.6 50.5 43.8 40.8 55.5 Peru Regions 1991 89.6 10.4 2.4 8.0 6.0 48.1 47.4 48.4 46.4 61.5 1994 84.2 15.8 2.2 13.7 6.5 48.3 47.6 48.3 46.4 62.8 National 1994 84.6 15.4 2.1 13.3 6.1 48.5 47.9 48.6 46.5 64.2 2000 85.6 14.4 2.0 12.4 7.0 50.0 48.9 49.1 46.5 61.0 Trinidad & Tobago 1992 100.0 42.6 42.6 42.6 43.3 Uruguay 1989 75.8 24.2 4.2 20.0 19.6 48.1 44.5 45.9 45.9 50.7 1995 71.5 28.5 3.3 25.1 21.7 47.0 46.1 47.8 47.8 46.9 2000 68.4 31.6 3.4 28.2 23.7 46.3 45.9 47.6 47.6 47.6 RB de Venezuela 1989 100.0 36.9 36.9 36.9 36.0 1995 92.9 7.1 2.1 5.0 45.3 41.9 41.9 39.8 61.4 1998 92.8 7.2 1.8 5.4 46.3 44.4 44.4 44.6 54.7 Note: Nonlabor income = capital and profits + transfers. Transfers = pensions + other public and private transfers. Source: Authors' calculations based on microdata from household surveys. 305 TABLE A.22 Years of education of adults ages 25-65, by equivalized income quintile, age, and gender By age and gender By equivalized income quintiles (25-65) (10-20) (21-30) (31-40) (41-50) (51-60) (61+) Country 1 2 3 4 5 Average Female Male Female Male Female Male Female Male Female Male Female Male Female Male Argentina 1992 7.5 8.0 8.7 9.8 12.2 9.5 9.4 9.6 7.9 7.6 11.0 10.9 10.2 10.0 9.3 9.4 8.2 8.9 7.2 7.9 1996 7.3 8.1 9.0 10.0 12.8 9.8 9.8 9.9 7.7 7.4 11.3 10.8 10.5 10.3 9.7 9.8 8.7 8.9 7.3 8.2 2001 7.3 8.3 9.2 10.6 13.4 10.1 10.3 10.2 7.9 7.5 11.7 11.0 11.0 10.6 10.2 9.9 9.1 9.6 7.3 8.2 Bolivia Urban 1992 6.3 7.3 8.2 9.3 11.5 8.7 7.8 9.7 7.6 7.7 9.7 10.9 8.3 9.9 6.8 9.3 6.1 8.2 4.3 7.1 1996 6.5 7.6 8.1 9.7 12.4 9.1 8.1 10.3 7.6 7.7 9.7 11.2 8.7 10.5 7.6 10.3 5.8 9.1 4.6 7.3 National 1996 2.8 4.7 6.3 7.6 10.7 6.7 5.6 7.6 6.3 6.6 7.8 9.2 6.3 8.2 5.0 7.2 3.1 5.8 2.4 4.1 1999 2.8 4.9 6.6 8.9 11.4 7.3 6.2 8.3 6.8 7.1 8.9 10.4 7.2 8.9 5.4 7.7 4.0 6.6 2.4 4.1 Brazil 1990 1.9 2.9 4.1 5.5 8.9 5.1 5.1 5.2 4.4 3.8 6.8 6.3 5.8 5.8 4.2 4.5 3.1 3.7 2.2 2.5 1995 2.3 3.4 4.5 6.1 9.7 5.6 5.6 5.6 4.9 4.2 7.0 6.4 6.5 6.3 5.3 5.3 3.7 4.1 2.5 3.0 2001 3.0 4.2 5.3 6.9 10.4 6.4 6.5 6.3 6.0 5.4 8.0 7.3 7.2 6.7 6.3 6.3 4.8 5.0 3.0 3.4 Chile ON 1990 6.6 7.2 7.9 9.2 11.8 8.8 8.6 9.0 8.3 8.0 10.4 10.2 9.4 9.8 7.9 8.5 6.7 7.3 5.4 6.0 1996 6.7 7.7 8.8 10.0 12.3 9.3 9.2 9.5 8.3 8.0 11.1 11.1 10.1 10.1 8.6 9.3 7.0 7.5 5.7 6.2 2000 7.4 8.3 9.2 10.6 13.2 10.0 9.8 10.1 8.4 8.2 11.6 11.5 10.6 10.7 9.8 10.2 7.7 8.3 6.0 6.5 Colombia Bogota 1992 6.8 7.1 7.9 9.4 12.4 9.0 8.7 9.4 7.0 6.5 10.0 10.3 9.4 9.9 7.5 8.7 6.4 7.7 5.4 6.6 1996 6.3 7.2 7.8 9.1 12.2 8.8 8.6 9.1 7.6 7.1 10.1 9.8 9.2 9.7 8.1 8.9 6.7 7.6 4.9 6.1 National 1996 3.7 4.8 5.8 6.9 10.1 6.6 6.6 6.7 6.2 5.6 8.2 7.8 7.4 7.4 5.9 6.4 4.6 4.9 3.5 3.9 1999 4.4 5.2 6.1 7.3 10.5 7.0 7.1 7.1 6.6 6.1 8.7 8.3 7.8 7.7 6.7 7.0 5.0 5.7 3.8 4.3 Costa Rica 1990 5.0 6.2 6.8 7.8 11.1 7.6 7.6 7.8 6.8 6.6 9.2 8.9 8.4 8.6 6.6 7.1 5.0 5.8 4.1 4.1 1995 5.6 6.5 7.3 8.4 11.6 8.2 8.2 8.3 7.1 6.8 9.3 9.2 9.1 9.1 7.8 8.0 6.2 6.5 4.5 4.7 2000 5.4 6.6 7.4 8.4 11.6 8.2 8.2 8.3 7.1 6.7 9.2 8.8 9.0 9.0 8.4 8.5 6.2 7.0 4.5 4.5 Dominican Rep. 1995 5.2 6.1 7.0 8.0 10.3 7.5 7.3 7.5 7.1 6.5 9.2 8.4 8.1 8.3 6.5 7.1 4.6 5.6 3.3 4.1 Ecuador 1994 4.2 5.2 6.7 7.5 10.6 7.1 6.8 7.3 6.7 5.3 8.9 8.9 7.6 7.9 5.9 6.8 4.3 4.8 3.2 4.2 1998 4.8 6.1 7.1 8.2 11.4 7.8 7.5 7.9 6.9 5.7 9.3 9.3 8.6 8.7 6.8 7.7 4.8 5.9 3.4 3.9 El Salvador 1991 2.4 3.0 4.1 5.4 8.5 5.0 4.6 5.5 5.7 5.6 6.7 7.2 5.2 6.4 3.8 4.9 2.8 3.4 1.8 2.5 1995 2.1 2.8 4.1 5.8 9.4 5.3 4.9 5.8 5.3 4.9 7.0 7.3 5.7 6.6 3.9 5.3 3.2 3.9 1.9 2.3 2000 3.8 4.4 5.3 6.5 9.0 6.1 5.8 6.7 5.7 5.6 8.0 8.2 6.6 7.4 5.1 6.5 3.8 4.9 2.3 2.9 Guatemala 2000 1.4 2.2 2.8 4.1 8.5 4.2 3.4 4.7 3.9 4.2 4.6 5.9 3.8 5.3 3.2 4.5 2.0 3.0 1.7 2.2 Honduras 1990 1.8 2.1 2.8 4.0 7.5 4.0 3.9 4.1 4.5 4.1 5.5 5.4 4.5 4.7 3.1 3.5 2.0 2.6 1.8 2.1 1995 2.5 3.1 4.1 5.4 8.0 4.9 4.7 5.0 5.1 4.7 6.4 6.2 5.4 5.7 4.0 4.7 2.6 3.1 1.9 2.3 1999 2.8 3.4 4.5 5.9 9.1 5.5 5.1 5.2 5.1 4.7 6.6 6.1 5.9 5.7 4.7 5.1 3.3 3.6 1.8 2.2 Jamaica 1990 7.2 7.3 7.7 8.1 9.4 8.1 8.2 8.2 8.3 8.1 10.2 9.9 9.0 8.6 7.2 7.3 6.4 6.6 5.6 5.8 1996 8.5 9.0 9.0 8.9 10.5 9.3 9.5 9.2 8.4 8.2 10.8 10.6 10.1 9.8 9.4 8.9 8.0 7.3 6.5 5.6 1999 8.5 8.5 8.7 9.0 9.7 9.0 9.2 9.1 8.1 7.9 10.3 9.9 9.9 9.6 9.1 9.0 7.3 8.2 6.7 5.6 Mexico 1992 2.7 4.1 5.1 6.6 10.0 6.1 5.6 6.5 6.5 6.3 7.6 8.3 6.1 7.2 4.5 5.7 3.9 4.6 2.5 3.1 1996 3.4 4.6 5.8 7.2 10.7 6.7 6.3 7.2 6.6 6.6 8.4 8.8 6.9 8.0 5.2 6.4 3.8 4.7 2.7 3.2 2000 3.5 5.3 6.8 8.1 11.6 7.4 6.9 8.0 7.0 6.9 9.0 9.5 7.5 8.6 6.5 7.8 4.7 6.2 3.0 3.5 Nicaragua 1993 2.3 3.5 4.5 5.6 7.3 4.9 4.4 4.8 4.3 4.0 5.9 5.7 4.9 5.5 3.3 4.0 2.4 2.6 1.8 1.9 1998 2.7 3.6 4.7 5.6 6.1 5.2 4.9 5.2 5.0 4.4 6.5 5.8 5.4 6.0 4.4 5.0 2.7 3.4 2.1 2.0 Panama 1991 4.7 6.5 7.8 9.2 12.0 8.5 8.4 8.2 7.2 6.8 10.2 9.5 9.3 9.0 7.7 7.7 6.3 6.4 4.9 4.8 1995 5.3 7.1 8.3 9.8 12.6 9.1 9.0 8.7 7.2 6.7 10.2 9.6 9.9 9.6 8.4 8.4 7.1 6.8 5.0 5.2 2000 6.6 7.9 9.0 10.0 13.1 9.7 9.9 9.5 7.5 7.1 10.9 10.1 10.7 9.9 9.7 9.7 8.3 8.5 7.2 7.2 Paraguay 1995 3.7 4.6 5.5 6.8 9.8 6.4 6.2 6.6 5.5 5.2 7.6 7.6 6.8 7.2 5.8 6.3 4.3 5.1 3.2 4.0 1999 4.0 5.1 6.2 7.7 10.1 7.0 6.6 7.0 6.1 5.8 8.1 8.3 7.2 7.6 6.0 6.6 5.0 5.4 3.5 4.2 Peru Regions 1991 7.6 9.0 9.5 10.3 11.5 9.8 9.2 10.2 7.6 7.8 10.8 11.2 9.8 10.8 8.3 10.1 7.4 8.3 6.8 7.5 1994 6.8 8.5 9.3 10.7 12.5 9.9 9.6 10.1 7.8 7.8 11.0 11.4 10.5 10.9 8.7 9.9 7.7 8.1 6.8 7.7 National 1994 6.4 8.0 9.2 10.2 12.2 9.6 9.3 9.8 7.5 7.6 10.6 11.1 10.0 10.6 8.4 9.5 7.4 7.8 6.6 7.4 2000 5.9 7.5 9.1 10.4 12.5 9.4 8.7 10.0 7.5 7.4 10.8 11.2 9.6 10.5 7.9 9.7 6.5 8.8 4.9 6.7 Trinidad & Tobago 1992 6.6 6.8 7.6 8.3 10.3 8.2 8.1 7.9 7.7 7.3 9.8 9.2 8.8 8.3 7.2 7.6 6.1 6.3 5.1 6.3 Uruguay 1989 5.9 6.7 7.6 8.6 10.4 8.1 7.9 8.3 7.3 7.2 9.8 10.0 9.1 9.3 7.9 8.3 6.5 6.9 5.1 5.4 1995 6.5 7.1 8.0 9.2 11.6 8.7 8.7 8.7 7.7 7.4 10.5 10.1 9.8 9.5 8.8 8.8 7.4 7.6 5.5 5.7 2000 7.0 7.8 8.6 9.8 12.1 9.3 9.3 9.2 9.2 8.7 10.8 10.2 10.1 9.9 9.5 9.3 8.3 8.2 6.3 6.5 RB de Venezuela 1989 4.6 5.6 6.4 7.4 10.0 7.2 6.9 7.4 6.5 5.9 8.6 8.2 7.7 7.8 6.2 7.1 4.4 5.9 2.7 4.0 1995 5.2 5.9 6.7 7.7 9.9 7.4 7.3 7.4 7.1 6.3 9.0 8.4 8.2 7.9 6.9 7.2 5.2 6.2 3.2 4.1 1998 6.0 6.5 7.1 8.4 11.1 8.1 8.1 8.0 7.3 6.5 9.7 8.7 8.9 8.4 7.8 7.9 5.7 6.9 3.5 4.6 Source: Authors' calculations based on microdata from household surveys. INEQUALITY IN LATIN AMERICA TABLE A.23 Years of education of adults apgs 259iO, by equivalized income quintile and ajge Age (10-20) (21-30) (31-40) Country 1 2 3 4 S Mean 1 2 3 4 S Mean 1 2 3 4 S Mean Argentina 1992 6.9 7.4 7.7 8.3 8.5 7.7 8.5 9.3 10.4 11.4 13.1 10.9 7.8 8.5 9.5 10.8 12.9 10.2 1996 6.6 7.2 7.9 8.1 8.5 7.5 8.6 9.6 10.7 11.6 13.2 11.0 7.7 8.6 9.6 11.2 13.5 10.3 2001 6.8 7.4 7.8 8.4 8.6 7.6 8.6 9.6 10.7 12.1 13.7 11.3 7.7 9.0 9.9 11.4 14.3 10.8 Bolivia Urban 1992 7.2 7.5 7.8 8.2 8.3 7.7 9.1 9.1 10.1 10.7 12.0 10.4 6.7 7.6 8.5 9.8 12.0 9.0 1996 7.1 7.4 7.6 7.9 8.5 7.7 9.2 9.4 10.1 10.3 12.8 10.5 6.9 7.7 9.0 10.4 13.1 9.6 National 1996 S.0 6.2 6.7 7.2 8.0 6.5 4.6 7.2 8.2 9.2 11.3 8.6 3.4 5.2 6.8 8.6 11.7 7.4 1999 S.0 6.5 7.4 7.9 8.4 7.0 5.2 7.7 9.0 10.8 12.6 9.7 3.2 5.6 7.3 9.7 12.3 8.0 Brazil 1990 2.3 3.4 4.3 5.3 6.0 4.1 3.0 4.4 5.7 7.3 10.1 6.6 2.2 3.4 4.6 6.2 10.1 5.8 1995 2.7 3.8 4.8 5.7 6.6 4.5 3.4 4.8 6.1 7.6 10.3 6.7 2.7 4.1 S.S 7.0 10.7 6.4 2001 4.0 5.1 6.1 6.9 7.6 5.7 4.3 5.7 7.2 8.8 11.2 7.6 3.4 4.8 6.2 7.8 11.3 7.0 Chile 1990 7.3 7.8 8.2 8.8 9.0 8.1 8.1 9.0 10.0 11.0 12.9 10.3 7.2 8.0 8.9 10.3 12.7 9.6 1996 7.3 7.8 8.4 8.8 8.9 8.2 8.4 9.7 10.8 11.9 13.7 11.1 7.4 8.7 9.7 11.1 13.2 10.1 2000 7.6 8.0 8.5 8.8 8.9 8.3 9.2 10.2 11.3 12.4 14.2 11.6 8.0 9.1 10.2 11.6 13.9 10.6 Colombia Bogota 1992 6.0 6.6 6.7 7.5 7.3 6.8 8.3 8.9 9.4 10.4 12.4 10.1 7.1 7.5 8.6 9.9 13.2 9.7 1996 6.5 6.8 7.8 7.9 8.2 7.4 7.9 8.5 9.2 10.5 12.5 10.0 6.7 7.8 8.4 10.1 13.2 9.4 National 1996 4.6 5.3 6.0 6.8 7.3 5.9 5.2 6.3 7.4 8.6 10.8 8.1 4.2 5.4 6.6 7.7 11.0 7.4 1999 5.2 5.8 6.4 7.1 7.7 6.4 6.1 6.9 7.8 8.9 11.1 8.5 4.9 5.9 6.9 8.2 11.3 7.7 Costa Rica 1990 5.7 6.1 6.5 7.3 7.9 6.6 6.8 7.5 8.3 9.3 11.7 9.0 5.7 6.6 7.4 8.9 12.2 8.4 1995 5.9 6.3 6.8 7.4 8.1 6.8 6.6 7.3 8.1 9.7 11.9 9.1 6.3 7.3 8.2 9.3 12.6 9.0 2000 5.8 6.5 6.9 7.3 8.1 6.9 6.0 7.2 8.0 9.3 12.0 8.8 6.3 7.3 8.2 9.4 12.4 8.9 Dominican Rep. 1995 6.0 6.3 6.7 7.3 8.1 6.8 7.1 7.6 8.6 9.1 10.9 8.8 6.0 7.0 7.6 9.1 11.0 8.3 Ecuador 1994 S.S 6.1 6.6 6.9 7.8 6.5 6.2 7.2 8.8 9.5 11.2 8.9 4.4 5.9 7.5 8.4 11.7 7.8 1998 5.6 6.5 6.8 7.3 8.2 6.8 6.9 7.9 8.7 9.8 12.1 9.4 5.6 6.8 7.9 9.4 12.1 8.7 El Salvador 1991 4.5 4.9 S.S 6.4 7.3 5.7 3.9 4.4 6.0 7.8 10.4 7.0 2.8 3.6 4.7 6.3 9.6 5.7 1995 3.7 4.1 5.1 6.2 7.0 5.1 3.5 4.4 5.8 8.0 11.0 7.2 2.4 3.3 4.9 6.5 10.6 6.1 2000 4.7 5.1 5.6 6.4 7.3 5.7 5.8 6.4 7.2 8.5 10.8 8.1 4.4 5.2 6.2 7.2 9.6 5.7 Guatemala 2000 2.61 3.1 3.84 4.95 6.57 4.18 2.27 3.17 4.09 6.15 9.44 5.49 1.56 2.33 3.52 5.29 9.31 4.69 Honduras 1990 3.2 3.6 3.9 4.7 5.8 4.2 3.0 3.6 3.9 S.S 8.4 5.4 2.1 2.4 3.2 4.5 8.1 4.5 1995 4.0 4.2 4.6 5.3 6.2 4.9 3.8 4.5 5.7 6.8 8.7 6.3 3.0 3.6 4.6 6.1 8.9 5.6 1999 4.0 4.4 S.0 5.6 6.5 5.1 3.6 4.7 5.8 6.8 9.6 6.7 3.3 4.1 5.1 6.8 9.9 6.1 Jamaica 1990 8.0 8.0 8.1 8.6 8.8 8.2 9.7 9.0 10.0 9.7 10.8 10.0 7.5 7.3 8.1 8.8 9.7 8.5 1996 8.2 8.0 8.3 8.7 8.4 8.3 10.3 10.0 9.9 10.3 11.5 10.6 9.0 9.6 9.8 9.7 10.7 9.9 1999 7.7 7.8 8.3 8.5 8.5 8.1 9.3 9.5 10.0 10.2 10.5 10.0 9.0 9.3 9.3 9.9 10.6 9.7 Mexico 1992 4.9 6.0 6.7 7.2 7.7 6.4 4.5 6.2 7.2 8.7 11.1 8.0 3.0 4.5 5.6 7.5 10.8 6.6 1996 5.1 6.2 7.0 7.4 7.6 6.6 5.4 6.7 8.0 9.3 11.4 8.6 3.9 5.2 6.4 8.0 11.7 7.4 2000 5.6 6.6 7.2 7.7 8.3 7.0 5.6 7.6 8.7 9.7 12.5 9.2 4.2 6.1 7.6 9.2 12.0 8.0 Nicaragua 1993 3.0 4.2 4.4 4.9 S.S 4.4 3.3 4.8 6.1 6.7 8.2 6.1 2.4 4.0 4.6 6.6 8.2 5.6 1998 3.2 4.0 5.2 5.4 6.4 4.8 3.6 4.9 6.0 7.3 8.8 6.3 3.1 4.2 5.4 6.5 8.7 5.8 Panama 1991 5.8 6.7 7.4 7.7 8.2 7.0 6.9 8.7 9.7 10.9 12.4 10.0 5.6 7.1 8.4 10.3 12.8 9.3 1995 5.7 6.7 7.2 8.0 8.2 7.1 7.0 8.4 9.7 11.1 12.5 10.1 6.2 8.0 9.2 10.9 13.4 10.1 2000 6.2 6.9 7.6 8.1 8.3 7.3 7.6 8.8 10.2 11.2 13.5 10.6 7.3 8.7 9.7 11.0 14.0 10.3 Paraguay 1995 4.3 4.7 5.4 5.9 6.9 5.4 4.8 5.7 6.5 8.2 10.2 7.6 3.9 S.0 6.4 7.6 10.6 7.0 1999 4.8 5.4 6.1 6.7 7.7 6.1 5.4 6.3 7.5 9.2 11.3 8.5 4.5 5.8 7.1 8.4 10.9 7.7 Peru Regions 1991 6.9 7.6 8.0 8.1 8.2 7.7 9.2 10.3 11.1 11.8 12.1 11.1 7.8 9.6 10.2 10.9 12.8 10.5 1994 6.8 7.5 7.8 8.4 8.9 7.8 8.8 10.2 11.0 11.9 13.2 11.2 7.4 9.2 10.4 11.6 13.4 10.8 National 1994 6.5 7.2 7.5 8.2 8.7 7.6 8.0 10.0 10.6 11.6 12.9 10.9 6.9 8.5 10.4 11.1 13.1 10.4 2000 6.5 7.0 7.9 8.1 8.4 7.5 8.2 9.5 10.9 11.9 13.3 11.0 6.5 8.4 9.9 11.5 13.7 10.2 Trinidad & Tobago 1992 7.1 7.3 7.5 7.8 8.0 7.5 8.1 8.7 9.1 10.2 11.3 9.5 6.6 7.0 8.4 8.8 11.4 8.8 Uruguay 1989 6.3 7.1 7.6 7.8 8.2 7.3 7.6 8.5 9.7 10.7 11.7 9.9 6.7 7.9 9.0 10.0 11.7 9.2 1995 6.8 7.3 7.8 8.1 8.4 7.6 8.0 9.1 10.1 11.0 12.5 10.3 7.4 8.2 9.2 10.5 12.7 9.7 2000 7.9 8.8 9.1 9.8 10.1 8.9 8.1 9.3 10.4 11.2 13.1 10.5 7.5 8.3 9.5 10.8 13.1 10.0 RB de Venezuela 1989 5.2 5.8 6.2 6.8 7.5 6.2 6.5 7.2 7.7 8.6 10.5 8.4 S.0 6.1 7.1 8.2 10.7 7.8 1995 5.9 6.2 6.6 7.3 8.0 6.7 6.9 7.3 8.1 9.2 1 0.8 8.7 5.8 6.5 7.5 8.5 10.7 8.1 1998 6.0 6.4 6.9 7.3 8.2 6.9 7.2 7.9 8.6 9.7 11.8 9.2 6.6 7.3 7.7 9.2 11.7 8.7 Source: Authors' calculations based on microdata from household surveys. 308 STATISTICAL APPENDIX (41-50) (51-60) 1 2 3 4 5 Mean 1 2 3 4 5 Mean 7.4 7.7 8.2 9.4 12.0 9.3 6.5 7.1 7.8 8.4 10.9 8.5 7.2 7.8 8.7 9.8 13.2 9.7 6.5 7.0 7.6 8.7 11.5 8.6 7.2 8.1 8.9 10.7 13.6 10.0 6.1 7.1 8.2 9.1 12.3 9.1 5.4 6.9 7.0 8.9 11.0 8.0 4.1 5.2 6.2 7.3 10.4 7.1 6.0 7.7 6.9 9.5 11.8 8.8 4.3 4.7 5.7 8.4 11.5 7.4 2.4 3.9 6.3 6.9 10.3 6.3 1.5 2.6 3.8 5.3 9.5 4.5 2.3 4.9 6.1 8.5 10.8 6.7 1.8 3.1 4.2 5.9 9.6 5.3 1.4 2.1 3.3 4.5 8.2 4.4 1.0 1.5 2.4 3.5 6.6 3.4 1.8 2.8 3.9 5.6 9.6 5.3 1.2 1.8 2.8 3.9 7.8 3.9 2.6 3.9 4.9 6.6 10.5 6.3 1.6 2.3 3.2 4.7 8.9 4.8 5.7 6.4 7.1 8.5 11.5 8.2 4.7 5.0 5.7 7.3 10.2 7.0 6.0 7.0 8.5 9.6 12.0 9.0 4.4 5.2 6.3 7.6 10.5 7.3 7.2 8.2 8.9 10.7 13.4 10.0 5.0 5.9 6.7 8.2 11.6 8.0 5.4 6.0 7.0 8.8 12.0 8.1 5.4 5.3 5.7 7.6 10.1 7.0 5.7 6.6 7.4 8.6 12.1 8.5 4.4 5.2 5.7 6.4 10.6 7.0 3.1 4.2 5.1 6.4 9.7 6.1 2.5 3.2 3.8 4.6 8.0 4.7 4.0 4.7 5.4 6.9 10.5 6.7 3.1 3.5 4.1 5.1 8.9 5.2 4.4 5.5 5.8 6.5 10.4 6.8 2.9 4.2 4.1 5.1 8.9 5.3 5.2 6.1 6.9 7.8 11.4 7.8 4.2 4.4 5.1 6.0 9.8 6.2 5.8 6.9 7.4 8.4 11.5 8.4 3.7 4.4 5.3 6.5 10.1 6.4 4.5 5.3 6.2 6.9 10.1 6.8 3.4 3.9 4.2 5.5 8.3 5.2 3.8 4.5 5.1 6.5 10.1 6.4 3.1 3.0 4.3 4.3 7.7 4.6 4.2 5.4 6.2 7.2 11.4 7.3 3.1 3.8 4.3 5.5 9.2 5.4 1.9 2.6 3.5 4.4 7.7 4.3 1.3 1.5 2.2 3.2 6.0 3.1 1.7 2.4 3.6 4.9 8.3 4.5 1.2 1.5 2.2 4.1 7.0 3.5 3.6 3.8 4.8 5.7 8.2 5.5 2.3 3.0 3.4 4.1 7.6 4.4 1.07 2.36 2.32 3.15 8.19 3.94 0.56 0.69 1.12 1.95 6.7 2.54 1.5 1.6 2.2 3.0 7.1 3.2 0.9 1.2 1.5 2.0 5.2 2.2 2.3 2.4 3.3 4.8 7.6 4.4 1.3 1.6 2.0 3.0 5.8 2.9 2.6 2.9 4.0 5.3 8.8 5.2 1.3 2.1 2.7 4.1 7.5 3.8 5.8 7.0 6.6 7.1 8.3 7.1 6.3 5.7 6.7 6.8 7.4 6.7 8.0 8.4 8.0 8.6 10.2 8.9 6.9 7.6 8.2 7.4 9.3 8.0 8.2 8.2 8.6 8.4 9.5 8.7 7.9 7.1 7.2 7.2 7.4 7.3 2.2 3.3 4.0 5.3 9.3 5.1 1.4 2.6 3.3 4.2 8.2 4.3 2.7 3.8 5.0 6.2 10.1 5.8 1.8 2.3 3.4 4.4 7.9 4.2 3.3 4.3 6.2 7.5 11.6 7.2 1.7 3.0 4.5 5.3 10.5 5.5 1.7 3.0 3.6 4.7 5.9 3.9 1.5 1.7 2.3 2.7 4.7 2.7 2.3 3.2 4.2 5.1 8.1 4.9 1.7 1.9 2.6 2.8 6.1 3.2 4.1 5.6 7.1 8.1 11.7 7.9 3.2 4.4 5.7 7.0 10.4 6.5 4.6 6.2 7.2 9.2 12.4 8.7 3.3 4.8 6.1 7.6 11.6 7.3 6.1 7.5 8.5 9.8 13.1 9.7 5.0 6.3 6.9 7.8 12.0 8.4 3.4 4.1 5.2 6.4 9.6 6.1 2.6 3.4 3.7 4.5 8.4 4.7 3.5 4.6 5.6 6.9 10.3 6.5 2.9 3.4 4.9 5.6 8.2 5.4 6.7 8.8 9.0 10.1 10.4 9.2 7.1 7.1 7.4 7.8 9.2 7.8 5.7 7.7 8.8 10.2 12.1 9.3 5.3 6.3 7.2 8.2 10.5 8.0 5.6 7.0 8.5 10.0 11.9 9.0 5.3 5.8 7.0 7.7 10.3 7.8 5.4 6.7 8.9 10.1 12.0 8.9 4.0 5.6 7.1 8.2 10.7 7.7 5.8 6.0 6.7 7.6 9.7 7.5 6.2 5.3 5.1 5.7 7.8 6.2 5.7 7.0 7.6 8.7 10.3 8.1 4.4 5.3 6.1 6.9 8.9 6.7 6.4 7.2 8.2 9.6 12.0 8.8 5.1 5.8 6.7 7.8 10.3 7.5 7.1 8.0 8.7 10.1 12.3 9.4 5.7 6.8 7.2 8.5 11.1 8.3 3.9 5.0 5.6 6.9 9.7 6.6 2.7 3.6 4.2 5.1 8.1 5.1 4.8 5.4 6.4 7.5 9.5 7.1 3.3 4.0 4.6 5.6 8.5 5.7 5.7 6.1 6.6 8.0 11.0 7.9 4.2 4.1 5.3 6.3 9.4 6.3 309 TABLE A.24 Schooling in Latin mercan countries: shar-e of population over age 25, by education level and average years of schooling % of no schooling % of primary school completed % of high school completed Average schooling years 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 Argentina 12.0 8.3 7.1 5.7 5.8 19.9 30.6 33.0 34.6 30.1 1.8 2.4 3.7 7.2 11.9 5.0 5.9 6.6 7.8 8.5 Barbados 0.0 1.1 0.8 1.9 2.6 22.5 9.3 17.6 13.4 10.7 0.7 0.8 2.2 5.4 8.1 5.2 9.1 6.8 8.2 9.1 Bolivia 50.2 51.2 44.8 36.8 30.9 5.1 6.8 8.9 10.0 11.0 2.5 2.6 4.2 6.7 9.8 4.2 3.7 4.0 4.7 5.5 Brazil 43.2 42.6 32.9 22.4 21.2 11.8 19.4 4.9 12.4 11.5 1.4 1.4 3.4 4.9 5.7 2.8 2.9 3.0 3.8 4.6 Chile 20.2 12.4 9.4 5.8 5.3 22.1 14.2 12.8 10.7 9.6 1.4 2.6 4.9 8.4 10.7 5.0 5.5 6.0 7.1 7.9 Colombia 35.1 39.2 24.5 24.5 19.8 10.9 10.8 12.0 11.0 10.8 1.3 1.4 2.9 4.7 6.7 3.0 2.7 3.9 4.4 5.0 Costa Rica 17.7 20.6 14.5 10.8 9.4 21.0 12.1 16.0 14.0 13.6 1.8 2.1 5.7 9.4 12.7 3.9 3.6 4.7 5.6 6.0 Dominican Rep. 43.4 40.1 35.7 28.8 25.7 13.8 4.3 10.2 11.2 10.3 0.5 1.3 2.9 6.1 9.8 2.4 2.9 3.4 4.3 5.2 Ecuador 37.8 37.6 25.4 20.0 17.7 13.6 18.0 34.1 14.3 13.1 1.0 1.3 5.2 11.9 12.7 3.0 3.2 5.4 5.9 6.5 El Salvador 61.8 54.2 36.0 37.1 35.0 5.1 8.0 11.5 10.2 10.1 0.3 1.3 2.2 4.3 7.2 1.7 2.3 3.3 3.6 4.5 Guatemala 69.2 68.4 54.7 52.9 47.1 7.2 6.2 8.0 8.0 8.3 0.3 0.7 1.5 3.0 4.0 1.4 1.5 2.3 2.6 3.1 Guyana 13.7 12.2 8.1 9.6 7.3 23.0 21.5 20.2 16.0 14.6 0.4 0.7 1.2 1.4 2.7 3.5 4.0 4.7 5.4 6.0 Haiti 90.2 83.5 77.0 57.0 54.4 1.4 1.8 3.2 7.2 7.1 0.2 0.2 0.5 0.6 0.8 0.7 0.9 1.5 2.4 2.7 Honduras 60.9 61.9 49.0 31.9 25.9 6.9 7.9 8.6 11.5 12.4 0.4 0.6 1.2 3.1 4.4 1.7 1.7 2.3 3.7 4.1 Jamaica 18.8 5.1 3.2 3.5 3.3 20.8 24.0 22.1 17.7 15.1 0.3 0.7 1.4 1.8 2.8 2.5 3.0 3.6 4.6 5.2 Mexico 46.0 35.0 34.2 18.8 12.4 10.4 15.3 17.2 19.9 19.4 0.8 1.5 3.2 5.4 6.6 2.4 3.3 4.0 5.9 6.7 Nicaragua 59.0 53.9 48.9 41.3 31.7 9.1 7.9 8.7 9.0 9.5 1.7 3.1 3.8 5.4 6.0 2.1 2.6 2.9 3.6 4.4 Panama 28.0 24.9 18.3 12.9 11.4 24.4 16.4 23.2 21.6 21.0 1.8 2.9 5.7 11.4 13.5 4.3 4.6 5.9 7.3 7.9 Paraguay 26.4 19.6 14.2 8.7 9.8 16.8 10.3 15.4 20.7 20.7 1.0 1.7 2.7 6.0 6.6 3.4 3.7 4.6 5.8 5.7 Peru 42.8 35.0 24.0 20.5 13.8 11.7 16.1 17.2 10.3 8.8 1.8 3.1 6.9 9.6 15.3 3.0 3.9 5.4 5.9 7.3 Trinidad & Tobago 14.7 11.6 1.3 4.5 5.1 18.6 21.9 42.6 16.3 12.8 0.7 0.8 2.0 2.3 3.1 4.2 4.5 6.6 6.7 7.6 Uruguay 14.1 13.9 7.3 5.5 3.2 18.4 22.8 20.5 13.2 12.3 3.5 3.9 5.1 7.9 8.5 5.0 5.2 5.8 6.7 7.2 RB deVenezuela 49.1 47.1 23.5 21.2 15.7 15.7 9.5 10.6 12.2 12.6 1.0 1.8 4.8 8.0 12.2 2.5 2.9 4.9 4.9 5.6 Total 37.1 33.9 25.9 21.0 18.0 14.4 13.7 16.5 14.1 13.3 1.2 1.7 3.4 5.9 7.9 3.2 3.6 4.4 5.2 5.9 Middle East and 65.9 60.8 52.6 39.2 28.7 10.2 9.2 9.5 11.0 11.2 1.3 1.7 3.7 6.2 7.9 2.3 2.7 3.8 5.2 6.3 North Africa Sub-Saharan Africa 76.3 73.4 63.0 53.5 46.0 5.5 5.1 6.9 7.8 8.4 0.2 0.4 0.5 1.2 1.7 1.2 1.4 2.0 2.7 3.4 Latin America and 37.1 33.9 25.9 21.0 18.0 14.4 13.7 16.5 14.1 13.3 1.2 1.7 3.4 5.9 7.9 3.2 3.6 4.4 5.2 5.9 the Caribbean East Asia and 51.5 41.7 31.4 26.0 18.1 15.7 16.2 18.9 19.4 19.0 1.7 2.2 3.4 5.2 8.5 3.1 3.7 4.8 6.1 7.3 the Pacific South Asia 78.1 76.6 69.8 62.7 54.5 3.0 6.7 5.3 6.5 8.3 0.5 1.0 1.2 1.7 2.2 1.2 1.5 2.0 2.5 3.0 Advanced countries 5.8 5.1 4.5 3.1 3.3 34.8 32.3 23.4 19.9 17.1 3.1 4.2 6.9 9.4 12.5 6.6 7.2 8.2 8.8 9.5 World average 46.3 43.0 36.0 29.8 24.8 15.7 15.1 14.3 13.5 12.9 1.4 2.0 3.4 5.2 7.1 3.2 3.6 4.5 5.3 6.1 Source: Updated version of Barro and Lee 2000. STATISTICAL APPENDIX TABLE A.25 Distribution of years of education, Gini coefficients 1960 1965 1970 1975 1980 1985 1990 Latin America and the Caribbean 50.1 50.2 47.0 46.0 43.1 42.6 41.8 Africa 78.5 77.7 74.7 70.2 66.3 63.8 61.8 Asia 67.0 64.5 61.4 58.4 54.3 51.5 47.9 Eastern Europe 29.6 29.8 27.6 28.0 25.3 26.1 22.8 Developed countries 29.3 29.2 28.2 29.2 27.1 27.2 27.5 Average 54.2 53.4 50.8 49.1 45.9 44.6 42.8 LAC countries Argentina 34.4 34.9 31.1 32.5 29.4 31.8 27.3 Barbados 25.3 27.9 18.0 19.1 29.6 29.9 30.0 Bolivia 51.9 54.6 55.7 56.4 55.9 54.7 53.7 Brazil 60.0 56.4 50.8 42.9 44.5 44.4 39.3 Chile 36.5 36.6 33.0 33.2 31.5 31.2 31.3 Colombia 53.4 49.3 50.9 46.0 47.2 47.5 48.6 Costa Rica 39.9 40.5 41.0 39.2 40.6 41.6 42.6 Cuba 44.6 46.0 31.7 34.0 32.2 33.1 33.5 Ecuador 51.3 51.5 51.1 47.0 39.3 44.0 44.9 Guatemala 75.8 75.2 74.3 73.2 63.8 63.6 62.6 Guyana 32.3 34.9 32.7 35.0 32.5 33.5 34.0 Haiti 92.7 91.7 85.1 84.6 78.0 63.9 65.0 Honduras 67.2 65.3 62.3 59.0 56.9 48.4 46.8 Jamaica 35.2 38.6 28.9 31.3 31.7 32.6 33.0 Mexico 56.0 56.9 51.0 49.8 49.7 46.9 38.4 Nicaragua 70.0 67.9 66.1 64.1 63.0 60.8 58.7 Panama 43.7 46.2 47.4 46.4 38.0 39.6 33.9 Paraguay 41.6 41.5 39.6 38.9 38.0 39.1 39.8 Peru 56.8 58.0 50.4 50.3 42.5 43.7 43.1 Trinidad & Tobago 35.7 36.0 33.7 33.9 23.9 29.7 31.2 Uruguay 38.8 38.1 39.2 34.9 35.8 33.5 34.2 RB de Venezuela 58.3 55.9 61.2 59.4 44.0 44.6 47.2 LAC Latin America and the Caribbean Source: Authors' calculations based on Thomas and others 2002. 311 INEQUALITY IN LATIN AMERICA TABLE A.26 Distdbution of years of education by age grwp, Gini coedfients Age brackets Country (25-65) (10-20) (21-30) (31-40) (41-50) (51-60) (61+) Argentina 1992 22.7 21.1 18.8 20.9 23.3 23.6 24.5 1996 22.5 21.4 17.6 20.3 23.1 24.5 26.3 2001 22.2 22.2 17.2 20.1 22.5 25.1 28.3 Bolivia Urban 1992 31.7 22.8 19.7 29.3 36.6 42.9 51.4 1996 33.6 22.6 23.8 30.9 35.8 44.6 51.2 National 1996 46.4 27.4 32.6 42.1 49.7 60.1 69.2 1999 43.4 26.2 28.1 38.2 45.7 55.9 69.6 Brazil 1990 47.6 39.3 36.3 43.3 51.6 56.1 66.6 1995 45.2 36.4 34.7 40.0 48.1 55.6 65.2 2001 41.1 30.2 30.4 36.5 42.1 51.5 62.3 Chile 1990 29.5 21.6 20.7 25.4 32.3 37.0 45.4 1996 27.1 21.0 18.4 22.2 28.9 35.6 43.2 2000 24.2 20.2 16.1 20.1 24.0 32.2 40.3 Colombia Bogota 1992 29.3 24.4 21.8 26.7 32.9 36.2 42.9 1996 28.1 22.4 20.5 25.2 30.8 34.9 39.3 National 1996 38.3 29.2 28.9 34.5 41.4 45.9 50.3 1999 37.3 27.8 27.7 32.8 39.5 46.1 52.6 Costa Rica 1990 32.3 23.3 23.3 27.9 35.2 41.7 46.1 1995 30.0 22.4 23.8 25.4 31.2 38.5 44.0 2000 29.7 22.8 25.7 25.7 28.6 37.5 46.0 Dominican Rep. 1995 36.7 26.6 27.4 32.5 39.3 46.6 55.5 Ecuador 1994 38.4 25.3 26.2 35.0 42.1 46.4 54.8 1998 35.3 23.5 24.2 30.5 38.4 45.5 53.3 El Salvador 1991 52.3 32.5 40.9 46.8 54.7 63.3 71.2 1995 52.4 37.9 39.2 47.0 55.5 63.2 73.5 2000 47.3 33.9 34.1 42.7 49.5 57.6 69.4 Guatemala 2000 61.8 42.9 48.7 57.7 65.2 74.6 78.5 Honduras 1990 55.8 36.1 42.5 50.2 61.0 70.8 72.9 1995 48.5 31.0 35.3 43.3 51.9 65.3 69.0 1999 47.7 30.6 36.7 42.6 S0.0 61.4 71.7 Jamaica 1990 20.3 17.5 11.7 17.6 19.2 20.2 25.8 1996 16.8 17.6 11.0 12.0 17.7 21.3 23.4 1999 16.2 17.0 10.6 12.7 16.5 20.0 18.4 Mexico 1992 42.8 25.5 30.3 39.1 46.1 52.0 61.5 1996 40.0 25.0 26.6 35.6 44.3 52.6 62.8 2000 36.6 23.1 24.7 31.4 38.7 49.8 60.2 Nicaragua 1993 53.0 41.2 41.0 50.3 58.4 66.3 72.8 1998 48.3 37.6 37.4 42.9 53.7 63.4 69.5 Panama 1991 32.2 24.3 22.7 27.8 34.4 40.3 47.7 1995 29.6 23.7 22.1 25.2 31.9 38.4 45.9 2000 24.7 22.7 20.3 21.8 25.5 29.0 31.5 Paraguay 1995 35.8 28.8 28.0 33.1 37.2 42.2 49.6 1999 34.5 26.9 27.8 30.8 35.6 40.5 47.7 Peru Regions 1991 25.361 20.235 17.009 22.905 28.471 29.332 29.563 1994 26.433 20.512 18.318 23.037 29.272 31.815 31.237 National 1994 27.342 21.218 19.51 24.248 30 32.744 31.744 2000 30.049 22.158 19.816 25.939 31.782 39.161 44.779 Trinidad & Tobago 1992 26.9 22.0 19.7 25.1 28.0 29.4 33.4 Uruguay 1989 29.0 23.4 21.4 25.0 28.1 31.1 37.9 1995 26.3 20.7 18.3 21.5 25.5 29.6 36.0 2000 24.2 27.5 19.2 21.4 23.0 26.7 33.9 RB de Venezuela 1989 33.4 26.1 24.6 29.2 35.9 44.7 S9.9 1995 31.4 24.3 23.0 27.1 32.1 42.4 57.6 1998 30.9 24.1 24.0 27.0 31.7 39.9 55.4 Source: Authors' calculations based on microdata from household surveys. 312 STATISTICAL APPENDIX TABLE A.27 Adut self-reported literacy rate by quintile Country 1 2 3 4 5 Average Country 1 2 3 4 5 Average Argentina Honduras 1992 0.96 0.98 0.98 0.99 1.00 0.98 1990 0.51 0.55 0.66 0.76 0.90 0.70 1996 0.96 0.98 0.99 1.00 1.00 0.99 1995 0.61 0.65 0.77 0.84 0.92 0.77 2001 0.98 0.98 0.99 1.00 1.00 0.99 1999 0.65 0.72 0.81 0.89 0.95 0.82 Bolivia Jamaica Urban 1990 0.91 0.94 0.95 0.98 0.99 0.96 1992 0.85 0.90 0.92 0.94 0.97 0.92 1996 0.98 0.98 0.96 0.97 0.99 0.98 1996 0.88 0.89 0.92 0.95 0.98 0.93 1999 0.96 0.97 0.97 0.96 0.97 0.97 National Mexico 1996 0.60 0.76 0.85 0.90 0.96 0.82 1992 0.66 0.83 0.88 0.93 0.98 0.87 1999 0.58 0.77 0.88 0.91 0.97 0.83 1996 0.70 0.82 0.91 0.94 0.98 0.89 Brazil 2000 0.70 0.87 0.93 0.95 0.99 0.90 1990 0.51 0.68 0.82 0.91 0.97 0.81 Nicaragua 1995 0.59 0.73 0.85 0.93 0.98 0.84 1993 0.38 0.52 0.64 0.76 0.80 0.64 2001 0.67 0.79 0.87 0.95 0.99 0.87 1998 (a) 0.46 0.57 0.69 0.76 0.87 0.69 Chile 1998 (b) 0.60 0.67 0.78 0.83 0.91 0.77 1990 0.91 0.92 0.94 0.97 0.99 0.95 Panama 1996 0.90 0.93 0.96 0.98 0.99 0.96 1991 0.74 0.85 0.91 0.94 0.97 0.90 2000 0.92 0.94 0.96 0.98 1.00 0.96 1995 0.78 0.89 0.93 0.96 0.98 0.92 Colombia 2000 0.94 0.96 0.98 0.98 1.00 0.98 Bogota Paraguay 1992 0.96 0.98 0.98 0.99 0.99 0.98 1995 0.68 0.79 0.83 0.89 0.95 0.84 1996 0.96 0.98 0.99 0.99 0.99 0.98 1999 0.79 0.86 0.92 0.96 0.98 0.91 National Peru 1996 0.90 0.94 0.95 0.97 0.99 0.96 Regions 1999 0.91 0.93 0.96 0.96 0.99 0.96 1991 0.81 0.90 0.94 0.97 0.98 0.92 Costa Rica 1994 0.74 0.88 0.92 0.95 0.98 0.90 1990 0.70 0.83 0.83 0.88 0.96 0.85 National 1995 0.76 0.84 0.88 0.91 0.97 0.88 1994 0.74 0.87 0.90 0.94 0.97 0.90 2000 0.76 0.86 0.88 0.92 0.97 0.89 2000 0.82 0.89 0.94 0.96 0.99 0.93 Dominican Rep. Trinidad & Tobago 1995 0.83 0.88 0.90 0.93 0.97 0.91 1992 0.71 0.74 0.81 0.83 0.88 0.81 Ecuador Uruguay 1994 0.74 0.85 0.86 0.91 0.96 0.87 1989 0.88 0.91 0.95 0.96 0.96 0.94 1998 0.78 0.86 0.90 0.93 0.98 0.90 1995 0.91 0.93 0.96 0.97 0.99 0.95 El Salvador 2000 0.93 0.96 0.97 0.98 0.99 0.97 1991 0.49 0.57 0.68 0.76 0.88 0.69 RB de Venezuela 1995 0.52 0.60 0.72 0.84 0.93 0.75 1989 0.78 0.86 0.90 0.94 0.98 0.90 2000 0.64 0.71 0.79 0.86 0.93 0.80 1995 0.82 0.88 0.92 0.94 0.98 0.92 Guatemala 1998 0.86 0.90 0.93 0.95 0.99 0.93 2000 0.40 0.52 0.59 0.72 0.91 0.65 Note: Data for Nicaragua 1993 and 1998 (a): share of adults with less than three years of formal education; for Nicaragua 1998 (b): self-reported literacy rate. 5ource: Authors' calculations based on microdata from household surveys. 313 INEQUALITY IN LATIN AMERICA TABLE A.28 Educa ional snwctre of the adult populton by age and gender (population shares Population, 18-65 Population, 25-55 Males, 25-55 Females, 25-55 Working males, 25-55 Low Medium High Low Medium High Low Medium High Low Medium High Low Medium High Countty (i) (ii (iii) (iv) (v) (vi) (vii) (vill) (ix) (x) (xi) (xii) (xiii) (xiv) (xv) Argentina 1992 42.2 38.8 18.9 42.5 37.3 20.2 42.1 37.5 20.4 42.9 37.2 19.9 41.7 38.5 19.8 1996 42.0 39.4 18.6 41.2 37.8 21.0 41.2 38.4 20.4 41.2 37.3 21.6 42.3 38.1 19.6 2001 37.8 39.8 22.4 38.0 36.7 25.3 39.3 38.2 22.6 36.9 35.4 27.7 37.7 39.4 22.9 Bolivia Urban 1992 40.1 45.1 14.8 43.3 37.6 19.1 35.1 44.8 20.2 50.8 31.1 18.2 35.9 43.5 20.6 1996 41.4 37.2 21.5 42.7 30.6 26.7 34.2 33.6 32.2 50.0 28.1 21.9 35.3 34.3 30.4 National 1996 62.2 25.3 12.6 65.1 20.8 14.2 59.1 23.8 17.1 70.5 18.0 11.5 58.0 24.7 17.3 1999 54.1 30.0 15.8 59.0 22.6 18.4 52.7 25.9 21.5 64.8 19.6 15.6 52.1 26.9 21.0 Brazil 1990 76.4 17.6 6.0 75.8 16.5 7.8 76.1 16.1 7.7 75.4 16.8 7.8 76.1 16.1 7.7 1995 73.8 19.9 6.4 73.1 18.8 8.1 74.3 17.8 7.9 72.0 19.7 8.3 74.2 17.9 7.9 2001 64.6 27.9 7.5 66.1 24.7 9.1 67.9 23.6 8.5 64.5 25.8 9.8 67.8 23.8 8.4 Chile 1990 43.7 42.7 13.6 50.0 37.6 12.4 43.7 40.2 16.1 46.2 40.0 13.8 43.6 40.2 16.2 1996 37.7 45.9 16.4 38.5 44.7 16.9 37.6 44.0 18.3 39.3 45.2 15.5 37.5 44.2 18.4 2000 33.1 48.3 18.6 33.0 47.0 20.0 32.5 46.2 21.3 33.5 47.7 18.9 32.3 46.2 21.5 Colombia Bogota 1992 42.6 39.0 18.3 43.2 34.7 22.1 40.5 33.8 25.8 45.6 35.4 19.0 40.3 34.3 25.4 1996 43.5 41.1 15.5 45.0 36.9 18.1 44.1 36.6 19.3 45.8 37.2 17.0 43.5 37.6 19.0 National 1996 62.8 28.9 8.2 63.8 26.1 10.0 63.8 25.3 10.9 63.8 26.9 9.3 64.1 25.4 10.6 1999 57.8 31.9 10.4 59.6 28.1 12.4 60.0 26.7 13.3 59.1 29.3 11.6 61.5 26.1 12.3 Costa Rica 1990 59.4 29.4 11.2 59.7 27.5 12.8 59.3 27.5 13.1 59.9 27.5 12.6 59.3 27.4 13.3 1995 55.4 30.6 14.0 54.9 29.5 15.6 55.2 28.7 16.1 54.6 30.3 15.1 56.2 28.0 15.8 2000 55.2 30.7 14.1 58.6 27.6 13.9 58.5 27.8 13.7 58.6 27.4 14.1 59.6 27.5 12.9 Dominican Rep. 1995 57.9 30.6 11.5 59.7 26.8 13.5 60.3 25.7 14.0 59.1 27.8 13.0 60.2 25.7 14.1 Ecuador 1994 61.4 27.1 11.5 64.1 21.9 14.0 63.3 21.8 14.9 64.9 22.1 13.1 63.1 21.9 15.0 1998 56.5 30.3 13.2 56.9 27.1 16.1 56.3 26.6 17.1 57.4 27.6 15.0 55.7 26.8 17.5 El Salvador 1991 67.9 24.8 7.3 70.9 21.4 7.7 67.0 23.8 9.2 74.1 19.5 6.4 66.0 24.6 9.4 1995 67.0 26.0 7.1 69.1 22.4 8.5 65.0 24.9 10.1 72.4 20.5 7.2 60.5 28.2 11.4 2000 59.5 30.6 9.9 61.6 26.7 11.7 57.7 29.4 12.9 64.7 24.5 10.8 62.6 29.2 8.2 Guatemala 2000 79.8 14.9 5.4 81.1 12.6 6.4 77.9 13.4 8.7 83.8 11.9 4.3 74.4 15.6 9.9 Honduras 1990 84.0 12.8 3.2 80.2 14.9 5.0 80.0 13.5 6.5 80.4 16.1 3.6 80.0 13.3 6.6 1995 79.4 17.1 3.5 73.6 21.1 5.3 73.2 19.9 6.9 74.0 22.1 3.9 73.0 20.1 6.9 1999 77.8 17.7 4.6 73.1 20.5 6.5 73.5 18.5 8.1 72.7 22.1 5.2 72.7 22.1 5.2 Jamaica 1990 38.7 56.1 5.2 41.1 52.8 6.1 42.6 52.3 5.1 39.7 53.3 7.0 40.3 55.4 4.4 1999 31.1 63.1 5.8 27.8 65.7 6.6 26.7 67.5 5.9 28.8 64.0 7.2 22.6 71.9 5.5 Mexico 1992 62.5 28.1 9.4 67.6 22.1 10.4 64.3 22.0 13.8 70.6 22.2 7.2 64.4 22.0 13.6 1996 56.6 32.8 10.7 59.0 28.8 12.2 55.8 29.0 15.2 62.0 28.6 9.4 55.5 29.1 15.3 2000 50.3 35.8 14.0 51.6 33.9 14.5 48.1 33.5 18.4 54.8 34.1 11.1 47.1 34.0 19.0 Nicaragua 1993 77.7 18.7 3.6 77.2 17.7 5.2 76.6 16.9 6.5 77.7 18.3 4.0 74.1 18.5 7.4 1998 74.9 20.8 4.3 74.9 19.8 5.3 73.9 19.9 6.2 75.7 19.8 4.5 71.4 21.6 7.1 Panama 1991 49.0 37.5 13.5 49.3 34.2 16.4 51.4 32.9 15.7 47.4 35.4 17.1 51.5 32.9 15.6 1995 45.2 39.9 14.9 44.7 37.6 17.7 46.5 36.9 16.6 42.9 38.3 18.9 46.5 36.9 16.6 2000 39.0 43.0 18.0 38.5 40.7 20.7 40.7 40.8 18.5 36.5 40.6 22.9 37.2 43.0 19.8 Paraguay 1995 71.0 22.0 7.0 71.7 20.1 8.2 70.1 21.7 8.2 73.3 18.5 8.2 69.1 22.5 8.5 1999 65.1 27.0 7.9 67.4 23.5 9.2 65.5 25.8 8.7 69.2 21.2 9.6 60.7 29.4 9.9 Peru Regions 1991 34.2 48.2 17.6 36.8 41.5 21.8 32.2 42.8 25.1 41.1 40.2 18.6 29.6 44.2 26.2 1994 34.2 45.1 20.7 35.7 39.7 24.7 33.1 40.5 26.4 38.1 38.9 23.0 31.8 41.0 27.3 National 1994 37.7 43.3 19.0 39.2 38.2 22.6 36.5 39.4 24.1 41.8 37.1 21.1 34.6 40.0 25.4 2000 37.5 42.7 19.8 37.9 39.4 22.8 32.1 43.8 24.1 43.2 35.3 21.6 30.3 45.1 24.6 Trinidad & Tobago 1992 49.8 42.2 8.0 52.9 38.4 8.8 54.3 35.9 9.9 51.6 40.8 7.7 55.3 34.2 10.5 Uruguay 1989 53.5 35.3 11.3 51.3 35.9 12.8 49.7 36.6 13.7 52.8 35.3 12.0 49.6 36.6 13.8 1995 45.4 41.4 13.2 43.4 40.8 15.8 43.7 41.6 14.7 43.1 40.2 16.7 43.8 41.6 14.6 2000 41.6 41.9 16.5 39.5 43.0 17.6 40.0 44.6 15.4 39.1 41.5 19.5 39.7 44.7 15.5 RB de Venezuela 1989 61.0 30.4 8.6 61.9 27.5 10.6 61.2 27.3 11.6 62.6 27.7 9.7 61.6 27.7 10.7 1998 52.6 33.6 13.8 52.7 32.3 15.0 54.1 32.1 13.9 51.5 32.4 16.1 53.3 32.5 14.2 Note: Low education = less than 8 years of schooling; medium = between 9 and 13 years of schooling; high = more than 14 years of schooling. Source: Authors' calculations based on microdata from household surveys. 314 STATISTICAL APPENDIX TABLE A.29 Distribution of hourly waps, Gin! coefficients Urban Men 25-55 Urban All Urban All workers workers Men 16-65 Men 25-55 Urban salaried Salaried salaried workers workers Primary job Primary job Primary job Primary job Primary job workers workers Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) Argentina 1992 38.2 38.2 38.0 38.0 38.9 39.8 38.3 36.1 36.1 1996 40.4 40.4 40.1 40.1 40.5 40.4 38.9 37.9 37.9 2001 43.3 43.3 43.2 43.2 44.6 43.8 42.6 41.1 41.1 Bolivia Urban 1992 56.9 56.9 57.0 57.0 54.2 50.9 48.2 56.1 56.1 1996 54.3 54.3 54.3 54.3 52.5 52.4 48.6 52.1 52.1 National 1996 57.0 53.8 56.9 53.8 56.8 57.2 47.9 50.7 51.4 1999 60.9 55.9 61.7 56.7 57.2 57.5 50.8 52.3 52.4 Brazil 1990 60.2 58.8 60.3 59.0 59.2 58.2 54.0 58.3 57.0 1995 58.8 57.7 59.1 58.0 59.1 58.2 53.9 55.9 55.1 2001 57.4 56.6 57.6 56.8 58.0 57.6 53.7 53.8 53.4 Chile 1990 56.8 56.0 56.8 56.0 57.7 57.1 50.4 1996 56.3 55.7 56.3 55.7 56.8 56.3 49.0 2000 55.8 55.4 55.8 55.4 57.5 57.5 49.5 Colombia Bogota 1992 51.5 51.5 51.5 51.5 51.5 50.3 48.5 48.0 48.0 1996 52.3 52.3 51.9 51.9 53.5 52.4 49.6 50.0 50.0 National 1996 52.2 50.3 51.7 49.5 51.8 51.7 45.9 46.2 46.8 1999 54.1 51.3 54.1 51.3 55.0 55.0 48.2 48.9 46.9 Costa Rica 1990 42.1 42.2 42.2 42.4 41.7 42.1 40.9 39.8 40.6 1995 42.6 41.7 42.8 41.8 41.6 42.3 39.4 40.1 39.9 2000 43.2 44.0 43.2 44.0 42.2 41.9 41.2 39.8 41.1 Dominican Rep. 1995 49.1 52.2 49.1 52.2 48.1 47.2 42.8 43.3 45.9 1997 45.6 46.3 45.6 46.4 44.7 46.0 41.7 43.6 44.0 Ecuador 1994 54.1 50.9 54.6 51.6 52.7 54.0 47.3 47.1 48.9 1998 54.3 53.0 53.3 52.2 50.9 51.7 44.1 47.5 47.6 El Salvador 1991 54.0 49.6 50.6 48.4 49.5 49.8 41.8 49.5 43.5 1995 49.2 48.6 47.6 47.4 45.3 45.2 39.1 45.3 44.0 2000 48.4 48.9 48.0 48.9 45.1 45.4 38.6 38.5 39.3 Guatemala 2000 56.9 55.3 56.6 56.1 54.7 56.3 51.6 56.9 55.3 Honduras 1990 56.2 55.7 56.1 54.7 54.7 55.1 47.9 52.6 51.8 1995 54.7 52.6 55.5 52.7 56.1 55.9 41.3 44.3 44.0 1999 56.3 53.0 57.7 54.3 57.5 57.8 47.1 49.6 48.0 Jamaica 1990 42.5 41.2 42.5 41.2 39.8 41.9 36.0 40.6 41.1 1999 43.0 45.5 43.0 45.5 39.6 41.9 45.5 42.9 45.8 Mexico 1992 55.6 53.2 56.3 54.1 56.7 55.3 48.5 49.5 48.1 1996 54.9 52.4 55.5 53.3 55.4 55.1 49.4 51.3 49.4 2000 54.1 51.2 53.9 51.6 54.0 53.2 49.0 49.6 47.8 Nicaragua 1993 53.2 50.4 53.1 50.3 55.7 56.1 47.1 46.6 45.2 1998 56.6 56.2 56.8 57.1 58.8 58.3 53.9 52.4 54.1 Panama 1991 48.1 48.4 47.7 46.4 45.7 1995 49.3 48.3 49.0 48.1 49.1 48.7 46.0 47.2 46.3 2000 50.0 48.8 50.8 49.6 51.7 51.0 46.2 45.6 44.7 Paraguay 1995 59.0 56.7 57.2 54.8 57.3 57.8 45.1 48.1 48.4 1999 52.7 50.6 53.1 51.2 52.5 52.2 44.1 45.7 45.5 Peru Regions 1991 50.9 48.3 50.5 48.4 47.6 45.5 39.4 44.3 43.8 1994 51.8 49.5 51.4 49.5 50.1 49.1 41.6 45.1 44.0 National 1994 51.6 49.3 51.2 49.1 50.0 49.0 41.3 45.1 43.8 2000 52.5 50.0 52.4 50.2 50.2 50.0 42.8 46.7 45.5 Trinidad & Tobago 1992 43.3 46.1 43.3 46.1 40.4 37.7 38.3 41.1 44.7 Uruguay 1989 43.4 43.4 42.0 42.0 40.7 38.3 35.9 37.8 37.8 1995 44.4 44.4 45.2 45.2 44.6 42.9 40.8 41.3 41.3 2000 44.7 44.7 45.3 45.3 44.7 42.7 39.7 40.9 40.9 RB de Venezuela 1989 37.1 36.5 37.1 36.5 36.5 35.9 33.7 33.4 32.2 199S 43.8 43.2 43.8 43.2 43.0 42.2 35.6 38.9 36.7 1998 46.6 44.9 46.6 44.9 46.2 46.3 43.2 42.5 41.3 Source: Authors' calculations based on microdata from household surveys. 315 INEQUALITY IN LATIN AMERICA TABLE A.30 Ratio of hourly wages by educational groups, -nen ages 25-55 High/medium High/low Medium/low High/medium High/low Medium/low Country (i) (ii) (iii) Country (i) (ii) (iii) Argentina Guatemala 1992 1.9 2.5 1.4 2000 2.4 5.0 2.1 1996 2.0 2.6 1.3 Honduras 2001 2.1 2.8 1.4 1990 2.1 4.9 2.3 Bolivia 1995 3.0 5.5 1.8 Urban 1999 1.9 3.9 2.1 1992 1.9 2.5 1.3 Jamaica 1996 2.2 2.9 1.3 1990 1.7 2.0 1.1 National 1999 1.9 2.6 1.4 1996 2.1 3.8 1.7 Mexico 1999 2.1 3.7 1.7 1992 2.2 3.9 1.8 Brazil 1996 2.2 4.2 1.9 1990 2.2 5.8 2.6 2000 2.7 4.2 1.5 1995 2.6 6.2 2.4 Nicaragua 2001 2.8 6.5 2.3 1993 1.7 3.2 1.9 Chile 1998 2.5 4.7 1.8 1990 2.8 3.8 1.4 Panama 1996 2.8 4.6 1.7 1991 2.2 3.3 1.5 2000 3.1 5.2 1.7 1995 2.3 3.8 1.7 Colombia 2000 2.3 3.6 1.6 Bogota Paraguay 1992 2.5 4.0 1.6 1995 2.0 3.6 1.8 1996 3.2 4.6 1.4 1999 2.1 2.9 1.4 National Peru 1996 2.6 4.8 1.8 Regions 1999 2.5 4.7 1.9 1991 1.3 1.9 1.5 Costa Rica 1994 1.6 2.4 1.5 1990 2.1 3.2 1.5 National 1995 2.1 3.0 1.4 1994 1.6 2.3 1.5 2000 2.0 2.9 1.5 2000 1.9 3.0 1.5 Dominican Rep. Trinidad & Tobago 1995 2.0 2.7 1.3 1992 2.0 2.6 1.3 1997 1.9 2.5 1.3 Uruguay Ecuador 1989 1.6 2.2 1.4 1994 2.0 2.6 1.3 1995 1.9 2.9 1.5 1998 1.8 3.0 1.7 2000 1.9 2.7 1.4 El Salvador RB de Venezuela 1991 2.0 3.4 1.6 1989 1.7 2.5 1.4 1995 2.1 3.4 1.6 1995 1.5 2.4 1.6 2000 2.0 3.1 1.6 1998 2.0 2.7 1.4 Source: Authors' calculations based on microdata from household surveys. 316 TABLE A.31 Marginal returns on completing a given educational level: Mincer equation coefficients on education All workers Urban salaried workers Men Women Men Women Primary Secondary College Primary Secondary College Primary Secondary College Primary Secondary College Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) Argentina 1992 0.179 0.395 0.540 -0.037 0.385 0.403 0.155 0.437 0.526 0.017 0.387 0.223 1996 0.122 0.346 0.721 0.006 0.290 0.601 0.108 0.328 0.722 0.095 0.226 0.523 2001 0.217 0.402 0.758 0.138 0.404 0.630 0.262 0.353 0.745 0.130 0.390 0.580 Bolivia Urban 1992 0.384 0.002 0.631 0.190 0.195 0.594 0.302 0.058 0.680 0.014 0.660 0.670 1996 0.077 0.138 0.908 0.050 0.247 0.991 -0.010 0.149 1.073 0.043 0.518 0.889 National 1996 0.283 0.177 0.882 0.201 0.244 0.843 0.010 0.166 1.025 0.097 0.442 0.905 1999 0.129 0.413 0.766 0.184 0.664 0.782 0.264 0.073 0.800 0.114 0.621 0.753 Brazil 1990 0.505 0.415 0.847 0.509 0.500 0.933 0.493 0.478 0.870 0.525 0.545 0.929 1995 0.459 0.412 0.885 0.459 0.594 1.030 0.458 0.426 0.941 0.399 0.485 0.818 2001 0.399 0.392 0.942 0.329 0.376 0.903 0.382 0.395 0.996 0.297 0.437 0.961 Chile 1990 0.107 0.383 0.808 -0.007 0.716 0.952 0.109 0.380 0.835 -0.024 0.591 0.661 1996 0.057 0.425 0.966 0.099 0.652 1.091 0.040 0.445 0.934 0.005 0.510 0.747 2000 0.084 0.388 0.942 0.115 0.473 0.971 0.088 0.398 0.932 0.015 0.405 0.813 Colombia Bogota 1992 0.141 0.661 0.891 0.262 0.640 0.804 0.094 0.743 0.874 0.469 0.579 0.824 1996 0.088 0.453 0.990 0.169 0.576 0.966 0.096 0.473 1.020 0.098 0.446 0.911 National 1996 0.204 0.478 0.840 0.255 0.598 0.777 0.134 0.400 0.943 0.128 0.471 0.793 1999 0.202 0.429 0.905 0.227 0.521 0.859 0.094 0.407 0.875 0.054 0.360 0.833 Costa Rica 1990 0.142 0.440 0.694 0.211 0.505 0.668 0.130 0.523 0.670 0.014 0.536 0.583 1995 0.157 0.368 0.688 0.196 0.428 0.716 0.178 0.352 0.694 0.361 0.441 0.643 2000 0.077 0.380 0.665 0.023 0.461 0.694 O.OSS 0.440 0.748 0.155 0.327 0.694 Dominican Rep. 1995 0.216 0.114 0.862 0.392 0.485 1.198 0.266 0.033 0.904 0.393 0.345 0.815 1997 0.246 0.107 0,711 0.280 0.553 0.955 0.169 0.181 0.746 0.129 0.570 0.611 Ecuador 1994 0.212 0.399 0.558 0.008 0.641 0.721 0.144 0.529 0.515 0.196 0.571 0.496 1998 0.306 0.378 0.645 0.124 0.649 0.567 -0.050 0.470 0.691 -0.253 0.721 0.411 El Salvador 1991 0.235 0.351 0.539 0.185 0.537 0.538 0.176 0.436 0.556 0.268 0.563 0.458 1995 0.156 0.433 0.668 0.159 0.737 0.729 0.196 0.382 0.622 0.333 0.762 0.667 2000 0.141 0.351 0.531 0.234 0.221 0.589 0.203 0.265 0.408 0.298 0.262 0.459 Guatemala 2000 0.356 0.652 0.498 0.307 0.876 0.515 0.370 0.693 0.552 0.193 0.898 0.509 Honduras 1990 0.368 0.622 0.714 0.349 1.177 0.655 0.376 0.580 0.738 0.342 1.248 0.630 1995 0.243 0.616 0.851 0.384 0.892 0.478 0.207 0.527 0.745 0.321 0.919 0.447 1999 0.330 0.655 0.659 0.409 0.797 0.587 0.331 0.626 0.765 0.391 0.789 0.732 Jamaica 1990 0.762 0.116 1.000 -0.032 0.386 0.534 0.231 0.280 0.946 0.378 0.431 0.430 1999 0.000 0.296 0.943 0.005 0.478 1.385 0.431 0.090 0.983 0.071 0.490 1.226 Mexico 1992 0.326 0.504 0.755 0.217 1.051 0.173 0.239 0.573 0.683 0.221 0.957 0.206 1996 0.303 0.686 0.575 0.283 0.858 0.389 0.306 0.641 0.649 0.285 0.898 0.383 2000 0.347 0.449 0.840 0.195 0.663 0.416 0.259 0.459 0.846 0.210 0.856 0.533 Nicaragua 1993 0.190 0.296 0.435 0.098 0.288 0.379 0.236 0.219 0.576 0.043 0.345 0.502 1998 0.350 0.290 0.771 0.136 0.236 0.823 0.269 0.418 0.696 0.222 0.354 0.803 Panama 1991 0.081 0.377 0.805 0.273 0.815 0.743 0.193 0.509 0.750 0.255 0.814 0.698 1995 0.232 0.424 0.851 0.223 0.799 0.797 0.271 0.472 0.841 0.193 0.752 0.804 2000 0.182 0.365 0.956 0.256 0.534 0.712 0.126 0.310 0.936 0.159 0.430 0.768 Paraguay 1995 0.193 0.543 0.822 0.163 0.703 0.647 0.269 0.473 0.723 0.325 0.673 0.633 1999 0.250 0.425 0.837 0.335 0.439 0.983 0.333 0.349 0.821 0.401 0.502 0.706 Peru Regions 1991 0.300 0.139 0.360 0.047 0.285 0.458 0.118 0.342 0.314 -0.167 0.217 0.407 1994 -0.005 0.332 0.513 0.151 0.299 0.410 0.125 0.400 0.510 -0.224 0.410 0.567 National 1994 0.016 0.321 0.496 0.180 0.286 0.407 0.077 0.391 0.492 -0.096 0.397 0.554 2000 0.156 0.370 0.637 0.021 0.368 0.727 0.167 0.189 0.690 0.078 0.356 0.572 Trinidad & Tobago 1992 0.208 0.252 0.601 -0.027 0.682 0.965 0.152 0.379 0.563 0.309 0.583 0.881 Uruguay 1989 0.088 0.338 0.600 -0.010 0.433 0.544 0.088 0.336 0.599 -0.012 0.437 0.529 1995 0.140 0.426 0.626 0.088 0.608 0.627 0.129 0.462 0.593 0.054 0.604 0.520 2000 0.150 0.403 0.689 0.097 0.511 0.688 0.215 0.407 0.681 0.082 0.478 0.580 RB de Venezuela 1989 0.257 0.338 0.504 0.327 0.521 0.480 0.09S 0.322 0.522 0.151 0.422 0.498 1995 0.215 0.355 0.444 0.172 0.278 0.371 0.098 0.261 0.483 0.240 0.389 0.543 1998 0.192 0.347 0.510 0.092 0.246 0.408 -0.020 0.418 0.638 -0.018 0.638 0.524 Notes: Regressions are estimated by Heckman's two-stage procedure. Dependent variable: logarithm of hourly wage from primary job of individuals ages 25-SS. Explanatory variables: educational dummies, age, age squared, regional dummies, and urban/rural dummy. Selection equation: same variables plus number of children and school attendance. Source: Authors' calculations based on microdata from household surveys. 317 INEQUALITY IN LATIN AMERICA TABLE A.32 Dispersion in unobservables and gender wage gap Dispersion in unobservables Dispersion in unobservables All workers Urban salaried Gender wage gap All workers Urban salaried Gender wage gap Urban salaried Urban salaried Men Women Men Women workers Men Women Men Women workers Country (i) (ii) (iii) (iv) (v) Country (i) (ii) (iii) (iv) (v) Argentina Guatemala 1992 0.571 0.596 0.532 0.507 0.867 2000 0.804 0.942 0.753 0.921 0.757 1996 0.600 0.621 0.553 0.518 0.909 Honduras 2001 0.659 0.652 0.601 0.543 0.874 1990 0.851 0.884 0.620 0.654 0.777 Bolivia 1995 0.862 0.813 0.549 0.590 0.817 Urban 1999 0.967 0.902 0.628 0.625 0.834 1992 0.764 0.915 0.676 0.695 0.765 Jamaica 1996 0.736 1.013 0.663 0.686 0.801 1990 0.638 0.560 0.524 0.572 0.599 National 1999 0.615 0.803 0.608 0.689 0.817 1996 0.953 0.944 0.658 0.702 0.792 Mexico 1999 1.009 1.131 0.689 0.764 0.778 1992 0.783 0.821 0.644 0.625 0.824 Brazil 1996 0.790 0.834 0.662 0.635 1.074 1990 0.806 0.784 0.735 0.702 0.636 2000 0.738 0.930 0.609 0.633 0.960 1995 0.770 0.927 0.689 0.665 0.625 Nicaragua 2001 0.748 0.727 0.660 0.636 0.693 1993 0.930 0.795 0.684 0.633 0.828 Chile 1998 0.960 0.831 0.769 0.703 0.779 1990 0.780 0.914 0.662 0.618 0.699 Panama 1996 0.735 1.010 0.624 0.625 0.754 1991 0.756 0.659 0.587 0.602 0.798 2000 0.727 0.802 0.577 0.574 0.777 1995 0.665 0.673 0.583 0.593 0.743 Colombia 2000 0.698 0.658 0.587 0.556 0.824 Bogota Paraguay 1992 0.856 0.836 0.750 0.774 0.825 1995 0.898 0.929 0.603 0.657 0.812 1996 0.718 0.670 0.597 0.536 0.802 1999 0.815 0.831 0.585 0.585 0.853 National Peru 1996 0.795 0.742 0.553 0.519 0.822 Regions 1999 0.875 0.809 0.608 0.574 0.887 1991 0.786 1.051 0.676 0.664 0.835 Costa Rica 1994 0.770 0.857 0.609 0.609 0.869 1990 0.634 0.671 0.562 0.514 0.770 National 1995 0.634 0.678 0.525 0.535 0.789 1994 0.784 0.856 0.628 0.601 0.881 2000 0.629 0.689 0.553 0.554 0.845 2000 0.867 0.918 0.663 0.749 0.903 Dominican Rep. Trinidad & Tobago 1995 0.716 0.953 0.628 0.603 0.825 1992 0.637 0.700 0.573 0.566 0.858 1997 0.664 0.929 0.595 0.606 0.829 Uruguay Ecuador 1989 0.604 0.667 0.535 0.601 0.752 1994 0.853 0.940 0.683 0.752 0.754 1995 0.667 0.688 0.585 0.600 0.768 1998 0.838 0.947 0.661 0.700 0.798 2000 0.662 0.657 0.574 0.588 0.789 El Salvador RB de Venezuela 1991 0.849 0.806 0.601 0.597 0.806 1989 0.552 0.547 0.476 0.426 0.832 1995 0.606 0.716 0.539 0.627 0.782 1995 0.722 0.863 0.593 0.599 0.858 2000 0.691 0.830 0.628 0.596 0.815 1998 0.814 0.928 0.667 0.629 0.855 Notes: Dispersion in unobservables = standard deviation of error term in Mincer equations of table A.31. Gender wage gap = ratio of the mean of the counterfactual wage that men would earn if they were paid like women over the actual average wage for men. Higher values in column (v) mean a narrower gender gap in hourly wages. Source: Authors' calculations based on microdata from household surveys. 318 STATISTICAL APPENDIX TABLE A.33 Distdibution .o1 hurly wages by educational group, mlen ages 25-55, Gini coefficients Educational groups Educational groups Low Medium High Total Low Medium High Total Country (i) () (iii) (iv) Country (i) (ii) (iii) (iv) Argentina Guatemala 1992 31.1 34.0 41.5 39.8 2000 47.2 43.2 46.6 56.3 1996 32.2 33.8 40.6 40.4 Honduras 2001 35.3 36.6 42.5 43.8 1990 51.0 40.9 39.8 55.1 Bolivia 1995 51.1 41.1 54.6 56.0 Urban 1999 55.3 47.6 42.5 57.8 1992 46.9 48.0 47.1 50.9 Jamaica 1996 40.0 49.2 49.1 52.4 1990 44.2 40.7 19.9 41.9 National 1999 29.7 42.9 33.9 41.9 1996 53.3 48.9 48.8 57.2 Mexico 1999 58.7 44.1 48.6 57.5 1992 48.4 48.6 45.9 55.3 Brazil 1996 45.6 49.8 47.0 55.1 1990 49.7 47.0 41.9 58.2 2000 45.0 42.2 45.2 53.2 1995 48.7 46.7 44.9 58.2 Nicaragua 2001 45.6 47.6 48.0 57.6 1993 54.2 48.4 51.3 56.1 Chile 1998 53.1 49.0 58.9 58.3 1990 50.3 48.2 53.7 57.1 Panama 1996 43.1 46.5 53.0 56.3 1991 38.2 39.0 42.3 46.4 2000 39.8 45.8 55.6 57.5 1995 37.4 41.2 44.2 48.7 Colombia 2000 39.7 44.9 46.4 51.0 Bogota Paraguay 1992 39.0 42.5 39.7 50.3 1995 55.3 51.4 45.9 57.8 1996 39.0 40.1 44.7 52.4 1999 53.4 40.4 43.8 52.2 National Peru 1996 42.2 42.5 42.7 51.7 Regions 1999 44.9 47.6 45.4 55.0 1991 43.0 46.0 40.9 45.5 Costa Rica 1994 44.0 47.3 45.9 49.1 1990 34.5 34.7 36.7 42.1 National 1995 34.8 34.7 37.0 42.3 1994 46.0 46.7 45.5 49.0 2000 33.2 38.1 36.7 41.9 2000 48.7 43.1 46.6 50.0 Dominican Rep. Trinidad & Tobago 1995 41.2 44.5 45.5 47.2 1992 32.2 31.9 38.3 37.7 1997 42.6 41.7 43.3 46.0 Uruguay Ecuador 1989 31.8 36.1 41.3 38.3 1994 51.7 46.2 48.2 54.0 1995 34.0 38.7 43.4 42.9 1998 44.3 50.1 45.5 51.7 2000 34.9 38.3 43.3 42.7 El Salvador RB de Venezuela 1991 46.6 41.6 41.7 49.8 1989 30.8 32.1 33.2 35.9 1995 38.7 37.2 37.0 45.2 1995 39.1 39.5 36.5 42.2 2000 40.4 41.6 42.8 45.4 1998 40.7 40.8 44.2 46.3 Source: Authors' calculations based on microdata from household surveys. 319 INEQUALITY IN LATIN AMERICA TABLE A.34 Wages and hours worked, by education level Correlation between Correlation between hours worked and hours worked and hourly wages Hours worked hourly wages Hours worked Urban Education Urban Education salaried salaried All workers workers Low Medium High Total All workers workers Low Medium High Total Country (i) (ii) (iii) (iv) (v) (vi) Country (i) (ii) (iii) (iv) (v) (vi) Argentina Honduras 1992 -0.1700* -0.1617* 44.6 44.7 40.0 43.7 1990 -0.1451* -0.2773* 46.7 45.0 43.2 46.4 1996 -0.2131* -0.1835* 42.2 45.6 41.5 43.3 1995 -0.1330* -0.2341* 48.0 46.8 46.4 47.8 2001 -0.1916* -0.1758* 40.2 44.0 40.2 41.6 1999 -0.1336* -0.2261* 47.0 47.1 48.7 47.1 Bolivia Jamaica Urban 1990 -0.0872* -0.1035* 38.7 40.1 38.7 39.6 1992 -0.1355* -0.1359* 50.8 48.7 41.1 48.3 1999 -0.0666* -0.0525 41.0 42.9 41.5 42.4 1996 -0.1872* -0.2472* 48.9 48.1 41.9 47.2 Mexico National 1992 -0.1267* -0.1787* 48.1 46.6 44.4 47.3 1996 -0.1776* -0.2479* 48.6 47.9 41.4 47.5 1996 -0.1290* -0.1437* 47.0 47.0 44.7 46.8 1999 -0.2094* -0.2484* 49.6 50.0 42.2 48.4 2000 -0.1361* -0.2006* 48.3 48.2 44.0 47.6 Brazil Nicaragua 1990 -0.0718* -0.1368* 44.5 41.7 41.1 43.8 1993 -0.1563* -0.1677* 49.1 46.3 46.2 48.4 1995 -0.0915* -0.1356* 44.7 42.5 42.1 44.0 1998 -0.1456* -0.1756* 49.9 48.4 46.7 49.4 2001 -0.1081* -0.1276* 44.5 43.1 42.2 43.9 Panama Chile 1991 -0.1164* -0.1713* 42.3 42.5 41.7 42.3 1990 -0.1581* -0.2322* 50.4 50.5 45.6 49.6 1995 -0.1078* -0.1462* 43.1 43.7 43.3 43.4 2000 -0.1507* -0.1856* 48.0 49.0 46.0 48.0 2000 -0.1237* -0.1705* 42.2 43.7 43.1 43.1 Colombia Paraguay Bogota 1995 -0.1642* -0.2746* 41.8 45.0 40.9 42.4 1992 -0.1986* -0.2142* 49.1 46.8 44.8 47.4 1999 -0.1591* -0.1960* 45.9 47.6 43.5 46.2 1996 -0.1809* -0.2157* 49.3 48.2 45.6 48.3 Peru National Regions 1996 -0.1777* -0.2390* 48.2 47.1 44.2 47.5 1991 -0.2170* -0.2321* 47.0 46.2 43.5 45.9 1999 -0.1280* -0.2709* 47.5 46.9 44.1 46.9 1994 -0.2417* -0.2098* 47.6 48.4 44.0 47.1 Costa Rica National 1990 -0.1655* -0.1853* 47.3 47.7 44.6 47.1 1994 -0.2371* -0.2169* 47.2 48.3 44.1 46.9 1995 -0.1653* -0.1662* 45.8 47.7 45.3 46.3 2000 -0.2137* -0.2085* 46.1 49.9 46.3 47.7 2000 -0.1795* -0.2255* 46.7 48.1 45.6 46.9 Trinidad & Tobago Dominican Rep. 1992 -0.1493* -0.1904* 38.7 40.8 41.1 39.9 1995 -0.1144* -0.1526* 44.1 43.2 42.8 43.7 Uruguay 1997 -0.1841* -0.2221* 43.4 43.2 41.2 43.1 1989 -0.0583* -0.1221* 46.3 47.0 44.3 46.3 Ecuador 1995 -0.1139* -0.1101* 45.5 45.9 43.1 45.3 1994 -0.1153* -0.1359* 44.0 46.4 43.6 44.5 2000 -0.0927* -0.1095* 43.2 45.2 41.7 43.8 1998 -0.1057* -0.1866* 44.7 49.3 45.9 46.1 RB de Venezuela El Salvador 1989 -0.1180* -0.1835* 42.9 41.8 40.5 42.4 1991 -0.0582* -0.2361* 48.1 48.2 43.2 47.7 1995 -0.1344* -0.2186* 42.6 42.2 40.8 42.3 1995 -0.1575* -0.3035* 48.0 47.3 42.7 47.4 1998 -0.1359* -0.1832* 41.8 42.2 40.8 41.8 2000 -0.1551* -0.1718* 44.3 46.6 42.7 44.8 Guatemala 2000 -0.1603* -0.2060* 51.0 50.3 46.7 50.7 *Significant at 5%. Source: Authors' calculations based on microdata from household surveys. 320 STATISTICAL APPENDIX TABLE A.35 Between- and within-group inequality in the dktribution of hourly wages, Thlul decomposiions Education Gender Age Urban/rural Country Between Within Between Within Between Within Between Within Argentina 1992 18.3 81.7 0.3 99.7 0.2 99.8 1996 22.6 77.4 0.0 100.0 0.5 99.5 2001 24.4 75.6 0.0 100.0 0.7 99.3 Bolivia Urban 1992 9.8 90.2 0.7 99.3 0.2 99.8 1996 16.4 83.6 1.5 98.5 1.7 98.3 National 1996 19.7 80.3 0.3 99.7 0.3 99.7 10.4 89.6 1999 15.9 84.1 0.0 100.0 0.3 99.7 11.0 89.0 Brazil 1990 35.0 65.0 1.2 98.8 0.6 99.4 8.3 91.7 1995 33.6 66.4 1.2 98.8 1.4 98.6 5.7 94.3 2001 32.3 67.7 0.6 99.4 1.7 98.3 4.2 95.8 Chile 1990 16.1 83.9 1.0 99.0 1.7 98.3 1.8 98.2 1996 24.1 75.9 0.4 99.6 2.1 97.9 3.7 96.3 2000 24.7 75.3 1.3 98.7 1.8 98.2 2.7 97.3 Colombia Bogota 1992 32.7 67.3 1.6 98.4 0.7 99.3 1996 36.9 63.1 1.5 98.5 1.7 98.3 National 1996 32.2 67.8 0.1 99.9 0.5 99.5 8.4 91.6 1999 28.7 71.3 0.0 100.0 0.8 99.2 5.1 94.9 Costa Rica 1990 28.0 72.0 0.1 99.9 0.9 99.1 9.8 90.2 1995 25.9 74.1 0.0 100.0 0.6 99.4 5.3 94.7 2000 24.2 75.8 0.3 99.7 0.8 99.2 7.2 92.8 Dominican Rep. 1995 17.0 83.0 0.2 99.8 1.0 99.0 3.5 96.5 Ecuador 1994 12.1 87.9 0.6 99.4 1.4 98.6 2.6 97.4 1998 17.5 82.5 0.7 99.3 0.5 99.5 6.9 93.1 El Salvador 1991 16.9 83.1 0.3 99.7 0.3 99.7 8.5 91.5 1995 29.6 70.4 3.1 96.9 0.3 99.7 8.9 91.1 2000 14.9 85.1 1.5 98.5 0.6 99.4 5.7 94.3 Guatemala 2000 32.8 67.2 1.4 98.6 0.9 99.1 9.6 90.4 Honduras 1990 26.6 73.4 0.1 99.9 0.4 99.6 11.1 88.9 1995 22.6 77.4 1.0 99.0 1.6 98.4 3.7 96.3 1999 16.6 83.4 0.4 99.6 0.6 99.4 4.4 95.6 Jamaica 1990 4.8 95.2 6.5 93.5 1.1 98.9 0.6 99.4 1999 12.4 87.6 0.3 99.7 1.4 98.6 5.0 95.0 Mexico 1992 24.2 75.8 0.5 99.5 1.2 98.8 5.5 94.5 1996 25.0 75.0 0.5 99.S 0.3 99.7 5.5 94.5 2000 31.0 69.0 1.2 98.8 2.6 97.4 6.0 94.0 Nicaragua 1993 11.1 88.9 0.3 99.7 1.9 98.1 5.6 94.4 1998 19.7 80.3 1.1 98.9 0.4 99.6 5.2 94.8 Panama 1991 27.6 72.4 0.2 99.8 2.2 97.8 1995 25.0 75.0 0.4 99.6 3.0 97.0 6.9 93.1 2000 18.9 81.1 0.0 100.0 1.4 98.6 5.6 94.4 Paraguay 1995 15.0 85.0 0.3 99.7 1.2 98.8 5.3 94.7 1999 16.2 83.8 0.4 99.6 0.4 99.6 5.6 94.4 Peru 1991 5.7 94.3 0.0 100.0 0.1 99.9 4.7 95.3 1994 13.9 86.1 0.8 99.2 1.2 98.8 5.3 94.7 2000 14.7 85.3 0.1 99.9 0.7 99.3 8.0 92.0 Trinidad & Tobago 1992 27.0 73.0 1.0 99.0 1.8 98.2 0.6 99.4 Uruguay 1989 14.5 85.5 3.0 97.0 0.9 99.1 1995 20.2 79.8 0.7 99.3 0.8 99.2 2000 17.3 82.7 0.3 99.7 1.5 98.5 RB de Venezuela 1989 21.7 78.3 0.9 99.1 1.4 98.6 4.0 96.0 1995 10.7 89.3 0.3 99.7 0.6 99.4 2.2 97.8 1998 14.1 85.9 0.8 99.2 0.8 99.2 2.0 98.0 Source: Authors' calculations based on microdata from household surveys. 321 INEQUALITY IN LATIN AMERICA TABLE A.36 Emphoyment and unemployment, by education bvfel X adults employed % adults unemployed Duration of unemployment (months) Education Gender Education Gender Education Low Medium High Female Male Total Low Medium High Female Male Total Low Medium High Total Country (i) (ii) (iD) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii) (xiv) (xv) (xvi) Argentina 1992 57.1 63.0 74.3 45.0 81.5 62.7 4.1 4.6 3.5 3.2 5.3 4.2 3.1 4.4 4.7 3.9 1996 52.6 57.7 70.9 41.6 74.4 57.3 12.1 12.1 8.6 10.5 12.7 11.6 6.7 8.7 10.0 8.0 2001 52.2 54.7 72.5 45.0 70.0 57.7 13.9 14.0 8.0 9.6 16.2 12.6 6.2 7.2 8.8 7.0 Bolivia Urban 1992 65.0 54.7 79.8 48.7 78.0 62.8 2.9 4.3 2.6 2.9 4.0 3.4 5.7 6.7 7.4 6.4 1996 74.3 62.2 68.8 58.6 80.2 68.6 1.9 3.9 3.8 2.9 3.3 3.1 3.7 3.3 5.1 3.9 National 1996 82.4 65.9 72.0 67.5 87.6 76.9 1.0 3.4 3.4 1.7 2.1 1.9 3.1 3.1 5.1 3.6 1999 79.1 62.3 71.1 62.5 84.1 72.8 1.5 5.3 3.0 2.9 2.8 2.9 4.1 3.3 10.1 4.7 Brazil 1990 63.9 71.4 85.0 46.6 87.6 66.5 2.2 3.4 1.4 1.5 3.3 2.4 3.6 4.5 S.0 3.9 1995 68.4 72.1 85.1 55.2 86.2 70.2 3.8 S.S 2.3 3.9 4.3 4.1 0.6 0.7 1.0 0.6 2001 64.9 69.1 82.8 53.8 82.1 67.4 5.9 8.8 3.7 6.9 6.2 6.6 0.7 0.7 0.7 0.7 Chile 1990 SO.S SS.0 69.5 34.1 78.0 SS.0 4.4 5.6 3.7 3.6 6.2 4.8 1996 54.9 60.6 69.0 39.7 81.2 59.8 3.0 4.3 2.6 3.1 4.0 3.5 1.9 2.7 2.6 2.4 2000 52.2 57.7 68.4 41.0 76.0 57.9 6.2 7.6 4.5 5.4 7.8 6.6 3.6 3.7 4.6 3.8 Colombia Bogota 1992 64.1 64.2 81.9 53.4 84.2 67.4 4.8 6.4 4.2 5.9 4.6 5.3 6.5 8.5 8.3 7.7 1996 62.2 65.4 76.2 52.1 81.4 65.7 7.1 7.5 5.6 7.2 6.8 7.0 5.6 6.6 8.7 6.4 National 1996 60.2 61.7 80.3 43.2 83.6 62.3 4.7 8.5 5.2 6.0 5.6 5.8 5.9 7.5 8.2 6.7 1999 59.1 56.5 75.6 43.3 78.5 60.0 8.1 15.2 10.2 11.1 10.1 10.6 7.5 9.2 10.1 8.6 Costa Rica 1990 57.0 62.6 72.2 34.3 87.5 60.3 2.0 3.2 2.0 1.7 2.9 2.3 3.3 4.8 5.6 4.1 1995 57.8 62.6 73.6 37.2 86.9 61.5 2.3 3.7 2.4 2.2 3.3 2.7 3.0 4.0 5.9 3.8 2000 57.6 64.2 76.1 39.5 86.1 62.2 3.3 3.1 1.8 2.6 3.5 3.0 3.9 5.1 6.8 4.6 Dominican Rep. 1995 55.9 SS.0 72.5 34.6 81.9 57.5 8.7 12.7 9.6 11.7 8.3 10.0 0.9 1.0 2.5 1.1 1997 57.8 58.6 72.4 36.3 83.6 59.5 8.9 13.6 12.6 12.9 8.0 10.5 1.1 1.4 1.5 1.3 Ecuador 1994 73.7 69.9 81.3 56.6 90.3 73.4 0.9 2.8 2.5 1.4 1.7 1.6 1.2 2.3 3.4 2.1 1998 74.2 69.7 81.6 58.2 90.3 73.9 1.9 3.6 2.5 2.1 2.9 2.5 1.9 1.9 3.5 2.1 El Salvador 1991 58.1 63.6 69.0 41.7 82.8 60.3 4.7 6.6 4.6 3.5 7.2 5.2 4.6 7.0 8.7 5.9 1995 60.7 63.5 76.3 45.2 83.4 62.5 3.9 6.3 3.2 2.6 6.8 4.5 3.3 5.6 7.8 4.3 2000 61.5 63.0 72.8 49.2 79.6 63.1 3.9 6.2 3.6 2.0 7.7 4.6 1.0 1.6 1.8 1.3 Guatemala 2000 66.1 70.5 83.7 46.2 92.2 67.7 0.8 2.5 1.2 0.8 1.3 1.1 1.2 2.1 3.8 1.7 Honduras 1990 61.6 63.2 66.8 36.5 89.6 62.0 2.5 8.3 7.3 2.9 4.0 3.4 4.3 8.1 7.7 5.7 1995 61.7 64.7 73.9 37.5 90.4 62.6 1.6 3.7 2.4 1.3 2.8 2.0 3.8 6.4 7.3 4.7 1999 66.4 69.9 75.0 47.6 90.5 67.4 2.1 3.2 2.8 1.6 3.1 2.3 2.2 4.7 3.9 2.9 Jamaica 1990 68.3 63.4 79.7 53.4 78.9 65.9 7.2 13.6 3.7 14.9 6.4 10.7 9.7 9.6 9.5 9.6 1 999 69.2 72.4 86.8 59.2 82.9 70.6 5.4 8.5 1.9 9.4 S.0 7.3 8.6 8.6 9.6 8.6 Mexico 1992 52.4 60.1 69.6 32.7 81.5 56.2 1.5 2.7 2.7 1.0 3.0 2.0 1996 59.6 64.0 73.8 42.2 85.1 62.5 1.8 3.7 3.7 1.5 3.8 2.6 2000 61.9 65.7 73.1 43.2 86.7 63.8 0.8 1.7 1.6 0.6 1.9 1.2 Nicaragua 1993 50.3 54.1 70.9 35.2 70.3 51.7 12.8 16.2 15.2 9.3 18.1 13.5 5.2 6.5 7.6 5.7 1998 56.1 54.7 63.0 35.8 76.4 56.1 5.9 9.4 9.6 5.6 7.9 6.8 2.0 2.0 2.0 2.0 Panama 1991 53.4 51.9 67.5 34.5 76.0 54.8 6.6 15.4 11.4 9.9 11.2 10.5 10.5 14.2 19.4 13.8 1995 56.9 58.8 74.0 39.5 81.1 60.2 7.0 13.2 9.1 10.0 9.6 9.8 8.7 10.4 13.2 10.2 2000 56.9 57.1 75.2 40.3 79.8 60.3 5.9 9.5 5.6 5.6 9.0 7.4 6.4 8.0 11.1 7.9 Paraguay 1995 79.2 76.9 88.8 66.1 92.9 79.4 1.5 3.7 2.0 1.8 2.2 2.0 5.3 S.0 7.1 5.3 1999 66.7 68.7 83.4 50.9 86.5 68.6 2.9 5.3 2.6 3.1 4.0 3.5 5.1 6.8 6.7 5.9 Peru Regions 1991 68.4 61.0 73.6 52.5 80.2 65.6 1.5 4.0 3.3 2.4 3.2 2.8 1994 69.3 59.7 71.2 51.3 81.3 65.4 1.6 4.1 4.1 2.3 3.9 3.0 National 1994 71.0 61.2 71.8 52.5 82.9 66.9 1.4 3.9 4.1 2.1 3.6 2.8 2000 75.1 67.4 73.6 60.0 84.1 71.6 1.7 4.5 4.8 3.1 3.9 3.5 1.4 2.3 2.3 2.2 Trinidad & Tobago 1992 47.0 53.9 77.1 36.2 69.1 52.4 13.3 14.7 4.7 10.1 16.3 13.2 Uruguay 1989 60.3 70.7 79.1 49.9 85.2 66.1 3.4 6.5 4.6 4.9 4.3 4.6 5.8 7.1 9.3 6.8 1995 59.9 70.8 80.9 53.0 83.3 67.2 S.S 8.2 3.5 6.9 5.8 6.3 5.9 6.5 8.2 6.4 2000 59.3 70.1 74.0 54.0 80.0 66.3 9.7 10.3 8.8 10.6 8.8 9.8 7.0 7.7 7.1 7.3 RB de Venezuela 1989 56.4 58.8 73.4 36.7 80.3 58.5 6.3 6.1 4.4 3.0 9.0 6.0 7.0 8.0 6.9 7.3 1995 60.0 62.7 75.5 41.1 82.4 61.8 6.0 7.6 6.4 6.0 7.3 6.6 9.5 12.3 17.2 11.0 1998 62.7 65.0 72.7 47.5 82.2 64.9 7.2 8.8 6.9 6.6 8.8 7.7 8.6 10.6 13.7 9.9 Note: Data for Brazil, 1995 and 2001: duration refers to average months since workers left their last jobs. Source: Authors' calculations based on microdata from household surveys. 322 STATISTICAL APPENDIX TABLE A.37 Distributions of income soumces, by equivalized household income quintile Capital income, rents, and profits Pensions Transfers exciuding pensions Country 1 2 3 4 5 1 2 3 4 5 1 2 3 4 s Argentina 1992 0.9 1.3 2.3 11.9 83.6 9.0 18.1 19.0 21.2 32.6 3.3 6.7 14.4 28.1 47.5 1996 1.0 0.6 5.1 9.1 84.2 5.0 11.2 17.1 21.4 45.4 5.8 8.7 12.5 29.7 43.4 2001 1.5 4.6 7.4 16.3 70.3 2.7 9.0 17.0 26.5 44.8 8.1 9.3 12.3 23.4 46.9 Bol ivia Urban 1992 4.9 8.0 14.5 19.0 53.8 3.7 4.8 6.7 15.7 69.0 1996 4.1 6.7 6.1 11.5 71.7 5.1 6.6 8.7 25.9 53.6 7.7 11.3 14.6 19.2 47.1 National 1996 0.4 2.6 7.2 9.4 80.4 0.4 4.9 10.1 20.1 64.6 2.1 8.6 14.5 21.0 53.8 1999 0.5 2.0 5.9 15.7 75.8 0.0 1.9 5.8 14.4 77.9 1.7 6.4 14.2 22.0 55.7 Brazil 1990 1.4 3.2 5.8 10.9 78.7 3.4 6.1 8.4 15.5 66.6 0.0 3.8 2.9 15.9 77.5 1995 0.5 1.5 3.8 9.6 84.6 2.6 7.3 11.4 16.2 62.6 7.6 10.6 13.4 19.9 48.5 2001 3.2 3.4 4.5 10.5 78.4 2.0 6.0 11.6 17.4 62.9 9.7 10.9 10.8 18.0 50.5 Chile 1990 0.0 0.0 1.1 9.7 89.1 4.7 9.3 15.3 25.2 45.5 23.4 22.5 20.9 18.0 15.2 1996 2.8 8.1 16.6 25.8 46.7 28.7 26.6 22.2 16.7 5.8 2000 2.4 7.1 14.6 23.8 52.1 33.0 29.0 20.8 12.3 4.9 Colombia 1999 2.0 3.2 5.5 13.5 75.7 0.4 3.4 7.8 16.6 71.9 5.4 8.5 10.3 15.9 59.9 Costa Rica 1995 5.4 8.6 13.0 19.8 53.2 10.7 10.4 10.7 17.3 51.0 2000 6.6 8.7 10.3 18.2 56.2 11.0 10.5 10.6 16.2 51.7 Dominican Rep. 1997 4.4 6.4 8.8 16.4 63.9 3.6 6.5 6.0 9.8 74.1 6.2 9.7 12.3 23.3 48.5 Ecuador 1994 2.4 4.1 11.5 17.1 64.9 1998 1.2 3.1 4.5 8.7 82.5 El Salvador 1991 1.1 2.5 5.3 9.9 81.2 1.6 4.2 9.4 21.9 62.8 4.7 8.7 13.3 20.7 52.5 2000 2.1 3.2 4.0 9.9 80.8 1.4 6.3 11.4 17.9 63.0 11.1 13.8 14.6 17.1 43.4 Guatemala 2000 0.1 0.2 1.4 2.3 95.9 0.7 2.3 5.9 12.9 78.2 2.1 5.4 11.4 16.0 65.0 Mexico 1992 1.8 3.8 5.8 17.2 71.4 1.5 11.6 13.4 21.0 52.5 5.0 11.4 14.5 15.3 53.8 1996 1.4 4.0 5.3 9.1 80.2 1.8 8.5 14.8 24.0 50.9 3.4 11.5 13.1 22.6 49.4 2000 1.6 1.8 6.0 7.1 83.4 1.3 5.9 9.9 12.3 70.5 10.5 11.0 14.5 18.9 45.0l Nicaragual 1993 1.8 13.3 4.5 6.0 74.4 8.9 12.9 17.2 19.5 41.4 3.1 4.9 9.5 15.3 67.2l 1998 0.2 0.8 1.3 4.1 93.6 3.9 3.2 10.7 19.0 63.2 3.7 7.8 10.4 16.5 61.5l Panama 1991 4.7 5.4 7.3 13.2 69.4 0.7 4.6 9.5 21.1 64.1 18.4 20.7 17.7 17.4 25.9 1995 2.3 3.4 7.3 12.7 74.3 0.8 4.4 9.5 20.5 64.8 8.2 10.6 13.3 21.5 46.4l 2000 0.7 2.4 8.4 9.1 79.3 0.7 3.6 7.8 19.0 69.0 12.1 13.7 15.3 19.9 38.9 Paraguay 1995 0.1 0.9 2.7 8.2 88.1 0.2 10.6 7.5 16.5 65.3 5.1 7.5 17.4 18.8 51.2 1999 1.1 1.8 3.9 11.4 81.7 0.5 1.9 12.0 19.2 66.4 6.2 14.4 17.8 17.0 44.7 Peru 1994 0.8 1.8 5.8 5.8 85.8 1.8 5.8 15.7 22.1 54.6 5.4 10.8 16.2 22.3 45.4 2000 0.8 2.9 5.6 12.3 78.5 0.4 3.5 9.3 22.6 64.1 4.7 9.5 13.6 18.8 53.4 Uruguay 1995 1.1 2.6 6.2 13.1 76.9 5.7 13.1 17.7 22.6 40.9 12.3 15.1 18.6 22.9 31.1 2000 0.6 2.2 5.2 10.8 81.2 3.7 9.6 15.5 23.3 47.9 13.7 16.3 20.1 23.0 26.8 RB de Venezuela 1995 5.0 5.8 10.9 11.5 66.8 10.5 11.7 15.0 17.7 45.1 1998 3.6 7.4 8.9 12.1 68.1 8.7 10.8 15.8 22.4 42.3 Source: Authors' calculations based on microdata from household surveys. 32 3 INEQUALITY IN LATIN AMERICA TABLE A.38 TABLE A.39 Distribution of operational holdings of agnicultural land Distribution of operational holdings of agricultural land in Latin worldwide: Gin! coefficients America: Gini coefficients Deininger & Olinto UNDP Deininger & Olinto UNDP Region (i) (ii) 1950-1979 1980-1994 1981 Latin America 0.81 0.74 Middle East and North Africa 0.67 0.56 Antigua 074 North America 0.64 Antiua 0.74 Sub-Saharan Africa 0.61 0.51 Bahamas 0.86 0.87 Western Europe 0.57 Barbamas 0.90 0.87 East Asia & South Asia 0.56 0.52 Belize 0.72 0.71 Bolivia 0.77 Note: Column (i) shows averages for the 1950-1994 period; Brazil 0.83 0.85 0.86 Column (ii) shows values around 1981. Colombia 0.85 0.77 0.70 Source: Deininger and Olinto 2000 and UNDP 1993. Costa Rica 0.81 Chile 0.64 Dominican Rep. 0.80 0.70 Ecuador 0.86 0.69 El Salvador 0.83 Grenada 0.78 0.74 0.69 Guatemala 0.86 Guyana 0.68 Honduras 0.75 0.64 Jamaica 0.81 0.81 Mexico 0.59 Nicaragua 0.80 Panama 0.71 0.87 0.84 Paraguay 0.86 0.78 0.94 Peru 0.94 0.61 Puerto Rico 0.73 0.77 Suriname 0.73 Trinidad & Tobago 0.68 Uruguay 0.82 0.80 0.84 RB de Venezuela 0.92 Note: The values for each country correspond in column (i) to the first value in the 1950-1979 period and in column (ii) to the most recent observation in the 1980-1994 period. Source: Authors' calculations based on Deininger and Olinto 2002 and UNDP 1993. 324 l | | | | R | | | u u u u - - -~~~~~~ST TITIC L PPE DI TABLE A.40 Housing ownership, number of rooms, and persons per room, by quinitile Ownership of housing Number of rooms Persons per room Country 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean Argentina 1992 0.67 0.72 0.74 0.73 0.71 0.72 2.6 2.6 2.8 3.0 3.2 2.9 2.0 1.4 1.4 1.2 1.0 1.4 1996 0.65 0.71 0.71 0.71 0.74 0.71 2.7 2.7 2.8 3.0 3.2 2.9 2.1 1.6 1.4 1.2 0.9 1.4 2001 0.65 0.71 0.70 0.73 0.75 0.72 2.5 2.6 2.7 2.9 3.3 2.9 2.3 1.7 1.4 1.1 0.9 1.4 Bolivia Urban 1992 1.6 1.8 1.8 2.0 2.2 1.9 3.9 3.6 3.3 2.8 2.1 3.0 1996 0.50 0.50 0.47 0.50 0.56 0.51 1.6 1.8 1.8 2.0 2.3 2.0 3.8 3.4 3.1 2.6 1.9 2.8 National 1996 0.89 0.71 0.60 0.53 0.57 0.66 1.3 1.4 1.7 1.7 2.1 1.7 3.8 3.8 3.4 2.9 2.2 3.1 1999 0.88 0.70 0.55 0.54 0.61 0.65 1.4 1.6 1.7 1.8 2.0 1.7 3.5 3.3 3.2 2.7 2.0 2.9 Brazil 1990 0.69 0.68 0.66 0.68 0.70 0.68 2.1 2.0 2.0 2.1 2.1 2.1 2.6 2.4 2.3 2.1 1.8 2.2 1995 0.62 0.65 0.67 0.68 0.71 0.67 2.1 2.0 2.0 2.0 2.1 2.0 2.7 2.3 2.1 2.0 1.7 2.1 2001 0.62 0.65 0.68 0.69 0.73 0.68 2.0 2.0 1.8 1.9 2.0 1.9 2.6 2.2 1.9 1.8 1.6 2.0 Chile 1996 0.53 0.56 0.60 0.64 0.56 0.58 2.1 2.3 2.5 2.6 3.0 2.5 2.4 2.1 1.8 1.6 1.2 1.8 2000 0.61 0.63 0.67 0.68 0.65 0.65 2.3 2.4 2.5 2.7 3.0 2.6 2.2 2.0 1.7 1.5 1.1 1.7 Colombia 1996 0.73 0.63 0.58 0.58 0.63 0.63 2.8 2.9 3.1 3.2 3.8 3.2 2.1 2.0 1.7 1.5 1.1 1.6 1999 0.71 0.60 0.55 0.57 0.60 0.60 2.8 2.9 3.0 3.3 3.8 3.2 2.1 2.0 1.7 1.5 1.0 1.6 Dominican Rep. 1995 0.82 0.79 0.75 0.75 0.68 0.76 1997 0.89 0.83 0.79 0.71 0.65 0.76 2.9 3.2 3.2 3.4 3.9 3.4 1.8 1.6 1.5 1.3 1.0 1.4 Ecuador 1994 0.77 0.73 0.66 0.66 0.63 0.68 1.6 1.8 2.0 2.0 2.4 2.0 3.6 3.6 3.2 2.7 1.9 2.9 1998 0.79 0.68 0.64 0.62 0.59 0.66 1.7 1.9 1.9 2.1 2.4 2.0 3.2 3.3 3.0 2.6 1.8 2.7 El Salvador 1991 0.61 0.55 0.58 0.60 0.67 0.61 1.2 1.3 1.4 1.6 2.1 1.6 5.0 4.6 4.1 3.3 2.2 3.7 1995 0.65 0.60 0.62 0.64 0.71 0.65 1.3 1.3 1.5 1.7 2.2 1.6 4.6 4.4 3.7 3.0 2.1 3.4 2000 0.69 0.65 0.65 0.63 0.70 0.66 1.8 1.8 1.9 2.0 2.3 2.0 3.7 3.3 3.1 2.8 1.9 2.8 Guatemala 2000 0.82 0.72 0.69 0.64 0.66 0.70 1.4 1.4 1.6 1.7 2.2 1.7 4.8 4.5 4.0 3.4 2.3 3.7 Honduras 1990 0.84 0.76 0.71 0.64 0.62 0.71 1.2 1.2 1.3 1.5 2.0 1.5 5.7 5.4 4.9 4.2 2.7 4.4 1995 0.83 0.74 0.69 0.69 0.69 0.72 1.3 1.4 1.5 1.6 2.1 1.6 5.3 4.8 4.2 3.7 2.4 4.0 1999 0.80 0.76 0.65 0.62 0.63 0.69 1.4 1.6 1.7 1.8 2.2 1.8 4.8 4.2 3.8 3.3 2.2 3.6 Jamaica 1990 0.80 0.76 0.73 0.71 0.55 0.69 1996 0.64 0.65 0.65 0.62 0.53 0.60 2.9 2.8 2.9 2.8 2.9 2.9 2.3 1.9 1.8 1.S 1.0 1.6 1999 0.64 0.54 0.64 0.57 0.49 0.56 2.6 2.7 2.9 2.8 2.6 2.7 2.4 1.8 1.6 1.5 1.1 1.6 Mexico 1992 0.51 0.56 0.57 0.60 0.68 0.59 1.5 1.7 1.8 2.0 2.1 1.9 4.2 3.6 3.1 2.6 2.0 3.0 1996 0.48 0.56 0.63 0.66 0.67 0.61 1.6 1.7 1.9 2.0 2.1 1.9 4.0 3.4 2.9 2.4 1.9 2.8 2000 0.48 0.63 0.63 0.66 0.69 0.62 1.7 1.8 1.9 2.0 2.2 1.9 3.5 2.9 2.5 2.2 1.7 2.5 Paraguay 1995 0.89 0.82 0.77 0.75 0.67 0.77 1.8 1.8 1.9 2.1 2.3 2.0 3.8 3.2 2.8 2.4 1.7 2.7 1999 0.89 0.84 0.78 0.81 0.75 0.81 2.0 1.9 2.1 2.2 2.4 2.1 3.4 2.9 2.5 2.2 1.6 2.4 Peru Regions 1991 0.81 0.79 0.73 0.71 0.66 0.74 2.9 3.3 3.4 3.7 3.7 3.4 2.4 2.1 2.1 1.9 1.6 2.0 1994 0.78 0.71 0.68 0.70 0.70 0.71 2.8 3.1 3.4 3.7 4.5 3.6 2.7 2.2 1.9 1.6 1.2 1.9 National 1994 0.77 0.71 0.67 0.69 0.69 0.70 2.7 3.0 3.3 3.6 4.4 3.4 2.6 2.4 2.0 1.7 1.2 2.0 2000 0.84 0.74 0.73 0.72 0.73 0.75 2.7 3.0 3.2 3.7 4.4 3.5 2.4 2.2 2.0 1.7 1.2 1.9 Trinidad & Tobago 1992 0.26 0.34 0.38 0.39 0.51 0.39 2.4 2.5 2.6 2.7 2.7 2.6 2.6 2.3 2.0 1.7 1.2 1.9 Uruguay 1989 0.62 0.65 0.68 0.67 0.73 0.67 1.9 1.9 2.0 2.1 2.2 2.0 2.3 1.8 1.8 1.6 1.4 1.7 1995 0.64 0.68 0.70 0.71 0.75 0.70 2.0 2.0 2.0 2.0 2.0 2.0 2.2 1.8 1.7 1.6 1.4 1.7 2000 0.57 0.65 0.68 0.70 0.76 0.68 2.0 2.0 2.0 2.0 2.0 2.0 2.4 1.9 1.6 1.5 1.3 1.7 RB de Venezuela 1989 0.83 0.81 0.75 0.73 0.69 0.75 2.5 2.6 2.7 2.8 2.9 2.7 3.1 2.7 2.4 2.0 1.5 2.2 1995 0.86 0.85 0.82 0.80 0.79 0.82 2.3 2.3 2.4 2.4 2.4 2.4 2.8 2.7 2.3 2.0 1.7 2.3 1998 0.84 0.84 0.83 0.81 0.80 0.82 2.2 2.2 2.3 2.4 2.3 2.3 2.8 2.6 2.4 2.1 1.8 2.3 Note: In most cases, "rooms" refer to the number of rooms used only by the household, excluding the kitchen and bathrooms. Source: Authors' calculations based on microdata from household surveys. 325 INEQUALITY IN LATIN AMERICA TABLE A.41 Number of children under age 12 per household, by parental Income and education Parental income quintile Parental education 1 2 3 4 5 Low Medium High Country (i) (ii) (iii) (iv) (v) Mean (vi) (vii) (viii) Argentina 1992 1.80 1.70 1.68 1.44 1.32 1.59 1.87 1.53 1.15 1996 1.80 1.76 1.55 1.28 1.20 1.52 1.90 1.34 1.09 2001 1.89 1.68 1.35 1.25 1.04 1.44 1.96 1.31 0.93 Bolivia Urban 1992 2.11 2.27 2.18 2.10 1.94 2.12 2.37 2.08 1.65 1996 1.89 2.13 1.91 2.00 1.63 1.91 2.15 1.92 1.51 National 1996 2.09 2.13 2.26 2.09 1.93 2.10 2.27 2.04 1.56 1999 2.61 2.18 2.03 1.86 1.61 2.06 2.47 1.76 1.39 Brazil 1990 2.13 1.90 1.72 1.55 1.42 1.74 1.90 1.36 1.23 1995 1.89 1.66 1.47 1.32 1.18 1.50 1.65 1.18 0.99 2001 1.63 1.41 1.27 1.13 0.97 1.28 1.43 1.04 0.83 Chile 1990 1.59 1.59 1.51 1.38 1.43 1.50 1.55 1.50 1.36 1996 1.55 1.52 1.46 1.35 1.32 1.44 1.48 1.48 1.28 2000 1.44 1.41 1.37 1.24 1.30 1.35 1.45 1.35 1.24 Colombia Bogota 1992 1.67 1.56 1.35 1.14 1.20 1.38 1.61 1.36 1.03 1996 1.28 1.29 1.36 1.19 1.11 1.25 1.43 1.20 0.94 National 1996 1.97 1.67 1.61 1.50 1.29 1.61 1.76 1.40 1.14 1999 1.77 1.70 1.56 1.37 1.24 1.53 1.69 1.35 1.05 Costa Rica 1990 2.06 2.00 1.94 1.72 1.88 1.92 2.02 1.75 1.64 1995 2.05 1.77 1.80 1.67 1.51 1.76 1.93 1.60 1.40 2000 1.86 1.68 1.64 1.57 1.30 1.61 1.75 1.49 1.23 Dominican Rep. 1995 1.51 1.64 1.69 1.62 1.58 1.61 1.61 1.51 1.56 1997 1.54 1.61 1.57 1.68 1.53 1.59 1.64 1.52 1.40 Ecuador 1994 2.30 2.05 2.12 1.86 1.56 1.98 2.29 1.56 1.39 1998 2.07 2.21 1.86 1.77 1.55 1.89 2.17 1.63 1.48 El Salvador 1991 2.43 2.32 2.20 2.05 1.82 2.16 2.40 1.79 1.42 1995 2.28 2.15 1.86 1.82 1.64 1.95 2.17 1.60 1.47 2000 1.90 1.81 1.83 1.76 1.65 1.82 2.05 1.56 1.36 Guatemala 2000 2.98 2.97 2.81 2.41 1.95 2.62 2.87 2.04 1.64 Honduras 1990 3.00 2.85 2.68 2.53 2.28 2.67 2.85 1.90 1.77 1995 2.75 2.71 2.40 2.46 1.94 2.45 2.64 1.91 1.70 1999 2.70 2.47 2.19 2.08 1.80 2.25 2.50 1.72 1.46 Jamaica 1990 1.62 1.24 1.10 1.41 1.53 1.38 1.36 1.40 1.15 1996 1.45 1.33 1.05 1.29 1.40 1.31 1.27 1.35 0.86 1999 1.18 1.24 1.10 1.19 1.04 1.15 1.20 1.17 0.86 Mexico 1992 2.59 2.22 2.14 1.93 1.75 2.13 2.37 1.77 1.58 1996 2.32 2.14 1.85 1.67 1.51 1.90 2.19 1.69 1.35 2000 2.15 1.81 1.62 1.63 1.40 1.72 1.94 1.55 1.30 Nicaragua 1993 3.11 2.60 2.66 2.39 2.16 2.58 2.85 1.93 1.71 1998 2.61 2.60 2.44 2.28 2.05 2.40 2.64 1.87 1.56 Panama 1991 1.86 1.70 1.73 1.51 1.38 1.64 1.89 1.52 1.32 1995 1.84 1.65 1.61 1.30 1.24 1.53 1.81 1.46 1.20 2000 1.76 1.55 1.48 1.23 1.17 1.44 1.67 1.40 0.98 Paraguay 1995 2.70 2.60 2.11 2.08 1.95 2.29 2.50 1.81 1.61 1999 2.48 2.27 2.16 2.04 1.72 2.14 2.41 1.86 1.52 Peru Regions 1991 2.30 1.97 2.04 2.06 1.76 2.02 2.29 2.00 1.65 1994 2.57 2.34 2.10 1.94 1.65 2.12 2.68 1.99 1.49 National 1994 2.59 2.43 2.15 2.09 1.71 2.19 2.70 2.06 1.54 2000 2.48 2.25 1.93 2.01 1.68 2.07 2.45 1.97 1.63 Trinidad & Tobago 1992 1.96 1.52 1.59 1.62 1.67 1.67 1.84 1.63 1.42 Uruguay 1989 1.48 1.47 1.46 1.26 1.42 1.42 1.53 1.33 1.32 1995 1.45 1.38 1.28 1.28 1.30 1.34 1.47 1.30 1.14 2000 1.33 1.33 1.26 1.11 1.11 1.23 1.46 1.15 0.92 RB de Venezuela 1989 2.35 2.23 2.10 1.84 1.65 2.03 2.28 1.73 1.38 1995 2.13 2.22 1.96 1.72 1.47 1.90 2.16 1.63 1.08 1998 1.99 1.95 1.80 1.74 1.50 1.80 2.03 1.65 1.27 Source: Authors' calculations based on microdata from household surveys. 326 STATISTICAL APPENDIX TABLE A.42 Household size, by income quintile and education of household head Equivalized income quintile Education of household head 1 2 3 4 5 Mean Low Medium High Country (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) Argentina 1992 4.39 3.36 3.64 3.37 2.89 3.46 3.50 3.58 3.12 1996 4.75 3.86 3.45 3.18 2.73 3.47 3.64 3.42 3.06 2001 5.00 3.89 3.46 2.96 2.63 3.42 3.63 3.35 2.97 Bolivia Urban 1992 5.15 5.11 4.67 4.36 3.81 4.56 4.75 4.44 4.23 1996 4.74 4.67 4.42 4.07 3.61 4.26 4.42 4.17 3.99 National 1996 4.38 4.43 4.53 4.15 3.61 4.19 4.22 4.19 3.96 1999 4.47 4.74 4.64 4.19 3.63 4.29 4.44 4.19 3.78 Brazil 1990 4.95 4.43 4.13 3.96 3.44 4.11 4.25 3.56 3.41 1995 5.00 4.14 3.81 3.66 3.24 3.89 4.01 3.46 3.26 2001 4.80 4.07 3.41 3.33 3.00 3.62 3.74 3.33 3.06 Chile 1990 4.66 4.37 4.03 3.89 3.53 4.06 4.17 3.99 3.71 1996 4.42 4.30 3.95 3.82 3.43 3.95 4.02 3.95 3.70 2000 4.52 4.28 3.90 3.65 3.24 3.86 3.96 3.87 3.57 Colombia Bogota 1992 4.79 4.33 4.29 3.74 3.36 4.04 4.34 3.83 3.45 1996 4.46 4.22 3.80 3.75 3.38 3.88 4.00 3.74 3.53 National 1996 4.87 4.74 4.43 4.02 3.52 4.25 4.36 3.90 3.67 1999 4.87 4.82 4.39 4.09 3.37 4.23 4.39 3.83 3.47 Costa Rica 1990 4.75 4.71 4.48 4.32 3.71 4.36 4.47 4.06 3.98 1995 4.56 4.52 4.31 4.02 3.51 4.15 4.25 3.94 3.78 2000 4.31 4.40 4.24 4.05 3.54 4.08 4.09 4.16 3.69 Dominican Rep. 1995 5.07 4.64 4.43 4.17 3.76 4.38 4.31 3.89 4.10 1997 4.77 4.55 4.31 3.96 3.60 4.19 4.20 3.90 4.08 Ecuador 1994 5.10 5.40 5.10 4.78 3.88 4.79 5.05 4.08 3.94 1998 4.72 5.23 4.80 4.47 3.71 4.53 4.75 4.09 4.01 El Salvador 1991 5.60 5.33 5.03 4.54 3.88 4.79 4.95 4.24 3.92 1995 5.31 5.26 4.83 4.45 3.86 4.68 4.84 4.12 3.92 2000 5.11 4.77 4.62 4.27 3.53 4.39 4.53 3.93 3.85 Guatemala 2000 5.90 5.76 5.44 5.02 4.17 5.18 5.36 4.27 4.18 Honduras 1990 6.19 6.07 5.77 5.39 4.36 5.47 S.S0 4.32 4.17 1995 6.20 5.95 5.57 5.12 4.34 5.35 5.36 4.45 4.19 1999 5.80 5.59 5.39 4.99 4.14 5.11 5.18 4.18 4.13 Jamaica 1990 6.07 5.47 4.64 4.50 2.92 4.44 4.17 3.82 3.48 1996 5.67 4.58 4.56 3.85 2.58 3.97 3.59 3.79 2.95 1999 5.14 4.30 4.34 3.72 2.47 3.76 3.59 3.27 3.00 Mexico 1992 5.75 5.38 4.93 4.54 3.83 4.79 5.04 4.06 3.99 1996 5.67 5.18 4.84 4.25 3.62 4.60 4.85 4.07 3.78 2000 5.11 4.68 4.27 4.07 3.43 4.24 4.37 3.91 3.67 Nicaragua 1993 6.71 6.28 5.97 5.43 4.68 5.72 5.79 4.61 4.52 1998 5.97 5.75 5.64 5.10 4.71 5.39 5.56 4.74 4.38 Panama 1991 4.62 4.82 4.74 4.06 3.42 4.26 4.42 4.06 3.74 1995 4.59 4.65 4.18 3.88 3.26 4.03 4.16 3.95 3.57 2000 4.36 4.34 4.10 3.63 3.11 3.84 4.03 3.79 3.37 Paraguay 1995 5.82 5.11 4.75 4.41 3.76 4.68 4.83 4.18 4.02 1999 5.73 4.99 4.66 4.32 3.69 4.58 4.80 4.12 4.06 Peru Regions 1991 5.42 5.48 5.50 5.35 4.73 5.28 5.39 4.91 4.74 1994 5.82 5.49 5.13 5.89 4.35 5.09 5.45 4.88 4.50 National 1994 5.77 5.67 5.22 4.91 4.34 5.13 5.46 4.91 4.49 2000 5.24 5.21 5.17 5.04 4.38 4.98 5.09 4.93 4.69 Trinidad & Tobago 1992 5.54 5.10 4.78 4.26 3.09 4.37 4.20 3.87 3.41 Uruguay 1989 3.89 3.35 3.25 3.15 2.86 3.27 3.19 3.45 3.35 1995 4.11 3.41 3.18 3.03 2.72 3.23 3.13 3.46 3.20 2000 4.49 3.53 3.13 2.81 2.54 3.17 3.14 3.33 2.95 RB de Venezuela 1989 6.45 5.87 5.36 4.96 3.95 5.16 5.30 4.43 4.07 1995 S.S9 5.52 5.03 4.65 3.93 4.86 5.11 4.37 3.81 1998 5.37 5.22 4.88 4.61 3.86 4.72 4.98 4.43 3.83 Source: Authors' calculations based on microdata from household surveys. 327 INEQUALITY IN LATIN AMERICA TABLE A.43 Conelations between spouses: education, wages, and hours worked Hours worked Hours worked Years of Hourly Years of Hourly education wages All Workers education wages All Workers Country (i) (ii) (iii) (iv) Country (i) (ii) (iii) (iv) Argentina Honduras 1992 0.6427* 0.5061* 0.0648* 0.0562* 1990 0.7355* 0.3095* 0.0001 0.1527* 1996 0.6525* 0.4600* 0.0793* 0.0715* 1995 0.7184* 0.2364* 0.0198 0.1472* 2001 0.6623* 0.4709* 0.0815* 0.0946* 1999 0.7298* 0.0787* 0.0303 0.1262* Bolivia Jamaica Urban 1990 0.7121* 0.7662* 0.0273 0.3356* 1992 0.7516* 0.2248* 0.0638* 0.2122* 1996 0.6898* 0.5984* 0.1581* 0.0983 1996 0.6749* 0.3529* 0.0805* 0.2027* 1999 0.7125* 0.4922* 0.1412* 0.1981* National Mexico 1996 0.7584* 0.1698* 0.1264* 0.2265* 1992 0.7017* 0.3234* -0.0273* 0.1329* 1999 0.8119* 0.1335* 0.1369* 0.2769* 1996 0.6985* 0.2268* 0.0325* 0.1201* Brazil 2000 0.7366* 0.3241 * 0.0761 * 0.0849* 1990 0.7452* 0.4493* 0.0985* 0.1594* Nicaragua 1995 0.7300* 0.5077* 0.1594* 0.1467* 1993 0.7239* 0.2155* -0.0227 0.2060* 2001 0.7151* 0.3466* 0.1386* 0.1485* 1998 0.6623* 0.6166* 0.0419* 0.1893* Chile Panama 1990 0.7390* 0.4770* 0.0970* 0.3341* 1991 0.7458* 0.4696* 0.2068* 0.2262* 1996 0.7392* 0.4672* 0.0981* 0.3479* 1995 0.7262* 0.3263* 0.1739* 0.1734* 2000 0.7388* 0.4579* 0.1081* 0.3130* 2000 0.6585* 0.5420* 0.1277* 0.1401* Colombia Paraguay Bogota 1995 0.7303* 0.2241* 0.0662* 0.2433* 1992 0.7783* 0.6230* 0.1713* 0.3009* 1999 0.7280* 0.3111* 0.1100* 0.2183* 1996 0.7596* 0.5002* 0.1522* 0.1824* Peru National Regions 1996 0.7538* 0.4354* 0.0817* 0.2006* 1991 0.7370* 0.2337* 0.1889* 0.2101* 1999 0.7384* 0.2066* 0.0763* 0.2034* 1994 0.7162* 0.4383* 0.1571* 0.1938* Costa Rica National 1990 0.6735* 0.3569* 0.0959* 0.1403* 1994 0.7073* 0.3079* 0.1260* 0.1770* 1995 0.6436* 0.4362* 0.0862* 0.1793* 2000 0.7178* 0.3858* 0.1053* 0.1780* 2000 0.6406* 0.3951* 0.0745* 0.1004* Trinidad & Tobago Dominican Rep. 1992 0.5856* 0.5178* 0.1514* 0.3559* 1995 0.6938* 0.3940* 0.0597* 0.2368* Uruguay 1997 0.6680* 0.4168* 0.0333 0.1432* 1989 0.5287* 0.4347* 0.2316* 0.1469* Ecuador 1995 0.6193* 0.3601* 0.2322* 0.1480* 1994 0.7710* 0.2378* 0.1356* 0.1883* 2000 0.5953* 0.3073* 0.2377* 0.1677* 1998 0.7689* 0.2863* 0.1889* 0.2689* RB de Venezuela El Salvador 1989 0.7197* 0.3516* 0.0376* 0.1552* 1991 0.6977* 0.2501* 0.0441* 0.1388* 1995 0.7036* 0.2386* 0.0147 0.1232* 1995 0.7169* 0.4435* 0.0581* 0.2261* 1998 0.6921* 0.1353* 0.0291* 0.1267* 2000 0.7118* 0.2896* 0.0673* 0.1681* Guatemala 2000 0.7419* 0.3163* 0.0665* 0.1546* *Significant at 5%. Source: Authors' calculations based on microdata from household surveys. 328 STATISTICAL APPENDIX TABLE A.44 Share of children ages 10-14 who work (percent) Equivalized household income quintile Equivalized household income quintile Country 1 2 3 4 5 Average Country 1 2 3 4 5 Average Argentina Guatemala 1992 1.6 1.5 3.5 1.9 0.3 1.8 2000 36.2 27.4 28.4 27.5 15.4 27.7 1996 0.9 1.6 0.9 2.1 0.1 1.1 Honduras 2001 0.9 0.4 0.5 0.1 0.1 0.5 1990 16.1 14.9 13.1 9.9 11.2 13.2 Bolivia 1995 14.3 13.4 9.9 7.6 9.4 11.2 Urban 1999 16.2 15.7 17.3 14.8 12.5 15.5 1992 7.2 9.0 8.9 10.2 9.4 8.8 Jamaica 1996 8.7 10.4 13.1 13.7 8.5 10.9 1990 0.7 2.3 1.8 2.3 1.3 1.7 National 1996 0.8 0.0 0.9 0.0 1.5 0.6 1996 61.7 26.8 18.6 17.5 12.1 29.3 1999 0.0 1.5 0.0 1.2 3.6 1.0 1999 60.3 32.0 16.9 20.2 10.0 29.1 Mexico Brazil 1992 9.0 5.9 4.1 4.2 2.5 5.6 1990 25.0 21.3 15.0 11.6 7.2 17.3 1996 14.6 10.2 7.5 4.7 2.9 8.9 1995 30.5 20.8 15.6 11.4 6.8 18.8 2000 9.3 7.1 4.5 4.7 4.8 6.5 2001 19.1 11.5 9.1 6.8 4.4 11.6 Nicaragua Chile 1993 9.2 10.5 8.4 7.5 9.4 9.0 1990 1.2 0.7 2.1 1.3 0.9 1.2 1998 18.1 11.1 10.8 10.3 9.4 12.2 1996 1.2 1.9 2.4 1.0 1.6 1.6 Panama 2000 1.2 1.9 0.6 1.0 0.3 1.1 1995 5.2 2.0 1.8 1.5 3.1 2.9 Colombia 2000 2.9 2.0 0.6 0.8 1.5 1.8 Bogota Paraguay 1992 6.3 4.2 3.5 0.8 6.1 4.4 1995 49.9 35.4 30.6 26.2 20.0 34.3 1996 4.0 2.7 5.1 1.2 7.0 4.0 1999 19.5 12.7 9.6 7.0 9.6 12.3 National Peru 1996 16.0 10.8 11.6 8.4 7.7 11.3 Regions 1999 10.7 8.0 7.6 5.8 5.8 7.9 1991 24.2 13.0 9.2 7.7 6.9 12.8 Costa Rica 1994 36.4 18.2 9.1 8.7 6.4 17.5 1990 8.0 6.5 6.6 7.1 1.2 6.1 National 1995 10.6 11.3 7.5 6.2 3.0 8.3 1994 38.8 21.1 12.5 9.7 8.5 20.0 2000 4.3 4.0 4.3 2.5 1.4 3.5 2000 52.7 26.1 19.2 14.5 9.0 26.8 Dominican Rep. Uruguay 1995 5.0 3.4 5.6 5.5 3.5 4.6 1989 9.2 9.8 4.3 5.9 2.0 6.7 1997 7.3 3.2 3.3 5.2 4.7 4.9 1995 9.7 12.4 7.8 5.5 2.2 8.3 Ecuador 2000 7.2 3.8 1.3 2.8 1.4 3.9 1994 35.1 26.1 31.9 30.0 21.3 29.4 RB de Venezuela 1998 46.3 32.7 33.5 28.7 22.7 33.8 1989 5.5 3.8 3.4 2.9 1.2 3.7 El Salvador 1995 3.6 3.6 3.7 2.4 2.7 3.3 1991 14.3 16.8 16.6 15.7 12.5 15.3 1998 3.6 3.4 4.2 3.5 1.7 3.4 1995 12.7 11.3 11.4 9.6 6.4 10.7 2000 10.5 8.6 8.9 7.3 5.0 8.4 Source: Authors' calculations based on microdata from household surveys. 329 INEQUALITY IN LATIN AMERICA TABLE A.45 Wages for prine-age men and household per capita income: ratio of nonwhites to whites Wages, prime-age men Per capita income Total Urban Total Urban Count,y (i) (ii) (iii) (iv) Bolivia 1999 0.534 0.686 0.478 0.637 Brazil 1990 0.515 0.540 0.454 0.480 1995 0.518 0.545 0.446 0.470 2001 0.511 0.527 0.457 0.474 Guatemala 2000 0.524 0.530 0.448 0.512 Peru 2000 0.438 0.485 0.415 0.497 Trinidad & Tobago 1992 0.241 0.223 0.358 0.414 Source: Authors' calculations based on microdata from household surveys. 330 STATISTICAL APPENDIX TABLE A.46 Interpersonal and institutional trust in Latin America, by age, education, subjective income, and inequality perception, 1996-2001 Interpersonal Institutional trust trust Age 18-30 0.423 0.196 31-55 0.426 0.202 56-65 0.446 0.207 more than 65 0.466 0.225 Education No school 0.436 0.221 Some primary 0.437 0.210 Some secondary 0.403 0.158 Some tertiary 0.400 0.181 Subjective income Great difficulties to finance living costs 0.394 0.168 Not enough to finance living costs 0.417 0.183 Finance living cost 0.444 0.217 Finance living cost, and save 0.469 0.261 Income distribution Very unfair 0.390 0.159 perception Unfair 0.435 0.188 Fair 0.510 0.362 Very fair 0.508 0.422 Source: Authors' calculations based on microdata from Latinobar6metro. 331 TABLE A.47 School enrollment rates, by age and income quintile 3 to 5 years old 6 to 12 years old 13 to 17 years old 18 to 23 years old Country 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean Argentina 1992 0.22 0.34 0.29 0.43 0.51 0.34 0.97 0.99 0.98 0.99 0.99 0.98 0.73 0.74 0.77 0.81 0.94 0.78 0.33 0.34 0.35 0.44 0.54 0.41 1996 0.22 0.26 0.35 0.42 0.48 0.32 0.98 0.98 0.99 1.00 1.00 0.99 0.67 0.76 0.85 0.80 0.97 0.79 0.26 0.29 0.41 0.49 0.62 0.42 2001 0.34 0.43 0.44 0.54 0.54 0.44 0.97 0.98 1.00 1.00 0.99 0.99 0.87 0.91 0.95 0.95 0.99 0.92 0.36 0.36 0.45 0.57 0.72 0.49 Bolivia 1996 0.89 0.96 0.96 0.97 0.99 0.95 0.39 0.59 0.65 0.66 0.77 0.61 0.08 0.28 0.35 0.36 0.49 0.33 1999 0.40 0.38 0.60 0.53 0.74 0.50 0.87 0.96 0.97 0.96 0.97 0.94 0.41 0.62 0.76 0.75 0.89 0.68 0.13 0.40 0.32 0.48 0.62 0.42 Brazil 1990 0.25 0.29 0.36 0.47 0.64 0.38 0.70 0.79 0.87 0.92 0.96 0.83 0.52 0.56 0.65 0.75 0.84 0.65 0.14 0.16 0.21 0.26 0.40 0.24 1995 0.24 0.28 0.32 0.39 0.60 0.34 0.81 0.89 0.93 0.96 0.98 0.90 0.64 0.68 0.74 0.82 0.91 0.74 0.20 0.23 0.24 0.30 0.48 0.29 2001 0.36 0.40 0.44 0.52 0.72 0.45 0.93 0.95 0.97 0.98 0.99 0.96 0.81 0.83 0.87 0.91 0.96 0.86 0.32 0.30 0.31 0.34 0.55 0.36 Chile 1990 0.24 0.23 0.29 0.35 0.48 0.31 0.96 0.96 0.98 0.97 0.99 0.97 0.79 0.81 0.81 0.88 0.94 0.83 0.25 0.22 0.25 0.32 0.52 0.31 1996 0.30 0.32 0.41 0.43 0.63 0.40 0.96 0.98 0.99 0.99 1.00 0.98 0.82 0.84 0.90 0.95 0.97 0.89 0.27 0.29 0.35 0.43 0.63 0.39 2000 0.37 0.39 0.45 0.50 0.65 0.46 0.98 0.98 0.99 0.99 1.00 0.99 0.87 0.89 0.93 0.96 0.98 0.92 0.26 0.31 0.38 0.45 0.64 0.40 Colombia 1996 0.16 0.18 0.23 0.26 0.26 0.21 0.86 0.90 0.93 0.96 0.98 0.91 0.64 0.73 0.74 0.81 0.83 0.75 0.20 0.25 0.26 0.29 0.47 0.31 1999 0.18 0.18 0.21 0.27 0.29 0.22 0.88 0.90 0.93 0.96 0.97 0.92 0.66 0.72 0.74 0.81 0.85 0.75 0.22 0.22 0.23 0.28 0.46 0.29 Costa Rica 1990 0.83 0.89 0.88 0.88 0.94 0.88 0.45 0.53 0.56 0.60 0.82 0.58 0.16 0.15 0.17 0.24 0.42 0.23 1995 0.22 0.25 0.35 0.37 0.43 0.32 0.93 0.95 0.96 0.98 0.99 0.96 0.58 0.61 0.62 0.69 0.89 0.67 0.22 0.17 0.22 0.28 0.46 0.28 2000 0.28 0.38 0.41 0.36 0.56 0.38 0.94 0.96 0.97 0.97 0.98 0.96 0.S9 0.63 0.63 0.70 0.84 0.66 0.20 0.25 0.27 0.35 0.SS 0.34 Dominican Rep. 1995 0.26 0.30 0.25 0.31 0.51 0.32 0.92 0.94 0.94 0.9S 0.96 0.94 0.85 0.87 0.88 0.89 0.91 0.88 0.37 0.38 0.35 0.37 0.47 0.39 Ecuador 1994 0.22 0.23 0.30 0.32 0.48 0.30 0.78 0.82 0.91 0.91 0.9S 0.86 0.47 0.57 0.60 0.68 0.77 0.61 0.22 0.26 0.24 0.30 0.39 0.29 1998 0.18 0.25 0.33 0.36 0.47 0.31 0.91 0.92 0.95 0.98 0.99 0.94 0.50 0.65 0.68 0.74 0.85 0.68 0.19 0.24 0.21 0.29 0.48 0.29 El Salvador 1991 0.12 0.14 0.20 0.35 0.54 0.24 0.64 0.69 0.76 0.86 0.92 0.75 0.47 0.51 0.S9 0.66 0.74 0.S9 0.14 0.14 0.19 0.26 0.39 0.24 1995 0.21 0.23 0.31 0.46 0.70 0.35 0.74 0.77 0.84 0.91 0.96 0.83 0.53 0.58 0.64 0.71 0.80 0.64 0.17 0.14 0.19 0.27 0.43 0.25 2000 0.22 0.21 0.26 0.33 0.47 0.29 0.79 0.83 0.86 0.92 0.97 0.86 0.61 0.68 0.71 0.77 0.81 0.70 0.24 0.21 0.22 0.23 0.42 0.27 Guatemala 2000 0.06 0.08 0.08 0.13 0.29 0.11 0.67 0.71 0.78 0.86 0.93 0.77 0.42 0.44 0.49 0.57 0.75 0.53 0.09 0.12 0.13 0.19 0.40 0.21 Honduras 1990 0.73 0.70 0.78 0.88 0.91 0.79 0.38 0.33 0.40 0.49 0.63 0.45 0.04 0.07 0.09 0.16 0.31 0.15 1995 0.83 0.84 0.89 0.91 0.96 0.88 0.40 0.41 0.45 0.53 0.64 0.49 0.06 0.10 0.08 0.19 0.31 0.16 1999 0.78 0.84 0.85 0.91 0.92 0.85 0.40 0.43 0.52 0.58 0.67 0.52 0.07 0.13 0.18 0.18 0.34 0.20 Jamaica 1990 0.79 0.72 0.75 0.75 0.84 0.77 0.98 0.99 0.98 0.96 1.00 0.98 0.81 0.77 0.85 0.81 0.79 0.80 0.05 0.02 0.03 0.04 0.03 0.04 1996 0.85 0.73 0.73 0.87 0.90 0.82 0.99 0.99 0.99 1.00 1.00 1.00 0.88 0.87 0.84 0.92 0.94 0.89 0.14 0.08 0.09 0.12 0.14 0.12 1999 0.96 1.00 0.96 0.97 0.98 0.97 0.99 0.99 1.00 1.00 1.00 0.99 0.87 0.86 0.88 0.95 0.96 0.90 0.13 0.14 0.29 0.18 0.18 0.19 Mexico 1992 0.39 0.53 0.66 0.78 0.83 0.62 0.86 0.94 0.96 0.96 0.97 0.93 0.48 0.54 0.65 0.63 0.84 0.61 0.12 0.15 0.19 0.24 0.40 0.23 1996 0.55 0.75 0.84 0.88 0.95 0.77 0.91 0.96 0.97 0.98 0.99 0.96 0.47 0.58 0.67 0.74 0.88 0.65 0.07 0.14 0.21 0.28 0.46 0.23 2000 0.67 0.84 0.94 0.92 0.92 0.85 0.93 0.95 0.98 0.98 1.00 0.96 0.57 0.63 0.70 0.78 0.90 0.70 0.16 0.15 0.29 0.28 0.52 0.29 Nicaragua 1993 0.11 0.19 0.26 0.33 0.54 0.28 0.93 0.95 0.98 0.98 0.99 0.97 0.54 0.70 0.74 0.78 0.86 0.72 0.21 0.23 0.20 0.29 0.41 0.28 1998 0.22 0.37 0.38 0.40 0.54 0.37 0.73 0.81 0.88 0.94 0.95 0.85 0.45 0.58 0.66 0.66 0.79 0.62 0.13 0.20 0.31 0.28 0.42 0.28 Panama 1991 0.30 0.43 0.43 0.58 0.76 0.47 0.93 0.95 0.98 0.97 0.98 0.96 0.56 0.76 0.83 0.87 0.88 0.76 0.14 0.24 0.32 0.38 0.53 0.33 1995 0.29 0.46 0.63 0.79 0.83 0.56 0.95 0.96 0.98 0.99 1.00 0.97 0.63 0.78 0.84 0.92 0.87 0.79 0.16 0.26 0.30 0.37 0.50 0.33 2000 0.52 0.58 0.79 0.88 0.91 0.69 0.96 0.97 0.99 0.99 1.00 0.98 0.70 0.82 0.85 0.91 0.96 0.83 0.22 0.29 0.34 0.44 0.65 0.39 Paraguay 1995 0.02 0.07 0.12 0.14 0.17 0.10 0.84 0.89 0.89 0.91 0.96 0.89 0.48 0.56 0.67 0.64 0.76 0.61 0.10 0.14 0.21 0.21 0.35 0.22 1999 0.11 0.17 0.16 0.17 0.17 0.15 0.90 0.93 0.97 0.98 0.99 0.95 0.60 0.69 0.75 0.83 0.84 0.73 0.13 0.27 0.21 0.31 0.49 0.30 Peru Regions 1991 0.99 0.99 0.99 1.00 1.00 0.99 0.87 0.94 0.96 0.93 0.95 0.93 1994 0.99 0.99 1.00 1.00 1.00 0.99 0.88 0.91 0.94 0.93 0.96 0.92 0.41 0.42 0.46 0.44 0.56 0.46 National 1994 0.98 0.99 0.99 0.99 0.99 0.99 0.84 0.88 0.90 0.92 0.95 0.89 0.35 0.37 0.45 0.41 0.56 0.43 2000 0.63 0.59 0.70 0.71 0.80 0.67 0.99 0.99 0.99 0.98 1.00 0.99 0.86 0.88 0.93 0.92 0.96 0.91 0.25 0.35 0.31 0.45 0.56 0.39 Trinidad & Tobago 1992 0.53 0.62 0.74 0.70 0.73 0.66 0.98 0.96 1.00 1.00 0.98 0.98 0.69 0.76 0.77 0.89 0.93 0.79 0.07 0.19 0.18 0.24 0.30 0.19 Uruguay 1989 0.24 0.39 0.56 0.67 0.85 0.50 0.96 0.98 0.98 0.98 0.99 0.98 0.68 0.76 0.78 0.89 0.92 0.78 0.17 0.23 0.29 0.36 0.50 0.32 1995 0.36 0.50 0.59 0.73 0.86 0.57 0.98 0.99 1.00 0.99 1.00 0.99 0.66 0.71 0.82 0.85 0.95 0.77 0.18 0.23 0.30 0.39 0.59 0.34 2000 0.51 0.60 0.70 0.80 0.91 0.65 0.98 0.99 0.99 1.00 1.00 0.99 0.68 0.78 0.86 0.92 0.98 0.81 0.17 0.24 0.36 0.42 0.68 0.36 RB de Venezuela 1989 0.25 0.32 0.34 0.42 0.55 0.36 0.86 0.91 0.93 0.96 0.97 0.92 0.65 0.72 0.74 0.75 0.84 0.73 0.24 0.24 0.24 0.30 0.43 0.29 1995 0.45 0.45 0.50 0.58 0.67 0.51 0.94 0.95 0.97 0.98 0.97 0.96 0.75 0.75 0.77 0.80 0.89 0.78 0.28 0.29 0.30 0.36 0.48 0.35 1998 0.45 0.46 0.50 0.59 0.68 0.52 0.94 0.96 0.96 0.97 0.99 0.96 0.72 0.75 0.76 0.80 0.91 0.78 0.27 0.28 0.30 0.37 0.54 0.35 Note: Some country statistics in the first panel refer to schooling for children age 5 and older, since no information is recorded for younger children. See http://www.depeco.econo.unip.edu.ar/cedlas/wb/ for details. Source: Authors' calculations based on microdata from household surveys. INEQUALITY IN LATIN AMERICA TABLE A.48 Distubution of conditional probabilities of school attendance, Gini coefficients Primary Secondary College College all Primary Secondary College College all Country (i) (ii) (iii) (iv) Country (i) (ii) (iii) (iv) Argentina Guatemala 1992 0.6 12.4 15.0 37.3 2000 6.2 23.4 28.2 73.0 1996 0.7 12.7 16.9 38.9 Honduras 2001 1.1 4.5 13.5 33.6 1990 3.4 29.2 26.4 77.1 Bolivia 1995 5.1 23.9 24.3 74.5 Urban 1999 5.2 24.9 27.1 74.7 1992 0.9 6.3 15.5 28.3 Mexico 1996 0.5 11.6 23.0 33.0 1992 3.2 21.7 15.7 53.1 National 1996 3.0 19.3 20.3 56.3 1996 1.1 10.7 26.3 43.5 2000 2.3 18.6 18.7 48.8 1999 1.1 8.6 28.7 38.6 Nicaragua Brazil 1993 7.2 13.5 29.4 63.6 1990 15.2 20.4 38.1 65.2 1998 5.7 13.7 25.3 58.9 1995 3.2 24.6 40.1 69.8 Panama 2001 1.0 26.2 50.1 69.7 1991 2.0 14.7 35.8 55.2 Chile 1995 1.7 14.0 32.0 50.1 1990 6.9 9.1 26.9 47.6 2000 0.9 10.3 26.7 44.3 1996 0.3 7.2 22.7 43.5 Paraguay 2000 0.4 5.8 25.6 46.5 1995 1.6 15.6 27.7 47.9 Colombia 1999 1.0 11.4 32.3 54.8 Bogota Peru 1992 2.8 3.5 17.8 34.1 Regions 1996 1.9 7.0 23.1 38.5 1991 0.7 5.0 25.6 30.5 National 1994 1.0 5.1 27.5 35.6 1996 2.9 8.3 24.1 49.9 National 1999 2.3 8.4 29.8 49.6 1994 1.0 7.1 25.1 33.8 Costa Rica 2000 0.9 5.5 30.5 40.1 1990 2.0 24.5 19.4 53.9 Trinidad & Tobago 1995 1.3 22.1 18.2 51.9 1992 1.2 11.9 58.4 66.7 2000 1.5 19.2 20.8 51.8 Uruguay Dominican Rep. 1989 2.1 12.5 36.2 58.9 1995 0.8 8.6 20.4 46.9 1995 1.5 14.0 18.9 48.5 Ecuador 2000 1.4 12.8 17.4 49.1 1994 3.2 24.1 24.3 48.9 RB de Venezuela 1998 7.5 17.2 30.6 54.6 1989 3.2 13.5 24.0 49.4 El Salvador 1995 3.7 11.8 17.7 41.8 1991 6.1 10.8 20.1 61.8 1998 3.3 11.9 19.8 41.5 1995 11.4 8.9 17.1 55.9 2000 4.8 8.6 21.6 54.0 Note: (i) children between 9 and 12; (ii) youth between 15 and 17 who finished primary school; (iii) youth between 19 and 21 who finished high school; (iv) all youth between 19 and 21. Source: Authors' calculations based on microdata from household surveys. 334 STATISTICAL APPENDIX TABLE A.49 Index of intergenerational educational mobility, by age groups 13-19 20-25 13-19 20-25 Country (i) (i) Country (i) (ii) Argentina Honduras 1992 0.922 0.851 1990 0.841 0.727 1996 0.904 0.834 1995 0.853 0.746 2001 0.907 0.814 1999 0.840 0.728 Bolivia Jamaica Urban 1990 0.990 0.929 1992 0.898 0.908 1996 0.990 0.941 1996 0.892 0.900 1999 0.984 0.973 National Mexico 1996 0.828 0.799 1992 0.905 0.842 1999 0.838 0.799 1996 0.905 0.846 Brazil 2000 0.868 0.768 1990 0.827 0.763 Nicaragua 1995 0.808 0.762 1993 0.860 0.855 2001 0.844 0.795 1998 0.828 0.811 Chile Panama 1990 0.918 0.862 1991 0.867 0.822 1996 0.914 0.823 1995 0.850 0.798 2000 0.922 0.834 2000 0.893 0.856 Colombia Paraguay Bogota 1995 0.846 0.768 1992 0.911 0.829 1999 0.851 0.762 1996 0.930 0.838 Peru National Regions 1996 0.845 0.794 1991 0.938 0.905 1999 0.842 0.812 1994 0.923 0.900 Costa Rica National 1990 0.854 0.806 1994 0.917 0.912 1995 0.853 0.778 2000 0.898 0.874 2000 0.856 0.766 Trinidad & Tobago Dominican Rep. 1992 0.964 0.944 1995 0.885 0.871 Uruguay Ecuador 1989 0.923 0.880 1994 0.852 0.815 1995 0.907 0.842 1998 0.824 0.782 2000 0.900 0.820 El Salvador RB de Venezuela 1991 0.865 0.786 1989 0.831 0.799 1995 0.844 0.774 1995 0.857 0.816 2000 0.819 0.793 1998 0.843 0.788 Guatemala 2000 0.799 0.735 Source: Authors' calculations based on microdata from household surveys. 335 INEQUALITY IN LATIN AMERICA TABLE A.S0 Health status measures and health services indicators, by quintiles of socioeconomic status a. Under-S mortality and children underweight Under-S mortality rates (per thousand) Children underweight rates Country/region 1 2 3 4 5 Average CI 1 2 3 4 5 Average Cl Bolivia 146.5 114.9 104.0 47.8 32.0 99.1 -0.22 16.9 9.8 6.3 3.9 3.1 9.0 -0.31 Brazil 98.9 56.0 39.2 26.7 33.3 56.7 -0.26 11.5 5.1 2.8 1.9 3.0 5.7 -0.21 Colombia 52.1 37.1 30.7 34.9 23.6 37.4 -0.13 14.7 9.4 6.7 3.2 3.0 8.4 -0.29 Dominican Rep. 89.9 73.0 60.1 37.3 26.6 61.0 -0.21 12.9 6.6 3.2 1.7 1.0 5.9 -0.42 Guatemala 89.1 102.9 82.0 60.7 37.9 79.2 -0.12 35.1 33.2 28.5 16.2 7.3 26.6 -0.19 Haiti 163.3 150.1 137.1 130.6 105.6 140.6 -0.07 38.9 29.8 26.8 22.2 10.2 27.5 -0.17 Nicaragua 68.8 66.6 52.5 48.5 29.7 56.0 -0.12 18.4 14.2 11.4 6.5 3.9 12.2 -0.23 Paraguay 57.2 50.0 59.0 39.4 20.1 46.6 -0.13 5.9 4.3 4.0 1.8 0.8 3.7 -0.28 Peru 110.0 76.2 48.0 44.1 22.1 68.4 -0.25 16.7 7.8 4.6 1.9 1.4 7.8 -0.40 Latin America and the Caribbean 97.3 80.8 68.1 52.2 36.8 71.7 -0.17 19.0 13.4 10.5 6.6 3.7 11.9 -0.28 East Asia, Pacific 84.0 62.9 53.7 41.1 27.1 57.1 -0.19 Central Asia 82.5 64.5 69.8 57.5 40.2 64.9 -0.09 17.8 14.2 11.0 7.8 6.5 12.1 -0.19 Middle East, 140.6 117.8 92.2 80.1 50.4 100.3 -0.17 30.1 26.5 22.0 19.6 13.7 22.7 -0.19 North Africa South Asia 144.2 152.6 136.1 110.8 71.7 126.6 -0.11 56.8 52.9 49.5 42.2 29.2 46.7 -0.11 Sub-Saharan Africa 191.7 190.9 174.3 156.6 112.4 168.4 -0.09 36.2 32.9 29.5 26.2 18.1 28.9 -0.13 All countries 148.3 140.8 126.8 110.0 77.4 124.2 -0.12 32.2 28.3 24.9 21.2 14.6 24.8 -0.17 b. Complete immunization coverage rates and prevalence of diarrhea Complete immunization coverage rates Prevalence of diarrhea (%) Country/region 1 2 3 4 5 Average Cl 1 2 3 4 5 Average Cl Bolivia 21.8 24.9 21.0 33.4 30.6 25.5 0.08 21.8 19.8 20.5 17.9 11.7 19.2 -0.07 Brazil 56.6 74.0 84.9 83.1 73.8 72.5 0.07 18.3 12.9 12.7 9.3 7.4 13.1 -0.16 Colombia 53.8 66.9 68.1 70.6 74.1 65.5 0.06 18.4 19.8 16.8 14.9 10.0 16.7 -0.09 Dominican Rep. 28.0 30.2 46.9 42.6 51.7 38.7 0.12 17.9 16.4 17.8 14.1 10.1 15.7 -0.08 Guatemala 41.2 43.0 47.1 38.3 42.5 42.6 0.00 22.8 21.5 23.3 17.7 16.0 20.9 -0.06 Haiti 18.8 20.1 35.3 37.9 44.1 30.2 0.17 30.9 27.1 24.4 31.6 20.4 27.4 -0.04 Nicaragua 61.0 74.6 75.3 85.7 73.1 72.6 0.05 16.1 14.0 14.2 14.4 8.7 14.0 -0.07 Paraguay 20.2 30.8 36.4 40.7 53.0 34.2 0.18 9.8 8.5 9.2 7.4 4.6 8.1 -0.11 Peru 55.3 63.8 63.5 71.7 66.0 63.0 0.04 21.4 20.3 18.6 14.1 9.3 17.9 -0.11 Latin America and the Caribbean 39.6 47.6 53.2 56.0 56.5 49.4 0.09 19.7 17.8 17.5 15.7 10.9 17.0 -0.09 East Asia, Pacific 48.3 56.8 60.3 64.6 72.9 59.3 0.08 10.5 9.9 9.9 8.6 6.3 9.3 -0.08 Central Asia 64.2 67.9 71.8 75.7 77.4 70.9 0.04 19.0 15.6 15.0 14.6 13.7 15.8 -0.02 Middle East, 42.2 53.3 62.5 73.2 81.1 61.0 0.17 21.0 20.3 19.1 17.2 14.7 18.7 -0.06 North Africa South Asia 29.8 31.4 41.6 49.8 64.4 42.0 0.17 17.0 14.4 14.3 15.3 12.4 14.9 -0.04 Sub-Saharan Africa 33.6 42.0 44.4 53.1 66.9 47.3 0.17 24.5 23.3 22.5 22.6 18.2 22.3 -0.05 All countries 38.3 45.8 50.3 57.2 66.6 50.7 0.14 21.2 19.6 19.1 18.5 14.8 18.9 -0.05 c. Basic antenatal care and attended delivery Basic antenatal care rates (to a medically trained person) Attended delivery rate (by a medically trained person) Country/region 1 2 3 4 5 Average Cl 1 2 3 4 5 Average Cl Bolivia 38.8 57.8 70.4 88.6 95.3 65.1 0.17 19.8 44.8 67.7 87.9 97.9 56.7 0.28 Brazil 67.5 87.7 93.4 96.9 98.1 85.6 0.08 71.6 88.7 95.7 97.7 98.6 87.7 0.07 Colombia 62.3 81.1 89.8 95.4 95.9 82.5 0.09 60.6 85.2 92.8 98.9 98.1 84.5 0.09 Dominican Rep. 96.1 98.2 99.0 99.2 99.9 98.3 0.01 88.6 96.9 97.3 98.4 97.8 95.3 0.02 Guatemala 34.6 41.1 49.3 72.2 90.0 52.5 0.19 9.3 16.1 31.1 62.8 91.5 34.8 0.42 Haiti 44.3 60.0 72.3 83.7 91.0 67.7 0.14 24.0 37.3 47.4 60.7 78.2 46.3 0.21 Nicaragua 67.0 80.9 86.9 89.0 96.0 81.5 0.07 32.9 58.8 79.8 86.0 92.3 64.6 0.19 Paraguay 69.5 79.5 85.6 94.8 98.5 83.9 0.07 41.2 49.9 69.0 87.9 98.1 66.0 0.18 Peru 37.3 64.8 79.1 87.7 96.0 67.3 0.17 13.7 48.0 75.1 90.3 96.6 56.4 0.31 Latin America and the Caribbean 57.5 72.3 80.6 89.7 95.6 76.0 0.11 40.2 58.4 72.9 85.6 94.3 65.8 0.20 East Asia, Pacific 64.9 80.7 86.9 91.4 96.2 81.9 0.08 30.5 53.0 68.4 80.6 93.4 60.8 0.22 Central Asia 78.2 84.7 86.8 93.3 96.3 86.9 0.05 82.7 92.3 95.1 98.6 99.7 92.8 0.04 Middle East, 13.7 21.1 33.4 49.3 73.0 35.2 0.32 12.8 21.7 37.7 58.6 82.2 38.5 0.36 North Africa South Asia 16.8 23.2 28.8 43.0 70.9 34.6 0.30 5.3 8.1 11.7 21.9 49.3 17.7 0.46 Sub-Saharan Africa 61.1 69.5 74.9 84.2 93.6 75.7 0.10 24.6 32.9 41.2 59.2 82.1 46.2 0.26 All countries SS.0 64.8 71.1 80.6 91.0 70.8 0.13 31.2 42.1 51.6 66.2 84.0 52.5 0.25 Cl Concentration index Source: Demographic and Health Surveys (DHS) 2002. 336 STATISTICAL APPENDIX TABLE A.51 Disabled people as a share of population in Latin Amienican and Caribbean countries Count'y Percent Year Data instrument Belize 6.6 1991 Census Brazil 16.0 2001 Census Chile 4.3 1996 Census Colombia 1.8 1993 Census Costa Rica 9.3 1998 Survey Ecuador 13.2 1996 Census El Salvador 1.9 1992 Census Jamaica 4.8 1991 Census Mexico 2.3 2000 Census Nicaragua 12.3 1995 Census Panama 1.5 1990 Census Paraguay 1.0 1992 Census Peru 13.1 1993 Census St. Vincent and the 7.2 1991 Census Grenadines Uruguay 16.0 1992 Survey TABLE A.52 Overmew of disability types, share of total disabled population Auditory Country Visual and speech Physical Mental Other Source Barbados 23.8 14.1 26.1 12.9 28.3 BARNOD Chile 26.9 32.9 21.2 13.1 5.9 MIDEPLAN Colombia 48.0 31.7 24.0 17.0 NA Down Syndrome Corporation Costa Rica 26.5 4.8 18.9 8.1 41.7 National Council of Rehabilitation and Special Education El Salvador 22.2 21.0 29.4 16.2 NA Salvadoran Institute of Disability Rehabilitation Mexico 28.6 21.0 44.9 14.6 0.7 XII General Population and Housing Census Nicaragua 63.1 14.1 8.5 4.0 9.6 National Autonomous University of Nicaragua Paraguay 11.0 20.2 19.5 13.2 36.1 Ministry of Public Health and Social Well-being Peru 9.0 20.0 49.0 22.0 NA CONADIS Uruguay 11.6 16.0 45.0 19.0 NA National Commission of Disability 337 INEQUALITY IN LATIN AMERICA TABLE A.53 Malapportionment in Latin America, 1999 Chamber Country Lower Upper Argentina 0.14 0.49 Belize 0.08 n/a Bolivia 0.17 0.38 Brazil 0.09 0.40 Chile 0.15 0.31 Colombia 0.13 0.00 Costa Rica 0.02 n/a Dominican Rep. 0.08 0.38 Ecuador 0.20 n/a El Salvador 0.07 n/a Guatemala 0.06 n/a Honduras 0.04 n/a Mexico 0.06 0.23 Nicaragua 0.06 n/a Panama 0.06 n/a Paraguay 0.04 0.00 Peru 0.00 n/a Uruguay 0.03 0.00 RB de Venezuela 0.07 0.33 Latin America 0.08 0.25 United States 0.01 0.36 Industrial democracies 0.04 0.18 World (without Latin America) 0.06 0.18 Note: Percentage of political seats allocated to districts that would not receive those seats in the case of perfect apportionment, 1999. Source: Samuels and Snyder 2001. 338 STATISTICAL APPENDIX TABLE A.54 Cnme victimization across quintiles 1 2 3 4 5 Argentina 34.1 37.7 34.5 40.4 41.2 Bolivia 33.0 32.9 37.8 37.7 30.7 Brazil 34.1 34.5 32.0 40.5 45.8 Colombia 29.4 34.3 34.9 39.4 42.2 Costa Rica 33.7 35.5 36.0 43.2 35.0 Chile 27.8 32.2 27.2 33.2 33.6 Ecuador 42.0 39.7 45.5 42.6 43.0 El Salvador 45.3 38.5 47.5 41.6 59.8 Guatemala 54.8 50.9 52.5 58.9 58.5 Honduras 28.4 27.8 39.7 44.3 41.4 Mexico 40.3 39.1 44.5 48.2 47.6 Nicaragua 29.7 32.9 34.9 40.9 42.2 Panama 25.9 26.4 34.1 29.6 26.1 Paraguay 27.2 32.1 37.5 38.9 32.9 Peru 34.9 33.8 35.4 44.0 39.3 Uruguay 20.1 17.5 23.9 31.5 31.8 RB de Venezuela 37.9 42.3 47.0 45.8 53.8 Spain 9.4 13.3 15.2 17.3 18.0 Note: Percent answering yes to the question "Have you or any member of your family been assaulted, robbed, or victimized in any way during the past 12 months?" Quintiles of a socioeconomic index constructed from the ownership of durable goods and household characteristics. Source: Gaviria and Pages 1999. 339 TABLE A.55 Access to water, hygienic restrooms, electricity, and telephone Water Hygienic restrooms Electricity Telephone Country 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean Argentina 1992 0.93 0.97 0.96 0.98 1.00 0.97 0.74 0.85 0.87 0.91 0.97 0.88 1996 0.94 0.96 0.98 0.99 1.00 0.98 0.75 0.87 0.91 0.96 0.99 0.91 2001 0.96 0.98 0.99 1.00 1.00 0.99 0.60 0.81 0.87 0.96 0.99 0.87 0.98 0.99 1.00 1.00 1.00 1.00 Bolivia Urban 1992 0.75 0.77 0.82 0.87 0.94 0.84 0.69 0.68 0.72 0.76 0.85 0.76 0.92 0.93 0.94 0.95 0.98 0.95 1996 0.78 0.83 0.89 0.92 0.95 0.88 0.60 0.69 0.80 0.87 0.94 0.79 0.92 0.95 0.96 0.98 0.99 0.96 National 1996 0.45 0.60 0.75 0.84 0.91 0.72 0.26 0.40 0.60 0.74 0.86 0.59 0.21 0.55 0.77 0.85 0.93 0.68 1999 0.20 0.58 0.75 0.81 0.90 0.66 0.24 0.55 0.75 0.83 0.90 0.67 0.22 0.63 0.85 0.90 0.95 0.72 0.01 0.08 0.18 0.31 0.58 0.25 Brazil 1990 0.17 0.35 0.56 0.73 0.87 0.56 0.60 0.82 0.93 0.98 0.99 0.88 1995 0.23 0.41 0.58 0.75 0.89 0.60 0.73 0.88 0.95 0.99 1.00 0.92 0.02 0.04 0.10 0.23 0.61 0.22 2001 0.33 0.52 0.64 0.79 0.90 0.67 0.86 0.95 0.97 0.99 1.00 0.96 0.17 0.37 0.52 0.76 0.93 0.59 Chile 1996 0.79 0.88 0.92 0.96 0.98 0.91 0.55 0.70 0.81 0.90 0.97 0.80 0.89 0.95 0.96 0.98 0.99 0.96 2000 0.83 0.91 0.94 0.97 0.99 0.93 0.65 0.78 0.86 0.92 0.98 0.85 0.95 0.97 0.98 0.99 1.00 0.98 0.29 0.44 0.59 0.75 0.91 0.62 Colombia Bogota 1992 0.97 0.98 0.99 0.99 1.00 0.99 0.99 0.99 1.00 0.99 1.00 0.99 0.74 0.81 0.83 0.89 0.97 0.86 1996 0.98 0.99 1.00 1.00 1.00 0.99 0.97 0.99 1.00 1.00 1.00 0.99 1.00 0.99 0.99 1.00 0.99 0.99 0.88 0.89 0.92 0.93 0.94 0.91 National 1996 0.60 0.78 0.86 0.91 0.96 0.84 0.54 0.76 0.84 0.90 0.97 0.82 0.81 0.91 0.94 0.96 0.98 0.92 0.13 0.24 0.35 0.51 0.71 0.41 1999 0.70 0.80 0.86 0.92 0.96 0.86 0.65 0.77 0.85 0.91 0.96 0.84 0.87 0.93 0.95 0.96 0.98 0.94 0.24 0.37 0.50 0.63 0.80 0.53 Ecuador 1994 0.40 0.46 0.59 0.65 0.80 0.59 0.42 0.53 0.68 0.75 0.90 0.67 0.70 0.86 0.90 0.93 0.96 0.88 0.OS 0.04 0.13 0.15 0.37 0.16 1998 0.5S 0.62 0.68 0.78 0.86 0.71 O.SS 0.68 0.76 0.83 0.96 0.77 0.81 0.92 0.95 0.97 0.99 0.93 0.06 0.11 0.16 0.26 0.56 0.25 El Salvador 1991 0.19 0.27 0.40 O.SS 0.76 0.46 0.OS 0.12 0.23 0.38 0.64 0.31 0.42 0.55 0.69 0.81 0.91 0.70 0.01 0.01 0.03 0.06 0.23 0.08 1995 0.22 0.29 0.38 0.56 0.77 0.47 0.08 0.13 0.25 0.43 0.71 0.35 0.51 0.64 0.76 0.88 0.96 0.77 0.02 0.03 0.07 0.16 0.41 0.16 2000 0.22 0.29 0.40 0.54 0.77 0.47 0.68 0.77 0.87 0.92 0.98 0.86 0.17 0.21 0.28 0.39 0.63 0.36 Guatemala 2000 0.57 0.60 0.68 0.75 0.92 0.72 0.08 0.17 0.23 0.38 0.74 0.35 0.49 0.64 0.76 0.84 0.93 0.75 0.01 0.04 0.07 0.15 0.48 0.18 Honduras 1990 0.43 0.50 0.59 0.69 0.85 0.63 0.05 0.09 0.16 0.30 0.62 0.27 0.15 0.22 0.35 0.59 0.80 0.45 1995 0.73 0.70 0.81 0.84 0.92 0.82 0.27 0.36 0.43 0.53 0.76 0.49 0.33 0.53 0.69 0.77 0.87 0.66 1999 0.79 0.83 0.89 0.93 0.97 0.89 0.26 0.31 0.48 0.59 0.82 0.51 0.41 0.60 0.75 0.90 0.94 0.74 Jamaica 1990 0.33 0.38 0.49 0.54 0.65 0.50 0.27 0.36 0.45 0.42 0.60 0.45 0.52 0.50 0.62 0.67 0.76 0.64 0.03 0.04 0.05 0.08 0.14 0.08 1996 0.46 0.47 0.50 0.57 0.71 0.57 0.32 0.34 0.46 0.40 0.65 0.47 0.70 0.64 0.69 0.70 0.87 0.74 0.19 0.18 0.16 0.21 0.40 0.25 1999 0.54 0.44 0.65 0.62 0.72 0.61 0.32 0.40 0.52 0.50 0.68 0.51 0.74 0.69 0.80 0.83 0.84 0.79 0.26 0.28 0.34 0.40 0.48 0.37 Mexico 1992 0.63 0.81 0.88 0.93 0.96 0.86 0.28 0.54 0.70 0.80 0.91 0.68 0.76 0.93 0.96 0.98 0.99 0.93 0.03 0.08 0.14 0.27 0.54 0.24 1996 0.80 0.90 0.93 0.96 0.98 0.92 0.32 0.53 0.71 0.84 0.93 0.70 0.85 0.95 0.98 0.99 1.00 0.96 0.05 0.10 0.21 0.35 0.60 0.29 2000 0.74 0.88 0.95 0.96 0.98 0.91 0.35 0.70 0.85 0.90 0.96 0.78 0.92 0.98 1.00 1.00 1.00 0.98 0.06 0.20 0.33 0.51 0.74 0.40 Paraguay 1995 0.687 0.71 0.79 0.85 0.95 0.81 0.09 0.24 0.48 0.66 0.87 0.50 0.42 0.64 0.82 0.90 0.97 0.78 0.00 0.01 0.04 0.10 0.37 0.12 1999 0.708 0.78 0.85 0.92 0.97 0.86 0.16 0.39 0.60 0.75 0.89 0.60 0.73 0.88 0.94 0.97 0.97 0.91 0.02 0.06 0.14 0.34 0.59 0.26 Peru Regions 1991 0.56 0.74 0.77 0.82 0.88 0.76 0.67 0.81 0.87 0.90 0.93 0.84 0.50 0.79 0.89 0.93 0.97 0.82 0.076 0.093 0.101 0.189 0.248 0.145 1994 0.55 0.71 0.78 0.85 0.91 0.77 0.55 0.76 0.82 0.92 0.96 0.81 0.49 0.78 0.87 0.91 0.95 0.81 0.01 0.048 0.095 0.208 0.389 0.163 National 1994 0.43 0.63 0.71 0.80 0.87 0.70 0.50 0.71 0.77 0.89 0.94 0.78 0.39 0.68 0.80 0.88 0.92 0.75 0.008 0.026 0.085 0.176 0.36 0.145 2000 0.51 0.64 0.79 0.85 0.93 0.75 0.60 0.72 0.85 0.90 0.97 0.82 0.53 0.66 0.84 0.89 0.96 0.78 0.033 0.072 0.199 0.358 0.623 0.271 Trinidad & Tobago 1992 0.60 0.62 0.61 0.81 0.88 0.73 0.43 0.48 0.56 0.68 0.83 0.63 0.86 0.93 0.98 0.95 0.96 0.94 0.261 0.227 0.359 0.508 0.6 0.42 Uruguay 1989 0.84 0.93 0.96 0.98 0.99 0.95 0.89 0.96 0.99 1.00 1.00 0.97 1995 0.99 0.99 1.00 1.00 1.00 1.00 0.71 0.86 0.93 0.97 0.99 0.91 0.96 0.99 1.00 1.00 1.00 0.99 2000 0.99 0.99 1.00 1.00 1.00 1.00 0.76 0.89 0.95 0.99 1.00 0.93 0.97 0.99 1.00 1.00 1.00 0.99 RB de Venezuela 1989 0.82 0.88 0.92 0.95 0.97 0.92 0.70 0.80 0.86 0.91 0.96 0.86 0.92 0.95 0.97 0.98 0.99 0.97 1995 0.88 0.90 0.92 0.96 0.98 0.93 0.75 0.80 0.87 0.92 0.97 0.87 0.98 0.98 0.99 0.99 1.00 0.99 0.15 0.17 0.24 0.33 0.52 0.30 1998 0.87 0.91 0.93 0.96 0.97 0.93 0.759 0.822 0.885 0.936 0.969 0.882 0.97 0.98 0.99 0.99 0.99 0.98 0.19 0.23 0.31 0.39 0.61 0.36 Notes: The variable water refers to the availability of a source of water in the house or lot. The variable restroom is equal to 1 when the household has a restroom with a toilet connected to the sewage system or to a septic tank. The variable electricity includes all sources of electricity. Telephone includes fixed and cellular phones. Source: Authors' calculations based on microdata from household surveys. TABLE A.56 Access to water, hygienic restrooms, electricity, and telephone: urban areas only Water Hygienic restrooms Electricity Telephone Country 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean 1 2 3 4 5 Mean Argentina 1992 0.93 0.97 0.96 0.98 1.00 0.97 0.74 0.85 0.87 0.91 0.97 0.88 1996 0.94 0.96 0.98 0.99 1.00 0.98 0.75 0.87 0.91 0.96 0.99 0.91 2001 0.96 0.98 0.99 1.00 1.00 0.99 0.60 0.81 0.87 0.96 0.99 0.87 0.98 0.99 1.00 1.00 1.00 1.00 Bolivia Urban 1992 0.75 0.77 0.82 0.87 0.94 0.84 0.69 0.68 0.72 0.76 0.85 0.76 0.92 0.93 0.94 0.95 0.98 0.95 1996 0.78 0.83 0.89 0.92 0.95 0.88 0.60 0.69 0.80 0.87 0.94 0.79 0.92 0.95 0.96 0.98 0.99 0.96 National 1996 0.79 0.78 0.85 0.90 0.94 0.88 0.62 0.59 0.70 0.84 0.92 0.79 0.86 0.93 0.94 0.96 0.98 0.96 1999 0.86 0.91 0.87 0.89 0.94 0.91 0.67 0.76 0.83 0.87 0.93 0.86 0.98 0.97 0.97 0.98 0.99 0.98 0.12 0.17 0.22 0.35 0.63 0.39 Brazil 1990 0.32 0.49 0.65 0.79 0.90 0.70 0.89 0.95 0.98 1.00 1.00 0.98 1995 0.38 0.53 0.66 0.79 0.91 0.71 0.94 0.97 0.99 1.00 1.00 0.99 0.03 0.05 0.11 0.25 0.62 0.26 2001 0.48 0.62 0.72 0.83 0.92 0.76 0.97 0.99 0.99 1.00 1.00 0.99 0.24 0.43 0.58 0.79 0.94 0.67 Chile 1996 0.96 0.98 0.99 0.99 0.99 0.99 0.76 0.85 0.91 0.95 0.98 0.91 0.98 0.99 1.00 1.00 1.00 0.99 2000 0.97 0.98 0.99 0.99 1.00 0.99 0.84 0.90 0.93 0.96 0.99 0.94 0.99 1.00 1.00 1.00 1.00 1.00 0.36 0.51 0.65 0.78 0.93 0.69 Colombia Bogota 1992 0.97 0.98 0.99 0.99 1.00 0.99 0.99 0.99 1.00 0.99 1.00 0.99 0.74 0.81 0.83 0.89 0.97 0.86 1996 0.98 0.99 1.00 1.00 1.00 0.99 0.97 0.99 1.00 1.00 1.00 0.99 1.00 0.99 0.99 1.00 0.99 0.99 0.88 0.89 0.92 0.93 0.94 0.91 National 1996 0.9S 0.96 0.98 0.98 1.00 0.98 0.89 0.93 0.96 0.97 0.99 0.96 0.99 0.98 0.99 0.99 0.99 0.99 0.38 0.42 O.S0 0.63 0.78 0.61 1999 0.96 0.98 0.98 0.99 0.99 0.98 0.90 0.94 0.97 0.98 0.99 0.97 0.99 0.99 1.00 0.99 1.00 0.99 0.56 0.62 0.69 0.76 0.88 0.74 Ecuador 1994 0.62 0.62 0.67 0.75 0.85 0.74 0.79 0.78 0.84 0.89 0.97 0.88 0.98 0.99 1.00 1.00 1.00 0.99 0.14 0.10 0.18 0.21 0.44 0.26 1998 0.77 0.77 0.78 0.87 0.90 0.84 0.86 0.85 0.86 0.90 0.98 0.91 0.99 0.99 1.00 1.00 1.00 1.00 0.18 0.17 0.21 0.31 0.61 0.36 El Salvador 1991 0.40 0.50 0.62 0.73 0.89 0.72 0.21 0.31 0.46 0.58 0.79 0.58 0.78 0.86 0.92 0.96 0.99 0.94 0.02 0.03 0.06 0.09 0.28 0.14 1995 0.44 0.48 0.55 0.68 0.83 0.67 0.26 0.31 0.44 0.57 0.79 0.57 0.83 0.90 0.93 0.96 0.99 0.95 0.07 0.07 0.12 0.21 0.47 0.26 2000 0.54 0.56 0.61 0.69 0.85 0.69 0.92 0.92 0.95 0.96 0.99 0.96 0.37 0.36 0.39 0.47 0.68 0.50 Guatemala 2000 0.81 0.77 0.84 0.84 0.97 0.89 0.40 0.47 0.52 0.58 0.85 0.67 0.80 0.88 0.90 0.94 0.98 0.94 0.06 0.11 0.15 0.24 0.58 0.35 Honduras 1990 0.67 0.72 0.76 0.79 0.90 0.82 0.20 0.31 0.36 0.45 0.75 0.54 0.65 0.75 0.77 0.88 0.95 0.86 1995 0.83 0.79 0.81 0.86 0.94 0.87 0.50 0.47 0.52 0.63 0.83 0.65 0.78 0.86 0.93 0.94 0.98 0.93 1999 0.90 0.88 0.94 0.94 0.99 0.95 0.52 0.49 0.60 0.69 0.88 0.71 0.90 0.89 0.95 0.99 1.00 0.97 Jamaica 1990 0.60 0.79 0.80 0.83 0.81 0.79 0.50 0.62 0.72 0.62 0.74 0.67 0.64 0.68 0.76 0.79 0.82 0.77 0.10 0.10 0.06 0.10 0.20 0.13 1996 0.72 0.88 0.87 0.79 0.88 0.84 0.56 0.60 0.78 0.61 0.80 0.70 0.83 0.75 0.89 0.77 0.93 0.86 0.33 0.36 0.30 0.33 0.58 0.43 1999 0.83 0.77 0.80 0.87 0.93 0.87 0.55 0.65 0.67 0.72 0.83 0.73 0.83 0.78 0.80 0.86 0.88 0.85 0.42 0.51 0.49 0.59 0.65 0.57 Mexico 1992 0.85 0.89 0.91 0.95 0.97 0.93 0.54 0.70 0.79 0.86 0.95 0.82 0.96 0.98 0.99 0.99 0.99 0.99 0.09 0.11 0.17 0.30 0.57 0.31 1996 0.90 0.93 0.95 0.97 0.99 0.96 0.56 0.69 0.81 0.89 0.96 0.83 0.95 0.99 0.99 1.00 1.00 0.99 0.11 0.15 0.25 0.39 0.64 0.37 2000 0.86 0.92 0.97 0.97 0.99 0.96 0.67 0.83 0.92 0.93 0.97 0.91 0.95 1.00 1.00 1.00 1.00 0.99 0.12 0.26 0.37 0.53 0.76 0.49 Paraguay 1995 0.68 0.81 0.87 0.90 0.95 0.89 0.30 0.46 0.65 0.78 0.92 0.75 0.78 0.92 0.96 0.97 1.00 0.96 0.02 0.04 0.07 0.14 0.42 0.21 1999 0.89 0.87 0.88 0.95 0.99 0.93 0.49 0.62 0.72 0.87 0.94 0.81 0.95 0.97 0.96 1.00 0.98 0.98 0.09 0.11 0.19 0.42 0.65 0.39 Peru Regions 1991 0.81 0.83 0.82 0.86 0.89 0.85 0.85 0.89 0.92 0.93 0.94 0.92 0.85 0.92 0.94 0.98 0.98 0.95 0.18 0.12 0.12 0.21 0.26 0.18 1994 0.80 0.81 0.86 0.88 0.94 0.87 0.84 0.85 0.89 0.95 0.98 0.92 0.83 0.92 0.96 0.97 0.98 0.95 0.03 0.06 0.11 0.23 0.41 0.21 National 1994 0.75 0.79 0.84 0.86 0.92 0.86 0.80 0.85 0.86 0.94 0.97 0.91 0.78 0.89 0.95 0.96 0.97 0.94 0.03 0.04 0.11 0.20 0.39 0.20 2000 0.87 0.83 0.88 0.92 0.96 0.91 0.86 0.90 0.97 0.97 0.98 0.96 0.86 0.89 0.97 0.97 0.99 0.96 0.12 0.13 0.27 0.42 0.66 0.40 Trinidad & Tobago 1992 0.64 0.76 0.68 0.83 0.96 0.81 0.59 0.65 0.70 0.74 0.91 0.75 0.86 0.93 0.98 0.95 0.96 0.94 0.37 0.35 0.50 0.54 0.71 0.53 Uruguay 1989 0.84 0.93 0.96 0.98 0.99 0.95 0.89 0.96 0.99 1.00 1.00 0.97 1995 0.99 0.99 1.00 1.00 1.00 1.00 0.71 0.86 0.93 0.97 0.99 0.91 0.96 0.99 1.00 1.00 1.00 0.99 2000 0.99 0.99 1.00 1.00 1.00 1.00 0.76 0.89 0.95 0.99 1.00 0.93 0.97 0.99 1.00 1.00 1.00 0.99 RB de Venezuela 1989 0.98 0.99 1.00 1.00 1.00 1.00 0.97 0.98 0.99 0.99 1.00 0.99 0.99 1.00 1.00 1.00 1.00 1.00 1995 0.96 0.99 1.00 1.00 1.00 1.00 0.92 0.98 0.99 0.99 1.00 0.99 1.00 0.99 1.00 0.99 1.00 1.00 0.34 0.25 0.31 0.39 0.67 0.49 1998 0.96 0.99 0.99 0.99 1.00 0.99 0.97 0.98 0.99 1.00 1.00 0.99 1.00 1.00 0.99 0.99 1.00 0.99 0.32 0.28 0.36 0.43 0.72 0.54 Notes: Water refers to the availability of a source of water in the house or lot. The variable restroom is equal to 1 when the household has a restroom with a toilet connected to the sewage system or to a septic tank. The variable electricity includes all sources of electricity. Telephone includes fixed and cellular phones. Source: Authors' calculations based on microdata from household surveys. INEQUALITY IN LATIN AMERICA FIGURE A.1 Equivalized household income a. South American countries b. Central American and the Caribbean countries Gini coefficient Gini coefficient 60 - 60 - 55- 55- 50- 50- 45 - 45 - 40- I I i 40- Early 1990s Mid-1990s Early 2000s Early 1990s Mid-1990s Early 2000s -|- Argentina -_- Bolivia Brazil Chile | Costa Rica El Salvador Honduras -|- Colombia - - Peru _ Uruguay Venezuela | Jamaica -e- Mexico + Nicaragua c Panama Source: Authors' calculations based on microdata from household surveys. FIGURE A.2 Equivalized household income Argentina, Mexico, and Brazil Gini coefficient 65 - 60 - 55- 50- 45- Brazil 40 p"IF, ico Early 19 90s sArgentina Mid-1990s Early 2000s Source: Authors' calculations based on microdata from household surveys. 344 STATISTICAL APPFNDIX FIGURE A.3 Aggregate welfare measures a. Argentina b. Brazil Social welfare index Social welfare index 130- 120 - 2 ~~~~~~~~~~~~~~~~~~~~120- 100 110 80 - 100 - 60 I I 90- 1992 1996 1998 2001 1990 1995 2001 c. Chile d. Costa Rica Social welfare index Social welfare index 170- 1 ~~~~~~~~~~~~~~~~~120- 150 - 130 -J 110 11011 100 90 I I 90 1990 1996 2000 1990 1995 2000 e. El Salvador f. Mexico Social welfare index Social welfare index 130- 120 -120- 110 -j 110 100 90 - 100 80 90 1991 1995 2000 1992 1996 2000 g. Panama h. Peru Social welfare index Social welfare index 140 - 130 - 130 - 120 - 120 -J 110 110 - 100 -] . 100 -. 90 90 1991 1995 2000 1991 1994 2000 i. Uruguay j. RB de Venezuela Social welfare index Social welfare index 130 - 120 - 120 - 110 _ 110 - 100 - 1-00 - : 90 - 90- 80- 1989 1995 2000 1989 1995 1998 |-- Per capita GDP - Sen - Atk(1) Atk(2) I Notes: See Lambert 1993 for details on the aggregate welfare functions. Atk(e) refers to the function proposed by Atkinson 1970, i.e., a CES function with parameter equal to e. First observation for each country = 100. Source: Authors' calculations based on microdata from household surveys and per capita GDP from World Bank 2000b and World Development Indicators Database. 345 INEQUALITY IN LATIN AMERICA FIGURE A.4 Poverty headcount ratio: share of individuals who live on less than US$2 a day, late 1990s Percent 80 - 70- 60- 50- 40- 30- 20 - 10 0 ~~~~\ Source: Authors' calculations based on Szekely 2001. FIGURE A.5 Relative income of each percentile of within-class distributions a. Argentina, 2001 b. Brazil, 2001 Income in pesos Income in reais 600 900 - 800- 500- 700- 400 - 600- 500- 300 - 400- 200 -300- 100 - ~~~~~~~~~~~~~200- 100~~~~~~~~~~~~~~~~~0 0-I 5% 25% 50% 75% 95% 5% 25% 50% 75% 95% -4- Capitalists -_- Professionals/executives Petty entrepreneurs -- Formal workers I-- nformalworkers Note: Income relative to the median of the income distribution for petty entrepreneurs. See text of box 2.1 for definitions of classes. Source: Authors' calculations based on microdata from Encuesta Permanente de Hogares (EPH) and Pesquisa Nacional por Amostra de Domicilios (PNAD). 346 STATISTICAL APPENDIX FIGURE A.6 Difference in average hours of work between unskilled and skilled workers Mexico Colombia Nicaragua Paraguay Brazil '_:_i:::_:::_i_ I Chile :___._,. El Salvador Uruguay Costa Rica RB de Venezuela Argentina. .I Peru*I Jamaica ~ Panama | Cl Around 1990 m Around 2000 Ecuador Honduras -2 -1 0 1 2 3 4 5 Hours Source: Authors' calculations based on microdata from household surveys. 347 INEQUALITY IN LATIN AMERI(:A FIGURE A.7 Difference in unemployment rates between unskilled and skilled workers Percentage points 6 - | ED Around 1990 M Around 2000 5- 4- 3- 2- 1 ~C 3 - 2- l | | rl | | SL _l_ _ 2 - ,< 02 v eo -3 - -4 -5 -6 Note: Unemployment rates in this figure refer to the ratio of the unemployed to the total adult population. Source: Authors' calculations based on microdata from household surveys. FIGURE A.8 Difference in housing ownership between top and bottom quintiles, by percentage points Percentage points 22- FE Around 1990 m Around 2000| 18- 14- 10 16 - I ] F-i 2- -2 A . -6 -10- -14- -18- -22- -26 - Source: Authors' calculations based on microdata from household surveys. 348 STATISTICAL APPENDIX FIGURE A.9 Distribution of conditional probabilities of attending college for youths, ages 19-21 Gini coefficient 80 - |] Ar.o,nd 1990 M Aro.nd 2000 60- so 40 Source: Authors' calculations based on microdata from household surveys. FIGURE A.10 Index of intergenerational educational mobility for youths, ages 13-19 Index 1.00 IEl Around 1990 U Around 2000 0.95- 0.90 0.85- Source: Authors' calcul ations bsdomiratfomhousehold surveys. 349 Bibliography Background Papers References The following papers or notes were commissioned for this Note: The word processed describes informally reproduced study, and are as yet unpublished manuscripts. Copies can works that may not be commonly available through be obtained by contacting the Office of the Chief Economist, libraries. Latin America and the Caribbean, The World Bank, or directly from the authors. Abente, Diego. 1995. "A Party System in Transition: The Case of Paraguay." In Scott Mainwaring and Timothy R. Ames, Barry. 2003. "The State, Civil Society, and Inequality." Scully, eds., Building Democratic Institutions: Party Systems Baiocchi, Gianpaolo. 2003. "After Dependency: New in Latin America. Stanford, CA: Stanford University Press. Approaches to (New) Inequalities in Latin America and Abers, Rebecca Naeara. 1996. "From Ideas to Practice: The the Caribbean in the Sociological Literature." Partido dos Trabalhadores and Participatory Governance Bird, Richard. 2003. "Taxation in Latin America: The Bal- in Brazil." Latin American Perspectives 23 (4): 35-53. ance between Equity, Efficiency, and Sustainability." . 1998. "From Clientelism to Cooperation: Local Cunningham, Wendy, and Joyce P. Jacobsen. 2003. "Income Government, Participatory Policy, and Civic Organizing Inequality Within and Across Racial and Ethnic Groups in Porto Alegre, Brazil." Politics andSociety 26 (4): 511-37. in Latin America." _ . 2000a. "Inventing Local Democracy: Neighborhood Cunningham, Wendy, and Mauricio Santamaria. 2003. Organizing and Participatory Policymaking in Porto "Labor Markets as a Source of Income Inequality." Alegre, Brazil." PhD diss. University of California, Los Fiszbein, Ariel, and Sebastian Galiani. 2003a. "Does Infla- Angeles. tion Increase Wage Inequality? Some Preliminary Evi- _ . 2000b. Inventing Local Democracy. Grassroots Politics dence from Argentina." in Brazil. Boulder, CO: Lynne Rienner Publishers. Fiszbein, Ariel, and Sebastian Galiani. 2003b. "A Brief Note Acemoglu, Daron, and James Robinson. 2000. "Why did on the Evolution of the Income Distribution in Argentina: the West Extend the Franchise?" Quarterly Journal of Eco- 1950-2000." nomics 115: 1167-99. Halac, Marina, and Sergio Schmukler. 2003. "Distributional Acemoglu, Daron, Simon Johnson, and James Robinson. Effects of Crises: The Role of Financial Transfers." 2001. "Colonial Origins of Comparative Development: Heller, Patrick, andJames Mahoney. 2003. "The Resilience and An Empirical Investigation." American Economic Review Transformability of Social Inequality in Latin America." 91: 1369-401. Malone, Mary. 2002. "The State and Inequality in Latin _ . 2002. "Reversal of Fortune: Geography and Insti- America." tutions in the Making of the Modern World Income Dis- Maldonado, Alberto. 2003. "Nota sobre la Experiencia de tribution." QuarterlyJournal of Economics 118: 1231-94. la Ciudad Bogota, Colombia." Acemoglu, Daron, Simon Johnson, James Robinson, and Tendler, Judith. 2003. "The Fear of Education." Yunyong Thaicharoen. 2002. "Institutional Causes, 351 INEQUALITY IN LATIN AMERICA Macroeconomic Symptoms: Volatility, Crises and Ahmad, Ehtisham, and Nicholas Stern. 1991. The Theory and Growth." National Bureau of Economic Research Work- Practice of Tax Reform in Developing Countries. Cambridge: ing Paper 9124. Cambridge, MA. Cambridge University Press. Acosta, M. 1999. "Overcoming the Discrimination against Ahuja, Vinod, Benu Bidani, Francisco Ferreira, and Michael Women in Mexico: A Task for Sisyphus." In J. Mendez, Walton. 1997. Everyone's Miracle? Revisiting poverty and G. O'Donnell, and P. Pinheiro, eds., The Rule of Law and inequality in East Asia. Washington, D.C.: World Bank. the Underprivileged in Latin America. Notre Dame, IN: Uni- Ajwad, M. I., and Q. Wodon. 2002. "Who Benefits from versity of Notre Dame Press. Increased Access to Public Services at the Local Level? Adato, M. 2000. "Final Report: The Impact of PROGRESA A Marginal Benefit Incidence Analysis for Education on Community Social Relationships." International Food and Basic Infrastructure" (2: 155-75). In Devarajan, S. Policy Research Institute, Washington, D.C. and E H. Rogers, eds., World Bank Economists' Forum. Adato, M., D. Coady, and M. Ruel. 2000. "Final Report: An Washington, D.C.: World Bank. Operations Evaluation of PROGRESA from the Per- _ . 2003. "Benefit and Marginal Benefit Incidence spective of Beneficiaries, Promotors, School Directors, and Analysis for Education and Health Services in Latin Health Staff." International Food Policy Research Insti- America." World Bank, Washington, D.C. Processed. tute, Washington, D.C. _ . Forthcoming. "Marginal Benefit Incidence Analy- Adato, M., B. de la Briere, D. Mindek, and A. Quisumbing. sis Using a Single Cross-section of Data." Applied Economics 2000. "Final Report: The Impact of PROGRESA on Letters. Women's Status and Intra-household Relations." Inter- Akerlof, George, and Rachel Kranton. 2000. "Economics and national Food Policy Research Institute, Washington, D.C. Identity." Quarterly Journal of Economics 115 (3): 715-53. Agarwal, Bina. 1994. A Field of One's Own: Gender and Land Alarc6n, D., and T. McKinley. 1998. "Increasing Wage Rights in South Asia. Cambridge, U.K.: Cambridge Uni- Inequality and Trade Liberalization in Mexico." In A. Berry, versity Press. ed., Poverty, EconomicReform, andlncome Distribution in Latin Aghion, Philippe, Philippe Bacchetta, and Abhijit Banerjee. America. Boulder, CO: Lynne Rienner Publishers, Inc. 2001. "A Corporate Balance Sheet Approach to Currency Albernaz, Angela, Francisco Ferreira, and Creso Franco. Crises." Discussion Paper 3092. Centre for Economic 2002. "Qualidade e Equidade no Ensino Fundamental Policy Research, London. Brasileiro." Pesquisa e Planejamento Econ6mico 32 (3): Aghion, Philppe, and Patrick Bolton. 1997. "A Theory of 453-76. Trickle-Down Growth and Development." Review of Alesina, Alberto, and Dani Rodrik. 1994. "Distributive Pol- Economic Studies 64: 151-72. itics and Economic Growth." Quarterly Journal of Economics Aghion, Philippe, Eve Caroli, and Cecilia Garcia-Pefialosa. 109: 465-89. 1999. "Inequality and Economic Growth: The Perspec- Alesina, Alberto, and G. Angeletos. 2003. "Fairness and tive of the New Growth Theories." Journal of Economic Redistribution: US versus Europe." National Bureau of Literature 37: 1615-60. Economic Research Working Paper 9502. Cambridge, Aghion, Philippe, and Simon Commander. 1999. "On the MA. Dynamics of Inequality in the Transition." Economics of Alesina, Alberto, Reza Baqir, and William Easterly. 1999. Transition 7 (2): 275-98. "Public Goods and Ethnic Divisions." Quarterly Journal Agier, Michel. 1995. "Racism, Culture, and Black Identity of Economics 114 (4): 1243-84. in Brazil." Bulletin of Latin American Research 14 (3): Alesina,Alberto,EdwardGlaeser,andBruceSacerdote. 2001. 245-64. "Why Doesn't the US Have a European-Style Welfare Agier, Michel, Olivier Barbary, Odile Hoffman, Pedro Quintin, System?" National Bureau of Economic Research Work- Hector Fabio Ramfrez, and Fernando Urrea. 2000. "Espa- ing Paper 8524. Cambridge, MA. cios Culturales, Movilidad, y Urbanizaci6n, Dinamicas Cul- Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, turales e Identidades en las Poblaciones Afro-Colombianas Sergio Kurlat, and Roumain Wacziag. 2002. "Fractional- del Pacifico Sur y Cali: Una Perspectiva Integrada." Cidse- ization." National Bureau of Economic Research Working IRD-COLCIENCIAS Project, Cali and Paris. Paper 9411. Cambridge, MA. 352 BIBLIOGRAPHY Altimir, 0. 1986. "Estimaciones de la distribuci6n del Case of Colombia." World Bank, Washington, D.C. ingreso en la Argentina, 1953-1980." Desarrollo Economico Processed. 25 (100). Angrist, Joshua David, Eric Bettinger, Erik Bloom, .1994. "Income Distribution and Poverty Through Elizabeth King, and Michael Kremer. 2002. "Vouchers Crisis and Adjustment." ECLAC Review 52: 7-31. for Private Schooling in Colombia: Evidence from a .1996a. "Economic Development and Social Equity: Randomized Natural Experiment." American Economic A Latin American Perspective." Journal of Interamerican Review 92 (5): 1535-58. Studies and World Affairs 38 (2/3): 47-7 1. Appadurai, Arjun. 2004. "The Capacity to Aspire: Culture .1996b. "Cambios de la desigualdad y la pobreza en and the Terms of Recognition." In Vijayendra Rao and la America Latina." El Trimestre Economico 241 (61): 1. Michael Walton, eds. 2004. Culture and Public Action. .1998a. "Income Distribution and Poverty through Stanford, CA: Stanford University Press. Crisis and Adjustment." In A. Berry, ed., Poverty, Economic Arbache, Jorge S. 2002. "Do Unions Always Decrease Wage Reform, and Income Distribution in Latin America. Boulder, Dispersion? The Case of Brazilian Manufacturing."Jour- CO: Lynne Rienner Publishers, Inc. nal of Labor Research 20 (3): 425-36. . 1998b. "Inequality, Employment, and Poverty in Archer, Ronald P. 1990. "The Transition from Traditional to Latin America: An Overview." In V. Tokman and G. BrokerClientelism in Colombia: Political Stabilityand Social O'Donnell, eds. Poverty and Inequality in Latin America: Unrest." Working Paper 140. Kellogg Institute for Inter- Issues and New Challenges. Notre Dame, IN: University of national Studies, University of Notre Dame. Notre Dame Press. Arellano, Manuel and Stephen R. Bond. 1991. "Some Tests Altimir, O., and L. Beccaria. 2001. "Persistent Dete- of Specification for Panel Data: Monte Carlo Evidence and rioration of Income Distribution in Argentina." an Application to Employment Equations." Review of Desarrollo Econdmico: Revista de Ciencias Sociales 40 Economic Studies 58 (2): 277-97. (160): 589-618. Arias, Omar, Gustavo Yamada, and Luis Tejerina. 2002. Edu- Altonji, Joseph, and Rebecca Blank. 1999. "Race and Gender cation, Family Background, and Racial Earnings Inequality in the Labor Market." In 0. Ashenfelter and David Card, in Brazil. Washington, D.C.: Inter-American Develop- eds., Handbook of Labor Economics, Volume 3C. Amsterdam: ment Bank. Elsevier. Armingeon, K. 2002. "The Effects of Negotiation Democ- Ames, Barry. 2001. The Deadlock of Democracy in Brazil. Ann racy: A Comparative Analysis." European Journal of Polit- Arbor, MI: University of Michigan Press. ical Research 41 (1): 81-105. _____. 2003. "The State, Civil Society, and Inequality." Arocha, Jaime. 1998. "Inclusion of Afro-Colombians: Background Paper for the World Bank project on inequal- Unreachable National Goals?" Latin American Perspectives ity in Latin America and the Caribbean. University of 25 (3): 70-88. Pittsburgh, Pittsburgh, PA. Processed. Atkinson, Anthony B. 1970. "On the Measurement of Amiel, Yoram and Frank Cowell. 1997. "The Measurement Inequality."Journal of Economic Theory 2: 244-63. of Poverty: an Experimental Questionnaire Investigation." . 1997. "Bringing Income Distribution in from the Empirical Economics 22 (4): 571-88. Cold." EconomicJournal 107: 291-321. Andersen, L. 2001. "Social Mobility in Latin America: Links _ . 2000. "Countries and the Redistributive Impact of with Adolescent Schooling." Working Paper R-433. the Government Budget." World Institute for Develop- Inter-American Development Bank Research Network, ment Economics Research Working Paper 202. Helsinki: Washington, D.C. WIDER. Angell, Alan, Pamela Lowden, and Rosemary Thorp. 2001. _ . 2003. "Income Inequality in OECD Countries: Decentralizing Development: The Political Economy of Insti- Data and Explanations." Working Paper 881. Munich: tutional Change in Colombia and Chile. New York: Oxford CESifo. University Press. Atkinson, Anthony B. and A. Brandolini. 2001. "Promise Angel-Urdinola, Diego F. 2003. "Minimum Wage Increases and Pitfalls in the Use of'Secondary' Data-sets."Journal Can Have an Adverse Impact on Wage Inequality: The of Economic Literature 39 (3). 353 INEQUALITY IN LATIN AMERICA Atkinson, Anthony B. and Joseph E. Stiglitz. 1980. Lectures Health, and Nutrition Program)." International Food on Public Economics. London: McGraw-Hill. Policy Research Institute, Washington, D.C. Atkinson, Anthony B., A. Brandolini, and T. Smeeding. . 1999b. "A Preliminary Evaluation of the Sample 2002. "Producing Time Series Data for Income Distrib- Sizes Used for the Evaluation of the Education, Health, ution: Sources, Methods, and Techniques." Luxemborg and Nutrition Program (PROGRESA) of Mexico." Inter- Income Study Working Paper 295. Syracuse, NY: national Food Policy Research Institute, Washington, University of Syracuse. D.C. Baer, Werner. 1986. "Growth with Inequality: The Cases of Behrman, J., and J. Hoddinott. 2000. "An Evaluation of the Brazil and Mexico." Latin American Research Review 21 (2): Impact of PROGRESA on Pre-school Child Height." 197-207. International Food Policy Research Institute, Washington, Baiocchi, Gianpaolo. 2001. "Participation, Activism, D.C. and Politics: the Porto Alegre Experiment and Delib- Behrman, J., Piyali Sengupta, and P. Todd. 2000a. "Final erative Democratic Theory." Politics and Society 29 (1): Report: The Impact of PROGRESA on Achievement Test 43-72. Scores in the First Year." International Food Policy Baker, J. L. 2003. "Protecting the Poor through Social Pro- Research Institute, Washington, D.C. tection Programs." In Q. Wodon, ed., "Public Spending, . 2000b. "Progressing Through PROGRESA: An Poverty, and Inequality in Latin America." World Bank, Impact Assessment of a School Subsidy Experiment." Washington, D.C. International Food Policy Research Institute, Washington, Bakewell, Peter. 1997. A History of Latin America, New York: D.C. Blackwell Publishing. Behrman, J., A. Gaviria, and M. Szekely. 2001. "Intergen- Baland, Jean-Marie and James A. Robinson. 2003. "Land and erational Schooling Mobility in Latin America. " Economi'a Power." Centre for Economic Policy Research Discussion 2 (3): 1-44. Paper 3800. CEPR, London. BehrmanJere, Nancy Birdsall, and Miguel Szekely. 2001. "Eco- Baldacci, Emanuele, Luiz de Melo, and Gabriela Inchauste. nomic Policy and Wage Differentials in Latin America." 2002. "Financial Crises, Poverty, and Income Distribution." Inter-American Development Bank, Washington, D.C. Working Paper 02/04. International Monetary Fund, Behrman, J., B. Davis, D. Levy, and E. Skoufias. 1998. "A Washington, D.C. Preliminary Evaluation of the Selection of Beneficiary Banco de Espafia. 1999. Circular 9/1999. Madrid. Households in the Education, Health, and Nutrition Banerjee, Abhijit V., and Andrew F. Newman. 1993. "Occu- Program (PROGRESA) of Mexico." Report submitted to pational Choice and the Process of Development." Jour- PROGRESA. International Food Policy Research Institute, nal of Political Economy 101 (2): 274-98. Washington, D.C. Baron, James N., and Michael T. Hannan. 1994. "The Benabou, Roland. 2000. "Unequal Societies: Income Dis- Impact of Economics on Contemporary Sociology."Jour- tribution and the Social Contract." American Economic nal of Economic Literature 32 (3): 1111-46. Review 90 (1): 96-129. Barro, Robert J., and Jong-Wha Lee. 2001. "International Bernanke, Ben S., and Refet S. Gurkaynak. 2002. "Is Growth Data on Educational Attainment: Updates and Implica- Exogenous? Taking Mankiw, Romer, and Weil Seriously." tions." Oxford Economic Papers 53 (3): 541-63. National Bureau of Economic Research Working Paper Barrett, P. 2001. "Labour Policy, Labour-Business Relations, 8365. Cambridge, MA. and the Transition to Democracy in Chile." Journal of Berry, A. 1998. Poverty, Economic Reform, and Income Distrib- Latin American Studies 33: 561-97. ution in Latin America. Boulder, CO: Lynne Rienner Pub- Bates, Robert H. 1981. Markets and States in Tropical Africa. lishers, Inc. Berkeley: University of California Press. Berry, A., and J. Tenjo. 1998. "Trade Liberalization, Labor Becker, Gary. 1971. The Economics of Discrimination (second Reform, and Income Distribution in Colombia." In A. edition). Chicago: University of Chicago Press. Berry, ed., Poverty, Economic Reform, and Income Distribu- Behrman, J., and P. E. Todd. 1999a. "Randomness in the tion in Latin America. Boulder, CO: Lynne Rienner Experimental Samples of PROGRESA (Education, Publishers, Inc. 354 BIBLIOGRAPHY Bertola, Luis, and Gabriel Porcile. 2002. Rich and Impover- David Card, eds., Handbook of Labor Economics, Volume 3A. ished Cousins: Economic Performance and Income Distribution Amsterdam: Elsevier. in Southern Settler Societies. Montevideo: Universidad de la Boix, Carles. 1998. Political Parties, Growth and Equity: Con- Repuiblica. servative and Social Democratic Economic Strategies in the World Besley, Timothy, and Stephen Coate. 1991. "Public Provi- Economy. Cambridge, U.K.: Cambridge University Press. sion of Private Goods and the Redistribution of Income." Bordo, Michael, Barry Eichengreen, Daniela Klingebiel, and American Economic Review 81 (4): 979-984. Maria Soledad Martfnez Peria. 2001. "Financial Crises: Biewen, M. 2002. "Bootstrap Inference for Inequality, Mobil- Lessons from the Last 120 Years." Economic Policy 16 (32): ity, and Poverty Measurement. "Journal of Econometrics 108: 51-82. 317-42. Borjas, GeorgeJ., and Stephen G. Bronars. 1989. "Consumer Binder, A. 1991. Reform of the Penal System in Latin America. Discrimination and Self-employment." Journal of Politi- Arlington, VA: National Council of State Courts. cal Economy 97 (3): 581-605. Binswanger, Hans P., and Klaus Deininger. 1995. "World Boschi, R. 1999. "Decentralization, Clientelism, and Social Bank Land Policy: Evolution and Current Challenges." Capital in Urban Governing: Comparing Belo Horizonte World Bank, Washington, D.C. and Salvador." Dados: Revista de Ciencias Sociais 42 (4): Binswanger, Hans P., Klaus Deininger, and Gershon Feder. 655-90. 1995. "Power, Distortions, Revolt and Reform in Agri- Bouloukos, A., and B. Dakin. 2001. "Toward a Universal cultural Land Relations." In Handbook of Development Eco- Declaration of the Rule of Law: Implications for Crimi- nomics, Volume 3. Amsterdam: North-Holland Publishing nal Justice and Sustainable Development." International Company. Journal of Comparative Sociology 42 (1/2): 145-62. Bird, Richard 1992. "Tax Reform in Latin America: A Bourdieu, Pierre 1984. Distinction: A Critique of the Review of Some Recent Experiences." Latin American Judgement of Taste. Cambridge, MA: Harvard University Research Review 27: 7-34. Press. Bird, Richard, ed. 1991. More Taxing than Taxes? The Tax- _ . 1986. "The Forms of Capital." In John Richardson, like Effects of Non-Tax Policies in LDCs. San Francisco: ICS ed., Handbook of Theory and Research for the Sociology of Press. Education. New York: Greenwood Press. Bird, Richard, and 0. Oldman. 1968. "Tax Research and Tax _ . 1990. The Logic of Practice. Stanford, CA: Stanford Reform in Latin America: A Survey and Commentary." University Press. Latin American Research Review 3: 5-28. _ . 1998. Practical Reason. Stanford, CA: Stanford Uni- Bird, R., and De Wulf, L. 1973. "Taxation and Income Dis- versity Press. tribution in Latin America: A Critical Review of Empir- Bourguignon, Franqois. 2002. "The Growth Elasticity of ical studies." International Monetary Fund Staff Papers 20 Poverty Reduction: Explaining Heterogeneity across (3): 639-82. Countries and Time Periods." In T. Eichler and S. Bird, Richard, and Barbara Miller. 1989. "The Incidence of Turnovsky, eds., Growth and Inequality. Cambridge, MA: Indirect Taxes on Low-Income Households in Jamaica." Massachusetts Institute of Technology Press. Economic Development and Cultural Change 37 (2): 393-409. Bourguignon, Franc,ois, and Thierry Verdier. 2000. "Oli- Bird, Richard, and Enid Slack. 2002. "Land and Property garchy, Democracy, Inequality and Growth." Journal of Taxation: A Review." Paper prepared for World Bank Development Economics 62 (2): 285-313. Workshop, "Regional Workshop on Land Issues in Latin Bourguignon, Franc,ois, and Christian Morrison. 2002. America and the Caribbean," Pachuca, Mexico. "Inequality among World Citizens: 1820-1992." Amer- Birdsall, Nancy, and Juan L. Londoino. 1997. "Asset ican Economic Review 92 (4): 727-44. Inequality Matters: An Assessment of the World Bank's Bourguignon, Francois, and Satya Chakravarty. 2003. "The Approach to Poverty Reduction." American Economic Measurement of Multidimensional Poverty." Journal of Review 82 (2): 32-37. Economic Inequality 1 (1): 25-49. Blau, Francine, and Lawrence Kahn. 1999. "Institutions and Bourguignon, Francois, Francisco H. G. Ferreira, and Laws in the Labor Market." In Orley Ashenfelter and Nora Lustig. 1998. "The Microeconomics of Income 355 INEQUALITY IN LATIN AMERICA Distribution Dynamics in East Asia and Latin America." Buscaglia, E., and T. Ulen. 1997. "A Quantitative Assess- Research Proposal 638-18. World Bank, Washington, D.C. ment of the Efficiency of the Judicial Sector in Latin Bourguignon, Francois, Francisco H.G. Ferreira, and America." International Review of Law andEconomics 17(2). Phillippe Leite. 2003. "Beyond Oaxaca-Blinder: Account- Calder6n, Fernando, and Alicia Szmukler. 2004. "Political ing for Differences in Household Income Distributions Culture and Development." In Vijayendra Rao and across Countries." Policy Research Paper 2828. World Michael Walton, eds., Culture and Public Action. Stanford, Bank, Washington, D.C. CA: Stanford University Press. ___ . 2003. "Conditional Cash Transfers, Schooling, and Calder6n, Cesar, and Alberto Chong. 2004. "Volume and Child Labor: Micro-Simulating Brazil's Bolsa Escola Pro- Quality of Infrastructure and the Distribution of Income: gram." World Bank Economic Review 17 (2): 229-54. An Empirical Investigation." Review of Income and Wealth Bourguignon, Francois, Francisco H. G. Ferreira, and Marta 50 (1): 87. Menendez. 2003. "Inequality of Outcomes and Inequal- Calder6n, Cesar, and Luis Serven. 2003. "Macroeconomic ity of Opportunities in Brazil." Policy Research Paper Dimensions of Infrastructure in Latin America." World 3174. World Bank, Washington, D.C. Bank, Washington, D.C., and the Central Bank of Chile, Boyland, Delia. 1996. "Taxation and Transition: The Poli- Santiago. Processed. tics of the 1990 Chilean Tax Reform." Latin American Calder6n, Cesar, William Easterly, and Luis Serven. 2002. Research Review 31 (1). "How did Latin America's Infrastructure Fare in the Era of Bravo, D., D. Contreras, and 1. Millan. 2001. "The Distrib- Macroeconomic Crises?" Central Bank of Chile Working utional Impact of Social Expenditure: Chile 1990-1998." Paper 285. World Bank, Washington, D.C. In World Bank, Chile's High Growth Economy: Poverty and Calvo, Ernesto. 2001. "Transformation of Labor Markets Income Distribution 1987-1998. World Bank Country in Latin America: Collective Bargaining and Industrial Study 73-114. Washington, D.C.: World Bank. Wage Disparities." Desarrollo Economico: Revista de Cien- Bresser Pereira, Luiz Carlos. 1998. Reforma do Estadopara a cias Sociales 41 (163): 395-410. Cidadania: A Reforma Gerencial Brasileira no Perspectiva Calvo, Ernesto, Juan-Carlos Torre, and Mariela Szwarcberg. Internacional. Sao Paulo: Editora 34. 2002. The New Welfare Alliance. Buenos Aires: Universi- Bruhn, K. 1997. "The Seven-Month Itch? Neoliberal Politics, dad di Tella, Department of Political Science. Popular Movements, and the Left in Mexico." In Chalmers, Calvo, Guillermo. 1998. "Capital Flows and Capital- D., Carlos M. Vilas, Katherine Hite, Scott B. Martin, Market Crises: The Simple Economics of a Sudden Stop." Kerianne Piester, and Monique Segarra, eds. The New Pol- Journal of Applied Economics 1 (1): 3 5-54. itics of Inequality in Latin America: Rethinking Participation Calvo, Guillermo, and Carmen Reinhart. 2000. "When and Representation. New York: Oxford University Press. Capital Inflows Come to a Sudden Stop: Consequences Buhmann, B., G. Rainwater, G. Schmaus, and T. Smeeding. and Policy Options." In P. Kenen and A. Swoboda, eds., 1988. "Equivalence Scales, Well-being, Inequality and Key Issues in Reform of the International Monetary and Finan- Poverty: Sensitivity Estimates Across Ten Countries cial System. Washington, D.C.: International Monetary Using the Luxembourg Income Study Database." Review Fund. of Income and Wealth 34: 115-42. Camargo, Jose, and Ferreira, Francisco. 2001. "O Beneficio Bunwaree, S. 2001. "The Marginal in the Miracle: Human Social Unico: Una proposta de reforma da politica social Capital in Mauritius." Internationaljournal of Educational no Brasil." Discusi6n Paper 443, Departamento de Development 21 (3): 257-7 1. Economia, Pontificia Universidad Cat6lica, Rio deJaneiro. Burki, Shahid Javed, and Guillermo Perry. 1998. Beyond the Cameron, D. 1984. "Social Democracy, Corporatism, Labor Washington Consensus: Institutions Matter. Washington D.C.: Quiescence, and the Representation of Economic Inter- World Bank. ests in Advanced Capitalist Democracies." In John Burniaux, Jean-Michel, Thai-Thanh Dang, Douglas Fore, Goldthorpe, ed., Order and Conflict in Contemporary Michael Forster, Marco Mira d'Ercole, and Howard Oxley. Capitalism. Oxford: Oxford University Press. 1998. "Income Distribution and Poverty in Selected Campbell, Tim E. J. 2003. The Quiet Revolution: Decentral- OECD Countries." OECD Economics Department Work- ization and the Rise of Political Participation in Latin American ing Paper 189. Paris. Cities. Pittsburgh: University of Pittsburgh Press. 356 BIBLIOGRAPHY Caprio, Gerard, and Daniela Klingebiel. 1997. "Bank Insol- Rethinking Participation and Representation. New York: vency: Bad Luck, Bad Policy, or Bad Banking?" In M. Oxford University Press. Bruno and B. Pleskovic, eds., Annual World Bank Con- Chang, Roberto, and Andres Velasco. 2001. "A Model of ference on Development Economics 1996. Washington, D.C: Financial Crises in Emerging Markets." QuarterlyJournal World Bank. of Economics 116 (2): 489-5 17. Card, David, and John E. DiNardo. 2002. "Skill Biased Tech- Chavas, Jean-Paul. 2001. "Structural Change in Agricultural nological Change and Rising Wage Inequality: Some Prob- Production: Economics, Technology and Policy." In B. lems and Puzzles." National Bureau of Economic Research Gardner and G. Rausser, eds., Handbook of Agricultural Working Paper 8769. Cambridge, MA. Economics, Volume 1. Amsterdam and New York: Elsevier. Cardoso, Eliana. 1992. "Inflation and Poverty." National Chen, S., Datt, G., and Martin Ravallion. 1995. "Is Poverty Bureau of Economic Research Working Paper 4006. Increasing in the Developing World?" World Bank Policy Cambridge, MA. Research Paper 40 (updated version). Washington, D.C. Cardoso, Eliana, and A. Helwege. 1992. Latin America's Chen, S., and M. Ravallion. 2001. "How Did the World's Economy: Diversity, Trends and Conflicts. Cambridge, MA: Poorest Fare in the 1990s?: Methodology." Global Massachusetts Institute of Technology Press. Poverty Monitoring Database. World Bank, Washington, Cardoso, Fernando Henrique. 1973. "Associated-Dependent D.C. Available at http://www.worldbank.org/research/ Development: Theoretical and Practical Implications." In povmonitor/method.htm. Alfred Stepan, ed. Authoritarian Brazil: Origins, Policies, Chevigny, P. 1999. "Defining the Role of the Police in Latin and Future. New Haven: Yale University Press. America." In Mendez, J., G. O'Donnell, and P. Pinheiro, Carey, J., and M. Shugart. 1995. "Incentives to Cultivate a eds. TheRule of Lawand the Underprivileged in Latin America. Personal Vote: A Rank Ordering of Electoral Formulas." Notre Dame, IN: University of Notre Dame Press. Electoral Studies 14 (4): 417-39. Chong, Alberto, and Florencio L6pez-de-Silanes. 2003. "Pri- Carnoy, Martin, 2002. "Is Latin American Education vatization in Latin America: What Is the True Record?" Preparing its Workforce for 21st Century Economies?" Inter-American Development Bank, Washington, D.C. Background paper for De Ferranti and others. 2003. Clos- Processed. ing the Gap in Education and Technology. Washington, D.C.: Chu, Ke-young, Hamid Davoodi, and Sanjeev Gupta. 2000. World Bank. "Income Distribution and Tax and Government Social Carter, Michael R., Bradford L. Barham, and Dina Mesbah. Spending Policies in Developing Countries." United 1996. "Agricultural Export Booms and the Rural Poor Nations University and World Institute for Development in Chile, Guatemala, and Paraguay." Latin American Economics Research Working Paper 214. Helsinki. Research Review 31 (1): 33-65. Chubb, Judith. 1983. Patronage, Power and Poverty in Sou- Carvalho, Jose Murilo de. 1997. "Mandonismo, Coronelismo, thern Italy: A Tale of Two Cities. Cambridge: Cambridge Clientelismo: Uma Discussao Conceitual." Dados: Revista University Press. de Cie^ncias Sociais 40 (2). Clert, C. 2000. "Social Exclusion, Gender, and the Chilean Castell6, Amparo, and Rafael Domenech. 2002. "Human Government's Anti-Poverty Strategy: Priorities and Meth- Capital Inequality and Economic Growth: Some New ods in Question." In E. Gacitua-Mario and C. Sojo, eds. Evidence." EconomicJournal 112: C187-200. Social Exclusion and Poverty Reduction in Latin America. Castells, Manuel. 1997. The Power of Identity. The Informa- Washington, D.C.: World Bank and FLACSO. tion Age, Economy, Society, and Culture (Volume 2). Clert, C. and Q. Wodon. 2002. "The Targeting of Govern- Malden, MA and Oxford: Blackwell Publishing. ment Programs in Chile: A Quantitative and Qualitative CEDLAS. 2003. Estad/sticas distributivas de la Argentina. Assessment." In Chile's High Growth Economy: Poverty and La Plata, Argentina: Universidad Nacional de La Income Distribution 1987-1998. World Bank Country Plata, Centro de Estudios Distributivos, Laborales y Study 141-66. Washington, D.C.: World Bank. Sociales. Coady, David. 2000. "Final Report: The Application Chalmers, D., Carlos M. Vilas, Katherine Hite, Scott B. of Social Cost-Benefit Analysis to the Evaluation of Martin, Kerianne Piester, and Monique Segarra, eds. PROGRESA." International Food Policy Research 1997. The New Politics of Inequality in Latin America: Institute, Washington, D.C. 357 INEQUALITY IN LATIN AMERICA ____ . 2001. "Evaluation of the Distributional Power of Salvador." Williams College, Department of Economics, PROGRESA's Cash Transfers in Mexico." International Williamstown, MA. Food Policy Research Institute, Food Consumption Nutri- Contreras, Dante, and Andres G6mez-Lobo. 2000. Privatiza- tion Division Discussion Paper 117. Washington, D.C. tion of Telecommunications and Electricity in Chile: How Coady, David, and R. L. Harris. 2000. "A General Equi- did the Poor Fare?" University of Chile, Department of librium Analysis of the Welfare Impact of PROGRESA Economics, Santiago. Processed. Transfers." International Food Policy Research Institute, Contreras, Dante, Oscar Larrafiaga, Julie Litchfield, and Washington, D.C. Alberto Valdes. 2001. "Poverty and Income Distribution Coady, David P., and Susan Parker. 2002. "A Cost- in Chile, 1987-1998: New Evidence." Cuadernos de Effectiveness Analysis of Demand and Supply-Side Edu- Economia 38 (114): 191-208. cation Interventions: the Case of PROGRESA in Mexico." Coppedge, Michael. 1994. Strong Parties and Lame Ducks: International Food Policy Research Institute, Food Presidential Partyarchy and Factionalism in Venezuela. Consumption Nutrition Division Discussion Paper 127. Stanford, CA: Stanford University Press. Washington, D.C. Cornelius, Wayne A. 1977. "Leaders, Followers, and Official Coady, David, R. Perez, and H. Vera-Llamas. 2002. "A Cost Patrons in Urban Mexico." In Steffen W. Schmidt, James Analysis of PROGRESA." International Food Policy C. Scott, Carl Lande, and Laura Guasti, eds., Friends, Research Institute. Washington, D.C. Processed. Followers, and Factions: A Reader in Political Clientelism. Coady, David, Margaret Grosh, and John Hoddinott. 2003. Berkeley: University of California Press. "Targeting Outcomes Redux." International Food Policy Corrales, J., and 1. Cisneros. 1999. "Corporatism, Trade Research Institute, Food Consumption Nutrition Divi- Liberalization, and Sectoral Responses: The Case of sion Discussion Paper 144. Washington, D.C. Venezuela, 1989-1999." World Development 27 (12): Coatsworth, John H. 1981. Growth against Development: The 2099-122. Economic Impact of Railroads in Porfirian Mexico. De Kalb, Cortez, W. 2001. "What Is Behind Increasing Wage Inequal- IL: Northern Illinois University Press. ity in Mexico?" World Development 29 (11): 1905-22. Cohen, Yousef. 1989. The Manipulation of Consent: The State Cotler, Julio. 1978. Clases, Estado y Nacion en el Perd. Lima: and Working-Class Consciousness in Brazil. Pittsburgh: Uni- IEP. versity of Pittsburgh Press. Coudouel, A.,J. Hentschel, and Q. Wodon. 2002. "Poverty Coleman, J. 1990. Foundations of Social Theory. Cambridge, Measurement and Analysis." In J. Klugman, ed., A MA: Harvard University Press. Sourcebook for Poverty Reduction Strategies. Volume 1: Core Collier, David, and Ruth Berins Collier. 1991. Shaping the Techniques and Crosscutting Issues. Washington, D.C.: World Political Arena: CriticalJunctures, the Labor Movement, and Bank. Regime Dynamics in Latin America. Princeton, NJ: Prince- Cowell, Frank. 1995. Measuring Inequality. Hemel Hemp- ton University Press. stead: Harvester Wheatsheaf. Conaghan, Catherine M. 1995. "Politicians against Parties: _ . 2000. "Measurement of Inequality." In Francois Discord and Disconnection in Ecuador's Party System." Bourguignon and Anthony B. Atkinson, eds., Handbook of In Scott Mainwaring and Timothy R. Scully, eds., Build- Income Distribution. Amsterdam: North-Holland Publish- ing Democratic Institutions: Party Systems in Latin America. ing Company. Stanford, CA: Stanford University Press. Cox, G., and M. Shugart. 1991. "Proportionality, Dispro- Conning, Jonathan, and James A. Robinson. 2002. "Land portionality, and Electoral Systems: Comment." Electoral Reform and the Political Organization of Agriculture." Studies 10 (4): 348-52. Centre for Economic Policy Research Discussion Paper . 1996. "In the Absence of Vote Pooling: Nomina- 3204. London. tion and Vote Allocation Errors in Colombia." Electoral Conning,J., and M. Kevane. 2002. "Community-based Tar- Studies 14 (4): 441-60. geting Mechanisms for Social Safety Nets: A Critical Cragg, Michael Ian, and Mario Epelbaum. 1996. "Why Has Review." World Development 30 (3): 375-94. Wage Dispersion Grown in Mexico? Is It the Incidence Conning, Jonathan, Pedro Olinto, and A. Trigueros. of Reforms of the Growing Demand for Skills?"Journal 2000. "Managing Economic Insecurity in Rural El of Development Economics 51 (1): 99-116. 358 BIBLIOGRAPHY Criscuolo, Alberto. 2002 "Crafting Capitalism: Artisan Asso- Deaton, Angus, and S. Zaidi. 2002. "Guidelines for Con- ciations and Small Business Development in the Third structing Consumption Aggregates for Welfare Analysis." Italy." Master's Thesis, Massachusetts Institute of World Bank Living Standards Measurement Study Technology, Department of Urban Studies and Planning. Working Paper 135. Washington, D.C. Cambridge, MA. Deere, C., and M. Leon. 2001a. Empowering Women: Land and Cubberley, Ellwood P. 1920. The History of Education: Edu- Property Rights in Latin America. Pittsburgh, PA: Uni- cational Practice and Progress Considered as a Phase of the versity of Pittsburgh Press. Development and Spread of Western Civilization. Boston: _ . 2001b. "Institutional Reform of Agriculture Houghton Mifflin. under Neoliberalism: The Impact of the Women's and Cunningham, Wendy. 2000. "Breadwinner versus Care- Indigenous Movements." Latin American Research Review giver." In Elizabeth Katz and Maria Correia, eds., The Eco- 36 (2): 31-64. nomics of Gender in Mexico: Market, Family, State. De Ferranti, David, Guillermo Perry, Indermit Gill, and Luis Washington, D.C.: World Bank. Serven. 2000. Securing our Future in a Global Economy. _____. 2001. "Sectoral Allocation by Gender of Latin Washington, D.C.: World Bank. American Workers over the Liberalization Period of the De Ferranti, David, Guillermo E. Perry, Daniel Lederman, 1990s." World Bank Policy Research Working Paper and William E. Maloney. 2002. From Natural Resources to 2742. Washington, D.C. the Knowledge Economy: Trade andJob Quality. Washington Cunningham, Wendy, and William Maloney. 2000. "Mea- D.C.: World Bank. suring Vulnerability: Who Suffered in the 1995 Mexican De Ferranti, David, Guillermo Perry, Indermit Gill, Jose Crisis?" World Bank, Washington, D.C. Processed. Luis Guasch, William F. Maloney, Carolina Sanchez- .2003. "Heterogeneity among Mexico's Microenter- Paramo, and Norbert Schady. 2003. Closing the Gap in prises: An Application of Factor and Cluster Analysis." Education and Technology. Washington, D.C: World Bank. Economic Development and Cultural Change 50 (1): 131-56. Deininger, Klaus, and Pedro Olinto. 2000. "Asset Distrib- Cunningham, Wendy, andJoyce P. Jacobsen. 2003. "Income ution, Inequality, and Growth." World Bank Policy Inequality Within and Across Racial and Ethnic Groups Research Paper 2375. Washington, D.C. in Latin America." World Bank, Washington, D.C. Deininger, Klaus, and Lyn Squire. 1996. "A New Dataset: Processed. Measuring Income Inequality." World Bank Economic Cunningham, Wendy, and Mauricio Santamarfa. 2003. Review 10 (3): 565-91. "Labor Markets as a Source of Income Inequality." World _ . 1998. "New Ways of Looking at Old Issues: Bank, Washington, D.C. Processed. Inequality and Growth."Journal of Development Economics Damiani, Octavio. 2003. "Effects on Employment, Wages, 57 (2): 257-85. and Labor Standards of Nontraditional Export Crops in De Janvry, Alain, and Elisabeth Sadoulet. 1989. "A Study Northeast Brazil." Latin American Research Review 38 (1): in Resistance to Institutional Change: The Lost Game of 83-112. Latin American Land Reform." World Development 17 (9): David, M., M. Dirven, and F. Vogelgesang. 2000. "The 1397-407. Impact of the New Economic Model on Latin America's _ . 2000. "Growth, Poverty and Inequality in Latin Agriculture." World Development 28 (9): 167 3-88. America: a Causal Analysis, 1970-1994." Review of Income Davidson, R., and J. Y Duclos. 2000. "Statistical Inference and Wealth 46 (3): 267-87. for Stochastic Dominance and for the Measurement of _ . 2002a. "Comments on Political and Equity Poverty and Inequality." Econometrica 68 (6): 1435-64. Aspects of Land Rights." Paper presented at the Regional Davies, J. B., and Anthony F. Shorrocks. 2000. "The Workshop on Land Issues in Latin America and the Distribution of Wealth." In A. B. Atkinson and F. Caribbean, Pachuca, Mexico (May 19-22). Bourguignon, eds., Handbook of Income Distribution. _ . 2002b. "Land Reforms in Latin America: Ten Amsterdam: North-Holland Publishing Company. Lessons toward a Contemporary Agenda." Paper presented Deaton, Angus. 1997. The Analysis of Household Surveys: at the Regional Workshop on Land Issues in Latin Microeconomic Analysisfor Development Policy. Washington, America and the Caribbean, Pachuca, Mexico (May D.C.: World Bank. 19-22). 359 INEQUALITY IN LATIN AMERICA _____. 2003. "Targeting and Calibrating Educational Di Tella, R., S. Galiani, and E. Schargrodsky. 2002. "Crime Grants: Focus on Poverty or on Non-Enrollment Risk?" Victimization and Income Distribution." Inter-American University of California, Berkeley. Processed. Development Bank, Washington, D.C. De la Torre, Augusto, Eduardo Levy Yeyati, and Sergio Diwan, Ishac. 2001. "Debt as Sweat: Labor, Financial Crises, Schmukler. 2003. "Living and Dying with Hard Pegs: and the Globalization of Capital." World Bank, The Rise and Fall of Argentina's Currency Board." Washington, D.C. Processed. Economia 3 (2): 43-102. _ . 2002. "The Labor Share During Financial Crises: De Luca, M., M. Jones, and M. Tula. 2002. "Back Rooms New Results." World Bank, Washington, D.C. Processed. or Ballot Boxes? Candidate Nomination in Argentina." Dollar, David, and Aart Kraay. 2002a. "Growth is Good for Comparative Political Studies 35 (4): 413-36. the Poor." Journal of Economic Growth 7: 195-225. De Luna-Martinez, Jose. 2000. "Management and Resolu- _ . 2002b. "Institutions, Trade and Growth." Carnegie- tion of Banking Crises: Lessons from the Republic of Rochester Conference Series on Public Policy. University Korea and Mexico." World Bank Discussion Paper 413. of Rochester Simon School of Business, Rochester, NY Washington, D.C.: World Bank. and Carnegie-Mellon University, Pittsburgh, PA. Demery, Lionel. 2003. "Analyzing the Incidence of Public Doner, R. 1992. "Limits of State Strength: Toward an Insti- Spending." In Bourguignon, FranIois and L. Pereira da tutional View of Economic Development." World Politics Silva, eds. The Impact of Economic Policies on Poverty and 44 (3): 398-431. Income Distribution: Evaluation Techniques and Tools. Dr6ze,Jean, and Amartya Sen. 1996. IndianDevelopment: Selected Washington, D.C.: World Bank. Regional Perspectives. New Delhi: Oxford University Press. Deruyttere, Anne. 2001. Pueblos indigenas, globalizacion y Dupor, Bill, and Wen-Fang Liu. 2003. "Jealousy and Equi- desarrollo con identidad: algunas reflexiones de estrategia. librium Overconsumption." American Economic Review 93 Washington, D.C.: Inter-American Development Bank. (1): 423-28. De Soto, Hernando. 1989. The Other Path: The Invisible Duryea, S., and C. Pages. 2002. "Human Capital Policies: Revolution in the Third World. New York: Harper and Row. What They Can and Cannot Do for Productivity and ___ . 2000. The Mystery of Capital. New York: Basic Books. Poverty Reduction in Latin America." Inter-American De Sousa, Santos B. 1998. "Participatory Budgeting in Porto Development Bank Working Paper 468. Washington, D.C. Alegre: Toward a Redistributive Democracy." Politics and Dziobec, Claudia, and Ceyla Pazarbasioglu. 1997. "Lessons Society 26 (4). from Systemic Bank Restructuring: A Survey of 24 Coun- Diaz-Alejandro, Carlos F. 1970. Essays on the Economic His- tries." International Monetary Fund Working Paper tory of the Argentine Republic. New Haven, CT: Yale 97/161. Washington, D.C. University Press. Easterly, William, and Stanley Fischer. 2001. "Inflation and Diaz-Cayeros, A. 1996. "Federal Resource Allocations the Poor." Journal of Money, Credit, and Banking 33 (2): Under a Dominant Party: Fiscal Transfers in Mexico." 160-78. Paper prepared for the 92nd American Political Science Easterly, William, and Ross Levine. 2002. "Tropics, Germs Association Meeting, San Francisco. and Crops: How Institutions Influence Economic Devel- Diaz-Cayeros, A., B. Magaloni, and B. Weingast. 2002. opment." Center for Global Development and Institute "Federalism and Democratization in Mexico." Paper pre- for International Economics, Washington, D.C. Processed. pared for the 2002 Annual Meeting of the American Easterly, William, and L. Serven. 2003. The Limits of Stabi- Political Science Association, Washington, D.C. lization: Infrastructure, Public Deficits and Growth in Latin Di Gropello, Emanuela. 2003. "Background Paper on Decen- America. Stanford, CA and Washington D.C.: Stanford tralization in Education for the Latin America and University Press and World Bank. Caribbean companion to the 2004 World Development Ebrill, Liam P., Michael Keen, Jean-Paul Bodin, and Vic- Report." World Bank, Washington, D.C. Processed. toria Summers. 2001. The Modern VAT Washington, D.C: DiNardo, John, Nicole Fortin, and Thomas Lemieux. 1996. International Monetary Fund. "Labor Market Institutions and the Distribution of Wages ECLAC (Economic Commission for Latin America and the 1973-1992: A Semiparametric Approach." Econometrica Caribbean). 1996. Anuario Estad/sticopara America Latina 64 (5): 1001-44. y el Caribe. Santiago: ECLAC. 360 B I BLi 0, RAP HY _ 2000. Panorama Social, 1999-2000. Santiago: Poor." Paper presented at the fifth meeting of the ECLAC. LACEA/IDB/World Bank network on inequality and ______ * 2001. Panorama Social 2000-2001. Santiago: poverty, Madrid. Also available from the Grupo de Anali- ECLAC. sis para el Desarrollo (GRADE), Lima, Peru. Edwards, John, and Donald Winkler. 2002. "Education, Eth- Eskeland, Gunnar S. and Filmer, Deon. 2000. "Does Decen- nicity, and Poverty: A Study with Applications to the tralization Improve Learning? Autonomy and Parental Maya of Guatemala." Paper presented at the meeting on Participation in Argentine Schools." World Bank, Etnicidad, Raza, Genero en America Latina, Grupo de Washington, D.C. Processed. Analisis para el Desarrollo, Lima, Peru. _ . 2002. "Autonomy, Participation and Learning in Elbers, Chris, Jean 0. Lanjouw, Peter Lanjouw, and Argentine Schools: Findings and their Implications for Phillippe G. Leite. 2003. "Poverty and Inequality in Decentralization." World Bank, Washington, D.C. Brazil: New Evidence from combined PPV-PNAD Processed. data." World Bank, Washington, D.C. Processed. Estache, Antonio. 2002. "Argentina's 1990s Utilities Elwan, Ann. 1999. "Poverty and Disability: A Survey of the Privatization: A Cure or a Disease?" World Bank, Literature," World Bank Social Protection Discussion Washington, D.C. Processed. paper 9932. Washington, D.C. _ . 2003. "On Latin America's Privatization and its Engel, Eduard, Alexander Galetovic, and Claudio Raddatz. Distributional Effects." World Bank, Washington, D.C. 1998. "Taxes and Income Distribution in Chile: Some Processed. Unpleasant Redistributive Arithmetic." National Bureau Estache, Antonio, Vivien Foster, and Quentin T. Wodon. of Economic Research Working Paper 6828. Cambridge, 2002. Accountingfor Poverty in Infrastructure Reform: Learn- MA. ing from Latin America's Experience. World Bank Develop- Engerman, Stanley L., and Kenneth L. Sokoloff. 1997. ment Studies Series. Washington, D.C.: World Bank. "Factor Endowments, Institutions, and Differential Paths Esteban, J., and D. Ray. 1994. "On the Measurement of of Growth among New World Economies." In Stephen Polarization." Econometrica 62 (4): 819-51. Haber, ed., How Latin America Fell Behind. Stanford, CA: Esteban, J., C. Gradin, and D. Ray. 1999. "Extension of a Stanford University Press. Measure of Polarization, with an Application to the In- .2000. "Institutions, Factor Endowments, and Paths come Distribution of Five OECD Countries." Working of Development in the New World." Journal of Economic Paper Series 24. La Corufia, Spain: Instituto de Estudios Perspectives 3: 217-32. Economicos de Galicia Pedro Barrie de la Maza. ____ . 2002. "Factor Endowments, Inequality, and Paths Evans, P. 1979. Dependent Development: The Alliance of Multi- of Development among New World Economies." national, State, and Local Capital in Brazil. Princeton, NJ: Economia 3: 41-109. Princeton University Press. Engerman, Stanley L., and Robert E. Gallman, eds. The Cam- _ . 1995. Embedded Autonomy: States and Industrial bridge Economic History of the United States. Volume 1: The Transformation. Princeton, NJ: Princeton University Colonial Era. New York: Cambridge University Press. Press. Engerman, Stanley L., Stephen Haber, and Kenneth L. Evans, P., B. Rueschemeyer, and T. Skocpol, eds. 1985. Sokoloff. 2000. "Inequality, Institutions, and Differential Bringing the State Back in. Cambridge: Cambridge Uni- Paths of Growth among New World Economies." In versity Press. Claude Menard, ed., Institutions, Contracts, and Organiza- Faguet, Jean-Paul. 2000. "Does Decentralization Increase tion. Cheltenham: Edward Elgar. Responsiveness to Local Needs? Evidence from Bolivia." Ennis, Huberto, and Santiago Pinto. 2002. "Privatization World Bank Policy Research Working Paper 2516. and Income Distribution in Argentina." In John Nellis Washington, D.C. and Nancy Birdsall, eds., A Glass Half Full: Assessing the Fajnzylber, Pablo, Daniel Lederman, and Norman Loayza. Distributional Impact of Privatization. Washington, D.C.: 1998. Determinants of Crime Rates in Latin America and Center for Global Development. the World: An Empirical Assessment. World Bank Latin Escobal, Javier, and Carmen Ponce. 2002. "The Benefits of America and Caribbean Studies: Viewpoints. Washington, Rural Roads: Enhancing Income Opportunities for the D.C.: World Bank. 361 INEQUALITY IN LATIN AMERICA ___ . 2000. "Crime and Victimization: An Economic Per- Filmer, Deon, Jeffrey S. Hammer, and Lant Pritchett. 2000. spective." Economfa 1 (1): 219-78. "Health Policy in Poor Countries: Weak Links in the Fay, Marianne, Danny Leipziger, Quentin Wodon, and Tito Chain." World Bank Research Observer 15 (2): 199-224. Yepes. 2003. "Achieving the Millennium Development . 2002. "Weak Links in the Chain II: A Prescription Goals: The Role of Infrastructure." World Bank Policy for Health Policy in Poor Countries." World Bank Research Research Working Paper 3163. Washington, D.C. Observer 17 (1): 47-66. Fay, Marianne, Tito Yepes, and Vivien Foster. 2002. "Asset Finan, Frederico, Elisabeth Sadoulet, and Alain de Janvry. Inequality in Developing Countries: The Case of 2002. "Measuring the Poverty Reduction Potential of Housing." World Bank, Washington, D.C. Processed. Land in Rural Mexico." University of California, Berkeley. Felix, D. 1982. "Income Distribution Trends in Mexico and Processed. the Kuznets Curves." In Hewlett and Weinert, eds., Brazil Fiszbein, A., P. Giovagnoli, and I. Aduriz. 2002. "Argentina's and Mexico: Patterns in Recent Development. Philadelphia: Crisis and Its Impact on Household Welfare." World Bank Institute for the Study of Human Issues. Working Paper. Washington, D.C. Fernandez, R., N. Gunerm, and J. Knowles. 2001. "Love and Fl6rez, Carmen Elisa, Carlos Medina, and Fernando Urrea. Money: A Theoretical and Empirical Analysis of House- 2001. "Understanding the Cost of Social Exclusion Due hold Sorting and Inequality." New York University, New to Race or Ethnic Background in Latin American and York. Processed. Caribbean Countries." Inter-American Development Ferreira, Francisco H. G. 2001. "Education for the Masses? The Bank, Washington, D.C. Interaction Between Wealth, Educational, and Political Food and Agriculture Organization and World Bank. 2003. Inequalities." Economics of Transition 9 (2): 533-52. La Nueva Ruralidad en Europa y su Interes para Am6rica Ferreira, Francisco H. G., and Julie Litchfield. 1996. "Grow- Latina. Rome: FAO. ing Apart: Inequality and Poverty Trends in Brazil in the Forbes, Kristin J. 2000. "A Reassessment of the Relation- 1980s." LSE STICERD, Distributional Analysis Research ship Between Inequality and Growth." American Economic Programme Discussion Paper 23. London. Review 90 (4): 869-87. . 1999. "Calm After the Storms: Income Distribu- Foster, James, and Amartya K. Sen. 1997. "On Economic tion and Welfare in Chile, 1987-94." The World Bank Inequality after a Quarter of a Century." Annex to Economic Review 13 (3): 509-38. Amartya K. Sen, On Economic Inequality (expanded edi- Ferreira, Francisco H. G., and Philippe George Leite. 2004. tion). Oxford, U.K.: Clarendon Press. "Educational Expansion and Income Distribution. A Foster, Vivien. 2002. "Ten Years of Water Service Reform in Micro-Simulation for Ceara." In R. van der Hoven, and Latin America: Towards an Anglo-French Model." In Paul A. Shorrocks, eds. Growth, Inequality and Poverty. New Seidenstat, David Haarmeyer, and Simon Hakim, eds., York: Oxford University Press. Reinventing Water and Wastewater Systems: Global Lessons for Ferreira, Francisco H. G., Giovanna Prennushi, and Martin ImprovingManagement. New York: John Wiley and Sons, Inc. Ravallion. 1999. "Protecting the Poor from Macroeconomic _ . 2003. "Background Paper on Utilities for the Latin Shocks: An Agenda for Action in a Crisis and Beyond." America and Caribbean Companion to the 2004 World World Bank Working Paper 2160. Washington, D.C. Development Report." World Bank, Washington, D.C. Fields, G. 1989. "A compendium of data on inequality and Processed. poverty for the Developing World." Cornell University, Fox, J. 1993. The Politics of Food in Mexico. Ithaca and London: Ithaca, NY Cornell University Press. Filgueira, F., and Papad6pulos, J. 1997. "Putting Conser- _ . 1994. "The Difficult Transition from Clientelism vatism to Good Use? Long Crisis and Vetoed Alternatives to Citizenship: Lessons from Mexico." World Politics 46: in Uruguay." In D. Chalmers, Carlos M. Vilas, Katherine 151-84. Hite, Scott B. Martin, Kerianne Piester, and Monique Foxley, Alejandro T., and Claudio Sapelli. 1999. "Chile's Segarra, eds., The New Politics of Inequality in Latin Political Economy in the 1990s: Some Governance America: Rethinking Participation and Representation. New Issues." In Guillermo Perry and Danny M. Leipziger, eds., York: Oxford University Press. Chile: Recent Policy Lessons and Emerging Challenges. World 362 BIBLIOGRAPHll Bank Institute Development Studies. Washington, D.C.: Galor, Oded, and Joseph Zeira. 1993. "Income Distribu- World Bank. tion and Macroeconomics." Review of Economic Studies 60: Frankel, Jeffrey, and Sergio Schmukler. 2000. "Country 35-52. Funds and Asymmetric Information." InternationalJour- Ganuza E., L. Taylor, and S. Morley. 1998. Politica macro- nal of Finance and Economics 5 (3): 177-95. economica y pobreza en America Latina y el Caribe. PNUD- Freije, Samuel, and Luis Rivas. 2002. "Privatization, Inequality, CEAPL-IDB. Madrid: Ediciones Mundi-Prensa. and Welfare: Evidence from Nicaragua. Instituto de Garcfa Aracil, Adela, and Donald Winkler. 2002. "Ethnic- Estudios Superiores Administraci6n." Caracas: IESA. ity, Gender, and Education in Ecuador." Paper presented Freire, Paulo. 1970. Pedagogy of the Oppressed. New York: The at the meeting on Etnicidad, Raza, Genero en America Continuum Publishing Company. Latina, Grupo de Analisis para el Desarrollo, Lima, Peru. Friedman, Jed, and James Levinsohn. 2002. "The Distribu- Garret6n, Antonio. 1989. The Chilean Political System. tional Impacts of Indonesia's Crisis on Household Welfare: Boston, MA: Unwin Hyman. A 'Rapid Response' Methodology." World Bank Economic Gasparini, Leonardo. 2002. "On the Measurement of Review 16 (3): 397-423. Unfairness: An Application to High School Attendance Friedman, Milton. 1956. Studies in the Quantity Theory of in Argentina." Social Choice and Welfare 19 (4): 795-810. Money. Chicago: University of Chicago Press. Gasparini, L., and W. Sosa Escudero. 2001. "Assessing Fry, P. 1999. "Color and the Rule of Law in Brazil." In J. aggregate welfare: growth and inequality in Argentina." Mendez, G. O'Donnell, and P. Pinheiro, eds., The Rule of Cuadernos de Economla 38 (113). Law and the Underprivileged in Latin America. Notre Dame, Gasparini, Leonardo, M. Marchionni, and W. Sosa Escudero. IN: University of Notre Dame Press. 2001. La distribuci6n del ingreso en la Argentina. Buenos Gacitua-Mari6, Estanislao, and Carlos Sojo, eds., with Aires: Editorial Triunfar. Shelton H. Davis. 2001. Social Exclusion and Poverty Reduc- Gates, Paul W. 1968. History of Public Land Law Development. tion in Latin America and the Caribbean. Washington, D.C.: Washington, D.C.: Zenger Publishing Company, Inc. World Bank. Gauri, Varun, and Ayesha Vawda. 2003. "Vouchers for Basic Gacitua-Mari6, Estanislao, and Quentin Wodon. Forth- Education in Developing Countries: A Principal-agent coming. "Social Exclusion in Latin America and the Perspective." World Bank Policy Research Working Caribbean." World Bank, Washington, D.C. Paper 3005. Washington, D.C. Galal, Ahmed, and MaryShirley, eds. 1994.DoesPrivatization Gavin, M., and R. Perotti. 1997. Fiscal Policy in Latin Deliver? Highlightsfroma WorldBank Conference. EDI Devel- America. Cambridge, MA: MIT Press for National opment Studies. Washington, D.C.: World Bank. Bureau of Economic Research. Galasso, Emanuela, and Martin Ravallion. 2003. "Decen- Gaviria, A., and C. Pages. 1999. Patterns of Crime Victim- tralized Targeting of an Anti-Poverty Program." World ization in Latin America. Washington, D.C.: Inter- Bank, Washington, D.C. Processed. American Development Bank. Galenson, David W. 1996. "The Settlement and Growth of Gelbach, Jonathan, and Lant H. Pritchett. 1997. "More for the Colonies: Population, Labor and Economic Develop- the Poor is Less for the Poor: The Politics of Targeting." ment." In Stanley L. Engerman and Robert E. Gallman, World Bank Policy Research Working Paper 1799. eds., The Cambridge Economic History of the United States. Washington, D.C. Volume 1: The Colonial Era. New York: Cambridge GenderStats database. http://devdata.worldbank.org/ University Press. genderstats/. World Bank, Washington, D.C. Galiani, Sebastian, and Pablo Sanguinetti. 2000. "Wage Gemmell, N., and 0. Morrissey. 2002. "The Impact of Inequality and Trade Liberalization: Evidence from Taxation on Inequality and Poverty: A Review of Empir- Argentina." Buenos Aires: Universidad Torcuato Di Tella. ical Methods and Evidence." Nottingham, U.K.: Galiani, Sebastian, Paul Gertler, and Ernesto Schargrodsky. University of Nottingham. 2002. "Water for Life: The Impact of the Privatization Gertler, P. J. 2000. Final report: The Impact of PROGRESA of Water Services on Child Mortality." Buenos Aires: on Health. Washington, D.C.: International Food Policy Universidad Torcuato di Tella. Research Institute. 363 INEQUALITY IN LATIN AMERICA Gillespie, S., and L. Haddad. 2001. "Attacking the Double GonzMlez, E. 2001. Paraguay: determinantes regionales en la Burden of Malnutrition in Asia and the Pacific." Asian inequidad de ingresos. ~Son desiguales las regiones del pa is? Development Bank Nutrition and Development Series, Asunci6n: Centro de AnAlisis y Difusi6n de Economia Paper 4. Manila. Also available from the International Paraguaya. Food Policy Research Institute, Washington, D.C. Gonzalez, Luis E. 1991. Political Structures and Democracy Glaeser, Edward, Jos6 Scheinkman, and Andres Schleifer. in Uruguay. Notre Dame, IN: Notre Dame University 2002. "The Injustice of Inequality." Journal of Monetary Press. Economics 50 (1): 199-224. Gottschalk, Peter and Enrico Spolaore. 2002. "On the Eval- Glewwe, Paul W., Margaret Grosh, Hanan Jacoby, and uation of Economic Mobility." Review of Economic Studies Marlaine Lockheed. 1995. "An Eclectic Approach to 69: 191-208. Estimating the Determinants of Achievements in Gottschalk, P., and T. Smeeding. 2000. "Empirical Evidence Jamaican Primary Education." World Bank Economic Review on Income Inequality in Industrial Countries." In Anthony 9 (2): 23 1-58. B. Atkinson and FranI;ois Bourguignon, eds., Handbook of Glewwe, P., and G. Hall. 1998. "Are Some Groups More Income Distribution. Amsterdam: North-Holland Publishing Vulnerable to Macroeconomic Shocks Than Others? Company. Hypothesis Tests Based on Panel Data from Peru. "Journal Governance Indicators Database. http://www.worldbank.orgl of Development Economics 56: 181-206. wbi/governance/govdata2002/. World Bank Institute, Goldin, C. 1987. "Work and Ideology in the Maya High- World Bank, Washington, D.C. lands of Guatemala: Economic Beliefs in the Context of Government Finance Statistics database. Statistics Depart- Occupational Change. " Economic Development and Cultural ment, International Monetary Fund, Washington, D.C. Change 41 (1): 103-23. Graham, Carol, and Stefano Pettinato. 2002. Happiness and Goldin, Claudia. 1999. "Egalitarianism and the Returns to Hardship: Opportunity and Insecurity in New Market Economies. Education during the Great Transformation of American Washington, D.C.: Brookings Institution Press. Education."Journal of Political Economy 107 (6): S65-94. Graham, Richard, Thomas E. Skidmore, Aline Helg, and ____ 2001. "The Human-Capital Century and Ameni- Alan Knight. 1990. The Ideas of Race in Latin America can Leadership: Virtues of the Past." Journal of Economic 1870-1 940. Austin: University of Texas Press. History 61 (2): 263-92. Graham, Richard. 1990. Patronage and Politics in Nineteenth- Goldin, Claudia, and Lawrence F. Katz. 1999. "The Returns Century Brazil. Stanford, CA: Stanford University Press. to Skill across the Twentieth Century United States." Grant, E. 2001. "Social Capital and Community Strategies: National Bureau of Economic Research Working Paper Neighborhood Development in Guatemala City." Devel- 7 126. Cambridge, MA. opment and Change 3 2 (5): 97 5-97. Gollin, D. 2002. "Getting Income Shares Right."Journal of Gray-Molina, George. 2002. "The Offspring of 1952: Political Economy 1 10 (2). Poverty, Exclusion and the Promise of Popular Participa- G6mez-Lobo, Andr6s and Dante Contreras. 2000. "Subsidy tion. " Conference paper prepared for "The Bolivian Rev- Policies for the Utility Industries: A Comparison of the olution at 50: Comparative Views on Social, Economic, and Chilean and Colombian Water Subsidy Schemes." Uni- Political Change." Harvard University, Cambridge, MA. versity of Chile, Department of Economics, Santiago. Grindle, Merilee S. 1977. Bureaucrats, Politicians, and Peas- Processed. ants in Mexico: A Case Study in Public Policy. Berkeley: Uni- Gonzaga, Gustavo M., Naercio Menezes Filho, and Cristina versity of California Press. Terra. 2002. "Trade Liberalization and Evolution of Skills _____. 2002. "Despite the Odds: Contentious Politics and Earnings Differentials in Brazil." Discussion Paper 463. Education Reform." Harvard University, Ca-mbridge, MA. Pontificia Universidade Cat6lica, Department of Eco- Processed. nomics. Rio de Janeiro. Grosh, Margaret. 1994. Administering Targeted Social Programs GonzMles-Vega, C., and V. Cespedes. 1993. "Costa Rica." In in Latin America: From Platitudes to Practice. Washington, Rottemberg, ed., The Political Economy of Poverty, Equity and D.C.: World Bank. Growth. Costa Rica and Uruguay. Washington, D.C.: World Haber, Stephen H. 2002a. Political Institutions and Banking Bank. Systems: Lessons from the Economic Histories of Mexico and the 3 64 BIBLIOGRAPHY United States, 1790-1914. Stanford, CA: Stanford Hartlyn, Jonathan. 1998. The Strugglefor Democratic Politics University Department of Political Science. in the Dominican Republic. Chapel Hill: University of _______ ed. 2002b. Crony Capitalism and Economic Growth North Carolina Press. in Latin America: Theory and Evidence. Stanford, CA: Hausman, R. and Rigobon, R., eds. 1993. Government Spend- University of California, Hoover Institution. ing and Income Distribution in Latin America. Washington, Haber, Stephen, Noel Maurer, and Armando Razo. 2003. The D.C.: Inter-American Development Bank. Politics of Property Rights: Political Instability, Credible Com- Heckman, James, and Carmen Pages-Serra. 2000. "The mitments, and Economic Growth in Mexico: 1876-1929. New Cost of Job Security Regulation: Evidence from Latin York: Cambridge University Press. American Labor Markets." Economia 1 (1): 109-54. Haddad, L., H. Alderman, S. Appleton, L. Song, and Y Heller, Patrick. 2001. "Moving the State: the Politics of Yohannes. 2002. "Reducing Child Undernutrition: How Democratic Decentralization in Kerala, South Africa, and Far Does Income Growth Take Us?" International Food Porto Alegre." Politics & Society 29 (1): 131-63. Policy Research Institute, Food Consumption Nutrition Hellman, Joel, and Daniel Kaufmann. 2003. "The Inequal- Division Discussion Paper. Washington, D.C. ity of Influence." World Bank, Washington, D.C. Hager, B. 2000. The Rule of Law: A Lexicon for Policy Makers. Processed. Missoula, MT: Mansfield Center for Pacific Affairs. Hentschel, J., J. 0. Lanjouw, Peter Lanjouw, and J. Poggi. Haggard, Stephan, and Robert Kaufman. 1990. "The Polit- 2000. "Combining Census and Survey Data to Trace the ical Economy of Inflation and Stabilization in Middle- Spatial Dimension of Poverty: A Case Study of Ecuador." Income Countries." World Bank Policy Research World Bank Economic Review 14: 147-65. Working Paper 444. Washington, D.C. Henriques, Ricardo. 2001. "Desigualdade Racial no Brasil: Haggard, Stephan, Chung H. Lee, and Sylvia Maxfield. Evolucao das Condic,6es de Vida na Decada de 90." 1993. The Politics of Finance in Developing Countries. Ithaca, Discussion paper 807. Instituto de Pesquisa Economica NY: Cornell University Press. Aplicada, Rio de Janeiro, Brasil. Hagopian, F. 1996. Traditional Politics and Regime Change in Hicks, N., and Q. Wodon. 2001. "Social Protection for the Brazil. Cambridge: Cambridge University Press. Poor in Latin America." ECLAC Review 73: 93-113. _____. 2000. "Political Development, Revisited." Com- Hiskey, J. 1999. "Does Democracy Matter? Electoral Com- parative Political Studies 33 (6/7): 800-911. petition and Local Development in Mexico." PhD diss. Halac, Marina, and Sergio L. Schmukler. 2003. "Distribu- University of Pittsburgh, Pittsburgh, PA. tional Effects of Crises: The Role of Financial Transfers." Hoddinott, J., E. Skoufias, and R. Washburn. 2000. "The Working Paper 3173. World Bank, Washington, D.C. Impact of PROGRESA on Consumption: A Final Hamann, A. Javier, and Alessandro Prati. 2002. "Why Do Report." Report submitted to PROGRESA. International Many Disinflations Fail? The Importance of Luck, Food Policy Research Institute, Washington, D.C. Timing, and Political Institutions." International Mon- Hoggarth, Glenn, Ricardo Reis, and Victoria Saporta. etary Fund Working Paper 02/228. Washington, D.C. 2002. "Costs of Banking System Instability: Some Empir- Handa, S., M.-C. Huerta, R. Perez, and B. Straffon. 2000. ical Evidence." Journal of Banking and Finance 26 (5): "Final Report: Poverty, Inequality, and 'Spill-over' in 825-5 5. Mexico's Education, Health, and Nutrition Program." Holston, J., and T. Caldeira. 1998. "Democracy, Law, and Report submitted to PROGRESA. Washington, D.C.: Violence: Disjunctions of Brazilian Citizenship." In International Food Policy Research Institute. Felipe Aguero and J. Stark, eds., Fault Lines of Democracy Harberger, Arnold C. 2003. "Reflections on Distributional in Post-Transition Latin America. Miami: North-South Considerations and the Public Finances." Prepared for Prac- Center Press. tical Issues of Tax Policy in Developing Countries (course), Holzer, Harry, and David Neumark. 2000. "What Does April 28-May 1, 2003. World Bank, Washington, D.C. Affirmative Action Do?" Industrial and Labor Relations Harrison, A. 2002. "Has Globalization Eroded Labor's Share? Review 53 (2), 240-70. Some Cross-Country Evidence." University of California, Honohan, Patrick, and Daniela Klingebiel. 2000. "Control- Berkeley, and National Bureau of Economic Research, ling the Fiscal Costs of Banking Crises." World Bank Cambridge, MA. Policy Research Working Paper 2441. Washington, D.C. 365 INEQUALITY IN LATIN AMERICA Hsieh, Chang-Tai, and Miguel Urquiola. 2003. "When Jayasuriya, R., and Q. Wodon. 2003. "Improving Health and Schools Compete, How Do They Compete? An Education Outcomes in Latin America: Spending More Assessment of Chile's Nationwide School Voucher Pro- or Spending Better?" In Q. Wodon, ed., "Public Spend- gram." National Bureau of Economic Research Working ing, Poverty, and Inequality in Latin America." World Paper 10008. Cambridge, MA. Bank, Washington, D.C. Processed. Hutchcroft, P. 1997. The Politics of Privilege: Assessing the Jimenez, Emmanuel. 2000. "Equity and the Decentraliza- Impact of Rents, Corruption, and Clientelism on Third tion of Social Services." In Shahid Javed Burki and World Development. Political Studies 45 (3): 639-58. Guillermo E. Perry, eds., Decentralization andAccountability Inter-American Development Bank. 1998. Facing Up to of the Public Sector. Proceedings of the 1999 Annual World Inequality in Latin America. Baltimore, MD: Johns Bank Conference on Development in Latin America and Hopkins University Press. the Caribbean. Washington, D.C.: World Bank. ___ . 2002. "The Privatization Paradox." Latin American Jimenez, Emmanuel, and Sawada, Y 1998. "Do Commu- Economic Policies Newsletter 18. nity School Managed Schools Work? An Evaluation of El ICRG (International Country Risk Guide). Various years. East Salvador's EDUCO Program." World Bank Economic Syracuse, NY: PRS Group, Inc. Review 13 (3): 415-41. IFPRI (International Food Policy Research Institute). 2001. Jones, M. 1993. "The Political Consequences of Electoral "Evaluation System for the Pilot Phase of Nicaraguan Laws in Latin America and the Caribbean." Electoral 'Red de Protecci6n,' Social Impact Evaluation." Report Studies 12 (1): 59-75. to the Red de Protecci6n Social, International Food Policy . 1994. "Presidential Election Laws and Multipar- Research Institute. Washington, D.C. tism in Latin America." Political Research Quarterly 47 (1): _____. 2002. "Evaluation System for the Pilot Phase of 41-57. the Nicaraguan Red de Protecci6n Social: Impact . 1995. Electoral Laws and the Survival of Presidential Evaluation." International Food Policy Research Institute. Democracies. Notre Dame, IN: University of Notre Dame Washington, D.C. Press. _____. 2003. "An Evaluation of Geographic Targeting in _ . 1997. "A Guide to the Electoral Systems of the Bolsa Alimentacao in Brazil." International Food Policy Americas: An Update." Electoral Studies 16 (1): 13-15. Research Institute. Washington, D.C. Jones M., P. Sanguinetti, and M. Tommasi. 2000. "Politics, ILO (International Labour Organization). 2000. World Labour Institutions, and Fiscal Performance in a Federal System: Report 2000. Geneva. an Analysis of the Argentine Provinces."Journal of Devel- Itzigsohn, Jose. 2000. Developing Poverty: The State, Labor opment Economics 61 (2): 305-33. Market Deregulation and the Informal Economy in Costa Rica Juhn, Chinhui, Kevin Murphy, and Brooks Pierce. 1993. and the Dominican Republic. University Park, PA: "Wage Inequality and the Rise in Returns to Skill." Pennsylvania State University Press. Journal of Political Economy 101 (3): 410-42. Jacobsen, Joyce. 2002. "What About Us? Men's Issues in Kabeer, Naila. 1994. Reversed Realities: Gender Hierarchies in Development." World Bank, Washington, D.C. Development Thought. London and New York: Verso Press. Processed. Kabeer, Naila, and R. Subrahmanian. 2000. Institutions, Jakee, K., and S. Turner. 2002. "The Welfare State as a Fiscal Relations and Outcomes. London: Zed Books. Commons: Problems of Incentives Versus Problems and Kanbur, Ravi. 2000. "Income Distribution and Development." Cognitions." Public Finance Review 30: 481-508. In Anthony B. Atkinson and Fran,cois Bourguignon, eds. Jalan, Jyotsna, and Martin Ravallion. 1999. "Income Gains Handbook of Income Distribution. Amsterdam: North-Holland to the Poor from Workfare: Estimates from Argentina's Publishing Company. Trabajar Program." World Bank Working Paper 2149. Karl, Terry Lynn. 1995. Paradox ofPlenty: OilBooms andPetro- Washington, D.C. States. Berkeley: University of California Press. . 2003. "Estimating the Benefit Incidence of an Katzenstein, PeterJ. 1984. Corporatism and Change: Austria, Antipoverty Program by Propensity Score Matching." Switzerland, and the Politics of Industry. Ithaca, NY Cornell Journal of Business and Economic Statistics 21 (1): 19-30. University Press. 366 B I 0 (( R A PH Y .1985. Small States in World Markets: Industrial Policy Knack, S., and P. Keefer. 1997. "Does Social Capital Have in Europe. Ithaca, NY: Cornell University Press. an Economic Payoff? A Cross-country Investigation." The Kaufman, R. 1974. "The Patron-Client Concept and QuarterlyJournal of Economics 112 (4): 1251-88. Macro-Politics: Prospects and Problems." Comparative Knaul, Felicia Marie. 2001. "The Impact of Child Labor and Studies in Society and History 16 (3): 284-308. School Dropout on Human Capital: Gender Differences in Kaufmann, Daniel, and Aart Kraay. 2002. "Growth with- Mexico." In Maria Correia and Elizabeth Katz, eds., The Eco- out Governance." Economia 3 (1): 169-229. nomics of Gender in Mexico. Washington D.C.: World Bank. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. Kohli, Atul. 1986. The State and Development in the Third 2003. Governance Matters III: Governance Indicators for World. Princeton, NJ: Princeton University Press. 1996-2002. Washington D.C.: World Bank Institute. Kohli, A., M. Altfeld, S. Lotfian, and R. Mardon. 1984. Kaufmann, Daniel, Francesca Recanatini, and Sergiy Bilet- "Inequality in the Third World: An Assessment of sky. 2002. "Assessing Governance: Diagnostic Tools and Competing Explanations." Comparative Political Studies 17: Applied Methods for Capacity Building and Action Learn- 259-68. ing." Discussion draft. World Bank, Washington, D.C. Korpi, W. 1980. "Social Policy and Distributional Conflict Keefer, Philip. 2002. "Clientelism, Credibility and in the Capitalist Democracies: A Preliminary Compara- Democracy." World Bank Development Research Group, tive Framework." West European Politics, 3 (3): 296-315. Washington D.C. Processed. Krueger, Alan. 2002. "Putting Development Dollars to Use Keefer, Philip, and Stephen Knack 2000. "Polarization, South of the Border." The New York Times (May 2). Politics, and Property Rights: Links Between Inequality Krugman, Paul 1979. "A Model of Balance of Payments and Growth." World Bank Policy Research Working Crises." Journal of Money, Credit, and Banking 11 (3): Paper 2418. Washington, D.C. 311-25. Kelley, Jonathan, and Herbert Klein. 1981. Revolution and _ . 1999. "Balance Sheets, the Transfer Problem and theRebirthoflnequality:ATheotyAppliedtotheNationalRevo- Financial Crises." In P. Isard, A. Razin, and A.K. Rose, lution in Bolivia. Berkeley: University of California Press. eds., International Finance and Financial Crises: Essays in Kenworthy, L. 2002. "Corporatism and Unemployment in Honor of Robert P. Flood. Dordrecht, The Netherlands: the 1980s and 1990s." American Sociological Review 67 (3): Kluwer Academic Publishers and Washington, D.C.: 367-88. International Monetary Fund. Key Indicators of the Labour Market database. http://www.ilo. Kurer, 0. 1993. "Clientelism, Corruption, and the Alloca- org/public/english/employment/strat/kilm/. Employment tion of Resources." Public Choice 77 (2): 259-73. Sector, International Labour Organization, Geneva. _ . 1996. "The Political Foundations of Economic Kim, Dae-Il, and Robert H. Topel. 1995. "Labor Markets Development Policies." Journal of Development Studies and Economic Growth: Lessons from Korea's Industrial- 32 (5): 645-68. ization, 1970-1990." In Richard B. Freeman and Kurtz, MarcusJ. 1999. "Social Origins of Central American Lawrence F. Katz, eds., Differences and Changes in Wage Democracy." Journal of InterAmerican Studies and World Structures. Chicago: University of Chicago Press. Affairs 41 (1): 87-96. King, Elizabeth M., and Berk Ozler. 1998. "What's Decen- LABORSTA database. http://laborsta.ilo.org/. International tralization Got to Do with Learning? The Case of Labour Office, Geneva, Switzerland. Nicaragua's Autonomy Reform." Series on Impact Laeven, Luc, and Giovanni Majnoni. 2002. "Loan Loss Pro- Evaluation of Education Reforms. World Bank Devel- visioning and Economic Slowdowns: Too Much, Too opment Economics Research Group Working Paper 9. Late?" World Bank Policy Research Working Paper 2749. Washington, D.C. Washington, D.C. Kitschelt, H. 2000. "Linkages between Citizens and Politi- Lam, David. 2001. "The Impact of Race on Earnings and cians in Democratic Polities." Comparative Political Studies Human Capital in Brazil, South Africa and the United 33 (6/7): 845-79. States." Paper presented at the meeting of the Network Klein, Hebert 1992. Bolivia: The Evolution of a Multi-Ethnic on Inequality and Poverty in LACEA, Montevideo, Society. Oxford: Oxford University Press. Uruguay. 367 INEQUALITY IN LATIN AMERICA Lambert, P. 1993. The Distribution and Redistribution of Income. Levy Yeyati, Eduardo, Sergio Schmukler, and Neeltje Van Manchester: Manchester University Press. Horen. 2003. "The Price of Inconvertible Deposits: The Lang, James. 1975. Conquest and Commerce; Spain and Eng- Stock Market Boom during the Argentine Crisis." land in the Americas. New York: Academic Press. World Bank, Washington, D.C. Processed. Lanjouw, Peter, and Martin Ravallion. 1995. "Poverty and Li, H. 2000. "Political Economy of Income Distribution: household size." EconomicJournal 105: 1415-34. A Comparative Study of Taiwan and Mexico." Policy _____. 1999. "Benefit Incidence, Public Spending StudiesJournal 28 (2): 275-291. Reforms, and the Timing of Program Capture." World Lindauer, David, and Lant Pritchett. 2002. "What's the Big Bank Economic Review 13: 257-73. Idea? Three Generations of Development Advice." La Porta, Rafael, and Florencio L6pez-de-Silanes. 1999. "The Economi'a 3 (1): 1-39. Benefits of Privatization: Evidence from Mexico." Quar- Lindert, P. 2002. "Why the Welfare State Looks Like a Free terly Journal of Economics 114 (4): 1193-242. Lunch." Working Paper 02-7. University of California, La Porta, Rafael, Florencio L6pez-de-Silanes, Andres Department of Economics. Davis, CA. Schleifer, and Robert Vishny. 1997. "Trust in large orga- Linz, J., and A. Valenzuela. 1994. The Failure of Presidential nizations." American Economic Revieu' 87 (2): 333-38. Democracy. Baltimore: Johns Hopkins University Press. Larrea, C. 1998. "Structural Adjustment, Income Distrib- Litchfield, Julie. 1999. "Inequality: Methods and Tools." ution, and Employment in Ecuador." In A. Berry, ed., Available online at America. Boulder, CO: Lynne Rienner Publishers. Lledo, V., A. Schneider, and M. Moore. 2003. Pro-poor Tax Latinobar6metro. Various years. Informe-Resumen: La Democ- Reform in Latin America: A Critical Survey and Policy racia y la Economfa. Santiago. Recommendations. Sussex: University of Sussex, Institute of .2003. Informe-Resumen: La Democracia y la Economla. Development Studies. Santiago. Loayza, Norman, Pablo Fajnzylber, and Cesar Calder6n. Laurell, A. 2001. "Health Reform in Mexico: The Promo- 2002. "Economic Growth in Latin America and the tion of Inequality." InternationalJournal of Health Services Caribbean: Stylized Facts, Explanations, and Forecasts." 31 (2): 291-321. World Bank, Washington, D.C. Processed. Leal, Victor Nunes. 1948. "Coronelismo, Enxada e Voto." Lokshin, Michael, and Martin Ravallion. 2000. "Welfare Rio de Janeiro: Revista Forense. Impacts of the 1998 Financial Crisis in Russia and the Leamer, Edward E., Hugo Maul, Sergio Rodriguez, and Peter Response of the Public Safety Net." Economics of Transi- K. Schott. 1999. "Does Natural Resource Abundance tion 8 (2): 269-95. Increase Latin American Income Inequality?" Journal of Londofio,Juan-Luis. 1995. Distribucidn dellngresoy Desarrollo Development Economics 59: 3-42. Economico: Colombia en el Siglo XX. Bogota: TM Editores. Lederman Daniel, Norman Loayza, and Ana Maria Menendez. Londonio, Juan-Luis, and Miguel Szekely. 1997. "Persistent 2002. "Violent Crime: Does Social Capital Matter?" Eco- Poverty and Excess Inequality: Latin America, 1970-1995." nomic Development and Cultural Change 50 (3): 509-39. Office of the Chief Economist, Working Paper series 375, Le Grand, J. 1991. Equity and Choice: An Essay in Economics and Inter-American Development Bank, Washington, D.C. Applied Philosophy. New York: Harper Collins Academic. L6pez, Humberto. 2003. "Macroeconomics and Inequality." Levenson, Alec R., and William E Maloney. 1998. "The World Bank, Washington D.C. Processed. Informal Sector, Firm Dynamics, and Institutional Par- L6pez, Ram6n, and Alberto Valdes. 2000a. "Fighting Rural ticipation." World Bank Policy Research Working Paper Poverty in Latin America: New Evidence of the Effects of 1988. Washington, D.C. Education, Demographics, and Access to Land." Economic Levine, Daniel H. 1973. Conflict and Political Change in and Development and Cultural Change 49 (1): 197-211. Venezuela. Princeton, NJ: Princeton University Press. , eds. 2000b. Rural Poverty in Latin America. New Levy, S. 1994. "La pobreza en Mexico." In E Velez, ed., York: St. Martin's Press. Causas ypoliticaspara combatir lapobreza en Mexico. Mexico L6pez-Acevedo, Gladys, and Angel Salinas. 2000. "How City: ITAM and FCE. Mexico's Financial Crisis Affected Income Distribution." 368 BIBLIOGRAPHY World Bank Policy Research Working Paper 2406. Unions." Inter-American Development Bank Working Washington, D.C. Paper 456. Washington D.C. ____ . 2002. "The Distribution of Mexico's Public McEwan, Patrick and Wilson Jimenez. 2002. "Estudiantes Spending on Education." World Bank, Washington, D.C. Indigenas y Acceso a la Educaci6n Primaria en Bolivia." Processed. Paper presented at the meeting on Etnicidad, Raza, L6pez-Calix, J., L. Alcazar, and E. Wachtenheim. 2003. Genero en America Latina, Grupo de Analisis para el "Improving the Efficiency of Public Spending in Peru." In Desarrollo, Lima, Peru. Q. Wodon, ed., "Public Spending, Poverty, and Inequal- MacIsaac, DonnaJ. and Harry Anthony Patrinos. 1995. "Labor ity in Latin America." World Bank, Washington, D.C. Market Discrimination against Indigenous People in Processed. Peru. "Journal of Development Studies 32 (2): 218-33. L6pez-Calva, Luis Felipe, and Juan Rosell6n. 2002. "Priva- McKenzie, David, and Dilip Mookherjee. 2003. "The Dis- tization and Inequality: The Mexican Case." Puebla, tributive Impact of Privatization in Latin America: An Mexico: Universidad de las Americas. Overview of Evidence from Four Countries." Economia L6pez-Maya, M. 1997. "The Rise of Causa R in Venezuela." 3 (2): 161-233. In D. Chalmers, Carlos M. Vilas, Katherine Hite, Scott Maddison, A. 1992. The Political Economy of Poverty, Equity B. Martin, Kerianne Piester, and Monique Segarra, eds. and Growth: Brazil and Mexico. New York: Oxford The New Politics of Inequality in Latin America: Rethinking University Press. Participation and Representation. New York: Oxford Mahoney, James. 2001. The Legacies of Liberalism: Path Depen- University Press. dence and Political Regimes in Central America. Baltimore: Lora, Eduardo. 2001. "Structural Reforms in Latin America: Johns Hopkins University Press. What Has Been Reformed and How to Measure It." Inter- Mahoney, James, and Vom Hau. 2002. "Indigenous American Development Bank Working Paper 348. People, Colonialism, and Social Development in Washington, D.C. Spanish America." Brown University, Providence, RI. Loury, Glenn C. 2002. The Anatomy of Racial Inequality. Processed. Cambridge, MA: Harvard University Press. Mainwaring, S. 2000. Rethinking Party Systems in the Third Lucas, Robert E. 1988. "On the Mechanics of Economic Wave of Democratization: The Case of Brazil. Stanford, CA: Development."Journal of Monetary Economics 22 (1): 3-42. Stanford University Press. Lustig, Nora. 1994. "Solidarity as a Strategy of Poverty Mainwaring, S., and M. Shugart. 1995. Building Democratic Alleviation." In Wayne Cornelius, Ann L. Craig, and Institutions: Party Systems in Latin America. Stanford, CA: Jonathan Fox, eds., Transforming State-Society Relations in Stanford University Press. Mexico: The National Solidarity Strategy. San Diego: Center _ . 1997. "Juan Linz, Presidentialism, and Democ- for US-Mexican Studies. racy: A Critical Appraisal." Comparative Politics 29 (4): ______ ed. 1995. Coping with Austerity: Poverty and Inequal- 449-471. ity in Latin America. Washington, D.C.: Brookings Malloy, J. 1977. Authoritarianism and Corporatism in Latin Institution. America. Pittsburgh, PA: University of Pittsburgh Press. .1999. Poverty and Inequality in Latin America. Santafe _ . 1979. The Politics of Social Security in Brazil. Pitts- de Bogota: Tercer Mundo S.A. burgh, PA: University of Pittsburgh Press. _____. 2000. "Crises and the Poor: Socially Responsible Malloy, James M., and Eduardo Gamarra. 1988. Revolution Macroeconomics." Economia 1 (1): 1-30. and Reaction: Bolivia, 1964-1985. New Brunswick, NJ: McAdam, D., S. Tarrow, and C. Tilly. 2001. Dynamics of Transaction Books. Contention. Cambridge: Cambridge University Press. Maloney, William E 2003. "Informality Revisited." World McClintock, Cynthia. 1998. Revolutionary Movements in Latin Bank, Washington D.C. Processed. America. El Salvador's FMLN and Peru's Shining Path. Maloney, William, and Jairo Nufiez. 2000. "Measuring the Washington D.C.: United States Institute of Peace Press. Impact of Minimum Wages: Evidence from Latin Maceira, Daniel, and Marfa Victoria Murillo 2001. "Social America." Working Paper 2597. World Bank, Wash- Sector Reform in Latin America and the Role of ington, D.C. Processed. 369 INEQUALITY IN LATIN ANIERICA Maluccio, J. 2003. "Education and Child Labor: Experi- Socioeconomica Nacional." Ministerio de Planificacion y mental Evidence from a Nicaraguan Conditional Cash Cooperaci6n. Santiago, Chile. Transfer Program." International Food Policy Research Milanovic, B. 2000. "The Median Voter Hypothesis, Institute, Washington, D.C. Processed. Income Inequality and Income Redistribution: An Manuelyan Atinc, Tamar, and Michael Walton. 1998. Social Empirical Test with the Required Data." European Jour- Consequences ofthe East Asian Financial Crisis. Washington, nal of Political Economy 16 (3): 367-410. D.C.: World Bank. .2002. "True World Income Distribution, 1988 and Marcel, Mario. 1999. "Effectiveness of the State and 1993: First Calculation Based on Household Surveys Development Lessons from the Chilean Experience." In Alone." EconomicJournal 112 (476): 51-92. Guillermo Perry and Danny M. Leipziger, eds., Chile: Recent Milanovic, B., and S. Yltzhaki. 2002. "Decomposing World Policy Lessons andEmergingChallenges. World Bank Institute Income Distribution: Does the World Have a Middle Development Studies. Washington, D.C.: World Bank. Class?" The Review of Income and Wealth 48 (2): 155-78. Martinez-Vazquez, J. 2001. "The Impact of Budgets on the Mills, J., and S. Zandvakili. 1997. "Statistical Inference via Poor: Tax and Benefit Incidence." Teaching Module 10, Bootstrapping for Measures of Inequality." Journal of prepared for the World Bank Institute. Washington, Applied Econometrics 12: 133-50. D.C.: World Bank. Miraftab, E 1997. "Flirting with the Enemy: Challenges Martz, John D. 1983. "Populist Leadership and the Party Faced by NGOs in Development and Empowerment." Caudillo: Ecuador and the CFP, 1962-8 1. " Studies in Com- Habitat International 21 (4): 361-75. parative International Development 18 (3): 22-49. MNPTSG (Mesoamerica Nutrition Program Targeting Matthews, M. 2001. "Institutions, Interests, Energy, and the Study Group). 2002. "Targeting Performance of Three Environment: Policymaking in Corporatism, Pluralism, Large-Scale Nutrition Oriented Social Programs in and Beyond." Policy Studies Journal 29 (3): 407-13. Central America and Mexico." Food and Nutrition Bulletin Meerman,J. 1980. "The Incidence of Sales and Excise Taxes, 23(2): 162-171. The United Nations University Press, or Where Do We Put the Transfers?" Journal of Political Tokyo. Economy 88 (6): 1242-48. Molina, O., and M. Rhodes. 2002. "Corporatism: The Past, Melo, Marcus Andr. 1998. "When Institutions Matter: The Present, and Future of a Concept." Annual Review of Polit- Politics of Administrative, Social Security, and Tax ical Science 5: 305-31. Reforms in Brazil." Paper delivered at the Latin American Molinar Horcasitas, Juan, and Jeffrey Weldon. 1994. Studies Association meeting, Chicago, Illinois. "Electoral Determinants and Consequences of National Mendez, Juan E., Guillermo O'Donnell, and Paulo Sergio Solidarity." In Wayne Cornelius, Ann Craig, and Jonathan Pinheiro, eds. 1999. The (Un)Rule of Law and the Under- Fox, eds., Transforming State-Society Relations in Mexico: privileged in Latin America. Notre Dame, IN: University The National Solidarity Strategy. La Jolla: Center for US- of Notre Dame Press. Mexican Studies, University of California, San Diego. Menezes-Filho, Naercio. 2001. "Educa,co e desigualdade." Mondino, Guillermo, Esteban Fernandez Medrano, and In M. Lisboa and N. Menezes-Filho, eds., Microeconomia Luciano Laspina. 2002. "Argentina's Neopopulism e Sociedade no Brasil. Sao Paolo: Editora Contracapa. Voyage." Latin Source Monthly Report (December). Menocal, A. R. 2001. "Old Habits Die Hard? A Statistical Morgan, Edmund S. 1975. American Slavery, American Exploration of the Politicisation of PROGRESA, Mexico's Freedom: The Ordeal of Colonial Virginia. New York: W.W. Latest Federal Poverty-Alleviation Programme Under the Norton and Company. Zedillo Administration."Journal of Latin American Studies Morley, S. 2001. The Income Distribution Problem in Latin 33 (3): 513-38. America and the Caribbean. Santiago: Economic Commis- Middlebrook, Kevin. 1995. The Paradox of Revolution: Labor, sion for Latin America and the Caribbean. the State. and Authoritarianism in Mexico. Baltimore: Johns Morley, S., and D. Coady. 2003. From Social Assistance to Social Hopkins University Press. Development: A Revieu' of Targeted Education Subsidies in MIDEPLAN (Ministerio de Planificaci6n y Cooperaci6n). Developing Countries. Washington, D.C.: International 1998. "Resultados de la VII Encuesta de Caracterizacion Food Policy Research Institute. 370 BIBLIOGRAPHY Morley, S., and R. Vos. 1997. Poverty and Dualistic Growth Oaxaca, R.L. 1973. "Male-Female Wage Differentials in in Paraguay. Washington, D.C.: Inter-American Develop- Urban Labor Markets." International Economic Review ment Bank. 14 (3): 693-709. Morrison, Christian. 2000. "Inequality in Historical Obstfeld, M. 1996. "Models of Currency Crises with Self- Perspective." In Anthony B. Atkinson and Franc,ois fulfilling Features." European Economic Review 40: 1037-47. Bourguignon, eds., Handbook of Income Distribution. Ocampo, J. A., Maria Jose Perez, Camilo Tovar, and Francisco Amsterdam: North-Holland Publishing Company. Lasso. 1998. "Macroeconomia, ajuste estructural y Murillo, Maria Victoria. 1997. "Union Politics, Market- equidad en Colombia: 1978-1996." In Ganuza, Taylor, Oriented Reforms, and the Reshaping of Argentine and Morley, eds., Politica macroecon6mica y pobreza en Corporatism." In D. Chalmers, Carlos M. Vilas, Katherine Amirica Latina y el Caribe. New York: United Nations Hite, Scott B. Martin, Kerianne Piester, and Monique Development Programme. Segarra, eds., The New Politics of Inequality in Latin O'Donnell, Guillermo. 1973. Modernization and Bureaucratic- America: Rethinking Participation and Representation. New Authoritarianism: Studies in South American Politics. York: Oxford University Press. Berkeley: University of California Press. ____ . 2002. "Political Bias in Policy Convergence: . 1998a. "Horizontal Accountability in New Democ- Privatization Choices in Latin America." World Politics racies." Journal of Democracy 9 (3): 112-26. 54 (4): 462-93. . 1998b. "Poverty and Inequality in Latin America: Murrugarra, E. 2003. "Targeted Social Assistance Programs Some Political Reflections." In V. Tokman and G. O'Don- in Uruguay." In Quentin Wodon, ed., Public Spending, nell, eds. Poverty and Inequality in Latin America: Issues and Poverty, and Inequality in Latin America. Washington, D.C.: New Challenges. Notre Dame, IN: The University of Notre World Bank. Dame Press. Musgrave, R. 1959. The Theory of Public Finance. New York: _ . 1999a. "Democratic Theory and Comparative Pol- McGraw-Hill Book Company. itics." University of Notre Dame, Kellogg Institute of Neri, Marcelo, and Jose Marcio Camargo. 1999. "Structural International Studies. Notre Dame, IN. Reforms, Macroeconomic Fluctuations and Income . 1999b. "Polyarchies and the (Un)Rule of Law in Distribution in Brazil." Reformas Econ6micas series 39. Latin America." In Juan E. M6ndez, Guillermo Economic Commission for Latin America and the O'Donnell, and Paulo Sergio Pinheiro, eds., The (Un)Rule Caribbean, Santiago. of Law and the Underprivileged in Latin America. Notre Neri, Marcelo, Gustavo Gonzaga, and Jose Marcio Camargo. Dame, IN: University of Notre Dame Press. 2001. "Salario Minimo, Efeito-Farol e Pobreza." Revista Oliveira Barbosa, Maria Ligia de. 2002. "Gender and Color de Economia Politica 21 (2): 78-90. Sao Paulo. Differences at School in Brazil: Teachers and Students' Nielson, D., and M. Shugart. 1999. "Constitutional Change Evaluations." (Unpublished manuscript.) in Colombia: Policy Adjustment through Institutional Olson, M. 1982. The Rise and Decline of Nations: Economic Reform." Comparative Political Studies 32 (3): 313-41. Growth, Stagflation, and Social Rigidities. New Haven, CT: North, Douglass C. 1990. Institutions, Institutional Change, Yale University Press. and Economic Performance. Cambridge: Cambridge Uni- OECD (Organisation for Economic Co-operation and Devel- versity Press. opment). 2001. Knowledge and Skills for Life: First Results Nozick, R. 1974. Anarchy, State, and Utopia. Oxford: Basil from the OECD Programme for International Student Assess- Blackwell Publishing. ment (PISA) 2000, Paris. Nylen, W. 1997. "Reconstructing the Workers' Party (PT): O'Rourke, Kevin H., and Jeffery G. Williamson. 2002. From Lessons from Northeastern Brazil." In D. Chalmers, Malthus to Ohlin: Trade, Growth and Distribution Since Carlos M. Vilas, Katherine Hite, Scott B. Martin, 1500. National Bureau of Economic Research Working Kerianne Piester, and Monique Segarra, eds., The New Paper 8955. Cambridge, MA. Politics of Inequality in Latin America: Rethinking Partici- Ortega, Daniel, and Francisco Rodriguez. 2002. "Openness pation and Representation. New York: Oxford University and Factor Shares." Office of Economic and Financial Press. Advisors, National Assembly, Caracas, Venezuela. 371 INEQUALITY IN LATIN AMERICA Paes De Barros, Ricardo, and Rosane Mendon,a. 1997. "A Perotti, Roberto. 1996. "Growth, Income Distribution, and Absor,ao de Mao-de-Obra no Setor de Servic,os." Dados: Democracy: What the Data Say." Journal of Economic Revista de Ciencias Sociais 40 (1). Growth 1: 149-87. _____. 1998. "The Impact of Three Institutional Innova- Perry, Guillermo. 2003. "Can Fiscal Rules Help Reduce tions in Brazilian Education." In W. D. Savedoff, ed., Orga- Macroeconomic Volatility in the Latin America and nization matters: Agency Problems in Health and Education. Caribbean Region?" World Bank Working Paper 3080. Washington, D.C.: Inter-American Development Bank. Washington, D.C. Paes de Barros, Ricardo, and M. Foguel. 2000. "Focaliza,cao Perry, Guillermo, and Luis Serven. 2002. "La anatomia de dos Gastos Publicos Sociais e erradica,cao da pobreza no una crisis multiple: que tenia Argentina de especial y que Brasil." In Ricardo Henriques, ed., Desigualdade e Pobreza podemos aprender de ella." Desarrollo Econdmico 42 (167): no Brasil. Rio de Janeiro: IPEA. 323-75. Paes de Barros, Ricardo, Rosane Mendonc,a, D. Santos, and Persson, Torsten, and Guido Tabellini. 1994. "Is Inequality G. Quintaes. 2001. "Determinantes do desempenho edu- Harmful for Growth? Theory and Evidence." American cacional noBrasil."PesquisaePlanejamentoEconomico 31 (1). Economic Review 84 (3): 600-21. Page, B., and Simmons J. 2000. What Government Can Do: Piketty, Thomas, and Emmanuel Saez. 2001. "Income Dealing with Poverty and Inequality. Chicago: University Inequality in the United States, 1913-1988." National of Chicago Press. Bureau of Economic Research Working Paper 8467. Paige, Jeffrey M. 1997. Coffee and Power: Revolution and the Cambridge, MA. Rise of Democracy in Central America, Cambridge, MA: Pinheiro, P. 1997. "Popular Responses to State-Sponsored Harvard University Press. Violence in Brazil." In D. Chalmers, Carlos M. Vilas, Parker, S., and E. Skoufias. 2000. "The Impact of PRO- Katherine Hite, Scott B. Martin, Kerianne Piester, and GRESA on Work, Leisure, and Time Allocation." Inter- Monique Segarra, eds., The New Politics of Inequality in national Food Policy Research Institute, Washington, D.C. Latin America: Rethinking Participation and Representation. Patrinos, Harry Anthony. 1997. "Differences in Education New York: Oxford University Press. and Earnings across Ethnic Groups in Guatemala." Quar- . 1999. "Introduction: The Rule of Law and the terly Review of Economics and Finance 37 (4): 809-21. Underprivileged in Latin America." In J. Mendez, _____. 2002. A Review of Demand-side Financing Initiatives G. O'Donnell, and P. Pinheiro, eds., The Rule of Law and in Education. Washington, D.C.: World Bank. the Underprivileged in Latin America. Notre Dame, IN: Patrinos, Harry Anthony, and Eduardo Velez. 1996. "Costs University of Notre Dame Press. and Benefits of Bilingual Education in Guatemala: A Par- Pirez, P. 2002. "Buenos Aires: Fragmentation and tial Analysis." World Bank Working Paper HCDWP 74. Privatization of the Metropolitan City." Environment and Washington, D.C. Urbanization 14 (1): 145-58. Patrinos, Harry Anthony, Eduardo Velez, and George Pissarides, Christopher A. 1990. Equilibrium Unemployment Psacharopoulos. 1994. "Language, Education, and Earn- Theory. Oxford, U.K.: Basil Blackwell Publishing. ings in Asunci6n, Paraguay." Journal of Developing Areas Political Risk Service Group, Inc. 2003. International 29: 57-68. Country Risk Guide. East Syracuse, NY: The PRS Patroni, V. 2001. "The Decline and Fall of Corporatism? Group, Inc. Labour Legislation Reform in Mexico and Argentina POLITY IV database. http://www.cidcm.umd.edu/inscr/ during the 1990s." CanadianJournal of Political Science polity/. Integrated Network for Societal Conflict Research 34 (2): 249-74. (INSCR) program at the Center for International Devel- Paxson, Christina, and Norbert Schady. 2002. "The Alloca- opment and Conflict Management (CIDCM), University tion and Impact of Social Funds: Spending on School of Maryland, College Park, MD. Infrastructure in Peru." World Bank Economic Review 16(2): Pontusson, J, D. Rueda, and C. Way. 2002. "Comparative 297-310. Political Economy of Wage Distribution: The Role of Pero, Valeria. 2003. "Intergenerational Social Mobility in Partisanship and Labour Market Institutions." British Brazil." Instituto de Economia-UFRJ, Rio de Janeiro. Journal of Political Science 32: 281-308. 372 BIBLIOGRAPHY Portes, Alejandro, and Ruben G. Rumbaut. 2001a. Legacies: Quintero, Rafael. 1980. El Mito del Populismo en el Ecuador. The Story of the Immigrant Second Generation. Berkeley: Quito: Editorial Universitaria. University of California Press. Randall, LauraJ. 1978. An Economic History ofArgentina in the _______ 2001b. Ethnicities: Children of Immigrants in America. Twentieth Century. New York: Columbia University Press. Berkeley: University of California Press. Rao, Vijayendra, and Michael Walton. 2004. "Introduction." Portes, A., and K. Hoffman. 2003. "Latin American Class In Rao, Vijayendra and Michael Walton, eds. Culture and Structures: Their Composition and Change During the Public Action. Stanford, CA: Stanford University Press. Neoliberal Era." Latin American Research Review 38 (1): Ratliff, W., and E. Buscaglia. 1997. "Judicial Reform: The 41-82. Neglected Priority in Latin America." Annals of the Powell, John Duncan 1970. "Peasant Society and Clientelistic American Academy of Political Science 550: 59-71. Practices." American Political ScienceReview 64(2): 411-25. Ravallion, Martin. 1997. "Can High-inequality Developing _ . 1977. "Peasant Society and Clientelist Politics." In Countries Escape Absolute Poverty?" Policy Research Steffen W. Schmidt,James C. Scott, Carl Lande, and Laura Working Paper 1775. World Bank, Washington, D.C. Guasti, eds., Friends, Followers, and Factions: A Reader in _ . 2000. "Are the Poor Protected from Budget Cuts? Political Clientelism. Berkeley: University of California Press. Theory and Evidence from Argentina." Policy Research Powers, Nancy R. 2001. Grassroots Expectations of Democracy Working Paper 2391. World Bank, Washington, D.C. and Economy: Argentina in Comparative Perspective. (Also forthcoming in theJournal of Applied Economics.) Pittsburgh, PA: University of Pittsburgh Press. _ . 2002. "Who Is Protected? On the Incidence of Fiscal Prillaman, W. 2000. The Judiciary and Democratic Decay in Adjustment." World Bank, Washington, D.C. Processed. Latin America: Declining Confidence in the Rule of Law. Ravallion, Martin, and Shaohua Chen. 1997. "What Can New Westport, CT: Praeger Press. Survey Data Tell Us about Recent Changes in Distribution Pritchett, Lant. 2001. "Where Has All the Education Gone?" and Poverty?" World Bank EconomicReview 11 (2): 357-82. World Bank Economic Review 15 (3): 367-9 1. Rawlings, Laura, and Gloria M. Rubio. 2003. "Evaluating Prud'homme, R. 1990. "Decentralization of Expenditures the Impact of Conditional Cash Transfer Programs: or Taxes: The Case of France." In R. J. Bennett, ed., Lessons from Latin America." World Bank Policy Decentralization, Local Governments, and Markets. Oxford, Research Working Paper 3119. Washington, D.C. U.K.: Clarendon Press. Rees, H., and Shah, A. 1986: "An Empirical Analysis of Self- Psacharopoulos, George, and ZafirisTzannatos. 1992. Women's employment in the UK." Journal of Applied Econometrics Employment and Pay in Latin America. Washington, 1 (1): 95-108. D.C.: World Bank. Rehren, A. 1996. "Corruption and Local Politics in Chile." Psacharopoulos, George, and Harry Patrinos. 1994. Indige- Crime Law and Social Change 25 (4): 323-34. nous People and Poverty in Latin America: An Empirical Reich, G. M. 1998. "The 1988 Constitution a Decade Later: Analysis. Washington, D.C.: World Bank. Ugly Compromises Reconsidered." Journal of Inter- Psacharopoulos, George, Samuel Morley, Ariel Fiszbein, American Studies and World Affairs 40 (4): 5-24. Haeduck Lee, and William C. Wood. 1995. "Poverty and Reis, Elisa. 1999. "Elite Perceptions of Poverty: Brazil." IDS Income Inequality in Latin America during the 1980s." Bulletin 30 (2). Review of Income and Wealth 41 (3): 245-64. Reis, E. J., P. Tafner, and L. 0. Reiff. 2001. "Distribuicao de _____. 1997. Poverty and Income Distribution in Latin Riqueza Imobiliaria e de Renda no Brasil: 1992-1999." America: The Story of the 1980s. Technical Paper 351. Seminarios DIMAC N 75. IPEA, Rio deJaneiro. Washington, D.C.: World Bank. Remmer, K., and E. Wibbels. 2000. "The Subnational Putnam, R. 1993. Making Democracy Work: Civic Traditions Politics of Economic Adjustment: Provincial Politics and in Modern Italy. Princeton, NJ: Princeton University Press. Fiscal Performance in Argentina." Comparative Political Putnam, R. D., S. J. Pharr, and R. J. Dalton. 2000. "Intro- Studies 33 (4): 419-5 1. duction: What's Troubling the Trilateral Democracies?" Rey, Juan Carlos. 1990. "El Papel de los Partidos en la In R. D. Putnam and S. Pharr, eds., Disaffected Democra- Instauraci6n y el Mantenimiento de la Democracia en cies. Princeton, NJ: Princeton University Press. Venezuela." Fundaci6n IDEA, Caracas. Processed. 373 INEQUALITY IN LATIN AMERICA Reynolds, Clark W. 1970. The Mexican Economy: Twentieth- Latin America. Notre Dame, IN: University of Notre Century Structure and Growth. New Haven, CT: Yale Dame Press. University Press. Rodrik, Dani. 1999. "Where Did All the Growth Go? Exter- Rinne, Jeffrey James. 2001. "Redesigning the State in Latin nal Shocks, Social Conflict and Growth Collapses."Journal America: Pundits, Policymakers, and Organized Labor in of Economic Growth 4 (4): 385-412. Argentina and Brazil." PhD diss. Princeton University. _ . 2003. "Growth Strategies." Paper prepared for the Princeton, NJ. Handbook on Economic Growth. Cambridge, MA: Harvard Rivera, J.A., G. Rodriguez, T. Shamah, J. L. Rosado, E. University Press. Casanueva, I. Maulen, G. Toussaint, and A. Garcia- Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. Aranda. 2000. "Implementation, Monitoring, and 2002. "Institutions Rule: The Primacy of Institutions Evaluation of the Nutrition Component of the Mexican over Geography and Integration in Economic Develop- Social Programme (PROGRESA)." Food and Nutrition ment." Harvard University, Cambridge, MA. Processed. Bulletin 21 (1): 35-42. Roemer, John E. 1996. Theories of Distributive Justice. Riveros, Luis 1998. "Chile's Structural Adjustment: Rele- Cambridge, MA: Harvard University Press. vant Policy Lessons for Latin America." In A. Berry, ed., _ . 1998. Equality of Opportunity. Cambridge, MA: Poverty, Economic Reform, and Income Distribution in Latin Harvard University Press. America. Boulder, CO: Lynne Rienner Publishers. Rojas, F. 2000. "The Political Context of Decentralization Robbins, Donald. 1994. "Relative Wage Structure in Chile, in Latin America." In ShahidJaved Burki and Guillermo 1957-1992: Changes in the Structure of Demand for E. Perry, eds., Decentralization and Accountability of the Schooling." Estudios de Economica 21: 49-78 Public Sector. Washington, D.C.: World Bank. Robbins, Donald, and T. H. Gindling. 1999. "Trade Liber- Rosado,J. L.,J. Rivera, G. Lopez, and L. Solano. 2000. "Devel- alization and the Relative Wages for More Skilled Work- opment, Production, and Quality Control of Nutritional ers in Costa Rica." Review of Development Economics 3 (2): Supplements for a National Supplementation Program in 140-54. Mexico." Food and Nutrition Bulletin 21 (1): 30-34. Roberts, K. 1995. "Neoliberalism and the Transformation Rose, R., and D. Shin. 2001. "Democratization Backwards: of Populism in Latin America: The Peruvian Case." World The Problem of Third-Wave Democracies." British Jour- Politics 48 (1): 82-116. nal of Political Science 31 (2): 331-54. _ . 1997. "Rethinking Economic Alternatives: Left Rothschild, Emma. 2001. Economic Sentiments: Adam Smith, Parties and the Articulation of Popular Demands in Chile Condorcet. and the Enlightenment. Cambridge, MA, and and Peru." In D. Chalmers, Carlos M. Vilas, Katherine London: Harvard University Press. Hite, Scott B. Martin, Kerianne Piester, and Monique Rueschemeyer, Dietrich, Evelyne Huber Stephens, and John Segarra, eds., The New Politics of Inequality in Latin Amer- D. Stephens. 1992. Capitalist Development and Democracy. ica: Rethinking Participation and Representation. New York: Chicago: University of Chicago Press. Oxford University Press. Rutherford, T., M. Light, and F. Barrera. 2002. Equity and Robinson, James. 2002. "Politician-Proof Policy?" Back- Efficiency Costs of Raising Tax Revenue in Colombia. Bogota: ground paper for the 2004 World Development Report. Fedesarrollo. Washington, D.C.: World Bank. Saavedra, Jaime, and Maria de los Angeles Cardenas. 2002. Robinson, James, and Thierry Verdier. 2002. "The Politi- "Access, Efficiency, and Equity of the Peruvian Educa- cal Economy of Clientelism." Centre for Economic Policy tional System: An Analysis of Indicators Using the LSMS Research Working Paper 3205. London. and the School Census." Paper presented at the meeting Robles, M. 1999. "Pobreza y distribuci6n del ingreso en on Etnicidad, Raza, Genero en America Latina, Grupo de Paraguay." Asunci6n: Programa MECOVI-DGEEC. Analisis para el Desarrollo, Lima, Peru. Rodley, N. 1999. "Torture and Conditions of Detention in Sabate, R., and A. Schneider. 2003. "Taxation, Account- Latin America." In J. Mendez, G. O'Donnell, and P. ability, and the Poor." University of Sussex, Institute of Pinheiro, eds., The Rule of Law and the Underprivileged in Development Studies Seminar Report. Sussex. 374 B I B LI OG R APHY Sahn, David, and D. Stifel. 2003. "Progress Towards the and Institutions in Comparative Politics). Ann Arbor: Uni- Millennium Development Goals in Africa." World versity of Michigan Press. Development 31 (1): 23-52. Scheper-Hughes, Nancy. 1992. Death Without Weeping: The Sahn, E. David, Paul Dorosh, and Stephen Younger. 1997. Violence of Everyday Life in Brazil. Berkeley: University of Structural Adjustment Reconsidered: Economic Policy and California Press. Poverty in Africa. Cambridge, U.K., and New York: Schiff, Maurice. 2003. "The Inefficiency of Inequality, with Cambridge University Press. Implications for Development Institutions." World Safa, Helen. 1998. "Introduction." Latin American Perspec- Bank, Washington, D.C. Processed. tives 100 (25): 3-20. Schiff, Maurice and Alberto Valdes. 1992. The Plundering of Sala-i-Martin, X. 2002. "The World Distribution of Income Agriculture in Developing Countries. Washington D.C.: (Estimated from Individual Country Distributions)." World Bank. National Bureau of Economic Research Working Paper Schmidt, Steffen W. 1977. "The Transformation of Clien- 8933. Cambridge, MA. telism in Rural Colombia." In Steffen W. Schmidt, James Salinas de Gortari, C. 1978. "Political Participation, Public C. Scott, Carl Lande, and Laura Guasti, eds., Friends, Investment, and System Support: A Study of Three Rural Followers, and Factions: A Reader in Political Clientelism. Communities in Central Mexico." PhD diss. Harvard Berkeley: University of California Press. University, Cambridge, MA. Schmitter, Philippe C., and Wolfgang Streeck, eds. 1971. Samuels, D. 2002. "Pork Barreling is not Credit Claiming or Interest Conflict and Political Change in Brazil. Stanford, CA: Advertising: Campaign Finance and the Sources of the Per- Stanford University Press. sonal Vote in Brazil."Journal of Politics 64 (3): 845-63. _ . 1974. "Still the Century of Corporatism?" Review Samuels, David, and Richard Snyder. 2001. "Devaluing the of Politics 36 (1): 85-131. vote in Latin America."JournalofDemocracy 12(1): 146-59. _ . 1985. Private Interest Government and Public Policy. . 2002. "The Value of a Vote: Malapportionment Beverly Hills and London: Sage Publications. in Comparative Perspective." British Journal of Political Schneider, B. 1991. Politics within the State: Elite Bureaucrats Science 31 (4): 651-71. and Industrial Policy in Authoritarian Brazil. Pittsburgh, Sanchez-Paramo, Carolina, and Norbert R. Schady. 2002. PA: University of Pittsburgh Press. "Off and Running? The Rising Demand for Skilled _ . 2002. "Why is Mexican business so organized?" Workers in Latin America." World Bank Policy Research Latin American Research Review 37 (1): 77-118. Working Paper 3015. Washington, D.C. Schultz, T. P. 2000a. "School Subsidies for the Poor: Sanguinetti, Pablo, Rodrigo Arim, and Juan Pantano. 2001. Evaluating a Mexican Strategy for Reducing Poverty." "Changes in Production Structure and Employment International Food Policy Research Institute, Food Structure and Relative Wages in Argentina and Uruguay." Consumption Nutrition Division Discussion Paper 102. Universidad Torcuato Di Tella, Buenos Aires. Washington, D.C. Santamarfa, Mauricio. 2000. "External Trade, Skill, _ . 2000b. "Final Report: The Impact of PROGRESA Technology and the Recent Increase of Income Inequal- on School Enrollments." International Food Policy ity in Colombia." PhD diss. Georgetown University, Research Institute, Washington, D.C. Washington, D.C. _ . 2000c. "Impact of PROGRESA on School Atten- Scalon, M. C., and C.A.C. Ribeiro. 2001. "Mobilidade de dance Rates in the Sampled Population." International classe em perspectiva comparada." Dados: Revista de Food Policy Research Institute, Washington, D.C. Cie'ncias Sociais 44 (1). _ . 2000d. "Preliminary Evidence of PROGRESA's Schady, N. 2000. "The Political Economy of Expenditures by Impact on School Enrollments from 1997/98 to 1998/99." the Peruvian Social Fund (FONDCODES) 1991-1995." International Food Policy Research Institute, Washington, American Political Science Review 94 (2): 289-304. D.C. Schamis, Hector E. 2002. Re-Forming the State: The Politics of Scobie, James. 1964. Revolution on the Pampas: A Social History Privatization in Latin America and Europe (Interests, Identities, of Argentine Wheat. Austin, TX: Texas University Press. 375 INEQUALITY IN LATIN ANMERICA Scott, John. 2003. "Public Spending and Inequality of Shome, Parthasarathi. 1999. "Taxation in Latin America: Opportunities in Mexico: 1970-2000." In Q. Wodon, Structural Trends and Impact of Administration." ed., "Public Spending, Poverty, and Inequality in Latin International Monetary Fund Working Paper 99/19. America." World Bank, Washington, D.C. Processed. Washington, D.C. Scott, James. 1998. Seeing Like a State: How Certain Schemes Shrader, E. 2001. "Methodologies to Measure the Gender to Improve the Human Condition Have Failed. New Haven, Dimensions of Crime and Violence." World Bank, CT: Yale University Press. Washington, D.C. SDStats (Social Development Indicators) database. http://esd. Shugart, M. 2001. "Electoral 'Efficiency' and the Move to worldbank.org/sdstats/. World Bank, Washington, D.C. Mixed-Member Systems." Electoral Studies 20 (2): 173-93. Segarra, M. "Redefining the Public/Private Mix: NGOs and Shugart, M. 1992. "Leaders, Rank-and-File, and Con- the Emergency Social Investment Fund in Ecuador." In stituents: Electoral Reform in Colombia and Venezuela." D. Chalmers, Carlos M. Vilas, Katherine Hite, Scott B. Electoral Studies 11 (1): 21-45. Martin, Kerianne Piester, and Monique Segarra, eds. The _ . 1995. "The Electoral Cycle and Institutional New Politics of Inequality in Latin America: Rethinking Sources of Divided Presidential Government." American Participation and Representation. New York: Oxford Political Science Review 89 (2): 327-43. University Press. . 1998. "The Inverse Relationship between Party Seligmann, Linda. 1995. Between Reform and Revolution: Polit- Strength and Executive Strength: A Theory of Politicians' ical Struggles in the Peruvian Andes. 1969-1991. Stanford, Constitutional Choices." BritishJournal of Political Science CA: Stanford University Press. 28: 1-29. Sen, Amartya. 1973. (Expanded edition 1997.) On Economic Shugart, M., and R. Taagepera. 1994. "Plurality Versus Inequality. Oxford: Clarendon Press. Majority Election of Presidents: A Proposal for a Double ______. 1985a. Commodities and Capabilities. Amsterdam: Complement Rule." Comparative Political Studies 27 (3): Elsevier. 323-48. _ . 1985b. "Well-being, Agency and Freedom. The Siaens, Corinne, and Quentin Wodon. 2003. "Public Social Dewey Lectures, 1984." Journal of Philosophy 82 (4): Spending and Inequality in Latin America: Gini Income 169-221. Elasticities." World Bank, Washington, D.C. Processed. _____. 1992. Inequality Reexamined. Cambridge, MA: Silva, Nelson do V., and D. Roditi. 1988. "Et plus ,a Harvard University Press. change.Tendencias hist6ricas da fluidez social no Brasil." _____. 1999. Development as Freedom. New York: Knopf Dados: Revista de Cie^ncias Sociais 29 (3). Press. Skoufias, Emmanuel. 2001. "PROGRESA and its Impacts ______ 2000. "Social Justice and the Distribution of on the Human Capital and Welfare of Households in Rural Income." In Francois Bourguignon and Anthony B. Mexico: A Synthesis of the Results of an Evaluation by Atkinson, eds. Handbook of Income Distribution. IFPRI." International Food Policy Research Institute, Amsterdam: North-Holland Publishing Company. Washington, D.C. Shadlen, K. 2000. "Neoliberalism, Corporatism, and Small Skoufias, Emmanuel, and B. McClafferty. 2001. "Is Business Political Activism in Contemporary Mexico." PROGRESA Working? Summary of the Results of an Latin American Research Review 35 (2): 73-106. Evaluation by IFPRI." International Food Policy Research Shah, A., and J. Whalley. 1990. "An Alternative View of Tax Institute, Washington, D.C. Incidence Analysis for Developing Countries." World Skoufias, Emmanuel, and S. W. Parker. 2001. "Conditional Bank Working Paper 462. Washington, D.C. cash Transfers and their Impact on Child Work and Shefner, J. 2001. "Coalitions and Clientelism in Mexico." Schooling: Evidence from the PROGRESA program in Theory and Society 30 (5): 593-628. Mexico." Economla 2 (1): 45-96. Shefter, Martin. 1994. Political Parties and the State: The Skoufias, Emmanuel, B. Davis, andJ. Behrman. 1999. "Final American Historical Experience. Princeton, NJ: Princeton Report: An Evaluation of the Selection of Beneficiary University Press. Households in the Education, Health, and Nutrition Pro- Sheng, Andrew, ed. 1996. Bank Restructuring: Lessonsfrom the gram (PROGRESA) of Mexico." International Food 1980s. Washington, D.C.: World Bank. Policy Research Institute, Washington, D.C. 376 BIBLIOGRAPHY Skoufias, Emmanuel, B. Davis, and S. de la Vega. 2001. Stephan, Cookie White, and Walter G. Stephan. 2000. "The "Targeting the Poor in Mexico: An Evaluation of the Measurement of Racial and Ethnic Identity." International Selection of Beneficiary Households into PROGRESA." Journal of Intercultural Relations 24: 541-52. World Development 29 (10): 1769-84. Stern, Nicholas H. Forthcoming. Poor People and Economic Smeeding, T., and Grodner, A. 2000. "Changing Income Opportunity. Cambridge, MA: Massachusetts Institute of Inequality in OECD Countries: Updated Results from Technology Press. the Luxembourg Income Study (LIS)." Luxemborg Stokes, Susan C. 1995. Cultures in Conflict: SocialMovements and Income Study Working Paper 252. Syracuse University, the State in Peru. Berkeley: University of California Press. Maxwell School of Citizenship and Public Affairs, Stotsky, Janet, and Asegedech WoldeMariam. 2002. "Cen- Syracuse, NY. tral American Tax Reform: Trends and Possibilities." Smith, J., and K. Subbarao. 2000. "What Role for Safety International Monetary Fund Working Paper 02/227. Net Transfers in Very Low-income Countries?" World Washington, D.C. Bank Discussion Paper. Washington, D.C. Sumarto, Sudarno, Asep Suryahadi, and Lant Pritchett. 2003. Smith, W., and R. Korzeniewicz. 1997. Politics. Social Change, "Safety Nets or Safety Ropes? Dynamic Benefit Incidence andEconomic Restructuring in Latin America. Miami: North- Analysis of Two Crisis Programs in Indonesia." World South Center Press. Development 31 (7): 1257-77. Snow, Peter G., and Gary Wynia. 1990. "Argentina." In Sutil, J. 1999. "Judicial Reforms in Latin America: Good Howard J. Wiarda and Harvey F. Kline, eds., Latin News for the Underprivileged?" In J. Mendez, G. American Politics and Development. 3rd ed. Boulder, CO: O'Donnell, and P. Pinheiro, eds., The Rule of Law and the Westview Press. Underprivileged in Latin America. Notre Dame, IN: Soares, Sergei. 2000. "O Perfil da Discriminac,o no Mer- University of Notre Dame Press. cado de Trabalho: Homens Negros, Mulheres Brancas e Szekely, M. 2001. "The 1990s in Latin America: Another Mulheres Negras." Discussion Paper 769. Instituto de Decade of Persistent Inequality, but with Somewhat Pesquisa Econ6mica Aplicada, Brasilia, Brazil. Lower Poverty." Inter-American Development Bank Sosa Escudero, W, and L. Gasparini. 2000. "A Note on the Working Paper 454. Washington, D.C. Statistical Significance of Changes in Inequality." Szekely, M., and M. Hilgert. 1999. "What's Behind the Economica 46 (1): 111-22. Inequality We Measure: An Investigation Using Latin Sosa Escudero, W., and M. Marchionni. 2000. "Household American Data." Inter-American Development Bank Structure, Gender, and Economic Determinants of Working Paper 409. Washington, D.C. School Attendance in Argentina." In Argentina: Poor _ . 2001. "What Drives Differences in Inequality People in a Rich Country. Vol, 11. Washington, D.C.: World Across Countries?" Inter-American Development Bank Bank. Working Paper 439. Washington, D.C. Sour, Laura. 2003. "Return of the Enforcement Spending on Szekely, M., Nora Lustig, M. Cumpa, andJ. Mejia. 2000. "Do VAT Revenues in Mexico." Mexico City: CIDE. We Know How Much Poverty There Is?" Inter-American Stame, N. 1999. "Small and Medium Enterprise Aid Pro- Development Bank Working Paper 437. Washington, grams: Intangible Effects and Evaluation Practice." Eval- D.C. uation and Program Planning 22 (1): 105-1 1. Tabatabai, H. 1996. Statistics on Poverty and Income Distrib- Steinmo, S. 1993. Taxation and Democracy: Swedish, British, ution: An ILO Compendium of Data. Geneva: International and American Approaches to Financing the Modern State, New Labour Office. Haven, CT: Yale University Press. Talvi, Ernesto, and Vegh, Carlos. 2000. "Tax Base Variabil- . 2002. The Evolution of Policy Ideas: Tax Policy in the ity and Procyclical Fiscal Policy." National Bureau of 20th Century. Boulder, CO: University of Colorado. Economic Research Working Paper 7499. Cambridge, MA. Stepan, A. 1978. The State and Society: Peru in Comparative Tavares, J., and Wacziarg, R. 2001. "How Democracy Affects Perspective. Princeton, NJ: Princeton University Press. Growth." European Economic Revieu' 45 (8): 1341-78. . 2000. "Brazil's Decentralized Federalism: Bringing Telles, Edward, and Nelson Lim. 1999. "Interessa Quem Government Closer to Citizens?" Daedalus 129 (2): Reponde a Questao Sobre Classificacao Racial e Desigual- 145-69. dade de Renda no Brasil?" Estudio Afro-Asiaticos 35: 7-27. 3 77 INEQUALITY IN LATIN ANMERICA Tendler, Judith. 1997. Good Government in the Tropics. Tsebelis, George. 1995. "Decision Making in Political Sys- Baltimore: Johns Hopkins University Press. tems: Veto Players in Presidentialism, Parliamentarism, _____. 2001. "Transforming Local Economies: Lessons Multicameralism, and Multipartyism." BritishJournal of from the Northeast Brazilian Experience." Massachusetts Political Science 25 (3): 289-325. Institute of Technology, Department of Urban Studies and _ . 1999. "Veto Players and Law Production in Planning, Cambridge, MA. Processed. Parliamentary Democracies: An Empirical Analysis." ______. 2002. "Small Firms, the Informal Sector and the American Political Science Review 93 (3): 591-609. Devil's Deal." IDS Bulletin 33 (3). _ . Forthcoming. Veto Players: How Political Institutions Teranishi,Juro. 1997. "Sectoral Resource Transfers, Conflict Work. Princeton, NJ: Princeton University Press. and Macrostability in Economic Development: A Com- Ungar, M. 2002. Elusive Reform: Democracy and the Rule of Law parative Analysis." In Masahiko Aoki, Hyung-Ki Kim, in Latin America. Boulder, CO: Lynne Rienner Publishers, and Masahiro Okuno-Fujiwara, eds. The Role of Govern- Inc. ment in East Asian Economic Development. Oxford and New United Nations Development Programme, Population York: Oxford University Press. Division. 1993. Human Development Report 1993. New Teruel, G., and B. Davis. 2000. "Final Report: An Evalua- York: United Nations Development Programme and tion of the Impact of PROGRESA Cash Payments on Pri- Oxford University Press. vate Inter-household Transfers." International Food _ . 2000. World Population Prospects: The 2000 Revision, Policy Research Institute, Washington, D.C. Volume 1. New York: United Nations. Teulings, C., J. Hartog, and S. Bazen. "Corporatism or United Nations Statistics Division. 2002. Population and Competition? Labour Contracts, Institutions, and Wage Vital Statistics Report, Statistical papers, Series A, 54 (4). Structures in International Comparison." EconomicJournal http://unstats.un.org/unsd/seriesa l 11 (475): F802-04. Valenzuela, Arturo. 1977. Political Brokers in Chile. Durham, Thiesenhusen, W. 1995. Broken Promises: Agrarian Reform and NC: Duke University Press. the Latin American Campesino. Boulder, CO: Westview Press. Valle Silva, Nelson do. 2001. "Race, Poverty, and Social Exclu- Thomas, D. 1999. "Comments on Acosta." In J. Mendez, sion in Brazil." In Estanislao Gacitiua-Mari6 and Carlos G. O'Donnell, and P. Pinheiro, eds., The Rule of Law and Sojo, eds., Social Exclusion and Poverty Reduction in Latin the Underprivileged in Latin America. Notre Dame, IN: Uni- America and the Caribbean. Washington, D.C: World Bank. versity of Notre Dame Press. Van Cott, D. 2000. "Party System Development and Indige- Thomas, Vinod, Yan Wang, and Xibo Fan. 2002. "A New nous Populations in Latin America: The Bolivian Case." Dataset on Inequality and Education: Gini and Theil Party Politics 6 (2): 155-74. Indices of Schooling for 140 countries: 1960-2000." Velasco, Andres. 1996. "Fixed Exchange Rates: Credibility, Washington, D.C. World Bank. Flexibility and Multiplicity." European Economic Review 40: Tilly, Charles. 1999. Durable Inequality. Berkeley: Univer- 1023-35. sity of California Press. Velazquez, Ernesto. 1982. "Los Mecanismos del Subdesar- Tobin, J. 1970. "On Limiting the Domain of Inequality." rollo en una Regi6n Marginada de Colombia: Estudio de Journal of Law and Economics 13: 263-77. la Evoluci6n de las Condiciones de Vida de la Poblaci6n Tokman, V., and G. O'Donnell. 1998. Poverty and Inequal- Nariniense-1982." Bogota, Camara de Representantes. ity in Latin America: Issues and Neu Challenges. Notre Mimeo. Dame, IN: The University of Notre Dame Press. Velez, C. E. 1995. Gasto Social y Desigualdad: Logros y Torres-Saillant, Silvia. 1998. "The Tribulations of Blackness: Extravios. Santa Fe de Bogota: Misi6n Social, Departa- Stages in Dominican Racial Identity." Latin American mento Nacional de Planeaci6n. Perspectives 25 (3): 126-46. V6lez, C. E., and V. Foster. 2003. "Public Social Spending Trejos, J. 1999. "Reformas Econ6micas y Distribuci6n del in Colombia: Incidence and Sector Priorities in the Ingreso en Costa Rica." Reformas Econ6micas series 37. 1990s." In Q. Wodon, ed., "Public Spending, Poverty, Economic Commision for Latin America and the and Inequality in Latin America." World Bank, Caribbean, Santiago. Washington, D.C. Processed. 378 B I B LI 00; R A P H Y' Vianna, M. L. 1998. Americanizafdo (Perversa) da Seguridade Peeters, Corinne Siaens, and Shlomo Yitzhaki. 2000. Social no Brasil: Estrate'gias de Bem-estar e Politicas Publi- Poverty and Policy in Latin America and the Caribbean. cas. Rio de Janeiro: Editora Revan. Washington, D.C.: World Bank. Vicepresidencia de la Republica de Colombia. 2002. Wodon, Quentin, Rodrigo Castro-Ferinandez, Kihoon Lee, "Observatory on Human Rights in Colombia." Bulletin Gladys L6pez-Acevedo, Corinne Siaens, Carlos Sobrado, 22. Presidential Program on Human Rights, Bogota. Jean-Philippe Tre. 2001. "Poverty in Latin America: von Amsberg, J., Peter Lanjouw, and K. Nead. 2003. Trends (1986-1998) and Determinants." Cuadernos de "The Poverty Targeting ofSocial Spending in Brazil." In Q. Economla 38 (114): 127-53. Wodon, ed., "Public Spending, Poverty, and Inequality in Wodon, Quentin, N. Hicks, B. Ryan, and G. Gonzalez. Latin America." World Bank, Washington, D.C. Processed. 2003. "Are Governments Pro-Poor but Short-Sighted? Wagstaff, A. 2000. "Socioeconomic Inequality in Child Targeted and Social Spending for the Poor during Booms Mortality: Comparisons Across Nine Developing Coun- and Busts." In Quentin Wodon, ed., "Public Spending, tries." Bulletin of the World Health Organization 78 (1): Poverty, and Inequality in Latin America." World Bank, 19-29. Washington, D.C. Processed. .. 2001. "Inequalities in Health in Developing Coun- Wodon, Quentin, and William Maloney. 2000. "Self- tries: Swimming Against the Tide?" World Bank Work- employmentasanExplanationforHighInequality in Latin ing Paper 2795. Washington, D.C. America." World Bank, Washington, D.C. Processed. Wagstaff, A., and N. Watanabe. 2000. Socioeconomic Inequal- Wolfson, M. 1994. "When Inequalities Diverge." The ities in Child Malnutrition in the Developing World. World American Economic Review. 84 (2): 353-58. Bank Working Paper 2434. Washington, D.C. Wood, Adrian. 1997. "Openness and Wage Inequality in Weisman, S. 2002. The Great Tax Wars. New York: Simon Developing Countries: The Latin American Challenge to and Schuster. East Asia Conventional Wisdom." The World Bank Weyland, Kurt. 1995. "Social Movements and the State: the Economic Review 11 (1): 33-57. Politics of Health Reform in Brazil." World Development Woodward, Ralph L. 1966. Class Privilege and Economic 23 (10): 1699-712. Development: The Consulado de Comercio of Guatemala . 1996a. "Neopopulism and Neoliberalism in Latin 1793-1871. Chapel Hill: University of North Carolina America: Unexpected Affinities." Studies in Comparative Press. International Development. 31 (3): 3-31. World Bank. 1990. World Development Report 1990: Poverty. . 1996b. Democracy uwithout Equity: Failures of Reform New York: Oxford University Press. in Brazil. Pittsburgh, PA.: University of Pittsburgh Press. _ . 1993. The East Asian Miracle. New York: Oxford . 1997. "Growth with Equity in Chile's New University Press. Democracy?" Latin American Research Review 32 (1). _ . 1995. World Development Report 1995: Workers in Whalley, J. 1984. "Regression or Progression: The Taxing an Integrating World. New York: Oxford University Question of Incidence Analysis." Canadian Journal of Press. Economics 17 (4): 654-82. _ . 1997. World Development Report 1997: The State in Williamson, Jeffery G. 1999. "Real Wages, Inequality and a Changing World. New York: Oxford University Press. Globalization in Latin America before 1940." Revista de _ . 1999. World Development Report: Entering the 21st Historia Economica 17: 101-42. Century: The Changing Development Landscape. New York: Willis E., C. Garman, and S. Haggard. 1999. "The Politics Oxford University Press. of Decentralization in Latin America." Latin American _ . 2000a. Argentina: Poor People in a Rich Country. Research Review 34 (1): 7-56. Washington, D.C.: World Bank. Wodon, Quentin, ed. 2003. "Public Spending, Poverty, and _ . 2000b. World Development Report 2000/01: Attack- Inequality in Latin America." World Bank, Washington, ing Poverty. New York: Oxford University Press. D.C. Processed. _ .2000c. Bolivia: From Patronage to a Professional State, Wodon, Quentin T., Robert Ayres, Matias Barenstein, Institutional and Governance Review. Washington D.C.: Norman Hicks, Kihoon Lee, William F Maloney, Pia World Bank. 379 INEQUALITY IN LATIN AMERICA _____. 2000d. Making Transition Workfor Everyone: Poverty . 2003e. Caribbean Economic Overview 2002: Macro- and Inequality in Europe and Central Asia. Washington economic Volatility, Household Vulnerability and Institutional D.C.: World Bank. and Policy Responses. Washington D.C.: World Bank. . 2000e. Reforming Public Institutions and Strengthen- . 2003f. Jamaica: The Road to Sustained Growth and ing Governance: A World Bank Strategy. Washington, Poverty Reduction, Country Economic Memorandum 2003. D.C.: World Bank, Public Sector Group, Poverty Washington D.C.: World Bank. Reduction and Economic Management (PREM) _ . 2003g. Country Assistance Strategy for the Dominican Network. Republic. Washington D.C.: World Bank. . 2000f. Ecuador Gender Review: Issues and Recommen- _ . 2003h. Colombia: Land Policy in Transition. dations. Washington D.C.: World Bank. Washington D.C: World Bank. . 2001a. Household Risk, Self-Insurance, and Coping . 2004 "Minimum Wages in Latin America: The Strategies in Urban Argentina. Report 22426-AR. Impact on Employment, Inequality and Poverty." Latin Washington D.C.: World Bank. America and Caribbean Region, Washington, D.C., ___ . 2001b. Mexico: Land Policy: A Decade after the Ejido World Bank. Mimeographed. Reform. Washington D.C.: World Bank. . Forthcoming. El Salvador: Poverty Assessment. . 2001c. Peru-Institutional and Governance Review. Washington D.C.: World Bank. Washington D.C.: World Bank. World Development Indicators Database. http://www. . 2002a. A Review of Gender Issues in the Dominican worldbank.org/data/wdi2003/. World Bank, Washington, Republic, Haiti, andJamaica. Report 21866. Washington D.C. D.C.: World Bank. World Economic Forum. 2003. Global Competitiveness Report . 2002b. Guatemala Poverty Assessment. Washington 2002-2003. Peter Cornelius, Klaus Schwab, and Michael D.C.: World Bank. E. Porter, eds. New York: Oxford University Press. . 2002c. Brazil Gender Review: Issues and Recommen- WIDER (World Institute for Development Economics dations. Washington D.C.: World Bank. Research). 2000. UNU/WIDER-UNDP World Income . 2003a. World Development Report 2004: Making Database, Version 1.0 (September). Services Work for Poor People. New York: Oxford Univer- Wossmann, L. 2000. "Schooling Resources, Educational sity Press. Institutions and Student Performance: The International . 2003b. Colombia: The Economic Foundation of Peace. Evidence." Kiel Institute of World Economics Working Washington D.C.: World Bank. Paper 983, Kiel, Germany. _____ .2003c. Land Policyfor Growth andPoverty Reduction. Yaschine, Iliana. 1999. "The Changing Anti-poverty Washington D.C.: World Bank. Agenda: What can the Mexican Case Tell Us?" IDS . 2003d. Brazil: Strategies for Poverty Reduction in Bulletin 30 (2): 47-60. Ceard. Washington D.C.: World Bank. 380