Report No. 22950-BR BrazH The New Growth Agenda (In Two Volumes) Volume I1: Detailed Report December 31, 2002 Brazil Country Management Unit Latin America and the Caribbean Region C: 0 Z Cj) >< * C: 0~~ 1~~0 r r m; Document of the World Bank .,~~~~~~~~~~~~~~~~~ F' '' "1~ R ! l S l ! l !0 ~~~-~ / Ag\! CURRENCY AND EXCHANGE RATES (R$IUS$) Currency Unit - Real (R$) December 2000: R$1.95 December 2001: R$2.65 December 2002: R$3.52 WEIGHTS AND MEASURES The Metric System is used throughout the report. FiISCAL YEAR January 1 to December 31 ACRONNYWS AND ABBREVDATIONS CPMF Temporary Tax on Financial Transactions EMU European Monetary Union FDI Foreign Direct Investment FIAS Foreign Investment Advisory Service GDP Gross Domestic Product ICT Information and Communications Technology INPI National Institute of Intelectual Property (Instituto Nacional de Propriedade Intelectual) IT Information Technology OECD Organization for Economics Cooperation and Development R&D Research and Development RGPS General Regime for Social Security (Regime Geral da Previdencia Social) RJU Pension Regime for Government Workers (Regime Jur[dico Unico) SOE State Owned Enterprise TFP Total Factor Productivity Vice President LCR: David de Ferranti Director LCC5C: Vinod Thomas Director LCSPR: Ernesto May Lead Economist: Joachim von Amsberg Task Managers: Indermit S. Gill and Mark R. Thomas ACKNOWLEDGMENTS This report was prepared by a team consisting of Indermit Gill and Mark Roland Thomas (Joint Task Managers), Norman Loayza, Marc-Andreas Muendler, Luis Serven, Claudia Paz Sepulveda, Naercio Menezes-Filho, Carl Dahlman, Anuja Adhar Utz, Jean-Eric Aubert, Christine Zhen-Wei Qiang, Harry Patrinos, Hoon Sahib Soh and the task managers. Authors' affiliations are given in Volume II, Chapter 1. The report was prepared under the overall guidance of Gobind Nankani (Country Director until September 2001), Suman Bery (Lead Economist), Joachim von Amsberg (Lead Economist), and Ernesto May (Sector Director). Peer reviewers for this task are Armando Castelar (National Development Bank of Brazil-BNDES), Mansoor Dailami (World Bank Institute), and Gene Grossman (Princeton University). The report has benefited from the work and insight of many people. The initial scope and form of the report were the subjects of input from distinguished participants at a workshop in Rio de Janeiro in November 2000. We would like to acknowledge important contributions at this workshop from Edward Amadeo and Joachim Levy (MF), Armando Castelar (BNDES), Pedro Cavalcanti Ferreira (FGV), Jose Guilherme Almeida dos Reis (MPOG, now MF), Paulo Eduardo Rocha Brant (MDIC), Simon Schwartzman (IBGE), Karl Voltaire, Toshi Masuoka, Mary Iskenderian, Flavio Guimaraes, and Atul Mehta (IFC), and Joachim von Amsberg. The work received thoughtful contributions at the Concept Paper stage from the peer reviewers, Paul Beckerman, Sara Calvo, Homi Kharas, Danny Leipziger, Guillermo Perry, and David Rosenblatt. At the Review Draft stage, the team received another round of excellent comments from the three peer reviewers, Andreas Blom, Mauricio Carrizosa, Luis Coirolo, Lauritz Holm- Nielsen, Sanjay Kathuria, Guillermo Perry, Sergio Margulis, Gobind Nankani, Roberto Rocha, Alberto Rodriguez, Vinod Thomas, Dorte Vemer, and Joachim von Amsberg. The report also benefited at various points from discussions with William Easterly. We were privileged to receive the insights of a stellar Quality Enhancement Review (QER) panel consisting of Gene Grossman, Homi Kharas, and Ricardo Hausmann. Our discussions with these three economists in May 2001 have had a profound impact on the final content and form of the report, and we are grateful for their support and the effort it entailed. The team sincerely thanks Sara Calvo and Vera Wilhelm for their enthusiasm and efficiency in conducting the QER, and the PREM Anchor for financial support. Most importantly, we wish to acknowledge the continued collaboration of BNDES with World Bank economic growth efforts, and in particular the support and astute commentary of Armando Castelar, Director of Research at BNDES. We acknowledge the discussions with and comments from representatives of the Federal Government of Brazil. The findings and views expressed here are exclusively those of the WnrIdI Rank. BRAZIL THE NEW GROWTH AGENDA Table Of Contents EXECUTIVE SUMMARY OF VOLUME II ............. ............................i Q i $02 1111 9.1:1 J 4I,6=@*l 06^1 1iv TA l. INFORMING BRAZILIAN ECONOMIC POLICY .................... ................ I Foreword .................................1 Background .................................2 fI. ECONOMic GROWT11: FROM THEORY TO PRAC TI1CE .................................... 4 Fifty Years is Not Enough .................................4 Policy as a Determinant of TFP .................................7 Ill. THE NEW GROWTH. AGENDA .................................... 9 Knowledge and Innovation in the International Economy .................................. 9 The Importance of Public Policy .................................9 IV. PossIBI.E CONSTRAINTS ON BRAZILIAN GROWTH .................................... I I History and Hypotheses ................................. 11 V. BRINGING DATA TO THE QUESTI ION ........... ......................... 1 2 Data Analysis as Description ................................. 12 Choosing Episodes to Understand Policy Effects ................................. 14 VI. A ROADMAP TO TlE REST OF TI-IIS VOLUME .................................... 15 A,ItetMT1 9390I2IaIIL :T,A I I I. INTRODUCTION .................................... 20 it. STYLIZED FACTS ............................................ 20 1. Long-Run Growth Trends in Brazil, Latin America, and the World .................... 20 2. Sector Composition of Output Growth in Brazil ........................................ 21 3. Growth, Saving, and Investment in Brazil ........................................ 24 4. Growth Accounting in Brazil ........................................ 28 1If. IMPLICATIONS FOR BRAZIL FROM A CROSS-COUNTRY PERSPECTIVE ......... 31 Econometric Methodology ............................................ 32 Results ............................................ 34 IV. AREAS FOR IMPROVEMEN1 ................................................ 39 Brazil Compared to Latin America, East Asia, and the OECD .............................. 39 V. St!MMARY OF CONCLUSIONS ................................................ 41 murl T IVt 0MA9 k A.1111 I 3 06- I. IN-rRODUCTION. ....... 50 I1. OVERVIEW OF EARLIER STUDIFS ...................................... 54 Ill. FRAMEWORK AND DATA ...................................... 58 Estimating TFP .................................. 58 Data .................................. 59 IV. PATTFRNS OFTFP GROWTH IN BRAZIIIAN INDUSTRY ............................... 63 Overview .................................. 63 Details by Sector .................................. 66 Details by Region .................................. 68 V. TFP GROWTH AND RESOUIRCE A-LLOCATION ............................................. 69 TFP and Exposure to Foreign Trade .................................. 73 TFP and Knowledge Embodied in Input Quality . ................................. 76 Resource Allocation ..... 76 Physical Investment ..... 77 VI. CONCLUDING REMARKS ..... 78 El S ID ILi10] *0 e .11IT .1L I D I I 0 I. GROWTH ACROSS STATES AND TNCOMF GROUPS ....................................... 84 One Dimension of the Quality of Growth: The Equality of Growth ........................ 84 Measuring Patterns of Growth in Brazilian Household Data ................................. 85 Hypotheses ........................................ 86 IT. DESCRIPTIVE RESULTS ............................................ 89 Introduction ......................................... 89 Four Periods ........................................ 89 State Income Growth ........................................ 90 Convergence ........................................ 93 Income Growth among the Poor ........................................ 95 Descriptive Conclusions ........................................ 101 111. ANALYTICAL RESLTTS ............................................ 102 Introduction ........................................ 102 Average Income Growth ........................................ 105 Income Growth among the Poor ........................................ 107 The "Equality of Growth" and Possible Tradeoffs ........................................ 107 Distinguishing between Levels of Education ........................................ 109 Location and Intemational Trade ........................................ 110 Migration ......................................... 112 IV. POLICY IMPLICATIONS ............................................1 112 A Retum to the Hypotheses ........................................ 112 Main Messages ........................................ 115 Closing Remarks ........................................ 117 I. INTRODUCTION ............................................ 120 II. BRAZIL AND THE KNOWLEDGE REVOLUTION ........................................... 121 The Knowledge Revolution and Global Competition ........................................ 121 Key Elements of the Knowledge-Based Economy ........................................ 123 An Overview of Brazil's Situation ........................................ 123 II. BRAZIL'S INNOVATION SYSTEM AND POLICY ........................................... 127 Technology and Economic Performance ........................................ 127 The Innovation System ........................................ 128 Government Policies ........................................ 132 Policy Guidelines ........................................ 133 IV. THF [CT SECTOR IN BRAZII ............................................. 134 Recent Reforms and Privatization and Liberalization Developments .................. 134 Industrial Policy and Regulatory Issues ........................................ 139 The Future of Information Infrastructure: Main Challenges ................................. 140 V. CONCLUSIONS ............................................ 142 ANNEXES . .144 Annex 1A: Knowledge Assessment Methodology .144 Annex 1 B: Country Selection .............................................. 146 Annex IC: Variables Used in the Standard 15-variable Scorecards .................... 147 Annex ID: Other Variables Available in the KAM .............................................. 148 Annex 1 E: Knowledge Assessment Scorecards .............................................. 150 Annex 1 F: More Detailed Scorecards .............................................. 152 Annex 1G: Data for the Knowledg ssessment Methodology.160 1. INTRODUCTION ................................................... 162 What is "Human Capital"? .............................................. 163 What are "Human Capital Policies"? .............................................. 164 The Plan of this Paper and Main Conclusions .............................................. 164 1I. BRAZIL.: THE ANATOMY OF AN INTANGIBI.E CAPITAI." LAGGARD ......... 167 Education .............................................. 167 Health .............................................. 169 Brazil's Human Capital Scorecard .............................................. 170 fll. LESSONS FROM rilE LITERATURE ................................................... 172 Human Capital in the Theory of Economic Growth ............................................. 172 Single Country Estimations .............................................. 175 Cross-Country Estimations .............................................. 178 Other Issues .............................................. 180 IV. SCHOOLING AND GROWTH IN BRAZIL ................................................... 184 Brazilian schooling and GDP during the last half-century ................................... 184 Some Econometrics .............................................. 185 V. EVIDENCE FROM THREE "EPISODES" ................................................... 188 United States, 1901 -2000: America's "Human Capital Century" .......................... 189 Korea, 1951-2000: A Hundred Years in Fifty? .............................................. 192 India 1976-2000: Growth and Higher Education Investments Amidst Illiteracy ... 195 VI. POLICY IMPL1CA'IONS FOR BRAZIL ................................................... 199 VII. SUMMARY AND CONCLUSIONS ................................................... 205 Summary Table of Human Capital Policy Recommendations ............................. 208 I. INT RODUCTION ................................................... 212 11. OVERALL. PUBLIC GOVERNANCE MEASUTRES ........................................... 213 III. MEASLIREMENIS OF UNDERLYING DETERMINANTS ...................2.............. 16 Court Performance .............................................. 218 Regulatory Burden for Start-up Entrepreneurs .............................................. 219 IV. GOVERNANCE AN) ENTREPRENEURSHIP ..................................................2 20 V. GOVERNANCE AND FOREIGN DIRECT INVESTMENT ................................... 225 The Size of FDI Inflows .............................................. 226 The Sustainability of FDI .............................................. 227 The Sector Distribution of FDI .............................................. 230 The Industry Distribution of FDI ............................................... 232 Research on Govemance and FDI .............................................. 234 V I. POI.ICY IMPLICATIONS ...................................................2 39 Annex .............................................. 247 EXECUTIVE SUMMARY OF VOLUME II 1. This is the second of two volumes, and collects together seven background papers written for the Wodd Bank's Brazilian growth study. The first volume summarizes the overall conclusions for policy drawn from these background papers and other relevant research, as well as giving a historical account of the driving forces behind Brazilian economic growth since the 1960s. This summary limits itself to the content of this volume. Hypotheses 2. Chapter 1 surveys economic growth theory and describes the factors emphasized by the theory that will drive the report's recommendations. The report considers growth determinants traditionally emphasized, such as the level of physical investment (private and public), saving, macroeconomic balance, government consumption, and health and education. But the report also emphasizes factors that newer theories suggest may be important: the environment for knowledge creation and the adoption of new technologies, the analysis of tertiary versus basic education, and the dynamic effects of foreign direct investment. The chapter then outlines a list of candidate hypotheses to explain the seemingly sluggish response of the Brazilian economy to the successful stabilization of the mid-I 990s. It should be noted that these are hypotheses: their relevance or lack thereof is to be determined by appealing to empirical obsenration: * Cost of capital and crowding out of private investment by the public sector borrowing requirement? * Human capital deficiencies, particularly relating to health, primary, secondary, or tertiary education and training? * Deficiencies in the flow of knowledge and innovation, related to intellectual property rights and/or use of foreign technology? * Overly rigid and costly formal-sector labor-market regulations driving up the cost of labor, including lengthy and unpredictable labor court delays? * High tariffs and relative isolation from international trade flows? * High public-sector investment composition combined with the low-productivity of public-sector investments and/or a lack of investment in infrastructure? Poor climate for private investment, including regulatory and administrative barriers and an inefficient judicial system? * Macroeconomic instability, including rollover risk of external debt, remaining uncertainty surrounding public debt sustainability and inflation? * Constraints related to demographics and migration, including the decrease in migratory flows from rural to metropolitan areas, and from the Northeast to the South and Southeast regions? * High and distortionary taxes, including cascading taxes on exports, a financial transactions tax, and an overall tax burden greater than 30 percent of GDP? 3. It is not the case that every chapter introduces evidence on every one of the above topics. However, most chapters touch upon several, allowing the report to build up a "body of evidence" on Brazilian growth, its determinants, and its main constraints. Table 1 at the end of this summary describes the ways in which chapters shed light on hypotheses. After the introduction of the main concepts and framework in chapter 1, the remaining chapters fall into two parts. Part I (chapters 2- 4) consists of data analysis, while Part II (chapters 5-7) focuses on particular areas of policy suggested as important by Part I. Part I: New Data 4. Chapter 2 uses aggregate data-time series data for Brazil since 1960, and panel data for an international data set-to analyze Brazil's macroeconomic growth determinants. Time series data suggest that saving is uncorrelated with Brazilian economic growth. And capital accumulation does not seem to have been the driving factor behind Brazilian economic growth in the period: economic growth precedes (in a time-series sense) both private and public investment. A growth accounting exercise corroborates this result: the fall in growth between the 1970s and now is only accounted for in small part by a fall in physical capital formation, while human capital formation increased in the 1980s and 1990s, leaving TFP changes to explain the lion's share of the reduction in growth. 5. Cross-country regression techniques fail to explain more than half of the decline in Brazilian economic growth between the 1990s and the 1970s, and 1990s Brazil is an outlier in this analysis. World economic conditions, inflation (in the early 1990s), higher initial income (cf. convergence theory), a reliance on primary exports, and terms of trade shocks all reduced Brazilian growth in the 1990s. These effects were not fully offset by those factors that did improve, in particular human capital formation, trade opening, and improved macroeconomic stability. The chapter concludes that despite these areas of unmistakable progress, Brazil has fallen further behind the OECD and East Asian economies in important areas: human capital, financial depth, govemance, and public infrastructure. 6. Chapter 3 examines the causes of Brazilian industrial productivity growth using a new panel data set constructed from the Brazilian annual industrial survey. At the firm level, Brazilian TFP growth was on average negative in the 1980s before recovering modestly in the 1990s. Performance differed widely by industry, with the automotive and electrical equipment industries showing the strongest productivity growth in the 1990s. Less technological sectors-food, textiles, leather, and non-metallic minerals-showed the worst performance and actually declined in productivity during the 1990s. It is worthy of mention that the industrial composition of the manufacturing sector in the North and Northeast of Brazil led to lower TFP growth in these regions relative to the South and Southeast. 7. TFP and TFP growth regressions across firms within industries suggest that trade opening may have contributed to raise average TFP by about 6 percent. Market penetration by foreign competitors raised the level of TFP, though not its growth rate. On the other hand, export orientation does not raise TFP in the Brazilian data; indeed this relationship is negative. Other findings suggest the importance of technological innovation: higher shares of IT in physical capital raised both TFP and its growth rate. To complement this result, firms employing more skilled labor, in the form of a higher white-to-blue collar ratio, showed faster TFP growth, suggesting that Brazilian human capital investments contribute to TFP growth. Foreign machinery, on the other hand, had no significant effect, questioning the role of this as a conduit of embodied technological change in Brazilian growth. Given evidence from firms elsewhere on the importance of international knowledge flows (Chapters 5 and 7), this result raises the concern that Brazil's integration into international production is not generating all its potential possible benefits, perhaps owing to a bias towards regional trade, perhaps owing to impediments to technology transfer. Finally, new firms possessed higher and faster-growing TFP, suggesting that measures to ease their entry may be a source of future productivity gains for Brazil. 8. Chapter 4 uses stacked annual household data sets (PNAD, 1981-98) to analyze growth of average household income and of average household income among the poorest 25 percent of the population. These data reveal that only since the real plan stabilization has significant convergence occurred between the poorer and richer Brazilian states, and that a large part of the convergence pattern is owing to the pro-poor impact of eradicating inflation. International evidence suggests that any growth strategy that jeopardizes this central achievement is likely to fail. This chapter suggests that, furthermore, it is likely to hurt the poor most. 9. Comparison of the income growth of the average household with that of the average poor household reveals that the poor do benefit from general income growth, although their incomes do not rise proportionally with the rest: that is, a general income rise of one percent corresponds to an income rise of less than one percent for the poor. Moreover, this effect tends to increase as growth increases, implying that higher growth has tended to raise income inequality between households. These effects have been attenuated in the late 1990s, however. 10. Panel regression analysis of differences in household income growth by state and age cohort reveals the primacy of education as an influence on income growth in Brazil. However, the effect of average education on the income growth of the poor is nil: improvements in basic education among the poor have not (yet) significantly fed into income growth. Further analysis suggests that investments in upper basic education are central both to increasing income growth and attacking inequality. 11. Finally, other factors are important growth determinants for household income. Local infrastructure (e.g., electrification, trash collection) plays a significant role in subsequent income growth. And international trade and political stability also help income growth. These factors vary in their importance to the poor relative to the average. Electrification seems to have been particularly important for the income growth of the poor. iv Part II: Policy Foci 12. Chapter 5 investigates Brazil's positioning to take advantage of the knowledge economy along four dimensions: (1) economic incentives and institutions, (2) the innovation system, (3) human capital, and (4) information infrastructure. International benchmarking reveals that Brazil is particularly badly placed with respect to (2), its innovation system, the consequence of low business R&D, few patent applications, and low revealed comparative advantage in technology. This is despite high spending on tertiary education in Brazil, and a high index of entrepreneurship. On (1), Brazil's otherwise average ranking on economic incentives and institutions is reduced by a low score for govemment effectiveness, poor relative performance on corruption and rule of law, and low trade volumes. On (3), Brazil's suffers from having few technical workers, low tertiary enrollment, and poor labor relations. Analysis of (4) reveals that Brazil has a relatively strong information infrastructure. 13. The authors conclude that, in order to increase its benefit from international knowledge flows and technological innovation, Brazil could explore ways to increase the recruitment of researchers by enterprises, and more generally the R&D links between enterprises and the university sector (e.g., through industrial R&D PhDs). And increased efforts to improve the climate surrounding the protection of intellectual property rights are recommended. 14. However, increased public funding for the tertiary education sector need not follow as an implication of this analysis; indeed Chapter 6, which develops a Brazilian human capital growth strategy, illustrates that the opposite holds for Brazil. The chapter surveys the economic literature on human capital and highly successful episodes in the USA and South Korea, and concludes that such "leapfrogging" in human capital investments do not constitute a viable strategy. Oft-cited examples, such as India's technology initiatives, do not withstand scrutiny. While there is some evidence of knowledge related spillovers, there is little proof of growth-related externalities associated with higher education. And at current levels of graduation from high school, public initiatives to expand higher education do not qualify as non-elitist. 15. Brazil has made an "epidemiological transition"-its health related achievements reflect middle-income status-yet has still to make what may analogously be termed a "pedagogical transition." Chapter 6 recommends a heightened focus on increasing upper primary school completion and secondary school enrollment. Brazil will face declining primary enrollments for demographic reasons, and has the opportunity to plan for a strategy of gradually increasing the resources it devotes to secondary education: a challenge as both primary and secondary education are sub-national responsibilities. New international evidence, corroborated by evidence from Brazil, also suggests increasing pre-primary interventions. In this way, Brazil may at last embark on the "pedagogical transition" that its recent successes in primary education have made it possible to envision. 16. The growth effects of such a strategy will not be immediate. Correlations using time series data for Brazil confirm international evidence: short-term correlations between human capital and growth are as weak as long-term associations are strong. Countries should view human capital policies as instruments of long-term growth, not as devices to spark short-term growth spurts. v 17. Chapter 7 investigates the role of public governance in nurturing private-sector activity, and uses two lenses to examine this facet of the investment climate. The first lens is the regulatory burden on entrepreneurship, in particular pertaining to start-up firms. The second is public policy effects in foreign direct investment (FDI). Both are particularly relevant for innovation, technological progress, and productivity growth, given earlier findings of this volume. And Chapter 7 illustrates that measures of the quality of public governance suggest concern in both these areas. 18. The chapter goes on to enumerate regulations and procedures that impede start-ups. Although financial costs are not great in intemational comparison, Brazil imposes long time delays on its entrepreneurs, increasing the uncertainty and thus the risk of investing. Many of these delays are associated with dealing with the three levels of the federal structure t various ends such as registration, site licenses, and taxation; experience from the US suggests that such complexity is unnecessary. A further area where attention is recommended is the judicial system, where the evidence suggests that increased capital budget, technology, and managerial activism are more promising approaches than traditional reforms emphasizing staff numbers, remuneration, and training. 19. Chapter 7 also illustrates the link between public govemance factors and FDI flows. Despite large flows into a limited number of non-tradable service sectors (notably telecom and financial services), there are large untapped efficiency gains in such sectors as food and retailing, which FDI could help to capture. Moreover, governance considerations are shown to be a factor in Brazil's low volume of export-oriented FDI, which is more sensitive to the investment climate than is market-seeking FDI. Improved governance may therefore also be seen as one tool to improve Brazil's extemal accounts and hence its long-term sustainable growth rate. Levelheaded Hopes 20. By focusing on a few key areas suggested by data analysis as important, the report hopes to avoid the pitfall of recommending "everything all at once." While it is impossible to quantify the growth effects of every policy intervention, the report nonetheless suggests a list of priority actions focusing on the investment climate (short term, physical investment), the innovation system (medium term, TFP growth), and secondaiy education (long term, human capital). 21. Returning to Chapter 1, this volume cautions against Panglossian expectations of the benefits that will flow from a comprehensive economic growth program along the lines described above. Since 1980, very few large economies have sustained growth averages above 5 percent. In Latin America, only Chile has sustained decadal growth averages above 6 percent. With the significant microeconomic reforms suggested here, and conditional on continued macroeconomic stability (low inflation and measures to reduce total public liabilities over the long term), we estimate that Brazil could attain sustained GDP growth in the vicinity of 6 percent. Table El: Investigation Strategy Volume - Part I-New Data, Part 3-Pollicy Foci Chapter 2 3 4 5 6 7 Countries Companies Households Innovation Skills Institutions -Possible Growth Constraint Cost of Brazil's high capital costl Causality between Have capitaHntensive Firms' assessment of Capital owing to fiscal deficits and savings or investment firms grown less or capital costs as an high spreads, reduces and growth larger firms more? obstade to investing investment and thus growth Human Capital Brazlrs historically low health Correlabons between Unk betwen sklls . Does ntal education Description of Brazils Detaied assessment Firms' assessment of (HC) or education levels constrain human capital -mix and tf- gr6wUi A by state-& age cohort HC in the context of ofBrazll's tIC mIxn . human capital as an growth through both measures and growth - ;- cause growt?. = knowbdge flows an Inirnatonal ight obstade to invesfing Innovation and investment :. k f an n a lih o t investing Knowledge Brazirs economic isolabtion Effects of foreign Descriptbof *o - Have HC policies FDI composition re and innovation system capital or intemnediates bottenecksln the favored knowledge- other FDI recipients. reduces knowledge flows, on productivity Innovation system based growth? FI Usurvey evidence innovation,and thus TFP - _ -_ . . . . Labor Brazil's labor oode and dual Shrinking versus Does formal sector extent Assessment of Firms assessment of Regulation labor market deter growing firms' labor affect income growth? intemational evidence labor regulations as an Investment and productivity costs on training initiatives obstacle to investment grawth . . Macro-Trade Brazirs tariffs, bureaucracy Trade and tariffs Eflects of export Effect of measures of Description of Brazirs Extent of export- and taxation have a net indicators and their orientatfon or foRnign trade connectedness on trade regime and its oriented FDI negative effect on the trade relation to growth penetratin on TFP state growth effects on innovation acoDunt limiting growth ._...._._. Investment Public investment and/or Does public/private Do resources flow to _Local Infrastructure Extent of public versus Firms assessment of Composition rent-seeking in private composition in Brazil less productive uses? affects household private innovaton infrastructure as an & investment divert resources affect growth? Income growth? activies obstade to Investing Infrastructure to unproductive uses Investment Bureaucratic barriers to Do govemance Do new firms show Does policy or politcal Descripton of Do companies reporti Climate entrepreneurship hinder measures matter In higher productivity? instability at the state investment dimate greater obstacles than economic dynamism and growth regressions? Do small frmns -grow? level explain growth? instituons overseis? constrictthe SME sector . m Macro Remaining uncertainty in FGranger causaton Time pattern of fli- Time patter of Firms cite InstabilIty-; Stability economic expectations between stability and level productivity household Income as a maln obstacbe to deters tong-term planning growth?. growth- growth - Investing? ; and investment .._...___... ._ .._._._ Demographics Congestion, crime and other Has growth slowed more and Migration constraints in large cities in immigration recipient have curtailed productivity states? _ growth through mgration Taxation High overal tax burden, hlgh Tax and spending Do firms report taxes as marginal taxes, and measures in growth a main obstade to distortions impede growth regressions invesEng? Isignificantty Dark Yellow (13): strong evidence for reform. Light Yellow (11): weak or partial evidence for reform. White (15): opposite, contradictory, or no evidence for reform. 1. CONCEPTS, FRAMEWORK, ROADMAP Prepared by Mark Roland Thomas and Indermit Gill Foreword 1.1 This volume is the second of two, and comprises the original background research that underpins Volume I, the policy summary. Volume I draws on the work in this volume, but also on additional sources where these inform the policy choices faced by the authorities. 1.2 This report overall represents an attempt to apply the insights of economic theory and empirical studies to suggest directions for economic policy makers in Brazil. It almost goes without saying that this is an enterprise of some magnitude, so let us first limit the scope and ambitiousness of the task: 1.3 First, any attempt to use theory and data to cast light on questions of economic growth policy is subject two fundamental constraints: (a) the scarcity of agreement regarding the appropriateness of competing theoretical models and thus ultimately the relative importance of different causal factors, and (b) the scarcity of reliable data. Neither of these constraints having been lifted for the study at hand, the aim is to add evidence, inform, provoke, and refocus attention around certain policy issues, rather than provide categorical answers to sweeping questions. We do not wish to add hubris to the list of pitfalls. 1.4 Second, this volume will not address all the public policy areas that impinge on Brazil's growth performance. The report's overall form is the result of consultations that took place during its inception, and which were specifically designed to circumscribe its range. Policy areas were excluded according to three basic criteria: (a) areas that were already the subject of recent existing work or work currently under way, (b) areas where the state of advancement of knowledge was deemed to be sufficient, or where the basic directions of Brazilian public policy were clearly established, and (c) the expressed interest of the study's Brazilian audience.' 1.5 There is therefore a long list of subjects that are not explicitly dwelt upon by the present volume, among them fiscal sustainability and the associated vulnerability to market volatility, tax policy, competition policy, privatization, and infrastructure provision and regulation. By way of example, the first of these, fiscal sustainability, has been the focus of many recent studies The audience is understood primarily to be decision makers in the Federal Govemment, but the report aims to be of interest also to a wider audience that includes other govemment levels, Brazilian civil society, academics, professional economists with an interest in Brazil, and development practitioners in other countries. 2 conducted on behalf of the Brazilian authorities by the World Bank and others, and has therefore been placed beyond the scope of the present study. That is not to say that it should not be a primary concern of the authorities in the pursuit of economic growth-external vulnerability is indeed still the main constraint on Brazil's macroeconomic performance. But rather than say a little that is already known about a wide array of topics, the present study aims to add knowledge in a few areas of microeconomic policy. Background New Brazil, New Economy? 1.6 Much has changed in the past decade in Brazil and in the world economy. It is still somewhat an open question to what extent the three fundamental Brazilian macroeconomic reforms of the 1990s-trade opening, fiscal stabilization, and the eradication of inflation-will be sufficient in the longer term to generate GDP growth rates of the order of 6 percent, say.2 The hypothesis of this report is that there are significant improvements that are open to Brazil in its new, more stable, environment. Indeed, some of these improvements-greater innovation associated with longer-term investments, greater investment in skills and training, the reform of institutions goveming the private sector-have only been conceivable under the clearer ground rules that macro stabilization has brought about. If there has been some debate over how "new" the "new world economy" really is, there can be little debate that the "new Brazilian economy" of the late 1990s does indeed represent a new departure. 1.7 At the same time as the Brazilian landscape has changed, so have the world economy and economists' perceptions of it. "Globalization" is a term that is rarely precisely defined, but in the present context we understand it to mean increased international trade, increased international knowledge flows (in particular pertaining to technological innovations), and greater capital mobility. (It is usually more enlightening to consider these three changes in incentives and constraints as separate entities, since each entails distinct consequences.) As these changes have set in, economists' models of growth have shifted their emphasis from factor accumulation towards the hamessing of knowledge and innovation to enhance productivity. A key factor in this context is the openness of economies, both to trade in goods and services and to foreign investment. 1.8 This report therefore focuses on these ingredients of economic growth, and asks what conclusions for Brazil's public policies might look different in the new environment. The report examines Brazil and other countries' recent experiences and looks forward to the economic circumstances that Brazil can reasonably expect to obtain in the coming years. Given these, it designs policy recommendations that are founded upon Brazilian experience and realities, and are reasonable given Brazil's political and economic constraints. 2 At the time of writing, most observers put Brazil's potential GDP growth rate at about 4.5 to 5 percent, that is, higher growth than this would lead either to inflation or to a balance of payments crisis, and thus not be sustained. Brazil is expected to grow under 3 percent in 2001 owing to an energy shortage. 3 Economic Approach 1.9 How have economists studied the causes of economic growth? Broadly, one discems two types of study that encompass the majority of the research program since Harrod, Domar, and Solow invented the modem form of the subject in the middle of the 20" century. The first category is "within-country" inquiries in developed countries. Solow's 1956 paper on the USA is an obvious example, and there are of course many others. The second category is cross-country comparisons, which have been undertaken for the developed and the developing world. The debate surrounding economic convergence is one instance of this category. We echo the second approach in chapter 2, but in the rest of this volume, the emphasis will rather be on developing a less common product: a within-country enquiry for a developing country with the focus on govemment action. What is the context of such action in Brazil in 2001 ? 1.10 The Brazilian economy underwent three structural changes during the 1990s, changes that progressively transformed Brazil from inward-oriented, inflation-prone, and crisis-vulnerable, to open, price-stable, and economically well managed. The task is not complete, but any assessment must recognize how far Brazil has come since 1990. First, the Collor reforms of 1990-92 put in place measures opening Brazilian industry to more extemal competitive pressures: Brazil embraced Mercosul and its average extemal tariff fell from 51 percent in 1987 to about 14 percent in 1993 (it is still about this level at the time of writing). Next, in 1994, came the Piano Real, well documented elsewhere (see Cardoso and Helwege, 1999, for a succinct but full treatment). Finally, in January 1999, came the forced but successfully navigated abandonment of the exchange-rate anchor in favor of macroeconomic management centered on inflation targeting and a necessarily greater degree of credibility of fiscal and monetary management. 1.11 Against this backdrop, and in particular with the combination of stability, openness, and exchange-rate depreciation, many Brazilians have expressed dissatisfaction with the country's economic growth performance, currently forecast to remain around the 4 percent mark for the next two to three years (and this with a high degree of uncertainty). Recent macro performance has suggested that current account imbalances make higher growth than this unsustainable even in the medium term. In recent years, other middle-income countres have been able to post significantly higher growth rates-above 6 percent over a number of years in the cases of Chile or some of the East Asian economies-without incurring such extemal imbalances. 1.12 This prognosis provokes a range of questions: What is Brazil's potential growth rate with the "right" reforms? Is the main obstacle still the high base interest rate arising from fiscal imbalance? Has Brazil really reformed adequately in areas such as trade and infrastructure regulation? Should Brazil intensify industrial policy to create competitive advantage in higher-technology sectors? What should be Brazil's reaction to the challenges of the "new economy" of knowledge and innovation represented by the intemet? Which areas of education and training promise the highest growth returns in this new environment? How can the country be sure that the right policies for growth do not militate against the poorest groups of society? These are the classes of question this report proposes to address. 1.13 We close this introductory section by setting expectations, regarding both the report and the Brazilian economy. First, as already emphasized, the report cannot answer the most sweeping questions about prioritizing public reforms in Brazil. Is tax reform more important than labor-market 4 reform for economic growth? We do not know. Both are no doubt necessary for the country to fulfill its potential. There is neither the data nor the theoretical consensus to provide a convincing case in favor or against any particular quantification of such priorities. Where the report will focus, however, it will bring to light new evidence specific to Brazil, interpret this evidence within a framework appropriate for Brazil, and make proposals that are realistic for Brazil. 1.14 Second, with regard to the performance of the Brazilian economy, a quick look at international patterns since 1980 suggests a levelheaded discussion of growth targets for Brazil. Very few large economies have sustained decadal growth averages above 5 percent, and even when considering these, comparisons across space and time must of course be made with some skepticism. Table 1.1 displays decade growth averages for 1980-90,1985-95, and 1990-2000 for some star performers in addition to some regional comparisons for Brazil. In Latin America, only Chile sustained average growth rates above 6 percent. Even outside the region such cases are rare, and it is unclear how much one can generalize from, say, China to Brazil. Even with fairly significant microeconomic reforms, we estimate that Brazil should not count on sustained growth rates beyond the vicinity of 6 percent. Table 1.1 Selected Decade GDP Growth Averages @ ' I L j ,, om-In _ .M!. V 1 _ China 9.4 10.0 9.7 Korea 7.8 8.5 6.4 Indonesia 6.4 7.0 4.6 Ireland 3.1 4.0 6.8 Chile 4.2 7.4 6.4 Argentina -0.9 2.5 4.4 Mexico 2.2 2.5 3.3 Brazil 2.9 2.9 1.8 Source: Wodd Bank (Live Database) 1.15 The next two sections build the conceptual framework by distilling current thinking among economists working on economic growth, and casting the debate in a Brazilian light. The two sections after those discuss some hypotheses about current constraints on Brazilian economic growth and outline how one might then bring data to these questions within our framework. The final section gives a roadmap to the remaining chapters. Fifty Years is Not Enough 1.16 Robert Solow, in the 1956 paper that gave birth to the modem subject of economic growth and won the Nobel Prize for Economics, showed how growth through capital deepening could simply be viewed as a social choice between consumption and saving (and investment). This view is undoubtedly still relevant for Brazil today, exhibiting as Brazil does quite low savings rates, private and public. Solow's principal insight, however, lay in what his steady-state model did not explain. Long-run growth in the Solow growth model peters out in the absence of exogenous technological progress, which is needed to augment the level of output attainable with a fixed level 5 of factor inputs: labor, capital, and, in later incarnations of this model, human capital. Since the experience of the developed world and parts of the developing world in the 19" and 20e centuries suggests that long-run economic growth is not petering out, Solow's discussion in fact focuses attention on such increases in so-called "total factor productivity" (TFP) as the source of sustained economic growth.3 1.17 Solow's current discourse reflects the two fundamental insights that his model generates: first, that TFP should be the object of economists' attention; second, that over the horizon of most political cycles, the assumption that the economy has reached a steady state in which the nation's capital stock would remain constant in the absence of TFP growth is almost certainly wrong. "If they mean anything at all, those many right-hand side variables in growth regressions are determinants of TFP. But then they should be selected with that function in mind, and TFP (or its growth rate) should be the left-side variable..." "... many countries, much of the time, are nowhere near steady-state growth. This suggests that comparative studies should focus less on 'The" growth rate and more on comparing and understanding whole time paths." Robert M. Solow, 26 February 2001 Keynote speech made at the World Bank conference: "What have we learned from a decade of empirical research on growth?" 1.18 Both these insights formed the basis of the departures of Romer (1986), Lucas (1988), and Mankiw, Romer, and Weil (1992). These authors augmented Solow's treatment of physical capital deepening with (at least) three observations. First, that technological progress could be viewed as a by-product of production or of investment itself (leaming by doing), and that this assumption was sufficient to give long-run growth as a steady state phenomenon. Second, that human capital accumulation also increases the productivity of the other factors (capital and labor), and that simply incorporating this into Solow's model and examining its convergence to steady state gave much improved predictive performance. And third, that these ingredients could be combined to give a view of human capital accumulation that is also driven by learning by doing. These authors thus shifted attention onto human capital and thus from a policy perspective onto education, workforce characteristics, training, labor-market relations, and on-the-job learning as significant determinants of economies' growth performance. These factors will be addressed by the present study in the Brazilian context, and in particular by chapter 6. 1.19 Other authors have analyzed episodes of economic growth in countries in the light of the theories referred to above. Young (1995) characterizes post-war growth in the East Asian economies. Hong Kong's growth can be attributed more to sustained improvements in TFP than that of almost any other country. Jorgenson and Stiroh (2000) also make the case that the strong performance of the US economy in the late 1990s was significantly drven by technological progress increasing TFP. These cases could be contrasted with, say, Indian or Russian postwar 3 Grossman and Helpman (1991) point out that 'classical writers such as Mill and Marx had speculated that standards of living could not rise indefinitely unless advances in technology served to augment the productivity of resources." 6 growth: between 1970 and 1990 TFP in these economies did not improve, and economic growth was entirely accounted for by capital deepening (Young, 1993). Japan or South Korea seem to have been intermediate cases, with moderate TFP growth allied with fast capital (and human capital) deepening. Turning to Brazil, it is of note that Abreu and Verner (1997) find that Brazilian growth during the period 1930-94 could be explained predominantly as a phenomenon of physical capital accumulation: human capital and total factor productivity did not explain a large part of Brazil's rapid growth during the period. We return to this matter in Chapter 2. 1.20 This discussion raises the possibility of "different kinds" of per capita economic growth. A schematic representation is shown in Figure 1.1. The issue in distinguishing between these scenarios is long-run sustainability. If the above typology is correct, then a quick look at what has happened during the 1990s to Russia and Japan is salutary. The reasons why economic growth in the absence of TFP growth may not be sustainable echoes Solow. Without TFP growth, fast economic growth is not an equilibrium phenomenon. It is possible that economic growth occurs "off the equilibrium path"-if past impediments are removed or if structural parameters such as the savings rate or human capital stock change. But unless such changes occur continually, such growth should be expected to peter out. Another way of seeing the argument is through the prism of a small open economy. Growth in output per worker that is not founded in TFP growth must follow from capital deepening. Ceteris paribus in an open economy (in particular holding the domestic savings rate constant), this implies greater capital inflows. These in tum imply a higher interest rate than the international rate or an appreciating exchange rate. Neither is an equilibrium state. Figure 1.1 Schematic Forms of Economic Growth Composition 'HongKong'_, i 'Japan' - T7- EW 'Russia' 0 w.s 1_~ K S *7 0 Labor Growth U Capital Deepening 0 TFP Growth 1.21 It is the premise of this report that the Brazil of the new millennium will only return to the growth rates it enjoyed during the "miracle years" if it finds ways to enhance TFP through technology innovation or adaptation, and through increasing the quality of its human capital stock. Turning to cross country evidence that includes the developing countries, Easterly and Levine (2001) echo Solow and list five stylized facts about growth: * Factor accumulation accounts for much less of the differences between countries' growth rates than does TFP * Countries have tended to diverge economically rather than converge 7 * Growth is not persistent over time: "Much of the world is characterized by miracles and disasters... and not by countries with stable long-run growth rates." * Factors of production and hence economic activity tend to be highly spatially concentrated * National policies influence long-run growth: "In models that emphasize TFP growth, national policies that enhance the efficiency of capital or labor or alter the rate of technological change can accelerate long-run growth." 1.22 This report will also present some cross-country findings (chapter 2) in a descriptive spirit, as background to the more in-depth analysis of episodes in one country, which Solow recommends. As many commentators have pointed out, cross-country growth regressions are plagued by issues of reverse causality (from growth to the supposedly independent variables on the right-hand side), omitted variable bias, and difficulties of interpretation. Nonetheless, performing within country analysis without looking at international evidence always begs the question whether the latter broadly corroborates the former. In particular, chapter 2 will focus on the fifth of Easterly and Levine's stylized facts above, and examine, through regression analysis, the descriptive partial correlation between national policies and economic growth. After this initial presentation of international evidence, the other two empirical chapters of the report will dwell on the recent growth experience of Brazil, again attempting to trace possible policy effects. The final three chapters then turn to more detailed discussions of the policies themselves, with attention to the Brazilian context, yet drawing mainly on international lessons. Policy as a Determinant of TFP 1.23 The discussion so far has led us to formulate public policies as determinants of TFP growth, which in turn it is argued is the source of sustainable long-run economic growth. Yet how are we to conceptualize the effects of public policies on TFP? 1.24 While "policy as a determinant of TFP" has perhaps not explicitly been the message of much of the economic growth literature, it is easy to adapt modern growth theory to this end. First, policy may affect TFP through its effects on innovation in the economy: we turn to this below. Second, public policy may affect the returns to factors in the economy, affecting the investment decisions of agents and thereby TFP. Third, policy may be thought of as directly creating certain factors, such as in the case of public education. 1.25 Klenow (1998) has argued that the evidence of within country growth experience favors the view that ideas and thus innovation are behind TFP growth.4 It is worth emphasizing that such innovation need not necessarily be '"echnological" but may for instance be embodied in improvements to managerial techniques. McKinsey Global Institute (1999) recently completed a detailed analysis of Brazilian productivity based on case studies in selected industries, and came to the conclusion that the major inefficiencies reside in the lack of penetration of international best practice in management methods, partly owing to a lack of competition in certain sectors, whether 4Barro (2000) gives a theoretical account of how R&D may be built into a growth accounting framework, in the context of Grossman and Helpman style models of innovation-led growth. 8 domestic (suggesting changes to competition policy)5 or foreign (suggesting further trade opening).6 1.26 Many models of endogenous growth, in particular those of Romer (e.g., 1986), give a critical role for policy in determining the location of activities and subsequent agglomeration effects. In Romers words, policy can be "hypercritical": U. . small changes in [choices] can now imply much bigger changes in equilibrium outcomes. So, if you build models that have complementarity plus fixed costs, you could have, in the aggregate, forms of increasing retums and positive feedback which make policy much more effective." 1.27 Concluding that policy may have a critical role is not the same as detecting its effects. Indeed, in cross-country studies issues of endogeneity make such inferences of causal relations impossible to make. This leads to Solow and others' recommendation of the study of country "episodes" or experiences across periods of time. This will be the spirit of chapters 3-7 of the present volume. 1.28 Perhaps the clearest exposition of a class of models in which public policy and institutions can have these kinds of critical effects have been developed by Grossman and Helpman (many of their models are collected in Grossman and Helpman, 1991). Their discussions emphasize technological innovation and adaptation (or imitation) as equilibrium activities undertaken by firms who gain monopolistic rents from the fruits of their research and development activities. Note that the monopolistic competition that these models assume may follow from similar assumptions to Romer's fixed costs and increasing retums (cf. Models of economic geography and path dependency such as Krugman, 1991 a, 1991 b). Within this class of models, economic growth is driven by the level of innovative (and adaptive) activities in the economy, but this innovation is an equilibrium choice by firms rather than simply a by-product of investment or production. The role for policy may in tum depend on the scope of knowledge spillover effects (e.g., national or international in scope), on human capital and other resource endowments, and on country size. Perhaps more important, the types of policy that govemment's may contemplate may differ from those suggested by the models so far, as we discuss in the next section. 5 A recent OECD (2000) assessment of Brazilian competition policy was, however, broadly positive, and suggested that Brazil was not an outlier in this dimension with respect to many benter-performing developing economies. 6 It is of note that despite its drastically reducing trade barriers in the early 1990s, Brazil's trade regime is still not particularly open by international comparison. Average tariffs have in fact risen slightly since the mid-1990s, owing to renegotiations within Mercosul (although it should be recognized that Brazil does not practice the 'tariff spikes' - extremely high tariffs against narrow groups of products - practiced by many industrial countries, including some of Brazil's trading partners). International trade as a fraction of GDP is still low in Brazil by intemational standards (controlling for other factors such as country size). 9 Knowledge and Innovation in the International Economy 1.29 If innovation, and thus flows of knowledge, drive growth, we must still follow the advice of Robert Solow to "put TFP on the left-hand side of our regressions." The determinants of TFP may now look different, however. 1.30 The first important point to emphasize is that knowledge is in many cases a public good: its use by one innovator does not preclude its use by another. How can emerging economies such as Brazil maximize the flows of knowledge into their economies? 1.31 If knowledge is (at least to some extent) a public good, innovation is nonetheless not a free good, and Grossman and Helpman's models recognize this explicitly. Indeed, it is an empirical fact and a theoretical assumption in the models that innovation is human-capital intensive. This poses a policy quandary for developing countries, in which human capital is typically scarce. Although one class of policy recommendations centers on how to make human capital less scarce in Brazil (see chapter 6), this parameter can only change relatively slowly over time as, say education reforms take approximately a generation to work their way fully into the labor market. One question facing Brazil is therefore whether its scarce human capital is employed to a greater extent in more or less innovative or knowledge generating activities. 1.32 Fortunately for countries scarce in human capital, innovation and thus productivity growth can also be imported via foreign direct investment, foreign machinery, imported intermediates, and even competition in export markets and greater competition in domestic markets from abroad. The vectors of this information flow are thus trade, foreign investment, foreign competition, and information and communication technology. In the new economy, these areas of policy are placed under the spotlight. The Importance of Public Policy 1.33 In this framework, a country's comparative advantage is dynamic: no longer does it follow directly from its endowments of factors such as labor and natural resources. Rather, these endowments set the initial conditions that are but one parameter facing the country. Govemment policies towards different sectors of the economy or factors may induce implicit or explicit subsidies or taxes, and these may permanently affect the pattern of specializatiqn of the country. As Grossman and Helpman (1991) note, aHistorical data also link various govemment policies with exceptional growth performance. Countries with high shares of govemment consumption in GDP have on average grown more slowly than others, whereas those with high rates of govemment investment have tended to grow more rapidly. High marginal tax rates are associated with slow growth in output, holding constant the average rate of taxation. Kuznets (1988) tries to identify the common features in the successful growth performances of Japan, Taiwan, and South Korea, and concludes that all three countries have pursued a policy of encouraging the corporate sector and of removing regulatory restrictions on business activity. Finally, several researchers 10 find a strong relationship between a countrys trade policy regime and its dynamic performance." 1.34 Table 1.2 illustrates the central policy emphases that the three main waves of economic growth theory since Solow, and briefly outlined in this section, have suggested. Each in some way builds upon and generalizes the earlier work. Accordingly, the report plans to examine policy options using all of these "lenses." But the role of policy in the new growth agenda is richer than before, and the costs of the wrong policy choices have likely been raised. What can or should govemment do? 1.35 It should be stressed that economic theory does not provide straight answers to the riddles of economic growth and how to attain it through government policy. Indeed, in the more sophisticated models of the new growth theories-incorporating knowledge spillovers both national and international in scope, small and large countries, and sectors with and without innovation- almost any desired result can be generated by tailoring the assumptions. The onus thus shifts back onto the data and the judicious interpretation of country experience to suggest the right mix of public interventions. Table 1.2 Policies Emphasized by Different Classes of Growth Model -Fin 1 . II Exogenous Growth Saving rates Total factor productivity may increase from Solow (1956) Capital deepening investment in basic research Shell (1967) Exogenous technological Suggests government investment versus progress govemment consumption Basic research High fiscal deficits will reduce growth Growth rates determined by propensity to save and marginal capital product Forced saving may increase growth but not necessarily welfare Endogenous Growth Leaming by doing Importance of education for growth Romer (1986) Human Capital Possible gains from promoting investments in Lucas (1988) Economies of scale human or physical capital Mankiw, Romer & Weil Investment externalites High marginal tax rates reduce growth (1992) Gains from openness Persistent effects of good or bad policy Technological Growth Imperfect competition Past production patterns may affect current level Grossman & Helpman Investment in R&D of innovation (1991,1993) Investment externalities Importance of international knowledge spillovers Aghion & Hewitt (1992) Imitation for developing countries Dynamic comparative advantage R&D activities divert resources First mover advantage Gains from openness Strategic role played by workforce skills I Possibility of strategic trade and investment policy History and Hypotheses 1.36 Brazil was by some measures the fastest growing economy of the 20th Century. It was by all measures one of the world's most dynamic economies in the 1960s and 1970s, prompting references to the "Brazilian Economic Miracle" of that period. What caused economic growth to cease? Are the same factors responsible for Brazil's seemingly sluggish response to stabilization in the 1990s? And today, what are Brazil's 'binding constraints' with regard to economic growth? 22. Reflecting the discussion so far, this report considers growth determinants traditionally emphasized, such as the level of physical investment (private and public), saving, macroeconomic balance, govemment consumption, and health and education. But we also emphasizes factors that the newer theories suggest may be important: the environment for knowledge creation and the adoption of new technologies, the analysis of tertiary versus basic education, and the dynamic effects of foreign direct investment. Observers differ in their rankings, but the following is a list of hypotheses worthy of discussion. It should be noted that these are hypotheses: their relevance or lack thereof is to be determined by appealing to empirical observation: * Cost of capital and crowding out of private investment by the public sector borrowing requirement? * Human capital deficiencies, particularly relating to health, primary, secondary, or tertiary education and training? * Deficiencies in the flow of knowledge and innovation, related to intellectual property rights and/or use of foreign technology? * Overly rigid and costly formal-sector labor-market regulations driving up the cost of labor, including lengthy and unpredictable labor court delays? * High tariffs and relative isolation from international trade flows? High public-sector investment composition combined with the low-productivity of public-sector investments and/or a lack of investment in infrastructure? * Poor climate for private investment, including regulatory and administrative barriers and an inefficient judicial system? * Macroeconomic instability, including rollover risk of extemal debt, remaining uncertainty surrounding public debt sustainability and inflation? * Constraints related to demographics and migration, including the decrease in migratory flows from rural to metropolitan areas, and from the Northeast to the South and Southeast regions? * High and distortionary taxes, including cascading taxes on exports, a financial transactions tax, and an overall tax burden greater than 30 percent of GDP? 12 Data Analysis as Description 1.37 In this volume we present data-whether from countries, firms, households, or episodic accounts-in a descriptive spirit. To use an analogy that we owe to Paul Romer, we are not so much testing models as building up evidence in the way of a detective solving a crime. The evidence we amass will be at best suggestive and circumstantial. We hope it will nonetheless provide the platform for thinking about sensible policies for Brazil's economic growth. Country Data 1.38 An initial question is: where does Brazil fit in the growth data? As Figure 1.2 illustrates, Brazil has been through periods of both high and low growth during the past 50 years. Until 1980, Brazil was generally cited as a positive example for developing countries to emulate when designing growth policies. For the period 1966-76, Brazilian GDP growth averaged above 9 percent. Two external events changed the story. The first was the 1973 oil shock, which put in place underlying imbalances that would cost Brazil dearly by 1982, when the first debt crisis pulled Mexico and then Brazil into brutally severe recessions. As has been well documented, the remainder of the 1980s and the early 1990s were a period of stagnation for Brazil, with per capita growth rates around zero averaging over the period. The establishment of price stability since 1993 has allowed the country to retum to positive growth territory, but there are no signs as yet of the potential to grow at the sustained levels of 6 percent and above that some of the East Asian economies, Chile, and certain European economies such as Spain, Portugal, and Ireland have exhibited since 1980. Figure 1.2 Two Periods of Brazilian Economic Growth in Context Kore I S n a i I~~~~~ 13 _X= - - ~~~~~~~~~~~~~~~~~Brazil Source: World Bank (Live Database) 1.39 A second question might be: where does Brazilian policy fit in the data? We recognize limitations in our ability to answer this question quantitatively: policy comparisons across countries are not readily available. Moreover, policy coefficients from regressions are not enough in themselves to draw conclusions, owing to the fact that both policy and income growth are jointly determined. Nonetheless, in the spirit of analysis as description alluded to at the beginning of this section, it is desirable to analyze Brazilian policy and growth through the lens of some cross- comparisons. Company Data 1.40 A further source of information on the processes of innovation and growth is firms themselves. For example, Olley and Pakes (1996) find that for the US telecom industry, 'the productivity increases that followed deregulation were primarily a result of a reallocation of capital towards more productive establishments." Findings of this nature for Brazil would suggest the mechanisms of recent productivity growth, important for understanding how this growth might be increased. State Data 1.41 A further feature of growth emphasized by Easterly and Levine is the concentration of activity, both internationally and within countries. This certainly characterizes Brazil: with a combined population of 49 million, or 30 percent of Brazil's total, the two states of Sao Paulo and Rio de Janeiro account for about half of the country's GDP. Table 1.3 shows indices of population and economic activity by region for Brazil in 1998. Table 1.3 Indicators of Concentration in Brazil's Regions Southeast 43 58 63 65 Northeast 28 13 10 13 South 15 17 21 11 North 7 4 _ 4 __2 Center-West 7 7 2 _ _ 9. (column percent, may not add to 100 owing to rounding.) Source: IBGE 14 1.42 Economic concentration is of course closely related to the issue of the distribution of income, which may be viewed (World Bank, 2000a) as one important dimension of the "quality of growth." Are there tradeoffs between quantity and quality? Are there public policies that promote more equitable growth? While these are not the central issues of this report, it is appropriate in a treatment of Brazilian growth that some attention is paid to matters of distribution. 1.43 If growth is unevenly distributed in Brazil geographically, people can of course migrate. Indeed Brazil's economic history has been marked by waves of migration of the sertanejos of the Northeast to the industrial centers of the Southeast. This migration continues to some extent but recent research (Fiess and Verner, 2001) suggests that the relationship between the regions in this regard has changed in the past 20 years, with migratory flows even returning towards the Northeastem cities from the Southeast. 1.44 If workers only migrate slowly owing to cultural or economic barriers (such as the fixed costs), capital could also migrate, potentially more rapidly. Yet it is clear that in Brazil this does not occur to any great degree: why not? The economics literature suggests two possible (and not mutually exclusive) explanations: local policies and complementary inputs (i.e., agglomeration). Unfortunately, the state of the art in econometrics cannot satisfactorily distinguish between these two hypotheses, so from the policy perspective the appropriate advice would seemingly be to hedge and assume that both are at work. The challenge is then to define what is "good policy," and at a second stage to ask in what activities the country and its regions may have comparative advantage in which to specialize, in the hope of then benefiting from the concentration of economic activity in Brazil (or in its poorer regions). Other Work 1.45 Finally in the context of factors that affect the income distribution, it is worth situating the present study in the context of two other imminent World Bank studies that to some extent delineate its boundaries, particularly with regard to questions of income distribution. The first will address reforms to Brazil's labor market regulations (forthcoming, 2002), the second the causes and consequences of Brazil's high levels of inequality (forthcoming, 2001). The present volume does not, therefore, propose an in depth treatment of either of these issues, despite their obvious relevance to economic growth. Choosing Episodes to Understand Policy Effects 1.46 Despite the desirability of presenting and analyzing data as a descriptive device, ambiguity about causality and the ever-present problem of omitted variables will always mean that such evidence is unconvincing in isolation. The second approach we wish to follow is then to focus on country episodes as evidence of what works and does not in the various areas the report isolates as important for growth in the new Brazilian economy: the nexus of innovation, trade, skills, productivity, and investment. 15 Knowledge and Innovation 1.47 Again, the central hypothesis is that innovation, adaptation, or in general the use of knowledge as an input in production is central to increasing the rate of growth of TFP. Recent work at the World Bank emphasizes four "pillars" that support this process: economic incentives and institutions; education and skills; information infrastructure; and the innovation climate. It is not enough to focus on any subset of these without the others to foster the optimal environment for knowledge-based economic growth. Human Capital and Innovation 1.48 The interaction between innovation and human capital is a particularly thomy one for policy- makers in developing economies, since they are operating under competing constraints and objectives. On the one hand, human capital is scarce, despite the investments that many countries including Brazil are making in this area. On the other, governments often aim to foster at least two types of activity through public policies: the first consists of poles of industrial activity or national champions, often pursued through explicit industrial policies and subsidies; the second is research and development itself, often pursued through subsidies to higher education and science and technology initiatives. It is far from clear in an environment of scarce human capital that all these objectives can be met at once. The development of a growth strategy therefore calls for a setting of priorities among these aims. The right balance of policies for education, training, and industrial or science and technology policy should then follow from this strategy. The Investment Climate 1.49 Finally, much technological progress is undoubtedly embodied in new investment in physical capital. Physical investment is thus a necessary condition for innovation to lead to growth. 7 Yet capital formation is also sensitive to the institutional environment, both private and public, in which it is embedded. The recent passing of a new corporate law in Brazil, which among other objectives brings in line with intemational practice the levels of protection afforded to minority shareholders in Brazilian firms, reflects this realization. The role of public institutions-local government, the courts, property rights, bureaucratic impediments to business-is less well understood in the Brazilian context.8 This therefore forms another area in which the present report aims to develop recommendations. 1.50 The preceding discussion defines the locus of our analysis. The next few paragraphs describe the chapters to come in this volume. Chapters 2, 3, and 4 present and analyze new data. Chapters 5, 6, and 7 present and analyze episodes and areas of policy. Volume I draws on this 7 If not a sufficient condition: witness the earlier discussion relating to TFP, and Easterly and Levine, 2001. 8 There are some significant exceptions to this statement. Castelar (2001), for example, provides an in depth analysis of the effects of the judicial system on economic growth in Brazil. 16 and other work, and within the framework described earlier in this section presents policy proposals to increase economic growth in Brazil. 1.51 In chapter 2, Norman Loayza9 presents new cross country evidence and macroeconomic data from Brazil. He looks at growth patterns using time series and panel data that has been updated and expanded for the purpose of this report, in light of the framework described above. This situates Brazil's experience in a global context and establishes some correlates of growth, including aspects of govemment policy. 1.52 In chapter 3, Marc-Andreas Muendler,'° Luis Serv6n,"1 and Claudia Paz Sepulveda'2 use Brazilian firm data to try to understand recent productivity growth. The data are a newly constructed panel from 1986-98 for Brazilian firms (collected annually by the Brazilian Statistical Institute"3) and the authors examine by sector the importance of factors such as export orientation, firm size, capital intensity, technology, and human capital in determining changes in total factor- productivity during the period, with particular emphasis on the 1990s since the real plan. 1.53 In chapter 4, Naercio Menezes-Filho'4 and Mark Roland Thomas'" analyze Brazilian state economic growth for the period 1981-98. The authors analyze both the quantity and one aspect of the quality of recent Brazilian growth: its incidence among the poor. They use repeated cross- section household data"8 to analyze variations in income growth across states and age cohorts for the poor and non-poor. Their methodology gives a novel description of the patterns of Brazilian growth, establishes the correlates of pro-poor growth, and allows an investigation of spatial impediments to income growth (c.f. Jalan and Ravallion, 2000). 1.54 In chapter 5, Carl Dahlman, Anuja Adhar Utz, Jean-Eric Aubert,'7 and Christine Zhen-Wei Qiang"8 ask the question: how can Brazil increase technological innovation and gain access to the most dynamic markets in the global economy? This chapter examines improvements needed according to the four "pillars of the knowledge economy" described above: economic incentives and institutions; information infrastructure; human resources; and the innovation system. The authors highlight the lessons of international experience particularly in the areas they categorize as the innovation system. 9 Central Bank of Chile and the World Bank. 10 University of Califomia at Berkeley. " Office of the Chief Economist, Latin America and the Caribbean Region, The World Bank. 12 Economic Policy Unit, Latin America and the Caribbean Region, The World Bank. 3 Instituto Brasileiro de Geografia e Estatisticas: IBGE. 14 Faculdade de Economia e Administracao, University of Sao Paulo. 15 Brazil Unit, Latin America and the Caribbean Region, The World Bank. 16 Stacked annual data from the national household survey (Pesquisa Nacional por Amostra de Domicflios: PNAD). 17 World Bank Institute. la Telecom Policy Division, The World Bank. 17 1.55 In chapter 6, Indermit Gill"9 and Harry Patrinos" ask: what type of human capital investments will promote skills formation for innovation and competitiveness while at the same time ensuring equitable distribution of the gains from economic growth in Brazil? The authors examine policies and public expenditures towards health, education, and training, using international experience in the OECD and East Asia to develop a strategy appropriate for Brazil in this area. They develop policy proposals that take into account both Brazil's short-term constraints and the significant progress the country has made in improving health and basic education in recent years. 1.56 In chapter 7, Hoon Soh2" asks: how can Brazil improve its institutions to encourage sustained capital investment? The author discusses the effects of public govemance on domestic investment, and on the size and composition of FDI. This leads to a characterization of the dimensions of the investment climate that public policy should focus on in order to increase both the quantity and the allocative efficiency of capital investments in Brazil. References Abreu, Marcelo and Dorte Verner (1997) Long-Term Brazilian Economic Growth: 1930-94, Paris: OECD. Aghion, Philippe and Peter Hewitt (1992) "A Model of Growth through Creative Destruction," Econometrica 60:2 pp. 323-51. Barro, Robert (2000) "Notes on Growth Accounting," NBER Working Paper 6654. Cardoso, Eliana and Ann Helwege (1999) "Currency Crises in the 1990s: The Case of Brazil," IPEA Serie SeminArios #13/99, Rio de Janeiro. Easterly, William and Ross Levine (2000) "It's Not Factor Accumulation: Stylized Facts and Growth Models," working paper, World Bank. Easterly, William, Norman Loayza, and Peter Montiel (1997) "Has Latin America's Post Reform Growth Been Disappointing?" Joumal of Intemational Economics, 43:3-4 pp. 287-311. Fiess, Norbert and Dorte Verner (2001) "Migration in Brazil," World Bank, mimeo. Grossman, Gene M. and Elhanan Helpman (1991) Innovation and Growth in the Global Economy, Cambridge, MA: MIT Press. Grossman, Gene M. and Elhanan Helpman (1993) "Endogenous Innovation in the Theory. of Growth," NBER Working Paper No. 4527, Cambridge MA. Jalan, Jyotsna and Martin Ravallion (2001) "Geographic Poverty Traps? A Micro Model of Consumption Growth in Rural China," working paper, World Bank. 19 Human Development Department, Latin America and the Caribbean Region, The World Bank. 20 Education Network, The World Bank. 21 Africa Technical Department, The World Bank. 18 Jorgenson, Dale W. and Kevin J. Stiroh (2000) "Raising the Speed Limit: US Economic Growth in the Information Age," Brookings Papers on Economic Activity, vol. 0, no. 1, pp. 125-211. Klenow, Peter (1998) "Ideas versus Rival Human Capital: Industry Evidence on Growth Models," Journal of Monetary Economics 40 pp. 2-23. Lucas, Robert E. (1988) "On the Mechanics of Economic Development," Joumal of Monetary Economics, 22:1 pp. 3-42. Mankiw, N. Gregory, David Romer, and David N. Weil (1992) "A Contribution to the Empirics of Economic Growth," Quarterly Joumal of Economics 107:2 pp. 407-37. McKinsey and Company, 1999, Produtividade no Brasil, A Chave do Desenvolvimento Acelerado, Rio de Janeiro: Campus. OECD, 2000, Competition Policy and Regulatory reform in Brazil: A Progress Report. Olley, G. and A. Pakes (1996) "The dynamics of productivity in the telecommunication equipment industry," Econometrica 64:6 pp. 1263-97. Romer, Paul M. (1986) "Increasing Returns and Long-Run Growth," Joumal of Political Economy 94 pp. 1002-37. Shell, Karl (1967) "A Model of Inventive Activity and Capital Accumulation," in K. Shell (ed.) Essays on the Theory of Optimal Economic Growth, Cambridge: MIT Press. Solow, Robert M. (1956) "A Contribution to the Theory of Economic Growth," Quarterly Joumal of Economics 70 pp. 65-94. World Bank (2000a) The Quality of Growth, Washington DC: World Bank. World Bank (forthcoming 2001) "The Causes and Effects of Inequality in Brazil." World Bank (forthcoming 2002) "Brazilian Labor Market Regulation." Young, Alwyn (1993) "Lessons from the East Asian NlCs: A Contrarian View," European Economic Review 38 pp. 964-73. Young, Alwyn (1995) "The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience," Quarterly Joumal of Economics, Vol. 110:3 pp. 641-680. 19 PART I NEW DATA 20 2. AGGREGATE GROWTH IN BRAZIL Prepared by Norman Loayza with the collaboration of Cesar Calder6n tiny 2.1 The objective of this study is to characterize Brazil's economic growth performance in the last four decades. For this, we take a macroeconomic (aggregate) perspective, using regional and world trends as benchmarks for comparison. 2.2 Section II describes the main stylized facts of growth in Brazil. We first review the long-run growth trends in this country and Latin America by decades from the 1960s to the 90s. Then, we examine the sectoral composition of growth in Brazil to determine the extent of its structural transformation. Next, we study the dynamic relationship between saving, investment, and growth, using a VAR methodology applied previously by Attanasio et al.(2000) in a cross-country panel setting. Finally, we attempt a decomposition of growth in Brazil into its sources related to capital accumulation, expansion of the labor force, and total factor productivity growth. 2.3 In section III, we attempt to explain the economic growth performance in Brazil from a cross- country perspective. We follow the approach in Barro and Lee (1994) and Easterly, Loayza, and Montiel (1997), which consists of linking aggregate economic, political, and social variables to growth rates in per capita GDP for a large sample of countries. The estimated model is then used to forecast the growth rate in the country and examine whether its performance has been close to expected values. 2.4 In section IV, we present some economic and social areas of relatively weak performance in Brazil. We argue that improvement in these areas will be accompanied by higher economic growth in the country. Section V presents some concluding remarks. 1. Long-Run Growth Trends in Brazil, Latin America, and the World 2.5 Figure 1 presents the per capita real GDP growth rate in Brazil by decade averages. For comparison purposes, it also presents the growth rates of the median countries in Latin America and the Caribbean (LAC) and the world, respectively. While in the 1960s the growth performance in Brazil was very similar to the typical LAC and world countries, in the 1970s Brazil's growth rate surged to become twice bigger than the world and regional growth rates. This outstanding performance was not to last: Brazil's growth rate decreased drastically in the 1980s and fell even 21 further in the 1 990s. The typical country in LAC did worse than Brazil in the 1 980s (reminding us why this time is called "Latin America's lost decade"); however, while the region recovered in the 1990s, Brazil did not. 2.6 The decline in Brazil's growth rate is an important stylized fact and, as such, must be analyzed. We do this in section 1II, where we take a cross-country-regression approach to explain the changes in growth rates by decades. In this empirical exercise, we mainly consider domestic conditions that drive output growth; however, we also take into account the changing intemational conditions in recent decades. From Figure 1, it is noteworthy the declining growth trend in the world wfth respect to the 1960s and 1970s. This is bound to have some role in explaining why the growth rate in Brazil was smaller in the last two decades. Figure 2.1 Per Capita Income Growth 121 7% - 'r 6%- 2 5% - . 4% 0 (93% ,*,.2% '~ .- CL 1%0 - ~T 1 i 0% U -1% ~60s 70s ~C e 1% EBrazil OLAC OWorld 2. Sector Composition of OuRput GirowthO in BraiZl 2.7 Table 1 presents the average output growth rates of agriculture, industry, and services by decades in Brazil. For industry and services, we also present growth rates by further disaggregated sectors. Industry and services grew strongly in the 1960s and 70s, and agriculture and livestock joined them as a highly growing sector in the 1970s. However, all major sectors decelerated in the last twenty years. The rise and decline of growth in Brazil was a phenomenon shared by all areas in the economy. The fact that the high growth of the 1970s was balanced across sectors but also short-lived may indicate that it was mainly fueled by an expansion in aggregate demand, possibly coming from public sources. We return to this issue in section 1II. 22 Table 2.1 Output Growth Rates Sector 65-99 65-69 70-79 80-89 90-99 Agriculture and Livestock 4.60% 2.25% 8.35% 3.56% 2.46% Industry 4.32% 9.21% 9.17% 1.91% 0.73% Mining and Quarrying 5.12% 5.80% 6.65% 7.51% 0.92% Manufacturing 4.17% 9.34% 9.15% 1.52% 0.34% Construction 4.06% 8.22% 8.79% 0.91% 0.82% Gas, Electricity and Water 7.40% 6.76% 11.64% 6.66% 4.17% Services 4.96% 8.37% 8.35% 3.59% 2.07% Wholesale and retail trade 4.06% 6.53% 7.80% 1.77% 1.61% Transport 5.48% 2.53% 11.04% 3.77% 2.81% Communications 11.69% 4.56% 12.50% 13.09% 12.32% Banking 4.48% 5.05% 12.58% 2.99% -2.36% Public Administration 3.98% 6.96% 7.34% 1.93% 1.49% Other Services 8.72% 21.88% 13.14% 4.34% 3.41% GDP 4.51% 7.24% 8.09% 2.81% 1.70% Source: Instituto Brasileiro de Geograf(a e Estatfstica (IBGE) Diretoria de Pesquisas, Departamento de Contas Nacionais. 2.8 The decline in industrial growth in the 1980s and 1990s was mostly due to the fall in manufacturing, construction, and, less severely, public utilities. Mining and quarrying remained strong in the 1980s but fell with the others in the 1990s. Considering that mining activity has always had a small share of value added in Brazil (see below), its pattem of growth indicates that resource extraction was not the reason behind the expansion and decline of growth in the country. In the service sector, the only area whose growth rate remained high in the last two decades was communications; the others, including, trade, transport, and banking decelerated considerably with respect to the 1970s. 2.9 Table la and Figure 2 show the corresponding sectoral shares in GDP over the last forty years. The share of agriculture decreased gradually from about 11% in the early 1960s to less than 8% in the late 1990s. The share of industry was roughly 40% from the 1960s to the 80s, but then declined to one third of GDP in the 1990s. Within industry, manufacturing remains the strongest sector, but over the years it has lost some ground to construction and utilities. Mining and quarrying has remained a small sector reaching slightly over 1% of GDP in Brazil. The share of services was also roughly the same in the first three decades at about 50% of GDP; in the 1990s, it rose to about 60% at the expense of both agriculture and industry. Within services, banking has become the dominant sector, displacing wholesale and retail trade whose share in GDP has fallen by half since the mid 1960s. Communications has continuously gained a larger share although still remains a small sector. 23 Table 2.1a Structure of Production Sector 65-99 65-69 70-79 80-89 90-99 Agriculture and Livestock 9.4 11.3 10.2 9.3 7.6 Industry 375 38.4 39.4 39.6 33.1 Mining and Quarrying 1.3 1.2 1.0 1.9 1.1 Manufacturing 27.4 29.4 30.5 29.1 21.5 Construction 6.9 6.5 6.4 6.5 8.0 Gas, Electricity and Water 1.9 1.3 1.4 2.1 2.5 Services 53.1 50.2 50.4 51.1 59.3 Wholesale and retail trade 10.5 14.8 13.5 8.5 7.5 Transport 3.7 4.1 3.9 3.8 3.1 Communications 1.1 0.7 0.8 1.0 1.6 Banking 9.8 7.8 7.6 11.4 11.6 Public Administration 9.8 10.0 8.9 7.0 13.5 Other Services 18.2 12.8 15.7 19.5 22.0 Total Value Added 100.0 100.0 100.0 100.0 100.0 Source: Instituto Brasileiro de Geografia e Estatistica (IBGE) Diretoria de Pesquisas, Departamento de Contas Nacionais. Figure 2.2 Composition of GDP, 1960-99 100% - 80% - 0~ O) 60% 20% 0% o% .,.,,,,, CD. . ,,,,,,,,,,0 , to - .0 ( 0, ,D, cO coD co c I - co r- c o co a) ED C o Co a) Co Co co C Co Co Co CD Co Co U Agriculture * Industry o Services 24 3. Growth, Saving, and Investment in Brazil 2.10 Figure 3 shows the annual per capita GDP growth rate and the saving and investment ratios to GDP, for the period 1965-1998. Although there is a fair degree of over-time variation in the investment ratio, its average level seems to have remained similar over the years. The saving ratio also remained stable until the mid-1 990s, but then it dropped drastically, producing a large current account deficit. We expect that in the coming years there will be a correction in either investment or saving ratios to close the extemal gap. Most likely, it will be the saving ratio, which will increase back to its historical levels once the growth rate recovers. Figure 2.3 Saving, Investment, and Growth 0.30 - 0.12 Xo.10- 1 0 t 9 -~~~~~~~~~0.102 o 0.25 - 0.06 0~~~~~~~~~~~~~~~~00 0.20 -.o 0.04 l 0.02 0 ~+0.15 -vOg - -0.02 0.10 ~~~~~~~~~~~~~~~-0.04 C5 0.05 0.006 -0.08 1966 1970 1974 1978 1982 1986 1990 1994 1998 ---Saving - -Investment - Growth 2.11 Contrary to the long-run stability of the investment (and, until recently, saving) rates, the growth rate experienced a level drop in the 1980s (see also Figure 1). This is the case even considering a few years of strong performance in the latter period (such as 1993,1994). When we decompose investment, we notice a slight reduction of its public component in the 1990s and concurrent increase in private investment (see Figure 3a). However, the compositional changes in investment are modest and there remains the indication --confirmed in the econometric exercises that follow-- that the drop of growth cannot be explained by changes in the level or composition of investment. 2.12 We now begin to explore the dynamic relationship between the growth rate and the saving and investment ratios. Following Attanasio, Picci, and Scorcu (2000), we study these relationships by running VAR systems on annual data. We consider two bivariate systems, namely, Growth- Saving and Growth-Investment. The VARs include two lags of each variable (further lags do not enter significantly in the regressions and are, thus, excluded in the final estimated system.) Table 2 presents the results. The dynamic relationship between saving and growth in Brazil appears to be not significant at annual frequencies according to the Granger-causality and forecasting tests. Judging only by the sign of the coefficients, past growth is positively related with future saving, while saving rises in anticipation of future lower growth. The former result can be explained by resorting to the theories of subsistence consumption and habit formation, while the latter follows from a precautionary motive (the "saving-for-a-rainy-day" effect.) However, neither effect is statistically significant. 25 Figure 2.3a Public and Private Investment and Growth 0.25 0.12 0.10 ,q 0.20 ~/ / - - - 0.08 0.06 0.0 I I I I I I I IaII I II -01 0.04 0.15 n ~~~~~~~~~~~~~~~~~0.00~ 0.10- IL -0.02 ~ to 0.004 B 0.05 -~~~~~~~~~~0.06w 0. ~~~~~~~~~~~~~~~~~t-0.08 1970 1974 1978 1982 1986 1990 1994 1998 -.-Private Inv - -Public Inv Growth 2.13 The dynamic relationship between investment and growth at annual frequencies is more interesting. Past growth has a positive link with current investment. This can be explained by considering that past growth creates incentives to new investment by making future growth more likely. On the other hand, past investment has a surprising negative link with current growth. Apparently this result contradicts the cross-country evidence, which finds a positive effect of investment on growth. However, the two results are not necessarily contradictory given that the dynamic relationship estimated here considers only short-run effects (with lags of at most two years) while the cross-country analysis focuses on long periods. Attanasio et al. (2000) and Blomstrom, Lipsey, and Zejan (1996) also find a negative (short-run) link between past investment and current growth. They explain it either by considering that investment is limited by saving, which anticipates growth negatively, or by taking into account that growth behaves cyclically, with high growth and investment preceding low growth. 2.14 Next we consider the dynamic relationship between growth and, respectively, private and public investment. The first result is qualitatively the same as for total domestic investment: a rise in growth precedes an increase in both private and public investment. The second resuit entails some difference between the two components of investment: while an increase in private investment is associated with a future fall in the growth rate, there is no significant effect coming from public investment. Regarding the fit of the regression, the dynamic model can explain public investment (R-square = 0.64) better than private investment (R-square = 0.39). 26 Table 2: Saving, Investment, and Growth in Brazil VAR Estimation, Annual Data 1960-99 Saving Growth Investment Growth Growth (-1) 0.1369' 0.3627*- 0.1296*- 0.3495*- (Std. Error) (0.0734) (0.1134) (0.0490) (0.1191) Growth (-2) -0.0454 0.2133 0.1424-- 0.3619-- (Std. Error) (0.0915) (0.1586) (0.0811) (0.1487) Saving (-1) 0.6994*- 0.0973 (Std. Error) (0.1802) (0-3352) Saving (-2) -0.1349 -0.2661 (Std. Error) (0.1477) (0.2686) Investment (-1) 0.5501- -0.5747 (Std. Error) (0.1502) (0.4032) Investment (-2) -0.1020 -0.0602 (Std. Error) (0.1203) (0.2703) Sum Growth Coeff. 0.0916 0.5760 0.2720 0.7115 (p-value) (0.3497) (0.0005) (0.0008) (0.0000) Sum Saving Coeff. 0.5645 -0.1688 (p-value) (0.0000) (0.5429) Sum Investment Coeff. 0.4481 -0.6349 (p-value) (0.0006) (0.0701) Granger Causality Tests [p-values] - From Growth: 3.4916 15.9774 14.3115 20.2590 (0.1745) (0.0003) (0.0008) (0.0000) - From Saving 22.4516 1.1228 (0.0000) (0.5704) - From Investment 15.4742 3.2893 (0.0004) (0.1931) R Squared 0.4446 0.2606 0.6002 0.3238 Observations 32 32 37 37 Note: Saving and Investment expressed as ratios to GDP. Growth rate is the real per capita GDP growth rate. 27 Table 2.2a: Private Investment, Public Investment, and Growth In Brazil VAR Estimation, Annual Data 1960-99 Private Investment Growth Public Investment Growth Growth (-1) 0.1345-- 0.3549- 0.0014 0.3357- (Std. Error) (0.0684) (0.1220) (0.0334) (0.1356) Growth (-2) 0.1209- 0.3579*- 0.0959-- 0.1583 (Std. Error) (0.0844) (0.1644) (0.0326) (0.1770) Private Investment (-1) 0.2637- -0.8249- (Std. Error) (0.1792) (0.4045) Private Investment (-2) 0.0229 0.1791 (Std. Error) (0.1180) (0.4329) Public Investment (-1) 0.5316- 1.1281- (Std. Error) (0.2549) (0.5619) Public Investment (-2) 0.3570-- -1.4295- (Std. Error) (0.1330) (0.8090) Sum Growth Coeff. 0.2553 0.7128 0.0973 0.4940 (p-value) (0.0055) (0.0001) (0.0084) (0.0084) Sum Private Inv Coeff. 0.2866 -0.6458 (p-value) (0.1107) (0.1106) Sum Public Inv Coeff. 0.8886 -0.3015 (p-value) (0.0000) (0.5011) Granger Causality Tests [p-values] - From Growth: 8.2287 17.8452 10.1705 9.1141 (0.0163) (0.0001) (0.0062) (0.0105) - From Private Investment 2.6707 4.8057 (0.2631) (0.0905) - From Public Investment 100.4527 4.0463 (0.0000) (0.1322) R Squared 0.3875 0.3185 0.6449 0.3253 Observations 28 28 28 28 Note: Saving and Investment expressed as ratios to GOP. Growth rate is the real per capita GOP growth rate. 28 4. Growth Accounting in Brazil 2.15 The last exercise on stylized facts is a Solow-style decomposition of per capita output growth into the contributions of capital and productivity growth. We use two methods to derive measures of total factor productivity growth per capita (or the Solow residual). Both are residuals from aggregate production functions. The first measure (TFP1) builds on the neoclassical production function with physical capital K, labor L, and the level of total factor productivity A. Assuming a Cobb-Douglas production function, we have, Y = AKaL-a 2.16 To solve for the growth rate of productivity, we take logs and time derivatives. Assuming a capital share am0.3 and solving for the growth rate of productivity, we have, TFP1 = GdpGrowth - 0.3 * CapGrowth - 0.7 * LaborForceGrowth The first measure of total factor productivity growth ignores human capital accumulation. The next measure includes human capital, H, in the aggregate production function. We use the average years of schooling in the adult population (Barro and Lee 2000) as proxy for the human capital stock in the economy. We follow Mankiw (1995) by augmenting the neoclassical production function as follows: Y = AKaHyL!-a-r 2.17 We assume that o0.3 and y-0.5 (as in Mankiw 1995), since about two thirds of labor income can be considered return to human capital. We then solve for the second measure of growth in TFP (TFP2), TFP2 = GdpGrowth - 0.3 * CapGrowth - 0.5 * SchoolGrowth - 0.2 * LaborForceGrowth where SchoolGrowth is the growth rate in average years of schooling. 2.18 Table 3 presents the growth accounting results. Figure 4 shows the growth decomposition considering only the contributions of the labor force, physical capital, and TFP1, while Figure 5 adds the contribution of human capital. Both growth accounting exercises give the same qualitative result: total factor productivity expanded strongly in the 1960s and 70s but then fell drastically in the 1980s and did not recover in the 90s. The contribution of physical capital enlarged in the 1970s and, although it fell in the next two decades, remained positive throughout the period. The contribution of human capital expanded in the 1980s to a level even bigger than that of physical capital. The positive contributions of labor and physical and human capital reflect, by construction, the positive growth that their respective stocks experienced in the last few decades (see first panel of Table 3). 2.19 The results from growth accounting corroborate the suggestion from previous exercises that capital accumulation is not the main factor driving economic growth in Brazil. However, before rejecting the "capital fundamentalism" hypothesis, we should consider that not all investment that is measured by national accounts represents an increase in productive capital (see Pritchett 2001). 29 Besides, there are also good reasons to doubt the quality of investmentXdata during periods of high and accelerating inflation. Also, the value of ao0.3 assumed for these decompositions may not be appropriate for Brazil: Brazilian national accounts data yield a ratio of a-0.5, and Castelar et. al. (2001) for example, compute these decompositions using both ac0.3 and a-0.7. But in any case, variations in TFP growth account for a significant fraction of the variation in economic growth rates since the 1970s. Table 2.3: Growth Accounting In Brazil, 1960-99 (in average annual growth rate per decade) Physical Human Labor Total Factor Productivity Output Capital Capital Force TFP1 TFP2 A. Growth Rates 60s 5.89% 4.75% 3.39% 3.06% ... 70s 8.05% 10.70% 3.64% 3.45% ... 80s 1.44% 5.96% 5.50% 3.16% ... 90s 1.98% 4.09% 4.29% 2.15% B. Contribution to Output Growth (TFP1 = Solow Residual) 60s 5.89% 1.42% ... 2.15% 2.32% 70s 8.05% 3.21% ... 2.41% 2.43% 80s 1.44% 1.79% ... 2.21% -2.57% 90s 1.98% 1.23% ... 1.51% -0.75% C. Contribution to Output Growth (TFP2 = Education-Augmented Solow Residual) 60s 5.89% 1.42% 1.69% 0.61% ... 2.15% 70s 8.05% 3.21% 1.82% 0.69% ... 2.33% 80s 1.44% 1.79% 2.75% 0.63% ... -3.74% 90s 1.98% 1.23% 2.15% 0.43% ... -1.82% 30 Figure 2.4 Growth Accounting for Brazil, 1960-99 10% 8% _ 6% - 4% - 2%- 0% - -- -2% 60s 70s 90S -4% I Growth in Total Factor Productivity * Contribution from Labor Force Growth o3 Contribution from Physical Capital Growth Figure 2.5 Growth Accounting for Brazil, 1960-99 10% 5% -5% IGrowth in Total Factor Productivity * Contribution from Labor Force Growth * Contribution from Human Capital Growth O Contribution from Physical Capital Growth 31 0 11 u R [, " [@12z Zl1 tX CA I X 6111 ON OWN N S t 2.20 In this section, we attempt to explain the economic growth performance in Brazil from a cross-country perspective. We follow the approach in Barro and Lee (1994) and Easterly, Loayza, and Montiel (1997), which consists of linking aggregate economic, political, and social variables to growth rates in per capita GDP for a large sample of countries. The estimated model is then used to project the growth rate in the country and examine whether its performance has been close to expected values. 2.21 The regression equation to be estimated is the following: Yi,. Yi,-, =Q`Yi,Y-i + fAXi, +A,U + 7i +E1, (1) where y is per capita output, X is a set variables postulated as growth determinants, ,Au is a period- specific effect, and m7i represents unobserved country-specific factors, and eis the regression residual. The subscripts i and t refer to country and time period, respectively. The sample consists of an unbalanced panel of 73 countries for the years 1960-98. In order to smooth out transitory fluctuations, we work with decade averages of all variables. Therefore, for each country we have at most 4 periods and at least 3 consecutive periods (which is the minimum required to run the instrumental variable procedure outlined below.) 2.22 The growth regression equation (1) is dynamic, in the sense that it can be rewritten as a lagged-dependent variable model. The inclusion of the initial level of per capita output (yj,.,) follows from the neoclassical growth model and captures the transitional convergence effect. The time- specific effect, ,u1 allows to control for international conditions that change over time and affect the growth performance of countries in the sample. The term ?7i accounts for unobserved country- specific factors that both drive growth and are potentially correlated with the explanatory variables. 2.23 There is a large variety of economic and social variables that can be proposed as growth determinants, X. We use the variables that are most popular in the empirical growth literature given both their quality as indicators of development in specific areas and their data availability. The list of explanatory variables is the following (see Appendix 1 for details on definitions and sources), (a) Initial level of per capita GDP -- to capture transitional convergence (b) The average years of schooling of the adult population -- to proxy for human capital in the working force (c) The ratio of domestic credit to the private sector to GDP -- to measure financial development (d) The share of primary in total exports -- to measure trade orientation and dependence on commodity markets 32 (e) The ratio of government consumption to GDP -- to measure the burden of government size and taxation to private activity (f) The black market premium on foreign exchange --to proxy for relative price distortions and govemment intervention in external markets (g) The inflation rate -- to proxy for the lack of fiscal and monetary discipline (h) A binary variable for whether the country is landlocked or not -- to indicate the difficulty in accessing trade markets (i) A qualitative variable measuring the overall quality of governance in the country, including the efficiency and honesty of the bureaucracy, the rule of law, and the peaceful resolution of conflicts 0 Terms of trade shocks -- to account for the effect of intemational conditions on the country's trade markets. 2.24 The proposed growth regression poses some challenges for estimation. The first is the presence of unobserved period- and country-specific effects. While the inclusion of decade-specific dummy variables can account for the period effects, the common methods to deal wfth country- specific effects ("within" or differences estimators) are inappropriate given the dynamic nature of the regression. The second challenge is that most explanatory variables are likely to be jointly endogenous with economic growth. That is, we need to control for the biases resulting from simultaneous or reverse causation. In the following paragraphs we outline the econometric methodology we use to control for country-specific effects and joint endogeneity in a dynamic model of panel data. Econometric Methodology 2.25 We use the Generalized-Method-of-Moments (GMM) estimators developed for dynamic models of panel data that were introduced by Holtz-Eakin, Newey, and Rosen (1988), Arellano and Bond (1991), and Arellano and Bover (1995). After accounting for the time-specific effects, we can rewrite equation (1) as follows, Y=i t layo1 + # Xj + 7, + (2) In order to eliminate the country-specific effect, we take first-differences of equation (2), Yi,,-Y i-. = d Yi - Yi,a-2) + -(Xi,-xi,. ) + (-o i., ) (3) 2.26 The use of instruments is required to deal with (1) the likely endogeneity of the explanatory varables, and, (2) the problem that, by construction, the new error term, eit - iS correlated with the lagged dependent variable, Yi t-l - Yi t-2 . This instrumental-variable method does not allow the Xvariables to be fully endogenous. Given that it relies on past values of all variables as instruments, the method only allows current and future values of the explanatory variables to be affected by the error term. Thus, under the assumptions that (a) the error term, e, 33 is not serially correlated, and (b) the explanatory variables, X, are weakly exogenous (i.e., the explanatory variables are assumed to be uncorrelated with future realizations of the error term), the GMM dynamic panel estimator uses the following moment conditions. E[yi.ts (-j -6'it-)] = 0 fors>2;t=3,...,T (4) E[Xi,,-, .(i,-i, -,)] = 0 fors>2;t=3,...,T (5) 2.27 The GMM estimator based on these conditions is known as the difference estimator. Notwithstanding its advantages with respect to simpler panel data estimators, there are important statistical shortcomings with the difference estimator. Alonso-Borrego and Arellano (1996) and Blundell and Bond (1997) show that when the explanatory variables are persistent over time, lagged levels of these variables are weak instruments for the regression equation in differences. Instrument weakness influences the asymptotic and small-sample performance of the difference estimator. Asymptotically, the variance of the coefficients rises. In small samples, Monte Carlo experiments show that the weakness of the instruments can produce biased coefficients.2 2.28 To reduce the potential biases and imprecision associated with the usual difference estimator, we use a new estimator that combines in a system the regression in differences with the regression in levels (developed in Arellano and Bover 1995 and Blundell and Bond 1997). The instruments for the regression in differences are the same as above. The instruments for the regression in levels are the lagged differences of the corresponding variables. These are appropriate instruments under the following additional assumption: although there may be correlation between the levels of the right-hand side variables and the country-specific effect in equation (2), there is no correlation between the differences of these variables and the country- specific effect. This assumption results from the following stationarity property, E[yi,t+p 7i ] = E[yi,t+q * ] (6) and E[Xi,t+ p l7i ] = E[Xi,t+q .?i ] for all p and q 2.29 The additional moment conditions for the second part of the system (the regression in levels) are:23 E(yi,,-.- y,2) (77i + ie, )j= 0 (7) EK(xi,, - X,,-2 ). (,7i + iet .)]= 0 (8) 22 An additional problem with the simple difference estimator relates to measurement error: differencing may exacerbate the bias due to errors in variables by decreasing the signal-to-noise ratio (see Griliches and Hausman, 1986). 23 Given that lagged levels are used as instruments in the differences specification, only the most recent difference is used as instrument in the levels specification. Using other lagged differences would result in redundant moment conditions. (see Arellano and Bover 1995). 34 Thus, we use the moment conditions presented in equations (4), (5), (7), and (8) and employ a GMM procedure to generate consistent and efficient parameter estimates. 2.30 Using the moment conditions presented in equations (4), (5), (7), and (8), we employ a Generalized Method of Moments (GMM) procedure to generate consistent estimates of the parameters of interest and their asymptotic variance-covariance (Arellano and Bond 1991, and Arellano and Bover 1995). These are given by the following formulas: 6 = (X'ZQ'-Z'X')- X'ZQ-'Z'y (4) AVAR(d) = (X'ZQ-'Z'X)' (5) where Ois the vector of parameters of interest (a, /3), y is the dependent variable stacked first in differences and then in levels, X is the explanatory-variable matrix including the lagged dependent variable (yt.1, )X) stacked first in differences and then in levels, Zis the matrix of instruments derived from the moment conditions, and Q is a consistent estimate of the variance-covariance matrix of the moment conditions. 24 2.31 Consistency of the GMM estimator depends on the validity of the instruments. To address this issue we consider a specification test of the Sargan type. This test of over-identifying restrictions examines the overall validity of the instruments by analyzing the sample analog of the moment conditions used in the estimation process. Results 2.32 Table 4 presents the estimation results. The Sargan test cannot reject the null hypothesis of correct specification of our model. The estimated coefficients on all explanatory variables have the expected sign and are statistically significant. First, economic growth is affected by economic characteristics of development. Thus, the growth rate rises with a lower initial level of output (relative convergence effect), better-educated labor force, deeper financial markets, and a smaller share of primary (commodity) exports. Second, economic growth is shaped by the country's type of govemment. Consequently, the growth rate rises with better governance, smaller government size, less inflation, and shorter black-market premium (less relative price distortions). Third, the country's natural endowment and current international conditions also determine economic growth. Thus, the growth rate is higher in countries that have access to the sea and increases with positive terms of trade shocks. The negative and significant coefficients on the period dummy variables indicate that the 1980s and 1990s were less propitious for growth throughout the world than the 1970s were. 24 In practice, Arellano and Bond (1991) suggest the following two-step procedure to obtain consistent and efficient GMM estimates. First, assume that the residuals, ,,, are independent and homoskedastic both across countries and over time. This assumption corresponds to a specific weighting matrix that is used to produce first-step coefficient estimates. Then, construct a consistent estimate of the variance-covariance matrix of the moment conditions with the residuals obtained in the first step, and use this matrix to re-estimate the parameters of interest (i.e. second-step estimates). Asymptotically, the second-step estimates are superior to the first-step ones in so far as efficiency is concemed. 35 Table 4: Determinants of Economic Growth Estimation Technique: Arellano and Bover (1995) GMM-IV System Estimator Dependent Variable: Growth Rate of GDP per Capita Constant 50.95035 (4.2287) Initial GDP per capita -1.43649 ** (in logs) (0.1750) Average Years of Schooling 0.791679 (in logs) (0.3614) Terms of Trade Shocks 6.728587 ** (log difference of the terms of trade) (1.7153) Domestic Credit to Private Sector 0.510894 ** (as ratio to GDP, in logs) (0.1897) Government Consumption -2.22178 ** (as ratio to GDP, in logs) (0.4797) Inflation -2.00755 (log of 1 + inflation rate) (0.6484) Black Market Premium -4.78281 (in log of 1 + bmp) (0.5575) Primary Export Orientation -1.11538 ** (as ratio to total exports, in logs) (0.1976) Landlocked -0.9456 ** (dummy: 1 = landlocked) (0.3069) Governance Index 4.438432 ** (2.0726) Dummy 80s vs. 70s -1.22523 ** (0.1608) Dummy 90s vs. 70s -1.49846 ** (0.2417) Sargan Test (p-value) 0.165 (Ho: Instruments are valid) Number of Countries/Observations 65/116 Numbers ih parentheses represent standard errors. * (**) Significant at the 10 (5) percent level. In column [1], inflation is proxied by log(1+inf), whereas in column [2], it is proxied by inf/(1+inf). 36 2.33 Our regression model can be used to explain the changes over time in economic growth for any country in the sample. We cannot, however, explain the levels of growth given that we do not estimate the unobserved country-specific effects (although we control for them).25 We are interested in assessing the extent to which our model can account for the different growth performance in the 1980s and 1990s with respect to the 1970s. We present the results for Brazil and a few other Latin American countries in Table 5. The accuracy of the projection depends on the country (although in all cases the projected and actual growth differences have the same sign). For Chile, Colombia, Mexico, and Peru, the projected growth difference approximates closely its actual value. For Argentina and Venezuela, at least one period is projected accurately. It is for Brazil and Ecuador that the model performs poorly. In both cases, the actual drop in growth rates in the 1980s and 1990s with respect to the 1970s was larger than what the model predicted. This is particularly the case for the comparisons related to the 1990s --the growth residuals are found in the 10% tail of the distribution. 2.34 In Brazil, the drop in the growth rate from the 1970s to, respectively, the 1980s and 1990s was 5.55 and 5.68 (there is little growth difference between the latter two decades). We can explain 70% and 35% of these amounts, respectively. In table 6, we assess the contribution of each explanatory variable to the projected difference in the growth rate. Regarding the growth deceleration from the 1970s to the 80s, less favorable international conditions and the rise in inflation and price distortions played the biggest role. Decreasing returns and the forces behind conditional convergence also contributed to the growth slowdown. As mentioned before, explaining the poor growth performance of the 1990s vis-a-vis the 1970s is more difficult. Conditional convergence, a reduction of world growth, and the presence of inflation (at the beginning of the decade) contributed to the fall in Brazilian growth rates. However, the reduction of price distortions, favorable terms of trade, a better-educated labor force, and a higher trade orientation toward manufactured products predicted an increase in the growth rate of the 1990s with respect to the 70s. 2.35 Explaining why the growth rate in the 1970s and even 1960s was so significantly larger than in more recent decades remains an area for future research. In this investigation, it will be important to assess the extent to which the growth rate in the 1960s and 1970s was due to temporary factors. Looking to replicate the policies of those years may be a mistake if the growth experienced then was exceptional or maybe even the source for the ensuing slowdown. 25 For the same reason, cross-country level comparisons are not feasible with our model. 37 Table 2.5: Comparison of Actual and Projected Growth Changes for Selected Latin American Countries Actual Projected Residual Argentina 80s vs. 70s -3.8761 -2.3769 -1.4992 90s vs. 70s 1.4214 1.6462 -0.2247 Brazil 80s vs. 70s -5.55 -3.86 -1.68 90s vs. 70s -5.68 -2.08 -3.60* Chile 80s vs. 70s 1.27 1.37 -0.10 90s vs. 70s 3.82 2.68 1.14 Colombia 80s vs. 70s -2.10 -2.17 0.07 90s vs. 70s -2.33 -2.26 -0.08 Ecuador 80s vs. 70s -6.04 -3.23 -2.82 90s vs. 70s -5.28 -1.21 -4.07** Mexico 80s vs. 70s -3.89 -4.03 0.14 90s vs. 70s -1.73 -1.28 -0.45 Peru 80s vs. 70s -2.91 -2.95 0.04 90s vs. 70s 1.43 0.02 1.41 Venezuela 80s vs. 70s -3.77 -2.96 -0.81 90s vs. 70s -0.18 -3.08 2.90 Notes: The standard deviation for the residuals are 1.9221 for the 80/70 period and 1.919 for the 90/70 period. ("J indicates that the residual is different from zero at the 10 (5) percent significance level. 38 Table 2.6: Sources of Growth Sources 80s /70s 90s 70s Actual Growth -5.55 -5.68 Projected Growth -3.86 -2.08 Initial Income per capita -0.82 -0.73 Avg. Years Schooling 0.05 0.20 Terms of Trade Shocks -0.04 0.27 Domestic Credit to Private Sector -0.10 0.00 Govemment Size -0.10 -0.07 Inflation Rate -1.24 -1.15 Black Market Premium -0.75 0.31 Primary Exports 0.35 0.58 Time Dummies -1.23 -1.50 39 Brazil Compared to Latin America, East Asia, and the OECD 2.36 The empirical exercise presented in the previous section can help explain the growth performance in Brazil by drawing from the cross-country evidence. This approach requires long series of data for a large sample of countries. Thus, data availability constrains the quantity and type of variables that can be used as growth determinants in econometric exercises. In this section, we consider a wider variety of economic and social indicators given that our interest is not to conduct econometric estimation but to evaluate qualitatively the economic and social areas where Brazil needs most improvement. In the selection of these economic and social indicators, we are implicitly assuming that they contribute to economic growth. For each indicator we compare Brazil with the typical country in Latin America and the Caribbean (LAC), the seven rapidly growing East Asian countries (EAP7), and the industrialized countries (OECD). 2.37 We have selected indicators in the following areas: human capital, government-induced distortions, financial development, governance, and public infrastructure. Table 7 presents the indicators and their respective values for Brazil, LAC, EAP7, and the OECD, corresponding to the last years of the 1990s. Figures 6-10 present these values graphically and, in addition, show the evolution of the corresponding gaps between Brazil and the three regions in the 1990s. Examining the evolution of these gaps will help assess whether Brazil is making progress in the factors that drive economic growth. The main results are presented in the following paragraphs. 2.38 Human capital. Brazil lags behind LAC, the EAP7, and the OECD in both years of schooling and life expectancy. The gaps with respect to LAC and OECD are not closing, and the gap with respect to the EAP7 has actually expanded in the 1990s (see Figure 6). 2.39 Govemment-induced distortions. By 1998, Brazil already had made much progress in controlling inflation and reducing the black market premium (see Figure 7). In fact, following the trend in Latin America, inflation was sharply reduced and the price distortions on foreign exchange were cut down considerably. Thus, the gaps with respect to the EAP7 and OECD basically disappeared by the end of the 1990s. These improvements in monetary management have been strengthened by the implementation of inflation targeting and flexible exchange rates in 1999. 2.40 Financial Development. Although Brazil is similar to the typical LAC country regarding the depth and private-sector orientation of the financial system, it lags considerably behind the EAP7 and OECD. Although the gap with respect to the OECD closed somewhat, the difference with the EAP7 rose notably (see Figure 8). 2.41 Govemance. As with financial development, Brazil's quality of governance appears to be close to the typical LAC country but still far from the EAP7 and OECD. It is particularly worrisome that the Brazilian gap with respect to these regions has increased in the 1990s (see Figure 9). In the areas of government corruption and law and order, the gap expansion has been due to the deterioration of these indicators in Brazil. 40 Table 2.7: Challenges for Growth Brazil LAC OECD EAP7 L Human Capital Average Years of Schooling, 1999 - Primary 3.36 3.97 5.50 4.90 - Secondary 0.92 1.49 3.29 2.86 - Tertiary 0.28 0.36 0.64 0.45 - Total 4.56 5.74 9.54 8.12 Ufe Expectancy 67.0 70.7 77.5 72.3 It. Government-induced Distortions Black Market Premium 0.0114 0.0184 0.0000 0.0095 Inflation 0.0769 0.0721 0.0148 0.0354 111. Finance Liquid Liabilities 0.2864 0.3337 0.7238 0.8958 Private Sector Credit 0.3102 0.2777 0.8107 1.3577 IV. Governance [0-11 Corruption 0.5000 0.5000 0.8333 0.6667 Bureaucratic Quality 0.6667 0.5000 1.0000 0.8333 Law and Order 0.5000 0.5000 1.0000 0.8333 V. Infrastructure Main Telephone Lines (thousands) 17445 559 5721 4863 (lines per 100 workers) 23 23 110 49 Energy (in MW) 62972 1490 23347 19723 (in MW per 100 people) 0.08 0.08 0.39 0.14 Paved Roads (in km.) 173166 6385 135531 52405 (in km. per 100 people) 0.23 0.18 2.28 0.19 VI. Capital Stock per worker (in 1990 PPP US$ thousands) 32.4 20.2 101.4 53.2 Note: We present the median for all the variables 41 2.42 Infrastructure. In the areas of telecommunications, energy generating capacity, and transportation, the difference between Brazil and the EAP7 and, particularly, the OECD are quite large. The gap with respect to the industrialized countries has not closed in the 1990s, and that with respect to the EAP7 has actually expanded (see Figure 10). Brazil's progress in these areas has been just as slow as in the typical Latin American country. 2.43 Capital accumulation has not been the main driving force for growth in Brazil. The large growth differences between the 1970s and the following decades have not been matched by comparable changes in capital accumulation. Furthermore, we present evidence that investment follows rather than precedes output growth. 2.44 Less favorable international conditions, domestic macroeconomic instability-mostly brought about by inflation and price distortions-and diminishing retums explain about haff of the growth deceleration in Brazil between the last two decades and the 1970s. 2.45 The reasons behind the extraordinary growth performance in Brazil in the 1970s are not fully understood in this study and merit further research. 2.46 Regarding the progress Brazil made in the 1990s on the factors that drive growth, the most important one was in the area of macroeconomic stability. The sharp reduction of inflation and the elimination of price distortions, particularly in extemal markets, have set the ground for future growth. The implementation of a monetary regime based on inflation targeting and flexible exchange rates in Brazil is the keystone for macroeconomic stability. 2.47 Regarding other areas of structural reform, Brazil's is still largely underdeveloped. In the areas of human capital, financial depth, govemance, and public infrastructure, Brazil not only lags behind the OECD and East Asian countries, but its gap with respect to them has mostly expanded in the 1990s. 42 Figure 2.6 Human Capital Education Average Years of Schooling, Total 10 1 120 CY)~~~~~~~~~~~~~~~~~1 -8 -~~~~~~~~~~~~~~~~ i ot 7 tt ~~~~~~~100 -[3-1 R1-13-3-E3-1 8 (I6 )_ xi80x *x* 0 g 5 ~ ~ ~ ~ ~ ~ ~ ~ ~ 5 40 4 3 _- = 60 ^ K Ki 12 - t 500 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 _Brazil A- AMER- 13 -IND - - EAP7 Life Expectancy Life Expectancy at Birth (years) X 80 -" 776 * 103 - 78 101 76[a ( 74 - ~~~~~~~~~~97- 72- 95 - .~70 - 93 - ~.68 9 co 89 - x 64~~~~~~~~~~ 87 -_________ Wl 62 85 -i 60 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 -4--Brazil A AMER - i FIND - 0- EAP7 43 Figure 2.7 Government Induced Distortions Black Market Premium Black Market Premium 0.080 135 0, 0.070 - 130 0.060 125 125 E 0 050 - B~~~~ ~~~~~~ 120 \. 0.050 115 (0 2 0.040 110 0.030 - 105 2 0.020 95o 0.010 _ _ 90 . 0.000 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 _Brazil -*-AMER - 0 IND - EAP7 Inflation Rate Intlation Rate 0.14 3500 0.12 3000 "500- * 0.10 ci250 °00C - t2000 \ \ _ 0.06 8 * 0.04 - 500- 0.02 0.0, -; 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Brazil LAC OECD EAP7 0.04 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ _rzl M R 3WID-*EP 44 Figure 2.8 Financial Development Finance: Uquid Liabilities Liquid Liabilities as a ratio to GDP 10 1.6 18 1.4 - 160 - 1.2- 1, 4 .120 -0 0.8 -10~~~~~~~~~~~~~~~~~~~~~s y06 _ _0 - _ = _ __ - x 40 :3 0.2 _ 20 - cT 1990 1991 1992 1993 1994 1995 1996 1997 1998 Brazil LAC OECD EAP7 _Brazil )K AMER - la IND - *- EAP7 Finance: Domestic Credit to the Private Sector Domestic Credit as a ratio to GDP 0* 1.0 180- 0.9 ,160 r: 0.8 '140 C0.7 -! 120- 0.6 - -i- 0 + 0.5 .2- 80 i 0.4. 60- ~03 X40- 002 - .2 20 0.1 _ __O IL 0.00 1990 1991 1992 1993 1994 1995 1996 1997 1998 Brazil LAC OECD EAP7 _ Brazil --AMER - 13 IND - 4- EAP7 45 Figure 2.9 Governance Indicators (high scores mean better performance) Corruption 0.9 Corruption 0.8 - 120 0.7 - 8o110 0.6 - ~ ~ ~ ~ ~ ~ ~ I 0.6- ,jiOO Ei - 8l ' 83-E 3-9-E 0.5 90 0.4 - 80 _ 0.3 70 0.2 - 60 0t~~~~~~~~~~~~~~~~~~~~~~~6 - _ _1 _9 _-_ _ 0.1 50 0.0 I , 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 Brazil -)--AMER- 3 IND - *- EAP7 Bureaucratic Quality 1.0 Bureaucratic Quality 110 0.9 _ 0.8 - 00 -9--- El--E--E B 8 - E3 0.7 - 90 0.6 *80. 0.5- _- 'a 0.4 ~~~ ~~~~~~~~~~~~70 - 0.4-_1 - -^0- 0.3 =E60 - 0.1 40 0.0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 _Brazil -EAMER - 3 IND - *- EAP7 Law and Order 1.0 Law and Order 0.9 0.8 -§00- - 8 - S 8 8 E3 8 a El 0.7 90 0.6 80 0.5 X 70 0.4 - 0° 60 - 0.3 - co 50 - _ 0.2 - 40 / 0.1- 11|11 1|1 , 30 0.0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Brazil LAC OECD EAP7 _ Brazil -IN AMER - 3 IND - *- EAP7 46 Figure 2.10 Infrastructure Telecommunications, 1996498 Main Telephone Lines per 1000 workers 140- 120 . - - 0 80 'o 80 o 0 60 C60 4*0 eS 40 -40 220 20 _ _ _ O . . . . . . .0 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 Brazil LAC OECD - Brazil W AMER - a tIND - 4- EAP7 Energy Generating Capacity, 1996-98 Energy Generating Capacity per 1000 workers 0.50 120 0.45 o ll ep oo44o - ~~~~~~~~~~goo f * - 0.40 .I 0~~~~~~~~~~~~~~~~8 a0.30 0.25 _ 0.20 . _ -40 3 0.15 * 0.10 2 0.05 0- 0.00 1990 1991 1992 1993 1994 1995 1996 1997 Brazil LAC OECD . . _Brazil -*AMER - 1G IND - 4- EAP7 Paved Roads, 1996-98 Paved Roads (in km. per 1000 workers) 120 3.0 2.5 .* !1 a . - X80 o 2.00 !60 I 1.5 840 | 0.5 t 20 . . ,,, , . - * o ^ ° o 0.5 0 0.0 1990 1991 1992 1993 1994 1995 1996 1997 Brazil LAC OECD - Brazil - -AMER - G IND - 4- EAP7 47 References Arellano, Manuel and Bond, Stephen. "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations," Review of Economic Studies 1991, 58(2), pp. 277-297. Arellano, Manuel and Bover, Olympia. "Another Look at the Instrumental-Variable Estimation of Error-Components Models," Journal of Econometrics 1995, 68(1), pp. 29-52. Attanasio, Orazio P., Picci, Lucio, and Scorcu, Antonello E. "Saving, Growth, and Investment: A Macroeconomic Analysis using a Panel of Countries," Review of Economics and Statistics, 2000, 82(2), pp. 182-211. Barro, Robert J. and Lee, Jong-Wha. "Sources of Economic Growth," Camegie-Rochester Series on Public Policy, 40:1-57, June 1994. Barro, Robert J., and Lee, Jong-Wha. "Intemational Data on Educational Attainment Updates and Implications," National Bureau of Economic Research Working Paper Series No. 7911: 1 - 36, September 2000. Beck, Thorsten, Demirguc-Kunt, Asli, and Levine, Ross. "A New Database on Financial Development and Structure," The World Bank Economic Review, 14(3), pp. 597-605, September 2000. Blomstrom, Magnus, Lipsey, Robert E. and Zejan, Mario. "Is Fixed Investment the Key to Economic Growth?" Quarterly Journal of Economics 111:269-276, February 1996. Canning, David. "A Database of World Stocks of Infrastructure," The World Bank Economic Review, 12:529-547, September 1998. Castelar, Armando, Indermit Gill, Luis Serv6n, and Mark Thomas. uBrazilian Economic Growth, 1900-2000: Lessons and Policy Implications", paper prepared for the GDN- LACEA Conference on Economic Growth, Rio de Janeiro, December 2001. Easterly, William, Loayza, Norman and Montiel, Peter. "Has Latin America's Post Reform Growth Been Disappointing?" Joumal of Intemational Economics 1997, 43, pp. 287-311. Gallup, John L. and Sachs, Jeffrey D. "Geography and Economic Development," National Bureau of Economic Research Working Paper Series No. 6849: 1-81, December 1998. Holtz-Eakin, Douglas, Newey, Whitney and. Rosen, Harvey S. "Estimating Vector Autoregressions with Panel Data," Econometrica, 56:1371-1395, November 1988. Intemational Currency Analysis, Inc. World Currency Yearbook, Brooklyn, NY: Intemational Currency Analysis, Inc. Various Issues. International Road Federation. World Road Statistics, Geneva; Washington, DC: International Road Federation. Various Issues. Intemational Telecommunication Union. World Telecommunication Development Report, Geneva, Switzerland: International Telecommunication Union. Various Issues. Kaufmann, Daniel, Kraay, Aart, and Zoido-Lobaton, Pablo. "Aggregating govemance indicators," The World Bank Policy Research Working Paper 2195, October 1999. 48 Kaufmann, Daniel, Kraay, Aart, and Zoido-Lobaton, Pablo. "Governance Matters," World Bank Policy Research Working Paper 2196, October 1999. Mankiw, N. Gregory. "Growth of Nations," Brookings Papers on Economic Activity, 1: 275-329, 1995. Marshall, Monty G., and Jaggers, Keith. "Polity IV Project - Political Regime Characteristics and Transitions, 1800-1999: Dataset Users Manual," University of Maryland, College Park, Mimeo. Wood, Adran. "Global Trends in Real Exchange Rates, 1960 to 1984," World Bank Discussion Paper No. 35,1988. World Bank. World Development Indicators, Washington, DC: The World Bank, 1997. 49 Appendix 2.1: Variables and Sources Variable Source Output Real per capita GDP (in 1990 PPP US$) Summers and Heston (1991), The World Bank (2000) Share of Agriculture In Total Value Added (%) The World Bank (2000) 'World Development Indicators'. Share of Industry In Total Value Added (%) The Worid Bank (2000) 'World Development Indicators'. Share of Services in Total Value Added (%) The World Bank (2000) 'World Development Indicators". Physical and Human Capital Domestic Capital Stock Summers and Heston (1991), The World Bank (2000) Investment as a ratio to GDP The World Bank (2000) World Development Indicators". Labor Force, Total The World Bank (2000) "World Development Indicators". Average Years of Schooling Barro and Lee (2000) Life Expectancy at birth (years) The World Bank (2000) "World Development Indicators". Extemal Sector Primary Exports (% Total Exports) The World Bank (2000) "World Development Indicators". Terms of Trade Shocks The World Bank (2000) "World Development Indicators". Finance Liquid Liabilfties Beck, Demirguc-Kunt and Levine (2000) Domestic Credit to the Private Sector Beck, Demirguc-Kunt and Levine (2000) Government-Induced Distortions Inflation Rate Intemational Monetary Fund - Intemational Financial Statistics Black Market Premium (BMP) Wood (1988), Intemational Currency Analysis (various issues) Democracy and Govemance Corruption Intemational Country Risk Guide (ICRG) Bureaucratic Quality. Intemational Country Risk Guide (ICRG) Law and Order Intemational Country Risk Guide (ICRG) Infrastructure Main Telephone Lines Canning (1998), Intemational Telecommunications Union Energy Generating Capacity (in MW) Canning (1998), United Nations Paved Roads (in km.) Canning (1998), IntematIonal Road Federation Govemment Govemment Consumption (% GDP) Flxed Factors Dummy for Landlocked Countries Gallup and Sachs (1998) 50 3. PRODUCTIVITY GROWTH IN BRAZILIAN INDUSTRY Prepared by Marc Muendler, Luis Serven, and Claudia Sepulveda 3.1 Gains in total factor productivity - that is, the ability to produce more outputs from the same inputs - have long been recognized as a major driving force of improvements in income and welfare. Indeed, it has been argued that productivity growth is the main factor behind the observed differences in long-term per capita income growth across countries.26 3.2 In many developing countries the industrial sector has traditionally been protected from foreign competition and, as a result, has performed poorly in terms of productivity growth. Brazil has been no exception in this regard, and until the late 1980s its industrial sector was largely sheltered from foreign competition in a government-led import-substitution industrialization model. During the last decade, however, Brazil implemented a series of reforms aimed at enhancing the role of market forces and building an export-oriented economy. The reforms started in 1988 with trade liberalization, the Mercosur trade agreement in 1991, and the privatization program initiated in the early 1990s and accelerated in 1996. The drive towards reform, however, was hampered by persistent high inflation and macroeconomic instability during the late 1980s and much of the 1990s. 3.3 During this period, Brazil's industrial sector lost ground in terms of value added and employment relative to aggregate economic activity. Industrial GDP declined from 45 percent of total GDP in the early 1980s to 38 percent in 1990 and under 30 percent in 1999 (Figure 3.1). Similarly, industrial employment decreased by 26 percent in the period 1990-99 (Figure 3.2). Nevertheless, the economic reforms of the 1990s generated renewed interest in the evolution of productivity in industry, and existing studies tend to find a marked productivity improvement during the 1990s.27 26 see for example Klenow and Rodriguez-Clare (1997) and Easterly and Levine (2000). 27 Recent studies focusing on the industrial sector include Hay (1997), Bonelli and Fonseca (1998), Rossi and Ferreira (1999) and Gomes (2001). 51 Figure 3.1 Composition of GDP, 1960-99 100% 80%- ( 60% _ 40% .C 20% 0% .. to co CO D co r- r-- co X 0C 0) O O0 0e 0n 0) a) as 0) 0, 0) 0) ) 0 0) 0 * Agriculture * Industry O Services Figure 3.2 Employment in Industry (million) 12 - 10 8 6 - 4O- 2- 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 - Industry- -Sample 3.4 This paper takes a fresh look at the evolution and determinants of total factor productivity (henceforth TFP) growth in Brazil's industrial sector during the 1980s and 1990s. The objective is to identify the factors that contributed to shape the observed TFP performance during that period. The analysis focuses on four main issues: the impact of trade liberalization, the contribution of knowledge embodied in input quality, the role of resource reallocation across firms and sectors, and the factors and constraints shaping firms' investment decisions. 3.5 Unlike most previous empirical studies of TFP in Brazil, which make use of economy- or industry-wide aggregates, this paper breaks new ground by performing the first systematic exploration of a large firm-level data set available in Brazil, the Pesquisa Industrial Annual. On a priori grounds, microeconomic data offer two important advantages over aggregate data from the perspective of studying TFP. First, aggregate TFP patterns may conceal widely diverging trends among different subsectors and/or types of firms. For example, TFP performance may vary systematically with firm-specific characteristics (e.g., size, composition of input mix etc), and these 52 relations would be masked in aggregate data. Second, measures of TFP growth based on aggregate data mix together firm-level TFP growth with the effects on TFP of reallocation of inputs and outputs across microeconomic units with different productivity levels (through channels such as changing market shares of incumbent firms or entry and exit of firms).28 With disaggregated data it is possible to disentangle the contribution to overall TFP performance of these two conceptually distinct forces. 3.6 In the case of Brazil, however, the microeconomic data also pose major challenges concerning the coverage, consistency and reliability of the information. PIA presents methodological breaks that hamper comparability of the data across different time periods, lacks information on the utilization of productive inputs -- even on capital stocks after 1995 -- and much of its time coverage corresponds to years of extreme inflation, which complicates considerably the construction of key variables for the analysis. While we have made a major effort to correct some of these shortcomings, and we believe that we have been largely successful in limiting the extent of measurement error embedded in our TFP estimates, the empirical results reported below should be taken with a considerable dose of caution. 3.7 The paper draws extensively from Muendler (2001 a, b). It is organized as follows. In Section 2 we provide an overview of a selected group of earlier studies of productivity growth in Brazil. Section 3 lays out the methodological approach to the estimation of total factor productivity with firm-level data, and summarzes the main features, as well as problems, of our data set. Section 4 describes the pattems of the resulting TFP estimates over time, across industrial sectors and geographical regions. Section 5 examines the factors behind the observed TFP performance to shed light on the issues under investigation.. Section 6 presents some concluding remarks. 28 This reallocation effect may be particularly important for the evolution of aggregate TFP over short periods; see Jorgenson (1 990). 53 Table 3.1 Some Recent Studies of TFP Growth In Brazil Total Factor Labor Productivity Authors Aggregation Level Period Productivity Growth Growth (annual) (annual) Bacha and GDP 1940-00 0.32 Bonelli (2001) 1970-80 0.09 1980-91 -2.28 1991-00 1.73 Industry (excluding 1950-99 3.38 construction) 1980-91 0.47 1991-99 _ 4.80 Bonelli and GDP 1980-89 0.10 Fonseca (1998) 1990-97 0.20 Transformation 1989-89 -0.73 Industries 1990-97 3.38 General Industry 1992-97 9.59 Textiles 6.20 Non-metallic mineral 11.46 products De Gregorio and GDP 1960-90 0.80 Lee (1999) 1980-89 -1.40 Elias (1990) GDP 1940-90 0.80 Gomes (2001) GDP 1975-98 0.21-0.3 1975-89 0.56 0.14-0.12 Industry 1990-98 3.2-4.07 0.39-0.46 Hofman (2000) GDP 1950-89 2.60 3.90 1990-98 0.10 0.40 1992-98 0.70 1.00 Industry 1976-89 0.37 1986-98 0.78-1.37 1976-98 3.76 1986-98 3.5-4.7 54 3.8 The evolution of productivity in Brazil has been the focus of a number of empirical studies. Some are concerned with productivity at the aggregate (GDP) level, while others focus on the industrial sector as a whole, or on specific subsectors. Among the former group, several studies offer cross-country results and therefore allow a comparative perspective on Brazil's productivity performance. 3.9 Table 3.1 summarizes the results of some recent empirical analyses. Among aggregate- level studies, De Gregorio and Lee (1999) estimated a TFP growth rate for the period 1960-90 of 0.80 percent (below the average in Latin America). When they considered the decade of the 1980s only, however, they found that the Brazilian economy experienced a negative TFP growth of -1.40 percent. In tum, Bonelli and Fonseca (1998) found average TFP growth of 0.1 percent in the period 1980-89, and 0.2 percent over 1990-97. Using data for a much longer period, Hofman (2000) estimates a TFP growth rate of 2.60 percent for Brazil in 1950-89. This contrasts with the earlier results in Elias (1990), who finds a much lower TFP growth rate (0.8) over a similarly long period, 1940-90. Over the 1990s, however, Hofman finds that TFP growth slowed down to 0.10 percent (similar to the results of Bonelli and Fonseca), although with a considerable acceleration (to 0.70 percent) after 1992. 3.10 Finally, two of the most recent studies of aggregate TFP, by Teixeira da Silva (2001) and Gomes (2001), find a similar pattem of fall and recovery in spite of using different methodologies and data. According to the results of Teixeira da Silva, TFP fell at an annual rate of -0.70 percent over 1980-92, and then rose at an average rate of 0.9 percent over 1993-2000. Moreover, aggregate labor productivity is found to display a very similar pattem. In turn, Gomes finds a similar tumaround in TFP growth, from 0.6-0.7 percent in 1985-89 to 4.7-5.7 percent in 1991-98, with the exact figure depending on the specific method used. 3.11 In tum, studies of TFP in Brazil's industry vary considerably in terms of coverage. Among the earliest studies, Braga and Rossi (1988) analyzed factor productivity for 21 sectors in industry for the period 1970-83 using a translog production function. They found that 10 of the 21 sectors had experienced negative TFP growth rate in this period." Bonelli (1992) used a growth accounting method to estimate TFP growth in manufacturing between 1975-85 using data from the industrial census of 1970,1975 and 1980. He found average TFP growth in manufacturing of 0.80 percent yearly. In his study, only leather, pharmaceuticals and furniture presented negative TFP growth rates. 3.12 The economic reforms of the 1990s renewed interest in estimates of TFP in manufacturing. Bonelli and Fonseca (1998) estimate TFP growth for overall manufacturing and for 21 individual sectors. At the aggregate manufacturing level, they calculate a negative TFP growth rate of -0.73 percent in the 1980s. For the 1990s, in contrast, they find positive TFP growth of 3.4 percent per annum. Likewise, Rossi and Ferreira (1999) calculate TFP and labor productivity for 16 industrial sectors in the period 1985-97. They find that TFP and labor productivity growth accelerated for 29 Another early study is that of Pinheiro (1989), which is based on firm-level data from the 1970 and 1980 census. 55 virtually all sectors after 1991. On average, TFP grew at 2.2 percent in 1991-97, while in 1985-90 it had declined at an annual rate of -2.5. Finally, the already-cited study by Gomes (2001) also reports estimates of aggregate industrial TFP that display a similar acceleration in the recent period: according to this study, annual TFP growth rose from 0.4 percent in 1976-89 to 3.4-4.4 percent in 1991-98. 3.13 How does this performance compare with that of other countries? The development in recent years of several large international data sets has made it possible to compare the growth experience of a substantial number of countries by employing a transparent methodology. A review of these studies is presented in Table 3.2. 3.14 Elias (1990) presents TFP growth for a group of seven Latin American countries for the period 1940-90. Among them Brazil shows TFP growth below the Latin America average, and below Chile's and Mexico's TFP growth. De Gregorio and Lee (1999) examine TFP growth for a larger sample of Latin American countries (21) for the years 1960-90, a shorter time span than Elias (1990). They found for Brazil a TFP growth higher than the average TFP growth for Latin America. 3.15 Across the time dimension, Latin America TFP grew in the 1960s at 1.9 percent per annum, in the 1970s at 0.7 percent and in the 1980s at -2.0 percent per annum. On the other hand, Brazil's TFP growth in the 1960s, 1970s and 1980s was1.5, 2.5 and -1.4 percent respectively. To sum up, Brazil's TFP growth (decline) was above (below) Latin America's TFP growth (decline) in the three decades covered by De Gregorio and Lee study, with the exception of the 1960-69 period. Indeed in the 1970s Brazil's TFP growth was almost three times the average of Latin America. 3.16 Collins and Bosworth (1996) present estimates of TFP growth for the East Asian countries between 1960-94. All East Asian countries, with the exception of the Philippines, experienced TFP growth above 1.0 percent per annum. In the same period TFP in the United States grew at 0.3 percent per year, and in the other industrialized countries at 1.1 percent per year. Moreover, TFP growth in the East Asian countries did not follow the pattern across time observed in Latin America. In all three decades from the 1960s to the 1990s, TFP growth is positive, with the exception of Philippines. 3.17 Nehru and Dhareshwar (1994), using a sample of 83 industrial and developing countries for the period 1960-87, obtain an average TFP growth for Brazil, Chile and Mexico of 1.39, 0.37 and 0.68 percent per annum respectively. These figures are higher than the ones observed in the studies of Elias (1990) and De Gregorio and Lee (1999) with a smaller sample of countries. Moreover, Brazil's TFP growth is higher than the average TFP growth for the OECD and Latin America countries in the same period. 3.18 It is interesting to compare the performance of industrial TFP in Brazil with other countries. Jorgensen and Stiroh (2000) report TFP growth for 37 industries in the United States using a growth accounting methodology for the period 1958-96. Industry-wide TFP grew 0.48 percent per year over this period. However, industry TFP growth ranged from 1.97 percent in Electronic and Electric Equipment to -0.44 percent in Printing and Publishing. Many industries (9) experienced negative productivity growth for a period of nearly 40 years. 56 3.19 Table 3.2 Intemational Studies of TFP Growth Total Factor Aggregation Productivity Authors Level Period Growth (annual) Collins and GDP Bosworth Latin America 1960-94 0.2 (1996) East Asia 1.1 Taiwan 2.0 Korea 1.5 United States 0.3 De Gregorio GDP and Lee Latin America 1960-90 0.1 (1999) Argentina -0.5 Brazil 0.8 Chile 0.9 Mexico 0.5 Latin America 1960-69 1.9 Argentina 0.7 Brazil 1.5 Chile 1.6 Mexico 2.3 Latin America 1970-79 0.7 Argentina 0.6 Brazil 2..5 Chile 0.5 Mexico 1.2 Lafin America 1980-89 -2.0 Argentina -2.6 Brazil -.1.4 Chile 0.6 Mexico -.1.8 Elfas (1990) GDP Latin America 1940-900 1.2 Argentina 0.5 Chile 1.4 Colombia 0.8 Mexico 1.1 Nehru and GDP Dhareshwar Latin America 1960-87 0.13 (1994) Brazil 1.39 Chile 0.37 Mexico 0.68 I OECD 0.76 57 3.20 Liang and Jorgenson (1999) present industry TFP growth for Taiwan, one of the Asian newly industrialized economies, in the period 1961-93. Liang and Jorgenson measure TFP growth using a translog production function. TFP growth rate of the manufacturing sector increased from 0.2 percent per annum during 1961-82 to 0.55 percent during 1982-93, and 0.32 percent for the whole period. In the period 1961-82, the highest TFP growth was in Electrical Machinery and Electronics with a 5.44 percent per annum followed by textiles and food. The laggards were wood and fumiture products with -12.35 percent per annum and Paper and Printing with -8.66 percent. 3.21 In Latin America, the World Bank (1999) presents TFP growth for Manufacturing in Mexico during the period 1993-97. Between 1993 and 1995 TFP growth accelerated from 0.6 percent per annum to 13.8 percent. Subsequently TFP growth rates declined 1.3 percent in 1995-96 to -3.9 percent in 1996-97. TFP growth estimated at the two-digit industry show two groups. The first group - Food Textiles and Apparel, Wood Product and Fumiture and Clay and Cement Products has lower TFP growth than the industry average. The second group, Paper and Printing, Chemicals, Metals, Machinery and Other Industries has higher rates of TFP growth. 3.22 On the whole, the international evidence suggests that (i) productivity growth in Brazil has lagged behind that of East Asian economies in recent decades, but was roughly on par with that in the rest of Latin America; (ii) Brazil, like other developing economies, experienced a TFP acceleration in the 1990s, after its disappointing performance during the 1980s; and (iii) industrial TFP growth in Brazil, as well as in other developed and developing countries, varies widely across sectors, and negative TFP growth rates at the sector level are often encountered in existing studies. 3.23 Several of the Brazil studies offer hypotheses - and a few studies actually test some of them -- regarding the reasons behind this time path of TFP. We should note that one likely reason behind the recovery of TFP in the 1990s in Brazil (as well as other Latin American economies) is the procyclicality of productivity estimates, that tend to follow the pattern of recession in the 1980s and recovery in the 1990s experienced by many economies in the region. 3.24 This is particularly clear from the annual TFP growth figures underlying the period-average TFP growth rates in Brazil to which we have referred so far. These figures display large swings from year to year, tracking closely the economys annual growth performance. For example, most estimates show abrupt TFP declines in the severe recessions of 1983 and, especially, 1990 (Teixeira da Silva 2001, Gomes 2001, Bonelli and Fonseca 1998).3 Indeed, estimated TFP growth is strongly correlated with the growth rate of the corresponding value added aggregate, whether at the economy-wide level (GDP) or the industrial level (industrial GDP). For example, Gomes (2001) reports that the correlation between his estimates of aggregate TFP growth and observed GDP growth is in the range .72-.76, while for industrial TFP the correlation is even higher, .86-.88. This procyclicality reflects in part measurement error regarding the utilization of labor and capital, for which no good proxies exist in the Brazilian case, as well as the possible action of economies of scale unaccounted for in the methodologies underlying the TFP estimates in the available studies. I 3.25 But there are also other factors at play behind this dynamic pattem. First, the decline in macroeconomic instability in the 1990s relative to the preceding decade, particularly reflected in 30 For this reason, Gomes (2001) excludes the year 1990 when computing period averages of TFP growth. 58 the containment of inflation after 1994, likely played an important role in the recovery of TFP, by facilitating investment decisions and restoring the informational value of relative prices to guide resource allocation, as argued by Teixeira da Silva (2001) and McKinsey (1998). 3.26 Secondly, the structural reforms of the 1990s, that strengthened the role of domestic and foreign competition, have also been underscored by several studies, which have particularly emphasized the impact on TFP of the lowering of foreign trade barriers that took place over the last decade (Hay 1997, McKinsey 1998, Rossi and Ferreira 1999, Gomes 2001). 3.27 The decline in public infrastructure investment as part of the fiscal crunch of the 1980s, and the action of labor regulations constraining firms' ability to shed labor (and hence their willingness to employ it) are also among the reasons that have been offered to explain the productivity slowdown of the 1980s. These are explored in some detail in McKinsey (1988), which concludes that they-did not play a major role in the observed time path of TFP in Brazil. We shall revisit these issues in section 5 below. 3.28 Against the background provided by these earlier studies, this paper brings original evidence on the evolution of TFP in Brazil's industrial sector, based on the a large firm-level data set that has remained virtually unexplored so far. This section outlines the analytical framework employed to construct these new TFP estimates and summarizes the main features of our data, as well their potential shortcomings. Estimating TFP 3.29 The analytical approach underlying our estimations of TFP is described in detail in Muendler (2001a); here we just summarize the main features. The approach is based on the estimation of sector-specific Cobb-Douglas production functions relating output to capital, intermediate materials and skilled and unskilled labor: Yij = fki, + ibtl, + ,6wht + /5 mmit + Wil + Ei (1) 3.30 Here ydenotes real output, k is capital, m are materials and I is labor input, with the superindices bland wh respectively denoting the unskilled and skilled categories (more precisely, blue and white-collar in our case). All variables are expressed in logs, and the subscripts i and t refer to the firm and period, respectively. (Log) TFP is measured by the composite error term, which consists of a serially corelated productivity index wobserved by the firm, and an unobservable random technology shock e. 3.31 Estimation of (1) using firm-level data poses two well-known problems. The first one is that firms can choose their inputs on the basis of the observed productivity index w, which renders the right-hand side variables endogenous and requires an instrumental variable procedure for consistent estimation of the fs. The second, related problem is that firms' entry and exit decision will likewise depend on their level of productivity; this means that at any given time the available sample of firms suffers from selection bias - due to the exit of the least productive firms and the 59 entry / survival of the most productive ones. The estimation procedure adopted here follows that of Olley and Pakes (1996) and is designed to overcome the simultaneity and selection problems when estimating production functions. 3.32 In this manner, we can recover from estimation of (1) firm-level values of TFP and hence its growth rate. To arrive at sector- or industry-wide TFP growth estimates, which often are of direct interest, it is important to note that aggregate TFP change between two periods can be decomposed into four ingredients:31 (i) the TFP growth of incumbent firms, given their size (i.e., given their respective shares in total output); (ii) the change in size of incumbent firms, given their respective TFP levels; (iii) the TFP level of new entrants; (iv) the TFP level of exiting firms. We shall return to this decomposition below. 3.33 To assess empirically the factors behind observed TFP performance, we rely mainly on a regression framework relating TFP, as estimated from (1) above, to a set of explanatory variables that summarize firm and sector characterstics. The basic specification is of the form oil = x11S + aj + uj, (2) where x is a vector of explanatory variables, Sare the parameters to be estimated, ais a firm- specific time-invariant effect and u is a random disturbance. In the experiments reported below, parameter estimates for equation (2) were obtained using fixed-effects regressions, since Hausman tests unambiguously rejected the random effects specification. 3.34 In addition to (2), we also estimated specifications with the dependent variable in first differences, to assess the effects of different factors on TFP growth. In these equations the explanatory variables were lagged one period, since the objective is to examine the role of firm and sector features on subsequent TFP growth performance. For both the level and difference regressions we checked for residual autocorrelation using the technique developed by Baltagi and Wu (1999), devised to control for time gaps in the sample (i.e., the unavailability of PIA in 1991) in unbalanced panels. In general we found no evidence of serial correlation. Data 3.35 Our source of firm-level data is the Pesquisa Industrial Anual (henceforth PIA), an annual survey of manufacturing firms" administered by the Brazilian census bureau since 1986, and available until 1998 (with the exception of 1991, in which no survey was conducted). The survey is biased towards medium and large firms, and thus cannot be viewed as representative of the manufacturing sector as a whole, which includes a very large number of small firms. The features of PIA are described in detail elsewhere (Muendler 2001 b), so that here we will limit the discussion to a few key issues. 3' The decomposition we use follows along the lines of Baily, Hulten and Campbell (1992). 32 PIA contains also plant-level data, although much less comprehensive than the firm-level data used here. 60 3.36 Sample: The sampling method, and thus the representativity of the sample vis-a-vis the universe of larger industrial firms, has changed over time. When the survey was first launched, it included all of the largest manufacturing firms, plus a random sample of medium-sized ones, and a non-random selection of (large) new entrants. Over the years, all surviving originally-surveyed firms remained in the sample, while the inclusion of new entrants was discontinued in 1993. Exit of old firms and absence of new entrants must have surely reduced the representativity of this evolving sample. Further, in 1996, almost a third of the original sample was dropped from the PIA records due to a change in sampling method. As a result, the data before and after 1995 might not be strictly comparable. 3.37 Measurement issues: PIA contains income statement and balance sheet data, plus information on a few economic variables -- among them, number of workers and investment flows. From the perspective of TFP estimation along the lines described earlier, these data pose major measurement problems: * Only the number of workers (divided into white and blue-collar) is available, not hours worked. No information is available either regarding utilization of the capital stock - - although this problem is often encountered in developing as well as developed- country firm-level data. As a result, standard estimation procedures such as ours will incorrectly identify as TFP changes any output fluctuations derived from varying utilization of labor (and capital) inputs. As hours worked are strongly procyclical, our TFP estimates will tend to behave procyclically as well - although as we shall see below this seems to be less of a problem with our data than in earlier studies.3 * Brazil's high inflation environment during most of the period of analysis (the inflation rate averaged 820% over 1986-94) introduces considerable difficulties for the calculation of real output and intermediate input series from the information in PIA. These series must be constructed from income statements, which are reported in nominal terms. Hence the task requires adequate price deflators for outputs and intermediate inputs; in the absence of firm-level price data, it is customary to use some kind of sector-level price indices. In our case, the measurement error generated by the use of these 'approximate' deflators is likely to be magnified by the high degree of relative price volatility that usually accompanies extreme inflation34, as well as by the fluctuation in output sales, input purchases and inflation over the year. Considerable effort was devoted to the construction of adequate sector-specific price indices for input purchases, and a number of alternative deflators were considered for output.3 But in spite of our efforts it is 33 There are reasons other than measurement error why productivity should be procyclical; see Basu and Fernald (2000). However, in our data these are very likely of secondary importance relative to the measurement error mentioned in the text. 34 For the Brazilian case, see Fava and Cyrillo (1999). 35 The results reported in this draft make use of the sectoral WPI. See Muendler (2001 b) for extensive details on the efforts made to construct appropriate deflators for output, intermediate inputs and capital. 61 nevertheless possible that the resulting measures of firm-level real output and inputs may still contain substantial measurement error. * Construction of the capital stock poses similar difficulties. Up to 1994, our capital stock figures are mainly based on balance sheet data (which typically include an adjustment for inflation based on government-mandated price indices that tended to understate actual inflation). The situation is more complicated after 1995, because PIA ceases to include balance sheet data and capital stock figures for later years must be constructed on the basis of investment flows. As with value added, considerable efforts were made to arrive at suitable deflators for investment and the capital stock; yet measurement error is still likely to remain in the constructed capital stock series. 3.38 In summary, these measurement problems" imply that the TFP estimates reported below should be taken with considerable caution. But it is also worth noting that in the context of regressions such as (2) much of the measurement problem error is likely to wash out. As it is well- known, measurement error in the dependent variable should not affect the consistency of the parameter estimates provided the measurement error is uncorrelated with the right-hand side variables. If this latter condition is met, the inferences drawn from estimation of equations like (2) regarding the determinants of TFP should remain valid even if TFP itself is measured with error. 3.39 Sample selection and aggregation: Firms in the sample were grouped into sectors according to their 2-digit level industrial classification -- the so-called Nivel 50 classification -- which comprises a total of 27 industrial sectors accounting for roughly two-thirds of total industrial value added in the late 1990s.37 For most of the empirical analysis, however, only thirteen sectors possessing the largest number of firm-year observations" were retained. In the late 1990s, these thirteen sectors accounted for about half of the value added by Nivel 50 industry, or about one-third of value added in overall industry (Figure 3.3). Equation (1) was separately estimated for each of these thirteen sectors. Appendix Table Al presents the detailed list of industrial sectors in Nivel 50 as well as those retained for the analysis. 36 We should also note that, like many other firm-level studies, we also lack information on input quality -- other than the crude distinction between blue- and white-collar workers. One should keep in mind that changing input quality has been shown to be a major factor behind TFP growth; see Jorgenson (1990) and Gu and Ho (2000) for some illustrative figures for the US and Canada. 37 Nivel 50 excludes 5 sectors included in overall industry: Minerals, Oil and gas, Other industries, Utilities and Construction. The latter is by far the largest, accounting for nearly 10 percent of GDP. 38 2,500 at a minimum. 62 Figure 3.3 Industrial Aggregates (Percent of Total Industrial Value Added) 100 90 80 70 60- 50- 40- 30- 20 100 1995 1996 1997 1998 1999 * PIA 0 Retained Sample Figure 3.4 Industry-wide TFP, 1986-98 1.3 1.2 1.1 1 - ~~~~~~~~3*.- -H -U- 0.9 0.8 0.7 86 87 88 89 90 91 92 93 94 95 96 97 98 -U-log TFP Stayers, Entry ant Exit (All lindustry) ---Log TFP Gomes (2001) -+--Log TFP Stayers, Entry and Exit (Industry 13) 63 Overview 3.40 We next review the TFP estimates obtained from PIA using the methodology described in the preceding section.39 We first review the industry-wide results and then examine their sectoral and regional breakdown. Figure 3.4 plots the estimated (log) TFP series for all (Nivel 50) industry, as well as that corresponding to the 13 sectors under consideration. In addition, for the purpose of comparison the figure presents also the industry-wide log TFP series estimated by Gomes (2001) on the basis of data from PIM (Pesquisa Industrial MensaP°). For ease of comparability, all three series are rebased to equal 1 in 1986. 3.41 The time pattern of our firm-level data estimates of TFP is qualitatively similar to that found by most of the studies summarized in section 2 above: TFP experiences a declining trend in the late 1980s, and then recovers in the 1990s (note that the year 1991 is omitted because no PIA survey was conducted). However, the amplitude of this cycle is very small, and on the whole the figure suggests that Brazil's industry-wide TFP has shown little change over the period of analysis. This conclusion is further strengthened if we ignore the year 1990, in which the severe recession that accompanied the Collor plan results in a largely artificial collapse of measured TFP due to the sharp decline in labor and capital utilization. 3.42 As Figure 3.4 also shows, until 1995 the PIA-based TFP estimates display a time pattem roughly similar to that of Gomes' (2001) PIM-based estimates, but the volatility of the former is much smaller than that of the latter. This suggests that the PIA-based estimates may be less subject to measurement error, and hence display less procyclicality, than the estimates obtained by Gomes. After 1995, however, the two sources yield divergent TFP estimates, with those of Gomes showing a steep increase that is absent from the estimates obtained here. 3.43 As noted earlier, aggregate TFP growth reflects resource reallocation across firms in the form of exit by old firms, entry by new ones, and changes in the productivity of staying firms (with this latter component in turn including both productivity changes at the firm level and changes in the relative size of staying firms).4" Using the PIA firm-level information we can separate the effects of entry and exit from the change in the TFP of the group of staying firms. The results of this decomposition are displayed in Figure 3.6 which compares the time path of aggregate TFP for the 39 Any TFP decomposition is influenced by the methodology used. Therefore one should view these results with cautious, since each one of these methodologies has strengths and weaknesses. 40 Unlike PIA, whose unit of observation is the firm, PIM is based on production-line data (although the study by Gomes cited in the text is based on aggregated PIM information, rather than the raw data). While PIM's sample size is considerably more limited that that of PIA, PIM offers the advantage that it contains information on hours worked, so that utilization can be taken into consideration in the analysis. See Gomes (2001) for details. 41 The importance of these reallocation effects for aggregate measures of TFP has been amply documented - e.g., by Jorgenson (1990) at the macroeconomic level and by Olley and Pakes (1996) at the firm level. 64 13 sectors of analysis (already shown in Figure 4) with the one that results from ignoring entrants and exiters. The two senes move very closely together; the biggest gap between them arises in 1994, but its magnitude is virtually negligible. This is essentially a reflection of the limited role of entry in the PIA sample. Figure 3.5 Brazil Industry-wide TFP 13 Sectors, All firms and Stayers, 1986-98 1.04 1.03 1.02 1.01A 0 ~~~~~~~~~~~~~~~ /~~~~~~~~~~~~~~ ;'^- ~~~~~~~~~~ \-.- 0.99 0.98 0.97 0.98 87 88 89 90 92 93 94 95 96 97 98 -rn-Log TFP Stayers, Entry and Exit (13 Selected Sectors) - i- Log TFP Stayerst (Industry 13) 3.44 In sum, our estimates agree with previous results in that they show a recovery of TFP in the 1990s, but disagree in, that they also show much less TFP variation over time than earlier estimates. Table 3.5 condenses the information in Figure 3.5 in terms of the average TFP growth rate over the 1980s and 1990s resulting from the PIA data; for comparison the figures from Gomes (2001) are also presented. In the 1980s TFP declined, at an annual rate of 0.8 percent (0.3 if 1990 is ignored) for the 13 sectors of analysis, and 1.6 percent for overall industry (1.3 without 1990). In the 1990s, however, TFP grew at positive (albeit modest) rates, of 0.3 percent per annum for the combined 13 industrial sectors on which we focus, and 0.2 percent for overall industry. Table 3.5 Brazil Industry-wide annual TFP Growth, 1987-98 1987-89 -0.29 -1.30 1 .67 1987-90 -0.81 -1.62 -2.49 1991-98 0.29 0.19 5.99 1987-98 0.10 -0.33 2.12 65 Figure 3.6 Decomposition of TFP Growth among Stayers by Sector, 1987-98 DECOMPOSiTION OF TFP GROWTH AMONG STAYERS 1987-98 98 97 96 9$ 94 9:3 92 90 . 89 88 87 1 . . z -0.2 4.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 *OTotIaTFPGrowth Osectorsize OSectorTFPGrowth *FirrnSze EFirm TPGrowth 3.45 As mentioned earlier, the evolution of aggregate productivity of staying firms reflects two forces: the time path of firm-level TFP, and the changes in firms' relative size. It is useful to examine the separate contributions of these two factors to the observed pattem of TFP over the sample period. The decomposition is provided by Figure 3.6, which for each year in the sample shows the contribution to TFP growth from firm-level TFP and changing firm- and sector size. It is apparent that firm size changes actually detracted from aggregate productivity in most years, with the only exceptions of 1992 and 1996, while sector size changes had a more mixed contribution to observed TFP growth. This seems to suggest that resource reallocation proceeded in a way harmful to aggregate productivity; we shall return to this issue later on. 66 Details by Sector 3.46 The sectoral patterns of TFP growth underlying these aggregate figures from PIA are summarized in Table 3.6. To save space, we only report average TFP growth rates over the same subperiods as in the preceding table; the figures include only the group of staying firms for the 13 sectors under consideration. The annual detail by sector can be found in Table A.2 in the Appendix. In general, it shows a fairly homogeneous time profile across sectors, with the majority of them experiencing a drop in TFP around 1989-90, a recovery in 1992-93 and a more diverse pattern thereafter. 3.47 According to the sector-specific TFP growth rates in Table 3.6, the period up to 1990 was characterized by a TFP slump in virtually all sectors under consideration, with the only exception of Other Metallic Products, which exhibited positive TFP growth during the late 1980s. In this period, the worst performance corresponded to Apparel and Wood and Furniture, for which the estimated TFP decline is particularly large. Again, exclusion of 1990 would considerably reduce the magnitude of some of these seeming TFP declines - but would not reverse them. In the 1990s, in contrast, 9 out of 13 sectors experienced positive TFP growth, at especially high rates in Electrical Equipment and Vehicles and Parts; these two sectors are also the ones that display fastest TFP growth over the entire period of analysis. Four sectors continued to experience a TFP slump in the 1990s: Food, Textiles, Leather and Non-Metallic Minerals. 3.48 This pattern is broadly similar to that reported by Rossi and Ferreira (1999) using aggregate PIA data (as opposed to the firm-level data used here). While their sector disaggregation differs somewhat from ours, for the sectoral aggregates common to both studies they also report a uniform TFP decline in 1985-90, followed (in their case) by a uniform increase in 1991-97. Further, in their results the Vehicles and Parts sector is also the performance leader in the most recent subperiod, with the Electrical Equipment sector among the top performers as well.42 The rapid productivity gains achieved by the Vehicles and Parts sector in the 1990s are also documented in McKinsey (1998), which traces them to the developments in trade opening in the 1990s, to be reviewed below. 42 They also find rapid TFP gains during 1991-97 in the Chemical sector, which is not included in our analysis. We should note, however, that our sectoral TFP performance rankings are fairly different from those reported by Gomes (2001), who uses using National Accounts data on industrial value added by sector. He reports negative TFP growth over 1990-98 in 8 out of our 13 sectors. This might reflect his inclusion of the year 1990 in the period of analysis. In that year, most sectors display a TFP decline, which basically reflects the deep recession that took place. Nevertheless, Gomes also finds that the Automobiles sector was the at the top in terms of TFP growth in the 1990s. 67 Table 3.6 TFP Growth bySctor Non Metallic Mineral Products -7.77 -1.16 -3.56 Other Metallic Products 1.17 1.15 1.16 Manufacturing and Maintenance of Machinery -5.42 3.06 -0.02 Electrical Equipment -3.32 12.07 6.48 Wood and FurnHure -12.94 0.43 -4.43 Paper, Pulp and Cardboard -0.02 2.83 1.80 Plastics -5.76 0.14 -2.01 Textiles -1.25 -2.06 -1.77 Apparel -16.73 6.05 -2.23 Leather Products and Footwear -0.79 -2.34 -1.78 Processed and Edible Products -7.31 0.03 -2.64 Food and Beverages -3.06 -4.10 -3.72 Vehicles and Parts -3.34 6.95 3.21 68 Details by Region 3.49 Finally, we examine the patterns of TFP growth from a regional perspective. For this purpose, we divide the country into three geographical regions: Sao Paulo, the South and the rest of the Southeast, and the Northeast. It is important to note from the outset that these regions differ considerably in terms of size of their industrial sectors. In 1997, Sao Paulo contributed 44 percent of all-Brazil industrial value added. Adding the South and the rest of the South East, the figure rises to 85 percent. The Northeast region accounts for the remaining 15 percent. Further, the three regions differ also considerably in terms of the sector composition of their industries. Since not all sectors are significantly represented in every region, we focus only on industry aggregates. Figure 3.7 Brazil Industry-wide TFP by Region 13 Sectors, Stayers, 1986-98 I 05 r0s~ ~ ~ ~~~~~~~~... ........ . 0 95 09 Tt 0.85 0.8 86 87 8a 89 90 92 93 94 95 98 97 98 |=Brazil --SaoPaulo - NorthandNorth-East --*--SouthandSouthEastexciudin SaoPauto 3.50 Figure 3.7 plots the time pattem of TFP across the three regions, as well as the all-Brazil average for comparison. The figures correspond to the 13-sector aggregate, and include only staying firms. Until 1990, the profile is fairly homogeneous across regions - they all experience declining TFP. In the 1990s, however, a large gap opens between the Northeast, which experiences a further abrupt decline in TFP in 1994-95, and the other two regions, which show rising TFP, especially rapid in the South and Southeast region. This contrasting regional performance likely reflects the limited presence in the Northeast of the industrial sectors that were the TFP performance leaders in the 1990s (such as Cars and Parts and Electrical Equipment).43 43 Indeed, f we perform the same exercise but including all industrial sectors - rather than the 13 in the text - the estimated TFP performance of the Northeast compares much more favorably with that of the other regions. 69 3.51 We next study the policy and structural factors shaping TFP performance and, more broadly, firm performance in Brazil's industry, using firm-level data from PIA and the firm-level TFP estimates constructed as described in section 3 above. The ultimate objective is to assess how selected aspects of the policy and regulatory framework have impacted on the observed evolution of Brazilian industrial firms over the last decade. 3.52 Specifically, we focus on four broad questions: Trade barriers: how has the changing exposure to foreign competition affected firm productivity? Did the decline in protection witnessed during the 1990s encourage TFP improvements? Knowledge: what has been the role of knowledge embodied in physical and human capital in promoting firm productivity? Has the latter been positively affected by the availability and use of foreign equipment-which may embody the latest technology available in world markets-information equipment and human capital? * Resource allocation: How do firm size and TFP relate to each other? Do newer firms bring in higher-productivity techniques? Does higher capital intensity of production result in higher or lower TFP? Over the sample period, did the regulatory environment result in resource reallocation towards the most productive firms? How did the rules and regulations that constrain labor adjustment affect firm performance? Did they pose a significant burden to those firms seeking to downsize their operation? . Physical capital and investment: was insufficient investment an obstacle for productivity improvement? What factors constrained firms' ability to expand their capital stocks? 3.53 To address these issues, we rely mainly on the results from the multivariate regressions introduced in section 3 above. To avoid biases resulting from omitted regressors, we include simultaneously in the regressions a set of explanatory variables that attempt to capture the various factors affecting TFP. 3.54 Thus, to reflect the changing exposure of domestic producers to foreign competition, we include in the regressions the value added-weighted nominal tariff rates by sector constructed by Kume, Piani and Souza (2000); the real exchange rate; the sector-specific degree of local market penetration by foreign exporters (defined as the ratio of imports over total final uses for each sectors goods, based on Ramos and Zonnenschain 2000), and the firm-specific ratio of exports to total sales. 3.55 Next, regarding the impact of embodied knowledge, we examine the effects on TFP of the composition of capital and labor inputs. Regarding capital, we focus on the ratio of used machinery 70 to total and the ratio of imported machinery to the total. For labor we examine the effects of the white / blue collar composition of the labor force, as a proxy for its skilled / unskilled composition." 3.56 Regarding resource allocation, the available regressors are the capital / labor ratio, the age of the firm and its size. The latter is measured by both value added and output, to capture possible differential effects across the two variables. Further, for both size and age we allow for nonlinear effects by including quadratic terms in the regressions.45 3.57 In addition to these basic regressions, we also performed additional empirical experiments to assess the determinants of firm size change, the effects of labor regulations, and the determinants of investment. These experiments are described in detail in Muendler (2001 b); below we limit ourselves to a brief summary of the results. 3.58 Not all of the regressors listed above are available in each sample year. Specifically, the composition of the capital stock (foreign and used machinery, as well as computers) is available in 1986-95. In turn, the export ratio to sales is unavailable prior to 1989. Thus we the only observations for which all candidate regressors are simultaneously available are 1989-95.4' For this reason, below we present two sets of regression results, corresponding respectively to the periods 1989-98 (thus excluding the variables describing the composition of capital) and 1989-95 (including them). Descriptive statistics for the dependent and independent variables in the altemative regression samples and specifications are shown in Table 3.7. 3.59 The parameter estimates from the regressions appear in Table 3.8. We defer discussion of the individual coefficients to the subsections that follow, but here we note that on the whole the regressions possess a satisfactory explanatory power given the huge sample sizes, especially in the case of the equations with TFP level as the dependent variable. Nevertheless, the R-squared reported at the bottom of the table suggests that unobserved factors may be responsible for much of the variation in TFP and its growth rate. In any case, the estimated parameters are in all cases highly significant jointly, as implied by the corresponding Wald statistics. Finally, we should note that these regression results should be interpreted with some caution, as we cannot rule out the possibility of reverse causation from TFP to some of the regressors - e.g., high tariffs might result from successful lobbying by low-productivity firms, rather than the other way round. 44 For the years 1996-98 PIA also includes information on the ratio of foreign materials to total materials used. In these years, however, the composition of the capital stock is not available. Nevertheless, we performed some additional regressions using only 1996-98 data to assess the effect of the composition of intermediate goods on TFP and its growth rate. The results were never significant, and hence we opt for not reporting these regressions in the text. 45 In addition, the regressions also include regional dummies, which were invariably insignificant and hence are not reported below, and the inflation rate, to control for cyclical factors - rather than just the effect of inflation itself. For this reason we regard its estimated coefficient as uninformative and ignore it in the discussion below. 46 Ignoring the information on the composition of intermediates available for 1996-98 only. 71 Table 3.7 Descriptive Statistics for alternative samples TFP TFP growth 1989-98 1989-95 1989-98 1989-95 Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Trade__ _ _ Tarff 0.2288 0.1470 0.2125 0.1315 0.2110 0.1384 0.1978 0.1186 Real Exchange Rate 1.0006 0.1299 1.0444 0.1396 1.030 0.1362 1.0655 0.1366 Exports to sales Ratio 0.5627 0.1370 0.0587 0.1361 0.058 0.1358 0.0590 0.1334 Foreign Penetration 0.0651 0.0635 0.0611 0.0587 0.0644 0.0612 0.0621 0..0594 Knowledge Used Machinery to Total 0.0586 0.1505 0.0579 0.1479 Machinery Imported Machinery to Total 0.0467 0.1191 0.0489 0.1209 Machinery I Computer Equipment to 0.0070 0.0124 0.0071 0.0120 Total Machinery Skilled Employees to Total 0.2514 0.1748 0.2614 0.1767 0.2521 0.1739 0.2585 0.1744 Workers Resource Allocation Age 25.5432 16.7961 27.3608 16.54 27.0833 16.8392 27.8095 16.5874 Size (Output) 5.49E+07 2.74E+08 6.33E+07 3.04E+08 6.06E+07 3.02E+08 6.73E+07 3.27E+08 Phsical Investment _ _ Capital-Labor Ratio 9841.54 25046.65 9232.446 24747.6 9529..551 24372.42 9080.968 24157.72 Total Factor Productivity TFP 5.944 2.0232 6.009 2.0440 TFP Growth _ 0.0438 0.3863 0.0514 0.3774 72 Table 3.8 Regression Coefficients from TFP Level and Growth Equations TFP TFP growth 1989-98 1989-95 1989-98 1989-95 Trade : Tariff -0.1688- -0.0543" -0.1449- -0.2299" Real Exchange Rate -0.0580" -0.0106 -0.9248" -0.9809 Exports to sales Ratio -0.0741" -0.1477' -0.1239' -0.2057- Foreign Penetration 1.6274" 0.9048- -0.7590" -0.6825" Knowledge . Used Machinery to Total 0.0023 -0.0010 Machinery Imported Machinery to Total 0.0557 0.0268 Machinery Computer Equipment to 2.0329" 1.3266- Total Machinery Skilled Employees to Total -0.0057 0.0317 0.0642" 0.1157" Workers Resource Allocation __. Age -0.0301 -0.0049 -0.0333" -0.0519- Age Squared -0.0001" -0.0002- 0.0000 0.0001 Size (Output) 2.19E-09" 2.41 E-09" 1.47E-09- 1.70E-09 Size Squared -1.95E-1 9" -1.95E-1 9" .1.25E-19" -1.35E-19-- Physical Investment . Capital-Labor Ratio -8.05E-07" 1.92E-06" 7.97E-07o 3.25E-06" R2within 0.1373 0.1018 0.0653 0.0838 No of Observations 30913 15739 20679 12164 Wald Test of 0.0000 0.0000 0.0000 0.0000 Ioint significance of all parameters (p value) Note: *'significant at 5 percent level, *significant at 10 percent level 73 TFP and Exposure to Foreign Trade 3.60 Until the late 1980s, Brazil's industrialization had been effectively based on high import barriers sheltering domestic producers from foreign competition. The process of trade liberalization-involving the elimination of non-tariff barriers and the reduction of tariffs-started in the late 1980s and gained momentum in the 1990s, although it suffered a partial reversal in the latter part of the decade. Thus, the external liberalization coincided with the privatization and domestic deregulation that also took place in the 1990s. 3.61 Trade liberalization proceeded in four stages (Kume, Piani and Souza 2000). In 1987-89, the reforms sought primarily the removal of redundant tariffs, which were quite numerous due to the pervasive non-tariff barriers. As'a result, the (value-weighted) average effective tariff declined from 68 to 39 percent. In the second stage (up to 1993), non-tariff barriers were removed -- leaving tariffs and the real exchange rate as the main protection instruments -- and tariffs reduced, bringing the average effective tariff down to 15%. Next, in 1994 with the advent of the Real plan the process of tariff reduction sped up, and the average effective tariff fell to 12%. Finally, in 1995-98 a partial reversal took place. First some rates were raised - on automobiles, consumer durables, textiles - and then a generalized 3% increase was imposed, so that the effective rate rose back to an average of 15%. Figure 3.8 Average Nominal Tariff and Import Penetration All Industry, 1989-98 a Foreign Penetr. ---6- Nominal Tariff - Real Exch. Adj. Tariff .7 .6 - .5- .4- .3- .2- 1989 1992 1995 1998 Calendar Year 74 3.62 Figure 3.8 plots the time path of the average nominal tariff, with and without adjustment for the protection accorded by the level of the real exchange rate, as well as the degree of foreign market penetration (as defined above), with all variables corresponding to the overall industry average. The steep decline in protection after 1989 is apparent from the figure, as is the partial reversal of the reform after 1995 and the steady increase in foreign penetration. 3.63 This general framework conceals significant variation in the experience with trade liberalization across sectors. One interesting experience is that of the automobile and car parts industry, that prior to the liberalization enjoyed one of the highest nominal protection rates, as well as the highest effective protection rate throughout the reform period (Kume, Piani and Souza 2000). As figure 3.9 shows, in this case the pattern of abrupt tariff decline and reversal is particularly striking, as is the steady increase in foreign penetration in spite of the partial rollback of the tariff cuts in the late 1990s. Figure 3.9 Average Nominal Tariff and Import Penetration Auto and Car Parts, 1989-98. e Foreign Penetr. - Nominal Tariff - Real Exch. Adj. Tariff .7 - .6 .5- .4- .3- .2- 1989 1992 1995 1998 Calendar Year 3.64 A few studies have focused on the impact of this reduction in trade barriers on Brazil's TFP performance. Hay (1997) examines the impact of nominal and effective protection rates on the TFP of a group of large industrial firms. He finds strong negative effects of both variables, as well as from protection accorded by depreciated real exchange rates, and concludes that both the elimination of non-tariff barriers at the end of the 1980s and the reduction in tariffs in the 1990s had strong positive effects on the level of TFP - along with the domestic deregulation and privatization. In turn, Rossi and Ferreira (1999) likewise find a strong negative impact of nominal and effective tariffs on the growth rate of aggregate TFP, and conclude that the process of trade opening was a 75 key factor behind the observed TFP recovery in the 1 990s. Finally, the report by McKinsey (1998) also emphasizes tariff reductions as the main force behind the observed productivity improvements in the automobile and auto parts sectors, and underscores the obstacle that high tariffs on capital goods (particularly telecommunication equipment) posed to investment and plant modernization at the beginning of the reform period. 3.65 The first group of coefficients in Table 3.8 allow us to assess the impact on TFP and its growth rate of this opening up to foreign competition.47 On a prioui grounds, we expect protection via tariffs and/or the real exchange rate to have a negative impact on TFP, while market penetration by foreigners and contact with export markets should be expected to have a positive effect. 3.66 Starting with tariff levels, we find that higher tariffs are invariably associated with lower TFP levels and growth rates in both sample periods. In the case of TFP levels, however, the association is significant only in the longer sample. In the case of TFP growth rates, the coefficients are significant in both samples.. They suggest a positive effect on TFP growth around 0.2 percent for each point decline in tariff rates. Thus, a 30-percentage point decline in tariff rates (roughly the observed decline in the average tariff rate over the reform period) would yield, other things equal, a 6 percentage point acceleration in TFP growth. 3.67 In turn, the real exchange rate also carries a uniformly negative coefficient, both in the regressions of TFP level and growth rate, although like with tariffs the coefficient is insignificant in the level regression for the 1989-95 sample. More depreciated (=higher) real exchange rates allow domestic firms additional protection from foreign competition. They also tend to raise the real cost of imported intermediates and machinery. On the whole, the regression results suggest that the net result is that a higher real exchange rate allows firms to get by with lower and/or more slowly growing TFP.8 3.68 Foreign penetration in local markets has a significant positive effect on the TFP level in both sample periods (columns 1-2). However, it has a negative impact on the subsequent growth rate of TFP (columns 3-4), which seems somewhat puzzling. 3.69 Finally, in the case of export orientation we find a negative and significant association with TFP and its growth rate in both sample periods, contrary to expectations. We should note, however, that in this experiment we are holding constant a number of key variables - skill intensity, machinery use, age and size of the firm - which other studies have found to be significantly associated with export orientation. Thus, the interpretation of this result is somewhat unclear.49 47 In addition, trade opening allows enhanced access to foreign capital, which may have an impact on productivity HI its quality is above that of domestic equipment. We return to this issue in the next subsection. 48 However, caution is necessary regarding this result. The reason is that the real exchange rate measure included in the regressions lacks cross-sectional variation and only displays time variation. Thus, it may partly capture the effects of other aggregate shocks affecting Brazilian industry, in addition to the effects of the real exchange rate itself. 49 Moreover, additional work in progress by Muendler (forthcoming) using a different econometric approach finds that export orientation and TFP show a positive association among Brazilian manufacturing firms. 76 TFP and Knowledge Embodied in Input Quality 3.70 We turn to the second broad issue, namely the role of knowledge embodied in physical and human capital. To explore this question, we focus on the variables describing the composition of the physical capital stock - specifically, the ratios of used machinery, imported machinery and computer equipment to total machinery -- and the white / blue collar composition of the labor force. It is worth noting that the foreign machinery ratio reflects access to foreign markets on the input side, as opposed to the output side access discussed in the preceding subsection. On a priori grounds, we would expect all these input quality / embodied knowledge measures to be positively related to TFP performance - with the exception of the used machinery ratio, which should exert a negative effect. 3.71 Of all the variables just listed, only the skill ratio is available in both samples under study. Table 3.8 (second group of coefficients) shows that it has a positive and sianificant effect on TFP growth in both periods, and no significant effect on the TFP level in either period. 3.72 Regarding the composition of the capital stock (available for 1989-95 only), the ratio of computers to machinery is significantly associated with higher TFP and TFP growth, and its coefficient is of large magnitude.' In contrast, neither the used nor the foreign machinery ratio exhibit any significant effect on TFP or its growth rate. The latter result, in particular, suggests that quality differences between domestic and foreign machinery are inconsequential from the perspective of productivity.51 Finally, the same conclusion applies to foreign intermediates: their ratio to total intermediate inputs not significantly associated with TFP level or TFP growth. Resource Allocation 3.73 We now tum to the role of resource allocation in the observed pattems of TFP. In particular, we are interested in four questions: (i) does productivity varies systematically with firm age and size - i.e., do newer / smaller firms bring in higher-productivity techniques? (ii) does capital intensity bear any relationship to TFP? (iii) did more productive firms tend to expand further than less productive ones? and (iv) do labor regulations pose a significant burden on firms' operation? 3.74 To answer the first two questions we can look at the third group of estimates in Table 3.8. Regarding age, in general we find that older firms exhibit significantly lower TFP as well as TFP growth. In the case of the TFP level, the effect is more than proportional - it becomes stronger as firms become older. Its magnitude seems considerable: other things equal, a one-year increase in age lowers TFP growth by 3-5 percent (note than mean age is around 25-30 years). In interpreting this result, however, it is important to keep in mind that other features that vary systematically with firm age - e.g., firm size - are being held constant. 3.75 Regarding firm size, which in the regressions is measured both by output and value added, in both cases we find a strongly significant effect, which in the case of value added is nonlinear: larger firms exhibit higher productivity and faster productivity growth, although beyond a certain 50 The computer / machinery ratio is generally very small. Its sample mean is just 0.6 percent. 51 As noted earlier, the same result was found for foreign intermediates over the shorter 1996-98 sample. 77 size the relationship changes sign. Near the sample mean, a 10-percent increase in size as measured by value added is associated with an increase in TFP of about 0.5-0.8 percentage points, and an increase in TFP growth of roughly the same magnitude. In turn, an increase in size as measured by output yields a more modest (but still positive) impact on TFP. 3.76 As for the impact of capital intensity on TFP performance, Table 3.8 shows that it exerts a positive and highly significant effect on TFP growth in both sample periods. Its impact on the level of TFP, however, is less clear - it is positive in the 1989-95 sample, when the variables describing the capital stock composition are available, and negative in 1989-98 - when they are not. 3.77 We next focus on changes in firm size. In section 4 above we saw that, contrary to expectations, the contribution of firm size changes to aggregate TFP was negative in several years. In other words, firm size change acted as a drag on aggregate productivity. It is worth inspecting this result more closely and asking whether, other things equal, low productivity prompted firm size increases over the sample period. If the question were answered in the affirmative, the suggested implication would be that the overall regulatory regime tended to encourage inefficiency. 3.78 To answer this question, we ran fixed-effects regressions with firm size change (as measured by value added) as the dependent variable, and with the same explanatory variables (other than size) as the TFP regressions summarized above, including as additional regressors the profit rate and the TFP level. The details are given in Muendler (2001a). On the whole, these regressions had a high explanatory power (with R-square coefficients around .20). The main result they yield is that size change is strongly positively affected by firm TFP levels. Thus, other things equal, more productive firms did tend to expand faster over the sample period. Other things, of course, did not remain equal, and must lie behind the negative contribution of firm size changes to aggregate TFP growth that arises from Figure 3.6 above.' 3.79 Finally, we also examined the effect of labor regulations on firm performance using additional regression experiments. Specifically, we tried to assess indirectly the burden posed by regulations on dismissals on firms' ability to adjust to changing conditions, by looking at the share of labor costs in total costs for those firms that were shrinking. The constraints on labor shedding should be reflected in rising shares of labor costs for those firms. The main finding from these experiments (see Muendler 2001 a) is that shrinking output is indeed associated with higher cost shares of labor, after controlling for other firm characteristics, which seems consistent with the above reasoning. However, it is difficult to establish the direction of causality underlying this association, and hence it cannot be viewed as conclusive. Physical Investment 3.80 Given the result above that capital intensity encourages TFP growth, it is of interest to investigate the factors that shape firms' investment decisions. To do this, our final set of empirical experiments involved the estimation of investment equations relating fixed capital formation to a set of real and financial variables; see Muendler (2001 a) for details. Among the real regressors, we 52 More precisely, while the unconditional correlation between size change and TFP level is negative in the sample, the conditional correlation (that is, controlling for the other factors mentioned in the text) is significantly positive. 78 added to the variables used so far the firm's TFP level and its capital / output ratio. In turn, the financial variables included the profit and credit ratios to output and a set of indicators of the financial structure of the firm. 3.81 The two key results from this analysis were: (i) TFP affects investment negatively - which might be interpreted as reflecting a 'catch-up' effect: firms invest precisely to raise productivity, and slow down their investment once productivity has reached a high enough level; and (ii) investment is systematically affected by financial variables, such as the profit ratio, the ratio of credit to output, the ratio of long-term credit to the total, and the credit / equity ratio. The first three of these variables are positively associated with investment; the fourth is negatively related to it. It is tempting to conclude that this reflects the action of financial constraints on firm investment; as before, however, it is difficult to establish causality, and hence the empirical association of investment with these variables has to be interpreted with considerable caution. 3.82 The pattem of TFP performance of Brazil's industrial sector has been the focus of several empirical studies in recent years. This study represents a significant addition to existing analyses in that it performs the first systematic exploration of TFP in Brazil making use of a large firm-level data set. We have focused on the evolution of productivity in 13 major industrial sectors over the period 1986-98, and examined the contribution of various factors to the observed TFP performance of firms in these industries. The major findings from the study can be summarized in six points. 3.83 On the whole, our results show that industry-wide TFP has shown little change since the late 1980s. Like in previous studies, we do find a cycle of initial productivity stagnation followed by recovery, but its magnitude is quite small. Qualitatively, the time path of TFP broadly matches that of Brazil's aggregate economic performance, suggesting that the reduction in macroeconomic instability in the 1990s played a role in the TFP recovery, as pointed out by a number of previous studies. Yet the estimated annual growth rate of TFP in the 1990s, in the 0.2-0.3 range, was still fairly modest. 3.84 While there were considerable differences in TFP growth across industrial sectors over the sample period, on the whole we find that the pattem of decline first and recovery later was fairly uniform, so that the majority of sectors experienced falling TFP in the 1980s and rising TFP in the 1990s. Comparison with eadier studies is hampered by differences in time coverage and sector disaggregation. Nevertheless, our findings agree with some previous studies in identifying the automotive and electrical equipment sectors among the top productivity performers after 1990. 3.85 The increased exposure of domestic producers to foreign competition resulting from the process of trade opening of the 1990s had a strong favorable impact on TFP. This is in agreement with the conclusions from previous studies. Our empirical estimates suggest that the 30 percentage-point reduction in the average nominal tariff may have contributed to raise industry- average TFP growth by as much as 6 percent over what it would have otherwise been. Market penetration by foreign competitors also appears to have raised the TFP level, although not its growth rate. 79 3.86 Some aspects of firms' input mix bear a systematic relationship to TFP, reflecting the contribution of knowledge to productivity. Thus, higher shares of IT equipment in total physical capital raise both TFP and its growth rate. Likewise, the skilled / unskilled composition of the labor force (proxied by its white / blue collar composition) also has a positive impact on TFP growth - which suggests that human capital investment may contribute significantly to raising TFP. In contrast, more intensive use of foreign machinery (relative to total machinery) does not seem to be associated with better TFP performance, and the same applies to the breakdown of machinery into new and used equipment. 3.87 Unfortunately, our data set does not allow proper study of the impact of new firm entry on overall TFP, but we do find that newer firms typically possess higher, and faster-growing, productivity levels. This suggests that measures that facilitate entry of new firms - by making the regulations goveming entry simpler and'more transparent and reducing red tape - could go along way towards raising aggregate TFP growth. In turn, TFP and its growth rate are positively associated with firm size - although beyond a certain size threshold the relationship is reversed. It is somewhat puzzling that over our sample period resource reallocation among existing firms (i.e., changes in their relative size) contributed little, if at all, to improve aggregate TFP performance. Nevertheless, we do find that, other things equal, TFP has a positive effect on size change - i.e., more productive firms tend to expand more. In this sense, resource reallocation proceeded in the 'right' direction. 3.88 Finally, the capital / labor ratio is positively related to TFP growth. Thus, capital deepening tends to speed up productivity. This suggests that raising investment could have a significant payoff in terms of TFP in Brazil's industrial sector. In this regard, we also find evidence that firm investment is systematically related to financial cohditions - profits, long-term credit availability, and firms' debt/equity ratio. While the evidence is-not conclusive, it could be consistent with the view that physical investment is constrained by financial market imperfections. References Baily, M., C. Hulten and D. Campbell (1992) "Productivity dynamics in manufacturing plants," Brookings Papers on Economic Activity (microeconom ics), 187-267. Basu, S. and J. Femald (2000) "Why is productivity procyclical? Why do we care?" NBER Working Paper 7940. Baltagi and Wu (1999) "Unequally spaced panel data regressions with AR(1) disturbances," Econometric Theory, pp. 814-23. Bonelli and Fonseca (1998) "Ganhos de produtividade e eficiencia: novos resultados para a economia brasileira," IPEA Texto para Discussao 557. Braga, H. and Rossi, J (1988) "Produtividade total dos fatores de produ,co na industria brasileira: mensura,co e decomposi,ao de sua taxa de crescimento," IPEA Texto para Discussao 12. Collins, S. and Bosworth, B. (1996) " Economic Growth in East Asia: Accumulation versus Assimilation," Brookings Papers in"Economic Activity, 2, pp. 135-203. 80 De Gregorio, J. and Lee, J. (1999) "Economic growth in Latin America: Source and Prospects," mimeo, December. Easterly and Levine (2000) "It's not factor accumulation: stylized facts and growth models," mimeo, The World Bank. Elfas, V. (1990) Sources of Growth: a study of seven Latin American economies, San Francisco: IDS Press. Fava, V. and D. Cyrillo (1999) "Inflacao e dispersao de precos relativos: Qual a dire,co de causalidade?" Economia Aplicada/Brazilian Journal of Applied Economics 437-57. Gomes, V. (2001) "Fatos sobre a produtividade", mimeo, IPEA. Gu, W. and M. Ho (2000) "A comparison of industrial productivity growth in Canada and the United States," American Economic Review 90(2), 172-175. Hay, D. (1997) uThe post 1990 Brazilian trade liberalization and the performance of large manufacturing firms," IPEA Texto para Discussao 523. Hoffman, A. (2000) "Economic growth and performance in Latin America", S6rie Reformas Econ6micas 54, ECLAC. Jorgenson, D. (1990) "Productivity and economic growth," in E. Bemdt and J. Triplett, eds., Fifty years of economic measurement. University of Chicago Press. Jorgenson, D. and K. Stiroh (2000) "US Economic growth at the industry level," American Economic Review 90(2), 161-167. Klenow, P. and A. Rodriguez-Clare (1997) "The neoclassical revival in growth economics: has it gone too far?" NBER Macroeconomics Annual. Kume, H., Piani, G. and Souza, C. (2000) "A politica brasileira de importa9co no periodo 1987-98: descri,co e avalia,co," mimeo, May. Liang, C. and Jorgenson, D. (1999) "Productivity Growth in Taiwan's Manufacturing Industry, 1961- 93" in Economic Efficiency and Productivity Growth in the Asia-Pacific Region, ed. by Tsu- Tan Fu, Cliff J. Huang and C.A. Knox Lovell, Northampton, MA: E. Elgar, McKinsey (1998) Productivity-The Key to an Accelerated Development Path for Brazil, Sao Paulo. Muendler, M. (2001a) "Productivity change among large Brazilian manufacturers," mimeo, U.C. Berkeley. Muendler, M. (2001b) "The Pesquisa Industrial Annual 1986-1998: a detective's report," mimeo, U.C. Berkeley. Nehru, V. and Dhareshwar, A. (1994) "New Estimates of Total Factor Productivity Growth for Developing and Industrial Countries," World Bank Policy Research Working Paper # 1313, June Olley, G. and A. Pakes (1996) "The dynamics of productivity in the telecommunication equipment industry," Econometrica 64(6), 1263-97. 81 Pinheiro, A. C. (1989) "An Inquiry into the Causes of Total Factor Productivity Growth in Developing Countries: The Case of Brazilian Manufacturing, 1970-80", Ph D dissertation, UC Berkeley. Ramos,R. and Zonnenschain, C. (2000) "The performance of the Brazilian imports and exports based on a system of national accounts: 1980-98," Paper presented at the 13 th. International Conference on Input-Output Techniques, Macerata, Italy. Rossi, J. and Ferreira, P. (1999) 'Evolu,co da produtividade industrial brasileira e abertura comercial," IPEA, Texto para Discussao 651. Teixeira da Silva, T. (2001) "Estimando o Produto potencial brasileiro", Banco Central do Brasil Trabalhos para Discuss&o 17. World Bank (1999) uExport Dynamics and Productivity: Analysis of Mexican Manufacturing in the 1990s," Report No. 19864 ME. 82 Table Al Manufacturing sectors at Nivel 50 Nivel 50 Code Description 4 Non-metallic mineral products 5 Ferrous metal products 6 Non-ferrous metal products 7 Other metal products 8 Machinery 10 Electrical equipment 11 Electronic equipment 12 Automobiles 13 Other vehicles 14 Wood and furniture 15 Paper, pulp, and cardboard 16 Rubber products 17 Non-petrochemical chemical elements 18 Fuels, petrochemicals, and resins 1 9 Chemicals 20 Pharmaceuticals and perfumes 21 Plastics 22 Natural and artificial textiles 23 Apparel 24 Leather products and footwear 25 Coffee products 26 Processed edible products 27 Meat and poultry 28 Processed dairy products, 29 Sugar 30 Vegetable oil 31 Other food products and beverages Note: The 14 sectors shown in boldface were retained for the firm-level analysis. For this purpose, sectors 12 and 13 were combined into one single sector. 83 Table 3.A2 TFP growth by sector 1987-98 Manufacturing Non and Leather Processed Metallic Other Maintenance Wood Paper Pulp Products and Food Vehicles Mineral Metallic of Electrical and and and Edible and and Year Products Products Machinery Equipment Fumiture Cardboard Plastics Textiles Apparel Footwear Products Beverages Parts 87 -11.68 5.99 6.32 16.31 1.77 18.35 8.59 5.35 -3.18 8.55 -6.42 -2.24 7.95 88 -1.77 7.72 1.03 10.22 -0.32 16.52 -9.40 8.42 -3.26 -8.92 6.44 2.24 2.47 89 -5.86 1.61 -17.25 -7.84 -38.59 -13.52 -7.73 -11.83 -63.40 -7.89 -12.61 19.20 7.89 90 -11.76 -10.62 -11.78 -31.98 -14.61 -21.43 -14.50 -6.96 2.93 5.10 -16.65 -31.42 -31.66 92 3.25 -4.05 3.13 23.53 7.00 1.34 3.79 1.02 10.89 11.23 -1.38 -9.15 21.73 93 10.29 2.97 20.98 22.51 -5.47 -8.58 11.25 9.51 16.18 10.22 25.97 2.30 27.09 94 -17.43 -6.96 -17.15 -8.43 6.30 23.04 -14.75 -13.34 10.23 -27.43 -22.37 -7.53 -5.99 95 -7.54 -3.55 -19.61 9.04 -21.66 -5.00 -5.21 -9.45 4.89 -5.44 -2.59 -4.61 -1.07 96 -1.49 13.70 11.90 20.68 8.53 1.74 -0.17 -8.85 -20.47 -13.81 -4.68 -14.33 2.59 97 4.53 1.12 15.69 8.44 3.99 4.81 6.60 5.07 14.31 4.25 8.14 -0.42' 5.65 98 0.28 4.80 6.52 8.74 4.29 2.49 -0.54 1.59 6.29 4.61 -2.86 5.02 -1.35 84 4. HOUSEHOLD INCOME GROWTH AND ITS DISTRIBUTION Prepared by Naercio Aquino Menezes-Filho and Mark Roland Thomas One Dimension of the Quality of Growth: The Equality of Growth 4.1 An old debate, recently invigorated in Brazil as elsewhere, centers upon the question of whether economic growth per se is the correct focus of public policy efforts and discussions, or indeed whether economic growth even corresponds at all with underlying concepts of public welfare. At one end of the spectrum stand those who claim that globalization and capitalism have evolved to serve a plutocracy of multinational corporations, their shareholders, and the politically elites. For proponents of this view, economic growth is at best an irrelevance, the wrong measure, at worst a symptom of the repression of the needy. At the other end of this spectrum lie observers who point out the high correlation between measures of income such as GDP per capita and other pertinent measures of the human condition: health indicators, education and literacy, even the environment. For these observers, despite its philosophical oversimplification, economic growth is a satisfactory "sufficient statistic" for welfare for most purposes of policy design. 4.2 An intermediate position is that most commonly taken by development organizations and politicians of the center, that growth should be a principal but not the sole aim of public action, and in particular that there is sufficient divergence between income growth and some other welfare components to warrant separate foci on the "multiple facets" of welfare (and, by extension, poverty). 4.3 This paper focuses on the quantity of growth and one further dimension: its distribution, in particular, how much economic growth accrues to the poor. This is neither to claim that this is the only other dimension that matters, nor that there are necessarily tradeoffs between distribution and growth. Our position will be pragmatic: that one sensible objective might be to reduce the poverty gap, and that this being the case, the relevant prism through which to pass measures of income growth is how the poor fared as a subset of the total. 4.4 This returns us to the earlier question of whether this prism is likely to change ones conclusions much from those given by taking population-wide economic growth as the objective. Dollar and Kray (2000) have recently found that growth in GDP per capita across countries is more or less unrelated to changes in measures of their inequality. This finding can be interpreted in two ways: it refutes the arguments of those who claim that growth an irrelevant measure, that it is systematically regressive because it is generated and captured by elites; but it also suggests that 85 the distribution of growth might be an interesting second statistic to examine, since it is essentially "orthogonal," that is unrelated to, the level of growth. We will treat this as an open question for Brazil, and to answer it will compare patterns in aggregate income growth with those in its distribution. The paper will then go further, by examining the causes underlying both growth and its distribution. We shall retum to possible explanations of growth and its pattem later in this section. 4.5 As a final prefatory remark, we wish to emphasize a distinction. In debates about growth, inequality, and the welfare of the poor, the latter two are sometimes confused with one another. Let us try to avoid such misunderstandings. Economic growth may or may not come with rising inequality; the premise that it did would not imply that, if consulted, the poor would opt for less growth (and hence, under the premise, less inequality). The relevant question in this light might rather be: did the income of the poor rise (and by how much)? Responding to this question, we shall mostly restrict attention to one measure in addition to overall per-capita income growth: the percentage growth in the per-capita income of the poorest 25 percentP of the national population.5 This measure has the practical advantage that it is computable from repeated cross-sections of household data. We may then proceed in a computationally straighfforward way from this measure to one of the change in inequality, without confusing the concept of poor-income growth with the related but distinct concept of inequality. Measuring Patterns of Growth in Brazilian Household Data 4.6 This paper makes use of stacked household data from the years 1982-98.5 The data come from Brazil's Annual National Household Survey (Pesquisa Nacional por Amostra de Domicilios: PNAD). The approach has two main advantages. First, PNAD has a stratified sample design that makes it representative at the state level. This is useful in itself: it is interesting to compare growth in household incomes across different states, to assess aspects of state performance. Second, using household data to compile income growth measures allows one to focus on income growth among the poor (as well as among the whole population) as discussed above, and subsequently to calculate partial correlation with factors that vary at the household or the local level. 4.7 The PNAD has some fairly well documented shortcomings (see Ferreira, Lanjouw, and Neri, 1998, for the most complete discussion of this topic). In particular, its measures of income are rather partial, since the questionnaire does pay much heed to assessing home production and non- market income (important in rural areas). However, its one overriding advantage for present 53 This definition corresponds approximately to the definHion of the poverty headcount coming from the food poverty line commonly used in Brazil (see Ferreira, Lanjouw, and Neri, 1998). In 1999 the poverty rate using this line was 23 percent. The analysis was also performed using the poorest 50 percent instead. The results were not materially changed. This is one finding of this study, albeit a secondary one. 54There are contexts, some particularly relevant for Brazil, when other measures may be more salient. For example, some authors (e.g., Sanchez and Nunez, 2000) have suggested that certain measures of income spread or inequality are most useful for predicting rates of crime and violence. Inequality may also have specific social and economic effects that work through political channels, if plutocracies comer executive power. These effects are worthy of study in themselves, and indeed are the objects of studies in the World Bank and Brazil; they will not be addresses by this study. 55 In the context of explaining historical patterns in Brazil, it would have been nice to have comparable data from the 1970s. Unfortunately, such data are not available. 86 purposes is its comparability across time. Furthermore, the pattems obtained from the PNAD are consistent with the regional accounts (see Azzoni, Menezes-Filho and Menezes, 2001), suggesting at least that the conclusions in their general form are robust to any design faults specific to the PNAD survey. 4.8 The measure of income growth we use is change in log per capita household income.' For each year that the survey was performed (1982-98 with the exception of 1994) two such measures are constructed: change in mean income for the population, and change in mean income of the poorest (as measured in the earlier year) 25 percent of the population. These two measures are then treated as the "left-hand side" variable in the regression analysis. 4.9 The implicit unit of analysis for most of the analysis is the state. There are two parts to the presentation of results, the first descriptive, the second analytical. The descriptive part presents state growth rates in four periods: 1981-85, 1985-89,1989-93, and 1993-98. These periods are characterized briefly below. For the analytical section, the unit of observation is an age cohort within a state. This allows us to make use of greater variation in important dimensions such as educational attainment, health, employment patterns, and access to infrastructure. We do not have repeated observations of the same households over time (panel data), but by aggregating within state-cohort cells, we do have repeated observations of cohort means over time (quasi-panel data: see Deaton, 1980). The analysis then uses panel estimation techniques to derive consistent estimates of partial correlations between state-cohort characteristics and the state-cohort income growth. In this sense, the analysis in the second part of the paper tries to answer the question: what parts of Brazilian state-level characteristics or policies coincided with income growth within the periods studied? Hypotheses 4.10 To frame the answer to this question, it is necessary to start from some hypotheses about what determines or constrains household income growth in Brazil and its states. There are many such hypotheses; many are not testable using the data and methods of this paper. The following are hypotheses that the paper addresses, with notes on the evidence that would support each. Education Constraints 4.11 It has been said in reference to Brazil's stellar economic growth performance during the 1960s and 1970s, labeled Brazil's economic miracle, that the real miracle was how the country grew at such rates with so little education among its population. Most other examples of sustained growth among developed economies have been preceded by or accompanied by a fast educational expansion. Brazil's low level of education among its workforce (it ranks behind most of its South American neighbors in most of the statistics) are often cited by investing companies as an '9 For changes of about 1 0 percent or less, this measure is approximately equal to the percentage change in income. Using log-differences has certain analytical advantages that make it preferable, however. Parts of the analysis were performed using percentage changes in income, without making a great difference. The main difference between the two is that log-differences deflate large (greater than ten percent, say) income growth figures. The results suggest that this distortion does not do too much harm to the main conclusions of this analysis. 87 impediment to economic activity. So it is natural to posit this as one of the constraints holding back present-day Brazil. 4.12 Were this the case, one would expect to see states and cohorts with more education among the workforce showing significantly higher levels of income growth. The hypothesis therefore has a clear counterpart in the data. Other work that investigates the role of education in Brazil's growth agenda is by Gill and Patrinos (2001). - The Investment Climate: Political and Policy Uncertainty, and Governance 4.13 Policy uncertainty has been cited by firms in many survey as the number one obstacle to investing in Brazil. It is sometimes difficult to ascertain exactly what this uncertainty consists of: exchange rates, economic uncertainty, unpredictable legal rulings, and so on may all form parts. For present purposes, we shall focus on the educational level of the public sector in a state, the frequency of ideological shifts in government, and the volatility of voting patterns as measures of uncertainty that may impede economic growth. Assuming that these are at least partial proxies for the uncertainty that upsets entrepreneurs, the hypothesis therefore has a counterpart in the data. Infrastructure Bottlenecks 4.14 Public investment in infrastructure has fallen amid the fiscal constraints of the 1980s and 1990s in Brazil, and the necessary private investment to take its place has not always materialized amid imperfect regulation and scarce credit. To what extent the extemalities associated with infrastructure provision are truly a constraint on income generation is an open question, however: Brazil's infrastructure is not worse overall than its neighbors in many categories. There is, however, great variation in infrastructure between the states (e.g., electrification, water, paved roads, and public services such as garbage collection). If infrastructure were a constraining factor, one would expect to see the states best endowed in these dimensions growing faster, ceteris paribus. Labor Rigidities 4.15 Brazil's labor code is onerous by intemational standards. Payroll taxes are high in comparison with its neighbors, legal regulations applying to all formal labor contracts are numerous, and the rulings of labor courts are slow, unpredictable, and reputed systematically to favor workers versus employers. Again, the extent to which these factors truly impede growth is unclear, but were this the case, then under certain assumptions, one would expect to find lower income growth rates in states and cohorts with higher levels of formal labor contracts. 4.16 The proportion of contracts that are formal is of course endogenously determined as a function of their associated costs and benefits, so the line of analysis suggested in the previous paragraph can only be pursued with some skepticism. However, nearly all the explicit costs of formality are federally imposed, and thus invariant across states and cohorts. Variation in the proportion of contracts that are formal must therefore be owing to other factors, some of which will be controlled for in the analysis: education and employment composition by sector, for example. The assumption needed is thus that remaining variation in the extent of formal contracts is exogenous, meaning that it is not correlated with other unobservable factors that are themselves 88 correlated with income growth. We shall return to this discussion at the end of the paper when discussing conclusions. Climatic Constraints 4.17 A long strand of economic literature posits that some countries are geographically doomed to lower growth rates owing to climatic conditions that impede economic activity. The most common evidence cited is the negative coefficient on "tropical' dummy variables in cross country growth regressions. We therefore investigate whether rainfall, latitude or altitude are associated with differences in income growth rates. Agglomeration and Spatial Factors 4.18 A related hypothesis is that poor areas remain poor owing to low aggregate demand (so- called trading extemalities, discussed theoretically in Diamond, 1982). These effects are therefore also "geographical" (see Jalan and Ravallion, 2000, for a fascinating application to China), but would show up in data as low growth rates associated with low local measures of economic activity (e.g., state GDP). We shall not enter into a discussion of the complications that arise in constructing the correct econometric estimates of these effects here: these are discussed at the end in an appendix. Access to Markets and Trade Volume 4.19 Brazil's economy, despite lowered tariffs in the 1990s, remains relatively closed. There is also variation in the distance to major markets among the states. If integration were constraining growth, we would expect to see those states furthest from markets, and those trading least, growing more slowly (ceteris paribus). In the data we use distance from the sea as a proxy for market access, and interstate trading volumes as a proxy for integration into the national economy, to attempt to test for these effects. Migration 4.20 Brazil's economic miracle included processes of urbanization, transformation of the productive base from traditional industries towards manufacturing, and migration, particulary from the Northeast to the Southeast of the country (see Gordon, 2001). Recently these migratory flows have ceased and even reversed (Fiess and Verner, 2001). It is an interesting question, though one hard to answer convincingly, to what extent this has been a causal factor in Brazil's economic slowdown. 4.21 The question is hard to answer because opportunities for income growth cause and are caused by migration, and so most attempts to assess causality are doomed to ambiguity. We shall touch upon the matter in the latter part of this paper and present some partial correlation between migration and income growth, without offering a satisfying solution to the problem of ascertaining causality. 89 Introduction 4.22 This section takes a step back from the hypotheses discussed in the previous one, and first presents the overall paftems in the data. The main issues we are interested in describing before a discussion of underlying causes are the following: * Income Growth over Time. How does income growth behave in different periods since 1982? What are the relative performances of the states and regions? * Convergence. Is there a tendency for poor states to catch up with richer ones? How fast? * Income Growth Among the Poor. To what extent does income growth accrue to the poor? How does this vary by state and region? * Changes in Inequality. What are the effects of income growth, and its accrual to the poor, on inequality? Has it risen or fallen? Is there a "tradeoff" between income and inequality? Four Periods 4.23 For convenience we divide the full period 1981-98 into four approximately five-year sub- periods: the early eighties including and following the debt crisis (1981-85); the late eighties including and following the cruzado plan (1985-89); the early nineties including and following the Collor plan (1989-93); and the late nineties including and following the realplan and stabilization (1993-98). We give a very brief description of each of these periods below. 4.24 It is probably well-known to the reader that Brazil has endured a series of macroeconomic crises over the past two decades that have created a great deal of volatility in its economic aggregates and two very deep recessions. Two remarks are apposite. First, there is no doubt that economic volatility perse has reduced economic growth. The Brazilian economy (with Latin America as a whole) stagnated throughout the 1980s, prompting the "lost decade" moniker. The underlying cause of instability was and remains fiscal. While it is true that repeated bouts of runaway inflation57 were both cause and effect of high inflationary expectations," the original inflation that led to the expectations was the result of govemment's recourse to monetizing debt. The need for this was itself the consequence of structural deficits in Brazilian public spending. Brazilian policy-makers are more than aware of this, and public policy continues to focus on fiscal imbalances as public enemy number one. For the purposes of this paper, the effects of economic 57 The PNAD data used for this paper have been deflated as carefully as possible using new regional price deflators devised by Azzoni and Menezes-Filho (2000). 5 The interaction between inflation and inflationary expectations is by now well-documented. The original paper by Barro and Gordon (1983) is perhaps still the clearest theoretical exposition. 90 volatility on growth will not be the focus: this effect is essentially national in scope, and we are here focusing in variation across states as the source of identification of economic growth effects. 4.25 Furthermore, the aim here is not to focus on the crises in themselves, but to try to identify correlates of economic growth. The periods have therefore been constructed to try and avoid their ending in a crisis. Despite this, it will be seen that there are many cases of states experiencing negative income growth in the 1980s and early 1990s, though far fewer in the later 1990s. 4.26 The early eighties period incorporates the debt crisis of early 1982, and the extremely severe recession of 1981-82 that accompanied it. Related work (Neri and Thomas, 1999)59 has documented the acute effects on household labor incomes of the macroeconomic conditions at the time: falls in real household income of 20 and 30 percent were the average. The two years from 1983 to 1985 then represented a period of relative recovery, although national economic growth rates were never sustained at levels above about 2 percent. 4.27 The period 1985-89 comprised a "heterodox" stabilization effort, the cruzado plan, which by introducing a new currency (the cruzado) eliminated inflation for a time and led to a boom in 1986- 87. This boom relied on price controls and other restrictions that proved unsustainable, however, and 1988 saw another recession, though nothing like as severe as that of 1981-82. 4.28 The period of 1989-93 followed the drafting of the new constitution of 1988, which paved the way for democratic elections in 1990. Inflation had retumed, and the Collor plan of 1991 was another failed attempt to eradicate it for good. A steep recession in 1991-92 preceded the impeachment of President Collor in 1992 and a leveling of the economy in 1992-93. 4.29 The final period covers the stabilization achieved in 1993-94, prior to the introduction of the dollar-pegged real. 1994-96 was a period of expansion, which also saw an appreciable decline in the poverty rate (World Bank, 2000). By 1998 the economy was entering a phase of stalling, owing to contagion effects from the events of 1997 in the East Asian economies. Growth continued to be positive, though at a lower pace than the early part of the real plan, until further deterioration in emerging market asset prices prompted by Russia's renegotiation of its debt forced the float of the real in January 1999, which lies outside the period studied in this paper. State Income Growth 4.30 Figure 1 shows state household-income growth (vertical axis) versus initial household per- capita income (horizontal axis) for the whole period, 1981-98, and show the high divergence between states' growth rates. Only 19 of Brazil's 27 federated units are shown on the diagrams. First, the Federal District comprising Brasflia and neighboring towns is not a state. Second, the states of Brazil's Northem region" have been excluded. These states are special cases in Brazil owing to their "frontiere characteristics. They have exhibited extremely variable growth rates, mainly 59 This work is not strictly comparable to the present analysis since it uses the monthly employment survey (PME), which is restricted to six metropolitan areas in Brazil. 60 Rond6nia, Acre, Roraima, Amapa, Para, and Amazonas. The state of Tocantins, which separated from Goias only in 1991, has been merged with Goias for the purpose of comparability across time. 91 owing to waves of immigration and exploitation of natural resources (e.g., oil in Roraima). Their inclusion would obscure more than it would reveal. Figure 4.1 Initial Income and Income Growth, 1981-98 Pi co 0) PB 0 RN 5 .5 - GO SC E RS 0 M 0 S C 0 P c 0v AL E BA 0- 3 4 5 6 initial mean log income 1981 4.31 There is some evidence of states falling into regional blocks.61' Brazil's regions are to some extent economically homogeneous clusters. The Southeast (SP, RJ, MG, ES) is richest and forms the industrial powerhouse of the country. The South (RS, SC, PR) is also relatively wealthy but more agricultural and less industrial than the Southeast. The Center West (MT, MS, GO, DF) is poorer than the first two, and relies mainly on natural resources and agriculture. Its infrastructure is far less developed than that of the South and Southeast of the country. The Northeast (PI, PB, RN, CE, MA, SE, PE, AL, BA) is much poorer. The Northeast's 28 percent of the national population account for only 13 percent of GDP. Its interior contains a large semiarid expanse, the SertAo, within which Brazil's most extreme problems of rural poverty are found. Infrastructure and social indicators are also much less advanced in the Northeast. 4.32 Where the states differ to a greater extent within regions is in their public policies. Evidence of clustering into regional growth performance is therefore evidence that initial conditions or conditions that change only slowly (levels of industrialization, some social indicators such as adult literacy) are responsible for a large part of growth differentials. In this context, differences between 61 Brazil has standardized definitions of its regions. The Southeast comprises Sao Paulo (SP), Rio de Janeiro (RJ), Minas Gerais (MG), and Espfrito Santo (ES). The South comprises Rio Grande do Sul (RS), Santa Catarina (SC), and ParanA (PR). The Center-West comprises Mato Grosso (MT), Mato Grosso do Sul (MS), Goias (GO) and the Federal District (DF). The Northeast comprises Bahia (BA), Sergipe (SE), Alagoas (AL), Pernambuco (PE), Paraiba (PB), Rio Grande do Norte (RN), Ceara (CE), Piauf (PI), and Maranhao (MA). 92 seemingly similar states within the same region are notable, although here a more analytical approach is then called for to try to disentangle the causes. 4.33 The South, Southeast, and Center-West all show less within region variation than the Northeast in Figure 1. In the Southeast, for example, the states of Rio de Janeiro and Sao Paulo are neighbors on the graph. Parana, Santa Catarina, and Rio Grande do Sul, all had about equal income in 1981 and showed similar growth rates over the period. This suggests that for these regions initial conditions and national considerations dominated state policies in determining economic growth. 4.34 The Northeast is another story. The region contains Brazil's fastest growing state (within the 19 we are discussing)-Piaui-and its slowest-Bahia. Despite starting from similar initial income, the states of the Northeast displayed widely divergent growth rates over the period. What can explain this? We will argue that at least part of the reason is state public policy. 4.35 It could be argued that the preceding discussion is a little misleading, since so much of it hinges on what happened under past state govemments, which have since changed, rendering such backward-looking exercises irrelevant. The next section remedies this by focusing on the four periods individually, and by emphasizing more recent events. Differences Across Periods 4.36 Figure 2 illustrates growth versus initial income for states in each of the four periods separately. The periods do not resemble one another graphically very much, supporting the argument of the previous paragraph. The PNAD data confirm, at the microeconomic level, the macroeconomic observation that the 1980s were a "lost decade". For 1981-85, four states record negative or zero real per-capita household income growth in the data. For 1985-89 the story is worse: the 19 states are splft about in half between those that record negative or zero per-capita income growth and those that record positive growth. And the recession of 1991-92 creates even worse effects in the third period, when only seven states show positive growth. Positive growth retums to the micro data just as it does to the macro data only after 1993. The dispersion between states is also slightly higher in 1989-93 than it is in 1993-98. 93 Figure 4.2 Initial Income and Economic Growth in Four Periods 1981485 1985-89 GO .2 SP RS RN Ms ~~~~~~~~~~~GO SC E " MS~~~~~M MA - Pi MG .2- - Pl PR BA SCE 2N O- tOtif; 8 sPP 8 MA AL M,MA RS , -.2- GO _~~~~~~~~~R SE Gu Ms c BEE SP .3- RSAL CE ES -.4- ,-,,.1 lPB Initial mean log Income 19891 Initial mean log Income 1985 1989-93 1993-98 .2 -Pi .5 PB PB coS at N~~~~~~~~~~~~~R -PB 4.37~~~C Som oftergoa lseigta a paetfrtewo epeidslssowthnub - . C E PR SC P~ ,,RN ES RJ MA~' periods. MA AL MG BA hs MGO R,I-.--_ BA ~ -.2 GOSE PR- a a~~~~~~~~~~~~~~~P. Ms SC E ~~~~~~~~~~~~~~~~~~~E RS N 3. 4 5B2 4- Initial mean log Inoome 1989 -Initial mean log Income 1993 4.37 Some of the regional clustering that was apparent for the whole period is less so within sub- periods. A closer look at the data reveals that for the bulk of the growth over the whole period is accounted for by the growth that occurs in the fourth period. It is also in this period that the regional clustering is most apparent, even for the Northeast, though this region still shows greater within- region variation in growth than the other three. Convergence A Brief Discussion with Regard to Brazilian States 4.38 A now extensive literature discusses the concept of economic convergence between countries, that is, the notion that, for reasons of technological imitation, factor mobility, and higher retums to scarce factors such as capital (physical and human), poor countries ought theoretically to be able to "catch up" with rich countries. Needless to say, the cross-country evidence is mixed, and there are many reasons why 'theoretically" may accord with inappropriate theory. 94 4.39 Nonetheless, the theories that imply convergence between countries ought to apply more readily to states within the same country. Obstacles to factor mobility and to technological imitation are fewer. And many institutional parameters are fixed at the country level and, therefore, to a first approximation are fixed across states, bringing the implicit cetenis paribus assumption behind growth comparisons a little closer to reality (see Barro and Sala-i-Martin, 1991, for an analysis of convergence applied to states and regions in the USA and Europe over the long run). 4.40 An analysis of convergence itself is not the object of this study, but Figure 2 shows quite clearly that if it has occurred at all in Brazil since 1981, this has in fact only been since 1993, or possibly, if one reads the third graph charitably, since 1989.' In the three earlier periods there is simply no visual (let alone statistically significant) relationship between initial levels of income and subsequent income growth. Since the real plan, however, there is a striking relationship: poorer states (particularly those in the Northeast) have exhibited faster income growth. This is no doubt in part because greater macroeconomic stability allows longer-term planning and investment, clearly a necessary component of the convergence story. It is also in part because the costs of high inflation had been disproportionately bom by the poor (who could not protect themselves against inflation through asset switching as effectively as richer households). The elimination of inflation in 1993-95 therefore generated relative gains for the poor. 4.41 Which of these two explanations fits the data best can be suggested by dividing the latter period into two: 1993-96 and 1996-98. The gains from the eradication of inflation were concentrated in the first of these. The gains from greater stability and long term investment only began to be felt in the second. Figure 3 shows what happened to convergence in these periods. Figure 4.3 Splitting the Fourth Period in Two 1993-96 1996-98 .5- MA P -PE .15 F. \RJ MS a 4 - SP _ CE GO PICEPE ESS 0b S ^>, GO = 0 MG SC .2- SP GO.05- RS B2M PP R PR pi ES M G GO Ms -. 2 - ~~~~~~~~~~~~~~~~~~~~~~-.35 .1 . 0 .1 .2 .3 0 4 .1 .2} mean log Income growth 1981-B5 mean log income growth 1985-89 1989-1 993 1993-98 Poverty.05 RS BA .2- m ~~~Ms GO MA CB P C o0 RN P ~- MG PR 0C E BA RJ MT0 o -..~~~~~~~ ~RSo -~~~- .~~~~~ AL - . _ PE -.05R 0 A' ~~~~~~~~~RN SCP AMA ES cr MT P SE CE P ~~~e -.1 S pi ~~~~~~~~~~~Rd -.2 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ -.15 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .A -.2 02.2..4 .5 .6 mean log income growth 1989-93 rman log income growth 1993-98 Poverty Reduction 4.49 If the poor-income growth coincides with broader income growth, and if inequality changes are (recently) unrelated to income growth, then we would expect poverty reduction to occur in states where higher income growth occurred. Figures 7 and 8 show that this is indeed the case. The three state with the highest income growth in the period, the small Northeastern states of Piauf, Paraiba and Rio Grande do Norte, also saw the largest falls in the poverty headcount rate.65 Sao Paulo and Rio de Janeiro had the lowest falls in the poverty rate, and were among the five lowest rates of economic growth during the period. 65The poverty headcount referred to here uses a food-poverty line, and is often also referred to as the extreme poverty rate (or taxa de indigencia). 99 Figure 4.7 Growth and Poverty Reduction, 1981-98 PB .3 RN 0) OD ,.2 MA .2G SE GO AL S > PR o PE MS PR BA RS RJ SP 0 0 .5 1 mean log income growth 1981-98 4.50 The relation between growth and poverty reduction has been stronger during the 1990s than it was during the 1980s. This shows itself as a steeper line of best fit through the states in the later two graphs in Figure 8. The magnitudes of these slopes are, respectively, 0.19, 0.28, 0.32, 0.30. If we interpret these slopes as (negative) elasticities of poverty to income growth, these numbers suggest that an extra one percent of income growth translates into a 0.3 percentage point fall in the poverty headcount rate. This estimate is consistent with Neri (1999), who uses a similar methodology to calculate this elasticity for Brazil." 66 Neri's point estimate is that a 1.0 percent increase in income growth leads to a 0.8 percent decrease in poverty. At a headcount poverty rate of about 25 percent, this implies a percentage point poverty decrease of 0.2. Our point estimate is therefore a little higher than Neri's, though it falls well within his 90 percent confidence interval. 100 Figure 4.8 Growth and Poverty Reduction in Four Periods 1981-85 1985-89 PS GO RN -.05 ~MA -.05 - 0 .1 0 .AL R -PR MG0. ES ~~~~~~~~~~~~~~~~~~iSP o ~~~MA ______________ .1- RS .2-R & RJ & PB M EBA ,.05 -.05 ESXLI -.41. 0 . 2 .2 .3 0 4 .5 .2 mean log Income growth 1981 -89 mean log Income growth 1985-89 1989-93 1993-98 Pi .2 PB - ~~~~~~~~~~~AL CQ 9~ .05 SBE CE~~~ C~~~i .15 ~~~BA MARN - CDteat Cihteecpino ergp, an h ddto fEsiioSnt wihbodrh c ES - P 0 ~~~~~~~0 - de~~~~~~~~~e Jaer to CEPEt) hsgou fsae a,snc 93 enbt oo-noegot n 0 GO, Ms S e a s Mn i a cE M -.-- .1 ~~~~~~~ ~~~~~~MS MKG > o PE R BA ~ ~ ~ ~ ~ ~ ~ ~ ~~~- SC SP -.1 ~~~~~~~~~~~~~~~~.05 A S -.4 -.2 0 .2 23.4 .5 .8 mean log Inomrre growth 1989-93 meoan log Inconme growth 1993-98 4.51 It is also interesting to compare the recent relationship between the income growth of the poor and poverty reduction, portrayed in Figure 9 for the fourth period. Of course one expects a close relationship between these two measures, the first of which is close in definition to a reduction in the "poverty gap" measure of poverty." The figure shows a the Brazilian states falling into three distinct groups during the period 1993-98. The first group comprises the states of the Northeast, with the exception of Sergipe, and the addition of Espirito Santo (which borders the Northeast and indeed contains areas with more in common with, say, Bahia to the north, than Rio de Janeiro to the south). This group of states has, since 1993, seen both poor-income growth and higher poverty headcount falls than the other two groups. At the other extreme, the second group comprises S&o Paulo plus the Southern states of ParanA, Santa Catarina, and Rio Grande do Sul. This group has performed worst in the period, both in terms of poor-income growth and of poverty reduction. The third group comprise Rio de Janeiro, Minas Gerais, Mato Grosso, Mato Grosso do Sul, GoiAs, plus Sergipe, and is an intermediate case in both dimensions. 67 That is, P1 as opposed to P., the poverty headcount rate. It is not quhte analogous, since income accruing to an individual after they have crossed the poverty line would be included in our measure of poor-income growth but would not technically constitute a reduction of the poverty gap. The reader may judge this to be a fairly casuistic distinction, and we would not disagree. 101 Figure 4.9 Poverty Reduction and Income Growth among the Poor, 1993-98 .2- AL co PB 0) BA M AN C') .15 - 0) o Pi 0 . / E CEPE MT .1 - SC SP .05 - RS .2 .3 .4 .5 log income growth (poor) 93-98 Descriptive Conclusions 4.52 Policy conclusions cannot be based on the descriptive methodology used so far in this paper, but we underline the following observations, which are of relevance to public policy. Although there was an apparent tradeoff in Brazilian household data between economic growth and inequality until the real plan, this tradeoff diminished or disappeared in the data after 1993. Economic growth is highly correlated with the growth of the incomes of the poor, and thus to poverty reduction. * However, income growth has not accrued proportionately to the poor. For one percent of income growth over the whole period (1981-98), the poor's income rose by a little over half a percent. This ratio has risen to about 0.9 since the real plan, but some of this improvement may not be sustained fully beyond the period of inflation eradication. * For the period 1993-98 we estimate an income elasticity of poverty a little greater than unity. The point estimate is about 1.2, although again because of the strong poverty reduction of 1994-95, the present value of this parameter may be a little smaller than this point estimate. * Almost all the significant growth in real household incomes in the period has occurred since 1993. Economic stability is clearly a sine qua non for improved household welfare, whether of the poor or of the population at large. 102 * Over the same period, there has been unconditional convergence in states' incomes. Moreover, there are strongly apparent "regional fixed effects" in growth suggesting the importance of initial conditions, economic structure, and also national policies in determining growth, seemingly dominating state policy parameters in their effects. * The preceding point should be qualified by the case of the Northeast region, whose states have shown a wider divergence of performances over the whole period and over recent years. One possible explanation may be the dependence in certain cases on a small number of crops, which adds volatility to the Northeastern states' economies, although there is no doubt much else going on in the data. Accepting this partial explanation might suggest these states' taking specific measures to insure against these risks. Introduction 4.53 We now tum to more structural explanations of the above patterns. For this section, as mentioned in the early part of the paper, we use more of the richness of the data than the descriptive section above. First, we pool growth observations for each year (1981 -97, since we are using first differences), state (of which we use 19), and age cohort (of which we define 10). 4.54 The aim of this section is to establish which state- and cohort-level variables are correlated with economic growth, pro-poor economic growth and, at a later stage, the difference between the two, which corresponds with our difference-in-differences equality measure described above. 4.55 We attempt to be as exhaustive as the data permit in including exogenous right-hand side variables in the analysis, before tuming to a discussion of some possible policy-relevant interpretations of the results. Some variables were constructed from the PNAD while others were imported from other datasets. Table 1 gives a summary of the main variables used. Table 2 describes the evolution through the period of certain of them. 4.56 For education, the average number of completed years of education was used as the simplest "summary statistic." At a second stage of analysis, the proportion of the population that had completed some primary, upper primary (middle school, or grades 5-8)61, secondary, and tertiary (college or university) was also included. For health status, infant mortality per thousand births, and life expectancy at birth were available at the state level, though not at the cohort level.69 4.57 Several variables capturing the effects of infrastructure provision were included: proportion of households served by running water, sewerage, garbage collection, and electrification. In 6 Called segundo ciclo do ensino fundamental in Brazil. 69 These variables are not discussed in the analysis since they did not appear significantly in any estimation. 103 practice, these variables are highly collinear, and the results reported later restrict to trash collection and electrification, the variables that showed the most marked effects. 4.58 Three variables were included to capture the influence exerted by government on economic activity and income generation. The first is the average education level of public sector workers in the state: this variable aims to proxy the quality of public governance. It may not be a very accurate measure of this quality, but we include it in all specifications. The other two variables aim to measure political instability at the state level. The first of these two is the sum across political parties of the squares of their changes in vote shares in the previous two state gubernatorial elections. The second simply takes the value one if the most recent state gubernatorial election delivered a change of political coalition in power. The former variable aims to capture political uncertainty, while the latter aims to measure policy uncertainty. 4.59 We include variables describing the composition of economic activity and the nature of labor markets. For the former, the proportion of households with members working respectively in industry, services and agriculture was calculated. For the latter, the proportion of household heads working in the formal sector70, and the proportion of households in urban areas, were calculated. 4.60 Next variables were added to the data to capture the effects of trade (inter-state), migration, climate and geographical location (fixed at the state level). For trade, the measure used was the. logarithm of the combined volume of exports plus imports. This datum was available only for 1985 and 1995: the 1985 value was used for observations in the 1980s, the 1995 for the 1990s. For migration, the proportion of inhabitants who had migrated from outside the state was available for the years 1989 and 1997. The 1989 value was sued for the 1980s, the 1997 value for the 1990s. This creates a clear problem of endogeneity of this variable in income determination, since economic growth would be expected to cause migration: this variable is excluded from most regressions in this paper and is discussed separately. Average temperature, altitude, distance from the sea, and latitude were included where the specification allowed for identification of state-fixed effects (see discussion below). 4.61 All the results we report in this section come from most efficient and consistent estimation technique available.7' Similar results were generated from simpler but less econometrically defensible estimation techniques such as ordinary least squares with state-cohort fixed effects, although where there was divergence in results we weaken our conclusions accordingly. 70 The definition of the formal sector in Brazil is standard, and is taken to mean that the worker has an 'employment booklet that is signed by his or her employer (carteira assinada). The booklet implies that the employment relationship is registered at the labor ministry and that the employee is therefore eligible for a certain set of employment-related benefits such as social security, severance pay, and unemployment insurance. 71 All reported estimates were estimated using first-difference or quasi-first difference generalized method of moments (GMM) estimation. In first difference specifications, variables that are not strictly exogenous are instrumented with lagged levels of right-hand side variables, following the methodology proposed by Arellano and Bond (1991). 104 Table 4.1 Summary of Variables Used Standard Name Description Mean Deviation Dependent Variables NB. "cellr refers to by-state by-cohort groupings. Delta Difference in mean log income in cell over the year 0.0317 0.231 Delta25 Difference in mean log income of the poorest quartile 0.0215 0.231 Quality Difference in differences: Delta25-Delta -0.0119 0.159 Weakly Exogenous Variables Educ Average completed years of schooling of household head 3.51 1.57 Primary Fraction of cell with completed primary education 0.691 0.189 Middle School Fraction of cell with completed lower secondary education 0.154 0.125 High School Fraction of cell with completed high school education 0.103 0.076 College Fraction of cell that attended college 0.052 0.042 Public Admin Average years of education of state's public employees 7.14 2.62 Political Sum of squares of changes in parties' vote shares 0.411 0.287 Policy Indicator of change in political leaning of govemor 0.529 0.499 Elecric Percentage of electrification in the state 0.773 0.169 Trash Penetration rate of garbage collection in the state 0.500 0.200 Industry Fraction of households in cell with industry employee 0.190 0.108 Agriculture Fraction of households in cell with agricultural employee 0.326 0.136 Services Fraction of households in cell with services employee 0.366 0.139 Urban Fraction of cell that lives in an urban area 0.659 0.144 Formnal Fraction of households where head formally employed 0.216 0.150 Rain Annual rainfall (cm) 103.1 24.6 Latitude Average Latitude of the State (degrees) 13.34 7.10 Altitude Average Altitude of the State (m) 328.6 164.9 Distance Average distance from the sea (km) 226.1 204.2 Migration Fraction of popualation that immigrated from out of state 0.146 0.109 Table 4.2 Evolution of Selected Variables i -Ji *I,' .* . . Log Income 7 _ 4.41 4.48 4.52 4.50 4.91 Log Poor Income 3.51 3.56 3.48 3.44 3.84 Education 3.22 3.36 3.47 3.66 3.90 Primary 0.779 0.738 0.691 0.658 0.583 Middle School 0.106 0.125 0.152 0.176 0.212 High School 0.0640 0.0910 0.105 0.114 0.142 Tertiary 0.0506 0.0454 0.0524 0.0525 0.063 Public Admin 6.99 6.62 6.91 7.15 8.06 Elecric 0.620 0.694 0.772 0.840 0.900 Trash 0.336 0.423 0.477 0.585 0.673 Industry 0.258 0.202 0.196 0.160 0.128 Agricufture 0.354 0.379 0.332 0.300 0.268 Services 0.376 0.403 0.428 0.309 0.287 Urban 0.615 0.627 0.640 0.703 0.709 Formal 0.294 0.273 0.258 0.161 0.117 105 Average Income Growth First Difference Estimates Controllingfor State-Fixed Effects 4.62 The first column of Table 3 reports coefficients from the regression of observations on overall income growth, pooled over the whole period, on the right-hand side variables. The corresponding OLS specification in first-differences explained 67 percent of the variation in income growth, a high proportion for quasi-panel data. The significant effects are from average years of education, the political and policy variables, the -penetration rate of garbage collection, and the level of employment in industry and services (that is, in relative terms, agriculture fared poorly). 4.63 Many studies find positive relationships between the level of education and income (whether analyzing households, countries, or other levels of aggregation): all of the literature on returns to education falls within this category, for instance. Yet, in the instance of our education coefficient, this is not what we are reporting, but rather a significant positive relationship between the initial level of education (of an age cohort in a state) and the subsequent rate of growth of income (of that age-cohort in that state). It is far less obvious that we should expect to find such a robust effect in the Brazilian data (the coefficient is also large: approximately 0.1, implying an extra year of average education completed increases subsequent income growth by ten percent). 4.64 Similarly, it is worthy of note that political and policy uncertainty both have the expected signs and are significant. States with uncertain political outcomes, or where the governorship moved between different points on the ideological spectrum, grew more slowly. 4.65' The effect of the variable trash is interesting. This variable is a proxy for public provision of infrastructure. The interpretation of this effect is of a direct effect of the level of infrastructure on productivity growth. This effect could also be viewed as an indicator of geographical effects: the same people have higher income growth from living in a richer location. Finally, the effects on growth of the proportion of household members working in industry and services are not robust across time-periods or across specifications, so we will not dwell on these at this point. 4.66 The second column of Table 3 reports the coefficients from the same estimation restricting to the later two sub-periods described earlier, covering 1989-98.n The education effect is almost identical, while the infrastructure effect (from trash) is still significant though smaller. There is also now a significant effect from the other infrastructure variable, the level of electrification of the state. Political uncertainty increases its negative effect, while policy uncertainty does not appear significant. There is now a higher significant positive effect from the proportion of employment in services, as opposed to industry or agriculture.73 This difference in the behavior of the sector variables across the two periods (1981-98 versus 1989-98) may reflect the true underlying differences in productivity growth across time. It has been reported in other countries that much 72 It would be desirable from the point of view of contemporary relevance to restrict even further to, say, the period 1993-98, but this is not feasible econometrically since the methodology being used relies on having enough variation in the regressors across time to identify effects. 73 These differences were not robust across specifications (OLS with state-fixed effects, OLS in first differences). 106 recent productivity growth has occurred in services relative to manufacturing industry. Finally, there is a significant positive effect from the proportion of workers in the formal sector.74 Table 4.3 Coefficients (absolute Z-values 7' from Income Growth Equations Period 1981-98 1989-98 1981.98 1989-98 Dependant Variable Delta Delta Delta25 Delta25 Coefficients Educ 0.103* 0.104* -0.0076 -0.024 (7.20) (6.38) (0.51) (1.29) Public Admin 0.0068** 0.0034 -0.0002 -0.0020 (3.37) (1.36) (0.10) (0.82) Political -0.044** -0.055* -0.042** -0.051 ** (2.27) (2.93) (2.23) (2.63) Policy -0.013** -0.0017 -0.0062 -0.0002 (2.61) (0.16) (1.09) (0.02) Electric 0.126 0.343** 0.095 0.402* (1.44) (3.12) (1.10) (3.68) Trash 0.361 ** 0.605* 0.194* 0.264* (4.04) (4.79) (1.71) (1.73) Industry 0.177* 0.361 ** 0.106 0.183 (1.95) (1.98) (1.09) (0.93) Agricufture 0.152 0.334** -0.021 0.092 (1.38) (2.37) (0.22) (0.63) Services 0.209** 0.591 ** -0.080 0.074 (1.97) (4.00) (0.75) (0.51) Urban -0.038 -0.324** -0.071 -0.223 (0.32) (2.06) (0.59) (1.39) Formal -0.083 0.085 -0.156 -0.144 (0.74) (0.54) (1.28) (0.79) Lagged depvar 0.334** 0.327** 0.1 53** 0.233** (12.76) (9.31) (4.19) (5.32) R squared from OLS 0.67 0.52 0.63 0.43 Number of obs. 2890 1853 2890 1853 Number of groups 190 190 1 90 190 Mean no. periods 15.2 9.8 15.2 9.8 74This effect was absent from the OLS specifications and also from the longer time period. 75 All stafistics are calculated from robust standard errors. 107 Income Growth among the Poor First Difference Estimation Controlling for State-Fixed Effects 4.67 Column IlIl of Table 3 reports coefficient estimates using the sample from the whole period restricting to the poor.76 The results are strikingly different from column 1. 4.68 First, education does not appear to be significantly correlated with income growth among the poor. We repeat here that education is undoubtedly correlated with income for this sample of households, but not necessarily with its subsequent rate of growth. Political uncertainty, however, appears wfth almost an identical negative coefficient to that in column 1. Next, the infrastructure effects are attenuated, though similar in pattem to column 1. Finally, there are no marked effects from the sector-composition of employment. All in all, for the period 1981-98, income growth among the poor seems harder to explain than it is among the population at large. 4.69 Column IV reports the same results for the period 1989-98. The education effect is again absent, while the effects of political uncertainty are heightened. The infrastructure variables seem to swap roles relative to column 11, with electrification (electric) highly significant, trash less so. There is no significant effect among the sector variables, although the highest point estimate is on the industrial employment variable. 4.70 Taken together, these results suggest that there is far less systematic correlation between state-cohort characteristics and economic growth among the poor than there is in the economy as a whole. This is in itself an interesting finding. One may now ask is whether the differences between the coefficients for the two groups are significant. Since the difference between income growth among the poor and among the whole population is precisely our measure of the equality of growth from the previous section, this amounts to regressing our equality measure on the same variables. The "Equality of Growth" and Possible Tradeoffs 4.71 This brings us back to the notion discussed earlier, and common in economic debates since Kuznets, that there may be tradeoffs between the pursuit of economic growth as an end in itself, and the distribution, or equality, of that growth. To the extent that individual variables appear with a significantly smaller coefficient in poor income growth regressions than in general income growth regressions, these variables may be thought of as micro-level incidences of such tradeoffs. 4.72 The first column of Table 4 shows the results from regressing growth equality on the RHS variables. As expected given the results so far, educational attainment appears with a highly significant negative coefficient. For the period 1981-98, improvements in education across the Brazilian states seem to have benefited the population as a whole more than they have benefited the poor. On the other hand, industrial employment appears with a significantly better" positive coefficient than employment in either agriculture or services. Over the whole period industrial 76 The definition of poor in this section for the purposes of calculating income growth is the same as that used in the previous section, namely the poorest quartile of individuals in the data. 108 employment was significantly correlated with income growth (Table 3), and moreover its differential effects relative to agriculture and services have been greater among the poor. There is also a small but significant effect of policy stability on the income distribution. Finally, the quality of public administration seems also to have had a slightly inequitable incidence on growth, though the point estimate of this effect is very small. 4.73 The corresponding results restricting to the later period of 1989-98 are reported in Table 4, column 11. The results are similar, suggesting that the effects reported in the last paragraph are not solely owing to the 1980s period in the data. In 1989-98, education again appears to have greater growth effects on average than among the poor. The main differences between the 1990s and the whole period lie in the sector and infrastructure effects. Services now appear with a significantly larger negative coefficient, that is, it appears that services income growth has not benefited the poor relative to the whole population. Finally the variable trash, proxying for local infrastructure, now appears with a significant negative coefficient. Table 4.4 Coefficients (absolute Z-values) from Equality Equations Period 1981-98 1989-98 1981-98 1989-98 Dependant Variable Equality Equality Relative Equality Relative Equality Coefficients Educ 0. 114t* l 0. 135* -0.113* 0. 144** (7.93) (7.44) (6.00) (6.05) Public Admin -0.0052** -0.0040 -0.0044* -0.0051 (2.62) (1.59) (1.86) (1.60) Political 0.0135 0.0136 -0.037 -0.0087 (0.65) (0.65) (1.01) (0.24) Policy 0.01 17** 0.0059 -0.015** -0.0008 (2.21) (0.48) (2.00) (0.05) Electric -0.151 -0.131 -0.129 0.052 (1.62) (1.15) (1.08) (0.29) Trash -0.124 -0.281 ** -0.271 '0.590* (1.21) (2.17) (1.85) (2.74) Industry -0.118 -0.358* -0.145 -0.312 (1.31) (1.72) (1.05) (1.23) Agriculture -0.299** -0.469** -0.667** -0.887** (3.46) (3.35) (4.25) (3.44) Services -0.313** -0.559** -0.431 ** -0.608* (2.78) (3.71) (3.09) (3.04) Urban -0.068 0.093 -0.309* -0.110 (0.50) (0.46) (.65) (0.42) Formal -0.066 -0.059 -0.063 -0.129 (0.58) (0.34) (0.38) (0.52) Lagged depvar 0.170** 0.168** 0.139** 0.214* (4.60) (3.93) (3.13) (3.93) R squared from OLS 0.28 0.26 1 0.24 0.23 Number of obs. 2707 1685 2707 1685 Number of groups 171 171 171 171 Mean no. periods 15.8 9.9 15.8 9.9 109 The "Relative Poor" 4.74 Columns ll and IV of Table 4 explore a further concept of income distribution: that of the equality of growth interpreted as its incidence among the poorest 25 percent within each state. This changes the emphasis from the income distribution at a national level to the distribution at the level of the individual states. The person at the 25 percent point of the income distribution in Sao Paulo, for example, is above the extreme poverty line used for this and other studies.77 On the other hand, much more than 25 percent of the population of, say, Ceara fall below this poverty line (indeed more than half do). As a convenient shorthand for this state-level concept of income distribution, we shall label people falling within the poorest quartile of their state as the "relative poor" as opposed to the "absolute poore whom we define nationally. It is worthwhile to check whether the same broad results hold if one interprets poverty as "relative" rather than "absolute" in this sense. 4.75 The answer seems to be that the results do not change very much, with two exceptions. As column IlIl (Table 4) shows, for the whole period 1981-98, the policy uncertainty variable now appears with a significant negative coefficient. Also, relative to column 1, agricultural employment was significantly less pro-poor (in our newly relative terms) than was industry. 4.76 For the period 1989-98 the picture is not dissimilar, even if the coefficients alter somewhat. Education and infrastructure (through garbage collection as opposed to electrification) again appear with negative coefficients. Agricultural employment again seems to be the precursor of inequitable growth, with an even higher negative coefficient than for the whole period. Distinguishing between Levels of Education 4.77 Since the results regarding education of the analysis so far are robust and also important from a policy perspective, we now proceed to assess differences by education level. Table 5 reports coefficients from equations that included measures of the development of different levels of the education system (primary, lower secondary, high school, and college). Since we are mainly interested in the differential effects of the various levels of education, the coefficients we report are the differences between the effects of secondary, high school, and college attendance measures and those related to no education or just some primary school attendance. Moreover, for digestibility we restrict to the period 1989-98 and omit mention of all the other coefficients. The results for the other coefficients are anyway very similar to those reported in Tables 3 and 4, and we wish to concentrate the remainder of this section on the analysis pertaining to education. 4.78 Column I reports the results of the general income equation estimation, while column 11 restricts to the (absolute) poor. Column IlIl then reports the point estimates of the effects of the different levels of education on our measure of equality. Column IV (analogous to columns IlIl and IV of Table 4) investigates the notion of relative equality defined above. 77The poverty line is defined in absolute terms at R$65 (measured in Sao Paulo in 1997), and adjusted for regional price variations (Azzoni and Menezes-Filho, 2000). Since the poverty rate measured this way varies a great deal across states, there is a significant difference between the concepts of absolute and relative poverty as defined in the text. 110 4.79 Where the results are significant, they are what one might expect. College enrollments, despite their positive (though statistically insignificant) effect on income growth, increase subsequent income inequality in these data. This may not seem surprising, but taking the data at face value nonetheless reinforces the message that policy makers be aware of the distributive consequences of investments in tertiary education. 4.80 Beyond this, the results of splitting education by levels are quite clear. Upper primary education enrollments correlate with subsequent improvements in the income distribution (the coefficients are relative to the omitted category of primary education enrollment). Secondary and tertiary enrollments correlate with'subsequent deterioration in the income distribution, whether measured according to a national or a state-relative definition. There is a general pattern from positive distributive impacts towards negative as one moves up through the levels, as one would expect. Table 4.5 Analyzing Levels of Education Period 1989-98 1989-98 1989-98 1989-98 Dependant Variable Delta Delta25 Quality "Relative" Equality Coefficients Secondary -0.018 -0.031 0.241 0.219 (0.46) (0.68) (1.97) (1.39) High school -0.023 -0.0059 -0.391 ** -0.291 ** (0.36) (0.08) (3.49) (2.39) College 0.160 -0.143 -0.396* -0.278 (0.80) (0.83) (1.68) (1.11) ... other coefficients not reported R squared (OLS-FD) 0.47 0.42 0.18 0.17 Number of obs. 1845 1845 1685 1685 Number of groups 190 190 171 171 Mean no. periods 9.7 9.8 9.9 9.9 Location and International Trade Augmenting the Model with State-Fixed Variables: Quasi-First Differences 4.81 This section addresses three more of the hypotheses enumerated at the beginning of the paper. These concern trade, migration, and geographical location (climate, latitude, and altitude). 4.82 Since these variables are fixed in time, or are so for practical purposes because they are only available for certain years in our data, the estimation strategy of the previous section, which relies on taking first differences of the data in consecutive years to control for all state-fixed effects (which simultaneously rules out identifying them), is inapplicable. However, following the approach of Holtz-Eakin, Newey and Rosen (1988), a workable alternative is at hand. The main element is that by allowing state-fixed effects to be non-stationary (time-variant), the model is generalized to encompass the state-fixed effects model as a testable restricted form. This more general specification is known as Quasi-First Difference estimation (QFD). If the restricted stationary fixed 111 effects form is rejected (as it is in our data), this allows the effects of other time-invariant variables (such as geographical location) to be identified by the estimation. 4.83 Table 6 gives the coefficients from a QFD specification including location and trade variables. Rather than repeat the list of all coefficients, we focus here on those variables that this specification is designed to identify. Column I shows the estimation of average income, column 11 average income among the poor, column Ill the difference-in-differences definition of the uequality" of growth. The period of estimation is 1984-98 owing to the use of lags of the variables as instruments in the estimation. Columns IV, V, and VI show the equivalent results restricting to the 1990s (1989-98). Table 4.6 Coefricients (absolute t-statistics) from QFD Growth Equations Period 1984-98 1984-98 1984-98 1989-98 1989-98 1989-98 Dep. Variable Delta Delta25 Equality Delta Delta25 Equality Coefficients Temp 0.116 0.0157 -0.0142 -0.0048 -0.0231 -0.0470 (0.66) (0.94) (085 (0.16) (0.86) (1.60) Rain -0.0017 0.0018 0.0080 0.0545* 0.0645* 0.0074 (0.10) (0.10) (0.48) (1.66) (1.71) (0.17) Latitude 0.0133 0.0006 -0.0043 0.038 0.0455 0.0044 (0.83) (0.03) (0.31) (1.31) (1.57) (0.15) Altitude 0.0005 0.0010 -0.0073 0.0057 -0.0033 -0.0165* (0.10) (0.21) (1.39) (0.74) (0.48) (1.82) Distance -0.0072* -0.0016 0.0103** -0.0187** 0.0001 0.0281** from the sea (1.77) (0.41) (2.61) (2.96) (0.01) (3.53) Number of 2550 2550 2550 1695 1695 1695 obs.__ _ _ _ _ _ Number of 171 171 171 171 171 171 groups Mean no. 14.9 14.9 14.9 9.9 9.9 9.9 periods 4.84 There is evidence of trade and climate effects in income growth overall, though not among the poor. Most noticeable is the negative effect of distance from the sea on income growth, an effect that became more marked in the 1990s. We interpret this effect as possibly one of intemational trade, and it suggests that the importance of trade for income growth has increased for Brazil, a result that is intuitive given the lowering of tariff barriers in the early 1990s. However, the incomes of the poor do not show sensitivity to trade access, giving some succor to the view that opening to trade has not helped the poor. The result is confirmed in the inequality regressions (columns Ill and VI): the positive coefficient on distance in our interpretation implies that access to markets has increased inequality in Brazilian household income, with the effect larger and more significant in the 1990s. These conclusions are highly tentative and will form the focus of further investigation, since distance from the sea in Brazil may be capturing factors other than access to foreign markets, given the proximity of all the main metropolitan centers to the coast. 4.85 There is a weak effect of rainfall on income in the 1990s (column IV), in the direction one might expect (higher rainfall raises income growth). This effect is approximately equal for poor and 112 non-poor alike (columns IV and V), reflected in the absence of any relation between rainfall and inequality (column VI). Tuming this around, this suggests that low rainfall hurts the poor proportionately (or even a little more than the average if one uses the point estimates, although the difference between them is not at all statistically significant), justifying public interventions in times of drought on the equity grounds. We would not make too much of this result: it probably did not require sophisticated econometrics to draw this conclusion. Migration 4.86 Estimates of the effect of migration in these data suffer from two problems. First, migration figures have only been calculated for two years of the data, so much of the time variation in this variable remains unobserved. Second, as a consequence, it is not possible to structure a convincing specification that sorts out the particularly severe problems of endogeneity that this variable poses. Migration is clearly a choice that is made in response to incentives, part of which are the relative expected economic opportunities offered by point of departure and destination. This induces correlation between future economic activity and inward migration, even in the absence of a causal effect running from migration to subsequent growth. 4.87 We nevertheless ran one specification of the regressions including the migration variable, and found that migration into a state is not correlated with subsequent growth in average household per capita income in the period. A Return to the Hypotheses 4.88 Let us now return to the policy hypotheses we listed at the beginning of this paper. How strongly have they been supported by the empirical analysis presented in the last section? Part of the aim of the present study is to attempt to offer some priorities in terms of public action, so we impose our own interpretation on the extent to which the evidence on each question is convincing in this context. Education Constraints: Strong Evidence 4.89 Education is playing a large role in the story of present-day Brazilian growth. The pattern of growth in the 1960s and 1970s, based upon industrialization and urbanization, may have allowed for growth without education, but all the evidence is that those days are gone. 4.90 The strongest effect found in the data is for upper primary education. Primary enrollment is now nearly universal in Brazil, so the strongest implication of the education results is for states to concentrate on keeping primary school children in school longer. This will probably entail increasing the quality of primary education while expanding access to upper primary and secondary. Importantly, the analysis of poor-income growth suggests that this will be one of the only ways of increasing the pro-poor pattern of growth and reducing the growth-inequality tradeoff that has been apparent in Brazil for most of the last two decades. 113 The Investment Climate: Strong Evidence 4.91 The coefficient on policy uncertainty has been robustly significant, as has that on the educational level of public employees. The policy conclusions to draw from such evidence are a little less obvious than in the case of education. It would be unhelpful to suggest that policies never be changed. And there are obviously limits to the educational levels one would expect to find at certain levels of local govemment. 4.92 Nonetheless, the evidence accords so strongly with what investors and entrepreneurs actually say, that it is hard to avoid the conclusion that uncertainty and public govemance are serious issues. Other recent work has examined issues of public governance in relation to private sector activity, and here we defer to these authors for a detailed assessment of Brazil's policies (see Soh, 2001, and Foreign Investment Advisory Service, 2001). Infrastructure Bottlenecks: Strong Evidence 4.93 Our data are not ideally suited to identifying infrastructure growth effects, since the measures of infrastructure we might expect to have the strongest growth effects-roads, ports, agricultural irrigation-are absent. Nonetheless, electrification does seem to have a strong impact on subsequent household income growth, an impact that is at least larger on the income of the poor than on the whole population. Trash collection also has a strong positive impact, though in the case of this variable the effect is lower on poor-income growth than on average. Labor Rigidities: No Evidence 4.94 The interpretation of the coefficients on our formalvariable is anyway problematic, as we discussed in the earlier section on hypotheses, but since this variable does not seem to play a significant role in determining subsequent income growth, we shall not dwell on the matter. Brazil's labor code is the subject of much debate, and will probably also be modified in the years to come. It may affect Brazil at a national level. But variation between income growth rates across states do not reveal any effect at the state level. Climatic Constraints: Weak Evidence 4.95 A long strand of economic IHerature posits that some countries are geographically doomed to lower growth rates owing to climatic conditions that impede economic activity. The most common evidence cited is the negative coefficient on "tropical" dummy variables in cross country growth regressions. We therefore investigate whether rainfall, latitude or altitude are associated with differences in income growth rates. 4.96 There is also evidence of the effect of rainfall on economic growth. This is not surprising: rainfall affects the Brazilian economy through its large agricultural sector (and more recently through the availability of hydroelectric energy, although this effect was probably absent in the period covered by our analysis). Rainfall effects are equally or even slightly more strongly felt among the poor than on average. 114 Domestic Interstate Trade: No Evidence 4.97 The volume of trade with other Brazilian states showed no relation with subsequent income growth in our data. Since barriers to trade between states are not great, this would not seem to be a concem for policy makers. International Trade and Access to Markets: Weak Evidence 4.98 Brazil's economy, despite lowered tariffs in the 1990s, remains relatively closed. There is also variation in the distance to major markets among the states. If integration were constraining growth, we would expect to see those states furthest from markets, and those trading least, growing more slowly (ceteris paribus). In the data we use distance from the sea as a proxy for market access to attempt to test for the effects of trade.78 4.99 Distance from the sea is significantly associated with lower economic growth on average, though not among the poor. Thus, if one accepts the plausibility of distance from the sea as a proxy for market access, one is left with the plausible if worrying conclusion that trade promotes economic growth in Brazil, but that the poor tend not to benefit from the growth so generated. This would suggest that arguments favoring trade liberalization are right, as are concerns about protecting those whose economic activities suffer rather than benefit from increased openness, paying attention to social safety nets, etc. 4.100 These tests are clearly not perfect, however, and we would be unwilling to emphasize policy conclusions based only on one regression result. There is, however, a wealth of other evidence on this question both worldwide and for Brazil, with perhaps the most convincing to be found in Moreira and Correa (1998). Cohen (2001) also suggests that increased trade openness in the absence of flexible labor markets increases inequality. Agglomeration and Spatial Factors: Weak Evidence 4.101 A principal finding of our research is of slow but significant unconditional convergence between Brazilian states in the late 1990s. Our first-order results therefore do not support the notion that spatial poverty traps are an inescapable fact in Brazil. 4.102 Within states, there is of course considerable variation in local conditions, and here our results, discussed above, for infrastructure variables may be seen as evidence of spatial poverty traps (in the spirit of Jalan and Ravallion, 2000). However, we prefer to interpret these results as supporting infrastructure provision per se, rather than emphasizing the spatial interpretation of these coefficients. 4.103 Finally, in the 1990s there has been a negative relationship between the level of urbanization (the number of households classified in the PNAD as in urban areas) and the rate of economic growth. Combined with empirical evidence that migration to Brazil's largest metropolitan poles is subsiding (Fiess and Verner, 2001), this evidence accords with the notion that urban 78 We also experimented with interstate trading volumes as a proxy for integration into the national economy, but there was no evidence that this played a significant role in growth. 115 problems-considerations of security and congestion in particular-are having a significant economic impact in today's Brazil. This, combined with the findings in favor of convergence (albeit not rapid), constitutes the opposite of the "agglomeration" result-higher economic growth in cities owing to network extemalities-that we were expecting. Sector Composition: No Evidence 4.104 Industrial growth has seemed to favor poor-income growth slightly more than growth in services and agriculture, although the effects are not very robust to the estimation specification used. Otherwise there were not systematic differences between growth rates associated with industry, services, or agriculture. Migration: No Evidence 4.105 Brazil's economic miracle was founded upon urbanization, the transformation of the productive base from traditional industries towards manufacturing, and migration, particularly from the Northeast to the Southeast of the country (see Gordon, 2001). Recently these migratory flows have diminished and in some cases reversed (Fiess and Vemer, 2001). It is an interesting question, though hard to answer convincingly, to what extent this has been a causal factor in Brazil's economic slowdown. 4.106 The question is hard to answer because opportunities for income growth cause and are caused by migration, and so most attempts to assess causality are doomed to ambiguity. We analyzed partial correlation between migration and income growth, without addressing causality. But the results show no correlation between migration into a state within a cohort and subsequent income growth for that cohort in the state. In the absence of such correlation, the question of causality is moot. The result casts doubt on any public policy that places too heavy a reliance on migration to alleviate the problems of the poor. Main Messages 4.107 Policy recommendations should only follow from a greater body of evidence, of which the present study represents a small part. What follows can therefore be but suggestive. 4.108 Convergence has only occurred to an appreciable extent since stabilization. It is already fairly well accepted in Brazil that the eradication of inflation held great benefits for the poor, and this is a large part of the explanation for the stronger convergence during the years 1993-96 in the state data. The stable environment that followed has also led to steadier economic growth, and the data in all periods show that there is a clear correlation between income growth widely defined and the income growth of the poor. 4.109 Income growth among the poor is harder to explain than among the population at large. Whereas education, infrastructure, and political variables (in initial levels) are correlated with subsequent general income growth at the state-cohort level, the relationships are much harder to detect for the poor. This is not related to sample size, since the point estimates of the coefficients 116 are lower in magnitude in the latter case, and averaging within cells implies that the two sets of regressions are run on sets of observations of very similar size. 4.110 The differences in coefficients in tum suggest that there may be policy tradeoffs between growth and equality. The tradeoff that stands out most in our data is that associated with investments in education. The education variable used for most of the analysis (average completed years) does not necessarily reflect the success or failure of public education policies to increase the human capital of the poor, since it may be responding more over time to changes in levels of schooling in the middle income brackets, at the national level. But the results suggest the need for vigilance to ensure that public efforts in education do not simply benefit the upper three-quarters of the income distribution. In particular, they suggest that making sure that the poor stay in school beyond the first few years of primary school is central for growth, the reduction of inequality, and the reduction of poverty. Other investments in education have so far had more inequitable effects. 4.111 Infrastructure investments are also important determinants of growth, and the two we have focused on in differ in their effects on the poor. Electrification has been a gradual process in rural areas, but its marginal effects appear to be as strong for the poor as for the average household. Trash collection also has positive growth effects, although these are less equitable. There is evidence that this source of inequity has become more acute in the 1990s, as growth has picked up. Certainly the conclusion should not be that Brazil lessen its commitment to extend basic infrastructure to its population. But again, vigilance may pay off in setting priorities if the most benign distributive effects are desired. 4.112 It bears repeating that some of the conclusions from the PNAD about economic growth over the period 1981-98 are extremely robust. The general level of education appears to have had a very strong relationship with subsequent economic growth. This relationship appears with the same broad magnitude regardless of the estimation technique or the period. Moreover, variables reflecting political and policy uncertainty or variability detract significantly from growth at the state level. This effect echoes arguments that are often made in policy discussions, but as far as we know there are not many such strong empirical confirmations of the effects. Flgurs 4.10 Representation of Effects of RHS Vaviables (Response of Income Growth to a One Sandard Deviation Improvement In the Variable) I __ __. Educ Trnnh Electric Urban PolttIcd Oiotnnco Ran 0 average 0 poor 117 Closing Remarks 4.113 The approach used here is new in several ways. First, it isolates the effects of economic activity on labor income, the most important income source for the average citizen, rather than national income aggregates, which may correspond less closely with the welfare of individuals. Second, we have presented a deeper analysis of the distributional dimension of the growth process in Brazil than has been undertaken until now. This approach may be tailored in future work to address, for example, the concerns of individual states about their economic performance. It turns out that the income dynamics of the poor, though similar in the aggregate to those of the whole population, show distinct microeconomic effects (see Figure 10). This suggests that, while to a first approximation growth is good for all, there are tradeoffs in policy composition between growth and how lfair" that growth is perceived to be. 4.114 This said, these results should inform rather than form public policy in the absence of more detailed assessments of policies and experiences. The PNAD data have too many shortcomings, and econometric results do not imply policy recipes. Nonetheless, we believe that the analysis of the pro-poor incidence of income. growth is an important step. References Arellano M. and S. Bond (1991) "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations," Review of Economic Studies, Vol. 58, No. 2, pp. 277-297. Azzoni, C., N. Menezes-Filho and T. Menezes (2001) "Convergence with micro data," University of Sao Paulo, mimeo. Azzoni, C., N. Menezes-Filho, T. Menezes and R. Silveira-Neto (2001) "Geography, Public Investments and Income Growth Across Brazilian States," University of Sao Paulo, mimeo. Azzoni,C. and T. Menezes (2000) "Um fndice de Custo de Vida Comparativo para as Principais Cidades Brasileiras: 1981-1999," Estudos Econ6micos, vol. 30(1) pp.165-186. Barro R. and D. Gordon (1983) "A Positive Theory of Monetary Policy in a Natural Rate Model," Joumal of Political Economy 91(4) pp. 589-610. Barro R. and X. Sala-l-Martin (1991) "Convergence across States and Regions," Brookings Papers on Economic Activity 1 pp. 107-182. Bourgignon, F. F. Ferreira and P. Leite (2001) "Prices, Preferences or Endowments? Accounting for Excess Inequality in Brazil," working paper, Pontificada Universidade Cat6lica, Rio de Janeiro. Cohen, Daniel (2001) uFear of Globalization: The Human Capital Nexus," working paper, OECD. Diamond, Peter (1982) "Aggregate Demand Management in Search Equilibrium," Joumal of Political Economy 90(5) pp. 881-894. 118 Dollar, David and Aart Kray (2000) "Growth is Good for the Poor," working paper, World Bank. Fabio Sanchez and Jairo Nunez (2000) "Crime and Income Distribution in Colombia: A 20 Year History," working paper, Universidad de Los Andes. Ferreira F., P. Lanjouw and M. Neri (1998) "The Urban Poor in Brazil in 1996: A New Poverty Profile using PPV, PNAD and Census Data," working paper, World Bank. Fiess, Norbert and Dorte Vemer (2001) "Migration in Brazil," World Bank, mimeo. Gordon, Lincoln (2001) Brazil's Second Chance, En Route to the First World: Washington DC: Brookings. Holtz-Eakin, Douglas, Whitney Newey and Harvey Rosen (1988) "Estimating Vector Autoregressions with Panel Data," Econometrica, 56:1371-1395. Jalan J. and M. Ravallion (2000) "Geographical Poverty Traps? A Micro Model of Consumption Growth in Rural China," working paper, World Bank. Moreira, M.M. and P.G. Correa (1998) "A First Look at the Impact of Trade Liberalization on Brazilian Manufacturing Industry," World Development 26(10) pp. 1859-74. Neri, M. (1999) "A Note on the Growth Sensitivity of Poverty across Brazilian States," IPEA, Rio de Janeiro. Neri, Marcelo C. and Mark R. Thomas (2000) 'Growth and Recessions: An Episodic Analysis of their Effects on Employment, Income, and Poverty in Metropolitan Brazil since 1982," working paper, Funda,co Getulio Vargas, Rio de Janeiro. World Bank (2000) "Attacking Brazil's Poverty," Washington DC World Bank. 119 PART II POLICY FOCI 120 5. BRAZIL AND THE KNOWLEDGE ECONOMY Prepared by Carl Dahiman, Anuja Adhar Utz, Jean-Eric Aubert, and Christine Zhen-Wei Qiang 5.1 Growth in Brazil over the last two decades has been slower than the world average. Unlike Brazil, large economies like India and China, and most East Asian and Pacific developing economies have significantly increased their shares of global GDP (Table 5.1).79 In trade, Brazil has lost market share in exports (Table 5.2),80 and its export structure is based on commodities, while most the rest of the world's trade is primarily in manufactured products (Table 5.3). 5.2 Now that Brazil is regaining macroeconomic stability, its strategic focus is shifting to the longer term. This is important because the global context has changed in fundamental ways. We are in the midst of what could be called a "knowledge revolution," 81 with major implications for Brazil's strategy: these will be outlined in this chapter. 5.3 The next section lays out some trends of the knowledge revolution and their implications for Brazil. It argues that to make effective use of knowledge for its development, Brazil can make improvements in four inter-dependant areas: the economic incentive and institutional regime, education and skills, the information and communications infrastructure, and the innovation system. The third section analyses Brazil's innovation system in more detail. The fourth section examines the information and communication infrastructure. 7Average GDP growth for East Asia and Pacific developing economies was 8.0 percent during 1980-1990 and 7.5 percent during 1990-1999. 80 According to cross country ratings of competitiveness done annually by IMD, a management institute based in Switzerland, Brazil's rank varied between 33 and 37 during the past few years. In 2000 it was ranked in 34 among 47 countries, up from 37 in 1998 and 35 in 1999. See IMD, The Wold Competitiveness Yearbook 2000. 81 The terms "new economy", 'digital economy" and 'knowledge economy" are currently used interchangeably, and usually focus on the creation and use of ICTs. We have chosen a broad definition of the knowledge-based economy that includes not only ICTs and industries dependent on R&D, but also covers the use of technical, policy, and social knowledge for economic activities. However, this not as broad as other definitions, for example, that of the 'knowledge society," which includes spiritual, social, intellectual and philosophical knowledge. 121 Table 5.1 Growth of Gross Domestic Product, Brazil versus other Countries 1980-1999 .;g .L .! :!.. Low Income 4.7 3.2 India 5.8 6.0 Lower Middle Income 4.2 3.4 China 4.2 7.2 Upper Middle Income 2.6 3.6 Brazil 2.7 3.0 High Income 3.4 2.3 USA 3.6 3.3 World 3.4 2.5 Source: WDI 2001, Table 4.1, pp.194-6. Table 5. 2 Brazil's Share of World Markets: 1980-1996 Basic 6.1 6.9 5.4 5.1 Industrialized 0.81 1.21 |.80 0.79 Total 1.3 1.7 | 1.1 1.0 I Source: Horta e Souza (2000) cited in Frischtak 2001, p. 5. Table 5.3 Composition of Merchandise Exports, Brazil and other Countries 1999 Food Agricultural Primary Fuels Ores and Manufactured Products Minerals Goods Low Income 15 3 20 3 52 India 17 2 0 2 76 Lower Middle 9 2 15 5 62 Income China 6 1 2 2 88 Upper Middle 10 2 10 4 73 Income Brazil 29 5 1 10 54 High Income 7 2 3 2 82 US 8 2 5 3 79 WORLD 8 2 5 3 79 Source, WDI 2001, Table 4.5, pp. 210-212 The Knowledge Revolution and Global Competition 5.4 A new consideration for policy makers is the speed of changes in production and dissemination of knowledge made possible by increasing scientific understanding and advances in information and communications technologies (lCT). The cost of voice transmission circuits has dropped by a factor of 10,000 over the last 20 years and computing power per dollar invested has risen by a factor of 10,000 over the same perod. The ICT revolution, in tum, is increasing the power of electronic networks as research tools, permitting a larger set of new technologies. This is also having a major impact on business and social interactions. 122 5.5 The rapid development and spread of knowledge facilitated by technical progress and the ICT revolution is creating a more competitive and interdependent world. The share of world trade (exports and imports) in world GDP increased from 28 percent in 1970 to 52 percent in 1999 (World Bank, 2001 ). Beyond trade, there is greater interdependency and competition through foreign direct investment (FDI). Sales in 1999 by foreign affiliates of transnational corporations were estimated at $13.6 trillion, roughly twice the $6.9 trillion of world exports. Globalization is even higher when account is taken of international sourcing of production inputs and inter-firm alliances, including the intemationalization of R&D and technology licensing activities. Globalization is largely being driven by investments in R&D made by transnational companies. Estimates are that between 75 percent and 80 percent of all global R&D is being made by transnational corporations. 5.6 The knowledge revolution also requires large investments in intangibles such as marketing, design, and distribution. It is estimated that in advanced OECD countries, investments in intangibles, including R&D (about 2.2 percent of GDP) plus patents and licensing, design, marketing, education, training, and software may represent as much as measured GDI.3 5.7 Even in industries characterized as low- or medium-technology, technical knowledge and the associated organizational changes increasingly provide the edge in productivity and product differentiation. Traditional industries, such as textiles, cement and steel now use new technical knowledge and information systems to improve design and quality, production processes, and the efficiency of marketing and distribution. Likewise, in agriculture, better understanding of plant reproduction and growth, genetic engineering, and techniques for harvesting, storing, transporting and distribution affect the value of plant and animal products. Transportation, distribution, finance, insurance, health and education themselves are becoming increasingly knowledge intensive. What matters more than technology creation is the effective use of technology and knowledge throughout the economy. 5.8 While, in principle, the globalization of trade, finance, and information flows may make it easier to narrow knowledge gaps across countries, the accelerating pace of change and the difficulties of many developing countries in getting started may bring about the opposite-a widening "knowledge divide." Capital and other resources could flow to those countries with stronger knowledge bases, reinforcing inequality (World Bank, 1999c). Brazil's challenge is to develop a strategy that maximizes the potential benefits and minimizes the risks. 82 The direct role of technology in this process is reflected in changing pattems of trade. Between 1976 and 1996, the share of high- and medium-technology products increased from 33 percent to 54 percent of total goods traded. The share of primary commodities fell from 34 percent to just 13 percent, while that of resource-based products remained constant. These trends have major implications for developing countries, which are primarily exporters of primary commodities. Both the share and the prices of their exports have been falling for five decades. 83 For the OECD as a whole, investments in a subset of these intangibles-R&D, public education, and software-is 7.9 percent of GDP: comparable with average investment in plant and equipment (8.6 percent of GDP). 123 Key Elements of the Knowledge-Based Economy 5.9 The effective use of knowledge and information is becoming the most important factor for productivity, the creation of wealth, and the improvement of social welfare." This does not imply that a country such as Brazil must simply rush to develop ICTs and high technology. We define a knowledge-based economy as one that encourages its organizations and people to acquire, create, disseminate, and use knowledge more effectively for greater economic and social development. The fundamental pillars of a knowledge-based economy are: * An economic and institutional regime that provides incentives for the efficient use of existing knowledge, the creation of new knowledge, and entrepreneurship; * An educated and skilled population that can create and use knowledge; * A dynamic information infrastructure that facilitates the communication, dissemination, and processing of information; * An effective innovation system centered on firms, and comprising a network of research centers, universities, think tanks, consultants, and other organizations that tap the stock of global knowledge, assimilate and adapt it to local needs, and create new knowledge and technology. 5.10 These elements are interdependent. It is not enough just to improve domestic research and development (R&D), or education, or the information and communication infrastructure without also improving the overall system of economic incentives and institutions. These affect the efficiency with which resources are put to use, the extent to which the country draws on expanding stocks of global knowledge, and the overall flexibility of the economy to restructure, both to take advantage of new knowledge and to improve performance in traditional sectors. An Overview of Brazil's Situation 5.11 This section presents a preliminary assessment of Brazil's situation using a benchmarking tool developed by WBI with indicators for the various elements of the framework. Annex 1A explains the benchmarking tool and the indicators, which are available for 92 countries and 66 indicators. To give the assessment some country specificity, Brazil's position is compared not only to that of the US (perhaps the leading knowledge-based economy) but also to that of three other Latin American countries (Argentina, Chile and Mexico), as well as to two other large developing economies (China and India)." 5.12 Brazil is behind the US in almost all indicators except the flexibility of its people. But it may surprise some readers that on many variables Brazil is also behind the Latin American countries, 84 All economies are knowledge-based. What has emerged recently, however, is that rapidly growing economies are becoming increasingly dependent on the effective creation, acquisition, distribution, and use of knowledge. 85 The analysis relies on the standard 15-variable scorecards presented in Annex IE as well as the more detailed scorecards on each pillar of the knowledge economy, as highlighted in Annex 1 F that compare Brazil with other Latin American economies such as Argentina, Chile and Mexico, as well as with China, India and the US. 124 and that in some variables it lags China and India, much poorer countries. For Brazil to take advantage of the knowledge revolution it will need to develop a comprehensive strategy. Updating the Economic Incentive and Institutional Regime 5.13 Clearly one of Brazil's weaknesses has been macroeconomic instability. This appears to have been brought under control in the last decade, so Brazil can now focus on the structural and incentive issues that constrain its potential to use knowledge more effectively. 5.14 A striking characteristic of Brazil is that it is a relatively closed economy. The ratio of exports plus imports to GDP is one of the lowest among all countries. In 1999 this ratio was 22 percent in Brazil, less than half the world average of 52 percent. Brazil is not well integrated into the global knowledge economy, and this is not just because of its size. This ratio is much higher in larger economies such as Germany (57 percent,) the UK (53 percent), France (50), China (41 percent), and even India (27 percent).86 Besides not having a large traded sector, Brazil still has relatively high tariff and non-tariff barriers, reducing pressure from international competition. According to some qualitative indices, domestic competition is quite high, implying that the main issue in the overall economic regime seems to be low exposure to international competition: Brazilian firms do not face as much pressure as those in other countries to improve their technology and productivity to keep up with the rapidly changing intemational environment. 5.15 Brazil ranks in the mid-range on the soundness of its banks and adequate regulations and supervision of financial institutions, and is better than other large developing countries such as China and India. Saving and investment rates are also mid-range, but interest rates are high, which constrains the financing of enterprises. On protection of intellectual property, the legislative regime in Brazil appears to be above average. However, as will be seen below, there are significant issues regarding their implementation and enforcement. 5.16 On institutional variables such as the regulatory framework, rule of law, government effectiveness, and corruption Brazil also falls in the mid- or lower mid-range. A stronger rating appears in voice and accountability as compared to say, Mexico or Argentina and even more than China or India, a tribute to the resurgence of its democratic tradition. Upgrading Education and Skills 5.17 Besides the relative isolation of the economy from the global economy, a second weakness of Brazil's knowledge structure is the low educational attainment and skill level of the population. For an economy at its per capita income, Brazil's literacy rate is low (at 85 percent of the population aged over 15 years). More striking is Brazil's secondary enrollment rate, just 62 percent. This compares poorly, for example, with China, a country with less than one fifth of Brazil's per capita income, but where secondary enrollment is already 70 percent. Progress has been made in recent years in Brazil, with a doubling of graduates-a trend that reflects both increased enrolments and reduced dropouts. 86 The only major country with a lower ratio than Brazil is Japan, with just 19 percent. The ratio for the US is 24 percent. 125 5.18 Brazilian enrollment rates in tertiary education are satisfactory and higher even than in China, although China is fast expanding tertiary education and may catch up with Brazil in less than five years. One way China has been able to expand its higher education so rapidly is because it now covers one third of the costs of higher education through student tuition fees. The article in Brazil's constitution mandating free public higher education obviously makes it difficult to expand higher education. In addition, the system is highly regressive in that as a rule the children of the more well-to-do gain access. Brazil's relative spending pattems in primary, secondary, and tertiary education are shown in Table 5.4. Brazil is not getting the best from its tertiary education sector. Table 5.4 Expenditures per Student - Brazil and Comparison Countries at imary 859 (12.6) 2115 (4.1) 3308 (2.1) I 3769 (2.9) Secondary 1002 (10.8) 2292 (3.8) 3518 (1.9) 5507 (2.0) University 10791 8775 6844 10893 Source: C. Haddad, op.cit., and OECD a Education at a Glance", 2000. a/ ratios between university, secondary and primary public expenditures in parenthesis, from Frischtak, 2001, p. 25.Y7 5.19 Brazil's great strength among the indicators on this component of the framework is the flexibility of people to adjust to new challenges. However, without a strong education and skill base, the effects of this asset will be limited. The country thus faces a major challenge in developing an efficient and equitable secondary and higher education strategy. 5.20 Brazil needs a greater share of professional and technical workers in the population (currently a lower share than in Chile or Mexico). But to do this it will have to devise ways to use private sources of finance. This may imply a greater focus on the provision of basic education through to secondary level. The financial constraints embodied in the constitution represent the main challenge in tertiary education. Improving tertiary quality means not only expanding the public system by tapping private and corporate resources, but also encouraging the expansion of private universities. This will involve a heightened public role of quality control over the overall system.0 A further option is to explore the potential of using ICT more actively in the provision of education, not just for teacher training, which appears to be the main area of current focus, but also for higher education, as is being done in OECD countries and in China, Mexico, and India. Greater use of ICT may permit expansion of access while maintaining quality. With greater private finance and provision, the government could focus more on the provision of grants and loans for poor students. 5.21 Regarding financing, public expenditure on education is higher as a percentage of GDP in Brazil than in other Latin American countries such as Argentina, Chile and Mexico, as well as that 87 Frischtak notes that Brazilian spending on higher education may be biased upwards by costs of hospitals and personnel associated with universities. However, this may also be true for some comparison countries. 88 The government has made major strides to improve the quality of tertiary education through several means: a large national comparative exam for graduates, an institutional evaluation program and an annual national tertiary education census. These measures add up, making Brazil a leader in tertiary education quality assurance, not just among developing countries. See 'Higher Education in Brazil: Challenges and Options," Human Development Department, Latin America and the Caribbean Region, The World Bank, August 17, 2001. 126 of large developing countries such as China and India, and equal to that of the US. Another positive feature is that its national culture seems to be more open to foreign influence than these countries, with the exception of Chile. Improving the Innovation System 5.22 As can be seen from the scorecards in Annex 1 E, the Brazilian innovation system is weak in virtually all areas, with the exception of entrepreneurship, in spite of govemment efforts. Brazil spends 0.8 percent of GDP on R&D (2.2 percent for OECD countries). It is striking that Brazil has fewer scientists and engineers engaged in R&D per million population than Chile and Mexico, and fewer also than China. Moreover, in Brazil, most R&D is performed in public institutes and state owned enterprises, and relatively little is done by the private sector. This is in contrast with Korea, for example, where private firms do 80 percent of R&D. 5.23 As pointed out by Frischtak (2001), there are some large government programs in aeronautics (Embraer), satellites (CBERS), deep-water oil exploration (Petrobras), tropical agriculture (Embrapa), and biotechnology (Genoma), distinguished by world-class technology development. By and large, however, private firms have not developed or adapted knowledge intensively, despite the strong entrepreneurial capability in Brazil. 5.24 R&D output is also low when measured in terms of patents or technical publications, compared with Argentina, Chile, China or India (though higher than Mexico). This suggests that the efficiency of the domestic R&D efforts could improve. 5.25 One way in which Brazil has obtained access to global knowledge has been through FDI, and until the 1990s, Brazil was the largest FDi host among developing countries. China's opening has now made it more successful in attracting FDI relative to the size of its economy. Large FDI inflows have been one of the drivers of China's modemization (Dahlman and Aubert, 2001). 5.26 Finally, the venture capital market is also not as well developed as in Chile, or indeed China or India, which limits the start-up of new companies, especially high technology companies without a track record or many assets for collateral. 5.27 Thus, whilst Brazil could improve the management and efficiency of its R&D effort, the main issues seem to link back to the broader policy regime that does not put sufficient pressure on firms to keep up with global best practices and new knowledge. Efforts may also be needed in the near future to attract "high-quality" (export-oriented) FDI to help access global knowledge. Expanding the Information Infrastructure 5.28 Brazil has accomplished considerable progress in developing its information infrastructure in recent years with the privatization of Telebras and the liberalization strategy orchestrated by the National Telecommunications Agency (Anatel). "Teledensity" remains lower than in advanced economies (fixed line teledensity-23.1 lines per 100 inhabitants-is one third of US and half of Spain, while mobile service density reached 14 accesses per 100 people in 2000). Brazil has fewer intemet hosts per 10,000 persons than Argentina or Mexico. The Internet is thus still out of reach for the majority of the population-less than 10 percent of the population has access to it-owing 127 to a series of factors: lack of infrastructure, high prices, and low educational levels. Promoting universal access to telecommunications-through infrastructure development and tariff lowering- requires appropriate deregulation, intensified competition and the attraction of foreign investors. 5.29 Computer penetration is also lower in Brazil than in Argentina, Chile or Mexico, but higher than that of China and India. The ICT skills of the population could be developed, which may also increase internet content. In terms of the percentage of companies that use the internet for e- commerce, Brazil ranks lower than Argentina and Mexico and slightly higher than Chile. 5.30 The contribution of innovation and technology to Brazilian economic development has been limited. Competitiveness has declined slightly over the last two decades. Overall productivity is low compared to other major developing economies and the innovation system contains serious gaps. Brazil does not take advantage of the entrepreneurship of its population, nor from significant investments in its science base. Improvement requires a multi-pronged, long-term effort. University-industry interactions need to be improved by various means. Appropriate financial and technical support should be provided to enterprise-particularly to SMEs-to stimulate its innovation and R&D efforts. Incentives of various forms (export promotion, competition, standards), as well as the intellectual property regime, could be strengthened. More generally, the technical culture of the population could be raised in the context of a comprehensive education policy. Technology and Economic Performance 5.31 The position of Brazil in international trade has declined slightly over the past two decades. After an improvement in the early eighties, there was a decline in the second part of the eighties and stagnation since then (Table 5.2). Brazil has shifted its export composition away from basic products towards industrialized goods, a trend that is consistent with world trade. However, its comparative advantage has been accentuated in resource-based sectors, while there has been a loss of competitiveness in technology intensive sectors. Falls in market shares are noticeable in electronics, textiles and transport equipment, as well as in processed agricultural products. Gains in market shares were in steel, non-ferrous metals, chemicals and petrochemicals (Table 5.5). 5.32 There has been a moderate contribution of technology to competitiveness, insufficient to keep up with intemational trends, judging by international benchmarks. Brazil appears to have low overall productivity and low labor productivity, regardless of sector (agriculture, manufacturing, and services).90 89 This section is largely based on the information provided at the WBI Knowledge for Development Policy Forum for Brazil, China and India (Wilton Park, U.K. March 2001), and notably by Frischtak (2001). go World Competitiveness Report, IMD, 2000. Out of the 47 countries listed in the report, Brazil's rankings are 41 for the overall productivity, 38 for labor productivity, and respectively 37, 38, and 38 for agricufture, manufacturing and services (related GDP (PPP) per person employed. 128 Table 5.5 World Market Shares of Brazilian Exports Major Product Groups Basic Products 6.1 5.1 Agricuhure and livestock 6.2 3.9 Mining 6.1 8.3 Soybeans and related products 5.9 4.8 Industrialized Products 0.81 0.79 Non-metallic minerals 0.6 0.8 Steel 1.3 3.5 Non-ferrous metal 0.4 1.4 Mechanical Products 0.6 0.6 Electric products/electronics 0.5 0.2 Transport material 0.8 0.7 Wood and Fumiture 0.9 1.1 Pulp and Paper 1.2 1.5 Rubber 0.9 1.5 Chemicals 0.3 0.6 Petrochemicals 0.4 0.6 Pharmaceuticals 0.5 0.4 Textiles 1.2 0.8 Food Products 2.8 2.6 Cut and Processed Meat 1.6 1.5 Other 0.3 0.3 Total 1.3 1.0 Source: same as Table 5.2 The Innovation System 5.33 Brazil's investment in science and technology, dating from prior to the Second World War, is not large measured in percent of GDP, but it is substantial in absolute terms. Total S&T expenditures reached $10.7 billion in 1997 (versus $6.2 billion in 1993). Overall R&D spending-at about 35 percent to 40 percent of total S&T expenditures-is 0.8 percent of GDP.9" International comparisons of R&D efforts per capita are given below (Table 5.6). 91 Brazilian statistics in science and technology do not follow OECD standards. Overall S&T investment figures are available (of which the content varies from one country to another). Then estimations are made for R&D that vary depending on the source. There is some convergence around the figure of 0.8 percent of GDP. There are also problems in the estimation of R&D personnel: there are no credible figures for "Full Time Equivalent" (FTE) researchers. IMD's figure of 21,500 (2001) appears low, though many R&D personnel work in the university sector and therefore devote much of their time to teaching. 129 Table 5.6 R&D Expenditures and Personnel per Capita, 1998 Total Expenditures US 842.5 4 Japan 969.9 3 China 5.3 40 Brazil 37.2 31 India 2.4 43 Business Expenditures US 807.92 _ 1 Japan 698.76 | _3 China 2.38 40 Brazil 14.89 28 India 0.47 43 Total R&D personnel nationwide per capita (per 1000 people) - US 3.73 I 18 Japan 7.09 3 China 0.60 33 Brazil 0.137 41 India 0.135 42 Total R&D personnel in Business per capita (per 1000 people) US 2,962 9 Japan 4.65 3 China 0.25 22 Brazil 0.04 40 India 0.038 41 Source: IMD, 2000 5.34 The total number of scientists and engineers working in R&D is estimated to be 48,800. The number appears to be considerably lower when counted in full time equivalent (FTE): 21,500 (Table 5.6). Two other issues deserve mention. First, there is a strong regional concentration of S&T potential in a few areas, notably Sao Paulo, which concentrates about 50 percent of total S&T expenditures and personnel. Second, business contributes modestly to the overall R&D effort. About 30 percent is funded by business, the rest by the govemment. Of concem is the very low number of researchers in the business sector (6,800 in FTE). Seventy percent of the S&T workforce is in the university and public sector. 5.35 At the same time, Brazil benefits from the most entrepreneurial population in the world, according to a recently compiled index of entrepreneurial activity (Table 5.7), combining data of business creation-Brazil being first in the world with one adult out of eight creating his or her own business-and data of adult employees working in firms with less than 42 months of existence- Brazil being third behind Korea and the US. Even more surprising is that this dynamism is taking place in a climate that is not very favorable to business creation and development (issues of regulatory burden, tax overload, etc.-see chapter 7). 130 Table 5.7 Index of Entrepreneurial Activity - . S *. . . . . . S . S . ..S 16 113.6 112.6 1 10.9 1 7.9 j7.9 7.7 6.3 15.6 .15.1 4.7 14.5 Source: Global Entrepreneurship Monitor 2000. 5.36 However, the links between S&T investment and entrepreneurial capability are not well established. Brazil neither takes full advantage of its investments in science, nor of its entrepreneurial capability. The innovation system thus presents definite strengths and gaps. 5.37 Some innovation successes-of worldwide significance-are noticeable, notably in heavy engineering (aviation and oil offshore exploration), and in agriculture. In all cases the public sector has played a crucial role. The state owned enterprises-Embraer and Petrobras-have been the main engineers of successful developments in aeronautics and oil exploration. In the agriculture sector, Embrapa, the public research institute, has been the key platform. Promising developments in life sciences (genetic research) are also noticeable, though they have yet to be transformed into commercial successes. 5.38 As a whole, these successes remain isolated. The inventive performance of the country, measured in terms of patents deposited in international offices such as the USPTO or WIPO, remains mediocre (though increasing), and is not commensurate with S&T investments (Table 5.8). Table 5.8 Patent Requests to WIPO, 2000, Number and Percent .511111j 5n WMS .MY S=TV *.ETO .M FM 38171 112039 9402 1 5538 3601 3071 12587 11701 11627 1600 11514 1579 161 1 42.0 13.2 10.3 6.1 4.0 3.4 2.8 1.9 1.8 1.8 | 1.7 0.64 0.18 Source: WIPO, cited by Frischtak 5.39 The relatively low effort of the business sector in R&D affects the innovative capacity of the economy. Problematic is the low number of R&D personnel employed by enterprises. This can be explained by several factors. First, hyperinflation did not encourage such long-term investment, and these effects persist in Brazil's financial system today. Secondly, the specialization of Brazilian industry tends to be concentrated in intermediate products of relatively low technological content. Thirdly, low recruitment of qualified scientists and engineers seems due to the domestic orientation of most enterprises. Exports represent only 5 percent of domestic firms' sales and 12.5 percent of the turnover of foreign-owned firms. 5.40 Nevertheless, innovation surveys made in the Sao Paulo area92-the most dynamic part of the economy-show that many firms innovate (about 25 percent have introduced technologically 92 'Technological Innovation in Brazilian Industry: An assessment Based on the Sao Paulo Innovation Survey," Ruy Quadros et al. Intemational Joumal of Technological Forecasting and Social Change, vol. 67, no. 2, May 2001. The survey is made on bases comparable to those implemented in the EU, following the methodologies of the OECD and Eurostat. 131 modified products or processes over the period 1996-98).93 There is a concentration of the innovation effort in intermediate goods (e.g. chemicals) and conventional engineering sectors. A large proportion of innovations are driven by demands and ideas coming from suppliers and clients rather than by science and research activities (a feature observable in advanced economies too, though in a lower proportion). 5.41 There are signs that a dynamic small and medium enterprise (SME) segment is emerging in the Brazilian economy. A stock of over 10,000 "high productivity" firms in the range of 20 to 500 employees have been identified,4 and the number of exporting companies has been increasing regularly over the years in all size categories (although 450 firms were responsible for nearly three quarters of manufactured exports). They benefit from a number of export and SME supporting programs, the latter spearheaded by Sebrae- the well-funded Brazilian Service for Support of SMEs-in conjunction with private actors and social organizations. They also take advantage of 45 industry-funded National Centers of Technology (Cenatecs) and of a number of services provided by universities, which have spread in recent years (such as "dial technology" offices). 5.42 Incubators-to support utechno-entrepreneurs"-are also spreading. More than 135 structures have been established in 17 states. They have so far attracted a limited number of enterprises (1,100), contributing modestly to employment creation. Although future prospects are encouraging (200 incubators are expected to be established at the end of 2001), this illustrates one issue affecting Brazil's innovation capability: the inadequacy of technical culture among the population. On one hand, the majority of entrepreneurs do not pursue technical ventures. On the other, the scientists do not manifest the same entrepreneurial spirit as the population. Only 462 PhDs were working in business incubators at the end of 2000, less than 1.7 percent of a pool of some 27,700 PhDs located in university departments and research institutions.95 5.43 Some of these incubators are part of larger, complex structures such as technology parks, (e.g. in Rio de Janeiro, Pemambuco, Rio Grande do Sul, Ceara and Brasilia). Most are incipient, but a first cluster of S&T parks is beginning to consolidate around the city of Campinas in the State of Sao Paulo, where 13 research universities and research institutions are responsible for 16 percent of the country's scientific output, and generate 9 percent of the national GDP. 5.44 Traditionally MNCs have been the main vector of technological and managerial modemization of the Brazilian economy. Despite the growth in FDI flows and continuing investor confidence, the potential for FDI to operate as a knowledge precursor seems to be receding.96 In particular, FDI is increasingly directed to services and other non-tradable goods, where competitive pressures are smaller, where the knowledge gap not as pronounced as in the manufacturing sector, and where competitive advantage centers on inexpensive cost of capital. Cumulative FDI flows in 1996-99 were 19.2 percent for industry and 79.2 percent for services. There are significant ,obstacles that discourage high-quality FDI, including a restrictive regime on technology transfer 93 A lower proportion than in advanced economies (50 percent in average) 94 See "Valor, December 21, 2000, data generated by the Ministry of Finance (cited by Frischtak). 95 As noted by Frischtak, op cit. 9 See Frischtak (2001). 132 (systematically requiring govemment licenses), and mediocre protection of property rights.97 A well functioning IPR regime, fully in line with TRIPS agreements, would significantly increase FDI.98 5.45 A problematic intellectual property rights regime affects the innovation climate in Brazil. Laws in the concerned areas (inventions, industrial designs, trade marks, software, etc) are, in general, in compliance with international standards. But the conditions goveming their application are inadequate, e.g. the patent approval process is slow (five years on average). Their enforcement is also mediocre, due to judicial back-logs and a dearth of judges adequately trained in intellectual property law.99 As for foreign patents, the Brazilian law permits the issuance of compulsory licenses in cases where patent holders choose to supply the Brazilian market through imports rather than local production."° Moreover, there are no incentives to stimulate inventions coming from R&D work performed in universities or in businesses funded by the federal government to encourage innovation (incentives such as the Bayh-Dole act in the US, which allows benefiting entities to retain title to inventions under federally-funded research programs and encourages their commercialization, notably through SMES). Government Policies 5.46 Over the years, the government has put in place a series of schemes, some with the support of the World Bank (PADCT I and 11, which invested a total of US$ 470 million in 4,500 projects over the period 1986-97), which have been strongly science-oriented (focused on building research infrastructure and training of scientific personnel), rather than technology-oriented. They have contributed to the development of important programs and institutions (such as the National Council of S&T Development, CNPq, and the Fund for study and projects, FINEP). Meanwhile some state govemments have taken similar initiatives, principally Sao Paulo with the establishment of a science foundation (FAPESP). The impact on the scientific potential of the country has been substantial. For instance the number of PhDs amounted to 2,200 per year in 2000 versus about 500 in 1980. Brazil has nearly tripled its share of world publications (from 0.44 percent in 1981 to 1.2 percent in 2000). 5.47 In the late 1990s, a new series of more technology-oriented measures were put in place with the support of PADCT. A reform and investment program was implemented with a total planned expenditure of some $360 million for the period 1997-2001. The program included a series of schemes to support R&D cooperative projects (between the govemment/university sector and the business sector), basic technological research and various "sectoral" services (metrology and standards, IPR regime, evaluation and monitoring, etc). These measures, if effectively implemented, should begin bearing fruit. 97 As noted in the FIAS report (2001). The report indicates that, in a recent survey, 40 percent of foreign companies said they did not use their most advanced technology or processes in their Brazilian-based operations. 98 The potential increase would be 7.4% of current FDI flows according to a recent econometric study. See the World Bank report on Global Economic Prospects 2002, chapter 5 (July 2001). 99 See FIAS (2001), notably pp. 187 to 199. 1° Although the most aggressive use of the law has been in pharmaceuticals (and is estimated to have helped fight AIDS), the law refers to all patents, whatever sector is considered. See GEP 2002, op. cit. (box 5.2). 133 5.48 To spur the innovative capability of the economy, the govemment has recently decided to establish seven sector funds (using resources brought about by a tax on services provided by recently privatized enterprises), though the conditions of use and management of such funds remain unclear. The decision has also been taken to establish a public venture capital fund. In order to reinforce the strategic capacity of the govemment, a Strategic Study Center has also been created in the Ministry of Science and Technology. Policy Guidelines 5.49 A series of well-targeted and vigorous measures are needed to improve Brazil's innovation climate. A first priority would be to act on those elements of the incentive regime that affect the innovation climate. In particular, it is recommended to: Improve the IPR regime by strengthening enforcement and accelerating the approval process. This requires increased resources for INPI (the national institute in charge of intellectual property), as well as a transformation of the review procedures (e.g. systematically involving foreign partners concerning foreign technology applications). * Further reduce obstacles to firm creation (bureaucracy, regulations, tax burden, etc) particularly for new technology-based companies, and obstacles to the availability of long term finance. This is a key area for improvement (see the 2000 World Competitiveness Report), and is crucial for innovation ventures, which are by nature uncertain and only have a long-term pay off. * Stimulate exports and encourage enterprises to improve their product scopes and quality. In particular, technical norms and standards should be reviewed and made more demanding, placing them on par with international competition. The establishment of a permanent commission at the Federal level for this purpose is a possibility, which would work in close contact with State authorities to reduce tariff and non-tariff barriers. * Eliminate regulatory obstacles that prevent adequate use of foreign investments and related technology transfer (by suppressing all govemment licenses and authorizations as has been done in Chile). 5.50 A second set of measures should address the interaction between the university and public research sector on the one hand, and the business sector on the other. Three actions are recommended: * Establish a powerful IPR-related incentive instrument along the lines of the US Bayh- Dole act. In addition, the conditions of licensing by the university and public sector of their inventions to the private sector should be adapted. * Consider financial measures to facilitate-for the benefit of SMEs-the recruitment of researchers, as well as the contracting of R&D work to universities and other research institutions. These schemes have proven to be useful and efficient in 134 advanced economies (cf. French, German, and Dutch experiences). They consist of subsidizing a proportion of the cost of salaries of researchers recruited by the SMEs, or of the cost of contracts of SMEs with university or public laboratories. The subsidy can take the form of a tax deduction. Establish industrial R&D PhD programs that are jointly developed by the university and business sectors (as done in several OECD countries-see for example the Danish experience, which allows the granting of PhDs to individuals pursuing full time research work in enterprises). 5.51 It is important to design the new sector funds under MCT efficiently (these will pertain to technologies and industries of the Brazilian economy, such as oil, biotechnology, etc.). The scope of these funds-which will become a major renewable source of public funding for technology development-is not yet clear; they range from the support of basic research to technical research of a more applied nature (although at the pre-competitive stage). The lessons of international experience can be brought into the design and implementation of these incentives. In particular, a clear distinction should be made between the support of basic research, which should be based on "excellence" criteria, and the support of technical/industrial research, to be based on "relevance" criteria (patent prospects, potential profits, etc). 5.52 Finally, there is the possibility of expanding schemes for technical (and commercial) assistance, information, and training that address the needs of SMEs: the number of supporting structures (such as Cenatecs) is small. This could be done at the local and regional level and the government could encourage states to provide funding. Mechanisms adopted in a number of countries (Germany, France, Spain) consist of developing programs jointly with local authorities and associations of entrepreneurs and engineers, on the principle that local authorities match the funding provided by the central govemment. 5.53 It is clear that improving the overall innovative capability of the economy will take time and will not bear significant results before a decade or so. A number of measures must be taken in close coordination between the Federal and State governments. Finally, as is usual with knowledge and innovation policies, there is need for inter-ministerial action. For these reasons, the establishment of an appropriate coordinating body at the top level of the govemment could help catalyze and oversee policies. _ LW M a_to w Ki n M I 1311, WIN Recent Reforms and Privatization and Liberalization Developments 5.54 The Brazilian telecom sector was a state monopoly until July 1998, when the privatization of Telebras occurred. Until this date the market was characterized by a lack of investment, poor quality of service, and high costs."' The plan announced by Brazilian regulator Anatel (Agencia 101 Prior to 1972, there were more than 900 telecommunications companies scattered across Brazil. By 1975, Telebras and its operating subsidiaries (collectively known as the 'Telebras system") had acquired the vast majority of these companies, creating a monopoly in telecom services. The Telebras monopoly lasted until 1975, when the Brazilian govemment proposed a reform of the regulatory system. In 1995, Congress approved Constitutional 135 Nacional de Telecomunicagoes) in September 1998, for the implementation of the final stages of privatization, paved the way for a deep transformation of the country's telecom sector. The first stage of Anatel liberalization strategy is to permit competition through a number of "mirror" companies by allowing a new service provider in each of the service areas dominated by the former Telebras companies. This will be followed up by full competition in 2002, when there will be no geographical or service restrictions. Efficient Instruments and Clear Goals It was the introduction of the 7ha General Law of Te1--crrimt rn1c8'r! '> (Lei Sorn! de Tefecomtrntcagbes LGT) In 19 97 that provided for haoth Co, mll. icr and to c'.atfz of Ta1ebras. The 3dmnidisl atively independent and flnarcialiy atiforomous regij!ator Anatol was the instrument to guide Brazi towards telecommunrdations reform. In =adtion to Anatsl, The ' o'inP' of Communications determines overall pcsdpes to be followed and CADE ensures fair CO -:Xkii!on withi!n aiU business snctors 'n 3rmi,. Angiefe- fundamental job is to regulate, grant concessions and oversee elecommunication services in Brazil. All procedural steps, proposals and dseisions are :uomit,c'1 to widespread ;crLTiry, aimed at *econciling the public interest with the particular interests of inR pr',ae sector and oihe3r -nmernts of society. Vith regard, to services, Aratsis goal is to ansure universal access to the pcpl !etion, vini. -nandatorv expansion and quality goals. LGT classifies telecorr! ieices IC_rd I,_ to the judicial status under which they are provided, pubric or p0vate. This classification uwas IMp!errented so that the governmen could ensure that operators providing services underihe public regime fulfilled th.ei srOaI ratar' and r.cr ,,,I, abigerione 5.55 The success of liberalization of the telecom sector and the privatization of Telebras is visible today in the explosion of new services available. In July 1998, when 27 state-owned telephone companies were privatized, there were 20.2 million hard-line telephone accesses. The 34 hard-line telephone operators have since installed (up to December of 1999) 7.6 million new lines, an expansion of more than 37 percent in a year and a half. Likewise, the number of public telephones, during this same period, rose from 547,000 to 740,000. Quality also improved during this period, such as the network digitalization index, the average time to get a dial tone, the local and long- distance phone rates and the number of orders placed for repair services per 100 public telephones. More than 271,000 phones purchased through expansion plans that had not been delivered in December of 1998 were installed by May of 1999. 5.56 In the area of wireless phone service, the evolution has also been considerable. The launch of prepaid services increased demand, and increased competition is driving down costs. In 1994, 800,000 people owned cell phones, 5.6 million were in use in July of 1998, and by December 1999 15 million cell phones were in operation in Brazil. An increasing proportion of the Brazilian population is benefiting from access to telecom resources. Fixed-line teledensity in 2000 was 23.1 lines per 100 inhabitants, approximately half of Spain's and one-third of the USA's, and is expected to reach 32.6 by 2005, while mobile service density, after experiencing an extremely fast Amendment No. 8 ending the monopoly of the state-controlled telecom companies and representing the first step toward profound reforms. In 1998, the cellular interests of Telebras were broken up into separate companies. Eight cellular providers were formed, each operating in one of eight cellular regions, along with three fixed-line service providers. The government completed the privatizaton of Telebras in July 1998. 136 expansion, is projected to grow from 14 to 32.6 accesses per 100 people between 2000 and 2005 (Table 5.9 below)'1. Table 5.9 Brazil: "Teledensity" 1995-2005 a/ !I :.*]. . II, _ I! ; ....... L Fixed Line 9.3 13.6 23.1 32.6 I 20.0 7.1 Mobile 0.9 4.5 14.0 32.6 73.1 18.4 Source: 'Infrastructure and Universalization of Telecommunications Services," presentation to the Policy Forum on Using Knowledge for Development, WiRton Park, March, 2001. a/GI and G2 - average annual rates of growth between 1995-2000 and 1998-2005. b/Ministry of communications projections. Limited Access to the Internet 5.57 Informational constraints affect most people in the country. In view of the revolutionary nature of the internet-and its capacity to subsume most media-when it comes to accessing and exchanging information, as well as the far more limited informational content of the widely disseminated open TV (and radio), most discussion around the issue of access centers on the use of internet resources by the population.'03 5.58 The intemet remains outside the reach of most people in Brazil. Though estimates vary, partly due to definitional issues, approximately 9-10 million people may be considered "connected" to the internet, that is 54-60 people per 1000 population, which compares unfavorably with Mexico (95 per 1000), Chile (156/1000) or Argentina (174/1000).'° Considering there are some 129 million people aged 10 and above, 7-8 percent of the population has access to broad informational resources. Ibope, a public opinion research institute which tracks internet use in the country, concluded that 84 percent of users are in the A/B socio-economic category (upper and middle class), 14 percent in lower middle class (class C), and only 4 percent in classes D and E, which make up most of the population. According to these data, the internet is still overwhelmingly an elite instrument in Brazil, the potential of which could and most likely is being used to reinforce existing concentration of income and wealth. 5.59 A combination of lack of (quality) infrastructure, high prices of critical services and equipment (particularly relative to families' disposable income), and low educational levels which impede the use of the intemet, are restrictive factors explaining why less than 10 percent of the "eligible" population is connected. 102See 'Infrastructure and Universalization of Telecommunications Services," presentation at the Wilton Park seminar, and Sociedade da Informa,ao no BrasR - Livro Verde (2000), chapter 3. It is worth noting that total telecom access paths per 100 inhabitants-37.1 in 2000-was still little over one-half the average for OECD countries in 1988 (72.1). See OECD (2001). 103 As noted in the paper by Frischtak, op cit., from which the data are extracted. 104See Eric Portero (2000), except for Brazil. 137 5.60 Despite these limitations, Brazil is Latin America's largest Internet market, representing over 50 percent of all Internet users and nearly 80 percent of all e-commerce activities in the region. These statistics, combined with the fact that the Brazilian intemet market is growing at over 50 percent a year, present a lucrative proposition to many international operators. Explosive demand is also taking its toll on Brazilian Intemet Service Providers (ISPs), not only in terms of catering adequately, but also in making adequate provision for increased bandwidth requirements. The advent of free subscriptions, introduced by banks and e-commerce companies in January 2000, threatens to fuel the fire. Today there are seven backbone networks in Brazil that have been constructed by private entities. In addition, Embratel has built a commercial backbone linking several of Brazil's largest cities. 5.61 In terms of e-commerce, 27 percent of the Brazilian Internet users currently make purchases on line. This figure represents 80 percent of the entire Latin American region's e-Commerce activity. Recent surveys suggest that over the next few years, the Latin American e-commerce market will explode, and more than 50 percent of Brazilian users will make purchases online. Conversely, a number of barriers will affect the future prospects of e-commerce in Brazil. These include low-PC penetration, low-credit card ownership, and poor product fulfillment infrastructure. High Potential and Investment 5.62 The new regulatory conditions, coupled with the sheer size of the country, presents unrivalled opportunities in the country's telecom sector. Brazil is an enormous market. Brazil is among the largest telecom markets in the world. With its rapid economic growth, Brazilian telecom has great potential for development: it is estimated that telecom revenues in Brazil could reach US$78 billion in 201 O."' Commercial activities in Brazil that use the Internet are expanding considerably, to the point that they now make up more than half of the Latin American market, in terms of users and in volume of transactions. Investments, which in 1998 alone exceeded $13 billion, coupled with the arrival of new technologies, equipment, and information services, have placed Brazil at the leading edge of the production, storage, use, and dissemination of voice, data, text, sound, and image formation. Increased investment has had a positive impact on prices, quality of service, and the range of services readily available. The projected magnitude of investments in this sector for the next 5 years is shown in Table 5.10. Table 5.10 Investment Projections through 2005 (US$ billion) Fixed 6 _ 5. 4.4 4.1 3.9 29. Mobile 2. 3.1 3.7 4 4.< 4.5 21.' ass Communications 1. 2. 2 2.2 2.M 2.2 12.4 Total 10.9 10.! 10. 10.6 10. 10.6 64.11 Source: Anatel 2000 105 Pyramid Research, 2000. 138 5.63 In the Intemet area, booming corporate and consumer use is attracting some of the world's largest ISPs, along with several Latin American providers with financial backing from US investors. Among the foreign companies are AOL, which is investing heavily in the launch of a number of local web sites in Brazil, and Telef6nica, which recently purchased Nutec, Brazil's second largest ISP (for about US$500 million). Instruments to Stimulate R&D 5.64 In comparison with the rest of Latin America, Brazil does have a small sophisticated technological base and a contingent of qualified human resources. The Secretariat of Computer and Automation Policies (Secretaria de Poifticas em Informatica e Automa9ao-SEPIN) of the Ministry of Science and Technology (MCT) is in charge of managing all aspects related to the Law of Fiscal Incentives in Computer Technology (Law 8248). A close evaluation reveals that, during the period of 1993 to 1999, the results obtained with the incentives provided by this Law are quite remarkable: In 1999,183 requests for fiscal incentives were approved, filed by 263 companies. Investments in R&D also rose significantly-a total of R$600 million was spent in 1999 in this area by companies taking advantage of the incentives available through legislation. Of this total, R$225 million was spent in partnerships of private corporations with universities and research centers. In the same year, R$15 million resulted from the payment of charges established through the Computer Technology Law. With this money, the govemment was able to strengthen projects for the qualification and development of human resources in the area of R&D, to expand the national Research Network and bolster the exportation of software. 5.65 There are some challenges: * There is a shortage of researchers; * In terms of distrbution, the Brazilian scientific community is concentrated, in large part, in universities and research centers-only 20-30 percent work in industry. * Technology transfer from academic circles to the Brazilian industrial sector is low; * Brazil's education system and skills bottlenecks will, over the longer term, present the largest threat to the future of the ICT development. Table 5.11 ICT Trade_Balan ce8 (U9$Umilionn) mports 126 357 008 omputer and perpherals 662 070 015 elecommunication 1464 287 1993 xports 82 53 92 omputer and peripherals 89 329 7 elecommunication 3 24 255 Deficit 3744 4804 -4416 Source: SEPIN/MCT Shortage of Engineers and Scientists 139 Industrial Policy and Regulatory Issues 5.66 The Brazilian telecom regulator Anatel has enhanced the competitive environment in the course of sector reform. However, some operators have noted that the pace of Anatets regulation process is too slow to keep pace with the demands of the market. Both fixed and mobile operators find that Anate!s regulation, which is targeted to protect consumers' interest, does not consider investors' business reality and hence hinders private investment. Cable operators have stated that Anatets upcoming regulations on bi-directional services have lagged the market's needs."08 Specific regulatory issues include the following areas. Failure of C-Band and D-Band Licensing 5.67 On February 2nd, 2001, due to a lack of bidders, Anatel postponed the upcoming C-band PCS license auction, which promises efficient Internet and data transmission over cellular phones. Having suffered a number of delays due to lawsuits and injunctions, the failure of the C-band PCS auction will postpone the entrance of a third competitor in Brazil's major markets until January 2002. Anatel's failure to assess market reality and existing service penetration levels contributed to the lack of interest in the C-band license auction. Following an extensive consultation period on everything except pricing, Anatel determined the value of each license using a formula that failed to take into account several market factors, decreasing the attractiveness of the investment at Anatel's minimum established prices. 5.68 The latest wireless (D-band) auctions failed to attract new market players. Telecom Italia (TI) was the only multinational company to participate, bidding in all three regions and winning two of them (including Sao Paulo state). The absence of new investors has more to do with global investors' recent loss of faith in high-technology shares than in a lack of interest. However, Anatel has also raised service-quality standards, including tough customer-satisfaction targets (measured by complaint volume). These factors, along with high prces for the concessions in general, make it harder for companies to enter the Brazilian market. Interconnection Rates 5.69 Interconnection rates in Brazil do not reflect costs: they represent approximately 50 percent of the user tariffs, although for certain distances and times, the interconnection rates are higher than the user tariffs. The regulation of interconnection rates by Anatel tends to limit abusive rates. However, the absence of competition in the local access market limits negotiation of rates. These factors represent a limitation for the development of the telecom market. Fines 5.70 Fines are up to R$50 Million for violation of Universal Service Obligation; R$50 Million for jeopardizing competition; R$50 Million for violation of Quality Obligations; R$50 Million for violation 108 Anatel's imminent decision to allow bi-directional Intemet services over coaxial cable systems will enable Brazil to offer a variety of access technologies: traditional copper, TDM, broad-band wireless, coaxial cable and fiber optics. 140 of User Rights. The fines only have caps and the caps are too high. There are no clear definitions of violations and rules to determine the amount of fines according to the degree of the violation. Government versus Private Sector 5.71 Since its privatization in 1998, Brazil's telecom infrastructure has been modernizing quickly. Heavy investments have served as stimuli for the development of the Internet and related industries. However, companies are beginning to worry that government may introduce new taxation on the sector. The concern has been widely shared by e-business executives regarding the authorities' intention to introduce a new form of taxation on electronic transactions. The levy could be similar to the existing 0.3 percent financial-transactions tax and assessed at the clearing stage of payments. The city of Rio de Janeiro is already charging telecom firms R$1 for each kilometer of fiber-optic network installed. This, coming on top of an already heavy corporate tax burden, could damage business prospects associated with e-commerce. The Future of Information Infrastructure: Main Challenges Equipment Cost 5.72 Equipment costs have been traditionally one of the binding constraints to web connection. More recently, entry of new producers (Compaq and Dell, among others), with gains in scale and productivity (and continuing competition from gray market producers/imports, which supply an estimated 50-60 percent of demand) has led to a fall in prices. Nonetheless the cost and availability of credit and the income distribution both limit demand. As a result, the rate of PC penetration was still low at 26.3 per 1000 habitants in 1 999.107 5.73 Equally important, with few national backbones, the connection costs to the IP network charged by the dominant provider (the privatized Embratel, which controls an estimated 70-80 percent of the internet traffic) costs a multiple of equivalent services in other countries, where competition is more intense (Table 5.13). Table 5.12 Product and Services Pric om arison: Brazil versus USA (US$) Fixed line subscription a/ 10.20 16.00 -36.3 Minute in mobile b/ 0.23 0.35 -34.0 Pentium 3/800 Mhz/64 mega memory 1008 924 17 Printer HP deskiet 640 C 152 99 54 Microsoft Office 2000 standard 438 468 -6.4 Connection of 1 Mbps to IP c/ 7200 700-1200 500-900 Source: 'Valor", 12-14 January 2001. Notes: a/ in the US all local calls are included, whereas in Brazil, each additional call is charged above a certain number, which implies a significantly higher total price than the US; b/ average tariff for Telesp Celular, in S.Paulo; c/ prices for UUNet in the US, and for Embratel, according to Brazilian providers (for Embratel, the price ratio is 2.5 to 3.0). 07 By comparison, Mexico's penetration rate was 37.3; Argentina's 39.2; Chile, 54.1; Spain's 22.1; and the US, 406.7. See Ministry of Science and Technology (September 2000). 141 Backboning and Access 5.74 High-speed communications depends increasingly more on the utilization of fiber optic infrastructure. The infrastructure in the North and Northeast regions of Brazil is not readily available. This vast area will still depend, for some time, on satellite-based means of communication, either geostationary or low orbiting, to meet their communication needs (including voice telephone). With taxes on local calls currently averaging 28.6 percent, the average cost for local calls in Brazil remains among the region's highest (see Table 5.13), and dedicated connections are prohibitively expensive. Without the adoption of altemative tariff plans (such as unlimited local calls or byte-based rates) for internet usage, dial-up Internet access, which is currently the dominant mode of accessing the Internet, will remain too costly to most consumers. Table 5.13 Telephone Tariffs in Selected Countries of the Americas Region, 1998 (US$) Residential Business 1=@ ~I I ~ E J Local call per capita Argentina 150 12.8 150 36.4 0.1 1.9 Brazil 43 6.7 43 11.6 0.09 1.7 Canada 42 13.2 58 37.8 .. 0.8 Chile 159 16.3 159 16.3 0.12 4.2 Peru 151 14.8 151 16.2 0.08 7.3 United States 44 19.9 70 41.3 0.09 0.8 Central America 121 6 159 9.9 0.06 4.4 outh America 186 7 247 12.3 0.07 2.8 aribbean 51 9.2 65 19.8 0.07 3.7 ericas 107 8.3 145 16.4 0.07 3.3 Source: Americas Telecommunication Indicators 2000, ITU 5.75 On Brazil's current trajectory, its low-income households and less developed regions will be left behind. It will take a great deal of government-support funding to bring affordable community Internet access centers and to provide basic education/training on how to use the Internet. Policy & Regulation 5.76 Continued growth in consumer and business adoption of e-commerce depends on several factors: online user privacy, electronic contracts legitimacy, industry-specific insurance provision and antitrust regulations (limiting mergers and acquisitions among telecom and internet service providers). With the liberalization of various market segments and the change in the nature of services offered, there is a need for transparent policy and rules on these issues. Lack of Content in Portuguese 5.77 The volume of information available in networks is an indicator of a nation's knowledge capacity. In Brazil, despite emerging local content, the internet remains a predominantly English- speaking medium and, therefore, places content restrictions on the vast majority of Brazilians. 142 5.78 Growth in Brazil over the last two decades has been slower than the world average. In order to catch up, Brazil could make effective use of knowledge for development. This does not just imply that Brazil must simply rush to develop ICTs and high technology. It could, however, make improvements in four key areas: the economic incentive and institutional regime, education and skills, the information and communications infrastructure, and the innovation system. 5.79 These elements are interdependent. It is not enough to improve domestic R&D, or education, or the information and communication infrastructure, without also improving the overall system of economic incentives and institutions. These affect the efficiency with which resources are put to use, the extent to which the country draws on expanding stocks of global knowledge, and the flexibility of the economy to restructure. For Brazil to take advantage of the knowledge revolution to accelerate its rate of growth and to increase its welfare it will need to develop a comprehensive strategy encompassing reforms in all four pillars of the knowledge economy. Updating the Economic Incentive and Institutional Regime 5.80 Brazil is not very well integrated into the global knowledge economy, and this is not just because of its size. Besides having a small traded sector, Brazil still has relatively high tariff and non-tariff barriers, and this reduces pressure from intemational competition. The main issue in the overall economic regime thus seems to be low exposure to international competition: Brazilian firms do not face as much pressure as those in other countries to improve their technology and productivity to keep up with the rapidly changing intemational environment. 5.81 Brazil ranks in the mid-range on the soundness of its banks, has adequate regulations and supervision of financial institutions, and is better than other large developing countries such as China and India. Savings and investment rates are also in the mid-range, but interest rates are high which constrains the financing of enterprises. On protection of intellectual property, the legislative regime in Brazil appears to be above average. However, there are significant issues regarding their implementation and enforcement. Upgrading Education and Skills 5.82 Another feature of Brazil's knowledge structure is the low educational attainment and skill level of the population, especially at the secondary level. On the other hand, Brazilian enrollment rates in tertiary education are satisfactory when compared to that of large developing countries. Brazil's great strength is the flexibility of its people to adjust to new challenges. However, without a strong education and skill base, the effects of this asset will be limited. The country thus faces a major challenge in furthering the reforms already undertaken and in developing an efficient and equitable secondary and higher education strategy. Improving the Innovation System 5.83 Brazil could improve its innovation system to take advantage of the tremendous entrepreneurial spirit of its population. It spends 0.8 percent of GDP on R&D, average for middle- 143 income countries but lower than the average of 2.2 percent for OECD countries. The output of the domestic R&D system is low when measured in terms of patents or technical publications, compared to other countries in the region. The export performance in R&D intensive sectors is deteriorating. There are some large govemment programs such as aeronautics, deep-water oil exploration, tropical agriculture, and biotechnology, that are distinguished by world-class technology development. By and large, private firms have not developed or adapted knowledge intensively, despite the strong entrepreneurial capability in Brazil. 5.84 Thus while Brazil could improve the management and efficiency of its R&D effort, the main issues seem to link back to the broader policy regime (trade, competition etc.) that does not put sufficient pressure on firms to keep up with rapidly evolving global practices and new knowledge. Efforts may also be needed in the near future to attract high quality FDI to help access global knowledge. (Until the 1990s, Brazil was the largest FDI host among developing countries, although its position has been eroded recently by China). For this, as well as to stimulate domestic innovation efforts, the Intellectual Property Rights regime could be improved and better enforced. Measures to stimulate university-industry interactions would also improve the innovation climate. Expanding the Information Infrastructure 5.85 Brazil has made considerable progress in developing its information infrastructure in recent years, with the privatization of Telebras and the liberalization strategy orchestrated by the National Telecommunications Agency (Anatel. 'Teledensity" remains lower than in advanced economies. Brazil also has lower intemet host levels than Argentina or Mexico, and computer penetration is also low. The Internet is thus still out of reach for the majority of the population--less than 10 percent of the population have access to it-owing to a series of factors: lack of infrastructure, high prices for connection, critical services and equipment, and low educational levels. Promoting universal access to telecommunications-through further infrastructure development and tariff lowering-requires appropriate deregulation, intensified competition, and renewed attraction of foreign investors. At the same time, the ICT skills of the population could also be developed. References Dahlman, Carl and Jean-Eric Aubert (2001), "China and the Knowledge Economy: Seizing the 21w' Century," Washington, DC: World Bank. FIAS (2001), Legal, Policy and Administrative Barriers to Investment in Brazil, Vol. ll. Frischtak, Claudio (2000), "Using Knowledge for Development: The Brazilian Experience." Horta, M.H. and C.F. de Souza (June 2000), "A Inser,co das Exporta,6es Brasileiras: Analise Setorial no Periodo 1980/96," IPEA, Texto Para Discussao no. 736, Rio de Janeiro. IMD (2000), "The World Competitiveness Yearbook," Lausanne, Switzerland. Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobaton, "Govemance Matters," The World Bank, October 1999. OECD (1997), "National Innovation Systems," free brochure, Paris: OECD. OECD (1999), "Technology, Productivity and Job Creation-Best Policy Practices," Paris: OECD. 144 Quadros, Rui et al, 2001, "Technological Innovation in Brazilian Industry: An assessment Based on the Sao Paulo Innovation Survey, Intemational Joumal of Technological Forecasting and Social Change, vol. 67, no. 2, (May). World Bank (1999), 'World Development Report 1998/99: Knowledge for Development," Washington, DC: World Bank. World Bank (2001, a), 'World Development Indicators," Washington, DC: World Bank World Bank (2001, b) Higher Education in Brazil: Challenges and Options, Latin America and the Caribbean Region World Bank (2001,c), Global Economic Prospects 2002, Draft (09/07/01) Annex 1A: Knowledge Assessment Methodology WBI's program on Knowledge for Development uses a knowledge assessment methodology (KAM) which consists of a set of 66 structural and qualitative variables that benchmark how an economy compares with its neighbors, competitors, or countries it wishes to emulate. It helps to identify the problems and opportunities that a country faces, and where it may need to focus policy attention or future investments. The comparison for the 66 variables is undertaken for a group of 92 countries, which includes most of the developed OECD economies and about 60 developing economies. A list of these countries is presented in Annex 1 B. The set of 66 variables serve as proxies for the four areas that are critical in the development of a knowledge-based economy, namely: * An economic and institutional regime to provide incentives for the efficient use of existing and new knowledge and the flourishing of entrepreneurship. * An educated and skilled population to create, share and use knowledge. * A dynamic information infrastructure to facilitate the effective communication, dissemination, and processing of information. * An efficient innovation system of firms, research centers, universities, consultants and other organizations to tap into the growing stock of global knowledge, assimilate and adapt it to local needs, and create new technology. In addition, the KAM includes several variables that track the overall performance of the economy. These variables help to illustrate how well an economy is actually using knowledge for its overall economic and social development. The team has also developed an interactive Intemet-based KAM exercise which can facilitate comparison of a particular country with (a) all countries in the sample (in this case 92 countries); (b) all countries within a particular range of human development; and (c) other countries in the Region. In addition, regional comparisons can also be undertaken to see how a particular region compares to another. The exercise also includes a "help" feature (on the upper right corner of the screen) that is useful in navigating through the exercise (http://wwwl.worldbank.org/gdln/kam.htm). 145 Knowledge Assessment Scorecards and the Interactive KAM As a large set of variables is unwieldy, simplified uknowledge assessment scorecards" consisting of 15 variables have been developed. In the interactive KAM exercise, the 15 variable scorecards are referred to as the "standard" scorecards. However, participants are not just limited to the 15 variables that are used in the standard scorecards, but can choose other variables from the 66- variable data set and create their own scorecards. Annex 1C presents a list of variables used in the standard scorecards; Annex ID presents the rest of the variables used in the KAM exercise. Annex 1 E presents the standard 15-variable scorecards; Annex 1 F presents the scorecards by each pillar of the knowledge economy. Annex 1 G provides the data used for the scorecards. The 15-variable standard scorecards presented in Annex 1 E represent the four pillars of the knowledge-based economy (economic incentive and institutional regime, education, innovation system, and information infrastructure), and include some performance variables as well. Thus, these scorecards attempt to capture the essence of a country's preparedness for the knowledge- based economy. Each of the variables used in the scorecards is normalized on a scale of zero to 10, so that the highest value is rated 10 and the lowest, zero. An economy should not necessarily aim for a score of 10 on all variables. Some variables reflect performance; others reflect trade-offs that characterize different development strategies. Still others reflect the particular structural characteristics of an economy. The normalized variables are put on star diagrams to graphically illustrate and facilitate comparisons among countries. 146 Annex 1 B: Country Selection G7 Europe East Asia South Asia Transition Economies Canada Denmark China Bangladesh Belarus France Finland Hong Kong Bhutan Bulgaria Germany Ireland Indonesia India Czech Republic Italy Netherlands Korea Nepal Estonia Japan Norway Malaysia Pakistan Hungary UK Spain Mongolia Sri Lanka Kazakhstan USA Sweden Philippines Latvia Singapore Lithuania Taiwan Poland Thailand Romania Vietnam Russia Slovakia Slovenia Turkey Ukraine Uzbekistan 7 7 11 6 16 Latin America and Middle East North Africa Sub-Saharan Africa Caribbean Bahrain Algeria Botswana Argentina Egypt Morocco Cameroon Barbados Iran Tunisia Djibouti Bolivia Jordan Eritrea Brazil Kuwait Ethiopia Chile Lebanon Ghana Colombia Oman Kenya Costa Rica Qatar Madagascar Dominican Rep. Saudi Arabia Mauritania El Salvador Syria Mauritius Jamaica UAE Mozambique M6xico Yemen Nigeria Paraguay South Africa Peru Tanzania Uruguay Uganda Zimbabwe 14 12 3 16 Total: 92 countries 147 Annex IC: Variables Used in the Standard 15-variable Scorecards Performance Indicators 1. Average annual GDP growth 1990-99 (%) (World Development Indicators, 2001) 2. Human development index 1999 (Human Development Report, UNDP, 2001) Economic Incentives 3. Gross capital formation as a % of GDP (Annual Average Growth, 1990-1999) (World Development Indicators, 2001) 4. Tariff and non-tariff barriers 2001 (Heritage Foundation, 2001) Institutional Regime 5. Rule of law (World Bank Institute, 1999) 6. Control of corruption (World Bank Institute, 1999) Innovation System 10. FDI as percentage of GDP 1990-99 (SIMA database 2001) 11. Total expenditure for R&D as a percentage of GNI 1987-1997 (World Development Indicators, 2001) 12. High technology exports as a percentage of manufactured exports 1999 (World Development Indicators, 2001) Education and Human Resources 7. Adult literacy rate (% age 15 and above) 1999 (Human Development Report, UNDP, 2001) 8. Secondary enrolment 1997 (World Development Indicators, 2001) 9. Tertiary enrolment 1997 (World Development Indicators, 2001) Information Infrastructure 13. Telephone per 1,000 persons, 1999 (telephone mainlines + mobile phones) (International Telecommunication Union, 2000) 14. Computers per 1,000 persons, 1999 (International Telecommunication Union, 2000) 15. Internet hosts per 10,000 persons, 2000 (International Telecommunication Union, 2000) 148 Annex ID: Other Variables Available in the KAM Performance Indicators 1. Gender development index 1999 (Human Development Report, UNDP, 2001) 2. Poverty index 1999 (Human Development Report, UNDP, 2001) 3. Composite ICRG risk rating 2000 (World Development Indicators, 2001) 4. Unemployment rate, % of total labor force 1996-98 (World Development Indicators, 2001) 5. Productivity growth (% change of GDP per person employed) 2000 (IMD World Competitiveness Yearbook, 2001) Economic Incentives 6. Overall central government budget deficit as % of GDP, 1998 (World Development Indicators, 2001) 7. Trade as % of GDP, 1999 (World Development Indicators, 2001) 8. Intellectual Property is well protected (WEF Global Competitiveness Report, 2000) 9. Soundness of banks (WEF Global Competitiveness Report, 2000) 10. Adequate regulations & supervision of financial institutions (IMD World Competitiveness Yearbook, 2001) 11. Local competition (WEF Global Competitiveness Report, 2000) 12. Protection of property rights (WEF Global Competitiveness Report, 2000) Institutional Regime 13. Regulatory framework (WBI, 1999) 14. Government Effectiveness WBI, 1999) 15. Voice and accountability (WBI, 1999) 16. Political stability (WBI, 1999) 17. Press freedom 2001 (Freedom House, 2001) Innovation System 18. Technology Assessment Index (Human Development Report, UNDP, 2001) 19. Royalty and license fees payments (millions) (1999) (World Development Indicators, 2001) 20. Scientists and engineers in R&D per million 1987-97 (World Development Indicators, 2001) 21. Research collaboration between companies and universities (WEF Global Competitiveness Report, 2000) 22. Entrepreneurship among managers (IMD World Competitiveness Yearbook, 2001) 23. Easy to start a new business (WEF Global Competitiveness Report, 2000) 24. Availability of venture capital (WEF Global Competitiveness Report, 2000) 25. Number of technical papers per million people 1997 (World Development Indicators, 2001) 26. Patent Applications granted by the USPTO (per million pop.) 2000 (USPTO) 27. Private sector spending on R&D (WEF Global Competitiveness Report, 2000) 149 Human Resources 28. Primary Pupil-teacher ratio, pupils per teacher,1998 (2001 SIMA database) 29. Life expectancy at birth, years, 1999 (2001 SIMA database) 30. Management/worker relations (WEF Global Competitiveness Report, 2000) 31. Flexibility of people to adapt to new challenges (IMD World Competitiveness Yearbook, 2001) 32. Public spending on education as % of GDP 1999 (World Development Indicators, 2001) 33. Professional and technical workers as % of the labor force 1999 (ILO, 2000) 34. 8th grade achievement in mathematics (TIMMS 1999) 35. 8th grade achievement in science (TIMMS 1999) 36. National culture is open to foreign influence (IMD World Competitiveness Yearbook, 2001) 37. Companies invest heavily to attract, motivate and retain staff (WEF Global Competitiveness Report, 2000) 38. Management education is locally available in first class business schools (WEF Global Competitiveness Report, 2000) 39. Well educated people do not emigrate abroad (IMD Wodd Competitiveness Yearbook, 2001) 40. University education meets the needs of a competitive economy (IMD World Competitiveness Yearbook, 2001) Information Infrastructure 41. Telephones per 1,000 people, 1999 (Intemational Telecommunication Union, 2000) 42. Mobile phones per 1,000 people, 1999 (Intemational Telecommunication Union, 2000) 43. TV Sets per 1,000 people, 1999 (World Development Indicators, 2001) 44. Radios per 1,000 people, 1999 (World Development Indicators, 2001) 45. Daily newspapers per 1,000 people, 1996 (World Development Indicators, 2001) 46. Investment in telecom as % of GDP 1998 (IMD World Competitiveness Yearbook, 2001) 47. Rating of computer processing power as % of total worldwide MIPS 1998 (IMD World Competitiveness Yearbook, 2001) 48. Intemational telecommunications: cost of call to US in $ per 3 minutes, 1999 (World Development Indicators, 2001) 49. Information Society Index 2000 (IDC 2000) 50. Percentage of Companies that use the Intemet for electronic commerce (WEF Global Competitiveness Report, 2000) 51. ICT Expenditures as a % of GDP 1999 (World Development Indicators, 2001) 150 Annex 1 E: Knowledge Assessment Scorecards Standard 15-variable Scorecards Argentina Average Annual GDP growth 1990-99 (%) (2001 WDI -, I Human Development Index 1999 (2001 UNDP) [Inl Internet hosts per 10, 0 people, 200,(2~ ITU), ,- 10 -, Gross Capital Formation as % of GDP (Average) 1990-1999 [In] Computers per 1,000 persons 1999 (2000 ITU) - \ Tariff & nontariff barriers 2001 (2001 HerItage Fdn) [ln) Telephone per 1000 people 1999 (telephone - Rule of Law (1 999 WEI) Tertiary enrolment 1997(2001IRI) 'V , l' Control of Corruption (1 999 WBI) Secondary Enrolblent 1997(2001 W?)". ".FDI as % of GDP 1990-1 999 (2001 SIMA) Adut Iteracy rate (% age I5 and above) 1999 (2001 UNDP) - Ttal axpendture for R&D as % of GN, 1987-1997(2001 WDI) High-Technology exports as % of manufactured exports, 1999 (2001 WDI) BrazU Average Annual GDP growlh 1 990-99 (%)2001 W -- _Human Developmert Index 19 (2001 UNDP) [lnj Internet hosts per 10,000 people, 2000, (2000 ITU)- 10 Gross Capial Formation as%ot ODP(Average)1990-1 999 [n] Computers per 1,000 persons 1999 (2000 ITU)/ : Tariff & nontarift barriers 2001 (2001 Heriage Fdn) [In] Telephone per 1000 people 1999 (teephone l-- - -jRule of Law (1 999 VW) Tertiary enrolhient 1 997 (2001 WD)I - -y Cortrol of Corruption (1 999 WE?) Secondary Enrollment 1997 (2001 WDI) rFDI as % of GODP 1990-1999 (2001 SIMA) Adult lIteracy rate (% age 15 and above) 19 (2001 UN Total expenditure for R&D as % of GM, 1987-1997 (2001 WDI) High-Technology exports as % of manufactured exports, 1999 (2001 WDI) Chile Average Annual GDP growth 1990-99 (%) (2001 WDQ, Human Development Index 1999 (2001 UNDP) [ln] Internet hosts per 10,000 people, 2000, (2000 ITU), ,Gross Capial Formation as % of GP ( Average) 1990-1999 In] Computers per 1,000 persons 1999 (2000Tu) t - Tarilt & nortarif barriers 2001 (2001 Heritage Fdn) [In) Telephone per 1000 people 1999 (telphone '. - - - j Rule of Law (1 999 WE?) Tertiary enrollmert 1997 (2001 WI) Cortrol of Corrupton (1 999 WBI) Secondary Enrollmert 1997 (2001 WDI) ' 'P,FDI as of GOP 1990-1 999 (2001 SIMA) Adult rteracy rate (% age 15 and above) 1999 (2001 UNDP) Total expenditure for R&D as % ot GNI, 1987-1997(2001 WMI)) High-Technology exports as % of manufactured exports, 1999 (2001 WI)? China Average Annual GDOP growth 1990-99 (%) (2001 WDQ 7 Human Developmert Index 1999 (2001 UNDP) In] Irternet hosts per 10,00 people, 2000, (200 ITU),, - G,rossCaptatForretionas%ofGDP(Average)1990-1999 [In] Computers per 1,000 persons 1999 (2000 ITU)( / " Tarlff & nortariff barriers 2001 (2001 Heriage Fdn) [In] Telephone per 1000 people 1999 (telephone... - -)Rule of Law (1999 WE?) Tertiary enrollment 1997 (2001 W?)D Control of Corruption (1 999 WV) Secondary Enrollment 1997 (2001 W)) 'r- PFDI as of GODP 1990-1 999 (2001 SIMA) Adult Iteracy rate (% age 1 5 and above) 1999 (2001 UNDP) Total expenditure for R&D as % of GNI, 1987-1997 (2001 WI) High-Technology exports as % of manufactured exports, 1999 (2001 WDI) 151 India Average Annual GDP growth 1990-99 (1) (2001 WMI) Human Development Index 1999 (2001 UNDP) [ilbnte)net hosts per 100 people, 2000, (2000 ITU) i0 . Gross Capial Formation as % of GDP ( Average) 1990-1999 [n) Computers per 1,000 persons 1999 (2000 ITU) 'Tariff 8 nontariff barriers 2001 (2001 Heritage Fdn) [In] Telephone per 1 000 people 1999 (telephone..( I1 i~ Rule of Law (1 999 MDI Tertiary enrollment 1997 (2001 WMI) Cordrol of Corruption (1999 W8I) Secondary Enrollment 1997 (2001 WDI) FDI as % of GDP 1990-1999 (2001 SIMA) AduK literacy rate (% age 1 S and above) 1999 (2001 UNDP) Total expenditure for R&D as % of ONI, 1987-1997 (2001 WMIl) High-Technology exports as % of manufactured exports, 1999 (2001 WI) Mexico Averag Annual GDP growth 1990-99 (%) (2001 WVDI, Human Development Index 1999 (2001 UNDP) [in] Intemethosts per 10,000 people, 2000, (2000 ITU) ": 0J -;, Gross Capital Formation as % of GDP ( Average) 1990-1999 [In] Computers per 1,000 persons 1999 (2000 ITU) / % ,- TarIff 8 nontariff barriers 2001 (2001 Heritage Fdn) (in] Telephone per 1000 people 1999 (telephone ... / CRoL,t Rule of Law (1999 WBI) Tertiary enrollment 1997 (2001WDI) k / ; Control of Corruption (1999 WEI) Secondary Enrogment 1997 (2001 WOO V \ Fl as # of GDP 1990-1999 (2001 SIMA) AduK literacy rate (% age IS and above) 1999 (2001 Tatal expenditure for R&D as % of GNI, 1987-1997 (2001 WMIl) High-Technology exports as % of manufactured exports, 1999 (2001 WDI) USA . Avere Annual GDP growth 1990-99 (%) (2001 WMI) Human Development Index 1999 (2001 UNDP) [In] Internet hosts per 10,000 people, 2000, (2000 ITU / ,Gross CapItal Formation as % of GDP ( Average) 1990-1999 [h) Computers per 1,000 persons 1999 (2000 ITU) f . Tariff & nontariff barriers 2001 (200 Heriage Fdn) [In] Telephone per 1000 people 1999 (telephone ... Rule of Law (1999 WEN) Tertiary enrolment 1997 (2001 WVVl) W I ) Control of Corruption (1999 WEI) Secondary Enrollment 1997 (2001 WFD) I / FO as % of GDP 1990-1999 (2001 SIMA) Adk teracy rate (% age 15 and -above) 1999 (20 UND ) - Total expenditure for R&D as % of GM, 1987-1997 (2001 WVl) High-Technology exports as % of manufactured exports, 1999 (2001 WMI) 152 Annex 1 F: More Detailed Scorecards Economic Incentive and Institutional Scorecards Argaotina Intelledual Property Is well protected (2000 WEF) Tariff & nortaitt barrlers 2001 (2001 Hertage Fdr) 0 -- Soudness of banks (2000 WEF) Trade as iS of GODP. 1999 2001 WDM) \ N/ Adequate regulons a supervision of kinancial Instilons Overal Certri Oovt budget delict as %S of OW 1998 / /4- -\Locad competition (2000 WEF) Gross Capital Formation as % of GODP ( Average) P- -- Protecton of Property Rights (2000 WEF) Press freedom 20011 (Freedomn House 2001) ~ 7Regulaory framework (1 999 VYBD Cortrol of CorrupIon (I 999 WED -Rule of Law (1 999 MED Poitical stabilty (1999 WED Government Ef fectveness (1999 WEB) Voice and accounrability (1999 WED Brazil ktelectuel Property is well protected (2000 WEF) Tariff & nortrltfbaniers 2001 (2001 Hertage Fdn) - * '--Soundness of banks (2000 WEF) Trade as % of GODP, 1999 (2001 W)l) - ' - Adequate regultions 8 supervisIon of financl bittutlons Overal Certra Govl budget defll as % of OW 1998 .. r Local competition (2000 VWf) Gross Capital Formation as S of GOP ( Average) . . .f- Protection of Property RIghts (2000 WVEF) Press freedom 2001 (Freedom House 2001) " ' " "Regulatorytramework(1999 WI) Control of Corruption (1999 WS) I - Rtue of Law (1 999 WI) Political staIEty (1999 WV )i Governmert Effectiveness (1999 WB) Voice and accourtabtliy (1999 WB) Chi0a teUectual Property is wet proected (2000 WEF) Tarill 8 norarttl barriers 2001 (2001 Hertage Fdn) -- Soundness of banks (2000 WV) Trade as # of GOP, 1999(2001 - Adequate regulations 8 superion of financia Institons Overd Certra Gov't budget deficit as iS of ODP 1998 (.. ODI P |; \ Local competitIon (2000 WEF) Gross Catal Formen as S of ODP ( Averaqe)...( - ) Protection of Property Rlghts (2000 WEF) Press freedom 2001 (Freedom House 2001) '; / /Regulatory framework (1999 WM) Cortro of Corruption (1999 WED *- Rule of Law (1999 WB) PoEtcd stabity (1999 Et) Goverrnment Etfectiveness (1999 WBI) Voice and accountablity (1999 WED Ch8na irteedual Property is wel protected (2000 WVEF) Tarwit & nortariff barriers 2001 (2001 Heritage Fdn)- - Soundness of banks (2000 WEF) Trade as iS of GODP 1999 (2001 WED, \ ./ Adequate regulations 8 supervision of financl Istiuton Overal Centrl Govt budget deftit as S of GODP 1998 t x S\Locat compettion (2000 WEF) Gross Capidl Formaion as % of ODP (Average) ... ! J Protecton of Property RIghts (2000 WEF) Press freedom 2001 (Freedom House 2001)\ ;Regulatory frarework (1999 WE) Contrd of Corrupiun (1999fWElL - .j \ u otLa w (1999 WB0 Polcal staty (1999 WEI) Government EFfct-iveness (1999 WED) Voice and accountabllty (1999 WBI) 153 Intelectuad Property is wel proected (2000 WF) Tariff £ nordariff barriers 2001 (2001 Herkage Fdn) - Soundness of baks (2000 WF) Trade as % of ODP, 1999 (2001 UOI) . E | X s * Adequate regulations & supervsion of fnancatW hstlto Overd Cerirl Govt budget defid es % of GDP, 1998 ... ,\Local competdtion (2000 WH) Gross Capial Formation as % of GDP (Average) . . . ( -- - Protection of Property Rigtts (2000 Wv Press freedom 2001 (Freedom House 2001 ) \ Regulatory framework (1999 %AS Control of Corruptlon (1999 WB) '-Y Rue of Law (1o999 WBI) Poiica stably (1999 W - - Government Efectiess (11999 1S) Voice and accouretaby (1999 Wet) Mexico helctual Property Is wet proteded (2000 WEF) Twiff & nortafiff barriers 2001 (2001 Heriage Fdn) 10 ' Soundness of banks (2000 WVF) Trade as S of GDP, i1999 (2001 WDI) ,c '; / Adequate regulations & supervision of fnanchl hstlituons Overel Certrel Govt budget defcit as % of GDP, 1 998m L\ocal cmpetiton (2000 WU) Gross Capital Formation as % of GDP ( Average) ... Proteion of Property Rigtts (2000 EF) Press freedom 2001 (Freedom House 2001 ) ' 'Regutory amework (1999 VAB? Cortrolof Corruption (1 999E) '.I RueofLaw(1999 Val) Potical stabilly (1999 WE?) Government Effectiveness (1999 VE?) Voice and accourtabilty (1 999 VWE?) USA trtelectual Property Is waeD protected (2000 WEF) Tarifft nortarlff barriers 2001 (2001 Heriage Fdn). - Soundness of bats (2000 WM) Trade as % o1 GDP, 1999 (2001 , *.Adequate regultions & supervtsion of fina tidutlons Overdl Central Govt budget defid as % of GDP, 1998 .. . Lo ml cpetition (2000 WV) Gross Cptl Formaion as 16 of GOP ( Average) . . Protecion of Property Rights (2000 WVF) Press freedom 2001 (Freedom House 2001)', ' / tRegulatory framework (1999 WE?) Control of Corruption (1999 WB0 - Rul of Law (1999 WB0 Poiticad stably (1999 WE?) - - Governmet EffecIveness (1999 WEE) Voice and accountabIiy (1999 WE0 154 Innovation System Scorecards Argentina Entrepreneurship among Managers (2001 iMD) Research collboration between collparnO5 and urnversitles:\ 0 ' Easy to start a new business (2000 WE) On] Scientists and egineers In R&D per million 1987-97 y ' ity of Verture cap/te (2000 WEE) Toal expenditre for R&D as % of ONI 1987-1997 ( [On] Number of technical papers per million peopie 1997 [inl Royaiy and iicense fees payments S millions 1999 1[N''' [In] Patent applications granted by the USPTO 2000 (per FDI as % of GDP 1990-1999 (2001 SIMA) | High-Technoiogy exports as % of manufactured exports, Technology Assessment index (2001 UNDP) Private sector spending on R&D (2000 WEE) Brazll Entrepreneurship among Managers (2001 IMD) Research collaboration between companies and universties . --' -- ;. Easyto start a new business (2000 WEF) [inl Scentists and engineers In R&D per millIon 1987-97 . A / Availablty of Venture captal (2000 WEE) Toal expendture for R&D as % of ONI 1987-1997 ( O On] Number of technidcal papers per miiion people 1997 ln] Royaiy and ilcense fees payments 5 millions 1 93 '' n] Patent applications granted by the USPTO 2000 (per FDI as % of GDP 1990-1999(2001 SiMA) / lgh-Technology exports as % of manufactured exports, Technology Assessment Index (2001 UNDP) Private sector spending on R&D (2000 WEF) Chile Entrepreneurship among Managers (2001 iMD) Research colaboration between companies and universties s , Easy to start a new business (2000 WEV ) [in] Sciertsts and engineers In R&D per million 1987-97 ..Availbility of Venture capil (2000 WEE) Total expenditure for R&D as%of ON,1987-1997 ( - On] Number of technical papers per miron peopie 1997... [n] Royaty and lcense fees payments S milDons 1999 ¼ (In] Patent applications granted by the USPTO 2000 (per FDI as % of GDP 1990-1999 (2t1 SIMA) -" High-Technology exports as % of manumfacued exports, Technology Assessment Index (2001 UNDP) Private sector spenrdng on R&D (2000 WEF) China Entrepreneurship among Managers (2001 IMD) Research collaboratlon between companies and universiies ; 10 ;fasy to start a new business (2000 WEF) [In] Scientists and engineers In R&D per millin 1987-97 / Aviablity of Verdure capital (2000 WEE) Total expendture for R&D as % of ON, 1987-1997 o / technical papers per million people 1997 . (in] Royalty and license fees payments 5 millions 1 ; applicat gans darded by the USPTO 2000 (per FDI as % of GD 1990-1999 (2001 SilvA) o ] \ ..>tgh-Tehowgy exports as # of manufactured exports, Technoiogy Assessmernrt ndex (2001 UNDP) Private secor spending on R&D (2000 WEF) 155 hda Erdrepreneurship mong Managers (2001 NiD) Research cotaboration between companies and universlies , -' O-T Easy to start a new business (2000 WEF) [inl Sdentists and enneers In R&D per milon 1987-97 ...( /. \Avaiiebiy ot Verture captal (2000 WEF) Total expendture for R&D as % of ONI, 1987-1997 ... ( [In] Number of technica papers per miion people 1997 [in] Royaety and lcense fees paymenrds $ millions 1999 ... ,I Y .; n] Paternt aplcations grarded by the USPTO 2000 (per FDt as % of GODP 1990-1999(2001 SIMA) j - - Hgh-Technology exports as 6 of manufactured exports,... Technology Assessmerd Index (2001 UNDP) Private sector spending on R&D (2000 WEF) Meaco Entrepreneurship among Managers (2001 MD) Research colaboration between companies and universitles ... Easy to start a new business (2000 WEM) (in] Scientists and angineers In R&D per milflon 1987-97 ..c . Avalabdy of Verture capital (2000 WEF) Total expendiure for R&D as % of ON, 1987-1997 W . -inl Number of technical papers per mcilon people 1997 [In] Royaty and Ocense fees payments $ mllions 1999... V [in] Paterd applications granded by the USPTO 2000 (per ED as of GDP 1990-1999(2001 SC4A) - / ' Hgh-Technology exports as of manuftatured exports, Technology Assessmetrd hex (2001 UNDP) Private sector spendig on R&D (2000 AEF) USA Ertrepreneurship among Managers (2001 NMO) Research collaboradIon between compaies and universties - - Esy to start a new business (2000 WE) [inl Scentitds and engheers In R&D per muCon 1987-97 . Avalabiity of Venture capdal (2000 WEF) Total expendhtre for R&D as % of Ni, 1987-1997 ( ¶ [i]h Number of technical papers per millon people 1997 [In] Royalty end lcense fees payments $ miions 1999 e . [in] Paterd applications granded by the USPTO 2000 (per FDOes % of GOP 1990-1999 (2001 SIMA) exports as ¶ of manufactured exports, Technology Assessmerd Index (2001 UNDP) Prvate sector sperndig on R&D (2000 WEF) 156 Education Scorecards Argentina Public spending on education as % of GDP, 1999 ODI 2001,_ ., Professional and technical workers as % of the labor force (2000 Flexibiity of people to adapt to new challenges (200 IMD) 10 O8th grade achievement in mathematics (TIMSS 1999) Managemeritworker relations (2000 WE-) -' ..8th grade achievement in science (TIMSS 1999) Uta expectancy at birth, years,1999 (2001SIMA) ,4 - - National cuture is open to foreign influence (2001 IMD) Primary Pupiteacher ratio, pupils per teacher, 1998 L 'Companies invest heaviyto attract, motivate, and retain Tertiary enrolment 1997 (2001 Mnen education is locally avaalabie In first-class Secondary Enrollment 1997 (2001 WI Well educated people do not emigrate abroad (2001 IMD) Adut literacy rate (% age 15 and above) 1999 (2001 UNDP) University education meets the needs of a competitive economy Brazil Pubic spendng on education as % of GDP, 1999 M Dl 2001) - Professional and technial workers as % of the labor force (2000 FiexbIbty of peope to adapt to new challenges (200 MD)IMS i th grade achievement in mathematics (TIMSS 1999) Management/worker relations (2000WEE) / :8th grade achievement in science (TIMSS 1999) Lie expectancy at birth, years,1999 (2001SIMA) . NationalcultureIsopentoforeigntnfluence(2001 IMD) Primary Pupilteacher ratio, pupils per teacher, 1998 ... V / Companies Invest heavily to attract, motivate, and retain Tertiary enrollment 1 997 (2001 WDI) Management education Is localy available in first-dass Secondary Enrollment 1997 (2001 WDI) - .| - Wet educated people do not emigrate abroad (2001 IMD) Adult literacy rate (% age 15 and above) 19S9 (2001 UNDP) University educaion meets the needs of a competitve economy ... Chile Public spending on education as % of GODP, 1999 sWDl 2001)_ Professional and technical workers as % of the labor force (2000 Flexlbulity of peopie to adapt to new challenges (2001MD, 10 77 5th grade achievement in mathematics (TIMSS 1999) Management/worker redaions (2000 WEE)1 , 8th grade achlevement In science (TMSS 1999) Ute expectancy at birth, years,1 999 (2001SMA) ( Naionalculureis open to foreign influence (2001 [MD) Primary Pupil-teacher ratio, pupils per teacher, 1998 V CompaniesInest heavily to attrad, motivate, and retain Tertiary enrollment 1997(2001 WDI) Management education is localy available in first-dass Secondary Enrolment 1997 (2001 WD0. educated people do not emigrae abroad (2001 MD) Adult lteracy rate (% age 15 and above) 1968(2001 UNDP) University education meas the needs of a compeitbve economy China Public spending on educatdon as % of GDP, 1999 -WOI 2001) Professional end technical workers as % of the labor force (2000 Flexiblity of people to adapt to new challenges (2001 ND),- / -8th grade achievement In mathematics (TIMSS 1999) Management/worker relations (2000 WEF) . 8th grade achievement in science (TIMSS 1999) Life expectancy at birth, years,1 999 (2001 SIMA) National culture is open to foreign influence (2001 IMD) PrmaryPupil-teacherratio,pupilsperteacher, 1998 ... m Civte. ompaniesinesheavilytoatrac,motivae,and retain Tertiary enirollment 1 997 (2001 infOsI)as . Tertby enrintnt997(201WDI *, ,' v \:. ,Management educaionIs locally avaiableinfret-class Secondary Enrolment 1997(2001 WDI) - .| \ Wel educaed peopie do not emigrae abroad (2001 [MD) Adult iteracy rate (% age 15 and above) 1999 (2001 UNOP) University educaion meets the needs o a compettive economy 157 India Public spending on education as % of GDP, 1999 Pr20M pofessional and technical workers as % of the labor force (2000 Flexiblity of people to adapt to new challenges (2D CMD) 0 I >8th grade achievement In mathematics (TIMSS 1999) Managemerttworker reltons (2000 WF) '' 8th grade achievement In science (TlMSS 1999) Lifeexpedancyadbirth,years,1999(2001 S1MA) f > \ ji _ Netionalcutitureisopentoforelgnlnftuence(2001 MD) Primary Pup-teacher radio, pupis per teacher, 1998 - i Comparies invest heavily to attract, motivate, and retain Tertiary enrollment 1997 (2001 i)N; / e/Mnagement educadion is localy available in first-iass Secondary Enrolment 1997(2001 WDi) -I 5Wel educated people do not emigrate abroad (2001 IMD) Adult Iteracy rate (% age 15 and above) 1999(2001 UNDP) Universdiy education meets the needs of a competitive economy Mexico Public spending on education as % of GDP, 1999 tWDl 2001),._ Professional and technical workers as iS of the labor force (200O0 Flexibilty of people to adapt to new challenges (200 aMD) - O' --? 8th grade achievement in mathematics (T115 1999) Managementw'orker relatIons (2000 WEF),- - 8th grade achievemert In science (TiMSS 1999) Life expectancy at birh, years,1 999 (2001 SIMA) r , & - National cuture is open to foreign influence (2001 MD) Primary Pupil-teacher ratio, puplis per teacher, 1 998 Companies invest heavily to attract, motivate, and retain Tertiary enrollment 1997 (2001 WDi) , Manament education is localy avaiable in first-class SecodaryEnrolmenI99 W200 '- V ell educated people do not emnigrate abroad (201 IMD) Adutiteracy rate(% ageanda nabove) i 999(200 P Unrversity education meets the needs of a compett economy USA Pubiic spending on education as S% of GDP, 1999 (W"I 2001) Professional and technial workers as iS of the labor force (2000 Flexibility of peopae to adapt to new chalenges (200 l[MD) \0 / 8th grade achievement in mathemdics (TSfS 1999) Managementiworker relations (2000 WF)F W8ih grade achievement in science (TlMSS 1999) Life expectancy at birth, years,1 999 (2001 SMAA) iu 0 "S 1/ National cuthre is open to foreign inftuence (2001 MD) Primary Pupil-teacher radio, pupils per teacher, 1998 Companies invest heavIly to attract, motivae, and retain Tertiary enrollment 1997 (2001 WDi) ' /\ ' /Management education Is locally available In first-class Secondary Ernrollment 1997(200 WDI * W eii educaed peopie do not emigrate abroad (2001 lD) Aduit lteracy rate (iS age 15 and above) 1999 (200- UNP) University educaion meets the needs of a competitive economy 158 Information Infrastructure Scorecards Argentia [In] Deity newspapers per 1,000 people, 1996 (2001 WDI iI_>hnvestment mntelecom as% of GOP 1998 (2001 IMD) ln] Radios per 1,000 people, 1999 (2001 W 0 / Rating of compuer processing power as %S of tdal ... [in] TV Sets per 1,000 people, 1999 (2001 WAOO i: [n fOlernet hosts per 10,000peopl, 2000, (2000ITU) [In] Computers per 1,000 persons 1999 (2000 ITt) t - :1 emmatonal telecommuiactlons, cost of call to US ($S3 Pn] Mobile phone per 1,000 people, 1999(2000 TJ) /tformatlon Society Index (DC) 2000 [In] Telephone per 1,000 people, 1999(2000 ITU) / E-Commerce (2000 WEF) ln] Telephone per 1000 people 1999 (telephone malnlnes+moble jCT Expendiure as % of GODP 1999 (2001 WMD Brazil [in] Daly newspapers per 1,000 people, 1 9S (2001 wV. - ,Investment In teecom as % of GODP 1998 (2001 MD) [In] Radiosper1,000people,1999(2001 WI') 0 1 71 ^)RatIngocomptAerprocessingpoweras%ofttal... [In] TV Sets per 1,000 people, 1999 (2001 WI)m f. -. p1] I nternet hosts per 10,000 people, 2000, (2000 ITU) [In] Computers per 1,000 persons 1999 (2000 ITU) ( -- ) -- fr) Internationd telecommunications, cost of caD to US ($13 Vn] Mobile phone per 1,000 people, 1999 (2000 ITU) f; 19v 'InformatIon Society Index (lDC) 2000 [In] Telephone per 1,000 people, 1999 (2000 ITU) ' I E-Commerce (2000 WEF) [On] Telephone per 1000 people 1999 (telephone mannes.mobte I iCT Expendture as % of GOP 1999 (2001 WOD Chile [In] Daily newspapers per 1,000 people, 1Y96 (2001 WDIl Investment In telecom as % of GO 1998 (2001 IMD) [In] Radios per 1,000 people,1999 (2001 WD(,D 0 |°| ,.Rat ot compiter processing power as S oft ta ... [in] TV Sets per 1,000 people, 1999 (2001 W0.; Inlernet hosts per 10,000 people, 2000, (2000 MU) In] Computers per 1,000 persons 1999 (2000 ITU) \-,- ernatlonal telecornications, cost of cal to US ($13. [In] MobBe phone per 1,000 people, 1999 (2000 ITU) ormaton Society Index (DC) 2000 [In] Telephone per 1,000 people, 1999 (2000 ITU)* .. - "IE-Commerce (2000 WEF) [In] Telephone per 1000 people 1999 (telephone malI;lnes+moble ... ICT Expenditure as % of GODP 1999 (2001 WDN China [in] Daily newspapers per 1000 people, 1996 (2001 WMg_ hvestment In telecom as iS of GOP 1998 (2001 IMD) [InlRadiosper1,000people,1999(2001WDl 0 ,Ratlngofcomputerprocessingpoweras %oftotad [In] TV Sets per 1,000 people, 1999 (2001 WOO -- [in] Internet hosts per 10,000 people, 2000, (2000 ITU) [n] Computers per 1,000 persons 1999 (2000 ITU) Interational telecontmeuncatloss, cost of cl to LlS ($/3. [In] Mobile phone per 1,000 people, 1999 (2000 M,U) Informaton Society Index (DC) 2000 [in] Telphone per 1,000 people, 1999 (2000 ITU) 1 .i- E-Commerce (2000 WEM) [In] Telephone per 1000 people 1999 (telphone malnes+mobile ... ICT Expendture as iS of GDP 1999 (2001 WOOI) 159 India [n] Daily newspapers per 1,000 people,1996 (2001 W - Investment ntelecom as% of ODP 1998(2001 MD) [in] Radios per 1,000 people,1999 (2001 WO, ,. j° 7 -,Rating of computer processIng power as % of total ... [In] TV Sets per 1,000 people, 1999 (20 0 1 V) t, - ertet hosts per 10,000 people, 2000, (2000 ITU) [In] Computers per 1,000 persons 1999 (2000 ITU) ( 4 rdemtonal telecommunications, cost of cell to US (/3. l iMoble phone per 1,000 people, 1999 (2000 ITU) t; , / \ , I Intornation Society Index (IDC) 2000 [In] Telephone per 1,000 people, 1999 (2000 ITU) . ' E-Commerce (2000 W ) [In] Telephone per 1000 people 1999 (telephone malnrines+mobile ... CT Expendrure as % of GD 1999 (2001 WDI) Mexico [Inl Daily newspapers per 1,000 people, 1996 (2001 WDI) Investment In telecom as %S of GDP 1998 (2001 MD) [Iln Radios per 1,000 people, 1999 (2001 10I Rating of computer processIng power as iS of totl ... [In] TV Sets per 1,000 people, 1999 (2001 WO)n Internet hosts per 10,000 people, 2000, (2000 ITU) [In] Computers per 1,000 persons 1999 (2000 rTU) 3*tr- International telecommurIcatIons, cost of call to US ($13 [Inl Mobie phone per 1,000 people, 1999 (2000 ITU) '1hformation Society Index CDC) 2000 [In] Telephone per 1,000 people, 1999 (2000 ITU) - E-Commerce (2000 WEF) [In] Teephone per 1000 people 1999 (telephone malnDnes+moble ... ICT Expenditure as S of GOP 1999 (2001 WDI) USA [in] Daily newspapers per 1,000 people, 1996 (2001 WDI) _ Jpvestment In telecom as iS of GOP 1998 (2001 MD) [h] Radios per 1,000 peopie, 1999 (2001 Wo) R0 Rating of computer processing power as iS of total [In] TV Sets per 1,000 people, 1999 (2001 W'A) >. / [h] internet hosts per 10,000 people, 2000, (2000 MU) [n] Computers per 1,000 persons 1999 (2000 ITU) t- -- N -- terntional telecommunications, cost of caD to US ($/3 [In] obile phone per 1,000 people, 1999 (2000 rW '4 ormation Socidy Index (IDC) 2000 [On] Telephone per 1,000 people, 1999 (2000 T) (2000 WEF) [In] Telephone per 1000 people 1999 (telephone masntnas.moblle ... CT Expenditure as S of GOP 1999 (2001 WOO) KsiEldllll at! s iX|t ||||||| cri L !i a 0 h% a Si RI a _ O X ~~1 01 n 11 m o I2 t-iItIXi3; i @ cq ~~~lX s. 6.0 9. 9 e a) ~ ~ ~ ~ ~ !;a'-z-11t ti; 00q-Qq E ~ ~ ~~A --- Iz ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 9~q a) a~~~~~~~~~~V l 0 I m- J111- 1 VM IR I I io 0 a-- A3jiM FIN~a x~~~~~~ . a) o C3 161 KAM Data Corresponding to Innovation Systems (in] Patent ru High- On) graland Tecthl-y [nl) RyWty Sofentah hin Numrber by de exporG as and Doan" TOW of tacihnkol USPTO % of teen exper4biure a nW s t oraton Enbepreneu Eay to Aaabblity Papers pew 2000 (pa rnrsnacn P&bats ector Tacnobgy FOI as % of paymets 5 R&D as % of R&D per bt_e rapmong Sta w ofVentLf mOlon milton ad -ospoes, pendtgon Aaennt GDPP 1990. nilons ONI 1997- nilon 1917 1o anysdara counr Maran bruness I pila rh 199 OD.) 1999 R&D Index 199 1999 1997 97 uwdvasta lArpetna - S.O3 3.8 3.2 4.085 0.994 8 2.9 0.381 2.515 0.094 O.38 8.492 3.4 tszlt 8.357 4 3.1 3.189 10.514 13 29 OS11 1.401 7.158 0.81 5.124 3.9 Ch4t 8.704 4.8 3.8 4.055 0.72 4 2.7 0.357 5.221 SSS1 0.98 8.098 3.2 Clina 5.934 4.2 3.1 2.109 0.122 17 2.9 0.299 4.092 6.670 0.86 6118 4.3 0Indh8 4.782 4 3.6 2247 0.123 8 2.8 0.201 0.392 8.790 0.73 5.004 S.8 Mexico 4.883 3.5 2.9 3.032 0.708 21 2.5 0.589 2.19 8.319 03 5..388 3.8 USA 7.107 599 0.4 8.399 5,98 I 35 5.5 0.733 1.153 .A494 2.83 8.210 44 KAM Data Corresponding to Information Infrastructure T.iphon. p. rD Rg d g o paspt 1999 (in oono r 1w (tolpyhn Tdophona [In Mbi. lb,] [tnl TV p (ing [SI] rI, d Intotnuwtto|| ,no.nh pr 1,000 Phone p. Crrnpid Set p. (I] Refo. [Il Daly paoas % of bt per fIetommwnk,b Irnornation ICT sailo psopba. 3,90 PO 1,000 1.0D0 p.r 1,90 nquponsr Ien_sdh In tow 10.000 eto, at Or Society Experftruras phor_) 1959(29ra PDPI. pa. poptb. p-spt. pr O1.0D to'os 91 edwbIde peopb, 0m9to0USS/S rt IdgODC( ) %Co1OPI Countrka 000ITU ITU) I999 1sw 3999 1999 P.P* ss dof OPs3 PsAP INS 00DDO | dn)bt 9 | 23O E.Co,nera 398 | t n 931 5.29 9 35098 s.e 8.24 4.12 7 . s 3978 2 .8 I 0S I 4.7 S. an, s7d2 5203 10 3552 915 60Ni6 s6s 0.r 1 1 36 11 42 1T _ I s. ziw -.O6.098 s.399 5.410 I545 5.872 4ws , .565 0.15 3.5 2.9 32 1 38.9 Chk.-- - 5.103 4.711 4.01 2.501 5577 0.9 n 1.9i 1.F V = 1 1 18T 1 * itdb &570 37 34 1.253 94 4 317 4.759 io T 0.69 0.278 1 4. 2 7.4 Me1has 99ss 4924 4958 3185 5587 4.78 4575 O.S7 102 1 3.i44 1 3 1 44 I 1 4 USA 6.970 6,551 15W 92 1 - .TS 0412 1 . 2 - T1Y 162 6. HUMAN CAPITAL POLICIES FOR GROWTH The Implications of Theory and Evidence for Brazil Prepared by Indermit S. Gill and Harry A. Patrinos 108 6.1 Pick up any report by intemational development agencies and it is difficult to escape strong pronouncements on the importance of one or another aspect of "human capital"-the generic term used for the skills and energies of humans-especially education. Here is a sample: 'Investment in education is a key to sustained growth, not only because it contributes directly through productivity, but because it also reduces income inequality. '1 'l am not for a moment suggesting that primary education and secondary education are not at the very essence of development ... [but that is] not enough .... If you are going to advance the issue of poverty and development in developing countries .... the key ... is higher education...""1 uHigher education has never been more important to the future of the developing world as it is right now. It cannot guarantee rapid economic development - but sustained progress is impossible without it." "' "Today, more than ever before in human history, the weaith - or poverty - of nations depends on the quality of higher education.' 112 6.2 Some months ago, we set out to study what human capftal policies are most needed in Brazil, especially the issue of the relative emphases on pre-primary, primary, secondary and tertiary education. Based on the last three assertions, we might have reasoned that if higher education is the key for even a low-income country, it definitely should be the highest priority for an upper-middle income country such as Brazil. We might then have checked whether Brazil fares 108 We would like to thank Avani Parekh for research assistance, and Andreas Blom, Norbert Fiess and Mark Thomas for useful inputs and comments. 109 Nancy Birdsall et al. (1997): Chapter 4 in Pathways to Growth: Comparing East Asia and Latin America, Inter- American Development Bank. '11 James Wolfensohn, President of the World Bank, cited in Chapter 5 of World Bank (2001): Constructing Knowledge Societies: New Challenges for Tertiary Education. 1" World Bank and UNESCO (2000): A Report of the Task Force on Higher Education and Society. 112 Malcolm Gillis, President of Rice University, 12 February 1999; quoted in World Bank and UNESCO (2000): A Report of the Task Force on Higher Education and Society. 163 badly relative to other Latin American countries in tertiary enrollment and found that it does; besides, some of these countries-such as Peru and Ecuador-actually are much poorer. We could then have concluded that today, more than ever before, the key for improving the lot of Brazil's people-especially its poor-is increased access to and quality of higher education. 6.3 We would almost certainly have been wrong. What we did instead was to examine the evidence. We examined the literature on economic growth and human capital, and studied the experience of countries such as the US and South Korea that have followed generally effective education and training policies and have also grown rapidly, conditional upon attained per capita income levels or resource endowments. We then juxtaposed the lessons learned from these and other countries against Brazil's current state of affairs-high debt levels that constrain govemment spending, a high level of income inequality, looking to grow its way out of trouble-and came to a different conclusion. We conclude instead that to increase economic growth, the highest priority for Brazil over the next decade may be to rapidly expand secondary education (grades 5-11) of acceptable quality-subsidizing training any more is not likely to help much, and the time for greater subsidies for higher education will come later. This paper explains why we reach these conclusions. What is "Human Capital"? 6.4 For some purposes, human capital can usefully be classified into tangible and intangible components, following the taxonomy adopted by David (2001). Tangible human capital are the more directly observable traits, such as health, longevity and physiological condition such as strength and eyesight. Much of the discussion of the relationship between human capital and growth, however, generally concems intangible human capital whose formation, through study, training and work, is costly. This intangible human capital can be subdivided further into psycho- motor capability, cognitive capability, and procedural capability."3 6.5 During the last half-century the importance of intangible forms of capital has increased relative to tangible forms of capital, both human and physical. In the US, for example, the ratio of intangible to tangible capital stock increased from 0.5 in 1930 to more than 1 by the end of the century (David, 2001). Of the intangible stock, human capital formed through education and training investments is believed to now represent about three-fourths of the total non-tangible stock. We can surmise that intangible human capital stocks are both smaller and relatively less important in middle and low income countries such as Brazil and India, but also that their importance grows rapidly with economic growth. 6.6 This paper therefore is mostly devoted to a discussion of these intangible aspects of human capital. That is, we focus in this paper on education and-to a lesser extent-on post-school investments in human capital. Therefore, this paper shares the bias of much of the existing literature on economic growth in treating human capital as synonymous with the intangible attributes acquired by people through study, training and experience. But this bias does not make this effort hopelessly flawed as a policy discourse on Brazil. One can reasonably make the 113 David (2001) further subdivides procedural capabilities into four categories: flexibility and retrainability, problem- solving skills and leadership, creativity and innovativeness, and social capability. 164 argument that over the last half-century Brazil has done somewhat better in the more tangible areas of health and longevity-though it remains behind on male health and survivability-than it has in fostering the formation of intangible human capital most needed by a middle-income country. What are "Human Capital Policies"? 6.7 Given the breadth of the definition of "human capital", it is difficult to find policies that do not influence the acquisition or formation of human capital, and hence be excluded from the set of "human capital policies". David (2001, page 67) suggests a practical solution, defining human capital policies as 'those which are introduced with the explicit intention of affecting the levels, diversity and relative availability of human capabilities". In this regard, a previously underemphasized point stressed by Heckman (1999), David (2001) and other economists is that the human capital formation effects of the general fiscal system (how taxes are raised, how subsidies are structured) should definitely be considered an important part of human capital policies. The other components are more obvious: government policies to promote health and longevity, measures to foster education and training at all levels and for all age groups, and regulations and interventions to promote acquisition of skills as a by-product of work. We do not directly address government incentives for research and development in this paper. 6.8 The aim of this paper is to identify what a government that faces a fairly tight budget constraint and with similarly limited administrative capabilities can do to ensure that the country's human resources (at least) not act to dampen economic growth and preferably provide a boost to economic growth. We recognize that this is rather confining: there may be good reasons to improve education and health levels in their own right, or for non-economic reasons such as national pride or social welfare, regardless of their growth effects. In this paper, we restrict the discussion to economic reasons for promoting human capital formation mainly for tractability reasons, but also because economic growth is and should be an important policy objective of poorer nations."4 6.9 Our main interest here is therefore on human capital policies that promote growth. This paper is written with Brazil in mind, though much of what we write may also be applicable in other middle-income countries. Being economists, we use the only method we know to address this question, viz., an economic approach. This paper may indeed be summarized as "lessons from the dismal science for human capital policies in Brazil and middle-income countries." The Plan of this Paper and Main Conclusions 6.10 In the sections that follow, we zigzag between the lessons of international experience on the one side and findings for Brazil on the other. In the next section, we briefly describe some human capital indicators in Brazil and the health, education and training systems. Section 3 then summarizes the theoretical characterization of human capital in the economic growth literature, and presents the main findings from empirical investigations using cross-country data. Note here that much of the estimations of the relationship between human capital and growth is correlative in nature, but what is required for policy is evidence on whether and which human capital investments 114 Those who need convincing should read Easterly (2001). 165 lead to higher growth. Section 4 zigs back to a quick look at long-term time-series evidence for Brazil based on a short paper by Fiess and Gill (2001) that attempts to decipher the direction of causation. Section 5 then studies the relevant aspects of education policies in three countries-the US, South Korea and India-during the last century. Based on the findings of the previous three sections. In section 6, pulling together the policy lessons leamed in sections 3-5 and relying on the additional insights provided in David (2001), Heckman (1999) and Heckman and Klenow (1997), we conclude with some policy guidance for fostering human capital formation and economic growth in Brazil. 6.11 Because this is a long paper, we will not withhold the main messages till the end. But we state them here only briefly and perhaps more forcefully than the evidence warrants, in the hope of encouraging or even provoking the reader to take a look at the rest of the paper. Essentially, our trek through the theory, intemational evidence, and assessment of Brazil's condition over the last decade or so leaves us with the following conclusions- * On the benefits of rapidly increasing human capital investments. The evidence that human capital investments rapidly lead to economic growth is weak at best. Put another way, there is little in the way of evidence from cross-country analysis and country studies that increased flows lead to higher growth rates. From a policy point of view, this means that massive investments in education, training, and health would probably not raise economic growth rates quickly. * On the benefits of having large human capital stocks. There is evidence that countries that have grown steadily and impressively have higher human capital stocks. Put another way, there is some evidence that countries with higher stocks of human capital-both tangible and intangible-grow faster. From a policy point of view, these two findings mean that countries should view human capital policies as instruments of long-term growth, not as devices for sparking economic growth spurts. * On Brazil's education levels and growth since the 1950s. Simple econometric tests for Brazil's own experience seem to support these findings: That is, short-term correlations between human capital and economic growth are as weak as long- term associations are strong. * On the importance of investments in health versus education and training. The pattem of human capital indicators for Brazil suggests while Brazil made the "epidemiological transition" during the 1980s, it still has to make what can analogously be termed a "pedagogical transition". Put differently, while Brazil made the transition to middle-income country status in its health-related achievements, it still has education-related indicators (such as secondary and tertiary enrollment ratios and the qualifications of teachers at all levels) that are more likely to be found in low-income countries. While Brazil's health levels can still improve-a lot in the case of adult male health, for example-the emphasis of human capital policies over the next decade should be on education. 166 On the potential for "leapfrogging" in human capital investment. While countries can "leapfrog technologies", the experience of the US in the last century and South Korea during the last fifty years indicates that there is little choice for today's middle and low income countries but to build up human capital stocks in basically the same sequence as earlier developers-investing in primary education first, then in secondary education, and finally in tertiary education. Altemative strategies such as building a stock of workers with higher education (e.g., India's technology initiatives) do not stand up under closer scrutiny. * On investments in tertiary versus lower levels of education. Again, while there is some evidence of knowledge- or R&D-related extemalities (also called spillovers or leaks), there is little proof of growth-related externalities associated with higher education. Middle-income countries such as Brazil that face tight budget constraints should pursue human capital formation aggressively but not in a frenzied manner. It is doubfful that the unavailability of workers with higher education is a constraint on growth, so frenzied investments in colleges and universities are unlikely to unleash growth. Besides, at current levels of graduation from high school, public initiatives to expand higher education do not qualify as non-elitist. * On the importance of secondary school investments. In Brazil, if there was just one thing the policymakers should do, it would be to focus their attention on expanding the access to secondary education of acceptable quality, starting with what we term Junior secondary schooling" (grades 5-10). * On gradually shifting resources from primary to "adjacent" levels. Brazil has achieved remarkable progress in primary school enrollments over the last decade but now- like many middle-income countries-faces declining primary school enrollments because of demographic reasons (the fertility in 2000 was half its 1980 level of 4 births per woman). The challenge for countries such as Brazil where primary and secondary education are sub-national responsibilities is to shift resources from primary to secondary levels. New evidence from industrialized countries- corroborated by studies in Brazil-also points to high economic retums and equity- enhancing effects of pre-primary schooling or early childhood interventions, so some of the resources saved at the primary level can also be transferred to pre- primary interventions. On school versus post-schooling investments in skills. Job-specific training has a high rate of return when employers and workers determine where, when and for what workers should be trained. In sharp contrast to such private training, the rate of return to public training initiatives is essentially zero. To foster training that yields high retums, the two things middle-income country governments could do are: (a) raise education levels of the workforce, since education and private sector training are strongly correlated; and (b) properly regulate labor markets so that private sector incentives are not distorted. 167 6.12 In this section, we provide a brief description of long-term trends in national income and selected aspects of human capital in Brazil, especially education and health. One of the aims here is to identify episodes in the half-century since 1950 that are especially revealing of the relationship between economic growth and human capital formation in Brazil Education 6.13 The Brazilian education system is divided into three levels: fundamental (grades 1-8), intermediate (grades 9-11) and higher. It has expanded quite rapidly over time. In 1964 there were about 10 million students in the system, by 2000 there were more than four times as many. In 1990, there were 4 million preschool students, about 30 million at the elementary level, about 4 million at the (senior) secondary level, and about 2 million at the tertiary level. The notable features of the Brazilian systems are high repetition and dropout at primary and secondary levels, low secondary enrollments (about 40 percent of the eligible population), and an over-subscription to higher education. Table 1 Basic Characteristics of the Brazilian Education System 1 This paper's Grade Length Average No. of Estimated cost Legal terminology years to in RS per student responsibility complete" (year Primary 1-4 4 5.4 517 Ensino Education Municipality Fundamental Lower Secondary 5-8 4 4.9 637 and State Education (shared) Ensino m6dio Upper Secondary 9-11 3 3.7 661 State Education I Superior Tertiary 12-17 4-64 4.5 13,654 Federal Education I_ Source: Blom et al (2001). The above unit costs do not take into account the high rates of dropping out or repetition. Adjusting for these problems would especially increase unit cost for the lower levels of education. The cost for tertiary is calculated only for federal universities, which normally are considered to be the most expensive. 6.14 Brazil spent 5.1 percent of its GDP on education in 1995, considerably more than Argentina (3.5 percent), Mexico (4.9 percent), Korea (3.7 percent), and India (3.2 percent). By 2000, Brazil had increased this ratio to 5.5%. All public education is free, and some private education also receives public support. What stands out is the extraordinarily low priority given in the Brazilian budget to secondary education, and the high priority given to higher education (see Figure below). 168 Figure 6.1 Brazil Spends a Small Fraction of Its Education Budget on Secondary Education 50 140- - . 0 Btazil Mexio A ln K-rex Indla Couitew: LAC Big Thrao. Karco, and IndIa 1OPr3 P1ay & Pvlwy rnSendy OTe,a & 011 6.15 Given the findings in Blom, et al (2001)- reproduced below-that rates of retuum to secondary education have declined over the last two decades while those to higher education have fallen, it may seem that what Brazil most needs is an expansion of higher education funding. We will examine this issue in each of the subsequent sections. Here it should suffice to note that higher private rates of retum to education do not by themselves constitute a case for a reallocation of public subsidies towards higher education, also noting that Blom et al (2001) do not make this assertion. Figure 6.2 Brazil; Returns to Schooling in 1982 and 1998 20% 16%I 12% 0 E 0% Primary Lower Upper Tertiary (4 ysch) secondary (8 secondary (1 5 years) ysch) (11 ysch) Source: Blom, Verner and Holm-Nielsen (2001). Note: The reported coefficient is the OLS-estimate of the return to one completed year of schooling at the indicated level. The wage regressions are based on wage earners only, and only in the six largest metropolitan areas, and include control variables. ysch stands for years of completed schooling 169 Health 6.16 Trends in health status in Brazil are encouraging, though the Brazilian people now are said to "have the misfortune to suffer from both the health problems of the industrialized First World and the preventable diseases of the Third World"."' Children are healthier now than in the past and more of them are getting healthier faster, 'because of improvements in purchasing power, matemal education, utilization of health services, community infrastructure and water supply, and individual behavior." (World Bank, 1998). Despite these recent improvements, sharp regional inequalities persist: child health, access to maternal education, health care and water are poorer in the Northeast, and worst in the rural Northeast. Endemic parasitic diseases continue to threaten rural and remote areas in all parts of Brazil. Basic Indicators 6.17 Brazil has experienced a sharp decline in fertility over the last two decades, which "has in all likelihood reduced the prevalence of the factors associated with risks to childhood health.... Diarrhea among children declined sharply and diseases preventable through vaccination are largely under control in Brazil. An epidemiological transition is said to occur in a country when, as a result of modernization and development, infant mortality and fertility decline, life expectancy increases, and infectious and parasitic diseases are no longer the leading causes of death. As early as 1980, "post-transitional" conditions became the leading causes of death in every region of Brazil. By 1980, cardiovascular disease became, and remains till today, the leading cause of death in each major region and in almost all states."' 6.18 Brazil has also made progress in combating HIV/AIDS. The changes in incidence of AIDS in response to government initiatives shows the benefits of "intangible human capital": data from the Health Ministry indicates that while in 1984, 83 percent of AIDS sufferers were university graduates by 1994 almost 70 percent were illiterate or had at best completed primary school. The Health System 6.19 Brazil's system meets the basic economic axiom of separating the financing from the provision of services." The government health care system, called the Sistema Unico de Saude (SUS), contracts out much of inpatient care and a substantial portion of outpatient services to a network of non-profit and private for-profit hospitals and clinics. The govemment manages and owns only 31 percent of the hospital beds it supports. Despite the efficient structure of the health system, under-financing of the public system is believed to result in regional inequality, arbitrary rationing, and a perceived decline in quality. These are not unusual problems for a middle-income country. 115 Director of Brazil's national center of epidemiology, Dr. Eduardo Levcovitz. 116 These are not diseases of the rich alone. Indeed, both the prevalence of these diseases and of the risk factors associated with them, such as smoking, poor diet and lack of exercise, are generally higher among the poor, who are relatively uneducated and are less likely to be reached by information campaigns. There is also increasing incidence of tuberculosis and leprosy (World Bank, 1998). 170 BrEzil's Basic Health Indicators Have Improved Steadily Since 1960 180 1 160- 140- 120- 60- .40 - - - 1 , ._: _ _ Intant Mortality Under-Five Mortality Crude Birth Rate Cnrde Death Rate Fertility Rate 101960 01970 01980 01990 02000 Figure 6.3 Brazfl's Human CapiaW Scorecard 6.20 It is useful to compare Brazil's human capital performance with other countries in Latin America and the Caribbean. Overall, compared with other countries in Latin America and the Caribbean, Brazil still does poorly in human capital indicators."17 But this underperformance is not uniform, and should be kept in mind while prioritizing investments. o Child (under-five) mortality is marginally worse-Brazil under-performs by about 6 percent compared with the average Latin American country. o Female health indicators (survival to 65 years, and years spent in poor health) are not much worse than the average-these are areas in which Brazil has made considerable progress over the last two decades. o Brazilian males fare worse: longevity and quality of life indicators are about 10 percent worse than for Latin America as a whole. o Brazil has caught up with Latin America in pre-primary and net primary education enrollments. o By far the worst performance is for net secondary school enrollment-Brazil is a full 40 percent behind the average Latin American country. 117 All numbers are from the 2001 World Development Indicators of the World Bank. This comparison is meaningful because Brazil's per capHa income (adjusted for purchasing power) is roughly average for the region. 171 o Brazil also does poorly in higher education, lagging behind other countries by about 13 percent. Brazil Lags Behind Latin American Countries Most in Secondary and- to a lesser degree-in Tertiary Education and Male Health & Longevity .0 0 .5 0 - o t°/ ' f-1 0 0 . .20 0- .25 0 - .30 0 -35 0 - -40 G - _ ;- Incomo USMortalily SorvIcaS165 SurvIco185- PoorHealth. PoorHouith- Pro-Pr,inry Prlmory-N.1 Secondary- T-rlary. M F M F Not Groos Figure 6.4 6.21 A one-line summary of this section would be that while Brazil made the epidemiological transition during the 1980s, it has yet to make the analogous transition in education. The aim of policymakers could then be to implement a set of policy measures that enable Brazil to make this "pedagogical transition" to middle-income country education levels during the next decade. 6.22 If this were a study of human development, the policy priorities would be somewhat obvious. But our aim here is to determine whether and which human capital investments would most likely help increase economic growth rates in Brazil. We examine what theory and the experience of other countries can tell us about the importance of human capital for growth, and the types of human capital investments that will be most effective in increasing growth in a middle income country such as Brazil. We tum to this next. 172 6.23 This section presents the result of a quick search through the heap of conjectures, arguments, hypotheses, models, statistics, and regression estimates generated by economists on the subject of economic growth in the hope of finding something on what all this implies for human capital policy. Compared with the researchers that have contributed to this body of work, who were often attempting to explain worldwide growth patterns, the question is somewhat narrower for us: the policymakers we assumed to be our readers were assumed to be from Brazil, a country that is no longer poor, that has sometimes grown rapidly, but has always lagged behind other middle- income countries in human capital development. What follows is the result of this search, organized to be of some help to policymakers in middle-income developing countries. 6.24 In,this section, then, we first briefly survey the treatment of human capital in theories of economic growth developed in the last half-century, since the seminal papers on economic growth by Swan (1953) and Solow (1953), to the explicit recognition of human capital in influential papers by Uzawa (1965) and Lucas (1988), and more recent work by Romer (1989), Grossman and Helpman (1991), and others linking investments in human capital, R&D, knowledge extemalities, and economic growth. Our aim here is not to comprehensively survey the treatment of human capital in this literature (for that we recommend Topel 1999 and Patrinos 2001). The objective here is to uncover the policy implications, viz., strong clues in the literature for defining a useful role for government. The main premise in the theory is a common sense one-that human capital is an important correlate of growth-and the main policy implication is that because social returns to investments in human capital exceed private returns, there is a case for subsidization of this form of investment because the competitive solution results in under-investment in human capital. 6.25 Later in this section we survey the empirical literature on human capital and growth that has followed these theoretical developments. Again, our aim here is to summarize the policy implications of this literature. This consists of some single-country estimations but mainly of a flood of multi-country growth regressions, which began with Barro (1991), Mankiw, Romer and Weil (1992) and others who found a significant correlation of education with growth. These findings were challenged by Benhabib and Spiegel (1994), Pritchett (1997) and others, and efforts also began to ensure consistency between these cross-country macroeconomic estimations and established within-country microeconomic findings on the rates of return to human capital (see, e.g., Bils and Klenow, 2000). This debate continues unabated even as we write this. But there may be some light at the end of this decade-long tunnel: initiated by Wolff and Gittleman (1993) and others, developed further by Baumol and Wolff (1997) and others, and helpfully summarized by Krueger and Lindahl (1998), Topel (1999), Wolff (2000) and Temple (2001), this line of thinking may be now at a state of knowledge that could yield reliable clues for policy formulation. Human Capital in the Theory of Economic Growth 6.26 Influential ideas often have deceptively humble origins. Modern growth economics had a simple but surprising start. Solow (1953), in an attempt to understand the determinants of US growth, developed a simple model of the economy in which output depended on capital and labor. When he tested how well the model fit the data, however, Solow found that the growth of measured 173 capital and labor accounted for only a fraction of the growth in output. The rest was unexplained or captured by the "residual". The efforts of growth economists for the last fifty years can be summarized as theoretical and statistical attempts to chip away at this residual so that the explanatory power of economic theory is improved to a point that economists are no longer embarrassed by their ignorance about the most basic economic issue: the growth of income (see Easterly, 2001, for an enjoyable and insighfful account of this "quesf'). 6.27 Economists began work on the effort to explain the Solow residual right away. To simplify somewhat, the work basically headed in two directions. The first priority was a relative no-brainer even for economists to determine and so started right away: capital and labor inputs were obviously not as well-measured as output, so better measures had to be found. Perhaps this would explain more or even all of the growth of nations. The quality of labor-not just its quantity measured in hours worked-clearly mattered. More educated or better trained workers produced more, and produced for longer as they became healthier and longer-lived. Human capital had made its entrance. Education played a central role, in large part because it was much easier to measure than other forms of human capital."' So, as we will see in more detail in the next section, education became a favorite of empirical growth economists. 6.28 The second strand was related but more subtle, and hence took longer to gather steam. The main implication of this formulation was that beyond a point, people can not be made better off just by giving them more machines to work with or by making them work harder. This could only be done by organizing production better, so that more could be produced with the same amount of capital, labor and other inputs (such as energy). That is, technological progress had to occur to make economies grow over long periods. The question economists began to grapple with was whether and how technical change could be speeded up. Concepts such as knowledge, innovation, R&D, and patents swirled around this body of work. Knowledge was central here- education obviously mattered too but played a supportive rather than central role. The concepts lent themselves better to theorizing than measurement, and so, not coincidentally, knowledge and innovation became the favorites of theoretical growth economists. 6.29 After Uzawa (1965) who recognized the possibility of technological progress being determined "within the system" rather than being fate, the progress in growth theory essentially halted for two decades. Then Romer (1986), Lucas (1988) and others sparked a revival, motivated in large part by the failure of standard neoclassical growth theory to explain the failure poorer economies to catch up to the richer ones. The main culprit-according to these theorists-was the characterization of technical change in standard growth theory as largely exogenous to the country's own efforts, when in fact it depended on factors that could be influenced or were endogenous. By the early 1990s, "new growth theory" was in full bloom."19 6.30 The neoclassical view of economic growth is that as physical and human capital are accumulated, their incremental contribution to output diminishes. The implication for developing countries is that since they have smaller endowments of physical and human capital, they will grow 118 A decade or so later Becker (1973) and Mincer (1974) would argue that post-school investments in human capital- proxied by years of experience-were perhaps a larger fraction of human capital than education. "9 Comprehensive reviews of new growth theory are provided elsewhere (Amable and Guellec 1992; Brander 1992; Behrman 1990; Verspagen 1992; Sala-i-Martin 1990; Klenow and Rodriquez-Clare 1997; Temple 1999b). 174 faster than rich countries for the same level of investment in physical and human capital assets. Eventually, poor economies will catch up with rich economies and per capita incomes will converge. While a central element of economic growth theories is technological change, basic neo- classical theory assumed that it is exogenous to the economic process (Solow 1956,1957,1970). The newer growth theories that emerged in the late 1980s are differentiated from each other by the accumulated factor which acts as the source of growth, for example: physical capital (with learning by doing or complementarities); technology (research and development); human capital; infrastructure; and public services. Thus, long-run growth is still determined by technological change, but it is endogenously-driven. This in tum may be brought about by a variety of reasons, such as leaming-by-doing (Romer 1986), external effects of human capital formation (Lucas 1988), production externalities of public expenditures (Barro 1990) and quality improvements through the invention of new products (Grossman and Helpman 1991). 6.31 New growth theories underscore the distinction between capital-based models and ideas- based models (Romer 1993a, 1993b). Capital-based models assert that growth is the result of capital accumulation. Ideas-based models give emphasis to a factor that opens up new investment opportunities, alternatively known as innovation, invention or technological change. Ideas drive both growth in income and capital accumulation. The discovery of ideas is treated as being endogenous rather than exogenous in the new theories. Ideas are not the same as physical or human capital. But unlike neo-classical theory, ideas are something other than freely available public goods. In industrialized countries they come about as a result of intentional attempts to make discoveries. Ideas arrive in developing countries as a result of attempts to transfer them, to the extent that this is encouraged by market incentives. 6.32 In an equilibrium growth model of endogenous technological change, long-run growth is driven primarily by the accumulation of knowledge (Romer, 1986). This theory postulates a positive externality in the production of knowledge because knowledge is assumed not to be perfectly patentable or kept secret. The major new assumption of this theory is that knowledge grows without bound. One adaptation adds human capital with extemalities in the sense that the average level of human capital affects a workers productivity in addition to the effect of her own human capital (Lucas 1988). In this sense human capital is "twice blessed;" because of its inherent productivity and because interactions among well-educated people further increases their efficiency.'20 But, in any case, the distinction between education-related and knowledge externalities is an important one for policy and should be kept in mind. 6.33 Other theorists emphasized labor-augmenting spillovers from human capital investments. Once a given level of knowledge is achieved, it is much easier to acquire further knowledge or a sharp change in production possibilities is induced so that there is a critical mass in human capital attainment. As a result of such threshold externalities, countries with high human capital investment relative to their per capita incomes can experience periods of high-sustained growth. 120 In "old" neo-classical theory, the production function is Y = F (K, L), where output (Y) is a function of capital (K) and labor (L). In 'new" neo-classical theory, models have been developed from the following basic formulation: Y = f (T, L, T), where L is a conventional production factor such as labor, T represents the stock of investment in technological change, and the bold T indicates a general volume that is available to all firms in the economy. In Romer (1986), for example, it indicates the sum of all individual Ts, while in Lucas (1988) it is the average level of human capital (Verspagen 1992). 175 Therefore, there could be high payoffs in terms of growth to human capital investments at higher levels (Azariadis and Drazen 1990). Developing countries may experience low growth rates until a crucial (perhaps human capital level) threshold is passed (Tallman and Wang 1992). 6.34 In addition to the role of increasing returns to scale, new growth theory emphasizes the leaming by doing effects of human capital in the production of technology-intensive products and the dynamic spillover effects of the growth of the export sector, which acts as a leading sector in the diffusion of modem technology across other sectors and industries. Growth theorists also distinguish between education, especially at primary and secondary levels, and knowledge, acquired through work, production, research, and education. Table 6.2 Policy Implications in New Growth Theory Author Component Policy Implication Romer (1986) leaming-by-doing Invest in R&D Lucas (1988) extemal effects (spillovers) Invest in human capital Azariadis and Drazen (1990) threshold levels Invest in human capital Barro (1990) production externalities Public expenditures Young (1991) invention Research/production experience Grossman and Helpman (1 991) innovation Trade, openness, encourage knowledge flows 6.35 Different roles for the market and the government are specified in the new growth models. The need for government intervention is justified and a role for monopoly power to promote innovation is provided. Amable and Guellec's (1992) review of new growth theory shows that most models provide a basis for public intervention in order to reach the optimal equilibrium. The implication from Lucas' model is that human capital investments are likely to be below the socially optimal level unless there is market intervention in the form of a subsidy for accumulating additional human capital (Tallman and Wang 1992). Individuals do not take extemal effects into account, making subsidies for human capital accumulation necessary. Tallman and Wang (1992) argue that this may help explain the rapid development of the Newly Industrialized Countries. Single Country Estimations Human Capital Matters, But Not So Much 6.36 Most of the empirical work conducted to validate new growth theory thus far has relied on cross-country evidence (which we survey next)."2' These cross-country studies, however, are often plagued by the measurement problems associated with the use of such data (Tallman and Wang 1992). In many cases it is not known how the data was collected, how reliable it is, or how consistent it may be. A more intensive examination of specific case studies may be a promising path toward understanding the role of human capital in economic development. Some interesting 121 See, e.g., Romer 1989,1990; Barro 1991; Lau, Jamison and Louat 1991; Scott 1991; Benhabib and Spiegel 1992; Barro and Lee 1993; Gundlach 1993; Knight, Loayza and Villanueva 1993; Baffes and Shah 1993). 176 work using new growth theory has examined the determinants of the within-country economic growth (which we now examine), though this work has not got the same attention as cross-country work. 6.37 As noted earlier, just as the relatively amorphous concept of knowledge was the favorite of new growth theorists, education has been the favorite component of human capital for empirical investigators, presumably because of data availability reasons. Attempts to determine the contribution of education to economic growth began in the 1960s. Denison (1967) tried to identify the contribution of different factors of production to the growth of national income or gross national product of the United States between 1910 and 1960. Increases in the quantity of labor and physical capital did not explain the increase in gross national product. There was a large "residual factor." Denison (1967) suggested that improvements in the quality of the labor force, including increased education, were important, and Denison (1985) found that changes in schooling accounted for almost 30% of the change in per capita income between 1929 and 1980. Several more efforts were made to refine or update Denison's estimates. And indeed the refinement continues till today: Recently, for example, Jorgenson and Stiroh (2000) estimated that labor quality growth accounted for nearly 15 percent of labor productivity growth for the 1959-1998 period in the United States. 6.38 A case study of Singapore and Hong Kong by Young (1992) and a later paper by Young- (1994) that expanded this exercise to include the other two East Asian fast-growers-Taiwan and South Korea-also used educational attainments to account for labor quality changes. Singapore and Hong Kong is used to develop some insights into the growth process and to evaluate the empirical validity of existing models of endogenous growth (Young 1992). The study finds that both capital and human capital-adjusted labor have grown considerably faster in Singapore, and that while technical change has contributed substantially to economic growth in Hong Kong, its contribution to growth in Singapore is next to nil. The difference in initial conditions is also important: Hong Kong had a better-educated labor force and industrial elite, and this is significant in explaining subsequent differences in the performance of the two economies. Hong Kong's superior labor force in terms of education and higher rate of TFP growth, supports endogenous technical change models which emphasize the supply of human capital as determining the ability of an economy to absorb new technologies. 6.39 An examination of the growth process using new growth theory in South Korea between 1967 and 1986 indicated that the growth of human capital and the diffusion of skills-which are emphasized in new growth models-seem to have been important for growth in Korea (Sengupta 1991). In the case of Indonesia, evidence for 1971-1990 shows that primary and secondary schooling has very important effects on overall growth: a one percent increase in the ratio between enrollment in secondary school to total enrollment (primary and secondary schools) would contribute to the increase in real GDP per capita by 0.5 percent (Juoro 1993).'22 Using data for Taiwan for the period 1965 to 1986, Tallman and Wang (1990) investigate an open economy 122 In an influential paper, Murphy, ShleHer and Vishny (1991) show that countries with higher proportions of engineering college majors grow faster than countries with a higher proportion of law majors. Jouro (1993) also verifies that engineering graduates influence growth positively in Indonesia, while law graduates have a negative effect. 177 model with endogenous growth through human capital accumulation. The evidence supports the theoretical suggestion that the inclusion of a human capital measure in the labor input variable increases its importance considerably. In an analysis of industrial growth in Coastal China, Wang and Mody (1993) examine the external effects of human capital. They argue that foreign investment and exposure to foreign knowledge has stimulated industrial growth in coastal China. Education is shown to have an important impact, although more so in tandem with knowledge acquired through intemational links (see also Chen and Feng, 2000). Table 6.3 Human Capital In New Growth Theory: Single Country Estimates Country Theoretical Concept Empirical Result Study Austria External effects of human capital Consistent Winter-Ebmer (1992) Brazil Threshold levels of education Consistent Lau, Jamison, Liu and Rivkin (1993) China Extemal effects of human capital Consistent Wang and Mody (1993) Guatemala External effects of human capital Somewhat consistent Sakellariou (195 Israel Knowledge Consistent Bregman and Marom (1993) South Korea Spillovers of human capital Consistent Sengupta (1991) South Korea Product variety Somewhat consistent Feenstra, Madani, Yang and and Taiwan Liang (1999) Taiwan Endogenous technological growth Consistent Tallman and Wang (1990) Taiwan Endogenous growth (education and Consistent Chuang 2000 exports) 6.40 The endogenous growth theory literature is used to specify a model of economic growth for Israel (Bregman and Marom 1993). The empirical analysis goes beyond neo-classical models to include physical infrastructure, human capital and the degree of openness of the economy. The returns to physical infrastructure (54-63 percent) and human capital (15-24 percent) are found to be greater than the returns to business capital (14 percent). Also, greater openness to foreign trade leads to a positive contribution to productivity. Openness in foreign trade is hypothesized to lead to greater growth because of increased competition, enhanced efficiency, and the exchange of ideas and technology, thus promoting the accumulation of knowledge and human capital. Using Lucas' (1988) model of the extemal effects of human capital formation, Winter-Ebmer (1992) examines the impact of human capital on wages in Austria. In the Lucas model the human capital investment of an individual has extemal benefits for his co-workers in the form of higher productivity, which should in tum have an impact on wages. The results are not inconsistent with the Lucas model. Estimating a number of different regressions, the author finds that the only consistently significant variable explaining industry wage premiums is average years of schooling.123 123 A microeconomic approach is used in order to test Lucas' basic assumption of the extemal effects of human capital. Extemal effects are specific microeconomic phenomena and cannot easily be tested using cross-sections or time-series of countries. This procedure obviously has the advantage that - especially relevant for the study of human capital effects - years of schooling are comparable across industries within one country but not across different countries all over the world (see below). The approach is two-step: internal effects of human capital on pay are controlled for using wage regressions for individual data. External effects on pay are present if the 178 6.41 Finally, an examination of economic growth over time in Brazil using state-level information for the 1970s and early 1980s-a period of rapid growth-estimated that education explains about 20 percent of this "miracle" growth (Lau et aL. 1993). Among the reasons put forward to explain such a high contribution is that there is a threshold level of minimum average education somewhere between three and four years of schooling in order for education to begin to have an impact, and Brazil crossed this threshold some years before this growth spurt.124 In more technical terms, there is a range where there are sharply increasing returns to intangible capital as proposed by Romer (1989). The authors caution, however, that it may be a long time before education has an effect on output (Lau et aL. 1993). Cross-Country Estimations Human Capital Stocks Matter, Flows Seem Not to 6.42 Getting human capital data for a panel of countries puts an additional strain on researchers: the numbers have to be comparable across countries. The result is that most cross-country regressions use schooling enrollment data to proxy intangible human capital investments, education levels to measure the stock-equivalents, and life expectancy data to proxy tangible human capital stocks. 6.43 There is considerable confusion about what the results of cross-country analysis are. Take a look at three assessments of the "consensus" from the literature, all circa 1997: 'in the burgeoning literature on the determinants of growth rates across countries, the importance of education is a robust factor, proving relatively insensitive to changes in either specification or sample composition."125 ratherjarring is the repeated finding, in these intemational data, that changes in the estimated levels of schooling or human capital do not contribute positively to growth, at least measured over the 1965-85 period.425 'Cross national data show no association between the increases in human capital attributable to rising educational attainment of the labor force and the rate of growth of output per worker. This average amount of human capital of a reference group (say, industries) is determining average wages in this group (or more precisely interindustry wage differentials) over and above the effect of individual characteristics on individual pay. In the first step intemal effects of education are filtered out using wage functions for individuals. In the second step the resufting industry wage premiums are regressed on industry-specific characteristics and, above all, on average human capital in the industry to account for external effects of human capital. The author tries to separate internal returns to education from external ones. 124 Lau, Jamison and Louat (1991) also find that a threshold of four years of primary schooling has a noticeable effect on economic growth, using cross-country data. 125 Birdsall, Ross, and Sabot (1997). 126 Griliches 1997, 'Education, Human Capital and Growth: A Personal Perspective", Journal of Labor Economics 15, S330-342. 179 implies the association of educational capital growth with conventional measures of TFP is large ... and negative.Qn2 6.44 Empirical studies by Romer (1989), Azariadis and Drazen (1990), and Barro (1991) were among the first to attempt to bring cross-country evidence to bear upon the implications generated by externality-based models of economic growth. The main findings of these studies was that economic growth in a broad cross-section of countries over the period 1960-1985 depends positively upon measures that are claimed to represent initial levels of human capital. Mankiw, Romer, and Weil (1992), in another influential paper, found that investments in human capital contribute towards explaining growth in an analysis of cross-country differences in growth. 6.45 The initial results confirming education's favorable impact on economic growth were soon questioned. Benhabib and Spiegel (1994) found little impact of education on economic growth, and the results were confirmed by Wolff (1996) and Baumol and Wolff (1997). Pritchett (1997) also [ound that in cross-national data there is no association between increases in human capital attributable to rising educational attainment, of the labor force and the rate of growth of output per Norker.'28 Wolff (2000) estimated three models on the role of education in economic growth and Found, for the most part, that educational levels, the growth in educational attainment, and interaction effects between schooling and R&D, are not significant determinants of country labor productivity growth.'29 3.46 The pendulum appears to be swinging back towards the initial result-that increases in education are correlated with increases in economic growth-recently. A positive correlation between economic growth and increases in educational attainment is estimated by Temple [1 999a), who argues that this effect was always in the data, but is hidden by unrepresentative Dbservations. His correction for measurement error is to remove outliers from the dataset. Then, ising the same data as Benhabib and Spiegel (1994) but applying "least trimmed squares" and successively eliminating 14 outliers, he finds that human capital is significantly correlated with growth. Most recently, Barro (2001) again looks at the impact of education on growth, Jistinguishing between the quantity of education (measured by years of school attainment) and 127 Pritchett 2000, revised version of a 1997 paper. 28 Nevertheless, Pritchett argues that the impact of education varies by country, going on to postulate various reasons for this finding. First, institutional/governance can be significantly perverse so that educational capital accumulation lowered economic growth. Second, perhaps the marginal returns to education fell rapidly as the supply expanded while demand for educated labor was stagnant. Third, educational quality could have been so low that years of schooling have created no human capital. 29 Wolff (2000) concludes with a review of possible reasons for this finding. Poor quality of education data-based on compilations of country responses to UNESCO questionnaires-is dismissed as the reason. The second possible reason is comparability across nations, but this is not viewed as seriously biasing the results. A third possibility is specification errors, but several different specifications yield the same results. The fourth is that the causal relation between productivity and schooling may be the reverse of the one assumed - that is, schooling might respond to income levels rather than productivity growth to educational levels. This may account for some of the results, especially given the fact that some countries have developed elite systems where education is more of a luxury good. The fifth possibility is the one that Wolff believes is correct: formal education perse may not have much relevance to productivity growth among advanced industrial countries. Rather, it may be the case that only some forms of schooling and training are related to growth. 180 quality of education (measured by scores on intemationally comparable examinations). His model is applied to one hundred countries, estimating the determination of the growth rate of real per capita GDP, measured over three ten-year periods between 1965 and 1995. On the effects of education, Barro (2001) finds that the quantity of schooling, measured as the school attainment of males at secondary and higher levels, has a positive and statistically significant relationship with growth.'30 On the quality of education, Barro (2001) finds that science scores have a statistically significant positive effect on growth. The implication is that a one-standard-deviation increase in scores would raise growth rate on impact by 1.0 percent a year. By contrast, a one-standard- deviation increase in school attainment would increase growth rate by 0.2 percent a year. Thus, he concludes that quality and quantity of education matter, but quality matters a lot more. 6.47 Finally, on what level of education matters, Wolff and Gittleman (1993) report while generally that the data seem to be consistent with the proposition that increases in higher education increase growth rates in high-income countries (but the effect diminishes over time), increases in secondary education for middle-income countries, and increases in primary school performance for low income countries. They conclude that for "developing countries, it appears that primary emphasis should be given to improvement of both the quality and quantity of primary and secondary education. These two levels are likely to be the main ingredients in promoting their economic growth for the foreseeable future." 6.48 When one stands back and considers these findings, they seem unsurprising. What seems more surprising is that it has taken economists two decades to leam this (and some are still not sure). Other Issues Reconciling Macro- and Micro-economic Evidence, Externalities, and Causality Reconciling the evidence 6.49 All of the above was confirmation at an essentially qualitative level that education matters for output growth. What the cross-country estimates tell us is different from what microeconomic studies have shown to be the case in many countries. The rate of return to a year schooling, for example, is about 10 percent, and even higher in low-income countries where these rates of return can reach 20 percent. 6.50 In a recent review of the literature, with a special focus on labor markets, Topel (1999) highlights the difficulties with establishing a relationship between education and economic growth. He documents the limitations of the macro approach to this question, suggesting that beyond establishing the premise that education matters, the channels through which education affects 130 The implication is that an additional year of schooling raises the growth rate by 0.44 percent a year, implying a social rate of return to male secondary and higher education of 7 percent. Female attainment at the secondary level is statistically insignificant. Male primary schooling is insignificant, but primary is a pre-requisite for secondary education, so it remains important. Female primary schooling is positive but statistically insignificant; however, when fertility is held constant, it becomes significant. Hence, female primary schooling promotes growth, at least indirectly. 181 growth are not touched upon. His estimations find significant "social" returns to education that are as large or larger than private retums. Using within-country changes in education and productivity, he finds that a one-year increase in the average years of schooling for a country's labor force raises output per worker by between 5 and 15 percent. Since the macro literature offers few clues about policies to pursue that will allow human capital accumulation to contribute to economic growth, Topel (1999) suggests that policies that artificially compress wage differentials also reduce the returns to post-schooling investment. That is, low retums to schooling deter investments in human capital. The implication is that supporters of greater educational investments should be pleased, not perturbed, to see high and increasing rates of return to schooling. 6.51 In an unpublished paper, Mamuneas, Sawides and Stengos (2000) try to reconcile the evidence between cross-country growth results and micro estimates, using total factor productivity estimates to construct estimates of the elasticity of output with respect to human capital. They find that for many developing countries, years of schooling have no effect on aggregate output. Constructing an estimate of the social rate of return to human capital from aggregate data on output and inputs, they find substantial differences in rates of return according to a country's stage of economic development. High-income countries have the highest rates while those for low and middle income are frequently insignificant and even negative. Their estimates of the social rate of return for developed economies are consistent with those from micro studies (private rate of retum) while those for developing economies are lower than private rates of retum. 6.52 Krueger and Lindahl (1998) also offer evidence to suggest that the impact of education on economic growth is as high as the retums to education estimated in microeconomic studies. They argue that evidence finding no relation between education and economic growth is a spurious result, coming about because of a high degree of measurement error in education data. They account for the error and find that the effect of changes in educational attainment - as opposed to the initial level of education, as is used in most of the macro growth literature - on income growth in cross-country data is at least as great as microeconomic estimates of the rate of return to years of schooling. Externalities 6.53 It has been difficult to estimate externalities associated with human capital, which would provide a rationale for subsidization of human capital investment and hence a role for public policy. 6.54 Heckman and Klenow (1997) estimate the externality by comparing the schooling coefficient from cross-country Mincer regressions with those from cross-individual Mincer regressions. When they take into account differences in technology, social retums are similar to private retums. Rauch (1993) examines the effect of average education on wages and finds significant externalities. Other attempts to estimate social retums include Krueger and Lindahl (2000), Bils and Klenow (1997), Wolfe and Zuvekas 1997 and Haveman and Wolfe (1984). But previous estimates show that social goals can be at least partially achieved via education and its extemal effects. The total annual value of the non-marketed effects of schooling are about the same as the annual, marketed eamings-based effects of one more year of schooling; i.e., the annual value of incremental schooling reported in standard human capital estimates might capture only about one-half of the total value of an additional year of schooling (Haveman and Wolfe 1984;). Lochner (1999) 182 estimates the social benefits of high school graduation through reduced crime amount to almost 20 percent of the private retum. 6.55 A recent review by Venniker (2000) finds that evidence is scarce and inconclusive, providing support for human capital externalities that is neither strong nor undisputed. Studies that estimate extemalities in the form of individual's human capital enhancing the productivity of other factors of production through channels that are not internalized by the individual (similar to Lucas, 1998) provide evidence that is ambiguous. Some estimates give negative values, others give high estimates. One way to summarize this may be that while there is anecdotal evidence of large knowledge spillovers through production (see, e.g., Easterly, 2001), the spillovers from education per se may be small. But there is no obvious policy implication that can be drawn. 6.56 Patrinos and Psacharopoulos (2001) provide evidence that if these externalities are not considered, social returns to all levels of education are considerably lower than private returns. The most noteworthy result is that while private returns to secondary schooling are lower than returns to higher education, this order is reversed when these returns are deflated by public costs of education (see Figure 5). 131 Causality 6.57 Much of what has been discussed relates to the correlation between human capital and growth. Some researchers attempt to determine causality-the distinction between initial human capital and concurrent investments in human capital in part addresses this weakness. But there have been more concerted efforts, a notable one being Bils and Klenow (2001). They find that higher initial levels of schooling enrollment is associated with faster subsequent growth in GDP per capita or worker for a broad cross-section of countries between 1960 and 1990. Then, constraining the cross-country estimations to be consistent with microeconomic evidence on rates of return to schooling and experience for individuals in 52 countries, they attempt to determine whether the data are more consistent with schooling causing growth, or with growth causing schooling. The conclusion is that the causation is more from growth-especially expected growth-to schooling than the other way round. 131 Additionally, for policy purposes it is important to determine at what level externalities are greatest. Psacharopoulos (1994) asks whether the externalities at lower or higher schooling levels are more important for economic development: 'Can one weigh two very elusive items: (a) the positive externalities associated with a university graduate discovering a new vaccine; and (b) the negative externalities associated with 30 percent of the population being illiterate for their entire lifetimes?" 183 Social and Private Rates of Return to Education, By Level 30- 15- ' X 25- /;', Primary Social Secondary Social Tertiary Social Primary Private Secondary Private Tertiary Private a Latin America and the Caribbean * Asia 0 OECD] Figure 5 Where do we go from here? 6.58 Perhaps a fair one-line summary of the results of dozens of such studies is that higher initial human capital stocks raise subsequent growth rates, but current investments in human capital appear not to have a significant growth-enhancing effect immediately or in the near future." This is not very different from the results of Lau et al (1993) for Brazil during its growth spurt in the 1970s and is essentially the same conclusion reached through time series analysis of Brazilian data for 1951-1999 by Fiess and Gill (2001) that is discussed in the next section. 6.59 Solow (2000) and others who have followed this debate for more than a decade advise that-for both theoretical reasons such as the assumptions regarding technical progress and practical considerations such as the dubious quality of aggregate data on human capital-these cross-country estimations should be augmented by within-country empirical investigations. So we now examine what time series macroeconomic data from Brazil can tell us about the relationship between education (the best measured component of human capital) and economic growth. 132 Though Krueger and Lindahl (1998) claim to find the opposite for OECD countries, they get this result only when they consider 10 to 20 year lags. This is in fact similar in spirit to the finding that "stocks matter, not flows". 184 Brazilian schooling and GDP during the last half-century 6.60 School attainment is measured as the average years of completed schooling of 20 year-olds in any year. Growth is measured as annual increments in real GDP; real GDP figures are from IBGE, the Brazilian Institute of Geography and Statistics. Since we do not have schooling data as far back as 1950, we construct a the time series on schooling using household survey data from stacked household surveys for 1998 and 1999.t" We impute schooling for each year between 1950 and 1999 from these data. The exercise assumes that individuals complete their school education by the age of 20. Thus, the average schooling level of a persons aged 69 years in the 1999 survey is used to approximate the average schooling level of people who were 20 year-olds in 1950, school attainments of a person aged 68 (67) in 1999 (1998) are equivalent to the school attainments of a 20 years old in 1951, and so on. The two series are displayed in Figure 6. Real Output and Schooling 7.2 1 .12e+CJ9 -,--_ - 7. 9.6e+08 6.4 ae.08 5.6 6.4e+08 4.8 4.8e+08 4.0 3.2e+08 , - 1.6e+08 L / ...... - 3.2 O0- - __ . . .. . 2.4 1950 1955 1960 1966 1970 1975 1980 1985 1990 1995 Figure 6.6 133This section is based entirely on a note by Fiess and Gill (2001). 134The PNAD (Pesquisa NacionalporAmostra de Domicilios) is the Brazilian annual National Household Survey. PNAD data come from annual national household surveys of about 100,000 households, performed in the third quarter of each year. The survey is conducted by IBGE. It began at national level in 1971 and underwent major revision between 1990 and 1992. This revision makes it difficult to obtain fully compatible data before and after 1992, and since we do compare across decades this is important to keep in mind. The survey contains extensive information on personal characteristics, including information on income, labor force participation, and educational attainment and attendance. 185 Some Econometrics'35 6.61 We use a bivariate time series model to assess the causality between school attainment and growth in Brazil between 1950 and 1999. The econometric technique used here is called a Granger causality test, which essentially examines whether changes in schooling levels and GDP (a) are correlated with each other; and (b) display causality, e.g., increases in schooling levels systematically precede increases in GDP-in which case schooling is said to 'Granger cause" economic growth. 6.62 Since we find strong evidence that schooling and real GDP are non-stationary, we test for Granger causality within an error correction framework, i.e., we first determine the order of cointegration between the two series following Johansen (1988) and then augment short-run dynamic equations for changes in school attainment and growth with the derived error correction terms from the cointegration regressions. The VAR models for the cointegration analysis includes a constant in the cointegration space and 3 lags for school attainment and GDP. This specification proves sufficient to produce random errors.'36 The X .test indicates one significant cointegrating vector (Table 4). Table 6.41'7 Null Altemative Lag: 4 95% 90% Hypothesis Hypothesis With Constant CrHfical Value Criftical Value Ho: rank = r X t,e test r = 0 r > 0 23.84- 20.17 17.79 r 1 r>1 6.56 9.10 7.50 'Rejection at the 5% level of signHficance 6.63 Normalizing the cointegration vectors on the 1st element yields the following estimates for 1 and a (Table 5): 135 This section can be skipped without prejudice to understanding the rest of the paper. '3 The model specHfication is in the Appendix along wih tests for long-run exclusion, stationarity and weak exogeneity. Both variables appear to be non-stationary and the diagnostics on the residuals of the system show the absence of autocorrelation and indicate normality. A sensitivity analysis for dffferent lag lengths. There is some evidence for weak exogeneity of school attainment. 137 This result comes about by starting at the top of Table 1 and moving downwards until Ho cannot be rejected. As this is the case in the second row, the analysis maintains the Hoof one or less cointegrating vectors; this implies the existence of exactly one cointegrating vector in the data. 186 Table 6.5 3 (1) za (2) t-stat. School 1.000 A school 0.209 1.955 GDP -1.866 AGDP 0.112 4.446 Constant 31.755 Note: A indicates a variable in first differences. 6.64 The adjustment coefficient of A school is borderline significant and a formal test indicates that weak exogeneity for school attainment cannot be ruled out. (The LR test is: x2 (1) = 2.40, p- value = 0.12.). This indicates that in the long-run school attainment is driving growth and growth is the only variable adjusting to disturbances in the long-run equilibrium. 6.65 To assess Granger causality while taking account of cointegration between school attainment and real output, we estimate a short-run dynamic equation for growth and changes in school attainment, augmented with the lagged error correction terms from Table 5, column 1. The choice of a lag length of 2 for the short-run models is determined by the specification of the VAR model for the long-run analysis. The findings are reported in Table 6. Table 6.6 2 2 E A School,, E A GDP1, ecm1.1 n=l n=l A School, F-test: F-test: t-test: 1.10, p = 0.27 F(2,41) =1.34, p = 0.27 F(2,41) = 0.32, p = 0.73 A GDP, F-test: F-test: t-test: 3.91, p = 0.00 F(2,41) = 0.69, p = 0.51 F(2,26) = 2.40, p = 0.11 Note: A variable x, is said to Granger cause a variable y, if either its lagged Values (A x,-,) or the error correction term, ecRm.,, enter significantly the short-equation for A yt. 6.66 We find no short-run causality between school attainment and growth. Since the error correction term is only significant and positive (coefficient: 0.11) in the growth equation, we conclude that school attainment causes growth in the long run, but growth does not cause school attainment, either in the short or the long run. Schooling seems to help, but only in the long-term 6.67 Time series data for Brazil over the last fifty years appear to be consistent with the main result of cross-country data analysis: countries with higher levels of education, other things being equal, seem to grow faster; but increases in education levels do not result in higher growth rates. There is an apparent inconsistency here-how can stocks and flows of the same variable (education) have such different effects on growth. The paradox can be resolved by bringing in the time dimension of this relationship. The international evidence seems to indicate that human capital-proxied often by schooling levels-pays rewards in terms of higher growth rates only over long time horizons. Today's flows are tomorrow's "initial conditions" or stocks. 187 6.68 The results reported for Brazil in this section suggest (the data do not allow a stronger statement) that higher investments in education do raise growth rates over the long run. But schooling increases do not translate into immediate increases in growth rates; if anything, higher growth rates-by raising the retums to schooling and hence encouraging greater private investments in schooling or by enabling governments to spend more on education-appear to raise schooling levels (see Figure from Blom et al, 2001 '3). The policy lesson here may be that Brazil should maintain steady investments in education and other complementary forms of human capital, and not expect immediate or short-term growth dividends from such investments. Over longer periods, when combined with other pro-growth policies, these human capital investments will pay off. When it comes to human capital, patience seems to be the watchword. Lau et. al (1993) come to a similar conclusion from their study of Brazilian growth during the 1970s (see section 3). Metropolitan Brazil: The Returns to Schooling and GDP growth 14.0% 10 8 136o- - 6 2 -2 12.4% -4 1982 1984 1986 1988 1990 1992 1994 1996 1998 |-- Return (left scale) -0-- GDP groMh (t+1) in % (right scade) Source: Blom et al (2001), based on PME-data and for GDP-growth: World Bank database SIMA. Note: The returns to schooling are given by the coefficient to years of schooling in an earnings regression with control variables. Figure 6.7 6.69 Once again, however, we should stress the unreliability of such exercises as evidence on their own. But these results appear to be generally supported by analysis from cross-country data. Even if we take these results seriously, however, these findings are not enough to inform policy. For example, it is less than obvious at what level policy efforts should be focused-education levels increase if some people acquire a lot more education, or if many people acquire a little more education. It is to these questions that we now tum. Human capital formation before, during, and after reasonably long periods of economic growth may be especially instructive for determining 38 Blom et al (2001) find that a relationship between GDP growth and retums to schooling is discernible, but only after 1986. The simple correlation between the two is 0 for the whole period 1982-1998, but 0.5 for the period after 1986. After 1986, GDP-growth seems to be related to the average returns to schooling in the subsequent year. That is, GDP-growth, ceteris paribus, increases the reward to human capital. 188 policy. The call for the study of "growth episodes" by Romer (2001) actually conforms closely to the policy practice of providing "good practice" cases in economic reports, and this is what we tum to next. 6.70 In this section, we study three "episodes" of remarkable growth in per capita GDP: the United States (henceforth US) between 1901 and 2000, the Republic of Korea (henceforth Korea) between 1951 and 2000, and India between 1976 and 2000. The emphasis is on verifying or qualifying the policy conclusions drawn from the previous two sections, and not on describing or analyzing the growth episodes in more general terms. 6.71 The choice of countries for closer study is deliberate, and the periods chosen are deliberately long as well-100 years, 50 years and 25 years respectively. We have chosen to contrast currently democratic countries at various stages of economic growth, from the US at a per capita income of about $30,000 and Korea at $10,000, to Brazil at $4,500 and India at $450 (see table below). More interestingly, the choice conforms with "best practice" defined using a criterion that economic growth analysts may find interesting: these are countries exhibiting high economic growth during the last two decades or more, conditional on levels of output of or inputs to growth (along the lines of Hanushek and Kimko, 2000). Since 1980, The US was one of the fastest growers conditional upon initial per capita income levels, * Korea was a fast grower conditional upon initial natural resources per capita, India was a fast grower conditional upon initial physical and infrastructure capital per capita and human capital, and 6.72 We contrast these conditionally fast growers countries with Brazil, which was a fast grower conditional upon initial human capital endowments.'39 139 For support for this selection see, e.g., Hanushek and Kimko (2000), figures 2 and 3, page 1203. The countries are first ranked according to their average annual growth rates of real GDP between 1960-1990, and then re-ranked by conditional growth rates allowing for initial income and initial human capital. India and the US jump from the 25-50 percentile group to 75-90 percentile group, Brazil also moves up more than 20 percentage points from the 50-75 percentile group to the 75-90 percentile group, and South Korea stays in the 90-1 00 percentile group. Another country that moves up strongly in ranking is Mexico; examples of countries that move the other way are Germany, Japan, Israel, and Panama. 189 Table 6.7 Growth Episodes and Human Capital Formation in Four Countries, 1970-2000 World USA Korea Brazil India Unit Real GDP, 1999 30220 8710 405 750 445 Current $ Billions Real GDP/Capita, 1999 5055 31915 8685 4470 450 Current $ Growth of Real GDP Annual % 1971-1980 3.6 2.8 7.6 8.5 3.0 1981-1990 3.0 3.2 7.7 1.5 5.9 1991-2000 2.4 3.2 6.1 2.6 5.8 Growth of GDP/Capita Annual % 1971-1980 1.7 1.8F 5.7 . 5.9 0.7 1981-1990 1.3 F F 2. . 6.4 -0.4 3.61 1991 -2000 1.3L _ 2.2 6.0 1.2 3 . 91 Education Level Years 1960 8.7 1970 10.1 1980 11.9 1990 1995 12.2 10.1 4.3 4.2 Not, -.. -- . . - _ Shaded.area Indicalntes pc 1 -ii of s ustaIried G P -gowth Sources World Bank staff estimates for GDP, Barro and Lee (2000) for education United States, 1901-2000: America's "Human Capital Century" 6.73 At the beginning of the 20' century, the leading European countries such as Germany, France, and the UK, and the US all had secondary school enrollment ratios that were around 20-25 percent, not too different from Brazil's own net enrollment ratio at the beginning of the 2151 century. Barely 10 percent of US youth were high school graduates in 1910. Shaped by what Goldin (2001) and others have termed a "Republican ideology" and facilitated by new world endowments, the US devised an education system that made it the leader in post-elementary education within three decades. By the late-1 930s, the US was three or four decades ahead of Britain and France in post- elementary education, and these gaps persisted until even after the 1950s (see Graph and Goldin, 2001). By the mid-1930s, 'the median youth had a high school diploma" (Goldin, 1999) so that public initiatives to expand access to higher education-such as the GI Bills for retuming soldiers in the 1940s and 1950s-can safely be classified as non-elitist, at least outside USA's deep South. While post-World War II Europe struggled to expand secondary enrollments, the US was well on its way to ensuring robust tertiary education enrollments. 190 Secondary School Enrollment Ratios In the Mid-1950s 80- 70- / 60- so - 13FulTime 30 - 20 - 10- 0X - - USA Germany UK Denmark Sweden Ireland France Belgium Austria Norway N. Zealand Figusre 6.8 6.74 By 1970, many countries had caught up with the US secondary enrollment levels, but it took some of them another three decades to achieve the gross tertiary enrollment ratio of more than 70 percent that the US had attained by the end of the century. The 20th century has been nicknamed the "American century'-for some of the same reasons, it is also the "human capital century" as in Goldin (2001). Since 1950, the US economy has undoubtedly benefited from increases in both educational attainment and research intensity: one growth accounting exercise (Jones, 2000) concluded that 30 percent of US growth between 1950 and 1993 was due to the rise in educational attainment and 50 percent to the rise in research intensity, facilitated undoubtedly by the high educational stocks in the US.1'4 6.75 What distinguished the US secondary education system from its European counterpart was that it "was open, infinitely forgiving, lacking universal standards, and academic yet practical," while the European system "was closed, unforgiving, with uniform standards, and academic for some and industrial for others. One was egalitarian, the other was elite." (Goldin, 2001). Another distinguishing feature was decentralization. The high school expansion in the US was not led by the federal government, and even state compulsory education laws were a secondary factor. 6.76 The pattern of the rates of return to education in the US is also instructive. The relative wages of more skilled workers have over time fallen, with three periods during which they fell 140 Labor quantity is responsible for the remainder. 191 sharply: between the late 1910s and early 1920s, during the 1940s and-for college graduates only-during the 1970s (Goldin and Katz, 2001). But education levels increased even during these periods. Current issues in the US-quality and equality 6.77 The attention to secondary education quality-or "universal standards"-in the US is a relatively modem phenomenon, and by some measures the US appears to have traded quantity for quality. US students are stereotyped as doing considerably worse in international tests than their European and East Asian counterparts (Franciosi, 2001). This may be one aspect in which a virtue of the US system handicaps it: the forgiving and egalitarian nature of secondary education in the US implies that a relatively large population of disadvantaged youth take the test in the US than in other countries (Berliner and Biddle, 1995). This should imply that US students do as well as students in less egalitarian systems at lower grades, and this difference should increase for higher grades. In fact, this appears to be true. Franciosi (2001) reports that at age 10, fewer than one-third of countries outperforms the US students in math and science tests, by age 14, this ratio is close to half, and by the last year of high school, this ratio is more than two-thirds. In any case, standardized test scores seem to be a poor predictor of the outcomes that matter. Heckman (1999) cites evidence that the link between test scores, especially in early years of schooling, and later outcomes is at best weak. 6.78 After attaining high levels of primary, secondary and tertiary enrollment rates, motivated by equality concerns and spurred studies showing the effectiveness of early childhood interventions, the US has launched pre-primary programs such as Head Start for disadvantaged children. The beneficial effects are estimated to be high and lasting (Heckman, 1999): Treated when they were 4-5 years old, disadvantaged subnormal IQ youth assigned to the Perry Preschool program fared significantly better even 30 years later than comparable non-participants, retuming about $7 in higher earnings and/or lower crime for every dollar spent. There are few public programs that have such high benefit-to-cost ratios. The US experience with Head Start also shows that such interventions have to be well funded to be successful, a strong argument for govemments to be selective in where and how to intervene. Heckman (1999) reports positive results for interventions for primary school students, but the benefit-to-cost ratios are considerably lower. The downward trend continues with interventions for older youth-public training programs often fail to meet simple cost-benefit criteria. 6.79 The third issue is competition at the primary and secondary level, motivated again by equality of opportunity considerations, and spurred by the example of US higher education, which is widely regarded as the world's best. The US primary and secondary education system is decentralized down to the district level. Since the rich are more mobile, they can move to districts that have better schools (which tend also to have higher house prices and rents). There is thus a concern that the US system of education does not serve the poorest sections of society well. Greater competition between school districts is suggested as a solution, but such reforms are opposed by public teachers' unions. Hoxby (1998) reports large gains in costs, parental involvement and student performance when public schools face greater competition from private religious and non-religious schools-higher levels of achievement are produced at lower cost. Countries such as Brazil that have decentralized education systems can benefit from the US experience. 192 Table 6.8: Effects of competition among US school districts Costs - Per pupil spending 17% decrease Public school costs - Public schools' per pupil spending 0% change Student achievement - Test scores 3 percentile point increase Student achievement - Wages 4% increase Student achievement - Educational attainment 0.4 additional years Public school student achievement - Test scores 8 percentile point increase Public school student achievement - Wages 12% increase Public school student achievement - Educational attainment 12% increase in college graduation Parental involvement - Parent's visits to schools 30% increase Korea, 1951-2000: A Hundred Years in Fifty? 6.80 In the last haf -century, South Korea has grown faster than any other country. Without attributing causation, it is also remarkable that during this time the average education level in Korea has increased from less than four years to more than ten, and is now higher than some West European countries.1' In three decades, it achieved almost universal primary and secondary education coverage, and now has a tertiary education sector that is as large as those in many developed countries (Kim, 2001). Korea's unprecedented growth in education is not just an outcome of economic growth itself but, rather, can be attributed to special features of the Korean growth strategies that were pursued with both equity and outward-orientation. In tum, Korea's economic growth has beneffted from the education level of its human resources, which have played a key role in absorbing advanced technology from the developed countries."' 6.81 The noteworthy features of Korean education are (a) accelerated pace of expansion, (b) egalitarianism; and (c) 'conservative sequencing' in that primary enrollment levels were high before a massive secondary education initiative was undertaken, and secondary enrollment rates were high before a massive tertiary education expansion. In the second two aspects, one could 41 There have been many studies on how Korea achieved high, sustained economic growth. Factors including high savings and investment ratios, a well-educated labor force and well-directed export-oriented development strategies have been cited as responsible for Korea's success. Several aKtempts to investigate the ingredients of Korea's economic growth have followed the growth accounting approach, which, by decomposing the growth of output into a variety of inputs, gives a measure of human capital's contribution to output growth. In general, the growth accounting estimates show that the contribution to growth of human capital as a productive factor has been small in the Korean economy: one study finds that the increase in human capital explains 0.7 percent of the annual growth rate during 1963-82, or less than 10 percent of total growth. But this may be an underestimate because growth accounting does not explain where the growth of inputs and, especially, total factor productivity come from. '4 On the other hand, human development itself has undoubtedly benefited from the strong demand for education triggered by the growth of income. However, Korea's unprecedented growth of education indicates that the accomplishment in human development is not just an outcome of economic growth itself but, rather, can be attributed to special features of the Korean growth strategies that were pursued with both equity and outward- orientation. 193 argue that South Korea's experience was uncannily similar to that of the US during the last century."' Accelerated expansion, not "leapfrogging" 6.82 The 1950s and 1960s were a period of emphasis on primary education in Korea, with policy efforts concentrated on expanding basic education opportunities. The rapid pace of expansion necessitated that much of education be "low cos"-double shifts, large class size, and growing public discontent marked this phase. By the mid 1960s, the shift to better quality primary and a rapid expansion of (vocational) secondary education had begun, but the big increase in the number of high schools did not occur until 1975-1980 (when high schools increased by almost 300 and enrollment by half a million-see Kim 2001). In the mid 1970s, the govemment started to pay attention to access to colleges by all groups, especially through improved quality of instruction in secondary schools for all income classes. By the late 1980s, despite concems of a possible uover- education" of Koreans, the pace of educational progress has not slowed down. 6.83 It is almost tautological that a country cannot have higher enrollment ratios at the secondary school level than at the primary level, and similarly, greater ratios at the tertiary level than at the secondary level. What is exceptional about Korea is how quickly the gains in enrollment were attained: Korea accomplished in about three decades (1 950-1980) what took the US six decades after 1900. Even more remarkable, perhaps, is the careful sequencing. Rapid expansion at the secondary school level occurred between 1970 and 1990, afterachieving (between 1950 and 1970) universal enrollment at the primary school level. Again, the big push at the tertiary education level occurred between 1990 and 2000, after near-universal secondary enrollment had been achieved and-using the criteria suggested by Goldin (2001) for the US-public initiatives to encourage higher education enrollments could safely be classified as "non-elitist '.Y4" While this does not mean that no expansion took place at the higher levels before universal enrollment was attained at the lower levels; the point is that public policy aimed at a sequenced expansion; '4 The main difference was in the degree of decentralization, the Korean approach being considerably more centralized than that of the US. 144 A combination of high growth and equitable income distribution is another factor in Korea's progress in human capital. Korea has had relatively equal income distribution, which has enabled the majority of the people to access private education. 194 By 2000, Korea had Universal Primary and Secondary School Enrollment 120 100- 1970 Pr1mary 2 I D~~~~~ High School 20 1970 1980 1990 2000 Figure 6.9 Authoritarian, but egalitarian 6.84 Between 1950 and 1970, the share of primary education (grades 1-6) remained at about 70- 75% of the budget, starting to decline to about 50% in the 1970s, only after universal primary enrollment had been achieved. At this time (in 1968), entrance restrictions to middle schools were also abolished to make access universal and a system of random allocation of students to middle and high schools was instituted to make the system more egalitarian. The pattern still remains egalitarian in its most basic features: over the last three decades, the share of public school enrollments has remained high and even increased at the primary level and middle school levels, but the share of private institutions in enrollment has increased at the tertiary level (see Figure 10). 195 Higher Education Has Become Privatized Over Time In Korea, But Primary Education Has Remained Public 100- go-, 8-70- j 60- I50- [OP 2K 4I -0r0- 20~~~~~~~~~~~~~~~~~~~~~~~~~~~000 30- 20- o' i _ t7 - - ' Fi¢ Primary Middle School High School Junior College University Figure 6.10 India 1976-2000: Growth and Hfigher Education Investments Amidst Illiteracy 6.85 India is sometimes held up as a counter-example to the advice given by economists to developing countries that they move sequentially from an emphasis on primary education to one on secondary education, and only then on higher education. After all, ignoring this advice, India spent a large chunk of its public money on higher education, especially on technical institutes, and-if the popular press is to be believed-is the better for it. India, it is said, has become a technology leader in the world; never mind that almost half of India is still illiterate.'45 Cities like Bangalore and Hyderabad are touted as the Silicon Valleys of tomorrow; India has a new growth engine in information technology, and will be pulled by it to higher and higher levels of living standards. 6.86 To determine if this is indeed the case, we surveyed the evidence on the premier public technology institutes, focusing on the Indian Institutes of Technology, and on India's rapidly growing information technology sector, especially in and around Bangalore. Indian Institutes of Technology-human capital investments for other countries? 6.87 In 1946, India's Ministry of Education convened.a special committee to find a way to train brght young Indians in engineering. The Sarkar Committee recommended that four institutions be 45 Illiteracy rates in India for adult males and females (aged 15 and over) in 1997 were 33 percent and 61 percent respectively. 196 built, one each in the north, south, east and west of the country. The schools were intended, in the words of future Prime Minister Nehru, to "provide scientists and technologists of the highest caliber who would engage in research, design, and development to help build the nation towards self- reliance in her technological needs." The planners drew on Massachusetts Institute of Technology as a model and on UNESCO for funds to build the first campus near Calcutta. Five other campuses followed, in Kanpur, Delhi, Bombay, Madras and, most recently, Guwahati. At various times, the United States, Britain, the former Soviet Union and Germany have provided financial backing and technical support. Table 6.9 India's Institutes of Technology IIT Campus Established Undergraduates Graduates PhDs Links Kharagpur, West Bengal 1951 2700 students total http://www.iitkg.emnetJin Kanpur, Uttar Pradesh 1959 1250 850 58 http:/Awww.iitk.ac.in/ Bombay, Maharashtra 1959 450 711 203 http:/wwwI.iib.ernet.in/ Madras, Tamil Nadu 1959 no info available http://www.iitm.emet.in/ Delhi 1961 1501 1040 488 hftp://www.iitd.ac.in/ Guhawati, Assam 1995 no info available http://www.iitg.ernet.in/ Location to be decided Na na na na na 6.88 The IlTs have also been teaming up with industry on development. IIT-Kharagpur patents a dozen new products each year. Companies such as Intel and Philips Electronics, which are big recruiters at the IlTs, have funded endowments and scholarships. They have even bankrolled computer and electronics laboratories in order to maintain standards. IBM opened a research center on the campus of IiT-Delhi in 1998-and with sponsored projects that are collaborations between IIT faculty and industry. The Indian govemment pays most of the $3,000-six times the annual per capita income of the country-it costs annually to educate each student. One view has it that India has been able to produce world-class universities at surprisingly little cost, and by nurturing the schools, the country will reap huge rewards as the graduates invest in India and draw it further into the circle of global trade and prosperity. 6.89 This cost advantage has been a significant cause of brain drain. About 40 percent of IIT graduates are believed leave the country and not retum. In 1998, Indian engineers were running more than 775 technology companies in Califomia's Silicon Valley that accounted for $3.6 billion in sales and 16,600 jobs (Saxenian, 2000). Further, it is estimated that 40 percent of Silicon Valley start-ups are Indian-spawned, and of those, half are by IIT graduates. It is hoped that at least some of these will return to India or at least bring business to India. But the connections between these non-resident Indians and their home country rarely extend beyond holiday visits. But, it is said, despite the fact that many graduates never return to India, the IlTs spawn an interest in investing in India both by IIT alumni and by companies interested in the IT phenomenon in India. Bangalore, a city in Southern India, has attracted a lot of this business. 197 The Lore of Bangalore 6.90 Prior to 1984, the Indian software industry operated within the framework of a highly regulated, autarkic model of import substitution-led industrialization (ISI) and the ideology of self- reliance that guided the Indian economy. This policy regime stifled entrepreneurs and isolated the India from the global economy. As a result, efforts to promote software exports during the period never took off. Policies that permitted the import of state-of-the-art computers in exchange for a guarantee to export a certain amount of software were not enthusiastically received. Import procedures were cumbersome, duties were high and obtaining foreign exchange for business expenses was difficult. 6.91 The post-1984 policy changes were crucial to the growth of the Indian software industry because they allowed domestic producers to exploit domestic resources in global markets. India's greatest asset is a large, educated, English-speaking workforce that is willing to work at relatively low wages. In spite of widespread illiteracy, India boasts thousands of educated engineers who have remained either under-employed or unemployed for decades. Few countries can match India's combination of low-wage, high-skill workers. In 1994, wages for software programmers and systems analysts in India were less than one-tenth of those for their United States counterparts, and lower even than other developing countries like Mexico.141 6.92 The performance of India's IT industry during the 1990s has been impressive, particularly compared with other sectors of the Indian economy. The sectors compound annual growth rate for 1994-1999 exceeded 40 percent, compared with 6.6 percent for the economy as a whole. This strong growth was led by the software industry, which in 1999 accounted for 65 percent of India's total IT revenues. Total software revenues of $3.9 billion in 1999 were close to four times those of IT hardware manufacturing and grew more than 55 percent per year in the late 1990s. Moreover, the software industry's growth was driven primarily by exports. While the domestic market for software has grown in absolute terms, software exports account for a large and increasing share of total industry revenue. This export success is particularly striking for an industry that remained peripheral to world markets throughout most of the 1980s. A report by McKinsey & Co. has forecast that by 2008, India's IT industry will generate $87 billion in annual revenues, $225 billion in market value, and 2.2 million jobs. 6.93 One may never really know it because of the hype, but Bangalore and other technology centers are still just small oases in India's human capital desert. The total employment in the Bangalore software industry, for example, is about 200,000, a miniscule 0.05 percent of India's labor force of 440 million, and less than the number of Indians granted temporary worker visas in the US in the last decade (see Figure 11). India's software revenues are less than 1 percent of an estimated world software market of some $300-$500 billion. In spite of the sectors rapid export ,4 Programmers in India are increasingly aware of the global demand for their skills and the higher compensation available in developed economies. Many aspire to work for foreign companies for opportunities for overseas jobs. The United States has been a major beneficiary. A recent study of the H-1 B visa program, which grants temporary work authorization to skilled foreign persons, reports that Indian H-1 Bs grew steadily from 1989, becoming the largest category in 1994, doubling in size by 1996 and quintupling by 1999. Indians accounted for 47% of all visas issued in 1999, implying 55,000 Indian workers in 1999 alone, and a total of almost 200,000 between 1989 and 1999. The next largest groups of H-1 visa holders were from the UK and China, but each was 6% or less of total visas granted. 198 expansion between 1985 and 1995, from $28 million to $480 million, India's share of total world IT exports has remained only 0.5 percent. Moreover, India's 0.5 percent share of world IT exports in 1995 was less the country's 0.6 percent share of world aggregate exports in the same year. In spite of the evidence that a handful of Indian software companies are gradually moving up the value chain and gaining international recognition for their quality and performance, the industry as a whole remains significantly less productive than its global competitors. The annual revenue per employee in India was $15,000-20,000, a fraction of the $100,000 per employee in other software producers such as Israel and Ireland. Such numbers suggest that policy lessons from India's IT experience should be drawn with caution (Saxenian, 2000). Despite the Press, India's Information Technology Ssctor Is Still Miniscule 0.9- - - -. 0.7- ________ ____ _ _ 0.65 &0.5 _ it 0.4- /_. -- - - 02 - 0.1 - I Bangalore IT Share of Indian India's Software Revenues as India's Share of World rT India's Share of AJI World Labor Force Share of World Software Exports Exports Sales Figure 6.11 Lessonsffrom India for other developing counties 6.94 The lessons for other low- and middle-income countries from India's investments in the IlTs are difficult to determine. In attempting to learn lessons, three points are relevant. First, while the growth and outcomes of the IlTs are impressive, it is important to note that the IlTs are not leading "research centers"-not a bad thing for a country with a per capita income of $500 that is considerably behind the world's technology frontiers. Second, the IlTs were developed in close collaboration with institutions in countries that are (or were) the world's technology leaders, and have maintained these links over the years. Finally, luck was on India's side-the timing of these investments in human capital happened to be just right. IIT graduates were able to benefit from the telecommunications revolution and now-by facilitating links between US and Indian firms-may be able to repay India for the large subsidies they have privately benefited from. 199 6.95 The lessons from the Indian high technology "miracle" are somewhat easier to draw. First, one should be realistic "about the limits of software as a development strategy for India. Bangalore is not Silicon Valley and IT is not going to solve all of India's problems. IT is still a very small piece of the Indian output and exports...." (Saxenian, 2000). Second, the success of Bangalore almost certainly owes as much to good fortune as to good govemment. While the central govemment did lower import duties on computer software and hardware and adopt a policy of export development in the mid-1 980s, the government's noninterference may have also been responsible for the rapid growth of software exports than interventions. Some observers believe that had the govemment officials interfered too much, they may have prevented the provision of labor-intensive, low value- added programming services provided on-site to Western firms by Indian contractors-called "bodyshopping"-which later could be shifted offshore to places like Bangalore and Hyderabad. Finally, at least in the beginning, private initiatives (e.g., Texas Instruments' telecommunications center) overcame the main obstacle to growth of this offshore industry-lack of adequate telecommunications infrastructure. Sen (1994) observes that "until 1991-92, there was virtually no policy support at all for the software sector." It was only later (in the early 1990s) that government- aided technology parks-akin to export-processing zones-were set up that combined telecommunications infrastructure with tax incentives for software exports. 6.96 Over the last two decades, Brazil has made extraordinarily rapid progress in some aspects of human capital-notably in expanding access to primary education and improving its quality, and in dealing with some threatening healthcare issues such as HIV/AIDS. By sheer coincidence-it is certainly just that-these have also been years of poor and erratic economic growth. Even if it were not a coincidence, this would not be a strong argument for slowing down investments in human capital: a healthier, longer-lived, more educated citizenry is a goal in itself, at par if not higher than the objective of higher incomes and greater wealth. In fact, the evidence summarized in this paper-that human capital investments do not generally lead to economic growth spurts, that there may in fact be evidence that growth spurts lead to greater human capital accumulation, but there is reason to believe that sustained accumulation of human capital leads to higher growth over longer periods-is consistent with a close relationship between human capital and growth in Brazil since the 1960s (see Figure from Blom et al below). Even if economic growth were Brazil's only objective, these experiences call for a strengthened human capital effort. 6.97 But over the next decade, Brazil will be hobbled by its debt service obligations. It should take these obligations seriously-paying down its public debt is more likely to help in jump-starting growth than even large and well-conceived investments in human capital. Paying off public debt will, however, leave little fiscal room for investment, in human or non-human capital, which means that Brazil has to be more selective than other countries in picking these investments. Based on the experience of some other countries and on Brazil's own experience, this paper suggests actions most likely to pay off in terms of economic growth while improving income distribution (or at least not exacerbating inequality). For public investment, these are to maintain efforts to provide quality primary education, an increased effort to improve access and quality of junior secondary education for everyone, and selected pre-primary interventions to improve health and learning outcomes of the relatively disadvantaged The aim of regulatory reform should be to provide stronger incentives for Brazilian individuals and firms to invest in human 200 capital and for the private sector to shoulder more responsibility in its finance and provision. Regulatory reform priorities are labor policy changes to encourage private training initiatives and the reform of the higher education system to ease restrictions to entry by private providers-including foreign Institutions-and greater autonomy for all institutions so they strengthen links with industry. We discuss these in more detail now. More Attention to Secondary-Especially Junior Secondary-Education 6.98 Perhaps the most important recommendation of this paper is that Brazil should dedicate the next decade to improving junior secondary or middle school education. There is some ambiguity regarding the definition of "secondary schooling": for Brazil, this should roughly be interpreted as grades 5-11, i.e., the second cycle of ensino fundamental (grades 5-8) and all of the ensino m6dio (grades 9-1 1). The most important rule is that the government initiative be "non-elitist": the majority of people should already be qualified to benefit from it. This is what disqualifies higher education from being a candidate for a large increase in public investment. The net secondary enrollment rate in Brazil is less than 40 percent, so it is likely-because of high dropout and repetition rates-that less than a quarter of Brazilian youth have high school diplomas. The reasons for this recommendation are: * Brazil's net secondary enrollment ratio in 2000 is roughly the same as the US level in the 1920s. Brazil lags behind the average Latin American country by about 40 percent in net enrollment ratios: that is, while the LAC average is about 50 percent, Brazil net ratio is 35 percent. Of all human capital indicators, this is by far the largest shortfall. * Brazil spends relatively less on secondary education than countries such as Argentina, Mexico and South Korea, and relatively more than these countries on higher education. * There is some evidence from cross-country data that improvements in secondary education access and quality yield the greatest payoff for the average middle- income country (e.g., Wolff and Gittleman, 1993). Since Brazil's secondary education performance has been worse than middle-income countries on average, this could mean that the effort should initially be focused at the junior secondary level. * While improvements in secondary education will not lead to immediate increases in economic growth, econometric evidence indicates that it is unlikely that any investment in human capital-e.g., higher education-will do so. So Brazil may as well be patient and egalitarian, instead of being impatient and elitist. It will probably not suffer a loss in short-term growth performance, and will almost certainly earn a premium over the long term. * Besides, there is little evidence of significant positive externalities associated with higher education (e.g., Heckman, 1999). The high and rising private retums to university education observed in Brazil call for measures to facilitate greater private provision and financing, and not to allocate scarce public funds to higher 201 education. The immediate goal should be to shift resources from tertiary levels to junior secondary education. Due to a sharp drop in fertility in Brazil during the last two decades, enrollments in primary education (grades 1-4) have begun to gradually decline. In a few years, resources will have to be shifted out of primary education. It is important that a large part of these resources be shifted then to secondary education, including grades 9-11. 6.99 Even if Brazil did not have a heavy fiscal burden, it should encourage the private sector to share the responsibility for this expansion of secondary education. As it turns out, Brazil already has a vibrant private sector (see box below)-it should not be crowded out by the public effort. The secondary education shortfall is so large, and the benefits of competition so well established, that a strong private secondary education sector will improve both access and quality of education. Private Post-elementary Education in Brazil Brazil already has a sizable and vibrant private sector in education. In Sao Paulo state alone, the private education sector caters for approximately 15% of all students-with nearly 20% at high school level. The largest school chain in Brazil is UNIP/Objetivo, which caters for more than 500,000 students. Many private schools operate more than one "unit," frequently vertically integrating into other levels of schooling, including kindergarten, or taking over other schools that are not so successful. The Pitagoras school chain, for example, a private educational institution, is based in Belo Horizonte, Minas Gerais State, in southeast Brazil. It was established in 1966. Pitagoras enrolls more than 85,000 students in 190 schools located all across the country. The COC - 'Curso Oswaldo Cruz" - school chain has three schools, 63 franchises and 26,000 students. (Tooley, 2000) Private schools and universities do not receive any direct subsidy from govemment. However, they are offered a number of tax incentives, which together amount to an indirect subsidy. If institutions are not-for- profit, then they are not required to pay taxes. According to the Education Ministry, however, the most significant 'indirect subsidy' comes from the possibility of families deducting expenses from taxation in terms of tutoring, and companies deducting from taxation their expenses putting employers through courses, and also if they fund scholarships to private schools. (Tooley) Nevertheless, a 1997 survey by GRUPO - the Sao Paulo State Private Schools Association - showed that 90% of its schools are established as limited companies, while only 10% are not-for-profit foundations and philanthropic entities (Tooley, 2000). There is small, but growing tendency, for private schools to seek funds from overseas investors. These include schools in Sao Paulo State such as Porto Seguro, (investment from Germany), Miguel Cervantes, (Spain), and Dante Alighieri, (Italy). These are known as very good schools with a large number of students. These potentially provide some role models of extemal investment in Brazilian education. (Tooley, 2000) 202 Maintain Attention to Early Childhood-Especially Preprimary-Interventions 6.100 Given Brazil's debt servicing burden and tight budget constraints, it has to be especially picky in deciding in whom to invest public money. Early childhood interventions qualify as investments with the highest retums: * Evaluations in the US suggest that early childhood interventions may have among the highest payoffs in terms of greater student achievement, higher eamings and lower crime rates. The experience of the US indicates that pre-school interventions have been successful in altering the social skills and motivation of children * Assessments from Brazil indicate that preschool attendance has significant beneficial effects in terms of learning and eamings(World Bank, 2000), even controlling for family background and other factors. Two years of preschool increase schooling attainment by about one year, with some evidence that this effect is greater for poorer households. Two years of preschool reduces grade repetition by 6-10% (greater for poorer households), thus increasing the efficiency of education expenditures. Two years of preschool increase male eamings by between 4 and 12 percent, again with some indication of a higher increase for poorer families. * The benefits of these investments for growth and poverty reduction will obviously be felt over the long term but, again, it is unlikely that there are any human capital investments, which will raise economic growth in the short term. Improve Regulatory Framework-But Reduce Public Spending-for Higher Education 6.101 This third area is a difficult political issue-as explained in Schwartzman (1998)-but it has to be tackled if Brazil is to put its scarce public resources to good use. The reason why this reform is difficult is that it would be best for the country, over the short to medium term, to reduce public funding for higher education. This should naturally be preceded by institutional reform to allow even greater private participation in tertiary, especially university, education. The rising rates of return to higher education are at once discouraging---because it means that the supply of higher education has not kept pace with demand-and encouraging, because it means that there is likely to be considerable willingness to pay for it. But one thing is clear: Brazil's higher education system may be the most elitist govemment program in existence. 203 Higher Education In Brazil Is Elitist: 1998-1999 Numbaro 7- caB e 7 ------ r - 0 / 0 3-z 2- 0- Fower than 3 MInimum 3-10 MInimum Salades 10-20 Minimum 20-50 Mlnimum More than 50 Minimum Salades Salaries Salaries Salaries Incomfe Level Figure 6.12 6.102 The priorities for reform for the Brazilian Higher Education Sector, detailed in World Bank (1999) are: * Removal of the constitutional mandate that all publicly provided education must be free of charge. In most other countries, the rule appears to be that schooling that is legislated as compulsory also be provided free of charge by the govemment. * Changing the funding formula for tertiary education to encourage competition among universities, and to increase the incentives to tap other revenue sources including current students, industry and alumni. * Streamlining the procedures for starting and running private institutions of higher leaming, including universities. See Table 10 for a contrast between Spain and Chile, which demonstrates that countries can choose to promote private activity or they can make it more difficult. Chile has decided to regulate the private university system lightly, while Spain takes a more heavy-handed approach. The Chilean system seeks to instill innovation and competition in the system. Other countries provide further incentives, such as subsidized loans, concessions on land, and student financing support. * A reduction in the share of the education budget spent on higher education for the next decade or so, so as to allow an increase in secondary education funding without increasing the fiscal burden of public education. 204 Table 6.10 Establishment of Private Universities: The Cases of Spain and Chile Spain: Stringent regulations Chile: Liberal System Curriculum requirements for higher education considered Curriculum requirements for higher education are not in establishmenUrecognition of universities considered A new university must have educational structure needed A new university must have the educational structure to provide at minimum instruction leading to eight degrees needed to provide instruction leading to a professional at least three in second cycle with one in experimental career chosen from among 12 prescribed by law sciences or technical studies Before beginning its operations it must elaborate It may begin operations and continue to function prospectus defining areas of research activities without undertaking research activities of any kind Total number of teachers in each university cannot be less There is no minimum than that required to maintain ratio of 1:25 with respect to the number of students In first cycle 30% of teachers must hold doctoral degrees; There is no minimum in second cycle 70%; in third 100% In no case may number of teachers holding doctorates be There is no minimum less than 50% of total teaching staff 60% of teachers must be employed full time There is no requirement to employ full-time personnel Teachers in private universities may not belong to the There is no incompatibility active teaching staff of any public university Full and effective compliance with the principle of There is no requirement in this regard academic freedom must be guaranteed Formal commitment must be made to keep university and There is no requirement in this regard each division in operation for minimum period of time to allow students to complete studies Inauguration of new university authorized by competent New universities begin to function automatically, once education administration after it ascertains all the approval formalities for their plans and programs commitments fulfilled and Government accredits official are completed by an examining university. degrees to be conferred Report on teaching and research activities carried out in They have no reporting obligations. The 1990 context of university's multi-year program must be amendments to the law on universities established submitted annually to administration and Board of reporting obligations only for universities that opt for Universities the public accreditation system Public authorities shall periodically inspect compliance by Plans and programs of new universities subject to private universities with the standards that apply to them approval; students and graduates must be examined by traditional university for number of years Source: ECLAC-UNESCO (1992), Education and Knowledge: Basic Pillars of Changing Production Pattems with Social Equity (Santiago, Chile) Maintain current taxation policies-especially the deductibility of training and education expenses 6.103 The reasons for this recommendation are: * Taxes on physical capital actually reduce human capital investment decisions. The level of human capital investment declines when the (net of tax) interest rate increases, so lowering the tax on interest income can have a beneficial effect on human capital accumulation. But it is unlikely that this will be popular recommendation since it will be seen as a move favoring capital and hence rich people. 205 The current tax system favors human capital accumulation on the job versus pre- employment schooling since training expenses by employers are fully deductible in Brazil but those by individuals are not. But there is a prevailing belief that Brazilian firms do not sponsor enough on-the-job training, so this bias towards training is probably best left unaltered. * While a progressive tax on earnings like the one Brazil has discourages human capital accumulation, it would be unpopular-and, in the short term, inequality- increasing-to replace it with a flat rate consumption tax. * Changes in tax policy involve complex federal-state issues that are best left untouched. ILTJI OWR N EM II silsoI M 6.104 To conclude, the main findings and policy implications of this paper are: * The evidence on the effects of human capital formation on economic growth is far from overwhelming; accordingly-despite claims to the contrary-policies to increase investments in human capital formation should not be regarded as the silver bullet for increasing economic growth rates; investments in human capital are for the long-term: * Part of the problem is measurement related-human capital is difficult to measure, even when proxied only by education, its major component. * Human capital formation requires complementary inputs to pay off in terms of economic output. A key piece of evidence in this regard is that the coefficient for human capital falls when capital accumulation is controlled for. It also falls when initial technology and policy effects are controlled for. This is what one should expect, because human capital is strongly correlated with both capital and unmeasured/unspecified inputs. 6.105 Sustained and stable economic growth appears to be an important determinant of rapid human capital formation, rather than the other way round. For Brazil, just as recent stabilization efforts were aimed at "breaking the back of inflationary expectations", this finding means that efforts to strengthen "expectations of growth" through steady and sustained-even if modest- economic growth will increase human capital formation. * Cross-country evidence and analysis of growth episodes indicates that the causality between human capital and economic growth runs both ways, with evidence of economic growth encouraging human capital formation in the short and medium term somewhat stronger than the evidence for human capital formation "causing" economic growth. * Some of the cross-country evidence and the experience of countries like the US, Korea, and India indicates that episodes of strong human capital formation have 206 followed-not accompanied or preceded-periods of steady, sustained, economic growth. An important catalyst for human capital formation may be the reduction of volatility of real GDP growth. 6.106 Brazil has done better in improving health-related outcomes than education and training related indicators. The thrust of human capital policies should be to aggressively improve these "intangible" components of human capital in Brazil. * Health outcomes have improved quickly since the 1960s, and Brazil made the epidemiological transition in the 1980s. This progress permits a generalization: while progress in some areas-such as adult male health-is still needed, the problems of the "median Brazilian" are more those of upper-middle and high income countries than those afflicting people in low-income countries. * On the other hand, many education outcomes continue to lag even the average Latin American country, and in some cases (e.g., net secondary school enrollment rates) are as poor as some lower-middle-income and low-income countries. 6.107 The main thrust of Brazilian education policy over the next decade should be on access and quality of non-elitist post-elementary education, roughly grades 5-11. While the literature on human capital and growth has dwelt on the contribution of human capital relative to other factors such as capital accumulation and technical change, the question of policy interest in most countries is what level of education-elementary, secondary, and tertiary-governments should subsidize at the margin. * Adapting the successful American model of non-elitist post-elementary schooling to Brazil to determine good practice in public education for economic growth implies that tertiary education enrollment expansion will not qualify as "non-elitist" in aggregate terms at current secondary school enrollment ratios. * The evidence for middle-income countries (including the cross-state analysis for Brazil reported in Chapter 4) appears to indicate that access to non-elitist post- elementary schooling is most likely to foster long-term economic growth without exacerbating inequality seriously. Therefore, despite the increasing private rates of return to university education in urban Brazil, the approprate level for closest govemment attention should be secondary education (consistent with the argument in World Bank, 1999). 6.108 Policies to encourage post-schooling human capital formation, principally those by firms investing in job-specific skills of workers, may be as or more important-especially over the short- term-than initiatives to increase schooling enrollments. * Post-schooling investment in human capital is generally ignored by the empirical literature because of severe measurement problems. This may not be a fatal flaw, 207 since education levels and on-the-job training are net complements for many workers (e.g., firms spend more on the training of more educated workers). In any case, anecdotal evidence indicates that post-schooling investments in human capital earnings functions are lower in Brazil compared with countries such as the US and Korea. If it is confirmed that post-school human capital formation rates are lower, then Brazil should examine how labor market reforms may increase these investments. 6.109 Policies to encourage growth of the Brazilian "innovation sector" include reform of the financing of higher level science and technology education in Brazil to encourage closer university- industry linkages, along the lines urged in World Bank, 2000b. The argument is not for overwhelmingly private higher education in Brazil, but one for greater autonomy and competition that encourages cost-sharing with students and firms. The evidence that leads to this conclusion comes in part from the study of the growth episode in the US after the Bayh-Dole Act of 1980 that allowed individuals and institutions to patent inventions made with public support, some comparisons with the less formal industry-university linkages in Japan, and Japan's passage of the Technology Transfer Act in 1998 in an effort to mimic the strong incentives for R&D inherent in the US system. * The distinctive features of the higher education system of the US are decentralization, competition, regionalism, and the coupling of research and graduate education. But given the state of pre-university education in Brazil relative to the US, the case for large-scale subsidization of tertiary education are much weaker in Brazil until far-reaching reforms are carried out for university financing, and secondary education enrollment increases to a majority of the relevant population group. 208 Summary Table of Human Capital Policy Recommendations ;'. ! ~ :_ ii. * ' '; i' ' i L' Pre-primary Large benefits of well-funded public early Maintain attention to early childhood Education and childhood programs in US and Brazil, especially programs. Increase funding for targeted early Health for disadvantaged children: schooling attainment childhood-especially preschoo- programs Interventions up, grade repetition down, earnings up and for disadvantaged children crime down. Benefit-to-cost ratios of 5 to 8 in the US Primary Increased primary net enrollment ratios during Continue policies in place since the early Education the last two decades-Brazil now does not lag 1990s; prepare to reallocate expenditures behind other countries in the region away from primary education as demographic transition occurs over the next decade Secondary Brazil underspends on secondary schooling and Increase effort to improve secondary Education has low net enrollment ratio relative to Latin education outcomes over the next decade, America average, expansion will be non-elitist starting with junior secondary education, until given near universal primary school enrollment net secondary enrollment ratios reach 100 since 1995, evidence indicates junior secondary percent. Careful sequencing is recommended, education has greatest (long-term) effect on not leapfrogging. growth in Brazil Higher Brazil overspends on higher education but has Aggressive reforms to facilitate private sector Education low enrollment relative to Latin America average, involvement. Increase financial autonomy for increasing private returns to tertiary education public universities and reduce public spending, over the last decade, more govemment with saving diverted to junior secondary spending will be elitist given low high school education. Encourage foreign investment in completion ratios. for-profit postsecondary education. Enterprise- When financed and provided by the private Identify and change labor regulations (e.g., based training sector, training has among the highest return of severance laws) that create perverse all forms of human capital investments. When incentives for firms and workers to keep financed and provided by the public sector, the employment relationships short. Limit public returns are close to zero. training programs to disadvantaged workers. Research and Institutional autonomy for research universities, Institutional autonomy for universities, and a post-graduate public funding for research with a liberal law allowing private researchers to profit from study intellectual property rights law such as the Bayh- inventions/innovations made using public Dole Act in the US promotes profitable research funds university-industry links. Matemal and High rates of return to public programs Continue policies in place. Child Health Adult Male Brazil lags being other Latin American countries Investigate i special attention to this is needed Health in male longevity and morbidity Tax Policy Progressivity in income taxes discourages Continue policies in place for the next human capital accumulation decade-the only major cost of human capital investment that is not tax deductible is college tuition 209 References Acemoglou, Daron and Joshua Angrist. 2000. "How Large Are Human Capital Extemalities? Evidence from Compulsory Schooling Laws." MIT (mimeo). Barro, Robert J., and Jong-Wha Lee. 2000. "Intemational Data on Educational Attainment Updates and Implications," Working Paper 7911, National Bureau of Economic Research. Bils, Mark, and Peter Klenow. 2000. "Does Schooling Cause Growth?" in American Economic Review, Volume 90 No. 5, pages 1160-1183. Birdsall, Nancy, David Ross and Richard Sabot. 1997. "Education, Growth and Inequality" in Birdsall and Jaspersen, eds., Pathways to Growth, Johns Hopkins Press for the Inter-American Development Bank. Birdsall, Nancy, and Richard Sabot. 1996. "Opportunity Foregone: Education in Brazil", Washington DC: Inter-American Development Bank. Birdsall, Nancy and Juan Luis Londono. 1998. "No Tradeoff: Efficient Growth Via More Equal Human Capital Accumulation." In Birdsall, Graham and Sabot, eds., Beyond Trade Offs - Market Reform and Equitable Growth in Latin America. Inter-American Development Bank, Brookings Institution Press. Blom, Andreas, Lauritz Holm-Nielsen, and Dorte Vemer. 2001. "Education, Eamings, and Inequality 1982- 1998." World Bank mimeo. Branscomb, Lewis, Fumio Kodama, and Richard Florida. 1999. Industrializing Knowledge: University- Industry Linkages in Japan and the United States. MIT Press. Dahlman, Carl and Thomas Andersson, editors. 2000. Korea and the Knowledge-Based Economy: Making the Transition. World Bank and OECD, World Bank Institute. DeLong, J. Bradford. 2001. "India Since Independence: An Analytic Growth Narrative." First Preliminary Draft: Not Yet for Citation (http:/Avww.i-bradford-delong.netl) Fiess, Norbert, and Indermit Gill. 2001. "Schooling and Growth in Brazil: the Long and Short of it," work in progress, World Bank. Goldin, Claudia. 1999. "Egalitarianism and the Retums to Education During the Great Transformation of American Education," Joumal of Political Economy, Vol. 107, no. 6. Goldin, Claudia. 2001. "The Human Capital Century and American Leadership: Virtues of the Past," Working Paper 8239, National Bureau of Economic Research, Cambridge, MA. Griliches, Zvi. 1997. "Education, Human Capital, and Growth," in Joumal of Labor Economics 15, pages S330-S342. Heckman, James. 1998. "Rethinking Myths about Education and Training: Understanding the Sources of Skill Formation in a Modem Economy," mimeo, Presidential Address, Midwest Economics Association. Heckman, James, and Peter Klenow. 1997. "Human Capital Policy", mimeo, University of Chicago. Kremer, Michael. 1993. "The O-Ring Theory of Economic Development," in Quarterly Joumal of Economics, pages 551-574. Krueger, Alan and Mikael Lindahl. 1998. "Education for Growth in Sweden and the World", Working Paper No. 41 1, Princeton University Industrial Relations Section, Princeton NJ. Lau, Larry, 1994." 210 Lee, Jong-Wha. 1997. "Economic Growth and Human Development in the Republic of Korea: 1945-1992." UNDP Human Development Report Occasional Paper 24 (http://www.undp.orglhdro/oc24aa.htm). Lucas, Robert E., Jr. 1988. "On the Mechanics of Economic Development." Joumal of Monetary Economics 22, pages 3-42. Mankiw, N. Gregory, David Romer, and David Weil. 1992. "A Contribution to the Empirics of Economic Growth," The Quarterly Joumal of Economics, pages 407-437. O'Neill, Donal. 1995. "Education and Income Growth", Joumal of Political Economy, Volume 103, Issue 6, pages 1289-1301. Park, Y. B., D. Ross, and R. Sabot. 1996. "Educational Expansion and the Inequality of Pay in Brazil and Korea," in N. Birdsall and R. Sabot, eds., Opportunity Foregone: Education in Brazil, Johns Hopkins University Press for the Inter-American Development Bank. Patrinos, Harry A. 2000. "The Impact of Technology on Education Premiums," Draft, Human Development Network, World Bank. Patrinos, Harry A. 2001. "Technology, Education and Economic Growth," Draft, Human Development Network, World Bank. Patrinos, Harry Anthony. 2001. 'School Choice in Denmark." Human Development Network, World Bank (mimeo). Plank, David N. 1996. The Means of Our Salvation: Public Education in Brazil, 1930-1995. Westview Press. Pritchett, Lant, 2000. "Where has all the education gone?", Draft: Kennedy School of Government and the World Bank. Rodrigues, Alberto, and Carlos Herran. 2000. Educa,co Segundaria no Brasil: Chegou a Hora, IADB and World Bank. Saxenian, AnnaLee. 2000a. "Bangalore: The Silicon Valley of Asia?" Stanford Institute for Economic Policy Research and University of California at Berkeley. Saxenian, AnnaLee. 2000b. "From Brain Drain to Brain Circulation?" University of California at Berkeley (Forthcoming in Kenneth Kenniston and Deepak Kumar, eds., Bridging the Digital Divide: Lessons from India, Bangalore: National Institute of Advanced Study, 2000). Schwartzman, Simon. 1998. "Higher Education in Brazil: The Stakeholders," LCSHD Paper No. 28, Latin America and Caribbean Region, The World Bank, Washington DC. Solow, Robert M. 2000. 'Toward a Macroeconomics of the Medium Term," in Joumal of Economic Perspectives, Volume 14, Number 1, pages 151-158. Temple, Jonathan. 2001. "Generalizations that aren't? Evidence on education and growth," European Economic Review, forthcoming. Topel, Robert. 1999. Labor Markets and Economic Growth, Chapter 44 in 0. Ashenfelter and D. Card, eds., Handbook of Labor Economics, Volume 3, pages 2943-2984. Velloso, Joao. 1994. "Higher Education in Brazil: Trends and Recent Developments in Resource Allocation Policies," IIEP. Venniker, Richard. 2001. "Social Returns to Education: A Survey of Recent Literature on Human Capital Externalities." CPB (Netherlands Bureau for Economic Policy Analysis) Report 00/1 (htto://www.cDb.nl/eng/cpbrepor/2000 1/s3 4.Ddf). Woessmann, Ludger. 2001. Schooling Resources, Educational Institutions, and Student Performance: The Intemational Evidence." Kiel Institute of World Economics. 211 Wolff, Edward N. and Maury Gittleman. 1993. "The Role of Education in Productivity Convergence: Does Higher Education Matter?", in A. Szimai, B. van Ark, and D. Pilat (eds.) Explaining Economic Growth, Elsevier Science Publishers. Wolff, Edward N. 2000. "Human capital investment and economic growth: explorng the cross-country evidence," in Structural Change and Economic Dynamics 11, pages 433-472. World Bank, 2000a. Brazil: Early Childhood Development: A Focus on the Impact of Preschools, Washington DC. Wordd Bank, 1999. Brazil: Higher Education Sector Study, Washington DC. World Bank, 1998. Brazil: The Health System. Impact Evaluation Report No. 18142. Operations Evaluation Department. Washington DC. Young, Alwyn. 1995. 'The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience," in Quarterly Joumal of Economics, Volume 110, Issue 3, pages 641-680. 212 7. THE INVESTMENT CLIMATE IN INTERNATIONAL PERSPECTIVE Prepared by Hoon Sahib Soh 7.1 This paper analyzes the relationship between public governance and investment in Brazil. The first section evaluates the quality of public governance in Brazil, focusing on measurements that are relevant to private sector business activities. It describes the available governance measurements, compares Brazil's ratings with other countries, and suggests priority areas for improvement. The next two sections analyze the impact of governance on entrepreneurial investment and foreign direct investment, respectively. The section on entrepreneurial investment evaluates the regulatory burden faced by entrepreneurs starting up a new business in Brazil. The section on foreign direct investment analyzes the impact of FDI inflow for Brazil and how the quality of governance has affected this flow. We concentrate on these two aspects for two reasons. First, these are areas that are particularly important for productivity gains through innovation. Second, these are areas where it is possible, through surveys and quantitative methodology, to measure the impact of regulations on businesses. 7.2 Public governance has been defined as the "exercise of authority through formal and informal traditions and institutions for the common good."'47 This broad definition encompasses concepts such as transparency and accountability of public institutions, quality of policy making and public service provision, and effectiveness of legal and regulatory systems. Such concepts are inherently difficult to measure. Surveys, based on interviewees' assessments of govemance, have traditionally been the common method, with their obvious drawback of subjectivity. 7.3 A more objective method is to measure the underlying variables that determine overall govemance quality. These variables include govemment processes and institutional arrangements: examples are civil service pay and the size of government expenditure. Such variables have the advantage of providing policymakers with information on which they can directly target policies, and they are more conducive to objective as opposed to subjective methods of measurement. Their disadvantage is that interpreting the policy implications of these variables can be unclear. Decreases in court case delays can result from decreased demand owing to deteriorating judicial services or from improved efficiency of the court system. The policy implications are very different depending on the interpretation. 147 Thomas, Vinod et al., "The Quality of Growth," World Bank and Oxford University Press, 2000. 213 7.4 Tables Al and A2 (Annex) provide a list of currently available public governance measurements, which are organized into the two aforementioned groups, overall measures and measures of underlying determinants of govemance.quality. If a data set has a mixture of both types of measurements, then it is included in table 1A but marked by an asterisk. The tables also list data sets that measure public governance variables only for Brazil. 7.5 This section analyzes overall governance measurements, and compares Brazil to other countries. Govemance measurement for Brazil focuses on three broad dimensions: the legal system, the regulatory system, and government bureaucracy. All three areas affect the quality of policies, public service provision, and the overall investment climate. Table 1 summarizes the main results from the most widely used rating agencies. Table 7.1 Governance Ratings for Brazil Legal Regulatory Bureaucracy / Public Institutions Average compared to similar Poor ratings for general Average compared to similar LAC LAC countries regulatory burden but average countries Surveys judicial independence, for others Surveys independence of policy, GCR likelihood of out of court Surveys time spent dealing with institutional stability, police settlement, litigation cost, govemment bureaucracy, protection, pervasiveness of corruption regulatory burden, private irregular payments property rights, stringency of Chile has high ratings regulatory standards Poor rule of law, which N/A Poor bureaucratic quality, lowest ICRG assesses the law and order among compared countries I tradition of the country Poor contract enforceability; Poor policy enforceability; lowest Poor bureaucratic quality; lowest BERI lowest rating among compared rating among compared rating among compared countries countries countries _ 7.6 The Global Competitiveness Report (GCR) evaluates the general business environment of each country, where business environment is determined by legal, regulatory, and bureaucratic dimensions, as well as physical infrastructure, education and technology, and financial market development. Brazil ranks 32 among 57 developed and developing countries for its business environment. Similar upper middle income countries'" are ranked as follows: 24 for Chile; 44 for Argentina, 43 for Mexico, 28 for S. Korea, 30 for Malaysia, 31 for Hungary, 34 for Czech Republic, and 42 for Poland. Brazil's business environment ranking is average compared to other countries with similar per capita income. Chile has a relatively high ranking within this group: this country could be a good benchmark for Brazil. 7.7 Figure Al (Annex) presents GCR measurements of legal system quality: judicial independence, the likelihood of disputes being settled out of court, litigation cost, and corruption in 148Using the guidelines in World Development Report 2000, upper middle income is defined as 1998 GNP per capita between $3,031 and $9,360, where exchange rate is calculated using the World Bank Atlas method. 214 the legal system. Brazil compares fairly well with other Latin American countries as well as the East Asian countries represented by Korea and Malaysia. In particular, Brazil has measurements similar to Chile, and both countries have better measurements than Argentina and Mexico. The GCR data thus do not support the perception that the legal burden in Brazil is severe. Given that cost of litigation includes both the duration of the legal process as well as attomey fees, the perception that the court system in Brazil is beset by long delays is not supported by the GCR survey results, unless one assumes that such problems are just as prevalent in Chile or Korea. 7.8 Figure A2 (Annex) presents the GCR results for the regulatory system: time spent dealing with government bureaucracy, general burden of regulation, clear delineation and enforcement of private property rights, and stringency of regulatory standards. As with the legal system, the regulatory system measurements suggest that Brazil has regulatory quality comparable to similar countries, with one important exception. Among the countries compared, Brazil has the poorest rating for the general burden of regulations. In this respect, its rating is similar to the ratings of much poorer countries such as India and Mauritius. 7.9 The public institutions variables, as shown by figure A3 (Annex), indicate that Brazil received similar ratings to the other Latin American countries except for Chile. These measurements consist of independence of policy, institutional stability, police protection, and pervasiveness of irregular payments. Brazil has poorer ratings on all four of these measurements than Chile. Indeed, Chile has public institutions that are also higher in quality than those of East Asian and East European countries. 7.10 Figure A4 (Annex) presents three areas of public govemance rated by ICRG: rule of law, bureaucratic quality, and corruption. Again, Chile received relatively high ratings, while Brazil generally received ratings lower than similar East Asian and East European countries. The rule of law variable suggests that Brazil has a poor legal and regulatory environment overall: its rating is substantially lower than all the other countries except for Mexico. Furthermore, the bureaucratic quality variable for Brazil is the lowest in the group. 7.11 Figure 7.1 presents three measurements of public governance by BERI: contract enforceability, policy enforceability, and bureaucratic quality. Again, Brazil has poor ratings for these measurements, while Chile generally received ratings higher than similar middle-income countries. Indeed, Brazil has the lowest rating on these measurements in the group. As in the ICRG survey, Brazil's bureaucratic quality is the lowest of the group. 215 Figure 7.1: BERI Public Governance Measurements 4 - - -~~ ~ r -;. - v 3.5- 2. Contract enforceability 3- 2- I Policy continuity 0 03 Bureaucracy co~ ~~ U) 0 )757 7.12 Other agencies provide similar evaluations. Price-Waterhouse-Coopers (PWC) publishes country governance ratings based on interviews of firms, equity analysts, banks, and PWC employees. The overall rating, shown in Figure A6, is based on five categories: corruption; legal system; govemment macroeconomic and fiscal policies; accounting standards and practices (including corporate governance and information release); and regulatory regime. The rating is aggregated from the five categories with equal weights, where lower scores indicate higher quality. According to PWC, the quality of governance in Brazil is similar to Argentina but poorer than Mexico and Chile, and similar to the representative East Asian and East European countries. PWC's assessment is similar to GCR ratings presented above. Both studies suggest that the overall business environment of Brazil is comparable to similar countries. Again, Chile stands out as having a high quality of public govemance, with ratings similar to the United States. 7.13 Given that the overall rating is an aggregation of the ratings on five different subcategories, it is of interest to examine the subcategories. Figure A presents these for Brazil and other Latin American countries. For all five subcategories, Brazil has ratings similar to Argentina but lower than Mexico and Chile. In particular, Brazil has a much lower rating for the fourth category, accounting standards and practices, corporate governance and information releases. This paper will not delve into issues specific to accounting standards and corporate governance, but these issues are intricately related to public governance. For example, enforcement of shareholder rights depends on the quality of the legal and regulatory system, and the independence of firm activity from govemment interference will depend on transparent corporate governance structure.'49 7.14 The main conclusions from the various public governance assessments can be summarized as follows: 149 Refer to 'Reforming Public Institutions and Strengthening Governance," page 14. 216 * The evidence is mixed regarding the quality of Brazil's governance compared to similar middle-income countries, but Brazil consistently received ratings lower than the developed countries. o Compared to Latin American and other middle-income countries, Brazil received poor ratings from BERI and ICRG, but ratings from poor to average from GCR. BERI and ICRG survey experts, whereas GCR surveys business people and govemment officials. * Among the three subcategories, Brazil received a particularly low rating on its regulatory system. o The high regulatory burden is a theme repeated throughout the rest of this chapter, where evidence is presented to indicate that the regulatory burden acts as a major barrier to both entrepreneurial and foreign investment activities. It should be noted that this conclusion is mainly derived from the GCR results. * Among the middle-income countries, Chile stands out for high-quality public govemance. o Chile's ratings from PWC were similar to the ratings for United States. Similarly, Chile received high ratings for all categories from BERI and ICRG, and for bureaucracy and public institutions from GCR. Lessons drawn from Chile's public sector reforms might prove useful for Brazil (Box 1): these reforms reduced the discretionary power and increased the transparency of the public sector. 7.15 The above provide a direct assessment of public govemance. On the other hand, these indicators do not provide clear policy implications, as they measure broad phenomena affected by many factors. The problem remains to assess these underlying institutional factors directly. 7.16 One possible strategy is to use more finely disaggregated subjective assessments. The GCR, as described in the previous section, implements such a strategy. For example, instead of a single "rule of law" index, GCR provides multiple measurements such as judiciary independence, use of altemative dispute resolution mechanisms, litigation cost, and corruption in the court system. 7.17 An alternative strategy is directly to measure presumed determinants of govemance quality. One must form hypotheses about the relationship between a particular legal, regulatory, or institutional arrangement and govemance quality. Because govemance reform policies target these underlying arrangements, such measurements can more easily result in policy implications. If the hypothesis is accepted, these measurements also allow one to verify reform progress. 7.18 The following subsections presents two cross-country data sets that measure such determinants. The first data set measures court performance; the second measures regulatory burden for start-up entrepreneurs. 217 Box 7.1: Chile's Public Sector Reforms'5` It can be argued that the quality of its public institutions is a major reason why Chile consistently receives high marks for its governance. The public institutions have been shaped by the reforms carried out by the military regimes during the 1970s and 1980s and the democratic administrations of the 1990s. These reforms resulted in a public sector characterized by institutional stability, independence of policies, and a relative lack of corruption. One of the key achievements of the reforms has been the reduction of discretionary power within the public sphere. In one sense, these reforms were based on a preference for fostering the "credibility" as opposed to the "flexibility" of the public institutions. The discretionary power of the state was reduced by implementing a uniform tariff which limited private sector lobbying for import protection, banning nontariff barriers, privatizing social security and consolidating the budget procedures in order to limit the amount of lobbying on fiscal matters, and institutionalizing the independence of the Central Bank in order to end the influence of private sector interests on credit policies. 15 Institutional mechanisms were installed which made it difficult to reverse the reforms, some requiring a three-fifths majority in Congress for amendments. Changes would have to be discussed openly in Congress, thereby reducing the scope for lobbying and backroom political deals. The budget process was centralized in the Ministry of Finance, which took precedence over the sectoral ministries in the formulation and execution of the budget, and the executive branch rather than the legislature had the last word in the approval process. Congress could only reduce items of expenditure proposed by the administration, and were prohibited from altering revenue calculations, increasing expenditures, or reallocating resources between programs. Strict limits were placed on the time allocated to congressional discussion of the budget. Congress had only 60 days to agree on the amendments to the budget bill once received, after which the executive's proposal was automatically approved.152 The highly hierarchical and centralized nature of Chile's budgetary institution limited opportunities for discretionary influences on the budget process. Along with reductions in opportunities for discretionary influences, efforts were made to increase institutional accountability and transparency by increasing the emphasis on results and performance. Starting with a pilot program launched in 1993, public services produced specific goals and management targets based on strategic planning. Indicators were developed which allowed for effective internal and external monitoring of the institution's operations, and the performance indicators were eventually incorporated into the budgetary process. All public services were required to prepare an annual performance report indicating the degree to which budget, program, and management targets have been met. 15 Transparency was also maintained during the democratic period through constant consultations with the business community. Although never institutionalized, major business associations had frequent access to the Ministries. Policymakers and business leaders would often negotiate on draft legislation, where the business association would make comments based on their own technical evaluations. Identification with any single conglomerate, which could result in lobbying opportunities, was avoided by consulting with business associations that represented the interests of a broader spectrum of the business community. The drawback of this arrangement was that labor interest was clearly subordinated to business interest.154 50 Marcel (1999), Foxley T. and Sapelli (1999), and Silva (1997). 151 Foxley T. and Sapelli (1999), page 414. 152 Alesina et al. (1995) showed that Chile's budgetary institution had one of the most clearly established lines of authority and responsibility in Latin America. 153 Marcel (1999), page 318. 154 Silva (1997), pages 173-179. 218 Court Performance 7.19 The Legal Department of the World Bank has conducted a cross-country analysis of the performance of courts by measuring the determinants of the quality such as court delays and legal costs.155 The project reviewed the efficiency of courts in twelve developing and industrial countries. Efficiency was analyzed by measuring clearance rate, caseload, backlogs, and cost of supplying court services. Cost includes not only operational cost but also capital expenditure and opportunity cost. 7.20 Clearance rate, which is the percentage of filed cases that are resolved, measures the productivity of courts and judges. Figure A8 presents the data. The two industrialized countries in the sample have a clearance rate above 97 percent, while the figure for Brazil at around 90 percent is lower than Chile, Peru, and Singapore. The clearance rate of the Brazil court system is on the lower end of the range among similar countries but is higher than several countries, including Colombia, Hungary, Ukraine, and Ecuador. 7.21 The determinants of the clearance rate were analyzed using regression analysis. Increased capital spending, increased use of technology, and decreasing the proportion of time that judges spend on administrative tasks have positive and significant affects on the clearance rate. By contrast, the salary level of judicial personnel and the number of court staff do not have a significant effect on the clearance rate. These results raise concems about the effectiveness of traditional legal reform policies focusing on salary and staff. Instead, legal reforms should focus on increasing capital spending, technology, and administrative support for judges. 7.22 The time required to resolve cases is another factor in the quality of the court system, particularly for developing countries where delays in the legal process are often cited as a major problem."' Unfortunately, a cross country data set of the time required to resolve cases is not available. On the other hand, the number of pending cases affects the time it takes to resolve a case. Figure A9 presents data for the number of pending cases per judge for Brazil and other countries of similar and higher income levels. Along with Chile, Brazil clearly has substantially higher number of pending cases relative to countries in other regions. In particular, Peru has a low figure, which could have resulted from experiments with major legal reforms, including the use of temporary judges to resolve backlogged cases. 7.23 Another indirect measurement of the time required to resolve cases is the "backlog index." The backlog index is the number of cases pending at the start of the year divided by the number of cases resolved during the year.'7 A higher ratio implies longer civil case processing time. Figure 155 Dakolias (1 999a, 1 999b); The project chooses to focus on measurements of efficiency of the legal system because it provides a quantifiable means of comparing courts worldwide, but this does not mean that other dimensions of legal system, such as judicial independence and procedural transparency, are not important determinant the overall quality of the judicial system. Hence, if the study determines that certain variables are deemed insignificant determinants of judicial system efficiency, policymakers should keep in mind that such variables might still have significant affects on other aspects of judicial quality. 56 Pinheiro (1999), and Buscaglia and Dakolias (1996) 57 National Center for State Courts, Examining Court Delay, The Pace of Litigation in 26 Urban Trial Courts (1987) in Dakolias (1 999b) 219 7.2 presents the data. Among the compared countries, the court systems in Brasilia and Sao Paolo have the highest and third highest figures. Both the number of pending cases per judge and the backlog index indicate that court delays in Brazil are comparatively serious. Figure 7.2 World Bank Legal Department Backlog Index 2- 1.8 4. 1.6 ~ 1.4 - 1.2 , X 0.8 0.6 0.4 0.2 - O tD~~~~ O 2 ? I ° Cu .~~~ ~ 2 . 2 ct N Cu ~~~a E & co co 0 0 7.24 The World Bank Legal Department also analyzed the backlog as the dependent variable of a regression. The results indicate that the capital budget, technology, managerial activism, and cost per case are significant determinants of case processing time, while the number of court staff, salaries, and the level of training do not appear to be significant. In a related analysis, the results indicate that the more judges rely on managerial activism, the greater the capacity of budgetary resources to increase the supply of court services. Court systems with larger capital budgets and greater use of information technology have shorter expected duration of cases. Again, the study raises issues about the effectiveness of traditional legal reform policies that focuses on salary and staff numbers. Regulatory Burden for Start-up Entrepreneurs 7.25 As with the legal system, measurements of the quality of regulatory systems have relied mostly on subjective assessments, but a recent joint study by Djankov, La Porta, Lopez-de- Silanes, and Shleifer focuses on objective measurements. The study analyzed the entry of start-up firms in 75 countries, including Brazil, and collected data on required procedures, official time, and official cost. The advantage of relying on official figures is that the variables are easily measured but maintain objectivity; the disadvantage is that the actual regulatory burden faced by firms could differ from official figures. 7.26 The data suggests that the official cost of entry is high in most countries, and that countries with heavier regulation of entry have higher corruption and larger unofficial economies. Compared to other countries, Brazil requires a large number of days in order to complete the start-up procedures. A more detailed analysis of the cost of entry will be presented in the following section. 220 7.27 In summary, the cross-country analysis of the determinants of governance quality reveals: * Court delays are a serious problem in Brazil where, compared to other countries, the court system requires substantially higher number of days to resolve cases. * The capital budget, technology, managerial activism, and cost per case are significant determinants of case processing time, while the number of court staff, salaries, and the level of training do not appear to be significant. * The time required to complete regulatory procedures relating to business investments in Brazil is much higher than in other countries. 7.28 So far this chapter has analyzed governance as it pertains to private-sector activity on a general level, without focusing on any particular aspect of the Brazilian economy. We now turn to a more detailed analysis of the last point above-the impact of governance on investment activities- and focus on two types of investment that are particularly important for Brazil: entrepreneurial investment and foreign direct investment. 7.29 This section first presents survey results from two sources, the World Business Environment Survey (WBES) and the GCR. The section then presents an analysis of the data set collect by Djankov et al. (2000). 7.30 The World Bank conducts an on-going cross-country survey of the general business environment called the World Business Environment Survey (WBES). The project surveys at least 100 firms for each country and surveyed 200 in the state of Sao Paulo. In particular, the WBES asks firms' managers about the relative importance of various constraints to the operation and growth of their businesses. Figure All summarizes the survey results. Taxes and regulations, and political instability were considered the most serious obstacles to conducting business in Brazil. Other govemance related measurements such as the judiciary system, corruption, and crime receive ratings received lower ratings. The results suggest that regulatory burden is one of the most problematic constraints for business operations, but remains inconclusive because the survey combines the categories tax and regulatory burden. 7.31 Figure A12 presents the same survey data for Brazil versus other middle-income Latin- American countries. Tax and regulatory burden are cited as the most problematic constraint for Argentina as well, whereas political instability and inflation are cited as the most problematic for Chile and Mexico, respectively. Tax and regulatory burden for Brazil received the highest rating among all countries and across all categories. In the remaining survey results, the most striking difference between Brazil and non-Latin American countries is the high degree of importance assigned to political instability by Brazilian respondents. 7.32 The annual GCR survey includes questions concerning business environment constraints for new businesses. It asks for an assessment of the degree to which unfavorable economic and financial conditions, burdensome administrative procedures, and lack of legal protection are obstacles for new businesses. Figure A13 presents the results. Higher ratings are better. Brazil is 221 similar to other Latin American countries except for Chile, which in general has higher ratings. An interesting observation is that a lack of legal protection is considered by far the least important obstacle to starting up a new business in all the countries. Also, regulatory burden, measured by administrative procedures, is as much of an obstacle as the economic and financial conditions. 7.33 In general, these two firm level surveys fail to provide conclusive evidences on the importance of regulatory obstacles to starting up a new business in Brazil, although they hint at its importance. By contrast, the new data set collected by Djankov et al. (2000) on the regulatory burden for business start-ups in 75 countries strongly indicates that the official cost of entry for Brazil is high.'18 7.34 Figures A14 through A16 (Annex) present the cross-country data on the number of procedures, business days to complete the procedures, and cost of completing the procedures, to start up a new business. Brazil's official number of procedures, at 11, is approximately average compared to other large, middle-income Latin-American countries. Chile has the lowest number of procedures. For the two representative Asian countries in our sample, Korea has the same number of procedures as Brazil, whereas Malaysia with 7 has a lower number. Developed countries, represented by Ireland (chosen for its recent dynamism) and the US, require only 3 procedures. 7.35 The cost of starting up a new business is standardized for comparative purposes by dividing by per capita GDP. Figure Al 5 presents these figures. The figures ranges from a low of 0.5% for the US to a high of 55.5% for Mexico, although the next highest is 26.5% for Malaysia. Brazil at 13.4% has cost comparable to Argentina, Chile, Korea, and Ireland. The official cost of starting up a business in Brazil is similar to even the higher growth countries such as Korea and Ireland. Figure 7.3 Number of Days to Complete Procedures for New Businesses'59 90 80 70 60 50 40 30 20 - 10 0 -- X C R 0 y 158 The data set was kindly made available by Mr. Djankov. 159 Djankov et al. (2000). 222 7.36 Brazil compares most poorly with respect to the number of business days required to start up new businesses. Figure 7.3 presents the figures. With 82 total official days required to start up a new business, Brazil has the highest figure compared to all the other countries in our sample. Excluding Brazil, the other three Latin American countries have an average of 47 days, where Chile at 27 days has the lowest requirement. Korea requires 21 days and Malaysia requires 41 days, both substantially lower than Brazil. Ireland requires 11 days, while the US requires 2 days. 7.37 Summarizing, the analysis of the regulation of entry data indicates that more than the cost of procedures or the number of procedures, Brazilian entrepreneurs are most burdened by the time required to comply with government regulatory procedures concerning start-up businesses. 7.38 The source of the differences in the required procedures can be analyzed using Figure 7.4. These figures show the detailed start up procedures for Brazil, Chile, and Ireland, respectively. Chile, as is often the case with many govemance measurements, has the least obstructive procedure among the middle income Latin American countries, while Ireland is presented in order to show how procedures in a developed countries enjoying high growth differs from Brazil. 7.39 Focusing first on the time component, the figure for Brazil shows that most of the time required to start up a new business comes from the earlier half of the procedures. In fact, steps four and five alone, each requiring 22 business days, comprise 52% of the total number of days. Step four consists of registering with the state tax authorities, the coletoria estadual da fazenda, where registration is done at three different offices. Step five consists of registering with the municipality and receiving an operations permit, an Alvara de Funcionamento. Besides steps four and five, steps one through three each requires 10 business days. Step one consists of registering with the state commercial board, step two consists of ordering receipts and invoices from authorized printing companies, and step three consists of registering with the municipal taxpayers registry, and receiving approval for the authorized receipts and invoices. 7.40 In contrast to Brazil, none of the steps in the start up process for Chile requires more than 10 days. Step seven, obtaining an operations license from the municipality, requires 10 business days, while step two, publishing the deed in the official gazette, requires 4 days, and step three, registering in the commercial state registry, requires 4 business days. All the other steps require only 1 business day each. The contrast with Ireland is even more striking. The entire process requires only three steps. The first step, filing necessary documents such as the article of association with the Registrar of Companies, takes 10 business days, while the other two steps take 1 day each. Actually, the last step consisting of registering income, social insurance, and VAT with the Revenue Commissioner is performed online and hence takes less than a day. file applicafion with register wifth registrar notarize deeds -- Commercial Board print authorized make a oompany seae publish deeds receipts & invoices register inorporaion register register for -n empoynen pymnts inoprto eitrmunicipal taxes print invoicreceipts register for state taxes register lor taxes munidpal permit w munidpal inspection 0 recatinspeciopa rt 0.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. municipa \ - - - - registerem ployees patents or sodal security t ~ r; - r_- = ~ - ~ 1- - 0 approval of register for Isapre - employee registry book - ~~ registerfor l-l ~~~~~~~~~~~~~~open umemployment ~ accidert insurance - t T account - -- , - _ I registersociallntegraton -t .. . . . . . . . { ..... . '~ . ' _ . . . ......... ... J . . r ~ .. rn . . (PlS) pfrogr( m _program * . . ; * - . .- - _. - .- notify Ministry of Labor - | * | | | -. -. . 0 0 0 0 0 0 0 0 00;9 00' .. o p p o - t ) CODi Xvr ooo0oo0o M~~~~~~~~~~~~~~~O Q O 224 7.41 The comparative analysis suggests some specific ways of reducing the regulatory burden for Brazilian entrepreneurs starting up businesses. * In general, the start-up process should be streamlined and centralized, including through the use of online services. o For example, entrepreneurs in Brazil have to deal with up to 6 govemment agencies. In comparison, entrepreneurs in Ireland deal with 2 agencies, where one of the agencies is accessed through the intemet. * Efforts to streamline the regulations goveming the start-up process will no doubt be hindered by Brazil's multi-layered govemment organization. The fact that entrepreneurs have to deal with multiple agencies spread throughout various national and subnational governments clearly increases regulatory burden. In the United States, where the government is also organized according to a fiscal federalism structure, consistently requires the least amount of regulatory burden in the start-up process. Hence, it would seem that the process can be improved in Brazil despite its multi-layered government structure. * The tax registration process should be better integrated into the rest of the start-up regulatory process. o The time required to register for municipal and state taxes in Brazil, combining for 32 business days, is a burden that is disproportionately large compared to the situation in Chile and Ireland. In Chile, entrepreneurs register at the Single Tax Register (Rol Unico Tributario), which requires 1 business day, while in Ireland tax registration is performed online. 7.42 The results from the Djankov study are generally corroborated by a recent study'" conducted by the Foreign Investment Advisory Service (FIAS).16' The FIAS study similarly argues that the lengthy registration process characterized by multiple procedural requirements is a major impediment for the private sector in Brazil. It recommends establishing a single contact point for business registration by combining all company registration in one authority, merging the company and tax registration process, streamlining information exchange procedures among involved authorfties in order to reduce unnecessary duplicative efforts by firms, and reducing excessive documentation requirements. 7.43 In the previous section, it was noted that delays in the court system was often cited by experts as the most significant shortcoming of the legal system. The present section documented the delays caused by the regulatory system. These delays are symptoms of a governance structure unable to support quick resolution of business activities, thereby adding cost, uncertainty, and suppressing entrepreneurship. The inefficiency of the regulatory and legal system not only 160 "Legal, policy, and administrative barriers to investment in Brazil," FIAS, May 2001, chapter IlIl, sections A and B. 161 FIAS is a unit of the World Bank Group. 225 dampens domestic private sector activities, but it also hinders domestic firms from effectively competing in the intemational markets. 7.44 One caveat is that the figures from the Djankov study are official figures, and therefore the actual regulatory burden for Brazil relative to other countries could be different. On the other hand, given that irregular payments for licenses and permits are considered more common in Brazil compared to developed countries, as indicated by the GCR data, the actual regulatory burden for Brazil relative to developed countries is likely to be even greater than the official figures. 7.45 The World Bank's Global Development Finance presents some general statements conceming the relationship between capital flows and investment. They are listed below and may serve as useful starting points:1" * Long term capital flows are significantly associated with domestic investment, whereas the relationship between short term flows and investment is considerably weaker * FDI is clearly associated with increases in investment compared to portfolio flows 163 * In the short term, FDI through mergers and acquisitions may have less of an impact on domestic investment than "Greenfield" FDI, but the long term comparisons are ambiguous '6 * Domestic structural factors such as govemance, combined with others such as the level and quality of education, quality of physical infrastructure, and financial market development determine the host country's absorptive capacity, or the ability to convert the foreign capital into productive investments. 7.46 The evidence presented in this section shows that FDI is closely associated with domestic investment and that the quality of governance affects the amount of FDI inflow as well as its impact on domestic investment. The section first describes and analyzes the FDI inflow data for Brazil, then presents various qualitative and quantitative observations that support the existence of a positive relationship among govemance, FDI, and investment, then concludes with some policy implications. 162 Refer to World Bank's Global Development Finance 2001, chapter 3. 163 Bosworth and Collins (1999) finds that FDI and bank lending has a strong impact on domestic investment, where portfolio flows has a positive but insignificant impact. 164 The relationship between FDI and domestic investment has weakened in countries where the proportion of M&A has risen, such as countries in East Asia and Latin America. However, M&A might significantly increase long term productivity. 226 The Size of FDI Inflows 7.47 Worldwide foreign direct investment increased substantially over the recent years, but the share of FDI flowing to developing countries decreased from 35 percent in 1994 to 16 percent in 2000, reflecting a more competitive environment as well as the completion of many large scale privatization projects in developing countries. The ownership form of foreign investment has changed as well in the late 1990s, as M&A activities grew rapidly in both Latin America and Asia, and especially in Brazil. 7.48 Cross-border M&As, including the privatization of state-owned enterprises, rather than "Greenfield" investments have been the main driving force behind the latest rise in FDI. Although it is impossible to measure cross-border M&A's share of FDI inflow precisely because of measurement incompatibilities,165 if we assume that all cross-border M&As are financed through FDI, then the share reached over 80 percent in 1999. For developing countries, where "Greenfield" foreign investment is still dominant, the share is lower than the world average. Among the developing countries, the Latin American and Caribbean countries, especially Brazil and Argentina, dominate cross-border M&A sales. 7.49 For Brazil, FDI has increased significantly over recent years, as shown by Figure A19. In contrast, international portfolio equity flows have remained stagnant, raising speculation that weaknesses in the corporate govemance system in Brazil have held back development of domestic equity markets. Brazil has been successful in attracting non-portfolio FDI, but these flows' sustainability remains unproven, and we turn to this in the next section. 7.50 In addition to the size of FDI, the host country's degree of involvement in international production is another indicator of FDI performance. A comparatively low level of involvement in international production would indicate that the host country lags other countries in its ability to benefit from the intemational production system, which in turn indirectly indicates low levels of FDI. UNCTAD publishes a Transnationality Index, which measures the extent of a country's involvement in international production.'6 The index is an average of the following four measurements: FDI inflows as a percentage of gross fixed capital formation between 1995 and 1997; FDI inward stock as a percentage of GDP in 1997; value added of foreign affiliates as a percentage of GDP in 1997; and employment of foreign affiliates as a percentage of total employment in 1997. Among 30 developing countries, Brazil, at approximately 7 percent, ranks 25th, well below the average of 14 percent. Chile is ranked 7th at 17.5 percent, Mexico is ranked 16th at 11 percent, Argentina is ranked 18th at 10 percent, and China is ranked 14th at 12 percent. 165 Refer to World Investment Report, page 10, and Global Development Finance 2001. M&As financed locally or through the intemational capital market is not included in FDI data. FDI is reported on a net basis while M&A is reported on a gross basis. The total amount of an M&A transaction phased over several years is recorded in the first year in the FDI data. 166 UNCTAD, World Investment Report, 2000. 227 The Sustainability of FDI 7.51 Figure 7.5 shows that Brazil's recent strong FDI performance. It plots FDI inflow, as a percentage of GDP, during the 1990s for Brazil and comparable Latin American and Asian countries. The figure makes it clear that for most of the decade, compared to the size of its economy, Brazil lagged behind the other countries except for South Korea. The ratio of FDI to GDP for Brazil surpassed Mexico and Argentina only in 1998, then outpaced China in 1999. In particular, for the second half of the decade, Chile outpaces Brazil as well as other comparable countries in its ability to attract FDI.'67 The ratio of FDI to gross fixed capital formation paints a similar picture, and shows that Brazil lagged other comparable countries until the late 1990s. Figure 7.5 FDI as percentage of GDP 16~ ~ _ 14- 6 Brazil 12- -u- Brazil w.o. priv. 10/ - ChiArgentna 8 X X i-Medco --o Korea X cm N to co N w 0 ) 0) 0) 0 0a 0) 0) 0 ) 7.52 One of the issues that affects FDI sustainability is the degree to which Brazil's FDI relies on privatization. The recent increase in FDI coincided with the massive privatization program, as illustrated by Figure 7.6. 167 Chile's ability to attract proportionately more FDI could be the result of policies that specifically deal with foreign investors as well as the high quality of its broader regulatory, legal, and institutional environment, but the fact that a large proportion of its FDI targeted the mining and, in 1999, the utility sector suggest that the recent increases might not be sustainable. The figures for 2000 indicates a sizable decrease in the FDI inflow, although as a percentage of GDP the figure is still at the upper range among comparable developing countries. 228 Figure 7.6 Prazil FDI Inflow 40000 35000 A,; 30000 -. .... te 25000 Mohe 20000 otherfl 15000 U. ~ privetization 15000 10000 5000 0 U) a) t. CD Co 0 a) oD CD Ca) ax 0 - N c 7.53 On the other hand, the figure also shows that FDI unrelated to privatization has increased at a faster rate than FDI resulting from privatization. Privatization related FDI accounted for an average of approximately a quarter of the total FDI inflow during the second half of 1990s. While privatization related FDI decreased in 2000, non-privatization FDI increased unabated. It indicates that the recent increase in FDI inflow into Brazil was not merely the result of privatization of state owned enterprises, although it clearly benefited from the privatization program. It provides hope that the recent high levels of FDI could be maintained despite the inevitable decrease of privatization related FDI. 7.54 Regardless of its sustainability, it is clear that FDI has increased substantially over the recent years. Then the relevant question is whether this increase has been matched by a similar increase in domestic investment. Figure 7.7 plots FDI and gross fixed capital formation for Brazil over the 1990s. As before, the figure shows that FDI increased substantially over the recent years, while gross fixed capital formation increased at a decreasing rate, eventually almost flattening out over the last three years. During this period, gross fixed capital formation as a proportion of GDP steadily decreased, from a peak of 20.75 % in 1994 to 18.92 % in 1999. Hence, whether in absolute amount or proportionate to the size of the economy, investment as measured by gross fixed capital formation has not increased as much as the increase in FDI. 7.55 There are three possible explanations for the lack of growth in domestic investment in response to the increase in FDI: (1) FDI attracted by the privatization programs have largely been used to finance govemment consumption; (2) FDI has largely been financed through mergers and acquisitions which bring less resources to the host country compared to "Greenfield" FDI; (3) demand for investment funds is inelastic with respect to the cost of funds, and therefore an increase in supply of investment funds caused by an increase in FDI led to a relatively large decrease of the cost but a small increase in investment volume. In another words, investors are not price sensitive. 229 Figure 7.7 Domestic Investment and FDI 200 180 ' 160 140 -0 120 Cap. 100 ---FDI 80a FDI iIw.o.pr. 60- 40- 20 0 CD N W 7.56 The first hypothesis, although plausible, cannot by itself explain for the less than expected increase in domestic investment because, as discussed previously, privatization accounts for only a quarter of Brazil's total FDI inflow. 7.57 The second hypothesis could possibly account for the lack of response in domestic investment. If we assume that cross-border M&A results in small additions to the capital stock, then FDI composed largely of mergers and acquisitions would result in less than expected increases in domestic investment. Although FDI and cross-border M&A data are not completely compatible, Figure A23 suggests that a large proportion of FDI has been financed through M&As. On the other hand, even if Brazil's FDI inflow has been mostly accounted for by M&As, the explanatory power of the second hypothesis hinges on the assumption that FDI through M&A results in small additions to the capital stock. This assumption depends on two factors: the time horizon and the response of the seller in the cross-border M&A transaction. 7.58 Time horizon matters in the following sense. Even if cross-border M&A initially resufts in a change of ownership accompanied by only a small initial increase in the domestic capital stock, the longer run impact on investment is unknown. The acquiring foreign agent in a cross-border M&A transaction often follows up the initial investment with a series of additional investments. The assumption that cross-border M&A results in only a small increase in the capital stock compared to "Greenfield" FDI, could be valid in the short run but invalid in the long run.'68 Hence, this delay in impact could explain why FDI has increased without a corresponding increase in domestic investment. 168 For the present argument, the comparison of the two modes of FDI entry focuses on increases in domestic investment while ignoring other effects on areas such as employment, technology transfer, and domestic competition. In the long run, the differences in these other areas could have a greater impact on the host country's economy than increases in the domestic fixed capital formation. 230 7.59 The response of the domestic seller also determines FDI's impact on domestic investment. The domestic seller can use the proceeds from the transaction in several ways. If we assume that the domestic seller reinvests all the proceeds in the domestic economy in such a way that adds to the capital stock, then FDI would result in an equally large increase in domestic investment. 7.60 The third hypothesis is based on assumptions regarding how domestic investors would respond to exogenous increases in the supply of investment funds. In the extreme case where the demand for capital is perfectly inelastic, an increase in FDI inflow would be exactly matched by a decrease in the supply of domestic investment capital, resulting in a "crowding out' effect. If the objective of the policymaker is to increase domestic investment, then the proper policy response would depend on one's beliefs regarding the determinants of the elasticity of the domestic demand for investment funds. The policymaker would need to know the factors that increase the sensitivity of investors to the cost of capital. The Sector Distribution of FDI 7.61 Figure A24 shows the distribution of Brazil's FDI inflow into different sectors. For the latter half of 1990s, the service sector received approximately 80 percent of the total FDI inflow. By contrast, the stock of FDI in 1995 was 43 percent in the service sector, 55 percent in the manufacturing sector, and 2 percent in the agriculture and mining sector. The FDI share of the service sector is a more recent phenomenon. The services sectors in Argentina and Chile also receive a much larger share of FDi, whereas the service and manufacturing sectors in Korea receive approximately equal amounts, and the manufacturing sector in Thailand receives 70 percent of FDI. 7.62 The rapid increase of FDI inflow into the service sector was driven mainly by only two industries. By 2000, the telecommunications industry received 45 percent and the financial intermediation industry received 26 percent of all service sector FDI. 7.63 The fact that the manufacturing sector receives a much smaller share of FDI is a matter of concern for policymakers who are interested in export promotion. Although services are becoming increasingly more tradable, manufacturing has been the traditional engine of export growth for developing countries. In addition, many services still require proximity to the customers, limiting their capacity to contribute to exports. Finally, developing countries usually lack the higher levels of skill and technology necessary for export of services. The rapid growth of the NIEs (Newly Industrialized Economies) of East Asia and China are recent examples of manufacturing export based economic growth. In particular, China has enjoyed sustained high rates of growth based on surging foreign investment in its manufacturing export industries (Box 2). The experience of China seems particularly relevant for Brazil given that China also has a large domestic market and yet chose to pursue a growth strategy based on manufacturing export, with foreign firms playing a major role in the export manufacturing industries. 231 Box 7.2: FDI in China's domestic manufacturing export industries'69 Foreign investment in its manufacturing export industries has been one of the most important factors behind China's sustained high rates of growth in the recent years. China's economy grew at an average of approximately 10% in the 1990s, as it became the world's ninth largest exporter in 1999, up from twelfth in 1990. The sector composition of China's exports consists mostly of manufacturing products such as general and electrical machinery and textile products. The Chinese industrial development process has been characterized by the heavy involvement of foreign direct investment. Subsidiaries of foreign affiliates accounted for 46% of its exports, 52% of its imports, and 12% of its fixed capital formation in 1999. In addition to direct contributions to trade and investment, the economic integration of the foreign affiliates in the local economy has resulted in backward and forward linkages and the creation of industrial clusters. Foreign processing and assembly industries have led to the formation of local parts companies. Clusters of such companies have led processing industries to form similar clusters nearby, leading to a virtuous cycle. The large-scale involvement of foreign firms in the domestic export industry has been accompanied by continued efforts to improve the investment and trade environment. The average tariff rates have decreased from nearly 40% in 1988 to below 20% since 1998, import quotas and import licenses for items such as heavy machinery, electrical machinery, and optical instruments were eliminated in 1997, and certain foreign banks were allowed to operate local currency businesses in Pudong, Shanghai in 1998. Although still far from a market-based economy, China's campaign for membership into the WTO will apply further pressure for institutional and structural reforms, including improved market access and protection of intellectual property rights for foreign investors. 7.64 The fact that foreign direct investors heavily favor the service over the manufacturing sector also indicates that foreign investors mainly view investing in Brazil as a means of reaching the large domestic market. Service sectors such as the telecommunication and finance industries witnessed large increases in FDI as changes in regulations and policies allowed and encouraged foreign investment. In addition, foreign investors wishing to conduct business in these industries have little choice but to directly invest in Brazil given that these industries are not easily tradable. Hence, investors in these industries have a relatively high tolerance for the poor investment climate in Brazil because they have little other alternative means of accessing the domestic market. 7.65 On the other hand, FDI inflow into the manufacturing sector remained stagnant, reflecting foreign manufacturers' reluctance to view Brazil as an attractive manufacturing base. Intemational investors searching for a manufacturing export base have a lower tolerance for a country with a poor investment climate, given that they can always choose to invest in another country. They can access the domestic market of a country with a poor investment climate by exporting to the country as opposed to directly operating in the country. This is a choice not available in a non-tradable service sector. Hence, the level of FDI in the manufacturing sector is a better indictor of the host country's investment climate than the level of FDI in the service sector. 7.66 The fact that foreign firms in Brazil have a local market orientation is also suggested by another recent survey. AT Keamey surveyed senior executives from the world's largest multinational corporations in order to determine the reasons why they would invest in a given market. The survey indicated that domestic market access is a more important reason than low production costs for FDI in Brazil, whereas the two factors are similar in their importance for FDI in Mexico (Figure A25). Low production costs are cited by only 26% of survey respondents as 18 METI (2001). 232 important factors for Brazil compared to 68% citing market size. For Mexico, 43% cited low production cost and 43% cited market size. 7.67 Finally, the large share of FDI in the service sector, in as far as it reflects the fact that foreign investors are local market-oriented, could explain why 40% of foreign firms in Brazil do not u'se their frontier technology. The figure was reported in a study by SOBEET, and cited in FIAS (2001). Using data on Asian subsidiaries of Japanese multinational firms in the electronics industry, Belderbos et al. (2000) showed that FDI with a local market orientation has a higher local content than export-oriented FDI. They conjecture that local market oriented FDI uses more mature and standardized low-cost components procured from locally established suppliers. This may explain why host market-oriented foreign firms in Brazil tend not to use sophisticated technologies. The Industry Distribution of FIDI 7.68 If one assumes that FDI benefits domestic industry both directly through investment and indirectly through competition and transfer of technology, then FDI can be a component of a country's industrial policy. Figure 7.8 shows the industrial distribution of Brazil's FDI inflow. The horizontal axis is the industry's value added as a percent of GDP in 1995, the vertical axis is the industry's labor productivity indexed to the US in 1995, and the size of the circle reflects the industry's average share of FDI over the second half of the 1990s. The industries included in the figure are limited to those included in McKinsey (1998). Figure 7.7 Sector breakdown of FDI Labor 80 productivity Indexed 70 to us 60 o Automotive 50 * Meturilagy 0 Food Proc. 40 - 0 Retail 30 S Construction 20 W o~~~~~~~~~~~~ Telecom * Finance 10 0 Value added, 0 1 2 3 4 5 6 7 8 9 1 0 % ofGDP Note: Size of circle is proportionate to average share of FDI Source: McKinsev(l 998). World Bank SIMA 7.69 Industries with low productivity, large GDP share, and currently low levels of FDI are industries with the highest potential for increased FDI to have the greatest impact on overall economic growth. In the figure, the industries represented by small.circles located nearest to the bottom, right-hand corner most closely satisfy these criteria. Although there are no industries that satisfy all three criteria, the three most promising candidates are retail, food processing, and construction industries. The retail and food processing industries have relatively low productivity 233 but also small GDP shares, while the construction industry has a relatively large GDP share but a high productivity. The retail industry at 14 percent and the food processing industry at 18 percent have much lower labor productivity compared to their US counterparts. As indicated by the size of the circles, all three industries received on average a relatively small share of the FDI inflow between 1995 and 2000, capturing 3.37, 2.57, 0.44 percent of total FDI inflow, respectively. 7.70 Increased FDI into these lagging industries can stimulate higher domestic investment and greater competition, as evidenced by the Korea's experience with increased FDI into its retail industry (Box 3). In the case of Brazil, although the share of total FDI is still small, international investment in the retail sector has been present and growing as the sector restructures. Five out of the seven largest hypermarkets are at least partially foreign owned, although the next largest thirteen are domestic firms. Still, there are areas that can be improved as the retail sector remains relatively highly fragmented and the infrastructure for delivery services remains poor.170 Box 7.3 Foreign Direct Investment In the Korean retailing industry171 In January of 1996, the Republic of Korea lifted most of the previous legal restrictions on foreign investment in the retail and distribution industry, leading to the entrance of multinational firms and substantial changes in the market structure and the level of competition. Regulation that restricted the number and size of foreign retail stores was abolished, and limits on foreign ownership of commercial real estate were substantially eased. Starting with the Anglo-Dutch company Makro, which opened an innovative membership-only warehouse discount store through a joint venture, major foreign retailers such as Costco, Wal-Mart, Carrefour, Tesco, and Promodes have become active in the Korean domestic market since 1996, mainly focusing on the large volume and low price discount sales concept. The investment entry mode has changed over time as well: recent foreign entries have opted for greenfield investments, shifting away from previously favored schemes such as joint ventures and technical collaborations. Competition has increased as a result. Domestic competitors simultaneously reduced the prices of 200 to 400 items by an average of 2 to 3 percent in response the opening of the Makro store. The number of competitors has so rapidly increased that major foreign retail firms now operate 22 megastores in the Korean market. By contrast, the domestic discount chain leader, E-Mart, operates 18 outlets throughout the country. Partly in response to heightened foreign competition, domestic conglomerates such as Samsung aggressively expanded their shopping malls and hypermarkets in an effort to capture economies of scale. Direct and indirect technology transfers have benefited the domestic firms as well. Joint ventures and technical collaborations with existing local firms have been the preferred mode of entry for many of the earlier foreign entries, and local competitors such as E-mart have openly admitted to adapting Price Club's advanced management techniques. Foreign entries have forced domestic firms to improve their scale and operational efficiency, ultimately benefiting the consumers. Some domestic firms have managed to withstand the foreign competition and continue to maintain a competitive edge over foreign rivals. Foreign discounters still lag E-Mart in sales performance and efficiency, although major domestic distribution firms are under heavy competitive pressure from foreign competitors who have a edge in low financing cost. Small and medium sized domestic retailers with weak financing capabilities are especially vulnerable. 170 Fabiana Borges da Fonseca, 'Brazil Retail Food Sector Report 2000," USDA GAIN Report #BR0026, 2000; American Chamber of Commerce 2001. 171 World Investment Repor 1997; Business Korea, December 1999; and Business Korea February 1997. 234 7.71 In summary, the following statements can be made based on our analysis of the data: * FDI inflow for Brazil increased substantially over recent years, but whether such levels are sustainable is questionable. o The current slowdown in the world economy could lead to a reduction in the total supply of foreign investment, reducing. FDI in Brazil, unless Brazil is able to increase its relative attractiveness as a host country for FDI. The data do not indicate that the recent increase in FDl is mainly based on government privatization programs, although there has been a reliance on a small number of service industries. * The fact that the recent upsurge in FDI inflow has mainly benefited the non-tradable service sector while mostly bypassing the manufacturing sector suggests that the investment climate in Brazil still does not compare favorably with other countries. o It indicates that Brazil is a country where foreign investors mainly invest in order to gain access to the large domestic market, as opposed to a country where an internationally competitive investment climate attracts investors seeking to establish a manufacturing export base. The current sector mix of Brazil's FDI raises concem given that typical country examples of export-based growth have often been based on exports of manufactured goods, and foreign firms can often play a major role in exports. In addition, foreign firms which are host market- oriented typically use more outdated technology. * The retail and food processing industries are characterized by low productivity and a small share of FDI inflow, suggesting that increased FDI can result in a relatively large impact on growth. o Increased FDI promotes growth in the industry not only through increased direct investment, but also through increased competition and transfer of technology. 7.72 The conclusion that the investment climate discourages FDI inflow into the manufacturing sector in Brazil is supported by the data. The next section presents the available research regarding the relationship between public governance, a major determinant of the investment climate, and the inflow of FDI. Given that cheap labor, tax benefits and a large domestic market are often cited as important determinants of FDI, the challenge is to identify the governance effect apart from the effects of these other variables. Research on Governance and FDI 7.73 The general business environment is one of many factors that influences inflow of foreign direct investment. Past research, including a survey conducted by UNCTAD of senior executives of multinational corporations with affiliates in Brazil, suggests that the following three issues are the most important determinants of FDI:172 size of the internal market; growth of the market; and the 72 UNCTC 1992, UNCTAD 2000. 235 stability of the business environment. Business environment mainly refers to the macroeconomic policy regime, and the social and political environment. These three factors are interrelated. For instance, a large domestic market alone might not be a sufficient enough incentive for certain foreign investors. Lack of freedom to operate intemally with a minimum of bureaucratic interference could deter foreign investment regardless of the size of the market. The important point to take away from the survey is that govemance is as important a factor as host country market size and growth for foreign investors. 7.74 The survey results from UNCTAD can be compared to the survey conducted by AT Kearney. The survey shows that Brazil is among the top five destination of choice for global corporations, and the outlook for Brazil has improved since the currency devaluation. The survey provides insight into the factors that lead multinationals to invest in Latin America in general and Brazil specifically. Figure A28 shows that macroeconomic stabilization is by far the most important factor behind investment decisions in Latin America, but it also shows that improved rule of law is a significant factor in encouraging foreign investment, where its importance is comparable to the free trade agreement and the strengthening of Mercosur. 7.75 Both the UNCTAD and AT Kearney surveys suggest that govemance has an important influence on foreign investment. This conclusion is supported by the data presented in World Bank's Global Development Finance (GDF, 2001). Using sources such as the European Round Table of Industrialists 2000 and the UNCTAD (1999), GDF 2001 presents evidences that the investment climate and govemance are positively related to FDI. Figure A29 presents the evidence. The first panel shows that improvements in the investment climate are associated with higher FDI growth. The investment climate index ranges from 1 to 6, and the GDF estimates are based on European Round Table of Industrialists 2000. The second panel shows a plot of the FDI growth rate against a govemance indicator. The governance indicator is an aggregated indicator from a World Bank study conducted by Kaufmann, Kraay, and Zoido-Lobotan (1999). It aggregates govemance measurements of voice and accountability, political stability, government effectiveness, regulatory framework, rule of law, and corruption. Again, the data indicates that there is a positive relationship between public governance and the FDI growth rate. 7.76 Given that the surveys of business executives indicate that govemance is a major concern in their foreign investment decisions, it is not surprising that some multinationals explicitly account for govemance quality in their foreign investment decision making process. For example, a positive assessment of the country's political and economic system was a major element of Intel's decision to construct a $300 million semiconductor plant in Costa Rica. Intel gave high marks to Costa Rica for its stable and supportive govemment and its transparent legal system (Box 4). 236 Box 7.4: Intel's investment in Costa Rica'73 The US semiconductor manufacturer Intel, with annual revenues of more than $25 billion, announced in November of 1996 plans to construct a $300 million assembly and testing plant in Costa Rica, a plant with up to 2000 (mostly Costa Rican) employees, over alternative sites in Brazil, Chile, Indonesia and Mexico. The decision by the large multinational corporation to invest in a relatively small country with limited domestic electronics and other high technology sectors attracted much attention in the foreign investment community. The Intel decision was notable because Costa Rica did not enter into a bidding war with other potential locations and offered little special subsidies other than the standard incentives available to other foreign firms that operate in the country's EPZs. This contrasts Brazil's subsidies to foreign automobile manufacturers which some have criticized as lowering national welfare.174 The governance quality of the country was a major factor in Intel's decision to make a large investment in Costa Rica. Costa Rica was a stable and democratic country committed to economic openness and which actively encouraged foreign investment, and the legal system was relatively transparent. The willingness of the president of Costa Rica to meet with Intel officials in order to pledge full support readily made it apparent the government's commitment. Intel officials met with representatives from international accounting firms KPMG, Peat Marwick, Price Waterhouse, and Emst and Young as well as Citibank to assess the transparency and reliability of the country's legal and financial institutions. Meetings were held with DSC Communications, the largest US electronics firm operating in Costa Rica, and several other foreign manufacturers in order to discuss the country's business climate as well as regulations regarding foreign transactions of US dollars. 7.77 Research also suggests that FDI is more sensitive to the governance quality when compared to other forms of capital flow. Hoekman and Saggi (1999) cite corruption, unnecessary regulatory requirements, complicated or nontransparent administrative procedures, and insufficient protection of physical and intellectual property rights as substantial obstacles to investment. Their study indicates that corruption has a greater impact on FDI than on debt or portfolio equity investment. One explanation for the greater sensitivity is that FDI requires a more direct interaction with local officials, suppliers, competitors, and clients. This section now turns to the existing research on the estimated size of this impact. 7.78 The Opacity Index, a govemance rating system constructed by Price-Waterhouse-Coopers (PWC), was described in a previous section, "Measurement of the Determinants of Governance Quality." PWC uses the Opacity Index in their recent analysis of the effect of governance on foreign direct investment (PWC, 2001). The study implements a regression analysis based on Wei (2000a). The dependent variable is the log of FDI, and the independent variables include the Opacity Index in addition to other variables controlling for tax rate, size of economy, and labor cost. Based on these estimates, the results indicate that Brazil can increase the annual inflow of FDI by 141%, or $40.3 billion, if it improved its govemance to US standards. The magnitude of the impact is clearly very large. 7.79 An interesting feature of this study is the inclusion of two independent variables indicating the presence of FDI incentive and restriction policies, respectively. The FDI incentive variable indicates the existence of policies such as tax concessions and export processing zones designed to promote foreign investment, while the FDI restriction variable indicates policies such as the 173 Spar (1998), also MIGA Investment Promotion Toolkit. 174 Hanson (2001) 237 presence of restrictions on foreign ownership and policies that excluded foreign firms in certain strategic sectors. The variables were constructed based on information from country reports published by the company, and each variable takes on a value from zero to four. The FDI incentive indicator variable was a significant determinant of FDI, while the restrictions indicator variables was insignificant. More importantly for the purpose of this report, the opacity variable was still statistically significant even with the inclusion of the incentive and restriction variables in the regression. The overall governance quality is a significant determinant of FDI even when specific FDI policies are taken into account. 7.80 Empirical studies assessing the quantitative impact of public governance quality on FDI flows (Wei 2000, Harms 2000, Lehmann 1999, Drabek and Payne 2000, Gastanaga et al. 1998, and Lee and Mansfield 1996) generally estimate a significant effect. Gastanaga et. al (1998) found that FDI was lower in countries with greater bureaucratic delay and higher nationalization risk, while Lee and Mansfield (1996) found that perceptions conceming the strength of intellectual property rights affected both the volume and composition of US foreign direct investment. 7.81 Wei (2000b) conducted a study comparing the effect of corruption on FDI, comparing its effect with cheap labor and tax benefits. The study uses GCR variables measuring corruption and regulatory burden as independent variables in the regressions on FDI. The results indicate that China could increase its FDI by 218% if it reduced corruption and red tape to Singapore's levels. 7.82 Drabek and Payne (2000) examined the impact of transparency on FDI inflows. The transparency variable is a country ranking based on a composite index constructed from the International Country Risk Guide (ICRG), where the underlying variables are the level of corruption, law and order, bureaucratic quality, contract viability, and the risk of government expropriation of private assets. The ranking ranges from 8.5 to 38. Using data from 52 developed and developing countries over the period 1991 to 1995, the study found that a one-point increase in the level of transparency was associated with an average of 40 percent increase in FDI inflows, although the expected increase varies widely among the countries. Unfortunately, the sample does not include Brazil, but the results indicate that Argentina could increase its FDI by 51%, Indonesia by 12%, and Chile by 7% if their rankings were increased by one point. 7.83 Of course, these numbers are merely illustrative, and comparison of the impact across the studies is difficult due to differences in definition and measurement methodology. Nevertheless, together the studies indicate that improving governance quality has a large impact on FDI inflows. 7.84 Governance also has an effect on the degree to which FDI is transformed into productive investment as opposed to consumption or an outflow of capital. The relationship between FDI and investment depehds on the country's absorptive capacity. The absorptive capacity depends on factors among which public governance is a major component. The World Bank's Global Development Finance (GDF) 2001 cites political stability, health of the financial system, efficiency of government services, and the degree of corruption as well as the macroeconomic policy framework as factors which determines absorptive capacity. 7.85 Empirical researchers have only begun to explore this relationship. Borensztein, De Gregorio, and Lee (1998) finds that FDI's positive effect on investment increases with the stock of 238 human capital. GDF 2001 illustrates this relationship with a plot of foreign direct investment and school enrollment against investment-GDP ratio. The figure is reproduced as Figure 7.8. Figure 7.8 FDI, School Enrollment, and Investment (Perc < 20 ':_ _ 2 _ / H~~~~~~~igh 15 -M edium Foreign Direct lnvee1ment School Enrollment 7.86 Domestic investment is higher when FDI is greater and also when education levels are higher. When education levels are low, foreign direct investment has the least impact on domestic investment. The study indicates that the effectiveness of FDI at low levels of schooling maybe be close to zero. 7.87 Research suggests that the level of education could also have an impact on the extent of technological spillovers from foreign direct investment. Urata and Kawai (2000), using data on Japanese FDI to the rest of the world, found that the educational level of the host country was a significant explanatory variable for intra-firm technology transfer between parent firms and overseas affiliates. The level of education seems to be a key determinant of the host country's capacity to absorb technologies from the foreign firms. 7.88 In a similar manner, quality of governance is likely to affect the extent to which the host country benefits from technological spillovers from foreign firms. Studies indicate that the quality of govemance affects general technological deepening in developing countries. Clarke (2000) showed that the quality of govemance, as measured by the risk of expropriation and the rule of law, is correlated with R&D expenditures. Hence, improving the quality of governance would increase the capacity of the host country to absorb foreign technology. Increased R&D expenditure and increased educational attainment both reflect an upgrading of skills and knowledge of the local population. 7.89 The benefits that can be accrued from FDI also depend on its mode of entry. Previous sections have described the growing importance of cross-border M&A as an investment vehicle for foreign direct investors. Furthermore, it was suggested that greenfield investments could have a 239 larger impact on domestic investment compared to M&As in the short run. In addition, FDI dominated by M&A could adversely affect the level of competition by increasing the concentration of market power, leading to abuse of market power and lower consumer welfare. For example, studies based on developing and developed countries indicate that multinational corporation activity and the concentration of producers in host country industries are positively correlated."75. On the other hand, the relationship between FDI and market concentration depends on several post-entry effects. And there is no consensus on how increased market power concentration would affect the level of investment. Nevertheless, the strengthening of competition policy is considered a key component of any FDI liberalization policy. 7.90 In summary, we have made the following arguments based on the available research: . Both survey and empirical research indicate that overall governance quality is a significant determinant of foreign direct investment in Brazil. o This relationship holds even when specific FDI policies and regulations are taken into account. Furthermore, compared to other forms of capital flow such as debt and portfolio investment, FDI seems to be more sensitive to the quality of governance. * Govemance affects not only the amount of FDI but also the impact of FDI on domestic investment through its affect on domestic absorptive capacity. o Factors which determine the host country's business climate, such as its political stability, the health of the financial system, the efficiency of govemment services, the degree of corruption, and the macroeconomic policy framework, seem to effect its absorptive capacity. The education level is one variable for which the data shows a clear relationship to absorptive capacity. FDI has the least impact on domestic investment when the level of education is lowest. In addition, the quality of governance affects the extent of technological spillovers from foreign firms. There is no consensus on how FDI affects market concentration in the host country, nor on how this affects aggregate investment, in comparison to cross-border M&A. 7.91 Governance, FDI, and domestic investment interact in a complex manner over the long run. Nonetheless, govemance is a significant determinant of the other two variables. Policies to improve the quality of public governance should be an integral part of any country's strategy to promote FDI, domestic investment, and growth. This is especially for countries such as Brazil where key components of its govemance trail other similar countries. Also, growing intemational competition for FDI means that foreign investors can choose from a growing number of potential investment locations. Countries can no longer afford to neglect improving their investment climate. 175 World Investment Report, 1997, page 137. In particular, refer to Wilmore (1989) for a study of Brazil. 240 7.92 In addition to maintaining high levels of FDI, Brazil has an opportunity to maximize the benefits of FDI by improving its investment climate through its public governance policies. The evidence suggests that the benefits from FDI are small under a system characterized by regulatory and legal impediments to private economic activities, a poorly educated work force, and macroeconomic instability. In addition, governance should be improved in order to maximize the benefits of technological spillovers from foreign firms. The benefits of improving governance are of course not limited to its impact on FDI, as it will also improve the investment propensity and productivity of domestic private firms as well. 7.93 The recent FIAS report on Brazil offers policy recommendations for improving the investment climate for foreign investors."76 In general, it recommends eliminating counter- productive red tape and streamlining administrative procedures in order to reduce investment cost, risk, and uncertainty. Some of the specific recommendations are to eliminate the excessive and redundant bureaucracy involved in site development, visas, and various permits and licenses, and to review and reform the corporate tax system in order to reduce its excessive complexity. 7.94 With regards to the form of FDI entry, given the lack of strong empirical evidence comparing greenfield and M&A FDI, the appropriate policy is to strengthen competition policy while avoiding policies which favors one form of entry over the other. On the other hand, countries such as Korea have implemented policies that favor greenfield over M&A forms of FDI.'7 7.95 An important difference between Brazil and the manufacturing export-based economies of East Asia is the sector composition of FDI. Improved governance-more efficient courts, reduced administrative barriers-could encourage FDI in the manufacturing sector to form the basis for export growth in Brazil. As part of an overall industrial policy, FDI can also be an effective tool in developing inefficient industries such as non-tradable service industries, which are naturally protected from foreign competition. For Brazil, encouraging FDI in the retail and food processing industries could have a high impact on economic growth given the low productivity and low levels of FDI in these industries. 7.96 Improving the business climate will be crucial for developing countries to compete for FDI flows. The underlying macroeconomic stability and openness to trade will need to be maintained, but in addition Brazil will need to improve its regulatory framework, reduce corruption and bureaucratic constraints, and improve the legal and administrative system that supports the enforcement of contracts. 176 FIAS (2001), page xxiii. 177 Refer to World Investment Report (2000), page 150. As a result of the Asian financial crisis in 1997, Korea changed its FDI and M&A policies in order to attract foreign investment in Ks troubled firms. Restrictions on foreign acquisition of domestic shares, M&As, and land were abolished. However, the new investment policy still slightly favors greenfield over M&A investment. Most of the newly introduced measures, such as the creation of foreign investment zones and tax incentives, are geared towards greenfield investment. The tax regime also favors greenfield over M&As by allowing reduction of taxes on corporate income, acquisitions, registration, and property and land. 241 7.97 The main conclusion is that the quality of public governance is a significant determinant of both private sector investment and FDI. Policies to improve the public govemance, traditionally a somewhat neglected area in Brazil, must be given more attention. 7.98 Given its importance, measurements of governance that provide policymakers with greater objectivity and specificity assume an added importance. Measurements of the determinants of govemance quality, based on the underlying govemment processes, are more effective in this manner as opposed to general govemance quality measurements. For example, data on the number of court staff, salaries, and technology allowed the World Bank Legal Department to determine the key determinants of legal system inefficiency. 7.99 In addition, general govemance quality measures are typically based on surveys, and are therefore difficult to assess because the results depend on the cultural background of respondents. Cross-state surveys may improve matters, as cultural differences across states within a country are generally less severe than between countries. Examples of such surveys are the World Bank FACS survey for India, and the IDESP survey of judiciary quality described in Pinheiro (1999). 7.100 Despite the potential shortcomings of survey methodology, the major international governance rating institutions consistently give Brazil a low rating for the quality of its regulatory system. The ratings on the legal system and government bureaucracy are mixed when compared to similar Latin American countries, but Brazil compares poorly against developed countries. Although there are obvious overlaps among the three categories, the survey results suggest that regulatory burden seems to be a particularly acute problem for Brazil. 7.101 The comparative measurements of the underlying legal and regulatory arrangements indicate that both court delays and the time required to complete the start-up regulatory procedures in Brazil are substantially higher than in other countries. Such delays not only represent a real cost to entrepreneurs, but since they occur before the outcome of processes is known, they also represent added uncertainty and thus financial risk, especially for those competing in the global markets. The more detailed analysis of Brazil's new business registration process showed that the delays were caused by the multiplicity of procedures required by the three levels of government. 7.102 Perhaps more so than domestic entrepreneurs, foreign direct investors, particularly when they are interested in establishing a manufacturing export base, are sensitive to the quality of the governance of the host country. The suspicion that foreign investors still do not consider Brazil's public govemance to be a high enough quality is further corroborated by the fact that FDI inflow has been heavily skewed towards the non-tradable service sector, and by survey results. 7.103 With regards to the legal system, the capital budget, technology, and managerial activism are promising areas of policy intervention for reducing court delays, as opposed to the more traditional areas such as the number of court staff, salaries, and the level of training. With regards to the regulatory system, simplifying and centralizing the business start-up process and tax registration across levels of government should result in a reduction of regulatory burden. 7.104 Improving public governance should also be a policy priority for promoting exports given its potential to substantially increase the flow of FDI into the manufacturing sector. Meanwhile, promoting FDI into non-tradable service industries such as retail and food processing should have 242 a high impact given the low productivity and current low share of FDI of these industries. The current trend towards a stable and liberalized set of regulations regarding the repatriation of foreign capital and profit could be maintained. Particularly important in the context of these industries, the excessive bureaucracy involved in site development, visas, and various permits and licenses could be reduced, and the corporate tax system could be simplified. 7.105 There is need for further research, and one compelling choice would be to analyze the govemance constraints faced by the informal sector. The sector is a substantial part of the Brazilian economy and suffers from low investment. One World Bank study has found that ineffective and discretionary administration of tax and regulations, weak rule of law, and corruption increases the size of the unofficial economy. In addition, the study shows that countries with larger unofficial economies tend to grow more slowly.171 Hence, from a policy perspective, encouraging movement from the informal to the formal sector by lowering the cost imposed by an inefficient legal and regulatory system could have a significant impact on investment and economic growth. References General Alesina, Alberto, 'The Political Economy of High and Low Growth," World Bank Annual Bank Conference on Development Economics, 1997. -, Ricardo Hausmann, Rudolf Hommes, and Ernesto Stein, "Budget Institutions and Fiscal Performance in Latin America," Inter-American Development Bank, Working Paper Series 394,1999. Aron, Janine, 'Growth and Institutions: A Review of the Evidence," World Bank Research Observer, 15(1) February 2000. Beck, Thorsten, and George Clarke, Alberto Groff, Philip Keefer, Patrick Walsh, "New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions," World Bank Working Paper No. 2283. Brunetti, Aymo, and Gregory Kisunko, Beatrice Weder, "Institutional Obstacles to Doing Business," World Bank Policy Research Working Paper No. 1759, April 1997. Buscaglia and Dakolias, Judicial Reform in Latin American Courts: The Experience in Argentina and Ecuador, World Bank, 1996 -----, "How Businesses See Government," IFC Discussion paper No. 33,1998. Chong, Alberto and Cesar Calderon, "Causality and Feedback Between Institutional Measures and Economic Growth," 12(1) March 2000. Dakolias, Maria, "Court Performance around the World," World Bank Technical Paper No.430, 1999a. 178 Johnson, Simon et. al (1999) 243 -, "Comparative International Study of Court Performance Indicators," World Bank Legal Department, 1999b. Fabricius, Michael, "The Impact of Economic Security on Bank Deposits and Investment," IMF Working Paper No. WP/98/98. Foreign Investment Advisory Service (FIAS), 'Legal, Policy, and Administrative Barriers to Investment in Brazil," World Bank, 2001. Foxley T., Alejandro, and Claudio Sapelli, "Chile's Political Economy in the 1990s: Some Governance Issues," in Guillermo Perry and Danny M. Leipziger, Chile. Recent Policy Lessons and Emerging Challenges, World Bank, 1999. Gastanaga, Victor M., Jeffrey B. Nugent, and Bistra Pashamova, "Host Country Reforms and FDI Inflows: How Much Difference do they Make?" World Development 26(7) 1998. Haque, Nadeem U., Manmohan S. Kumar, Nelson Mark, and Donald J. Mathieson, 'The Economic Content of Indicators of Developing Country Creditworthiness," IMF Working Paper No. WPI96/9. Hellman, Joel S., and Geraint Jones, Daniel Kaufmann, Mark Schankerman, "Measuring Govemance, Corruption, and State Capture," World Bank Working Paper No. 2312, April 2000. Johnson, Simon, and Daniel Kaufmann, Pablo Zoido-Lobaton, "Regulatory Discretion and the Unofficial Economy," AEA Papers and Proceedings, May 1998. -, "Corruption, Public Finances, and the Unofficial Economy," World Bank Policy Research Working Paper No. 2169, August 1999. La Porta, Rafael and Florencio Lopez-de-Silanes, Andrei Shleifer, "Trust in Large Organizations," AEA Papers and Proceedings, May 1997. -, "Law and Finance," Joumal of Political Economy, 106(6) 1998. -----, 'The Quality of Government," Joumal of Law Economics and Organization, 15 (April 1998). Kaufmann, Daniel, Aart Kraay, and Pablo Zoido-Lobaton, "Aggregating Governance Indicators," World Bank Policy Research Working Paper No. 2195, October 1999. -----, "Govemance Matters," World Bank Policy Research Working Paper No. 2196, October 1999. Knack, Stephen, "Social Capital and the Quality of Govemment: Evidence from the United States, World Bank Working Paper No. 2504, Knack, Stephen, and Philip Keefer, "Institutions and Economic Performance: Cross-Country Tests Using Alternative Institutional Measures," Economics and Politics 7(3) November 1995. -----, "Does Social Capital Have an Economic Payoff? A Cross-Country Investigation," Quarterly Joumal of Economics, November 1997. Manning, Nick, and Ranjana Mukherjee, Omer Gokcekus, "Public Officials and their Institutional Environment: An Analytical Model for Assessing the Impact of Institutional Change on Public Sector Performance," World Bank Working Paper No. 2427, August 2000. 244 Marcel, Mario, "Effectiveness of the State and Development Lessons from the Chilean Experience," in Guillermo Perry and Danny M. Leipziger, Chile. Recent Policy Lessons and Emerging Challenges, World Bank, 1999. Messick, Richard E., "Judicial Reform and Economic Development: A Survey of the Issues," World Bank Research Observer, 14(1) February 1999. Perry, Guillermo and Danny M. Leipziger, ed., "Chile. Recent Policy Lessons and Emerging Challenges," World Bank, 1999. Pfeffermann, Guy P., Gregory V. Kisunko, and Mariusz A. Sumlinski, 'Trends in Private Investment in Developing Countries," IFC Discussion Paper No. 37,1999. Poirson, Helene, "Economic Security, Private Investment, and Growth in Developing Countries," IMF Working Paper No. WP/98/4. PriceWaterhouseCoopers, 'Opacity Index," January 2001. Rauch, James E., and Peter B. Evans, "Bureaucratic Structure and Bureaucratic Performance in Less Developed Countries," Joumal of Public Economics 75 (2000). Ritzen, Jo, William Easterly, and Michael Woolcock, 'On "Good" Politicians and "Bad" Policies: Social Cohesion, Institutions, and Growth," World Bank Working Paper No. 2448. Rodrik, Dani, "Institutions for High-Quality Growth: What They are and How to Acquire Them," NBER Working Paper No. W7540, February 2000. Shiavo-Campo, Salvatore and Giulio de Tommaso, Amitabha Mukheree, "An International Statistical Survey of Government Employment and Wages," World Bank Policy Research Working Paper No. 1806. -----, "Govemment Employment and Pay in Global Perspective," World Bank Policy Research Working Paper No. 1771. Silva, Eduardo, "Business Elites and the State in Chile," in Sylvia Maxfield and Ben Ross Schneider, ed., Business and the State in Developing Countries, Comell University Press, 1997. Stone, Andrew, Brian Levy and Ricardo Paredes, 'The Legal and Regulatory Environment for Business Transactions in Brazil and Chile," World Bank Policy Research Working Paper No. WPS 891, April 1992. World Bank Public Sector Group, "Reforming Public Institutions and Strengthening Govemance," World Bank, Public Sector Board, Poverty Reduction and Economic Management, 2000. World Bank DEC and PREM Network, "Toward more operationally relevant indicators of govemance," World Bank DEC and PREM Network, PREM Notes, December 2000. World Economic Forum, Global Competitiveness Report 2000, Geneva: World Economic Forum, 2000. 245 Governance and entrepreneurship Alston, Lee J., Gary D. Libecap, and Robert Schneider "The Determinants and Impact of Property Rights: Land Titles on the Brazilian Frontier" Joumal of Law, Economics, and Organization 12(1):25-61 (1996) Pinheiro, Armando Castelar and Celia Cabral, "Credit Markets in Brazil: The Role of Judicial Enforcement and Other Institutions," Escola de Pos-Graduac,o em Economia da Fundac,o Getulio Vargas: Centro de Estudos de Reforma do Estado. Texto para Discussao (Brazil); No. 019:1-57, January 1999. Djankov, Simeon and Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Shleifer, 'The Regulation of Entry," NBER Working Paper No. W7892, September 2000. Governance and FDI A. T. Kearney, "FDI Confidence Index," 3(1) January 2000. Belderbos, Rene, Giovanni Capannelli, and Kyoji Fukao, "Local Content of Japanese Electronics Manufacturing," The Role of Foreign Direct Investment in East Asian Economic Development, The University of Chicago Press, 2000. Borenztein, Eduardo, Jose De Gregorio and Jong-Wha Lee, "How Does Foreign Investment Affect Economic Growth?" Joumal of Intemational Economics, 1996. Bosworth, Barry, and Susan M. Collins, "Capital Flows to Developing Economies: Implications for Saving and Investment," Brookings Papers on Economic Activity 1:1999, 1999. Clarke, George R. G., "How the Quality of Institutions Affects Technological Deepening in Developing Countries," World Bank Development Research Group Working Paper, 2000. Drabek, Zdenek and Warren Payne, 'The Impact of Transparency on Foreign Direct Investment," working paper, WTO and Economic Consulting Services, 2000. European Round Table of Industrialists, Improved Investment Conditions: Third Survey on Improvements in Conditions for Investment in the Developing World, Brussels, 2000 Foreign Investment Advisory Service, "Legal, Policy, and Administrative Barriers to Investment in Brazil," World Bank Group, May 2001. Gastanaga, Victor., Jeffrey B. Nugent and Bistra Pashamova, "Host country reforms and FDI inflows: How much difference do they make?" World Development, 26(7), 1998. Hanson, Gordon H., "Should countries promote foreign direct investment?" UNCTAD, G-24 Discussion Paper Series No. 9, February 2001. Harms, Phillip, "International Investment, Political Risk and Growth," Boston : Kluwer Academic, 2000. Lee, Jeong-Yeon and Edwin Mansfield, "Intellectual property protection and US foreign direct investment," Review of Economics and Statistics, 78, 1996. Lehmann, Alexander, "Country Risks and the Investment Activity of US Multinationals in Developing Countries," IMF Working Paper WP/99/133, October 1999. 246 Ministry of Economy, Trade and Industry, "White Paper on Intemational Trade 2001," 18 May 2001. McKinsey Global Institute, "Productivity - The Key to an Accelerated Development Path for Brazil," McKinsey & Company, 1998. PricewaterhouseCoopers, "Investigating the Costs of Opacity: Deterred Foreign Direct Investment," April 2001. Smarzynska, Beata K. and Shang-Jin Wei, "Corruption and the Composition of Foreign Direct Investment: Firm-Level Evidence," World Bank Working Paper, June 2000. United Nations Conference on Trade and Development (UNCTAD), FDI Determinants and TNC Strategies: The Case of Brazil, United Nations, 2000. United Nations Center on Transnational Corporations (UNCTC), The Determinants of Foreign Direct Investment. A Survey of the Evidence, United Nations, 1992. Urata, Shujiro and Hiroki Kawai, "Intrafirm Technology Transfer by Japanese Manufacturing Firms in Asia," The Role of Foreign Direct Investment in East Asian Economic Development, The University of Chicago Press, 2000. Wei, Shang-jin, 'Why Does China Attract So Little Foreign Direct Investment?" Harvard University and NBER Working Paper, 1998. -----, "How Taxing is Corruption on International Investors?" Review of Economics and Statistics, 2(1) 2000,1-11. [2000a] -, "Sizing up Foreign Direct Investment in China and India," Stanford University Center for Research on Economic Development and Policy Reform Working Paper No. 85, December 2000. [2000b] ----- and Beata Smarzynska, 'Corruption and the Composition of Foreign Direct Investment: Firm- Level Evidence," World Bank Policy Research Working Paper No. 2360, June 2000. World Bank, Global Development Finance 2001. 247 Annex Table Al Measurements of Overall Public Governance Quality ! . L : . .5 'S * S.........., l| | |! 1 , L &I~~~~~~~~~~~~~~~~I Business Risk Business Environment Bureaucratic delays, Survey of 50 mostly Yes Service Risk Intelligence contract enforceability experts developed (BERI) nationalization rsk, countries policy stability Intemational Political Risk Services Govemrnment corruption, law Survey of 140 developed Yes Country Risk Guide (PRS) Group and order tradition, experts and developing (ICRG) bureaucratic quality countries Freedom in the Freedom House Political freedom, civil Survey of 172 developed Yes World liberty experts and developing countres Global World Economic Independence of civil Survey of 57 developed Yes Competitiveness Forum service from politics, businessmen, and developing Survey competence of public sector govemment countries personnel, tax evasion officials Country Risk Standard and Poor's Survey of 106 developed Yes Review DRI / McGraw-Hill experts and developing countres Central European Wall Street Joumal Survey of 27 transition No Economic Review businessmen economies Transition Report European Bank for Survey of 26 transition No Reconstruction and experts economies Development Country Risk Economist Intelligence Survey of 114 developed Yes Service, Country Unit experts and developing Forecast countries Economic Freedom Heritage Foundation I Survey of 154 developed Yes Index Wall Street Joumal expeds and developing ___________________ __________________ _ _ _countries Asia Intelligence Political Economic Survey of 11 Asian No Risk Consultancy experts countries World Institute of Survey of 46 prmarily Competitiveness Management businessmen developed Yearbook Development countries Private Sector World Bank (World Policy and judicial Survey of Survey Development Report unpredictability, quality of businessmen 1997) govemment services, corruption and red tape Country Policy and World Bank Property rights and rule Survey of Institutional based govemance, quality experts Assessments of budgetary management, transparency, corruption Firm Analysis and World Bank Number of visits by Survey of 3 countries No Competitiveness govemment official, businessmen (India, Bolivia, Survey (FACS) business environment Morroco) assessment World Bank World Bank Time spent on regulation Survey of Economic Survey compliance, frequency of businessmen (WBES) ' payments to public officials Regulatory Cost World Bank Assessment of regulatory Survey of Ukraine No Survey burden, time required to businessmen register firm Quality of Brazil's IDESP (cited by Slowness, faimess, and Survey of Brazil only Yes judiciary system Armenio Castelar costs of courts businessmen Pinheiro (1999)) _ 248 Legal and World Bank (Stone et Time spent on regulation Survey of Brazil and Chile Yes regulatory al. (1992)) compliance, difficulty of businessmen environment regulation compliance Opacity Index Pfcewaterhouse- Corruption, legal system, Survey of 35 countfes Yes Coopers regulatory regime businessmen Corruption Worfd Bank (Pradhan Corruption that distorts Survey of et al. (2000)) formulation of laws and businessmen regulations World Value World Bank (cited by Degree of trust of others Survey of 30 countries Surveys La Porta et al. (1997), individuals Knack and Keefer (1997)) * Contains both quality and determinants of quality measurements. Table A2 Measurements of Determinants of Public Governance Quality Data Source Variables Methodology Coverage Brazil (Partial List) Data Civil service Schiavo-Campo et al. Number of govemment Collected by 80 to 100 Yes employment and (1997, 1997) employees, average authors countfies pay govemment wages Webefian Evans and Rauch Index of meritocratic hiring, Survey of 35 developing Yes comparative state (1999, 2000) intemal promotion index experts countries data prolect Polity98 Project: Constraints on chief Regime executives power, political characteristics participation Database of World Bank (Beck et Electoral rules, type of 177 countries Yes Political Institutions al. (2000)) political system, party composition of the opposition and govemment coalitions Political Constraint Henisz (2000) Number veto power in Collected by 140 countries Yes Index govemment, party author alignment across govemment branches Judicial Indicators World Bank Cases per judge, budget, Collected by 11 countries Yes (Buscaglia et al., clearance rate depadment (1999)) Regulation of entry Djankov et al. (2000) Number of procedures, Collected by 75 countries Yes official time, and official authors costs for starting up a business Public expenditure World Bank-IMF joint Budget formulation, Collected by staff management project execution, reporting economists assessment BNPP (Bank- World Bank (Manning Rule credibility (ex., Survey of public 8 developing No Netherlands et al. (2000)) assessment of budget sector countries and 6 Partnership management guidelines), employees in progress Program) Surveys policy credibility, resource adequacy Land titles Alston et al. (1996) Land titles, access to credit Survey of Brazil Yes I____________________I_______ _ landholders I_ I 249 Figure Al. Global Competitiveness Report - Legal System 8 7 .3 Judicial indep. 6- 5 m Use of courts 4 a~~~~~~~~~~~~~~~~ Litigation cost 3- 2- Corruption in legal F system 0 C~~~~~ co l 0 l l aRstadW m ° E y ffi c ~~~m a. 2 C., ~ ~ C : Figure A2. Global Competitiveness Report - Regulatory System 7- 6 a oTirre with governnent bureaucracy 5 m Burden of regulation 4 - 3 m rotection of property rights 2- 03 Regulatory standards o CD a) -a Figure A3. Global Competitiveness Report - Public Institutions 7 6 E3 ~~~~~~~~~~~Independence of policies 5 * Institutional stability 4- 3 [3~~~~~~~~~~~~~~ Pblice protection 1 ~~~~~~~0 trregular payrrents 0 C ) 250 Figure A4. ICRG 7 6 .o Ritule of Law 5:- 4 F Bureaucratc quality 3 :' 2 M Corruption 80 70 '. d ,,. co 0. | . | .' ,''''1:0' Figure A5. BERI 4 -_ _ _ _ _ _ _ _ _ _ _ 3.5 -~ Contract enf orceabir 3 - 2.5- 2 a Pblicy continuity 1.5 1 o~~~~~~~~~~~0Bureaucracy 0.5 0 co~ CU 0 co 0 (I)~~~~I Figure A6. PricewaterhouseCoopers - Opacity Index 80 70- 60 40- 40~~~~~~~~~~~~~~~4 30 - 201 0 co~~ IM~~~~~ 251 Figure A7. PricewaterhouseCoopers - Subcomponents of Opacity Index 4- 3.5- 3 - 2.5 - oBrazil 2 -. U~~~~~~~~~~~~I Argentina 1.5 - 0 ~~~~~~~~~~~~Chile 0.5 -c 0 C L E A R Figure A8. World Bank Legal Department - Clearance Rate 110 105- 100 - 95- 907 f 800 0 7000 E , 6000 'a. E cE m 0 0 5000~~~~~~~~~~~0 4000 3000- 2000 - I 1000 -Li 0 - ~ ~ ~ ~ 0 ~ 0 252 Figure Al 0 World Bank Legal Department - Backlog Index 2 1.6 1.4 1.2 -- 0.8 - . , .:. -. l 0 .6 - _|,§ g"_' ,l 0.4 0 - co 2 E a 0 cu c E co 0 8 ' = d 2 c, '°,) Figure All. WBES - General Business Constraints, Brazil 4 3.5 - 3 - 0 C ~~~~~~~~~~~~~~~~~~~~~~C Figure Al12. WBES - General Business Constraints, Brazil and LAC 4- 3.5- 3 -1 2.5 O~~~~~~~~~~~~~~~~~~3Brazil 7 0~~~ Argentina 2- 1.5- QGChile 1~ 0.5- 0 o 2 02 ; V. ~ ~ ~ ~~, 0 0 ~~ ~~0 IV CL ~~~~~4, 253 Figure A13. Global Competitiveness Report - Obstacles to New Businesses 7 - 6 o Legal protection 61- E=onorNc conditions 4- 3 - D3 Financial condftions 2 0~~~~~~~~~~~~~~[ Adnunistrative conditions 0 I L . 0 co 0 CU co cCD0) ~ e . o0 Figure Al14. Djankov et al. (2000) - Number of Procedures for New Businesses 20- 1 8- 1 6- 1 4 -1 1 2 - 10 '5 ° 8- 6- o~~~~~~~~~~~~~~~~o c Figure Al 5. Djankov et al. (2000) - Direct Cost for New Businesses as Fraction of GDP Per Capita 0.6- 0.5 0.4- 0.3- 0.2- 0~~~~~~~~~~ 8 ~~~ V V co~~~( 21~~~~~ < e 0 5 - r" C ( 31 Oq M -4 M (0 register with CD Commercial Board > print authorized . receipts & invoices 0 . 0 0 0 0 0 0 0 0 register for Brazil . municipal taxes CO Argentina CD register for state taxes CD \>> a7 ~~~~~Mexico |.ktdi"- CD~ ~ ~ ~ ~ ~ ~~~~ municipal permit Chl 7 . 3 _. Chile _ ' ::0 -' | a7~~~~~~~~~~~0- CD receive municipal c'f7 . 1 inspection a7Korea jj v cn On register employees Malaysia _ , 0 for social security c .3 approval of Poland t - wa employee registry book 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C0 . . | a7~~~c Ireland ( a open umemployment c (FGTS) account L CD register social integration 5.Z CD ~~~~~~~~~~z (PIS) program a. CD 0< notify Ministry of Labor CD a°) A) A° a°7 CD s M b a7~ ~~ ~~~~~~~~~~~~~~~~~~~~~C 0 0 0C0 file application with registrar notarize deeds make a company seal LI publish deeds register for taxes & employment payments incorporation register . (0n 1 \ TtT1 CD print invoice/receipts -' 90 ~~~~~~~~~~~~~~~~~~~~-4 CD register for taxes C= a.~~~~~~~~~~~~~~~~~~~~~~~~~~C . C1 wn. municipal inspection CD CD~~~~~~~~~~~~~~~~~~~~~~C CA . . ~ ~ ~~~~~~~~~~~~~~(D -\\cno Ci) cn patente municipal r. CDc co U register for Isapre CL~~~~~~~~~~~~~~i Ci,~~~~~~~~~~~~~~~C 0.~~~~~~~~~~~~~~~ . .0, 0) CD . - reise b 6 : . OD . . .acd tn 0 0 0 Q . . . ..' Q - - ,. 0 9 999 .~~~~~~~~~0 0 -. ' . -.d.. r > O O O ,0 0 .0 ,0 .0 > I ' I ' I ' r'., 1 1 256 Figure A19. FDI and Portfolio Equity Flow 35000 - 30000 25000 - 20000 - . Pbrtfolio equity 15000 - 10000 - 5000 , 0 Clj Cl) to o CO N co a) 0) 0) 0) 0) 0) 0C 0) 0) 0) 0) 0) 0) 0 0 0) 0) 0 0 Figure A20. FDI as percentage of GDP 16 14- 12-- Brazil 12- --w- Brazil w.o. priv. 10 - Argentina 8 / W MAco 6 " - ggt- Chile 4 i Z' i - ~ -.- 9 «JJa Korea /t / -,- f = China 2 -- 0 o _ N 0 L U0 CD 9 co 0) 0) a) a)a0) 0 a) a 0 0 ) 0) 0 a 0) 0 M a) a) a) a) a) 0 0_ _ Figure A21. Brazil FDI Inflow 40000 35000 - 30000 - 25000 - mother 20000 - te 20000 "; lspnstization 15000 - 10000 5000 0 0) 0 0) 0) 0 0 _ _ r c _ N 257 Figure A22. Domestic Investment and FDI 200 - 180 160- 140- 120 | Cap. 100 - F ,DI 80 - -. + FDI w.o. pr. 40 - 20 LO co c a) a) 0) a) C) 0C a) Figure A23. Brazil FDI and cross-border M&A 35- 30- 25- 15 -.M&A (B$) v0 -.-,-. a) a) a) a) a) v0) O vM o ou X O O vO a) a) a 5~~~ - a) a) a) CO a) D a) CD a) a) Figure A24. Brazil FDI by Sectors 35000 - 30000 - 25000- 20000 - c 0Mning &Agri. . Sersices 15000 - D Manufacturing 10000 5000- 0~ CD WO a) 0 co co-a) 0) 0 a) a) a~~C) CY) 0 - - ~~~~~~~~~clJ 258 Figure A25. AT Kearney - Drivers of FDI into Mexico and Brazil 100 90 4 80 r , Market potential 70- 60 U aNrket size 50- 40- 03 Low production 30- costs 20- 10 _ 0. Brazil Mexico Figure A26. Sector breakdown of FDI inflow Labor 80 produdcivty, Indexed 70- to US 60 - oAutormtive 50 , 9* Mturilagy 0 Food Proc. O Retail 30 - - Construction 20 -O Telecom 10 - ) Finance 0 Value added, 0 1 2 3 4 5 6 7 8 9 10 %ofGDP Note: Size of circle proportionate to average share of FDI Source: McKinsev(1998). World Bank SIMA 259 Figure A27. FDI component as share of total in select countries (period average) 95-98 r 7eg Poland 90-94 95-98-J' . _ = Mexico 90-94 ; C#, |_ [ aEQUTY CAPIT AE *REINV EARNINGS 95-98 - -OTHER CA PITAL 90-94 95-98 - Argentina 90-94 00/0 20% 40%h 60% 80% 100% Source: IMF Balance of Payments Statistics Figure A28. AT Kearney - Top Decision-Making Factors for Long Term Investment into LAC 100 90 80 W 70 60* 40 H 10 I 0~~~~~~~~~~ Figure A29. World Bank Global Development Finance (GDF) fL~~~~S 200 lindex] -50 _ _ _ _ _ _ _ __w_C L L EuFoWepn Governance indicator 4,3 Li i~L CsO C'. W_~'.