42851 rev. Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Edited byCem Mete This report is part of a series undertaken by the Europe and Central Asia Region of the World Bank. The series covers the following countries: Albania Lithuania Armenia Moldova Azerbaijan Montenegro Belarus Poland Bosnia and Herzegovina Romania Bulgaria Russian Federation Croatia Serbia Czech Republic Slovak Republic Estonia Slovenia FYR Macedonia Tajikistan Georgia Turkey Hungary Turkmenistan Kazakhstan Ukraine Kyrgyz Republic Uzbekistan Latvia ECONOMIC IMPLICATIONS OF CHRONIC ILLNESS AND DISABILITY IN EASTERN EUROPE AND THE FORMER SOVIET UNION ECONOMIC IMPLICATIONS OF CHRONIC ILLNESS AND DISABILITY IN EASTERN EUROPE AND THE FORMER SOVIET UNION Edited by Cem Mete 2008 The International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org All rights reserved. 1 2 3 4 11 10 09 08 This volume is a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and / or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete informa- tion to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA: telephone: 978-750-8400; fax: 978-750-4470; Internet: www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. ISBN-10: 0-8213-7337-4 ISBN-13: 978-0-8213-7337-8 e-ISBN-10: 0-8213-7338-2 e-ISBN-13: 978-0-8213-7338-5 DOI: 10.1596/978-0-8213-7337-8 Library of Congress Cataloging-in-Publication data has been requested. Contents The Report Team and Acknowledgments xi Key Findings and Future Research Directions xiii PART I A REGIONAL OVERVIEW 1 1. Introduction 3 Cem Mete with Jeanine Braithwaite and Pia Helene Schneider Regional Context 4 Different Definitions, Different Prevalence Rates 5 Main Causes of Disability in the Region 9 Employment and Disability 9 Heterogeneity among the Disabled 12 Earnings Disadvantage of the Disabled Is Larger in Transition Countries 12 Poverty and Disability 14 Does Employment Protect the Disabled from being Poor? 16 Health Shocks, Employment, and Poverty 17 Social Protection Transfers and the Disabled 17 Disabled Children's Limited Opportunities to Build Human Capital 19 Other Nonmonetary Costs of Disability: Caring for the Disabled 22 Appendix I: Disability-Related Questions in Available Household Survey Data 25 Appendix II: Basic Information on Time-Use Surveys 28 Appendix III: Proxies for Disability and Chronic Conditions Used in the Remaining Chapters 29 Notes 30 v vi Contents PART II COUNTRY STUDIES 33 2. Measurement of Disability and Linkages with Welfare, Employment, and Schooling 35 Kinnon Scott and Cem Mete Introduction 35 Measuring Disability 36 Incidence of Limited Physical Functioning and Health 41 To What Extent are Disabled Individuals Poor, Less Educated, and Out of the Labor Force? 48 Conclusion 56 Appendix I: Health, Disability, and Physical Functioning Questions in URPS, Waves 1, 2, and 3 58 Notes 65 References 65 3. The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 67 Cem Mete, Huan Ni, and Kinnon Scott Introduction 67 Bosnia and Herzegovina Context 68 Data 70 Empirical Framework 70 Results 72 Conclusions 80 Notes 81 References 82 4. Health Disabilities and Labor Productivity in Russia in 2004 85 T. Paul Schultz Introduction 85 Institutional Change and Uncertainties of the Transition: Mental Disabilities 87 Descriptive Statistics of the Russian Survey Population 89 A Conceptual Framework to Guide the Econometric Analysis 98 Empirical Findings 100 Conclusions 108 Appendix 110 Notes 114 References 116 Contents vii 5. The Implications of Poor Health Status on Employment in Romania 119 Cem Mete and Shirley H. Liu Introduction 119 The Context: Romania in Transition 121 Data and Descriptive Trends 122 Results 123 Conclusions 129 Appendix 131 Notes 135 References 135 Box Chapter 1 1.1 Defining Disability 6 Figures Chapter 1 1 WHO Definition of Disability 7 2 Various Definitions of Disability Incidence in Uzbekistan 7 3 Prevalence of Disability by Age Group 8 4 Prevalence of Chronic Illness by Age Group 8 5 Probability of being an Employee/Wage Employee 10 6 Employment Rates of Disabled and Nondisabled Individuals 11 7 Educational Attainment and Employment of Those with Congenital Disability Versus Other Disability 13 8 Disabled and Chronically Ill Wage Employees Earn Less 13 9 Relationship between Household Wealth and Disability at Various Life Stages 14 10 Disability Rates by Consumption-Based Poverty Status 15 11 Disability Rates by Asset-Based Poverty Status 16 12 Percentage of Disabled and Nondisabled Individuals in the Poorest Quintile of the Consumption Distribution 16 13 Persons Receiving Disability Benefits per 100,000, in 2003 18 14 Distribution of Disability Pension Beneficiaries by Household Consumption 19 15 Distribution of Disability Pension Beneficiaries by Household Consumption 20 16 Enrollment Rates of 16- to 18-Year-Olds 21 17 Percentage of Individuals Spending Time Assisting Family Adult, and Time Spent 23 18 Who Takes Care of Disabled Household Members? 24 viii Contents Chapter 2 1 Incidence of Disability Using Various Definitions in Uzbekistan 37 Chapter 4 1 Economic Activity of Russian Population, 2004 90 2 Proportion of Russian Population with a Pension, 2004 91 Chapter 5 1 Quality of Water and Public Health Services in 21 Transition Countries 120 2 Quality of Water and Public Health Services in 21 Transition Countries 121 Tables Chapter 2 1 Indicators of Physical Functioning 38 2 Characteristics of the Panel and Full Samples 41 3 Incidence of No and Full Limitations by Domain of Physical Functioning 42 4 Correlation of Scores for Domains of Physical Functioning 43 5 Disability and Physical Functioning 44 6 Alternative Measures of Disability and Physical Functioning 45 7 Correlations of Alternative Measures of Disability to Official Disability Status 46 8 Probability of Having Official Disability Status 47 9 Consumption and Disability: Log of per Capita Consumption Is the Dependent Variable, OLS Coefficients, and Standard Errors 49 10 Gross Enrollment Rates by Economic Region in Uzbekistan. 2002 to 2003 50 11 Probability of School Enrollment: Children Age 7­14 51 12 Probability of School Enrollment, Ages 15­18 52 13 Years of Schooling Obtained 54 14 Probability of Being Economically Active 55 Chapter 3 1 Correlation among Indicators of Health and Disability 72 2 Household Heads: Sample Statistics, Means, and Standard Deviations 73 3 The Effect of a Change in Household Head's Health on His or Her Hours Worked 74 4 Effect of a Change in Household Head's Health on Labor Supply and Earnings 75 5 Comparison of Household Situations before and after Health Shocks 76 6 Effect of a Change in Household Head's Health on Per-Capita Household Consumption 77 7 Marginal Effects of Changes in Household Heads' Health on their Children's Schooling 78 8 Effects of Decomposed Changes in Household Head's Health on Change in His or Her Hours Worked 79 Chapter 4 1 Russian Longitudinal Monitoring Survey for 2004: Labor Market Contents ix Participation, Hours and Wages by Age and Sex 90 2 Indicators of Poor Health, Classified Disabilities, Chronic Health Conditions, by Age and Sex 92 3 Health-Related Characteristics and Behaviors, by Age and Sex 94 4A Labor Force Productivity Determinants, Including Current Self-Evaluated Health Status: Males 101 4B Labor Force Productivity Determinants, Including Current Self-Evaluated Health Status: Females 102 5 Labor Force Productivity Determinants of Health Status, with and without Inputs 104 6A Labor Force Productivity Determinants, Including Medical Care, Smoking, and Alcohol: Males 107 6B Labor Force Productivity Determinants, Including Medical Care, Smoking, and Alcohol: Females 107 A1 Definitions and Sources of Variables and Sample Statistics from Russian Longitudinal Monitoring Survey, 2004 110 A2 Joint F-Test of Significance of Identifying Instruments in First-Stage Regressions 113 A3 Hausman Specification Tests (t or F) of the Exogeneity of Dependent Variables 113 Chapter 5 1 Means and Standard Deviations of the Variables Examined in the Models of Health and Employment Status 123 2 Health Environment Variables by Household Wealth 124 3 Predictors of Less-than-Good Health Status 128 4 Predictors of Employment 129 A1 Predictors of Less-than-Good Health Status of Males 131 A2 Predictors of Less-than-Good Health Status of Females 132 A3 Predictors of Employment of Males 133 A4 Predictors of Employment of Females 134 The Report Team and Acknowledgments This report was prepared by a team led by Cem schemes. Stefania Rodica Cnobloch provided Mete and comprising Kinnon Scott (coauthor assistance with the analysis of all data sets that of chapters 2 and 3), Huan Ni (coauthor of were used in the overview chapter, covering a chapter 3), T. Paul Schultz (author of chapter wide range from household budget surveys, to 4), and Shirley Liu (coauthor of chapter 5). Jea- living standard measurement surveys, to inte- nine Braithwaite provided valuable input into grated household surveys, to time-use surveys. the overview chapter on the distribution of dis- Daniel Mont provided constructive suggestions ability pension beneficiaries, and secured funds throughout the project, based on his in-depth to support the piloting of a disability survey knowledge of the disability literature in indus- instrument that provided important insights trialized countries. He also worked closely with into our understanding of the challenges Kinnon Scott on the design of the piloted dis- involved in defining and measuring disabilities. ability survey questionnaire. Pia Helene Schneider also contributed to the The team benefited from detailed comments overview chapter, focusing on the main causes from Martin Raiser, Philip O'Keefe, Akiko of disability in the Eastern Europe and Central Maeda, Jane Falkingham, Marianne Fay, Arup Asian countries, and policy implications of the Banerji, Vedat Rmljak, Anthony Ody, Sally M. observed trends. Lucian Pop provided useful Zeijlon, Mamta Murthi, and Eluned Roberts- advice on the nature of social protection pro- Schweitzer. At the beginning of the project, a grams and data sets in the region, building on critical issue that the team sought guidance on his experience in analyzing the (pro-poor) tar- was how to sharpen the focus of the proposed geting performance of various social assistance work, since very few (quantitative) papers have xi xii Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union been written on the broader topic of the eco- of work, the advice that emerged from these nomics of disability using data from Eastern meetings was that the World Bank should focus European and Central Asian countries. Thus, on two key issues in the short term: the relation- one can legitimately argue that there is a need to ship between poverty and disability--because of investigate each and every subtopic in an in- the World Bank's mission and also considering depth manner. Priorities had to be established, how little we know about this two-way not only considering the importance of the issues relationship--and the broadly defined topic of and the comparative advantage of the World "service delivery"--because a large share of Bank Bank, but also taking into account time, resource, projects deal with service delivery of one type or and data constraints. In addition to the guidance another. This particular report focuses exclu- from Chief Economist Pradeep Mitra's office, a sively on the poverty-disability relationship and disability conference hosted by the World Bank various extensions of it, including the linkages in late 2004 entitled "Disability and Develop- among disability and employment, school enroll- ment: Setting a Research Agenda," turned out to ments, and time-use patterns of adults. The work be particularly relevant for this purpose. This was carried under the general direction of conference brought together an esteemed group Pradeep Mitra (chief economist, Europe and of researchers and policy makers to discuss possi- Central Asia Region) and Arup Banerji (manager, ble contributions that the World Bank can make Human Development Economics, Europe and in this area. When it comes to the prioritization Central Asia Region). Key Findings and Future Research Directions A concise summary of the findings of this the pros and cons of each proxy for disability to research, along with a list of priority areas that make the most use of available information to may be tackled by future studies, are presented guide policy makers. below. In the Eastern European and Central Asian context, where most countries are well advanced in terms of fertility transition, the Different Definitions of Disability most common type of disability is restriction of movement, and a large share of the disability Alternative definitions of disability--such as a burden is due to noncommunicable diseases country's official disability classification, self- and injuries. The composition of the disabled reported disability status, functional disability population has economic implications, because assessment, Activities of Daily Living Index, individuals with movement restrictions are the Instrumental Activities of Daily Living Index, most disadvantaged group in terms of employ- self-reported chronic illness, and self-reported ment prospects, along with those with congen- health status--are highly correlated with one ital disabilities. The aging transition countries another. Also, the cross-country evidence con- may be able to contain one type of financial firms the sharp age gradient in reporting of burden by being restrictive in granting disabil- health ailments, especially for the reporting of ity benefits to the elderly. But the functional chronic illnesses. But different definitions lead limitations increase steeply by age with impli- to significantly different estimates of the preva- cations for employment and productivity, as lence of disability. It is important to understand discussed next. xiii xiv Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Linkages with Employment, Earnings, impact of disability on employment accumulates and Poverty over time after the start of disability, which has implications for the design of social protection The linkages between disability and economic programs. Also, in contrast to what is observed in and social outcomes of interest tend to be Organisation for Economic Co-operation and stronger in transition countries when compared Development (OECD) countries, the employ- with industrialized countries. Despite having ment rates of disabled and nondisabled individu- experienced respectable economic growth rates als are not correlated in transition economies. starting from the late 1990s, poor population Thus general-purpose pro-employment policies health status and rising health inequalities may not necessarily improve the employment emerge as main obstacles for equitable and sus- rates of the disabled in the transition economy tainable economic growth and poverty reduc- context. This divergence in employment trends tion in the region. is driven by the presence of a large informal sec- Disabled adults are much less likely to work tor in transition countries, as discussed in more when compared with nondisabled adults in all detail by the overview chapter. transition countries considered here. This The poor are more likely to be disabled. This ranges from a high of 60 percentage points less finding is robust across countries, and is visible likely to work in Moldova, to a low of 20 per- both when poverty is measured via a household centage points in Bosnia and Herzegovina. The consumption aggregate and a household assets disabled and chronically ill also earn substan- index. Having said that, the variation in the dis- tially less than others: the earnings gap is larger ability rate based on poverty status is not very for those who categorize themselves as "dis- large in some countries. abled" compared to those who report having There is evidence that employment protects chronic illnesses only. Furthermore, this analy- the disabled from being poor. Of the countries sis shows that simple associations tend to down- with relevant data, Romania is the only excep- play the linkage between poor-health/disability tion to this rule, and in that country, being a and employment because instrumental-variable wage employee does not remove the poverty estimates that attempt to single out causal disadvantage that affects disabled individuals. effects produce larger estimates. Heterogeneity within the disabled deserves attention as well. For example this report shows Disability Benefits and the Poor that other things being equal, adults with con- genital disabilities are less likely to be employed. In most OECD and Europe and Central Asia This may be because those with congenital dis- (ECA) countries, disability benefits as a percent- abilities are exposed to the disadvantages of age of GDP have increased since 1990. How- being disabled (in terms of intrahousehold ever, there is wide variation in the share of resource allocation, access to quality education individuals qualifying for disability benefits in etc) for a longer duration of time. transition countries, with Croatia, Poland, Even though the disabled who are employed Hungary, and Estonia reporting about twice as work less than others, the difference is less than many beneficiaries than the European Union five hours a week in Moldova and Bosnia and (EU) average, and poorer transition countries Herzegovina. However, it is sizable in Poland, (Kyrgyzstan, Tajikistan, Uzbekistan, and Roma- at about nine hours per week. nia) reporting less than half of the EU average. In contrast to the trends observed in some Disability pensions are well targeted to the industrialized countries, in at least one transition poor in most countries of the region. Two low- economy (Bosnia and Herzegovina), the negative income Commonwealth of Independent States Key Findings and Future Research Directions xv (CIS) countries, Tajikistan and Georgia, are well as those without tertiary education, spend exceptions to the rule, with almost uniform dis- more time assisting adult household members. tribution of disability pension beneficiaries Among households that report provision of regardless of household consumption. Thus adult care, time spent for this purpose is much there is room for improvement in the coverage higher in Romania (at more than 80 minutes per of disability benefits and the pro-poor targeting day on average) than in Estonia, Hungary, and performance of benefits in poor transition coun- the United Kingdom (all below 55 minutes per tries, where the official disability rates tend to day on average). Thus, it could be that in poor be particularly small. countries, the nonmonetary costs of disability The sizable discrepancy between the official are higher, although it is not possible to make disability rate and other definitions of disability sweeping conclusions on this topic because can make different demographic groups more comparable time-use surveys have not been vulnerable if their poor health status is not rec- implemented in other Eastern European and ognized as a disability that triggers support in Central Asian countries. In urban Romania, the form of social assistance. In particular, there Hungary and Estonia, time-use patterns are is evidence that the elderly and females are less closer to those observed in Netherlands and the likely to receive official disability status, after United Kingdom, and thus continued urbaniza- taking into account other individual character- tion may lead to a convergence across countries istics, including levels of functional limitations. in time spent on home care. This research shows that households are Finally, disabled children are significantly unable to cope with major deteriorations in the less likely to enroll in school. Neither Millen- health of the head of the household (as measured nium Development Goals (MDGs) nor the by the individual moving from nondisabled to Education for All Initiative can succeed in the disabled status over time). But household con- absence of a renewed commitment to disabled sumption is not sensitive to the gradual deterio- children's schooling outcomes. Children's rations in activities of daily living, or to the onset human capital accumulation is also sensitive to of a new chronic disease. At the individual level, the deterioration of the health status of their considering the evidence that the employment parents: There is some empirical evidence that consequences of being disabled worsen over children are more likely to drop out of school if time, there is a need to examine both the dura- their parents experience health shocks. tion of disability compensation, as well as the capacity (in terms of skills) and incentives for the individual to reenter the labor force. Future Research The policy implications of certain findings Nonmonetary Costs of Disability and require further consideration. One example is Chronic Illness the finding that poor households are not able to fully absorb the income loss caused by major There are significant nonmonetary costs of dis- health shocks to the head of the household. Even ability as well. Nondisabled individuals who live though universal catastrophic health insurance in a household that has at least one disabled schemes can be considered in such cases, it is not individual spend two to five times (in Hungary clear if such universal insurance programs can be and Estonia, respectively) as much time assist- implemented successfully in poor countries ing adult family members, compared to nondis- where a large segment of the population is abled individuals who live in a household employed in the informal sector (making it diffi- without anyone who is disabled. Females, as cult to collect insurance premiums from them). xvi Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Another example is the finding that in Bosnia research would be to calculate the costs and and Herzegovina, after the onset of the disabil- benefits of alternative preventive interventions ity, the decline in the number of hours worked in a way that can be compared to some of the and earnings becomes more severe over time statistics on the direct costs (such as loss of (the opposite trend is observed in the U.S.). income) and the indirect costs (such as the value This research revealed that--as opposed to of time devoted for the care of the disabled) of what is observed in OECD countries--the disability that are presented in this report. It also employment rates of disabled and the nondis- would be beneficial to compare the payoffs from abled individuals are not correlated in transition preventive interventions in developing coun- economies. The implication is that without tries to those in industrialized countries. institutional and legislative reforms (that also The researchers' ability to further this line of consider the informal sector) the markets are work will depend on the availability of relevant unlikely to fix this particular development chal- data sets. As discussed later in this report, there lenge. The solution will have to rely on a set of has been some progress in the way disabilities, factors including overall improvements in the chronic illnesses, and restrictions on "activities economy (which, at the very least, would serve of daily living" are captured in surveys. The early to increase the resources available for public efforts have primarily focused on ensuring the service delivery and social protection pro- presence of "correct" disability questions in the grams), changes in the duration of disability census data. This approach would improve the benefits, implementation of training programs estimates of the prevalence of disability in devel- to facilitate the transition from one type of job oping countries, but the limited scope of a typi- to another, ensuring the existence of incentives cal census questionnaire would be of little use to for disabled individuals to go back to work, enhance our understanding of key relationships addressing workplace discrimination, etc. of interest. Thus, one should not underestimate In some cases there may be tradeoffs between the potential of improving the design of house- efficiency and equity, and the way such tradeoffs hold surveys to inform policy makers. This work are tackled may be especially important for shows that increasing the sample size of a survey developing countries with limited resources. from 12,387 to 32,337 produced remarkably This research demonstrates the significant similar estimates of the prevalence of disabilities enrollment disadvantage of disabled children, and thus the inclusion of "correct" set of ques- although the solution to this challenge (which tions in standard LSMS-type household surveys might include training for teachers to enhance can produce valuable information despite rela- the benefits of an integrated teaching environ- tively small sample sizes. ment, or in some cases might require special- Finally, it is useful to point out the topics that ized education arrangements) will probably would benefit from elaborate analysis but are need to be formulated separately for rural and outside the scope of this particular research urban areas, taking into account the numbers of project. These include the social integration of disabled children involved. More generally, the disabled individuals, the status and shortcom- service delivery arrangements for disabled chil- ings of institutionalized care in the region, alter- dren require further research in the developing- native home care and community care models, country context. transport and infrastructure, detailed sectoral Another useful avenue for policy-relevant perspectives, and discrimination. PART I A REGIONAL OVERVIEW 1 CHAPTER 1 Introduction Cem Mete with Jeanine Braithwaite and Pia Helene Schneider* Disability is an important issue for the countries the life expectancy of Russian males during the of Eastern Europe and the former Soviet early 1990s (Bobadilla, Costello, and Mitchell Union, in large part because a significant por- 1997; Cornia and Paniccia 2000). tion of the population is either in poor health or In this context, the economic costs of poor disabled, which has implications for labor force health status and disabilities did not receive participation rates and productivity. Especially much attention because in a high-unemploy- in aging transition countries, the sustainability ment environment with an abundant supply of of social protection programs is also a concern skilled labor, the bottlenecks in the labor mar- due to the projected increases in the share of kets were (and are) considered to be in the disabled populations. domain of labor demand, not labor supply During the first phase of the transition from (World Bank 2005). Yet this situation is rapidly socialism to market economy, poor health status, changing because many transition countries disability, and premature mortality of individuals have experienced respectable economic growth received attention primarily as indicators of rates starting from the late 1990s, and the labor reduced living standards in the region. The markets are starting to tighten in many cases. inequality implications of the transition process For example, in 8 of 20 transition countries were also highlighted, in particular through the studied by the World Bank (2005), the unem- analysis of the abrupt and significant decline of ployment rate was at or below the EU-15 aver- * Cem Mete is a senior economist at the Europe and Central Asia region of the World Bank. Jeanine Braith- waite is a senior social protection economist at the Human Development Network of the World Bank. Pia Helene Schneider is a senior health economist at the Europe and Central Asia region of the World Bank. 3 4 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union age of 9 percent in 2003, even though employ- are only mentioned in the broad discussion of ment levels in many transition countries remain "vulnerable groups." Another is the way in below the EU-15 average of 65 percent.1 which redistributive policies are emphasized Failing to deal with disability issues or deal- instead of "unlocking the economic potential of ing with them in an inefficient manner can be the disabled individuals" (ILO 2002). very costly--especially for the rapidly aging This report aims to fill in the knowledge gap transition countries that aim to reach the Mille- in this field by analyzing cross-country data on nium Development Goals (MDGs). Poverty basic indicators, and by carrying out more reduction and universal primary school enroll- detailed empirical analysis on causal relationships ment MDGs seem particularly at stake, yet at of interest, including the impact of disability on this stage the empirical knowledge base in this employment, wages, poverty, and children's area is extremely weak. In particular, there is a school enrollments--focusing on four transition remarkable absence of quantitative information countries with household survey data sets that on the key linkages among disability and allow more elaborate econometric analyses. This employment, earnings, poverty, and children's is a tightly focused effort, leaving out a number of school outcomes.2 important topics that researchers may want to This report argues that it is timely to bring tackle in the future. The excluded topics include the economic costs of disability to the forefront the social integration of disabled individuals, the of development policy because of the large status and shortcomings of institutionalized care impact poor health status and disabilities have in the region, alternative home care and commu- on employment, poverty, children's schooling, nity care models, transport and infrastructure, and time spent in caring for disabled individu- detailed sectoral perspectives,3 discrimination,4 als, especially by adult females (which in turn and cost-benefit analysis of prevention against inhibits higher female labor force participation certain types of disability.5 prospects). In fact, the evidence provided here suggests that the linkages between disability and economic and social outcomes of interest are Regional Context stronger in transition countries when compared with industrialized countries. As a result, poor Under the Soviet/Yugoslav system, disabled indi- health status and disability emerge as major viduals were both protected and isolated from the obstacles to equitable and sustainable economic general population. Disability was one of the very growth in the region. few acceptable reasons for an adult not to work, In recent years, there has been some recogni- but disability was viewed in a narrow, medical tion of the need to discuss disability issues in way, and its study was (and is) termed defektlogia strategy documents such as Poverty Reduction in Russian--the study of defects. Parents were Strategy Papers (PRSPs) and country assistance encouraged to place children in residential insti- strategies (CASs). But in the absence of basic tutions, as it was thought that institutions could empirical evidence on the living conditions and do a better job of raising disabled children than behavior of disabled individuals, it is a challenge could parents. Noninstitutionalized children to formulate concrete steps to tackle this partic- with disabilities were typically segregated in spe- ular economic development problem. In fact, it cial schools and disability was highly stigmatized. is a challenge to define the magnitude and char- However, adults with disabilities were encour- acteristics of the problem. Not surprisingly then, aged to join collectives of persons with the same there is some dissatisfaction in the way disabled medically defined disability, such as associations populations are covered in the existing strategy for the blind and deaf. Mental disability was even documents. One criticism is the way the disabled more highly stigmatized. Introduction 5 The impact of transition on disability was increased prevalence of informal sector employ- pronounced. In many Eastern European and ment, which lacks the regulations and social pro- former Soviet Union countries, the number of tection benefits that come with formal sector (officially recognized) disabled individuals employment. Still, there are some differences increased significantly between 1991 and 1997, among the countries in the region in terms of reflecting several factors, including the prefer- available resources to tackle the disability- ence of employers to avoid paying severance pay related challenges due to geographical position- to fired workers--instead placing them on dis- ing--in particular, some transition countries are ability rolls--and the sharp deterioration in either EU members or on the EU membership health indicators (particularly for adult men) and path, while others will become neighbors of EU disruptions in the health system. At the same members. There are even differences in the time, financing for residential institutions was extent to which they are exceptions to the facts devolved to localities, without specific revenue outlined above (for example, fertility rates are sources, thus resulting in chronic underfinanc- relatively high in Tajikistan, so for that country, ing for such institutions--which, ceteris paribus, aging is not an issue in the medium term). may have reduced the number of institutional- ized individuals through demand- and supply- side effects. With the freeing of civil society, Different Definitions, Different disabled persons' organizations began to form, Prevalence Rates in some cases out of the old Soviet collectives, in other cases from exposure to international non- Capturing the incidence of disability is difficult. governmental organizations (NGOs), and in The World Health Organization (WHO) esti- some cases from the grass roots, including from mates that about 10 percent of the world's popu- parent-teacher organizations. lation experiences some form of physical, Other facts that characterize most Eastern mental, or intellectual disability.8 Industrialized European and former Soviet Union countries countries with aging populations tend to report are a tradition of universal health care coverage higher disability rates, partly because of better on the positive side, but unsustainable or col- data on the disabled, and partly because these lapsing health systems and widespread informal countries can afford to (officially) acknowledge consultation fees on the other. They are also and provide disability benefits to a larger share well-advanced in terms of demographic transi- of their populations. The average disability tion (and as a result face all the challenges of prevalence in OECD countries is 14 percent, of "aging populations," including the old-age-dis- which one-third are severely disabled. Northern ability burden), but poor--unlike industrial European countries and Portugal report the countries that also face the same demographic highest disability prevalence.9 situation.6 They have educated populations, but Alternative definitions of disability provide preventive health behavior is not on par with significantly different estimates of the preva- what is observed in Western societies. There is lence of disability.10 For example, 3.8 percent of an increased prevalence of depression and men- the population aged 7 and older in Uzbekistan is tal health cases--especially, but not exclusively, officially considered disabled. Yet almost 12 per- in post-conflict areas such as Bosnia and Herze- cent of individuals in that age group have at least govina and Serbia. There have been significant one serious difficulty or a full limitation in phys- changes in the labor market environment in a ical functioning (figure 2, discussed in more relatively short period of time, with increases in detail in chapter 2).11 the share of private sector employment and ser- The most commonly encountered type of vice sector employment,7 along with the disability is movement restrictions, the least 6 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union BOX 1.1 Defining Disability Disability is an umbrella term that can refer to quite different health ailments, depending on the context. Alternative approaches to measuring disability include diagnosis-based assessments (e.g., "Does anyone in this household have epilepsy?"); Activities of Daily Living (e.g., "Do you have trouble dressing or bathing yourself?"); Instrumental Activities of Daily Living (e.g., "Do you have trouble maintaining the household?"); participation/social-role questions (e.g., "Do you have a mental or physical impairment that limits the amount or type of work you can do?"); or functional questions (e.g.,"Do you have difficulties concentrating, remembering, or making de- cisions?"). Administrative data, such as those reported by the Transmonee database, are some- times used for making cross-country comparisons of disability, but the wealthier countries with better administrative recordkeeping capabilities routinely come up as the ones with the highest disability rates. Depending on the purpose of the study or policy intervention in question, it is natural to work with different definitions of disability. In practice, availability of data also influences the disability defi- nition that is used. One line of research focusing on international comparisons in OECD countries makes extensive use of the Activities of Daily Living (ADL) and Instrumental Activities of Daily Liv- ing (IADL) restrictions, distinguishing among "severe disability," meaning individuals with one or more ADL restrictions; "moderate disability," meaning individuals without an ADL restriction but experiencing IADL limitations; and "little or no disability," meaning no ADL or IADL limitations (see Jacobzone, Cambois, and Robine [2001] and the references cited by the authors). More comprehensive but perhaps less empirically oriented definitions are proposed by the WHO. The 1980 International Classification of Functioning, Disabilities, and Health (ICF) makes the distinction among disorder, impairment, disability, and handicap (WHO 1980). The 2002 ICF revised the definition, with a major difference being the linkages to the environment in which an individual functions--be that the physical, institutional, or cultural environment--and linkages to "involvement in life situations." Despite the additional challenges they pose for measurement and standardization, the more comprehensive definitions of disability seem to have contributed to the formulation of recent strategy documents such as the Community-Based Rehabilitation approach advocated by the ILO-UNESCO-WHO Joint Position Paper (2004). For an in-depth dis- cussion on definitions in the context of social science, see Freedman, Martin, and Schoeni (2004) and OECD (2003). common ones are hearing and communications, this trend, their impact is unlikely to be large while vision and learning fall somewhere in enough to undo the aging effect. between.12 One implication of this finding is Even though different disability proxies lead that aging populations can expect the preva- to significantly different disability prevalence lence of disability to increase substantially over estimates, they are correlated with one another. time. Even if medical advances, positive changes Furthermore, it is possible to make generaliza- in preventive health behavior, and improve- tions about different "groups" of poor ments in health care service delivery slow down health/disability variables and their relation- Introduction 7 FIGURE 1 WHO Definition of Disability Health condition (disorder/disease) Impairment Activity Participation Contextual factors A. Environmental B. Personal C. Institutional ship with socioeconomic characteristics and Instead, each disability indicator has strengths poverty. This review of available evidence and weaknesses, which make some indicators reveals that it is undesirable to categorically better suited for the analysis of certain issues, favor one disability indicator over the others. but not others. Through an improved under- FIGURE 2 Various Definitions of Disability Incidence in Uzbekistan 70 60 50 tn 40 cer pe 30 20 10 0 7­16 17­26 27­36 37­46 47­56 57­66 66+ age groups one full limitation one serious diff. or limit own disability assess. own func. disability assess. chronic official disability Source: Authors' calculations based on URPS data sets described in chapter 2. 8 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union standing of the reasons why various indicators in figure 3 is that disability rates in Poland, the produce the trends that they do, one can wealthiest transition country with relevant data, improve the way we identify the most vulnera- display the sharpest age gradient, leading to ble groups in the population. much higher disability rates among the elderly, The cross-country evidence confirms the as compared to poorer transition countries. sharp age gradient in the reporting of health ail- This finding may have to do with the fact that in ments, especially for the reporting of chronic wealthier countries, the disability benefits are illnesses (figures 3 and 4). An interesting trend highly concentrated among people over age 50 FIGURE 3 Prevalence of Disability by Age Group 35 30 25 tn 20 cer pe 15 10 5 0 16­25 26­35 36­45 46­55 56­65 66+ age groups Bosnia-Herzegovina Bulgaria Georgia Moldova Poland Romania Source: Authors' calculations based on household survey data sets listed in appendix 1. FIGURE 4 Prevalence of Chronic Illness by Age Group 80 70 60 50 tn 40 cer pe30 20 10 0 16-25 26-35 36-45 46-55 56-65 66+ age groups Bosnia-Herzegovina Bulgaria Kyrgyz Republic Moldova Romania Tajikistan Source: Authors' calculations based on household survey data sets listed in appendix 1. Introduction 9 (OECD 2003), while in poorer developing policy focus on nutritional health and environ- countries, working-age adults may be favored mental interventions, ensuring access to iodized for the granting of disability benefits (chapter salt and protection against lead contamination, as 2), affecting whether the surveyed individuals well as a focus on reproductive health care. The identify themselves as disabled when approached 2002 Turkey disabilities survey reveals payoffs to by interviewers. early interventions because more than 40 percent of speech and mental disabilities are congenital, and between 20 and 25 percent of orthopedic, Main Causes of Disability in the Region sight, and hearing disabilities are congenital.17 As countries pass through the health transition, a larger share of the disability burden is due to Employment and Disability noncommunicable diseases and injuries. Accord- ing to the Global Burden of Disease project, the Poor health status/disability is likely to be more main causes leading to disability among men in detrimental for labor force participation in tran- Eastern European and Central Asian countries sition countries as compared to industrialized are neuropsychiatric conditions13 (35 percent of countries because the health systems in many Years Lost to Disability, or YLD), unintentional transition economies are experiencing serious injuries due to such things as falls and traffic problems with service delivery, quality of care, accidents (12.5 percent), sense organ (vision or and even availability of medicines and equip- hearing) diseases (8.3 percent), cardiovascular ment.18 As a result, some health conditions that diseases (8 percent), and musculoskeletal dis- today do not have much of an impact on the eases (6.6 percent). The statistics are similar for daily functioning of individuals in industrialized women.14 These trends are reflected in percent- countries may still be a cause for concern in age-of-deaths-by-cause statistics, where the transition economies.19 Furthermore, manufac- share of noncommunicable diseases among tran- turing and agriculture sector jobs, which tend to sition countries is consistently high, at between be more demanding physically and also more 75 percent and 85 percent.15 prone to work conditions that may cause dis- For individuals 45 and older, neuropsychi- ability, still dominate the economic environ- atric diseases, diseases of the sense organs, and ment in transition countries. But it is also cardiovascular and musculoskeletal diseases are possible that the spread of medical advances the main cause for YLD. Among adults between across countries, and a shift toward service sec- ages 15 to 44, on the other hand, the cause pat- tor employment over time in transition coun- tern of disability reveals the importance of men- tries, may counteract these effects. Which set of tal health. About 30 percent of the ECA total of factors dominates at this point in time? YLD among men is due to disease and injuries The only transition country (Poland) that incurred at ages 15­29, and 27 percent in the was included in the cross-country analysis of most productive ages 30­44. At similar disabil- OECD (2003) is also the one where the relative ity rates, the main cause for YLD among women employment rate of disabled over nondisabled aged 15­44 is depressive disorder, injuries, and people is lowest, at around 0.3. In contrast, the maternal conditions. For men in this age YLD corresponding ratios in Switzerland, Mexico, is mainly due to depressive disorder, alcohol Korea, France, Norway, Canada, and Sweden abuse, musculoskeletal conditions, and injuries. are above 0.7. The empirical evidence presented For children aged 0 to 4, YLD is mainly due next shows that disabled adults in transition to iodine deficiency, lead-caused mental retarda- countries are indeed severely handicapped in tion, or birth traumas,16 suggesting that health terms of participation in the labor force. 10 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 5 Probability of being an Employee/Wage Employee employees wage-employees Poland Russian Federation Romania Bulgaria Bosnia and Herzegovina Georgia Moldova Kyrgyz Republic Tajikistan 60 40 20 0 20 40 60 chronic condition disability Source: Authors' calculations based on household survey data sets listed in appendix 1. Note: Urban sample. The employment gap of the disabled individuals is presented as absolute values of percentage points. Disabled adults are much less likely to work centage points and female participation by 6 per- when compared with nondisabled adults in all centage points. This effect is larger if one con- countries considered here, ranging from a high siders worsening of health as captured by the of 60 percentage points less likely to work in official disability classification, reducing work- Moldova, to a low of 20 percentage points in ing by 40 percentage points for men and 22 per- Bosnia and Herzegovina (figure 5). Evidence centage points for women. Another important from Uzbekistan, discussed in chapter 2, reveals finding that emerges from the analysis of Russ- that while all disability indicators considered are ian data sets is that simple associations tend to negatively associated with employment, having downplay the linkage between poor health/dis- the official status of being disabled is the indica- ability and employment, since instrumental- tor linked to the largest decline in the probabil- variable estimates that attempt to single out ity of employment (a 52 percentage point causal relationships tend to be substantially reduction), followed by having a full limitation larger than the conventional ordinary least in any one of the six physical functioning squares (OLS) estimates (chapter 4). The analy- domains (a 24 percentage point reduction). sis of Romanian data sets reported in chapter 5 Perhaps the strong correlation between offi- reveals that the relationship between health ail- cial disability status and employment is to be ments and employment is stronger for those who expected, since those who are officially disabled are above 40 years old, reducing the probability may risk losing benefits if they work. In the of employment by 57 percentage points. Russian Federation, a one-unit deterioration in The disabled who are employed work less health (say, from good to average health) reduces than others, but the difference is less than five male participation in the labor force by 15 per- hours a week in Moldova and Bosnia and Introduction 11 Herzegovina, although it is sizable--at about nomic growth performance across the board in nine hours--in Poland. Further analysis of the the region, the employment prospects of dis- data from Bosnia and Herzegovina--which abled individuals cannot be entrusted to the mar- focuses on health shocks that occur between kets under the existing institutional frameworks. survey waves in an attempt to single out causal The challenge, of course, would be to develop a relationships, and takes into account main indi- supportive environment for the employment of vidual and household characteristics--reveal the disabled without introducing new rigidities that newly occurring disability leads to an eight- to employment legislation, which can slow the hour decrease in weekly employment, on aver- speed of economic recovery and poverty reduc- age. Deteriorating health status, as captured by tion in a region that experienced significant an ADL index, has a smaller impact, estimated declines in living standards during the early to at a five-hour decline in weekly hours of work mid-1990s.21 (see chapter 3). In Russia, the deterioration of Similarly, looking at the trends in three coun- health leads to a 9 percentage point decrease in tries with data at two points in time (figure 6), it hours worked, both for men and women (chap- is possible to see a case where the disabled ter 4). employment rate declined, even though the There is no correlation between the employ- overall employment rate rose. This trend is ment rates of disabled individuals and nondis- driven by nonwage employment, however. In all abled individuals across transition countries.20 three countries considered here, if the wage This finding contradicts the trends observed in employment of the nondisabled rises over time, the OECD countries, where employment rates so does the wage employment of the disabled of disabled and nondisabled individuals are (not reported). The presence of significant strongly correlated (OECD 2003). Thus, in the informal sector employment in the transition transition economy context, it is difficult to countries may thus explain why transition coun- argue that general employment-promoting poli- tries are different from OECD countries in this cies would automatically foster the employment respect. More generally, how formal and infor- of special groups in the short term. Even during mal labor markets respond to certain policy the second phase of transition, with strong eco- interventions and the resulting movements to FIGURE 6 Employment Rates of Disabled and Nondisabled Individuals 1998­2003 70 60 50 tn 40 cer pe 30 20 10 0 nondisabled disabled nondisabled disabled nondisabled disabled Georgia Moldova Romania 2003 1998 Source: Authors' calculations based on household survey data sets listed in appendix 1. 12 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union and from formal and informal employment (in cated disabled are also less likely to report "over addition to dropping out of the labor force) 90 percent reduction in ability to work and par- determine whether the disabled will benefit ticipate in social functions" (12 percent among from an economic boom.22 those with tertiary education, 43.4 percent among those with secondary education, and 44.6 percent among those with primary education or Heterogeneity among the Disabled less). This might be partly because the types of work that individuals with higher education can It is useful to recognize the heterogeneity undertake tend to be less physically demanding. among the disabled population in this context, The severity of the disability, approximated both because the diverging trends in transition by the three-category classification used by countries and OECD countries may be influ- some transition countries, reveals the expected enced by the composition of the disabled, and trends in terms of the likelihood of employ- because some of the disabled may be more dis- ment. In Moldova, only 5.8 percent of the most advantaged in terms of employment compared severely disabled are employed, with this per- to the rest. Few household surveys contain nec- centage increasing to 10.7 percent and 17.9 per- essary information on different types of disabil- cent as the severity of disability decreases. ity and have sufficient sample sizes to allow for Similarly in Poland, the employment rate of meaningful empirical analyses, but some those who rank their disability as "considerable" insights emerge from available data. is 8.5 percent, followed by 24.7 percent and 36.8 In Bosnia and Herzegovina, 26 percent of the percent for those who consider themselves as disabled report hearing or visual limitations, 38 moderately and slightly disabled, respectively. percent report mobility limitations, and the The timing of the disability deserves atten- remaining 36 percent report war-related, learn- tion as well. Other things being equal, those ing, or other disabilities. The age group under with congenital disabilities will be exposed to consideration matters: for example, the war- the disadvantages of being disabled (in terms of related, learning, and other category makes up intrahousehold decision making and resource 75 percent of all disabilities among the 24- to allocation; more limited access to education and 65-year-olds in Bosnia. Those with mobility health services; and perhaps to limited social limitations emerge as the most disadvantaged in interactions) for a longer duration of time, and terms of employment prospects, with an thus they may be more vulnerable later in life. employment rate of 9.8 percent, followed by Indeed, figure 7 shows that adults with congen- war/learning/other (17.5 percent employment ital disabilities are less likely to be employed. rate) and hearing/visually disabled (44.4 percent This relationship can be driven in part by this employment rate). group's educational disadvantage, however, In Bulgaria, a survey question that inquires since the same figure shows that adults with about the extent to which disabilities result in congenital disabilities are much less likely to reduced ability to work and participate in social have completed tertiary education compared to functions reveals that disabilities reported by the others who are disabled. elderly (aged 66 and older) are much more likely to lead to "over 90 percent reduced ability," with 43.6 percent of respondents choosing this Earnings Disadvantage of the Disabled option. The percentage drops dramatically by Is Larger in Transition Countries age, with 24 percent of those aged 24 to 65, and 28 percent of those younger than 24 saying they The OECD (2003) reports that there is little dif- have "over 90 percent reduced ability." The edu- ference in work incomes between disabled and Introduction 13 FIGURE 7 Educational Attainment and Employment of Those with Congenital Disability Versus Other Disability ages 24 to 65 70 60 50 tn 40 cer 30 pe 20 10 0 primary secondary tertiary employed primary secondary tertiary employed or less or less Bosnia and Herzegovina Tajikistan non-congenital congenital Source: Authors' calculations based on household survey data sets listed in appendix 1. nondisabled persons in many industrialized wages being substantially larger for the disabled countries--exceptions are the United States, compared to the chronically ill (figure 8). Sweden, and Portugal, where the earnings of dis- The analysis of Russian data that aims to sin- abled employees are at or below 70 percent of gle out the causal relationship between disabili- the earnings of nondisabled employees. The dis- ties and wages (reported in chapter 4) finds that abled and the chronically ill earn less than others the relationship between poor health status/dis- in transition countries, with the reduction in ability and hourly wages is not as strong as the FIGURE 8 Disabled and Chronically Ill Wage Employees Earn Less Moldova Bulgaria Romania Poland 0 10 20 tn cer pe 30 40 50 disability chronic condition Source: Authors' calculations based on household survey data sets listed in appendix 1. 14 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union relationship between disability and employment, ment and the reduced earnings capacity. One although for males, simple OLS estimates sug- difficulty in interpreting the linkages between gest that employees who report poor health sta- poverty and disability is that the former is a tus earn 13 percent less than others, after taking household-level indicator of living conditions, into account individual, household, and commu- while disability is an individual-level event (with nity characteristics that are thought to influence implications for the broader household). For earnings. A one-step worsening in subjective example, if a significant portion of the disabled health status ranking (say from very good to move in with their wealthier children or parents, good) leads to a 14 percent decrease in wages, then the observed disability-poverty relationship while a one-step worsening in the disability may not capture the decline in living standards ranking leads to a 30 percent decrease in wages. for the disabled or the extended family members. For Russia, there is not a robust relationship The discussion of the relationship between between chronic illness and wages, although household socioeconomic characteristics and depression is associated with lower labor force disability is also affected by the stage of life participation and somewhat lower wages for under consideration (figure 9). For children, it females. In fact, the impact of disability on wages is easier to argue that the observed correlations is also larger for females if one considers self- between socioeconomic characteristics and assessed health status or disability status. poor health/disability are causal, since their earnings potential can be ignored in most cases. On the measurement side, higher mortality Poverty and Disability rates of disabled children (or shorter lifespans of the disabled elderly, for that matter) can result Poverty can lead to disability and poor chronic in smaller disability rates in the population over- health conditions through a number of mecha- all. For example, at the time of the 2004 Roma- nisms, including exposure to malnutrition in nia Reproduction Health Survey, the mortality early life, lack of access to adequate health care, rate among children who were reported by their and exposure to unsafe environmental condi- mothers to have had any disability was 28 per- tions either at work or at home. Disability and cent, as opposed to 1.6 percent for the remain- poor health conditions can also lead to poverty, ing children. In industrialized countries, the not only because of the financial implications of disabled children's survival chances are likely to seeking care and securing medication, but also be better, thus contributing to higher disability because of the decreased likelihood of employ- rates ceteris paribus. FIGURE 9 Relationship between Household Wealth and Disability at Various Life Stages Environmental Prenatal period: effects health behavior and outcomes Community Human capital investment period: characteristics e.g., schooling outcomes of disabled children Health service delivery infrastructure e.g., social networks, ability to find employment Household wealth/ Institutions Socioeconomic status Adulthood: e.g., employment e.g., employment effect on Macroeconomic household wealth environment Introduction 15 The poor are more likely to be disabled. The linkage is confirmed, if anything, in a more explicit disability gradient is quite steep in Bulgaria and manner (figures 10 and 11). Bosnia and Herzegovina, the two countries that Finally, the social stigma associated with being also have the highest rates of disability among disabled can be more severe for poor households, the countries that are considered here. The and thus they may be less likely to consider cer- poverty-disability linkage is sensitive to at least tain limitations as disability. This would lead to three sets of factors, one being the way disability an underestimation of the poverty gradient in the is defined. Chapter 2 in this report finds that the prevalence of disability. Also, as highlighted by disability indicator that is most closely associated Amartya Sen's "capabilities approach" to the with reduced per-capita household consumption study of poverty and inequalities, even if a dis- in Uzbekistan is having at least one serious diffi- abled person and a physically fit person have the culty or full limitation in any of the six physical same income and physical goods, the disabled functioning domains (vision, hearing, move- person is likely to live a much more restricted or ment, learning, communication, and self-care)-- difficult life. Thus, if the objective is to measure compared to other indicators of disability, such the extent to which living standards vary between as reporting of chronic illness, official disability households with disabled individuals and those status, and an ADL index. without any disabled individuals, then even the Similarly, the definition of poverty matters. The lack of a poverty-disability relationship would not standard, consumption-based poverty definition mean that the living standards of the two groups produces a strong correlation with disability in two are the same, on average. Nevertheless, these out of five countries considered, displaying a more findings suggest the need to pay special attention subtle relationship for other countries (still in the to the identification and economic and social expected direction). If one uses an asset-based integration of disabled individuals who are poor proxy to measure poverty--which may capture in the context of developing countries with sig- longer-term welfare--then the poverty-disability nificant resource constraints.23 FIGURE 10 Disability Rates by Consumption-Based Poverty Status 12 8 tn cer pe 4 0 poorest 2 3 4 wealthiest Bosnia and Herzegovina Bulgaria Georgia Moldova Romania Source: Authors' calculations based on household survey data sets listed in appendix 1. Note: Survey questions were "Do you have a disability?" in Romania; "Do you get disability allowance?" in Moldova; "Do you have a disabled status 1, 2, or 3?" in Georgia; "Do you have a recognized disability group?" in Bulgaria; and "Do you consider yourself disabled?" in Bosnia and Herzegovina. 16 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 11 Disability Rates by Asset-Based Poverty Status 15 12 tn 9 cer pe 6 3 0 poorest 2 3 4 wealthiest Bosnia and Herzegovina Bulgaria Georgia Moldova Romania Source: Authors' calculations based on household survey data sets listed in appendix 1. Does Employment Protect the Disabled the nondisabled (figure 12). Available evidence from being Poor? indeed suggests that employment protects the disabled from being poor. One exception to this Among employees, the likelihood of being in rule is Romania, where a significant portion of the poorest quintile of the consumption distri- the disabled wage earners are among the poor- bution is about the same for the disabled and est quintile, so being a wage employee does not FIGURE 12 Percentage of Disabled and Nondisabled Individuals in the Poorest Quintile of the Consumption Distribution separately for wage earners and self-employed wage earners self-employed Romania Bulgaria Bosnia and Herzegovina Georgia Moldova 30 20 10 0 10 20 30 disabled not disabled Source: Authors' calculations based on household survey data sets listed in appendix 1. Introduction 17 remove the economic disadvantage of disabled Social Protection Transfers and the individuals.24 Disabled As discussed previously, there is room for signif- Health Shocks, Employment, and Poverty icant improvement in the labor market environ- ment in transition countries if disabled Households are often unable to cope when there individuals are to contribute to, and take advan- is a major deterioration in the health status of tage of, economic growth through gainful the head of the household. Analysis of longitudi- employment. One such improvement would be nal data from Bosnia and Herzegovina (reported the formalization of informal sector employ- in chapter 3) reveals that per-capita household ment, which contributes to the particularly consumption decreases by 7.8 percentage points, dreadful labor market outcomes for the disabled. on average, if a household head becomes dis- The curbing of the informal sector would also abled. Similarly, a worsening of ADL leads to a boost a government's tax base, in turn generating 4.3 percentage point decrease in per-capita much-needed resources for the social protection household consumption. But household con- system, which will have to remain as a key policy sumption is not sensitive to the arrival of a new device to improve the living standards of dis- chronic disease. abled people who are unable to participate in the Furthermore, by making a distinction labor force. But what are the levels of disability between health shocks that happened within the benefits, and to what extent are these benefits last year and health shocks that occurred three targeted to the most vulnerable groups? or four years ago, one can see that employment The share of individuals that qualify for dis- protection legislation for the disabled seems ability benefits varies significantly across coun- effective in preventing an immediate decline in tries, with Croatia, Poland, Hungary, and weekly hours of work, although disabilities that Estonia reporting about twice as many benefici- began two to three years ago are associated with aries as the EU average. Mental and physical a startling 17 hours less work per week. This is impairments tend to be covered by sickness ben- in contrast to the situation in countries with efits paid by health insurance funds before indi- weaker employment protection legislation but viduals qualify for disability benefits. The lower unemployment rates, where a sharp initial SHARE study25 found that the large differences decline in hours of work is followed by a recov- in disability insurance enrollment across coun- ery period. (Presumably after an initial abrupt tries is not necessarily due to differences in adjustment, the disabled individuals eventually demographics and health status, but rather to start working more, though perhaps in a differ- institutional effects that create different enroll- ent field.) In the case of Bosnia and Herzegov- ment incentives. Such effects include easier ina, the existing legislation offers temporary enrollment and eligibility rules, and more gen- relief, but after a one-year period the disabled erous disability benefits in some countries than individuals face an inhospitable labor market in others.26 environment where formal employment oppor- One trend that is worth mentioning in the tunities are scarce and informal employment is Eastern European and Central Asian region is often physically demanding. Overall improve- that while life expectancy at birth is lowest in ments in labor demand would help in easing the poor Central Asian countries (and also Russia longer-term disadvantages of the disabled indi- and Azerbaijan), these are the countries that viduals, and targeted government interventions have the lowest share of persons receiving dis- that provide training and matching for new jobs ability benefits (figure 13). While in theory such can be effective in this context. a trend can occur if some populations live 18 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 13 Persons Receiving Disability Benefits per 100,000, in 2003 10,000 8,000 6,000 4,000 2,000 0 ania Rep. Rep. CIS Rep. CroatiaPoland GermanyHungaryEstoniaNorway rlands TFYR Finland Latvia Lithu Sweden BelarusaverageUkrainemenia Ar Portugal Moldova Belgium Tajikistan Romania Nethe Slovak Czech EU Switzerland Azerbaijan Kyrgyz Uzbekistan Source: WHO: Health for All database. http://www.euro.who.int/hfadb. shorter but healthier lives (in the sense of spend- avoid an excessive burden on public finances.30 ing few years with disability or chronic illness), The discrepancy between the official disability in practice this trend is likely to be driven by the rate and other definitions of disability can make fact that poor countries cannot afford to grant different demographic groups more vulnerable social assistance benefits to a larger share of if their poor health status is not recognized as a their populations. disability that triggers support in the form of In most OECD and ECA countries, the cost social assistance. For example, the gap between of disability benefits as a percentage of GDP has the official disability rate and physical function- increased since 1990. The exceptions are coun- ing limitations is particularly severe for the eld- tries where sick-leave benefits and old-age pen- erly in Uzbekistan: The official disability rate sions serve as an alternative to disability increases modestly by age, remaining at around insurance. In 1999, spending on disability bene- 10 percent among those who are older than 66 fits ranged from 0.2 percent of GDP in Korea, years. In contrast, the share of individuals with to 3.28 percent of GDP in Poland. Spending for at least one full limitation increases from about all disability-related programs surpassed 4 per- 5 percent for the 7 to 16 age group, to 65 per- cent of GDP in Norway, the Netherlands, Swe- cent among those who are 66 and over. den, and Poland.27 In Slovakia, the growing Not only do those who are of working age number of disabled has led to an increase in dis- have an advantage in receiving official disability ability expenditures from 1.6 percent of GDP in status, but males also are more likely to have 1990 to 2.3 percent in 2001,28 which is compa- official disability status after taking into account rable to the EU average. ECA countries spend a physical-functioning limitations and basic indi- similar share of GDP on disability benefits as vidual, household, and community characteris- OECD countries. In 2000, Lithuania spent 1.3 tics. Furthermore, individuals who live in percent of GDP for disability benefits, which is different regions (with otherwise comparable considerably more than Mexico or Korea.29 characteristics) face significantly different prob- Attaining official disability status is what abilities of receiving official disability status, sig- matters for receiving public (though not neces- naling variations in the way disability status is sarily private) transfers. Yet the share of the pop- granted at the local levels. The observed dis- ulation that is officially considered disabled can crepancies deserve attention both because be manipulated through the adoption of strin- females tend to earn significantly less than males gent or flexible eligibility criteria, perhaps to with similar characteristics and female-headed Introduction 19 households are more likely to be poor in many countries, both the coverage of disability bene- countries,31 and also because the disadvantage fits and the targeting performance of benefits for the elderly in being officially recognized as are in need of significant improvements. disabled will become difficult to ignore as the share of elderly increases over time. Disability pensions are well targeted to the Disabled Children's Limited poor in most countries of the region (figure 14). Opportunities to Build Human Capital Two low-income CIS countries, Tajikistan and Georgia, are exceptions to the rule, with almost Disabled children's limited access to public ser- uniform distribution of disability benefits, vices contributes to undesirable employment regardless of household consumption. These and wealth outcomes when they become adults. two countries are among the poorest in the Both demand- and supply-side factors influence region, with $2.15 per day poverty rates of 74 the human capital accumulation of children, percent and 52 percent, respectively. In these including the characteristics of the community; countries, too, a substantial portion of the dis- social norms; physical access to and affordability ability pensions reach the poor, but not the of public services; rationing of secondary or extreme poor. The disability pensions also make higher education opportunities through selec- up a significant share of the consumption in tion, quantity, and quality of teachers; labor lower quintiles of wealth distribution, serving to market conditions and (perceived) returns to improve the consumption ranking of some human capital; access to credit; household char- households that receive these transfers (figure acteristics; and the characteristics of the child 15). However, an alternative way to describe the (Schultz 1961; Becker 1981). Empirical applica- observed trends would be that in poor transition tions of this human capital accumulation model FIGURE 14 Distribution of Disability Pension Beneficiaries by Household Consumption before disability pensions 100 90 80 70 60 tn 50 cer 40 pe 30 20 10 0 Lithuania Poland Serbia & Bosnia Albania Bulgaria Romania Russian Tajikistan Moldova Uzbeki- Georgia Azer- Belarus Monte- and Fed. stan baijan negro Herze- govina EU-8 SEE Middle- Low-income CIS Low-middle income income CIS poorest 2 3 4 wealthiest Source: Authors' calculations based on household survey data sets listed in appendix 1. 20 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 15 Distribution of Disability Pension Beneficiaries by Household Consumption after disability pensions 100 90 80 70 tn 60 cer 50 pe 40 30 20 10 0 Lithuania Poland Serbia & Bosnia Albania Bulgaria Romania Russian Tajikistan Moldova Uzbeki- Georgia Azer- Belarus Monte- and Fed. stan baijan negro Herze- govina EU-8 SEE Middle- Low-income CIS Low-middle income income CIS poorest 2 3 4 wealthiest Source: Authors' calculations based on household survey data sets listed in appendix 1. are often used to explain how the female disad- bers with different characteristics and skills. For vantage in school enrollments emerges in devel- example, there may be cases where parents oping countries and what can be done to reduce invest in nondisabled children's schooling in the the gender gap in school enrollments (Lewis belief that total returns for their investments and Lockheed 2006; Lloyd, Mete, and Sathar will be higher this way (and perhaps altruistic 2005; Schultz 2001; King and Hill 1993). parents can provide better living conditions for The same conceptual framework can be the whole family through redistribution of employed to highlight the challenges involved in funds in the future)--a notion that is formalized providing better living conditions for disabled by Becker (1981). Previous findings that children. For example, in this context, physical demonstrate the significant employment disad- access to schooling refers to both the extent to vantage of disabled adults would serve to rein- which school buildings are designed to take into force the motivation for underinvesting in account the needs of disabled children, and also disabled children's human capital. Having said whether education for disabled children is pro- all this, to what extent are disabled children less vided in an integrated manner: separate educa- likely to enroll in school? tional paths often mean reduced access to The enrollment gap between disabled and schooling for the "vulnerable group," be it the nondisabled children is surprising to few. Nei- disabled children in transition countries or ther the MDGs nor the Education for All Ini- female children in countries where single-sex tiative can succeed in the absence of a renewed education is the norm. commitment to improve disabled children's On the household side, a key factor that schooling outcomes.32 In countries where pri- influences the demand for schooling is how mary school enrollments are already high, resources are shared among household mem- which is the case for most transition countries, Introduction 21 further gains in enrollment cannot be achieved worst decline in health care utilization during without an emphasis on the schooling of dis- the 1990s, and the poor experienced the worst abled children. For example, as of 2002, enroll- drops within these countries, with some recov- ment rates of disabled children between the ages ery after 1999 (World Bank 2005). In rural areas of 7 and 15 were 81 percent in Bulgaria, 58 per- of Romania, for example, poor individuals are cent in Moldova, and 59 percent in Romania, much less likely to have a hospital or a health while the enrollment rates of nondisabled chil- center available in their locality (6.9 percent ver- dren were 96 percent, 97 percent, and 93 per- sus 11.3 percent for the wealthiest quintile).33 cent, respectively. Similarly, figure 16 confirms Access to health care, in turn, has an influence the sizable enrollment gap for disabled children on the health status of individuals--the Roman- between the ages of 16 and 18. These findings ian data sets reveal that this effect is especially do not appear to be sensitive to the definition of visible among the elderly. disability at early levels of schooling. A multi- Similarly, there may be payoffs to distin- variate analysis of the determinants of school guishing between urban and rural areas when it enrollments in Uzbekistan, reported in chapter comes to the public provision of schooling 2, finds that most disability proxies have the opportunities for disabled children.34 In urban expected large effect on school enrollments for areas, the large number of disabled children 7- to 14-year-olds, although for older children, would make it feasible to invest in school infra- official disability status is the only indicator that structure and teachers to provide a better learn- has a large impact on enrollments (at over 40 ing environment to this group of children. In percentage points' drop in the probability of rural areas such economies of scale do not exist school enrollment). and, at least in some transition countries, the The observed relationship between disability current practice is to provide training through and poverty, which was documented previously, teachers' visits to disabled children's houses. arises from the interaction of a variety of fac- The effectiveness of this approach is yet to be tors, not only because of limited access to edu- evaluated, although at the very least, oversight cation, but also because of limited access to and monitoring of such programs may lead to health care. The ex-socialist countries faced the some improvements in learning outcomes. FIGURE 16 Enrollment Rates of 16- to 18-Year-Olds 80 60 tn cer 40 pe 20 0 Moldova Georgia Bulgaria Romania not disabled disabled Source: Authors' calculations based on household survey data sets listed in appendix 1. Note: The sample of disabled children aged 16 to 18 is too small to produce enrollment rate estimates for this group using the Bosnia and Herzegovina data sets. 22 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union In addition to ensuring better access to edu- and Herzegovina, children ages 11 to 15 whose cation and health services, in poor developing parents experience health shocks are 14 per- countries there is room to improve the quality centage points more likely to drop out of school of life and productivity of disabled individuals during the four-year time period between the by relatively straightforward interventions, such Wave 1 and Wave 4 surveys (chapter 3). In this as provision of eyeglasses, hearing aids, or case, the effect is visible only for male children wheelchairs to those in need. In Uzbekistan, 55 though, and it is larger for the children of heads percent of those with vision problems do not of households who become chronically ill wear glasses or contact lenses. The situation is between survey waves, compared to deteriorat- even more serious for hearing problems, since ing ADL (corresponding increases in the likeli- 96 percent of those who report having such hood of dropping out of school are 15 problems do not wear a hearing aid. The poor- percentage points and 9 percentage points, est two quintiles of the consumption distribu- respectively). tion are more likely to fall into these categories, The analysis of comparable time-use survey especially when it comes to vision problems, data from three transition countries (Romania, although the differences among various con- Hungary, and Estonia) and two developed sumption quintiles are not large.35 Another countries (the Netherlands and the United example to support this point comes from Tajik- Kingdom) expands on this framework by docu- istan, where the lack of wheelchairs, crutches, menting the prevalence of home care for the and prosthetics appears to be critical. For exam- disabled, the extent of cross-country variation, ple, one study assesses that in Dushanbe, the and the extent to which certain individuals (for capital city, 200 out of 1,100 registered children example females) end up playing a more signifi- with disabilities require a wheelchair but cannot cant role in this aspect of life. The main features afford one.36 of these time-use data sets are presented in appendix 2. In transition countries, the likelihood of Other Nonmonetary Costs of Disability: spending time assisting adult household mem- Caring for the Disabled bers is not more than it is in the Netherlands and the United Kingdom. But among those who The true costs of disability would be underesti- report providing care to adult household mem- mated if one focused solely on the implications bers, those who live in the two poorest countries of disability on the employment status of indi- with time-use survey data (Romania and Hun- viduals and on household living conditions, as gary) spend much more time on this activity (fig- captured by per-capita household consumption. ure 17).37 One possible explanation for this trend Indeed, disability prevents children from is that because the Netherlands and the United attending school, and furthermore, when a par- Kingdom are further along in the demographic ent becomes disabled, his or her children's transition, provision of "some help" to elderly schooling outcomes often suffer. adults is common.38 Yet the amount of time spent In particular, male children can be sum- for this purpose does not have to be as much as it moned to work and make up lost income due to is in poorer transition economies because the dis- the disability or poor health condition of the abled in wealthier countries are more likely to adult (which can be significant, as discussed ear- benefit from state-of-the-art health care and sup- lier), while female children may be required to portive equipment, which would enable them to provide more time either directly assisting the function more independently.39 disabled household member(s) or helping with In urban Romania, Hungary, and Estonia, other household chores. For example, in Bosnia time-use patterns are closer to those observed in Introduction 23 FIGURE 17 Percentage of Individuals Spending Time Assisting Family Adult, and Time Spent 7 90 6 80 5 70 tn 4 cer 60 pe 3 utes/day)n (mi 50 2 time 1 40 0 30 Romania Hungary Estonia Netherlands United Kingdom percentage time (min/day) Source: Authors' calculations based on household survey data sets listed in appendix 2. the Netherlands and the United Kingdom, and who are similarly educated and experienced, and thus continued urbanization (and further eco- there is a 6 percent return for an additional year nomic development) may lead to a convergence of schooling in both countries (Yemtsov, across countries in time spent on home care. Cnobloch, and Mete 2006). The adult care The percentage of individuals who report assist- responsibilities of females may also explain the ing adult household members is similar in urban relatively low employment rates of females in and rural areas, but rural residents report spend- transition countries when compared to OECD ing much more time in caring for adult house- countries. hold members (about 1.4 times more compared The remaining chapters investigate some of to urban residents in all three transition coun- the topics highlighted here in a more detailed tries considered here). manner. Making use of a household survey Females and less-educated household mem- designed specifically for this purpose, chapter 2 bers undertake a disproportionate share of the focuses on two basic but extremely important adult care responsibilities at home. Interest- research questions: how can one go about meas- ingly, time spent on care does not vary much by uring disability; and how are the alternative dis- household wealth (figure 18). These facts hold ability indicators related to one another and to true in general, although Estonia's case is unique poverty? The remaining three chapters focus on in that graduates with a basic education are less uncovering the causal impact of poor health sta- likely to report taking care of the disabled, tus and disability on employment, earnings, and although those who do carry out this activity poverty. To achieve this goal, chapter 3 makes spend almost twice the time on it as compared use of a unique four-wave panel survey that col- to graduates with a tertiary education. These lected information on household characteristics findings can be explained by the lower earnings (including household consumption patterns) potential of females and less-educated individu- and individuals' health and disability status. als: For example, in Romania and Hungary, This analysis links emergence of disabilities females earn about 20 percent less than males (and variations in the ADL index) to changes in 24 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 18 Who Takes Care of Disabled Household Members? ratios likelihood of providing care to time spent for this activity disabled household members (conditional on providing care) United Kingdom Netherlands Estonia Hungary Romania 4 3 2 1 0 1 2 3 basic/tertiary education female/male poorest/wealthiest quintile Source: Authors' calculations based on household survey data sets listed in appendix 2. employment, earnings, and household con- also be useful to compare the trends docu- sumption. The econometric strategy adopted in mented here for transition countries with those chapters 4 and 5 relies on the modeling of health that prevail in other developing countries with status and labor outcomes jointly in an instru- similar GDP per capita (but with different mental variables setup to display the impact of demographic compositions), although such health status on labor productivity. Perhaps the work does not seem feasible at this time due to most striking finding of this work is the robust data limitations. Appendix 3 lists the alternative and large economic effect of poor health and definitions of disability used by each chapter of disability in the region, compared to what is the report, to evaluate the robustness of the key observed in industrialized countries. It would findings. Introduction Appendix I Disability-Related Questions in Available Household Survey Data Albania Armenia Armenia Azerbaijan Belarus BiH Bulgaria Croatia Georgia Hungary Kazakhstan Kyrgyz Rep. Lithuania Type of survey LSMS HBS DHS HBS LSMS LSMS HBS HBS HBS HBS HBS HBS LSMS HBS HBS Year of survey 2002 2001­03 2000 2001­02 1998­2002 2001­03 1995 & 2001 2003 1998 1998­2002 1998­2002 2001­03 1998 2000­02 2000 1. Direct question on disability 1. Do you have a disability? Y Y Y 2. Do you need any kind of assistance in daily activities? 3. Type of disability? Y 4. Level of disability (I/II/III)? Y Y Y 5. Since when do you have it? Y Y 6. How much do you spend on it? Y 2. Self-evaluated health status 1. Qualitative question on self-evaluation of health status Y Y Y Y Y Y 2. Respondent identifies socio- economic status as "disabled" 3. Chronic illness 1. Do you have chronic illness/disability? Y Y Y Y Y Y Y 2. Type of chronic illness Y Y Y Y Y Y 3. Does it affect your daily activities? Y Y Y Y 4. How long have you had it? Y Y 5. How much do you spend on it? Y Y Y Y Y 6. Days lost due to disability in last month? Y Y Y Y Y 4. ADL? Y Y 5. IADL? 6. Info on disability transfers 1. Info on total disability transfers Y Y Y Y Y Y Y Y Y 2. Info on transfers by type of disability Y Y 25 26 Albania Armenia Armenia Azerbaijan Belarus BiH Bulgaria Croatia Georgia Hungary Kazakhstan Kyrgyz Rep. Lithuania Type of survey LSMS HBS DHS HBS LSMS LSMS HBS HBS HBS HBS HBS HBS LSMS HBS HBS Year of survey 2002 2001­03 2000 2001­02 1998­2002 2001­03 1995 & 2001 2003 1998 1998­2002 1998­2002 2001­03 1998 2000­02 2000 Economic 7. Work-related disability 1. Work part time because of disability/sickness? 2. How many hours worked? Implications 3. Did not work at all because of disability/sickness? Y Y Y Y Y Y Y Y Y Y 4. When did person stop working? Y 8. Wealth Proxy of 1. Consumption aggregate Chronic 1.1 Already constructed Y Y Y Y Y Y Y Y Y 1.2 Needs to be constructed but Illness data available Y Y Y Y 9. Panel exists? N N N Y Y N N Y N Y Y and Moldova Poland Romania Russian Fed. Serbia Tajikistan Turkey Turkmenistan Ukraine Uzbekistan Disability Type of survey HBS HBS LFS HBS HBS LSMS NOBUS LSMS LSMS HBS Disability survey LSMS DHS HBS HBS Year of survey 1998­2003 1998­2002 1998­2002 1998­2002 1993­2003 2003 2002­03 1999 & 2003 2002 2002 1998 2000 1999­2003 2001 1. Direct question on disability in Eastern 1. Do you have a disability? Y Y Y Y Y 2. Do you need any kind of assistance in daily activities? Europe 3. Type of disability? Y Y Y Y 4. Level of disability (I/II/III)? Y and 5. Since when do you have it? Y Y Y Y 6. How much do you spend on it? the 2. Self-evaluated health status Former 1. Qualitative question on self-evaluation of health status Y Y Y Y 2. Respondent identifies socio- Soviet economic status as "disabled" Y 3. Chronic illness Union 1. Do you have chronic illness/disability? Y Y Y Y Y Y Y Y Y Introduction Moldova Poland Romania Russian Fed. Serbia Tajikistan Turkey Turkmenistan Ukraine Uzbekistan Type of survey HBS HBS LFS HBS HBS LSMS NOBUS LSMS LSMS HBS Disability survey LSMS DHS HBS HBS Year of survey 1998­2003 1998­2002 1998­2002 1998­2002 1993­2003 2003 2002­03 1999 & 2003 2002 2002 1998 2000 1999­2003 2001 2. Type of chronic illness Y Y Y Y Y 3. Does it affect your daily activities? Y Y Y Y 4. How long have you had it? Y 5. How much do you spend on it? Y Y Y Y 6. Days lost due to disability in last month? Y Y 4. ADL? Y Y 5. IADL? 6. Info on disability transfers 1. Info on total disability transfers Y Y Y Y Y Y Y Y Y 2. Info on transfers by type of disability Y 7. Work-related disability 1. Work part time because of disability/sickness? Y 2. How many hours worked? Y 3. Did not work at all because of disability/sickness? Y Y Y Y Y 4. When did person stop working? Y 8. Wealth Proxy 1. Consumption aggregate 1.1 Already constructed Y Y Y Y Y Y Y Y Y Y Y Y 1.2 Needs to be constructed but data available Y Y 9. Panel exists? Y Y Y Y Y N N N N 27 28 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Appendix II Basic Information onTime-Use Surveys Estonia Hungary Netherlands Romania United Kingdom April 1999­ September 1999­ 2000, August­ June­ Survey period April 2000 September 2000 weeks 31­45 September 2000 September 2001 Number of visits 1 4, once every 3 months 3 1 1 Survey instruments Household questionnaire Y Y Y Y Individual questionnaire Y Y Y Y Y EU individual questionnaire Y Self-completed diaries Y Y Y Y Y Weeklong worksheets Y Y Y Agricultural diary Y Subjective well-being questionnaire Y Final questionnaire Y Coverage 10+ y.o. individuals 15­84 y.o. one 12+ ind. per hh 10+ y.o. individuals 8+ y.o. Sample Households 2,581 1,813 7,607 6,414 Individuals 6,234 10,105 1,813 16,949 11,664 Diaries 5,724 43,172 16,285 indiv. * 2 days 19,898 Income variable available Incorig (net monthly hh income) ln categories Prd 2 and 4, categ. ln categories Income (total hh income) Prd 2 and 4, categ. Y Household assets PC Y Y Y Y Y Videocamera Y Y Y Y Videorecorder Y Y Y Y Y TV color color all color and black/white color Car Y Y Y car/light van car/light van Motorcycle Y Y Y Y Y Freezer Y Y Y Y Y Dishwasher Y Y Y n.a. Y Microwave Y Y Y Y Y Washing machine automatic and other ordinary and automatic all automatic and other Y CD player Y Y Y Refrigerator Y Y Bicycle Y Y Introduction 29 Appendix III Proxies for Disability and Chronic Conditions Used in the Remaining Chapters Chapter Proxies used to capture disability and chronic conditions Measurement of Disability and Linkages with Welfare, Official disability classification, self-reported disability status, functional Employment, and Schooling disability assessment, ADL index, IADL index, self-reported chronic illness, self-reported health status The Impact of Health Shocks on Employment, Earnings, and Official disability classification, self-reported disability status, ADL index, Household Consumption in Bosnia and Herzegovina self-reported chronic illness, self-reported health status Health Disabilities and Labor Productivity in the Russian Federation Official disability classification, self-reported chronic illness, self-reported health status The Implications of Poor Health Status on Employment in Romania Self-reported chronic illness, self-reported health status 30 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Notes 9. OECD: Transforming Disability into Ability. 2003. Paris. 1. In 2002, less than 55 percent of working-age 10. One shortcoming of household survey data is the adults in Bulgaria, Georgia, Hungary, Moldova, exclusion of the institutionalized population, Poland, Romania, Russia, and Tajikistan were which can lead to an underestimation of the link- involved in wage employment. The only coun- ages between being disabled and not participat- try in the region where wage employment was ing in the labor force. Even though the numbers relatively high (at 70 percent) was Belarus of institutionalized disabled are not large enough (Yemtsov, Cnobloch, and Mete 2006). to have a major influence on the key findings, the 2. Ideally one should go further than simple corre- analyses focusing on the elderly may be more lations and aim to single out causal effects, since sensitive to the exclusion of the institutionalized the design of key policy interventions requires population. For example, in Hungary, 44,000 this type of information--examples include pro- elderly were institutionalized in 2001--out of ductivity loss due to disability (with implications roughly 10 million individuals, of which 15 per- for returns to investments on prevention) and cent are 65 years or older (Daroczi 2005). In the poverty alleviation impact of disability insur- Uzbekistan, 34 boarding facilities for the dis- ance schemes. abled serve around 8,900 individuals: about 1,500 3. For example, the extent to which teachers are general-disabled adults; 5,600 adults with mental trained adequately to teach disabled children in disorders; and 1,800 disabled children (Japan an integrated learning environment. International Cooperation Agency (JICA) and 4. The report documents the earnings gap between Tahlil 2004). Available evidence suggests the disabled and nondisabled individuals, but it does exclusion of institutionalized children in house- not attempt to distinguish between the lower hold survey data can be ignored for most pur- productivity of disabled individuals versus dis- poses: the 2004 Romania Reproductive Health crimination against them. Survey data, for example, inquired about the 5. The report's findings provide some of the infor- whereabouts of children who do not live with mation needed for such work, but they need to their mothers and out of 5,275 children, only be supplemented with costs associated with spe- four were reported to live in special institutions cific interventions to be useful in a standard cost- for disabled children (345 children did not reside benefit framework. with their mothers for various reasons). 6. The implications of aging for public finances is a 11. Me and Mbogoni (2001) discuss disability-data major concern for industrialized countries, availability in developing countries and report a including questions like whether the health sta- disability prevalence rate of 3 percent. The tus of the elderly is improving sufficiently com- authors find that generic questions show higher pared to life expectancy gains. Jacobzone, prevalence rates compared to questions based on Cambois, and Robine (2001) argue that this is a checklist, and they argue that social stigma the case for OECD countries. The findings may attached to being disabled, as well as interviewer not necessarily apply to ECA countries, how- bias, might be contributing to the underreport- ever. Many OECD countries increased their ing of disabilities. Similarly, for 11 developing public health spending over the last decade and countries Filmer (2005) reports disability preva- individual health behavior improved for the bet- lence rates of less than 2 percent and for India, ter during the same period. Neither is necessar- the World Bank (2006) reports a 1.8 percent dis- ily the case for many ECA countries. ability rate based on census data, but argues that 7. Occupational structure may have an impact on the the actual figure is likely to be in the neighbor- prevalence of disabilities: For example Costa hood of 4 to 8 percent. International Labour (2000) finds that 7 percent of the decline in func- Organization (ILO) country studies for tional disabilities in the United States from the "Employment of People with Disabilities" cover early 20th century to the early 1990s is due to Ethiopia, Kenya, Mauritius, Sudan, Tanzania, shifts away from manual labor to white-collar jobs. Uganda, Cambodia, Fiji, India, Sri Lanka, and 8. World Health Organization. "World Report on Thailand, listing census question­based disabil- Disability and Rehabilitation." Concept paper. ity rates from 0.7 percent in Kenya to 3.36 per- http://www.who.int/disabilities/publications/dar cent in Mauritius. These studies caution the _ world_report_concept_note.pdf (accessed July reader about the likely underestimation of dis- 24, 2006). ability rates based on standard census questions. Introduction 31 12. See, for example, Turkey Disability Survey arising from the threat of lawsuits. On a related Report (2002) and chapter 2 of this report. topic, the insider-outsider dilemma is noted for 13. Mainly depressive disorders and alcohol abuse. other industrialized countries: A review of 14. World Health Organization. 2002. "Global Bur- employment trends in OECD countries finds den of Diseases." http://www.who.int/health that even though legislative approaches to info/bodgbd2002revised/en/index.html. employment promotion differ in many respects 15. U.S. Agency for International Development. (rights based, obligations based, incentives 2006. "Non-Communicable Diseases and based), all approaches tend to benefit people Injuries in Eastern Europe and Eurasia." already in employment much more than those 16. World Health Organization. 2002. "Global Bur- who are out of work and looking for a job den of Diseases." http://www.who.int/health (OECD 2003). info/bodgbd2002revised/en/index.html. 22. World Bank (2006) also observes a divergence in 17. The 2002 Turkey Disability Survey interviewed the employment rates of the disabled and the 97,433 households. About 2.6 percent of the nondisabled in India, noting that the employ- interviewed individuals were reported to be ment gap between the disabled and the nondis- orthopedically, or mentally disabled, or have abled has increased during the last 15 years. disabilities in seeing, hearing, or speaking, and 23. Alternative views have been expressed for indus- 9.7 percent were reported to have at least one trialized countries. For example, one of the pro- chronic illness. There is a sharp increase in "tra- posed reforms for disability programs by OECD ditional" disabilities after age 60, with males (2003) is to recognize the status of disability being more likely to report disability compared independent of the work and income situation of to females. On the other hand, the reporting of individuals. chronic illnesses is similar to those available for 24. Furthermore, such simple correlations do not many other countries in the sense that females rule out the possibility that disabled individuals are more likely to report chronic illness, and the who are members of nonpoor households to gender gap in reporting chronic illness becomes start with are able to find employment. especially pronounced after age 30. Unfortu- 25. Survey of Health, Aging, and Retirement in nately, this survey contains little information on Europe (SHARE). April 2005. www.mea.uni- household characteristics and the nondisabled mannheim.de. population. As a result, one cannot investigate 26. Where health status requirements are weak for the linkages between household characteristics disability benefit eligibility, disability insurance (including poverty) and disability using these may also serve as an alternative to exit the labor data. market into early retirement. 18. Examples include significant medicine shortages 27. Organisation for Economic Co-operation and in Romania in 2003; and not being able to afford Development. 2003. "Transforming Disability basic equipment such as wheelchairs, crutches, into Ability." and prosthetics in Tajikistan (JICA 2002). 28. International Labour Organization. 2005. 19. For example, Kahn (1998) finds that the diabet- "Social Protection Expenditure and Perfor- ics significantly increased their labor force par- mance Review. Slovak Republic." http://www.ilo ticipation in the U.S. between 1976 and 1992. .org/public/english/protection/secsoc/downloads/ 20. The correlation coefficient is -0.03. The coun- publ/sp_expenditure_slovakie.pdf. tries with relevant data are Bosnia and Herze- 29. United Nations Development Programme. govina, Bulgaria, Georgia, Moldova, Poland, 2001. "Progress for All. Lithuania Country and Romania. Assessment." http://www.undp.lt/files//Progress 21. Disability discrimination laws can have unin- For%20All.pdf. tended consequences: for example, legislation 30. Alternatively, the generosity of the disability that requires employers to accommodate dis- benefits can be reduced while keeping the share abled workers and outlaw discrimination against of disabled constant. the disabled in hiring, firing, and pay in the 31. World Bank (2006). United States is argued to have a negative influ- 32. Analyzing data from 11 developing countries ence on the labor force participation and earn- (with one transition country, Romania), Filmer ings of the disabled individuals (Acemoglu and (2005) also notes that in some countries the Angrist 2001; Beegle and Stock 2003) due to school attendance of disabled children is accommodation costs and firing and hiring costs extremely low as compared to the school atten- 32 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union dance of nondisabled children, and that when 2.5 hours. Such a high level of time spent for other individual and household characteristics care is consistent with the trend documented are taken into account, the gap between disabled here--namely that the amount of time spent for and nondisabled children widens in most cases. the care of disabled is disproportionately related 33. Exposure to humidity and cold weather is also to the wealth of a country. reported more often by the poor, as discussed by 38. In OECD countries, informal care is estimated chapter 5 in this report. to account for up to 80 percent of total care 34. There is a wealth of information on disabled (Jacobzone, Cambois, and Robine 2001). children's special needs, which include perspec- 39. Analyzing trends in old-age disability in the tives on the training of teachers to maintain United States, Freedman et al. (2004) document high-quality education standards for all. Such consistent declines on the order of 1 to 2.5 per- sectoral perspectives are beyond the scope of this cent per year for "difficulty with daily activities" particular report, however. and "help with daily activities." The authors also 35. Data source is the 2005 Uzbekistan Regional observe that the proportion of older persons Disability Survey. who receive help with bathing has declined at 36. Japan International Cooperation Agency Plan- the same time as the proportion who use only ning and Evaluation Department. March 2002. equipment (but not personal care) to bathe has "Country Profile on Disability, Republic of increased. The OECD (1998) observes a similar Tajikistan." decline in the disability rates of those between 37. Analyzing the living conditions of the disabled the ages of 65 to 80 years, and notes that the eco- in India, the World Bank (2006) finds that nomic impact of these trends depends on the around 45 percent of households with a disabled institutional arrangements for long-term care member report an adult missing work to care for and the public costs of formal home care, which the disabled household member, on average for differ widely among countries. PART II COUNTRY STUDIES 33 CHAPTER 2 Measurement of Disability and Linkages with Welfare, Employment, and Schooling The Case of Uzbekistan Kinnon Scott and Cem Mete* Introduction reach all of the MDGs; the Education for All initiative1 will, ultimately, be unsuccessful; and In both developing and developed countries, our ability to reduce poverty impeded. How- disability is believed to be a correlate of poverty, ever, the limited research available on disability with implications for individuals' access to ser- in the context of developing countries prevents vices, educational attainment, labor force par- a full understanding of the effect of disability on ticipation, and consumption. Indeed, a 1999 any of these initiatives. review of available evidence describes the Capturing the incidence of disability is diffi- poverty-disability linkage as a two-way relation- cult. In part this is due to the heterogeneity of ship: disability can lead to greater poverty, but definitions and the concepts that make up dis- poverty itself makes individuals more at risk of ability, ranging from the strictly medical to the becoming disabled (Elwan 1999). The study broader concepts of functioning (WHO 1980, also states, however, that there is little system- 2001). And, in part, this is due to the lack of rel- atic research on these issues in developing coun- evant and reliable data on disability. Administra- tries--largely due to the lack of data on tive records are often the only source of data, but populations with disabilities--and that the liter- these can contain serious biases and do not pro- ature relies heavily on anecdotal evidence and vide the additional information needed to study case studies. It is argued that without addressing the linkages between disability and various the disability issue, it will not be possible to aspects of welfare. Traditionally, household sur- * Kinnon Scott is a senior economist at the Development Research Group of the World Bank. Cem Mete is a senior economist at the Europe and Central Asia region of the World Bank. 35 36 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union veys have not been considered to be good ity. And these data do not provide additional sources of data on disability, not simply because information on the characteristics of those indi- the questions included have been inadequate, viduals and households affected by disability, but also because disability appeared to be a rela- which is needed to determine the connections tively rare event, not amenable to being captured that might exist between poverty and disability. in the small sample sizes of many household sur- Household-level data exist in too few cases. veys. Data from 35 developing countries in the To shed light on the poverty-disability nexus, early 1990s from censuses or surveys that as well as the measurement issues, disability attempted to measure disability show only two questions were added to one round of a panel developing countries with incidence rates above survey in Uzbekistan that also measures welfare. 5 percent, and the vast majority with rates below The data allow us to distinguish among the 2 percent (United Nations 2005).2 The WHO, effects of various measurement issues on the however, estimates that underreporting of dis- incidence estimates obtained, as well as to link ability is high and that the rate of disability in the incidence of disability with monetary meas- most developing countries is around 10 percent ures of welfare, human capital indicators, and (WHO 1980). Recent efforts to improve the way participation in the labor force. The paper is in which questions on disability are asked have organized in the following way. Section I con- shown that the rates are higher than previously tains discussion of issues surrounding the meas- imagined. Uganda substantially revised its 1991 urement of disability, the Uzbekistan context, disability question for the 2001 census: The inci- and the data set. Section II presents findings on dence rate went from around 5 percent among the relationships among the various disability the oldest age group to more than 20 percent measures. In Section III, a first assessment of (Martinho and Banda 2005). A 2005 Nicaraguan the relationship of disability and a variety of survey specifically developed to measure disabil- welfare and participation measures is carried ity gives an incidence rate of 11 percent (INEC out. The paper ends with some conclusions and 2003), while a comparable survey in Ecuador recommendations for future work on investi- and the census in Brazil give rates of 12.1 per- gating poverty-disability links. cent and 14.5 percent, respectively (Flores et al. 2005 and Bercovich 2004). Similarly, a 2005 Uzbekistan survey reveals the sizable variations I Measuring Disability between the official disability rate and survey- based disability indicators, which are described Measurement Issues in detail in the next section (figure 1). In partic- In an attempt to address the measurement issues ular, figure 1 shows that compared to other dis- associated with disability, the United Nations ability indicators, the official disability rate does formed the Washington Group on Disability not vary much by age. Statistics in 1999. The Washington Group has While disability itself is difficult to measure, been charged with developing a short set of it is even harder to capture the link between questions for inclusion in national censuses and welfare and disability. To do so requires not just an extended set of questions for inclusion in accurate data on the incidence of disabilities, but national household surveys that will provide also data on a range of key characteristics of robust and cross-country comparable data on individuals and their households. In many the incidence of disability. For both sets of ques- developing countries, the only data on disabili- tions, the Washington Group is focusing on ties come from administrative records, which three purposes for the data collection: (i) equal- can be affected by benefit levels and other pro- ization of opportunities, (ii) monitoring of gram characteristics unrelated to actual disabil- trends in physical functioning or participation Measurement of Disability and Linkages with Welfare, Employment, and Schooling 37 FIGURE 1 Incidence of Disability Using Various Definitions in Uzbekistan (by age group) Panel A: Incidence 100 90 80 70 tn 60 cer 50 pe 40 30 20 10 0 7­16 17­26 27­36 37­46 47­56 57­66 66+ age groups one full limit one serious diff. or limit own disability assess. own func. disability assess. chronic official disability Panel B: Mean scores (normalized) for physical functioning and activities of daily living, by age group 1 0.9 0.8 er 0.7 sco 0.6 n 0.5 mea 0.4 0.3 0.2 0.1 0 7­16 17­26 27­36 37­46 47­56 57­66 66+ age groups norm. domain score norm. ADL score Sources: Uzbekistan Regional Panel Survey, Waves 1, 2, and 3. Authors' calculations. of the population, and (iii) provision of services the household surveys, it is proposed to develop (Altman, Cambois, and Robine 2005).3 For the four "extended sets" of questions, one for each censuses, a set of six questions has been devel- of the purposes listed above, and one to address oped and cognitive testing and field testing is the environmental context. Testing of the under way in approximately 15 countries. For extended sets will also be carried out. 38 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Following the lead of the Washington ous indicators are listed in table 1; the actual Group, our focus in this work is on functioning, questions are listed in annex 1. not "disability," per se. What this means in real terms is that it is perfectly possible for two indi- Official Disability Status viduals with the exact same physical conditions Uzbekistan has several laws that cover the defi- to have different levels of functioning or "dis- nition of disability and the rights of those with ability." The simplest example of this relates to disabilities.6 The laws define three types of offi- vision. A person with limited vision could have cial disability categories. Individuals who have limited functioning or "disability." Yet if that lost their ability to work and depend on the care person was given a pair of appropriate eye- of others fall into category 1, while those who glasses or contact lenses, he or she might have have lost their ability to work but do not depend no limitation in physical functioning.4 on others to care for them are classified as cate- The questions used in the Uzbekistan work gory 2. The third category is for those who have are based on a draft household module from the partially lost their ability to work. The benefits Washington Group for monitoring the trends that accrue vary by category and include disabil- in physical functioning or participation of the ity pensions (of varying amounts), as well as a population.5 The questions refer to six areas of series of other benefits. functioning: vision, hearing, movement, learn- In 2003, 817,000 persons were officially con- ing, communication, and self-care. For hearing, sidered disabled in the country, or 3.2 percent of vision, and learning, a filter question was the population. This represented a substantial applied, so the incidence is measured with only increase from the 1996 rate of 2.4 percent (State one question. Persons who indicated some level Department of Statistics as cited in Japan Inter- of disability were, however, asked follow-up national Cooperation Agency [JICA] and Tahlil questions. For movement, communication, and 2004). A variety of reasons can be posited to self-care, filter questions--while asked--were not actually used as filter questions: all individ- TABLE 1 uals were asked two to four questions for each Indicators of Physical Functioning domain. However, for comparability across the Indicator six domains, only the filter question is used to determine incidence. Specific physical functioning Seeing (with glasses if uses them) In addition to the specific physical function- Hearing (with hearing aid if uses one) ing questions from Wave 3, in the previous Movement (walking up stairs, walking a kilometer) waves of the survey, individuals were asked if Learning (memorizing, focusing attention) they had "any sign of disability." This very open Self-care (dressing, washing, feeding, being alone) question allowed respondents to list any physical Communicating (understanding, being understood) or cognitive problems they had that were linked Physical functioning index (normalized) Have official disability status to disability. An additional indicator was con- Self-describe as having some level of disability structed from this question by using this self- Self-describe as having physical functioning limitation reported health data to assess whether the Have chronic illness individual has a limitation of some sort in a Activity of Daily Living domain of physical functioning. Individuals were Vigorous activity (lifting heavy objects, running) Moderate activity (moving a chair, carrying groceries) also asked about chronic ailments and whether Walking uphill they were officially considered disabled. Finally, Walking 100 meters individuals were asked six questions on Activities Bending, lifting, or stooping of Daily Living (ADL), and a normalized score Self-care (eating, dressing, bathing, using the toilet) ranging from 0 to 1 was constructed. The vari- ADL Index (normalized) Measurement of Disability and Linkages with Welfare, Employment, and Schooling 39 explain this dramatic increase, separate from a designed to allow measurement of welfare levels real rise in impaired physical functioning. First, and assess the factors that affect welfare. It is car- part of the rise may be attributable to the fact ried out in three regions (oblasts) of Uzbekistan: that disease, rather than physical functioning, Tashkent, Andijan, and Kashkadarya. As a panel plays such a large role in the definition of dis- survey, the URPS follows individuals over time ability in Uzbekistan. Tuberculosis, which is a to look at changes in welfare status, economic cause for disability status in Uzbekistan, has activities, and coping mechanisms of the individ- been on the rise in recent years as the health sys- uals and their households. The first wave of the tem has not been able to maintain itself in the survey was carried out in March 2005 (pre-plant- face of declining income and revenues. ing) and the second in October­November of Increased numbers of persons who have the same year (post-harvest). A third wave was been granted official disability status may also conducted in December 2005. be linked to incentives to obtain the benefit. It The specific purpose of the third wave of the is not clear if this is the case in Uzbekistan. URPS was to shed light on issues of disability While there is some evidence that disability measurement in Uzbekistan. The data make it pensions were better protected than other pen- possible to quantify and characterize the levels sions and social assistance programs in the sec- of disability in the three oblasts. The data also ond half of the 1990s (JICA and Tahlil 2004), provide an opportunity to investigate the link- from 2001 to 2005, monthly disability pensions ages between disability and welfare levels, labor as a percentage of average monthly wages force participation, and educational attainment. decreased from 60 percent to 40 percent The sample of the URPS-III was doubled (com- (World Bank 2006). pared to the first two waves) because, even An alternative hypothesis explaining the though the WHO estimates 10 percent of the increase in disability in recent years is that the population is disabled, it was thought that an population (or denominator of the ratio) is not over-sample would be useful in case disability measured with any precision. The last census turned out to be such a rare event that the was conducted prior to independence, and, even URPS I and II sample sizes would not be large in nontransition countries, population projec- enough. The over-sample only creates a cross- tions this far out from the original measurement sectional data set: for anything beyond a very are very noisy. In a transition economy with minimal description of the linkages between post-independence migration flows, there is disability and other welfare characteristics, the certainly the likelihood that any proportion that panel data is needed. has the national population in the denominator The panel was constructed by following indi- diverges from the true proportion. viduals, not just households. Wave 1 of the In short, various factors affect the official URPS was administered to 3,000 households, measure of disability--factors that are not equally divided among the three oblasts. If, at the always strictly correlated with objective meas- time of the Wave 2 interview, a person had left ures of limitations of physical functioning. the household he or she was in at the time of Household-level data may help to confirm or Wave 1, that person was followed to his or her contradict the official statistics and lead to a new residence and interviewed, along with all greater understanding of the links between dis- the other members in this new household. Indi- ability and other factors. viduals who emigrated or who moved to remote regions in the northwest of the country could The Data not, however, be followed. The rules were dif- The Uzbekistan Regional Panel Survey (URPS) ferent for Wave 3, given that the data collection is a Living Standards Measurement Study survey was almost immediately after the Wave 2 inter- 40 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union views and the focus was only on collecting dis- care. The questions were recoded so that 0 ability data. For Wave 3, only those individuals indicates no problem at all, while 3 indicates still residing in the same place as in Wave 2 were the person is unable to perform the activity in interviewed for the panel.7 The number of per- question. Second, an index or score was also sons for whom complete data exist for all three constructed that simply sums the score of the rounds (including constructed welfare meas- six components. This score is normalized to ures) is 12,387. This represents 80 percent of range from a value of 0, indicating that the per- the potential pool of respondents. son has no difficulty on any component of func- Given the concern that the sample size could tioning, to a 1, indicating that the person is be too small for some of the analysis, an addi- completely limited across the six domains of tional sample of 3,000 households was created functioning. and interviewed. The full Wave 3 sample of A person with a full limitation in any area of 6,170 households and 32,337 individuals is only physical functioning is someone who cannot do used here to look at overall incidence rates for the activity in question. In other words, cannot the physical functioning variables. It should be see (is blind), cannot communicate at all, cannot noted that while the panel sample is representa- wash him- or herself. Variables are constructed tive of the three included oblasts, neither it nor to identify individuals with a full limitation in the additional Wave 3 sample is representative any of the six areas. A slightly broader definition of the population of the country: results can of disability is captured by including individuals only be extrapolated to the three oblasts under with severe difficulties, rather than complete discussion here. limitations. We also look at the number of Table 2 provides summary statistics of the domains for which a person is fully limited, sampled individuals used in the analysis. The along with the number for which the person has first pair of columns shows the data for persons at least a serious difficulty in performing the ages 7 and older, while the second pair shows activities. Individuals also provided information data only for heads of households. The first col- on whether they had official disability status. umn of each pair refers to the full sample; the Finally, there is the self-reported variable that second to the panel sample. As would be examines whether a person has an illness or ail- expected, there are negligible differences ment linked to a disability, as well as informa- between the two samples. tion on chronic illnesses. Children under the age of 7 are excluded Measures of Physical Functioning, from the analysis. Clearly, babies would appear Disability, and Health quite disabled using the measures included here. The first key question to be answered by this Ideally, one would have liked to ask about small data is the extent to which different measures of children's development with a different set of disability lead to different incidence levels. questions or in relation to "average" or "nor- From the Wave 3 data, several variables that are mal" development. As this could not be done, hypothesized to measure physical functioning we have excluded children under the age of 7 and disability are constructed and investigated from the analysis. The age cutoff is partly driven here. First, for each domain of functioning, the by the fact that the list of physical activities respondent was asked to rate the level of diffi- under the domains of physical functioning culty he or she has, on a scale of 1 (no difficulty) should be within the abilities of the average 7 to 4 (unable or completely limited). The six fil- year old. Second, the data seem to indicate this ter questions are used here that pertain to each pattern, although it might have suggested a year of the domains of functioning: vision, hearing, or two older for the cutoff point. However, movement, learning, communication, and self- because an area of interest for this paper is the Measurement of Disability and Linkages with Welfare, Employment, and Schooling 41 TABLE 2 Characteristics of the Panel and Full Samples (means and standard deviations) Panel Full Sample Variable All Hhld Head All Hhld Head Seven and older Years of age 31.5 51.0 31.5 50.7 (18.6) (14.3) (18.5) (14.6) Female 0.5171 0.2732 0.5183 0.2757 (0.4997) (0.4457) (0.4997) (0.4469) In school 2004­05 0.3121 0.0074 -- -- (0.4634) (0.0859) -- -- In school 2005­06 0.2950 0.0051 0.2994 0.0088 (0.4560) (0.0714) (0.4580) (0.0933) In school either year 0.3336 0.0074 -- -- (0.4715) (0.0859) -- -- Per Capita Consumption, soum 458,166 560,712 -- -- (406,004) (540,525) -- -- Working age Economically active, Wave 1 0.6555 0.8488 -- -- (0.4752) (0.3584) -- -- Economically active, Wave 3 0.6448 0.8154 0.6215 0.7984 (0.4786) (0.3881) (0.4850) (0.4012) Economically active either wave 0.7619 0.9059 -- -- (0.4260) (0.2920) -- -- Employed, Wave 1 0.6209 0.8172 -- -- (0.4852) (0.3866) -- -- Employed, Wave 3 0.5610 0.7453 0.5391 0.7400 (0.4963) (0.4358) (0.4985) (0.4387) Employed either wave 0.8277 0.9326 -- -- (0.3777) (0.2509) -- -- Unemployed, Wave 1 0.0346 0.0316 -- -- (0.1829) (0.1749) -- -- Unemployed, Wave 3 0.0838 0.0701 0.0824 0.0585 (0.2771) (0.2554) (0.2750) (0.2347) Unemployed, either wave 0.1094 0.0925 -- -- (0.3121) (0.2898) -- -- Sources: URPS, Waves 1 and 3; authors' calculations Note: Includes individuals ages 7 and over. Labor force figures only for persons aged 16­55 (female) and 16­60 (male). Economically active refers to ILO's definition for a person being active, which includes persons working or actively looking for work (unemployed). link between functioning and schooling, it was table 3. Communication is the domain with the considered important to include all children of smallest percentage of the population having any primary school age in the sample. type of difficulty (holding a conversation, under- standing and making oneself understood), with only 5.5 percent of individuals affected in some II Incidence of Limited Physical way, and less than a half of a percent unable to Functioning and Health communicate at all. Hearing difficulties are the second least common type of difficulty, with 7.3 Incidence rates of limitations in the various percent of the population reporting some type of domains of physical functioning are presented in hearing problem. Less than one-third of a per- 42 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union cent report total deafness. Movement and learn- quintiles is more complex. For three indica- ing are the most commonly reported areas where tors--vision, movement, and learning--the problems exist: 18.8 percent of the population richest quintile reports substantially more prob- reports some restriction in their ability to walk a lems (22.8 percent of the top quintile suffers kilometer, and 16.3 percent report a learning dif- from vision problems of some degree, while ficulty of some sort. Full limitations in learning only 8.7 percent of those in the lowest quintile affect less than one-third of a percent of the pop- report any problems). Yet, the lowest quintile ulation, however. In contrast, almost 2 percent reports more individuals with full limitations in of the population reports an inability to walk 1 two of these three areas. They also report more kilometer. full limitations in two of the other three areas-- With the exception of communication, men hearing, communication, and self-care. We will have a lower incidence of physical functioning return to this apparent contradiction below. In limitations. Women are slightly more likely to terms of demonstrating a link between poverty have full limitations than men, but the differ- and disability, the simple tabulations provide, at ences are very small in most areas. The contrast best, only weak evidence of such a link.8 One between the lowest and highest consumption must also keep in mind that the overall inci- TABLE 3 Incidence of No and Full Limitations by Domain of Physical Functioning Variable All Female Male Quintile 1 Quintile 5 No problems with: Vision 0.8556 0.8351 0.8774 0.9137 0.7718 (0.3516) (0.3712) (0.3280) (0.2809) (0.4198) Hearing 0.9265 0.9176 0.9359 0.9358 0.9038 (0.2610) (0.2750) (0.2449) (0.2452) (0.2949) Movement (walk) 0.8122 0.7743 0.8525 0.8673 07593 (0.3906) (0.4181) (0.3547) (0.3393) (0.4276) Learning 0.8366 0.8085 0.8666 0.8741 0.7895 (0.3697) (0.3936) (0.3400) (0.3318) (0.4078) Communication 0.9449 0.9448 0.9450 0.9475 0.9514 (0.2282) (0.2283) (0.2281) (0.2230) (0.2150) Self-care (washing) 0.9174 0.9032 0.9326 0.9129 0.9229 (0.2752) (0.2957) (0.2508) (0.2820) (0.2668) Full limitations with: Vision 0.0030 0.0033 0.0028 0.0028 0.0012 (0.0551) (0.0572) (0.0528) (0.0530) (0.0343) Hearing 0.0030 0.0030 0.0031 0.0034 0.0049 (0.0549) (0.0547) (0.0552) (0.0530) (0.0702) Movement (walk) 0.0194 0.0256 0.0129 0.0172 0.0243 (0.1379) (0.1578) (0.1127) (0.1301) (0.1539) Learning 0.0042 0.0040 0.0043 0.0057 0.0034 (0.0644) (0.0632) (0.0658) (0.0751) (0.0583) Communication 0.0045 0.0055 0.0033 0.0064 0.0016 (0.0666) (0.0739) (0.0577) (0.0796) (0.0402) Self-care (washing) 0.0145 0.0175 0.0112 0.0175 0.0065 (0.1194) (0.1311) (0.1054) (0.1311) (0.0804) Sources: URPS Waves 1 and 3; authors' calculations. Note: Movement refers to the ability to walk 1 kilometer, learning is the ability to memorize and focus attention, communication entails understanding and carrying on a conversation and making oneself understood. Includes persons 7 years of age and older. Standard deviations in parentheses. Measurement of Disability and Linkages with Welfare, Employment, and Schooling 43 dence of full limitations is low, and some cau- ted from the sample, the correlation of age with tion is needed in interpreting these results. the domains of functioning drops substan- The measures of functioning by domain are tially--actually becoming insignificantly corre- all positively correlated with one another (table lated to communication and, somewhat 4). A priori, there is no need to expect high lev- surprisingly, negatively correlated with limita- els of correlation: an illness that leads to the tions in the ability to wash oneself.9 However, inability to see might have no effect on the per- per-capita consumption is only negatively cor- son's ability to walk a kilometer, and an accident related with two of the six domains of function- that might limit a person's ability to walk could ing. No information is available on health care well have no impact on his or her ability to com- infrastructure or environmental factors. municate. On the other hand, indicators of var- The official statistics (as reported by individ- ious functioning limitations can be highly uals) on disability are correlated with the vari- correlated if the incidences of these limitations ous measures of physical functioning, but the are driven by common factors such as aging, relationship is not completely straightforward. poverty, poor health care infrastructure, or Of course, there is no a priori cutoff point that unfavorable environmental conditions. exists for the physical functioning variables to The data provide some support for this determine if a person should be classified as hav- explanation: in fact, age alone is a strong driver ing disabilities. Here we compare official status of the observed correlation. If all persons over with several cutoff points or indicators con- age 54 (the retirement age for women) are omit- structed from the data set. TABLE 4 Correlation of Scores for Domains of Physical Functioning Vision Hearing Movement Learning Communication Self-Care Age Seven and older Vision 1.0000 Hearing 0.4020* 1.0000 Movement (walk) 0.4985* 0.4388* 1.0000 Learning 0.4546* 0.4214* 0.5297* 1.0000 Communication 0.2003* 0.3171* 0.3141* 0.3986* 1.0000 Self-care (washing) 0.2757* 0.2785* 0.5032* 0.3914* 0.4111* 1.0000 Age 0.4788* 0.3829* 0.5522* 0.4180* 0.1429* 0.2238* 1.0000 Consumption, pc 0.0912* 0.0428* 0.0670* 0.0625* 0.0111 0.0239 0.1127* Age (7­55) 0.3147* 0.1405* 0.3035* 0.2175* 0.0017 0.0810* 1.0000 Heads of Household Vision 1.0000 Hearing 0.3842* 1.0000 Movement (walk) 0.4781* 0.4263* 1.0000 Learning 0.4857* 0.4377* 0.5224* 1.0000 Communication 0.2925* 0.3556* 0.3360* 0.3530* 1.0000 Self-care (washing) 0.3326* 0.3677* 0.5716* 0.4068* 0.4080* 1.0000 Age 0.4284* 0.4181* 0.5541* 0.4205* 0.2374* 0.4328* 1.0000 Consumption, pc 0.0426 0.0245 0.0185 0.0380 0.0344 0.0498 0.0568* Age (7­55) 0.2382* 0.1053* 0.2635* 0.1450* 0.0744* 0.1757* 1.0000 Sources: URPS Waves1 and 3; authors' calculations. Note: Movement refers to the ability to walk 1 kilometer, learning is the ability to memorize and focus attention, communication entails understanding and carrying on a conversation and making oneself understood. Includes persons 7 years of age and older. Age (7­55) shows the correlation of the age variable to the six domains of functioning if the sample is restricted to persons 7 through 55. * Significant at the .01 level. 44 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Of the population age 7 and older, 3.8 per- of women do. This is despite the fact that 3.7 cent claim to have formal, official disability sta- percent of women report having a full limita- tus (and are eligible to receive or actually receive tion, compared to only 2.7 percent of men. It is benefits). This is very close to the official rate of not clear what this finding reflects. It may be 3.2 in 2003 if the upward trend noted in official that the limitations that men suffer are more statistics continued to 2005. And it is close to related to their ability to work and, because the the percentage of the panel sample that has at ability to work is an important part of the offi- least one full limitation in any of the six areas of cial disability definition, this may result in the functioning: 3.2 percent. However, almost 12 greater percentage of men receiving official dis- percent of all individuals age 7 and older have at ability status. On the other hand, the finding least one serious difficulty or a full limitation. may reflect a greater ability of men to obtain For all individuals over 7 years of age, just under disability status or there being a greater benefit a quarter of all those with official disability sta- to men derived from obtaining disability status. tus indicate having a full limitation in one or The differences between those in the bottom more domains of physical functioning. Clearly and top quintiles of the consumption distribu- other factors affect official disability status. This tion provide an inconsistent picture of poverty is investigated more fully below. and disability linkages, although welfare status There is some disparity between official dis- does seem to be linked to official disability sta- ability status and physical functioning between tus. Of those in the bottom quintile, only 3.2 males and females. As shown in tables 3 and 5, percent have official disability status, compared females are less likely to have no problems at all, to 5.1 percent in the top quintile. Yet the poorer and more likely to have full limitations and seri- quintile has a similar or slightly higher inci- ous difficulties than men. Yet 4.5 percent of dence of full limitations than the top quintile. males who are age 7 and older report having The population in the top quintile, however, official disability status, while only 3.3 percent appears to suffer more from partial limitations. TABLE 5 Disability and Physical Functioning (means and standard deviations) Variable All Female Male Quintile 1 Quintile 5 Official disability 0.0383 0.0326 0.0445 0.0321 0.0507 (0.1920) (0.1776) (0.2061) (0.1763) (0.2194) No full limitations 0.6811 0.6476 0.7167 0.7367 0.5921 (0.4661) (0.4778) (0.4507) (0.4405) (0.4916) One full limitation 0.0323 0.0370 0.0274 0.0348 0.0305 (0.1769) (0.1887) (0.1633) (0.1834) (0.1720) One serious difficulty or 0.1177 0.1381 0.0959 0.1069 0.1294 full limitation (0.3222) (0.3450) (0.2945) (0.3091) (0.3358) Number of full limitations 0.0486 0.0588 0.0376 0.0530 0.0419 (0.3066) (0.3461) (0.2576) (0.3156) (0.2659) Number of serious difficulties 0.2053 0.2399 0.1684 0.1898 0.2235 or full limitations (0.6858) (0.7357) (0.6260) (0.6644) (0.7089) Normalized domain score 0.0526 0.0609 0.0437 0.0433 0.0657 (0.1095) (0.1177) (0.0992) (0.1048) (0.1141) Sources: URPS Waves 1 and 3; authors' calculations. Note: Normalized domain score ranges from 0 (no difficulty in any domain) to 1 (`unable' in all six domains). Includes persons age 7 and older only. Measurement of Disability and Linkages with Welfare, Employment, and Schooling 45 One factor that could be confounding the measures. The other measures--chronic illness, analysis is biases found in self-reported health the two self-assessed disability measures, and data. There is evidence that individuals' self- the ADL measures--all show much higher inci- ranking of their health status is affected by their dence of disability or physical limitation than own socioeconomic characteristics (Lindeboom the official disability status. Interestingly, these and van Doorslaer 2004; Van Doorslaer and other measures are all higher among the richest Gerdthan 2003; Schultz and Tansel 2002) or group than the poorest one. It may be because may be used to justify inclusion in specific pro- these indicators are more subjective than the grams or receipt of benefits (Kerkhofs and Lin- physical functioning questions. deboom 1995). What is often found is that the The various measures of disability are highly rich report a greater incidence of disease and ill correlated, as would be hoped (see table 7). health than poorer groups. Regarding physical However, having one complete limitation in any functioning questions, it may well be that hav- of the six physical functioning domains has the ing a "full limitation"--being blind, being lowest correlation with official disability status. unable to walk a kilometer--is closest to an In other words, official disability status appears objective measure of health and is least affected to reflect something other than full limitations by self-reported health biases, while "partial" or in physical functioning. The alternative meas- "serious difficulties" are more subjective and ure with the highest correlation to official dis- thus, more prone to self-reporting biases. If this ability status is an individual's own assessment of is the case, the fact that the top quintile suffers having a disability. On the surface, this is not more from partial limitations than the poor may surprising as the two should be strongly linked: be a result of the measurement tool and not a a person with official disability status would be true reflection of disability. more likely to respond positively to a question In addition to the six domains of physical about having signs of a disability. Yet the range functioning, the survey allows the calculation of of "disabilities" reported by respondents is so a series of alternative measures of disability. broad--from paralysis and blindness to having Table 6 shows incidence rates using these other had a hernia or a hysterectomy to allergies and TABLE 6 Alternative Measures of Disability and Physical Functioning (means and standard deviations) Variable All Female Male Quintile 1 Quintile 5 Official disability status 0.0383 0.0326 0.0445 0.0321 0.0507 (0.1920) (0.1776) (0.2061) (0.1763) (0.2194) Chronic illness 0.1189 0.1278 0.1093 0.0686 0.2114 (0.3236) (0.3339) (0.3120) (0.2528) (0.4094) Own disability assessment (w2) 0.1031 0.1032 0.1029 0.0715 0.1453 (0.3040) (0.3043) (0.3038) (0.2577) (0.3525) Own functional disability assessment (w2) 0.0577 0.0552 0.0604 0.0495 0.0609 (0.2332) (0.2284) (0.2382) (0.2169) (0.2393) Normalized ADL score, w1 0.1218 0.1436 0.0981 0.1059 0.1539 (0.2856) (0.3057) (0.2600) (0.2757) (0.3057) Normalized ADL score, w2 0.1054 0.1192 0.0905 0.0767 0.1636 (0.2737) (0.2874) (0.2573) (0.2462) (0.3210) Sources: URPS Waves 1, 2, and 3; authors' calculations. Note: Includes only persons ages 7 and older. W1 and w2 refer to wave 1 and wave 2, respectively. 46 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE 7 Correlations of Alternative Measures of Disability to Official Disability Status Variable All Female Male Quintile 1 Quintile 5 No full limitations 0.2548* 0.2167* 0.3007* 0.2623* 0.2278* One full limitation 0.2249* 0.1925* 0.2664* 0.3407* 0.2086* One serious difficulty or full limitation 0.3364* 0.2709* 0.4211* 0.3741* 0.3157* Number of limitations 0.2378* 0.2213* 0.2741* 0.3761* 0.1818* Number of serious difficulties or full limitations 0.3277* 0.2556* 0.4187* 0.4053* 0.3126* Normalized domain score 0.3306* 0.2665* 0.4169* 0.3602* 0.3167* Chronic illness 0.3838* 0.3353* 0.4387* 0.4133* 0.3652* Own disability assessment (w2) 0.4858* 0.4623* 0.5108* 0.4637* 0.5096* Own functional disability assessment (w2) 0.3684* 0.3394* 0.3951* 0.3940* 0.3125* Normalized ADL score, w1 0.2781* 0.2037* 0.3759* 0.2883* 0.2869* Normalized ADL score, w2 0.3048* 0.2605* 0.3605* 0.3299* 0.3060* Sources: URPS Waves 1, 2, and 3; authors' calculations. *Significant at the .01 level. asthma--that it is hard to square it with a single itations but by other factors. (Perhaps limita- coherent definition of disability. tions in female decision making regarding their There are many reasons why official disabil- own affairs, including health, may lead to such ity status does not match limitations in the six discrepancies).11 domains of physical functioning. In the first To display the determinants of official dis- place, as discussed above, the definition of dis- ability status in a more comprehensive manner, ability in Uzbekistan is not based on physical we regress disability status first on the main functioning, per se, but on the actual limitation indicators of physical functioning: the normal- on the ability to work. Second, the definition ized domain score, whether a person has at least includes those with chronic diseases: having a one domain where he or she experiences serious chronic disease is positively correlated with difficulty or full limitation, his or her own dis- official disability status (.3838, significant at the ability assessment, self-reported chronic illness, .01 level), but even including the presence of and the normalized score of the ADL. We then chronic diseases does not eliminate the lack of add in characteristics of the individual that overlap between official and survey measures. might affect official disability status, primarily The third reason for the discrepancy may be those that might also be related to working, as that the survey data are self-reported and may the classification system in the country is highly well differ from a medical opinion. Finally, focused on one's ability to work, along with there may be incentives built into the benefit location variables to take into account differ- program for disability that affect who receives ences in the distance to the capital and other official status. administrative characteristics that may affect These descriptive statistics reveal other receipt of disability status and household wel- trends as well: for example, the correlation coef- fare levels. The best health-related predictors of ficients for females are always smaller than the official disability are: (i) having a serious diffi- correlation coefficients for males.10 This find- culty in one area of the six physical functioning ing suggests that for females, official disability domains, (ii) one's own assessment of disabling status is driven not so much by functioning lim- illnesses or injuries, and (iii) the limitations in Measurement of Disability and Linkages with Welfare, Employment, and Schooling 47 TABLE 8 Probability of Having Official Disability Status (marginal effects and [standard errors]) Alternative specifications Variables (1) (2) (3) Normalized domain score 0.001 0.013 0.016 [0.009] [0.007]* [0.006]** Chronic 0.007 0.005 0.005 [0.004]* [0.003]* [0.003]** One serious difficulty or full limitation 0.032 0.028 0.025 [0.010]*** [0.009]*** [0.008]*** Own disability assessment (w2) 0.134 0.106 0.112 [0.023]*** [0.021]*** [0.022]*** Own functional disability assessment (w2) 0.002 0.002 0.002 [0.002] [0.001] [0.001]** Normalized score of Activities of Daily Living 0.011 0.009 0.008 [0.003]*** [0.002]*** [0.002]*** Working age 0.010 0.009 [0.002]*** [0.001]*** Head of household 0.000 0.000 [0.001] [0.001] Female 0.006 0.005 [0.002]*** [0.001]*** Resides in Kashkadarya Oblast 0.001 [0.001] Resides in Andijan Oblast 0.008 [0.003]*** Per-capita consumption 0.0000017 [0.0000022] Observations 11,091 11,091 11,091 Sources: URPS Waves 1, 2, and 3; authors' calculations. Note: Dependent variable equal to 1 if individual has official disability status. Calculations include only persons 7 years of age or older. * Significant at 10%; ** significant at 5%; *** significant at 1%, standard errors in brackets. the ADL. Chronic illness and a score that cap- ing this status (after taking into account func- tures all problems in the six domains of func- tional and other disability assessments). Thus, tioning are not good predictors. Additionally, our data do not provide support for the being of working age increases the probability hypotheses that the poor are disadvantaged dur- of having official disability status. However, ing the process that leads to the granting of offi- being female lowers that likelihood: Although cial disability status either because the wealthy the effect is small, it is statistically significant. can afford to "purchase" disability status Finally, living in Andijan, rather than Tashkent, through informal payments and bribes; or increases the probability of having official dis- because the poor may have less access to the ability status. This statistically significant commissions that carry out the required med- regional effect may reflect differences in the ical examinations. implementation of central policies at the local The results of this survey have demonstrated level. Interestingly, the wealth proxy, per-capita that, indeed, different definitions of disability consumption, does not have a statistically sig- provide different incidence rates and, more nificant influence on the probability of receiv- important, they identify different people as hav- 48 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union ing a disability. A straight physical functioning services to a household member with disabili- definition, whereby a person is considered to ties. Untangling the causal links between dis- have a disability if he or she has serious difficulty ability and welfare is a complex task, and the in one domain or is unable to carry out the activ- data available for Uzbekistan are not adequate ity included in the domain, indicates almost 12 for this purpose. However, they can provide percent of the population age 7 and above has a insights on the extent to which welfare levels disability. If the definition of disability is and disability are linked, and the extent to restricted to individuals who have a full limitation which educational and labor force participation in one of the six domains (blindness, inability to are related to functional limitations, disabili- communicate, etc.), the incidence of disability is ties, and poor health. 3.2 percent. Official disability status, however, is held by 3.8 percent of the population. Welfare It is not just that the different measures give The first step is to determine whether there is different rates of disability; they also identify any link between welfare and disability. Welfare different individuals as suffering from disabili- here is measured by the log of per-capita house- ties. The question then is whether this shifting hold consumption. Household consumption is a definition affects our understanding of the links measure that incorporates not simply household between disability and human capital formation, expenditures, but also the value of home pro- labor force participation, and poverty. In the duction and the use value of durables and hous- next section we provide some preliminary ing. (See Deaton and Zaidi 2002 for the insights on this issue by looking at the relation- conceptual foundations of this measure.) We ships between the different measures of disabil- regress log per-capita consumption on individ- ity and individual characteristics. ual and household characteristics to determine the impact of individual disability on poverty. The results are shown in table 9. III To What Extent are Disabled Holding all else constant, having a serious Individuals Poor, Less Educated, and limitation or full limitation has a statistically sig- Out of the Labor Force? nificant negative effect on consumption, result- ing in a 4.8 percent decline after other factors Poverty may expose individuals to greater risks are taken into account. While the estimated of becoming disabled as the poor may lack ade- coefficients for other disability indicators have quate health care, be forced to live in unsafe the expected sign, none are statistically signifi- housing and environments, or work in danger- cant. The lack of a statistically significant offi- ous jobs. On the other hand, disability can lead cial disability status effect on consumption may to lower levels of well being if individuals are in part be because disability benefits (from the unable to participate in society in the same way social security system) help to offset the loss of a person without disabilities can. This is the income or costs associated with having disabili- concept of functional limitations. If an individ- ties. Another interpretation, however, may be ual cannot attend school, his or her opportuni- that older people with limited functioning do ties for future earnings are limited, and without not view their condition as "disability," but adequate personal and physical devices, an indi- instead see it as simply part of the aging process. vidual might be unable to participate in the These individuals are thus less likely to seek dis- labor market. It is not only the person with dis- ability status and, given the definitional link abilities who is affected. Other household between disability status and employment, members may have to restrict their own eco- might be less likely to obtain it if they did seek nomic activities to care for or provide personal it. And the aging process, or having multiple Measurement of Disability and Linkages with Welfare, Employment, and Schooling 49 TABLE 9 Consumption and Disability: Log of per Capita Consumption Is the Dependent Variable, OLS Coefficients, and Standard Errors (in brackets) Alternative Specifications Variables (1) (2) (3) (4) (5) (6) Years of age 0.001* 0.001* 0.001* 0.001** 0.001* 0.001** [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Female 0.004 0.004 0.004 0.005 0.004 0.005 [0.012] [0.012] [0.012] [0.012] [0.012] [0.012] Household size 0.049*** 0.049*** 0.049*** 0.049*** 0.049*** 0.049*** [0.003] [0.003] [0.003] [0.003] [0.004] [0.004] No. of children < 7 0.002 0.002 0.003 0.003 0.002 0.002 [0.007] [0.007] [0.007] [0.007] [0.007] [0.007] No. children 7­15 0.076*** 0.076*** 0.076*** 0.076*** 0.076*** 0.076*** [0.006] [0.006] [0.006] [0.006] [0.006] [0.006] No pens. age adult 0.034*** 0.034*** 0.034*** 0.035*** 0.034*** 0.035*** [0.009] [0.009] [0.009] [0.009] [0.009] [0.009] Head's yrs. educ. 0.039*** 0.039*** 0.039*** 0.038*** 0.039*** 0.039*** [0.002] [0.002] [0.002] [0.002] [0.002] [0.002] Female head hhld. 0.088*** 0.088*** 0.088*** 0.089*** 0.088*** 0.089*** [0.015] [0.015] [0.015] [0.015] [0.015] [0.015] Kashkadarya 0.530*** 0.529*** 0.530*** 0.530*** 0.530*** 0.529*** [0.016] [0.016] [0.016] [0.016] [0.016] [0.016] Andijan 0.634*** 0.634*** 0.634*** 0.636*** 0.635*** 0.635*** [0.015] [0.015] [0.015] [0.015] [0.015] [0.015] One full limitation 0.030 [0.035] Off. disability stat. 0.032 [0.030] One serious difficulty or full limitation 0.048** [0.020] Chronic 0.007 [0.019] ADLs, norm. score 0.035 [0.023] Constant 13.094*** 13.094*** 13.095*** 13.092*** 13.095*** 13.092*** [0.037] [0.037] [0.037] [0.037] [0.037] [0.037] Observations 10,854 10,854 10,854 10,854 10,854 10,854 R-squared 0.44 0.44 0.44 0.44 0.44 0.44 Sources: URPS Waves 1, 2, and 3; authors' calculations. Note: Calculations include only persons 7 years of age or older. * Significant at 10%; ** significant at 5%; *** significant at 1%. limitations in physical activities, is associated developing countries (Psacharopolous and with lower monetary welfare. Tzannatos 1992; Psacharopoulos 1994); for transition countries as they undertake reforms Education (Yemtsov, Cnobloch, and Mete 2006); and in the The investment in education has a high rate of specific case of Uzbekistan. The World Bank return for individuals. This is true in general for (2006) finds that higher education is associated 50 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union with roughly a 60 percent increase in earnings, cally significant. Official disability status has the compared to the attainment of basic education strongest effect: a 24 percentage point decline in Uzbekistan. To the extent that disabilities in enrollments. This is followed by chronic ill- prevent children from attending school, there ness, with an effect ranging from 10 to almost will be a long-term negative impact on their 12 percentage points. Having at least one seri- earnings ability and future welfare. ous difficulty or full limitation in functioning Enrollment levels in Uzbekistan are fairly lowers the probability of enrollment by 6 to 7 high at the basic education level. As shown in percentage points. Scott (2005), just over 90 percent of all 6- to 14- If one looks at children between the ages of year-olds were enrolled in school in 2003, and 15 and 18, other factors, such as the consump- this is consistent across economic regions and tion level of the household and the education urban and rural areas (see table 10). By age 15, level of the head of the household, become however, enrollment levels fall, and by age 18, important in explaining dropout rates. only two-thirds of children are still in school. However, having official disability status is The welfare level of the household, the educa- the only one of the four health/disability indi- tion of the head of household, and the geo- cators that is associated with a significant graphic location of the household all affect decline in school enrollment. The effect is dra- schooling. Here, in addition to these basic cor- matic: holding all else constant, official disabil- relates of schooling, we explore the impact that ity status is associated with a 42 to 49 disability has on the probability of children's percentage point drop in the probability of school enrollment. school enrollment. This may indicate that dis- Three of the four disability or health vari- ability has a much stronger effect on school ables used here (official disability status, one enrollment at older age levels, perhaps reflect- serious or full limitation, and chronic illness) ing fewer resources available for such individu- have statistically significant negative effects on als in the school system, or greater costs or school enrollment for 7- to 14-year-olds (table fewer benefits associated with enrollment. It 11). The coefficient on the fourth (one full lim- may also, however, reflect the incentive struc- itation) has the expected sign but is not statisti- tures of official disability status. It may be more TABLE 10 Gross Enrollment Rates by Economic Region in Uzbekistan, 2002 to 2003 Ages 6­14 Ages 15­18 2002 2003 2002 2003 Urban 88.2 91.4 59.0 69.0 Rural 89.4 91.5 58.3 65.4 Tashkent 89.0 89.3 68.8 68.7 Mirzachul 88.8 92.2 52.8 60.6 Ferghana 86.8 91.4 58.8 68.5 Northern 90.6 91.8 50.3 62.3 Central 89.8 92.3 59.2 67.2 Southern 90.5 92.1 58.1 66.5 National 89.0 91.5 58.5 66.5 Source: World Bank Living Standards Assessment (2007). Note: Data are from the national Household Budget Survey implemented by the State Statistical Committee, not administrative records. Measur TABLE 11 Probability of School Enrollment: Children Age 7­14 ement (marginal effects and [standard errors]) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) of Disability Age 0.0004 0.0007 0.0001 0.0003 0.0001 0.0004 0.0005 0.0002 0.0001 0.00004 0.0003 0.0002 0.0001 0.0000005 0.000006 [0.0007] [0.0004] [0.0005] [0.0002] [0.0003] [0.0006] [0.0004] [0.0004] [0.0001] [0.0001] [0.0004] [0.0002] [0.0002] [0.000001] [0.00001] Female 3E-05 0.0004 0.0011 0.0002 0.0016 0.0005 0.0001 0.0006 0.00003 0.0005 0.0002 0.0001 0.0007 0.0000005 0.0002 and [0.0033] [0.0019] [0.0023] [0.0010] [0.0019] [0.0026] [0.0012] [0.0015] [0.0002] [0.0009] [0.0016] [0.0005] [0.0009] [0.000002] [0.0003] Household size 0.0005 0.0003 0.0004 0.0001 0.0002 0.0001 0.00004 0.000004 0.0000002 0.000004 Linkages [0.0007] [0.0003] [0.0003] [0.0001] [0.0002] [0.0005] [0.0002] [0.0002] [0.0000007] [0.00003] No. of children <7 0.0001 0.0006 0.0006 0.0001 0.0001 0.0002 0.0003 0.0003 0.0000005 0.00002 [0.0009] [0.0005] [0.0006] [0.0001] [0.0003] [0.0006] [0.0003] [0.0003] [0.000001] [0.00004] with No. of child. 7­15 0.0019 0.001 0.0011 0.0002 0.0007 0.0008 0.0003 0.0004 0.0000006 0.0001 W [0.0015] [0.0009] [0.0008] [0.0002] [0.0007] [0.0007] [0.0004] [0.0004] [0.000002] [0.0001] elfar Head's educ., yrs 0.0005 0.0002 0.0003 0.0000003 0.00002 e, [0.0004] [0.0002] [0.0003] [0.0000008] [0.00004] Employment, PC consumption 0.00001* 0.000004 0.000004 0.00000001 0.0000008 [0.000006] [0.000003] [0.000002] [0.00000004] [0.000001] One full limit. 0.1017 0.1062 0.0673 [0.0624] [0.0670] [0.0473] and Official disability status 0.2278* 0.2444** 0.3109** Schooling [0.1207] [0.1224] [0.1431] One serious diffic. or full limit. 0.0672** 0.0609** 0.0123*** [0.0317] [0.0301] [0.0116] Chronic 0.0992** 0.1160** 0.1295** [0.0461] [0.0518] [0.0566] Observations 2,383 2,383 2,383 2,383 2,383 2,383 2,383 2,383 2,383 2,383 2,030 2,030 2,030 2,030 2,030 Source: URPS Waves 1 and 3. * Significant at 10%; ** significant at 5%; *** significant at 1%, standard errors in brackets. 51 TABLE 12 52 Probability of School Enrollment, Ages 15­18 (marginal effects and [standard errors]) Economic (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) Years of age 0.0929*** 0.0934*** 0.0950*** 0.0940*** 0.0929*** 0.0925*** 0.0932*** 0.0951*** 0.0941*** 0.0929*** 0.0890*** 0.0894*** 0.0911*** 0.0904*** 0.0893*** [0.0097] [0.0097] [0.0097] [0.0097] [0.0097] [0.0103] [0.0103] [0.0103] [0.0104] [0.0103] [0.0100] [0.0100] [0.0100] [0.0101] [0.0100] Implications Female 0.0024 0.0017 0.0024 0.0017 0.0033 0.001 0.0014 0.0055 0.0013 0.0001 0.0128 0.0129 0.0175 0.0125 0.0119 [0.0216] [0.0214] [0.0213] [0.0215] [0.0216] [0.0211] [0.0209] [0.0209] [0.0210] [0.0211] [0.0191] [0.0190] [0.0188] [0.0191] [0.0191] Household size 0.0092 0.0094 0.0097 0.0091 0.0098 0.0033 0.0033 0.0038 0.003 0.0036 [0.0070] [0.0070] [0.0070] [0.0070] [0.0071] [0.0063] [0.0063] [0.0062] [0.0063] [0.0063] of No. of child. <7 0.0139 0.0113 0.0113 0.0129 0.0128 0.015 0.0128 0.0123 0.0144 0.0141 Chronic [0.0155] [0.0154] [0.0152] [0.0154] [0.0154] [0.0133] [0.0130] [0.0128] [0.0131] [0.0132] No. of child. 7-15 0.0028 0.0023 0.0015 0.001 0.002 0.0115 0.0108 0.0098 0.0097 0.0107 [0.0121] [0.0120] [0.0120] [0.0121] [0.0121] [0.0111] [0.0110] [0.0109] [0.0111] [0.0111] Illness Head's educ., yrs. 0.0070** 0.0069** 0.0063** 0.0067** 0.0070** [0.0028] [0.0028] [0.0028] [0.0028] [0.0028] and Female head of hhld. 0.0421 0.0357 -0.0429 0.0385 0.0387 [0.0303] [0.0302] [0.0301] [0.0303] [0.0303] Disability PC consumption 0.0002*** 0.0002*** 0.0002*** 0.0002*** 0.0002*** [0.0001] [0.0001] [0.0001] [0.0001] [0.0001] One full limit. 0.3513 0.3138 0.2722 in [0.2739] [0.2590] [0.2672] Eastern Official disability status 0.4261** 0.4336** 0.4875*** [0.1876] [0.1800] [0.1807] Europe One serious diff. or full limit. 0.1811 0.1737 0.1595 [0.1239] [0.1213] [0.1221] Chronic 0.0486 0.061 0.0734 and [0.0669] [0.0680] [0.0694] the Former Observations 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 1,309 Source: URPS Waves 1 and 3. Soviet * Significant at 10%; ** significant at 5%; *** significant at 1%, standard errors in brackets. Union Measurement of Disability and Linkages with Welfare, Employment, and Schooling 53 advantageous for older children to obtain such the work force. Here we look at basic individual status, or it may simply be easier once a child and household characteristics and the effect reaches a certain age for such status to be they have on whether a person is active in the obtained. A more detailed understanding of labor market. The International Labour Orga- how disability benefits are conferred is needed nization's (ILO's) definition of being "economi- to clarify this point. cally" active is used here, which includes Another way of looking at the link between persons working or actively looking for work education and disability would be to assess the (unemployed).12 effect of disability on the number of years of The dependent variable takes the value 1 if schooling an individual obtains. To do this the person was active in either Wave 1 of the properly, information is needed on the indi- survey (March 2005) or Wave 3 of the survey vidual and his or her household and location (December 2005). It may well be that there is during the ages when schooling was (or would seasonality in the labor market: using two points normally have been) acquired, as well as infor- in time should prevent seasonality issues from mation about the onset of any health problems being confounded with disability or other issues or limitations. Unfortunately, the data set used that affect labor force participation. here does not provide all of this information. Table 14 shows that the probability of being In the interest of further exploring the link economically active is closely linked to the dis- between disability and education given the ability and health status of the individuals. All strong association shown in terms of enroll- four disability/poor health indicators consid- ment, we regress the number of years of edu- ered here have statistically significant coeffi- cation on age and gender and various cients at the 1 percent level. Official disability indicators of disability. Obviously, the results status is associated with a 52 percentage point should be used with extreme caution, but, as reduction in the probability of being econom- can be seen in table 13, there does seem to be ically active. Corresponding statistics for hav- further evidence that disability and educa- ing a full limitation, one serious difficulty or tional attainment are linked. Having a full lim- full limitation, and having a chronic illness are itation or having a serious difficulty is 37 percentage points; 24 percentage points; associated with fewer years of schooling. Hav- and 19 percentage points, respectively. These ing a full limitation reduces schooling by estimates should be viewed as descriptive asso- almost two and a half years, while having a ciations rather than causal linkages, however. serious difficulty or full limitation lowers it by As highlighted by Bound (1991) and discussed more than one year. Official disability status in detail by Mete and Schultz (2007),13 treat- has less of an effect (less than half a year in one ing disability/poor health indicators as exoge- specification) and chronic illness is not associ- nous variables in employment regressions ated with fewer years of schooling, as the pre- overlooks the (upward) bias that may arise vious analysis would suggest. because of self-reporting of heath variables. On the other hand, as shown previously, chil- Labor Force Participation dren who are disabled are much less likely to Household and individual welfare are clearly stay in school after basic education, and so the influenced by an individual's ability to work. presence of schooling attainment variables Being able to work can be affected by the level (also treated as exogenous variables here) in of education one obtains, the physical or mental the model may lead to an underestimation of ability required for the job, and the willingness the causal effect of disability on labor force of society to include persons with disabilities in participation. 54 TABLE 13 Economic Years of Schooling Obtained (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Implications Years of age 0.044*** 0.036*** 0.044*** 0.031*** 0.046*** 0.049*** 0.041*** 0.048*** 0.035*** 0.047*** [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] Female 0.317*** 0.303*** 0.323*** 0.273*** 0.320*** 0.364*** 0.351*** 0.372*** 0.319*** 0.362*** [0.074] [0.073] [0.074] [0.073] [0.074] [0.072] [0.071] [0.072] [0.071] [0.072] of Andijan 1.465*** 1.461*** 1.467*** 1.530*** 1.498*** Chronic [0.087] [0.087] [0.087] [0.087] [0.088] Kashkadarya 1.340*** 1.306*** 1.347*** 1.342*** 1.367*** [0.093] [0.091] [0.093] [0.091] [0.092] Illness One full limit. 2.462*** 2.400*** [0.346] [0.335] and Official disability status 0.344 0.408* Disability [0.227] [0.223] One serious diff. or full limit. 1.240*** 1.341*** [0.165] [0.160] Chronic 0.183 0.213 in Eastern [0.141] [0.137] Constant 12.710*** 12.468*** 12.703*** 12.353*** 12.735*** 13.812*** 13.561*** 13.807*** 13.451*** 13.806*** [0.120] [0.120] [0.121] [0.121] [0.123] [0.137] [0.135] [0.137] [0.135] [0.137] Europe Observations 7,519 7,519 7,519 7,519 7,519 7,519 7,519 7,519 7,519 7,519 and R-squared 0.08 0.10 0.08 0.10 0.08 0.14 0.16 0.14 0.16 0.14 the Sources: URPS Waves 1, 2, and 3; authors' calculations. Former Note: Includes individuals ages 19 and older. * Significant at 10%; ** significant at 5%; *** significant at 1%, standard errors in brackets. Soviet Union Measur TABLE 14 Probability of Being Economically Active ement (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Age 0.0039*** 0.0041*** 0.0057*** 0.0052*** 0.0050*** 0.0036*** 0.0038*** 0.0054*** 0.0049*** 0.0046*** of Disability [0.0006] [0.0007] [0.0007] [0.0007] [0.0007] [0.0006] [0.0006] [0.0007] [0.0007] [0.0007] Female 0.1814*** 0.1842*** 0.1937*** 0.1817*** 0.1812*** 0.1853*** 0.1877*** 0.1967*** 0.1847*** 0.1856*** [0.0129] [0.0129] [0.0128] [0.0128] [0.0128] [0.0131] [0.0131] [0.0129] [0.0130] [0.0130] and Education (yrs.) 0.0309*** 0.0293*** 0.0274*** 0.0281*** 0.0295*** 0.0351*** 0.0335*** 0.0313*** 0.0321*** 0.0327*** [0.0031] [0.0031] [0.0031] [0.0031] [0.0031] [0.0031] [0.0032] [0.0032] [0.0032] [0.0031] Linkages Head of hhld. 0.0661*** 0.0673*** 0.0698*** 0.0651*** 0.0712*** 0.0621*** 0.0638*** 0.0664*** 0.0619*** 0.0644*** [0.0170] [0.0171] [0.0170] [0.0170] [0.0167] [0.0187] [0.0187] [0.0187] [0.0187] [0.0185] Hhld. size 0.0094*** 0.0092** 0.0088** 0.0084** 0.0102*** with [0.0036] [0.0036] [0.0036] [0.0036] [0.0036] W No. of child. <7 0.0132* 0.0124 0.0090 0.0097 0.0118 elfar [0.0076] [0.0076] [0.0076] [0.0076] [0.0077] e, No. of child. 7­15 0.0226*** 0.0210*** 0.0162** 0.0186*** 0.0199*** Employment, [0.0067] [0.0067] [0.0067] [0.0067] [0.0067] No. of pension-age adults 0.0051 0.0048 0.0049 0.0048 0.0043 [0.0096] [0.0096] [0.0095] [0.0096] [0.0096] Andijan 0.1028*** 0.1027*** 0.1032*** 0.0971*** 0.0881*** and [0.0142] [0.0142] [0.0140] [0.0143] [0.0146] Schooling Kashkadarya 0.0321** 0.0329** 0.0289* 0.0332** 0.0207 [0.0152] [0.0152] [0.0152] [0.0151] [0.0155] One full limit. 0.3727*** 0.3638*** [0.0661] [0.0672] Official disability status 0.5201*** 0.5201*** [0.0379] [0.0388] One serious diff. or full limit. 0.2431*** 0.2249*** [0.0315] [0.0319] Chronic 0.1901*** 0.1694*** [0.0257] [0.0260] Observations 7,467 7,467 7,467 7,467 7,467 7,467 7,467 7,467 7,467 7,467 Sources: URPS Wave 1, 2, and 3; authors' calculations. 55 Note: Includes individuals ages 16­55 (female) and 16­60 (male). * Significant at 10%; ** significant at 5%; *** significant at 1%, standard errors in brackets. 56 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union IV Conclusion reduces the likelihood. Living in Andijan, rather than Tashkent, increases the probability of hav- The basic finding of this paper is that the rate of ing official disability status. The wealth proxy, disability found in a population will vary sub- per-capita consumption, does not have a statis- stantially, depending on the measure used. tically significant influence on the probability of More important, our understanding of the link- receiving official disability status. Thus, two ages among welfare, education, and labor force ways that the current system can be improved participation can also be affected by the choice emerge as: (i) improving females' access to of measure. information and to commissions that determine We find that almost 12 percent of all individ- disability levels, and (ii) ensuring consistent uals ages 7 and older have at least one serious implementation of nationally determined crite- difficulty or a full limitation in physical func- ria for granting disability status across different tioning. If one restricts the definition of disabil- regions. ity to individuals who have a full limitation on As mentioned previously, the extent to which one of the six domains (blindness or learning, disability is related to welfare, education, and for example), the incidence of disability is 3.2 labor force participation is sensitive to the dis- percent. Official disability status, however, is ability measure that is used. For example, offi- held by 3.8 percent of the population age 7 and cial disability status plays the most important older. While the official disability rate increases role in explaining school enrollment for 15- to by age, it remains at around 10 percent among 18-year-olds. (Official disability status is associ- those who are older than 66. In contrast, the ated with more than a 40 percentage point drop share of individuals with at least one full limita- in the probability of school enrollment. The tion increases from about 5 percent for the 7- effect of other disability indicators ranges from to-16 age group, to 65 percent among those a 6 percentage point to 35 percentage point who are 66 and over. This finding suggests that drop.) Similarly, official disability status is while developing countries with aging popula- strongly related to probability of being econom- tions may be able to contain one type of finan- ically active among adults (official disability sta- cial burden by being restrictive in granting tus is associated with a 52 percentage point disability benefits to the elderly, the functional reduction in the probability of being economi- limitations still increase steeply by age, with cally active, while the effect of other disability implications for employment and productivity. indicators ranges from a 19 percentage point Even though all disability indicators consid- reduction to a 37 percentage point reduction). ered here are correlated with one another, their In contrast, having a serious difficulty or full linkage with official disability status is of partic- limitation is most closely associated with reduc- ular interest because the officially disabled are tions in household consumption (resulting in a entitled to social protection benefits. The best 4.8 percent decline), while the effect of having health-related predictors of official disability official disability status remains statistically are: (i) having at least one serious difficulty in insignificant, and less than half in magnitude. one of the six physical functioning domains, (ii) This last finding is perhaps a natural result of a one's own assessment of disabling illnesses or process in which disability benefits are granted injuries, and (iii) limitations in the ADL. to the officially disabled based on a medical Chronic illness and a score that captures all evaluation (as opposed to considerations of problems in the six domains of functioning are functional disabilities as well). not good predictors. Additionally, being of Finally, the original panel sample (which col- working age increases the probability of having lected information on 12,387 individuals) and official disability status, while being female the full Wave 3 sample (which collected infor- Measurement of Disability and Linkages with Welfare, Employment, and Schooling 57 mation on 32,337 individuals) provide remark- questionnaires do not inquire about household ably similar estimates of the prevalence of dis- consumption patterns. Furthermore, given the abilities, regardless of the definition that is used. nature of census questionnaires, the implemen- This finding sheds doubt on the validity of the tation of a comprehensive set of questions on view that disability is a rare event that can be disabilities is out of question. Yet the Uzbek data captured best by the inclusion of appropriate shows that there is no single disability indicator questions in census questionnaires. Such an that overshadows other indicators for many approach, while useful for estimating prevalence questions that might interest policy makers. rates according to one or two disability defini- Instead, various disability indicators, together, tions, would be of very limited use for establish- provide insights on identifying the most vulner- ing linkages with poverty because census able populations. 58 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Appendix I Health, Disability, and Physical Functioning Questions in URPS, Waves 1, 2, and 3 Waves 1 and 2 CHRONIC · Does [NAME] suffer from a chronic illness, such as asthma, diabetes, palsy, etc.? Yes/No ACTIVITIES OF DAILY LIVING · Has [NAME]'s health limited his/her ability to perform vigorous activities such as lifting heavy objects, running, and for how long? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 · Has [NAME]'s health limited him/her from doing moderate activities such as moving a table or carrying groceries, and for how long? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 · Has [NAME]'s health limited his/her ability to walk uphill, and for how long? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 · Has [NAME]'s health limited him/her from walking 100 meters? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 · Has [NAME]'s health limited him/her from bending, lifting, or stooping, and for how long? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 · Has [NAME]'s health limited him/her from eating, dressing, bathing, or using the toilet, and for how long? Yes, for more than three months..........................1 Yes, for less than three months.............................2 No..........................................................................3 Measurement of Disability and Linkages with Welfare, Employment, and Schooling 59 Wave 2 DISABILITY-SELF ASSESSED AND OFFICIAL · Does [NAME] have any signs of disability? Bad hearing...............................................................1 Deaf...........................................................................2 Bad vision..................................................................3 Blind..........................................................................4 Adynamia/aspen9z....................................................5 Palsy ..........................................................................6 Absence of hands, legs..............................................7 Mental disease ..........................................................8 Consequence of tuberculosis, heart disease ............9 Other, specify _______________________________ None .......................................................................98 · Is [NAME] officially recognized by medical workers as disabled? Yes, and with written confirmation ......................1 Yes, BUT written confirmation not given ...........2 No..........................................................................3 Did not refer to officials .......................................4 Wave 3 60 PHYSICAL FUNCTIONING Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Measur ement of Disability and Linkages with W elfar e, Employment, and Schooling 61 62 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Measur ement of Disability and Linkages with W elfar e, Employment, and Schooling 63 64 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Measurement of Disability and Linkages with Welfare, Employment, and Schooling 65 Notes 9. This may reflect people's ability to adapt to cer- tain limitations if the condition is static, as it may 1. Rough estimates, however unreliable, are that well be if not age related. 40 million of the 115 million children not 10. The correlation coefficients for the poorest enrolled in school have a disability ("Education quintile tend to be larger than the ones for the for All: Including Children with Disabilities," wealthiest quintile (save for one exception). Education Notes, The World Bank, August 11. The World Bank (2006) reveals that only 43 per- 2003). cent of adult females have a final say on their 2. Similarly, analyzing household survey data from own health care and almost 25 percent of adult 10 developing country household surveys, females are not allowed to go to the health cen- Filmer (2004) reports less than 2 percent of the ter alone. population as disabled. 12. Open unemployment is relatively low in Uzbek- 3. This section is based largely on the paper by Alt- istan, at less then 5 percent (World Bank 2006). man, Cambois, and Robine, 2005. For more 13. See also Autor and Duggan 2006. details on the background, tasks, and progress to date of the Washington Group on Disability Statistics, see their web site at http://www.cdc References .gov/nchs/citygroup.htm. 4. Differences in community resources (such as Altman, Barbara M., Emmanuelle Cambois, and wheelchair accessible transport and services) or Jean-Marie Robine. 2005. "Extended Question individuals themselves (such as education) could Sets: Purposes, Characteristics, and Topic Ar- also affect functioning even if the disability were eas." Paper presented at the 5th meeting of the the same. In this work we are able to control for Washington Group on Disability Statistics, Rio assistive devices but not for community resources. de Janeiro, Brazil. 5. One shortcoming of household surveys is their Autor, David H., and Mark G. Duggan. 2006. "The inability to take into account disabled individu- Growth in the Social Security Disability Rolls: als who reside in institutions. In Uzbekistan, the A Fiscal Crisis Unfolding." Working Paper Se- Ministry of Labor and Social Protection oper- ries, No. 12436, National Bureau of Economic ates 34 boarding facilities for the disabled, serv- Research, Cambridge, Massachusetts. ing around 8,900 individuals (about 1,500 Bercovich, A. 2004. "People with Disability in general-disabled adults, 5,600 adults with men- Brazil: A Look at 2000 Census Results." Work- tal disorders, and 1,800 disabled children). See ing Paper, Instituto Brasileiro de Geografia e Es- Japan International Cooperation Agency tatística, Rio de Janeiro, Brazil. (JICA)/Tahlil 2004 for more information. Bound, John. 1991. "Self-Reported Versus Objective 6. Major pieces of legislation relating to disability Measures of Health in Retirement Models." in Uzbekistan are: the Law on Social Protection Journal of Human Resources, 26 (1): 106­38. for Disabled Persons, 1991 (amendments in Deaton, Angus, and Salman Zaidi. 2002. "Guide- 1998, 2001); Law on State Pension Security for lines for Constructing Consumption Aggre- Citizens, 1992; Law on Pension Security for gates for Welfare Analysis, Living Standards, Military Service Persons, 1990; Law of Employ- Measurement Study," Working Paper No. 135, ment of the Population, 1992; Labor Code of World Bank, Washington, DC. the Republic of Uzbekistan (JICA and Tahlil Elwan, Ann. 1999. "Poverty and Disability: A Survey 2004). of the Literature." Social Protection Discussion 7. This could be problematic if persons with dis- Paper Series No. 9932, World Bank, Washing- abilities are more prone to migration than oth- ton, DC. ers. The data does not, however, allow us to Filmer, Deon. 2004. "Disability, Poverty, and examine this issue. Schooling: Results from 10 Household Sur- 8. When other factors such as household structure veys." Presentation for conference on Disability and household head's gender, age, and schooling and Inclusive Development, World Bank, are controlled for, disability is associated with a Washington, DC. decline in per-capita consumption for all indica- Flores, R., D. Yepez, and M. Pramatarova. 2005. tors that are considered, although the coefficient "Ecuador: La discapacidad en cifras." Instituto Na- is significant in only one case. (This is reported cional de Estadísticas y Censos de Ecuador, Quito, in table 9 later in the main text.) Ecuador. 66 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Kerkhofs, M., and Maarten Lindeboom. 1995. "Sub- Psacharopolous, George, and Zafiris Tzannatos, eds. jective Health Measures and State Dependent 1992. Women's Employment and Pay in Latin Reporting Errors." Health Economics, 4: 221­35. America, Washington, DC: World Bank. Instituto Nacional de Estadísticas y Censos-Nicaragua. Scott, Kinnon. 2005. "An Update on Living Stan- 2003. "Encuestas Nicaragüense para personas con dards in Uzbekistan, 2000­2003." Background discapacidades, ENDIS," Managua Nicaragua. paper for the Uzbekistan Poverty Assessment, Japan International Cooperation Agency (JICA) and World Bank,Washington, DC. Tahlil Center for Social Research. 2004. "Re- Schultz, T. Paul, and Aysit Tansel. 1997. "Wage and view of Situation of Disabled People in the Re- Labor Supply Effects of Illness in Côte d'Ivoire public of Uzbekistan." Report prepared for and Ghana: Instrumental Variable Estimates for Japan International Cooperation Agency, Days Disabled." Journal of Development Econom- Tashkent, Uzbekistan. ics 53 (2): 251­86. Lindeboom, M., and Eddy van Doorslaer. 2004. United Nations. 2005. "Disability Statistics Data "Cut-Point Shift and Index-Shift in Self-Re- Base." Statistics Division, http://unstats.un.org/ ported Health." Tinbergen Institute Discussion unsd/demographic/sconcerns/disability/disab2 Paper TI-2003-042/3. .asp. Martinho, Maria, and Jeremiah Banda. 2005. "Unit- Van Doorslaer, Eddy, and Ulf. G. Gerdtham. 2002. ed Nations Collection of National Disability "Does Inequality in Self-Assessed Health Pre- Statistics from Population and Housing Cen- dict Inequality in Survival by Income? Evidence suses and Related National Surveys and Admin- from Swedish Data." Social Science and Medicine istrative Records." Presentation at the 4th 57: 1621­29. meeting of the Washington Group on Disabili- World Bank. 2003. "Education for All: Including ty Statistics, Rio de Janeiro, Brazil. Children with Disabilities." Education Notes, Mete, Cem, and T. Paul Schultz. 2007. "Health and Washington, DC. Labor Force Participation of the Elderly in Tai- World Bank. 2006. Uzbekistan Living Standards Up- wan." In Allocating Public and Private Resources date. Washington, DC: World Bank. across Generations, ed. Anne H. Gauthier, C.Y. World Health Organization. 2001. International Classi- Chrus, and Shripad Tuljapurkar, 163­200, New fication of Functioning, Disability, and Health, Gene- York: Springer. va, Switzerland: World Health Organization. Mont, Daniel. 2005. "Disability in Central Ameri- ------. 1980. International Classification of Impair- ca." Presentation at the 4th meeting of the ments, Disabilities, and Handicaps, Geneva, Washington Group on Disability Statistics, Rio Switzerland: World Health Organization. de Janeiro, Brazil. Yemtsov, Ruslan G., Stefania R. Cnobloch, and Cem Psacharopoulos, George. 1994. "Returns to Invest- Mete. 2006. "Evolution of the Predictors of ment in Education: A Global Update." World Earnings during Transition." Manuscript. Development 22 (9): 1325­43. Washington, DC: World Bank. CHAPTER 3 The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina Cem Mete, Huan Ni, Kinnon Scott* Introduction and enrollment patterns--instead of exclusively focusing on earnings and household consump- Ill health, disease, and disability can have signif- tion.2 Despite the potential significance of this icant effects on individual and household wel- line of research on policy making (through fare in developing countries. Households may instruments such as catastrophic health insur- be able to smooth the negative impact on house- ance), the evidence is scarce. This is especially hold consumption that comes from episodes of true in terms of the linkages between disability relatively minor illness (Townsend 1994; 1995), and poverty: the developing country literature but they are unlikely to be able to cope with the consists primarily of studies based on anecdotal occurrence of a major illness or disability in the evidence and case studies (as observed by Elwan same manner (Gertler and Gruber 2002).1 If we 1999, and Wagstaff 2005). are to understand the linkages between health The relationship between poor health/dis- shocks and poverty, we need to be able to distin- ability and poverty may be particularly strong in guish among different types of illnesses and dis- the transition economies. Common coping ability and their effects on consumption. In mechanisms may not work in these economies, addition, there may be payoffs to considering especially those where GDP levels declined by the impact of health shocks on a broader set of more than half during the first decade of transi- household wealth indicators--for example, by tion (World Bank 2003). Informal social protec- taking into account children's school attendance tion arrangements through networks of family * Cem Mete is a senior economist at the Europe and Central Asia region of the World Bank. Huan Ni is an assistant professor at the economics department of the Kennesaw State University. Kinnon Scott is a senior economist at the Development Research Group of the World Bank. 67 68 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union and friends that can provide relief for small or A key contribution of this work is that we temporary economic shocks are unlikely to be tackle the question of the presence or absence of effective in the face of such widespread and a relationship between health shocks and house- extreme deprivation. Nor is there much room hold consumption by focusing on several indi- for intrahousehold adjustments to be made to cators of poor health/disability: an "Activities of smooth consumption: Many transition coun- Daily Living" Index (ADL); self-reported tries have had low fertility rates and nuclear chronic disease; and self-reported disability sta- household living arrangements for decades (as tus information. By doing this, we seek to deter- opposed to high fertility rates and extended mine which disability/poor health indicators are family living arrangements observed in poor more closely related to household consumption. South Asian countries). Furthermore, during This analysis builds on Gertler and Gruber's the 1990s, the health care infrastructure crum- proposition that the measures of health that are bled, and thus disabled individuals in many tran- used may make a difference in the analysis of sition countries have not benefited as much health shock impacts, particularly the difference from the continuous advances in medical tech- between unexpected major illness versus minor nology that developed country residents enjoy illness. We also investigate whether the timing (Kahn 1998; Costa 2000). of the onset of disability or disease has an impact The link between poor health/disability and on the observed trends in employment, earn- employment outcomes would also be expected ings, and household consumption. The motiva- to be strong given that the transition economies tion for this extension to the basic empirical are characterized by high unemployment and a framework is the implications for the design of relative abundance of skilled labor. The disabled social protection programs: more specifically, is and ill are further constrained because the it possible to obtain insights into the optimal option of taking informal sector jobs (often in starting time and duration of such programs if the agriculture sector) is limited, both because the sole objective were to smooth household such work tends to be physically demanding and consumption? Finally, to better understand because informal sector employment, by defini- what is behind the observed variations in house- tion, is not bound by regulations and policies hold consumption, we investigate the impact of (such as tax incentives) designed to protect the newly acquired disability or disease on employ- vulnerable populations. ment and earnings. This chapter uses data from a four-wave panel survey of households in Bosnia and Herzegovina to assess the extent to which Bosnia and Herzegovina Context households are able to smooth the impact of unexpected health shocks on key welfare indica- Prior to the breakup of the former Yugoslavia tors. The factors that make individuals in tran- into six separate countries, the population of sition countries more vulnerable to health Bosnia and Herzegovina enjoyed a well- shocks clearly apply to the Bosnia and Herze- coordinated health system that had health indi- govina case. The health-welfare linkage may be cators similar to those in OECD countries especially strong given the fact that the country (World Bank 2005) and a reasonably high level has also experienced massive movements of of living. The 1992­95 war had two immediate population due to the war: more than 200,000 effects in the postwar period: per-capita income people are estimated to have died in the conflict fell from US$2,429 (in 1990) to US$456 by and close to 2 million people (half the prewar 1995, and the health care system, like many gov- population) was internally displaced or became ernment structures and services, became highly refugees. fragmented. As a result of the Dayton Peace The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 69 Accords, Bosnia and Herzegovina is a country to disabled veterans, as well as to their depend- made up of two entities--the Federation of ents and the dependents of soldiers killed in Bosnia and Herzegovina (FBiH); and the action. Finally, civilian victims of war are enti- Republika Srpska (RS)--and a central Bosnia tled to separate disability benefits. Each pro- and Herzegovina government of fairly limited gram provides significantly different benefits. (but increasingly expanding) authority. An addi- In particular, a disabled individual who is eligi- tional administrative unit, the District of Brcko, ble for veterans' benefits can be expected to has also existed since 2000. receive much more support than others who are The fragmentation of the political sphere similarly disabled. The following figures pro- translated into the fragmentation of the provi- vide a sense of the coverage and level of bene- sion of the health care system. Serious problems fits associated with each scheme. Almost affected the health care system in the postwar 100,000 individuals received benefits under period, weakening individuals' protection from contributory invalid pensions in 2001; war vet- health shocks. It was estimated that 30 to 40 erans and social assistance programs were the percent of the country's hospitals were next largest in terms of enrollments, with about destroyed or seriously damaged during the war, 85,000 and 36,000 beneficiaries, respectively; and that up to 30 percent of the prewar health followed by assistance to about 6,000 civil vic- care staff had emigrated or died in the conflict tims of war. The ratio of disability benefit to (World Bank 2005). The disjointed system that average wage was largest for the veterans' ben- resulted from the war led to low levels of risk efits (63 percent), followed by disability pen- pooling for health insurance funds (particularly sions (38 percent), civil victims of war (10 in the smaller cantons in the FBiH), unequal percent), and social assistance (6 percent). access to health care, lack of portability of health A recent study of individuals with disabilities, insurance, a large share of the population no their families, and others working to assist them, longer being covered by health insurance, and a demonstrates the impact of the conflict on the high rate of expenditures devoted to health health system (Prism 2006). Problems and care--12 percent of GDP, with one-third of this issues raised by the focus groups interviewed for being out-of­pocket expenditures (World Bank the qualitative study relate to the dropoff in ser- 2000, 2003). Even though individuals in Bosnia vice availability, the differential benefits for sim- and Herzegovina are covered by a mandatory ilar limitations or disabilities, and the cost of health insurance scheme, during the period health services. The prewar health care system under investigation, the health system did not appears to have been able to deliver a broader provide adequate coverage, and out-of-pocket range of services to individuals with health expenditures remained high due to a disconnect problems and disabilities. between the legislated entitlements and the If an employee becomes sick or disabled at governments' ability to collect taxes (European work, employers are required to assist them Observatory on Health Care Systems 2002). with rehabilitation and alternative job place- There are four types of disability support ments.3 There are also tax incentives to encour- systems in Bosnia and Herzegovina (see IOS age employers to employ individuals with 2003 for a more detailed description of the disabilities, although in an economic environ- social protection system). The social assistance ment where the share of informal sector system targets the poorest individuals who are employment is estimated to be 36 percent of disabled and unable to work. The social insur- total employment, and compliance with tax laws ance system offers protection to the disabled is believed to be low, the tax incentives are who contributed to the pension and invalid unlikely to produce the intended employment fund. The veterans system provides assistance outcomes for the disabled. 70 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Data ensure that only comparable variables are included in the analysis. A more detailed discus- The data used here are from the Bosnia and sion of the variables used is found in the follow- Herzegovina Living Standards Measurement ing sections. Study Survey (LSMS) and the follow-up panel survey called Living in Bosnia and Herzegovina (LBiHS). The LSMS was the first nationally Empirical Framework representative household survey capable of measuring welfare carried out in the newly The Model formed Bosnia and Herzegovina.4 The survey The model we use is as outlined by Gertler and was conducted by the three statistical organiza- Gruber (2002), where the change in labor sup- tions of the country--the State Agency for Sta- ply for individual i in community j is explained tistics, the Federation Institute for Statistics, by community fixed effects, j , changes in and the Republika Srpska Institute for Statis- health status, hij, individual and household tics.5 The first round of the LSMS was carried characteristics, Xijk, and a random error term ij. out in 2001, with a sample of 5,400 house- holds--3,000 in the Federation of Bosnia and Lij = +hij + j kX ijk + ij Herzegovina (FBiH) and 2,400 in the Repub- k lika Srpska (RS).6 A total of 16,976 people were Taking advantage of the presence of a four- interviewed. wave household survey (although the health- The purpose of the LSMS was to provide a related questions are only comparable in Waves money-metric measure of poverty for the coun- 1, 3, and 4), we can write the change in health try, to assess individuals' and households' access status as to and use of government services, and to docu- ment the population's economic activities and hij = hij 4 - hij3+ hij3- hij1. levels of human capital a few years after the war. The LSMS provides household-level data on a This gives a more flexible specification that wide range of individual and household charac- provides insights on the extent to which the tim- teristics that affect welfare.7 ing of the health shock matters. Making this dis- Three further rounds of data collection were tinction seems beneficial, because empirical carried out between 2002 and 2004 on a yearly evidence from the United States suggests that basis under the LBiH Survey. The LBiH disabled men experience sharp drops in formed a panel using a subsample of households expected annual earnings (caused mainly by a interviewed in the full LSMS of 2001. This reduction in hours, rather than changes in included all of the rural households from the wages) at around the time of onset of disability, original sample and approximately one-third of but that earnings recover relatively rapidly by the urban households for a total of 3,007 house- two years after the onset of disability--still, with holds. 8 The focus of the LBiH was on a nar- a long-term earnings loss of at around 12 per- rower set of issues--primarily labor and social cent per year (Charles 2003).9 protection--than the LSMS. Given the change in focus of the survey, not all of the same vari- Lij = +44hij +33hij + j k X ijk + ij ables, concepts, and indicators exist in the four k rounds of data collection. Nor, even when they are present, are the same definitions used. This We use a similar specification to explain complicates the analysis of health shocks under- changes in per-capita household consumption taken here. However, care has been taken to Cij over time. The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 71 Cij = +44hij +33hij +22hij Thomas 1995) or justifying inclusion in other j programs or receipt of benefits (Kerkhofs and + kX ijk + ij Lindeboom 1995). k These studies suggest the need to evaluate In an attempt to see if deterioration of the the robustness of the findings by considering head of the household's health has an impact on alternative proxies of poor health and disability. the schooling of children, we turn our attention Fortunately, the Bosnia and Herzegovina LSMS to children who were between ages 11 and 15 offers three quite different health measures, and who were enrolled in school at the time of which are comparable across survey waves: pres- the full LSMS (Wave 1). The dependent vari- ence of chronic illness, disability, and ADL. Six able, Sij, takes the value 1 if the child dropped ADL are included in the LSMS survey. How- out of school, and takes the value 0 if the child ever, only three of these were included in the remained enrolled in school. subsequent rounds of the LBiH survey. This is Sij = + 4 4hij + 33hij + 22hij + probably because an ADL index constructed by j these three variables explains 95 percent of the kX ijk + ij variation in an ADL index constructed with all k six activities. The ADL index is thus constructed Many of the studies of the impact of health from these three variables (whether the person's shocks have been limited to a single indicator health limited him or her from: vigorous activi- that happened to be captured by the data set ties such as lifting heavy objects, running, or used. The Bosnia and Herzegovina data set, in participating in strenuous sports; walking contrast, provides several indicators of disability uphill; or bending, lifting, or stooping). We use and poor health. We separately explore the link- two alternative specifications to capture changes ages of these various indicators with employ- in ADL. One indicator captures the changes in ment, earnings, and household consumption.10 all ADL (an index) between the first and fourth waves, while the other indicator focuses on Alternative Approaches to Measuring whether the head of the household acquired a Changes in Health Status limitation (for the first time) in the same time The model is designed to assess the impact of period. changes in health status on various facets of wel- Both the presence of chronic disease and dis- fare. However, measuring health status outside ability are captured with a simple question on of a clinical setting is fraught with difficulties. whether the person has a chronic disease (dis- Subjective assessments of health status (SAHS) ability). For disability, relying on such a simple have been shown to have predictive power for question may lead to underestimating the inci- morbidity and mortality (Mete 2005; van dence of disabilities. The format of the question Doorslaer and Gerdtham 2003; Idler and Kasl is expected to only capture those individuals 1995; McCallum et al. 1994). Yet, it has also with severe disabilities or those with access to been shown that an individual's self-ranking of disability benefits, i.e., have formal disability health status is affected by his or her own demo- status. While the WHO estimates that 10 per- graphic and cultural characteristics (Mete and cent of the population in developing countries Cnobloch 2006; Lindeboom and van Doorslaer suffers from disability (WHO 1980), results 2004; Van Doorslaer and Gerdthan 2003; from census and survey data with questions sim- Schultz and Tansel 2002). SAHS may also over- ilar to the ones used in the Bosnia and Herze- state the effect of health on the supply of labor govina panel show much lower estimates: data as individuals use health status as a way of justi- from 35 countries in the early 1990s from cen- fying labor decisions (Bound 1991; Straus and suses or surveys show only two countries with 72 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union incidence rates greater than 5 percent (United has a limitation in any of the three measures of Nations 2005). In all likelihood, different ques- ADL. The change shows the percentage of indi- tions would have captured a much larger group viduals who had a new limitation occur since the as improvements in questionnaire design have a first wave of the survey. Chronic illness is also significant impact on reported incidence.11 measured using a dummy variable, indicating Table 1 shows how these different indicators are the presence or absence of chronic illness. The related to one another. same is true of disability. For both of these last two variables, the change between Waves 1 and 4 indicates the percentage of individuals suffer- Results ing from a new chronic illness (disability) since the first wave of the survey. Analysis of the effects of health shocks on wel- fare is restricted by the nature of the panel data Household Head's Employment and set. As noted above, some of the definitions and Earnings variables varied by survey wave. The most con- Documenting the impact of health shocks on sistent approach is that followed here: looking employment prospects is critical both because at changes between the first wave of the LSMS those who become disabled need at least as (2001) and the fourth wave (2004). It is also pos- much earnings after the onset of a disability as sible to decompose the health condition before, and because being employed-- changes to those that occurred between Waves independent from the earnings from 1 and 3, and those that occurred between Waves employment--influences the quality of life for 3 and 4. This allows more insight into the effect individuals with disabilities. To be consistent of timing of shocks on welfare. with the previous literature, our discussion The means of the four health indicators used focuses on the heads of households (although are included in table 2. The first is the ADL we have explored trends in other adult house- index, which takes on a value of 0 to 1, indicat- hold members' employment and earnings as ing the share of limitations to ADL suffered by well; the results are mentioned in the text as the individual. The second ADL variable is a appropriate). Typically, a person is identified as dummy variable indicating whether the person the head of household by other household TABLE 1 Correlation among Indicators of Health and Disability Correlations (of levels) Wave 1 Chronic disease Disability ADL index Any limitation to daily activity Chronic disease 1 Disability 0.2174 1 ADL index 0.5744 0.2343 1 Any limitation to daily activity 0.5469 0.2009 0.8565 1 Correlations (of changes between Waves 1 and 4) Worsening of New chronic disease New disability ADL index New limitation to daily activity New chronic disease 1 New disability 0.0345 1 ADL index worsened 0.2456 0.1134 1 New limitation to daily activity 0.2022 0.0543 0.6801 1 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004). The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 73 TABLE 2 Household Heads: Sample Statistics, Means, and Standard Deviations Variables Period 1 (2001) Period 2 (2004) Change (2001­2004) Health status ADL index 0.367 0.451 0.108 (0.402) (0.435) (0.418) Any ADL 0.532 0.586 0.200 (0.499) (0.493) (0.400) Chronic disease 0.403 0.419 0.141 (0.491) (0.493) (0.348) Disability status 0.105 0.125 0.069 (0.306) (0.331) (0.254) Economic status Per-capita consumption (nonmedical) 3529.151 4262.487 760.695 (2220.522) (2294.596) (1908.045) Per-adult equivalent consumption (nonmedical ) 3665.542 4393.514 732.495 (2027.478) (2118.465) (1860.79) Head's monthly earnings 389.541 502.628 98.078 (361.523) (431.608) (389.613) Head's monthly earnings, per capita 121.700 153.133 26.645 (157.745) (136.446) (155.330) Head's monthly earnings, per adult equivalent 156.692 202.586 38.898 (179.318) (174.069) (181.480) Head's hours of work 40.034 37.862 -2.164 (17.222) (18.420) (21.993) Head not working 0.065 0.074 0.064 (0.246) (0.261) (0.245) Demographic characteristics Head's age 54.140 (14.081) Head is male 0.753 (0.431) Head has 0­4 years' education 0.262 (0.440) Head has 5­8 years' education 0.195 (0.396) Head has 9­12 years' education 0.435 (0.496) Head has 13+ years' education 0.108 (0.311) Head is married 0.694 0.681 (0.461) (0.466) Family structure Spouse's age conditional on head being married 48.344 (13.645) Household size 3.151 3.130 -0.077 (1.574) (1.608) (0.872) Children under age 6 0.248 0.166 -0.099 (0.562) (0.454) (0.456) Children between age 7 and 15 0.359 0.348 -0.038 (0.670) (0.671) (0.556) Household members older than 66 0.396 0.437 0.081 (0.649) (0.670) (0.389) Observations 2997 2416 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. Note: Household consumption excludes medical expenditure. 74 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union members because his or her earnings represent changes in earnings adjusted for annual the principal source of earned income in the inflation. household.12 Thus, a decline in this person's We regress the labor force variables on basic earnings and overall employment is expected to characteristics of the head of household: age, have a substantial impact on both the person's gender, marital status, and education level. own and the household's welfare. Additionally, we control for changes in the size We examine the impact of health shocks on and structure of the household. The results on three labor-related variables. The first specifica- hours worked are found in table 3. A compari- tion looks at the change in the number of hours son of the coefficients on each of the four health worked per week.13 The second dependent vari- indicators is found in table 4. able of interest is the probability of working for Declines in health are associated with a pay, a slightly restricted labor force participa- decline in the number of hours worked: all five tion model. In the third specification, we look at health measures show the same tendency, TABLE 3 The Effect of a Change in Household Head's Health on His or Her Hours Worked Alternative specifications (1) (2) (3) (4) (5) Head's ADL index worseneda 4.84 (2.058)*** Actual change in ADL indexb 5.642 (2.301)*** Newly acquired limitation in 1.018 daily activity (2.230) Newly acquired chronic disease 3.232 (2.792) Newly acquired disability 8.292 (4.540)* Head's age 0.213 0.301 0.583 0.242 0.469 (0.647) (0.645) (0.653) (0.640) (0.646) Head's age squared/100 0.260 0.328 0.608 0.293 0.562 0.653 (0.653) (0.661) (0.648) (0.657) Head is male 1.078 0.835 1.441 0.747 1.249 (4.114) (4.112) (4.255) (4.052) (4.042) Head has 0­4 years' education 8.731 8.672 11.711 10.004 11.693 (3.790)*** (3.789)*** (4.008)*** (3.715)*** (3.829)*** Head has 5­8 years' education 4.488 4.709 3.377 4.397 4.543 (3.041) (3.041) (3.219) (2.982) (2.991) Head has 9­12 years' education 4.964 4.788 4.314 4.739 4.669 (2.397)*** (2.396)** (2.508)* (2.378)*** (2.386)** Head is married 2.311 1.913 2.466 1.757 1.827 (3.666) (3.661) (3.853) (3.573) (3.554) Change in the number of children 2.487 2.497 2.422 2.451 2.201 younger than 15 (1.461)* (1.460)* (1.494) (1.444)* (1.462) Change in the number of adults 1.244 1.349 1.454 1.173 1.444 (1.206) (1.205) (1.262) (1.193) (1.210) Observations 656 656 656 673 653 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. Note: a/ This is a dummy variable taking on the value of 1 if the individual's ADL index worsened in any way over the time period, and 0 otherwise. b/ An increase in the value of this variable indicates a decline in health status. * Significant at .10 level; ** significant at .05 level; *** significant at .01 level. The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 75 TABLE 4 Effect of a Change in Household Head's Health on Labor Supply and Earnings Head's newly Head's newly Head's ADL Actual acquired acquired Head's newly ADL index change in limitation to chronic becoming worsened ADL index daily activity disease disabled Change in head's 4.840 5.642 1.018 3.232 8.292 working hours (2.058)*** (2.301)*** (2.230) (2.792) (4.540)* Observations 656 656 656 673 653 Change in head's labor 0.060 0.113 0.085 0.026 0.247 force participation (0.125) (0.115) (0.131) (0.155) (0.273) Marginal effects 0.004 0.007 0.005 0.002 0.013 Observations 1850 1850 1850 1880 1807 Change in head's 56.910 56.342 36.802 3.813 1.019 monthly earnings (35.959) (39.877) (34.068) (49.694) (79.828) Observations 517 517 517 530 518 Change in head's per-capita 38.661 48.569 32.661 7.484 12.032 monthly earnings (16.523)*** (18.286)*** (17.757)* (22.885) (36.804) Observations 517 517 517 530 518 Change in head's per adult 50.053 42.556 34.343 6.949 13.199 equivalent monthly earnings (20.275)*** (18.305)** (19.107)* (25.530) (41.044) Observations 517 517 517 530 518 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. * Significant at .10 level; ** significant at .05 level; *** significant at .01 level. although the results are not statistically signifi- hours worked. Yet none of the health shocks cant for two of them. A worsening in the ADL examined is closely linked to the probability of index decreases hours worked by almost five employment (table 4). This suggests that the hours per week. However, moving from no lim- legislation aimed at protecting jobs in the face itation to having any limitation in ADL has a of illness or disability is binding. Or, more gen- smaller effect, at about one hour per week, erally, the labor code is such that dismissal of which is not statistically significant at the 10 any employee is costly and time-consuming. percent level. The most substantial decline in Earnings are negatively and substantially hours worked is linked to the onset of disability. affected by a worsening in the ADL index or the For these individuals, the number of hours onset of a new limitation in ADL. Becoming declined by eight, or 20 percent of the average more seriously limited in ADL decreases the hours worked. The onset of a chronic illness is head of household's per-capita monthly earn- associated with a decline in hours worked, ings by approximately KM 39, from mean earn- although this is not a statistically significant ings of KM 89. A new limitation will decrease effect. This could be because many chronic ail- per-capita earnings by almost KM 32. The ments may have little impact on the ability to onset of chronic illness does not have a statisti- work and may more closely approximate the cally significant impact on earnings, again, per- minor illnesses that Gertler and Gruber (2002) haps reflecting the heterogeneity of chronic discuss.14 illness and the fact that not all will affect a per- The effects of health shocks vary not just by son's ability to work. Note that while the onset type of shock, but by type of labor force out- of a disability lowered the hours worked, it does come. As shown above, a worsening in ADLs or not lower earnings. This finding is consistent the onset of a disability lowers the number of with the possibility that disability labor legisla- 76 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union tion compensates for the negative effects of authors upon request; a similar trend is also deteriorating health status, at least in the short noted by Gertler and Gruber for Indonesia). term. Similarly, changes in household size and com- position before and after a health shock are neg- Household Consumption ligible. Increased receipt of remittances and The extent to which negative effects of health other private transfers is a further mechanism shocks on employment are reflected in con- for offsetting the impact of health shocks. sumption patterns of households depends on Unfortunately, the data on private and public whether the deterioration in the head of house- transfers are not comparable between Waves 1 hold's health triggers a response from other and 4 of the surveys, so we cannot estimate the household members in the form of increased changes in these variables. A simple comparison work hours and earnings. Other factors that of the difference in the receipt of public and pri- may prevent significant declines in per-capita vate transfers between households whose head household consumption include adjustment in is disabled or suffering ill health and all other the size and composition of households; households does show, however, a much higher increases in social protection or assistance ben- likelihood of receipt of transfers by the former. efits; and increases in transfers from friends and Close to 40 percent of all households whose relatives. head is disabled receive disability benefits, com- In the Bosnia and Herzegovina case, the data pared to 11 percent of those without a disabled reveal that other household members do not head of household.15 Other social assistance move between employment states in a manner (public transfers) does not seem to be targeted that can be linked to variations in the head of toward disability or sickness, but somewhat on household's health status (results available from worsening levels of ADL. In contrast, private TABLE 5 Comparison of Household Situations before and after Health Shocks ADL Index worsened? New ADL limitation? New chronic disease? New disability? No Yes No Yes No Yes No Yes Receive disability benefit 11.8% 17.3% 12.5% 18.1% 13.6% 13.4% 11.2% 37.0% Value of disability benefit 2,252.03 2,409.50 2,376.05 2,456.77 2,241.59 2,685.52 2,336.50 2,336.43 (2,156.69) (2,191.61) (2,168.98) (2,383.38) (2,161.36) (2,156.56) (2,197.60) (2,239.38) Receive non-disability social assistance benefit 34.1% 51.5% 39.7% 39.4% 38.2% 48.5% 39.4% 39.4% Value of benefit 2,369.07 2,486.62 2,339.09 2,545.24 2,460.77 2,309.76 2,481.43 1,822.66 (1,564.34) (1,722.33) (1,468.86) (1,901.73) (1,725.03) (1,253.92) (1,652.32) (1,416.13) Receive any social assistance benefit 42.4% 63.7% 48.4% 52.8% 48.0% 57.1% 46.9% 70.9% Value of benefit 2,517.48 2,657.08 2,521.27 2,739.93 2,581.21 2,594.43 2,631.11 2,203.39 (1,816.20) (1,949.41) (1,733.57) (2,204.91) (1,936.62) (1,604.21) (1,861.03) (2,015.39) Receive private transfers 9.8% 12.7% 16.6% 14.0% 14.3% 22.0% 14.3% 26.8% Value of private transfers 616.71 702.02 633.07 486.61 550.55 803.20 588.65 692.27 (714.27) (973.01) (716.54) (436.76) (556.07) (978.02) (643.90) (863.85) Household size 0.06 0.13 0.06 0.16 0.07 0.12 0.07 0.17 (0.78) (0.96) (0.83) (0.95) (0.84) (0.86) (0.83) (0.94) Observations 1,241 631 1,372 343 1,635 268 1,703 127 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. * Significant at .10 level; ** significant at .05 level; *** significant at .01 level. The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 77 transfers are targeted to households where the point decrease in per-capita household con- head of the household is disabled or suffers from sumption. Variations in the reporting of a chronic ailment. chronic illnesses do not influence per-capita The Bosnia and Herzegovina data show that household consumption. This finding is consis- such mechanisms fail to fully insure per-capita tent with the literature, which argues that household consumption against the onset of households might deal with minor health disability and deteriorating ADL. Per-capita events but are not able to cope with the eco- household consumption decreases by 7.8 per- nomic costs of major illnesses. Note that con- centage points, on average, if the head of sumption data used here do not include health household becomes disabled. There is no evi- expenditures. Previous work on disability in dence of differential effects for the poorest Bosnia and Herzegovina has shown that health individuals versus the rest.16 Similarly, worsen- expenditures by the disabled are higher than for ing of the ADL index leads to a 4.3 percentage others (Tsirunyan 2005): this spending may be TABLE 6 Effect of a Change in Household Head's Health on Per-Capita Household Consumption Alternative specifications (1) (2) (3) (4) (5) Head's ADL index worsened 0.043 (0.021)** Change in ADL index 0.003 0.024 Newly acquired limitation 0.022 to daily activity 0.026 Newly acquired chronic 0.013 disease 0.028 Newly occurred disability 0.078 (0.040)** Head's age 0.003 0.002 0.002 0.001 0.0001 0.006 0.006 0.006 0.006 0.006 Head's age squared/100 0.003 0.002 0.002 0.001 0.0003 0.005 0.005 0.005 0.005 0.005 Head is male 0.008 0.008 0.018 0.011 0.0110 0.037 0.037 0.038 0.036 0.037 Head: 0­4 years of education 0.013 0.011 0.019 0.001 0.0010 0.037 0.037 0.039 0.037 0.038 Head: 5­8 years of education 0.051 0.053 0.028 0.066 0.052 0.037 0.037 0.039 (0.037)* 0.037 Head: 9­12 years of education 0.038 0.039 0.026 0.050 0.046 0.033 0.033 0.035 0.033 0.033 Head: married 0.066 0.065 0.084 0.059 0.053 (0.035)* (0.035)* (0.036)*** (0.034)* 0.035 Change in the number of 0.182 0.182 0.19 0.178 0.175 children younger than 15 (0.018)*** (0.018)*** (0.018)*** (0.018)*** (0.018)*** Change in the number of 0.141 0.140 0.144 0.141 0.140 adults (0.014)*** (0.014)*** (0.015)*** (0.014)*** (0.014)*** Observations 1,843 1,843 1,843 1,873 1,801 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. Note: Consumption excludes medical expenditure. * Significant at .10 level; ** significant at .05 level; *** significant at .01 level. 78 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union crowding out consumption expenditures on reported), we see that a worsening of the house- other goods and services. hold head's ADL index increases the probability of male children dropping out of school by 9 Children's Schooling percentage points (the effect is statistically sig- To investigate the possibility that health shocks nificant at the 5 percent level). Similarly, if a influence not only household consumption lev- household head reports a new chronic disease els, but other aspects of household living stan- between Waves 1 and 4, boys' probability of dards and decision making as well, this section dropping out of school increases by 15 percent- presents predictors of school dropout (between age points (the effect is, however, only statisti- Wave 1 and Wave 4) among children ages 11 to cally significant at the 10 percent level). None 15. Table 7 shows that new chronic disease of the health shocks considered here has a sta- developed by the head of household is associ- tistically significant effect on girls' school ated with up to a 14 percentage point increase in enrollments. children's probability of dropping out of school. Boys are consistently about 5 percentage points DoesTiming of the Onset of Disability more likely to drop out of school than girls, or Poor Health Matter? although the difference is statistically significant By taking into account the time of onset of the in only one specification. When separate mod- health shock, we see that the labor market pro- els are estimated for males and females (not tections for those with disabilities may be short- TABLE 7 Marginal Effects of Changes in Household Heads' Health on their Children's Schooling (Ages 11 to 15) Alternative specifications (1) (2) (3) (4) (5) (6) Head's ADL index worsened 0.01 0.01 (0.03) (0.03) Actual change in head's ADL index 0.01 (0.03) Head's newly acquired limitation to 0.01 daily activity (0.03) Head's newly acquired chronic disease 0.13** 0.14** (0.07) (0.07) Head's newly occurred disability 0.06 0.06 (0.03) (0.03) Child's age 0.09 0.08 0.04 0.07 0.08 0.06 (0.38) (0.37) (0.38) (0.35) (0.37) (0.36) Child's age squared/100 0.00 0.02 0.16 0.03 0.02 0.07 (1.12) (1.11) (1.15) (1.05) (1.09) (1.08) Child is male 0.04 0.04 0.05* 0.04 0.04 0.04 (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Change in the number of children 0.02 0.02 0.01 0.02 0.02 0.02 younger than 15 (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Change in the number of adults 0.02 0.02 0.02 0.02 0.02 0.02 (0.03) (0.03) (0.02) (0.03) (0.03) 90.03) Observations 334 334 334 339 331 322 Sources: Bosnia and Herzegovina LSMS 2001 and LBiHS (Wave 4, 2004); authors' calculations. * Significant at .10 level; ** significant at .05 level; *** significant at .01 level. The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 79 lived and vary from those that exist for other abilities that began two to three years ago lead types of health shocks. Disabilities that occur to a decline of almost 17 hours of work per within the previous year do not have a signifi- week, close to half the average of a 40-hour cant, negative effect on hours worked, contrary work week (table 8). This drop in hours may to what one might expect. The sign of the esti- reflect an adjustment made in the workplace to mated coefficient is negative, but the impact is accommodate the person's disabilities and not statistically significant. But, if one looks at changed work ability, or it may simply reflect the impact of a disability that began two to three the person's inability to work more hours either years earlier, the picture is quite different. Dis- physically or because jobs are unavailable. TABLE 8 Effects of Decomposed Changes in Household Head's Health on Change in His or Her Hours Worked Alternative specifications (1) (2) (3) (4) (5) ADL index worsened: 8.38** 8.96** Wave 1 to 3 (2.80) (3.05) ADL index worsened: 9.54** 9.59** Wave 3 to 4 (3.20) (3.42) Newly acquired limitation 2.89 to ADL: Wave 1 to 3 (3.26) Newly acquired limitation 1.27 to ADL: Wave 3 to 4 (2.97) Newly acquired chronic 3.34 1.59 disease: Wave 1 to 3 (2.86) (3.08) Newly acquired chronic 5.59* 4.94 disease: Wave 3 to 4 (2.92) (3.11) Newly occurred disability: 17.70** 11.08 Wave 1 to 3 (8.94) (9.80) Newly occurred disability: 5.10 1.48 Wave 3 to 4 (5.23) (5.38) Head's age 0.28 0.20 0.17 0.44 0.35 (0.64) (0.73) (0.64) (0.65) (0.66) Head's age squared/100 0.34 0.11 0.24 0.55 0.47 (0.65) (0.74) (0.65) (0.66) (0.67) Head is male 0.26 5.36 1.17 0.76 0.02 (4.13) (4.80) (4.05) (4.13) (4.23) Head: 0­4 years of education 9.16** 13.05** 9.93** 12.02** 11.30** (3.78) (4.35) (3.71) (3.93) (4.03) Head: 5­8 years of education 4.63 4.41 4.28 4.68 5.24* (3.03) (3.61) (2.98) (3.04) (3.10) Head: 9­12 years of education 4.58* 4.09 4.87** 4.61* 4.84** (2.39) (2.78) (2.38) (2.40) (2.43) Head: married 1.51 5.36 2.06 1.49 1.62 (3.66) (4.41) (3.57) (3.66) (3.75) Change in the number of 2.57* 2.41 2.29 2.34 2.15 children under 15 (1.46) (1.63) (1.45) (1.48) (1.51) Change in the number of adults 1.33 1.26 1.09 1.71 1.67 (1.21) (1.40) (1.19) (1.24) (1.26) Sources: Bosnia and Herzegovina LSMS 2001, LBiHS (Wave 3, 2003, and Wave 4, 2004); authors' calculations. * Significant at .05 level; ** significant at .01 level. 80 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Surprisingly, for those whose ADL worsened distinguishes only between being employed or in the past year, the number of hours they not. Furthermore, by making a distinction worked increased, but the (expected) negative between health shocks that happened within the effect on hours of work occurs if ADL worsened last year and those that occurred three or four two to three years ago. It may be that individu- years ago, we see that the employment protec- als with a serious health shock, but not one that tion legislation for the disabled seems effective leads to formal disability status, have to work in preventing an immediate decline in weekly additional hours to partially make up for low- hours of work, but not for continued protection ered productivity. This is in contrast to those as disabilities that began two to three years ago who obtain disability status and are thus better are associated with a startling 17 fewer hours of protected from the loss of earnings or losing work per week. This is in contrast to the situa- their job in the short run. The data appear to tion in countries with weaker employment pro- show, however, that this temporary increase in tection legislation but lower unemployment hours is not sustainable in the medium term. rates, where a sharp initial decline in hours of In contrast to disability and worsening ADL, work is followed by a recovery period.18 In the the onset of chronic illness appears to be an Bosnia and Herzegovina case, the existing legis- event from which the individual is able to lation offers temporary relief, but after a one- recover. The negative effect of chronic disease year period the disabled individuals face an occurring recently (within one year) is statisti- inhospitable labor market environment where cally significant, lowering hours worked by just formal employment opportunities are scarce over five and a half. Yet, the effect of having a and informal employment is often physically chronic disease that began two to three years demanding. Overall improvements in labor earlier is not statistically significant. Thus, there demand would help to ease the longer-term dis- appears to be an ability to recover and adjust to advantages of disabled individuals and targeted the effect of chronic disease in the medium term. government interventions that provide training Taking into account the timing of health and matching for new jobs could be considered shocks on per-capita household consumption in this context, subject to thorough evaluations and children's schooling outcomes does not pro- of such programs on a pilot basis at first. duce statistically significant estimates on any of A natural next step in the analysis of the eco- the variables considered. While one of the two nomic implications of health shocks is to move ADL variables is associated with a reduction in away from what happens to the household head earnings if the limitations took place two to three and instead examine whether households them- years previously, the other ADL limitation vari- selves can deal with the deterioration of the able (the difference between ADL scores) is head of household's health. In Bosnia and linked to a positive change in earnings.17 Herzegovina, only major health shocks, as cap- tured by the onset of new disability, and to a lesser extent deterioration in ADL, have an Conclusions influence on per-capita household consump- tion, reducing it by 7.8 percentage points and This paper showed that individuals who experi- 4.3 percentage points, respectively. The way ence a health shock reduce their hours of work households (partially) smooth the effects of such significantly, although the likelihood of stop- health shocks also matters. In Bosnia and ping employment altogether is small. Thus it is Herzegovina, other adult household members possible to miss a large effect on employment-- do not respond to the head of the household's for example, onset of disability leads to an eight- deteriorating health condition (and reduced hour decline in weekly hours of work--if one number of work hours) by working more. This The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 81 may be because of lack of employment opportu- trast, using data from Indonesia, Gertler and nities in a high-unemployment environment, or Gruber (2002) find that households cannot because of the need to provide care to the dis- insure consumption against illness if one meas- ures health shocks as changes in an Activities of abled individual at home. Instead, in addition to Daily Limitations (ADL) index. social protection transfers, transfers from 2. Chetty and Looney (2005) show that evaluating friends and relatives come into play. At this the value of social safety nets in developing stage, public disability assistance is strongly countries by exclusively focusing on variations in biased in favor of war veterans, with much more household consumption levels may be mislead- ing, if households--especially those that are generous benefits to this group. Thus in terms close to a subsistence level of consumption to of the allocation of funds, the social protection start with--avoid a substantial consumption system is not designed to smooth the negative drop as a result of an unexpected shock by, for impact of nonwar-related disabilities (other example, taking children out of school or send- than the formal employment protection legisla- ing additional household members into the labor tion). This research highlights the need to bal- force. Thus, the authors argue that in such cases, social insurance could yield large welfare gains. ance the benefits designed for war veterans with Empirical evidence supporting this finding has those for other disabled individuals, since the started to emerge in a variety of countries, sug- short-term (formal) employment protection gesting that girls' schooling prospects may be alone is not sufficient to maintain the wealth particularly vulnerable to unexpected household levels of the latter group. shocks (see Lloyd, Mete, and Grant 2006 using data from Pakistan; Skoufias and Parker [forth- A by-product of health shocks to the house- coming] using data from Mexico). hold head that is often ignored is the implica- 3. The legal protection for disabled employees tions for the education prospects of children. states: "An employee who has been injured at The Bosnia and Herzegovina data provide weak work or has obtained a disease as a result of evidence on this, revealing that if a household work, cannot be dismissed by an employer while head reports a new chronic health condition, his he/she is medically unfit for work, regardless or whether the employee has made a contract with or her children are 14 percentage points more the employer for permanent or temporary likely to drop out of school compared to others. employment. If the Pensions and Invalids' Insur- In addition, male children's school outcomes are ance fund establishes him/her to have remaining influenced by a deterioration in household work capacity, the employer is obliged to offer head's ADL index in the expected direction. the employee a different position that is suitable to his/her remaining capacity... If an employer is However, this part of the analysis raises more not able to offer an appropriate alternative work questions than it answers, including why male position, then the employee has the right to children's school outcomes react to changes in claim benefits from the Pensions and Invalids the head of household's health, but female chil- Insurance." However, there are questions about dren's do not; and also why we see a statistically the extent to which the employment protection significant relationship when the health proxy is for the disabled is implemented in practice. 4. A Multiple Indicator Cluster survey was the first reporting chronic illnesses (and ADL in the case nationally representative household survey car- of male children), but fail to confirm such a link- ried out in the newly formed Bosnia and Herze- age when the health proxy is reporting disability govina. This was done by UNICEF in 2000, but status. focused largely on maternal and child health. 5. The first round of the LSMS received financial and technical support from the World Bank, Notes DfID, UNDP, the EU, and Sida. The subsequent waves forming a panel of a subset of the LSMS households was supported by DfID and was called 1. Townsend (1994) finds that the proportion of the Living in Bosnia and Herzegovina survey. the year an adult male is sick has no impact on 6. The sample excluded the Brcko district. consumption in three villages in India. In con- 82 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union 7. For more details, see State Agency for Statistics to completely sick lowers hours of work by et al. 2003. almost 31 hours per week. 8. The actual number of households interviewed 15. Another member of the household, not the varied by year as the panel followed individuals, head, is disabled. not households, and new household formation 16. Jalan and Ravallion (1999), for example, show that increased the sample, while deaths and emigra- in the poorest wealth decile, 40 percent of the tion decreased it. See "Field Work and Techni- income shock is passed onto current consump- cal Reports," (Birks and Sinclair, and the tion, while the richest third of households are Independent Bureau for Humanitarian Issues) protected from almost 90 percent of an income for further information on each wave of Living shock in rural China. The estimates for different in Bosnia and Herzegovina survey. consumption quintiles are not reported here; 9. Charles (2003) does not consider the extent to results available from the authors upon request. which households compensate for the initial 17. The estimates from these specifications are drop in earnings. available from the authors upon request. 10. The most commonly available question on 18. Presumably after an initial abrupt adjustment, health is whether a person was sick or injured in the disabled individuals eventually start working a specific reference period. But, as Gertler and more, perhaps through moving to a different Gruber (2002) demonstrate, not being able to field or sector. distinguish between minor and major illnesses is a drawback--one that can be partially overcome with an ADL index. Another indicator that may References be considered in this context is nutritional sta- tus--for example, Wagstaff (2005) uses a body mass index (BMI) as a proxy for health shocks. Beegle, Kathleen, and Wendy A. Stock. 2003. "The But, in this context, a nutritional measure is Labor Market Effects of Disability Discrimina- problematic. It is not a measure of disability: tion Laws." Journal of Human Resources 38 (4): many of the individuals who are suffering from 807­59. malnutrition may be functioning at reasonable Birks and Sinclair. 2005. "Field Work and Technical levels and their BMI may hit normal levels if and Report, Household Survey Panel Series Wave when they find employment (this is clearly not 4." Independent Bureau for Humanitarian Is- the case for most disabled individuals). Also, the sues Consultant Report for U.K. Department reverse causality issue, namely the possibility for International Development Project on La- that reduced consumption itself leads to lower bor and Social Policy in Bosnia and Herzegov- BMI rather than the other way around, is a more ina: The Development of Policies and Measures serious concern when BMI is used to capture for Social Mitigation. changes in health status. ------. 2003. "Field Work and Technical Report, 11. The revision of the Uganda census disability Household Survey Panel Series Wave 2." Con- question from the 1991 to the 2001 census sultant Report for U.K. Department for Inter- changed the estimate of disability among the national Development Project on Labor and oldest group from around 5 percent to more Social Policy in Bosnia and Herzegovina: The than 20 percent (Martinho and Banda 2005) and Development of Policies and Measures for So- a survey in Nicaragua in 2005 designed specifi- cial Mitigation. cally to measure disability gives an incidence rate Bound, John. 1991. "Self-Reported versus Objective of 11 percent (INEC 2003). Measures of Health in Retirement Models." 12. This was not, however, an official criterion of Journal of Human Resources 26 (1):106­38. being the head of the household: households Charles, Kerwin Kofi. 2003. "The Longitudinal were free to identify one member based on Structure of Earnings Losses among Work- whatever criteria they felt to be appropriate. Limited Disabled Workers." Journal of Human 13. The reference period for hours worked per week Resources 38 (3): 619­46. in each round of the survey is the week prior to Chetty, Raj, and Adam Looney. 2005. "Consump- the administration of the interview, not a fixed tion Smoothing and the Welfare Consequences reference period for all households. of Social Insurance in Developing Countries." 14. Gertler and Gruber (2002) document even Working Paper No. 11709, National Bureau of larger effects using Indonesia data, where mov- Economic Research, Cambridge, MA. ing from completely healthy in the ADL index Costa, Dora L. 2000. "Long-Term Declines in Dis- The Impact of Health Shocks on Employment, Earnings, and Household Consumption in Bosnia and Herzegovina 83 ability among Older Men: Medical Care, Public istrative Records." Presentation at the 4th Health, and Occupational Change." Working meeting of the Washington Group on Disabili- Paper No. 7605, National Bureau of Economic ty Statistics, Rio de Janeiro, Brazil. Research, Cambridge, MA. Mete, Cem, and Stefania R. Cnobloch. 2006. Crossley, Thomas F., and Steven Kennedy. 2002. "Socioeconomic Status and Health Outcomes. "The Reliability of Self-Assessed Health Sta- A Relationship in Disguise?" Manuscript. tus." Journal of Health Economics 21:643­58. World Bank, Washington, DC. Elwan, Ann. 1999. "Poverty and Disability: A Survey Mete, Cem. 2005. "Predictors of Elderly Mortality: of the Literature." Social Protection Discussion Health Status, Socioeconomic Characteristics Paper No. 9932, World Bank, Washington, DC. and Social Determinants of Health." Health European Observatory on Health Care Systems. Economics 14: 135­48. 2002. Health Care Systems in Transition: Bosnia Mont, Daniel. 2005. "Disability in Central Ameri- and Herzegovina. Copenhagen, Denmark: Euro- ca." Presentation at the 4th meeting of the pean Observatory on Health Care Systems. Washington Group on Disability Statistics, Rio Gertler, Paul, and Jonathan Gruber. 2002. "Insuring de Janeiro, Brazil. Consumption against Illness." American Eco- PRISM. 2006. "Social-Economic Status of Disabled nomic Review 92 (1):51­70. Persons." Final Report, World Bank, Sarajevo, Gertler, Paul, David I. Levine, and Enrico Moretti. Bosnia and Herzegovina. 2002. "Do Microfinance Programs Help Fami- Schultz, T. Paul, and Aysit Tansel. 1997. "Wage and lies Insure Consumption against Illness?" Un- Labor Supply Effects of Illness in Côte d'Ivoire published Manuscript. and Ghana: Instrumental Variable Estimates for Instituto Nacional de Estadísticas y Censos-Nicaragua. Days Disabled." Journal of Development Econom- 2003. "Encuestas Nicaragüense para personas ics 53 (2):251­86. con discapacidades, ENDIS." Managua Skoufias, Emmanuel, and Susan W. Parker. Forth- Nicaragua. coming. "Job Loss and Family Adjustments in IOS Partners. 2003. "Policies for Persons with Dis- Work and Schooling during the Mexican Peso abilities." Report prepared for PIU_PELRP, Crisis." Journal of Population Economics. FBiH Development and Employment Founda- State Agency for Statistics, Republika Srpska Insti- tion of Republika Srpska. tute for Statistics, Federation of Bosnia and Jalan, Jyotsna, and Martin Ravallion. 1999. "Are the Herzegovina Institute for Statistics. 2003. Living Poor Less Well Insured? Evidence on Vulnera- Standard Measurement Survey in Bosnia and bility to Income Risk in Rural China." Journal of Herzegovina. Sarajevo: State Agency for Statis- Development Economics 58: 61­81. tics, Republika Srpska Institute for Statistics, Kahn, Matthew E. 1998. "Health and Labor Market Federation of Bosnia and Herzegovina Institute Performance: The Case of Diabetes." Journal of for Statistics. Labor Economics 16 (4): 878­99. Straus, John, and Duncan Thomas. 1996. "Human Kerkhofs, Marcel, and Maarten Lindeboom. 1995. Resources: Empirical Modeling of Household "Subjective Health Measures and State Depen- and Family Decisions." in Handbook of Develop- dent Reporting Errors." Health Economics 4: ment Economics, ed. J. Behrman and T.N. Sri- 221­35. navasan, Amsterdam, the Netherlands. Lindeboom, Maarten, and Eddy van Doorslaer. Tsirunyan, Sasun. 2005. "Disability and Poverty in 2004. "Cut-Point Shift and Index-Shift in Self- Bosnia and Herzegovina: Findings Based on Reported Health." Discussion Paper TI-2003- Living Standards Measurement Survey." Con- 042/3, Tinbergen Institute. sultant Report Prepared for the World Bank. Lloyd, Cynthia B., Cem Mete, and Monica Grant. Sarajevo, Bosnia and Herzegovina. 2006. "The Implications of Changing Educa- Townsend, Robert M. 1994. "Risk and Insurance in tional and Family Circumstances for Children's Village India." Econometrica 62 (3): 539­91. Grade Progression in Rural Pakistan: ------. 1995. "Consumption Insurance: An Evalua- 1997­2004." Working Paper No. 209, Popula- tion of Risk-Bearing Systems in Low-Income tion Council Policy Research Division, New Economies." Journal of Economic Perspectives 9 York, New York. (3): 83­102. Martinho, Maria, and Jeremiah Banda. 2005. "Unit- United Nations. 2005. "Disability Statistics Data ed Nations Collection of National Disability Base." Statistics Division, http://unstats.un.org/ Statistics from Population and Housing Cen- unsd/demographic/sconcerns/disability/disab2 suses and Related National Surveys and Admin- .asp. 84 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Van Doorslaer, Eddy, and Ulf. G. Gerdtham. 2002. surance Technical Assistance Project, 2003." "Does Inequality in Self-Assessed Health Pre- Human Development Unit, Europe and Cen- dict Inequality in Survival by Income? Evidence tral Asia Region, Washington, D.C. from Swedish Data." Social Science and Medicine, ------. 2003. Uzbekistan Living Standards Assessment. 57: 1621­29. Vol. 2, Report No. 25923-UZ, Human Devel- Wagstaff, Adam. 2005. "The Economic Conse- opment Sector Unit, Europe and Central Asia quences of Health Shocks." Policy Research Region, Washington, DC: World Bank. Working Paper No. 3644, World Bank, Wash- ------. 2000. "Memorandum of the President of the ington, DC. International Development Association to the World Bank. 2005. "Bosnia and Herzegovina Case Executive Directors on a Country Assistance Study." in Review of Experiences of Family Medi- Strategy of the World Bank Group for Bosnia cine in Europe and Central Asia. Volume III, Re- and Herzegovina." Report No. 20592 Bosnia port No. 32354-ECA, Human Development and Herzegovina. Unit, Europe and Central Asia, Washington, World Health Organization. 1980. International Clas- DC: World Bank. sification of Impairments, Disabilities, and Handi- ------. 2003. "Bosnia and Herzegovina Social In- caps, Geneva: World Health Organization. CHAPTER 4 Health Disabilities and Labor Productivity in Russia in 2004 Health Consequences Beyond Premature Death T. Paul Schultz* Introduction reasons is likely to reduce the productivity of other members of the team, and thereby dimin- Health status can directly influence an individ- ish the employer's profits and reduce what the ual's capacity to work or the person's labor pro- firm is willing to pay for a worker who may be ductivity along three or more dimensions. unpredictably absent from work (Pauly et al. Health can affect the probability that a person 2002). Another spillover that is widely recog- participates in the labor force; health can influ- nized but rarely assessed arises from the burden ence the hours worked per week, month, or of health care borne by other household mem- year; and health can affect the worker's produc- bers, and might be approximated by the care- tivity, which is approximated here by the hourly giver's reduction of work in the market wage rate. Spillovers of a person's health on the economy, in-home production, and leisure economic output of society can also impact what activities. This within-household spillover on others can accomplish, although these spillovers time allocation might be assessed from labor are less often evaluated in empirical studies. For force surveys that collect health and labor sup- example, if the worker is engaged in a produc- ply information for all household members, tion team for which other workers within the assuming that the arrangement of households firm are less than perfect substitutes for the can be treated as exogenous (Bartel and Taub- worker, absenteeism from the job due to health man 1979; Thirumurthy et al. 2005). Finally, * T. Paul Schultz is a Malcolm K. Brachman Professor of Economics at the economics department of Yale University. I acknowledge the comments and suggestions of Cem Mete. The research assistance of Paul McGuire is greatly appreciated. The author's interpretation of the evidence and analysis are his own and not those of any other groups. 85 86 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union health status and disabilities are expected to substantial in survey-based health indicators. But influence interhousehold decisions determining finding such an instrument is not straightfor- whether an elderly or ill individual lives with an ward. Using more "objective" health conditions, adult child or other relative or is institutional- or preconditions in the past as a(n) indicator ized, when the individual is unable to perform (measurement for) of current disability or health his or her own ADL (Dostie and Leger 2005). If status, does not entirely solve the problem, information on health status is comprehensive although it may reduce it empirically (Stern and reliable, assessments of the economic con- 1998). Using lagged health status variables or sequences of health status and disabilities should fixed effects for the individual in the analysis of include not only physical disabilities and panel data over time is also not a complete chronic illnesses, but also mental conditions answer to the problem, if dynamic changes in with a permanent or transitory effect on eco- health are subject to accumulating measurement nomic and physical functioning in society (Sci- error (Griliches and Hausman 1986). ence, January 27, 2006). A second problem with drawing policy infer- A central problem in assessing the conse- ences in this context is that relatively "objective" quences of health on an individual's capacity to measures of health status, such as those that may work or social spillovers involves how to meas- be clinically observed by a health professional, ure health status as a condition that can be con- will not necessarily perfectly capture the health fidently treated as exogenous to the productive conditions that influence work capacity or the behavior of the individual, their household, and many related consequences of health. A body extended family members. As Bound (1991) mass index (BMI) that is considered obese, or a states, there are two problems in performing risk factor such as high blood pressure, may sig- such an assessment of health consequences that nificantly predict whether an elderly individual should be acknowledged from the outset. will work in the labor force or how productive Health status as self-reported by a respondent he or she will be, but these more objective indi- in a survey is likely to be influenced by the indi- cators of health will embody substantial classical vidual's preferences to work and other behav- measurement error. In addition, they will prob- ioral conditioning or cultural factors. This ably not reflect all health conditions relevant to could account for why self reports of poor a specific behavioral outcome. Both noisy meas- health status tend to rationalize an individual's ures of health status and an imperfect match own reduced labor supply decisions and related between the health indicator and the behavioral social arrangements of households, most likely outcome are likely to lead to an underestimate of contributing to an overestimate of the causal the causal impact of changes in imprecise meas- effect of health on capacity to work, or on the ures of health status on productive outcomes. costs of social spillovers. Instrumental variable estimates of self-reported This "endogeneity bias" associated with self- health effects on work capacity are likely to yield reported health status may be corrected if an different estimated effects of health, depending instrumental variable technique is employed and on which measures of health are considered and the specified instrument predicts the self- which instruments are specified to be influenc- reported health status, yet is not associated with ing the behavior of subpopulations--which the individual's current preferences for work or could have heterogeneous responses to health-- preconceptions of what differentiates between as hypothesized in the literature estimating local good and poor health. The instrumental variable average treatment effects (LATE) (Imbens and estimation techniques should also correct atten- Angrist 1994; Heckman 1997). For similar rea- uation bias caused by classical forms of measure- sons, different, more objective or externally val- ment error in health, which is typically idated measures of health will tend to involve Health Disabilities and Labor Productivity in Russia in 2004 87 less measurement error, and hence result in less circumstantial by modern economic statistical downward bias, when used to assess the conse- standards (Brainerd and Cutler 2004). With quences of poor health and disability on any hindsight, one is tempted to conclude that particular aspect of labor productivity or capac- women are less vulnerable than men to the ity to work. social-psychological adjustments required by Bound (1991) postulates that the effects of the Russian transition, or that the traditional self-reported health status on labor force attach- Russian habits of drinking and smoking exposed ment are upward biased, while more "objective" men--more than women--to the health risks health indicators yield downward biased esti- associated with these behaviors while coping mates, and together they might help to establish with these social and economic transitions. reasonable measures for true health effects. Many studies have sought to identify the Because it remains an empirical question whether causes of death that led to the unprecedented one source of bias is greater than the other, it is decline in life expectancy among Russian adults. useful to consider a variety of health status indi- These increases in mortality occurred primarily cators and instrumental variable strategies to from heart disease and strokes, lung diseases, assess the probable impacts of health disabilities alcohol consumption, and violence related to on individuals' economic performance. crime and accidents, while death rates due to cancer and childhood infections remained stable or declined gradually. From Russian time series Institutional Change and Uncertainties on death rates, which are thought to be reason- of the Transition: Mental Disabilities ably accurate, mortality increased sharply from 1992 to 1994, when the Soviet Union broke With the breakup of the former Soviet Union, apart and the Russian economy began to evi- adult mortality increased, especially among dence decline. Death rates then decreased from middle-aged Russian men, although the lack of 1995 to 1998, only to increase again after the improvement in male life expectancy after the devaluation crisis, when the Russian economy 1960s in Eastern European countries may have contracted further and economic inequality signaled growing health problems that preceded increased. Death rates for women age 30 to 55 the restructuring of the Soviet Union after 1989 followed the same pattern as for men, but with (Guo 1993). These extraordinary declines in life less amplitude. Similar, although less pro- expectancy among males can be decomposed nounced, increases in adult mortality occurred at into a sharp rise in deaths due to cardiovascular the start of the transition in neighboring states, disease, but also a disproportionate increase in such as Belarus and Ukraine, but not in Eastern deaths attributed to trauma, i.e., accidents, Europe or the Caucasus (Nolte et al. 2004). Gor- homicide, suicide, and alcohol poisoning. Some bachev's policies to raise the price and reduce have inferred that this mortality crisis is caused access to alcohol in Russia (1985­87) may have by the tensions and challenges posed by the contributed to the initial decline in mortality transition of the economy from centrally during the mid-1980s, but was not sustained planned to decentralized market oriented, with (World Bank 2005). The recent rise in oil prices a lack of accountability in governance; social has fueled the current economic expansion in disorder and dislocations; growing unemploy- Russia, although life expectancy for males may ment and unpredictability of life; and the large not have yet regained the levels achieved in the decline in income levels, at least as measured by 1970s or 1980s, or begun to catch up to that in Russian national income accounts. The assem- the European Union (EU). These studies of the bled evidence of social and economic causes for causes of mortality, however, do not assess mor- the increased mortality is convincing, but it is bidity or other aspects of health, nor do they link 88 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union these indicators of health and disability to their and measurement are recognized and corrected potential consequences on economic activity of in any analysis of the consequences of health sta- surviving individuals or the performance of the tus. Mete and Cnobloch (2006) show that moth- aggregate economy. An objective of this paper is ers' self-assessed reports of their children's to begin to explore the linkages between aspects health status in Uzbekistan are not closely of health and disability (as can be measured in a related to objective indicators of health, such as household representative survey) to various the child's stunting (i.e., short for age and sex), dimensions of economic productivity at the indi- wasting (i.e., low weight for height), anemia, vidual level. and hypertension. Lokshin and Ravallion (2005) There is little evidence that widespread mal- also find that self-assessed health status in Rus- nutrition or deterioration in diet was responsi- sia is not associated with income, whereas objec- ble for the rise in mortality. And the continued tive indicators of good health status often are gradual decline in mortality among children positively associated with permanent consump- under age 15, and the stability in mortality tion in the Russian Longitudinal Monitoring among the elderly, suggests the deterioration of Survey (RLMS). They caution researchers the Russian medical and public health system against relying directly on self-assessed health was not the primary cause for the rise in male status, and their findings suggest that instru- adult mortality. Environmental pollution may menting for self-assessed health status with have been a contributory factor, but not one more objective indicators can identify more reli- which could reasonably cause an upsurge in able components of health status, as concluded mortality among only men aged 15 to 65 (Nolte by Schultz and Tansel (1997). et al. 2004). One is left with nonquantified The assessment of the economic burden due social-psychological risk factors that resulted in to disabling diseases requires the estimation of stress, which may have been expressed in exces- how the condition independently affects the indi- sive alcohol consumption and risky behavior vidual's welfare, as well as his or her family and (Brainerd and Cutler 2004; Suhrcke et al. 2005). community. This assessment task is complicated Mental illness and disabilities could play a major by puzzles which probably stem from measure- role in the rise in mortality, and it might also ment difficulties. These are even more extreme affect morbidity and productivity, though this is in the case of health indicators derived from more difficult to document, especially in Russia, representative surveys (Strauss and Thomas where institutionalizing the mentally ill is com- 1995, 1998). Individuals of the same age and sex mon, and commitment to medical and psychi- in high-income countries report disability more atric institutions may have also served political often than individuals in low-income countries. objectives. This paper describes how health sta- This positive association between self reporting tus is related to disability within the noninstitu- a disability and increased longevity, and dimin- tionalized population, where disability is ished clinically confirmed morbidity, illustrates defined as being unable to participate in the the potential for subjective standards of what market economy and work for wages. It also people call "good" health to rise with economic considers how health status affects hours development faster than (actual) health condi- worked and the after-tax hourly wage rate. tions (Johansson 1991). Relatively little is known about the costs and Why is it Hard to Measure the effects of introducing preventive measures to Consequences of Disabilities? reduce the prevalence of disabling mental or Many studies confirm the divergence between behavioral illness, such as may have occurred in self-assessed and more objective indicators of Russia. The instrumental variable specified by health status, unless problems of endogeneity the researcher to predict the occurrence of ill- Health Disabilities and Labor Productivity in Russia in 2004 89 ness should replicate to the extent possible the pares to 64 percent of Russian females. Because pathway by which private and public resources an objective is to assess the consequences of could be employed to reduce these forms of ill- health on productivity, labor force participation, ness and disability and thereby improve health. hours worked, and wage rates, the distinction Evaluating policy options to prevent and ame- must also be drawn for those who report them- liorate the economic outcomes associated with selves working in the last 30 days, but do not disabilities requires monitoring, as well as pre- provide sufficient information about their jobs ventive and curative health interventions where to calculate an hourly wage and the number of their administrative and opportunity costs are hours worked. This "not reporting" category fully documented. This is not feasible in this could occur for a variety of reasons.3 Whatever study. the cause for not reporting, it is an issue for a The objective of this paper is more modest. substantial segment of the population in the A rudimentary framework is proposed to repre- RLMS for 2004, and the information is lacking sent how health is produced and health-related for nearly 15 percent of the males and close to behavior is determined, within which the eco- 10 percent of the females (figure 1). nomic consequences of adult disabilities are The subsequent analysis of the consequences assessed. Based on the 13th round of the RLMS, of poor health and disability on labor productiv- conducted in 2004, relationships are estimated ity focuses on four component outcomes: (1) that can be interpreted within this framework.1 overall rate of participation in the labor force First, descriptive ordinary least squares (OLS) (extensive margin of labor supply), (2) with par- regressions are reported between adult health ticipation in the "reporting" labor force, [and and productivity in Russia. Then, instrumental the logarithm of monthly earnings of the variable (IV) estimates are contrasted of these reporting group, which is the sum of] (3) the log same productivity-health relationships. These of the hourly wage (productivity), and (4) the are designed to mitigate specific sources of bias log of hours worked per month (intensive mar- in the estimation process due to error in meas- gin of labor supply) for persons with positive uring health status and the joint determination wages and hours. of health-related behavior, self-reported health status, and work behavior. Unobserved factors such as household endowments, preferences, Descriptive Statistics of the Russian and community infrastructure and prices may Survey Population also be responsible for causing the relationships between health and productivity.2 The nationally representative RLMS for 2004 Figure 1 illustrates that 42 percent of Russian samples about 4,000 households including females and 49 percent of Russian males age 15 2,953 men and 3,632 women between the ages to 24 are students. Only about 1 percent of of 25 and 64 for which the basic variables are those age 25 to 34 are students. Consequently, defined as tabulated in tables 1, 2, and 3.4 Table this study focuses on adults over the age of 25 to 1 reports the descriptive statistics on aspects of avoid dealing with the allocation of time labor productivity as inferred from this RLMS, between school and work, although both should in which columns 2 and 3 are the means for the probably be regarded as approximately of the total sample, and the last six columns report the same value in equilibrium, because the alloca- means for those in the reporting labor force tion of time to school or work is more or less (defined as those who report earnings and hours voluntary. For Russian males, 72 percent of on their primary or secondary jobs).5 Labor those between the ages of 25 and 64 report that force participation for men declines after ages they have worked in the last 30 days. This com- 25 to 34, whereas it peaks later for women, at 90 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union FIGURE 1 Economic Activity of Russian Population, 2004 0.9 0.8 0.7 n 0.6 latiou pop 0.5 of n 0.4 tior 0.3 oporp 0.2 0.1 0 15­24 25­34 35­44 45­54 55­64 65­74 75+ 15­24 25­34 35­44 45­54 55­64 65­74 75+ age female male student working, hours and earning reported working, without hours and earning reported Source: Authors' calculations from the 2004 RLMS. TABLE 1 Russian Longitudinal Monitoring Survey for 2004: Labor Market Participation, Hours and Wages by Age and Sex Proportion Total Proportion report Hours Hourly sample working in hours and per wage Monthly Log Log wage Log size 30 days earnings month rate earnings hours rate earnings Sample, sex and age [1] [2] [3] [4] [5] [6] [7] [8] [9] Males, ages: 15­24 495 .66 .54 190 35.7 6,261 5.20 3.27 8.47 25­34 969 .80 .66 187 45.3 7,537 5.18 3.49 8.67 35­44 814 .76 .64 189 45.2 7,511 5.18 3.43 8.61 45­54 736 .72 .60 187 39.4 6,827 5.19 3.30 8.49 55­64 434 .49 .41 166 43.2 5,890 5.04 3.37 8.41 65­74 391 .13 .12 164 24.0 4,061 5.08 2.96 8.04 75 or more 140 .029 .029 154 20.0 2800 5.00 2.88 7.88 Total 25­64 2,953 .72 .60 186 43.6 7,190 5.17 3.41 8.58 Females, ages: 15­24 618 .52 .47 164 27.7 3,994 5.02 2.99 8.01 25­34 1,071 .65 .58 162 31.0 4,567 5.03 3.09 8.12 35­44 903 .77 .67 166 30.3 4,734 5.06 3.08 8.15 45­54 1,021 .72 .63 164 36.6 4,741 5.02 3.14 8.16 55­64 637 .33 .31 163 27.0 4,202 5.03 3.04 8.07 65­74 783 .074 .066 141 24.6 3,272 4.87 2.89 7.76 75 or more 470 .002 .002 168 4.17 700 5.12 1.43 6.55 Total 25­64 3,632 .64 .57 164 32.3 4,636 5.04 3.10 8.14 Source: Authors' calculations from the RLMS 2004. Note: Variables defined and their sources in the survey questionnaire provided in appendix table A-1. The currency units are 2004 Rubles. The disability classifica- tion variable is defined in table A-1 and described in the text and in a footnote and is a recoding of the original variable which could not be directly employed. Health Disabilities and Labor Productivity in Russia in 2004 91 ages 35 to 44, when it marginally exceeds that hours are equally well compensated and work as for men, at 77 percent compared to 76 percent, many hours as those who do report earnings and respectively. The decline in participation after hours. This type of synthetic cohort aggrega- age 45 to 54 is substantial, dropping by a third tion of productivity also assumes that the time for men and by a half for women ages 55 to 64. of people spent outside of the labor force is not Russian women are eligible for pensions at age (economically) productive--an exaggeration to 55 and men at 60 (figure 2). be sure--but one that is implicit in national Hours worked by those in the paid labor income accounts. According to this heuristic force decrease for men [among paid workers] by calculation, the expected value of the monthly 11 percent from ages 25 to 34 to ages 55 to 64, earnings of a male Russian ages 25 to 34 is 6,030 and is essentially stable for women (table 1). rubles (i.e., .80* 7,537), and is 4,915 at ages 45 to Hourly wage rates for men also peak at ages 25 54, and then drops with the onset of retirement to 34, whereas they peak 20 years later for to 2,886 at ages 55 to 64.6 This decline in labor women. Thus, earnings for women in the paid market productivity could be due in part to a labor force increase until ages 45 to 54, while deterioration of health and vigor, or conversely, earnings for men fall steadily after ages 25 to 34. this decline in earnings could contribute to For both sexes, earnings are lower for those ages depression, which in turn induces them to 55 to 64 than for those ages 25 to 34, but the engage in unhealthy and unproductive behav- decline is 22 percent for men, and only 9 per- iors, such as working less, binge drinking, and cent for women. engaging in risky activities associated with mor- The product of the participation rate in the tality. Or declining productivity could be caused labor force (table 1, column 2) and the earnings by other factors changing across birth cohorts, of this group (column 6) may be interpreted as such as improvements among younger Russians an approximation of the average of the produc- in the quality and productivity of education. tive value of the time of a birth cohort supplied For Russian women, the life cycle decline in to the labor force, if one assumes that those in productivity occurs much later, after age 55, the labor force but not reporting earnings or when the labor force participation rate for FIGURE 2 Proportion of Russian Population with a Pension, 2004 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 15­24 25­34 35­44 45­54 55­64 65­74 75+ 15­24 25­34 35­44 45­54 55­64 65­74 75+ age female male Source: Authors' calculations from the 2004 RLMS. 92 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union women also drops sharply and they increasingly chronic health conditions, cross-tabulated by rely on pensions (figure 2). Thus, for women, age and sex. In the basic sample studied here, expected earnings increase 15 percent, from of persons between the ages of 25 and 64, the 2,969 rubles per month in 2004 for those age 25 likelihood that an individual reports any health to 34, to 3,414 rubles for those age 45 to 54, and problem in the last 30 days doubles from 22 then falls to 1,387 at age 55 to 64, when two- percent for men in the youngest ages to 44 per- thirds of these older women leave the labor cent for men in the oldest ages. Health prob- force and rely largely on pensions.7 The focus lems are reported more often for women than here is on the determinants of labor force par- men, increasing from 31 percent among the ticipation, working in a job for which earnings youngest group to 61 percent in the oldest and hours in the last 30 days are reported (i.e., a group (table 2, column 2). Self-assessed health wage job), and then the log of hours and log of status (from 0=very good, to 2=average, to the hourly wage rate among this final group in 4=very bad) increases for men over this same the "reporting" labor force.8 Hours are also age range from 1.5 to 2.1, while for women it expressed in logs to facilitate a parallel treat- is slightly worse, increasing from 1.6 to 2.2 ment to wages, which then allows an additive (column 1). decomposition of the log earnings into effects Nonetheless, the gender gap in life on proportionate variation in labor productivity expectancy has widened in Russia, with recent and labor supply. 9 age-specific mortality estimates indicating that Table 2 reports the sample averages for a on average, women lived 12 years longer than variety of measures of current and lagged men by 2000--perhaps the largest [recorded] TABLE 2 Indicators of Poor Health, Classified Disabilities, Chronic Health Conditions, by Age and Sex Health status Health Days missed (0­4, 0 = very problem (0­1 due to Disability Chronic disease condition (M20.6) good) in 30 days) illness classification Heart Lung Liver Sample, sex and age [1] [2] [3] [4] [5] [6] [7] Males, ages: 15­24 1.36 .202 .465 .044 .040 .034 .040 25­34 1.48 .224 .531 .048 .035 .037 .040 35­44 1.64 .272 .554 .095 .053 .043 .047 45­54 1.81 .294 .408 .222 .099 .053 .046 55­64 2.07 .441 .404 .373 .205 .092 .071 65­74 2.24 .550 .271 .413 .319 .141 .090 75 or more 2.55 .750 .157 1.059 .421 .158 .129 Total 25­64 1.70 .286 .488 .152 .081 .051 .048 Females, ages: 15­24 1.50 .295 .427 .039 .044 .023 .029 25­34 1.61 .313 .540 .043 .034 .029 .057 35­44 1.83 .352 .566 .068 .082 .042 .081 45­54 2.01 .504 .555 .153 .156 .066 .134 55­64 2.17 .612 .400 .268 .311 .078 .190 65­74 2.49 .753 .126 .515 .481 .093 .235 75 or more 2.79 .816 .021 .564 .519 .068 .241 Total 25­64 1.87 .429 .526 .120 .129 .051 .108 Source: Authors' calculations from the RLMS 2004. Note: Variables defined and their sources in the survey questionnaire provided in appendix table A-1. M20.6 refers to the source of the variable in the adult ques- tionnaire of the RLMS that asks about chronic illnesses.`` The disability classification is explained in table A-1 and text and footnote 11. It does not refer to percentage. Health Disabilities and Labor Productivity in Russia in 2004 93 discrepancy in death rates by gender that has 64 (table 2), or among those who are in the paid been noted in the world. Mortality of males labor force in these ages (not reported).10 The from 25 to 64 is fourfold higher than in the EU, administrative/medical classification of disabil- apparently due to cardiovascular illness, cancer, ity assigned to an individual distinguishes violence, and accidental death associated partic- among three disability groups (1 to 3, where 0 is ularly with alcohol consumption. For Russian not disabled), with the average numerical dis- women, mortality is roughly twice the EU level ability classification (0 to 3) increasing for men (Suhrcke et al. 2005; Nolte et al. 2004). After age 25­34 to 55­64 from .03 to .30, and increas- the Soviet Union separated, mortality in the ing for women from .03 to .22, across the same Russian Federation increased slowly and then age groups. 11 abruptly for adult men from 1992 to 1994, Specific chronic health conditions are whereas the increase was more moderate and reported more frequently among older individ- transitory for women. Meanwhile, mortality uals, and are more often reported by women among infants, children under 15, and elderly than by men, even for chronic lung diseases, over age 65 was stable or continued to decline despite the tendency for males to smoke four to gradually (World Bank 2005). The number of five times the number of cigarettes per day as days ill and unable to work in usual activities Russian women (table 3). A serious nervous dis- (not only labor force activity) in the last month order or depression in the year before the sur- averages about a half day for men and women, vey is self-reported by 14 percent of the men and this absenteeism level is not strongly related and 24 percent of the women between the ages to age among those between the ages of 25 and of 25 and 64, increasing sharply to age 35 to 44, Depression Chronic disease condition (M20.6) in last High blood Diagnosed Kidney Stomach Spinal Other 12 months pressure with diabetes [8] [9] [10] [11] [12] [13] [14] .051 .083 .071 .131 .131 .100 .006 .030 .090 .070 .093 .116 .184 .006 .037 .122 .121 .108 .146 .269 .012 .046 .184 .176 .172 .147 .354 .016 .071 .154 .199 .242 .145 .424 .044 .118 .203 .214 .290 .160 .545 .064 .080 .259 .207 .496 .175 .584 .058 .042 .132 .129 .139 .136 .285 .016 .081 .141 .062 .109 .219 .103 .006 .094 .120 .065 .157 .200 .191 .008 .100 .158 .143 .219 .243 .330 .017 .119 .192 .233 .280 .254 .532 .038 .186 .287 .270 .342 .269 .691 .110 .184 .305 .299 .381 .289 .752 .139 .137 .267 .274 .473 .240 .767 .145 .119 .179 .169 .240 .238 .409 .037 94 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union and thereafter increasing only gradually for the researcher. A negative self-selection might both genders. High blood pressure is more occur if individuals with latent health problems common with advancing age and higher for who are in worse objective health are more women than men, at 41 and 29 percent, respec- likely to see doctors for a medical diagnosis of tively, in the ages 25 to 64. It should be noted their problem in any specified retrospective that all of these health conditions, along with period, such as the last three months in the chronic health problems, may be more likely to RLMS. In addition, there is likely to be a sub- be diagnosed if the person consults medical pro- jective interpretation by the respondent as to fessionals and has regular contact with the whether his or her health condition constitutes health care system. The frequency of chronic a serious enough ailment to seek a medical illnesses and medically diagnosed diseases checkup, or to report in the RLMS as a chronic reported in a survey may therefore vary from health condition. It would seem likely that indi- actual prevalence rates. vidual heterogeneity in reporting health and The positive self-selection of individuals engaging in related behaviors would interject into the health care system may be greater for bias when estimating with single equation sta- those who assign a higher value to their health tistical methods (such as OLS) the causal effects and are more willing to make investments in among health behaviors, health outcomes, and their cumulated stock of good health human labor productivity (e.g., Rosenzweig and capital, many of which will not be observed by Schultz 1983). TABLE 3 Health-Related Characteristics and Behaviors, by Age and Sex Alcohol Medical Body Waist- Cohabiting Years of Cigarettes grams checkup Exercise mass to-hip Height and Sample, schooling per day per day 3 months frequency index ratio in cm married sex and age [1] [2] [3] [4] [5] [6] [7] [8] [9] Males, ages: 15­24 11.8 10.4 88.9 .168 .584 22.8 0.850 176.7 .337 25­34 12.4 12.0 107.2 .160 .496 24.7 0.881 176.4 .764 35­44 12.3 12.6 114.3 .125 .361 25.6 0.898 175.0 .854 45­54 12.2 11.9 104.5 .136 .333 26.1 0.914 173.2 .874 55­64 11.8 11.1 85.3 .118 .252 26.4 0.927 170.8 .885 65­74 9.46 6.48 62.2 .123 .487 26.5 0.938 168.4 .831 75 or more 9.52 2.67 36.20 .171 .360 26.5 0.938 167.6 .629 Total 25­64 12.2 12.01 105.2 .140 .382 25.6 0.900 174.4 .834 Females, ages: 15­24 12.8 3.04 38.30 .251 .434 22.4 0.774 164.0 .523 25­34 13.1 3.17 42.20 .229 .371 24.6 0.784 164.0 .763 35­44 12.7 2.89 43.20 .207 .342 27.0 0.803 162.7 .729 45­54 12.5 1.78 31.30 .211 .314 29.5 0.827 160.7 .689 55­64 12.1 1.08 23.30 .152 .446 30.0 0.844 159.7 .560 65­74 9.44 0.240 12.00 .105 .345 29.9 0.867 157.1 .377 75 or more 7.34 0.122 5.00 .070 .284 28.4 0.881 154.6 .143 Total 25­64 12.7 2.33 35.90 .205 .357 27.6 0.811 162.0 .701 Source: Authors' calculations from the RLMS 2004. Variables defined and their sources in the survey questionnaire provided in appendix table A-1. Note: The variable is described in footnote 14 and is not included in table A-1 because it was not used in the final model estimated, in part because it was unrelat- ed to all outcomes. Medical consultation in the last three months is the question and married is having a civil marriage, cohabiting is opposite sexes living together but not married to each other. Health Disabilities and Labor Productivity in Russia in 2004 95 Economic Determinants of Individual participate in the labor force and find employ- Labor Supply andWages ment, the researcher can replace the wage with Participation in the labor force and the number productive characteristics of individuals which of hours worked are generally assumed to be are generally positively correlated with their determined by the after-tax hourly wage oppor- wage offers, other things being equal, such as tunities available to individuals, and by other their years of schooling, years of potential expe- sources of income they have that are independ- rience after schooling, height, and weight for ent of their labor supply behavior and add to height (i.e., BMI). These productive character- their consumption opportunities. Transfer istics perform the function of "instrumental income, such as pensions and unemployment variables" for the potential wage offer, and benefits, are sources of other income which are therefore are expected to increase labor force affected by prior and current work behavior, and participation of the individual, but are not nec- possibly also by health status. Therefore, receipt essarily expected to be associated with increas- of these forms of transfer income is not strictly ing the number of hours worked, conditional on exogenous to the labor productivity outcomes the individual participating. examined here. Other income may, however, be For example, education is often associated generally associated with life cycle savings and with increased labor productivity and wage health. Controls are included for a smoothed rates, and is also positively associated with par- quadratic function in age plus a dummy variable ticipation but not always with hours worked that take the value of 1 when the individual among participants. Education is also closely reaches the pension-eligible age of 55 to 64 for linked to many indicators of health status and women and 60 to 64 for men, and is otherwise 0. good health outcomes, such as diminishing con- Nonearned income may represent income sumption of cigarettes and alcohol. Consump- from savings and physical wealth, and also may tion of alcohol by Russians is thought to be high arise from intra- and interhousehold private by international standards, and is often impli- transfers, as well as differential access to public cated in the frequency of deaths due to cardio- services and subsidies, such as community-pro- vascular disease, stroke, accidents, homicides, vided housing and social services. It is generally and suicides, but alcohol consumption has not expected that age eligibility for a pension or been treated as a factor in Russian labor produc- unearned income will decrease current labor tivity (Shkolnikov and Nemtsov 1997; Leon et supply, although some part of unearned income al. 1997; Ryan 1995).12 Illegal production of could be a choice, at least in terms of when an alcohol is common, especially in rural areas, and individual dissaves from their assets to boost thus self-reported alcohol consumption in sur- current income and consumption. veys, such as the RLMS, may be understated, Wage opportunities to work are expected to but nonetheless approximate consumption increase labor force participation because at the magnitudes. moment of entry into the labor force, this wage The safety net of social programs in Russia is effect does not exert an offsetting income or not precisely targeted to the least productive, wealth effect. However, in the determination of disabled, or poor, and may therefore not exert a hours worked conditional on participation, the differential or salient effect on who works and wage effect may be initially positive for part- how much they work.13 The actual receipt of time workers and possibly become negative pensions and other transfers are contingent on among workers whose income effect of the wage an individual's application and qualification, dominates the substitution effect (Killingsworth which may therefore depend on health and pro- 1983). Because the wage offered to individuals is ductivity, and also might be related to individual not observed for those who do not decide to preferences for leisure and subjective attitudes 96 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union toward health and work (Bound 1991). Thus, overstate the benefits of the care allocated ran- the actual receipt of public transfers or pensions domly in the population because those using the is viewed here as endogenous and not used to observed form of preventive care are likely to explain health and productivity outcomes. use other unobserved health services and the Unearned income is not directly reported, estimated relation would attribute the effect of but is constructed by subtracting from total the omitted variables to the observed variable. individual income (adult questionnaire, J60) the On the other hand, if the visit to the doctor was earnings of the worker on their primary (J10) for curative health care, it is likely that the cor- and secondary (J40) jobs, other earnings (J57), relation between the visit and health outcomes and transfer income in the form of pensions and productivity would be biased in a negative (J76) and unemployment benefits (J89). This direction, or downward (Rosenzweig and measure of unearned income is 418 rubles per Schultz 1983). In these cases, a community month for the average woman age 25 to 64, or questionnaire measure of the price of a doctor 11 percent of their total monthly income, and consultation visit or the time costs required of 438 rubles for men in these ages, which repre- the patient to gain access to doctors and medical sents 7 percent of their total income (appendix care should be used to predict who does and table A-1). As an exogenous constraint on who does not visit a doctor. As a result,these choice, "nonearned" income would decrease community instrumental variables would not be labor force participation and possibly hours correlated with an individual's desire for pre- worked among participants, and increase the ventive care, or the likelihood that the person is demand for normal goods, including preventive sick and in need of curative care. To determine health care services, other things being equal. medical care usage, the community question- Three health-related inputs are analyzed as naire of the RLMS provides the time required determinants of Russian labor productivity: to reach a hospital and a private doctor for each having a medical checkup in the last three of the 159 survey sample clusters. It also consid- months, consumption of grams of ethanol alco- ers whether the community has a social welfare hol per day, and number of cigarettes smoked office, and whether the office provides medical per day.14 What is more distinctive about this services (table A-1). analysis is that these inputs are treated as poten- Alcohol consumption is assessed in the sur- tially endogenous and correlated with the errors vey in five different forms and is converted here in the health outcome and productivity equa- into total grams of ethanol equivalent consumed tions. Thus, they are estimated using two stage per day in the last 30 days. Alcohol consump- least squares (2SLS), which are identified on the tion tends to peak for men and women during basis of exclusion restrictions. An indicator of the ages of 35 to 44, when men, on average, con- access to or use of medical care in the RLMS is sume 114 grams per day, while women consume whether the respondent has seen a doctor in the about a third as much, or 43 grams.15 It is last three months. About one in five women and believed that moderate alcohol consumption one in eight men indicate having had such a provides a protective effect against some cardio- medical consultation, although it is unclear vascular diseases and related mortality, whereas from the RLMS whether the visit is for a pre- higher levels of consumption are harmful. Binge ventive checkup, or due to an illness and the drinking, defined in some cases as five or more need for diagnosis and curative care. drinks (or 80 grams ethanol) on one occasion, As a form of preventive health care, this causes health problems, especially when the measure of utilization might be positively alcohol is consumed in the form of spirits that related to a person's good health status. Its cor- are absorbed more rapidly (Kauhanen et al. relation with health and productivity is likely to 1997). However, the composition or frequency Health Disabilities and Labor Productivity in Russia in 2004 97 of "binge" drinking, which may contribute to nous and identify it with the community price of health problems and violence, is not reported in low-quality cigarettes. Of the 159 sample cluster the RLMS, and thus only the quantity con- sites in the working sample, information on the sumed on average for the last 30 days is price of low-cost vodka (the dominant source of included. alcohol) and low-cost cigarettes is reported in The nonlinearity of the effect on health and 137 and 135 community questionnaires, respec- productivity is approximated by estimating the tively. In the sample sites not reporting their own effect of a linear and squared consumption term, prices for vodka and cigarettes, the prices in the for which significance levels must be evaluated closest neighboring sample sites are used in this jointly. Empirical evidence of this nonlinear analysis. effect of alcohol on health outcomes is widely Other characteristics possibly related to documented in the health literature, but alcohol health status and productivity are reported in consumption is generally assumed to be exoge- table 3 and appendix table A-1. Schooling has nous, even though it is widely recognized that been relatively constant at 12 years for Russians unobserved individual factors may affect both born after the Second World War, increasing alcohol consumption and other welfare out- slowly among younger adults, with women comes, including health and labor productivity. receiving slightly more years of schooling than Mullahy and Sindelar (1991) relate alcohol con- men, but occupying less well-paid professions sumption to labor force participation and and specializations, which could help to explain income in the United States, and show that the why women's wages are about one-third lower relationship of alcohol consumption to partici- than those of men (column 8, table 1. See also pation among middle-aged women and men is Glinskaya and Mroz 2000).16 I distinguish negative and larger for women, whereas the link among years of regular schooling through grade between alcohol and household income is 12, vocational school, technical post-secondary weaker, especially for women. Tekin (2002) school, and university and post-graduate educa- shows an inverse U-shaped effect of alcohol tion, as defined in appendix table A-1. consumption on employment (but not wages) in Adult height is one characteristics of individ- the RLMS for males, and a positive effect for uals which is significantly associated with lower females, though these patterns are weakened risks of mortality within age and sex groups, and when fixed effects for the individual are also associated with increased labor productivity included to correct for unobserved persistent (Waaler 1984; Fogel 2004; Schultz 2005). Adult individual heterogeneity. Baltagi and Gek- height is relatively stable for an individual from shecker (2006) use a panel of observations from age 25 to 50, and then declines only gradually as the RLMS to model the effect of changes in the vertebra become more compressed with later price of alcohol on the levels of consumption age. Adult height reflects genetic potential to be and lagged consumption, and identify negative tall, but is also responsive to developing in a consumption responses to prices. In this study, I more healthy uterine environment and early allow for a flexible quadratic effect of alcohol child development, including reduced exposure consumption on health and productivity out- to childhood infections (Barker 1994; Manton comes, assuming the local prices of low-quality 1986; Bengtsson et al. 2004). According to the vodka identify the endogenous consumption RLMS, height has increased markedly among level of alcohol. younger Russian men (table 3, column 8) age 25 Although there is less evidence of nonmonot- to 34 who are 176.4 centimeters tall, whereas onic effects of tobacco consumption on health or those who have survived and are over age 75 are productivity than alcohol, the same specification only 167.6 centimeters.17 Of this increase in is used to treat cigarette consumption as endoge- height of 8.8 centimeters in those Russians men 98 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union born in about 1975 compared to those born 50 the person is currently married, cohabiting but years earlier in about 1925, it appears that early unmarried, and divorced, where being single or childhoold nutritional and health conditions widowed is the excluded category. Table 3 have improved substantially. Among Russian reports how the percentage of married plus women in 2004 their height among those age 25 cohabiting varies by age and sex within the sam- to 34 is 9.4 centimeters larger than among ple. Finally, residents in rural areas represent women over the age of 75, and within the age about 24 percent of the RLMS sample, and an span represented in my sample women born in additional 6 percent reside in what are called 1975 are about 4.3 centimeters taller than those secret cities (posyolok gorodskogo tipa, or PGT), born in about 1925. A monotonic relationship which were omitted from standard maps until between height and relative risks of mortality is 1995, and might have special attributes such as reported in Norway in the 1970s and mortality pollution or a restrictive labor market with mil- risks appears to decline at an approximately lin- itary facilities or defense industries. The 70 per- ear rate with increased height for both men and cent of survey respondents residing in the women. Therefore, in this analysis height is remaining urban areas are the excluded category treated as a linear determinant of health as well in this case. Ten regions of the country are also as labor productivity in Russia for both men and included as controls, where the city of Moscow women, and when a quadratic term in height is is the excluded category (see table A-1). included in the subsequent regressions it was not statistically significantly different from zero (Waaler 1984; Fogel 1994). A Conceptual Framework to Guide the BMI, or weight in kilograms divided by Econometric Analysis height in meters squared, is considered "over- weight" following conventions established by Five classes of variables will be analyzed from the WHO if BMI exceeds 27 (table 3). Accord- the 2004 RLMS. The first set will be referred to ing to this indicator, Russian women are more as P, for the market productivity of the adult often overweight than men, with the proportion population between the ages of 25 and 64, rising to 70 percent of Russian women age 55 to which includes: (1) participation in the labor 64, and 39 percent of men in this age group. force in the last 30 days; (2) whether the indi- BMI as a factor determining health and labor vidual reports both earnings in this job and productivity is represented in the form of the hours worked in the job (or is in the "reporting" inverted U-shaped profile (Costa 1996) follow- labor force); (3) the natural logarithm of the ing Waaler (1984) and Fogel (2004), who fit rel- hourly wage in current rubles; and (4) the loga- ative mortality risks to BMI in a U-shaped rithm of the hours worked in the last 30 days. relationship. Here a quadratic function in BMI The last two variables are analyzed only for the is estimated, allowing the Russian data to sug- sample with positive hours and wages, or in the gest at what level increased BMI reduces indi- "reporting" labor force (table 1). vidual health and productivity. Another The second and third sets of variables indi- anthropometric indicator that is predictive of cate health status, and are divided into those cardiovascular health problems is the waist-to- that are measured contemporaneously and sub- hip ratio, for which Russian males exceed jectively, and those representing lagged stocks females (table 3). of health human capital. The current health sta- Marital status is often correlated with labor tus, HC, includes: (1) a self evaluation of cur- supply behavior and even wages. Assuming mar- rent health status from 1 to 5, with 1 being very ital status is exogenous to health and productiv- good and 5 being very bad; (2) whether the indi- ity, dummy variables are included for whether vidual has had a health problem in the last 30 Health Disabilities and Labor Productivity in Russia in 2004 99 days; and (3) the disability classification of the cluster, Cc, of which there are 161 sites distin- individual (0­3), which is an official, medically guished here, and the community question- confirmed designation, with a higher value indi- naire provides control variables for the: (1) cating someone is more severely disabled. The distance to the nearest hospital; (2) distance to third set of lagged health status indicators, HL, the nearest private doctor in the administrative may be measured with less subjective error than area; (3­4) whether there is no response to the current indicators of health status because these questions in the community site ques- they refer to chronic illnesses, diseases, or con- tionnaire, and the community (low-quality) ditions that are thought to have a persisting price for (5) vodka and (6) cigarettes.22 The effect on health, including: (1­7) chronic health third set of control variables, Cr, refers to the conditions or diseases; (8) nervous breakdown 10 regions of the Russian Federation for which or depression in the last year; and (9) high blood fixed effects are included, thereby estimating pressure, as summarized in table 2.18 A critical regional deviations from the omitted region, assumption made in this investigation is that which is the city of Moscow. All variables are HC may be treated as endogenous or measured defined in terms of the coded variables in the with error, whereas HL may be treated as RLMS in appendix table A-1. exogenous to the current health state, HC, or Reduced-form linear relationships could be economic productivity, P. estimated among the four productivity variables The fourth set of variables refer to health- and all of the control variables and the lagged related inputs or behaviors, I, which include: (1) chronic health conditions: whether the individual had seen a doctor in the last three months; (2) the number of cigarettes P = P (Ci ,Cc , Cr , HL, e1), (1) consumed per day: and (3) the grams of pure alcohol equivalent consumed per day in the last where e's represent errors that embody the 30 days (table 3).19 effect of omitted variables, errors in functional The fifth set of variables refers to three form, measurement error, and other forms of classes of controls, C. The first set of control presumably random variation. The allocation variables relates to the individual, Ci, such as: of inputs to increase health and productivity of (1) age, approximated by a quadratic function; the individual may respond to the individual's (2) dummy variables for the ages when regular heterogeneity in preferences for health, differ- pensions are available for women and men in ences in the unobserved quality of health care, Russia;20 (3) education, approximated by a or unobserved latent health endowments of the spline in different levels and types of schooling individual. This individual heterogeneity is (years of grade school, vocational, technical, or likely to contribute to a correlation between university training); (4) unearned income (i.e., HC and I and the residual error e1. Conse- total income minus earned and transfer quently, estimating directly in equation (1) the income); (5) height in centimeters; (6) BMI relationship between productivity due to cur- (i.e., weight in kilograms compared to height rent health status, HC, or health inputs and in meters squared), and its square; (7) waist-to- behavior, I, would not yield estimates of the hip ratio; (8­10) marital status (married, causal effects of these health conditions or cohabiting, divorced, with the excluded cate- health inputs for an average individual, as dis- gory being single), (11) resident of a rural or cussed below (Rosenzweig and Schultz 1983). urban-type settlement; or (12) in a secret city However, reduced-form equations can also be (i.e., PGT).21 The second set of control vari- estimated for current health input demands ables represents constraints due to the com- and current health status indicators using munity, as measured at the level of the sample OLS: 100 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union I = I (Ci ,Cc ,Cr , HL, e2), (2) cal significance of the identifying variables. The Durbin-Wu-Hausman specification test is then HC = HC (Ci ,Cc, Cr, HL , e3). (3) interpreted as a test of the exogeneity of HC in the health production process, and reported in To assess how current health inputs and cur- appendix table A-3. The first two rows of the rent health status affects productivity, structural male and female productivity estimates in table assumptions are required. It is assumed that 5 report the estimates of equation (4) based on community health care facilities and local the other potentially endogenous current health prices, Cc, as well as lagged chronic health con- status variables, "having a health problem in the ditions, HL, influence productive outcomes last 30 days," and "being assigned an ordered only through their impact on a single current disability classification," without health inputs health status indicator for the individual: as shown in tables 4A and 4B. To save space, only the coefficients estimated for these alterna- P = P ( Ci, Cr , HC*(Cc, HL), e4), (4) tive current health HC indicators are reported in table 5, because the other coefficients on the where HC*(Cc, HL) denotes that the effect of exogenous variables in tables 4A and 4B con- one of the HC variables on labor productivity trolling for the self-evaluation of health status may be estimated by two-stage least squares tend to closely parallel those for the other cur- (2SLS), relying on the exclusion restriction of rent health indicators. Cc and HL not directly entering the productiv- The productivity equations are then reesti- ity function. These health production and pro- mated, including as determinants the three ductivity functions (4) are first estimated by health-related input variables: OLS under the assumption that the current measures of health, HC, are exogenous. P = P (Ci, Cr , HC*(Cc, HL), (5) Because these closely related dimensions of I*(Cc, HL), e5), health are highly correlated, only one of these three health indicators is considered at a time. where the marginal effects in individual produc- These OLS estimates are reported to estimate tivity due to the use of health inputs is first esti- likely endogenous risk factors affecting health mated by OLS subject to the strong assumption and productivity. The second set of estimates that the health inputs are exogenously allocated relies on 2SLS, which treat current health meas- across individuals, and then by 2SLS, where ures as endogenous or correlated with classical estimates are identified by the exclusion restric- measurement error in e4, and identify these tions embodied in Cc and HL. The first-stage estimates by including Cc and HL as determi- predictive power of the instruments is reported nants of HC, but excluding them from directly in table A-2, and the Durbin-Wu-Hausman test entering equation (4). of the exogeneity of the inputs is reported in These OLS and 2SLS estimates of the coef- table A-3.23 ficients of HC are reported in the first row of tables 4A and 4B for males and females where the "self-evaluation of current health status" is Empirical Findings treated as the indicator of HC, excluding the potentially endogenous three health inputs: Familiar empirical regularities emerge in tables medical care, alcohol, and cigarettes. The F test 4A and 4B, some of which are analogous to of weak instruments from the first-stage esti- those found either in studies of labor supply--in mates of the determinants of HC is reported in terms of participation probabilities and hours appendix table A-2, indicating the joint statisti- worked (Killingsworth 1983), or in the litera- Health Disabilities and Labor Productivity in Russia in 2004 101 TABLE 4A Labor Force Productivity Determinants, Including Current Self-Evaluated Health Status: Males (Beneath coefficient is absolute value of t ratio in parentheses) Reporting hours ln hours in Working in 30 days and wage 30 days ln wage rate Dependent variable OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS estimation method 1 2 3 4 5 6 7 8 Variables: Self-evaluation of 0.0888 0.146 0.0766 0.125 0.0119 0.0684 0.132 0.0245 Health (0-4) (6.93) (5.33) (5.47) (4.17) (0.73) (1.69) (3.91) (0.29) Age (x10-1) 0.0149 0.0189 0.0728 0.0761 0.121 0.142 0.0798 0.0516 (0.20) (0.25) (0.87) (0.91) (1.39) (1.61) (0.44) (0.28) Age squared 0.0063 0.0055 0.0134 0.0127 0.0168 0.180 0.0172 0.0157 (x10-2) (0.68) 0.59 (1.32) (1.25) (1.58) (1.67) (0.78) (0.71) Pension age 0.298 0.298 0.200 0.200 0.166 0.157 0.0237 0.0124 (5.78) 5.76 (3.55) (3.54) (2.22) (2.09) (0.15) (0.08) Schooling years Grade school 0.0141 0.0136 0.0156 0.0152 0.0035 0.0032 0.0232 0.0236 (2.06) (1.97) (2.08) (2.02) (0.39) (0.36) (1.27) (1.29) Vocational 0.0122 0.0124 0.0100 0.0102 0.0080 0.0074 0.0291 0.0236 (1.69) (1.72) (1.27) (1.29) (0.96) (0.87) (1.67) (1.29) Technical 0.023 0.0227 0.0168 0.0164 0.0097 0.0097 0.046 0.0459 (3.84) (3.75) (2.54) (2.48) (1.45) (1.43) (3.30) (3.28) University 0.0136 0.0131 0.0116 0.0112 0.0012 0.0007 0.0560 0.0565 (3.58) (3.75) (2.80) (2.68) (0.28) (0.18) (6.62) (6.66) Unearned income 0.0629 0.0595 0.114 0.111 0.0144 0.0241 0.0190 0.0320 (x10-4) (2.82) (2.65) (4.69) (4.55) (0.29) (0.47) (0.18) (0.30) Height (cm) 0.0020 0.0018 0.0025 0.0024 0.0010 0.0010 0.0051 0.0052 (1.63) (1.50) (1.92) (1.82) (0.76) (0.72) (1.80) (1.82) Married 0.153 0.152 (0.166) 0.166 0.0195 0.0216 0.158 0.156 (5.43) (5.39) (4.69) (5.38) (0.55) (0.61) (2.16) (2.12) Cohabiting 0.132 0.133 0.126 0.127 0.0017 0.0047 0.077 0.0730 (3.90) (3.92) (3.42) (3.44) (0.04) (0.11) (0.90) (0.85) Divorced 0.026 0.021 0.034 0.0298 0.065 0.065 0.034 0.035 (0.67) (0.53) (0.79) (0.69) (1.27) (1.25) (0.32) (0.32) Body Mass Index 0.0502 0.0397 0.0486 0.0397 0.024 0.0091 0.354 0.0484 (x10-1) (3.16) (2.40) (2.80) (2.20) (0.69) (0.47) (0.90) (1.19) BMI squared 0.0796 0.0615 0.731 0.0579 0.136 0.0059 0.0532 0.0769 (x10-2) (2.78) (2.07) (2.34) (1.79) (0.87) (0.17) (0.75) (1.05) Waist-hip ratio 0.309 0.275 0.362 0.332 0.136 0.107 0.0476 0.0854 (2.39) (2.09) (2.56) (2.33) (0.87) (0.68) (0.15) (0.26) Rural resident 0.110 0.116 0.199 0.204 0.134 0.0164 0.571 0.568 (5.49) (5.71) (9.06) (9.19) (0.56) (0.66) (11.1) (10.9) Secret city (PGT) 0.151 0.153 0.254 0.257 0.0455 0.0427 0.333 0.329 (4.25) (4.31) (6.54) (6.58) (0.95) (0.88) (3.33) (3.28) Intercept 0.0954 0.11 0.337 0.164 4.50 4.66 2.30 2.08 (0.27) (0.31) (0.88) (0.42) (10.8) (10.9) (2.65) (2.35) R squared 0.181 n.a. 0.175 n.a. 0.033 n.a 0.265 n.a. F 21.9 .21.1 21.00 20.5 2.10 2.15 22.4 21.7 p-value (0.0001) (0.0001) (0.0001) (0.0001) (0.0008) (0.0006) (0.0001) (0.0001) Note: All regressions include regional dummies as defined in table A-1. 102 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE 4B Labor Force Productivity Determinants, Including Current Self-Evaluated Health Status: Females (Beneath coefficient is absolute value of t ratio in parentheses) Reporting hours ln hours in Working in 30 days and wage 30 days ln wage rate Dependent variable OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS estimation method 1 2 3 4 5 6 7 8 Variables: Self-evaluation of 0.0487 0.0585 0.0563 0.0782 0.0046 0.0378 0.0501 0.0123 Health (0-4) (3.70) (1.89) (4.12) (2.43) (0.26) (0.86) (1.65) (0.16) Age (x10-1) 0.826 0.828 0.749 0.753 0.167 0.180 0.388 0.374 (10.9) (10.9) (9.54) (9.57) (1.55) (1.64) (2.06) (1.96) Age squared 0.101 0.101 0.0914 0.0913 0.0205 0.0214 0.0456 0.0446 (x10-2) (10.7) (2.50) (9.34) (9.34) (1.51) (1.57) (1.92) (1.88) Pension age 0.0976 0.0980 0.0746 0.0755 0.0457 0.0444 0.0959 0.0945 (2.49) (2.50) (1.83) (1.86) (0.88) (0.85) (1.06) (1.04) Schooling years Grade school 0.0188 0.0186 0.0215 0.0211 0.0117 0.0120 0.0599 0.0602 (2.85) (2.82) (3.15) (3.08) (1.16) (1.19) (3.40) (3.41) Vocational 0.0163 0.0166 0.0194 0.0199 0.0095 0.0104 0.0240 0.0230 (2.01) (2.03) (2.30) (2.35) (0.90) (0.98) (1.31) (1.25) Technical 0.0275 0.0275 0.0323 0.0322 0.0013 0.0011 0.0566 0.0566 (5.33) (5.32) (6.03) (6.00) (0.17) (0.17) (4.98) (4.97) University 0.0235 0.0235 0.0243 0.0241 0.0097 0.0098 0.0879 0.0880 (6.60) (6.57) (6.55) (6.49) (2.36) (2.38) (12.3) (12.3) Unearned income 0.140 0.140 0.261 0.262 0.0470 0.0466 0.162 0.162 (x10-4) (3.83) (3.83) (6.91) (6.91) (0.85) (0.84) (1.68) (1.67) Height (cm) 0.0024 0.0024 0.0026 0.0026 0.0032 0.0031 0.0054 0.0055 (1.97) (1.95) (2.06) (2.01) (2.05) (2.00) (2.02) (2.04) Married 0.0635 0.0644 0.0613 0.0632 0.064 0.0656 0.0032 0.0015 (3.03) (3.05) (2.81) (2.88) (2.46) (2.51) (0.07) (0.03) Cohabiting 0.121 0.122 0.102 0.102 0.0253 0.0258 0.0088 0.0093 (4.13) (4.14) (3.33) (3.35) (0.69) (0.70) (0.14) (0.15) Divorced 0.0236 0.0228 0.0152 0.0134 0.0026 0.0001 0.0180 0.0149 (0.87) (0.83) (0.54) (0.47) (0.08) 0.00 (0.32) (0.26) Body Mass Index 0.0281 0.0276 0.0295 0.0283 0.0017 0.0033 0.0381 0.0400 (x10-1) (2.91) (2.82) (2.94) (2.79) (0.13) (0.25) (1.69) (1.75) BMI squared 0.0400 0.0387 0.0422 0.0400 0.0022 0.0053 0.0545 0.0581 (x10-2) (2.51) (2.41) (2.57) (2.40) (0.10) (0.24) (1.46) (1.53) Waist-hip ratio 0.417 0.417 0.449 0.444 0.023 0.025 0.309 0.306 (3.52) (3.51) (3.16) (3.60) (0.15) (0.16) (1.15) (1.14) Rural resident 0.0576 0.0580 0.119 0.120 0.134 0.135 0.311 0.311 (2.94) (2.95) (5.83) (5.86) (5.30) (5.30) (7.04) (7.03) Secret city (PGT) 0.101 0.102 0.149 0.151 0.054 0.0570 0.256 0.253 (3.02) (3.04) (4.31) (4.35) (1.17) (1.23) (3.19) (3.14) Intercept 0.153 1.52 1.47 1.43 4.41 4.46 0.948 0.886 (5.25) (5.07) (4.86) (4.59) (11.1) (11.1) (1.37) (1.27) R squared 0.199 n.a. 0.178 n.a. 0.034 n.a 0.271 n.a. F 29.5 29.1 27.2 26.8 2.56 2.58 27.1 27.0 p-value (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) Note: All regressions include regional dummies as defined in table A-1. Health Disabilities and Labor Productivity in Russia in 2004 103 ture that estimates the covariates of log of wage at the age that is eligible for a pension also tend rates (Mincer 1974). Schooling is generally to work 16 log points fewer hours. An individual associated with higher wages, and consequently who receives 1,000 rubles more in unearned with greater labor force participation. One income per month is likely to participate less interpretation of these empirical regularities is often in the labor force. These nonearned that they are consistent with individual labor income effects are larger for women, 1.4 per- supply being upwardly sloped with respect to an centage points in the labor force and 2.6 per- individual's wage opportunities, and invest- centage points in the wage labor force, than they ments in schooling are associated in most stud- are for men, respectively. These gender differ- ies with increased wages and labor force ences in nonearned income effects on labor sup- participation, at least where there is a safety net ply by gender are also estimated in the United for the disabled, sick, unskilled, and unem- States and other high-income countries ployed (Welch 1997). (Killingsworth 1983). The private wage returns of individuals who As in some other populations, adult height have invested their time in schooling as students is associated with higher wages, participation, appear in columns 7 and 8 to be moderate in and hours, although this pattern is weakened Russia, increasing from about 2 percent per year in this framework by the inclusion of many in school grade completed for men, and 6 per- other long-term indicators of health that are cent for women. For university training, the inversely correlated with height (Schultz increase is 5 percent for men and 9 percent for 2005). An increment of 1 centimeter in height women. The smaller groups with vocational and is related to a half percent increase in wages of technical schooling receive lower wage returns men and women. Fitting a quadratic function and little impact on their labor force participa- to BMI reveals the expected inverted U-shaped tion or hours. However, the partial relationships pattern, which is jointly significant in these between schooling and standard measures of regressions for participation and log wages, but labor supply and earnings would be positive and suggests the maximum likelihood of labor larger if the self-evaluation of health status, as force participation occurs when an individual's well as the physical stature of the individual BMI is between 32 to 35 for men and women. proxied by height and BMI, were excluded from This level of BMI is conventionally regarded the reported wage equations, as in standard as being overweight and subject to elevated risk earnings functions that include only schooling of mortality after age 45 (Waaler 1984; Fogel and post-schooling experience (Mincer 1974). 1994; Costa 1996). Waist-to-hip ratio is also a Each additional year of schooling for men and sign of cardiovascular disease risk and is signif- women is associated a 1 to 2 percent increased icantly associated in Russia with reduced prob- likelihood of working in the labor force. As abilities of labor force participation.24 Being noted in table 1, without controls, male wages currently married or cohabiting is associated do not vary with age after controlling for these with men and women receiving 15 and 7 per- health indicators, whereas female wages rise and cent higher wages, respectively, and there is then fall over the life cycle, reaching their max- greater labor force participation for women imum value when women are in their early 40s. than men who are currently single or divorced. Participation in the labor force is 10 percentage Among Russian women who are currently points lower for women age 55 to 64, when they married or cohabiting versus single or are eligible for a pension, controlling for the divorced, wages are not significantly different, quadratic age life-cycle pattern, and about 30 although the women with partners participate percentage points lower for men age 60 to 64, less frequently in the labor force and in wage who are age eligible for general pensions. Men employment by about 6 and 12 percentage 104 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union points, respectively. Residents in rural areas ability classification indicator, also yield simi- report receiving much lower wage rates, ­.57 lar estimated effects of the other exogenous log points for men and ­.31 for women, with variables on current health status regardless of the secret cities (PGT) log wages ­.33 for men whether OLS or 2SLS estimates are reported, and ­.25 for men and women, respectively, they are not included in the summary table 5 to compared to the urban residents. Given the save space. much lower rural wages, labor force participa- The critical estimates for this paper are the tion rates are also lower in these areas, at 11 productivity effects of the health status vari- percentage points for men and 6 percentage ables, which are reported in the first row of points for women. All of these coefficients esti- tables 4A and 4B, and in table 5. When the self- mated for the covariates of labor productivity assessed evaluation of an individual's health is are not greatly affected by shifting from OLS larger and health status worsens one unit, say to 2SLS estimates for the self-assessed health from good to average health, the OLS esti- evaluation variable in tables 4A and 4B. mated association is for participation in the Because the alternative indicators of current labor force to decrease by 9 percent for men health status, namely, health problems and dis- and 5 percent for women. When this health TABLE 5 Labor Force Productivity Determinants of Health Status, with and without Inputs Working Reporting ln hours ln wage Dependent variable OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS estimation method 1 2 3 4 5 6 7 8 Sex specification: Male: Without input: 1 Health problems 0.0844 0.249 0.0667 0.195 0.0372 0.0933 0.0188 0.0891 in last 30 days (4.76) (5.48) (3.45) 3.95 (1.81) (1.68) (0.57) (0.77) 2 Disability classification 0.240 0.404 0.215 0.337 0.0218 0.648 0.320 0.329 (13.0) (7.16) (10.5) (5.47) (0.42) (2.87) (2.95) (0.73) With inputs: 3 Self-evaluation 0.0912 0.193 0.0784 0.141 0.0109 0.0715 0.128 0.0192 (7.12) (2.89) (5.60) (2.40) (0.67) (1.20) (3.79) (0.13) 4 Health problems 0.0883 0.256 0.0703 0.148 0.0383 0.0945 0.0400 0.0271 (4.99) (2.76) (3.64) (1.75) (1.86) (1.24) (0.93) (0.15) 5 Disability classification 0.255 0.425 0.226 0.384 0.0182 0.666 0.326 0.415 (13.70) (3.59) (11.00) (3.42) (0.35) (2.12) (3.01) (0.51) Females: Without inputs: 1 Health problems 0.0457 0.0933 0.0414 0.105 0.0197 0.024 0.039 0.0035 in the last 30 days (2.95) (2.33) (2.57) (2.53) (1.05) (0.48) (0.91) (0.04) 2 Disability classification 0.191 0.217 0.176 0.183 0.117 0.125 0.0508 0.0805 (10.5) (2.93) (9.30) (2.37) (2.71) (0.46) (0.68) (0.17) With inputs: 3 Self-evaluation 0.0528 0.0614 0.0597 0.0729 0.0067 0.0882 0.0457 0.0304 (4.06) (1.24) (4.42) (1.34) (0.38) (1.39) (1.51) (0.26) 4 Health problems 0.0469 0.107 0.0423 0.108 0.0214 0.0579 0.0121 0.0325 (3.07) (1.60) (2.66) (1.44) (1.13) (0.76) (0.37) (0.22) 5 Disability classification 0.194 0.327 0.167 0.238 0.118 0.345 0.0423 0.487 (10.8) (2.27) (8.89) (1.80) (2.75) (0.88) (0.57) (0.66) Health Disabilities and Labor Productivity in Russia in 2004 105 evaluation is treated as endogenous and reesti- current health status are estimated to exert mated using 2SLS, this decline is estimated to larger impacts on labor productivity, especially reduce male participation by 15 percent and on participation in the labor force, when treated female participation by 6 percent, with slightly as endogenous, measured with error, and iden- larger effects on working in a wage job tified by chronic health problems and local (columns 4-3). The OLS estimates suggest 13 access to medical services. Should these 2SLS percent lower wages for men who report their estimates that are substantially larger than con- health status as poorer, and an insignificant 5 ventional OLS estimates be viewed as reliable percent lower for women. In this case the 2SLS and reasonably precise? These results are con- estimates are insignificant for wages as well as sistent with the labor supply literature drawn for hours for both men and women. In table 5, mostly from the United States, in which labor row 1, reporting a health problem in the last 30 supply responds to economic constraints meas- days, without controlling for health-related ured by education and health mostly through inputs, is associated (OLS) with men working 8 variation in participation. Estimates of the percentage points less often, but when treated responsiveness of hours worked among those as endogenous and estimated by 2SLS, the already in the wage labor force are poorly impact increases to a 25 percentage point defined, especially for males in their prime reduction. Hours also decline, according to (Killingsworth 1983). OLS estimates, by 4 percentage points for men, and when estimated by 2SLS, the decline is 9 Estimated Effects of Health-Related percentage points, although the latter estimate Inputs: Medical Care, Alcohol, and is not significant. For women, having one more Cigarettes health problem is associated with a 5 percent- Having a medical checkup in the last three age point decrease in participation (OLS), and months is associated by OLS with a 9 percent- this becomes a 9 percentage point decrease age point increase in men participating in the according to 2SLS. No effects of women's labor force, and a 17 percentage point increase health problems are apparent on their wage rate for women in tables 6A and 6B. Seeing a doctor or on their hours of work. more frequently is directly associated with The government-assigned disability classifi- increased participation, although it is not evi- cation is rare and intended to designate those dently related to higher wages or hours of work who are unable to work or can only work with for men, and is even associated with lower wages difficulty and with assistance. A one-unit rates for women (OLS). However, when esti- increase in this variable, indicating worse mated by 2SLS, the identified predicted fre- health, is associated (OLS) with men working quency of medical checkups is unrelated to 24 percentage points less and women working working, hours, or wages, for either men or 19 percentage points less. But when estimated women. Specification tests in the case of med- by 2SLS, which should correct for classical ical checkups indicate that the local infrastruc- measurement errors and endogeneity, the ture and prices provide satisfactory levels of impact is to reduce working by 40 percentage explanatory power for the full population sam- points for men and 22 percentage points for ple, at better than 1 percent for men and women. Disability classification also reduces the women, but less power for the wage sample of work hours for males with a wage job, according persons reporting hours and earnings (table A- to the 2SLS estimates, but not for women, and 2). The Durbin-Wu-Hausman tests also con- does not appear to affect wages for those few firm the medical checkup coefficients estimated individuals classified as disabled and still work- by OLS and 2SLS differ significantly for partic- ing and reporting wages. All three measures of ipation and reporting hours and earnings 106 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union (p<.0001), but only for women's hours and greater participation, and the turning point of men's wages (table A-3). More frequent visits to the quadratic is after the individual consumes a doctor do not appear to be causally increasing 122 grams of alcohol per day for men, well participation in the labor force or productivity. above the sample mean for men of 105 grams The direct association (OLS) between medical (table A-1). The maximum participation for checkups and labor force participation may be women occurs at 124 grams per day, almost two due to individual heterogeneity, where those standard deviations above their sample mean of who prefer to see doctors more often are also 36 grams. By United States standards, these engaged in other behaviors designed to improve detrimental effects of alcohol are estimated for their health. Working may facilitate periodic men and women who are consuming on an aver- checkups if the medical care is encouraged by age day alcohol levels which would qualify as employers or even provided at the place of binge drinking. employment. Medical consultations and check- Cigarette smoking is associated in the pre- ups may increasingly require the provision of ferred 2SLS estimates in an inverted U-shaped bribes in the Russian market­oriented economy, relationship with labor force participation for and if linked to an employer these side payments men, where the positive initial effects of smok- may be moderated. ing on participation changes sign after smoking It is often hypothesized that moderate levels more than 12 cigarettes per day, roughly the of alcohol consumption may be physically pro- sample average. Among women, the negative tective from some risks of cardiovascular dis- effects of cigarettes on participation and work- ease and stroke, and may facilitate the ing for wages are more significant for the OLS formation of social relationships and networks estimates in table 6b than the 2SLS, but women that could improve work productivity. But also smoke fewer cigarettes per day than men, excessive alcohol consumption can lead to acci- averaging 2.3. dents, violence, disease, and health problems Including the three health-related inputs in (Russell 1987; Roman 1988; Mullahy and Sin- the second-stage productivity equations (5) delar 1991; French and Zavkin 1995; Tekin results in minor changes in the estimated effects 2002; and Baltagi and Gekshecker 2006). Thus, of the current health variables when they are a quadratic form for alcohol consumption may significant, as reported in tables 4A, 4B, and help to estimate nonmonotonic health conse- rows 3 to 5 in table 5. The reversal in sign of the quences of alcohol consumption, and a quad- benefits of medical care from the OLS to the ratic-form specification is also adopted for 2SLS estimates are consistent with the hypoth- cigarette consumption to assess nonlinearity esis that unobserved preferences for health lead (table 6). some individuals to invest selectively in more The power of the identifying instruments in healthy patterns of consumption and behavior, the first-stage equation explaining alcohol con- which are associated with improved productiv- sumption are jointly significant for both men ity, especially for women. This individual het- and women in the whole sample, and in the erogeneity in preferences contributes to the sample of individuals reporting positive hours positive OLS estimates in tables 6A and 6B. But and earnings (table A-2). The Durbin-Wu- the variation in demand for health care pre- Hausman test rejects the joint exogeneity of the dicted by the access to local health care infra- alcohol and alcohol-squared inputs for women's structure and doctors does not indicate that the participation in the labor force equation, and associated variation in periodic checkups with men's and women's wage participation (table A- these instruments contributes to improved pro- 3). Higher alcohol consumption is initially asso- ductivity of men or women (Rosenzweig and ciated in the preferred 2SLS estimates with Schultz 1983). 25 Health Disabilities and Labor Productivity in Russia in 2004 107 TABLE 6A Labor Force Productivity Determinants, Including Medical Care, Smoking, and Alcohol: Males Working in 30 days Reporting hours and wage ln hours in 30 days ln wage rate Dependent variable OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS estimation method 1 2 3 4 5 6 7 8 Medical checkup 0.0913 0.318 0.0685 0.242 0.0005 0.25 0.0163 1.03 in 3 months (4.05) (1.05) (2.79) (0.89) (0.02) (0.87) (0.33) (1.47) Cigarettes per day 0.0017 0.0555 0.0039 0.0445 0.0012 0.0251 0.0130 0.0909 (1.86) (2.01) (2.05) (1.81) (0.64) (1.42) (3.32) (2.13) Cigarettes 0.0094 0.178 0.0125 0.182 0.0026 0.0548 0.0293 0.246 squared (x10^-2) (1.82) (1.89) (2.21) (2.18) (0.46) (1.19) (2.54) (2.21) Alcohol per day 0.0104 0.203 0.0220 0.163 0.0267 0.0423 0.0078 0.755 squared (x10^-2) (0.89) (1.20) (1.73) (1.08) (1.44) (0.24) (0.20) (1.80) Alcohol 0.0035 0.0833 0.0073 0.0456 0.0068 0.0635 0.0021 0.197 squared (x10^-4) (1.77) (2.65) (3.44) (1.63) (1.52) (1.38) (0.22) (1.77) Joint significance (0.16) (0.13) (.086)* (.079)* (.072)* (0.36) (.0029)* (.080)* of cigarettes Joint significance (0.14) (0.003)* (.0006)* 0.213 (0.31) (.017)* (0.69) (0.19) of alcohol R squared 0.165 n.a. 0.174 n.a. 0.037 n.a. 0.264 n.a. F 18.2 7.36 18.2 10.5 2.06 1.47 19.3 9.7 p-value (0.0001) (0.0001) (0.0001) (0.0001) (0.0005) (0.0464) (0.0001) (0.0001) * Statistically significant jointly at the 10 percent level. TABLE 6B Labor Force Productivity Determinants, Including Medical Care, Smoking, and Alcohol: Females Working in 30 days Reporting hours and wage ln hours in 30 days ln wage rate Dependent variable OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS estimation method 1 2 3 4 5 6 7 8 Medical checkup 0.165 0.235 0.163 0.327 0.0191 0.0176 0.146 0.544 in 3 months (9.04) (1.08) (8.57) (1.35) (0.92) (0.09) (4.09) (1.44) Cigarettes per day 0.0018 0.0246 0.0036 0.0761 0.0032 0.0562 0.0030 0.0171 (1.86) (2.01) (2.05) (1.81) (0.74) (1.47) (0.79) (0.71) Cigarettes 0.0136 0.195 0.0067 0.251 0.0043 0.3430 0.0285 0.265 squared (x10^-2) (0.86) (0.75) (0.40) (0.87) (0.21) (1.73) (0.79) (0.71) Alcohol per day 0.0807 0.563 0.107 0.986 0.0081 0.170 0.0821 0.979 squared (x10^-2) (3.26) (1.79) (4.17) (2.80) (0.22) (0.64) (1.30) (1.93) Alcohol 0.0243 0.303 0.0350 0.399 0.0040 0.066 0.0110 0.136 squared (x10^-4) (2.76) (2.12) (3.83) (2.50) (0.25) (0.44) (0.40) (0.48) Joint significance (0048)* (0.66) (0.0041)* (0.158) (0.41) (0.22) (0.55) (0.56) of cigarettes Joint significance (.0050)* (0.105) (.0001)* (.019)* (0.97) (0.80) (.0041)* (.021)* of alcohol R squared 0.211 n.a. 0.198 n.a. 0.035 n.a 0.281 n.a. F 29.3 17.3 27.0 13.9 2.31 2.03 24.8 17.5 p-value (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0007) (0.0001) (0.0001) * Statististically significant jointly at the 10 percent level. 108 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Conclusions related inputs and behavior are in turn selected by individuals, based in part on their preferences Physical or mental inability to function nor- and their health status (which the statistician mally in society can be viewed as a disability. generally does not observe) it is necessary to The conditions that contribute to a person model the determinants of these basic health being thus unable to work or function in his or production processes by policies that can be her regular activities are specific to individuals, widely implemented and scaled up, and which as well as the household and social support net- are not subject to choice by the individual. Only work within which they function, and the when these policy and price variables can be threshold that qualifies a person to be called dis- appropriately viewed as exogenous to the health abled is shaped by culture, norms, and institu- and economic system can the researcher treat tions. This complex health-economy-culture these variables conditioning health outcomes as process determining who is reported to be dis- instrumental variables to identify exogenous abled may explain why low-income countries variation in current health status. Therefore, a generally report lower levels of disability than goal is to explain the underlying demands for do high-income countries, even though mortal- these inputs and behaviors that contribute to ity and medically documented morbidity is better health to assess without statistical bias higher in poor countries than in rich ones. This how these inputs improve health and enhance paper proposes to resolve this problem of objec- the productive capacity of a population. This tively measuring disability by employing investigation has made progress in understand- chronic and lagged health conditions of the ing the forms of health-related behavior that individual to predict (as instrumental variables) affect health and labor productivity in post- current health status indicators, which may oth- transition Russia, such as high levels of smoking erwise be more subjective and heterogeneous and drinking. Perhaps by reorienting the analy- than the individual's specific chronic diagnosed sis to focus more on variation over time in prices health preconditions. The study outlines a of these commodities and productive medical replicable methodology for assessing the eco- services and regionally distinct policies toward nomic and social consequences of health and alcohol consumption, and the use of preventive disabilities on individual productivity in the medical care and health education programs labor market, and the resulting estimation strat- that encourage healthier behaviors, it might be egy may also prove replicable across different possible to identify more clearly the pathways socioeconomic strata within a country, and pos- from policy instruments to survey measures of sibly even across cultures and countries in the health status, and to the size and productivity of world, when self-reported health indicators are the Russian labor force. not causally related to productivity. It is also Self-assessed health status questions from a important to control for other demographic and representative survey should be recognized as human capital characteristics that are not sub- measured with error and subject to individual ject to choice by adults over their mature work- choice and interpretation. This requires that the ing years in the economy. researcher attempt to explain these health status This study describes quantitatively the deter- responses accordingly as endogenous choices minants of economic activity and productivity subject to measurement error. This requires the in terms of observable health conditions and employment of at least an implicit structural human capital of the individual, and in terms of framework as outlined in this paper. In this his or her location and current access to medical analysis, it was assumed that chronic health con- care and health-related inputs, as well as past ditions could be treated as exogenous instru- health problems. However, because health- ments to predict the more subjective current Health Disabilities and Labor Productivity in Russia in 2004 109 responses to health status questions, and the emerged in the 1990s, during the first decade empirical results confirmed that this approach of the Russian transition from a centralized yielded reasonable patterns that were statisti- communist state to a partly decentralized mar- cally different from those obtained by conven- ket oriented economy (Guo 1993; Nolte et al. tional single-equation methods, i.e. OLS or 2004; World Bank 2005). The 2004 RLMS logit. In the case of participation in the labor survey data examined here did not explicitly force, these 2SLS estimates of the effects of link mortality to morbidity and health status of health on productivity were much larger than the sampled population. The RLMS samples conventional OLS estimates. Intervening with addresses and does not follow household mem- early health education, changing health-related bers who move or link their identities to death behaviors, and inducing the population to use registration. By exploiting the household panel effectively medical care to avoid the develop- feature of the RLMS, it might be possible to ment of chronic degenerative diseases has the trace the connections between actual adult potential to reduce what in the future will be the mortality between rounds of the survey, and past health problems and improve people's stock the predicted prior health status and disabili- of current health human capital. According to ties of that individual in earlier rounds of the the estimates reported in this paper, such an survey. This study instead analyzed a represen- improvement in current health status would tative sample of all middle-aged men and result in large increases in labor force participa- women to see whether their widely reported tion, work in wage jobs for which they report health behaviors and indicators of their cur- hours and earnings, and, to some extent, rent health status are individually associated increase hours and wage rates among those who with chronic health problems. Then these already are in the wage labor force. prior health conditions and current local com- Age-specific mortality among Russian males munity conditions as employed to predict cur- age 25 to 64 was four times the levels recorded rent health status and disabilities, and assess in Western Europe at the end of the 1990s, how this predicted variation in health impacts whereas Russian females' mortality was twice labor productivity, as measured by variation in West European levels. Half of this East-West labor force participation, hours worked, and mortality differential among prime aged adults wage rates. 110 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Appendix TABLE A1 Definitions and Sources of Variables and Sample Statistics from Russian Longitudinal Monitoring Survey, 2004a Alternative samples -- age 25 to 64 Class of variable\ Reporting positive hours and Variable definition All persons wages in last 30 days (Questionnaire Females Males Females Males source adult and [1] [2] [3] [4] community modules) Working sample size 3,429 2,708 1,999 1,699 ( Variables reported) Productivity (P) 1. Working in last 30 days 0.645 0.729 1.00 1.00 (J.7 & 30) 2. Reporting earnings and hours 0.583 0.627 1.00 1.00 (8 &10 or 38 & 40) 3. Ln hours in 30 days n.ab n.a 5.04 5.17 (8 & 38) [0.408] [0.366] 4. Ln wage rate in 30 days n.a n.a 3.10 3.41 (10/8 or 40/38) [0.819] [0.873] Health Status Current (HC) 1. Self-evaluation (0­4, Excellent 1.87 1.69 1.81 1.61 to Poor; M3) [0.615] [0.656] [0.553] [0.574 2. Any health problems in last 0.428 0.283 0.393 0.251 30 days (L5) [0.495] [0.451] [0.488] [0.434] 3. Assigned disability 0.0916 0.104 0.0275 0.023 classification (0­3; M 20.8) [0.410] [0.427] [0.212] [0.172] Health Lagged (HL) 1. Chronic heart disease 0.130 0.081 0.099 0.060 (M20.6.1) 2. Chronic lung disease 0.052 0.051 0.047 0.040 (M20.6.2) 3. Chronic liver disease 0.108 0.046 0.091 0.038 (M20.6.3) 4. Chronic kidney disease 0.116 0.042 0.104 0.031 (M20.6.4) 5. Chronic stomach disease 0.179 0.129 0.172 0.128 (M20.6.5) 6. Chronic spinal disease 0.170 0.129 0.166 0.126 (M20.6.6) 7. Other chronic disease 0.240 0.138 0.233 0.117 (M20.6.7) 8. Had serious nervous 0.235 0.137 0.215 0.112 disorder or depression in the last 12 months (M131) 9. Doctor diagnosed high blood 0.412 0.288 0.372 0.277 pressure (M58.1) Health Related Inputs/Behaviors (I) 1. Medical checkup in last 3 0.206 0.143 0.267 0.166 months (L26) Health Disabilities and Labor Productivity in Russia in 2004 111 Alternative samples -- age 25 to 64 Class of variable\ Reporting positive hours and Variable definition All persons wages in last 30 days Females Males Females Males [1] [2] [3] [4] 2. Alcohol consumption per 0.359 1.05 0.391 1.02 day in last 30 days in gr. [0.546] [1.17] [0.526] [1.02] ethanol ( x 10-2).c 3.Cigarettes consumed per dayd 2.33 12.0 2.28 11.9 [5.59] [10.4] [5.41] [10.6] Control - Independent Variables [C] Individual Variables [Ci] 1. Age in years ( x 10-1) 4.29 4.15 4.13 4.01 [1.11] [1.09] [0.97] [1.00] 2. Age in years squared 19.6 18.4 18.0 17.1 ( x 10-2) [9.69] [9.36] [8.11] [6.25] 3. Pension-eligible age 0.181 0.058 0.096 0.024 (women: 55­64; men: 60­64) 4. Education years completed (a) Grade school (0­12) 9.57 9.43 9.68 9.52 (J70.1) [1.18] [1.23] [0.983] [1.10] (b) Vocational (0­6) 0.563 0.898 0.565 0.903 (J7 2.2 +72.3) [1.04] [1.24] [1.04] [1.25] ( c ) Technical (0­9) 1.15 0.735 1.28 0.830 (J72.4) [1.58] [1.41] [1.61] [1.47] (d) University & graduate (0­13) 1.40 1.17 1.67 1.39 (J72.5 & J72.6) [2.35] [2.28] [2.51] [2.44] 5. Unearned income in last 30 0.0438 0.0418 0.0276 0.0208 days (x 10-4) [0.204] [0.351] [0.164] [0.175] (J60 - J10 ­ J40 - J76 - J89) 6. Height in cm. (R3) 162.0 174.0 162.0 175.0 [6.37] [6.91] [6.22] [6.89] 7. Currently married 0.587 0.703 0.586 0.748 (J72.17) 8. Cohabiting currently 0.111 0.132 0.104 0.132 (J72.17) 9. Divorced currently 0.137 0.072 0.154 0.052 (J72.17) 10. BMI (weight in 27.6 25.6 27.3 25.8 kg/height in MT squared) [5.87] [4.14] [5.69] [4.08] (R4 & R3) 11. BMI squared 7.94 6.71 7.78 6.80 (x10-2) [3.53] [2.28] [3.38] [2.25] 12. Waist/hip ratio 0.812 0.899 0.804 0.897 (R5 & R6) [0.072] [0.070] [0.069] [0.067] 13. Rural resident 0.230 0.260 0.178 0.188 (SETT_TYP) 14. PGT_secret city 0.062 0.061 0.046 0.042 (SETT_TYP) Community Variables (Cc) 1. Low price of vodka liter 110.0 110.0 110.0 110.0 (10-2 rubles: A1.90) [22.2] [22.1] [21.1] [20.8] 2. Low price of cigarettes pkg. 5.37 5.48 5.41 5.62 (rubles: A1.82) [4.41] [4.71] [4.36] [4.57] 112 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE A1 (continued) Definitions and Sources of Variables and Sample Statistics from Russian Longitudinal Monitoring Survey, 2004a Alternative samples -- age 25 to 64 Class of variable\ Reporting positive hours and Variable definition All persons wages in last 30 days Females Males Females Males [1] [2] [3] [4] 3. No community 0.109 0.107 0.120 0.133 questionnaire available 4. Hospital in area 0.744 0.733 0.765 0.754 (Q46) 5. Time to reach hospital in 5.15 5.81 3.57 3.62 hours (Q48) [15.2] [16.5] [11.2] [12.3] 6. Private doctor in area 0.640 0.623 0.682 0.676 (Q46) 7. Time to reach private doctor 25.0 26.3 21.6 18.4 (Q48) [70.0] [70.9] [66.0] [58.4] 8. No doctor time reported 0.032 0.037 0.017 0.017 9. Social welfare office in area 0.700 0.691 0.733 0.739 (Q79A) 10. Does social welfare office 0.668 0.656 0.697 0.692 provide medial services (Q80) Regional Variables (CR)-Omitted Region Moscow City 1. St. Petersburg 0.044 0.038 0.049 0.046 2. Moscow Oblast 0.043 0.038 0.041 0.038 3. North and North West 0.063 0.064 0.070 0.067 4. Central and Black Earth 0.131 0.130 0.141 0.141 5. Volga-Vaytski 0.175 0.181 0.168 0.171 6. North Cauzasian 0.131 0.137 0.105 0.110 7. Urals 0.136 0.137 0.151 0.149 8. West Siberia 0.094 0.093 0.093 0.082 9. East Siberia and Far East 0.094 0.093 0.087 0.086 Note: a. The standard deviations of binary variables are not reported because standard deviations are equal to (m(m-1)exp. 0.5. b. n.a. not available or defined. c. Ethanol consumption is based on the following weighted function: 0.05 beer + 0.11 table wine or champagne + 0.19 fortified wine + 0.40 vodka + 0.45 home-made liquor (samogen) + 0.20 other alcohol (Q80.1 - 80.6). No information is reported on "binge" drinking, often measured as having five or more drinks on one occasion, or 80g. ethanol or more. These binges may be more harmful to health than more moderate, uniform consumption, but this aspect cannot be assessed with these data. Alcohol consumption in grams of ethanol per day is divided by 100, and alcohol consumption squared is divided by 10,000 in the regressions, i.e., 100 squared. d. Tobacco consumption is summarized by the number of regular cigarettes consumed per day. Paperosi may be stronger tobacco, but are shorter or contain less to- bacco per cigarette than normal ones measured here. Cigarettes per day is divided by 10 in the regressions. Health Disabilities and Labor Productivity in Russia in 2004 113 TABLE A2 Joint F-Test of Significance of Identifying Instruments in First-Stage Regressions (p-values in parentheses) Dependent variable All persons age 25­64 Reporting hours and wage Males Females Males Females Self-evaluation 39.59 39.31 16.88 19.38 (0.0001) (0.0001) (0.0001) (0.0001) Health problems 26.07 31.32 14.01 16.56 (0.0001) (0.0001) (0.0001) (0.0001) Disability classification 17.50 11.41 5.29 2.66 (0.0001) (0.0001) (0.0001) (0.0001) Medical checkup 2.78 2.22 1.17 1.52 (0.0001) (0.0017) (0.28) (0.0687) Cigarette consumption 3.05 3.35 1.79 1.46 (0.0001) (0.0001) (0.0193) (0.0902) Cigarette squared 2.11 2.20 2.19 1.16 (0.0034) (0.0020) (0.0022) (0.28) Alcohol consumption 4.37 3.58 2.84 2.38 (0.0001) (0.0001) (0.0001) (0.0007) Alcohol squared 2.78 1.75 2.58 1.30 (0.0001) (0.0229) (0.0002) (0.17) TABLE A3 Hausman Specification Tests (t or F) of the Exogeneity of Dependent Variables* Working in 30 days Reporting hours & wage ln hours per month ln wage per hour Dependent variable Males Females Males Females Males Females Males Females Self-evaluation (0.0001) (0.0014) (0.0001) (0.0006) (0.121) (0.93) (0.0001) (0.084) Health problems (0.0054) (0.0263) (0.0416) (0.0842) (0.204) (0.35) (0.50) (0.55) Disability classification (0.0001) (0.0001) (0.0001) (0.0001) (0.684) (0.0073) (0.0037) (0.53) Medical checkups* (0.0001) (0.0001) (0.0009) (0.0001) (0.95) (0.37) (0.96) (0.0001) Cigarettes and (0.0596) (0.0065) (0.0253) (0.0072) (0.073) (0.39) (0.0005) (0.62) cigarettes squared* Alcohol and (0.34) (0.0059) (0.0008) (0.0003) (0.30) (0.99) (0.82) (0.0133) alcohol squared* Note: * Estimated jointly with other health-related inputs. 114 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Notes job as well. This group for whom the secondary job is relied on constitutes less than 2 percent of 1. This exploratory analysis does not identify pol- the paid labor force. icy priorities that would be most cost-effective 6. In 2004, US$1 equals about 29 rubles. For addi- means to reduce the prevalence of adult disabil- tional information on Russian social security ities or mitigate the economic consequences of programs, consult: www.ssa.gov/policy/docs/ these disabilities emerging during the transition progdesc/ssptw/2004-2005/europe/russia. health crisis. Nor does this study distinguish 7. If one assumed that those who report no earn- between mental and physical disabilities. ings or no hours, but nonetheless indicate they 2. The private and social costs of changing the worked in the last 30 days, received less com- instrumental variables constraining an individ- pensation per hour and worked fewer hours than ual's environment that are associated with the did those reporting their earnings and hours, the frequency of observed disabilities in Russia expected levels of cohort productivity for men could be a (the) focus of future study, based on and women would be lower than the above cal- more detailed local administrative health care culations suggest, but the magnitude of this sam- delivery data from Russia. Perhaps the repeated ple selection bias in reporting work, hours, and rounds of the RLMS during which health infra- earnings has not to my knowledge been assessed structure, insurance, employment conditions, in Russia. economic shocks, and disability pensions may 8. The distribution of wage rates tends to be have changed over time and across regions of skewed to right. When the wage is expressed in Russia would allow a researcher the opportunity logarithms, it is distributed more nearly to nor- to estimate the health consequences of macro mal. If errors in the wage function are normally and micro economic changes (Currie and distributed, OLS regression and maximum like- Madrian 1999). lihood methods are more attractive, and the 3. The person "not reporting" could work for an earnings function as conceptualized in the employer who is in arrears on his payments to human capital literature by Mincer (1974) is his workforce. If they completed the subsequent more readily interpreted. questions in the survey, they are attributed the 9. There is also the convenience of being able to actual payment received for actual hours decompose the marginal effects of explanatory worked, but some may have failed to respond to variables on log earnings into a linear model both questions. Alternatively, the individual may where the coefficients are the sum of those coef- be self-employed and cannot distinguish his or ficients in the analogous linear model for log her wages from returns to his or her capital and wages and log hours. reward for taking risks. The worker may also be 10. Although days missed or absenteeism has been paid in periods other than months, or does not emphasized in studies of Russia as being unusu- know or want to divulge his earnings or hours ally high, I did not find it particularly large, nor last month for fear he may owe taxes, perhaps was it responsive to control variables, or strongly because his work in the "black market" economy associated with any of the indicators of labor is unregulated, i.e., untaxed. supply and productivity examined here, with the 4. The survey questionnaires and sampling possible exception of hours of work in the OLS methodology are described and the data was estimates. For another perspective, see the downloaded from the Web site: http://www analysis by Suhrcke et al. 2005. .cpc.unc.edu/projects/rlms. The sample statis- 11. For example, among men age 55 to 64, 83 per- tics for the final estimation samples are reported cent have no disability (disability classification in appendix table A-1, which includes 2,708 men variable = 0), 5 percent have first-degree dis- and 3,429 women for whom all of the control ability (=1)--of whom about half work, 10 per- variables and instruments are defined or identi- cent have second-degree (=2), and 2 percent fied by missing variable dummies. have the most severe, third-degree disability 5. A few percent are included in the "paid labor (=3) and very few work. Note that survey code force" who do not report earnings and hours in 1 refers to third-degree disability, and survey their primary job, but do in their secondary job. code 3 refers to first-degree disability (adult Hours are then set to the sum of those hours questionnaire 20.7 and 20.8), which required a reported, and the hourly wage is taken from the recoding of the disability classification variable second job and assumed to apply to the primary analyzed here. Health Disabilities and Labor Productivity in Russia in 2004 115 12. However, Bobak and Marmot (1999) remind tion of alcohol, which was then associated with one that the causal evidence for the connection declines in adult mortality (Carcone 1994; between alcohol consumption and cardiovascu- Shkolnikov et al. 1997; Leon et al. 1997). I have lar mortality is ambiguous, and the officially calculated my own variable of total ethanol con- reported levels of alcohol consumption in Russia sumption from the five categories of consump- are relatively moderate by European standards, tion reported, rather than use the derived though probably underestimated given the variable in the survey file, which I could not extent of illegal production of spirits (Samogan) duplicate or otherwise explain its origins. in Russia (i.e., Trelm 1982). 16. Men age 25 to 64 have received more years of 13. The total budget for social expenditures of the vocational education than women, at .88 years federal government in Russia is estimated to be compared to .56, whereas women have received 13 percent of GDP, and 48 percent of this more years of technical and university educa- budget is allocated to general pensions (Tesliuc tion, 1.15 compared to .74 for technical school, and Zotova 2004: table 6). There are also special and 1.41 compared to 1.18 for university educa- pensions for the disabled, as well as for mater- tion. nity and sick leaves, all of which are funded by 17. Older age groups probably overstate the secular the centralized Social Insurance Fund, which is gains across birth cohorts in childhood health independent of contributions from employers or and nutrition, although selective survival favor- employees. Unemployment benefits are very ing the tall could operate in the opposite direc- small, representing only 0.1 percent of the tion to understate these height improvements budget for social expenditures in 2003, perhaps inferred from differences between young and because they have not always been adjusted for elderly age groups in a cross-sectional survey inflation (Foley 1997). Child allowances are such as the RLMS. widespread because they are not selectively 18. Diabetes was not significantly associated with granted to those with low incomes, but they have health status indicators or productivity of labor, declined in real value from 1998 to 2003 as a perhaps because of its infrequency (last column share of minimum subsistence levels (Tesliuc and table 2) , and is not included in the reported esti- Zotova 2004: figure 9). As noted above, the age mates. Nor is tuberculosis, which is reported to of eligibility for pensions is included in this be much rarer. analysis as an exogenous conditioning variable, 19. A question regarding individuals' self-reported holding constant also for a quadratic function in exercise was not significantly related to health age to approximate broader biological and status outcomes or productivity, and was not behavioral variation over the life cycle. included in the final model. 14. The amount of physical exercise an individual 20. In Russia, pensions for the elderly are routinely engages in is rated by the respondent, but it does available for women after age 54 and for men not clearly vary by gender or age (table 3), or any after age 59, although in contrast to some coun- of the three health status indicators or four pro- tries, receipt of the pension does not preclude ductivity measures, and is therefore not included working unless prepension incomes are very in the final specification of the model. high. Some pensions are awarded on "merit" 15. The grams consumed per day of various types because of disabilities, sickness, maternity leave, of alcoholic beverages are reported in the sur- unemployment, etc. vey and are converted by the author into 21. It might be argued that some of the controls are ethanol alcohol equivalents following the exam- endogenous to health and productivity of the ple of Baltagi and Gekshecher (2006), who individual, such as marital status, for one might adopted the conversion rates of Mullahy and imagine that healthy and productive individuals Sindelar (1991). I have used those mentioned in would be more likely to be married. Remaining the survey: 5 percent beer, 10 percent wine, 19 in a rural area and not migrating to the city percent fortified wine, 40 percent vodka, 45 where wages and living conditions might be eco- percent other home-produced spirits. The large nomically better could be explained by an indi- supplies of private illegal distilled spirits pro- vidual having a stronger than average preference duced in Russia have been estimated to increase for a rural lifestyle, which could also have a bear- when state prices of vodka are raised (Treml ing on his or her productivity or healthiness in 1982), and did so during Gorbachev's brief either location. The existence of the PGT cities 1985­88 campaign to reduce Russian consump- was acknowledged only after 1995, and these 116 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union hidden cities may have special industrial and References environmental problems for their population's health. Regular urban areas are the omitted cat- Baltagi, Badi H., and I. Gekshecker. 2006. "Rational egory, and health in the PGT cities appears Alcohol Addiction: Evidence from the Russian somewhat better than rural areas, but not as Longitudinal Monitoring Survey." Discussion good as urban areas. Paper No. 2134, IZA, Bonn, Germany. 22. In a few clusters, the community consumption Barker, D.J.P. 1994. Mothers, Babies, and Disease in questionnaire did not provide a price, in which Later Life. London: BMJ Publishing Group. case the average for the primary sampling unit, Bartel, A., and P. Taubman. 1979. "Health and La- of which there are 38 in the RLMS, is extrapo- bor Market Success: The Role of Various Dis- lated from an adjacent sample cluster to the site eases." Review of Economics and Statistics 61 (10): for which no price data were reported. 1­8. 23. The problem described in Bound (1991) is that Bengtsson, Tommy, et al. 2004. Life under Pressure: errors in measuring HC would bias the OLS Mortality and Living Standards in Europe and estimate of HC's effect on P toward 0 in esti- Asia, 1700­1950. Cambridge, MA: MIT Press. mating (4) by OLS, whereas the endogenous Bobak, Martin, and M. Marmot. 1999. "Alcohol and effects of preferences for leisure might be Mortality in Russia: Is it Different than Else- expected to lead to an overstatement of negative where?" Annals of Epidemiology 9 (6): 335­38. effect of disabilities on labor supply and produc- Bound, John. 1991. "Self-Reported versus Objective tivity indicators. The RLMS information on Measures of Health in Retirement Models." health input behavior is quite limited, and vari- Journal of Human Resources 26 (1): 106­38. ables reported by all respondents are unlikely to Brainerd, Elizabeth, and D.M. Cutler. 2004. "Au- measure community access to and prices of topsy on an Empire: Understanding Mortality health care, on which the endogenous demand in Russia and the Former Soviet Union." Work- for health-related inputs or behaviors might rea- ing Paper 10868, National Bureau of Economic sonably be determined and identified with satis- Research, Cambridge, MA. factory precision. The community program, Carcone, J.A. 1994. "Alcohol-Related Problems as policy, and price variations are, therefore, likely an Obstacle to the Development of Human to be "weak instruments" on which to base the Capital." Technical Paper No. 219, World otherwise desirable 2SLS estimates of the effects Bank, Washington, DC. of community and household variables on adult Costa, D. 1996. "Health and Labor Force Participa- health status and labor productivity. Such a case tion of Older Men, 1900­1991." Journal of Eco- of weak instruments is expected to yield esti- nomic History 56: 62­89. mates that are biased toward the OLS estimates. Currie, Janet, and B. Madrian. 1999. "Health, 24. As this final indicator of stature is not (to my Health Insurance, and the Labor Market." In knowledge) included in previous multivariate Handbook of Labor Economics, ed. O. Ashenfelter studies of labor productivity or health, I also and D. Card, 3309­3415. Amsterdam, the excluded the waist-hip ratio from the model and Netherlands: Elsevier Science BV. noted the estimated fit of BMI in the participa- Dostie, Benoit, and P.T. Leger. 2005. "The Living tion and wage functions as otherwise specified in Arrangement Dynamics of Sick, Elderly Indi- tables 4A and 4B implied that maximum produc- viduals." Journal of Human Resources 40 (4): tivity was associated with BMI values of 30 and 989­1014. 32 for men and women, respectively, which are Fogel, R.W. 1994. "Economic Growth, Population more in accord with the standard literature. Theory, and Physiology." American Economic Re- 25. A sign reversal in the estimated effect of medical view 84 (3): 369­95. care is analogous to seeing a doctor earlier in a Fogel, R.W. 2004. The Escape from Hunger and Pre- woman's pregnancy, which was associated (OLS) mature Death: Europe, America, and the Third with a lower birthweight outcome, but when the World, 1700­2100. New York: Cambridge Uni- timing of her first prenatal visit was estimated by versity Press. 2SLS and identified by local access to medical Foley, Mark C. 1997. "Labor Market Dynamics, Un- facilities, prices of medical care, and household employment Duration, and Multiple Job Hold- resources, earlier prenatal care was beneficial for ing in Russia during Economic Transition." the child's birthweight (Rosenzweig and Schultz Unpublished PhD dissertation, Yale University, 1983). New Haven, CT. Health Disabilities and Labor Productivity in Russia in 2004 117 French, Michael, and G.A. Zavkin. 1995. "Is Moder- Mincer, J. 1974. Schooling, Experience, and Earnings. ate Alcohol Use Related to Wages?" Journal of New York: Columbia University Press. Health Economics 14: 319­44. Mullahy, J., and J.L. Sindelar. 1991. "Gender Differ- Glinskaya, Elena, and T. Mroz. 2000. "The Gender ences in Labor Market Effects of Alcoholism." Gap in Wages in Russia from 1992 to 1995." American Economic Review 81: 161­65. Journal of Population Economics 13 (2): 353­86. Nolte, Ellen, M. McKee, and A. Gilmore. 2004. Griliches, Z., and J. Hausman. 1986. "Errors in Vari- "Mortality, Morbidity in Transition Countries ables in Panel Data." Journal of Econometrics 31 in the European Context." Background paper (1): 93­118. for European Population Forum 2004, London Guo, Guang. 1993. "Mortality Trends and Causes of School of Hygiene and Tropical Medicine. Death: A Comparison between Eastern and Pauly, Mark V., et al. 2002. "A General Model of the Western Europe, 1960s-1980s." European Jour- Impact of Absenteeism on Employer and Em- nal of Population 9: 287­312. ployee." Health Economics 11: 221­31. Heckman, J.J. 1997. "Instrumental Variables: A Pelkowski, J.M., and M.C. Berger. 2004. "The Im- Study of Implicit Behavioral Assumptions in pact of Health on Employment, Wages, and One Widely Used Estimator." Journal of Hu- Hours Worked over the Life Cycle." Quarterly man Resources 32 (3): 441­62. Review of Economics and Finance 44: 102­21. Imbens, G.W., and J.D. Angrist. 1994. "Identifica- Rosenzweig, Mark R., and T.P. Schultz. 1983. "Esti- tion and Estimation of Local Average Treat- mating a Household Production Function: ment Effects." Econometrica, 62: 467­76. Heterogeneity and the Demand for Health In- Johansson, S.R. 1991. "The Health Transaction: puts and the Effects on Birthweight." Journal of The Cultural Inflation of Morbidity during the Political Economy 91 (5): 723­46. Decline in Mortality." Health Transition Review Roman, P.M. 1988. "Biological Features of Women's 1 (1): 39­68. Alcohol Use." Public Health Reports, 103: 628­ Kauhanen, J., G.A. Kaplan, D.E. Goldberg, and J.T. 37. Salonen. 1997. "Beer Binging and Mortality: Ryan, M. 1995. "Alcoholism and Rising Mortality in Results from the Kuopio Ischaemic Heart Risk the Russian Federation." British Medical Journal Factors Study, a Prospective Population-Based 310: 646­48. Study." British Medical Journal 315: 846­51. Schultz, T.P. 2005. "Productive Benefits of Health: Killingsworth, Mark R. 1983. Labor Supply, Cam- Evidence from Low-Income Countries." In bridge, MA: Cambridge University Press. Health and Economic Growth: Findings and Policy Leon, D.A., L. Chenet, V. Shkolnikov, S. Zakharov, Implications, ed. G. Lopez-Casasnovas, B. J. Shapiro, and G. Rakhmanova. 1997. "Huge Rivera, and L. Currais, pp. 257­285 Cam- Variation in Russian Morality Rates 1984­94: bridge, MA: MIT Press. Artefact, Alcohol or What?" Lancet, 350, Schultz, T. Paul, and A. Tansel. 1997. "Wages and 383­388. Labor Supply Effects of Illness in Côte d'Ivoire Lokshin, Michael, and M. Ravallion. 2005. "Search- and Ghana: Instrumental Variable Estimates for ing for the Economic Gradient in Self Assessed Days Disabled." Journal of Development Econom- Health." Development Research Group, World ics 53 (2): 251­86. Bank, Washington, DC. Science. 2006. "Mental Health Disabilities in the Manton, K.G. 1986. "Past and Future Life Ex- World." January 27: 459. pectancy Increases in Later Ages: Their Impli- Shkolnikov, V.M., and A. Nemtsov. 1997. "The Anti- cations for the Linkage of Chronic Morbidity, Alcohol Campaign and Variation in Russian Disability, and Mortality." Journal of Gerontology Mortality." In Premature Death in the New Inde- 41: 672­81. pendent States, ed. J.L. Bobadilla, C.A. Castello, Mete, Cem, and S.R. Cnobloch. 2006. "Socio- and F. Michell, Washington, DC: National economic Status and Health Outcomes: A Rela- Academy Press. tionship in Disguise." Processed paper. World Stern, Steven. 1989 "Measuring the Effect of Dis- Bank, Washington, DC. ability on Labor Force Participation." Journal of Michell J., and R. Burkhauser. 1990. "Disentangling Human Resources 24 (3): 361­95. the Effects of Arthritis on Earnings: A Simulta- Strauss, John, and D. Thomas. 1995. "Human Re- neous Estimate of Wage Rates and Hours sources: Empirical Modeling of Household and Worked." Applied Economic Letters 22: 1291­ Family Decisions." In Vol. 3A of Handbook of 1310. Development Economics, ed. J. Behrman, and 118 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union T.N. Srinivasan, pp. 1883­2023 Amsterdam, ing Paper 11871, National Bureau of Economic the Netherlands: Elsevier. Research, Cambridge, MA. Suhrcke, Marc, L. Rocco, D. Urban, M. McKee, S. Treml, Vladimir G. 1982. Alcohol in the USSR: A Sta- Mazzuco, and A. Steinherr. 2005. "Economic tistical Study. Durham, NC: Duke Press Policy Consequences of Noncommunicable Diseases Studies. and Injuries in the Russian Federation." Euro- Waaler, H.T. 1984. "Height, Weight, and Mortality: pean Office for Investment for Health and De- The Norwegian experience," Acta Medica Scan- velopment, World Health Organization, dinava 77, suppl. no. 679. Venice, Italy. Welch, Finis. 1997. "Wages and Participation." Tekin, E. 2002. "Employment, Wages, and Alcohol Journal of Labor Economics 15 (1) pt. 2: Consumption in Russia." Working Paper No. S77­S106. 432, IZA, Bonn, Germany. World Bank. 2005. Dying Too Young. Washington, Tesliuc, Emil D., and E. Zotova. 2004. "A Social DC: World Bank. Safety Net for the Poor." Paper prepared for World Health Organization. 1992. World Health Re- conference on targeting social program in Rus- port 1991. Geneva, Switzerland: World Health sia, World Bank, Washington, DC. Organization. Thirumurthy, Harsha, J. G. Zivin, and M. Goldstein. Yemtsov, Ruslan G., S. R. Cnobloch, and C. Mete. 2005. "The Economic Impact of AIDS Treat- 2006. Evolution of the Predictors of Earnings dur- ment: Labor Supply in Western Kenya." Work- ing Transition. Washington, DC: World Bank. CHAPTER 5 The Implications of Poor Health Status on Employment in Romania Cem Mete and Shirley H. Liu* Introduction factors such as the accessibility and quality of health care. The economic collapse and emergence of The linkage between the health status and widespread poverty in the transition countries socioeconomic characteristics of adults is found to during the early 1990s was followed by be strong in transition economies (Gilmore, respectable economic growth rates and signifi- McKee, and Rose 2002; Kopp, Skrabski, and cant reductions in poverty starting from the Szedmak 2000; Bobak et al. 1998; Bobak et al. late 1990s. During this recovery period, a key 2000; Leinsalu 2002; Wroblewska 2002; Thomp- policy question is the extent to which health son, Miller, and Witter 2003). But this literature inequalities in these countries will be an obsta- might be overstating the impact of socioeconomic cle against equitable (and sustainable) eco- characteristics on health because the estimated nomic growth and poverty reduction. Yet little relationships are subject to omitted-variables bias: is known about the strength of the relationship none of the papers listed above consider house- between poor health status and employment in hold health environment (e.g., humidity, cold, this context. Furthermore, it is not clear if the noise, polluted air, poor water quality, etc.) or observed health inequalities are driven prima- availability of health facilities in the community as rily by factors that cannot be influenced by pol- determinants of healthiness.1 icy makers in the short term (such as schooling This approach is inconsistent with the theo- attainment and household wealth), or by other retical formulations of the determinants of * Cem Mete is a senior economist at the Europe and Central Asia region of the World Bank. Shirley Liu is an assistant professor at the economics department of the University of Miami. 119 120 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union health status, which acknowledge the possible stances such as "a poor household cannot afford role of local environmental factors (Strauss and heating and safe water on a regular basis, and Thomas 1998). Ignoring local environmental there is no hospital nearby (which might sym- conditions and public health infrastructure bolize quality and availability of treatment), and could be acceptable for studies focusing on the members of the household are in poor industrialized countries--where basic health health status" simply as a black-box relationship infrastructure is universally available and most between poverty (or low levels of schooling) and individuals are able to shield themselves from health outcomes, or as supportive evidence for adverse environmental conditions.2 This is not social determinants of health theory--as elabo- necessarily the case for transition countries, rated by Wilkinson and Marmot (1999) to where poverty is widespread and there are explain variation in health outcomes across growing concerns about deteriorating public industrialized countries--which states that an health, central heating, and water services individual's relative social standing has a direct (Lampietti and Meyer 2002; World Bank influence on his or her health status. 2003a). Indeed, BEEPS interviews with firms in The studies that use household survey data 21 transition countries reveal serious concerns from transition countries also tend to overlook about the quality of water and public health ser- the economic implications of poor health. vices, especially in poorer countries (figures 1 While both cross-country growth regression and 2).3 Thus, there is a possibility that the setup (Bloom, Canning, and Sevilla 2003) and existing studies end up interpreting circum- micro-level evidence from other developing FIGURE 1 Quality of Water and Public Health Services in 21 Transition Countries Reported percentages are obtained by the summation of "slightly bad," "bad," and "very bad" responses 90 80 70 60 tn 50 Romania cer pe 40 30 Romania 20 10 0 0 1,000 2,000 3,000 4,000 5,000 6,000 2000 GDP per capita (constant 1995 US$) less-than-good water less-than-good public health services linear (less-than-good water) linear (less-than-good public health services) Source: 1999­2000 BEEPS survey interviews of 3,878 firms. The Implications of Poor Health Status on Employment in Romania 121 FIGURE 2 Quality of Water and Public Health Services in 21 Transition Countries Reported percentages are obtained by the summation "bad" and "very bad" responses 70 60 50 40 tn cer 30 pe Romania 20 Romania 10 0 0 1,000 2,000 3,000 4,000 5,000 6,000 2000 GDP per capita (constant 1995 US$) poor-quality water poor-quality public health services linear (poor-quality water) linear (poor-quality public health services) Source: 1999­2000 BEEPS survey interviews of 3,878 firms. countries (Schultz 2002) testify to the impor- and Schultz 2007), or by using panel data that tance of the population's health on the economy, would give one more options to tackle causality this relationship remains understudied in the issues (Hurd, McFadden, and Merrill 2001; transition economy context. Attanasio and Emmerson 2001; Mete 2004). The shortcomings in the literature arise not because these issues are considered to be unim- portant, but rather due to data limitations. This The Context: Romania in Transition research relies on two sets of data from Romania--the nationally representative 2000 The significant output collapse that Romania Living Conditions Survey, with a sample size of experienced in the early years of transition came 25,395 individuals, matched with community- to an end in 2000, with a real GDP growth rate level information from locality records of 1995.4 of 2 percent for that year, followed by yearly On the positive side, the variables that allow the GDP growth rates over 4 percent until 2003.5 formulation of a better-specified health func- The labor force participation rates declined in tion are also those that can be used as instru- the early 1990s, similar to the experience of mental variables to sort out the causal impact of other transition countries. If one focuses on indi- health on employment. But there is certainly viduals aged 15 to 64, the share of employed room for further improvement, either by using individuals declined from 72 percent in 1996 to a richer set of instrumental variables that cap- 68.5 percent in 2001 (which is slightly below the ture the childhood environment experienced by EU average of 69.2 percent).6 The health impli- the interviewed adults (see, for example, Mete cations of economic collapse were visible, lead- 122 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union ing to increases in mortality rates in the early Table 1 presents means and standard devia- and mid-1990s. But compared to the Russian tions of the variables included in the empirical experience, one- or two-year declines in life analysis, separately for younger individuals expectancy at birth can be considered to be mod- (defined as ages 25 to 39 here) and older individ- erate (for comprehensive reviews and country- uals (ages 40 to 59). About 16 percent of younger specific statistics, see Cornia and Paniccia 2000; individuals reported their health status to be less Bobadilla, Costello, and Mitchell 1997). than good, and about 71 percent of those who A health insurance fund was established in said their health was less than good worked. As 1997, and mandatory health insurance was intro- expected, older individuals were more likely to duced in 1999. The 2000 Living Conditions report "less-than-good" health status (41 per- Survey shows that about 75 percent of individu- cent), and a smaller percentage worked (62 per- als surveyed had enrolled in the health insurance cent). As for residing in dwellings with program. The Roma were less likely to be potentially undesirable health implications: enrolled (with a 34 percent enrollment rate), as exposure to excessive noise is most prevalent (31 well as the households headed by those with no percent), followed by polluted air/smells (17 per- schooling (with a 57 percent enrollment rate).7 cent), dwelling water not proper to drink/bad The 2003 Public/Private Transfers and Social color or taste (12 percent), cold (8 percent), and Capital Survey data reveal that 7.5 percent of excessive humidity (6 percent). households have some household members not As mentioned previously, the health impact registered with a family physician (10.4 percent of socioeconomic status variables might be in rural areas, 5.7 percent in urban areas). The overstated if socioeconomic status indicators are same survey reveals that the main reasons that correlated with health care facility availability make it difficult to get to a family physician is the and environmental condition indicators, and if distance to a medical facility (27.3 percent of these latter set of variables is excluded from the responses), followed by the long waiting time empirical model. While the next section evalu- during consultation (16.7 percent), and services ates the validity of this hypothesis using multi- being too expensive (15.8 percent). variate analysis, it is useful to display the variation of these variables by household wealth, as proxied by a household possessions index.8 Data and Descriptive Trends From this point forward, the term "wealthiest" refers to the wealthiest 25 percent of the popu- The Romania Living Conditions Surveys lation and the term "poorest" refers to the poor- (RLCS) are nationally representative household est 25 percent of the population. surveys that have been implemented annually Highlights from table 2 are as follows: in by the Romanian National Institute of Statistics rural areas, wealthier individuals are much more since 2000. The survey sample consisted of likely to have a hospital or a health center/poli- 10,521 households (25,395 individuals) in 2000. clinic available in their locality (availability is To incorporate information on public hospital 11.3 percent for the wealthiest, as opposed to and public health center/policlinic availability, 6.9 percent for the poorest). The urban resi- these data are linked to locality data files that dents have significantly better access to health contain information on 365 localities in Roma- facilities (with around 97 percent having a nia. The final data set contains information on health facility in their locality), regardless of individual-level (self-evaluated) health out- wealth status. comes and characteristics, household environ- Humid and cold environmental conditions ment (including exposure to cold, etc.), and are reported more often by the poor in general, community. but especially by the urban poor. About 14 per- The Implications of Poor Health Status on Employment in Romania 123 TABLE 1 Means and Standard Deviations of the Variables Examined in the Models of Health and Employment Status Ages 25 to 39 Ages 40 to 59 Mean Std dev. Mean Std dev. Health status (1 if reported health status as less than good) .159 .366 .412 .492 Health status (1 if reported chronic illness) .101 .30 .277 .447 Employment status (1 if working) .707 .455 .622 .485 Age 31.8 4.23 48.9 5.53 Gender (1 if male) .489 .499 .484 .499 Marital status (1 married) .715 .451 .829 .376 No schooling .010 .101 .011 .105 Primary .019 .138 .110 .313 Gymnasium (5­8) .119 .324 .264 .441 Vocational .230 .421 .237 .425 High school .486 .499 .215 .411 Post-high school or foreman .039 .192 .074 .261 College and higher .095 .293 .090 .286 Romanian .892 .310 .903 .295 Hungarian .068 .252 .068 .252 Roma .030 .172 .018 .134 Other .009 .094 .010 .099 Household possessions index .396 .162 .408 .166 Urban residence .538 .499 .531 .499 Never smoked in his/her life .554 .497 .617 .486 Availability of public hospital in locality .526 .499 .531 .499 Availability of public health center or polyclinic in locality .535 .499 .530 .499 Environment Humidity .067 .250 .055 .228 Cold .082 .274 .071 .256 Noise (traffic, commercial activity, industrial, etc.) .313 .464 .316 .465 Polluted air, smells .174 .379 .163 .369 Dwelling water not proper to drink/bad color or taste .123 .328 .119 .325 Number of observations 4,628 7,216 cent of urban poor report humidity in their reporting bias is likely to influence some other dwellings, and 15 percent report cold, com- indicators as well. For example, it may be that pared to 4.5 percent of wealthiest individuals the wealthy-poor gap would be even more pro- reporting humidity, and 6 percent reporting nounced if objective indicators of "exposure to cold. However, when considering noise and air cold weather" were available. pollution, this trend is reversed: wealthier indi- viduals report more complaints. Separate tabu- lations for urban and rural areas show Results significantly high noise and polluted air com- plaints in urban areas overall. An odd finding in Predictors of (Less-than-Good) Health this table is that "dwelling water undrinkable, Status bad color/taste" rates are most common among This section first provides a brief description of the wealthiest urban residents. This may be due findings from our preferred specification that to variations in acceptable standards of water includes household health environment vari- quality by socioeconomic status. In fact, such ables and health facility availability indicators. 124 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE 2 Health Environment Variables by Household Wealth Reported numbers are percentagesa Hospital and health Dwelling center/policlinic (HCP) water availability in the locality Polluted undrinkable, Wealth grouping Either Hospital HCP Humidity Cold Noise air bad color/taste Total 51.3 48.1 48.6 5.8 7.1 29.7 15.5 11.4 Poorest 25 percent 27.0 23.9 24.4 7.1 6.8 20.7 10.7 8.2 Middle 50 percent 57.9 54.5 55.1 5.6 7.7 31.0 17.0 11.7 Wealthiest 25 percent 70.0 67.5 67.5 4.0 4.4 42.9 18.1 16.7 Urban 97.1 95.9 94.4 7.7 10.1 42.5 23.3 16.3 Poorest 25 percent 94.5 92.8 92.0 13.7 14.8 36.9 22.4 14.4 Middle 50 percent 97.4 96.0 94.7 7.4 10.3 42.0 23.5 15.6 Wealthiest 25 percent 97.8 97.6 95.2 4.5 6.0 48.9 23.2 20.7 Rural 8.6 3.6 5.9 4.0 4.2 17.8 8.2 6.8 Poorest 25 percent 6.9 3.4 4.2 5.1 4.4 15.9 7.2 6.3 Middle 50 percent 9.6 3.7 6.7 3.4 4.5 17.4 9.0 7.0 Wealthiest 25 percent 11.3 4.1 9.3 2.9 1.2 30.5 7.2 8.4 a. The proxy for wealth is the household possessions index. The data for hospital and health center/policlinic availability come from administrative records. Other "health environment" information comes from the household survey questionnaire (as revealed by the household head). Next, we discuss the differences with a specifi- such as humidity, cold, noise, polluted cation that excludes these variables. All these air/smells, poor dwelling water (not proper to models are estimated separately for individuals drink/bad color or taste). between the ages of 25 to 39, and those between The signs of the estimated coefficients for the ages of 40 to 59. Finally, we turn to insights the "standard" variables are as typically found in from various alternative specifications that the literature, namely: (i) males are less likely to relax some of the assumptions associated with report less-than-good health; (ii) being married the basic model (estimating separate coeffi- is associated with reduced likelihood of report- cients for males and females, for urban and ing less-than-good health; (iii) those with a pri- rural residents, allowing interaction terms mary education or less are most likely to report between household wealth and health facility less-than-good health--the health impact of availability, etc). increased schooling is mixed for the age group 25 to 39, while a more steady trend exists for Basic (preferred) specification. The coefficient those aged 40 to 59; (iv) the Roma--who are estimates for the basic/preferred specification almost three times more likely to be poor com- are reported in table 3. The dependent variable pared to others in Romania (see World Bank takes the value 1 if an individual reports his or 2003b)--are less likely to report less-than-good her health status as less than good. It takes the health (one explanation for this counterintuitive value 0 otherwise. The explanatory variables are finding is cultural differences in interpreting age (and age-squared/100); gender; marital sta- what conditions qualify as less-than-good tus; schooling; ethnicity; household possessions health); (v) those who are wealthier are less index; urban/rural residence; an indicator for likely to report less-than-good health; (vi) urban having never smoked; availability of a public residence increases the likelihood of reporting hospital in the locality; availability of public less-than-good health, although the estimated health center/policlinic in the locality; and coefficient is not statistically significant for the exposure to undesirable environmental factors, 25 to 39 age group. The Implications of Poor Health Status on Employment in Romania 125 The estimates for other variables are as fol- probability of reporting less-than-good health is lows. If a person has never smoked, the likeli- 6.9 percentage points higher for the younger hood of reporting less-than-good health declines individuals and 9 percentage points higher for by 2.1 percentage points for younger individu- the older individuals. The estimated coefficients als, and by 3.3 percentage points for older indi- for "noise" and "polluted-air/smells" are smaller. viduals. The estimated coefficients are statistically significant at 10 and 5 percent levels How different are the coefficient estimates in an for younger and older individuals, respectively. underspecified model? The estimated relation- The larger smoking impact on health for older ship between socioeconomic status and health individuals might be in part because having may be biased if variations in health care avail- never smoked is more indicative of an older indi- ability and household environmental factors are vidual's permanent, long-term health investment not taken into account. In table 3, the figures in behavior. It could also be because younger and brackets are the coefficient estimates and |t|-sta- older individuals differ when it comes to the tistics for a model that includes all variables in the duration of, and exposure to, smoking; the aver- preferred specification except individual health age age for first-time smokers; and the likelihood behavior variables and health environment fac- (as well as timing) of quitting smoking. tors: never smoked; availability of public hospital Availability of a public hospital or a public in the locality; availability of public health cen- health center/policlinic in the locality is not a ter/policlinic in the locality; and presence of statistically significant predictor of health status humidity, cold, noise, polluted-air/smells, and of younger individuals. Accessibility of health bad dwelling water. As discussed previously, this care facilities may be a more important factor in is more or less the typical specification in the lit- determining the health status of the older pop- erature. How do the results compare for the vari- ulation, because the elderly are more likely to be ables that appear in both models? in poorer health status and have higher utiliza- Interestingly, the underspecified model pro- tion of health care facilities if in poor health. duces similar estimates for all variables in the Indeed, for those who are between the ages of model. The only difference to note is that in the 40 to 59, presence of a public hospital in the fully specified model, urban residence does not locality is associated with a 10.7 percentage have a statistically significant impact on health point reduction in the probability of reporting status of younger individuals, but in the under- less-than-good health. (This finding is driven specified model, the urban-residence effect is by the females in the sample, as discussed later.) statistically significant at the 1 percent level. The signs of the coefficients on "humidity," Thus, to the extent that the Romanian findings "cold," "noise," "polluted air/smells," and "bad can be generalized, it seems the shortcoming of dwelling water" are all in the expected direction. previous studies is not so much wrong estimates The only household health-environment vari- for the variables that are included in the models able that is not a statistically significant predictor (with the usual caveats about causality issues of health status (at the 10 percent level or better) that are often not tackled by studies that use is exposure to humidity. As for the remaining cross-sectional data sets), but rather the inabil- variables, the estimated coefficients are larger ity to provide insights about the role of health for the older individuals. Exposure to cold behavior, household health environment (cold, increases the probability of reporting less-than- humidity, etc.), and health facility availability on good health by 6.1 and 18.0 percentage points the health status of individuals. for younger and older individuals, respectively. Similarly, if the dwelling water is reported to be Alternative specifications. Further insights "not proper to drink/bad color or taste," the emerge from adopting more flexible specifica- 126 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union tions that: (i) consider the predictors of male coefficient estimates. For older individuals, the and female health status separately; (ii) consider impact of the household possessions index is urban and rural areas separately; (iii) allow for more pronounced in urban areas. In urban interaction terms between health facility avail- areas, cold, noise, and bad water are linked to ability indicators and household possessions less-than-good health reporting, while in rural index used as a proxy for household wealth; and areas, all household-environment variables (iv) use reported chronic illness instead of self- except air pollution have statistically significant evaluation of health status. The main findings impacts on health status in the expected are as follows. direction. Appendix tables A1 and A2 illustrate the esti- When interactions between the household mation results separately for males and females. possession index and health facility availability Rather than getting into a lengthy discussion indicators are introduced (not reported), we find about commonly observed gender differences in that hospital availability in a locality decreases self-reported health status, we will point out a the probability of less-than-good health more few interesting trends here. The variable "public for the wealthy. The interaction term between hospital availability" has a statistically significant the household possessions index and health cen- impact on health status only for females aged 40 ter/policlinic availability indicator does not have to 59. Older females who live in a locality that a statistically significant coefficient estimate. has a public hospital are 14.9 percentage points These findings are valid for both age groups less likely to report less-than-good health. As for considered by this study. The reason why the the two key household-environment variables wealthy benefit more from hospital availability with the largest impact on health status, expo- may have to do with health insurance enroll- sure to cold and poor dwelling water, the gender ment patterns, which reveal the disadvantaged differences are also visible. The impact of expo- situation of those with little or no schooling, as sure to cold on health status is considerably discussed previously in section 2. In this con- larger for males--8.3 and 20.2 percentage points text, the practice of informal payments and the for younger and older males, respectively, com- importance of "private connections" to solve pared to 4.3 and 15.8 percentage points for health problems in Romania deserve further younger and older females, respectively--while attention. Tabulations from the 2003 World the health implications of bad dwelling water are Bank Public/Private Transfers and Social Capi- larger for females--6.4 and 6.7 percentage tal Survey show that about 45 percent of house- points for younger and older males, respectively, holds that faced a health problem in the last year compared to 7.4 and 11.2 percentage points for reported offering "gifts" as part of the consulta- younger and older females. tion process (the probability of offering gifts When the basic model is estimated separately increased from 39 percent for the poorest quin- for rural and urban areas (not reported), the fol- tile of households to 51 percent for the wealthi- lowing key findings emerge. For younger indi- est quintile of households). The same survey viduals, poor-quality water has a negative reveals larger disparities when it comes to impact on health in both urban and in rural "knowing someone who can help solve a health areas, though the effect is larger in rural areas (a problem:" 34 percent of the poorest and 75 per- 9 percentage point versus 6 percentage point cent of the wealthiest quintiles of households increase in the likelihood of reporting less-than- replied positively to this question. good health). Humidity and air pollution in Finally, when the dependent variable takes urban areas are statistically significantly related the value 1 if the individual has a chronic illness to poor health. In contrast, in rural areas, cold (0 otherwise), we find that for younger individ- and noise variables have statistically significant uals, smoking behavior does not have a statisti- The Implications of Poor Health Status on Employment in Romania 127 cally significant coefficient estimate. Also for reporting of chronic illnesses as one of the deter- younger individuals, the only statistically signif- minants of self-reported health status. icant health-environment coefficient estimate is For brevity, here we discuss only the impact for "bad water." Availability of a health cen- of less-than-good health on employment. For ter/policlinic becomes statistically significant in the younger individuals, the less-than-good the expected direction. For older individuals, health coefficient is not statistically significant, everything looks similar except in this specifica- at the 10 percent level. But for those who are 40 tion: Among the household environment vari- to 59 years old, the causal impact of less-than- ables, only exposure to cold and noise have good health on employment is extremely large, statistically significant coefficients. reducing probability of employment by some- where between 41.2 percentage points and 56.7 The Impact of Less-than-Good Health percentage points, depending on the empirical on Probability of Being Employed specification that is used. This finding--i.e., a To single out the causal impact of (less-than- stronger linkage between health status and good) health status on employment, a two-stage employment for older individuals--is consistent conditional maximum likelihood (2SCML) with the literature (see the review by Currie and model is estimated, as proposed by Rivers and Madrian 1999). When these models are esti- Vuong (1998). The basic/preferred specification mated separately for males and females (appen- reported in the previous section serves as the dix tables A3 and A4), we find that the estimated first-stage equation. The inclusion of first-stage coefficient for less-than-good health is statisti- equation residuals in the second-stage specifica- cally significant at the 10 percent level for young tion allows for a specification test for the null males (but not for young females), reducing hypothesis that the health status variables are probability of employment by 41.4 percentage exogenous: the reported t-statistics show that points. For those between the ages of 40 and 59, this null hypothesis is rejected at the 1 percent the impact of less-than-good health on employ- level for the overall sample and for the female ment is larger for females, leading to a 58 per- sample, while it is rejected at the 5 percent level centage point decline in the probability of for the male sample. The set of instruments (the employment (as opposed to 54.3 percentage variables that are in the first-stage health equa- points for males). tion but not in the second-stage employment Putting the magnitude of the relationship equation) include: never smoked; availability of between health status and employment into con- public hospital in the locality; availability of pub- text is more difficult, though. Costa (1996) finds lic health center/policlinic in the locality; and that the labor force participation of men was exposure to humidity, cold, noise, polluted more responsive to BMI in the 1900s compared air/smells, bad dwelling water. The household to the present time. Because in many transition possessions index is also excluded from this countries a large share of jobs are in the indus- reduced-form employment model because the trial and agricultural sectors (as opposed to the direction of causality between employment and service sector), and because physical fitness is household wealth is unclear. Table 4 presents the arguably more important for industrial and agri- predictors of employment status, where the cultural sector jobs, one can expect a larger dependent variable takes the value 1 if the indi- health effect on employment in transition coun- vidual works, and the value 0 otherwise. Esti- tries. On the other hand, a large public sector mates are reported separately for the two age presence might weaken the relationship between groups (25 to 39 and 40 to 59). The estimates in health status and employment. If one reviews the brackets are from an alternative specification comprehensive list of studies and summary that (in the first-stage equation) considers the results documented by Currie and Madrian, a 128 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE 3 Predictors of Less-than-Good Health Status The dependent variable takes the value 1 if self-reported health status is less than good. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. The numbers in brackets [] are those from an alternative specification that does not include the health facility, health behavior, and health environment variables in bold. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Age .013 0.68 .071*** 3.47 [.013] [0.63] [.065***] [3.15] Age squared / 100 .004 0.13 .052** 2.51 [ .002] [0.06] [ .046**] [2.24] Gender (1 if male) .053*** 4.58 .123*** 8.68 [ .046***] [4.23] [ .108***] [8.65] Marital status (1 married) .031** 2.41 .042** 2.55 [ .031**] [2.36] [ .045***] [2.78] No schoolinga -- -- -- -- Primary .128*** 4.08 .161*** 2.84 [ .131***] [4.15] [ .160***] [2.83] Gymnasium (5­8) .166*** 5.95 .177*** 3.09 [ .171***] [6.08] [ .171***] [3.00] Vocational .209*** 6.42 .169*** 2.92 [ .213***] [6.50] [ .163***] [2.82] High school .320*** 6.89 .218*** 3.82 [ .326***] [6.98] [ .214***] [3.75] Post-high school or foreman .151*** 6.00 .235*** 4.16 [ .154***] [6.06] [ .222***] [3.91] College and higher .181*** 7.68 .288*** 5.33 [ .184***] [7.71] [ .283***] [5.22] Romaniana -- -- -- -- Hungarian .052** 2.51 .027 1.15 [ .056***] [2.70] [ .034] [1.45] Roma .070** 2.41 .075 1.64 [ .071**] [2.41] [ .054] [1.20] Other .056 1.02 .114* 1.86 [ .062] [1.13] [ .111*] [1.83] Household possessions index .076* 1.89 .289*** 6.13 [ .076*] [1.90] [ .289***] [6.20] Urban residence .045 1.58 .138*** 3.93 [.047***] [4.14] [.100***] [7.42] Never smoked in his/her life .021* 1.78 .033** 2.25 Availability of public hospital in locality .026 0.86 .107*** 3.03 Availability of public health center or a policlinic in locality .004 0.15 .029 0.99 Environment Humidity .032 1.42 .023 0.80 Cold .061*** 2.81 .180*** 6.91 Noise (traffic, commercial activity, industrial, etc.) .022* 1.82 .045*** 3.26 Polluted air, smells .025* 1.71 .029* 1.67 Dwelling water not proper to drink/bad color or taste .069*** 4.02 .090*** 4.72 F-test (p-value): Joint significance of the identifier variables (bold above) 0.0000 0.0000 Overall significance (p-value) 0.0000 0.0000 Number of observations 4,628 7,216 a. Significance levels: * -- 10%; ** -- 5%; *** -- 1%. The Implications of Poor Health Status on Employment in Romania 129 TABLE 4 Predictors of Employment The dependent variable takes the value 1 if the individual is employed. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. The estimates in brackets [] are from an alternative specification that (in the first-stage equation) considers the reporting of chronic illnesses as a determinant of self-reported health status. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Estimated less-than-good health .247 1.40 .567*** 6.59 [ .217] [6.81] [ .412***] [19.2] Residual from first-stage regression .087*** 4.75 .190*** 15.3 [ .025] [1.11] [ .091***] [6.01] Age .103*** 3.96 .090*** 4.23 [.103] [3.97] [.084***] [4.03] Age squared / 100 .137*** 3.35 .104*** 4.92 [ .138] [3.39] [ .102***] [4.82] Gender (1 if male) .185*** 11.5 .178*** 12.0 [.188] [13.2] [.196***] [15.4] Marital status (1 married) .064*** 3.67 .021 1.24 [.064] [3.93] [ .012] [0.74] No schoolinga -- -- -- -- Primary .193** 2.50 .081 1.40 [.201] [3.21] [.111**] [1.98] Gymnasium (5­8) .170* 1.89 .031 0.53 [.183] [2.82] [.065] [1.14] Vocational .264*** 2.96 .085 1.44 [.277] [4.39] [.121**] [2.13] High school .329*** 3.01 .158*** 2.68 [.349] [4.60] [.200***] [3.68] Post-high school or foreman .269*** 4.14 .173*** 2.91 [.272] [5.94] [.211***] [3.95] College and higher .309*** 4.56 .275*** 4.94 [.314] [7.06] [.311***] [6.59] Romaniana -- -- -- -- Hungarian .056** 2.00 .089*** 3.61 [.058] [2.20] [ .083***] [3.39] Roma .108** 2.31 .083* 1.76 [ .104] [2.36] [ .080*] [1.69] Other .232*** 3.02 .103 1.62 [ .231] [3.02] [ .087] [1.39] Urban residence .099*** 5.97 .044*** 2.87 [.097] [6.72] [ .058***] [4.24] Overall significance (p-value) 0.0000 0.0000 Number of observations 4,628 7,216 a. Significance levels: * -- 10%; ** -- 5%; *** -- 1%. 41.2 to 56.7 percentage point reduction in the Conclusions probability of work due to less-than-good health status is a large effect (although some studies This paper argues that a recent wave of studies report effects that come close to that. See for aiming to display correlates of self-reported example Bound et al. 1996) and thus the findings poor health status in transition countries have seem to support the first possibility.9 been too quick to dismiss the role of environ- 130 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union mental factors and health facility availability on significant relationship. Availability of a health health status. In the context of transition center/policlinic in the locality is not correlated economies, the identification of those who are with health status of males or females. Further- most vulnerable to adverse environmental con- more, in rural areas, the distribution of health ditions and those with limited access to health facilities favors the wealthy. As can be expected, facilities--jointly with the extent to which these urban-area residents have similar access to disadvantages are associated with poor health health facilities, regardless of differences in outcomes--provide indispensable information socioeconomic characteristics. Also, the wealth- for the design of health sector reforms that build ier individuals' health status is more responsive on comprehensive but often unsustainable pre- to having a hospital in the locality. These transition health service delivery systems. findings--when combined with survey statistics This study shows that while schooling attain- that reveal the prevalence of informal payments ment and household wealth are associated with in the health sector and the importance of "hav- reduced likelihood of reporting less-than-good ing contacts to solve health problems"--suggest health status, it is also true that the members of that a poor household's geographical proximity households that report exposure to cold weather to a public health facility, by itself, does not nec- and poor dwelling water are much more likely essarily mean this household has access to ade- to report less-than-good health (even after tak- quate public health services. ing into account a variety of individual, house- Many transition countries, including Roma- hold, and community characteristics). If one nia, have experienced respectable economic subscribes to the view that "socioeconomic growth rates starting from the late 1990s. But characteristics are all that matter, an individual's rising health inequalities in these countries social standing in the society is critical for his or could well be the main obstacle for equitable health," then the implication is that unless (and sustainable) economic growth and poverty socioeconomic differences are eliminated, reduction because the relationship between health inequalities are here to stay. But there are health status and employment is extremely other things that matter for transition strong: for those individuals who are between economies, and they should not be ignored. the ages of 40 and 59, we find that poor health While it may be difficult, unrealistic, and even status leads to somewhere between a 41.2 and undesirable to change the distribution of a pop- 56.7 percentage point decline in the probability ulation's socioeconomic characteristics in the of employment. Reflecting on the past, the pop- short term, one can certainly aim for policies ulation's deteriorating health status during tran- that improve the infrastructure for delivering sition might have contributed significantly to adequate heat and clean water to the general the decline in employment rates. For example, population. the employment rate in Romania in 2000 was The findings regarding the correlation 59 percent, which is considerably lower than the between health facility availability and self- EU average of 65 percent. As for the upcoming reported health status are less clear cut. A rela- challenges in the near future, the population's tionship between availability of a hospital in the poor health status emerges as a serious obstacle locality and health outcomes is found for older to achieving higher employment rates, and by females. But for males, there is no statistically extension, a more equal distribution of income. The Implications of Poor Health Status on Employment in Romania 131 Appendix TABLE A1 Predictors of Less-than-Good Health Status of Males The dependent variable takes the value 1 if self-reported health status is less than good. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Age .004 0.14 .062** 2.17 Age squared / 100 .008 0.19 .045 1.53 Marital status (1 married) .036** 2.15 .060** 2.39 No schoolinga -- -- -- -- Primary .128*** 4.41 .251*** 3.24 Gymnasium (5­8) .157*** 5.16 .251*** 3.05 Vocational .259*** 5.79 .257*** 2.92 High school .341*** 5.90 .278*** 3.44 Post-high school or foreman .136*** 4.72 .280*** 3.68 College and higher .165*** 6.24 .313*** 4.27 Romaniana -- -- -- -- Hungarian .055** 2.00 .017 0.51 Roma .040 1.01 .048 0.77 Other .029 0.39 .092 1.17 Household possessions index .022 0.41 .312*** 4.66 Urban residence .014 0.39 .127*** 2.67 Never smoked in his life .018 1.24 .039** 2.23 Availability of public hospital in locality .009 0.22 .064 1.35 Availability of public health center/policlinic in locality .001 0.02 .009 0.24 Environment Humidity .004 0.15 .011 0.29 Cold .083*** 2.68 .202*** 5.53 Noise (traffic, commercial activity, industrial, etc.) .023 1.38 .052*** 2.71 Polluted air, smells .027 1.36 .017 0.71 Dwelling water not proper to drink/bad color or taste .064*** 2.75 .067** 2.52 F-test (p-value): Joint significance of the identifier variables (bold above) 0.0019 0.0000 Overall significance (p-value) 0.0000 0.0000 Number of observations 2,262 3,493 a. Significance levels: * -- 10%; ** -- 5%; *** -- 1%. 132 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE A2 Predictors of Less-than-Good Health Status of Females The dependent variable takes the value 1 if self-reported health status is less than good. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Age .027 0.89 .075** 2.56 Age squared / 100 .021 0.45 .054* 1.83 Marital status (1 married) .021 1.01 .028 1.27 No schoolinga -- -- -- -- Primary .090 1.37 .082 1.06 Gymnasium (5­8) .162*** 3.14 .113 1.47 Vocational .168*** 3.13 .110 1.39 High school .280*** 3.70 .161** 2.07 Post-high school or foreman .169*** 3.95 .188** 2.32 College and higher .190*** 4.37 .262*** 3.38 Romaniana -- -- -- -- Hungarian .050 1.60 .038 1.13 Roma .097** 2.21 .089 1.36 Other .081 1.00 .144 1.53 Household possessions index .145** 2.39 .258*** 3.91 Urban residence .078* 1.82 .149*** 2.92 Never smoked in her life .032* 1.75 .022 0.89 Availability of public hospital in locality .048 1.04 .149*** 2.89 Availability of public health center/policlinic in locality .006 0.17 .049 1.15 Environment Humidity .070** 2.01 .058 1.39 Cold .043 1.39 .158*** 4.32 Noise (traffic, commercial activity, industrial, etc.) .022 1.26 .038* 1.94 Polluted air, smells .020 0.92 .042* 1.74 Dwelling water not proper to drink/bad color or taste .074*** 2.98 .112*** 4.18 F-test (p-value): Joint significance of the identifier variables (bold above) 0.0010 0.0000 Overall significance (p-value) 0.0000 0.0000 Number of observations 2,366 3,723 a. Significance levels: * -- 10%; ** -- 5%; *** -- 1%. The Implications of Poor Health Status on Employment in Romania 133 TABLE A3 Predictors of Employment of Males The dependent variable takes the value 1 if the individual is employed. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Estimated less-than-good health .414* 1.91 .543*** 5.19 Residual from first-stage regression .057** 2.38 .189*** 12.3 Age .048 1.50 .054** 2.04 Age squared / 100 .054 1.07 .067** 2.53 Marital status (1 married) .134*** 6.51 .134*** 5.69 No schoolinga -- -- -- -- Primary .157** 2.04 .049 0.66 Gymnasium (5­8) .120 1.17 .017 0.23 Vocational .185 1.63 .006 0.07 High school .213* 1.65 .034 0.42 Post-high school or foreman .176** 2.15 .023 0.28 College and higher .192** 2.17 .141* 1.92 Romaniana -- -- -- -- Hungarian .048 1.40 .079** 2.52 Roma .169*** 2.91 .162*** 2.59 Other .142 1.52 .112 1.47 Urban residence .058*** 2.86 .073*** 3.95 Overall significance (p-value) 0.0000 0.0000 Number of observations 2,262 3,493 a. Significance levels: * --- 10%; ** --- 5%; *** --- 1%. 134 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union TABLE A4 Predictors of Employment of Females The dependent variable takes the value 1 if the individual is employed. It takes the value 0 otherwise. Marginal effects from the probit model and |t|-ratios are reported below. Ages 25 to 39 Ages 40 to 59 Marginal eff. |t|-value Marginal eff. |t|-value Estimated less-than-good health .115 0.43 .580*** 4.60 Residual from first-stage regression .109*** 4.04 .169*** 9.37 Age .141*** 3.57 .133*** 4.30 Age squared / 100 .196*** 3.17 .149*** 4.84 Marital status (1 married) .033 1.18 .130*** 5.57 No schoolinga -- -- -- -- Primary .193 1.38 .084 1.01 Gymnasium (5­8) .196 1.32 .019 0.22 Vocational .310** 2.21 .136 1.60 High school .385** 2.26 .216** 2.52 Post-high school or foreman .353*** 3.22 .288*** 3.34 College and higher .413*** 3.68 .364*** 4.27 Romaniana -- -- -- -- Hungarian .060 1.41 .091*** 2.62 Roma .071 0.94 .039 0.58 Other .292** 2.52 .066 0.68 Urban residence .129*** 5.17 .020 0.86 Overall significance (p-value) 0.0000 0.0000 Number of observations 2,366 3,723 a. Significance levels: * -- 10%; ** -- 5%; *** -- 1%. The Implications of Poor Health Status on Employment in Romania 135 Notes Economic Research Working Paper Series 8241. Cambridge, MA. 1. One possibility is to employ econometric tech- Bloom, D., D. Canning, and J. Sevilla. 2004. "The niques that account for unobserved heterogene- Effect of Health on Economic Growth: A Pro- ity at the village level, to take into account health duction Function Approach." World Develop- environment in the community (see, for exam- ment 32 (1): 1­13. ple, Skoufias [1998], who focuses on determi- Bobadilla, J.L., C.A. Costello, and F. Mitchell. 1997. nants of child health in Romania relying on the Premature Death in the New Independent States. 1994 Integrated Household Survey data). But Washington, DC: National Academy Press. this empirical strategy does not take into account Bobak, M., H. Pikhart, C. Hertzman, R. Rose, and health environment at the household level, and M. Marmot. 1998. "Socioeconomic Factors, it is silent about the presence or lack of a rela- Perceived Control, and Self-Reported Health tionship between access to health care and in Russia. A Cross-Sectional Survey." Social Sci- health outcomes. ence & Medicine 47 (2): 269­79. 2. Indeed there is empirical evidence revealing the Bobak, M., H. Pikhart, R. Rose, C. Hertzman, and disconnect between access to health care and M. Marmot. 2000. "Socioeconomic Factors, mortality rates in industrialized countries (see Material Inequalities, and Perceived Control in Deaton 1999 for a review). Self-Rated Health: Cross-Sectional Data from 3. The only country transition that we excluded Seven Post-Communist Countries." Social Sci- from the BEEPs data set is Slovenia, which is a ence & Medicine 51: 1343­50. major outlier with a 2000 GDP per capita of Bound, J., M. Schoenbaum, and T. Waismann. 1996. $11,646 (although it is not an outlier in terms of "Race Differences in Labor Force Attachment overall trend, because a small share of firms in and Disability Status." The Gerontologist 36: Slovenia reported bad-quality water and public 311­21. health services). Cornia, G.A., and R. Paniccia. 2000. The Mortality 4. We prefer to use lagged community characteris- Crisis in Transitional Economies. New York: Ox- tics to model health outcomes, although note ford University Press. that none of the findings change if one uses Currie, J., and B.C. Madrian. 1999. "Health, Health locality records of 2000. Insurance, and the Labor Market." In Handbook 5. Real GDP had grown by about 4 percent, thanks of Labor Economics Volume 3C, ed. O.C. Ashenfel- to an election-year push in 1996, which did not ter and C. Card, 3309­3407. New York: Elsevier. turn out to be sustainable. Deaton, Angus. 1999. "Inequalities in Income and 6. Data source: Romania Labor Force Surveys Inequalities in Health." National Bureau of (1996 and 2002). Economic Research Working Paper Series 7. For more information, see 2003 Romania 7141. Cambridge, MA. Poverty Assessment (The World Bank). Gilmore, A.B.C., M. McKee, and R. Rose. 2002. 8. The household possessions index varies between "Determinants of and Inequalities in Self- 0 and 1, and it takes into account ownership of Perceived Health in Ukraine." Social Science & cooker, stove, refrigerator, freezer, washing Medicine 55: 2177­88. machine, sewing machine, vacuum, TV, radio, Hurd, M.D., D. McFadden, and A. Merrill. 2001. video, microwave, PC, cell phone, and car. "Predictors of Mortality among the Elderly." In 9. As Currie and Madrian (1999) observe, there is Themes in the Economics of Aging, ed. D. Wise, no consensus on the magnitude of effects in the 171­97. Chicago: University of Chicago Press. industrialized country literature--modeling and Kopp, M.S., A. Skrabski, and S. Szedmak. 2000. different approaches to measuring health status "Psychosocial Risk Factors, Inequality, and emerge as main obstacles against making com- Self-Rated Morbidity in a Changing Society." parisons. Social Science & Medicine 51: 1351­61. Lampietti, Julian A., and Anke S. Meyer. 2002. Cop- ing with the Cold. Heating Strategies for Eastern Europe and Central Asia's Urban Poor. Washing- References ton, DC: World Bank. Leinsalu, M. 2002. "Social Variation in Self-Rated Attanasio, O., and C. Emmerson. 2001. "Differen- Health in Estonia: A Cross-Sectional Study." tial Mortality in the UK." National Bureau of Social Science & Medicine 55: 847­61. 136 Economic Implications of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union Mete, Cem. 2005. "Predictors of Elderly Mortality: Skoufias, Emmanuel. 1998. "Determinants of Child Health Status, Socioeconomic Characteristics, Health during the Economic Transition in Ro- and Social Determinants of Health." Health mania." World Development 26 (11): 2045­56. Economics 14 (2): 135­48. Strauss, John, and Duncan Thomas. 1998. "Health, Mete, Cem, and T.P. Schultz. 2007. "Health and La- Nutrition, and Economic Development." Jour- bor Force Participation of the Elderly in Tai- nal of Economic Literature 36 (2), 766­817. wan." In Riding the Age Waves: Allocating Public Thompson, R., N. Miller, and S. Witter. 2003. and Private Resources across Generations, ed. A.H. "Health-Seeking Behavior and Rural/Urban Gauthier, C. Chu, and S. Tuljapurkar. New Variation in Kazakhstan." Health Economics 12: York: Kluwer. 553­64. Rivers, D., and Q. Vuong. 1988. "Limited Informa- Wilkinson, Richard G., and Michael Marmot. 1999. tion Estimators and Exogeneity Tests for Simul- Social Determinants of Health. New York: Oxford taneous Probit Models." Journal of Econometrics University Press. 39 (3): 347­66. World Bank. 2003a. Water Resources Management in Schultz, T.P. 2002. "Wage Gains Associated with South Eastern Europe. Vols. 1 and 2. Washing- Height as a Form of Health Human Capital." ton, DC: World Bank. American Economic Review 92 (2): 349­53. ------. 2003b. Romania Poverty Assessment. Vols. 1 Schultz, T.P., and A. Tansel. 1997. "Wage and Labor and 2. Washington, DC: World Bank. Supply Effects of Illness in Côte d'Ivoire and Wroblewska, W. 2002. "Women's Health Status in Ghana: Instrumental Variable Estimates for Poland in the Transition to a Market Econo- Days Disabled." Journal of Development Econom- my." Social Science & Medicine 54: 707­26. ics 53 (2): 251­86. Eco-Audit Environmental Benefits Statement The World Bank is committed to Saved: preserving endangered forests and natural resources. The Office of the Publisher has · 6 trees chosen to print Economic Implications of · 353 lbs. of solid waste Chronic Illness and Disability in Eastern Eu- · 2,135 gallons of water rope and the Former Soviet Union on recy- · 651 lbs. of net cled paper including 25% post-consumer greenhouse gases recycled fiber in accordance with the rec- · 4 million BTUs of total ommended standards for paper usage set energy by the Green Press Initiative, a nonprofit program supporting publishers in using fiber that is not sourced from endangered forests. For more information, visit www.greenpressinitiative.org. This study is part of a series undertaken by the Europe and Central Asia region of the World Bank. The series draws on original data, the World Bank's operational experience, and the extensive literature on the Region. Poverty, jobs, trade, migration, energy, and productivity will be among the topics covered. Disability is an important issue for the transition countries of Eastern Europe and the former Soviet Union. Not only is a significant portion of their population either in poor health or disabled--with implications for labor force participation and productivity--but their aging demo- graphics project an increase in the share of disabled people, raising concerns about the sustainability of social protection programs. Thus, if these heavily resource-strapped countries fail to deal in an efficient manner with disability and health issues in their population, they could face serious challenges to their efforts to achieve stronger economic growth and improved living standards. Because the economic drivers and costs of poor health status and disabilities in this region are not well documented, Economic Implica- tions of Chronic Illness and Disability in Eastern Europe and the Former Soviet Union aims to close this gap by leveraging household survey data from a large number of transition countries, analyzing the poverty- disability relationship and the linkages between disability and employ- ment, earnings, children's school enrollments, and adults' time-use patterns. Altogether, disability appears to have stronger negative effects on the economic and social well-being of the population in these countries as compared with industrialized countries. The main reasons are the prevalence of a large informal sector, the relatively weak targeting performance of the existing social assistance programs, and the lack of broad-based insurance mechanisms to protect individuals against loss of income due to unexpected illnesses. Addressing these weaknesses is the challenge facing policy makers and the population at large in the region, through the definition and enactment of a deep, well-coordinated, cross-sectoral reform agenda. This book will be useful for policy makers and development officials working to improve living standards in the Eastern Europe and the former Soviet Union. ISBN 978-0-8213-7337-8 SKU 17337