, ZAA -M . . . f 10' '} ' 4'6'[ 'i~' ' / < | ' ''-9 % < ' AA'S .< - .~f t ipJ . -.o - L. WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES Viewpoints SECURING OUR FUTURE IN A GLOBAL ECONOMY by David de Ferranti Guillermo E. Perry Indermit S. Gill Luis Serven with Francisco H. G. Ferreira Nadeem Ilahi William F Maloney Martin Rama THE WORLD BANK WASHINGTON, D.C. Copyright © 2000 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing June 2000 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attrituted in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they repre- sent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do nct imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such bo.sndaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy porrions for class- room use is granted through the Copyrighr Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. David de Ferranti is vice president and Guillermo E. Perry is chief economist in the World Bank's Latin America and the Caribbean Regional Office. Indermit S. Gill is lead economist and Luis Serven is lead specialist in the World Bank's Latin America and the Caribbean Regional Office. Francisco Ferreira is professor of economics at the Pontifical Catholic University-Rio de Janeiro, Brazil. Nadeem Ilahi is an econo- mist with the International Monetary Fund. William E Maloney is senior economist in the World Bank's Latin America and the Caribbean Regional Office, and Martin Rama is senior economist in the World Bank's Development Research Group. Cover design by Janelle Welch/The Magazine Group Cover photography: Globe: NASA Goddard Space Flight Center; inset photos, clockwise from top: Catherine de Torquat/Sygma; Bill Pogue/Stone; Robert Frerck/Stone; Margo Pinkerton/Liaison Agency. de Ferranti, David. Securing our future in a global economy Guillermo E. Perry. p. cm. - (World Bank Latin American and Caribbean Studies. Includes bibliographical references. ISBN 0-8213-4730-6 Library of Congress pending Contents Acknowledgments .................................................................. ix Acronyms and Abbreviations ................................................................. xi Chapter 1: Opportunity and Risk in a Globalized Latin America and the Caribbean ............................ 1 Latin America and the Caribbean's Performance Improved in the 1990s ..................................... I Opportunities and Risks ...................................................................... 1 Why Economic Insecurity? . ..................................................................... 3 This Report ......................................................................... .... 6 LAC's Volatility is High-But has not Risen in the 1990s .. 6 Dealing with Economic Insecurity Requires a Comprehensive Insurance Approach ........................... . 7 LAC's Volatility Arises from Multiple Sources-Domestic and Foreign .. 7 Governments Can Do Much to Reduce Volatility-Even in a Globalized Economy .......................... . 7 External Risks Can Be Reduced by Diversification and Liquidity Management .. ................. 8 Anticyclical Macroeconomic Policies Ease Adjustment to Shocks ....... ............... 8 Deeper and Stronger Financial Systems Are a Key Part of Social Protection Policies .. 8 Deep Crises are Particularly Damaging for the Poor .. 8 The Poor Try to Protect Their Long-Term Welfare in Crises-As Long as Their Assets Permit . . 9 New Income Support Programs for the Unemployed Need to be Established . ............................... 9 Administrative Capacity and Labor Policies are Key in the Choice of Instruments .................. 9 Self-Insurance for Slow-Reforming Economies ..................................................... 10 Unemployment Insurance for Advanced Reforming Economies .. 10 Public Works Programs Provide Insurance Support for Informal Sector Workers .. 10 Targeted Programs for the Poor Need to be Better Protected in Downturns .. 10 Save in Good Times to Finance Social Spending in Bad Times .................. ....................... 10 Supranational Action and the Role of the International Financial Institutions ........... ..11................. 1 Myths and Realities About Economic Volatility ..1............................ 11 Securing Our Future ................................................................. 11 Notes .. ............................................................... 11 Chapter 2: Economic Insecurity in Latin America and the Caribbean: The Stylized Facts ....................... 13 Methodological Considerations . ................................................................ 13 Aggregate Volatility in Latin America ............................................................. 15 Volatility in Aggregate Output ................................................................ 15 Volatility in Aggregate Consumption . .......................................................... 15 Differences in Volatility Across Countries Within the Region ............................................ 16 Regional Trends in Economic Volatility . ........................................................ 18 Transmission of Aggregate Volatility to the Labor Market ............................................... 21 Are Latin American Workers Facing More Risk? ..................................................... 23 Changes in Earnings Volatility . ............................................................... 23 Trends in Unemployment Rates . .............................................................. 24 Changes in Turnover Rates ................................................................... 26 Rise in Informality ..................................................................... 27 iii SECURING OUR FUTIURE IN A GLOBAL ECONOMY The Deterioration in Prospects of Older Workers ................................................... 31 Changes in Income Distribution ......... .................................................... 32 C onclusion . . . . . . .. . . . . . .. . .. . . .. . .. . .. . . .. . .. . . . . . .. . . . . . .. . .. . . . . . .. . .. . . .. . .. . .. . . .. . 32 N otes . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Chapter 3: Designing Social Policy When People Face Risk: A Conceptual Framework ......................... 37 The Need for Sound Analysis . .................................................................. 37 Approach and Implications . .................................................................... 38 Advantages of a Disciplined Approach ............................................................. 39 Clearer Rationale for Government Action ........................................ ............... 42 Useful Insights . ......................................................................... 42 A Powerful Tool for Organization . .............................................................. 43 C onclusion . . .. . . .. . .. . . . . . .. . . . . . .. . .. . . . . . .. . .. . . .. . .. . . .. . .. . . . . . .. . .. . . . . . . .. .. . . . . . 43 N otes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 4 Chapter 4: Macroeconomic Volatility in Latin America and the Caribbean: Causes and Remedies ................ 45 Sources of Aggregate Volatility in LAC ............................................................ 45 External Shocks ........................................................................... 45 Macroeconomic Policy Volatility ............................................................... 49 Absorption and Amplification of Shocks: The Importance of Financial Markets ............................... 50 Weak Links with World Financial Markets ....................................................... 52 Shallow Domestic Financial Markets ............... ............................................ 53 The Combination Increases the Likelihood of Economic Crises ......................................... 54 Other Amplification Mechanisms . .............................................................. 57 Summary ............................................................................. 61 Policies to Deal with Aggregate Volatility in LAC . ................................................ 61 Terms of Trade Risk ........................ ............................................... 65 Capital Flows ............................................................................. 66 The Financial System ....6................................................................. 67 Fiscal Policy ........................................................................... . 67 Monetary and Exchange Rate Policy ....... .................................................... 67 Supranational Action . ...................................................................... 68 A n n ex I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 N otes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 Chapter 5: The Response of LAC Households to Economic Shocks ...................................... 73 The Risk of Unemployment: Who is Most Affected? .................................................. 73 Are the Poor Most Likely to Become Unemployed? ................................................. 73 Are Older, Less-Educated Men More Likely to Become Unemployed? .................................... 73 Is There No Unemployment in the Informal Sector? ................................................ 74 Who Becomes Unemployed and for How Long? ................................................... 76 Household Responses to Income Shocks: Findings of Panel Studies ........................................ 76 Do Aggregate Shocks Hurt the Poor More than the Rich? ............................................ 78 Does Ownership of Assets Reduce Vulnerability to Aggregate Shocks? ............ ...................... 79 Do the Poor Engage in Self-Destructive or Myopic Coping? ........... ................................ 81 Are All Downturns the Same in their Effects? ...................................................... 83 Conclusion .......................................................... 85 Notes .......................................................... 86 Chapter 6: Helping Workers Deal with the Risk of Unemployment ....................................... 89 A Typology of Programs . .......................................................... 90 Main Findings ...................................................... . .. 92 Who Is Covered by these Programs? .......................................................... 95 How Much Do the Programs Cost? ...................... 96 What Are the Efficiency Effects and Insurance Benefits? ........................................... .. 97 What Are the Main Weaknesses of these Programs? ................................................. 98 Policy Implications ........................................................... 98 Self-Protection: The State of Labor Markets ......................................................... 99 Self-Insurance: Individual Savings Accounts ......................................................... 99 lv CONTENTS "Market-Type" Insurance: Risk-Pooling Programs ................................................... 101 Severance Pay Provisions as Insurance for Unemployment ............................................ 102 Public Works Programs as Insurance for Unemployment ............................................ 102 "Conventional" Unemployment Insurance ....................................................... 103 C onclusion ......................................................................... 107 Notes ......................................................................... 108 Chapter 7: Helping Poor Households Deal Better with Economic Crises .................................. 111 Social Spending Over Economic Cycles . .......................................................... 113 How Do Governments Vary Spending Over the Economic Cycle? ...................................... 113 The Importance of Political Factors . ........................................................... 114 The Quality of Social Services Over the Cycle .................................................... 116 Targeted Spending During Booms and Busts ........... ..................................... ...... 119 Designing Economic Policy Under Political Constraints ........................................ ...... 121 The Long-Term Goal of Social Policy Must be to Improve the Distribution of Assets ........................ 122 Targeted Programs Should be Permanent and Better Protected During Crises ............................ 122 Keeping Increases in Social Spending Moderate in Good Times is Important Too ........................... 122 International Financial Institutions Can Help Overcome Political Constraints to Insurance ................... 122 Conclusion .......................................................................... 123 Notes ..................................................................... 125 Bibliography ..... 127 Boxes Box 2.1 Defining and Measuring Insecurity .14 Box 2.2 Trends in Aggregate Volatility in Latin America and the Caribbean .20 Box 2.3 Why Might Workers Prefer to be Precarious or Unprotected? .31 Box 2.4 Income Mobility and Risk in Two Countries .34 Box 3.1 Market Insurance, Self-Insurance, and Self-Protection: Distinguishing Features and Examples .39 Box 3.2 A Theory of Comprehensive Insurance .40 Box 3.3 The Framework in "Real Life" Situations .41 Box 4.1 Excess Sensitivity to Disturbances: The Case of Chile .52 Box 4.2 Weak Financial Markets and Volatility: A Framework .59 Box 4.3 The Welfare Cost of Volatility and the Gains from International Risk-Sharing .62 Box 5.1 Informal Self-Employment: Precarious Workers or Voluntary Entrepreneurs? .75 Box 5.2 Data Sets and Methodologies Used .77 Box 5.3 Where Does Time Go During a Crisis? .83 Box 5.4 Do Families Mortgage Their Children's Future' Other Evidence .86 Box 6.1 Income Support Programs for the Unemployed: Main Features .91 Box 6.2 Data Sources and Methodology .94 Box 6.3 Labor Markets: Latin America's "Forgotten Reform .100 Box 6.4 Combining Mandated Self-Insurance and Market-Type Unemployment Insurance .102 Box 6.5 Who Really Pays for Unemployment Insurance? .105 Box 6.6 Some Guidelines for Countries that Have Unemployment Systems .107 Box 7.1 How Do Pro-Poor Governments Vary Spending Over the Economic Cycle? .114 Box 7.2 Data Sources and Classification into Targeted and Social Spending .115 Box 7.3 Social Spending Over Economic and Political Cycles in Latin America .116 Box 7.4 Colombia's Healthcare Reform . 118 Box 7.5 Mexico's Progresa Program: Works Well, But Would it Do as Well in Crises? .120 Box 7.6 Social Programs, Entitlements, and Countercyclicality in the U.S .124 Tables Table 1.1 Per Capita GDP Growth in Latin America (Percent). 2 Table 1.2 Per Capita Private Consumption Growth in Latin America (Percent). 3 Table 1.3 Survey Responses in Latin American Countries on Expected Changes in Living Standards (Percentages Unless Otherwise Noted). 4 Table 1.4 Survey Responses to Social Insurance-Related Questions in 14 Latin American Countries, by Socioeconomic Category (Percentages Unless Otherwise Noted) .............................. 4 Table 1.5 Economic Insecurity: Twelve Myths . ............................................ 12 v SECURING OUR FUTURE IN A GLOBAL ECONOMY Table 2.1 Long-Term Volatility in Latin America (Standard Deviations of Growth Rates, Percent) ..... ......... 16 Table 2.2 Volatility in Latin America Over Time (Standard Deviations of Growth Rates by Decade, Percent) ....... 21 Table 2.3 Real Wage Grown Volatility in Latin America and the Caribbean, Percent ........................ 24 Table 2.4 Average Unemployment Rates in Latin America and the Caribbean, Percent ....................... 25 Table 2.5 Labor Turnover, LAC and OECD Countries .................. I ........................... 28 Table 2.6 Income Inequality Measures by Country, 1986-96 .. 33 Table 3.1 Government Policies and Their Effect on Individual Comprehensive Insurance ........ ............. 44 Table 4.1 Volatility in Terms of Trade Shocks for Selected Latin American Countries ........ .. .............. 47 Table 4.2 Estimated Welfare Gains from Diversification (Latin America and the Caribbean, Percent of Private Annual Consumption) ................................................ 63 Table 4.3 International Portfolio Diversification (Ratios to Total Wealth) ................................. 64 Table 4.4 An Overview of Policies to Deal with Aggregate Volatility ............. .. .................... 65 Table 4.A. 1 Empirical Determinants of GDP Growth Volatility (Dependent Variable: Standard Deviation of GDP Growth, 1975-99) .......................................... ....... 69 Table 5.1 Unemployment Rates by Household Income and Consumption Quintile ......................... 74 Table 5.2 Unemployment Rates by Age, Education, and Gender ....................................... 74 Table 5.3 Annual Probability of Becoming Unemployed from Formal and Informal Work (Percent) .... ......... 75 Table 5.4 Income Variance of Formal Salaried vs. Self-Employed Workers (Theil Index) ...................... 76 Table 5.5 Unemployment Duration, in Years ......... ............... ......................... ... 76 Table 5.6 Metropolitan Brazil: Percentage Income Changes by Head's Wage Bracket ........ .. .......... ... 79 Table 5.7 Mexico: Proportional Income Change by Income Quintile, 1995-96 ............................. 80 Table 5.8 Rural El Salvador: A Dynamic Decomposition of Poverty Changes, 1995-97 ....... .. ............. 80 Table 6.1 How Various Income Support Programs for the Unemployed Work .......... .. ................. 92 Table 6.2 Contrasting the Brazilian and U.S. Unemployment Insurance Systems ........................... 93 Table 6.3 Income Support Programs for the Unemployed: Beneficiaries Across Population Groups .... .......... 95 'rable 6.4 Income Support Programs for the Unemployed: Costs per Beneficiary .. 97 Table 6.5 Income Support Programs for the Unemployed: Effects on Employment, Earnings, and Consumption .... 98 Table 6.6 Costs of Unemployment Insurance: Burden-Sharing Among Workers, Employers, and Government ..... 104 Table 6.7 Income Support Programs for the Unemployed: Summary of Findings and Policy Implications .... .... 106 Table 7.1 Targeted Public Spending per Poor Person in Argentina and Mexico, 1994-96 .................... 113 Table 7.2 Elasticities of Spending to Growth, by Type of Spending .................................... 116 Table 7.3 Changes in Latin American Social Protection Spending, 1970-95 (Broad Spending Categories) .... .... 117 Table 7.4 Main Characteristics of Targeted Conditional Transfers ...................................... 119 Figures Figure 1.1 Per Capita GDP Growth (Regional Medians) .. 2 Figure 1.2 Per Capita Private Consumption Growth (Regional Medians) .. 3 Figure 2.1 Long-Term Volatility of Real GDP Growth ....................... ....................... 15 Figure 2.2 Long-Term Volatility of Real Private Consumption Growth ...... ............. . 15 Figure 2.3 GDP Growth Volatility and Country Size ...................... .......................... 17 Figure 2.4 GDP Growth Volatility and Per Capita Income .................... ........................ 17 Figure 2.5 GDP Growth Volatility and Private Consumption Growth Volatility ..18............... I Figure 2.6 Volatility of Real GDP Growth by Decade (Regional Medians) ............ .. .................. 19 Figure 2.7 Volatility of Real Private Consumption Growth by Decade (Regional Medians) . . 19 Figure 2.8 Volatility of Growth of GDP and Private Consumption in Latin America (Ten-Year Window, Regional Medians) .............................................. .. 20 Figure 2.9 The Links Between Aggregate and Microeconomic Volatility ............... ................... 22 Figure 2.10 Long-Run Own Wage Elasticities, 1980-95, Chile .................. ....................... 22 Figure 2.11 Wage and Unemployment Okun Coefficients ..23 Figure 2.12 White- and Blue-Collar Wage Volatility, Mexico ..24 Figure 2.13 White- and Blue-Collar Wage Volatility, Argentina ..25 Figure 2.14 Unemployment Rate and Expected Duration, Montevideo ..26 Figure 2.15 Unemployment Rate and Expected Duration, Interior Uruguay .26 Figure 2.16a Unemployment Rate and Expected Duration, Greater Santiago ..27 Figure 2.16b Labor Turnover, LAC and OECD Countries ..28 Figure 2.17 Evolution of Turnover in Chile, 1980-95 ..28 Figure 2.18 Evolution of Turnover in Colombia, 1980-91 ..29 Figure 2.19 Expected Tenure in Current Job, Montevideo (Months) ..29 vi CONTENTS Figure 2.20 Expected Tenure in Current Job, Interior Uruguay (Months) .................................. 30 Figure 2.21 Expected Tenure on Job, Greater Santiago .30 Figure 2.22 Self-Employment versus Industrial Productivity, OECD and LAC .31 Figure 2.23 Growth and Income Mobility in Argentina .34 Figure 2.24 Growth and Income Mobility in Mexico .35 Figure 4.1 Volatility in Terms of Trade Growth (Regional Medians) ..................................... 46 Figure 4.2 Share in Total Exports of Four Most Important Commodities (Selected LAC Countries) .... ........... 46 Figure 4.3 Volatility of Terms of Trade Shocks by Decade (Regional Medians) .............................. 47 Figure 4.4 Median Private Gross Capital Flows to Latin America (Percent of GDP) .......................... 48 Figure 4.5 Spread of Foreign-Currency-Denominated Sovereign Debt Instruments (bps) (Selected Major Latin American Countries) .............................................. 48 Figure 4.6 Coefficient of Variation of Gross Private Capital Flows (Percent of GDP, Regional Medians) .... . ...... 49 Figure 4.7 Volatility of Reserve Money Growth (Regional Medians) ..................................... 50 Figure 4.8 Volatility of Public Consumption Growth (Regional Medians) ................................. 50 Figure 4.9 Monetary Volatility and Fiscal Volatility ................................................. 51 Figure 4.10 Fiscal Volatility and Terms of Trade Volatility .51 Figure 4.11 Chile's Excess Sensitivity to Shocks .52 Figure 4.12 Copper Prices and Chile's Current Account .53 Figure 4.13 Average Private Domestic Credit (Percent of GPD, Regional Medians) .54 Figure 4.14 Stock Market Capitalization and Turnover Ratios (Regional Medians in Percent, 1990-98) .54 Figure 4.15 Stock Market Capitalization and Turnover Ratios for Selected LAC Countries, 1995-98, Averages .55 Figure 4.16 Stock Market Illiquidity (Regression Coefficient of Absolute Price Changes on Trade Volume) .55 Figure 4.17 GDP Volatility and Credit Depth .56 Figure 4.18 Credit Crunches in Argentina .57 Figure 4.19 Credit Crunches in Brazil ................................... ........................ 57 Figure 4.20 Credit Crunches in Mexico .............................. 58 Figure 4.21 Fire Sales .............................. 60 Figure 4.22 Excess Vulnerability .............................. 60 Figure 4.23 Why is Latin America More Volatile than Industrial and East Asian Countries? .61 Figure 5.1 Income Gains and Losses in Seven Episodes of Volatility in Brazil ............................... 79 Figure 5.2 El Salvador: Cumulative Distributions of Real Income Per Capita, 1995 and 1997 .................. 81 Figure 5.3 Brazil: Moves Into and Out of Poverty, by Level of Education a. Into Poverty During Growth . ..................................................... 82 b. Into Poverty During Recession . ..................................................... 82 c. Out of Poverty During Growth ..................................................... 82 d. Out of Poverty During Recession . ................................................... 82 Figure 5.4 Change in Housework Time Due to Labor Force Entry ....................................... 83 Figure 5.5 Brazil: Probability of School Dropout . ................................................... 84 Figure 5.6 Brazil: Probability of Repeating a Grade at School .......................................... 84 Figure 5.7 Brazil: Probability of Child Aged 10-15 Starting Work ...................................... 84 Figure 6.1 Workers Pay for UI even Though the Contributions are Levied on Firms ........................ 105 Figure 6.2 Workers and Firms Share Ul Costs even Though Contributions are Levied on Firms .... ............ 105 Figure 7.1 Targeted Social Spending Over the Economic Cycle ........................................ 112 vii Acknowledgments HIS REPORT IS THE RESULT OF A COLLECTIVE EFFORT BY A WORLD BANK TEAM COORDI- nated by the Office of the Chief Economist for Latin America and the Caribbean region and led by Guillermo Perry. Main authors by chapter were Guillermo Perry (Chapter 1), Indermit S. Gill (Chapters 1, 3, 5, 6, and 7), Luis Serv6n (Chapters 1, 2 and 4), William F. Maloney (Chapters 2 and 5), Nadeem Ilahi (Chapters 3 and 7), Francisco H. G. Ferreira (Chapter 5), and Martin Rama (Chapter 6). At the World Bank, major contributions were made by Wendy Cunningham, Gabriel Gonzalez, Emily Gustafsson-Wright, Norman Hicks, William Jack, Pedro Olinto, Bernadette Ryan, Guilherme Sedlacek, Rashmi Shankar, Mark Thomas, and Quentin Wodon. Outside contributors included Carlos Arango (University of Illinois), Ricardo Caballero (MIT), Jonathan Conning (Williams College), Adriana D. Kugler (Universitat Pompeu Fabra), Donna MacIsaac (consultant), Alejandra Mizala (Universidad de Chile), Marcelo Neri (Fundacao Getulio Vargas), Martin Ravallion (Universite des Sciences Sociales, Toulouse), Dani Rodrik (Harvard University), Pilar Romaguera (Universidad de Chile), James M. Snyder (MIT), Alvaro Trigueros (Universidad Centroamericana, San Salvador), and Irene Yackovlev (MIT). Valuable comments were received from Ian Bannon, The report team was advised by a Steering Committee Suman Bery, Francois Bourguignon, Ariel Fiszbein, consisting of Robert Holzmann, Gobind T. Nankani, and Marcelo Giugale, Gillette Hall, Emmanuel Y Jimenez, Dani Rodrik. Xavier Coll, Ana-Maria Arriagada, Charles Steen Jorgensen, Homi Kharas, Kathie Krumm, Klaus Griffin, and Donald R. Winkler also provided guidance Schmidt-Hebbel, and Michael Walton. and support. The report also benefited from the comments and sug- However, any errors or omissions in the report are the gestions provided by participants at the Forum on Poverty, sole responsibility of the authors and should not be attrib- Vulnerability, and Inequality held in Lima in August uted to any of the above individuals or the institutions they 1999. Special thanks to Katherine Bain for facilitating the represent. report team's participation at the Forum. Valuable com- Finally, Diane Stamm, Christopher Neal, and Lee ments were also received at the Latin America Economists' Morrison made a superb effort to see this publication Retreat held in Baltimore in November 1999. through the final stages of production. ix Acronyms and Abbreviations ECLAC Economic Commission for Latin America and the Caribbean EDS Encuesta de Desarrollo Social ENEU National Urban Employment Survey ENH Encuesta National de Hogares EPH Permanent Household Survey FDI Foreign direct investment FGTS Fundo Garantia por Tempo de Servicio FHIS Social investment fund FONASA Fondo Nacional de Salud (formerly SERMENA) GDP Gross domestic product GNP Gross national product IA Individual savings accounts IBGE Brazilian Statistical Institute IDB Inter-American Development Bank IFI International financial institution IFPRI International Food Policy Research Institute ILO International Labour Organisation IMF International Monetary Fund ISAPRE Instituciones de Salud Previsional LAC Latin America and the Caribbean LSMS Living Standards Measurement Survey NIC Newly industrialized country OECD Organization for Economic Cooperation and Development PETI Programa de Erradicacao do Trabalho Infantil PME Monthly Employment Survey PNAD Pesquisa Nacional de Amostra de Domocilios PRAF II Programa de Asignacion Familiar, Second Phase PW Public works SP Mandatory severance pay SIMCE Sistema de Medicion de la Calidad de la Ensenanza SUS Sistema Unico de Saude TG Training for the unemployed UI Unemployment insurance xi CHAPTER 1 Opportunity and Risk in a Globalized Latin America and the Caribbean Latin America and the Caribbean's Performance Improved in the 1990s I N THE 1990S LATIN AMERICA AND THE CARIBBEAN (LAC) BEGAN TO RESURFACE FROM THE "LOST decade" of the 1980s. Real per capita income, as measured by gross domestic product (GDP) per per- son, grew at about 1.5 percent per year,1 after having declined in the 1980s. Granted, the pace of eco- nomic expansion in the 1990s was, like in earlier decades, still slower than that of the seven "East Asian miracle" countries. It also fell short of the growth rates achieved in the 1960s and 1970s in Latin America, during which real per capita GDP grew at over 2 percent per year. But this growth decline relative to the pre-1982 performance affected all world regions, industrial and developing-with the excep- tion of only South Asia (see Figure 1.1). The incipient growth recovery was punctuated by episodes of regionwide financial turmoil-such as Mexico's Tequila crisis of 1994-95 and the worldwide fallout from the East Asia and Russia crises in 1997 and 1998-and was uneven across the region. As Table 1.1 shows, the majority of Latin America's larger economies-those with populations above 1 million in 1995-shared in the resumption of growth relative to the 1980s.2 Chile, the earliest reformer in the region, achieved rapid growth, well above historical levels. Other reforming countries such as Argentina, Bolivia, El Salvador, and Peru also grew faster in the 1990s than in previous decades. At the other extreme, several countries that have lagged behind in structural reforms (Ecuador, Haiti, Jamaica, Paraguay, and Venezuela) Jamaica and Venezuela the only countries-among those witnessed a decline in per capita GDP relative to the for which the information is available-to experience a 1980s. Finally, Brazil's growth rate declined compared to decline in private consumption per person relative to the the 1980s, reflecting the adverse effects of macroeconomic 1980s (see Table 1.2). imbalances and financial market turbulence during much of the 1990s. Opportunities and Risks Rising incomes were duly reflected in improving living The improvement in LAC's economic fortunes followed a standards in the majority of LAC economies, as measured sustained reform effort by many countries in the region by per capita private consumption growth, which aimed at enhancing the role of market forces and increas- rebounded from the negative rates of the 1980s. As with ing the region's real and financial integration into the GDP, however, the performance of consumption fell short global economy. The incipient economic upturn of the of the pace witnessed prior to the debt crisis of 1982, and 1990s suggests that this strategy has started to generate remained considerably behind the pace of the East Asian new opportunities for LAC in the global scene, especially miracle economies (see Figure 1.2). Nevertheless, the for earlier and deeper reforming economies. In spite of bet- upturn in consumption reached most countries in the ter opportunities, however, perceptions of economic inse- region, with Chile and El Salvador the star performers, and curity run high in the region. Indeed, there is a widely held I SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 1.1 Per Capita GDP Growth (Regional Medians) 7- 6- 0 1960 _ 5 U 1970; 4- . . ..... 0000000019980s' v3 -._ 1 990!i U 2 - .il_^- ^^1 ~ ^^ i _ _ -2 Industrialized East Asia 7 Latin America and Middle East and South Asia Sub-Saharan Africa Other Easr Asia Economies the Caribbean North Africa and Pacific TABLE 1.1 view that economic insecurity3 has become so Fervasive Per Capita GDP Growth in Latin America that it could undermine social and political support for the (Percent) ongoing reform process, and even bring it to a halt.4 COUNTRY 1960s 1970s 1980s 1990s That insecurity is a major concern for large segrments of Argentina* 2.4 1.2 -2.4 3.0 LAC's population is vividly illustrated by recent opinion Bolivia 0.7 1.7 -2.5 1.8 surveys in the region. In a large cross-country survey Brazil 2.9 5.7 0.8 0.4 undertaken in 1999, for example, nearly two-thirds of Chile 2.0 0.6 2.5 4.7 Colombia 2.0 3.3 1.2 0.6 respondents said that their parents had lived better than Costa Rica 2.3 3.4 -0.8 2.0 them, while less than half thought that their children Dominican Republic 1.2 5.2 1.5 2.2 Ecuador 1.0 5.8 -0.3 -0.3 would have lives better than their own (see Table 1.-3). This El Salvador 2.4 1.2 -3.2 2.8 pessimistic view about the future affected not only coun- Guatemala 2.5 3.0 -1.6 1.4 Honduras 1.6 2.4 -0.7 0.3 tries experiencing major economic and social difficulties, Haiti -1.4 1.8 -1.5 -3.1 such as Ecuador or Venezuela, but also others that had seen Jamaica 2.6 -0.4 0.1 -0.2** JaMeica 2.6 30.4 0.0 10.2 a marked improvement in their economic performance in Mexico 3.4 3.3 0.0 1.5 Nicaragua 3.9 -3.2 -3.7 -0.5 the 1990s, such as Argentina, Mexico, and Peru. Indeed, Panama 4.8 1.9 -1.4 3.3 even in these countries, a relatively small percentage of Paraguay 1.7 4.9 0.8 -0.2 Peru 2.2 1.1 -2.4 1.9 respondents--43 percent in Argentina, 30 percent in Mex- Trinidad and Tobago 4.2 4.7 -1.0 1.3 ico, and 37 percent in Peru-anticipated a better future for Uruguay 0.2 2.3 -0.2 3.0 their children.5 Venezuela 1.2 0.4 -2.8 -0.2 Mean 2.1 2.4 -0.8 1.2 Along with this heightened concern about economic Median 2.2 2.3 -0.8 1.4 insecurity, there are also strong signs of unsatisfied demand Small Countries for social insurance. The same survey mentioned above Bahamas 5.3 -0.7 1.5 -2.1 Belize 2.2 4.2 2.6 1.7 found that three-quarters of the respondents favored Barbados* 5.7 2.9 1.7 0.2 increased spending on unemployment insurance. All even Guyana 1.1 0.8 -3.5 4.0 Mean 3.6 1.8 0.6 1.0 hgher number supported icreased spendig on social Median 3.7 1.9 1.6 1.0 security (see Table 1.4). Moreover, the extent of support for Unweighted average 2.3 2.3 -0.6 1.2 these programs varied little with the respondents' income Overall nedian 2.2 2.3 -0.7 1.4 and education level, or even with the economic perfor- Wei gheed average* 2.5 3.5 -0.1 .1. mance of the different countries. In high-performing *Weighted averages use 1995 population. Note Decades are defined as 1961-69, 1970-79, 1980-89, 1990-99. Sample period is Chile, for example, 85 percent of the respondents favored defined as 1961-98. Exceprions: Bahamnas (1961-95), Barbados (1961-95), Guyana increased unemployment insurance, and over 90 percent (1961-95), Peru (1966-99). For Argentina, Brazil, Chile, Colombia, Ecuador, Peru, and Venezuela, figures are updated to 1999. supported greater pension expenditures. 2 OPPORTUNITY AND RISK IN A GLOBALIZED LATIN AMERICA AND THE CARIBBEAN FIGURE 1.2 Per Capita Private Consumption Growth (Regional Medians) 7- 6 1960s 5- 1970s 4- 1980s rv 3-X ~ | * . . * 19905 3- 2-2 -2- Industrialized East Asia 7 Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Economies the Caribbean North Africa and Pacific TABLE 1.2 Why Economic Insecurity? Per Capita Private Consumption Growth in Latin Amenca What lies behind these perceptions of insecurity and social (Percent) insurance demands? There are several factors. On the one COUNTRY 1960s 197()s 1980s 1990s hand, the unprecedented severity and duration of the crisis Argenrina* 2.2 1.1 -1.7 4.0 unleashed in 1982-reflected in a sharp and long-lived Bolivia .4 .4 -1.3 1.0 decline in per capita incomes from which LAC has taken many Brazil 2.4 6.4 -0.6 1.9 Chile 4.7 0.2 0.5 5.8 years to recover-left a profound imprint across the region's Colombia 2.6 3.2 0.7 1.4 social fabric concerning the dangers of economic instability. Costa Rica 1.4 2.7 -1.7 1.2 Dominican Republic 3.1 4.6 1.0 0.9 On the other hand, the incipient recovery from the "lost Ecuador 1.8 3.9 -0. 1 0.2 decade" of thel980s came along with a radical change in eco- El Salvador 1.7 0.9 -3.4 5.2 Guatemala 1.9 2.6 -1.2 1.5 nomic strategy in many LAC economies-a shift away from Honduras 1.1 1.6 -0.1 0.2 the protected government-led development model of previ- Haiti -2.6 1.1 -0.7 Jamaica 1.0 -0.3 2.4 -2.8 ous decades, and toward a new paradigm of strengthened Mexico 2.8 2.5 0.0 1.1 Nicaragua 3.6 -3.5 -4.6 0.0 domestic and foreign market forces in the context of a global Panama .. ., 3.6 3.0 economy. Barriers sheltering domestic economies from global Paraguay 3.3 3.0 1.0 3.2 trade and financial trends were lowered, obstacles to compe- Peru 4.0 0.9 -2.1 1.5 Trinidad and Tobago 4.6 4.9 -2.1 1.5 tition in domestic markets were removed or substantially Uruguay -0.4 0.8 0.2 4.1 Venezuela .. 6.1 -1.6 -0.6 weakened across LAC, and governments reduced consider- Mean 2.2 2.3 -0.6 1.7 ably their direct involvement in economic activity. ,Median 2.3 2.5 -0.6 1.4 These reforms deserve much of the credit for LAC's Small Countries Bahamas .. 22.7 2.3 expanding economic opportunities in the 1990s. However, Belize .. .. -2.3 2.1 while the reforms assigned a greater role to the action of Barbados* .. .. .. 0.9 Guyana 1.3 -0.3 -3.6 5.8 domestic and global market forces, they also led to the weak- Mean 1.3 11.2 -1.2 2.9 ening of major components of the rudimentary and Median 1.3 21.2 -2.3 2.1 inequitable traditional social protection system. The weak- Unweighted average 2.1 3.3 -0.5 1.7 Overall median 2.2 2.5 -0.6 1.5 ening of extensne barriers to domestic and foreign competi- Weighted average** 2.3 3.7 -0.5 1.7 tion made it harder to sustain a generous provision of public *Consumption figures for Argentina and Barbados correspond to total, and not pri- sector jobs and stringent firing restrictions that had resulted vate, consumption. Argentina: 1961-98, Barbados: 1967-94. **Weighted averages use 1995 population. in virtual lifetime employment for formal sector workers. Note: Decades are defined as 1961-69, 1970-79, 1980-89, 1990-99. Sample period is Th r f h o defined as 1961-98. Exceptions: Bahamas (1978-87), Barbados (1967-94), Belize e remova o these old mechanisms icompatible (1981-98), Bolivia (1961-79,1981-98), Guyana (1961-88), Haiti (1966-90), Hon- with the new market-oriented economic model has not duras (1961-97), Nicaragua (1961-96), Panama (1981-98), Peru (1966-98), and Venezuela (1975-98). been matched by the development of a new social protec- 3 SECURING OUR FUTURE IN A GLOBAL ECONOMY TABLE 1.3 Survey Responses in Latin American Countries on Expected Changes in Living Standards (Percentages Unless Otherwise Noted) WHOLE SAMPLE COUNTRY N % ARGENTINA BOLIVIA BRAZIL CHILE COLOMBIA 14,839 t00 1,200 794 1,00() 1,200 1.200 Taking everything into consideration, would you say that your parents lived better, the same, or worse than how you live today? Better 9,081 61.2 63 51 64 45 78 Same 3,261 22 22 31 9 32 14 Worse 2,139 14.4 12 16 25 22 8 No answer 358 2.4 3 3 2 1 1 And regarding your children, do you believe that they will live better, rhe same, or worse than how you live today? Better 6,843 46.1 43 56 58 61 36 Same 3,071 20.7 22 20 12 22 .1 Worse 3,261 22 20 13 21 11 38 No answer 1,664 11.2 16 11 9 7 6 Source: Mirror on the Americas poll, 1999, Wa/l StreetJournal. TABLE 1.4 Sufvey Responses to Social Insurance-Related Questions in 14 Latin Amefican Countries, by Socioeconomic Category (Percentages Unless Otherwise Noted) WHOLE SAMPLE AGE EMPLOYMENT SELF- PRIVATE N 5N 18-29 30-49 50+ EMPLOYED GOVERNMENT SECTOR Unemployment Insurance No answer 551 3.7 3 4 4 4 5 3 Spend more 10,088 73.4 74 73 73 74 69 74 Spend less 2543 17.1 18 18 16 17 19 17 Don't know 857 5.8 5 5 7 5 7 6 Pensions No answer 172 1.2 1 1 1 1 2 1 Spend more 12,426 83.7 83 83 86 83 84 84 Spend less 1861 12.5 13 13 10 13 12 13 Don't know 380 2.6 3 3 3 3 2 2 Defense and the Armed forces No answer 623 4.2 4 5 5 4 6 3 Spend more 4810 32.4 33 31 34 34 28 29 Spend less 8359 56.3 58 57 53 56 59 61 Don't know 1047 7.1 5 7 9 6 7 7 Source: Mirrow on the Americas poll, 1999, Wall Street Joarnal. tion and insurance system more suited to the changed eco- In addition, the improving economic environment in nomic environment. Thus, while households and workers LAC relative to the 1980s may itself be partly responsible in LAC's reforming economies have gained access to new for the heightened social insurance demand. Perhaps para- economic opportunities, they may have been left more doxically, economic analysis-this report shows-suggests exposed to new risks as well. Growth resumption in the that in better times, when individuals have more to lose new economic environment entails faster job creation in and can afford more of the costs of protecting against risk, expanding industries, and thus new opportunities, but also they may also demand more effective protection and insur- job destruction in declining sectors, and hence new risks. ance mechanisms. All these factors make economic insecu- 4 OPPORTUNITY AND RISK IN A GLOBALIZED LATIN AMERICA AND THE CARIBBEAN COUNTRY COSTA RICA ECUADOR GUATEMALA MEXICO PANAMA PARAGUAY PERU URUGUAY VENEZUELA 1,00() 1,200 1 ,I1(10 1,200 1,0(0 6011 1,(45 1,200 1,200 56 67 57 43 52 75 80 59 70 26 20 31 35 24 14 12 18 19 14 10 11 20 21 6 6 20 10 5 2 1 3 3 5 2 3 2 52 34 51 30 48 48 37 46 53 25 21 23 23 20 26 19 22 17 11 29 17 41 21 13 26 19 19 12 16 8 6 12 13 19 13 12 EMPLOYMENT EDUCATION INCOME UNEMPLOYED RETIRED HOUSEWIFE STUDENT PRIMARY HIGH UNIVERSITY HIGH MIDDLE LOW 3 4 4 3 4 3 4 4 4 3 80 74 73 74 74 75 71 72 73 74 13 16 17 19 16 16 19 19 17 17 3 6 7 5 7 5 6 4 6 6 1 1 1 1 1 1 1 2 1 1 88 88 84 82 83 85 83 83 85 83 9 9 12 15 13 12 14 14 11 13 1 2 3 2 3 2 2 2 2 3 3 5 4 3 4 4 5 5 5 3 36 33 35 32 37 33 27 32 31 34 56 54 52 6(0 50 57 62 58 56 57 5 8 9 5 8 7 7 6 8 7 rity one of the major unresolved items in LAC's policy of moving down or being left behind-hence the concern agenda. with insecurity and inequality. Inequality stands as the region's other big pending The social costs of insecurity should not be downplayed. issue. Indeed, as noted earlier, inequality and insecurity are Uncertainty about future employment and income has a related. Increased economic opportunities tend to enhance direct adverse impact on welfare, because most households income mobility-the chances of moving up or down the and workers care not only about the level of their standard distribution ladder. Thus, these added opportunities for of living, but also about its certainty-as the survey evi- economic improvement may come along with greater risks dence above clearly illustrates. 5 SECURING OUR FUTURE IN A GLOBAL ECONOMY Further, economic uncertainty itself can hamper real households react to economic insecurity, the policy chal- income growth, a fact confirmed by extensive empirical lenges these responses present, and the policies that are research focusing on LAC and other regions.6 In essence, best suited for countries in the region are based on recent high degrees of uncertainty tend to discourage growth- empirical and theoretical research conducted at the World enhancing long-term commitments, such as investment in Bank and elsewhere, on economic volatility and social risk physical and human capital, as individuals attempt to management, drawing on the experiences of countries in retain extra flexibility in order to deal with a volatile envi- Latin America and other parts of the world.9 ronment.' As a result, the choice of investment projects This report begins by stating the facts concerning eco- and production technologies is biased by inefficient "short- nomic insecurity in LAC (Chapter 2). It then sets out a termism" that leads to a diminished growth potential for general analytical framework to help organize the various income and living standards. options available to individuals and governments for deal- Finally, there are reasons why economic insecurity is ing with economic insecurity (Chapter 3). Using this particularly damaging for the poorer segments of the pop- framework, the remaining chapters focus on measures to ulation. On the one hand, the poor often lack the means to deal with risks. First, the causes of macroeconomic or protect themselves from adverse income and employment aggregate volatility are examined and some remedies sug- shocks-means such as accumulated financial assets or gested (Chapter 4). This report then examines how these access to credit. For the very poor, this implies that unfa- risks affect individuals and households, and their responses vorable temporary shocks may result in drastic declines in to economic shocks (Chapter 5). The risk of becoming consumption, bringing it down below subsistence levels unemployed is of concern in the region and elsewhere, and and permanently damaging their well-being. On the other public responses to help workers deal with this risk take up hand, growth in income of the poor is primarily deter- a full chapter (Chapter 6). Finally, the subject of appropri- mined by overall economic growth (Dollar and Kraay ate social insurance and social protection against the risk of 2000).5 As economic volatility hampers aggregate growth, poverty is considered in some detail (Chapter 7). We sum- it also hurts the growth of income of the poor and their marize the findings of these chapters here. chances to rise out of poverty. In fact, a 1999 World Bank study shows that economic insecurity ranks high among LAC's Volatility is High-But has not Risen in the the concerns expressed by the poor in LAC and across the 1990s world (Narayan and others 1999). Like most developing regions, LAC suffers from high eco- nomic volatility, well above the levels experienced by This Report industrial economies. Furthermore, living standarcls-as The purpose of this report is to assess the extent, causes, and measured by per capita consumption-are more volatile effects of economic insecurity in LAC and identify policies than real incomes, a feature shared with much of the devel- and institutions that can help reduce the degree of insecu- oping world but not with most OECD countries. This rity faced by workers and households in the region, while reflects a lack of adequate instruments for consumption- allowing them to take advantage of the enhanced economic smoothing in developing countries. opportunities brought about by the reforms of recent years. Contrary to a widely held view, however, there is no evi- Insecurity is a broad topic, however, and this report can- dence that volatility has increased following the region- not cover all of its many aspects. Thus, the report leaves wide shift toward a market-oriented economy and the aside issues related to crime and violence and insecurity increased integration of LAC into global markets. On the caused by natural disasters, to focus on the specific issue of contrary, the volatility of income growth has declined in insecurity caused by economic fluctuations. Within this most of the region's economies, and in a number of them it narrower area, social security and pensions, which have has fallen below the levels of the 1970s. To a lesser extent, more to do with life-cycle considerations than economic the volatility of private consumption has also declined. In fluctuations, are also excluded from the discussion. addition, there is no evidence that the income and employ- This still leaves a wide range of issues to be addressed in ment uncertainty faced by the majority of workers and the chapters that follow. The analysis of how workers and households in the region has changed for the worse- 6 OPPORTUNITY AND RISK IN A GLOBALIZED LATIN AMERICA AND THE CARIBBEAN although economic insecurity must surely have risen for Although the 1990s have witnessed a decline relative to specific groups of workers adversely affected by the reforms the 1980s in the volatility arising from each of these in some countries, especially some elder workers in the for- sources, the region faces larger terms of trade volatility mal sector, whose skills may have been rendered obsolete than most other developing regions. Likewise, the volatil- by economic restructuring. ity of capital inflows and domestic policies is also higher in LAC than in industrial economies and the more stable Dealing with Economic Insecurity Requires a developing regions, such as the East Asian miracle Comprehensive Insurance Approach economies. There are three major options available for dealing with Further, the economic impact of disturbances is magni- risk: market-type insurance, which involves sharing of fied by the region's weak links with international financial risks among individuals (or countries); self-insurance, markets and the insufficient development of domestic which typically entails precautionary saving or accumula- financial systems, which lag behind those of other world tion of assets in good times to shelter consumption in bad regions. In theory, domestic and foreign financial markets times; and self-protection, which involves the adoption of should play a major role in facilitating risk diversification measures to reduce the likelihood of adverse shocks. In and easing the adjustment to shocks. In practice their general, effective risk-management strategies should imperfections make them have the opposite effect-they employ all three types of instruments. The more instru- amplify aggregate shocks and are themselves a source of ments are available, the better the chances of sheltering liv- volatility. ing standards from economic insecurity. Thus, for example, developing countries should be able This general perspective illustrates the role of public to diversify terms of trade risks by hedging in international policies in dealing with risk. Government interventions are financial markets. In practice, however, these markets are warranted by the presence of incomplete or imperfect not deep enough, and capital flows behave procyclically domestic insurance markets that lead individuals to costly with respect to trade shocks, amplifying the international and inefficient self-insurance or self-protection decisions, business cycle. More generally, in spite of recent progress such as distress sales of assets or reduced investments in with the development of new international financial human capital (for example, taking children out of school) instruments such as contingent credit lines, world markets at times of crisis. Government interventions, in the form of still offer few possibilities for risk diversification and insur- social insurance and social protection, attempt to remedy these ance against aggregate disturbances. Hence, as discussed market failures, augment the availability and scope of mar- below, there is a role for supranational policy actions aimed ket insurance and efficient instruments for self-insurance, at the creation of missing markets and the enhancement of and buttress the self-protection efforts by individuals that instruments for international risk diversification. pay off only if sustained for some time. As with other types of insurance, the design and implementation of these pol- Governments Can Do Much to Reduce Volatility-Even icy interventions must deal with the serious challenges in a Globalized Economy posed by adverse selection and moral hazard. Governments possess a broad range of possible measures to This conceptual framework also shows how increased reduce aggregate volatility, that can improve risk-sharing, demand for social insurance may indeed result from an enhance economywide self-insurance, and reduce the like- improvement-rather than a deterioration-in the eco- lihood of adverse aggregate shocks. Given the rudimentary nomic environment, a factor that might be partly behind state of international insurance markets, the main options the results of LAC opinion surveys mentioned above. left to governments involve self-insurance and self-protec- tion mechanisms. Many such mechanisms have already LAC's Volatility Arises from Multiple Sources- been adopted by various countries in the region, and all Domestic and Foreign entail economic costs. Thus, the policy mix best suited to LAC's macroeconomic volatility reflects both external dis- each economy is largely dependent on country-specific fac- turbances-in international goods and financial markets- tors shaping the cost-effectiveness of the various policy and volatile domestic fiscal and monetary policies. options. 7 SECURING OUR FUTURE IN A GLOBAL ECONOMY External Risks Can Be Reduced by Diversification and like in others, faces a fundamental tradeoff between credi- Liquidity Management bility and flexibility. Rigid exchange rate pegs without the Nevertheless, some clear general principles emerge from option of an independent monetary policy may enhance the analysis. To deal effectively with terms of trade volatil- credibility, but can also make adjustment to shocks more ity, countries can resort to risk diversification and hedging painful in the presence of inflexible labor markets or inad- in international commodity markets, self-insurance equate fiscal policy. Floating exchange rate arrangements through commodity stabilization funds, and self-protec- with active monetary policy may offer enhanced flexibility tion through trade diversification. Facilitating foreign to deal with shocks, but can erode credibility unless clear direct investment (FDI) is another way of diversifying and transparent rules for monetary policy-possibly con- risks, and FDI also yields other benefits such as innovation tingent on developments in world goods and financial mar- spillovers, enhanced corporate governance, and higher kets-are publicly announced and strictly followed by the investment. Allowing domestic investors to hold foreign authorities. Intermediate options such as adjustable pegs, assets also improves their own risk diversification strategy crawling pegs, and exchange rate bands probably offer the and increases the resilience of the economy as a whole. worst of both worlds without the advantages of either one. In turn, facing up to capital flow volatility in a context of limited international insurance possibilities requires holdings Deeper and Stronger Financial Systems Are a Key Part of liquid assets and a prudent debt management strategy, and of Social Protection Policies avoiding excessive short-term liabilities, "bunching" of Development of deeper capital markets and strong bank- repayments, and currency mismatches between assets and lia- ing systems is a major priority to allow them to play their bilities. Capital controls may offer another self-protection intended role of shock absorbers and hence mitigate the tool to limit exposure to international financial disturbances, economic impact of disturbances. Enhanced capital and but their effectiveness remains under debate. They may affect liquidity requirements for banks-perhaps set in a pro- the composition of flows-discouraging volatile short-run cyclical manner-under adequate supervision, and preven- transactions if properly designed-but they seem powerless tion of currency mismatches can go a long way toward to alter their volume beyond the near term. strengthening the banks, so that they can contribute effec- tively to self-insurance against shocks. Strong and deep Anticyclical Macroeconomic Policies Ease Adjustment to financial systems are of paramount importance to facilitate Shocks savings and market insurance against microeconomic risks. In most LAC economies, fiscal policy has failed to play its intended stabilization role. Governments have generally Deep Crises are Particularly Damagingfor the Poor adopted an expansionary stance in booms and a contrac- How are households affected by adverse economic condi- tionary stance in recessions. To some extent, this reflects tions, and how do they respond to crises? To answer these constraints from worl(d and domestic financial markets. It two questions, this report systematically used household has also resulted, however, both from the failure of govern- panel data for Argentina, Brazil, El Salvador, and Mexico ments to provide for bad times by saving in good times, in both rural and urban settings. Several findings emerge and the lack of a sufficiently diversified fiscal revenue base, that should make us reconsider some commonly held which in several countries in the region is excessively beliefs about how households respond and when, how, and biased toward natural resource revenues. Tackling these how much governments should help them. two issues should be a policy priority, along with the adop- First, economic contractions differ significantly in their tion of contingent fiscal rules that can facilitate the effects on poverty and human capital investments: in deep response to shocks and make it more transparent, and the recessions the poor suffer greater proportional losses in implementation of a prudent public debt management income than the wealthy. In moderate recessions, the oppo- strategy along the lines mentioned earlier. site appears to happen-in many cases, the greatest pro- Finally, adequate monetary and exchange rate policies portional income losses were borne by the rich, and some can also make an important contribution to the absorption groups often thought to suffer disproportionately-such as of shocks. The choice of specific policy rules in this area, the elderly or single mothers-do not appear to be espe- 8 OPPORTUNITY AND RISK IN A GLOBALIZED LATIN AMERICA AND THE CARIBBEAN cially badly affected, although this is not true in every cri- in employment contracts. In the old economic environ- sis and in every setting. For example, the findings differ ment, these schemes effectively pooled unemployment between countries or, for the same country, between rural risks over a greater population because consumers actually and urban areas. On the whole, however, the conventional subsidized potentially bankrupt firms through higher wisdom that the poor invariably are affected more severely prices. But with globalization and reduced barriers to during recessions needs to be qualified. trade, this is no longer possible: prices are determined by Second, the poor seem to have gained more during world markets, so the pooling of unemployment risk growth periods than is generally acknowledged. This does becomes restricted to the firm. These provisions have also not mean that the poor should not be helped; it merely proved to be contentious, complicated to enforce, and judi- implies that from the perspective of poverty alleviation, cially burdensome. growth-oriented policies must be given a high priority, With neither economic efficiency nor administrative regardless of concerns of high inequality in the region. ease to recommend mandated severance pay, some coun- tries in the region have moved away from it, and others are The Poor Try to Protect Their Long-Term Welfare in contemplating change. This report considers the experi- Crises-As Long as Their Assets Permit ence of both reformers and nonreformers in this regard, and Third, the poor-like those with more wealth-are reluc- also employs theoretical principles to provide guidance to tant to permanently compromise their family's future dur- countries in the region. In deciding whether to move ing economic crises perceived to be temporary. This is toward government-mandated self-insurance-through especially true of parental decisions about their children. schemes such as individual savings accounts to be accessed The poor do not, for example, frequently pull their chil- in case of unemployment-or to forms of unemployment dren out of school during bad times-although they do insurance that involve the pooling of risk, several factors when the recession is severe. But the fact that some educa- must be considered. tional and health outcomes are hurt during especially bad times may be as much the result of the government's Administrative Capacity and Labor Policies are Key in inability to maintain the quality of social services as the the Choice of Instruments household's decision to invest less during crises. The first critical issue is the administrative capacity of gov- Finally and unsurprisingly, access to "reserves"-such as ernment. While administrative capacity can always be assets and underused family labor-reduces a household's built over time, it does limit the options of government in vulnerability to shocks, in the sense of having to adjust the immediate future. However, a blend of practicality and through reduced consumption or critical investments such analytical rigor can help countries devise strategies that as schooling and health. Assets may be the key factor for efficiently bridge immediate action and long-term vision. explaining differences in the responses of poor versus rich The second critical issue is the nature of labor markets, households in large versus moderate economic contrac- which influences the level and nature of risks faced by tions. In brief or mild contractions, even the limited assets workers. The logical first step is to do more to reduce the of the poor can help weather the crisis; in more severe or likelihood of adverse employment shocks. Most LAC recurring crises, the poor may eventually exhaust their economies have high levels of informal employment, and assets and be forced to suffer drastic declines in their well- many have high rates of formal unemployment as well. being, with adverse long-term effects. Hence policies While these phenomena have diverse causes (for example, aimed at strengthening the human capital of the poor (edu- high rates of taxation, overregulated labor markets, poor cation, health) can enhance their self-insurance and self- macroeconomic policies that impede growth), labor policy protection efforts. changes are widely regarded as lagging other economic reforms in the region. For governments that wish to facili- New Income Support Programs for the Unemployed tate comprehensive insurance decisions by their workers Need to be Established and households in a rapidly changing global economy, The common form of public unemployment support in labor policies should receive a high priority on the reform much of LAC has been mandatory severance pay provisions agenda. 9 SECURING OUR FUTURE IN A GLOBAL ECONOMY Self-Insurance for Slow-Reforming Economies Targeted Programs for the Poor Need to be Bether Countries that have not yet pursued comprehensive eco- Protected in Downturns nomic-especially labor market-reforms may be better The region has improved the poverty impact of social advised to rely more on self-insurance-type schemes such as spending through reform over the last decade by, for exam- individual capitalization funds. Their "insurance funda- ple, replacing generalized subsidies with programs specifi- mentals" favor such schemes: self-insurance is the preferred cally designed to help the poor. However, during crises, option when losses are frequent, it is less demanding in spending on tightly targeted programs for the poor does terms of administrative capacity, and the schemes entail appear to suffer more than general social expenditures. low labor market efficiency costs and low fiscal costs. The Governments could do better to protect these programs weakness of these schemes is their low attractiveness to from cuts. Experience in the region and in the U.S. shows poorer workers, for whom forced saving may have high that a successful strategy requires explicitly accounting for costs, some of which could be lowered through risk-pool- political economy factors that make programs resilient to ing or government subsidies. both political and economic changes. Such factors may include deliberately building in some features t.hat have Unemployment Insurance for Advanced Reforming been associated with long-lived government interventions. Economies LAC economies that have reduced the risk of unemploy- Save in Good Times to Finance Social Spending in Bad ment through comprehensive economic and labor reforms Times should consider conventional unemployment insurance. Governments in the region do appear to have behaved in a While administrative considerations are always important, pro-poor manner in the most general terms, especially this capacity can generally be built. Carefully designed since the return of democracy to the region. While author- unemployment insurance schemes that involve pooling but itarian and democratic regimes in LAC appear to have keep efficiency losses low-for example, by keeping bene- responded similarly to economic crises-both cut social fits frugal and mimicking the market as much as possi- spending sharply and about equally-greater increases in ble-are likely to increase welfare. Besides helping work- social spending take place under democratic regimes. In ers deal with idiosyncratic risk, insurance schemes that fact, social spending increases only when there is both involve pooling of risk have-when designed well-also democratic rule and a nonshrinking economy. BurF this is shown their worth as "automatic fiscal stabilizers," which also where governments run the greatest danger of adding governments in the region have lacked. policy risk to economic risk. Well-intentioned govern- Political opposition to labor market flexibilization is ments or those under political pressure to sharply increase almost always related to the perception of higher unem- spending on social programs during growth episocles only ployment risks in downturns. Hence, a sequencing option to have to reduce spending in the next contraction both to overcome political constraints may be to adopt labor raise risk and sow the seeds of social discontent. market reforms simultaneously with efforts to strengthen There is considerable room for improving the design of unemployment insurance. targeted programs, especially how they relate to the eco- nomic cycle. While meeting many of the goals they were Public Works Programs Provide Insurance Support for designed to accomplish in both rural and urban settings, Informal Sector Workers targeted conditional transfer programs such as Mexico's Pro- Those who cannot be reached through such contributory gresa and Brazil's Bolsa Escola may not be particularly well schemes should be assisted through programs that implic- suited to assist those who become poor duing economic itly pool risks such as public works programs, and share downturns. Through their innovative links with human some other characteristics with good unemployment insur- capital accumulation, these programs may be better suited ance, principally keeping benefits frugal. Such schemes than earlier interventions for addressing structural poverty should be thought of as insurance-not emergency-pro- concerns. However, the broad political support they enjoy grams, the difference being that insurance programs are make them a resilient policy instrument to offset cyclical permanent while emergency programs are temporary. swings in the quality of education and health services. Even 10 OPPORTUNITY' AND RISK IN A GLOBALIZED LATIN AMERICA AND THE CARIBBEAN if more conventional instruments such as public works pro- Latin America and the Caribbean. Uncertainty about future grams-when designed well-are better safety nets, tar- living standards is a major concern for workers and house- geted conditional transfer programs offer a strong option to holds in the region, and the report shows that there are form the third leg of a comprehensive social safety net- good reasons for this. LAC, like other developing regions, is along with social security for the aged and income support subject to much larger economic fluctuations than indus- for the unemployed. trial economies, and has fewer instruments available to pro- tect consumption levels from economic shocks. Supranational Action and the Role of the International However, contrary perhaps to popular perception, the Financial Institutions trend toward globalization in the 1990s has not made mat- The global economy also poses risks that cannot be effec- ters worse. Growth has risen and volatility has declined in tively addressed by individual countries on their own. the majority of economies in the region, and several that Imperfections in international insurance and financial mar- pursued strong reform policies have enjoyed both higher kets prevent national economies from properly diversifying growth and lower volatility than in earlier decades, while terms of trade risks, and typically lead to a withdrawal of countries whose reform drive lagged behind have been financial support when it is most needed-that is, at the among the worst performers in the 1990s. The lesson is that time of adverse shocks. In this context, just as national with globalization, good policies can reap larger rewards governments have a major role to play because of incom- than before, but bad policies may be more severely punished. plete or imperfect domestic insurance markets, the imper- This report shows that to face the new situation, adequate fections of international markets provide a rationale for macroeconomic policies and structural reforms need to be supranational action. It would aim at improving the self- matched with the development of a social protection and insurance and self-protection choices available to individ- insurance system suited to LAC's changed economic ual countries that may entail too high a cost in terms of environment. economic efficiency and growth, and provide insurance by making available financing during bad times. IFIs can help Notes developing economies efficiently deal with risk by leading 1. The figures mentioned in this section refer to the regional the way in developing new financial markets and instru- median growth rate-that is the race of per capita GDP or con- ments, such as contingent credit lines and borrowing guar- sumption growth in the region's "typical" economy. 2. It is Important to note that available GDP growth data reach antees. In addition, by deploying their financial resources up o 1999 fotr the major eight Latin American economies anticyclically, that is, reducing their lending in good times (Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, and and increasing it in bad times, they can partially counter- Venezuela). The discussion in the text and the figures in the tables act the procyclical behavior of private capital flows and are based on this updated information. In contrast, private con- cushion the adjustment to disturbances. sumption growth information only reaches up to 1998. Preliminary forecasts for 1999 alter somewhat the performance of specific coun- tries, but paint a broadly similar regional picture. Myths and Realities About Economic Volatility 3. These concerns with insecurity may reflect in part concerns In the light of the report's analysis, Table 1.5 shows how with (in)equity, as not all groups of economic actors have shared some widely held notions about economic volatility and its equally in the upturn, and some specific groups may have lost out effects in LAC should be reassessed and qualified. In a nut- with the reforms, at least in the short term. shell, volatility, though still high, does not seem to have 4. The following discussion draws from Rodrik (1999). worsened, governments can do much to reduce it even in a 5. The just-released 1999-2000 poll from Latinbarometer shows a slightly less pessimistic picture, with 58 percent of the respondents global economy, and closer scrutiny reveals that households regionwide expressing the view that their parents lived better, and and governments do not always respond to economic 52 percent anticipating a better future for their children. shocks in stereotypical ways. 6. The Inter-American Development Bank (1995) presents a com- prehensive study of the causes and consequences of volatilty in Latin Securing Our Future America, as part of which the GDP growth cost of the region's "exces- sive" volatility (relative to industrial economies) is estimated at over 1 We hope this report will succeed in calling to the attention percent per year. Empirical studies with a cross-regional focus include of policymakers the problem of economic insecurity in Ramey and Ramey (1995) and Aizenmann and Marion (1993). 11 SECURING OUR FUT[ RE IN A GLOBAL ECONOMY TABLE 1.5 Economic Insecurity: Twelve Myths MYTH REALITY 1. Aggregate volatility has increased in LAC. Volatility of output and consumption is still much higher in LAC than in industrial countries, but it appears to have declined in the 1990s in many of the economies in the region. 2. Workers in LAC face higher uncertainty Microeconomic data show no conclusive pattern-likely a general improvement, but a possible now than ever before. deterioration for some groups of workers in specific countries. 3. A greater demand for social insurance is As countries become wealthier, demand for overall insurance may go up even if risk does not. Demand unequivocal proof of greater economic risks. for insurance involving risk-pooling may rise even if overall risk declines. 4. Globalization means that countries are Governments can do a lot to reduce volatility through policies such as trade diversification, comrnodity powerless to reduce aggregare risk. stabilization funds, precautionary fiscal targets, deepening of the financial sector, and strengthening bank- ing systems. 5. Expanding global financial markets leave IFls have a major role to play in the development of instruments and markets to facilitate international no room for supranational action. diversification of risks, and to ease the adjustment to shocks through countercyclical financing and contingent credit lines. 6. The rise of democracy in the region has Both autocratic and democratic governments in LAC reduce spending during economic downturrs. not helped the poor much. However, poverty-related programs have expanded much more under democratic governments. 7. Governments in the region have not been Governments have not been successfiul at protecting social spending in downturns. In part the reason is that pro-poor poverty-related programs may have grown too quickly in good times, to levels difficult to mainta.n in bad times. 8. The poor are always hurt more than the The poor are hurt more than the rich when economic contractions are deep and persistent. Moderate rich during economic contractions, fluctuations usually hurt the rich more than the poor, although even these smaller losses suffered by the poor may be socially troublesome. 9. Poorer families respond to economic crises The poor adiust co crises by trying to protecr their long-term interests to the excent thar their assets- in ways that are harmful to their longer- including human capital-permit. In particular, they do not pull their children out of school during term well-being. contractions, except when the downturns are long or deep. 10. OECD-type unemployment insurance is Countries that have raised growth and lowered unemployment through comprehensive economic reforms unsuitable for all LAC economies. should seriously consider these schemes; countries that are only beginning labor reform should view them as a longer-term goal. 11. The informal sector is a safety net for The intersecroral flow goes both ways. Informal sector workers often join the pool of the unemployed. unemployed formal workers. Informal workers never become unemployed. 12. Public works programs are just an Public works programs should be viewed as insurance for informal sector workers, and should be emergency device for times of crisis. maintained in good times-but their nonlabor content should be strongly procyclical. 7. For example, regarding physical investment in developing 9. Much of this research was carried out in the context ofa World countries, the adverse impact of uncertainity is documented by Servcn Bank regional study on "Social Risk Management in Latin America," and Solimano (1993), Pindyck and Solimano (1993), Aizenmann and conducted at the Office of the Chief Economist of the Latin Akmerica Marion (1993), and Serven (1999). and the Caribbean region under the supervision of Indermrir Gill. 8. It is worth noting that chis applies both to upturns and down- Other major contributions ro this report include Rodrik (1999), turns, and to recent years as well as the 1970s and 1980s. Caballero (2000), and Snyder and Yackovlev (2000). 12 CHAPTER 2 Economic Insecurlty in Latin America and the Caribbean: The Stylized Facts ATIN AMERICA SUFFERS FROM A HIGH DEGREE OF ECONOMIC VOLATILITY A HISTORY OF repeated booms and busts has made economic insecurity a major concern for workers and house- holds across the region, and especially for the poorer segments of the population, who are more exposed to the consequences of income and employment variability. Drawing on both macroeconomic and microeconomic information, this chapter sets the stage for the analysis in the rest of the report by reviewing the major trends in economic volatility in Latin America over the last three decades. The objective is to establish the facts concerning (a) Latin America's performance over time and relative to the international experience in terms of aggregate volatility-that is, the variability of key economic variables such as consumption and income; and (b) economic insecurity from the perspective of individual workers and households-that is, the fluctuations in employment, unemployment, and labor earnings. Methodological Considerations How does the economic risk faced by workers and households arise? The answer to this question provides the organizing framework for this chapter. Here, risk may be measured by the variability of the real earn- ings of employed workers; the level, incidence, and duration of unemployment; the rate of turnover in jobs or, inversely, the average job tenure; and the quality or pre- tions. Second, risk faced by households may arise from cariousness of available jobs (a concept often related by microeconomic or sectoral volatility-namely, such factors observers to informality). In addition, the inequality of as the changing allocation of resources across economic sec- income distribution could be seen as a measure of the risk tors and the rapid obsolescence of skills-unrelated to of faring poorly relative to others in society. This deserves aggregate disturbances (see Box 2.1). This chapter is orga- mention here because increased inequality may lie behind nized around the following three factors: aggregate risk, the concern with economic insecurity that appears to have the transmission of aggregate risk to households, and spread across the region. microeconomic risk. The risks faced by households arise from two sources. Assessing Latin America's performance along these First, they may reflect just aggregate volatility-itself due three dimensions poses serious methodological challenges. to external shocks from global goods or financial markets, Beyond basic national income aggregates, there are few volatile fiscal or monetary policies, and other factors. The broadly available labor market and microeconomic indica- speed and extent with which aggregate shocks are trans- tors that permit comparisons across countries or that can mitted to household income and employment outcomes give a regional perspective on the key issues. Some, such as depends on factors including the sensitivity of labor unemployment rates, differ sharply in magnitudes in ways demand to wages, and labor market policies and institu- that suggest differences in data collection or definitions. 13 SECURING OUR FUTURE IN A GLOBAL ECONOMY BOX 2.1 Mning artdIf Zasuring hise*i~ In this report we extensively use terms such as economic Strictly speaking, then, volatility and uncertainty are inseurit, unertwint, variabiy, and voit in theory, not exactly synonymous: the former refers to the overall these conepts are not identical, but in practice they are variation of a variable, while the latter refers only to the closely related. unpredictable part of that variation. In practice, however, the two usually go hand in hand: volatile variables are Definitions also had to predict. For this reason, this report focuses on 1 Economic insecurity refers to the uncertain environenet measuring volatility as a tough approximation to uncer- faced by workers and households due to erratic movements tainty and insecurity. in key economic quantities and prices, sch as empoy- ment, income, and real wages, These variables change, Measurement sometimes abruptly, fom one month or year to the next, How should volatility be measured? Idelly, we need some and the ucertaity surrounding their fiture values is the summry indicator of the extent and fequency with essence of economic insecurity. In the economics literature, which a variable teads to depart from its central trend. A this is commonly referred to as unranty or r nuber of such measures are available from statistical the- It is impportant to distinguish between wo kinds of ory, and in this report we use the standard deviation most risks. Aggregte or ommon risks affect equally most or all eco- often, which quantifies the extent to which a variable typ- nomic actors. For example, the risk posed by fluctuations in ically departs from its average or mean value. Since our worldwide economic activity is common to all developing variables of interest are in many cases expressed as per- i countries, while that posed by fltuations in domestic eco- centas, their standard deviation is also a percentage, nomic activity is common to all workers agnd firms in the While the standard deviation is the most commonly national economy. In contrast, other risks are idda used measure of dispersion, it iS by no means the only (equivalently, microeconomc) or idhey afct only one, ad other measures may be more apprpriate in spe- specific individuals or particul groups of econoric acts. cific contexts. For example, the coefficient of variation- For example, fluctuating demand fbr swel is primarily a defined as the standard deviation of a variable divided by risk specific to the steel industry and its workers; uncerain its mean-might be preferable when the mean and stan- world cofee prices are a source of idiosyncratc risk for cof- dard deviation tend to move together, as is usually the fee-exporting countries but not for the rest. case with variables that display rising or falling trends. Vola r variabilt-ers in turn to the varia- Other "robust" measures (such as the interquartile range) tion of a magnitude around some central tred (typicaly may be superior in the presence of infrequent, large devi- its average or median value), for exampe, the movement ations of a variable from its central value. Using some of of oil prices relative to their historical average. In some these alternatives rather than the standard deviation to cases, part of the variation of certain economic variables measure volatility would change quantitatively some of may be predictable; for example, prices of agricultural the empirical findings in the report, but would leave goods typically rise before the harvest and fal afterward. them qualitaively unchanged. Others, such as labor turnover, are available only for a few economies. In our framework, households could face countries. In many cases, therefore, the discussion has to be increased risk due to larger aggregate shocks, strengthened guided by what can be learned from a few case studies. channels of transmission, higher microeconomic risks, or a It is likewise difficult to identify the links between the combination of all three. Reforms may have affected all evolution of economic insecurity in Latin America over the three of these ingredients, but disentangling their impact last two decades and the process of economic and institu- is no easy matter. In many dimensions, the postreform his- tional reform undergone by many of the region's tory is too short to allow distinction between transitional 14 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS effects derived from intersectoral resource reallocation and the volatility of GDP growth in the typical (median) permanent impacts on economic volatility faced by house- industrial country over the last four decades was just above holds-a distinction that is critical for the design of social 2 percent. In contrast, it exceeded 4 percent in Latin Amer- safety nets. Moreover, radical reform policies are often ica, higher than the levels seen in the most stable develop- implemented along with stabilization measures, so that the ing regions-the seven East Asian miracle economies and permanent effects of the former are difficult to separate South Asia. Sub-Saharan Africa (which comprises mainly from the largely temporary consequences of the latter. low-income economies), the Middle East, and North Africa (largely consisting of oil-exporting economies whose per- Aggregate Volatility in Latin America formance tracks closely the vagaries of world oil prices), Like other developing regions, Latin America faces a high and the rest of East Asia2 experienced even higher GDP degree of volatility of the major economic aggregates growth volatility than Latin America. related to national income, expenditure, and consumption. Volatility in Aggregate Consumption Volatility in Aggregate Output Latin America also suffers high volatility in real private Over the last four decades, the volatility of real output consumption growth-an aggregate which provides a more growth as measured by the standard deviation of the accurate measure of the change in the standard of living of growth rate of real GDP' in Latin America has been twice the population of each region. Using the standard deviation as high as in industrial economies. Figure 2.1 shows that of consumption growth as the yardstick, Figure 2.2 shows FIGURE 2.1 Long-Tenm Volatility of Real GDP Growth 9- 8 [ Population-weighted average 7 - L Region median 6 3- 12 - einlmda Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific FIGURE 2.2 Long-Teffn Volatility of Real Private Consumption Growth 16 - 14 - E Population-weighted average 12 - * Regional median 10 8 8 4 7 02 Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacifit 1 5 SECURING OUJR FUTURE IN A GLOBAL ECONOMY that volatility in Latin America is three times higher than TABLE 2.1 in industrial economies, well above the levels of South Asia Long-Tenn Volatility in Latin America and on par with those witnessed in the Middle East and (Standard Deviations of Growth Rates, Percent) North Africa. PRIVATE COUNTRY GDP CONSUMPTION Dfflerences in Volatility Across Countries Within Argentina* 5.4 5.5 Bolivia 4.0 2..7 the Region Brazil 4.3 'i.l The LAC region comprises a large number of different Chile 5.2 11.5 economies, and their respective performance from the Colombia 2.4 2.7 Costa Rica 3.4 point of view of economic volatility has been equally Dominican Republic 5.5 ,.8 diverse (Table 2.1). The volatility of annual GDP growth Ecuador 4.8 2. 7 has been highest in Nicaragua (with a standard deviation Guatemala 2.7 62 exceeding 7 percent) and the Bahamas (8 percent), and Haiti 4.6 4 5 lowest in Guatemala and Colombia (with standard devia- Honduras 3.0 3.8 Jamaica 4.6 IC.I tions around 2 percent). These two countries, along with Mexico 3.7 4.0 Bolivia and Ecuador, the data of which do not yet reflect Nicaragua 7.5 12.4 Panama 4.6 10.4 the 1999-2000 crisis, also possess the best record in terms Paraguay 3.7 8.5 of long-term stability of real private consumption growth. Peru 5.5 6.3 At the other extreme, the largest variability in consump- Trinidad and Tobago 5.6 12.3 U ruguay 4.2 6.3 tion growth rates, in excess of 10 percent, was suffered by Venezuela 4.4 5.2 several of the smaller economies in the region-the Mean 4.5 6.5 Bahamas, Belize, Guyana, Nicaragua, and Trinidad and Median 4.6 5.5 Tobago-as well as Chile. Smaller countries The regional and country experiences in Figures 2.1 and Bahamas 8l1 1 8 Belize 3.7 11.8 2.2 and Table 2.1 illustrate three general points. First, Barbados* 4.6 4.8 lower-income economies typically suffer from higher Guyana 5.8 19.5 macroeconomic volatility. This is apparent from the fact Mean 5.5 12.4 macroeconomic ~~~~~~~~~~~~~~~~~~~~~Median 5.2 12.6 that in all developing regions volatility is considerably higher than in industrial economies. This largely reflects Unueighted Average 4.6 7.4 Overall median 4.6 6. 3 the fact that lower-income economies often possess less WeightedAverage55 4.2 5.1 diversified productive structures than more advanced *Cf,sumprion figures for Argentina and Barbados correspond to mrat, and nor pei- economies, which increases their exposure to risk; more- vare, consumption. Argentina:1961-98; Barbados:1967-94. *aWeigh,cd av-rages usc 1995 poplation. over, once shocks happen, lower-income countries are less Note: Sample Period: 1961-98. Exceptions: GDP: Bahamas (1961-95), Barbados ablc to weather them than richer countries, due to their (1961-95), Guyana (1961-95), Peru (1966-99) Foe Argentina, Brazil, Chle, Colom- bia, Ecuador, Peru, and Venezuela, figures are updated to 1999. For consuarnprion: more limited access to external financing and their less- Bahamas (1978-87); Barbados (1967). developed domestic financial systems. The second stylized fact, apparent from Table 2.1, is that with few exceptions smaller economies tend to suffer These two facts are summarized by Figures 2.3 and 2.4, higher volatility than bigger economies. Smaller which plot the volatility of GDP growth against per capita economies are typically much more open to trade than income and country size (as measured by the logarithm of larger ones, and yet they cannot diversify their production population) for a large number of countries. As can be seen, as much as the latter. This makes them more vulnerable to in each case a negative relation emerges. It is worth noting terms of trade shocks (Easterly and Kraay 1999). Moreover, that for industrial countries actual volatility tends to fall many of them are located in regions prone to hurricanes short of what could be expected on the basis of their and other natural disasters, as is the case of the Caribbean income and size alone-that is, in the figures they tend to subregion. cluster below the line of best fit. 16 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2 3 GDP Growth Volatility and Country Size 16 14 NC 12 C g 10 Y ~~~ ~ ~ ~ ~ ~ ~~~~~BHS #Q *MN*MW*CH 5) 8 * CNN ° 6 MLI MDG MWI 6N GUY- *TT A SDTDM4, OW *o CV 4 DN AC AR, IR KWTB *BEN JAM ALA L GAT EUL ALRB A PNA*CRD* CAOGAU C N IS Ym EA ISL *LUX *GAB SGP+R My* TC L TUNAA K *MYS *CD TEA 4 JPN *IRL HND 4CHEGTM *TWN *EP A EGY A*NK SAEN tED ALA *CAN *ITAA OLKA *NOR GAA 4860 C0L ABA 0 I I I I 10 12 14 16 18 20 22 Log of Population (1995) FIGURE 2.4 GDP Growth Volatility and Per Capita Income 0.16 0.14 MAR 0.12 N (;MR 0.1 > Bl DZA 0.08 C a a S E2NM~~~~~~~~~~~~~~WI IEl NC SE"N A 0.06 - CLG ADG CA F 0 *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~A PE* IT SNCG fl NAW KWT MMFFJI 4GARE LSOECH EN~~~~~~~~~~~ A 0.04 -EUORY EN NR MENEB A Cs AMLwar UMAL KO MSAU AB MYS * OVEX SDGP HEK AGRAC J.PN RM HND CEGYIND TNRIRL IAK ETWN PETV 0.02 RLKA Au NAV OF NCA R IA ROAGD GMA SAB AE A A 4 6 8 10 12 14 16 18 Log of Real Per Capita GDP 17 SECURING OUR Ft-TtJRE IN A GLOIBAL ECONOMY These relationships between volatility and country size industrial world, to protect their consumption fromf fluctua- and volatility and per capita income are weak, however. tions in income through mechanisms such as asset depletion, Together, the latter two variables account for less than 10 borrowing, or countercyclical public sector policies. The percent of the observed variation in volatility across coun- result is that the impact of economic fluctuations on the wel- tries, which implies that its causes have to be found else- fare of households is likely much more severe in Latin Amer- where. Indeed, Chapter 4 will show that the main roots of ica and other developing regions than in OECD economies. volatility lie in trade and financial shocks, policy volatility, and underdeveloped financial systems. These factors account Regional Trends in Economic Volatility for the bulk of macroeconomic volatility and-once they are World regions experienced marked changes in economic taken into consideration-income level and country size are volatility over the last four decades, and LAC has been no no longer significantly associated with aggregate volatility. exception (see Figure 2.6). In industrial countries (as well as The third fact apparent from the region and country Sub-Saharan Africa), median GDP and consumption comparisons above is that in LAC, as in almost all develop- growth volatility peaked in the 1970s, largely a reflection of ing regions, the volatility of consumption is typically larger the global impact of the oil shocks of the 1970s. In LAC, in than that of income or production (this is the case in 21 of turn, volatility increased further in the 1980s-as the debt the 25 economies in Table 2.1). This is in contrast with crisis and the ensuing macroeconomic and financial turmoil more developed economies, where the volatility of con- threw many of the region's economies into disarray. The rise sumption growth is similar to or smaller than that of real in the variability of macroeconomic aggregates was signifi- income growth. This phenomenon is summarized by Figure cant for LAC: the median standard deviation of real GDP 2.5, which presents volatility measures for both developing growth rose from 3.5 percent in the 1970s to just under 5 and industrial countries. Industrial countries typically clus- percent in the 1980s, while that of private consumption ter on or above the 45-degree line along which private con- growth rose to 6.1 percent, a level surpassed in that decade sumprion and real GDP growth are equally volatile. Devel- by only the low-income economies of Sub-Saharan Africa. oping economies, however, tend to cluster below the In the 1990s, however, the rising trend in aggregate 45-degree line, reflecting their higher consumption volatil- volatility in Latin America was partially reversed. Contrary ity, and LAC economies are no exception to this rule. perhaps to popular perception, the available information on This regional contrast reflects the more limited ability of GDP (that reaches up to 1999 for the region's largest consumers in developing economies, relative to those in the economies3) shows that the variability of real GDP growth FIGURE 2.5 GDP Growth Volatility and Private Consumption Growth Volatility 16 a Latin America and the Caribbean ,, 14 - |Industrial Others - 12 0 10 C x 2 6 0 0%, 2% 4% 6% 8%o 10% 12% 14% 16% Standard Deviation of Private Consumption Growth 18 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.6 Volatility of Real GDP Growth by Decade (Regional Medians) El 1960s 1980s 5 D 1970s 1990s A: - |1 LII 2 Industrial East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Economies Pacific 7 the Caribbean North Africa and Pacific FIGURE 2.7 Volatility of Real Private Consumption Growth by Decade (Regional Medians) 10 -_ 9- j 1960s 1980s 87- D 1970s 1990s 7- 6- 5 4- 3- 2 1 ,l1 1 0 T Industrial East Asia 7 Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Economies the Caribbean North Africa and Pacific declined substantially across the region, to levels comparable It is worth emphasizing that this cycle of rising eco- to (and in a number of countries, lower than) those witnessed nomic instability followed by declining economic instabil- in the 1970s. This decline in growth volatility was shared by ity in Latin America is readily apparent in the macro- other developing regions, with the major exception of the economic data, and is not a result of the breakdown of the East Asian miracle economies. In turn, the volatility of pri- period of analysis into subperiods (decades) used here. The vate consumption growth (on which data are only available same pattern arises with alternative period definitions,4 though 1988) declined as well relative to the 1980s, but to and even if annual rather than decade-based measures of a more limited extent than that of GDP growth. volatility are used, as shown in Box 2.2. 19 SECURING OUR FUTURE IN A GLOBAL ECONOMY TheS 3 de linein 7 GDP gothii'x'S volatiity5 6'1g'in"6i th 1990 Regard 'in}8g privt cosmto grwh pefomac afece moB E lg} st glg contism in<§g o@the> regio (2 ou of the 25 wa mor mied InS a nube ofcutisih ein( FI SGUE 2.8 6o - 4....... # 4 Ii L 3 2-------- p 0 3 -1--lC- 3 l - - 3 - ------ g g 03 i 3 i W '~~ ~,, The0 declin in 7 GDPgrwt voltiit in8 the 2 194 1986Reard8n privat conumpio growt performanc affected most countrie s; in theregion (20 out of the 25 was more mixed. In a n of countries in the region (1333Yea shown in TableKR 2.2)QK3WX3 with the exception of ag. fe,2§3 zw, Caribbean out of the 25 in Table 2.2),volatilitydecli ecnme (th Doiia_eulc at,adTiia eaie ote18s u oei oeo h ags Th eln nGPgrowth volatility in the 1990s thanairtheng7 sraverae consumption voatliy oasmuh, meroremodes ecolatiet ofGothe DomiGnPcand PRiaepulc Conumtion ind LTrinia Am ricatv tohe18sburseisme ftelagt (Ten-Year Window, Regional Median) 9-^--- rvt osmto ---w-................... -----. .-- 9---.----- .----.----------.-- --..-<---w. ............. -- .- 1970 197 197 197 198 18 92 18 96 188 19 92 19 96 19 Year 20 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN THE STYLIZED FACTS TABLE 2.2 Volatility in Latin America Over Time (Standard Deviations of Growth Rates by Decade, Percent) GDP PRIVATE CONSUMPTION COUNTRY 1960s 1970s 1980s 1990s 1960s 1970s 1980s 1990s Argentina* 5.4 4.3 5.6 5.5 4.3 5.2 5.9 5.7 Bolivia 6.2 2.5 2.9 1.0 .. .. 3.3 0.5 Brazil 3.5 3,2 4.6 3.0 3.6 4.9 4.4 4.9 Chile 2.5 6.7 6.4 3.5 12.0 16.8 9.4 3.4 Colombia 1.4 1.7 1.5 3.3 4.1 1.8 1.2 1.9 Costa Rica 3.1 1.9 4.5 2.4 4.5 3.7 6.1 3.2 Dominican Republic 8.8 4.4 2.7 4.4 10.8 6.7 4.5 8.7 Ecuador 2.0 5.3 4.5 3.4 2.1 2.0 2.5 0.3 El Salvador 2.8 3.1 5.7 1.9 4.2 6.3 6.0 6.9 Guatemala 2.0 1.6 2.7 (.8 1.1 1.4 2.4 1.0 Honduras 2.4 3.7 2.5 2.4 2.4 6.3 3.1 1.9 Haiti 4.3 3.8 2.9 6.4 4.3 4.8 4.4 Jamaica 2.3 7.1 4.3 2.2 3.9 9.5 7.6 16.5 Mexico 2.4 2.1 4.4 3.6 1.9 1.8 4.9 5.1 Nicaragua 3.3 12.2 5.4 2.3 4.9 11.7 15.8 14.4 Panama 1.6 3.0 6.5 2.6 .. .. 11.3 10.2 Paraguay 2.3 2.3 5.3 1.5 3.8 4.5 11.2 12.5 Peru 2.5 2.6 8.4 5.2 2.8 6.7 8.3 4.6 Trinidad and Tobago 3.5 4.8 5.7 6.8 6.6 8.7 14.0 17.6 Uruguay 2.7 2.6 6.6 2.8 5.6 3.1 9.4 5.4 Venezuela 3.3 2.8 4.8 5.0 .. 3.4 3.5 4.7 Mean 3.3 3.9 4.7 3.3 4.6 5.8 6.6 6.5 Median 2.7 3.1 4.6 3.0 4.1 4.9 5.9 5.0 Smaller countries Bahamas 0.8 12.9 5.7 1.8 .. 15.8 9.7 Belize 0.8 3.5 5.8 3.0 16.1 4.9 Barbados* 4.8 4.0 4.3 4.0 .. 4.8 Guyana 7.0 4.1 5.2 4.2 30.3 12.4 19.5 8.6 Mean 3.3 6.1 5.2 3.2 30.3 14.1 15.1 6.1 Median 2.8 4.1 5.4 3.5 30.3 14.1 16.1 4.9 Unweighted Average 3.3 4.3 4.8 3.3 6.0 6.6 7.7 6.4 Overall median 2.7 3.5 4.8 3.0 4.2 5.2 6.1 4.9 Weighted Average** 3.2 3.1 4.6 3.5 3.5 4.4 4.9 4.7 Notes: See Table 2.1. Decades are defined as 1961-69. 1970-79, 1980-89, 1990-99.Weighted averages use 1995 population. than that in median consumption volatility. In over half goods demand to product prices and, as a result, also the the countries, consumption volatility remained in the sensitivity of labor demand to wages (see Rodrik 1997, 1990s above the levels of the 1960s and 1970s. for example). This would imply that any given aggregate shock would lead to larger movements in wages or Transmission of Aggregate Volatility to the Labor employment than previously. Estimates of labor demand Market equations for Brazil, Chile, Colombia, Mexico, Peru, and Even if aggregate volatility had remained unchanged, Uruguay, however, do not suggest that this has been a reform and stabilization measures may have led to tighter strong effect to date.' As an example, Figure 2.10 plots linkages between macroeconomic shocks and labor markets the own wage elasticities for blue- and white-collar through multiple channels, some of which appear on the workers for Chile during 1980-1995, a period of left-hand side of Figure 2.9. increased protection (1984-87) and then increasing inte- Increased product competition brought about by mar- gration. Although the series is volatile, neither casual ket-oriented reforms may increase the sensitivity of observation nor statistical tests suggest any trend 21 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 2.9 when economic conditions demand it. Recent research on The Links Between Aggregate and Microeconomic Volatility labor market reform suggests that government-mandated reductions in firing costs had a negligible impact on labor Macro Volatility Global Integration demand in Brazil, Colombia, Peru, and Uruguay, but did reduce tenure to a greater or lesser degree.6 Though in all cases overall employment may rise as a result of these Channels of Communication Channels of Communication reforms, workers may be more exposed to layoffs than pre- Labor Demand Elasiciries Technological Change viously. The ambiguous impact on risk also arises From the Labor Market Reform Direct Investment Flows diminished power of labor unions that has occurred either Inflation Reduction because of greater competition or the political evoLution of the 1970s and 1980s. Weaker collective bargaining in Uruguay, for example, is associated with lower unemploy- ment, but also more wage volatility and higher labor Measures of Risk demand elasticities than in the past (Allen, Cassoni, and -Wage Volarility Labadie 1997; Cassoni 1999). -Unemployment The successful fight against high inflation rates in the -Turnover -Informaliry region, and the resulting reduction in the ability to adjust -Distributicn real wages through inflation, have led labor markets to adjust through employment instead. Given the absence of unemployment insurance in most countries, workers may increase during the period of study (see Fajnzylber and feel more at risk, even if the shocks to labor demand are no Maloney 2000). larger than before. Figure 2.11 suggests an inverse relation Trade liberalization, and the labor market reform mea- between how much real wages fall with a shock to CGDP and sures thought to be a necessary complement to it, have how much unemployment rises with the same GDP shock loosened the relationship between firm and employee. To (Gonzalez 1999). Studying the impact over the last 20 years start, in a highly competitive environment, the traditional reveals ambiguous although broadly similar results. As promise of a lifetime labor contract is simply less realistic inflation falls, Argentina, Mexico, Peru, and Venezuela do than in the past. More generally, competing firms need appear to adjust less through wages, but only Colombia and greater flexibility to reallocate or reduce their work forces Mexico adjust more through unemployment as predicted. FIGURE 2. 10 Long-Run Own Wage Elasticities, 1980-95, Chile 0.7 - - Blue Collar 0.6- -.6 -< White Collar 0.2 - /\ 0.1 0 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1)95 22 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.11 Wage and Unemployment Okun Coefficients 2.50 - *Bolivia * Mexico 2.00 n f Peru < 1.50 U * Paraguay * Costa Rica t : 1.00 - *Argentina * Uruguay O Brazil Venezuela 0.50 *- * Chile UNITED STATES * Panama 0.00 - Colombia -0.50- 0.00 0.20 0.40 o.60 0.80 Unemployment Okun Coefficients Brazil, Bolivia, and Uruguay show either counterintuitive tion of the inflationary surge experienced by many Latin or insignificant coefficients (Galindo and Maloney 2000). American economies in the 1980s, which resulted in a gen- Finally, as depicted by the right side of Figure 2.9, the eralized rise in the variability of real wages during that general reduction in global barriers to direct foreign decade, followed by a decline derived from the success of investmenr may also make it easier for foreign investors to inflation stabilization. As a result, in the 1990s real wage relocate production in response to modest movements in volatility was broadly similar to, or even lower than, in the wages.7 There is ample anecdotal evidence of such footloose 1970s. The decline in the variability of real wages in the industries, particularly in Central America, though the 1990s was particularly marked in Argentina, Bolivia, and findings of research on firm entry and exit behavior in Chile. Brazil and Peru also witnessed a substantial decline Chile and Colombia are ambiguous.8 More generally, being in wage volatility, although its average level in the 1990s more tightly linked to the process of technological advance was still high, reflecting the persistence of inflation in the may create more dynamic industrial sectors in the region, early years of the decade. but also require workers to retrain and find new jobs more It could be possible, however, that individual workers frequently. face higher idiosyncratic wage risk than disappears in aggregation. Calculating the variance of individual wages Are Latin American Workers Facing More Risk? across a year from panel labor force data, Arango and Despite the falling trend in macroeconomic volatility, the Maloney (2000a) find no trend in Mexico since 1987, and mixed evidence on changes in transmission mechanisms, a downward trend in Argentina, consistent with the aggre- and the possibility of direct microlinkages, the question gate data above. Since workers may be particularly con- remains as to whether, overall, workers are facing more risk cerned about negative shocks, Figure 2.12 transforms the than in the past. changes (by taking the squared values), but again does not reveal a worsening trend. In Mexico, the size of negative Cbanges in Earnings Volatility shocks for skilled workers rose and then fell. In Argentina Wage or earnings volatility captures a central dimension of it is difficult to tell because of cyclical fluctuations, but income volatility. At the aggregate level, Table 2.3 pre- there appears to be an upward trend. sents this information for 14 countries in the region. The Still, it may happen that even if income shocks for the cycle of rise and fall in volatility tracks closely that found mean or median worker change little or not at all, workers in the macroeconomic aggregates. This is mainly a reflec- could be more exposed to large catastrophic shocks that 23 SECURING OUR FUTURE IN A GLOBAL ECONOMY TABLE 2.3 (see Figure 2.13). This evidence suggests that workers are Real Wage Growth Volatility in Latin America and the Caribbean, not being hit particularly harder by catastrophic shocks Percent than prior to the reforms. ENTIRE COUNTRY 1970s 1980s 1990s PERIOD Trends in Unemployment Rates Argentina 15.8 15.0 2.1 12.2 In sum, workers are not facing higher volatility in real Bolivia 14.5 31.9 3.9 20.3 Brazil 8.7 26.2 10.2 16.8 wages. However, as suggested above, this may only reflect Chile 20.6 6.0 1.1 12.0 a new reality in that, because of a fall in inflation without Colombia 6.7 3.1 2.5 3.9 deep reforms of labor market institutions, adjustments Costa Rica 6.9 16.3 3.3 11.2 Ecuador 6.1 10.7 6.4 8.6 occur largely through fluctuations in unemployment. El Salvador 7.1 14.1 13.0 11.7 Since, in the absence of insurance, unemployment implies Guyana 5.9 15.4 13.3 11.8 Mexico 3.2 10.2 7.9 8.1 a catastrophic fall in income, this is clearly a central Paraguay 4.3 5.1 6.6 5.1 dimension of the risk that workers face. Table 2.4 presents Peru (CEPAL) 7.0 23.5 9.3 16.1 decade averages of the unemployment rate for 1]3 Latin Peru (1IL0) 11.2 28.9 10.2 19.3 Uruguay 6.4 9.1 4.3 7.5 American economies. The trend that the table reveals is Venezuela 3.3 6.2 10.6 8.0 broadly similar to that already identified from the macro- LAC Mean 8.5 14.8 7.0 11.5 economic aggregates. Unemployment rates rose in the LAC Median 6.9 14.1 6.6 11.7 1980s in almost all the countries in the table, with the exception of only Brazil and Mexico. In the 1990s, how- such measures will miss. One way to rectify this is to exam- ever, the rise was partially reversed: average unemployment ine changes in the shape of the full distribution of income declined in 7 out of 13 economies, although it did increase shocks, and not just those at some average such as the in some major economies such as Argentina. median. When we examine the 25th quantile (the point at The net result is that in the 1990s average unemploy- which 25 percent of income shocks are lower and 75 per- ment rates still remained above the levels of the 1970s in cent higher), we find trends similar to those found at the 7 of 12 economies. The increase was substantial in median for Mexico, and no significant trend in Argentina Argentina, Paraguay, and Venezuela, and the dlata for FIGURE 2 12 White- and Blue-Collar Wage Volatility, Mexico 0 High Schooling a Low Schooling .8 - .6 0 C3 C2 0 013 .4 - 0 O o~~~~~ .2 o 0 10 20 30 40 Cohort Annual Log Differences of Real Wages Squared 24 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.1 White- and Blue-Collar Wage Volatility, Argentina O High Schooling A Low Schooling .2 0 .15 .1- 0 0~ ~ ~~ ~ ~ ~~~~~~~~~~~ 934 939 944 949 954 959 964 969 974 979 Cohort Annual Log Differences Df Real Wages Squared Colombia and Uruguay conceal falls in the early 1990s that Argentina, Brazil, Mexico, and Uruguay experienced his- were dramatically reversed by the end of the decade: booms torically unusual periods of growth with limited formal in the nontradables sector, particularly construction, sector employment creation.9 This is especially striking in absorbed labor and partially obscured the dislocations in Uruguay in the late 1990s, where unemployment was at the restructuring tradables sector. The end of these booms levels similar to those of the debt crisis years in the early both displaced workers, and revealed the higher industrial 1980s, despite apparently healthy economic growth. sector unemployment. More generally-and perhaps cen- Behind these numbers lie numerous phenomena that are tral to the stated feeling of insecurity in the region- fundamentally transitory: the dislocations due to industrial TABLE 2 4 restructuring, the temporary real exchange rate overvalua- Average Unemployment Rates in Latin America and the tions accompanying necessary stabilization plans, contam- Cadbbean, Percent ination effects from the Tequila, Asian, and Brazilian crises, and in Colombia's case, a rise in rural violence. Nei- ENTIRE COUNTRY 1970s 1980s 1990s PERIOD ther theory, nor the experiences of the Asian newly indus- Argentina 4.1 4.8 12.3 6.9 trialized countries and of Chile suggest that openness Bolivia 6.3 7.7 4.8 6.3 implies higher long-term rates of unemployment. How- Brazil 6.6 5.4 5.4 5ever, lower inflation rates may imply more frequent adjust- Chile 10.5 14.4 7.0 1037 Colombia 9.6 11.3 10.7 10.6 ment through quantities that push up unemployment rates Costa Rica 5.0 6.9 5.3 5.9 during downturns, and hence imply more risk for workers. Guyana 0.5 0.3 n.a. 0.4 Honduras n.a. 4.5 3.4 3.7 In addition, the nature of unemployment may change Mexico 7.0 47 3.8 5.0 even if the level stays the same-either a higher propensity Paraguay 7.3 11.8 14.2 10.7 to become unemployed or longer unemployment spells Peru 7.4 7.4 8.5 7.7 Uruguay 9.2 10.6 9.8 9.9 may be interpreted as riskier by workers. Neither effect Venezuela 5.7 8.8 9.7 8.0 enjoys strong empirical support. Probit analysis using the LAC Mean 6.6 7.6 7.9 7.0 Argentine and Mexican panel data, for example, does not LAC Median 6.8 7.4 7.8 6.9 suggest secular increases in the probability of becoming 25 SECURING OUR FUTURE IN A GLOBAL ECONOMY unemployed. In both countries, skilled workers in non- related with rising unemployment, which makes it diffi- tradable or protected sectors show lower probabilities of cult to postulate a permanent movement. Further, in Chile, becoming unemployed, but also more difficulty in being the duration of unemployment increases during the rehired after job loss. Colombia shows more substantial restructuring period, but then falls essentially to its 1960s swings in hiring and firing with given movements in GDP, levels (Figure 2.16a)." and exit rates out of employment and unemployment rose by 1 percent (Kugler 1999). Changes in Turnover Rates Neither Mexican nor Brazilian panel data suggest an More generally, turnover among jobs, while a necessary increase in duration across the 1990s. However, using byproduct of the creative destruction that offers new aggregate labor flow data, Figures 2.14 and 2.15 suggest a opportunities to some workers, may also represent more possible rise in duration of unemployment in the interior uncertainty for others. As Table 2.5 shows, turniover is of Uruguay'" relative to the early 1990s, and maybe a rise higher (or the length of tenure lower) in LAC countries in Montevideo. However, duration is always inversely cor- compared to OECD countries. However, turnover depends FIGURE 2.14 Unemployment Rate and Expected Duration, Montevideo 18.0 - _ Unemployment ( _ L4.0- 4 + ~~~~~~~~~~Duration (months) 14.0- 10.0- 6.0- 2.0- 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 FIGURE 2.15 Unemployment Rate and Expected Duration, Interior Uruguay 14 - 1 Unemployment (X)| | + Duration (months) 6 2 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 26 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.16A Unemployment Rate and Expected Duraton, Greater Santiago a rate A length 22.2 -0 a a 1962 1969 1975 1982 1987 1990 1995 Year on education, per capita income, and other demographic/ Mexico, tradables industries show higher rates of growth variables. Thus, for example, younger workers turnover in Chile (Levensohn 1999). This suggests that, change jobs more frequently, and lower levels of education to the degree that trade liberalization expands the share can imply lower levels of firm specific capital, and hence of tradables in total output, it may lead to more churning higher voluntary separations. In fact, in Mexican enterprise in the job market. surveys, over 80 percent of separations were reported to be The aggregate labor flows data do find evidence of quits, not fires (Maloney 1999; Maloney and Ribeiro decreasing tenure in Uruguay (Figures 2.19 and 2.20). 1999). As Figure 2.16b suggests, once we adjust for these However, these movements again parallel the sharp rise in factors, the region does not show "conditionally" higher unemployment, and therefore it is difficult to argue that turnover. 12 they are permnanent: a sharp reduction in tenure also However, a finding of increased turnover across the appeared during the 1973-87 restructuring period in period of liberalization may imply increased risk. But, as Chile that, by 1995, had largely although not completely in the industrial country literature, there is only mixed reversed (Figure 2.21). evidence that either greater trade liberalization or expo- sure to technological change leads to greater turnover pRise in Informality overall, beyond that discussed on the impact of firing COSt Even if all the above indicators of stability and employ- reductions mentioned earlier.23 Figures 2.17 and 2.18 ment showed no change, there is a concern that the quality plot the evolution of turnover in the manufacturing sec- of jobs available has fallen. In particular, the share of work- tor in Colombia during 1980-91, and Chile during ers unprotected by labor legislation or benefits has risen as 1980-95, in both countries periods of increased trade apshare ofLatin America's workforce.International Labour protection and then may 'impyincr4 Disaggregated into Organisation and Inter-Ad rerican Development Bank turnover due to birth and death of firms (BD) and studies suggest that 80 out of 100 jobs created in the turnover due to readjustments by continuing firms 1990s were in the informal sector, and the Economic Com- (Cont), there is no obvious pattern across time in either mission for Latin America and the Caribbean (ECLAC) country (Fainzylber, Ribeiro, and Maloney 2000). How- reports thao the share of informal employment in the work- ever, as with involuntary separations in Argentina and force rose from 52 percent in 1990 to 58 percent in 1997. 27 SLCURLNG OUR FLTUltE IN A GLOBAL ECONOMY TABLE 2.5 Labor Turnover, LAC and OECD Countries LAC OECD 9 < 2 Years Seniority (Manufactures) 38.1 24.5 Average Tenure (ManufactuLres) 7.61 10.5 SoR-c Maloney (I 999b). FIGURE 2. 16B Labor Turnover, LAC and OECD Countries .1 Ven (re Ita E 05 -Spr Fin Bel E IlUK Swit Lux v) 0 Ger Fou P., AuHon Net Pan Fra Den Aui S n Jap -.05 Bra -.1 -4 -2 0 2 4 Adjusted Mean Tenure Seouce. Maloney 1999b. FIGURE 2.17 Evolution of Turnover in Chile, 1980-95 0 Total A New and Dying Firms E] Continuing Firms 0.31 0.04 1980 Year 1995 Chile: Gross Job Reallocation 28 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.18 Evolution of Turnover in Colombia, 198I91 0 Total A New and Dlying Firms 0 Continuing Firms 0.28 0.07 1980 Year 1991 Colombia: Gross job Reallocation Some subtlety is necessary in interpreting these trends. from 18 percent to 22 percent from 1987 to 1990 in Mex- Recent work at the World Bank suggests that the conven- ico, and 23 percent to 27 percent from 1988 to 1993 in tional view of the sector as a residual for displaced formal Argentina, at the same time that the premium self- sector workers is probably incomplete. In many ways infor- employment enjoyed over formal salaried work rose from 0 mal self-employment behaves more like an unregulated percent to 25 percent and 4 percent to 13 percent, respec- entrepreneurial sector where, as in industrial countries, the tively. The expansion of the sector makes sense if we risks of entrepreneurship and lack of protection under indi- believe that entrepreneurs prefer good times, such as the vidual labor codes are voluntarily taken on (see Box 2.3).'5 construction booms in both countries, to start new busi- After economic reforms, for example, informal self- nesses.16 This is fully consistent with interview data from employment as a share of the work force rose procyclically both countries that suggest that roughly 70 percent choose FIGURE 2.19 Expected Tenure in Current Job, Montevideo (Monthsl 110 - 90 70 30- - l 1982 1985 1988 1991 1994 1997 29 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 2.20 Expected Tenure in Current Job, Interior Uruguay (Months) 110 -- 90 -. 70 50 30O- 1986 1989 1992 1995 1998 FIGURE 2.21 Expected Tenure on Job, Greater Santiago 89.7 - 0 0 ar a a~~~~~~~~~~~~~~~~~~~~~~~~~~l ao a n~~~~~~~~ 00~~~~~~~~ \ OX 34 - O 1962 1969 1975 1982 1987 1990 1995 Year to be in the sector for reasons of independence and higher share in self-employment, most of which is unprotected, earnings, and are not looking for other jobs (Arango and is no higher than that of OECD countries or other Maloney 2000). regions. This also suggests that the strong negative relation- But developments in the sector in the later 1990s may ship between formal sector productivity and the share of also reflect undesirable increases in uncertainty. Informal the workforce in self-employment (Figure 2.22) is due to self-employment shows countercyclical behavior in Uruguay the growing attractiveness of salaried jobs relative to self- and perhaps in Peru and Mexico after 1992.17 Further, there employment over the course of development. When is a secular increase in the share of informal salaried workers adjusted for income and other demographic variables, in Mexico after 1992, and in Argentina and Uruguay after Figure 2.16b again suggests that the LAC region's labor 1995, at the same time that their relative incomes were 30 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS turnover, only time will tell whether these are permanent WhyVWluer5RflobesPN ftrb*sPtwo.*se~w~ increases or related to the low rates of job creation and high rates of unemployment experienced during restructuring. It isX atmions!-formay Xwores tO desire to be 'tupro- cetd if .te,ON S4 the pvided,bee ass The Deterioration in Prospects of Older Workers much as te implicit taxt iidfo them. Worhebea- Much of the reported dissatisfaction in the region may be due efis are financed either hrogh explt taxes on work- to workers laid off in the restructuring process and who have ea, o implicty as lower waes. leficenldy povied suffered substantial falls in income, or are experiencing diffi- beriehw-poor medical sen! a social security sys- culty finding jobs. The privatization process has led to a fall em see as haxkmpt and unliable, or an unemploy- in often privileged jobs in public employment from 15 per- meat ntsurtce syem substtially di&iren from one cent of the labor force in 1990 to 13 percent in 1998. More workers wouldchose7-provid incentives,for woifers generally, restructuring of the private sector, both in tradables toIworkoff thn book. In his intervies with workers in and nontradables, may have had the same effect. Studies find Guda ec, Robens (1989) 6nds that, "Many an increase in subcontracting across the period that could be infrmmns cited the deidutio made for welfre as a due to restructuring of firms where they now put their clean- disadvantage offorml employmet, parictia.ly since. ing and security services out to unprotected employees. Mex- the servicedey Vceived were poor." In addition to ico, for example, experienced a once-and-for-all 3 percentage avidingonlP izaxs and regtlation, inornm work- point increase in the share of the workforce in subcontracting e may he d ent "protection" (see. in the mid-1990s. This was accompanied by a fall in relative Amade, Glt and Neri (2000 adal.oey (19:. incomes that suggests a deterioration in "job quality." Older displaced workers may be particularly affected by their inability to requalify themselves or by the unwilling- falling. Uruguay also experienced a rise in what workers ness of firms to employ them. While in both the formal termed "unstable employment" during the 1990s. and informal sectors, workers with primary education In sum, there is evidence of an increase in uncovered gained during the 1990s in Argentina, Brazil, and Costa work in the region that may imply a degree of lack of pro- Rica relative to those with more schooling, there are two tection for a segment of the workforce. As with falls in job notable exceptions in the case of Argentina (Arias 1999 in FIGURE 2.22 Self-Employment versus Industrial Productiviy, OECD and LAC .461 Peru Bot EIS Col Ven Hon Gua Par Uru Arg Gre Pan Chi MehFurk CBr Kor Ita Por Spa New Ice Ire A6901 576 -NbN I I l ~ ~~ ~~~~ II 6.63332 Log of Industrial V.A./Worker 10.5322 31 SECURING OUJR FUTURE IN A GLOBAL ECONOMY Cunningham and Artecona 2000). The first is that of those is mixed. In Argentina (1988-97) and Brazil (1989-95)- starting new informal businesses; many are older workers both liberalizing economies-and in Costa Rica displaced from previous jobs. The second group is older (1989-95), the gap between educated and primary school informal sector workers who may have lost their market workers declined, especially for women. Further, Chile, the niche; repairing, for example, domestically produced cars bellwether country for the region, experienced a harsh or working in small print shops due to imports of new cars deterioration in wage dispersion in the late 1980s, but a and technologies. In Peru in the 1990s, older workers reversal in the 1990s as returns to higher education fell to experienced an increased probability of becoming informal historically normal levels.'9 due to reduced restrictions on firing workers, and in both What has not been carefully studied to date is whether, Argentina and Uruguay, older workers suffered from despite relatively constant inequality measures, there may lengthening spells of unemployment.'8 In the bottom be more movement of individuals within the distribution. income quintile group, the share of unemployed Thus, for example, there may be increased earnings mobil- Uruguayan workers over 40 years of age with unemploy- ity or-looking at the other side of the coin-greater risk, ment duration of over a year rose from 28 percent during as a more open economy generates job opportunities in new 1982-86 to about 40 percent during 1999-2000, a period industries and causes other industries to close. Box 2.4 with similar aggregate unemployment rates. In sum, for a suggests that this probably is not the case in Argentina or sizable segment of older displaced workers, the loss in Mexico across their periods of trade liberalization: there value of their human capital may have been substantial. was little or no trend increase in labor income miobility To a lesser degree, these difficult adjustments for mature during the 1990s. What does emerge is that adjustment to workers are likely to become a permanent feature of the crises through unemployment exacerbated by wage rigid- postreform landscape in Latin America. In the industrial ity in Argentina leads to greater downward mobility than countries, the rapid pace of innovation is thought to adjustment through real wages in Mexico made possible by require that workers retrain more than once during their leaving the exchange rate peg. A generalized fall in wages lifetime. The design of training and pension programs will leaves the relative positions of individuals in the income need to reflect this reality. distribution the same, whereas unemployment expeEienced by a few radically changes their position. Changes in Income Distribution It is also possible that it is not uncertainty per se, but Conclusion changes in income distribution, that are creating the per- As in other developing regions, macroeconomic volatility is ception of insecurity. In broad terms, inequality increased high in LAC, and this translates into volatility in aggregate during the 1990s (see Table 2.6), though there is a wide wage measures and unemployment rates. In most countries range of country experience. Inequality in Brazil and Mex- in the region growth volatility is lower today than it was in ico increased between 1986 and 1989 before leveling off or the 1980s, and major labor market aggregates, particularly decreasing until the late 1990s. Chile and Paraguay expe- wage volatility, and to a lesser extent unemployment, rienced increasing inequality, although in Chile the low appear to follow this trend. The evidence is mixed, however, baseline for 1986 may compromise comparability. In on whether volatility has become higher today than it was Argentina, Colombia, Ecuador, Uruguay, and Venezuela in the relatively normal 1970s, because the 1980s are inequality has been relatively stable, with some indexes rightly viewed as an unusually turbulent decade. suggesting an increase, others a decrease. And in Bolivia, What may be a central issue is the slow rate of job the Dominican Republic, and Honduras there is a decrease growth that has coexisted for relatively long periods with in inequality over time. Overall, from the aggregate healthy economic growth rates, most clearly in Argentina, indexes, it would be hard to argue that opening the econ- Colombia, Mexico, and Uruguay. This phenomenon is omy, as in Argentina, Brazil, Colombia, and Mexico, led to associated with lower job tenure, higher unemployment a permanent worsening of aggregate inequality. duration, growing levels of informality and insecure jobs, The ILO and ECLAC find evidence of increased wage and difficulty of reinsertion of laid-off older workers. 'raken dispersion throughout the region, but again, the evidence together, these developments suggest that employrnent- 32 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS TABLE 2.6 Income Inequality Measures by Country, 1986-96 THEIL GINI ATKINSON THEIL GINI ATKINSON Argentina Ecuador 1986 0.51 0.50 0.68 1986 0.47 0.49 0.46 1989 0.68 0.58 0.73 1989 0.36 0.44 0.33 1992 0.48 0.51 0.65 1992 0.48 0.50 0.44 1995 0.52 0.52 0.61 1995 0.53 0.51 0.46 1996 0.53 0.53 0.64 1996 0.50 0.51 0.54 Bolivia Honduras 1986 0.56 0.54 0.60 1986 0.64 0.59 0.62 1989 0.76 0.63 0.64 1989 0.8 0.6 0.58 1992 0.49 0.50 0.41 1992 0.63 0.57 0.58 1995 0.58 0.53 0.41 1995 0.67 0.58 0.57 1996 0.52 0.51 0.40 1996 0.62 0.55 0.53 Brazil Mexico 1986 0.74 0.59 0.52 1984 0.42 0.47 0.33 1989 0.89 0.64 0.59 1989 0.61 0.52 0.39 1992 0.71 0.59 0.59 1992 0.56 0.53 0.39 1995 0.74 0.61 0.59 1994 0.58 0.53 0.39 1996 0.75 0.61 o.60 1996 0.55 0.52 0.38 Chile Paraguay 1986 0.36 0.44 0.30 1986 0.41 0.47 0.33 1989 0.83 0.59 0.46 1989 0.32 0.43 0.27 1992 0.58 0.53 0.43 1992 0.35 0.44 0.34 1995 0.75 0.57 0.47 1995 0.45 0.49 0.35 1996 0.83 0.58 0.47 1996 0.47 0.47 0.32 Colombia Uruguay 1986 0.64 0.57 0.68 1981 0.35 0.43 0.29 1989 0.55 0.54 0.54 1989 0.36 0.42 0.27 1992 0.63 0.57 0.67 1992 0.33 0.42 0.27 1995 0.75 0.56 0.57 1995 0.33 0.43 0.29 1996 0.69 0.56 0.58 1996 0.35 0.44 0.30 Dominican Republic Venezuela 1986 0.53 0.53 0.66 1986 0.49 0.50 0.57 1989 0.53 0.52 0.43 1989 0.41 0.47 0.51 1992 0.54 0.51 0.41 1992 0.38 o.46 0.49 1995 0.62 0.55 0.62 1995 0.43 0.48 0.40 1996 0.42 0.48 0.35 1996 0.48 0.50 0.47 Soure: Wodon (2000). related uncertainty might have risen in some countries in low-inflation environment leaves workers exposed to both the region, and this may he a factor in the perceptions of catastrophic falls in income against which they are not well economic insecurity alluded to in Chapter 1. It is difficult insured, and downward mobility relative to the rest of soci- to establish whether these and other adverse developments ety. The growing infeasibility of lifetime labor contracts in documented here represent a permanent rise in economic the face of global competition, labor market reforms that risks faced by workers and households, or are merely the reduce firing costs, and weakening union power, all have led transitional costs of the extensive restructuring and stabi- to higher turnover rates. Workers in more exposed tradables lization policies implemented in the 1990s. sectors face higher probabilities of displacement, more However, even if these developments are transitory, and turnover, and more wage volatility than in the past. How- if aggregate volatility were to remain unchanged, there is ever, the data do not suggest that such risks affect the labor some evidence that the magnitude of labor market risk may force as a whole. In addition, they are likely to be higher in have risen-or its form may have changed-for certain the short term, while the economy completes its adjust- groups of workers. The need to adjust to shocks through ment to the increased role of global market forces, than in unemployment rather than through falling real wages in a the long term. 33 SECURING OUR FUTURE IN A GLOBAL ECONOMY Income Mobility and Risk In Two Counties linome mobility can be used as a measure of risk which * MoiiW is negatively Associated with growth in takes into accowt not only the abslue level of income Argentina, and positively associated with growtin of individush, but also their ranking in the ovealldis- Mexico. The positive correlation in Mexico is as tribution of income; that is, their relative deprivation. expect in tha growth provides opporunities, mda Figures 2.23 and 2.24 give the trend in the Gini index of thu results in a more dynaric and mobile labor market. labor income mobility for adult men in Argentina and The negative corlion in Argentiam may be due to the exico using panel data. The main findtings are: fiat that the Agentine labor marker adjusts to shocks b* Laor income mobility as measured by this index is throuh quantes (unemployment) ather than prices high in both countries at about 0.3 in Mexico and 0.4 (rea wages). In Mexico, adjusnments tale place through in Atgentina. It is higher in Argentina, in part prc. Employmen losses rst in more rerankis in because the length of time separating observations fr the distribution of income, and therefre higher mobil| the individuals in the panels is greater (a s ster ver- ity than we losses. The difrences betwveen Agentina sus a quaer). k both countries, the level of mobility and Mexico ma alterativey or additionally be due to suggests that traditional one-period measures of dirent exchange re regimes, with pegged exchange inequality areoverestimated. The resuls also suest rates leading to employment losses, while flexible that short-tem safty nets should help to offset fre- exchange rates lead to real wages losses. quent income losses. * Finally, althoug this is not shown in Fiures 2.23 * There is no trend toward highet or lower mobility and 2.24, it can be shown that income mobility is over tine. That is, contrary to what popular beliefs associaed with individual-level charaeristics such as would suest, the resuls do not suggest a large age (the young are more mobile) and education (the increase in risk over timne. less eduated are more mobile, at least in Mexico). FIGURE 2.23 Growth and Income Mobility in Argentina 0.6 - 0.4 5 ' 0.3 -___ _ ___ a 0.2 - - 0. 0 - -0.1I - - . - - - - - - - -_ _ __ _ -0.2_- I I 93-2 94-1 94-2 95-1 95-2 96-1 96-2 Year 34 ECONOMIC INSECURITY IN LATIN AMERICA AND THE CARIBBEAN: THE STYLIZED FACTS FIGURE 2.24 Growth and Income Mobility in Mexico 0.55 - 0.45 - Mobility A ~~~- Growth -0.25 -o 015- 0.05- -0.05- -0.15- 88-1 89-1 90-i 91-1 92-1 93-1 94-1 95-1 96-1 Year As Chapter 1 showed, many countries in LAC are facing 5. Fajnzylber and Maloney (2000) find only mixed evidence of an an incipient recovery, offering enhanced economic oppor- impact of trade variables in Chile, Colombia, and Mexico. Paes de tunities. These may entail increased risks for some groups Barros, Corseuil, and Gonzaga (1999) find no impact in Brazil; and of workers and households-but the available evidence Cassoni, Allen, and Labadie (1999) find a reduction in long-run elas- ticities (absolute value) in Uruguay. does not show a generalized increase in economic insecurity 6. Cassoni, Allen, and Labadie (1999, 2000) for Uruguay; Paes de in the 1990s. Nevertheless, both the increased economic Barros and others (1999) for Brazil; Cardenas and Bernal (1999), and risks that those groups may be facing, and the still high Kugler (1999) for Colombia; Saavedra (1999), and Saavedra and levels of aggregate volatility in the region, provide ample Torero (2000) for Peru. justification for rethinking and strengthening social pro- 7. In the U.K. and Germany there is evidence of strong effects on labor demand from the falling barriers to direct foreign investment tection measures. in Europe. 8. Hatzius (2000) finds evidence that the long-run labor demand Notes elasticity may have risen substantially across the period of increased 1. Gross national product (GNP) is in principle a better measure direct foreign investment. Fajnzylber, Ribeiro, and Maloney (1999), of national income than GDP. However, we use the latter because of however, find limited evidence that the own wage elasticity of firm greater availability of data. Using GNP instead would make the con- entry and exit, a component of which is foreign, has increased in trast between Latin America and other regions even starker, but Chile and Colombia with liberalization. would force us to work with a smaller country sample. For this rea- 9. For Brazil, see Gonzaga (1998); for others, World Bank son, we focus on GDP. calculations. 2. In Figure 2.1, as well as in other figures below, East Asia shows 10. Kugler (1999) for Colombia; Cassoni, Allen, and Labadie a large discrepancy between the regional median and the population- (2000) in World Bank poverty assessment for Uruguay. weighted value. This reflects the large weight of China in the latter 11. Derived by Gill, Haindl, Montenegro, and Sapelli (2000) value, and the fact that China experienced extremely large volatility using Haindl's (1996) methodology for generating labor flows from in the 1960s. repeated cross-sectional data. 3. Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, 12. See Maloney (1999). Plot is ofresiduals ofregressions ofshare and Venezuela. of self-employment and job tenure on industrial productivity, share 4. For example, the qualitative results are similar when using of young people in the work force, the level of education, and the instead an alternative breakdown guided by developments in the level of interest rates. world economy: 1960-72 (pre-oil shock); 1973-81 (oil shocks); 13. Dunne, Roberts, and Samuelson (1989) found no strong 1982-90 (debt crisis and its aftermath); and 1991-present (postdebt trend in plant-level turnover in U.S. manufacturing from 1963-82, crisis and reform period). a period of substantial tariff reduction and technological progress. 35 SECURING OUR FUTUJRE IN A GLOBAL ECONOMY Davis, Haltiwanger, and Schuh (1996) found no relationship Mexico outstripped those of formal sector workers during 1987-96. between U.S. job flows and either import penetration or export share, See Arias (1999), Saavedra (1999), and Cunningham and Arrecona although Klein, Schuh, and Triest (2000) found that the responsive- (2000). ness of job flows to the real exchange rate varies with the industries' 16. Cassoni, Allen, and Labadie (2000) for Uruguay; and discus- openness to international trade. sions with Jaime Saavedra (Peru). 14. Using firm level panel data for Latin America, Roberts and 17. Arias (1999) for Argentina; Saavedra (1998) for Pleru; and Tybour (1996) found high turnover in Chile, Colombia, and Morocco Cassoni, Allen, and Labadie (1999) for Uruguay. relative to the U.S. (Davis and Halriwanger 1992), but no obvious 18. Arias and Saavedra in Cunningham and Artecona (2000); relation with trade reform. Tybout (1996) did find very high exit Montenegro (1998) for Chile. rates following the Chilean liberalization. However, Roberts (1996) 19. These figures are from Wodon (2000). Most measures of found that in Colombia during 1983-85, average entry and exit income mobility in the literature have been developed independently actually rose with trade restrictions, compared to the previous period of the concept of inequality. Yet, there are links between mobility of relative openness. and inequality. Yitzhaki and Wodon (2000) have propose.] a Gini 15. It might also explain why, wich the exception of women in index of mobility to provide for an explicit link between inequality Brazil, wage gains of informal workers in Argentina, Brazil, and and mobility. 36 CHAPTER 3 Desi g ning Social Policy When Peop e Face Risk: A Conceptual Framework S ALREADY DISCUSSED, DURING THE 1990S THERE WERE ECONOMIC REFORMS IN LATIN America and the Caribbean which resulted in its rapid integration into world markets. There appears to be growing concern, however, that the social insurance and social protection mech- anisms existing in most LAC economies are inadequate to deal with heightened economic - t ~insecurity.' In the popular press and opinion polls, there are sentiments voiced in favor of expanding the role of government in countering growing economic insecurity through, for example, intro- duction or expansion of formal unemployment insurance programs, government-sponsored health insur- ance, and safety nets for those not covered by labor market-related programs. Governments appear to be puzzled about how best to help people manage the risks they face. In examining these claims more systematically, Chapter 2 found that, in many LAC countries, aggregate risk actually appeared to have declined in the 1990s relative to the 1980s, and even relative to the 1970s. Microeconomic risk also shows no clear trend; some indicators of volatility, such as real wage fluctuations, have indeed registered sharp declines. In most countries economic growth has picked up over the last decade. These developments-falling economic risk and increased wealth, combined with a clamor for greater social insurance-may appear to be contradictory. More carefiul study using a structured analytical frame- work, however, shows that this is not so. The economics of insurance indicate that, other things being equal, the demand for all types of insurance will rise as incomes (1972), which provides an elegant treatment of an individ- rise. The overall demand will also rise as the potential loss uial s optimal insurance decisions when faced with the faced by individuals becomes greater, and the demand for options of market insurance, self-insurance, and self pro- certain types of insurance may rise even when the world tection. This report attempts to systematically derive from becomes less risky. This chapter provides an overview of solid economic foundations the public policy implications this approach and illustrates its usefulness in formulating of the potential inability of individuals to insure or protect effective but minimalistic social policy strategies to deal themselves effectively (see Gill and Ilahi 2000). with socioeconomic risks. The framework used here allows us to address problems that preoccupy policymakers around the world; that is, The Need for Sound Analysis changes in the demand for insurance due to globalization, The main danger of approaching the problem of risk with- economic growth, or increased uncertainty, and the likely out sound analysis is that it results in serious confusion effects of social safety nets created in response to these about the role of government policy.2 There is considerable changes. The approach is versatile enough to distinguish analytical work on the economics of insurance, which stud- between the policy implications of economywide (aggre- ies how individuals and families react when faced with gate) and idiosyncratic (microeconomic) shocks, between risk. This report relies on the work of Ehrlich and Becker catastrophic (large and rare) and noncatastrophic (small and 37 SECURING OUR FUTURE IN A GLOBAL ECONOMY frequent) losses, and between good and bad instruments for loss. Both involve expenditures. Market and self-insurance insurance and protection against these shocks. The serve to transfer income from the good to the bad state of approach yields insights that can-with some additional the world, but do not reduce the likelihood that these trans- work-lead to rigorous strategy formulation at the country fers will be required. Self-protection, on the other hand, level.3 only reduces the probability of the bad state of the world, In this chapter, we illustrate how a theory of individual doing nothing to the size of the loss in the event it occurs insurance and self-protection can be extended to identify anyway. The critical difference between market insurance "market-augmenting" roles of government (Olson 2000). and self-insurance is that the former uses pooling to spread Under one rather strict interpretation, the public policy risk across individuals. analogs of the individual's insurance and self-protection Individuals or families attempt to smooth consumption problem are social insurance (government actions to aug- over the good and bad states of the world. If both market ment market insurance and self-insurance) and social protec- insurance and self-insurance opportunities are present, the tion (government actions to augment self-protection). individual sees them as substitutes (see Box 3.2). The pro- vision of market insurance likely will reduce self-insurance; Approach and Implications thus, for example, the availability of unemployment insur- A systematic approach to social policy formulation should ance will reduce precautionary saving. The problem of begin by understanding how individuals or families behave "moral hazard" results if the purchase of market insurance when confronted with risk. Fundamentally, there are two reduces self-protection; thus, for example, if unemployment actions that an individual or family can take: insure, that is, insurance is available, people may become more likely to transfer incomes from good to bad states; and self-protect, shirk.4 The most common outcome if moral hazard is acute that is, lower the likelihood that the bad state occurs. Nei- is that private insurance markets may not exist, or may ther is without cost. A comprehensive framework would involve prohibitively high premiums. allow for all types of insurance and self-protection deci- The key features of and insights obtained from this sions. Any constraints on individuals taking these actions framework are: effectively would be of social policy interest, and the prob- lem then becomes one of deciding whether and how gov- * Levels of risk, incomes, and prices or costs of risk- ernments can help remove these constraints. management instruments all are important in deter- In addition to clarifying basic concepts, a good analyti- mining how much individuals spend on insurance cal framework for risk management should have three and protection. attributes. First, it should cover all the major instruments * Market insurance and self-insurance are substitutes, for managing risk and be sensitive to the relationships in that greater availability or lower prices of one lead between these instruments. Second, it should afford guid- to reduced expenditures on the other. Self-insurance ance on how individual efforts to insure and protect against and self-protection are also substitutes. risk can be improved; that is, the circumstances that pro- * Market insurance and self-protection may be substi- vide cause for governments to intervene. Third, working tutes or complements; a lower price of self-protection through the structured framework should formalize exist- increases self-protection and lowers risks, hence ing thinking about the subject of risk but-even more reducing demand for both insurance and self-insur- important-yield insights additional to those that we ance. In overall equilibrium, however, lower risks began with. may also reduce the price of market insurance and, The "comprehensive insurance" approach is especially thus, lead to an increase in the demand for market well suited for these goals. As formalized by Ehrlich and insurance. Becker (1972), the insurance problem of the individual is * An increase in the difference between crisis and non- characterized as one of determining the levels of expendi- crisis income levels (the "income at risk") coul(d lead ture on market insurance, self-insurance, and self-protec- to an increase in demand for insurance. Thus, indi- tion (see Box 3.1). The premise is that individuals can viduals may be richer (in that their expected incomes either insure against loss or lower the probability of the are higher) but may still demand more insurance. 38 DESIGNING SOCIAL POLICY WHEN PEOPLE FACE RISK: A CONCEPTUAL NETWORK BOX 3 1 Market suramnce, Selfinsurance, and SefProtecton: DisinguIshing Featwes and Exampes Market insurance transfers income or resources from a Two examples may help clarify these concepts. An good state to bad but does not change probabilities of individual, faced with the likelihood of damage to his car good and bad states, it is available at an observable (mar- in an accident, may purchase automobile insurance (mar- ket) price, and al.vays involves pooling of risks. ket insurance); he may buy a stronger-hence more Self-insurance, like market insurance, also transfers expensive but otherwise identical-car (self-insurance); or resources from a good state to bad, and does not change he may drive more cautiously, even though this increases probabilities of good and bad states. It differs from mar- travel time (self-protection). Again, faced with a higher ket insurance in two ways: it has an imputed, not actual, probability of being unemployed, a person may try to price (called a "shadow price" by economists), and it does purchase market insurance, may self-insure by increasing not involve risk-pooling. savings over and above what she saves for relatively certain I Sef-protection is different from both market and self. needs such as education of children and retirement, or insurance in that it does not transfer resources from a good engage in self-protection by studying to qualify for a pro- state to bad, but lowers the probability of the bad state. fession in which unemployment rates are lower. It is often difficult to determine whether a decision Note, however, that all three types of actions involve should be classified as self-insurance or self-protection, costs. market insurance requires a premium to be paid; self- since many instruments do both. It can sometimes be dif- insurance implies costs (because, for example, a stronger car ficult even to classify informal insurance measures as mar- costs more whether or not the accident occurs); and self- ket insurance or self-insurance. In such cases, the key dis- protection involves monetary or other costs (for example, tinguishing feature should be the absence or existence of schooling involves tuition fees, and driving slowly or pooling, attending classes implies less time for other activities). * Relatively rare (and large) losses may be better self-protection are available but no self-insurance is insured through market insurance, and relatively possible. The individual would be worse off in this frequent (and moderate) losses through self-insur- case than where all three are available. The reason is ance. Thus, for example, as individuals face lower that for losses that are not rare, the individual would probabilities of becoming unemployed, they may still have to use market insurance. However, we demand less insurance overall, but may also choose know from the framework that market insurance is to have relatively more market insurance and less a less-preferred instrument than self-insurance for self-insurance. At the level of the aggregate econ- losses that occur frequently. Second, suppose that omy, as countries improve their economic manage- market insurance and self-insurance are available, ment and regulations and reduce the likelihood of but it is not possible to invest in self-protection. crises, there will be a shift away from self-insurance Individuals who are relatively efficient at self-pro- (for example, fiscal stabilization funds) toward mar- tection would be worse off because they cannot ket insurance (contingent credit arrangements with reduce the premium paid for market insurance by world financial markets or the international finan- reducing the risk they face through expenditures on cial institutions). self-protection. (See Box 3.3 for a fuller discussion of * Individuals enjoy higher welfare when all three these issues.) instruments (market insurance, self-insurance, and self-protection) are available than when one is miss- Advantages of a Disciplined Approach ing. This can be best explained by two examples. There are three advantages of a disciplined, organized, First, consider the case where market insurance and comprehensive approach to the problem of risk. First, it 39 SECURING OUR FUTURE IN A GLOBAL ECONOMY BOX 3.2 A Theory of Comprehensive Insurance In the Ehrlich and Becker (1972) characterization, there of substitution: are two states of the world: bad (state 0) and good (state - - pU0(c*,s*,r*) 1). The bad state occurs with probabilityp, and the good L' (e9) +I - P(l)(c*,s*,r*) (3) state with probability I-p. The endowed incomes (and hence the consumption) of the individual in the two Expenditures on self-protection reduce the probabilky states are, respectively, ! and l , Thus, the expected of the bad state. These expenditures are optimized at utility of the individual is level r* where the marginal gain from redurcig the prob- ability of loss equals the marginal loss in utility from U = (l-p)U(I')+pUV(I) (1) having to pay r* fbr it in each period: However, faced with risk, the individual may purchase market insurance that involes paying a premium for 7rfbr P 4 E * * 4 every peso of coverage, and paying s pesos if the bad state occurs, The individual also spends resources on self- There are three main results of tis characterization of insurance (c), and self-protection (j') to smooth income the individual's risk management decisions within a over states. Each peso spent on self-insurance reduces the comprehensive insurance model, which would be absent loss in the bad state according to a "loss finction" L(L` c), in treatmenrs that either take a piecemeal approach (for where Le is the difference between endowed incomes in example, examine only market insurance) or neglect to the two states. Each peso allocated to self-protectioa low- include prices. First, market insurance and self-insurance ers the probability of the bad state according to the func- are substitutes; fbr example, an increase in the price of tion p(r). Just as a lower 7r allows the individual to buy market insurance lowers the demand fbr it and increases more market insurance with a given budget, increased the demand fbr self-insurance. Second, the individual is marginal productivity of self-insurance and self-protec- likely to pTrefr market insurance over self-insurance for tion allows the individual to get more at a given cost. insuring relatively rare losses because the 'shadow price" The individual chooses s, c, and r to maximize the of self-insurance does not fall as the probability of loss expected utility function before, the state of the world is decreases, while the price of market insurnce does. revealed (that is, the framework is ex ante): Third, market insurance does not inevitably cause "'moral U - LI - p(p'e, rjULIll- c - - rj + p(pe,+r)U (2) hazard," that is, reduce selfprotection, because of two L(Le, c ) - J2 countervailin effects. On the one hand, market insur- ance reduces the p eie loss, and thereifre, creates a In the absence of market insurance, s is constrained to tendency toward lower self- protection. On the other zero, and the individual's choice is restricted to c and r. hand, by reducing the probability f the bW st#e, self-pro- Analogously, the model catn accommodate situations tection makes market insurance cheaper and, hence, where self-insurance or self-protection are not possible, increases the tendency to use the market fr insurance. that is, where c=O or r=-O, respectively Since the 1970s the literature on inisurance has con- The individual chooses the levels of market insurance centrated mostly on the problem of moral hazard. For (s*) and self-insurance (c*) where the price of market social policy, however the coprvehensi insrawn aspects insurance equals the shadow price of selfinsurance, and of the theory-wich have been largely neglccted-may they both equal the probability-weighted marginal rate be as or even more relevant. places individuals, households, and firms-not govern- absence of well-functioning markets (for example, prohib- ment-at the center and provides rationales for govern- itively high prices) or the inability of some to use these ment action that are not ad hoc but are based on the instruments even at relatively low prices (for example, 40 DESIGNING SOCIAL POLICY WHEN PEOPLE FACE RISK: A CONCEPTUAL NETWORK BOX 3.3 The Framework in 'Real Ufe" Situations Increased Risk: Heightened Economic Insecurity Proportional Increases in incomes in atl States: Consider the case where only the probability of the indi- Economic Growth vidual being in the bad state (p) goes up. This may charac- Economic growth can be simplistically characterized by a terize the concerns in Latin America and East Asia, where proportionate increase in 4 and I,; hence, the prospec- it is believed that there is now greater econornic insecurity. tive loss rises in the same proportion. Under quite gen- The effect would be to increase the demand for overaLl eral conditions, the demand for market insurance and insurance in absolute terms, but also to change the mix self-protection will increase. This example illustrates between market insurance, self-insurance, and self-protec- that-somewhat counterintuitively-an improvement in tion. following our framework, an increase in p results ia a wealth where incomes in both states go up proportionalLy relative decline in market insurance, no relative change in wiLl result in an increase in the demand for insurance. self-protection, and an increase in self-insurance. This exer- Better income prospects in the good state will have the cise shows the importance of prices: what happens to the same effect. The environment not becoming riskier and demand for market insurance and self-insurance depends on economic growth taking place-an unmistakably posi- whether the market price of insurance adjusts to the tive combination-should result in an increased demnand increase in probability. If it does, then the optimal level of for insurance, often associated with matters becoming market insurance would be lower and self-insurance higher. worse. But if the price does not increase to reflect increases inp, an "excess demand" for market insurance results, and demand increases in Noncrisis Income Levels and Likelihood for self-insurance does not increase as much. of Crises: Globalization Finally, consider the case of "globalization" as it is Increased Expected Income During Crises: Provision commonly stereotyped-when prospective income in of Safety Nets the good state increases (viewed somewhat pessimisti- Suppose that the expected income in the bad state cally, losses become more catastrophic), but so does the increases because of, for example, a guararntee by the gov- probability of the bad state (losses become more fre- ernment that everyone will get a minimum income in the quent)-that is, both C and p increase. Assuming that i bad state which is higher than Io. This reduces the the price of insurance adjusts to changes in prospective demand for matket insurance because the prospective loss probabilities, the outcome for market insurance would decreases, expenditures on self-protection fall for the same be armbiguous because increases in probabilities of reason, but its effect on self-insurance is uncertain because crises weakens the tendency toward market insurance, the fall in self-insurance due to the reduced tendency to but increases in income in good times strengthens it. insure may be offset by an increase in self-insurance The effect on self-protection would be ambiguous, but because it is preferred over market insurance as losses probably positive. This example illustrates the diffi- become less "catastrophic." This example illustrates that culty of predicting how complex phenomena such as individuals will not necessarily reduce self-insurance when globalization affect the demand for insurance. Note such a "safety net" is available, but it is more likely that also that it is more likely-given the findings of Chap- they will reduce self-protection. Thus, the provision of a ter 2-that globalization imiples that p is no higher i public works program will not necessarily reduce precau- than before (or even lower), but losses are larger when tionary saving by individuals, but would lower the effort crises in fact do occur. Viewed this way, globalization is to reduce the probability of being in the bad state by, for essentially the opposite of the "safety niets" example example, reducing investments in health and work skills. given above. 41 SECURING OUR FUTURE IN A GLOBAL ECONOMY poverty or low budgets). The analysis also yields not just a described above is the lack of pooling. This category menu of policies, but also some rules for establishing pri- would include mandatory saving schemes such as orities that are necessary for strategy formulation. Second, employee provident funds in Singapore and the relationships between instruments to deal with risk are Malaysia, and individual severance funds in countries not arbitrary, but are derived from structured analysis, such as Brazil and Colombia. yielding clearer insights into how changes in the economic * "Social protection" can be viewed as policies to aug- environment affect the demand for insurance. Third, the ment self-protection. The failure of markets to facil- approach provides a logical framework for organizing the itate self-protection by individuals or families that is tools of social risk management and their likely effects. optimal provides the rationale for government to intervene. The feature that distinguishes these inter- Clearer Rationale for Government Action ventions from the above two sets of policies is that With an approach that is individual-centered, the need for the aim of social protection policies would be to government arises only where markets fail and social pol- reduce the probability of occurrence of the loss, and icy formulation is based on minimalistic and not ad hoc not simply insure against it. Policies to facilitate the principles. The role of government here-driven by effi- acquisition of human capital (better health, educa- ciency concerns in an environment of risk-is to augment tion, and training) may constitute the core of social markets; that is, to facilitate insurance and self-protection protection. by providing instruments if markets for them do not exist (for example, in the case of unemployment insurance), or Useful Insights through interventions to improve the quality of instru- The framework yields useful insights into questions central ments if individuals are using inferior modes of insurance to determining the scope and design of government policy. (for example, savings in the form of one or two assets Three sets of implications are especially important. instead of a diversified portfolio). Following this line of Welfare is higher when more and better options for reasoning: insurance are available to individuals. As discussed * "Social insurance" can be viewed as a policy to aug- above, the availability of all three "insurance" instruments ment market insurance. Failure of markets to effi- (market insurance, self-insurance, and self-prctection) will ciently insure because some risks are uninsurable or improve welfare over a situation where one or more instru- cannot be diversified, for example, or because moral ments are not available. For example, making available hazard problems are insurmountable for private income support programs for the unemployed is likely to insurers, creates the rationale for social insurance be welfare-improving even when there are efficiency losses policies. Government actions that help individuals (though the magnitude of such losses can be reduced using and families deal better with risk by facilitating adequate instruments-see below). Making market insur- transfers from good states to bad through risk-pool- ance available would lower self-insurance, but would still ing would be classified as social insurance. This result in welfare improvements. would include income-support programs for the Moral hazard may not be an insurmountable problem unemployed (such as unemployment insurance) and if social insurance mimics the market as much as possi- disability insurance. ble. The introduction of market insurance is usually * Mandated savings schemes are policies to augment thought to lower self-protection and raise the probability self-insurance. The failure of markets to provide of occurrence of the bad state ("moral hazard"), but our "good" instruments for self-insurance is one rationale framework and common sense indicate that much can be for governments to intervene. Moral hazard prob- done to limit this adverse side effect. For example, unem- lems, such as the failure to save enough for retire- ployment insurance that successfully discriminates among ment in anticipation of a government bailout of the workers by their risk factors (for example, using informa- old-age poor, provide another justification for com- tion on employment history, skill, or sector of occupation pulsory saving. Again, the feature that distinguishes to set insurance premiums) can lower this negative rela- these policies from "social insurance" of the type tionship between market insurance and self-protection, 42 DESIGNING SOCIAL POLICY WHEN PEOPLE FACE RISK: A CONCEPTUAL NETWORK and even reverse it under certain circumstances. Therefore, A Powerful Toolfor Organization the appropriate policy question may not be whether to pro- The framework described above also helps in obtaining a vide unemployment insurance-especially as governments structured view of government policies and programs. The implement reforms that make these risks less frequent- policies and programs discussed in Chapters 4 to 7 should but how to best design it and to determine how govern- be viewed as government-sponsored actions to assist indi- ments can most effectively develop the capacity to imple- viduals and families attain insurance that is as comprehen- ment it. sive as possible under the circumstances that exist in LAC Financial market strengthening should be a central countries. Table 3.1 shows how some of these policies can component of social policy, because it can augment self- be classified according to whether they help individuals insurance, market insurance, and self-protection. Finan- attain more efficient insurance (through pooling), self- cial sector strengthening is one of the most important- insurance, or self-protection. but relatively underemphasized-policies for balanced, market-augmenting social risk management. There are Conclusion four reasons. This chapter proposes a relatively simple approach to the First, financial markets facilitate risk-sharing. In well- problem of risk, both in terms of individual decisionmak- developed financial markets, individuals and firms can buy ing and the possible role of government. The approach is and sell assets with different risk profiles, diversifying their quite general in that it includes the three major options sources of income, and thus reducing their exposure to available to individuals for dealing with risk: purchasing adverse shocks affecting their particular industry or firm. market insurance, self-insuring, and taking steps to lower Financial markets also provide the most efficient channel to the probability of incurring losses (self-protecting). The promptly redirect resources toward those firms and sectors role of government policy arises when some markets are temporarily hit by adverse disturbances, easing their missing and individuals cannot reach optimal levels of impact on income, employment, and welfare. insurance and self-protection. The government can aug- Second, self-insurance involves precautionary saving. ment individual or household efforts by providing market- Without a strong financial sector, the poor may end up sav- type insurance where markets fail (for example, unemploy- ing through "bad" instruments such as cattle and land, ment insurance), by facilitating individual insurance which are highly illiquid and the prices of which may fall efforts through more efficient forms of self-insurance (for sharply if the bad state of the world ("crisis") occurs.5 example, financial sector development and regulation), or Financial sector strengthening can encourage the use of by assisting or subsidizing self-protection (for example, "good" instruments by savers; this is especially crucial public education and health services). where social insurance mechanisms such as unemployment Using this approach, the chapter traced the implica- benefits are difficult to establish. tions of changes in the environment, such as increased risk Third, financial sector strengthening can result in low- or increased wealth, on the demand for market insurance, ering the probability of a crisis occurring, thus augment- self-insurance, and self-protection. Combined with the ing self-protection efforts by individuals and families. In possibility that markets are missing or do not operate effi- the countries of East Asia where the financial sector weak- ciently, these findings suggest how the demand for social nesses were a primary cause of the crises in the 1990s, this insurance and social protection may arise when such self-protection augmenting role of financial sector changes take place as countries grow or face more or less strengthening is especially important.6 risky external environments. Some of the findings were Fourth, financial sector strengthening will help create expected. Others run counter to widely held views. Three (more efficient) markets for insurance against catastrophic of these findings deserve mention. losses such as those due to poor health or natural disasters. First, the demand for social insurance can increase even Thus, private financial markets can provide life insurance, when the environment becomes less risky and countries disability insurance instruments, and insurance against become more prosperous. This finding is surprising when natural disasters, and can even contribute to insuring market- or government-provided insurance is analyzed in against macroeconomic crises. isolation, but is a natural outcome of analysis using a more 43 SECURING OUR FUTUlRE IN A GLOBAL ECONOMY TAOLE 3.1 Government Policies and Their Effect on Individual Comprehensive Insurance MARKET INSURANCE SELF-INSURANCE SELF-PROTECTION REDUCING REDUCING MICRO RISK AGGREGATE RISK Economywide Risks Stable macro policies Fiscal stabilization funds ' Foreign reserve holdings ' Financial sector reform N N V Deposit insurance v Risk of Becoming Unemployed Unemplovment insurance N Mandated severance N Individual severance funds v Public works programs al Training programs N N Risk of Becoming Poor Cash transfers N Conditional cash tranisfers ' Education reform v Health insturance N N' Financial sector reform V a. Policies that augment seJf-protection for individuals may be setf-insurance or market insurance ar the cosotry level. For example, fiscal stabilization funds ace self-insuransce (because they transfer resources from good states tO bad} for countries, though :hey qualify as self-protection augmentation here (because they reduce aggregate risk for individuals). Access to loteroamional Monetary Fund credit during bad times is market insurance for countries (international risk-pooling), bur is again self-protection augmentation at the indi- ciduat level. b. Although in theory training programs for the unemploved involve an element of self-protection, this elerment appears modest according to the available evidence, so that these programs operate mainly as insurance mechanisms. c. Examples include Boasa Earsa in Brazil sad Progpesa in Mexico. comprehensive (and more realistic) framework where indi- 2. See Hlolzmann and Jorgensen (1999) for an excellent effort to viduals self-insure and self-protect. reduce this confusion. Second, ideally social policy should aim to facilitate all 3. The framework also lends itself to analysis of risk reduction three types of actions that individuals take when confronted policies at the multilateral level, and the possible role of international agencies such as the World Bank and International Monetary Fund. with risk. This finding should help reassess the pros and cons 4. The presence of moral hazard can prevent private insurance of policies such as income support programs for the unem- markets from emerging for some risks, such as business failures or ployed (and among them, unemployment insurance) in loss of employment. Ehrlich and Becker (1972) reason that moral developing countries, focusing the debate on their likely effi- hazard is not inevitable, because in one aspect market insurance and ciency cost and the capacity of governments to contain it. self-protection are complements-increased self-protection increases Third, the role of policies in facilitating precautionary the marginal product of market insurance. That is, if self-protection saving the financiale asses (uch as fianciatl nrectrustrengthy or a lowered probability of the bad state is rewarded by market insur- saving in financial assets (such as financial sector strength- ance (in the form of lower premiums), market insurance and self- ening) has been underemphasized as a social policy instru- protection can indeed become complements, and moral hazard could ment. This finding is a natural outcome of an approach be eliminated. that begins with the individual and derives the problem of 5. It also follows that illiquidity of savers' assets hampers the real- government as a residual, but can easily be missed by location of financial resources toward sound firms in distress in times analyses where this order is reversed, of crisis, which augments the disruptive effects of shocks. 6. The combination of weak interlinkages wirh international cap- ital markets and lack of depth in Latin America's domestic capital Notes markets represent a source of adverse shocks to the region (in our 1. This chapter is based on Gill and Ilahi (2000), a background framework, a higher probability of the bad state, p), and an amplifi- paper commissioned for this report. cation mechanism for other shocks. 44 CHAPTER 4 Macroeconomic Volatility in Latin America and the Caribbean: Causes and Remedies Sources of Aggregate Volatility in LAC A GGREGATE VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN REFLECTS TWO MAIN FORCES: real and financial external shocks and macroeconomic policy instability.' External Shocks LAC is subject to large external disturbances from world goods and financial markets. These can be broadly classified as volatility in the terms of trade and in international capital flows. Figure 4.1 graphs the standard deviation of the rate of growth of the terms of trade across world regions over the last four decades. The figure shows that over the last two decades, LAC suffered terms of trade disturbances that were much greater than those affecting industrial economies and the East Asian miracle countries, and on par with those experienced by South Asia and the Middle East and North Africa. A key factor behind the large terms of trade variability is the high share of a few primary commodities- such as oil (Colombia, Ecuador, Mexico, Trinidad and Tobago, and Venezuela) and metals (Bolivia and Chile)- in the total exports of many of the region's economies. World commodity prices are highly volatile, and this volatility translates into large terms of trade fluctuations for commodity-exporting countries. Figure 4.2 shows the share of the four most important commodities2 in the total exports of selected LAC countries in 1995 and 1999 (or the latest available year). Export concentration This factor can be taken into consideration by looking at remains high in a number of countries, although a few- the volatility of terms of trade shocks, a concept that reflects notably Mexico-have succeeded in reducing it over the both the changes in the terms of trade and the degree of open- last decade. ness of the economy.3 Figure 4.3 offers a comparative per- Terms of trade volatility was particularly high during the spective across regions and decades on the volatility of terms 1970s (largely reflecting the first oil crisis), and declined of trade shocks. The pattern that emerges is similar to that in somewhat in the 1980s and more so in the 1990s, both in Figure 4.1, although Latin America now ranks higher than the LAC region and other world regions. The economic South Asia due to greater openness to trade. impact of terms of trade fluctuations, however, is determined Table 4.1 shows the standard deviation of terms of trade not only by their magnitude, but also by the degree of open- shocks for the major LAC economies over the last four ness to international trade of the economies. Like other parts decades.4 The table shows that this magnitude is generally of the world, LAC has considerably increased its openness to higher in smaller commodity-exporting economies (for global trade over time, and this trend-unless matched by a example, the Dominican Republic, Jamaica, Nicaragua, and parallel increase in diversification of trade-could have Trinidad and Tobago). By this measure, volatility declined raised the exposure of the region's economies to external since the 1970s in a majority of economies, although for trade disturbances. some, such as the Dominican Republic, Jarnaica, Paraguay, 45 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 4.1 Volatility in Terms of Trade Growth (Regional Medians) 20 I8_ - 1960s 1980s 0 18 16_ g 1970s * 1990s. i 14 - 1 2 u 10 8 6- 4 2 0 Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific FIGURE 4.2 Share in Total Exports of Four Most Important Commodities (Selected LAC Countries) Venezuela Uruguay Trinidad and Tobago Peru a E E E l0E1t9 Paraguay 199S Panama Nicaragua Mexico Jamaica Honduras Haiti Guyana Guatemala El Salvador Ecuador Dominican Republic Costa Rica Colombia Chile Brazil _ Bolivia Belize urB Argentina 0 10 20 30 40 50 60 70 80 90 Percent 46 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN: CAUSES AND REMEDIES FIGURE 4.3 Vojatility in Tenrs of Trade Shocks by Decade (Regional Medians) 5.00- 4.50- _ 960s 1980s 40oo - 1970s 1990s 3.50- 3.00- 2.50- 2,00- 1.50- 1.00 0.50- 0.00 i Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific and Trinidad and Tobago, volatility was actually higher in is also subject to external financial shocks, reflected in wide the 1990s than in the 1980s. swings in the volume and cost of foreign capital inflows. Fig- In addition to the real external shocks represented by ure 4.4 shows the annual pattern of private capital inflows to terms of trade changes, LAC, like other developing regions, Latin America as a percent of GDP. Capital inflows rose steadily in the late 1970s to peak at over 5 percent of GDP TABLE 4.1 in the early 1980s, declining sharply to 2 percent in the late Volatility in Terns of Trade Shocks for Selected Latin Amenican 1980s, followed by a strong recovery in the 1990s that was Countries interrupted only by the Tequila Crisis in 1995 and the East Asian Crisis in 1998. 1960s 1970s 1980s 1990s (%) (%) (%) ( These large fluctuations in capital flow volumes are Argentina 1.8 0.7 1.0 0.4 accompanied by similarly large fluctuations in their prices. Belize - - 11.5 2.3 Figure 4.5 shows the spreads on external public borrowing Bolivia 3.0 5.6 3 fou mao df Brazil 0.5 1.4 1.0 0.7 by four major Latin American economies, defined as the dif- Chile 1.9 5.4 2.4 2.5 ferential over world interest rates.5 The spreads display Colombia 0.8 2.1 1.7 0.7 huge fluctuations, with strong surges at times of external Costa Rica 0.9 4.4 4.1 1.8 Dominican Republic 1.8 2.8 3.7 6.5 crises (for example, Mexico's Tequila Crisis during Ecuador 0.4 7.2 3.3 3.7 1994-95), which signaled a generalized withdrawal of El Salvador 1.7 9.5 3.2 1.5 Guatemala 0.7 2.7 1.5 11 financing for LAC economies. Guyana 6.4 9.5 6.2 7.9 The swings in the volume and cost of external financing to Haiti 3.3 2.5 1.1 5.8 LAC reflect a combination of external and domestic factors,6 Honduras 1.3 4.4 2.5 3.1 Jamaica 2.6 3.0 4.8 9.0 which affect both the decisions of nonresidents to supply Mexico 0.3 0.9 2.0 0.9 financing to the domestic economy, and the decisions of resi- Nicaragua 1.5 5.1 18.1 6.2 Panama - - 0.9 0.9 dents (including the public sector) regarding whether to bor- Paraguay 0.8 4.9 2.6 7.1 row or lend abroad. The two critical ingredients are the Peru 1.5 4.1 1.5 1.1 expected return from holding assets domestically, relative to Trinidad and Tobago 3.2 12.0 4.0 7.3 Uruguay 2.5 3.0 2.0 1.2 holding them abroad, and the perceived riskiness of that Venezuela - 3.6 6.8 3.7 return. LAC Medians 1.6 4.1 2.6 2.3 For given risk perceptions, private capital inflows tend to LAC Averages 1.8 4.5 3.9 3.4 move in a direction opposite to OECD interest rates, declin- 47 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 4.4 Median Private Gross Capital Flows in Latin America (Percent of GDP) 7 - 6- 5- 4.54 2- 6- . .. ...... ..:. 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 FIGURE 4.5 Spread of Foreign-Currency-Denominated Sovereign Debt Instruments (bps) (Selected Major Latin American Countries) 3,500 - Argenrina 3,000 - Brazil 2,000 - Venezuela m1,0(0 L,, 4J,S 1,000 50 0_ - --r^ -*""''".... b' <' 0 - O N, 0 11 u ' .' ~ C > C C > > C^. > m~ 0< C> > C> C> C C > > t C> A> C ing at times of high rates (as in the early 1980s), and rising Importantly, however, the flows also reflect developments when interest rates decline (as in the early 1990s). Inflows in the destination economies, because botb risk and return are also react strongly to payments crises in specific countries, affected by domestic economic policies. The variability of such as Mexico, which cause investors to reassess risk and capital flows does not reflect just external shocks, blt is in often leads to a generalized drop in inflows across emerging part governed by forces endogenous to the receiving markets, in what has been judged as evidence of "financial economies.8 With this important caveat in mind, Figure 4.6 contagion.' 7 shows the variability of gross private capital flows, as mea- 48 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN C*AUSES AND REMEDIES sured by their coefficient of variation,9 across world regions observed in industrial economies (7 percent) and the East and time periods. By this measure, volatility of capital flows Asian miracle countries (10 percent), and was surpassed only has risen relative to the 1970s in all world regions, although by Sub-Saharan Africa. It should be noted, however, that in in most of them it peaked in the 1980s and declined in the the second half of the 1990s monetary volatility has contin- 1990s. In all three decades, LAC ranks above the industrial ued on a declining trend in Latin America. countries and the East Asian miracle countries in terms of Fiscal policy is also volatile in LAC. Figure 4.8 shows the capital flow volatility, although the difference between LAC volatility of real public consumption growth (as measured by and these regions has narrowed in the 1990s. its standard deviation) across decades and world regions.11 As in previous cases, LAC displays higher volatility than indus- Macroeconomic Policy Volatility trial countries and the East Asian miracle countries-but less External factors are not the only cause of volatility in Latin than most other developing regions. America. Macroeconomic policies must share some of the Fiscal volatility is also related to monetary instability, blame. Policy volatility partly reflects mistakes by policy- because inflationary responses to unsustainable fiscal imbal- makers, but to a greater extent it is the result of large exter- ances has traditionally been one of the primary causes of nal shocks in the presence of weak insurance and financial volatile monetary aggregates in the developing world- markets and policy institutions, which constrain room to including Latin America until the early 1990s. Figure 4.9 maneuver in macroeconomic management."° plots the volatility of money against that of public consump- Monetary policy volatility has been consistently high in tion-both policy variables-for a large sample of countries. 12 LAC. Over the last two decades, the region has stood out for A clear positive association between both variables emerges. the recurrence of extreme inflation episodes driven by mone- As noted earlier, however, macroeconomic policy volatility tary financing of unsustainable fiscal imbalances. Since the also reflects the effect of external shocks hitting domestic 1970s, as Figure 4.7 shows, the standard deviation of base economies. This is especially so in developing countries money growth has been higher in LAC than in most other where public scctors are heavily depcndent on commodity world regions. It peaked in the 1980s at over 20 percent revenues, as in many LAC economies. Terms of trade distur- annually, and declined in the 1990s to just under 16 percent. bances have an immediate impact on public revenues and are The latter figure, which reflects the extreme inflation clearly reflected in fiscal aggregates. This can be seen in Fig- episodes of the early part of the decade in a few countries- ure 4.10, which plots fiscal volatility against terms of trade notably Argentina, Brazil, and Peru-is far above the levels volatility for over 100 countries." Terms of trade fluctuations FIGURE 4.6 Coefficient of Variation of Gross Private Capital Flows (Percent of GDP, Regional Medians) 0.70- -- 0.60 - F-I 197 Os 060_ * 1980s 0.50 - 1990S 0.40- 0.30- O 20- 0.10 0.00 _ Industrial Economies East Asia and Latin Amcrica and Middle East and SoLoth Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific 49 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 4.7 Volatility of Reserve Money Growth (Regional Medians) 25 - 1960s 1980s 20- 1970s 1990s 15- 10 5 0 Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific FIGURE 4.8 Volatility of Public Consumption Growth (Regional Medians) 0.14 - 0.12 - 1960s 1980s 1970s 1990s* 0.1 i 0.08 ..0.06- 0.04- 0.02- 0 Industrial Economies East Asia and Latin America and Middle East and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pacific appear to be a major force behind fiscal volatility, accounting and policies that play an instrumental role in absorbing or for a full one-third of the cross-country variation. amplifying shocks. Among these shock absorbers and ampli- fiers, the domestic and world financial markers are perhaps Absorption and Amplification of Shocks: The the most important. Importance of Financial Markets International financial markets allow domestic agents to The magnitude of the impact of major economic shocks dis- sell risky income-generating instruments such as stocks and cussed above on aggregate income and employment in LAC bonds of domestic firms. In this manner, domestic agents can is determined by the functioning of markets, institutions, reduce their exposure to risks associated with income volatil- 50 Fiscal Volatility Log of Standard Deviation of Reserve Money Growth K C o o o .° o o ° q 0 - a c N io+_ _~ ~~~ -* , *K CL~~~~~~~ 3~~~~~~~~~~~~~~- -- p = + 0 \ *0 ~~~~~~~~~~~~~~~~~ *~~~~~~~ o~~~~~~~~~~~~~~~~~~~~~~ _~~~~~~~~~~~~~~~~~~~~ ava *l C~~~~ ~~~ ~~~~~a ~~~~~~*~ SECURING OUR FUTURE IN A GLOBAL ECONOMY ity, diminishing the latter and thereby reducing aggregate Thus, well-functioning domestic financial markets provide volatility. Further, world financial markets also play an impor- the efficient means for economywide self-insurance against tant role by supplying financing to ease adjustment to shocks aggregate risk-efficiency in the sense that it can achieve rel- after they have occurred so that, for example, a temporary atively quick reallocation of financing and avoid unduly long worsening of the terms of trade, such as a fall in the price of or widespread disruptions in production and consumption. oil or copper, does not force the economy into a sharp reces- In most LAC economies, however, weak links with world sion. Instead, such financing allows the shock to be weathered financial markets and poorly functioning or shallow domestic through a temporary increase in the current account deficit. financial markets greatly contribute to amplifying shocks Domestic financial markets also play a key role in the rather than helping absorb them. This dual financial weak- adjustment to shocks, fulfilling the dual purpose of facilitat- ness is at the core of LAC's macroeconomic volatility."5 ing both ex ante risk diversification among domestic agents in the capital market-hence the diversification is limited to Weak Links with World Financial Markets individual, not aggregate, risks"4-and ex post channeling That links to international financial markets are weak follows of resources toward sound firms and sectors hurt by shocks. from LAC's modest volume and large swings in private capi- BOX4,1 Excess Sensiti to Disturbances The Case of Chile The weakness of the financial links of LAG economies to With unhampred access to external financing, Chile world markets makes them overly sensitive to distur- would be able to smooth out tenporary copper price fluc- bances. Here we document the case of Chile, the economic tuations, and the swings in growh rates would be more fbrtunes of wbhich fluctuate widely with world cpper muted than those of copper prices. However, panel (b) prices. shows that ihe opposite happens. The panel compares the Panel (a) in Figure 4.11 plots the spot prie of copper fluctuations il GDP actually observed with those that fiom the London Metal Exchange and Cies quarterly would be dicated fom perfct smoothing (specifically, the GMP growth. The tesemblance between the two is stril- an value of the present value impact of the change in ig, with the only important exception beig the 1990 copper prices, as a share of GDP)."8 It is apparent from the growth slowdown and subsequent recover, which had a fure (from the different scales in the axes, in particular) purely domestic origin. that fluauations in GDP are an order of magnitude larger FiGMRE 4A I ChiWs Excess Senesiivt to Scks (a) Growth and copper price (b) Present value effct of terms of trade shocks -4000 15% 3500 1.0% 3000 +~~~~~~~~~~~~~~~~~~5 -S% / -0.5%¢ 2500 0 3 (::opper Price 1a00 2 -5% _ ~~Groth .. 5 0-10% t = fiii i -1.0% Growth (dev from mean) (-N 9N N N NN Saalm Orowth ftom 15F; opper pices (London Metal Ethange) fiom Datstre 52 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN: CAUSES AND REMEDIES tal flows, and the volatile interest rate spreads on sovereign tioning of domestic financial markets in most LAC countries debt discussed earlier. Even more striking is the fact that makes them part of the economic instability problem rather LAC borrowers tend to face much higher premiums, and than a solution to it. This is true both of banks and other higher return volatility, than private U.S. borrowers of simi- financial institutions. lar rating.'6 All these facts suggest that LAC's integration in Figure 4.13 shows that LAC still lags behind most world world financial markets is still limited, which hampers the regions in terms of banking system development, as measured ability of the region's economies-even the economically by the ratio of credit to the private sector to GDP.'9 While well-integrated ones such as Chile-to smooth the effects of there is a great deal of variation across countries in the region,20 temporary disturbances (see Box 4.1). on the whole the ability of LAC's banking systems to efficiently intermediate financial resources remains rather limited.21 Shallow Domestic Financial Markets Capital markets, in turn, have experienced a rapid expan- Despite the considerable progress made since the 1980s, LAC sion in Latin America over the last decade, but they remain financial markets remain shallow, and financial systems are small and illiquid relative to those in other regions. Figure still weak in many countries in the region."7 The poor func- 4.14 provides a comparative perspective on the size and than would be observed if Chile were able to resort to inter- crisis and experience fast domestic growth despite the large national financial markets to navigate the disturbances. international credit crunch suffered by emerging Figure 4.12 reinforces the conclusion that the fuinda- economnies. This is confirmed in panel (b), which demon- mental problem is one of restricted access to international strates that Chile used a large fraction of the "liquidity" i financial markets. Panel (a) shows that the price of copper given by the high price of copper to offset the decline in and Chile's current account deficit are positively associ- capital inflows, as the current account deficit at "normal" i ated-exactly the reverse of what would be observed under copper prices reached its highest level during that year. smoothing of disturbances. Most important, exactly the opposite occurred during the The 1995 Tequila Crisis appears to be the exception 1998-99 episode, as the price of copper plummeted (eras- that proves the rule because high copper prices gave the ing Chile's liquidity) at the precise time that international i Chilean economy enough 'liquidity" to ride through the financial markets tightened. .FIGURE 4,12 Copper Prices and Chil Cun'ent Account (a) Balance of Payments and Cooper Price (b) Current Account Deficit 15S * 140 8_I lID _ __ _ __ _ _ _ _ __-130 __ __7 110 0~~~~~~~~~~~~~~~~0 .90n 680 -15~~~~~~~~~~~~~~~~7 ~~~~~~~~~~~~~~~-6o -2 -15 19 87 189 1O 9t g19 199 149 1S9 g96 999 9899S 50 1986 1987 1988 1989 1990 1991 1992 1993 1994 199S 1996 1997 1998 199 86 87? 8, 81 9 909, 92 '91 '94 9 6 9? 98 ;99 0- Actual CA. b,IfkkiuNnoCADli aCuamA,coaDelidt - ca ft9pi lqflo, Ca Fie| Soa,Isrs Iritituto National de Elcadfgtica and Banco Centra dt Chile. 53 SECURING (ITR FiTUJRF iN A GLOBAL FCONOMY FIGURE 41 1 Average Private Domestic Credit (Percent of GDP, Regional Medians) 120 - .6o- 19705 1996,, 80 C~~ -60 40- 20 - 0 611li Industrial Economies East Asia and Latin America and Middle Easr and South Asia Sub-Saharan Africa Other East Asia Pacific 7 the Caribbean North Africa and Pac fic FIGURE 4 14 Stock Market Capitalization and Turnover Ratios (Regional Medians in Percent, 1990-98)) s0 - Market Capitalization 70 -Turnover Ratios 60 - , 50 . 40- 30- 20 10 0 ' Industrial Economies East Asia and Pacific 7 Latin America and South Asia Other Easc Asia andi Pacific the Caribbean turnover (measured respectively by market capitalization as end, Brazil possesses a fairly liquid market, but its overall size percent of GDP, and the ratio of value traded to market capi- is modest by international standards. talization) of equity markets across world regions.22 The clear message is that LAC is lagging behind the rest of the world in The Combination Increases the Likelihood of Economic both dimensions. Figure 4.15 shows that there is considerable Crises diversity among the major economies in the region. At one The imperfections of LAC's financial markets severely limit extreme, Chile's market size is at or above industrial-country their ability to diversify risk and reallocate financial resources levels, but its market liquidity is extremely low. At the other at times of distress. This tends to amplify and propagate 54 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN: CAUSES AND REMEDIES FIGURE 4.15 Stock Market Capitalization and Turnover Ratios for Seected LAC Countiles, 199-98, Averages 120 - Market Capitalization 100 *Turnover Ratio 80 C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 8 60 - 40- 20 0 Argentina Brazil Chile Colombia Ecuador Mexico Peru Venezuela adverse disturbances by creating wasteful contractions in sec- ables are measured by their averages over the last three tors most affected by shocks or relatively poorly connected to decades. A negative association between both variables is domestic and international financial pipelines. This propaga- obvious from the figure.25 The relationship, however, appears tion effect is particularly evident in the case of the banking nonlinear, as indicated by the solid line-of-best-fit.26 Thus, system. As adverse shocks put domestic firms in distress, lead- increased size of financial systems is associated with reduced ing some of them to bankruptcy, the credit portfolio of the economic volatility, but the association becomes less strong as banking system deteriorates, lowering the ability and willing- the financial system becomes very large.27 ness of banks to bear risk and channel financial resources effi- This implies a qualification regarding the stability- ciently. Some borrowers may be completely excluded from the enhancing role of financial markets-the danger posed by credit market, exacerbating the magnitude of the downturn. excessive indebtedness. As financial systems expand, so does When the banking system's balance sheet is already weak, this leverage, and with it the vulnerability of the financial system sequence of events can bring banks to the verge of financial to shocks also increases. Rapid expansion of financial systems, collapse, and take sound borrowers along with them. particularly if inadequately regulated and supervised, can also Weak capital markets also amplify the effects of shocks. In contribute to economic volatility, a factor that played a cru- a manner similar to the credit rationing effect of weak bank- ing systems just described, they result in what may be termed Stock Market Illiquidity "1equity rationing"; that iS, the inability of firms to raise funds "ntheequity ro "arket wis,the inabit pofifirtoise higd (Regression Coefficient of Absolute Price Changes on Trade in the equity market without incurring prohibitively high Volume) costs. Moreover, thin markets also result in large fluctuations in equity prices, as shown by Figure 4.16, which compares 1.0 the sensitivity of equity prices to trading volume in Chile23- 0.9 _ 0.8 the LAC economy with the largest stock market-and three 0.7 - industrial economies. The result is that firms are unable to 0.6 - diversify their risks well through equity markets.24 0.5 - The association between underdeveloped financial mar- 0.4- kets and economic instability is clearly brought out by inter- 0.3 - national evidence. As an illustration, Figure 4.17 plots the 0.2 - stock of private sector credit as a ratio to GDP against GDP 0.1 - growth volatility for a large number of countries. The vari- Chile Australia Norway United States 55 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGURE 4.17 GOP Volatiliy and Credit Depth 180 - 160 - 140 - U \~~~~ P, 120- o 100- 80 U ° 80- ; 60- WA W~~~~U 20 , .ji I _ __ 0 U~~~~ea WCfo 5% 10% 15% 20% 25% GDP Growth Volatility cial role in the East Asian crisis of 1997-98. It is ironic that enced a severe credit crunch following the 1995 Tequila Cri- several LAC economies have suffered at both ends of the spec- sis. Loans-especially new loans-collapsed early in the cri- trum of financial development: chronic financial repression sis, especially as the peso went into free fall, dragging down and underdevelopment first, followed by accelerated expan- the already weak balance sheets of Mexican banks. Thie severe sion and collapse of the banking system later. Deficient bank credit crunch amplified the magnitude of the crisis, and the monitoring and supervision played a major role in these collapse of the banking system imposed massive costs on the boom-bust episodes. economy and the public sector accounts. A second important qualification is that the causation may In Argentina, in contrast, the amplification developed from run the other way too: high economic volatility itself tends to the other side of banks' balance sheets. Figure 4. 18 shows that hamper financial market development. In a highly volatile in Argentina the major force behind the credit crunch was the environment, firms may not be willing to undertake the risks run on bank deposits, driven by depositors' fears that tight associated with extensive borrowing, nor may households external conditions would eventually result in the collapse of wish to save in financial assets (at least those available domes- the system of convertibility between Argentina's peso and the tically). Without appropriate policy action, the economy may U.S. dollar. The figures also illustrate how the financial tur- get stuck in a self-perpetuating vicious circle characterized by moil in world markets during 1998-99 resulted in new credit weak financial markets that amplify volatility, which in turn slowdowns in Argentina, Brazil, and Mexico. prevents further financial market development. Beyond their shock magnification effect, however, In the LAC context, the interplay between weak links domestic and external financial weaknesses are also sources with international financial markets and underdeveloped of instability themselves, because they raise the likelihood domestic financial systems may be key to the region's aggre- that as-yet unrealized disturbances will have a major dis- gate volatility (see Box 4.2). A closer look at recent crises in ruptive effect on the economy, triggering precautionary the region may help illustrate this. Figures 4.18, 4.19, and responses by the government or the private sector that 4.20 show credit crunches in three major countries in the anticipate the crisis. An example of this is the case in which region that have followed episodes of external distress. Most policymakers foresee a tightening of external financing, striking is the case of Mexico (Figure 4.20), which experi- which leads them to contract monetary and/or fiscal policy, 56 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN: CAUSES AND REMEDIES FIGURE 4.18 Credit Crunches in Argentina 80 - 70 _ | Credit to the Private Sector 60-| Deposits - - - - - - - - - 60 - C 50 c30 - 20- 10 0 _ l l ~ ~ I I I I I Janl994 Oct1994 July1995 ApriL1996 Janl997 Oct1997 July1998 April1999 FIGURE 4.19 Credit Crunches in Brazil 250 - 200 - IJ __~~------------------------__ C 150 - -5 0 100 50 - - | _ Credit to the Private Sector -Deposits O - I I I I I I I I I I l I I I I I I I I I I I ~O \7~ ON ON ~N '~ N ~N '0 N \0 '.0 N cNo 00 00 0 0 ON N ON N O ON f' g < -$ t g wR U $ D g $ t driving the economy into recession ahead of the feared system considerably raises the likelihood of success of such external tightening. Likewise, the private sector often speculative attacks and their economic cost. reacts to an anticipated tightening of the external financial bottleneck by running against domestic assets, thus dri- Other Amplification Mechanisms ving down asset prices and forcing a tightening of macro- Beyond the financial system, other policy and institutional economic policies-a scenario similar to the Argentine factors also play an important role in magnifying or con- episode of 1995. Needless to say, a weak domestic financial taining the economic impact of shocks. 57 SECURING OLR FUTURE IN A GLOBAL ECONOMY FIGURE 4.20 Credit Cnunches in Mexico 700 _ ____ G~~~~~~~~~~~~~~~~~~~~~~Cedit to the Private Setor c "' ^vs s s s s s s _ ~~~~~~~~~~~~~~~~~~Deposits 500 0 o 400 o 300 - 7: 200- 100 O- ~ ~ ~ ~ ~ ~ ~~~~~~ - I. I. I I III .S g k R t @ @ @ S' N 'O 'O Fiscal policy has traditionally been assigned an "auto- ible rates allow monetary independence and are best matic stabilizer" function, which consists of offset- for protecting the economy from real disturbances.30 ting shocks by expanding aggregate demand in the In recent years, LAC has witnessed a shift toward face of contractionary disturbances and, conversely, both ends of the exchange regime spectrum: rigid in the case of expansionary disturbances. In LAC (and pegs (for example, currency boards as in Argentina, in much of the developing world), however, fiscal and proposals for outright dollarization in Argentina policy is often procyclical, adding to the expansion and Ecuador), and freely floating arrangements during booms and to the contraction during reces- (Brazil, Chile, Colombia, Mexico, and Peru). The sions.8 Thus, a policy risk is inadvertently added to experience of developing countries, notably in LAC economic risk, amplifying the effects of economic and East Asia, over the last decade or so has added shocks rather than offsetting them. To some extent some important qualifications to this conventional this again reflects the operation of financing con- wisdom. First, the degree of monetary independence straints, since at times of adverse shocks govern- allowed by flexible exchange rates may be limited in ments face sharp reductions in their access to exter- practice if firms and banks hold large unhedged lia- nal financing or large increases in its cost.29 bilities in foreign currency, because under such con- Procyclicality also reflects the failure of governments ditions exchange rate fluctuations can have large to provide for bad times by increasing their saving effects on firms and banks' net worth, as in the East during good times, when revenues are high. A stark Asia crisis.31 Second, the ability of flexible rates to example of this failure has been the frequent mis- ease the adjustment to real disturbances depends on management of resource booms in countries whose the credibility attached by the private sector to mon- pLublic sector is heavily dependent on natural etary policy and on the extent of inflationary inertia resource revenues. due, for example, to formal or informal indexation. * Exchange rate and monetary policy also shape the econ- Lack of credibility and widespread indexation may omy's ability to weather shocks. The conventional erode much of the real effect of nominal devaluation prescription is that pegged exclhange rates provide through additional inflation.32 Third, hard pegs may the best insulation against financial shocks, and flex- enhance financial stability and policy credibility. 58 MACROECONOMIC VOLATILITY IN LATIN AMERICA AND THE CARIBBEAN: CAUSES AND REMEDIES BOX 4.2 IWak Funancial Markets and VolatiOly: A Framework The interplay between weak international financial links financing (flat at the discount factor 1 until the maxi- and underdeveloped domestic financial systems may be at mum amount AtAt is reached, at which point supply the root of much of LAC's aggregate volatility. Here we becomes vertical), is enough to meet the needs of dis- present a stylized framework outlining the key mecha- tressed firms (the solid line that becomes vertical at the nism, drawing from work by Caballero and Krishna- economy's total number of projects, set at unity). Thus, murty (1999). equilibrium is reached at L = 1, so that the domestic cost T- e Basic Setup of funds equals the world interest rate, and distressed The Xasic Setup firms pledge only a fraction of their assets to the "inter- i Consider a schematic timeline such that at date 0, which d h t i a corresponds to "normal" times, investment decisions are made and agents plan toward the "bright future" of date eral In panel (b), however, international collateral falls 2. Much of this planning has to do with anticipating and short of the needs of distressed firms. The result is a fre sale of domestic assets, with the cost of domestic funds preventing a crisis that can happen in the near future at period 1. Weak international links imply that the coun- jumping to L > 1, and only a fraction AtAt of all projects try may have a hard time persuading foreign financiers getting financed. that they will share the gains in a relatively bright future (period 2) if they help to avert the period 1 crisis. Weak International Link* Shallow Domestic In this context, a crisis is a situation in which the Markets, and Excess Vulnerability economy needs substantial external funds (to repay debt If, unlike in Figure 4.21, domestic financial markets are or undertake new investments) but does not have suffi- also imperfect-in the sense that distressed firms can- cient international collateral to obtain them. To make not fully pledge their assets to the domestic intermedi- this insufficiency of collateral clear, assume that nontrad- aries holding the internationally-acceptable collat- able date 2 assets (for example, buildings that would be eral-the latter's incentive to hoard and supply completed at date 2), denoted An, are of no interest to international liquidity is lessened. In the model, the foreigners-they cannot be used as collateral abroad. domestic price of this liquidity L falls. This situation is I However, they can be used as collateral to borrow domes- shown in Figure 4.22 by assuming that only a fraction tically, at a discount factor L, a maximum of APL- This An < 1 of domestic collateral can be pledged. This shifts collateral is held by the "distressed>' firms.. collateral is held by the "distressed" firms. down the effective demand schedule (solid line) in panel In turn, other domestic firms or individuals hold (a), leading to a decline in L relative to Figure 4.21. internationally acceptable collateral, denoted At, which Frictions ins dornestic financial markets now distort the includes items like the output of firms in the tradable return to holding international collateral, and as a result goods sector at date 2, plus foreign currency assets, plus less will be held by the intermediaries supplying inter- perhaps some domestic assets attractive to foreigners, such as telecom firms. Assuming that the international national liquidity discount factor equals 1-or the interest rate equals 0- This is shown in panel (b) of Figure 4.22 as an inward the most the country can borrow abroad at date 1. is AtAt, shift in the supply of international liquidity. The reduced where At < 1, because of imperfect access to the world supply means that the economy will experience more fre- capital market. quent fire sales and more severe distress in the face of international disturbances. The economy becomes too Weak International Links and Fire Sales vulnerable to external shocks due to the undervaluation Figure 4.21 shows the equilibrium in the economy's of international liquidity created by domestic financial financial market. In panel (a), the supply of international market imperfections. 59 SECURING OtUR FUTUJRE IN A GLOBAL ECONOMY :m~~~~~~~~~~ HdM! N1 iVMW2200-e SOR16..l 1...... ; Ads i0ski i g t a 33 gi gS0f 3fX W .M9Ig.8llt.gitsilg}( l}s gWe 0 sg } Wg Ws gS atx W W 10 X 0 W 0 tW~ ~ ~ ~~~~~~~~~~~~~~i r - U 0-f jl,3s,§Se' :t. n -r X .n.. . u. - * ' .Hi ~W * c>5ylsae¢x .. -3 }:: - I3i NM1 ;'''';'W'W; W;W WW ; }W I;WWWWWWW4W W; SsSO(RW EMPg < ' ^;4--0,U'-<-55 2.3 2.7 0.6 8.0 2.7 2.7 2.3 11.3 Total 1.1 1.5 2.8 6.5 0.9 1.4 3.0 13.0 NA= Insufficient data. Source: Arango and Maloney (2000). Maximum Likeihood Estimations following Kalbfieisch and Lawless (1985). Consistent with more traditional views of the informal that one commonly cited "safety net," the informal sector sector, the likelihood of becoming an informal sector itself, is less comprehensive than often thought. worker is found to be two to three times as high as the like- But without evidence that the differences in Table 5.4 lihood of entering formal employment after being unem- are indicative of a higher income variance of any particular ployed. However, what is striking is that there are also large flows from informality into unemployment. In par- BOX 51 , . I~~~nfofmal Self-Employmerk: Precarious VVokers or Voluntary ticular, in Argentina, it is twice as likely for primary- school-educated informal sector workers to become unem- Enrepreneurs? ployed as their formal sector counterparts. Only some of Do the findings that the informal often find themselves this effect disappears when we compensate for the fact that unemployed provide additional evidence that informal fewer people leave formal employment relative to informal work is especially precarious? Perhaps not. As Chapter sector jobs. This suggests that what is traditionally consid- 2 suggested, there is evidence that up to 70 percent of ered the reserve sector for the formal sector unemployed the people working in'the informal sector may be itself generates many unemployed. In fact, tabulations of doing so voluntarily. Informal self-employment has the sector of origin of the unemployed sector in Argentina behaved procyclically for long periods in Argentina, and Mexico suggest that only 36 percent and 25 percent, Chile, and Mexico, and enterprise surveys suggest that respectively, of those currently unemployed (who previ- less than one-third of business owners in Argentina and ously held jobs) were in the formal sector. The remainder Mexico entered the sector involuntarily. were from either informal self-employment or informal Levenson and Maloney (1998) argue further that salaried employment. small firms everywhere show high rates of mortality In addition, the informal salaried sector is often thought and higher income variance. The fact that they may to be a "supercompetitive" sector where a laid-off worker either be able to avoid taxation and regulation (Loayza can instantaneously find another job. And in fact, in Mex- 1998), or that they generally do not benefit from for- ico and Argentina, those entering unemployment from the mal contracting, risk-pooling, and other institutions informal sector do spend between 22 and 35 percent less that larger formal sector firms can avail of, means that time, respectively, in unemployment than formal sector small firms will tend to be disproportionately informal. workers (see Box 5.1). But it is also true that the income Combining these two elements leads to a finding that variance among self-employed workers is significantly is common in the literature: informal firms show very greater than in the formal sector (see Table 5.4). However, high rates of failure and income volatility, and informal given the substantially higher incidence of unemployment, workers show high rates of turnover. But under this the difference in durations does not seem so large that we interpretation, this is voluntarily accepted risk and can conclude that informal workers are somehow of less does not reflect a "precarious" business environment in concern than formal workers from the point of view of the sense the term is frequently-used. designing an income security program. It also suggests 75 SECURING OUR S'UL1TURE IN A GLOBAL ECONOMY TABLE 5.4 Income Variance of Formal Salaried vs. Self-Employed Workers (Theil Index) ARGENTINA BOLIVIA CHILE COLOMBIA URUGUAY VENE:ZUELA Formal Salaried .295 .430 .411 .433 .350 .2 64 Seif-Employment .484 .819 .867 .972 .499 i70 S,(r,-\Vodon, N Maloney, and Barensrein (2O0U). individual in one sector versus the other, this cannot be Salvador (Conning, Olinto, and Trigueros 2000), for the Llsed as evidence of "precariotLsness"; it could as likely be 1995 Mexican Tequila Crisis (Cunningham and Maloney evidence of greater heterogeneity in the informal relative to 2000), and for various boom and bust episodes in B3razil in the formal sector. This should not imply, therefore, that the 1980s and 1990s (Neri and Thomas 2000).' those in informal activities should be the focus of an In interpreting the results of these three studies, readers employment security program. are advised to keep in mind an important shortcoming. All three studies use incomes, not consumption expenditures, Who Becomes Unemployed andfor How Long? to analyze the effects of aggregate shocks. For a study of Tables 5.3 and 5.5 shed some light on who becomes household responses to shocks, this is no small shortcom- unemployed and for how long. In both Argentina and ing. (See Box 5.2 for a discussion of methodologies used in Mexico, people with more schooling tend to become these studies.) If shocks are not perfectly foreseen, or if cap- unemployed less frequently, but remain unemployed ital markets are imperfect, a (constrained) consurnption- longer. This is consistent with more firm specific human smoothing household will respond by adjusting consump- capital leading to both lower separation rates and longer tion levels. But even an unforeseen negative shock ought to job searches. No clear pattern by age is shared across coun- lead to a (less than proportional) decline in consumnption, tries. In Mexico, older workers are more likely to become and an unexpected positive shock should entail a (less than unemployed, and for longer periods. In Argentina, the proportional) increase in consumption.6 Since welfare ulti- young are far more likely to become unemployed and, mately derives from consumption, rather than income, this among the less skilled, for longer periods. In sum, a blan- implies that income variations overstate welfare variations ket statement about who especially needs income protec- in all cases (and in both directions). The magnitude of the tion cannot be made easily. overstatement will, however, decrease with the degree of imperfection in capital markets and with how binding any Household Responses to Income Shocks: subsistence constraint is. Findings of Panel Studies The studies generated a wealth of detailed, country-spe- This section discusses the main results of the impact of cific information that should be valuable to those with a shocks on households and their coping strategies, obtained special interest in these countries, but there are also results for the agricultural production crisis of 1997 in rural El of more general interest. We focus on what we call the four TABLE 5.5 Unemployment Duration, in Years MEXICO ARGENTINA AGE HIGH SCHOOL LOW SCHOOL HIGH SCHOOL LOW SCHOOL <22 NA. 0.10 0.73 0.58 22-35 (.19 0.12 1.20 0.49 36-55 ().17 0.15 0.80 0.51 >55 0.20 0.14 0.64 0.39 Total 0.18 0.14 0.82 0.47 NA = Not applicable. Scarce: Arango and Maloney 120). Maximam Likelihood Estimarions following Kalbtleish a-d Lawless (198S). 76 THE RESPONSE OF LAC HOUSEHOLDS TO ECONOMIC SCHOCKS BOX 5.2 Dab Sets and Metdolges Used This box describes the data sets and empirical approaches interviewed quarterly, and stay in the panel for five con- used by the case studies commissioned for this report and secutive quarters. sumrnarized in this chapter. To account for the possibility that "vulnerability" might depend on the initial rank of the household in the Rural El Salvador welfare distribution, results are presented for unweighted Conning, Olinto, and Trigueros (2000) investigate the and weighted-where different weights are attached to effects of a downturn in agricultural activity in El Sal- income levels depending on the place in the distribution. vador in 1997, using a panel of 489 rural households sur- Three quantile regressions (for the first quintile, the veyed in 1995 and 1997 by the Universidad Cen- median, and the fourth quintile) are estimated, with pro- troamericana. The authors use these data to quantify the portional changes in household income as the dependent incidence of the impact by income groups, to disaggre- variable. A set of demographic, educational, and occupa- gate it by occupational category, and to investigate which tional dummy variables, three of which capture house- household characteristics were associated with differences hold behavior in response to the shock (head enters labor in the mragnitude of the income shock. The existence of force, spouse enters labor force, and child enters labor 1; such characteristics would indicate that they play a role force), are used as independent variables. in risk management by the household, either ex ante (as insurance or self-protection), or ex post (as coping). Metropolitan Brazil The authors deploy two complementary techniques. Neri and Thomas (2000) identify the groups most First, poverty indexes are computed by subgroup, using affected by aggregate economic fluctuations in Brazil, one of the seven occupational categories the household and then investigate the nature of the household fits into: self-employed only, nonagricultural wage only, responses. They restrict their inquiry to urban areas, but agricultural wage only, agricultural wage and nonagri- span a longer period: from 1982 to 1999. The data set is cultural wage, nonagricultural wage and self-employ- drawn from the Monthly Employment Survey (PME), meat, agricultural wage and self-employment, and all carried out by the Brazilian Statistical Institute (IBGE) 1 three occupations. Second, a model of income determina- every month for Brazil's six largest metropolitan areas. tion, including household-specific and time-variant vari- The same households are interviewed for four consecutive i ables, household-specific and time-invariant variables, months, then excluded from the sample for eight household-specific and time-invariant unobserved effects, months, and then revisited for a further four consecutive and household-specific and time-variant shocks, was esti- months. This allows the authors to construct a series of mated to generate random-effects estimators. A specifica- panels, using four-month averages of household incomes tion test suggested that the hypothesis that the estimates per capita, to investigate the impact on households of were the same could not be rejected. The authors focus on seven periods of macroeconomic volatility: three booms the analysis of the random effects estirnates. and four recessions. The distribution used is a distribu- tion of household per capita income, per household head. Metropolitan Mexico Neri and Thomas also find that using current income Cunningham and Maloney (2000) identify the groups to rank the distribution would bias the results by exag- worst affected by the 1995 Tequila Crisis in Mexico, and gerating true mobility. They too use a proxy for perma- i study the results of labor force participation changes nent income, which is the value of the predicted income around the shock, as coping strategies. The data set used for each household head in a Mincerian (earnings) equa- is a panel of 21,262 households in 16 metropolitan areas tion with age, experience, gender, marital status, and in Mexico, from 1994 to 1997, drawn from the National employment sector. Average proportional income Urban Employment Survey (ENEU). Households are (continnes on next page) 77 SECURING OUR FUTURE IN A GLOBAL ECONOMY BOX 512 |.Continued changes for each quintile in each episode are calculated, canges in occupational status Finally, a diffrence-rn- then disaggregard depending on whether the thuehold differences approh is used to compare the probabilities head was a ifomal or informal sector emnployee or was of four types of household responses to changes in self employed in the first per,iod. The probability of empoyment status of the household heads+ spouse enters entering or exiting poverty in booms ad recessions is emnployment, child leaves school, child repeats grade, and conmputed by edcation of the household head and child enters enployment. stylized facts of aggregate income risk and household wel- ated a greater proportional income loss to the poorest quin- fare in Latin America: tile than to any other. In all other cases (1990-91, 1996-97, (1) Aggregate income volatility affects different ranges and 1998-99), the greatest proportional-and therefore of income distribution differently, depending on obviously also absolute-income losses were borne by the the country and on the episode. There is no dis- richest quintile. In fact, during the recession that followed cernible pattern that either the poorest or the rich- the failed stabilization attempt known as the "Collor Plan," est households persistently have a higher income which was based on a temporary seizure of financial assets, volatility than others. proportional losses declined consistently by income quintile. (2) The ownership of assets-such as land, education, Growth episodes also appeared to have been more benev- and surplus household labor-reduces the risk faced olent to the poor than is generally acknowledged. In two of by households. the three boom episodes considered (1984-85 and (3) The poor, like everyone else, appear to be reluctant 1986-87) proportional income gains also declined consis- to make irreversible divestments during bad times, tently by income quintile. The third episode, which fol- and this is especially true of decisions concerning lowed the succesfull stabilization of the Brazilian real dur- the education of their children. The evidence ing 1994-95, is best described as broadly neutral. Table 5.6 broadly suggests that school enrollment is reason- and Figures 5.1 .a and b summarize the results from Brazil. ably insensitive to aggregate economic fluctuations, Using a different methodology, similar results emerge although school perfomance is not. Child labor is for Mexico in 1995. Households that suffered average or generally procyclical rather than countercyclical. median losses were found to be evenly spread across all (4) From the experience of some countries, relatively wealth classes. But households in the poorest 40 percent of large crises (deep or long recessions) appear to have the population were less likely to suffer large negative qualitatively different effects on poverty and invest- losses and were overrepresented among those "suffering" ments in human capital than smaller shocks; for small losses (or even gaining) in the aftermath of the 1995 example, the poor are affected more than the rich crisis. Table 5.7 reports the actual results of those regres- when the shocks are big, but vice versa when the sions. In addition, some groups often thought to suffer dis- shocks are smaller. proportionately, such as the elderly and single mothers, do not appear to be particularly badly affected, echoing earlier Do Aggregate Shocks Hurt the Poor More than the Rich? findings from Peru by Hall and Glewwe (1998). Macroeconomic volatility-in particular, unexpected nega- But just as they did during 1982-83 in metropolitan tive aggregate income shocks-do not appear to dispropor- Brazil, the poorest households do, on some occasions, fare tionately affect the incomes of any particular range of the worse than richer households in terms of the relative income distribution. Specifically, we find no support for the income losses inflicted by a shock. This was found for the common claim that the poorest are always those most rural Salvadoran sample, where the mean proportional affected by economic fluctuations. Of the four Brazilian income loss during the aggregate shock suffered by the recessions studied, only the most severe (1982-83) gener- poorest 20 percent of the population was 32 percent; 18 78 THE RESPONSE OF LAC HOUSEHOLDS TO ECONOMIC SCHOCKS TABLE 5.6 Metropolitan Brazil: Percentage Income Changes by Head's Wage Bracket WAGE BRACKET GROWTH RECESSION 1984-85 1986-87 1994-95 1982-83 1990-91 1996-97 1998-99 1 (poorest) 8.8 31.0 15.7 -33.3 -11.7 -1.8 -3.9 2 6.7 19.4 17.1 -30.7 -12.5 -1.9 -5.5 3 6.6 14.9 16.9 -31.0 -18.9 -1.7 -4.9 4 4.6 12.6 18.0 -28.6 -26.0 -2.0 -6.3 5 (richest) 3.8 4.9 14.0 -27.1 -28.1 -5.2 -6.5 Source: Neri and Thomas (2000). percent for the second fifth; 2 percent for the third; 5 per- although the different settings to which they refer imply cent for the fourth; and the richest 20 percent actually that the assets in question differed in importance. experienced a 9 percent gain in income.7 Perhaps the starkest evidence refers to the smoothing Just as the Salvadoran example showed that the results effect of land ownership on the plight of rural households may differ among countries-or for the same country, during the 1997 agricultural downturn in El Salvador. between rural and urban areas-results also differ if distrib- Table 5.8 decomposes the total change in the poverty head- utional weights are attached to losses, for example, with count ratio9 for the seven occupational categories previ- higher weights attached to incomes of the poor. This quali- ously mentioned, into three effects: one due to an increase fication should be kept in mind when interpreting results in poverty within the existing subgroup, another due to such as those for Brazil and Mexico, and in reacting to claims changes in the population shares of each subgroup, and a that the "the poor are affected more severely during crises."8 third that accounts for interactions between the previous two terms. Given the choice of a relatively high poverty Does Ownership of Assets Reduce Vulnerability to line, the headcount did not change much overall, rising Aggregate Shocks? from 0.65 to 0.69.10 However, the poverty profile was The second stylized fact of covariate income risk and transformed. The self-employed only group, which gener- household welfare in LAC is that asset ownership decreases ally has no access to land, accounted for some 16 percent of a household's vulnerability to shocks (in the sense of reduc- the poor in 1995; two years later, the figure was 25 percent. ing its proportional income variation). The term "asset" is Interestingly, this did reflect a rise in the intragroup head- used in a broad sense to include land, education, the bene- count ratio, but was predominantly due to an increase in fits associated with formal employment, and underused the number of people who lost jobs in agriculture. Corre- family labor. All three studies uncovered evidence of this, spondingly, the poverty share of agricultural wage workers FIGURE 5.1 Income Gains and Losses in Seven Episodes of Volatility in Brazil 35 -0 - ----------- ~ 6-97 87 ;98-99 - 'C 25 --10 20 - 15 2 3 4 5 1 2 3 41 r ~~~~~~~~~~~~~~~~~~-20 10 ~~~~~~~~~~~~~~~~~~~-30-- 5 - 35- 0 -40 - 1 2 3 4 5 1 2 3 4 5 Wage Bracket Wage Bracket 79 SECURING OUR FUTURE IN A GLOBAL ECONOMY TABLE 5.7 Mexico: Proportional Income Change by Income Quintile, 1995-96 PERCENTAGE CHANGE ACROSS 5 QUARTERS 0.2 0.5 0.8 0.2 0.5 0.8 A B C D E F Quintile 1 0.033* 0.007 0.085** 0.091** 0.091** 0.281** Quintile 2 0.048** 0.016 0.006 0.053** 0.046** 0.112** Quintile 3 0.018 -0.0068 -0.0291 0.064** 0.034** 0.094** Quintile4 0.011 -0.021 0.006 0.027 0.012 0.032 Consumption -0.562** -0.260** 0.115** -0.382** -0.004 0.508** Denotes statistical significance at the 1 percent level. ** Denotes staristical significance at the 5 percent level. Note; The table reports percentage income changes relative to the richest 20 percent of the population. The omitted category is the top quintile, and the proportional change in its income is given by the constant term; other entries indicate differences with respecr to thar change. SoAace: Cunningham and Maloney (2000). TABLE 5 8 Rural El Salvador: A Dynamic Decomposition of Poverty Changes, 1995-1997 CONTRIBUTION TOWARD CHANGE IN POVERTY (SE) HEADCOUNT RATIO TOTAL SECTORAL POPULATION SHIFT INTERACTION Self-employed only 160 22 127 11 Nonagricultural wage workers only -23 29 -43 -8 Nonagricultural wage + self-employed 21 -3 24 -() Agricultural wage workers only -91 11 -98 -4 Agricultural wage workers + self-employed 49 27 21 1 Agricultural wage + nonagricultural wage -12 36 -35 -13 Agricultural and nonagricultural wage + self -5 -22 20 -3 Total 100 100 16 -16 fell from 18 percent to 11 percent, despite an increase in its 5.2, which suggests that those with higher earnings did own headcount ratio. not suffer income losses even during the agricultural down- The results for El Salvador indicate that-controlling turn (and may even have done better during the crisis), this for income-land ownership makes households more likely hints at the role of education as an important self-protec- rn keep children enrolled in school, and helps preserve the tion instrument.'2 productivity of labor during crises. The importance of land This possible self-protection role of education was found is confirmed by statistical analysis to estimate earnings in to be associated with a smaller probability of transition both periods. Land ownership appears to have played a into poverty and a larger transition rate out of poverty, more important role as a self-insurance strategy than as a both during recessions and growth spurts in Brazil. Figures direct determinant of earnings in good times: the effect of 5.3.a through 5.3.d illustrate that those associations were land ownership was not statistically significant in 1995, robust for all three growth episodes and all four recessions and only became important after the crisis. During good considered. times, it was access to off-farm employment, rather than A partial exception to this role of education is the case having a plot of land, that had the greatest (and most sig- of Mexico, where households headed by college-educated nificant) impact on household income."1 And both original males suffered somewhat larger proportional falls in access to nonagricultural employment and the ability to income as a consequence of the 1995 crisis than did those keep it after the crisis appear, in turn, to have been corre- with primary or secondary education. However, inclusion lated with that other asset crucial to the poor: education. of what the authors call "coping variables"-namely The results for rural El Salvador also indicate a strong entry into the labor force by the head, the spouse, or a and statistically significant effect of the years of schooling child-reduces that advantage of the uneducated, and it of the household head on income. Combined with Figure ceases to be significant at the median. It appears that for 80 THE RESPONSE OF LAC HOUSEHOLDS TO ECONOMIC SCHOCKS FIGURE 5.2 is particularly true of parental decisions about the school- E1 Salvador. Cumulative Distributions of Real Income Per Capita, ing of their children. 1995 and 1997 A number of recent studies have emphasized the risk d that, in addition to their temporary impacts, temporary negative income shocks might have permanent effects on the incomes of poor families. One of the main transmission .75 - >°mechanisms which these studies suggest for such "poverty hysteresis" is that parents might be forced to take their o children out of school to deploy them in income-generat- ing activities (see, for example, Lustig 1999, and IDB 2000). Even though the opportunity cost of schooling is likely to have declined, as the covariate shock reduces the .25 - potential earnings in the market for children's labor, the argument goes, a subsistence constraint may become bind- ing and necessitate a reallocation of the child's time away from schooling and toward work. The existence of a sub- o- l sistence constraint, and of irreversibilities in the educa- 0 5,000 10 000 tional production function, would thus lead to a rational Income Per Capita, 1997 Colones decision which might, nevertheless, imply a reduction in the lifetime earnings of the child. Mexico any apparent greater ability to weather shocks by Although the conceptual argument is plausible, empir- the less educated is due to their greater ability or will- ical tests have been scant. And they are needed: theoretical ingness to send household members previously out of the predictions are ambiguous because of opposing effects of labor force into it. If one thinks of underemployed family earning opportunities (which are lower during bad times) labor as an asset, this finding simply suggests that educa- and the need for income (which is higher during bad tion is a subsitute for it, as a self-insurance strategy. And times). Examining the effect of the four Brazilian reces- if leisure is a good thing, as Box 5.3 suggests, the conse- sions and three growth episodes on three child schooling quences for family welfare, particularly that of women, variables-dropout rates, grade repetition rates, and child may be substantial. labor participation rates-helps shed light on the nature of In fact, recourse to underused family labor is also found the effects sketched above. As Figures 5.5, 5.6, and 5.7 to be an important coping strategy in rural El Salvador. suggest, school enrollment is largely acyclical with respect This may reflect a greater proportion of women in the wage- to trend, but child labor and grade repetition are mildly contract labor force-where demand rationing was clearly procylical. in effect-and perhaps more important, a complementarity The economic cycle appears to have no overall effect on between greater reliance on owned land and unpaid family dropout rates. It does, however, have some effect on repeti- female labor. Just as in Mexico's urban setting it appears tion rates, a serious problem in the Brazilian education sys- that education and surpLus family labor are substitute assets tem. Upturns appear to increase repetition rates; this may in coping with a crisis, in the rural El Salvador setting, sur- be associated with the procyclical nature of child labor plus family labor and land are complementary assets, with a (which can be engaged in without dropping out of school measurable volatility-reducing effect. entirely, but at the cost of diminished performance). The result for child labor, illustrated in Figure 5.7, suggests Do the Poor Engage in Self-Destructive or Myopic Coping? that the effect of a lower opportunity cost of schooling dur- The third stylized fact of covariate income risk and house- ing recessions outweighs the other effects.'3 hold welfare in LAC is that the poor, like everyone else, However, a more detailed investigation revealed that for appear to be reluctant to jeopardize their (family's) future children of workers moving from formal sector jobs to in an irreversible way during a temporary downturn. This informal self-employment-which is more frequent dur- 81 SECURING OUR FUTURE IN A GLOBAL ECONOMY FIGtlRE 5.3 Brazil: Moves Into and Out of Poverty, by Level of Education [a] Into Poverty During Growth [b] Into Poverty During Recession 40 40- 30 30 W>> ~ ~ ~~~~ ~ ~~~~~~~~~~~~~~~~ 90 9-\> 0 91 C 0 1 0 20 -_ _ 10 1~~~~~~~~~~~~~~~~0 I none 1-3 4-7 8-11 12+ none 1-3 4-7 8-11 12+ Years of Schooling Years of Schooling [c] Out of Poverty During Growth [d] Out of Poverty During Recession 70 70 60 - - .-- - _ . ...-. -_ . - 60 50 - - - - -6 -__ 50 .-.- . - - _____ ___6 _ 40 ~ 40- 2 30 30 a. 84-85 96-97 20 20- 10 I00 none 1-3 4-7 8-11 12-+ none 1-3 4-7 8-11 12+ Years of Schooling Years of Schooling Soare: Neri and Thomas (2000). ing recessions-repetition rates increased substantially, The procyclicality of child labor is not a peculiarity of and there was also an increase in the work participation of metropolitan Brazil. In Mexico, the proportion of house- these workers' spouses. It is tempting to hypothesize that holds sending children to work during the 1995-96 cycle informality may worsen school performance by drawing on was consistently low (less than 4 percent) and marginally "surplus" household labor." Brazilian children-and prob- higher during the recovery than during the downturn (see ably those in most LAC countries-age 10 to 15 have been Box 5.4). The procyclical nature of child labor was more pro- entering the labor market at decreasing rates over the last nounced for the poorest quintile, where 3.8 percent of fami- two decades. This trend is clear, and the effects of aggre- lies added children to the work force in the recession, but 5.7 gate fluctuations are relatively minor. But they do exist and percent did during the recovery. In Chile, once again, labor are procyclical: entrance into the labor force for this age force participation of youth age 15 to 19 is procyclical, group has generally been higher in booms than during falling for both males and females during the 1982-84 recessions. recession. The same is true for men age 20 to 24.'5 Again, it 82 ITHE RESPONSE OF LAC HOUSEHIOLDS TO ECONOMIC SCHOCKS BOXD 5 e3 during recessions. Hence, the results have important policy WI3oreDoesNmeGoDurlngaCrisi? implications. Caution is necessary, however, until this is found to be true for other countries or settings (for example, Using the same Mexican panel data, Cunningham rural areas), and one must keep in mind that this procycli- (1999) and Parker and Skoufias (2000) find that wives cality result relates to child labor, not school enrollment. In do enter the job market if the husbancd loses his job. fact, the results for school enrollment are mixed: in Brazil, But when a woman enters the job market, who does the enrollment was acyclical while in Chile enrollment declined work she used to do? Or does she simply lose her during a severe recession. There is also some evidence that leisure time and work a "double shift"? ingredients for producing well-schooled children decline in Cunningham (1999) finds that for every additional quantity or quality during downturns (for example, because hour worked outside the house, women do one-half of lower public and private spending or because some hour less housework, in effect, working a "shift and a spouses join the labor force and have less time to spend with half." Other members of the family compensate some- their children). It is also too early to make strong claims what, but overall, the household work falls by about about the nature of the effect of aggregate income shocks on the same amount, This implies both that women the education of poor children. Nevertheless, the evidence become more "time poor" (that is, lose leisure time) uncovered so far can provide useful guidance to policymak- and that many necessary household tasks-such as rais- ers, as we outline in Chapter 7. irg children, ensuring healthy living conditions, and isvesting in socially usefil nerworks-may get less Are All Downturns the Same in Their Effects? attention:(see Figure 5.4). Discussions about policies to protect the vulnerable from the effects of aggregate volatility generally assume that all downturns are similar in their effects on poverty and was only for young adult women (age 20 to 24) that partic- human capital investments. In fact, the evidence from the ipation rates were found to be countercyclical. The results case studies for Brazil, Chile, El Salvador, and Mexico from Chile confirm both the countercyclical pattern for reveal that this is not true: longer, deeper recessions appear female labor force participation rates and the procyclical pat- to have results qualitatively different from shorter down- tern for child labor found by the other studies. turns. The poor suffered greater proportional losses in These results run counter to frequent claims that gov- income during severe recessions than the wealthy (but this ernment efforts to combat child labor should be stepped up was reversed at least for metropolitan Brazil and Mexico), FIGURE 5.4 Change in Housework rime Due to Labor Force Entry 10 5 - . i g 1R~~~~~~~~~~~~~~~~~~~~~~~~F7 Headl male| O _ E] wC-': r S S ; CL Teen son * Teen daughrer MOther male -I Other female -10 Head Female Wife Enters Husband Enters Enters -15 83 iFCtJRING OtlR F[tTL RIF IN A GLOBAL ECONOMY FIGURE 5 5 Brazil: Probability of School Dropout 8- 7 5 4 3 _ 2 - downrurn . . .. . .. . .... . P10 *tptulrn I _ _ trend 0 -IIIIII 1982 1984 1986 1988 1990 1992 1994 1996 1998 FIGURE 5.6 Brazil: Probability of Repeating a Grade at School 40 - - 35 - 30 --1 25 - 1 20 - 15 - 10 4& downturn n upturn - trend 0 - 1982 1984 1986 1988 1990 1992 1994 1996 1998 FIGURE 5.7 Brazil: Probability of Child Aged 10-15 Starting Work 14 4 12 10 8- 6- 4 * downturn 2 1 uipturn 2 trend 0- 1982 1984 1986 1988 1990 1992 1994 1996 1998 Source Neri and Thomas (2000). 84 THE RESPONSE OF LAC HOUSEHOLDS TO ECONOMIC SCIIOCKS and children were pulled out of school in Brazil and Chile changes) was substantial. Recessions like those of when the recession was deep, but not otherwise. If con- 1982-83 and 1990-91 in Brazil led to proportional firmed for other countries and settings, this finding has declines in mean incomes for some quintile groups on the important policy implications. order of 30 percent. In Mexico's 1995 crisis, the median The Brazilian recession of 1982-83 was more serious proportional income loss for all households was around 25 than the other three downturns (1990-91, 1996-97, and percent, and for those suffering "catastrophic losses" (that 1998-99) examined here. But it was also different in that is, in the first quintile group of the distribution of while in milder recessions the poorer 40 percent suffered changes), the figure was over 50 percent. In El Salvador, proportionally less than the wealthiest 40 percent of the the agricultural crisis of 1997 led some measures of the population-declines in income of about 12 percent, 2 severity of poverty to rise by about 37 percent among the percent, and 5 percent for the poor during 1990-91, rural self-employed. 1996-97, and 1998-99, respectively, as compared with If Latin Americans are risk-averse, as we suppose most 27 percent, 4 percent, and 6 percent, respectively-the people to be, this degree of household income volatility deeper recession resulted in a loss of 32 percent for the implies a considerable loss of social welfare. Reducing it poorest two-fifths as compared with 28 percent for the would make workers and their families better off, other richest two-fifths (see Table 5.1 and Figure 5.1). Again, things remaining the same. This argument and the mag- the Chilean recession of 1982-84, which was steeper and nitude of the variations at the household level provide a longer than other episodes of negative growth, was also powerful incentive to study the determinants of micro- different in that it was the only recession during which economic risk and the strategies adopted by households child enrollment declined. to reduce it and insure against it. The rationale for these findings may be that the poor The general conclusion of this chapter is that house- have smaller asset stocks (both in absolute terms and rel- holds largely respond rationally and sensibly both from ative to income flows) as compared with the rich, limit- the viewpoint of individual households and often even ing their ability to draw on their assets for prolonged from that of society: most notably, they attempt to shelter periods. Thus, while they behave similarly during mod- their children's schooling and leisure from economic erate downturns-drawing down some of their assets and shocks, both aggregate and idiosyncratic. This does not working longer hours but maintaining critical long-term mean that governments cannot do more to help house- investments such as education of children-longer down- holds deal better with income risk. Chapter 6 dicusses turns result in a divergence of behavior between the rich how countries in the region have sought to deal with the and the poor. Again, while the reduced earnings (or "sub- risk of unemployment, and finds that countries can better stitution effect") may offset the propensity to send chil- match income support programs for the unemployed with dren to work due to the subsistence constraint (or the type and level of risk, and the level of sophistication of "income effect") during short downturns, the latter may the available instruments for self-insurance. Chapter 7, dominate the former during downturns that are expected which addresses the issue of what governments have done to last longer. In any case, if confirmed for other countries and how they could improve public interventions, finds and settings, this finding, too, has important policy that while "targeted" social spending is often both pro- implications which are discussed in Chapter 7. cyclical and poorly targeted, broader social spending such as that on education and health is less subject to be cut Conclusion during economic downturns, but is often insufficient to Large macroeconomic volatility in LAC both causes and prevent a deterioration of performance indicators during obscures even more substantial variation in the incomes recessions. The findings of Chapter 5 actually imply that and employment status of individuals and families in the there is considerable scope for welfare gains from better region. Although lack of consumption data prevented policies for dealing with the risks of unemployment and better estimates of variations in household welfare, the aggregate economic fluctions that can lead to increased magnitude of the income variations reported here (and poverty. The final sections of Chapters 6 and 7 propose which constitute an upper bound to the true welfare how these gains can best be realized. 85 SECURING OUR FUTURE IN A GLOBAL ECONOMY BOX 5.4 Do Families Mortgage Their Chldren's Future? Other Evidence As with the evidence fom Brazil and Mexico, there is lit- child wages dominates the income effect over the eco- tie or only weak evidence from other countries in the nomic cycle). region that fiailies &ced with income shocks either put Cunningham (1 999), Parker and Skoufias (2000), and children to work or pull them out of school. The caimns Cunningham and Maloney (2000) focus on Mexico and are made nevertheless. study the impact of income shocks on child labor and For Peru, using Living Standards Measurement Sur- school attendance befre, during, and after the 1995 Cei- veys for 199 and 1997, Ilahi (1999) found that when sis, examining the effects of idiosyncratic shocks while mothers become unemployed, children allocate less time controlling for aggregate risk. All focus especiaLly on the Ito household chores, but there is litde or no change in effcts on children of involuntary loss of employment. schooling or child labor. Schady (2000) found no signifi- Parker and Skoufias focus especially on those due to ill- cant difrece between school enrollment rates in Peru ness, divorce, or other labor market reasons-of any of in the crisis year of 1991 and in the growth yeas of 1994 the pares. Their finding is that children-boys more so and 1997. Once again, acyclicality is observed in the than girls-are largely unafcted by household-specific broader context of a secular rise in enrolment, economic shocks, in periods of both aggregate downturns Like the Neri and Thomas (2000) paper discussed and recovery here, Duryea (1998) examirne the effects of shocks in Exploiting the dynamic possibilities of the panel, Cun- metropolitan Brazil. But the nature of shocks is idiosyn- ningham and Maloney do find the predicted effects on girls cratic rather than aggregate: unemployment of the father in resonse to the father's job loss, but ambiguous evidence lowers the probability of grade advancement of children for boys. Fmuher they find that while in less well-off fam- age 10 to 15 by about 4 percentage points, but Duryea ilies children do work and drop out more, there is little (1998) does not explore the reasons for this-fr exam-i evidence of credit constraints that would cause poorer fain- pie, whether this is due to a less favorable family envi- ilies to put children to work. However, they do find that ronment, reduced intputs such as books, or because the where a wife or husband enters informal self-employment, child has to begin or increase paid work. Recall that Neri children are more likely to work. As with the evidence and Thoomas (2000) fund no systemnatic evidence that from Brazil that the substitution effect dominates the earolment or tpetition or dropout declines during pen- income effect, starting a family tnicroenterprise may raise ods of aggreate downturns, but fund some evidence the value of a child's work time and encourage entry. that child labor increases during periods of recovery (in Finally, they find only ambiguous evidence of household economic parlance, the "sibstitution effct" of changes in adjustments lasting more than a quarter. L ..... . - ---- -- ------ ---- --- ,__ ,.......... .. .................... . . ............. -.. - - ---------- - .- P. Notes seven contiguous EPH cohorts were combined beginning in May 1. See Caballero (2000), Calvo (1991), and Rodrik (1999) for a 1993 and rotated every six months. sampling of possible explanantions on high volatility in the region. 5. This section is based on Ferreira and Gill (2000). 2. Cornia and others (1987) and the World Development Report 6. The consumption variations would only be equi-proportional 1990 played an important role in raising the profile of this issue in if households were certain that the shocks were permanent. the late 1980s. 7. There are a number of good reasons to believe that the nature 3. For a path-breaking exception, see Glewwe and Hall (1998). of this negative shock in 1997 did affect the rural poor more severely 4. Arango and Maloney (2000) use both the Mexican Household than other segments. We will discuss some of these below. But it Survey described in Box 5.2 and the Permanent Household survey should also be noted that the ranking of households in this stuay is (EPH) from Argentina. The EPH conducts extensive biannual inter- based on an average of each household's income in 1995 and 1997. views in Greater Buenos Aires and is structured so as to generate This is arguably a less robust proxy for permanent income than either panels that allow tracking a quarter of the sample across two years. the neighborhood average used by Cunningham and Maloney (2000), To generate a sufficiently large sample of roughly 5,700 observations, or the Mincerian predicted income used by Neri and Thomas (2000). 86 THE RESPONSE OF LAC HOUSEHOLDS TO ECONOMIC SCHOCKS 8. Cunningham and Maloney (2000) show that when permanent farming and nonfarming hours, while the loss of wage hours is dri- incomes are weighted to give more importance to the poor, the pro- ven principally by reduced agricultural wage employment. portional losses accruing to the least educated-and hence likely the 12. See Box 3.1 for definitions of alternative insurance instru- poorest-do indeed become greater than when that weight is 1. Like ments; see Ehrlich and Becker (1972), and Gill and Ilabi (2000) for them, other researchers making claims about who suffers dispropor- details. tionately during crises should make their welfare weights explicit. 13. See Ferreira and Gill (2000) for a fuller characterization of 9. See Conning, Olinto, and Trigueros (2000) for analogous these opposing effects. decompositions using other poverty indexes. 14. This suggests a caveat to our earlier consideration of surplus 10. Depth and severity, as measured by other poverty indexes, household labor as an asset with insurance value. It is likely that rose in a more pronounced manner. those family members were not completely idle. In fact, mothers 11. The "self-employment only" category into which many were most likely being a highly productive input into the education households fell in 1997 due to having lost jobs is not made up of their children. Their joining the labor force is not without cost mostly of the landless. In fact, most of these households are engaged and, even if the evidence is largely against the proposition that this at least in part in agricultural or animal husbandry activities. The leads to massive increases in dropout rates, the quality and pace of relationship between land ownership and poverty and vulnerability their children's education may suffer nonetheless. is complex. The main finding in Conning, Olinto, and Trigueros 15. Mizala and Romaguera (2000) report gender-disaggregated (2000) is how much household labor allocation changes in a crisis. time-series for labor force participation rates for people age 15 to 19 The large increase in self-employment hours is split evenly between and 20 to 24 in Chile, from 1976 to 1993. 87 CHAPTER 6 Helping Workers Deal with the Risk of Unemployment L ATIN AMERICA'S SHIFT AWAY FROM THE STATE-LED DEVELOPMENT MODEL TOWARD A MARKET- based economic paradigm has rendered obsolete key components of the old income protection system, such as layoff restrictions and the state as employer of last resort. On the one hand, inter- national competition entails the need for firms to use greater flexibility in managing their staffs, weakening the traditional lifetime relationship with their employees. On the other hand, pub- lic finance constraints, privatization of public enterprises, and renewed emphasis on financial performance of those remaining in public hands have all led to public sector downsizing and to a much-reduced role of the state as employer. All these developments, together with the potentially adverse employment effects of the economic reform process on some groups of workers documented in Chapter 2, have raised considerable interest in Latin America and the Caribbean in income support programs that could mitigate the effects of economic insecurity in general, and of job loss in particular. This chapter deals with the latter. As in many other policy areas, developed countries have served as the reference when thinking about the appropriate design of income support programs for the unemployed. Some analysts have discussed ways to adapt unemployment insurance programs to developing countries (Hamermesh 1992), and some have even tried to calibrate the parameters of such programs to the specific characteristics of those countries (Hopen- hayn and Nicolini 1999).1 However, it is not clear that unemployment insurance is the best-suited income support program under all circumstances. More recently, This chapter assesses a set of income support programs that there have been proposals to introduce individual unem- have been tried in the Latin American context. Rather than ployment savings accounts, whereby workers are forced to starting from a theoretically "ideal" program and adjusting it set aside money when at work, and are given access to these to the characteristics of a specific country, the chapter consid- savings in the event of job loss. While these programs do ers specific programs that are currently in operation. Until provide a more certain severance benefit to workers, these recently, the accumulated knowledge on income support pro- proposals are justified largely on theoretical grounds: indi- grams for the unemployed in developing countries was quite vidual savings accounts provide better incentives than con- limited.3 Hence, while this chapter draws from existing liter- ventional unemployment insurance to contribute to when ature, it is mainly based on studies commissioned by the employed and to search for a job when unemployed.2 But World Bank specially for this report. These studies deal with there are theoretical arguments against this proposal as the operation and effects of specific income support programs well. In particular, individual savings accounts do not pool for the unemployed in Argentina (Ravallion 2000), Brazil risk among individuals, and thus may be less efficient than (Cunningham 2000), Colombia (Kugler 2000), Mexico those that do so explicitly (such as formal unemployment (Wodon 2000), and Peru (MacIsaac and Rama 2000).4 insurance) or implicitly (such as income support programs The objective is not to generate an unambiguous "rank- financed from general tax revenues). ing" of these programs; all of them can be expected to have 89 SECHURING OiLR FlTURE IN A GLOBAL ECONO)MY both strengths and weaknesses under different objectives of training for the unemployed (TG), unemployment insurance policymakers, depending on the constraints they face. For (UI) and individual savings accounts (IA). The main features example, a program may do well at offsetting the losses of these programs are summarized in Box 6.1. The timing formal sector workers experience as a result of increased and nature of the payments involved in each of these five import competition or deregulation, but fail to reach infor- income support programs is summarized in Table 6.1. mal sector workers who risk poverty as a result of aggregate To interpret Table 6. 1, consider the following stylized fluctuations. Similarly, a program may have broad cover- labor market sequence. In period 0 the worker loses work, age, but also entail a large cost for the budget. Which pro- in period F he or she finds a new job, D is the maximum gram is best suited for a country depends especially on the duration of benefits allowed by the income support pro- state of labor markets (for example, the extent of informal gram, and R is the worker's retirement age. The table is employment and the frequency of joblessness) and the constructed under the assumption that F comes before D, administrative capacity of governments to implement dif- but this is not necessarily so in practice. The rows in the ferent public income support programs. table correspond to the five broad types of income support This said, it should also be kept in mind that govern- programs currently in operation in LAC. A zero indicates ments can overcome some of these constraints over time: that the program does not entail any payment, from or to labor market reforms can reduce the difference between for- the worker, in the corresponding period. Possible pay- mal and informal activities, and administrative capacity can ments include the salary received from a public works pro- be built. In going from what governments have done to gram (W), the training allowance (A), the amount: of sev- facilitate income support to the unemployed to how they erance pay received from the employer (S), the contribution could do better, therefore, we bring in a medium- and by worker and employer to an income-support program longer-term perspective as well. While some of the policy (-C), a benefit whose amount depends on past contribu- recommendations are based on specific theoretical premises, tions (+C), and an old-age pension (P). we believe that a blend of practicality and analytical rigor In the individual accounts program, the old-age pension can help countries devise strategies that efficiently bridge can be accrued by the portion of the forced savings that was immediate action ancl long-term vision. This chapter intro- not used as income support in periods of unemployment. duces programs of income support in five countrics, distin- While only workers cnrollcd with social security can par- guishing their key features; summarizes the evaluation of ticipate in the unemployment insurance and individual these programs; and using the main findings of these and accounts programs, the other three programs are in princi- other studies and the comprehensive insurance framework ple accessible to nonenrolled workers-hence the question outlined in Chapter 3 as an organizing device, discusses the marks in the postretirement columns of Table 6.1. Other main policy implications for LAC economies. question marks in this table reflect the possible inability or unwillingness of the social security administration to mon- A Typology of Programs itor whether beneficiaries are actually out of a job. At least five different types of income support programs for In this report, we take one example of each of the five the unemployed have been tried in LAC.5 All of them programs and examine how well it has helped to deal with involve a net transfer of resources to workers who lose their job loss. Public works are represented here by Argentina's jobs, but the amount, conditions, and sources of the trans- Trabajar program. This program was created during fer differ substantially across programs. Conceptually, some 1996-97, in response to a surge in unemployment in of these programs can be seen as a mere redistribution of 1995. Trabajar allocates funds across provinces based on resources, from taxpayers to the unemployed; others the distribution of the unemployed poor. Proposals to use amount to forced savings or self-insurance, made available the funds are made by municipalities and non-government in the event of job loss. Still others resemble market insur- organizations. These proposals are approved at the regional ance, with a premium being paid while at work and a claim level, based on a system of points related to poverty in the being made in the event of unemployment. area and the merits of the project. The government pays for The five types of income support programs used in the the costs of unskilled labor and the sponsoring units pay region are public works (PW), mandatory severance pay (SP), for materials and skilled labor. The wage of unskilled 90 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT BOX 6.1 Income Support Programs for Me Unemployed: Main Features Public works. This income support program was used by salary, and in other countries the gratuity formula is dif- Chile in the 1970s and 1980s, and was introduced more ferent when job separation is due to economic reasons. recently by Argentina, Brazil, Colombia, and Mexico. Employers are usually not mandated to set aside any The program provides low-wage jobs to all those who are resources to pay the end-of-service gratuity. willing to take them. In principle most of the jobs are in Individual accounts. This is a "funded" version of the activities that do not compete with the private sector, severance pay program. Workers have individual and the jobs can be physically demanding and typically accounts to which some percentage of their salary is last for only a few months. Therefore, it is likely that the transferred on a regular basis. In the event of job separa- program attracts unemployed workers and those out of tion, whether voluntary or involuntary, workers can draw the labor force, and not the currently employed. High resources from their accounts. Any resources left in these effort, short duration, and low pay imply that only the accounts at retirement can be used toward old-age pen- neediest participate. sions. Workers can also "borrow" from their accounts Training. To the extent that some training programs under specific circumstances. A program along these are specifically targeted to the unemployed, and provide an lines has existed in Brazil for more than three decades. income allowance to the trainees, they can be viewed as a More recently, Colombia has replaced its severance pay form of income support. Mexico has a large program along program with fully funded individual accounts of this these lines. Training programs for the unemployed have type. Unlike unemployment insurance and severance pay, some similarities to public works programs: resources are this program involves no net transfer of resources to provided by the government, beneficiaries do not need to workers who lose their jobs. be enrolled with any social security program to have access Unemployment insurance. This program, modeled to the training, and being enrolled in a training program after those in developed countries, can be found in is usually incompatible with having a job, so that there is Argentina, Barbados, Brazil, Ecuador, Uruguay, and some self-selection of the beneficiaries. In addition, these Venezuela. Workers and employers in the formal sector programs generate a "product" in the form of better skills, make regular contributions to a fund, generally managed although the market value of these skills is open to debate. by the country's social security administration. After a They are different from public works programs in that eli- specified contribution period, workers are entitled to an gibility rules generally apply-only individuals of speci- unemployment benefit in the event of job loss (but aot of fied age or education levels are allowed to participate, voluntary separation). The benefit is some percentage of Severance pay. This program can be found in most the workers' salary, typically declining over time. Bene- Latin American countries, where the labor code mandates fits are paid for as long as the worker remains unem- employers to pay an end-of-service gratuity to the work- ployed, up to a maximum of several months or years. ers they fire without a "justified" cause (that is, for However, monitoring whether beneficiaries take a job in nondisciplinary reasons). The gratuity is a multiple of the the informal sector is practically impossible, so somne of worker's salary; in some countries, the gratuity cannot the programs do not even attempt to make the payment exceed a specified amount or multiple of the worker's of benefits conditional on actually being unemployed. laborers is set at two-thirds of the average wage for the ing standards due to the 1982 economic crisis. The stated poorest decile in the capital city. In principle, there are no objective of the program is to improve the productivity of restrictions on the eligibility to participate in the program, unemployed workers to help them find jobs. The program but in practice there is rationing. provides publicly funded training and a subsistence Mexico's Probecat training program was created in 1984, allowance during the training, with a maximum duration in response to rising unemployment and deteriorating liv- of six months. The allowance is equal to the minimum 91 SECURING OUR FUTULRE IN A GLOBAL ECONOMY 'IABLE 6.1 How Various Income Support Programs for the Unemployed Work FINDS INCOME ANOTHER SUPPORT AT PERIOD EMPLOYED LOSES JOB UNEMPLOYED JOB ENDS UNEMPLOYED RETIREMENT -I 0 1 F D D+1 R Public works 0 0 W 0 0 0 ? Training for unemployed 0 0 A 0 0 0 7 Mandatory severance pay 0 S C, 0 0 0 ? Individual saving accounts -C +C +C 0 0 0 P+C Unemployment insurance -C ? B B? B? 0 P Note: The timing of events is at follows: in period 0 the worker loses his or her job; in period F he or she rakes a new job; D is the maximum duration of benefits; R is the retire- ment age. The table is constructed under the assumption that F comes before D, but this is not necessarily so in practice. The programs involve the following payments or transfers: W is the salary paid by the public works program; S is the amount of severance pay received by the worker; A is a training allowance; -C is the contribution by worker and employer to a program; +C is a contribution-defined income transfer; P is the old-age pension, with the question mark indicating that beneficiaries of the program are nor necessar- ily enrolled with social security; B is anemployment benefits, with the question mark indicating that the worker may not receive the benefit during that period. wage plus transportation costs and health insurance cover- low. It was part of a broader set of reforms, which included age. Training was initially carried out in schools and train- trade and financial liberalization. Under the new system, ing centers, but it was subsequently expanded to enter- workers have to contribute 9.3 percent of their salary to an prises as well. Participating enterprises are required to hire unemployment fund. They are entitled to their savings in at least 70 percent of the trainees. Beneficiaries are selected the event of termination, but can also "borrow" from them based on a system of points, and can get trained only once. for housing and education purposes while employed. In the For about a decade the program was small, but after several old system, workers could also borrow part of their sever- positive reviews the program was expanded by a factor of ance pay entitlement from their employers, but the value 10. Two criteria are used to evaluate the program: success of the loan was not appropriately adjusted for inflation. in providing income support (that is, an income transfer or Some specific groups of workers can opt out of the new sys- "social insurance" function) and effectiveness in reducing tem, and get a higher salary in exchange for not being cov- future incidence or duration of unemployment (that is, aid- ered by the program. ing "self-protection"). Brazil has the largest unemployment insurance program The mandatory severance pay program examined here is in the region, though it is small in comparison with those the one Peru had during the 1990s. Over the two previous in OECD countries with relatively frugal UI systems, such decades, this program had been part of an attempt to guar- as the U.S. system (see Table 6.2 and Gill, Dar, and antee job security to workers after probation. Job security Thomas, 1999). This program was created in 1986, as part regulations lost their teeth in the early 1990s and were for- of a policy package (the Cruzado plan) aimed at price stabi- mally abolished in 1993. But mandatory severance pay reg- lization. The program is funded by taxes levied by employ- ulations were not substantially eased. Although the formula ers. To be entitled to unemployment insurance benefits, a setting the amount to be paid in the event of "unjustified" worker has to be covered during 15 of the 24 months pre- dismissal was modified four times in the 1990s, the basic ceding job loss. Benefits are in the range of one to two min- structure remained the same. As of 1996, the employer had imum monthly salaries,6 depending on past contributions. to pay one month of salary per year of service, up to a max- Benefits are paid in monthly installments, up to a maxi- imum of 12 months. In 1997, mandatory severance pay was mum of 120 days, but they are not contingent on being out raised by 50 percent. Since then, the mandatory severance of work. Workers need to be present in person at social pay program of Peru has not been modified. security centers to collect their benefits. Colombia is one of the few LAC countries that suc- ceeded in replacing its mandatory severance pay program Main Findings with an individual accounts program. This change was The five income support programs can be assessed along made in 1990, at a time when the unemployment rate was several dimensions (see Box 6.2). Proposals to introduce 92 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT TABLE 6.2 Contrasting the Brazilian and the U.S. Unemployment Insurance Systems CHARACTERISTICS BRAZIL U.S. Administered by Federal government State and federal governments Number of Claimants 3-5 million per year 15-20 million per year Objectives (i) Alleviate hardships due to [oss of earnings (i) Alleviate hardships due to loss of earnings (ii) Automatic macro stabilization (ii) Automatic macro stabilization (iii) Improving quality of job matches (iii) Help stabilize employment by experience-rating of (iv) Making employers share burden of unemployment employers Tax Rate Federal tax = 0.65% of revenues of private firms, Federal tax = 0.8% of taxable payroll; State Tax varies 1.0% of revenues of public firms, and 1.0% of from 0-1 0% of taxable payroll according to employers' expenditures of nonprofit firms experience-rating Tax Base Firm revenues or expenditures (see box above) Federal: up to $7,000 of each employee's payroll States: at least $7,000 of each employee's payroll (80% of states had tax base above $7,000) Use of Funds 40% of collections transferred to uses other than paying Both principal and interest must be used to pay only Ul benefits through mandated transfer to national Ul claims development bank Benefit Replacement Rates 80% of reference wages up to R$220 per month; Between 32-57% up to benefit ceiling. Lowest ceiling 50% of wages between R$220-254 per month; is $180/week; highest is $390 ($575 with dependency 0% above this allowance) Reference Wage Average salary 3 months prior to dismissal Last salary before dismissal Duration of Benefits 3-5 payments depending on work history during Regular benefits up to 26 weeks; extended benefits up previous 3 years; can receive Ul payment package only to 13 additional weeks; additional temporary benefits once in 16 months during recessions Source: Gill, Dar, and Thomas (1999). "optimal" unemployment insurance, or individual sav- * Cost. How large are income transfers in the program ings accounts, usually focus on the incentives these pro- considered? What fraction of the total cost of the grams could provide for workers to actively search for program is actually received by the worker? Do jobs. In countries with high informality, however, it is workers "buy" income protection through lower also important to consider who these programs reach. In wages, or is the burden shifted to employers or addition, the burden of some of these programs does not taxpayers? fall only on their beneficiaries. Depending on who "pays" * Incentives. Do workers who are covered by the pro- for the benefits, and how, the programs can have different gram remain unemployed for longer periods than implications on efficiency and equity grounds. Finally, in those who are not? Do they find jobs with higher the absence of income support programs, some of the earnings, or jobs that are "better" in any other sense? unemployed could resort to their savings, or to transfers * Insurance. Do workers who are covered by income from relatives, to support their consumption. It is there- support programs display smoother consumption fore important to assess whether formal income support patterns than those who do not? Does the consump- programs really help smooth consumption, or simply tion of covered workers fall less, other things equal, replace other more informal self-insurance mechanisms. in the event of job loss? Based on these considerations, the findings on these five Not all these questions can be easily answered for all the income support programs can be summarized along four programs. This would require a vast amount of informa- dimensions: tion on the employment, earnings, and consumption his- * Coverage, How many workers are eligible to partic- tory of a representative sample of workers, and the data ipate in the income support program considered? available in the five countries usually do not deal with all How many actually benefit from an income transfer, of these variables at once. Moreover, information on the or have benefited from one recently? How does cov- individual characteristics of the workers (such as age and erage vary with wealth? education) is necessarily limited. Therefore, it is always 93 SECURING OUR FUTURE IN A GLOBAL ECONOMY Box 62 Data $Sorces and MethodO Agentina's Taba/r Anoa de Hoga*, for all the years between 1986 and The efftcts of the Tr4ba]r program aeassessed coibiot- 1997. Th attibute all the observed changes in job ing two househol surveys, One of them, the Ecs tenure to the weakening of lob security reguiations, Dearoli Soia MS, was carried out in 1997 and cov- although this was not the only reform or shock that took ered the population residing in localities with 5,000 or place over this period more inhabitants. Jalan and Ravallion (1999) constructed a conparison gr out of the EDS sample, using dmah. Colmias lniil Sai Acoutt s ing methods. The other household surve used to assess ITn the case of Colombia, a similar natural experiment is the effects of Traba* is a 1997 sample of its participants, provided by the 1990 laor market reform, which covering 3,500 househds. The infrmation used by replaced severance payments by individual accounts. The Ravadlion (2000) on Trab r is fom government dat used by Kugler (2000) to assess the efcts of this I eords. change are daw fom the June 1998, 1992, and 1996 rounds of the E£nct Nacioal de R es (ENH). This exoirols Poet survey was administered in the seven largest metropoli- The assessment of the Prba: progrm by Wodon and tan areas of Colormbia. June waves of the EN include Minowa (199 combines da from the 1993940 rouads special modues on infrmality, thus allowing the identi- of National Urban Employment Survy (ENEW and fication of workers who are covered by mandatoy sever- from a survey administered rhly at the same time to ance pay reulations (until 1990) or included in the indi- Ptvbecat trainees. The atter was desied to match the vidual accounts progrm (afterward). Given the nature of questions in the ENEU, so hat information the two the policy hange in 1990, the results obtained refer to 1 surveys is comparable. The two samples a combined by the effets of replacing seveace pay with individual Wodon and Mnowa (1999) usng matching methods. accounts, not to the specific efcts of the ltter. | Peru Mandatory Sevrwne Brzis Unemplmnt Insorance The study on the effects of mandatory severance pay in Brazil's unemploytmt insurnce program is studied by Peru by Macsaac and Rama (2000) uses panel data fiom Cunningham (200 using diaa fom a sample of males I rhe Living Standards Measurement Surve (LSMS) of and females of workin age who left a nonagricultural 1994 and 1997. The panel nature of the data allows con- job, spent at least one month unemployed, and found a trol fbr unobservable difeences across individuals. new job. This sample is drawn foim the Pesqsisa Naonal faIsaac and Rama combine several job indicators (years deAmosr deDomiio (PNAD) in all years from 1992 to of service, written conttact, enroIlment with social secu- 1997, excpt 1994, when there was no sumey. Each r4ity size of the establishment, and presence of a union) to round of the PNAD srey has over 360,000 observa- evalae whether a worker will get severane pay in the tions. Despite being quite restrictive, the criteria used by f event of unjustified dismissal, The survey also conais Cuningham led to a pooW sample of more than 24,000 infomation on earnings at the indivdual level, and con- individuals. The increase in the generosity of unemploy- I sunption at the household level. Saavedr and Torero ment insurance benefits that took place in 1994 serves as f (1999) analyze the impact of mandatory severance pay on a natural experimen, allowing a difference-in-differences turnover using data on job tenure fiom the Encue evaluation of the effects of participation in the program. possible to argue that differences in employment, earnings, acteristics of the workers (for example, talent) which are or consumption are not due to participation in a specific correlated with program participation. For these reasons income support program, but rather to unobservable char- the evaluations should not be considered definitive assess- 94 HELPING WORKERS DIAL WITH THF RISK OF UlNEMPLOYMENT ments. But the studies commissioned for this report, and not being covered by the individual accounts program some other recent papers, take advantage of multiple (Kugler 2000). observations for the same workers, or of changes in the reg- The relatively high coverage of the programs among ulatory framework, to identify some of the effects of the those at work is in sharp contrast with the relatively low programs. number of beneficiaries among the unemployed, as shown by the second column in Table 6.3. The interpretation of Who Is Covered by these Programs? the figures in this column is not straightforward because All five income support programs cover, in principle, a they compare a flow (the number of beneficiaries in a year) considerable portion of the labor force. The highest cover- with a stock (the average number of unemployed during age corresponds to Argentina's public works program, the same year). In countries where movements in and out because anyone willing to take a job at the prevailing wage of unemployment are frequent, as in Mexico, the total rate is supposedly allowed to do it. In practlce, however, number of people who are unemployed at any point over the coverage of the program is determined by the resources the year could be several times higher than the average available. Coverage is slightly lower for Mexico's training number of unemployed. Taking this into account, it would program, because eligibility rules apply. The first column be safe to conclude that at best, no more than 1 out of every in Table 6.3 shows that the other three programs reach a 10 unemployed workers benefits from the income support smaller but still sizable share of the labor force. The share programs considered. appears to be much lower in the case of Peru's mandatory There are several reasons for the discrepancy between severance pay program, but this is due mainly to the way relatively high legal coverage and relatively low actual cov- the denominator is defined: while coverage rates for Brazil erage. One of them is self-selection. Public works programs and Colombia refer only to workers in the largest urban pay low wages, offer little continuity, and require hard centers, the coverage rate for Peru refers to all private sec- work, so that only the neediest among the unemployed tor workers, including those in agriculture. The relatively apply. Because training programs are not as physically high coverage of the programs is consistent with casual evi- demanding, and have the potential to increase skills, they dence that workers do not opt out of them when given the could attract a larger number of candidates. The other choice. In Colombia, for example, as of 1995 only 1.5 per- three programs could lead to self-selection by employers. cent of workers in manufacturing and 0.6 percent of work- The latter should have no interest in extending contracts ers in commerce preferred a higher salary in exchange for beyond the probationary period, hence incurring addi- IABLE 6.i Income Support Programs for the Unemployed: Beneficiaries Across Population Groups WORKERS ACTUAL LEGALLY BENPFICIARIES PROGRAM AND COVERED BY AS PERCENT OF SHARE OF BENEFICIARIES IN POPULATION GROUP (N) COUNTRY THE PROGRAM UNEMPLOYED Poorest 2nd poorest Middle 2nd richest Richest Public works in Argentina In principle, all 7.5 78.6 15.3 3.5 2.1 0.4 Training in Mexico Eligible by age, education 29.4 69.9 15.5 8.1 5.0 1.5 Severance pay in Peru 21.2% of all private sector 3.6 4.7 9.5 28.6 33.,3 23.8 Unemployment insurance in Brazil 39.6% of urban workers 11.8 10.6 24.6 19.1 25.1 13.6 Individual accounts in Columbia 47.29% of urban workers Unknown 0.0 4.3 Not applicable 19.1 76.6 Noe: Coverage is based on legal enttltement. The unemployed include fiest-time job seeker-. Data foe Argentina are from Jalan and Raval-itn (1999) and Jones and Ravallion (1999) and refer to 1997: groups are population quinetles according to household earnings per capita, excluding benefits paid by the program. Data for Mexico are from Wodon and Mi,orwa ( 1999). Data for Peru are from Maclsaac and Rama (21010) and refer to 1994, private sector workers include farmers, the self-employed, and unpasd family workers; groups are sample quintiles according to household consumption per capita Data for Brazil are frem Cunningham (2000); coverage figures are for 1997; groups are defined based on earn- ings in last job, as of 1992, with cutoff points at 1, 2, 3. and 5 minimum wages. Data fot Colombia are from Kugler (20(0)0), and refer to 1992-96; groups are workers quartiles according to income in last )ob. 95 SECURING OU1R FLrUTRE IN A GLOBAL ECONOMY tional costs in terms of contributions or severance pay- programs reviewed in this paper have a higher ratio of ben- ments for workers they may not want to retain. efit to nonbenefit expenses. Another potential explanation for the discrepancy Income transfers are financed in two different ways. In between legal and actual coverage is the weakness of the case of Argentina's public works program and Mexico's enforcement capabilities. In the case of Peru, MacIsaac and training program, the funding comes mainly from the Rama (2000) construct a coverage score that combines four budget, hence from general taxation. In the other three criteria, in addition to legal entitlement: having a written cases, the transfers are funded by explicit or implicit taxes contract, being enrolled in social security, working in a on employment. In principle, taxes on employment could firm where at least one trade union operates, and working be either more or less distortionary than general taxation, in a large firm. It can be assumed that the likelihood of depending on the nature of the tax system in force. But all actually getting severance pay in the event of dismissal taxes carry a marginal burden, in the sense that they reduce increases with the number of criteria met by the worker. If economic efficiency. This burden should be factored in only those workers who meet at least one of these criteria when assessing programs such as Argentina's Trabajar or do get severance in practice, the coverage rate falls from Mexico's Probecat. More specifically, the value of the physi- 21.1 percent to 9.3 percent of the labor force. It drops to cal or human assets generated by the programs should not 5.2 percent if workers have to meet any two of the criteria. be high enough just to cover their cost, but also to cover In the case of Brazil, Cunningham (2000) reports that a the corresponding marginal tax burden. The last column in significant portion of the unemployed is entitled to unem- Table 6.4 suggests that in practice this is not the case. ployment benefits, but does not collect them. This could The employment tax is formally similar in the Brazil- be due to the lack of social security offices nearby. ian, Colombian, and Peruvian income support programs. The actual beneficiaries of income support programs In particular, contributing 9.3 percent of the workers' tend to be relatively wealthy, with Mexico's Probecat and salary to an individual accounts program amounts to set- Argentina's public works program the exception. Trahajar ting aside roughly one month of salary per year of work. is a poverty alleviation program targeted through unem- Therefore, in the mid-1990s the employment "taxes" in ployment, rather than an income support program for the the Colombian and Peruvian income support programs unemployed. In all of the other programs for which the were roughly the same. Some of the proponents of the indi- information is available, the poorest population group has vidual accounts system claim that an income support pro- the smallest number of beneficiaries. Beneficiaries tend to gram along the Colombian lines is less burdensome to be more numerous among middle- or upper-middle income employers than a program along the Peruvian lines. How- groups. Colombia's individual account program is the least ever, Peruvian employers could set resources aside on a pro-poor; more than three quarters of the beneficiaries can monthly basis, if they wished to. It is not at all obvious be found among the richest quarter of the urban population. that forcing them to do so would make them better off. However, savings accounts that are administered by third How Much Do the Programs Cost? parties (not employers or workers) do facilitate labor The average income transfer received by the beneficiaries of mobility, reduce legal claims, and provide workers with these programs ranges from roughly US$300 in Mexico to more certain benefits. US$1,300 in Argentina. The spending figures reported for A potentially more important difference between these two countries in the first column of Table 6.4 are income support programs relying on an employment tax is higher, because they also include other costs of the pro- related to the endogenous adjustment of wages. If a pro- grams. In the case of Argentina, only one-third of each dol- gram is valued by the workers, they should be willing to lar spent is paid to laborers, with the other two-thirds "pay" for it through lower wages. In principle, the net going to materials ancl skilled personnel in charge of the impact of a program on wages depends on its explicit or activities supported by the program (Ravallion 1999a). implicit employment tax, on how much the workers The ratio increases to roughly three-quarters in the case of "value" the benefits from the program, and on the wage Mexico, with the other quarter going to trainers' salaries elasticity of labor demand and supply. This net impact was and other related expenses. The other three income support evaluated using panel data estimates for Peru by Mac Isaac 96 HELPING WORKERS DEAL WITH THE RISK OF tlNEMPLOYMENT TABLE 6.4 Income Support Programs for the Unemployed: Cost per Beneficiary BURDEN ON EMPLOYERS AVERAGE SPENDING PER CHANGE IN VALUE OF ASSETS PROGRAM AND BENEFICIARY CONTRIBUTIONS EQUILIBRIUM BURDEN ON GENERATED PER COUNTRY (IN U.S. DOLLARS) OR PAYMENTS WAGE (IN %) TAXPAYERS BENEFICIARY Public works in Argentina 3,100 None Not applicable All Similar to spending Training in Mexico 393 None Not applicable All Insignificant Severance pay in Peru 760 1 monthly wage per Insignificant None Not applicable year, lump sum Unemployment insurance 664 0.65% of firm's Unknown None (the system Not applicable in Brazil revenue, monthly runs a surplus) Individual accounts in Unknown 9.3% of workers' Insignificant None Not applicable Colombia wage, monthly Nrte: Data foe Argentina are from Ravallion (1999a), Jalan and Ravallion (1999), and Jones and Ravallion (1999), and refer to 1997. Cost data for Mexico were kindly provided by Quentin Wodon; the assessment of the value of the assets is from Wodon and Minowa (1999). Data for Peru are from Maclsaac and Rama (2000): payment by employer refers to 1994, whereas change in equilibrium wage was eseimated on 1994-97 panel data. Data for Brazil are from Cunningham (2000); spending refers to 1995. Data for Colombia are from Kugler (2000): the change in the equilibrium wage was estimated using differences in differences for 199(1 and 1992-96 and Rama (2000), and using difference-in-differences esti- when they were substantially weakened. But there was mators for Colombia by Kugler (2000). In both cases, the mandatory severance pay in both periods, whereas in the net impact turned out to be statistically insignificant, meantime the economy was subject to many other eco- implying that the burden of these two programs falls nomic reforms and external shocks. In the case of Colom- entirely on employers. bia, the shorter unemployment spells reported by Kugler (2000) after 1990 could not be due to a change in the What Are the Efficiency Effects and Insurance Benefits? amount of the transfers received, or in the conditions Proposals to introduce "optimal" unemployment insurance attached to them. The only difference is that in the new or individual savings accounts often emphasize the distor- system the beneficiaries can keep the unused portion of tions to incentives created by conventional unemployment their transfer in their individual accounts, whereas in the insurance. Key among those distortions is the lower effort old system they would have had to put that portion into a devoted to a job search by those who collect unemploy- bank account, or found some other form of investment for ment benefits. On the other hand, it can be argued that it. It is difficult to believe that the difference in returns income support allows the unemployed to search for a between these two alternatives is large enough to justify a longer period, possibly leading to a better job match. More difference of three weeks in the duration of unemployment generally, income support programs for the unemployed spells. The same reasoning casts doubt on the allegedly could have effects on the duration of unemployment spells, higher earnings observed upon reemployment. on the earnings level subsequent to reemployment, and One of the few clear-cut results in Table 6.5 refers to even on nonpecuniary characteristics of the new jobs. The where people get jobs after unemployment ends. In the evidence available in this respect is limited. However, the case of Brazil, Cunningham (2000) finds that unemploy- results summarized in Table 6.5 suggest that all these ment beneficiaries are more likely to become self-employed incentive effects are weak. than nonbeneficiaries. This result is consistent with credit Some of the evidence on unemployment spells is diffi- rationing at the household level. Under this hypothesis, cult to interpret. In the case of Peru, the allegedly longer unemployment benefits would provide start-up capital, unemployment spells are derived from an analysis of and the most profitable use of this capital would be to run changes in job tenure over time. Saavedra and Torero an independent business. This choice would not be avail- (1999) show that job tenure was longer in the 1980s, when able to those who do not receive unemployment benefits. job security regulations were in force, than in the 1990s, This result is also consistent with the view that working in 97 SECURING OUR FUTURL IN A GLOBAL FC(ONOMY TABLE 6.5 Income Support Programs for the Unemployed: Effects on Employment, Eamings, and Consumption EFFECTS ON EFFECTS ON EFFECTS ON CONSIJMPTION INCOME DURATION OF EARNINGS ON SECTOR OF RELATIVE ro UJNEMPIOYMENT REEMPLOYMENT REEMPLOYMENT NONPARTICIPIANTS Public works in Argentina Unknown Unknown Unknown 25.9c higher income Training in Mexico Insignificant Insignificant Unknown Unknown Severance pay in Peru Longer Unknown Unknown Consumption rose 6.8% for beneficiar-ies; fell 16.9% for nonrbeneficiaries Unemployment insurance Insignificant Insignificant Self-employment Unknowi in Brazil more likelv Individual accounts Three weeks shorter 5.5% higher Unknown Unknown in Colombia Noote: Information for Argentina is from Jalan and Ravallion (1999), based on 1997 data on household income per capita. Information on Mexico is from Wadon and Miscwa (1999.) INfarmarian an Pero is from Ma,csaac and Ra-a (20011) far consumpesan. suing data for 1994-97. and Soa.odra and Torres- (1999) fo, athere; anmplovoe s1el Is are assme-d rob longer because the average job tenure was longer in the late 198Os, when job security regulations were in force, than in the late 199()s. Information oB Brazil is trDm CLsnriogham (21)()()). Information an Columbia is from Kogl-e (211011); un-mplopy-nc spells are shoter compared to chase ,f sorke-s -neirled to sooeroac pay, bhocnc nec-sa-ily -cp-.arad te chose who are nor covered by ar, income support program. the informal sector is not necessarily an inferior outcome, market changes in the region, we offer some summary but often a deliberate choice. observations specifically for the five programs surveyed Credit constraints at the household level could also above. First, nonlabor costs in Argentina's Trabajar appear underlie the apparent effectiveness of income support pro- to be high, so there is room for improving the effectiveness grams at providing insurance, which is suggested by the of public works programs as instruments for income trans- evidence from Argentina and Peru. If households could fers by lowering the non wage component, though this borrow when one of their members is confronted with tem- may jeopardize the quality and nature of the investments porary job loss, they should not experience a serious drop being made through the program (see also the following in consumption. In Peru, consumption per capita drops by section). Second, the training in Mexico's Prohecat seems more than 16 percent when one household member ineffective, so there may be potential saving if the share becomes unemployed and does not get severance; on the spent on training costs is reduced or redirected. Third, sev- other hand, consumption per capita increases by almost 7 erance pay in Peru appears to be excessively generous. percent if the unemployed member gets severance. This Fourth, individual savings accounts in Colombia seem to result suggests that the Peruvian program mandates sever- be used mainly by the wealthy, who are more likely to have ance payments that are too generous. Jalan and Ravallion voluntary savings anyway Fifth, Brazil's unemployment (1999) show that the foregone income from participating insurance scheme covers largely those who also have indi- in Argentina's Trabajar program amounts to only half of vidual severance accounts, thus providing insurance- the transfer received.7 For the average household, partici- although quite frugal for most workers-to those who also pation results in an increase of almost 26 percent in income benefit from mandatory severance laws. Finally, these pro- per capita, a figure quite close to the 23 percent difference grams cover little more than 10 percent of the unem- in consumption per capita between beneficiaries and non- ployed, implying that by themselves they fail to offer most beneficiaries in Peru. Unfortunately, there are no similar workers any insurance against job loss. estimates available for the other three income support pro- grams considered. Policy Implications There Is an increasing clamor for greater unemployment What Are the Aain Weaknesses of these Programs? insurance in the region. There is also a widely held view, Bcfore drawing the policy implications of these assess- however, that given the nature of labor markets and the ments in the broader context of macroeconomic and labor extent of administrative capacity in developing countries, 98 HfELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT they should hesitate before setting up formal unemploy- ment self-protection, or lower the probability of occurrence ment insurance systems. In fact, extensive informality of crises or shocks. Chapter 4 discussed the monetary, fis- results in even greater problems in administering benefit cal, and capital market policies that will help lower the payments because it is difficult for the Ul agency to deter- frequency and size of economic crises, including unem- mine if claimants are in fact unemployed: many may be ployment of workers. This section discusses another impor- working in the unregulated sector while drawing benefits. tant set of public policy measures in this regard: labor mar- It also creates problems in financing the Ul system because ket reforms. it will be impossible to make a large number of employers Each country has a unique social consensus on the desir- and employees pay their contributions. The recent experi- able balance between economic efficiency and social equity, ence of Argentina is symptomatic of both problems (see for and labor policies straddle both concerns. Different eco- example, Mazza 1999). These considerations have led nomic and political histories of countries can result in sig- observers to argue that middle-income countries in East nificant differences between the impact of seemingly iden- Asia and Latin America would be better off reforming their tical laws on wages and employment, so that the subset of mandated severance payment schemes rather than institut- binding laws-and hence the labor reform agenda-is ing formal Ul systems (see, for example, Edwards and country specific (see Box 6.3). Here we simply note that Manning 1999). Variants of Brazil's individualized sever- labor market reform has lagged behind other economic ance accounts (Fundo Garantia por Tempo de Servicio [FGTS]) reforms in most countries in the region, with only a few are sometimes recommended as a substitute for unemploy- exceptions such as Chile. In fact, it has been described as ment insurance. the "forgotten" economic reform in LAC (Edwards and Using the findings summarized in the last section and Lustig 1997), or a major comnponent of an unfinished the conceptual framework outlined in Chapter 3, this sec- agenda of 'second generation reforms" in the region tion reevaluates these and related propositions. This reeval- (Guasch 1999a). For governments that wish to facilitate uation is done not with the objective of recommending comprehensive insurance decisions by their workers and specific changes in existing mechanisms for income sup- households in a rapidly changing global economy, labor port for the unemployed-though there may be some market reform should get high priority. lessons to be learned-but to contribute to the general dis- cussion that will gather steam as countries in the region Self-insurance: Individual Savings Accounts reassess whether the mechanisms they employ are indeed Programs where a specified part of a worker's salary is set appropriate for the types of product, labor, and financial aside in an account-generally held in a government- markets they have, and the types of aggregate and micro- approved financial institution, sometimes with guaranteed economic risks faced by workers and firms. In particular, rates of interest as in the case of Brazil's Fundo Garantia por we provide some tools and techniques to policymakers to Tempo de Servicio-are a form of forced self-insurance. Since determine the suitability and adequacy of the alternative governments cannot credibly state that they will not "res- programs of income support for the unemployed by asking cue" people who did not save enough, people may not save the questions: Do individuals have effective instruments of enough on their own-hence the need to make the pro- comprehensive insurance against the risk of unemploy- gram compulsory (Coate 1995). The main drawback of ment, that is, market insurance, self-insurance, and self- these programs is that they do not involve pooling of protection? If not, are government programs providing the unemployment risks, and hence lead to lower consumption missing instruments? and investment by households than traditional unemploy- ment insurance programs that are more "pay-as-you-go" in Self-Protection: The State of Labor Markets nature (see Gill, Haindl, Montenegro, and Sapelli 2000 for The logical first step in examining whether enough is more on this distinction). While this may not lead to being done to help workers deal with unemployment reduced welfare for wealthier households, poorer workers shocks is to determine if more can be done to reduce the would suffer more. Their main strengths are that they min- probability of these shocks. In the terminology of Chapter imize disincentive effects on work that are unavoidable in 3, this constitutes the set of government actions to aug- programs that involve pooling, make severance benefits 99 SECURING OUR FUTURE IN A GLOBAL ECONOMY l ' ' r--- -'- -' '' -'-^-' '' - '' - '-- ''- ---'' > - {'-'' " '' >-'--'..''.''. .-'-. .--.''.---.-..-'. - - ---'-- --''--'---'--'-' -' - e- ---'-- - ----^^--. .'''-'-''. ' - - - ' -''- - -- -'- BOX 6.3 Labor Markets: Latin America's "Forgotten Refonnm" This box describes the four main lessons of a recent study Labor Outcomes Depd only Party on Labor Polies (Gill and Montenegro 2000) that quanitatively assesses Labor reform is neither necessary nor sufficient for the main labor market issues in Argentina, Brazil, and improving labor outcomes. As the experiences of Chile. Afgentina, Brazil, and Chile show, successfil stabiliza- E tion unaccompanied by changes in labor policy in general Labor Poli issues are Country Speiof improves labor outcomes, though it also unmasks micro- Ironically, the first general lesson is that regional gener- economic imbalances (public-private compensation dif- alizations abou0t labor policy are often pointless or mis- ferentials in Brazil, relative prices of labor and capital in leading. There are no shortcuts: implementable labor Argentina, and the gaps between the rich and poor in policies must be designed by understanding labor mar- Chile). Similarly, fiscal adjustment col also improve kets country by country. For examle, there is evidence employment and earnings outcomes, illustrating that that a large number of workers are paid exactly the legal actions such as putting government finances in order can minimum wage even in Brazil's "unregulated" sector, and improve outcomes in the private labor market, even if no adjustments in this wage are matched by salary adjust- labor reforms tke place. Further, moving fiom a period iments. It is possibleand indeed has been confirmed- of adjustmnent to sustained economic growth will that in some other countries minimum wages are not mprove earnings and employment outcomes, even if binding even in the regulated sector; but it would be there are no accompanying improvements in labor poli- unwise to assert either that mininum wage legislation is cies, However, this does not mean that outcomes coldd not imnportant for economic outcomes or that iQt is imnpor- not be better still if appropriate labor reforms are made, I tant for all or most countries. Again, the finding that sus- though it is harder to make this case persuasively. tained growth in Chile has not been associated with increasing earnings inequality does not guarantee similar Making Labor Pocies Better is Dfilt During I results for other countriesi it merely weakens claims by Good Times I people who question the desirability of growth-orienrted The first corollary of the above is that it is difficult to i labor policies instituted during the 1980s in Chile. carry out labor reform3s during economic booms. In I more certain for workers, and lower transactions costs. of government (because market insurance would be avail- These considerations present the challenge of balancing able only at a high price given that the chances of a worker labor market efficiency and poverty concerns. being unemployed are high). The framework set forth in Chapter 3 can help resolve While considerable variation in unemployment rates this dilemma. Following that framework, programs con- exists among LAC economies, the regionwide average was taining a significant mandatory self-insurance component estimated to be about 10 percent in 1999.8 The likelihood may be especially well suited for countries where workers of becoming unemployed is anything but rare in most face high risk, for example, those that have high average countries, therefore, especially for the young, the less edu- rates of unemployment. For workers in these countries-if cated, and women, as seen in Chapter 5. For these reasons, there were no public interventions to help the unem- mandated self-insurance may be well suited for countries ployed-comprehensive insurance against this risk would such as Argentina and Brazil that have not carried out entail a relatively large component of self-insurance rela- comprehensive economic-especially labor market- tive to market insurance. Mandating such self-insurance reforms. By the same token, they may be less suited for ensures time consistency of government behavior, but does countries such as Chile, where the risks of unemployment not impose forms of insurance on individuals that would be have been lowered through far-reaching economic reforms purchased only in relatively small amounts in the absence (see below). But even countries such as Argentina should 100 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT Argentina, for example, labor policies did not rise to the of the likely effects of policy refbrm can help to advance top of the reform agenda during 1990-94 even though the reforms. Labor reform is always difficult and-when E unemployment was increasing steadily, because of attempted-reforms are usually piecemeal. In every improving wage and employment conditions as a result country, there are well-entrenched labor interests, politi- of successful stabilization. When the Tequila Crisis hit cal risks for reformers are high, and proponents of reform Argentina, unemployment skyrocketed and labor such as employers or economists are often ineffective in reform came to the forefront of discussions, only to convincing people of the benefits of taking risks. To help recede again when economic growth resumed as fiscal policymakers focus their efforts and explain them to the and financial reforms advanced. Again, in Brazil, labor electorate, it is important to know whether labor policy reforms only briefly dominated the political landscape changes are necessary and, if so, which aspects of labor when unemployment rates rose to historic levels in policy are binding, which are irrelevant, and what are the mid- 1 998 after the economy slid into a recession. When likely benefits. Quantification of the benefits of labor pol- the economic slowdown proved to be less severe than icy reform-rather than relying only on economic anticipated and unemployment rates fell, labor policy growth-shows that Argentina probably has paid a high reform was moved off the Brazilian government's list of price for not reforming labor legislation between 1995 priorities. Finally, a decade of sustained growth and and 2000. For Brazil, again using a quantitative improving employment and earnings outcomes in Chile approach, the aspects of labor legislation that are the had the effect of prompting the reversal of labor reforms most important for outcomes can be isolated. Quantify- that may have made these outcomes possible in the first ing the concept of "precariousness" helps to determine place. that Chile may be better off leaving labor policies largely unaltered, and looking to other policy measures such as Quantifying Key Magnitudes can Facilitate Labor improved education quality to narrow the income gap Reform between the rich and the pOOr, The second corollary of the finding that labor outcomes b only partly depend on labor policies is that quantification not turn their backs to pooled-risk insurance schemes: because the loss of employment is often too large a shock, pooling offers immediate poverty-related advantages and, or may affect a large fraction of the population at the same over the longer term as labor reforms progress and unem- time. Morduch (1999) argues that informal insurance, for ployment rates fall, it becomes more and more preferred to example, through reciprocal transfers within the extended self-insurance (see Box 3.1). Guasch (1999b) proposes a family or community, may thus be the least effective when program that could address both short- and long-term con- help is most needed. Private, market-based arrangements siderations for countries embarking on labor reforms. A may fail because of informational and incentive problems study at the World Bank proposes to again address the involved in insuring against even rare risks, toward which suitability of mandated self-insurance as income support individuals and households have a rational inclination, ver- for unemployment (Vodopovic 2000). sus self-insurance (the "price" of which is the same for rare and frequent losses, and self-protection, for which expendi- "Market-Type" Insurance: Risk-Pooling Programs tures may yield only small payoff when the probability of While precise conditions for the introduction of public the bad state occurring is already low; see Chapter 3). programs are difficult to pinpoint, there are advantages of While concerns that publicly provided insurance will formal, public programs in addressing unemployment risk. displace some private rtansfers are justified (see, for exam- Informal insurance mechanisms may not be effective, ple, Cox and Jimenez 1998), its introduction may improve 101 SECURING OUR FUTURE IN A GLOBAL ECONOMY efficient because they can pool resources across larger BOX 6.4 Combining Mandated Self-insurance and Mar*et-Type . goups. Unemployment Insurance Severance Pay Provisions as Insurance for Guasch (1 999b) recommends combining individual Unemployment savings accounts-which have the weakness that they The most common form of public unemployment insur- involve no pooling of risk and are simply fbrced self- ance in most of Latin America has been mandatory sever- insurance-and conventional uremployment itisur- ance pay provisions, such as those evaluated in depth by ance-which runs the risk of abuse in economies with MacIsaac and Rama (2000) for Peru. Though not generally high rates of formal unemployment combined with I associated with "market insurance" that involves pooling high shares of informal employmett, of risks, these programs in fact do pool risks to provide The main characteristics of the proposed program. insurance in the event of unemployment (with the are: employer and/or all workers paying an "insurance pre- E* ach employee is assigned an individlual account int mium" through reduced salary and benefits while an eligible financial institution of his or her choice, employed). The problem is that because the employer is * Each month, the worker and employer dits a responsible for severance pay, the pooling is at the level of fixed fraction of wages inro the account. the firm, and hence the risk is spread over only a small * The money is invested in finamcial securities but group. In the old economic environment protected by with strong safeguards agalist loss of prtincip l trade barriers, the risks were effectively pooled over a value, f greater population because consumers effectively subsi- e The account is fully portable in the event of job sep- dized potentially bankrupt firms through higher prices. aration or rtirement. With globalization and reduced barriers to trade, however, * Access to finds is permitted only in the eve of I this is no longer possible because products must be sold at uinemployment or retiremnent, and monthy with- world prices. drawals are limited to a fraction of last monthly If this scheme had proved to be administratively salary. uncomplicated relative to other options, there might still * A part of the worker's/employer's contribution goes be a reason to recommend the use of severance pay provi- into a general fum-ministerd at the firin, sec- sions. But, as pointed out in Rama and Maloney (2000), tor, or economywide level-to comtpement the most of the grievances handled by labor courts in Latin accounts of workers who may not h rechd America are in fact related to disputes over severance pay. amounts that would allow cettain minimum Employment adjustments needed for economic reasons are amounts when separated firm work, rendered complicated, and workers are deterred from seek- * Minimum monthly withdrawals fbr a maximum ing better job matches voluntarily. These mandates may stipulated period are guaranteed only fbr certain therefore be the worst among public "market-type insur- types of workers (for example, heads of househod, ance" programs: they involve high moral hazard with little * firing with just cause should include dismissals by pooling of risks, and may discourage hiring in the first the employer due to economic reasons, and legisla- place and hence raise the risk of unemployment for those tion should be made clear and simple. looking for jobs. They exist in most countries in the region even today; these countries may be well advised to seri- ously reevaluate the suitability of these mandates in rela- both welfare and efficiency. For example, in poorer coun- tion to other alternatives for ensuring income support for tries the beneficiaries of private transfers are often the the unemployed. elderly, and keeping more income for themselves would enable the young-who may also be poor-to invest more Public Works Programs as Insurance for Unemployment in their own education and health and that of their chil- Again, though generally not regarded as "market-type" dren (Morduch 1999). Public systems may also be more insurance that involves pooling of risks and the charging of 102 HELPING WORKERS DEAL WITH TIHE RISK OF UNEMPLOYMENT premiums, public works programs of the type analyzed cost per transfer and in the cost per job created, with no above can in principle be treated as such.9 The question obvious implications for policy. In other words, these pro- addressed here is whether these programs have fulfilled this grams appear to aim at a combination of objectives- role in the LAC region and whether there are any lessons income smoothing, employment per se, provision of infra- for the future. structure-which makes difficult the comparison of The main strength of these programs is that if properly workfare to other income support programs. Training pro- designed as a "work guarantee" (low wages, no rationing, grams, such as Probecat, can be seen as a special case of such low nonlabor costs-see Ravallion 1999) they serve effec- programs where the investment is now in human capital tively as unemployment insurance for those who formerly rather than infrastructure. Since matcrials costs are low, the were employed (in formal or informal jobs) and for house- rate of transfer is very high. holds the coping strategy of which is for family members to begin working when the main earner becomes unem- "Conventional" Unemployment Insurance ployed."0 The experience in the region and outside shows One of the more attractive features of a well-designed that these programs are able to target the poor when unemployment insurance program is that it can simultane- designed specifically for this purpose. ously help offset (part of) both microeconomic and macro- Both Ravallion (2000) and Snyder and Yackovlev economic fluctuations. In a study of the political economy (2000) confirm that some leakage to the nonpoor makes for aspects of social insurance and transfer programs in the resilience in social programs during economic contractions. U.S. and Latin America, Snyder and Yackovlev (2000) con- But the results for Argentina's Trabajar suggest the pro- duct cross-section, time-series analysis on 45 program- gram was clearly subject to the same constraints in the groups in the U.S. during 1962-98. Parr of the analysis political economy that influenced the incidence of past fis- focuses explicitly on what happens during the economic cal contractions in Argentina. The program expanded into cycle and major political changes, contrasting spending poor areas when the budget increased, but it retreated from levels just prior to recessions with the levels during reces- poor areas when the budget was cut. It was the program's sions. One of the findings is that the only class of programs disbursements to nonpoor areas that were protected."1 which show a clearly countercyclical pattern of spending Further, as Maloney (2000) argues, there is a question are those classified as "income security programs" (for about the cost-effectiveness of these programs and, even example, social security, family assistance, food stamps, more fundamentally, the proper means to evaluate them. and unemployment insurance). Most of the other social Measured against other income protection programs con- protection functions exhibit no clear pattern, and some sidered here which seek primarily to transfer income to show evidence of procyclicality. Overall, spending on social households experiencing shocks, the emphasis on employ- protection program-groups appears countercyclical, but ment through infrastructure projects means that a large the most countercyclical program in the U.S., by far, is fraction of the funds earmarked for income protection may unemployment insurance.`' be diverted to materials and capital costs. In noncrisis Most observers would probably not find this surpris- periods, these projects may be socially valuable when eval- ing-as unemployment rises during a recession, unem- uated at the market rate of discount. During crises, how- ployment insurance expenditures should rise as well. But it ever, when poor families facing credit constraints strongly is not unreasonable to expect that in the fiscally con- discount the future, they represent a diversion of resources strained atmosphere of a recession, unemployment benefits away from present income transfers that is socially might be cut or eligibility constrained, so as not to "bust costly.12 the budget." Evidently, this does not occur in the U.S. In this regard, Chile and Mexico appear to place a high Instead, a 1 percent increase in unemployment leads to value on the transfer and less on the investment per se, so somewhat more than a 1 percent increase in unemploy- that they reach rates of transfer close to 70 percent (see ment spending. This "automatic stabilizer" function makes Wodon 2000). Argentina and Colombia seem to value the it worthwhile to examine unemployment insurance pro- project component more, so they transfer 40 percent or grams more closely, especially that of the U.S., which has less. This implies large differences across countries in the several other attractive features as well."4 103 SECURING OUR FUTURE IN A GLOBAL ECONOMY Traditional unemployment insurance is usually financed design. The U.S. federal government, for example, pays through conitributions by employers and/or employees, state governments for administration costs. Countries such though government subsidies-either to cover deficits or as France, Germany, and the U.K. also pay the administra- to fund programs such as means-tested unemployment tion costs of UI programs, while not subsidizing payments assistance-are usually significant. Table 6.7 shows the to UI claimants. Another exception may be the payment extent of burden-sharing in the financing of UI (Gill, Dar, by government of social security dues on beha]f of the and Thomas 1999, based on data reported in Tzannatos unemployed during the period he or she is eligible to and Roddis 1998). In only 10 countries is the burden receive unemployment insurance benefits-for example, in entirely borne by just one party. Brazil, where only Germany and Portugal-though in principle this could employers contribute to Ul financing, is the notable exam- also be financed from UJ contributions. It is easier to jus- ple of such countries in LAC. Burden-sharing between two tify government subsidies, for assistance, to those among parties is more prevalent, especially between employers the unemployed who are poor, determined through reliable and employees. In 13 countries, including Argentina, means-testing. Under such a system, the governmint pays Ecuador, and Venezuela, the burden is shared by both. In for modest benefits for those where the main eligibility cri- Argentina, the UI program is financed by a 1.5 percent teria are not proof of past contributions and current invol- payroll tax on employers and a 1 percent tax on workers' untary unemployment (as required for unemployment wages (Mazza 1999). The most prevalent means of financ- insurance), but proof of current poverty (which is funda- ing Ul systems is through contributions by all three par- mental) and of current involuntary unemployment (which ties, with 38 countries financing their UI system through is secondary). Governments in Austria, Finland, France, these means. In 17 of these 38 countries, the role of the and the U.K. finance unemployment assistance along these government is limited to just paying off any deficit.15 lines.'6 Under most scenarios, it is difficult to justify large-scale Table 6.6 provides information on who among workers permanent subsidization of unemployment insurance pro- or employers is legally responsible for contributing to the grams by the government. The insurance aspect of the scheme on behalf of workers. Who actually pays-In more scheme implies that benefits should be paid from contri- general economic terms-is a far more complicated ques- butions. One exception is, perhaps, the cost of administer- tion, the answer to which depends on the design of the pro- ing UI programs; it is reasonable to expect the government gram and the relative market power of workers and to pay the administration costs of a system that it has set employers, both of which are country-specific considera- up, especially if it wishes to significantly influence its tions (see Box 6.5). The relative market power of employ- TABLE 66 Costs of Unemployment Insurance: Burden-Sharing Among Workers, Employers, and Government PAID BY OECD LATIN AMERICA AND OTHER TOTAL NUMBER Worker only None None 0 Employer only Iceland Bangladesh, Brazil, Moldova 4 Government only Australia, New Zealand Chile, Estonia, Hong Kong, Tunisia 6 Employer and government Iraly Bulgaria, China, Georgia, Russia 5 Employer and worker Canada, Greece, Netherlands, Algeria, Argettina, Barbados, Ecuador, Hungary, Serbia 13 Sweden, U.S. and Monrenegro, South Africa, Venezuela Employer and/or worker; Belgium, Denmark, Ireland, Albania, Armenia, Azerbaijan, Belarus, Czech Rep., Egypt, 17 government pays any deficit Norway Iran, Lithuania, Poland, Romania, Turkmenistan, Ukraine, Uzbekistan All three; goverrinrent's Austria, Finland, France, Croatia, Cyprus, Guernsey, Israel, Kyrgyscan, Latvia, 21 contribution is nonresiduai Germany, Japan, Luxembourg, Liechtenstein, Malta, Slovak Rep, Slovenia, Uruguay Portugal, Spaii, Switzerland, U.K. Nate: Turkey and Mexico are the onlv OECD countries wichout formal unemployment insurance systems as defined in this paper. Source: Tzannatos and Roddis (1998), using data reported in Shanil Securiy Systers Throughout the World-1997, published by the U.S. Social Security Administration. 104 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT BOX 6.5 Who Really Pays for Unemployment Insurance? Who bears how much of the burden of contributions to FIGURE 6.1 formal unemployment insurance depends on two factors: Workers Pay for Ul even Though the Contributions are Levied on (a) the extent to which the design of the UI program dif- Films fers from what workers would have chosen for themselves as insurance against unemployment; and (b) to the extent Labor that some differences exist, Ul contributions will be Supply viewed as a tax, the incidence of which will depend on the demand and supply elasticities of labor (see Gill, Dar, and Thomas 1999). w -------------- If the government-sponsored Ul program is exactly WI --- --- -- --- --- --- . Labor what every worker would have chosen by themselves, w' then the cost will be borne entirely by the worker (who e will willingly accept a wage net of all UI contributions). If-as is likely-the government program is different Employment from the insurance against unemployment the worker would buy privately, the tax burden will in general be done if the contributions had been raised directly from shared by the worker and the employer. The higher the their paychecks. elasticity of demand for labor, other things being the Figure 6.2 shows the outcome of an employer contri- same, the larger the share of the tax borne by the worker. bution if the supply of labor is more sensitive to the The higher the elasticity of supply of labor (or the ease of FIGURE 6.2 becoming informal), the higher the share of the tax borne Workers and Finns Share Ul Costs even Though Contribtdions by the employer. Who actually pays (that is, bears the are Levied on Finns incidence) therefore depends not on whom the tax is wage levied but the design of the scheme and the relative mar- ket power of the worker vis-a-vis the employer. \ Figures 6.1 and 6.2 illustrate this point. Figure 6.1 Sup shows the effect of imposing a UI contribution on employers in a labor market in which the supply of labor wo ..... ...... is highly inelastic. This has the effect of reducing the w2 ----- labor demand of firms at any given wage (the demand Labor curve shifts down by the amount of the employer contri- bution). In the new equilibrium, after the imposition of the Ur contribution, workers receive a wage which is Ernployment lower by the amount of the contribution w1 (as opposed to the previous equilibrium of w ); employers end up wage. Now the incidence is shared. Rather than falling paying the same gross wage as they did before. In other all the way from w0 to wI, the supply response pulls the words, the fact that the contribution is nominally raised wage back up to w2. Workers still "pay" for a part (w.- on employers makes no difference to the outcome, which w2) of their UI contributions by receiving lower wages, is that workers end up paying, just as they would have while employers pay the difference (w2-w). ers and workers is not something the government should of central importance to government. In this regard, the try to influence, but the design of the program should be key question is how well the program's design "mimics the 105 SECtURING OLUR FUTURE IN A GLOBAL ECONOMY TABLE 6.7 Income Support Programs for the Unemployed: Summary of Findings and Policy ImplicaUons NATURE OF MEASURE INSTRUMENT ADVANTAGES AND DISADVANTAGES POLICY IMPLICATION Individual savings Self-insurance-no pooling of risk Low labor market efficiency costs, but welfare Should be considered by countr-ies that accounts reduction especially for poorer workers. have high unemployment, especially where labor reforms are only a distant possibility. Severance pay Pooling over small group- Almost no advantage. Little pooling of risk, Possibly the worst form of unernploy- globalization makes group even entails labor market inefficiency, makes labor ment support in a globalized economy. smaller relations contentious, and is administratively challenging. Public works and Market-type insurance Can reach informal sector workers and the Should be considered for a part of work training programs elements -implicit pooling of poor, bue can entail high leakages in the form force, but not a oiniversal scheme. Perma- risk of nonlabor costs when investment element is nent schemes allow for better balance made a priority. Training programs show less between consumption smoothing and leakage but also lower coverage potential, investment over the economic cycle. Unemployment Market-type insurance-explicit Most pooling of risk, can be used both to Should be considered by governments insurance pooling of risks address idiosyncratic and aggregate risk, and that have carried out compreherisive eco- hence, serve as a "automatic fiscal stabilizer." nomic reforms; labor market disincentivc Generally politically popular. May be admin- effects can be reduced by keeping benefits istratively demanding. frugal and "mimicking the marker" as much as possible in design. market." One of the principal features of private insurance stands out as a rare government program that is strongly markets is that the price reflects the degree of risk, even if countercyclical in nature. imperfectly. This is a noteworthy feature of the U.S. sys- There are several ways to move from the current systems tem, perhaps the only one that tries to match unemploy- of mandated severance pay to such unemployment systems. ment tax rates to risk through employer "experience-rat- Coloma (1996) proposes an unemployment insurance sys- ing," where rates of tax vary according to the frequency tem for Chile-which has severance pay provisions but is with which an employer's former workers have filed for considering the introduction of a new system-that uses unemployment benefits.17 Mimicking the market, and severance pay benefits as a "deductible." Under this pro- more particularly, the insurance that workers would choose posal, the unemployed would first have to draw down the to buy, is also essential to the long-run goal of covering the accumulated severance benefits, and only then have access informal sector. If wages fall to reflect the cost of insurance to unemployment insurance payments. The effort to make that workers do not want, then they have the incentive to the hybrid system resemble the structure of private insur- avoid the implicit "tax" and become informal (see Maloney ance schemes-where the insured are not paid the full loss 1998). but the loss minus a deductible-makes the proposal an The U.S. unemployment insurance program may be a improvement over a system of mandated severance pay. good model for LAC countries that arc considering such Following the line of reasoning developed in Chapter 3, systems: there are minimal mandated severance pay rules countries that have low unemployment risks because of under the general labor laws, the system mimics the mar- comprehensive economic reforms and strong information ket as much as countrywide public systems can, benefit systems, (for example, strong administrative data and reg- level and duration are modest, the rules are relatively uni- ular households surveys), have the "insurance fundlamen- form throughout the country even though states collect tals" that make for moves toward unemployment sapport taxes and pay benefits, and the design of the program in systems that pool risk to be welfare-increasing for its citi- general makes it an automatic fiscal stabilizer. While this zens, even when some efficiency losses are involved. For role may not be large in quantitative terms, the program countries such as Brazil, that already have minimalistic 1 06 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT unemployment insurance systems but may or may not have framework presented earlier in the report, do provide for these preconditions, Box 6.6 provides some tentative some broad but potentially useful guidelines. guidelines. Table 6.7 presents these findings, which draw from both theoretical considerations and the regional experience. They Conclusion should be viewed as a starting point for closer policy analy- This chapter summarized the experience in five countries sis from a country-specific perspective. With this objective, with five types of income support programs for the unem- the table casts each available policy instrument in the gen- ployed, and then drew some policy lessons. While it is dif- eral framework of Chapter 3 and notes their respective ficult to determine an unambiguous ranking of these pro- advantages and disadvantages. Based on this, the rightmost grams independent of country-specific circumstances, these column of the table outlines the conditions under which experiences, combined with the comprehensive insurance each instrument may represent an appropriate policy choice. BOX 6.6 Some Guidelines for Countdes that Have Unemployment Systems International experience appears to suggest the following made on grounds of administrative ease for both govern- lessons for LAG economies that currently have unem- ment and employers. Ideally, the base should be the same ! ployment insurance (Ul) systems, but which cover only a or similar to that used for other taxes collected. Thus, if small fraction of the workforce. the system is administered by states, the tax base should It does not appear necessary to extend the UI tax to be the same as that used for other state taxes. If the sys- iworkers. Having a levy only on employers is in line with rem is federally administered, there may be grounds to international experience in the OECD and middle-income make the tax base the same as that used for the main fed- countries. Besides, economic theory suggests that the final eral tax on employers, usually the social security tax. burden of UI taxes has less to do with who bears the ini- Experience-rating of employers can be a reasonable tial impact and more with the design of the scheme and medium-term goal. While experience-rating has many the elasticities of demand and supply of labor. attractive features, it is administratively demanding. Keep the role of government in Ulfinance minimal. Before making any decisions in this regard, countries in Governments should pay only for the costs of adminis- the region would be well advised to seek technical assis- tration of the UI system. This is in line with interna- tance from experts in the U.S., especially UI administra- tional experience, and in keeping with the principle of tors from states that have relatively recently and success- employers and employees together insuring workers fully instituted experience-rating. against drastic income loss during unemployment. Government financing of unemployment assistance Frugality of benefits should be maintained. Because could he considered. Most countries in the region have a of the high degree of informality, it is difficult for the UI high degree of informal employment, often synonymous agency to determine if claimants are in fact unemployed. with noncompliance with social security laws. The region Keeping a waiting time of about 30 days before benefits confronts the challenge of extending income support to commence and the benefit levels low-as Brazil has suc- those in the informal sector as economies are opened up to ceeded in doing so far-circumvents this problem some- the rigors of international competition. One option could what and also reduces work-related disincentives associ- be for the government to finance a system of unemploy- ated with all UI systems. Financing a more generous UI ment assistance with low, uniform benefits to those who system will also be difficult because it will be impossible are currently unemployed and who satisfy a means test, do to make a large number of employers and employees pay *ot qualify for unemployment insurance because their their contributions, employers have not paid UI taxes, but who can prove that Decisions on the tax base sbould be made on admin- they have contributed their social security dues for the istrative grounds. The decision on the tax base should be same length of time as required for UI eligibility. 107 SECURING OUR FUTURE IN A GLOBAL ECONOMY However, a major conclusion of this chapter is that in 1999a and 1999b; Wodon and Minowa, 1999), the literature is designing an effective strategy to help workers deal with much more scant. the risk of unemployment, administrative capacity should 4. The chapter also draws from a series of independently pro- duced studies dealing with one income support program in be an imoratutntverdigocenfgoe Argentina (Ravallion 1999a, 1999b, and 1999c; Jalan and Ravallion ment. Most countries in the region are capable of building 1999; Jones and Ravallion 1999). this capacity. The more important questions are whether 5. Sections 2 and 3 of this chapter are based on Rama and Mal- there are government actions that can rapidly lower the oney (2000). risk of unemployment, and what are the type of unem- 6. In 1999 the monthly minimum salary was less than $100. ployment support programs that are in demand but the 7. It follows that the public works program is not merely dis- supply of which is constrained. These questions can be placing other, roughly equivalent ways of generating income. Raval- posed as follows: lion does assume, however, that jobs displaced are lost and not taken posed as follows: by other unemployed. Relaxing this assumption would lead to * What are the measures that would augment the self- higher estimates of benefits to the target population. protection efforts of workers-that is, reforms that 8. Countries such as Argentina have rates as high as 15 percent, lower the risks of unemployment? Labor policy mea- while those in Chile are less than half these rates. sures are a leading candidate, though their impor- 9. Programs with large transfer elements can be treated in an insurance setting as market-type insurance with a negative loading tancet and nature hostpravetobe deid red by eachgontr factor, that is, those for which the price of insurance is better than *What are the most practical measures to auglment acurlyfi. actuarially fair. market-type insurance involving pooling of unemploy- 10. Montenegro and Gill (2000) and others find that in the 1980s ment risks? These measures invariably imply effi- a large share of the participants of Chile's Minimum Employment ciency costs but suit countries better and better as they Program were formerly our of the labor force. successfully implement self-protection-augmenting 11. Given the low wage rate offered, the direct benefits from the policies that lower the risk of unemployment. program are still likely to have favored the poor, even after the cuts. • What are the most practical measures to augment Thus, the design features of the program undoubtedly helped protect * What are the most ptactical measures to augment the poor from cuts. se/f insurance efforts of workers? These mneasures gen- 12. This diversion may be justified by appeals to the dignity of erally minimize labor market distortions and best work, or avoiding the adverse social consequences of mass unem- suit countries where self-protection augmenting ploymenr, although such considerations tend to receive less discus- efforts are underway, but may require that special sion when programs affecting middle-class workers are discussed. attention be paid to poorer workers. From the political economy perspective, governments may also be more willing to finance investment projects than simply transfers, or While the relative weight on each of the three measures even workfare projects that have low materials costs and are essen- will differ across the economies in the region, a sound pol- rially recurrent in nature (maintenance, cleaning, repairing). In this icy mix involves pursuing all three objectives so that work- case, benefits to the unemployed may be higher when packaged in an ers obtain not a full guarantee against all shocks, but com- investment project than simply as a transfer. prehensive insurance that allows them to seize the 13. Another important finding in Snyder and Yackovlev (2000) is opportunities presented by globalization, and to see that that there is a difference between programs that are targeted at the risk is a fact of life in a world that grows smaller every year. poor and those-such as unemployment insurance-that are nor; that is, spending on programs that are targeted at the poor is much more sensitive to party control in Congress than spending on non- Notes targeted programs. 1. Argentina, Barbados, Brazil, Ecuador, Uruguay, and Venezuela 14. Gill, Dar, and Thomas (1999) summarize the features rhat have unemployment insurance programs, though some of these are lead to this strong countercyclicality as (a) stringent legal restrictions quite limited in scope. that unemployment tax proceeds can be used only for paying unem- 2. Chile is currently debating the merits and demerits of intro- ploymenr benefits, (b) established rules by which the federal govern- ducing individual saving accounts for dealing with the risk of unem- ment provides loans to states whose UI trust funds are drying up, ployment. and (c) rules that the federal government uses its own UI trust fund 3. Only for public works programs was there a substantial litera- to extend unemployment benefits during long recessions-the max- ture available (Ravallion, 1990; Ravallion, Datt and Chaudhuri, imum duration can be doubled to 26 weeks. 1993; Dart and Ravallion, 1994). For other income support pro- 15. While in some countries the government may directly con- grams, and with a few exceptions (Cunningham, 1997; Kugler, tribute into the UI fund (for example, Israel, Japan, and Malta), 108 HELPING WORKERS DEAL WITH THE RISK OF UNEMPLOYMENT more often this may involve the government's financing of means- ance is time of contribution, while eligibility for assistance is based tested unemployment assistance programs (for example, Austria, on a means test that qualifies the recipient as needy. Finland, France, and the U.K.), social insurance programs (for 17. Note again that, given the design of the program, who actu- example, Germany and Portugal), or even active labor market or ally pays the contribution does not depend on whom the tax is social assistance programs (for example, Latvia and the Slovak levied. In essence, the U.S. system transfers the problem of deter- Republic). mining unemployment risk onto the employer. Thus employees who 16. The main difference between unemployment insurance and have filed for UI benefits more frequently are likely to be less attrac- unemployment assistance is that the eligibility condition for insur- tive to future employers. 109 CHAPTER 7 Helping Poor Households Deal Better with Economic Crises G IVEN THE FINDINGS IN CHAPTER 5 ON HOW HOUSEHOLDS RESPOND TO ECONOMIC CRISES, designing minimalistic and.effective interventions to help households-especially poor fam- ilies-deal better with crises is not easy. While more investigation is needed to confirm the findings of panel studies in Argentina, Brazil, El Salvador, and Mexico, what we found appears to suggest three fundamental points. • First, the poor, like the rest of the population, are reluctant to take actions that are not in their own long- term interest-such as withdrawing children from school during short or moderate downturns. But they have to draw down their assets like everyone else and-since reserve labor is a primary asset of such house- holds-greater labor force participation of secondary workers (mainly women) in the household is observed. Since this takes time away from household production, these changes are likely to affect the quality of education and health. * Second, steeper or longer downturns do have negative effects even on education enrollment and basic healthcare decisions. For example, children may be withdrawn from school or attendance reduced, and the incidence of child labor increases. That is, both quantity and quality of schooling and healthcare are reduced: as shocks become more serious, "good coping" appears to give way to "bad coping" as assets are exhausted. * Third, the poor do gain from economic growth episodes-in fact, good times appear to be more benefi- cial for the poor than the non-poor. These findings should influence policy choices of any gov- While the first two findings involve government actions ernment concerned with sustainable poverty reduction. Gov- that are treated in Chapter 4 (macroeconomic, financial sec- ernments should not be reluctant to carry out growth- tor, and capital markets policies) and Chapter 6 (labor pol- enhancing liberalization and reform that may mean icy), the third-spending on social services-is the focus of somewhat greater volatility during the transition, because this chapter. Protecting the quality of selected social ser- while short downturns may not hurt the poor much, vices that the poor need during economic crises is a difficult increased growth helps the poor a lot. Further, the findings task for even a determined government. This generally suggest that macroeconomic policies should be oriented not involves maintaining the level of spending per poor person to avoid downtums at all costs, but to prevent them from during economic downturns, which is doubly challenging becoming long or deep. In addition, the quality of social pro- because the fiscal envelope shrinks at the same time that the grams used by the poor should be smoothed over the cycle- number of poor increases. This is where the appeal of pro- protecting the quantity and quality of public education and grams that are well targeted to the poor is highest: even if health services used by the poor is critical in both long and governments cannot maintain social spending per poor per- short downturns. son at their normal levels, the adverse effects of the down- 111 SECURING OUR FUTURE IN A GLOBAL ECONOMY turn may be reduced if a subset of this spending that is used Briefly, what we find for governments contrasts with our mainly by the poor is maintained or even increased. findings in Chapter 5 on how households respond to eco- This chapter examines whether governments in Latin nomic risk. While poor households in LAC tend not to rely America have maintained social spending over the economic on "bad coping" over the economic cycle, for example, by cycle, distinguishing as much as possible between "general" sharply cutting investments in the education of their chil- social spending and its more targeted components. Two dren during downturns, governments in the region do studies of cyclical fluctuations in government spending in behave in ways that are shortsighted by sharply increasing Latin American countries commissioned for this report form spending when times are good, and decreasing critical the core of this section, but the section also uses more investments such as in education and health when times detailed examinations of public education in Chile, and are bad. This report provides some conjectures as why this health insurance in Argentina, Brazil, Chile, and Colombia. may be so, and suggests policies that can help make gov- We briefly study the characteristics of five poverty-targeted ernment behavior conform to the principles of effective programs in Brazil, Honduras, Mexico, and Nicaragua that insurance. aim to reduce current and future poverty by giving cash Before we discuss how governments in the region have transfers conditioned on health and education decisions by tried to help the poor deal with economic shocks, it is use- recipient households. Given the concerns that the poor may ful to briefly discuss what sound insurance principles reduce education and health investments when their income would require of governments. Figure 7.1 presents an ideal falls, these are considered prime candidates for the type of scenario where the targeted social spending per poor per- programs that should be protected or even expanded during son increases steadily or acyclically at the long-term rate of economic downturns. However, their suitability as an effec- growth of per capita income (which is subject to cyclical tive instrument for countering cyclical fluctuations in fluctuations). This implies, however, that the share of tar- income and human capital investments is not self-evident geted spending to total government spending or GDP from either their design or their track record. We evaluate must be strongly anticyclical. Maintaining a noncyclical whether they can in fact serve this function well, and pro- pattern of targeted social spending is a tall order for even pose some policy lessons based on our findings in the light the most pro-poor and determined government, but it may of evidence on how households respond, bringing in politi- be a good measure against which governments can judge cal economy considerations. their own performance. This chapter will show that gov- FIGURE 7 1 Targeted Social Spending Over the Economic Cycle Per Capita Income (Economic Cycle) - - ~~~~~~Targeted Spend ing per Poor Person (A(yclical) Share of Targeted Spending in Total Government Spending , ...-- ... --...... (Anticyclical) 1 1 2 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES TABLE 7.1 Targeted Public Spending per Poor Person in Argentina and Mexico, 1994-96 ARGENTINA MEXICO SHARE OF SHARE OF REAL TARGETED TARGETED REAL TARGETED TARGETED PER SOCIAL NUMBER SPENDING PER SOCIAL NUMBER SPENDING CAPITA SPENDING POVERTY OF POOR PER POOR CAPITA SPENDING POVERTY OF POOR PER POOR GDP IN GDP RATE PEOPLE PERSON GDP IN GDP RATE PEOPLE PERSON (1994 = 100) (%) (%) (MILLION) (1994=11)0) (1994 = 101)) (%1 (N) (MILLION) (1994=100) Level Level 1994 100 1.24 21.6 7.5 100 1994 100 1.36 46.95 42.04 100 1995 94.68 1.21 27.2 9.6 63.12 1996 95.07 1.23 60.93 56.51 67.30 Change Change 1994-95 -5.32% -27.88% 1994-96 -4.93% -23.70% Soarce: Hicks and Wodon (2001)). ernments in the region have found it hard both to keep tar- more closely how social spending by government varies geted spending per poor person from falling during bad over the economic cycle. Of special concern is how public economic times, and to keep it from rising "too fast" dur- spending that is targeted to the poor is affected by expan- ing good times. sions and recessions (see Box 7.1). Another commissioned paper, by Snyder and Yackovlev (2000), examines these fac- Social Spending Over Economic Cycles tors for the U S. and for Latin American countries within a For social spending that is targeted toward the poor to political economy framework. reduce the losses incurred by them because of an economic downturn (that is, for it to act as a complement to self- How Do Governments Vary Spending Over the insurance and a substitute for missing market insurance), Economic Cycle? governments should design social spending to be counter- In this section, we first summarize the findings for seven cyclical. In fact, we generally observe the opposite in Latin LAC countries regarding how the elasticity to growth of America. Table 7.1 summarizes the findings for both targeted spending for the poor behaves during booms and Argentina and Mexico during the 1994-96 recession in busts. The countries are Argentina, Chile, Costa Rica, the these two countries. As one might expect, the natural effect Dominican Republic, Honduras, Mexico, and Panama. of the recession is to lower GDP per capita and to increase They were chosen because data available were more reliable the headcount poverty ratio, as was observed in both cases. for them than for other LAC economies. The data cover the To increase social spending per poor person, the govern- 1980s and 1990s, generally between 1981 and 1997-98. ment should have therefore considerably increased its share The data are not refined enough to test whether govern- of GDP in targeted spending. Instead, that share fell in ments are "pro-poor" in the sense required by Wodon and both countries. The result was that targeted spending per others (2000) (see Box 7.2). Assuming that social spending poor person declined by about 28 percent in Argentina and (for example, on education and health) is more pro-poor 24 percent in Mexico (Hicks and Wodon 1999) during the than nonsocial spending, testing whether governments economic crisis. have been pro-poor can then be done using social spending We also recognize that the pattern observed by Hicks as a proxy for targeted spending. The share of social spend- and Wodon (1999) is almost certainly not the result of ing in total spending is found to increase during booms, ignorance on the part of governments. There are obviously and is not reduced during busts (see Table 7.2). This sug- some factors-both economic and political constraints- gests that these governments are "prosocial," that is, they that prevent them from following policies that are so obvi- make special efforts to protect social expenditures. ously in the interests of their citizens. In a paper commis- This should help to protect the poor during a crisis, but sioned for this report, Wodon and others (2000) analyze it is not enough. Despite efforts by government to main- 113 SECURING OUR FUTURE IN A GLOBAL ECONOMY How Do PoPoor ommonts Viry Spendig Over the EcooMic C According to Wodon aad others (2000) gvermments are T assess how gowth affects how much targeted pub- "pro-poor" if the groh elsicity of targeted public lic spentding reaches each poor person, this can be trans- I spending is at leas I during booms, and smler thao 1 formed to yield: during recessions. This asymmetry between booms and ,Sp SP busts is tested empiricatly using pael data on pubic DLog(-.--) Do -); expenditumres for seven Latin American countries. The DP GDP GD P , , DLog(-) DLqg( tesults suggest r governments are opoor or at least N N N tend to vary social spending over the cycle to conform to That is, the growth elasticity of targeted spending per this rule. But this is not enough to protect the poor dur- poor person is I pus the growth elasticity of the share of ing a recession. The reason is that during a downturn, the targeted spending in per capita GDP, ninus the growth economy (aid government spending) contracts at the elasticity of poverty. same time that the number of poor increases. To increase targeted spending per poor person dutiWng a To more formally understand why, denote the tota cisis, the lefthand side of the equation should be negative. targetd spendig for the poor by the government by SP, Wodon and others (2000) estimae the elasticity of poverty the headcount index of poverty by H, and the size of the reduction to growth to be rminus 1, that is, a 1 perent total population by N, so that the tareted public spend- increase in per capita GDP reduces the number of poor by inig per poor person is SPI(H*N). This can be expressed 1 percent. T maintain tagWed public spending per poor as a function of thee parameters: ( the targeted bud- person constant, theefr, the growth elasticity of the share getary spning as a share of GMP denotd by SPIGP of targeted sending in GDP must be less than -2. This is (b) the level of GDP per capita, denotd by GDN and a difficut task for any government. For the seven Latin (c) the inverse of the headcount index of poverty, denoted American countries studied, dhe observed elasticity during by 1/H: recessions is not statistically different from zero, so that a - :$P am 1 XP ] 1 percentage point negaive growth dAces targeted social IH * GP N s spenig per poor peron by 2 percentage points. tain targeted and social spending constant as a share of total The Importance of Political Factors spending during a crisis, a 1 percentage point decrease in The procyclicality of social spending in Latin America is GDP still reduces targeted public spending per poor person also confirmed by Snyder and Yackovlev (2000) using data by about 2 percentage points during a recession. Half of over roughly the same period as that analyzed by Wodon this impact (1 percentage point) is due to the reduction in and others (2000), but for 19 countries in the LAC region. GDP, which leads to reduced total spending even when the While they do not distinguish between targeted and gen- share of targeted spending in GDP remains constant. The eral social spending, they examine the influence of political other half is due to the increase in the number of poor peo- factors as well as economic cycles. One of their main find- pie due to the economic contraction.' ings is that while both authoritarian and democratic gov- Another test of the pro-poor tendency of governments is ernments behave similarly during recessions (cutting social to see how they spend during times when budgets are not spending per capita), the behavior during good times as tight, for example, during economic expansions. For appears to be more pro-poor under democracy (see Box 7.3 these seven countries, the elasticity of social spending to and Table 7.3). They also find that the relatively nontar- GDP growth is found to be larger than 2. Thus, in the geted parts of social spending (for example, higher educa- most general terms, governments expand spending on tion) tend to be more procyclical than those that help the social programs twice as fast as overall budgets during poor more (primary and secondary education), which is periods of economic growth. encouraging. 114 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES BOX 7.2 Data Sources and Classification into Targeted and Social Spending Although Wodon and others (2000) carefully construct side program targeted according to malnutrition rates and the data using both the General Financial Statistics of the providing cash stipends for nutrition and school enroll- International Monetary Fund and country-level data pro- ment-see following section), some general subsidies, and vided by governments, social spending information is expenditures for a number of smaller programs. more reliable than data for targeted spending. Targeted spending is not strictly comparable between countries Mexico because while some countries classify certain programs as The data are for the federal government only but, since targeted, others may classify similar programs as social, the decentralization process in Mexico started only in but not targeted. Social spending is more reliable because 1998, this is not of major concern. Social spending con- it includes all targeted expenditures as well as spending sists of spending for education, health and social security, for education and health (plus a few small items). labor, regional development, water, environment, and social assistance. Targeted spending is the sunm of social Argentina assistance and spending for labor, which includes pro- Annual GDP and budgetary data for 1980-97 are from grams such as Empleo Temporal (public works in rural the Ministry of the Economy. Apart from information on areas) and Probecat (job training in urban areas), rnen- total spending, the Ministry provides consistent series for tioned earlier. Health spending includes social security social spending (education, health, water and sanitation, expenditures, so that health as a share of social expendi- j social assistance, labor, housing, and other services for tures is overstated. urban areas excluding those expenditures allocated in the social security budget). Within social spending, the data Chile, Costa Rica, the Dominican Republic, and identify targeted spending as consisting of spending for Panama housing and urban development, social assistance, and The expenditure data for these four countries are from the I labor. This includes Trabajar, the public works program IMPFs Government Financial Statistics, combining the discussed in Chapter 6. The data include spending at fed- series for consolidated central government, state or provin- eral, provincial, and municipal levels. Health excludes cial govermnents, and local government where available. health expenditures allocated in the social security budget. When data on transfer payments from the central govern- ment to other levels of govemment are available, these 1 Honduras were added to the consolidated expenditures in education, The data are from the Ministry of Finance and are not health, and targeted spending. Targeted spending was cal- available in published form. The expenditures are for the culated by subtracting social security spending from I central government (but the level of decentralization is low "Social Security and Welfare" expenditures. This yields an in Honduras). Targeted expenditures exclude the expendi- approximate measure of spending that is targeted because rtures for the social investment fund, but include all other countries do, for example, grant noncontributory pensions expenditures directed specifically at the poor. Programs targeted at the poor. These countries were selected because included in targeted expenditures are PRAP II (a demand- of data quality considerations. Overall, the results suggest that governments do make spending per poor person falls despite their efforts. And efforts to protect the poor-or at least to protect social equally worrisome is that government behavior in expan- expenditures during crises-and that they increase these sions may be pro-poor but shortsighted-democratic gov- expenditures faster than economic growth during periods ernments expand too fast, perhaps responding to strong of expansions. Unfortunately, the findings also indicate political pressures to "make up" for their inadequacies that their efforts during contractions are not enough- during recessions. The finding in Chapter 5 that the poor 115 SECURING OUR FUTURE IN A GLOBAL ECONOMY TABLE 7.2 Elasticities of Spending to Growth, by Type of Spending ELASTICITY WITH RESPECT TO GROWTH OF TARGETED SPENDING NONTARGETED SPENDING EDUCATION SPENDING HEALTH SPENDING Targeted spending/GDP Overall 0.75* 0.31 0.35 0.24 In expansions 1.06* 0.55* 0.43* 0.55* In contractions 0.44 0.07 0.27 -0.04 Total spending/GDP In expansions -0.08 -0.08 -0.02 -0.07 In contractions 0.o4 0.04 0.23 0.18 Social/total spending In expansions 0.69* 0.69* 0.74* 0.75* In contractions 0.07 0.07 0.13 0.08 Targeted/social spending In expansions 0.46 -0.06 -0.29 -0.14 In contractions 0.32 -0.05 -0.09 -0.30* * Denotes significant at 10 percent level of significance or better. Otherwise the coefficient should be interpreted as zero elasticity. Note: These are elasticities of shares. A zero growth elasticity of the ratio of total spending to GDP implies that spending increases in proportion to GDP Source: Wodon and others (2000). register strong income gains during growth episodes also The Quality of Social Services Over the Cycle means that governments help them most when they least These findings suggest that the quality of social services need the help. consumed by the poor should be even more procyclical iB£)X 7.3 Social Spending Over Economic and Polital Cyles I Ltin merca Snyder and Yakovlev (2000) Conduct cross-seti not statistically diferent om zero-that is, per time-series (panel) regressions fbr 19 Latin AmericAn cata socia security spending is not prcyclical. contries firom 1980-96, for eight sendi variables * Breakig down education speig into three (totai social spending-consisting of social security edu- bad e riesmar secondary, an higher cation, heath, and housin g-n education pending at education-Snyder and Yakkovlev (2000) find primaryf secondary and tertiary levels). The independet icome elasticities of about I for prtmary and sec- variales are growth of GDP (curnxat and lagged), gov- ondary educaton, but a noticeably higher elasticity ernment deficit (lagged), regime type, and governing of about L. fbr higher education spending, which party "ideolo " The maii results ale is the least taqgeted of these categories. * The income elasticity of overall per capita social * Authoritarian and. democratic regimes appear to spending with respect to GDP is cleay positive, respon similarly to economic crises. Both cut but less than 1. social spending per capita, and about equally. But For the four broad categories, educatio%, health, there is a large difference by regime type on spend- housing, and social security they find that the ing changes during expansions: geater increases in income elasticity of spending on educatiom and spending take place under democratic true. Social bealth is about 1; the elasticity is also 1 ft rhousin spending increases only when theme Is both democ- but it is imprcisey estimaed. Interestingl ifr ratic rule and a nonshrinking economy. i soci security-which is probably ls tatgeted than Thee appears to be little effect of the executive public education and health care-the elasticity is bratnch's "ideology" or populist leanings. 116 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES TABLE 7.3 Changes in Latin American Social Protection Spending, 1970-95 (Broad Spending Categories) DEPENDENT VARIABLE = PERCENT CHANGE IN PER CAPITA SPENDING TOTAL SPENDING 4 BROAD CATEGORIES 4 BROAD CATEGORIES %A in per capita GDP .73* - - %A in per capita GDP x social security - .24 .25 %A in per capita GDP x education - .90*** %A in per capita GDP x health - .97*** %A in per capita GDP x housing -1 60* %A in per capita GDP X not soc. sec. - -1.07* Lagged deficit _.01* -._01* -l* New democratic regime .05*** .09** .08** Old dcmocratic regime .04 .06 .o6 # of observations 226 835 835 * Significant at the .10 level. ** = Significant at the .05 level. * Significant at the .01 level. Nste: Country-specific, fixed effects included In all specifications. Source: Snyder and Yackovlev (2000). than social spending because government spending on forces cuts in targeted programs intended to benefit disad- social services such as education and healthcare is cut at the vantaged students. same time that private capacity to pay for them declines. While the issue of cyclical fluctuations in education But there are mechanisms specific to education and health- quality requires much more study, it appears that during a care that may offset some of these effects. In education, for downturn there is a negative effect on the quality of educa- example, governments may reallocate spending from tion for the middle and upper-middle classes, who generally higher education toward primary and secondary education send their children to private subsidized and unsubsidized during downturns-the previous section discusses some schools. For poor children things may be even worse: they evidence that suggests this. In healthcare, the reforms in use public schools which may be even more vulnerable to countries that have strengthened health insurance for those expenditure cuts, and they benefit from special public pro- employed, as well as others, may provide some relief dur- grams that are threatened as well. The only group whose ing economic cycles. education quality may be unaffected by cyclical fluctuations There is no systematic evidence on this question of how is the wealthy. Economic volatility may thus, through its quality of social services varies with aggregate economic effects on government spending, make it harder to narrow shocks. Two studies commissioned for this report, Mizala the educational gap between the rich and poor. and Romaguera (2000) and Jack (2000), address this issue This is especially unfortunate because education has for public education and health, respectively, but the been found to be related to the ability of workers and fam- results should be regarded as preliminary. Mizala and ilies to withstand aggregate shocks (see Chapter 5). Many Romaguera (2000) approach the question by studying countries in the region have chosen to not redistribute changes in the quality of educational outcomes in Chile in assets such as land, focusing instead on improving the dis- the mid-1990s. They find that the quality of educational tribution of human capital assets such as education services, using two standardized school achievement test through aggressive public education initiatives for the scores as proxies, behaves procyclically.2 There are two pos- poor. This is in all likelihood the most sensible policy, but sible explanations. First, a downturn reduces private the rewards will be seen only after some time. Violent incomes for the wealthier households, thus reducing the cycles in public education spending and the quality of edu- demand for places in fee-charging private schools that tra- cation services push the rewards from public-education-as- ditionally have displayed higher educational attainment. redistribution policies even further into the future. Pro- Second, decreased educational spending affects schools, grams such as Mexico's Progresa and Brazil's Bolsa teacher incentives, and other inputs generally, but also Escola-if used as instruments to reduce this cyclicality in 117 SECURING OUR FUTURE IN A GLOBAL ECONOMY education quality-have the attractive feature that they -BO - 4--- .. ...... . l can reduce the amplitude of these quality swings for poor Iolnbias Healthcare Reform families. The usefulness of these programs is evaluated in some detail in the next section. Colombia's health sector reforms initiated in the early In Latin America, the regulatory role of governments is 1990s represent one of the most ambitious policy inter- becoming more complex, especially in two areas: the regula- ventions undertaken in Latin America. Before the tion of financial markets, and the regulation and public pro- reforms, Colombia had a centralized, budget-financed, vision of health services. A critical review of the recent expe- and poorly organized public health delivery system, riences with health insurance reform in Argentina, Brazil, and many infrmal sector workers and their families Chile, and Colombia can be found in Jack (2000). The study were uninsured. The general goal of the Colombian finds that the traditional approach of public health systems reforms was to ensure a basic level of coverage for all in LAC attempted to provide free universal coverage, moti- individuals, that could be inproved upon for those vated more by a concern for equity than for the efficiency of willing and able to pay more. the insurance arrangements available to households. This Although no formal voucher scheme exists, the was, in turn, caused by the highly unequal income distribu- scheme is equivalent to a two-level voucher system. tions prevalent throughout most of the region. Effectively, rnembers of one group of families (those During the 1980s and 1990s a number of governments with workers in the formal sector) receive a voucher for in the region, including the four studied by Jack (2000), insurance that covers a wide range of services at high sought to improve the efficiency of public health provision quality, while all others (many of whom are poor) by relying on or mimicking private insurance mechanisms, receive a voucher for a less generous packae of insur- albeit to varying degrees. In some cases, like that of the ance. Members in the first group are said to be in the obras sociales in Argentina, this was achieved by reforming "contributory regime," and those in the second are the focus of existing institutions. In others, entirely new referred to as participating in the "subsidized regime." institutions were created, such as Chile's Instituciones de The tax base consists of a payroll tax plus general Salud Previsional (ISAPRE). revenes. Participants in the contributory regime are Because health insurance and health care are almost require to pay a 12 percent payroll tax to help finance always integrated, the task of reducing the exposure of indi- health care. This tax is earmarked for health services viduals to health risks is intimately connected to the organi- provision. Participants in the subsidized regime also zation of health care delivery and financing mechanisms. contribute, but these contributions are means-tested. Colombia's health insurance reform appears to have been There has been a marked increase in formal coverage explicitly market-augmenting: the reform aimed at ensuring of the population, particularly among lower-income that those who could pay for coverage-employees in formal igroups. Overall, the proportion of individuals with sector jobs-were guaranteed access to quality healthcare, Iinsurane more than doubled during this period from while those from whom contributions are harder to collect- 24 percent to 57 percent, with the largest proportion- the unemployed, the self-employed and the poor-were ate gains among the poor-the lowest quintile group's guaranteed access to services as well, but of modest quality. coverage rate rose firom about 5 percent to 45 percent. The performance of such "dual-voucher" systems during i . . . aggregate shocks should be studied in greater detail. How- ever, with evidence that health system coverage for the poor On the whole, health insurance functions in LAC are increased from 5 percent to almost 50 percent because of the still covered by a dichotomous system. On the one hand, reform, the new system is almost certainly better for helping most countries now have a private or quasi-private mar- those affected adversely by aggregate fluctuations (see Box ket for actual insurance policies, with explicit premiums, 7.4). This does not rule out cyclical fluctuations in quality as coverage rules and deductibles, which gives access to the relatively high quality unsubsidized subsystem contracts varying degrees of reasonably high-quality services. On and the subsidized subsystem expands during downturns, the other, there remains almost everywhere a large public and vice versa during upturns. or publicly subsidized provider of health services, such as 118 IHELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES Brazil's Sistema Unico de Saude (SUS) or Chile's grams focused on the poor, some countries have tried to FONASA, which is quantity-rationed and provides low- make the programs more self-targeting, for example, by quality care. Reform of these health systems must con- using low-wage work as the targeting device (as in public tinue in most countries. Given the complexity of health works programs in Argentina, Brazil, and Chile). While insurance markets, governments throughout the region LAC followed countries such as India in these second-gen- would be well advised to invest in a greater understand- eration programs, the region has led the world in what can ing of the main design and regulatory principles that be considered the third generation of antipoverty programs need to be set in place. "targeted conditional transfers," which make means-tested cash transfers, but make them conditional on "socially Targeted Spending During Booms and Busts desirable behavior" of recipients. The five programs Marny countries in the region have steadily moved from reviewed in this chapter belong to this class of targeted using general subsidies (especially for food and fuel) as the programs that provide social assistance to poor families major instrument of support to poorer households, to pro- with children, on the condition that these families invest grams aimed at providing income transfers to the poor. in their education and health (Sedlacek, Ilahi, and Gustafs- Facing administrative difficulties in keeping these pro- son-Wright 2000). TABLE 7.4 Main Characteristics of Targeted Conditional Transfers PROGRESA PRAF-I1 RED BOLSA ESCOLA PETI INDICATOR (MEXICO) (HONDURAS) (NICARAGUA) (BRAZIL) (BRAZIL) Implementing Agency Federal National National Municipal/State Federal Objectives Education enrollment Yes Yes Yes Yes increase Health and nutrition Yes Yes Yes Yes improvement Child labor reduction Yes Yes Yes Yes Poverty alleviation Supply-side support Yes Yes Yes No Yes Current coverage 2.3 million families Under preparation Under preparation 200,000 families 131,000 children (1997) (1995) (1996) Size of monthly US$10 per person US$5 per person US$9.3 per person US$32-$65 per family US$12 per person education grant Geographical National National National Municipal National targeting level Beneficiary selection Income means-tested None Under preparation Income means-tested Income means-tested criteria and score Targeting outcome Low leakage, but high Low leakage, but high undercoverage undercoverage Outcomes - Improvements in Yes: enrollment Yes: lower dropout, No evaluation education increases promotion increases - Better health and nutrition Yes - Child Labor Mixed Mixed Suitability for expansion in crisis - Intensive Yes Yes Yes Yes Yes (more for old covered) - Extensive Difficult Difficult Difficult Difficult Difficult (new participants) Sourre: Sedlacek, Ilahi, and Gusrafsson-Wrighr (2000). 119 SECURING OUR FUTURE IN A GLOBAL ECONOMY BOX 7.5 Mexico's PVo Program: Works WellO But Would It Do as Well in C,ises Prors gives cash grants to poor famiies in rural areas househols in non-Pgtsa localities (Schultz 2000). ihe on the condition that their children attend school and inreases in enrollments were largest in the grades in visit heat centers regulady. The stated objective of the whih earollments wer lowest-between completing program is to reduce current and fiure poverty; the lat- elemeary school (grade 6) and starting junior second ary ter by increasing investments in children's human cao- school. These ects imply, fior example, that a 16-year- tal. This demand-side intervention is also accompanied old copleted on average 1.1 more years of schooling by sizable supply-side support in the form of increases in than a poor child in a community without Pesa. The teacher salaries and the supply of medicines. Prors interal rate of etu on Prgea grants is 9.2 percent. began in 1997 and today covers 2.6 million rural fmi- The program also improved health indicators (Gertler flies-about one-tenth of all families in Mexico-at a cost 2000). Clinic visits iPrgma localities were 18 percent of $800 mnilion, or 0.2 percent of GDP. Thee questions higher than in non-Prowa areas, the number of prenant are of primar conern fr this report, Does the program women making their irst visit in the first trimester i target well? Does the program improve child school and increased by about 8 percent, and prenatal care visits health outcomes? And can the program be alered to Increased by 5 percent. Participation lowers the probabil- serve the purpose of a social safety net in a world with ity of illns by 22 perent among children aged 0 to 2. risk? Evaluations of the program being carried out by the International food Policy Research Institute (IFPRJ) in Suitabi in Cnses collaboration with Progesa can provide answers to these While the targeting and outcomes are enicouraging, Pro- questionsm. gsa desip suggests that its ability to serve as an instnment of social insurance might be limited. It is rting usefil here to distinguish between the program's ability Prgs was found to be the most efective of the tareted to expand ntsiel and its ability to expand ext y. progams in Mexico-both in terms of selecting poor It would do better in the former-that is, it might be rel- localities and selecting poor househols within ther. atively simple to increase the amount of benefit distrib- Prgra is not effective, however, when it comnes to dis- ued to households already in the program during perif Iinguishing between localities in the middle of the scale. ods of economic crises so as to continue providig the As Pro exands into less-poor communities, selection incentive to beneficiary failies to keep their children in error is higher. It also did nor do well in selecting mod- school. However, any attempt to expand the program i etately poor households. Thus, as Prors expands into extensively-that is, to include households that experi- less marginal communities, leakages are likely to cor- ence transitory income or employment shocks-would pound at both the locality and the household evel require changing selection methods and criteria. In addi- (Skoufias, Davis, and Behrman 1999). tiotn, a"n major exansion would also necessitate defining exit rules-that is, how fimilies that "improve" after a daon" and H th Outcomes positive shock will be dopped fiom the rolls. Otherwise, Systematic evaluations of Prga have revealed signifi- the program wilU not be financially sustainable over the cant impacts on education and health. tinrolment rtes of long ran. c children in househols in P ea localities are higher t compared to the enrollment rates of childtren in similar AdEeSa gmmaunel 5fias, International Food Policy esearch Institute,I cornpard to te elitlimentrates o childen in imliat and U4i.er of the Progrrs Evaluairin Project. J. Table 7.4 summarizes the principal features. Broadly health, and nutrition (and hence a subsequent reduction in speaking, these programs have three objectives: the allevi- long-term poverty); and the reduction of child labor ation of poverty; improvements in educational attainment, (explicit in some of the programs such as Brazil's Programa 120 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES de Erradicacio do Trabalho Infantil (PETI), and implicit in nerability to risk in the long run-than in providing mar- others such as Mexico's Progresa, Honduras' PRAF II, and ket-type insurance during crises for all or many of those peo- Brazil's Bolsa Escola. The programs are demand-side inter- ple who fall into poverty. As the next section proposes, ventions, with some supply-side support. The largest of however, these programs may have the attributes that these programs is Progresa, which covers more than 2 mil- make transfer programs resilient to cuts over the economic lion households (or about 10 percent of total households in cycle. Political economy considerations may overturn their Mexico). purely economic drawbacks. Regarding their effects on poverty, human capital, and child labor, rigorous evaluations are scarce, but the pro- Designing Economic Policy Under Political grams appear to work well. Progresa has been systematically Constraints studied, though, and appears to have improved education, In examining the role of governments in assisting their health, and nutrition (see Box 7.5). The programs have low citizens in dealing with economic risks, we found three leakage to the non-poor. However, there is also consider- findings of note. First, social spending by LAC govern- able undercoverage of the eligible poor both because the ments is generally highly procyclical: even when the programs are relatively new and expansion has been cau- share of social spending in total budget outlays rises dur- tious, and because of the inevitable fiscal constraints facing ing bad times, total spending shrinks and headcount some of the programs such as Bolsa Escola. poverty increases, so that social spending per poor person Because eligibility requires having school-age children, is procyclical. Second, the spending-which includes the programs will exclude some of the poor even if all eli- most expenditure on social insurance and safety net pro- gible families are covered. This makes it relevant to ask if grams-is often poorly targeted. Third, the quality of it is the behavioral condition that leads to the observed social services-especially education-also behaves pro- gains, or if this is the effect of the income transfer making cyclically. These are not desirable characteristics of poli- the household somewhat better off. In determining this, cies for facilitating comprehensive insurance by individu- however, both administrative and political economy con- als and families. siderations are important. First, the additional condition- Insurance principles require that governments transfer ality may somewhat paradoxically lead to lower adminis- resources from good times to bad-"saving" during good trative costs: the programs anchor the monitoring system times, and "dissaving" during bad, or borrowing during in established schools and clinics, and therefore circum- bad times and repaying loans during expansions. It is clear vent the need for completely new administrative arrange- that this is not what has happened over the last two ments. Second, this conditionality may be key to their decades in Latin America. Whether governments are pre- political popularity, and may make them resilient to cuts vented from doing so by political and economic factors is even when budgets are being cut. important to understand. What is clear is that govern- For the purposes of this report, however, the critical ments seem to treat changes over the economic cycle as question is how well the programs can serve as a safety net permanent-being shortsighted when times are good, and over the economic cycle. They come up somewhat short in engaging in "bad coping" when times are bad by cutting this regard. The programs do not cover families that are down on critical investments such as education and non-poor in good times, but who fall into poverty during health. That is, governments respond to economic shocks uncan- a recession. Thus, while the amount of the cash transfer to nily like the stereotyped responses that poor households allegedly those already covered can be increased quickly when display. Ironically, in Chapter 5 we found that the poor incomes fall (thus being responsive on what can be called actually do not behave as stereotyped-in bad times that the "intensive margin of poverty"), the programs cannot appear to be temporary, (that is, short or mild recessions), by their design cater to the transient poor (and hence are the poor draw upon assets such as reserve labor, and do not unsuited on the 'extensive margin of poverty"). sharply cut investments in health and education. In the terminology of Chapter 3, these targeted condi- The factors that make governments poor practitioners tional transfer programs may therefore be more effective of the most basic insurance rules are worthy of closer in augmenting self-protection-decreasing household vul- study. In proposing policies, this section makes some sum- 121 SECURING OUR FUTURE IN A GLOBAL ECONOMY mary observations on this subject. There are four major decrease in per capita GDP. There are two reasons for the policy implications that follow from the analysis of how failure of targeted public spending to protect the poor. households respond to economic volatility (in Chapter 5) First, when GDP falls, even if targeted spending remains and what governments in the region have done to help. constant as a share of GDP, there will be less money avail- able to distribute to the poor through targeted programs. The Long-Term Goal of Social Policy Must be to Second, when GDP falls, poverty increases, which means Improve the Distribution of Assets that targeted spending for the poor must be distributed to First, since assets are crucial to enable households to self- a larger number of poor people. These two factors com- protect and self-insure against shocks, a better distribu- bined make targeted spending for the poor highly pro- tion of assets should reduce ex post variations and thus cyclical, which leads to a lack of protection during hard improve welfare. Our findings provide additional support times. The same is true for social spending. Our results to the already traditional emphasis on more and better suggest that additional efforts should be made to create education: in addition to the impact it has on income lev- effective countercyclical programs and safety nets to pro- els, education appears to reduce the vulnerability to tect the poor during crises. shocks and enable both rural and urban workers to cope With governments cutting targeted spending per poor better with them. person during economic crises, the finding that during expansions targeted spending per poor person increases by Targeted Programs Should be Permanent and Better more than 2 percentage points for every percentage point Protected During Crises increase in per capita GDP may seem like good news. Second, the temporal profile of social spending-especially However, for several reasons, this finding is not as encour- on targeted programs-needs considerable realignment. aging as it seems. There is evidence that the income of the Targeted social spending accounts for small shares of GDP, poor grows rapidly-generally even faster than that of the but the programs it makes possible can make such a large nonpoor-during growth episodes so that, as a rule, they difference to poor people affected by a negative shock that need government transfers the least in these times. In addi- governments should make an effort to protect them from tion, rapid increases during good times make the subse- the great budgetary pressures which arise during reces- quent cuts in spending during bad times seem much sions, and to design them as much as possible to be auto- worse, and may be politically destabilizing. Moderation in matically countercyclical. Unemployment insurance pro- spending during good times lowers the risk of large reduc- grams, public works guarantees, and poverty-targeted tions in spending during crises-especially if accompanied human development programs can all be designed to have by transferring resources from good times to bad. this property. Other budget items that deserve attention during recessions are those which relate to the quality of International Financial Institutions Can Help selected social services, such as the salaries of teachers and Overcome Political Constraints to Insurance primary healthcare workers, and the maintenance budgets Fourth, for democratic governments, the pressures to spend of the facilities with which they work. during economic recovery can be ignored only by risking loss of political power. Because economic and political Keeping Increases in Social Spending Moderate in Good cycles seldom coincide, it is equally difficult to ensure that Times is Important Too the savings during good times are spent only for the right Third, while the evidence is not definitive, there is enough things (social services and targeted programs) at the right to suggest that it may be as important not to increase social time (during economic crises)-the record of such self- spending during good times as rapidly as countries have, as insurance efforts by governments, such as fiscal stabiliza- it is to protect it during bad times. The empirical evidence tion funds, is patchy at best. for LAC summarized in this chapter indicates that despite Under these political constraints, governments that efforts to restrict cuts in social spending, targeted and gen- have taken appropriate self-protection measures through eral social spending per poor person are reduced during comprehensive reforms should adopt strategies that recessions by 2 percentage points for each percentage point involve a good measure of market insurance. Recall from 122 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES Chapter 3 that at the margin, rarer losses are better insured ties. The presence of these provides an efficiency- through market insurance than self-insurance, and Chapter based reason for government subsidy or direct pro- 4 applied these principles at the level of the country. In the vision. An additional equity-related rationale may absence of a well-developed market for insuring against arise if these activities, in addition to contributing to aggregate risk, the strategy that suggests itself is for gov- self-protection, also increase lifetime earnings. Edu- ernments to borrow during bad times to prop up social cation and health care qualify under this heading spending and repay during good times. The problem, of and, in practice, most "social spending" finances course, is that the private market for countercyclical these services. finance is also thin or nonexistent. Governments that have * Second, the provision of market-type insurance for risks carried out comprehensive economic reforms deserve access where markets may be missing or underdeveloped, to countercyclical finance from multilateral financial insti- and some scope for risk-pooling exists. Unemploy- tutions. For governments that have yet to carry out the ment insurance and public works guarantees are typ- required economic reforms-and face a high likelihood of ical examples. Public health services, either in the crises-the appropriate mechanisms for transferring form of direct provision or of cash subsidies to users resources would be of a more self-insurance nature. Recall or suppliers of private services, are another important from Chapters 3 and 4 that more frequent risks may be category of social insurance against idiosyncratic risks better insured against through self-insurance than through which may be unrelated to aggregate income risks. market-type insurance. Setting up programs that build up * Third, regulation of private insurers helps to extend reserves during good times which are strictly earmarked to insurance to many who would be excluded without he spent only for these purposes during bad times may be such rules. Additionally, other forms of regulation- the main viable option for such governments until they notably prudential regulation of financial intermedi- carry out comprehensive economic reforms. aries-may reduce aggregate risk in an economy and provide safer instruments of self-insurance to individ- Conclusion uals. Financial and capital market sector strengthen- The foregoing policy recommendations aim either to bet- ing may be the most seriously underemphasized ter enable households to self-insure and self-protect, or to instrument of social policy. improve the government's role in assisting them. When all In practice, almost every example of governmental is said and done, however, these are necessary steps largely action that successfully fulfills one of these three roles will because insurance markets are either missing or seriously also to some extent fulfill one or both of the others. In addi- imperfect. Ultimately, risk is best dealt with through a tion, many social insurance policies will also perform social combination of market insurance, self-insurance, and self- assistance (that is, redistribution from richer to poorer protection. Policymakers should recognize this, and note households). especially that the market for insurance with pooling of Many countries in the region have improved the poverty risk is highly prone to failure. The best solution will usu- impact of social spending through reform over the last ally be to correct and complement the market, rather than decade, for example, by replacing generalized subsidies with to replace it. Intelligent regulation is essential for this, be programs specifically designed to help the poor. Evidence on it in labor markets, financial markets, or health services. government spending over the cycle for several countries is Following the comprehensive framework outlined in consistent with the view that LAC governments are sincere Chapter 3 for understanding household behavior in the face about protecting social spending during downturns: spend- of risk, the absence of insurance markets would generally ing on education, health, housing, and social security gener- make households worse off. Governments may be able to ally does not fall by as much as GDP. However, social spend- improve matters through public action (see Gill and Ilahi ing per poor person does fall-roughly equally because of the 2000). This can be of three types: reduced overall budget and the increased number of poor * First, the provision of or subsidy to activities used by people during economic contractions. households to generate self-protection, but the produc- Social spending-while being generally pro-poor- tion of which is characterized by positive externali- directly benefits the nonpoor as well. Spending on more 123 SECURING OUR FUTURE IN A GLOBAL ECONOMY ~~~---X~ 7.-6~ ~ ^ ~ ~ ~ ^-~~ ^~~ ~~ ~~~~~-~ ~ ~--~t tightly targeted programs for the poor does appear to suf- ocidal Prorams, Entitlements, and C:ounterlcaflty in tim - fer more during crises. Governments could do better to protect these programs from cuts. Experience in the region and in the U.S. shows that a successful strategy requires Since the New Deal launched by President Fraklin explicitly accounting for political economy factors that Roosevelt in the 1930s, the U.S. has had mnan of the make programs resilient to both political and economic programs being considered in LAC. Some of the U.S. changes (see Box 7.6). Such factors may include deliber- I experience may be relevant for coutrites in the rgion. ately building in some features that have been associated Snyder and YckovIlev (200 provide a quantitative with longer-lived government interventions, such as analysis of social Spending in the U.S. during 1962487, designing and marketing them as countrywide programs using detailed spending series and controlling for both aimed at the poor, rather than programs targeted at partic- political and economic factors. Some of their results are- ular parts of the country. Overall, spending on social protection is quite There is room for improving the design of targeted pro- countercyclical in the U.S. The analysis deals grams, however, especially how they relate to the economic only with ongoing programs, though it pearsp cycle. While meeting many of the goals they were that extending the analysis to new progms will designed to accomplish in various settings (for example, make the spending appear somewhat mote pro- both rural and urban), targeted conditional transfer (TCT) cyclical. programs such as Mexico's Progresa and Brazil's Bolsa Escola * The most countercyclical program, by far, is do not seem to be especially well suited to assist those vul- unemployment insurance (see Chapter 6). Social nerable to poverty with cash assistance in economic down- securi is also relatively countercyclical. these turns. Through their innovative links with human capital programs are distinguishable from others in ha- accumulation, TCT programs may be better suited than X ing a strong 'enttitement" factor: because people earlier interventions to address structural poverty concerns, see them as something they have specifically con- and even to counter the cyclical swings in quality of edu- tributed to, they are difficult for poit'icians to cation and healthcare services. They look even better when alter over the cycle. political economy factors are explicitly considered-the * argeted and nontareted programs appear to be programs appear to have increasingly broad political sup- equally countercyclical. Prgrms can be quite port, which is rather rare for transfer programs. More con- well targeted and still be resilient over both eco- I ventional instruments such as public works programs- nonic and political cycles-that is, it is not true when designed well-may be better safety nets, but have that programs must help both the poor and non- not enjoyed the same degree of popular support in the poor to attain resilience in a democracy. Avoiding region. Based on these considerations, targeted conditional overt "welfare" labels, keepi eligibility flxble transfer programs should be viewed as a strong contender so that the transient poor also can benefit fiom for forming the third leg of a comprehensive and perma- the program (for example, food stamps) and aim- nent safety net-the first two being social security for the ing to help poor children rather than adults can elderly and disabled, and income support for the unem- keep support fbr the programs high among even ployed in both the formal and informal sectors. the nonpoor. In conclusion, governments in the region do appear to * Programs that aretaeted at (poor placs Appear have behaved in a pro-poor manner in the most general to fiare worse than nationwide programs targeted terms, especially since the rise of democracy in the last two at the poor. decades. While authoritarian and democratic regimes in * Which party is in control of Contgress clearly i LAC appear to have responded similarly to economic matters-all social protection progras grow crises-both cut social spending sharply and about faster under Democratic control. However, party equally-greater increases in spending take place under ontrol matters even more fbr targeted programs. democratic rule. However, this is also where governments run the greatest danger of adding policy risk to economic 124 HELPING POOR HOUSEHOLDS DEAL BETTER WITH ECONOMIC CRISES risk. Well-intentioned governments or those under politi- times and repay during good times, The main problem in cal pressure to sharply increase spending on social pro- this regard is that private markets for such instruments do grams during growth episodes, only to have to reduce not exist: short-term capital usually flows out of countries spending in the next contraction, both raise risk and sow during economic downturns, to return only in good times. the seeds of social discontent. Discipline on the part of both governments and inter- The obvious solution for governments is to rely less on national financial institutions can help countries deal bet- ex post coping and more on ex ante insurance-that is, ter with aggregate economic volatility. Countries that move resources from good states to bad. This report takes institute effective self-protection (that is, through compre- the view that the reason why governments have not been hensive economic reforms) and self-insurance (that is, doing this is neither ignorance nor indifference, but lies in through well-designed and efficiently run social programs) an interplay of political and economic factors. Self-insur- should be rewarded with credit at reasonable terms. These ance at the country level (for example, through fiscal stabi- loans should be repaid during good times. Lending by lization funds) is a difficult option for democratic govern- these institutions should therefore be both strongly coun- ments: saving during good times runs the risk of being tercyclical so that it serves as insurance, and discriminating punished by the electorate, and the funds may be used up so that it encourages self-protection by governments. by more short-sighted successors. There are two viable options. The first is to create a sense of entitlement among Notes the electorate for programs that have a genuine insurance 1. Wodon and others (2000) find that this is roughly propor- component, such as that displayed by unemployment insur- tional: a I percent fall in per capita GDP leads to a I percent rise in ance and social security in the U.S. The second is access to headcount poverty. 2. As measured by the SIMCE Mathematics, Spanish, and Gen- financial markets in a manner that serves the purpose of eral Knowledge tests, taken ar the fourth and eighth grades in all market insurance: governments can borrow during bad Chilean schools. 125 Bibliography The word processed describes informally reproduced works Blanchflower, D., and S. Burgess. 1996. "Job Creation and that may not be available through libraries. Job Destruction in Great Britain in the 1980s. Indus- trial and Labor Relations Review 50(1) :17-38. Aizenman, Joshua, and Nancy P. Marion. 1993. "Policy Caballero, Ricardo. 1999(a). "Structural Volatility in Mex- Uncertainty, Persistence and Growth." Review of Interna- ico: A Policy Report." MIT mimeo. Cambridge, MA: tional Economics 1 :145-163. Massachusetts Institute of Technology, October 28. Allen, Steven G., Adriana Cassoni, Gaston J. Labadie. . 1999(b). "Structural Volatility in Chile: A Policy 1994. "Labor Market Flexibility and Unemployment in Report." MIT mimeo. Cambridge, MA: Massachusetts Chile and Uruguay. "Estudios de Economia 21 (Special Institute of Technology, October 25. Issue) :127-46. Santiago, Chile, November. . 1999(c). "Structural Volatility in Argentina: A Amadeo, Edward, Indermit Gill, and Marcelo Neri. 2000. Policy Report." MIT mimeo. Cambridge, MA: Massa- "Do Labor Laws Matter? The Pressure Points in Brazil's chusetts Institute of Technology, October 17. Labor Legislation." In Indermit Gill and Claudio Mon- Caballero, Ricardo. 2000. "Aggregate Volatility in Latin tenegro, (eds.), Readdressing Latin America's Forgotten America: Diagnostic, Analysis, and Policy Recommen- Reform: Quantifying Labor Policy Challenges in Argentina, dations." Paper written for the Regional Study Dealing Brazil and Chile. Unpublished manuscript. Washing- with Economic Insecurity in Latin America and the ton, D.C.: World Bank. Caribbean. Office of the Chief Economist, Latin Amer- Arango, Carlos, and William F. Maloney. 2000a. "Trade ica and the Caribbean Region. Washington, D.C.: Reform and Income Insecurity in Mexico and World Bank. Argentina." Mimeo, Latin America and Caribbean Sec- Caballero, Ricardo, and Arvind Krishnamurthy. 1999. tor Management Unit for Poverty Reduction and Eco- "Emerging Market Crises: An Asset Market Perspec- nomic Management. Washington, D.C.: World Bank. tive." MIT mimeo. Cambridge, MA: Massachusetts . 2000b. "Unemployment Dynamics in Latin Amer- Institute of Technology, August. ica: Estimates of Continuous Time Markov Models for . 2000. "International Liquidity Management: Steril- Mexico and Argentina." Mimeo, Latin America and ization Policy in Illiquid Financial Markets." MIT Caribbean Sector Management Unit for Poverty Reduc- mimeo. Cambridge, MA: Massachusetts Institute of tion and Economic Management. Washington, D.C.: Technology, April. World Bank. Calvo, Guillermo, and Carmen M. Reinhart. 1999. "Capital Arias, Omar. 1999. "Marginalization of Men in Latin Flow Reversals, the Exchange Rate Debate, and Dollar- America." Mimeo, Latin America and Caribbean Sector ization." Finance and Development: A Quarterly Publication Management Unit for Poverty Reduction and Economic of the International Monetary Fund 36 :13-15, September. Management. Washington, D.C.: World Bank. Carleton, D. 1979. "Why New Firms Locate Where They Athanasoulis, Stefano, and Eric Van Wincoop. (Forthcom- Do: An Econometric Model." In W. Wheaton, (ed.), ing). "Growth Uncertainty and Risk-Sharing." Journal Interregional Movements and Regional Growth. Washing- of Monetary Economics. ton, D.C.: The Urban Institute. 127 SECURING OUJR FUTURE IN A GLOBAL ECONOMY Cardenas, Mauricio, and Raquel Bernal. 1998. "Labor Mar- Cunningham, Wendy, and William Maloney. 2000. "Mea- ket Regulation and Labor Demand in Colombia, suring Vulnerability: Who Suffered in the 1995 Mexi- 1976-1996." Mimeo, Project on Labor Demand and can Crisis?" Latin America and Caribbean Sector Man- Labor Market Reform. Washington, D.C.: Inter-Ameri- agement Unit for Poverty Reduction and Economic can Development Bank. Management. Washington, D.C.: World Bank. Cassoni, Adriana, Steven G. Allen, and Gaston J. Labadie. Datt, Gaurav, and Martin Ravallion. 1994. "Transfer Ben- 1999. "Union, Labor Market Regulation, and Employ- efits from Public-Works Employment: Evidence from ment in Uruguay." Mimeo, Inter-American Develop- Rural India." Economic Journal 104(427) :1346-1369, ment Bank. Washington, D.C. November. . 2000. "The Effects of Unions on Employment: Evi- Davis, S., and J. Haltiwanger. 1990. "Gross Job Creation, dence from an Unnatural Experiment in Uruguay." Gross Job Destruction and Employment Reallocation." National Bureau of Economic Research Working Paper QuarterlyJournal of Economics 107(3) :819-864. 7501. Cambridge, MA. Dollar, David, and Aart Kraay. 2000. "Growth Is Good for Coate, Stephen. 1995. "Altruism, the Samaritan's the Poor." Mimeo. Washington, D.C.: World Bank. Dilemma, and Government Transfer Policy." The Amer- Dunne, T., M. Roberts, and L. Samuelson. 1989. "Plant ican Economic Review 85 :46-5 7. Turnover and Gross Employment Flows in U.S. Manu- Coloma, Fernando C. 1996. "Seguro de Desempleo: Teoria, facturing."Journal of Labor Economics 7(1) :48-71. Evidencia y Una Propuesta." Cuadernos de Economia Ano Duryea, Suzanne. 1998. "Children's Advancement 33(99) :295-320, August. Through School in Brazil: The Role of Transitory Conning, Jonathan, P. Olinto, and A. Trigueros. 2000. Shocks to Household Income." Mimeo. Washington, "Managing Economic Insecurity in Rural El Salvador." D.C.: Inter-American Development Bank. World Bank and Williams College, Department of Eco- Easterly, William Russel, and Aart Kraay. 1999. "Small nomics, processed. States, Small Problems?" Policy Research Working Cox, Donald, and Emmanuel Jimenez. 1998. "Risk-Shar- Paper 2139. Washington, D.C.: World Bank, June. ing and Private Transfers: What About Urban House- Easterly, William Russel, Roumeen Islam, and Joseph holds?" Economic Development and Cultural Change 46(3) Stiglitz. 2000. "Shaken and Stirred: Volatility and :621-639. Macroeconomic Paradigms for Rich and Poor Coun- Cunningham, Wendy V. 1997. "Unemployment Insurance tries." Michael Bruno Memorial Lecture. Washington, in a Multi-Sectoral Labor Market: The Case of Brazil." D.C.: World Bank. Unpublished manuscript. Washington D.C.: World Eaton, Jonathan, and Gene M. Grossman. 1985. "Tariffs as Bank. Insurance: Optimal Commercial Policy when Domestic - 1999. "Breadwinner versus Caregiver: Labor Force Markets are Incomplete." Canadian Journal of Economics Participation and Sectoral Choice over the Mexican 18: 258-72. Business Cycle." World Bank draft. Washington D.C., Edwards, Alejandra Cox, and Chris Manning. 1999. "The October 28. Economics of Employment Protection and Unemploy- . 2000. "Unemployment Insurance in Brazil: Unem- ment Insurance Schemes: Reflections on Policy Options ployment Duration, Wages and Sectoral Choice." Paper for Thailand, Malaysia, the Philippines, and Indonesia." prepared for the Regional Study on Economic Insecu- Paper prepared for the OECD, presented at World rity, Office of the Chief Economist, Latin America and Bank-International Labour Organisation-JMOL/JIL Caribbean Region. Washington, D.C.: World Bank. Seminar on Economic Crisis, Employment, and Labour Cunningham, Wendy, and Raquel Artecona. 2000. "Labor Markets in East and South East Asia. Tokyo, October Market Outcomes: Recent Trends and the Role of Gen- 13-15. der in Three LAC Countries." Latin America and Edwards, Sebastian. 1999. "How Effective are Capital Con- Caribbean Sector Management Unit for Poverty Reduc- trols?"Journal of Economic Perspectives 13(4) :65-84. tion and Economic Management. Washington, D.C.: Edwards, Sebastian, and Miguel A. Savastano. 1999. World Bank. "Exchange Rates in Emerging Economies: What Do We 128 BIBLIOGRAPHY Know? What Do We Need to Know?" National Bureau Preliminary Report prepared as part of International of Economic Research, Working Paper Series No. 7228 Food Policy Research Institute's evaluation of the Pro- :1-72. Cambridge, MA. gresa program. Washington, D.C., March. Edwards, Sebastian, and Nora Lustig, (eds.). 1997. Labor Ghosh, Atish R., Anne Marie Gulde, and Holger C. Wolf. Markets in Latin America: Combining Social Protection with 1998. "Currency Boards: The Ultimate Fix?" Working Market Flexibility. Washington, D.C: Brookings Institu- Paper 98/8. Washington D.C.: International Monetary tion Press. Fund, January. Ehrlich, Isaac, and Gary Becker. 1972. "Market Insurance, Gill, Indermit, Amit Dar, and Mark Thomas. 1999. Self-insurance and Self-Protection." Journal of Political "Financing Unemployment Insurance: International Economy 80 :623-648. Experience and Lessons for Brazil." Unpublished manu- Eichengreen, Barry, and Ricardo Hausmann. 1999. script. Washington D.C.: World Bank. "Exchange Rates and Financial Fragility." National Gill, Indermit, and Claudio Montenegro, (eds.). 2000. Bureau of Economic Research, Working Paper Series Readdressing Latin America's Forgotten Reform: Quantifying No. 7418 :1-52. Cambridge, MA, November. Labor Policy Challenges in Argentina, Brazil, and Chile. Fajnzylber, Pablo, and William F. Maloney. 2000. "Labor Unpublished manuscript. Washington, D.C.: World Demand and Trade Liberalization in Latin America." Bank. Mimeo, Latin America and Caribbean Sector Manage- Gill, Indermit, and Nadeem Ilahi. 2000. "Economic Inse- ment Unit for Poverty Reduction and Economic Man- curity, Individual Behavior, and Social Policy." Paper agement. Washington, D.C.: World Bank. prepared for the Regional Study on Economic Insecu- . 2000. "Labor Demand in Colombia, Chile and Mex- rity, Office of the Chief Economist, Latin America and ico, Dynamic Panel Modeling." Mimeo, Latin America Caribbean Region. Washington D.C.: World Bank. and Caribbean Sector Management Unit for Poverty Gill, Indermit, Erik Haindl, Claudio Sapelli, and Claudio Reduction and Economic Management. Washington, Montenegro. 2000. "Chile: Has Employment Become D.C.: World Bank. More 'Precarious'?" In Indermit Gill and Claudio Mon- Fajnzylber, Pablo, Eduardo Ribeiro, and William F. Mal- tenegro, (eds.), Readdressing Latin America's Forgotten oney. 1999. "Turnover, Exit/Entrance Decisions and Reform: Quantifying Labor Policy Challenges in Argentina, Labor Demand in Chile." Mimeo, Latin America and Brazil, and Cbile. Unpublished manuscript. Washing- Caribbean Sector Management Unit for Poverty Reduc- ton, D.C.: World Bank. tion and Economic Management. Washington, D.C.: Gill, Indermit, Eric Haindl, Claudio Montenegro, and World Bank. Claudio Sapelli. 2000. "Has Employment in Chile Ferreira, Francisco H. G., and Indermit Gill. 2000. Become More Precarious?" In Readdressing Latin Amer- "Weathering Storms: Households, Governments and ica's Forgotten Reform: Quantifying Labor Policy Challenges Aggregate Income Shocks in Latin America." World in Argentina, Brazil and Chile. World Bank draft. Wash- Bank draft. Washington, D.C. ington, D.C. Frankel Jacob E., Sergio Schmukler, and Luis Serven. 2000. Glewwe, Paul, and Gillette Hall. 1998. "Are Some Groups "Verifying Exchange Rate Regimes." Mimeo. Washing- More Vulnerable to Macroeconomic Shocks Than Oth- ton, D.C.: World Bank. ers? Hypothesis Tests Based on Panel Data from Peru." Garibaldi, P. 1999. "Job Flow Dynamics and Firing Journal of Development Economics 56(1) :181-206. Restrictions." European Economic Review 42(1) :245-275. Gonzalez, Jose Antonio. 1999. "Labor Market Flexibility in Gavin, Michael, and Ricardo Hausmann. 1995. "The 13 Latin American Countries and the U.S." World Bank Macroeconomics of Capital Flows to Latin America: Latin American Viewpoints Series and Latin America Experience and Policy Issues." Inter-American Develop- and the Caribbean Regional Study. Washington, D.C. ment Bank, Office of the Chief Economist, Working Guasch, J. Luis. 1999a. Labor Market Refowm andJob Cre- Paper 310. Washington, D.C., October. ation: The Unfinished Agenda in Latin American and Gertler, Paul. 2000. "A Preliminary Evaluation of the Caribbean Countries. Directions in Development Series. Impact of Progresa on Health and Health Utilization." Washington, D.C.: World Bank. 129 SECURING OUR FUTURE IN A GLOBAL ECONOMY _- 1999b. "An Alternative to Traditional Unemploy- Jalan, Jyotsna, and Martin Ravallion. 1999. "Estimating ment Insurance Programs: A Liquidity-Based Approach the Benefit Incidence of an Antipoverty Program by Against the Risk of Earnings Losses." Unpublished Propensity Score Matching." Unpublished manuscript. manuscript. Washington, D.C.: World Bank. Washington, D.C.: World Bank. Haindl, Erik. 1985. "Un modelo para la determinacion de Jones, Polly, and Martin Ravallion. 1999. "The Trabajar flujos y parametros dinamicos en el mercado de trabajo." Program in Argentina." Unpublished manuLscript. Estudio de Economia 12, marzo. Santiago: Universidad de Washington, D.C.: World Bank. Chile. Kalbfleisch, J. D., and J. F. Lawless. 1985. "The Analysis Hamermesh, Daniel S. 1992. "Unemployment Insurance of Panel Data under a Markov Assumption." Journal of for Developing Countries." Policy Research Working the American Statistical Association 80(392) :863-871. Paper, No. 897. Washington, D.C.: World Bank, May. Kaminsky, Graciela, and Sergio Schmukler. 2000. "Sever- - 1993. Labor Demand. Princeton: University Press. ing International Financial Links: A New Look at Cap- Hatzius, Jan. 2000. "Foreign Direct Investment and Factor ital Controls." Mimeo. Washington, D.C.: World Bank, Demand Elasticities." European Economic Review 44 (1) April 17. :117-43. Klein, M., S. Schuh, and R. Triest. 2000. "Job Creation, Hicks Norman, and Quentin Wodon. 2000. "Economic Job Destruction and the Real Exchange Rate." National Shocks, Safety Nets, and Fiscal Constraints: Social Pro- Bureau of Economic Research, Working Paper 7466. tection for the Poor in Latin America. In XII Seminario Cambridge, MA, January. Regional de Politica Fiscal: Compendio de Documentos 2000. Konings, J. 1995. "Job Creation and job Destruction in Economic Commission for Latin America and the the U.K. Manufacturing Sector." Oxford Bulletin of Eco- Caribbean. Santiago, Chile, United Nations, 381-407. nomics and Statistics 57(l) :5-24. Holzmann, Robert, and Steen Jorgensen. 1999. "Social Pro- Kraays and Nran 57(1) :5-24. tection as Social Risk Management: Conceptual Under- Kraay, Aarr, Norman Loayza, Luis Serven, and Jaumre Ven- tection asocilRskMangemnt:Cocepuatura. 2000. "Country Portfolios." Mimeo. Washington, pinnings for the Social Protection Sector Strategy Paper." D.C.: World Bank. Journal of International Development 11(7) :1005-27. D.C.: Adriank. Hpenhyn, Hugo A., and Juan Pablo Nicolini. 1999. Kugler, Adriana D. 1999. "The Impact of Firing Cosso "Hopeterogenei andO.ial Unmloyment 1nsur Turnover and Unemployment: Evidence from the "Heterogeneity and Optimal Unemployment Insur- Colombian Labour Market Reform." International Tax ance." Paper presented at the Conference on Social Pro- ColiancLaournal Tax tection and Poverty. Washington, D.C.: Inter-American and PulcFnneJourna 6(3). Development Bank, February. . 2000. "Does Replacing Employer Insurance for Self- IDB (Inter-American Development Bank). 1995. Economic Insurance Improve the Performance of the Labor Mar- ID (ItrAeia Deeomn Bak) 195 Economic , ket? The Case of the Colombian Severance Payments and Social Progress in Latin America. Office of the Chief Economist. Washington D.C., January. Savings Accounts." Paper prepared for the Regional . 2000. Social Protection for Equity and Growth. Wash- Study on Economic Insecurity, Office of the ChieftEcon- ington, D.C. omist, Latin America and Caribbean Region. Washing- Ilahi, Nadeem. 1999. "Children's Work and Schooling ton, D.C.: World Bank. Under Shocks: Does Gender Matter? Evidence from the . 2000. "Job Security Provisions, Labor Market Flexi- Peru LSMS Panel Data." Background paper for Engen- bility and Non-Compliance: Evidence from Colombia." dering Development, Technical Report. Washington, Forthcoming in Daniel Hamermesh, James He,ckman D.C.: World Bank. and Carmen Pages, (eds.), Labor Market Regulation in International Labour Office. 1999. Decent Work and Protec- Latin America. Washington, D.C.: Inter-American tion for All: Priority of the Americas. Report of the Direc- Development Bank. tor General. Geneva. Leiderman, Leonardo, Ricardo Hausmann, and Gavin, Jack, William. 2000. "The Evolution of Health Insurance Michael. 1995. "Macroeconomics of Capital Flows to Institutions: Theory and Four Examples from Latin Latin America: Experience and Policy Issues." Working America." World Bank draft. Washington, D.C. Paper No. 310, Oct. 130 BIBLIOGRAPHY Levenson, Alec R., and William F. Maloney. 1998. "The Mizala, Alejandra, and Pilar Romaguera. 2000. "The Informal Sector, Firm Dynamics, and Institutional Par- Quality of Social Services: The Case of Chile." World ticipation." Policy Research Working Paper No. 1988. Bank draft. Washington, D.C. Washington, D.C.: World Bank. Montenegro, Claudio E. 1998. "The Structure of Wages in Levinsohn, J. 1999. "Employment Responses to Interna- Chile 1960-1996: An Application of Quantile Regres- tional Liberalization in Chile." Journal of International sion." Estudios de Economic 25(1) :71-98. Economics 47(2) :321-344. Montenegro, Claudio E., and Indermit Gill. 2000. "Chile's Lewis, Karen K. 1999. "Trying to Explain Home Bias in Public Work Programs: Did They Work Then and Equities and Consumption." TheJournal of Economic Lit- Would They Work Now?" Paper prepared for the erature 37(2). Regional Study on Economic Insecurity, Office of the Loayza, Norman. 1995. "The Economics of the Informal Sec- Chief Economist, Latin America and Caribbean Region. tor: A Simple Model and Some Empirical Evidence from Washington, D.C.: World Bank. Latin America." World Bank draft. Washington, D.C. Montiel, Peter, and Carmen M. Reinhart. 1999. "Do Cap- Lustig, Nora. 1999. "Crises and the Poor: Socially Respon- ital Controls and Macroeconomic Policies Influence the sible Economics." Poverty and Inequality Unit, Volume and Composition of Capital Flows? Evidence processed. Washington, D.C.: Inter-American Develop- from the 1990s." Journal of International Mone' and ment Bank. Finance 18(4) :619-35. MacIsaac, Donna, and Martin Rama. 2000. "Mandatory Morduch, Jonathan. 1999. "Between the State and the Severance Pay in Peru: An Assessment of its Coverage Market: Can Informal Insurance Patch the Safety Net?" and Effects Using Panel Data." Paper prepared for the World Bank Research Observer 14(2) :187-207, August. Regional Study on Economic Insecurity, Office of the Narayan, Deepa, Raj Patel, Kai Schafft, Anne Rademacher, Chief Economist, Latin America and Caribbean Region. and Sarah Koch-Schulte. 2000. Voices of the Poor: Can Washington, D.C.: World Bank. Anyone Hear Us? (World Bank.) New York: Oxford Uni- Maloney, William. 1999a. "Does Informality Imply Seg- versity Press. mentation in Urban Labor Markets? Evidence from Sec- Neri, Marcelo, and Mark Thomas. 2000. "Household toral Transitions in Mexico." World Bank Economic Responses to Labor-Market Shocks in Brazil: Review 13(2) :275-302, May. 1982-1999." World Bank draft. Washington, D.C. . 1999b. "Self-Employment and Labor Turnover: Obstfeld, Maurice. 1994. Foundations of International Cross-Country Evidence." Policy Research Working Macroeconomics. Cambridge, MA: MIT Press. Paper No. 2102. Washington, D.C.: World Bank. Olson, Mancur. 2000. Power and Prosperity: Outgrowing Com- . 2000. "How Should We Think about Emergency munist and Capitalist Dictatorships. New York: Basic Programs?" Unpublished manuscript. Washington, Books. D.C.: World Bank. Paes de Barros, Ricardo, Carlos Henrique Corseuil, and Maloney, William F, and Eduardo Ribeiro. 1999. "Effi- Gustavo Gonzaga. 1999. "Labor Market Regulations ciency Wage and Union Effects in Labor Demand and and the Demand for Labor in Brazil." Mimeo. Wash- Wage Structure in Mexico: An Application of Quantile ington, D.C.: Inter-American Development Bank. Analysis." Mimeo. Washington, D.C.: Inter-American Paes de Barros, Ricardo, Carlos Henrique Corseuil, Miguel Development Bank. Foguel, Gustavo Gonzaga, Ricardo Henriques, Rosane Marquez, Gustavo. 1999. "Labor Markets and Income Sup- Silva Pinto de Medonca, and Danieli Milton. 1998. port: What Did We Learn from the Crises?" Unpub- "Labor Market Regulations and the Demand for Labor lished manuscript. Washington, D.C.: Inter-American in Brazil." Mimeo, Project on Labor Demand and Labor Development Bank. Market Reform. Washington, D.C.: Inter-American Mazza, Jacqueline. 1999. Unemployment Insurance: Case Development Bank. Studies and Lessons for Latin America and the Caribbean. Pagnagariya, Arvind. 1999. "Trade Openness: Consequences RE2/SO2, Technical Study. Washington, D.C.: Inter- for the Elasticity of Demand for Labor and Wage Out- American Development Bank, May. comes." Mimeo, University of Maryland Regional Study. 131 SECURING OUR FUTURE IN A GLOBAL ECONOMY Parker, Susan W, and Emmanuel Skoufias. 2000. "Job Roberts, M. 1996. "Employment Flows and Producer Loss, Change in Marital Status, and the Allocation of Turnover." In M. Roberts and J. Tybout, (eds.), Industry Time within Families: Evidence from Urban Mexico." Evolution in Developing Countries. New York: Oxford Uni- Draft, Programa de Educaci6n, Salud y Alimentaci6n versity Press. and International Food Policy Research Institute. Roberts, M. J., and E. Skoufias. 1997. "The Long-Run Washington, D.C. Demand for Skilled and Unskilled Labor in Colombian Perry, Guillermo, and Daniel Lederman. 1999. "Adjust- Manufacturing Industries." Review of Economics and Sta- ments After Speculative Attacks in Latin America and tistics 79 :330-334. Asia: A Tale of Two Regions?" World Bank Latin Amer- Roberts, M. J., and J. R. Tybout. 1996. Industrial Evolution ican and Caribbean Studies, Viewpoints. Washington, in Developing Countries: Micro Patterns of Turnover. Produc- D.C.: World Bank. tivity, and Market Structure. New York: Oxford Univer- Pindyck, Robert S., and Andres Solimano. 1993. "Eco- sity Press. nomic Instability and Aggregate Investment." National Rodrik, Dani. 1997. Has Globalization Gone Too Far;) Wash- Bureau of Economic Research, Working Paper No. ington, D.C.: Institute for International Economics. 4380. Cambridge, MA, June. . 1999. "Why is There So Much Economic Insecurity Rama, Martin, and William Maloney. 2000. "Income Sup- in Latin America?" World Bank draft, processed. Wash- port Programs for the Unemployed in Latin America." ington, D.C., October. World Bank draft. Washington, D.C. Saavedra, Jaime. 1999. "Labor Demand of Formal Firms in Ramey, Garey, and Valerie A. Ramey. 1995. "Cross-Coun- Peru." Mimeo, Project on Labor Demand and Labor try Evidence on the Link Between Volatility and Market Reform. Washington, D.C.: Inter-American Growth." American Economic Revieaw 85 :1138-51. Development Bank. Ravallion, Martin. 1990. "On the Coverage of Public Saavedra, Jaime, and Maximo Torero. 1999. "Labor Market Employment Schemes for Poverty Alleviation."Journal Regulations and Employment in Peru." Unpublished of Development Economics 34(1-2) :57-79, November. manuscript, Grupo de analisis para el desarrollo. Lima, - 1999a. "Appraising Workfare." World Bank Research Peru. Observer 14(1) :31-48. . 2000. "Trade Liberalization, Labor Market Reforms _- 1999b. "Is More Targeting Consistent with Less and Labor Demand." Mimeo, Project on Labor Demand Spending?" International Tax and Public Finance. Wash- and Trade Liberalization, W. F. Maloney, (ed.), Washing- ington, D.C.: World Bank. ton D.C.: World Bank. - 1999c. "Monitoring Targeting Performance when Schady, Norbert. 2000. "The Impact of Economic Crises on Decentralized Allocations to the Poor are Unobserved." School Attendance: Evidence from Peru-A Research Unpublished manuscript. Washington D.C.: World Proposal." East Asia and Pacific Region, processed. Bank. Washington, D.C.: World Bank. . 2000. "Can Public Works Programs Protect the Schultz, T. Paul. 2000. "The Impact of Progresa on School Vulnerable during Macroeconomic Downturns?" Paper Enrollments." Final Report prepared as part of Interna- prepared for the Regional Study on Economic Insecu- tional Food Policy Research Institute's evaluation of the rity, Office of the Chief Economist, Latin America and Progresa program, April. Caribbean Region. Washington, D.C.: World Bank. Sedlacek, Guilherme, Nadeem Ilahi, and Emily Gustafs- . 2000. "The Incidence of Fiscal Contractions: Theory son-Wright. 2000. "Targeted Conditional Transfers in and Evidence for Argentina." Mimeo. Washington, Latin America: An Early Survey." World Bank draft. D.C.: World Bank. Washington, D.C. Ravallion, Martin, Gaurav Datt, and Shubham Chaudhuri. Serven, Luis. 1999. "Macroeconomic Uncertainty and Pri- 1993. "Does Maharashtra's Employment Guarantee vate Investment in Developing Countries: An Empirical Scheme Guarantee Employment? Effects of the 1988 Investigation." Mimeo. Washington, D.C.: World Bank. Wage Increase." Economic Development and Cultural . 2000. "Note on the Evolution of Macroeconomic Change 41(2) :251-275. Risk in Latin America." Mimeo, Latin America and 132 BIBLIOGRAPHY Caribbean Sector Management Unit for Poverty Reduc- Weller, Jurgen. 1999. "Crecimiento, empleo y equidad." tion and Economic Management. Washington, D.C.: Mimeo, Economic Commission for Latin America and World Bank. the Caribbean. Santiago, Chile. Serven, Luis, and Andres Solimano. 1993. Striving for Wodon Quentin. 2000. "Income Mobility and Risk Dur- Growth After Adjustment: The Role of Capital Formation. ing the Business Cycle." Mimeo, Latin American and Washington, D.C.: World Bank. the Caribbean. Washington D.C.: World Bank. Serven, Luis, and N. Gonzalez. 2000. "A Note on Aggre- . 2000. "Poverty and Policy in Latin America and the gate Employment/Wages in LAC." Mimeo, Latin Caribbean." Regional Study, Latin America and America and Caribbean Sector Management Unit for Caribbean Sector Management Unit for Poverty Reduc- Poverty Reduction and Economic Management. Wash- tion and Economic Management. Washington, D.C.: ington, D.C.: World Bank. World Bank. Skoufias, Emmanuel, Benjamin Davis, and Jere Behrman. . 2000. "Public Works Employment and Workfare 1999. "An Evaluation of the Selection of Beneficiary Programs: Optimizing the Timing of Benefits for Poverty Households in the Education, Health and Nutrition Reduction." Mimeo. Washington, D.C.: World Bank. Program (Progresa) of Mexico." Final Report prepared as Wodon, Quentin, and C. Siaens. 1999. "Training for the part of International Food Policy Research Institute's Urban Unemployed: A Reevaluation of Mexico's Probe- evaluation of the Progresa program, June. cat. In Government Programs and Poverty in Mexico, Report Snyder, James, and Irene Yackovlev. 2000. "Political and No. 19214-ME. Washington, D.C.: World Bank. Economic Determinants of Government Spending on Wodon, Quentin, Norman Hicks, Bernadette Ryan, and Social Protection Programs." MIT Department of Polit- Gabriel Gonzalez. 1999. "Are Governments Pro-poor? A ical Science and Economics. Paper prepared for the Test Based on Targeted and Social Spending During Regional Study on Economic Insecurity, Office of the Booms and Busts." World Bank draft. Washington, D.C. Chief Economist, Latin America and Caribbean Region. Wodon, Quentin, W. Maloney, and M. Barenstein. 2000. Washington, D.C.: World Bank. "Self-Employment as an Explanation for High Inequal- Tybout, J. 1996. "Chile, 1979-86: Trade Liberalization ity in Latin America." Processed. Washington, D.C.: and its Aftermath." In M. Roberts and J. Tybout, (eds.), World Bank. Industry Evolution in Developing Countries. New York: Wodon, Quentin, and Mari Minowa. 1999. "Training for Oxford University Press. the Urban Unemployed: A Reevaluation of Mexico's Tzannatos, Zafiris, and Suzanne Roddis. 1998. "Unem- Probecat." Paper prepared for the Regional Study on ployment Benefits." SP Discussion Paper No. 9813, Economic Insecurity, Office of the Chief Economist, Social Protection Unit, Human Development Network. Latin America and Caribbean Region. Washington, Washington, D.C.: World Bank. D.C.: World Bank. van Wincoop, Eric. 1999. "How Big are Potential Gains Wodon, Quentin, R. Ayres, M. Barenstein, N. Hicks, K. from International Risk Sharing?" Journal of Interna- Lee, W. Maloney, P. Peeters, C. Siaens, and S. Yitzhaki. tional Economics 47(1) :109-135. 2000. "Poverty and Policy in Latin America and The Vodopovic, Milan. 2000. "Income Support Systems for Caribbean." Washington, D.C.: World Bank. Unemployed Workers." Research Project (in Progress), World Bank. 1990. World Development Report 1990. New Human Development Network, Social Protection Unit. York: Oxford University Press. Washington, D.C.: World Bank. . 1997. The Long March, A Reform Agenda for Latin Vroman, Wayne. 1998. Topics in Unemployment Insurance America and the Caribbean in the Next Decade. World Financing. Kalamazoo, Michigan: W. E. Upjohn Insti- Bank Latin American and Caribbean Studies, View- tute for Employment Research. points. Washington, D.C. 133 THE WORLD BANK I 8 IS H Street, N.W. Washington, D.C. 20433, U.S.A. Telephonie: 202-477-l234 Facsimile: 202-477-6391 Telex: MCI 64145 WORLDBANK M\4CI 248423 WORl.DBANK NVlorld Wide 'Web: wwNv.\vorldban k.org/ E-mail: hooks@worjldbanik.org ISBN 0-8213-4730-6