DIRECTIONS IN DEVELOPMENT Safety Net Programs and Poverty Reduction Lessons from Cross-Country Experience K. SUBBARAO ANIRUDDHA BONNERJEE KENE EZEMENARI JEANINE BRAITHWAITE CAROL GRAHAM SONIYA CARVALHO ALAN THOMPSON V Kg,e _3v S, _ ~~~~~ A~~~~~~~ 9-. - B. k _ X ...' DIRECTIONS IN DEVELOPMENT Safety Net Programs and Poverty Reduction Lessons from Cross-Country Experience K. Subbarao Aniruddha Bonnerjee Jeanine Braithwaite Soniya Carvalho Kene Ezemenari Carol Graham Alan Thompson The World Bank Washington, D.C. C) 1997 The International Bank for Reconstruction and Development/ THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433 All rights reserved Manufactured in the United States of America First printing March 1997 The findings, interpretations, and conclusions expressed in this study are entirely those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. K. Subbarao and Jeanine Braithwaite are principal economist and human resources economist, respectively, in the Poverty, Gender, and Public Sector Management Department of the World Bank. Aniruddha Bonnerjee, Soniya Carvalho, Kene Ezemenari, and Alan Thompson are consultants at the World Bank. Carol Graham is a fellow at the Brookings Institution, she was a visiting fellow at the World Bank when the study was written. Cover photo: Curt Carnemark, The World Bank Library of Congress Cataloging-in-Puiblication Data Safety net programs and poverty reduction: lessons from cross-country experience / [prepared by] K. Subbarao ... [et al.]. p. cm. Includes bibliographical references. ISBN 0-8213-3890-0 1. Income maintenance programs-Case studies. 2. Economic assistance, Domestic-Case studies. I. Subbarao, K. HC79.15S24 1997 362.5'82-dc2l 97-3168 CIP Contents Foreword v Acknowledgments vii Abbreviations ix 1 Introduction 1 What Constitutes a Safety Net? 2 Who Needs Safety Nets? 3 Notes 13 2 Conceptual Issues in the Analysis of Social Assistance and Poverty-Targeted Programs 15 Targeting Costs 15 Targeting Mechanisms 20 The Choice between Cash and In-Kind Transfers 21 Conclusion 23 Notes 23 3 Cross-Country Patterns of Cash Transfers 25 Famnily Assistance 26 Social Assistance 35 Cash Transfers in Eastern Europe and Countries of the Former Soviet Union 39 Notes 42 4 Cross-Country Patterns of In-Kind Transfers 45 Food Subsidies 45 Housing Subsidies 55 Energy Subsidies 60 Conclusion 65 Notes 66 iii iv SAFETY NET PROGRAMS AND POVERTY REDUCTION 5 Income Generation Programs 69 Public Works 69 Livelihood Programs 85 Notes 91 6 Innovations in the Delivery and Organization of Social Assistance and Poverty-Targeted Programs 93 Alternative Delivery Models 93 Program Experiences under Public and Private Delivery 97 Lessons of Experience 111 Notes 114 7 Modes of Financing Social Assistance Programs 117 Sources of Financing and Trade-Offs 117 A Cross-Country Analysis of Funding 121 Transfer Spending, Social Services, and Budget Deficits 125 Fiscal Federalism, Decentralization, and Public Expenditure Management 126 The Specific Challenges of Financing in the Economies in Transition 129 Notes 133 8 The Political Economy of Social Assistance Programs 135 Protecting the Poor during Adjustment in Latin America and the Caribbean 138 Sub-Saharan Africa: Social Funds and Social Action Programs 144 Safety Net Issues in Transition Economies 148 Politics of Safety Nets: Some Generalizations 155 Notes 157 9 Choosing and Designing Programs 159 A Framework for Choosing among Programs 161 Appropriate Program Design 164 Conclusion 170 Note 170 References 171 Foreword There is now a widespread consensus on the need for safety nets as a key component of a poverty reduction strategy. Safety nets are basically income maintenance programs that protect a person or household against two adverse outcomes: a chronic incapacity to work and earn, and a decline in this capacity caused by imperfectly predictable life-cycle events (such as the sudden death of a bread- winner), sharp shortfalls in aggregate demand or expenditure shocks (through economic recession or transition), or very bad har- vests. Safety net programs serve two important roles: redistribution (such as transfers to disadvantaged groups) and insurance (such as drought relief). The ongoing structural adjustment programs and the unprece- dented economic and political transition of countries of the former Soviet Union and Eastern Europe have given rise to a renewed inter- est in social safety net programs. Economic crisis-whatever its sources-is most likely to impinge harshly on vulnerable groups in the short run, underscoring the need for countervailing measures. Moreover, the role of these policies in specific countries depends on the way the economy distributes resources in the absence of safety net policies. A propoor development path will place less demands on safety net policies. Over the past three decades several developing countries have embarked on a variety of programs, including cash transfers, subsi- dies in-kind, public works, and income-generation programs. Few studies have synthesized the lessons from widely differing country experiences. Moreover, there is little guidance on appropriate pro- gram design for countries and international donors supporting such safety net initiatives. This book fills the gap. It reviews the concep- tual issues in the choice of programs, synthesizes cross-country experience, and analyzes how country- and region-specific con- straints help explain why different approaches are successful in dif- ferent countries. v vi SAFETY NET PROGRAMS AND POVERTY REDUCTION It is hoped that World Bank staff, members of other international aid agencies, policymakers in client countries, and the academic community will benefit from this collaborative work of the Poverty and Social Policy Department. Ishrat Husain Director Poverty and Social Policy Department World Bank Acknowledgments This book was prepared by a team comprising K. Subbarao (team leader and main author), Aniruddha Bonnerjee, Jeanine Braithwaite, Soniya Carvalho, Kene Ezemenari, Carol Graham, and Alan Thompson. Nicholas Barr and Soonwon Kwon contributed back- ground papers. The study was carried out under the overall direction of Oey Astra Meesook and Ishrat Husain, the former manager and the director, Poverty Analysis and Social Policy Department. For helpful comments and suggestions, the authors wish to thank Petros Aklilu, Harold Alderman, Nicholas Barr, Bruce Fitzgerald, Louise Fox, Fredrick Golladay, Margaret Grosh, Lshrat Husain, Emmanuel Jimenez, Polly Jones, Steen Jorgensen, Shahid Khandker, Kathie Krumm, Soonwon Kwon, Peter Lanjouw, Samuel Lieberman, Kathy Lindert, William McGreevey, Oey Astra Meesook, Tawhid Nawaz, John Newman, Nick Prescott, Sandor Sipos, Elisabeth Stock, Dominique van de Walle, Jack van Holst Pellekaan, and Lawrence Wolff. Michael Lipton, professor of economics at the University of Sussex, served as an external peer reviewer. The study benefited greatly from his valuable comments and suggestions. Successive drafts of the paper were processed by Gay Santos and Maria Paz Felix. Ilyse Zable edited the book and Glenn McGrath laid it out. vii Abbreviations AGETIP Agence d'Execution des Travaux d'Interet Public Contre le Sous-Emploi AIDS Acquired immune deficiency syndrome AKRSP Aga Khan Rural Support Project APRA American Popular Revolutionary Alliance BKK Badon Kredit Kecamatan BRAC Bangladesh Rural Advancement Committee BRI Bank Rakyat Indonesia CDP Community Development Program DIRE Delegation a l'Insertion et le Reinsertion a l'Emploi EEP Export Enhancement Program ESF Emergency Social Fund FIS Fondo de Inversi6n Social GDP Gross domestic product GNP Gross national product ICDS Integrated Child Development Services 1CM Brazil's general value added tax IFAD International Fund for Agricultural Development IFPRI International Food Policy Research Institute ILO International Labour Organization IMF International Monetary Fund IRDP Integrated Rural Development Program JRY Jawahar Rojgar Yojuna LAC Latin America and the Caribbean MCHN Mother/child health and nutrition MEGS Maharashtra Employment Guarantee Scheme MENPROSIF Mendoza Provencial Program for Basic Social Infrastructure MMD Movement for Multi-Party Democracy MPU Microprojects Unit MYRYDA The Mysovre Resettlement and Development Agency NGO Nongovernmental organization NIA Philippines Communal Irrigation Project NBU National Bank of Ukraine .x x SAFETY NET PROGRAMS AND POVERTY REDUCTION OECD Organization for Economic Cooperation and Development PAIT Temporary Income Support Program PAU Urban Action Program PEM Minimum Employment Program POJH Occupational Program for Household Heads PPP Purchasing power parity PRADAN Professional Assistance for Development Action PRI Institutional Revolutionary Party PRONASOL National Solidarity Program RMP Rural Maintenance Program SAPs Social Action Programs Sedesol Secretariat of Social Development SEWA Self-Employed Women's Association SF Social fund SPARC Society for Promotion of Area Residence Centres SRF Social Recovery Fund SRP Social Recovery Project TINP Tamil Nadu Integrated Nutrition Project UNICEF United Nations Children's Fund UNIP United National Independence Party USAID United States Agency for International Development WFP World Food Program WGTFI Working Group in Targeted Food Interventions 1 Introduction The current proliferation of structural adjustment programs and macroeconomic stabilization measures has given rise to a renewed interest in social safety net programs.' The economic and political transition of the countries of the former Soviet Union and Eastern Europe has contributed to this renewed interest as well. Countries in this region are restructuring their economic systems at the same time that they are attempting to liberalize prices and stabilize external bal- ances, provide public services, and protect the well-being of their cit- izens. These countries have inherited systems of social protection that are no longer fiscally sustainable or able to protect economically and socially vulnerable groups. Public spending on social protection (social insurance and social assistance payments) as a share of gross domestic product (GDP) continues to be high, though many of the newly emerging working poor remain unprotected.2 Evidence shows that poverty increased by as much as 100 percent between 1987 and 1993-the years immediately preceding and following transition. The unskilled laborer was most likely to experience worsening poverty during this period. In low- and middle-income countries in the process of adjusting, formal safety nets and community and family support networks are increasingly unable to cope with the extent of need in the wake of events-such as the massive numbers of AIDS-orphaned children in Africa-unforeseen a decade ago.3 Regardless of a country's location, be it in Africa, Latin America, Asia, the former Soviet Union, or Eastern Europe, economic crisis and the adjustment or transition that follows have adverse short-term effects on the living standards of vulnerable groups. And shrinking budgets are severely restricting the amount of resources available for social assistance. There is clearly an acute need to protect the new poor and the chronically poor by introducing social assistance programs or restructuring existing ones in ways that make them fiscally, politically, and administratively sustainable. 1 2 SAFETY NET PROGRAMS AND POVERTY REDUCTION Added to the urgent need to revisit safety net issues is the ongoing debate over the optimal policy mix between the indirect growth- oriented path to poverty reduction and the direct path of targeted transfers to the poor. Should all scarce resources be allocated so as to maximize economic growth, the benefits of which will eventually trickle down to the poor, or should a part of these resources be trans- ferred to the poor to improve and maintain their living standards dur- ing the process of economic growth? A third option is to give the poor access to assets (such as land) that provide them with a source of eco- nomic growth based on their absolute advantagthat is, labor. Although diverting scarce resources directly to the poor compro- mises economic growth in the short run, the forgone cost of not accounting for the poor may compromise economic growth in the long-run. In order to survive, the poor may seek to minimize income or consumption risks, resort to criminal or marginalized activities, or be forced to accept extreme insecurity and risk of death. Moreover, denying the poor access to economic and educational opportunities accentuates inequality-an outcome likely to retard economic growth (Lipton 1995). In democratic societies neglect of the poor, particularly during reform, can also lead to political instability. Some conflict between direct poverty-targeted programs and growth-promoting activities is inevitable in most countries. But the dichotomy between the direct and indirect paths to poverty reduction may be minimized if a country adopts a growth-oriented adjustment program that is sufficiently broad-based to include all socioeconomic groups, particularly the poor, in productive activities. Finally, the path taken by a specific country, and the relative weight given to poverty- targeted programs, will depend on the level of institutional and administrative capacity, and on the quantity (and quality) of human capital. If the forgone cost of not including the poor-especially the accentuation of inequality in the short run-is a major consideration, then safety nets will need to be a strategic component of the path that is taken to poverty reduction. What Constitutes a Safety Net? Safety nets are programs that protect a person or household against two adverse outcomes: chronic incapacity to work and earn (chronic poverty) and a decline in this capacity from a marginal situation that provides minimal means for survival with few reserves (transient poverty). A chronic incapacity to work or earn will usually result from physical or mental disability, long-term illness, or old age. A decline in the capaci- ty to work is usually caused by imperfectly predictable life-cycle events (such as the birth of twins or a sudden death of a bread-winner); a INTRODUCnON 3 sharp fall in aggregate demand or expenditure shocks (through eco- nomic recession or transition, during unavoidable cutbacks in public spending, as a result of a decline in production in sectors from which workers are immobile); or poor harvests (due to drought, flood, or pests, especially when they affect prices and production over a wide area), which cause the rural poor to lose the usual sources of protection offered by informal transfers (see discussion below). There is a broad range of mechanisms for protecting individuals from acute deprivation or inadvertent declines in income. In some societies informal or community-based arrangements (private safety nets) help mitigate the adverse outcomes in welfare. In addition, pub- licly supported social safety nets also help the vulnerable. These include social services (health and education in particular), social insurance programs such as pensions, all publicly funded transfers (cash transfers such as family allowances and in-kind transfers such as food subsidies), and income-generation programs targeted to the poor (such as public works). This book focuses on publicly funded transfers aimed at the poor and on income-generation programs. Although social insurance programs constitute the most dominant form of cash transfer in most countries of Eastern European and the former Soviet Union, and provide relief to the poor in the formal sectors, these pro- grams are not addressed here because issues pertaining to pensions were the focus of a recent World Bank policy study (Fox 1994). Where necessary (as in the discussion on cash transfers in countries of the for- mer Soviet Union), the study will make reference to the interface between social insurance and social assistance programs. Who Needs Safety Nets? In every society some individuals are able to take advantage of emerging opportunities better than others, in the same way that some individuals are more vulnerable to chronic or transient declines in income. Individuals are likely to be at a disadvantage in either situation if they are: * Most prone to unpredictable, sharp falls in real spending power. * Least resilient to a reduction in real spending power. * Least likely to have formal insurance. * Least likely to be attractive to informal insurance groups (such as reciprocal savings or insurance societies), which seek dependable co-contributors. Thus the poor are most likely to be in need of safety nets. In the absence of publicly supported safety nets the poor are likely to turn their production and consumption behavior toward risk avoidance instead of toward income maximization. In other words, the poor often cannot afford to be entrepreneurial-and may remain underproductive 4 SAFETY NET PROGRAMS AND POVERTY REDUCTION and poor. In order to counteract this, cash and in-kind transfers, and poverty-targeted programs-the main focus of this study-should be examined in a broad developmental context, while ensuring that the poor also have access to the gains from economic growth. Cash and In-Kind Transfers in a Developmental Context Any cash or in-kind transfer requires real resources, and budgetary limitations imply that there are often difficult trade-offs between trans- fers and other programs supported by tax revenues. Therefore, it is important to coordinate transfer programs with broader developmen- tal goals. These programs must be formulated in the context of a poverty-reduction strategy, comprising initiatives for labor- demanding growth and expanded access to social services for the poor. Targeted transfers to provide temporary relief to the poor should never replace development programs, such as irrigation, education, or infrastructure development, that enable long-term escape from pover- ty. In some countries or regions investments in drought-proofing or in activities such as village-to-market roads can significantly stabilize the poor's employment possibilities and incomes, dramatically reducing the need for short-term poverty relief (Ahmed and Donovan 1992; Binswanger, Khandker, and Rosenzweig 1993). Thus the type and structure of demand for safety nets may vary with an economy's size, age structure, health, real average income, and type and extent of fluctuations in production, and may interact with other eco- nomic policies. A consequence is that as growth or redistribution increas- es a poor person's expected (average) annual real income, his or her capacity to save, insure, spread short-term risks, and meet both acute and chronic shocks without disaster also increases. In addition, governments opting for fast liberalization must factor in both the need for more spend- ing on safety nets in the short to medium run and the likelihood that this spending will drop in the longer run. Well-distributed real income reduces the need for safety nets and increases an individual's ability to cope with distress. Also, policies that spend scarce resources on irriga- tion, control of infectious illness, labor-intensive production, and land redistributions-whatever their advantages or drawbacks-save resources by reducing the need for publicly provided safety nets. Therefore, under no circumstances should safety net policies be framed as alternatives to investments in labor-demanding growth activities. Public Transfers and Private Safety Nets Valuable empirical information is now available demonstrating the importance of personal reciprocity and informal safety nets in devel- INTRODUCTION 5 oping countries and in some economies in transition (table 1.1). Traditional coping mechanisms reduce economic insecurity.4 Public transfers, therefore, should complement not only broader develop- mental activities, but also the prevailing traditional arrangements. The usefulness of public transfers for poverty relief is greatly reduced if public transfers induce reductions in private transfers (remittances, and so on) to the poor. But it is important not to exaggerate the role of private transfers in poverty reduction-as is commonly done by wealthy taxpayers to undermine public support for the poor. Private transfers could be minimal if an entire country or parts of a country are visited by a drought, or if transfers are made between groups that are poor.5 Cross-Country Patterns in Program Choice Many countries have experimented with a wide array of social assis- tance and poverty-targeted programs. The choice and scope of instru- ments varied a great deal across countries, depending on circum- stances. The degree of success in reaching the poor in a cost-effective way also varied with the extent and depth of poverty, the composition of target groups, the speed and sequencing of reforms, the availability of financing, administrative capacity, and political economy factors. Often, the same instrument produced quite different outcomes and effects on the poor in different countries and across regions in the same country. The situation prevailing from the mid-1980s to the mid-1990s is the result of the historical evolution of programs (table 1.2). Four patterns are worth noting: @ Program choice was to some extent influenced by the nature and availability of food aid. Thus in the 1950s and 1960s food aid (par- ticularly from the United States) led to the emergence of food-for- work programs, especially in South Asia. With the decline of food aid in the 1970s and 1980s, food-for-work programs also diminished in importance-now, only two countries have such programs.6 During the 1980s availability of direct nonfood assistance by (bilat- eral) donors and an eagerness to channel such assistance through nongovernmental channels led to the emergence of social funds. * A few countries sought to realize multiple objectives with one pro- gram. For example, public works programs that give staple food as wages seek to relieve poverty by ensuring that participants have access to food, are able to smooth their consumption, and build infrastructure-thus complementing economic growth. * Countries varied in the pervasiveness of interventions: some relied on one major program (such as family assistance in the countries of 6 SAFETY NET PROGRAMS AND POVERTY REDUCTION Table 1.1. The Importance of Private Transfers in Select Countries Country and segment National GNP per capita of the population Year (1993 US dollars) El Salvador (urban poor) 1976 1,320 Mexico (two villages) 1982 3,610 Peru 1985 1,490 India (rural) 1975-83 300 Indonesia (Java) 1982 740 Rural Urban Malaysia 1977-78 3,140 Philippines 1978 850 Kenya Rural 1974 Urban 1974 Urban (sample of recent migrants) 1968 270 Nairobi (urban poor) 1971 Nationwide 1974 Kyrgyz Republic 1993 850 Romania 1994 1,140 Rural Urban Russia 1992 2,340 United States 1979 24,740 - Data not available. a. Number in parentheses denotes percentage for households in lowest income quintile. the former Soviet Union), while others (such as India) experiment- ed with a rich array of programs, including food subsidies, nutrition programs, public works, and credit-based livelihood programs. * Labor-intensive public works (run either by governments or by pri- vate contractors) are more prevalent in Africa and Asia than else- where, whereas food transfers are prevalent in most regions. One difference between public works in South Asia and those in Africa is worth noting. In South Asia public works receive greater support, funding, and direct involvement of governments, while in Africa, works programs are driven largely by donors, and are short-lived.7 Within each program countries chose varying types of intervention. Food-related interventions ranged from open general food subsidies to more targeted programs, such as food stamps and feeding programs. The choice of a particular type of intervention is generally governed by the specific conditions in a country (such as its sociopolitical and eco- nomic history, its climate and agroecological zones, its limited admin- istrative capacity, and so on) and the specific advantages in a country INTRODUCTION 7 Average transfer Percentage of amount (percentage of households average income) Receiving Giving Receiving Giving Source 33 11 Kaufmann and Lindauer (1986) 16-21 Stark, Taylor, and Yitzhaki (1988) 22 23 2 1 Cox and Jimenez (1989) 93 8 Behrman and Deolalikar (1987) 31 72 10 8 44 45 20 3 19-30 33-47 11(46)' Butz and Stan (1982) 47 9 Kaufmann (1982) 19 2 3 Rempell and Lobdall (1978) 62 4 6 59 13 Johnson and Whitelaw (1974) 89 27 3 4 Knowles and Anker (1981) 13 11 41 16 Cox, Jimenez, and Jordan (1994) 7 - 2 Subbarao and Mehra (1995) 5 - 1 8 - 3 24 24 31 12.9 Cox, Eser, and Jimenez (1995) 15 - 1 Cox (1987) Source: Updated from Cox and Jimenez (1990). See the last column of the table for sources for different countries. The data for the Kyrgyz Republic, Romania, and Russia are taken from sources listed in the table. Data for GNP are taken from World Bank (1995m). (such as the availability of private transfers). Thus in South Asia food rations emerged as a result of postwar shortages, but evolved into other schemes that continued to provide in-kind transfers for the poor long after food shortages had abated. Variations in program choices and characteristics, total public spending as a share of GDP, and the amount of public spending on safety nets as a share of GDP across countries may be explained by variations in countries' objectives, historical circumstances, level of real purchasing power parity (PPP) GDP per capita, the inherent nature of constraints, and the degree of buoyancy of private transfers (Jimenez 1993). A broad, regional typology sets the high-income Organization for Economic Cooperation and Development (OECD) countries, where private transfers (as a fraction of average household income) are expected to be minimal,8 and which typically can be regarded as "high public spenders," at one end of the spectrum (table 1.3). At the other end of the spectrum are the low-income countries of Asia and Africa, where private transfers (as a fraction of average 8 SAFETY NET PROGRAMS AND POVERTY REDUCTION Table 1.2. Classification of Countries by Programs, Late 1980s to Mid-1990s Income distribution PPP GDP GNP per capita per capita Poorest Richest Puiblic Country (1993) (1993 int. $) 20% 20% works Suib-Saharan Africa Botswana 2,790 - - - x Ethiopia 100 - 8.6 41.3 x The Gambia 350 1,170 - - Ghana 430 1,970 7.0 44.1 x Kenya 270 1,290 3.4 61.8 x Lesotho 650 1,620 2.9 60.0 x Madagascar 220 670 - - Malawi 200 690 - - x Mali 270 520 - - Mozambique 90 550 - - x Nigeria 300 1,400 5.1 49.0 x Senegal 750 1,650 3.5 42.8 x Somalia - - - - Tanzania 90 580 2.4 62.7 x Zambia 380 1,040 5.6 49.7 Zimbabwe 520 2,000 4.0 62.3 South Africa 2,980 - 3.3 63.3 Middle East and North Africa Algeria 1,780 5,380 6.9 46.5 Egypt 660 3,780 - - Morocco 1,040 3,090 6.6 46.3 Tunisia 1,720 4,780 5.9 46.3 Jordan - 4,100 6.5 47.7 Souith Asia Bangladesh 220 1,290 9.5 38.6 x India 200 1,220 8.8 41.3 x Pakistan 430 2,170 8.4 39.7 Sri Lanka 600 2,990 8.9 39.3 x East Asia China 490 2,330 6.4 41.8 Indonesia 740 3,150 8.7 42.3 Philippines 850 2,670 6.5 47.8 x Thailand 2,110 6,269 6.1 50.7 x INTRODUCTION 9 Program types Credit-based Family assistance Food livelihood and/or cash Energy Houising transfers programs social assistance suibsidies suzbsidies Othersa SAP x SF x x x x SAP x x x SF/SAP x SF x x x SF x SF x x x x SAP x x x SF x x x x x x x x x SAP x SF Regional x x x x x x (Table continutes on following page.) 10 SAFETY NET PROGRAMS AND POVERTY REDUCTION Table 1.2 (cont.) Income distribution PPP GDP GNP per capita per capita Poorest Richest Public Country (1993) (1993 int. $) 20% 20% works Latin America and the Caribbean Bolivia 760 2,420 5.6 48.2 x Brazil 2,930 5,370 2.1 67.5 Chile 3,170 8,400 3.3 60.4 x Colombia 1,400 5,490 3.6 55.8 Costa Rica 2,150 5,520 4.0 50.8 Ecuador 1,200 4,240 - - Guatemala 1,100 3,350 2.1 63.0 Guyana - - - - Haiti 600 2,990 - - Honduras 600 1,910 2.7 63.5 x Jamaica 1,440 3,000 6.0 48.4 Mexico 3,610 6,810 4.1 55.9 Nicaragua 340 1,900 4.2 55.3 x Peru 1,490 3,220 4.9 51.4 Uruguay 3,830 6,380 - - Venezuela 2,840 8,130 4.8 49.5 Eitrope and Central Asia Bulgaria 1,140 4,100 8.4 39.3 Czech Republic 2,710 7,550 - - Hungary 3,350 6,050 10.9 34.4 Kyrgyz Republic 850 2,320 2.5 57.0 Poland 2,260 5,000 9.2 36.1 Romania 1,140 2,800 - - Russia 2,340 5,050 4.2 48.0 Ukraine 2,210 4,450 - - - Data not available. Note: Program types are based on cross-country reviews in World Bank (1995m); this is not an exhaustive classification of countries. household income) are significant. Public spending is predictably low- est in these regions. In between fall the middle- and low-income economies in transition, all maintaining unsustainably high levels of public spending on safety nets. As objectives, constraints, and country circumstances change, pro- grams also change, though only rarely are programs discontinued entirely. Thus most countries in transition are attempting to curtail open-ended transfers that they can no longer afford. Various targeting options are being explored. In some countries a particular program may generate a constraint and induce its own change. For example, an expensive and untargeted program might lead to fiscal tightness, which, in turn, leads to a change in its design. INTRODUCTION 11 Program type Credit-based Family assistance Food livelihood and/or cash Energy Houtsing transfers programs social assistance subsidies sutbsidies Othersa x x x x x x x x x x SF x SF x SF x SF x x SF x x x x SF x SF x x SF x x x x x x SF x x x x x x x x x x x x x x x x x x x x a. SAP is Social Action Program. SF is Social Fund. Source: World Bank (1995m). Objectives and Outline of the Book The main objectives of this book are to review the conceptual issues that arise when choosing among programs; analyze how country- and region-specific constraints help to explain why different approaches are successful in different countries and regions; synthesize the lessons learned from experience; and offer some promising best-practice examples. The study's scope is limited to social assistance and poverty- targeted programs, of which there are two kinds: programs with and without work requirements. Programs with no work requirements include two principal cash transfers (prevalent especially in countries 12 SAFETY NET PROGRAMS AND POVERTY REDUCTION Table 1.3. Regional Patterns in Safety Nets Pu7blicly suipported social secturity (contributory pensions) + means-tested public houtsing Coutntry type and medicare to all in need High-income OECD countriesa x Middle-income countries in Latin America High/middle-income countries in East Asia Middle-income countries in transition Low-income countries in transition Low-income countries in Asia Low-income countries in Africa Food price subsidies, supported rations Maternal (other than and nutrition Cotntry type subsidies) food stamps High-income OECD countriesa Middle-income countries in Latin America x High/middle-income countries in East Asia Middle-income countries in transition x Low-income countries in transition Low-income countries in Asia x x Low-income countries in Africa x x a. The United States is the only OECD country to offer a food stamp program. b. Data on private transfers are not available for OECD countries, but these are in transition) and three in-kind transfers (prevalent in most developing countries and in countries in transition). Programs with a work requirement include public works and credit-based income- generation programs (prevalent in many developing countries). The selection of programs and countries to be included in this review is limited by the availability of rigorous evaluations. The programs and countries chosen do not constitute a representative sample in a statis- tical sense. Chapter 2 outlines the issues pertaining to targeting and the accom- panying trade-offs between cost and efficiency. Chapters 3, 4, and 5 discuss country experience with cash transfers, in-kind transfers, and income-generation programs. Three common issues are: how success- fully the programs reached the poor (targeting and distributional effi- ciency), how cost-effectively they reached the poor, and whether or not the programs were able to avoid adverse incentives for work, labor supply, and private saving.9 Chapter 6 evaluates the design issues and delivery mechanisms generic to all programs, and develops a fourfold typology of program delivery. This chapter also examines how social funds-the most pop- INTRODUCTION 13 Publicly supported Privately, social sectrity filly-funded toformal sector social Food Family Social employees only secuirity stamps assistance assistance x x x x x x x x x x x x x Region/country characteristics Credit-based Putblic works, Range of pueblic livelihood food for Private spending (as percentage programs works transfers of GDP) Negligibleb 10-15 x Low 8-15 Moderate 2-5 Moderate 10-25 Moderate 10-15 x x High 2-5 x x High 1-2 believed to be negligible. Source: Authors' assessments. ular institution in some countries-generate different outcomes depending on which delivery mechanism is chosen. Chapter 7 ana- lyzes the issues involved in financing safety-net programs, reviewing the limited empirical information. Chapter 8 analyzes aspects of polit- ical economy through a country case-study approach. Countries stud- ied include Bolivia, Chile, El Salvador, Ghana, Mexico, Peru, Poland, Senegal, Ukraine, and Zambia. Chapter 9 sums up the principle lessons from cross-country experience, and delineates the issues that should be borne in mind when choosing or designing programs. Notes 1. Structural adjustment promotes incentive and regulatory reform, accompanied by a reduction in the role of the state. Macroeconomic stabilization is intended to ensure fis- cal and monetary discipline. 2. Social insurance is defined as benefits provided by the state, most often through specified contributions, to cover specific risks. It includes unemployment benefits, sick- ness and maternity benefits, and retirement, invalidity, and survivors' pensions. Social assistance refers to a range of benefits in cash or in-kind intended to provide protection for the most needy in society. Family assistance refers to all benefits in cash or in-kind targeted directly to children or families with children. 14 SAFETY NET PROGRAMS AND POVERTY REDUCTION 3. Numerous case studies in the household-coping literature have examined how dif- ferent shocks (environmental, economic, social, and political) affect, and at times under- mine, traditional coping mechanisms based on community and family networks. This literature examines how households deal ex ante with anticipated shocks and what ex post coping strategies they apply following an adverse event. In particular, see Platteau (1991); Agarwal (1991); regarding the economic impact of AIDS, see Ainsworth and Over (1992); see Alderman and Paxson (1992) for a discussion of issues pertaining to risk shar- ing and consumption smoothing. 4. Private transfers may take various forms. In Sub-Saharan Africa community-level saving societies contribute to pooling and spreading of income risks. In Mali urban asso- ciations called 'tontines" function as saving and insurance clubs, providing cash to members in difficult times (World Bank 1993a). There are similar groups in The Gambia and Burkina Faso (von Braun 1989; World Bank 1993b). Chapter 4 discusses the Islamic- based Zakat and Ushr system, which is a common form of safety net in parts of South Asia and some African countries. In Mongolia exchange of animals is a common form of private transfer (Subbarao and Ezemenari 1995). 5. Recent evidence from the Philippines illustrates this. Subbarao, Ahmed, and Teklu (1995) found that remittances from abroad were a significant source of household income in the relatively more developed regions of the country. In addition, within these regions a large proportion of the transfers went to upper-income groups. Domestic remittances, on the other hand, are predominant among the poorer income groups. The study concluded that foreign remittances accentuated income inequality among regions and households. 6. A food-for-work program, largely driven by P.L. 480 food aid from the United States, operated in India until the mid-1960s. The program was replaced by a regular employment program (now known as the Jawahar Rojgar Yojuna) with wages paid in cash rather than in-kind. As the composition of food aid changed from grains to infant formula, a program (introduced in the mid-1970s) aimed at improving child nutrition began to absorb the donor-supplied infant food. 7. A relatively lower emphasis in Africa than in Asia on labor-intensive public works programs, or other comparable programs for consumption smoothing, is probably the rea- son for the harsher impact of a drought in Africa than in South Asia. Thus a massive drought visited India in 1987, affecting nearly one-third of the population. Yet there was not a single drought-related mortality. This was because of the prevalence of consumption- smoothing public works programs, an administrative and institutional framework that is well-adapted (over time) to drought situations, and a free press. By contrast, a drought of much smaller magnitude in Ethiopia caused considerable distress and mortality in 1991. Moreover, rural infrastructure has remained in a very poor and neglected state in Africa- hurting the poor both as sellers (low product prices) and as buyers (high input prices). 8. As of now, no data exist on the share of private transfers in average household income for most OECD countries. 9. Not all of the issues could be addressed for each of the programs owing to pauci- ty of data. In particular, data limitations narrowed the scope of chapter 3 (cash transfers). 2 Conceptual Issues in the Analysis of Social Assistance and Poverty-Targeted Programs This chapter reviews the conceptual issues pertaining to public trans- fers and targeting, the impact of poverty-targeted programs on con- sumption decisions and labor market behavior, fiscal sustainability, and some political economy factors. The chapter suggests that the issue of whether to target or not depends on the costs and benefits of doing so. It then discusses how to target by describing the conceptual ideas behind the different targeting mechanisms. It concludes with a discussion of the trade-offs in delivering benefits in cash or in-kind. The rationale for targeting is that the social returns for a given level of transfers are higher for individuals or households at the lower end of the income distribution than at the upper end.1 To maximize the welfare effect of a transfer program, the appropriate target would be the population segment deemed poor according to some criterion. Thus the ability to measure poverty and identify the poor is essential for designing any targeted transfer program. Targeting gains depend on initial country conditions and the type of targeting mechanism adopted. Targeting the poorest and most vulner- able saves budgetary resources by confining benefits initially to those who need them most, provided the administrative costs of identifying the poor are not high. Saving budgetary resources allows for greater benefits for the poor, greater expenditures on other programs-possi- bly income-generation programs-or the reduction in public revenue collection through potentially distortionary taxes. Targeting Costs Targeting is beneficial when it increases the efficiency of antipoverty programs, but it also entails costs. Four types of costs stand out: administrative, incentive (deadweight), stigma, and political. 15 16 SAFETY NET PROGRAMS AND POVERTY REDUCTION Administrative Costs Administrative costs are those associated with identifying, reaching, and monitoring the target population. Targeting requires some mech- anism to distinguish the poor from the nonpoor so that the poor can avail themselves of benefits, and the nonpoor can be prevented from receiving benefits. Quantifying administrative costs by program is not easy, particularly if costs are shared across programs. Costs will also differ across program types. Following Besley and Kanbur (1993), the total revenue required to finance a program, R, can be broken down into three components: administrative costs, A, leakage to the nonpoor, L, and the effective transfer to the poor, P. Targeting efficiency is defined by the ratio P/(L+P). The proportion of administrative costs to total revenue rises according to how meticulously the poor and the nonpoor are sorted, as measured by the ratio P/(L+P). The trade-off between achieving finer targeting and curtailing administrative costs is often formulated in terms of type 1 (exclusion) and type 2 (inclusion) errors (table 2.1). The numbers in table 2.1 represent the population covered by a given poverty-targeted program (disaggregated according to poverty status). If it is assumed that all individuals covered by the program receive one dollar of program benefits, then P/(L+P) = 30/(20+30) = 0.6. Leakage is defined as the total percentage of nonpoor covered divided by the total percentage of people covered by the program (20/50=0.4); and coverage is defined as the percentage of the poor who are covered by the program divided by the total percentage of poor (30/40=0.75). This is often referred to as the participation rate. Lower levels of both inclusion errors and exclusion errors are preferable, though, as mentioned above, minimizing these errors can be costly. It is difficult to compare the trade-offs between type 1 and type 2 errors. The higher the priority given to raising the welfare of the poor, the greater the need to eliminate type 1 errors. Conversely, if more weight is given to saving budgetary resources, it is important to eliminate type 2 errors. Strict administration of a transfer program Table 2.1. Program Leakage and Coverage (percent) Total Covered Not covered population Poor Success Type 1 (exclusion) error 30 10 40 Nonpoor Type 2 (inclusion) error Success 20 40 60 Source: Grosh (1994). CONCEPTUAL ISSUES 17 may exclude not only some of the nonpoor but also some of the poor. At times, it may be prudent to deliberately allow leakage so as not to exclude any of the poor. Whether it is worth incurring the administrative costs to improve the targeting efficiency of a program depends not only on how well the targeting mechanism screens eligible persons but also on the value of the benefit delivered. Administrative costs are determined by the level of targeting (self-targeting is less costly than geographic targeting, and individual assessment is the most costly); the level of automation and information exchange (computers that hold data on all government programs reduce the cost of income testing an individual); the income/asset profile of the country (formal income is relatively easy to assess, whereas informal income and the value of land is difficult to assess); and the frequency of change in the benefit rate (the needs of a pensioner are likely to remain unchanged, whereas the needs of an unemployed family member could change frequently). A judgment must be made about the administrative costs of a benefit relative to the program costs. Grosh (1994) estimates (across various programs and countries) that targeted transfers cost about 3-8 percent more than a similar universal transfer. Administrative costs also vary with the type of mechanism used, the level of existing information, the institutional capacity of a country, and the local costs of personnel and equipment needed to carry out the targeting. Incentive Costs Targeting schemes may foster both positive and negative incentive effects. Three negative incentive effects that may outweigh the benefi- cial impact of targeting are distortions in the labor-leisure choice, migration to advantaged areas, and the unproductive use of resources or time (moral hazard problems). Targeted programs can also foster positive incentives: a food subsidy targeted to parents whose children attend school encourages higher school enrollments. The disincentive to work is one problem associated with programs targeted according to a means test (usually income). Agents may change their behavior by altering their labor supply and hence their income. For example, imagine a means test that proposes a monthly income of $350 as an eligibility threshold. Individuals with a monthly income less than $350 are eligible to receive $50 in benefits with no work requirement. Individuals with monthly incomes in the range $350-$399 (in the absence of the program) would be better off adjust- ing their labor supply to earn just less than $350, thus qualifying for the $50 income supplement. Their incomes would be higher than 18 SAFETY NET PROGRAMS AND POVERTY REDUCTION before, and they would be better off. But society as a whole would lose their output. Furthermore, program costs would go up since the num- ber of claimants would go up.2 Programs can also generate disincen- tive effects in the community-a generous food subsidy may lower community self-help efforts. To the extent that marginal tax rates affect incentives to work, and thus income, income is an endogenous variable and depends on the tax rate implicit in the targeted program (Besley and Kanbur 1993). With a 100 percent marginal tax rate, no one with an income below the poverty threshold would have an incentive to work-they would instead choose to receive government payments to pull them out of poverty. But if all those with incomes below the poverty line chose not to work, the financial burden of the transfer would rise substantially, since each person below the poverty line would require a transfer equal to the full poverty gap. This result implies that the marginal tax rates on the rich would have to be higher to finance the program's increased costs. In turn, the rich would now have fewer incentives to work (since their incomes are being taxed away), meaning less rev- enue would be available for the intervention. Thus labor supply effects are important, though only some targeted programs have adverse effects. Relocation is another potentially undesirable effect. If benefits are targeted geographically, people might have an incentive to migrate to areas covered by benefits. If the bulk of the movers are poor, then an increase in costs is justifiable, since more of the poor are being covered. But if better-off households also move, then program costs would increase without justification. If the relocation also causes problems, such as congestion in service delivery and a reduction in the number of benefits available per eligible person, the net costs to society could be large. The extent of the relocation effect depends on the size or amount of the benefits gained relative to the moving costs. If the rich have access to more efficient transportation and information networks, their unit costs of transportation may be low, and the relocation effect could be quite large. Incentive compatibility problems also give rise to undesirable pro- gram effects. For example, in a child nutrition program rations given to children are planned according to what children were fed before they entered the program. Two kinds of incentive compatibility prob- lems may arise. Mothers may feed children less (because they know they are being fed in nutrition programs) so that other members of the household can eat more (adverse incentive problem). Or, an inadver- tent moral hazard problem may arise because the nutritional status of the children does not improve. Thus these children must stay in the program for a longer period.3 CONCEPTUAL ISSUES 19 Dreze and Sen (1989) provide a rather drastic example of the adverse incentive problem. If food supplements in a nutrition program are based on children's nutritional status, mothers may have an incentive to gain eligibility by underfeeding their children before weighing them. Even though there is little empirical evidence to support this example, this type of behavior would generate a "deadweight" loss, since a wedge is created between private and social (collective) welfare. Stigma Costs Stigma is a possible negative consequence of means-tested govern- ment transfer programs. It arises from two main sources. First, recipi- ents may lose their self-esteem because they regard themselves as fail- ures who must rely on public support. In such cases needy people are unnecessarily screened out. Second, nonrecipients may have negative attitudes toward recipients, and recipients may be treated differently from nonrecipients by both program officials and the general public.4 Individuals not wanting to be classified as needy could refuse subsi- dies and depress the take-up rate. Stigma may not act as a deterrent to the nonpoor, particularly if the receipt of one transfer offers access to other transfers or privileges. Or, stigma may assist in targeting. Requiring welfare recipients to perform menial tasks may impose relatively smaller costs on the intended recip- ients who have very limited options anyway, while deterring impostors who may have other attractive options (Nichols and Zeckhauser 1982).5 Political Economy Costs Targeting raises interesting political economy issues. First, any inter- vention creates losers as well as gainers. The goal is to transfer resources from the gainers to the losers so that the losers are just com- pensated for their losses, while the gainers are still better off than they would have been without the targeted intervention. But the distribu- tion of political power in a country may not permit this goal to be real- ized. Second, the interests of the different players involved in admin- istering the program will shape how it is eventually implemented. The poorest strata of society are often the most difficult and costly to reach, and therefore intervention efforts tend to focus on the most vocal and organized groups, which are not necessarily poor. Third, very fine tar- geting may sometimes run counter to the interests of the poor. If a program is well-targeted and the poor are relatively disenfranchised, the program may have scant political support and be allotted a small budget. In contrast, a broadly targeted program may elicit greater political support and a larger budget.6 20 SAFETY NET PROGRAMS AND POVERTY REDUCTION Good targeting should increase take-up rates among intended ben- eficiaries and reduce leakage. It should minimize distortions, reduce program costs, and distribute scarce public resources fairly and effi- ciently. These objectives invariably imply difficult trade-offs. Leakage is minimized only at the expense of higher administrative costs. Finer targeting carries some benefits, but it also imposes costs and distor- tions, and may lead to loss of political support. The decision to target more finely, therefore, hinges on the balance between these benefits and costs, which vary from country to country depending on the eco- nomic, political, and social fabric of specific economies. These consid- erations imply that targeting mechanisms should provide incentives for the target group (and no others) to seek the subsidy. Targeting Mechanisms The mechanisms used to identify the poor can be classified according to three broad categories: individual targeting, geographic or indicator targeting, and self-targeting. Any of these types of targeting could be applied to the individual or the group. This flexibility is important because some of the information and incentive problems that seem daunting at the individual level may not be so at the group level. An example of an individual assessment is a means test that ascer- tains whether household income is below the cutoff point. The test uses either income, nutritional status, or type of dwelling (with or without electricity, toilet, and so on) as a criterion. If it is possible to conduct a "perfect" means test, cash transfers dominate other forms of interven- tions because they carry only an income effect and do not distort pri- vate consumption decisions at the margin. But because of the informa- tional and administrative constraints, means testing may not be perfect and may induce costly leakages and create adverse incentives.7 These considerations have led to a variety of schemes for indicator tar- geting, whereby transfers are made contingent on some correlates of poverty. Geographic and indicator targeting mechanisms grant eligibili- ty to groups of individuals on the basis of an easily identifiable shared characteristic. The idea is to find an indicator that is less costly to identi- fy (than income) but is sufficiently correlated with income to be useful for identifying the poor (Besley and Kanbur 1993). Although some leakage is inevitable, the cost of leakage can be offset in part by reduced admin- istrative costs. The indicators used for targeting should not be easily manipulable during the course of the intervention since adverse incen- tive effects could result. For example, individuals might fake attributes that would qualify them for assistance (such as ownership of land or other durable assets) or change their behavior to legitimately acquire such attributes. Indicators that are good predictors of income are land CONCEPTUAL ISSUES 21 ownership, house (floor area) or other durable asset ownership, number of children, level of education, and so on. Most of these indicators are easy to observe and difficult to manipulate in the short run. But if the number of indicators increases, the situation may degenerate to one in which every individual is "tagged" separately, enhancing administrative costs.8 Geographic targeting allocates resources to states, municipalities, or neighborhoods based on their average welfare leveL The headcount index is commonly used to establish a geographic ranking of poverty. Alternatively, combinations of region, residence, and household characteristics could be used, as in Colombia, where areas of poverty were first identified as part of the national development plan. Food subsidies were then targeted to households with children under five, or to pregnant or lactating mothers. Prospects for reaching the poor through regional and indicator targeting depend on a number of fac- tors, including local administrative infrastructure, the incentives fac- ing local administrators, their social relations with the poor, and the extent of political representation of the poor. Self-targeting occurs when a program is ostensibly available to all but is designed to discourage the nonpoor from participating. For example, participants can be requived to work, as in public employ- ment programs. Alternatively, a low-quality product could be subsi- dized, inducing the nonpoor to stay away from the program. If indi- viduals are heterogeneous with respect to ability and tastes, then in- kind transfers of goods or services preferred by the target group may help ward off impostors (Nichols and Zeckhauser 1982). For example, subsidies on coarse flour in Egypt accrued disproportionately to lower- income groups. It is not always easy, though, to obtain accurate infor- mation about the preference structure of individuals. Nevertheless, an important advantage of self-targeted programs is that they are less vul- nerable to admninistrative corruption and bureaucratic manipulation. One problem with all targeted approaches is that along with the nonpoor, some of the very poor may also be screened out. Self-targeted approaches are no exception. For instance, self-selection methods of targeting used in public employment schemes by lowering wage rates do screen out the nonpoor, but some of the poor may also be screened out in the process. The aged, the infirm, or the physically handicapped do not participate in such schemes. Thus countries depending largely on self-targeted public works programs must develop other programs to protect vulnerable groups that are unable to work. The Choice between Cash and In-Kind Transfers In theory, cash transfers are preferable to in-kind transfers because they do not distort individual consumption or production choices at 22 SAFETY NET PROGRAMS AND POVERTY REDUCTION the margin. But only rarely would actual country conditions allow this proposition to be true. Less-than-ideal circumstances (such as the costs involved in conducting means tests) often tilt the choice in favor of in- kind subsidies. There are situations when in-kind transfers are equivalent to cash transfers. For example, a ration of commodity x at a unit price r (where r < p is the price to the target group, and p is the market price) is equiv- alent to a lump-sum transfer of m dollars-that is, m = (p - r)x. This equivalence holds providing there are no restrictions on (or exorbitant costs to) the resale of the subsidized commodity. Even if the subsidized commodity cannot be physically resold (as in the case of education and health services), there may still be an equivalence to a lump-sum transfer under specific conditions. This equivalence can arise if house- holds prefer as much or more of the free commodity (inframarginal transfer), in which case the composition of consumption is not affect- ed (Gill, Jimenez, and Shalitz 1991). Where means tests are imperfect and costly, in-kind transfers may be preferable to cash transfers. When information is less than perfect, gov- ernments must develop mechanisms that enable members of the target group to select a consumption bundle different from the one offered to the nontarget group. Since governments cannot distinguish between individuals, the provision of subsidies with a self-selection mechanism will weed out impostors and constrain program costs (such as subsi- dizing coarse grains consumed only by the poor). While such in-kind subsidies may cause distortions at the margin, these deadweight losses may be outweighed by the decline in leakage. And in some countries with high levels of inflation in-kind transfers have higher real denomi- nations than nonindexed cash transfers. This difference occurs only because the program provider absorbs the cost of inflation when mak- ing an in-kind transfer, while, without indexation, the recipient absorbs that cost. For that reason governments may think they have greater control over the cost of cash transfers. In countries with weak institu- tional and administrative capacities it may be easier to implement in- kind subsidies than cash transfers. In-kind transfers can also be used to reduce intrahousehold discrimination against women or children. Thus information gaps, inflationary conditions, and the need to realize spe- cific policy goals (household food security, nutrition to children) may induce governments to opt for in-kind transfers.9 When in-kind subsidies are chosen, one question that arises imme- diately is which commodity to subsidize? In theory, the government should try to find commodities that form a significant part of the poor's (but not the nonpoor's) food expenditures. The chosen goods do not have to be inferior goods, but they must be goods that are consumed disproportionately by the poor when their prices drop. If these com- CONCEPTUAL ISSUES 23 modities are consumed only by households below the poverty line, leakage will be eliminated. But because it is difficult to isolate com- modities that only the poor consume, errors of exclusion may still arise. Conclusion Poverty-targeted interventions could be provided to the entire popu- lation or aimed at only certain individuals. Because of information gaps, perfect targeting is rarely achieved. The decisions of whether or not to target, which targeting mechanism to use, and how to target depend on the economic and opportunity costs of targeting, negative incentive effects, the size of the target group relative to the total popu- lation, the prevailing differences in consumption patterns between the poor and the nonpoor, the characteristics of the commodity or services selected for targeting (including the size of demand and income elas- ticities), and administrative capability and political economy consider- ations. In particular, minimizing leakage and reducing costs simulta- neously may involve trade-offs. It is important to be aware of these limitations, even if it is not possible to estimate the leakage and costs associated with targeted programs. Notes 1. This chapter draws from Barr (1995), Besley and Kanbur (1993), Lipton and Ravallion (1995), and Gill, Jimenez, and Shalitz (1991). 2. For a thorough treatment of the incentive effects on labor supply see Besley and Kanbur (1988) and Kanbur, Keen, and Tuomala (1994). 3. Kennedy and Alderman (1987) discuss the conceptual issues and empirical evidence associated with program leakage of this kind. Pinstrup-Andersen (1988a) pro- vides estimates of this type of leakage. 4. See Besley and Coate (1991) for additional details regarding the ramifications of stigma associated with targeting. 5. While the problem of stigma is much referred to in the literature on transfer pro- grams, there is very little empirical evidence on this issue. Ranney and Kushman (1987) examine this problem in the context of food stamps in the United States. 6. The history of food subsidies in Sri Lanka since 1977 is a case in point. With the intro- duction of targeting the poor lost the support of the middle class, and the real value of food stamps was allowed to depreciate. Similarly, a food subsidy directed to the poor in Colombia was so tightly targeted that it lacked an effective pohtical constituency and was dropped subsequently with a change of government. A broadly targeted program may foster links between the middle dass and the poor, though at a higher budgetary cost. 7. An interesting counterexample of a successful means test based on community identification is provided by the Antyodaya Program in Rajasthan, India, where the vil- lage elders decide which are the poorest ten families in the village. 8. Ravallion and Chao (1989) illustrate the intuitive proposition that the percentage gains from targeting using indicators are greater when budgetary resources are smaller. See also Glewwe and Kannan (1989) and Grosh and Baker (1995). 9. Basu (1996) argues that in-kind transfers also insulate nonrecipients against sup- ply pressures that cash transfers may create. j 3 Cross-Country Patterns of Cash Transfers This chapter reviews cross-country experience with providing cash transfers for poverty reduction, and explains differences in the type and scale of programs adopted among regions.1 The review highlights issues most relevant to the design and implementation of cash transfers. Cash benefit programs are frequently designed to help indi- viduals and families with incomes just below the poverty line. Often, the programs require that only small amounts of assistance go to the individual or family. A dearth of reliable household data limits the scope for determining how effectively cash transfer programs meet their objectives. Nevertheless, wherever possible, the implications of various programs and targeting mechanisms for reducing poverty are discussed. Common cash transfer programs include social assistance programs and family assistance programs. Social assistance refers to a range of benefits given in cash or in-kind intended to protect the most needy persons in society. Family assistance refers to all benefits given in cash or in-kind targeted directly to children or families with children. These programs are found in various forms in different countries. All have the same primary objective: poverty relief. But the secondary objec- tives vary widely Social assistance has the greatest impact when it is targeted to the poorest and most vulnerable. In most countries chil- dren greatly outnumber adults in households with incomes that are below the poverty level. Thus family assistance will help reduce poverty even when provided to all children. Family assistance and social assistance often have very different objectives. Consequently, seemingly similar programs can have differ- ent outcomes. It is possible to draw conclusions about the effectiveness of one scheme compared with another in terms of costs and effects on poverty-but it is wrong to attempt to rank the outcome indicators because of differing objectives. This point is important to many coun- tries of Eastern Europe and the former Soviet Union that seek to 25 26 SAFETY NET PROGRAMS AND POVERTY REDUCTION emulate Western European models without determining the best model for their specific situation and the best time scale for moving from one model to another. A cross-country comparison suggests that social assistance and fam- ily assistance are used in two different ways. OECD countries use social and family assistance as part of a broader social safety net pro- gram, which often features a social insurance program. Developing economies in Africa, South Asia (and part of India), and parts of Latin America rely more on social and family assistance as the primary transfer mechanism in their social safety net. Countries of Eastern Europe and the former Soviet Union are in a confused position because their inherited social safety nets were built around a social insurance program. Their starting point was therefore more akin to the OECD model. But the changes during transition have been great: loss of secure incomes, steeply rising unemployment, increasing inability to fund social insurance programs, and rising numbers of poor spread throughout society rather than in isolated and identifiable groups. Such changes have forced these countries to cut back on social insurance programs and rely more on social and family assistance programs. Because they now rely so much on social and family assistance, their position is more akin to that of developing countries. Designing the best system for the countries of Eastern European and the former Soviet Union is difficult because their economies are in transition. The problems of poverty are emerging in different sections of the community and are changing at different rates. Also, the inher- ited family assistance and social assistance schemes had different objectives. These countries will require a variety of long-term cash transfer systems for the poor. The type of system will depend on the speed of transition and on the emphasis given to different aspects of transition. The policy context of the chosen system will depend on the long-term prognosis for the economy in general and the social welfare system in particular. It would be helpful, therefore, in designing sys- tems for economies in transition, to compare objectives and outcomes in developed economies with those in developing or emerging market economies. Family Assistance Family assistance programs are usually tied to the number of children in a household. Family assistance can have two main financial objec- tives: to smooth consumption over the life cycle by maintaining per capita household income with each additional child or to maintain a level of income that allows one parent to care for the children full-time. CROSS-COUNTRY PATTERNS OF CASH TRANSFERS 27 In some economies family assistance also has other pronatal objectives. For example, given the aging populations and acute labor shortages, in the former Soviet Union, family assistance was employed to promote high fertility rates (Sipos 1994). Several mid- to high-income countries (such as Argentina, Brazil, Chile, and Uruguay) in Latin America have adopted family assistance programs. They usually help people covered by social insurance, mainly the nonpoor. Family assistance in Latin America accounted for between 3 and 14 percent of total social security spending in 1988 (Mesa-Lago 1990).2 In Latin America family assistance is intended primarily for con- sumption smoothing over the life cycle for households covered under social insurance. But spending on family assistance in these countries declined during the early 1980s (table 3.1). In some Latin American countries costs were pushed to intolerable levels by slack entitlement conditions and high benefits, combined with leakage and inequalities in benefits for different segments of the population. Also, in the 1980s economic crisis, followed by adjustment, eroded the level and quality of benefits. Mesa-Lago (1994:51) indicates, however, that expenditure levels may now be increasing. The Western European Model of Family Assistance Economies in transition are emulating the family assistance model of Western Europe. Thus it is important to understand this model's back- ground. Family assistance was introduced in Western Europe between the late 1930s and mid-1940s. All countries in the region offered some form of benefit by the mid-1950s. The initial intention was to provide easily targeted assistance to families (with children) who were most likely to be among the poorest groups in the post-depression and war years (Kamerman and Kahn 1988:370). Program design varied by country. Some countries paid a flat rate for each child. Others paid a Table 3.1. Family Allowance Expenditures in Latin America, 1961-86 (percentage of total social security) Coutntry 1961 1965 1970 1975 1983 1986 Argentina - - - 27.0 14.3 22.4 Bolivia 30.9 - - - 3.8 2.2 Brazil - - 9.2 - 3.3 3.7 Chile - 45.9 - - 10.3 7.0 Uruguay - - - 16.9 10.7 6.7 - Data not available. Note: Social security expenditure includes that for sickness, matemity, pensions, employment injury, family allowance, and unemployment. Source: ILO (1991) as cited in Mesa-Lago (1994). 28 SAFETY NET PROGRAMS AND POVERTY REDUCTION Box 3.1. Evolution of Family Assistance in the United Kingdom Family assistance was introduced in 1945 as a universal benefit designed to reach all families with more than one child. Assistance was given at a flat rate for the second child and subsequent children. These payments were intended to meet the needs of growing family expenditures, rather than to guarantee income when one adult was out of the labor market. That need changed as families became able to command two incomes; relative poverty became much more concentrated on single-parent fam- ilies, which increased sharply as a proportion of all families. Thus while family assistance remained in place, its real value fell, and a one-parent benefit was added to improve targeting. But during the 1980s the real value of family assistance declined. Increasing the level of assistance would have resulted in large cash transfers to the nonpoor, so a family credit was introduced, in which a family with low eamings could have its income supplemented up to a predetermined level. Family assistance in the United Kingdom has always been given without regard to income. And as such, it provides a continuing income source for the poor even if family income improves. Thus it creates an incentive to become self-sufficient, while also smoothing out potential poverty traps. The United Kingdom has gone from offering a simple uni- versal benefit aimed at supporting larger families to a complex group of benefits with a number of different secondary policy intentions, all with the aim of providing cash transfers to families with children. These ben- efits are increasingly targeted to the poorest group (single parents)-the poorest identified by a simple income test. Note: Family assistance is only one small part of the comprehensive social safety net in the United Kingdom, and this illustration should not be seen as an exam- ple of best practice. variable rate for the second and subsequent child. But they all had the same objective: protecting people in poverty. In time, the objective of family assistance changed to reflect changes in the nature of family poverty. These programs have worked well to reduce poverty (table 3.2). Moreover, these programs are only a small part of a much bigger social safety net system. Thus they should not be viewed in isolation from other social insurance programs. For example, family assistance in the United Kingdom must be judged by how it affects social assis- tance and general tax legislation (box 3.1). It is a mistake for any coun- try trying to develop family assistance to copy another country's model without considering the broader national social safety net pro- gram and its fiscal implications, particularly how revenues are raised to finance a program. CROSS-COUNTRY PATTERNS OF CASH TRANSFERS 29 Table 3.2. The Role of Public Transfers in Reducing the Poverty Gap among OECD Families, Mid-1970s to Mid-1980s (percentage of family poverty reduced) Families with children Single-parentfamilies Social Means-tested Child Social Means-tested Child insutrance programs allowances insutrance programs allowances Australia n.a. 87 13 n.a. 88 12 Canada 38 48 14 19 69 12 Germany, Fed. Rep. of 68 11 21 67 16 18 Norway 86 3 11 83 4 13 Sweden 52 37 11 45 45 10 Switzerland 93 7 n.a. 92 8 n.a. United Kingdom 38 38 24 15 63 22 United States 29 71 n.a. 7 93 n.a. n.a. Not applicable. Source: Smeeding, Torrey, and Rein (1988), tables 5.2 and 5.11. Except for Italy, all countries in Western Europe pay family assis- tance as a universal benefit without regard to income. But they all have separate systems with variable rates, depending on age or number of children. All countries, except for Denmark, have improved targeting and reduced expenditures as a percentage of GDP between 1980 and 1990 (table 3.3). Western Europe's family assistance benefits are much larger relative to GNP in the north (Belgium, Denmark, and France) than in the south (Greece, Spain, and Portugal), which has higher lev- els of poverty. This difference reflects the north's higher income levels, its better ability to provide transfers, and a greater willingness of the extended family to support assistance socially and politically. Note that a higher share of GNP need not imply higher benefits per person at risk. Table 3.3. Family Assistance Expenditures (percentage of GDP) Country 1980 1990 Belgium 3.1 2.2 Denmark 2.7 2.9 France 2.6 2.2 Greece 0.4 0.2 Italy 1.0 0.6 Netherlands 2.6 1.7 Portugal 0.9 0.8 Spain 0.5 0.1 United Kingdom 2.3 2.0 Source: Eurostat (1993). 30 SAFETY NET PROGRAMS AND POVERTY REDUCTION Family Assistance in Economies in Transition Prior to the transition the objective of family assistance programs in Eastern Europe and the former Soviet Union was to maintain a level of per capita income as family size increased. Assistance was also provided as a supplement to wages. Take-up of the allowance was almost universal; administration was handled by the enterprise. If both parents worked or if one parent worked for more than one employer, duplicate payments were common. Family assistance now accounts for 2-4 percent of GDP among countries in the former Soviet bloc, and up to 7 percent of household expenditures in the Kyrgyz Republic (table 3.4). After the transition many governments modified their near-univer- sal coverage (box 3.2). The Kyrgyz Republic discontinued family assis- tance to children older than age eighteen. It also introduced income testing for family assistance. In Moldova universal coverage was with- drawn, and family assistance was income tested. All countries now support families with children if either parent is out of work or if fam- ilies become too large to support themselves on two incomes. Over the course of the transition expenditures on family assistance (as a percentage of GDP) have varied among countries (table 3.4). Countries can be divided into three groups according to spending trends. In the first group of countries expenditures have either been maintained or increased. In the second group expenditures have been reduced to preindependence levels (figure 3.1). And in the third group (Belarus, Georgia, Lithuania, and Russia) expenditures increased until 1991-92, then decreased thereafter. In the latter two groups countries appeared to protect social insurance (pensions) to the detriment of social and family assistance, as fiscal constraints tightened. Moreover, linking family assistance to nonindexed minimum wages led to a Box 3.2. The Evolution of Family Assistance in Poland Poland inherited a program of family assistance that was paid by the enterprise as a part of compensation. In four years the benefit has changed from being available to all workers, regardless of income, to being available to only the poorest families in and out of work. The pro- gram was still in existence in 1991-92, and the value of assistance was increased in line with rising wages. At that time only families with at least one employer received benefits. In 1993 entitlement was extended to the unemployed, and the benefit was paid at a flat rate and not increased in line with either price or wage increases. In 1994 income testing was intro- duced, and only families with a low per capita income became eligible. Source: Transmonee data set. CROSS-COUNTRY PATTERNS OF CASH TRANSFERS 31 steady decline in expenditures. At the same time, poverty has increased among countries in this region. Generally, rural areas have been adversely affected.3 Hardest hit were individuals, such as young children and adults, the working poor and the unemployed, the dis- abled, and the aged without family support (Sipos 1994; Milanovic 1995a).4 In most countries with hyperinflation family assistance has not been indexed to inflation. Because the poor suffer most from infla- tion, family assistance has not adequately protected the people in greatest need. Data are lacking that will help to determine the impact Figure 3.1. Share of Expenditures on Family Allowance in Selected Countries in Transition, 1989-94 (percentage of GDP) Those that maintained and/or increased 5 4 H gary 3 >,w >_ ~~~~~Czech 4 ''Armenia Latvia 2 7ts"Z**__* _------ Estonia Poland 1 ~~~~~~~~~Ukraine 1989 1990 1991 1992 1993 1994 Systemic decline 4 " -_ - Slovakia " '~~~~-^,^ A~~~~Azerbaijan 2 Bulgaria 1 ''* ~~~~~Romania 0 1989 1990 1991 1992 1993 1994 Source: Transmonee data set. 32 SAFETY NET PROGRAMS AND POVERTY REDUCTION Table 3.4. Indicators of Poverty and Expenditures on Family Assistance in the Transition Economies Totalfertility rates Headcount ratio (most recent Pretransition Posttransition estimate) Region and cointry (1987-88) (1993-94) 1980-85 1988-93 Balkans and Poland 5 23a Bulgaria 2b 17a,c 2.0 1.5 Poland 6 20a,c 2.3 1.9 Romania 6b 32a 2.2 1.5 Central Eutrope 0.5 1 Czech Republic 0 <1 2.0 1.9 Hungary