37048 INTERNATIONALBANK FOR WORLD BANK R E T C N O E N STRUCTION PM AND DEVELO June 2006 No. 91 A regular series of notes highlighting recent lessons emerging from the operational and analytical program of the World Bank`s Latin America and Caribbean Region REDISTRIBUTING INCOME TO THE POOR AND THE RICH: PUBLIC TRANSFERS IN LATIN AMERICA AND THE CARIBBEAN Kathy Lindert, Emmanuel Skoufias and Joseph Shapiro A recent regional study by the World Bank1 measures the tion can they achieve? How well do they perform? Do they extent to which publicly-subsidized transfers in LatinAmerica reduce LAC's high inequality ­ or exacerbate it? How much do and the Caribbean (LAC) redistribute income. Four factors poor people benefit from these transfers? This study analyzes motivate our interest in this topic: poverty, inequality, percep- precisely these questions. tions of social injustice in LAC, and the significant share of public resources devoted to these trans- Our findings suggest that public trans- fers in several countries in the region. fers can be effective instruments to re- Clearly, social policy has other important distribute income to the poor. But they objectives besides redistribution, in- have not often done so. Indeed, Robin cluding (a) smoothing of income over Hood works in both directions in LAC, the life cycle in relation to people's with public transfers redistributing in- needs; (b) promoting the accumulation come to both the rich and the poor. of human capital; and (c) ensuring ad- equate protection against events such Public Spending on Transfers as sickness, disability, unemployment, or loss of income. These themes, how- Overall, public spending on transfers ever, are beyond the scope of this report, represents about 5.7% of GDP in LAC. which modestly asks: "how redistribu- About three quarters of this spending tive are public transfers in LatinAmerica (4.3% of GDP) finance public insurance and the Caribbean (LAC)?" benefits, with the remainder (1.4%) allo- cated to social assistance transfers. Although some would While there are few doubts about the importance of growth for object to the treatment of social insurance (e.g., pensions, poverty reduction, growth in LAC has been slow over the past unemployment insurance) as "transfers" because they are decade, and, barring a few exceptions, existing growth has (partially) financed by direct ear-marked contributions, most benefited the poor less than proportionally.2 Growth associ- social insurance programs in LAC incur significant deficits ated with progressive distributional changes will have a greater which are financed from general revenues. As such, our study impact in reducing poverty than without such changes. More- "nets out" average payroll contributions, so as to allow us to over, recent evidence suggests that poverty itself may be asses the redistributive impact of the "pure" public transfer impeding higher growth rates in LAC. While investments in from social insurance (the part financed by general revenues access to productive assets ­ such as education, land, prop- not direct contributions). erty rights and infrastructure ­ can reduce poverty and in- equality in the long run, asset-based strategies take time to Within the region, there is considerable variation in overall implement and improve welfare.3 spending and the composition of transfer instruments. Some "higher spenders" ­ such as Argentina, Brazil, Chile and What about the role of public transfers? Can they promote Colombia ­ devote significant resources (averaging 11.5% of equity faster and in an efficient manner? How much redistribu- GDP) to social protection transfers. The bulk of this spending 1 is allocated to social insurance, though these countries also social insurance transfers are regressive in all countries in our operate fairly sophisticated cash-based social assistance sample ­ and in some cases, so regressive that they exacerbate transfer schemes. At the other end of the spectrum, most inequality. Central American and Caribbean countries allocate relatively little to social protection transfers (averaging 1.9% of GDP), Regressivity applies to net social insurance transfers, which most of which goes mainly to in-kind social assistance trans- are subsidized by government budgets at the expense of all fers. taxpayers. Our analysis focuses on the incidence of net transfers for pensions ­ subtracting the share of benefits Redistributive Impact of PublicTransfers financed by direct contributions. The basic premise for this is that these "net pension subsidies" compete for tax-financed The redistributive power of 56 transfers in eight countries ­ resources with social assistance and other forms of social Argentina, Brazil, Chile, Colombia, the Dominican Republic, spending, with very different redistributive patterns. More- Guatemala, Mexico, and Peru ­ is measured using household over, the regressivity of net social insurance transfers has not survey data by their coverage, size, absolute incidence, simu- been helped by any significant degree of progressivity in tax lated impacts on poverty and inequality, and by their distribu- financing. Rather, these deficit-ridden and largely truncated tional characteristic, a statistic derived from taxation literature. social insurance schemes seem to have created a "Reverse The findings are summarized for social insurance and social Robin Hood" situation of taking from the poor and redistribut- assistance as follows. ing to the rich, thereby exacerbating LAC's high inequalities. The magnitude of these tax-financed social insurance benefits is huge. Net pension subsidies absorb about 5% of GDP in the higher spending countries, far higher than spending on social assistance, and higher than average public spending on education and health in these countries. Important opportunity costs indeed. The concern, then, is one of equity, efficiency and fiscal sustainability. The more recent emergence of social assistance only partially offsets this truncation of the welfare state in LAC. True: coverage of the poor can be impressive in some instances. True: absolute incidence is progres- sive overall, and highly progressive for certain types of social assistance programs (Figure 2). However, overall spending is low in many countries and unit subsidies The redistributive impacts of social insurance transfers are are very small, thus muting the redistributive, poverty and limited ­ and even regressive in most instances. Our findings inequality impacts of even the most targeted programs.4 As quantify the degree to which Latin America's "elite" captures such, broad coverage of social assistance has not fully com- public subsidies to existing social insurance schemes. Specifi- pensated for significant biases in the far more generous (net) cally, those in the top quintile of the population receive about social insurance subsidies to the rich. 60% of net social insurance transfers (Figure 1). This regressivity derives from two main "design" factors: a trunca- There is considerable variation in redistributive impacts tion in coverage and relatively generous unit benefits for those within the class of social assistance transfers. Too many are in the top quintile (Q5 in th Figures). The persistence of a historical "truncation" thwarts coverage of the poor by design. Specifically, a defining characteristic of social insurance programs in LAC is that they tie eligibility to membership in the formal labor market ­ which hampers cov- erage of the poor, who are largely em- ployed in the informal sector. Our analy- sis confirms this truncation in practice. Moreover, social insurance transfers are typically far more generous (a) for recipi- ents in the upper quintiles; and (b) than social assistance programs ­ even when contributions have been netted out. As a result of this truncation in coverage and these biases in the generosity of benefits, 2 regressive. These include scholarships and many food-based protection in LAC ­ arming them with additional evidence programs. Governments should reconsider these programs ­ on redistributive impacts, namely: or at least strengthen their design. They could look to the targeting mechanisms used by conditional cash transfers · Reducing pension deficits ­ and rethinking some highly- regressive social insurance programs ­ must be a top priority on the policy agendas in LAC. These findings are consistent with recommendations for social insurance from other World Bank publications,7 emphasizing: (a) the need to reduce regressive net subsidies for pensions by promoting fully-funded pension plans for those who can afford it (pillar 2 reforms) and (b) reserving subsidized pensions for the poor (better targeting of pillar 1) and moving away from the restriction of such benefits tied to formal sector employment. (CCTs) ­ with impressive rewards (Figure 3). The relatively · Improving the targeting of social assistance transfers is high progressivity of CCTs is likely driven by a clear definition also a priority in many cases ­ at least as a short-term of the poor as the target group and the explicit use of targeting measure, short of a more sweeping overhaul of social mechanisms to determine eligibility, and not on their condition- policy. As noted above, many social assistance transfers alities per se. Such design mechanisms could be built into are regressive, despite explicit objectives to assist the other social assistance programs, for example using a combina- poor. Yet this study ­ and others8 ­ clearly shows that tion of geographic targeting and individual assessment mecha- better targeting is feasible. The targeting mechanisms nisms to target needs-based scholarships. used by conditional cash transfers, for example, could be adopted by other social assistance programs. "Quo Vadis Latin America?"5 Given these findings ­ and recent trends in LAC ­ what is the future trajectory for · In cases where social assistance programs are well- LAC? Importantly, how will LAC's democracies reconcile targeted and otherwise effective, countries might also the competing forces of: (a) widespread perceptions of consider the possibility of increasing unit subsidies for dissatisfaction with the social injustice of LAC's high better redistributive and poverty impacts ­ perhaps with poverty and inequality (similar to those in Europe) ­ which, eventual savings from further reforms to reduce pension as our paper shows, is often exacerbated by the deficits. Yet this raises the issue of potential adverse regressivity of many public transfers (each with their own incentive effects. Although evidence of strong labor vested interests); (b) high and growing fiscal pressures, disincentives is absent for these programs in LAC, and exacerbated by high pensions deficits; and (c) a persis- certain design features would plausibly discourage such tently informal labor market (to a degree far higher than in effects, policy makers in LAC should more explicitly Europe)? incorporate design features to reduce potential work disincentives ­ and researchers should further investi- Barrientos (2004) suggests that one trajectory for LAC is in gate this potential. Moreover, any potential increase in the direction of more "liberal welfare regimes" (like the unit transfers should be weighed against other develop- United States), with increasing reliance on market welfare ment priorities. production combined with finely-targeted social assis- tance. On the other hand, there has been considerable debate in several countries (Argentina, Brazil, Uruguay) · But targeting and design improvements to individual about an alternative path, towards "minimum universal social assistance programs are not enough. An adequate income" schemes ­ more along the lines of those in Eu- mix of instruments is needed to weave a social safety net rope.6 Given the need for large fiscal resources to assure that (a) promotes other roles of social policy (promotion such universal minimums, such a scheme would require an of human capital, smoothing consumption, and ensuring overhaul of the welfare state and significant reallocations ­ adequate protection against shocks, in addition to redis- perhaps from the existing grossly inefficient and regressive tribution); and (b) covers a variety of circumstances, social insurance schemes. including programs tailored to the needs of specific vulnerable groups. Recognizing that redistribution is only one of many roles for social policy, the findings of this report reinforce the However, while simple arithmetic suggests that lower spending recommendations commonly made in the literature on social on social insurance leaves more room in the government 3 budget for spending on better targeted social assistance As Fiszbein (2004) so eloquently remarks, "the status quo is programs, serious consideration needs to be given to the clearly unsatisfactory ­ and citizens of Latin America are political feasibility and sustainability of such reallocations. saying so quite openly. The potential answers are not simple Although it is beyond the scope of this paper to explore the to design or implement, as they will require solving both underlying forces ­ political, economic and demographic ­ technical and political difficulties. In the end, the Latin behind the existing spending biases, there is clearly some American experience suggests that it is politically easier to barrier to lowering social insurance transfers (even where they `target' one percent of GDP for social assistance within a are highly inequitable) and raising social assistance transfers regressive tax and social protection system than to make (even where they are progressive) in many countries in LAC. benefits much more inclusive within a more progressive Economic and demographic factors may come into play ­ as system. The debate is in process, but the end result is still evidenced by the rise in social insurance transfers as a share of unclear. Quo vadis Latin America?" GDP with per capita incomes (greater scope for risk pooling with higher incomes) and aging populations ­ a pattern that is **************** not mirrored for social assistance. Some barriers likely arise from the different political constituencies served by these two classes of transfers, and the vested interests of more organized formal sector workers. Another possible factor is the possible perceived "legitimacy" conferred on social insurance programs by the fact that they link benefits to contributions (even if only partially).9 Even though they are highly regressive and financed to a large part by general tax revenues, social insurance schemes may claim more legitimacy because societies perceive them as "rightfully earned" through contributions. Perceptions of "legitimacy" could also explain ­ at least in Notes part -- the emerging popularity of conditional cash trans- fers, whereby societies perceive that the fact that beneficia- ries have to comply with a set of "co-responsibilities" 1 Lindert, Kathy, Emmanuel Skoufias and Joseph Shapiro. (human capital conditionalities) bestows a greater degree March 2006. "Redistributing Income to the Poor of legitimacy on these transfers than pure cash or in-kind and the Rich: Public Transfers in Latin America and the handouts. Indeed, conditional cash transfers are promising Caribbean." The World Bank. Download this Report from not only for their redistributive impacts, but also for their http://www.worldbank.org/lacpoverty demonstrated impacts on human capital and their ability to 2Perry, et. al. (2005). break the inter-generational transmission of poverty. 3De Ferranti et. al. (2004). 4 The exception is Argentina's Household Heads Program, Yet these "islands of success" should not remain islands ­ which combines good coverage of the poor with progressive part of a dual system of social assistance, along side an targeting and relatively high unit transfers. As a result, inefficient and regressive social insurance system as if it Argentina's program does yield important impacts on poverty were "the elephant in the room" that is not being effectively reduction and inequality. addressed. No country in LAC has (yet) been able to 5Fiszbein (2004). effectively and convincingly integrate social insurance and 6See, for example, Suplicy (2002). social assistance as part of a coherent system of social 7Gill et. al. (2005) and Holzmann and Palmer (eds, 2006). protection.10 The duality of these social insurance and 8See, for example, Coady et. al. (2004). social assistance systems creates a patchwork of inefficien- 9The authors would like to thank Augusto de la Torre for his cies arising from a lack of policy consistency and coordina- thoughtful suggestions on this point. tion, including: gaps and duplications in coverage, in- 10Fiszbein (2004). creased administrative costs from multiple registry and 11World Bank (2005). governance systems, incentives distortions, etc. Many countries have undertaken reforms to improve these sys- tems, but such reforms generally follow parallel tracks. A About the Authors more integrated approach could bring about the realloca- tions and efficiency gains necessary for more progressive and inclusive outcomes ­ as well as possible fiscal savings Kathy Lindert is the Country Sector Leader for Human that could be used for other development priorities, such as Development based in the World Bank's Brazil Office. investments in improving the quality of education which is Emmanuel Skoufias is a Senior Economist in the Poverty crucial for equalizing "opportunities" for redistribution in Group of the Latin America and the Caribbean Region and the long-run.11 Joseph Shapiro is studying at Oxford University in the UK. 4