33339 The Poverty Targeting of Social Spending in Brazil Joachim von Amsberg, Peter Lanjouw, and Kimberly Nead1 1This document is a background paper to the World Bank Report "Attacking Brazil's Poverty," Report No. 20475, June 30, 2000. The views expressed in this paper are those of the authors and should not be attributed to the World Bank or its Board of Executive Directors. This paper in- cludes special tabulations of the PPV provided by Mark Thomas. Leo Feler edited the final re- port. Table of Contents THE POVERTY TARGETING OF SOCIAL SPENDING IN BRAZIL.............................. 1 JOACHIM VON AMSBERG, PETER LANJOUW, AND KIMBERLY NEAD .......................................1 TABLE OF CONTENTS..................................................................................................... 2 TABLE OF FIGURES......................................................................................................... 3 1. INTRODUCTION........................................................................................................ 4 METHODOLOGY..................................................................................................................5 CAVEATS............................................................................................................................6 ORGANIZATION OF THEPAPER.............................................................................................7 2. BACKGROUND: URBAN POVERTY AND SOCIAL EXPENDITURES IN BRAZIL8 OVERVIEW OF URBANPOVERTY..........................................................................................8 A FEDERAL PERSPECTIVE ON SOCIAL SPENDING...................................................................9 OVERVIEW OF PUBLIC SOCIAL SPENDING IN BRAZIL...........................................................11 3. SPECIFIC AREAS OF SOCIAL SPENDING............................................................ 14 EDUCATION......................................................................................................................15 Education System Overview.......................................................................................... 15 Crèches....................................................................................................................... 16 Kindergarten ............................................................................................................... 17 Primary Education....................................................................................................... 17 Secondary Education.................................................................................................... 19 Higher Education......................................................................................................... 21 Adult and Professional Education ................................................................................. 22 HEALTH CARE..................................................................................................................23 Health Care System Overview....................................................................................... 23 The Poor's Access to Health Care................................................................................. 23 Basic Health Care and Prenatal Care ........................................................................... 25 Prenatal Care .............................................................................................................. 25 Targeting: The Incidence of Health Subsidies ............................................................... 25 Piso de Assistência Básica (PAB).................................................................................. 27 Benefits for the Poor .................................................................................................... 27 NUTRITIONPROGRAMS.....................................................................................................28 URBAN SERVICES..............................................................................................................30 Housing....................................................................................................................... 30 Transport .................................................................................................................... 31 Water and Sanitation.................................................................................................... 32 TRANSFER PROGRAMS......................................................................................................34 4. POVERTY INCIDENCE OF SOCIAL SPENDING................................................... 36 5. CONCLUSIONS........................................................................................................ 42 6. REFERENCES .......................................................................................................... 43 Table of Figures Figure 1 Poverty Headcount Ratio for NE and SE Urban Areas............................................................................8 Figure 2 Composition of the Urban Poor....................................................................................................................8 Figure 3 Composition of the National Consumption Quintiles ..............................................................................9 Figure 4 Incidence of Public Primary Education, National Quintiles .................................................................10 Figure 5 Incidence of Public Primary Education, Local Quintiles.......................................................................10 Figure 6 Composition of the National Consumption Quintiles ............................................................................11 Figure 6 Public Expenditures in 1995 by Program Area........................................................................................12 Figure 7 Share of Resources Originating from Federal, State, and Municipal Governments, by Program..13 Figure 8 Share of Resources Administered by Federal, State, and Municipal Governments, by Program...13 Figure 10 Share of Public Education Resources Managed at the Federal, State, and Local Levels ..............15 Figure 11 Share of Public Education Resources Originating at the Federal, State, and Local Levels ...........15 Figure 12 Share of the Poor in Public Student Population, Urban Brazil, by Type of Facility.......................15 Figure 13 Incidence of Public Education by Level, Urban....................................................................................15 Figure 14 Reach of Public Education Among the Urban Poor.............................................................................16 Figure 15 Crèche Attendance, 0 to 3 Years, Urban ................................................................................................16 Figure 16 Incidence of Children in Urban Public Crèche......................................................................................17 Figure 17 Kindergarten Attendance, Urban 4-6 Year Olds...................................................................................17 Figure 18 Incidence of Children in Public Kindergarten.......................................................................................18 Figure 19 Primary School Attendance, Urban 7-14 Year Olds ............................................................................18 Figure 20 Reasons for Not Attending Primary School, Urban Poor, 7-14 Years Old ......................................19 Figure 21 Reasons for Not Attending Primary School, Urban Non-Poor, 7-14 Years Old .............................19 Figure 22 Percentage of 15-19 Year Olds Attending Primary School, Urban Brazil.......................................19 Figure 23 Secondary School Attendance, Urban 15-19 Year Olds......................................................................20 Figure 24 Reason for Not Attending School, Urban Poor 15-17 Years Old ......................................................20 Figure 25 Reasons for Not Attending School, Urban Non-Poor 15-17 Years Old............................................20 Figure 26 Share of Public Secondary Students (Regional)....................................................................................20 Figure 27 Share of Public Secondary School Students (National).......................................................................20 Figure 28 Higher Education Attendance, Urban 20-24 Year Olds ......................................................................21 Figure 29 Incidence of Higher Education, Urban ...................................................................................................21 Figure 30 Incidence of Adult Education, Urban......................................................................................................22 Figure 31 Reasons for Not Seeking Medical Care, Urban Non-Poor..................................................................24 Figure 32 Reasons for Not Seeking Medical Care, Urban Poor...........................................................................24 Figure 33 Where Health Care is Obtained, by Consumption Quintile, Urban Brazil.......................................25 Figure 34 Incidence of Overall Public Health Care, Urban...................................................................................26 Figure 35 Incidence of Public Health Care by Type of Facility...........................................................................27 Figure 36 Incidence of Milk Distribution Programs ...............................................................................................28 Figure 37 Coverage with Regular School Lunches ................................................................................................28 Figure 38 Incidence of School Lunches ...................................................................................................................29 Figure 39 Share of Quintile Living in Favela ..........................................................................................................30 Figure 40 Housing Tenure and Poverty Rate, Urban Brazil..................................................................................31 Figure 41 Incidence by Mode of Transport, Urban Commuters...........................................................................31 Figure 42 Incidence of Public Transport by Quintile .............................................................................................32 Figure 43 Value Incidence of Vale Transporte, Urban Brazil..............................................................................32 Figure 44 Share of Quintiles living Without Access to Safe Water....................................................................33 Figure 45 Coverage with Sewage Network..............................................................................................................33 Figure 46 Incidence of Unemployment Insurance Benefits ..................................................................................34 Figure 47 Incidence of Pension Receipts .................................................................................................................34 Figure 48: Structure and Targeting of Federal Social Spending, 1997.................................................................37 Figure 49 Budget Cost Per Current Benefit to the Poor.........................................................................................39 Figure 50 Budget Cost Per Total Benefit to the Poor.............................................................................................39 Figure 51 Reach and Targeting of Social Programs ...............................................................................................41 The Poverty Targeting of Social Spending in Brazil Page 4 1. Introduction A broad array of government policies and provide immediate poverty relief. Moreo- programs has the potential to help Brazil's ver, much debate exists as to whether it is urban poor. There is no shortage of options more important to ensure that none of the from which to choose. Resources are more poor are excluded from program benefits or limited than ideas. How should government whether policies should aim to minimize funds and administrative capacity be allo- leakage of benefits to the non-poor. Each of cated if Brazil wants to alleviate urban pov- these choices entail value judgments, and erty effectively? this paper will not argue in favor of pursuing one priority versus another. Instead this This paper examines the possibilities for paper aims to describe the current and po- reducing urban poverty through three im- tential impact of government interventions portant categories of government interven- on Brazil's urban poor, to clarify the implicit tions: ensuring the urban poor's access to tradeoffs in favoring one intervention over effective education and health services, another, and to illuminate the ramifications ensuring their access to urban services and of certain policy and program design infrastructure such as clean water, sanita- choices. tion, transport and housing, and ensuring the availability of an adequate social safety net In the following sections, this paper exam- to protect the consumption levels of vulner- ines the options for helping the urban poor able groups through pensions, unemploy- through direct government interventions in ment insurance, nutrition programs, and the areas of education, health, water, sanita- guaranteed minimum income programs. tion, transport, housing, and social safety net Given time and space limitations, the analy- programs. For each alternative intervention, sis does not cover two admittedly critical the existing information is gathered and new aspects of any sustainable poverty reduction evidence is presented in the following areas: strategy: sound macroeconomic policies to support broad-based economic growth, and · Government expenditures. the indirect poverty impact achieved by · The share of Government subsidies or fostering economic growth through prudent program benefits currently received by investments in human capital and economic the urban poor. This is the intervention's infrastructure. "targeting". · The current extent of program coverage The comparison of policy interventions for and the urban poor's access to services. poverty reduction is complicated by the lack This is the program's "reach". of consensus on how to define poverty, as · The potential "reach" of the program well as by the lack of consensus on the pre- among the urban poor, as well as the cise policy objective even given a shared current and potential benefits of poverty definition of poverty. Should poverty be reduction for the urban poor. defined by low incomes, inadequate access · Net spending, or benefits per poor to basic services, degree of social exclusion, household or poor individual reached. or sub-standard outcomes such as shorter life expectancy and illiteracy? Even if a The above obviously constitutes an ambi- consensus can be reached regarding the tious agenda, one that this paper will only precise definition of poverty, "poverty alle- partially realize. First, it is not possible to viation" still encompasses several policy answer all of the above questions for every objectives, some of which may conflict. intervention because in some cases there are Another question intrinsic in the develop- no data. In other cases, data exist but have ment of poverty alleviation policies asks serious limitations. It should be emphasized whether the priority is to maximize long- that the data and the analysis presented in term, sustainable poverty reduction or to The Poverty Targeting of Social Spending in Brazil Page 5 this paper, particularly the estimates of pro- by the World Bank, and conducted in 1996- gram benefits, should be taken not as precise 97 by Brazil's national statistics agency, point estimates but as rough indications. IBGE, to assess the poverty targeting of Even where household survey data form the Government social spending in Brazil. basis of the analysis, it must be remembered that the resulting numbers are estimates of IBGE implemented the PPV or Survey on the real situation, with margins of error on Living Standards in 1996 and 1997, with either side, and that the path to a "number" assistance from the World Bank. The aim of involved numerous methodological choices.2 the PPV is to supplement the information already available through IBGE's annual While it is important to emphasize the limi- household survey, the PNAD, in order to tations of the analysis, it is also important to improve the data available for poverty highlight the recent improvements in the monitoring and policy analysis in Brazil. available information on Brazil's social The PNAD has a number of strengths. It programs and their impact on the poor. As utilizes a large sample and allows compari- described in the next subsection, new Bra- sons over time, due to continuity in the basic zilian data allow for more confident analysis core of the questionnaire. The PNAD's core than was possible even in the recent past. It has a strong focus on income and employ- is hoped that by bringing together the dispa- ment issues, and also contains questions rate strands of evidence that exist, a step will relating to education, housing ownership and have been taken toward clarifying the trade- conditions, migration, access to services, offs involved in emphasizing one program and ownership of durable goods. The sur- over another in an urban poverty strategy. vey also includes additional questions on special topics that vary from year to year. Methodology However, the PNAD provides little infor- mation on household expenditures and con- This paper focuses on the analysis of sumption patterns, health status and health spending incidence for the bottom quintile service utilization, and transportation usage. of the Brazilian population. The bottom Although it asks questions regarding house- quintile roughly corresponds to the poor as hold members' educational levels and they were recently identified by applying a school attendance, the PNAD does not dis- food-only poverty line of R$65 per capita tinguish between public and private school per month to household income data from attendance; this makes it an inadequate the 1996 PNAD household survey. This source of information on the distribution of poverty line produces a national headcount public education subsidies. poverty rate of 22.6% (Ferreira, Lanjouw, and Neri, 1998). A poverty line of R$130 The PPV was designed to fill some of the per capita per month corresponds roughly to data gaps left by the PNAD. It provides a the bottom two quintiles and results in a much more detailed picture of household national headcount poverty rate of 45.3%. expenditures and consumption, as well as utilization of various publicly subsidized This paper uses data from the Pesquisa So- services, particularly education, health, and bre Padrões de Vida (PPV), a household transportation. The questionnaire is much survey similar to the Living Standard Meas- longer, and requires multiple visits to each urement Survey supported in many countries household. This richer information comes at a price. To keep survey expenses within reason, the sample size is much smaller and 2 Some of the more significant and controversial of these the survey only covers the two most popu- are: 1) the choice of the poverty line, 2) the ways for which regional price differences are accounted, 3) the lous of Brazil's five regions, the Northeast selection of household income or consumption as the and Southeast. The Northeast and Southeast welfare measure, and 4) the method used to adjust total together account for 73% of Brazil's popu- household income/consumption to accurately reflect the living standards of people living in households of differ- lation and 80% of Brazil's poor. All results ent sizes and compositions. The Poverty Targeting of Social Spending in Brazil Page 6 presented in this paper are based on the Caveats analysis of these two regions only. Unfortunately, the IPEA government spending data and the PPV household level The PPV is representative for ten spatial data are for two different years, 1995 and units (the metropolitan areas of São Paulo, 1996-97, respectively. The inaccuracies Rio de Janeiro, Belo Horizonte, Salvador, introduced by using the PPV household Recife, and Fortaleza; the non-metropolitan consumption data for 1996-97 and the IPEA urban Northeast; the non-metropolitan urban spending data for 1995 are likely to be small Southeast; the rural Northeast; and the rural in most cases, since there is no reason to Southeast). This paper reports three types of believe a priori that the utilization of serv- results: individual results by spatial unit, ices by the various income and consumption aggregate results for all urban units (ex- groups changed significantly between the cluding only rural Northeast and rural two surveys. Southeast), and aggregate results for all units. It is important to note that the picture drawn by the use of the PPV household survey data A comprehensive picture of government and IPEA spending data may not reflect the spending that includes outlays at both the current policy environment, despite being national and subnational levels is essential to fairly up-to-date. For example, the Brazilian understanding equity and efficiency issues government has enacted significant policy in Brazil's social sectors. Financing and changes since 1995 in two of the most im- administration of programs is, to varying portant spending categories, education and degrees, decentralized. In many cases, the health.3 Both policy changes are likely to Constitution dictates the allocation of re- substantially improve the poverty targeting sources and assignment of expenditure and achieved by these public expenditures. administrative responsibilities. The states and municipalities play major roles in the In 1996 Brazil made major changes in edu- financing and provision of education, health, cation finance that should have the effect of sanitation, housing and urban development, directing a larger share of education re- social assistance, nutrition, sanitation, and sources toward the poor, since they increase mass transport. The only program areas the share of resources going to primary ver- where the state and local governments have sus other levels of education and equalize little involvement on both the financing and per student spending across public primary administration side are social security (pen- schools (which has historically been highly sions and retirement benefits) and labor unequal, with municipal schools in poor (mostly unemployment benefits). areas having much lower unit costs than state schools or municipal schools in rela- The IPEA Social Expenditure Study allows tively rich municipalities.) The year 1996 for such a comprehensive picture of gov- also brought equity improving reforms in the ernment spending. When combined with the health sector. The Ministry of Health PPV data on households' consumption pat- adopted a policy designed to ensure a mini- terns, including consumption of publicly mum amount of spending on basic health subsidized goods and services, the IPEA care in every municipality, the Piso Assis- spending data provides a reasonable infor- tencial Básico. This policy should help mation base from which to assess the distri- equalize the differences on basic public bution of benefits resulting from govern- health care spending between poorer and ment spending on various social programs. wealthier municipalities. The Piso Assisten- The IPEA study and the PPV together fa- cial Básico is expected to increase the cilitate the analysis of poverty targeting poor's access to quality health care and im- issues. prove the benefit incidence of health expen- 3See "Brazil: Social Spending in Selected States." The Poverty Targeting of Social Spending in Brazil Page 7 ditures for the poor. These policy reforms accurate analysis of poverty targeting in are discussed in the subsections below Brazil. dealing with education and health. For now, it suffices to recognize that Brazil is in the Organization of the Paper process of enacting several important re- forms that will most likely improve the pov- Throughout the paper, the benefit incidence erty impact and targeting of public expen- analysis only covers the Northeast and ditures on health and education. Southeast. For the purposes of this paper, the "national" income or consumption dis- Finally, the PPV data and the IPEA data on tribution refers to the combined population spending have intrinsic limitations. The of the Northeast and Southeast regions. PPV data only cover the Northeast and Southeast. This means that the benefit inci- A general discussion of urban poverty issues dence analysis and other information on the and social expenditures in Brazil follows in poor's access and utilization of various pub- Section 2 of this report. Section 2 also pres- licly provided services cannot be assumed to ents data on the distribution of poverty describe the situation in urban areas in the throughout different regions and provides North, Centerwest, and South. Even within information regarding the distribution of the Northeast and Southeast, the PPV's total public expenditures among different small sample size severely limits the ability consumption quintiles. to look at questions of targeting and access at the local level. At best, the data allow The incidence of spending (targeting) and estimates for each of the metropolitan areas, the program coverage by consumption quin- the rural Northeast, the rural Southeast, the tile (reach) are analyzed for most govern- non-metropolitan urban areas of the North- ment social spending programs that can be east, and the non-metropolitan urban areas adequately tracked with the data from the of the Southeast. In cases where an event is PPV. These programs include different relatively rare, such as the likelihood of a levels of education, health care, nutrition family member having received a vaccina- programs, public transport, water and sani- tion in the 10 days prior to the survey, or the tation services, pensions, and unemployment number of the poor attending higher educa- insurance. Section 3 contains the analysis of tion, the sample size poses even greater the main programs of social spending and restrictions on the ability to make accurate presents their coverage and targeting rates. estimates at a spatially disaggregated level. Unless otherwise noted, consumption quin- The IPEA data on social expenditures for tiles are constructed on the basis of the con- 1995 are limited by incomplete coverage of sumption distribution in the entire PPV. spending at the municipal level.4 The data are more extensive for the 186 municipali- Section 4 combines the analysis of program ties which are either state capitals or belong targeting with actual social spending data to to one of the metropolitan regions. generate a picture of the overall poverty targeting of social spending in Brazil. Com- In summary, the sources of data available bined with assumptions about the benefit- for the benefit incidence analysis presented cost ratios of different programs, Section 4 in this paper have limitation. This discus- presents an indicative ranking of the transfer sion of caveats, however, should not obscure effectiveness of social programs. the fact that these new sources present a major advancement over previously avail- Remarks about implications and possible able information. Even with the caveats, the further work conclude this paper. PPV and IPEA data allow for a reasonably 4 Brazil had 4,974 municipalities in 1995, with obvious implications for the challenges of collecting detailed data on social spending by each municipal government. The Poverty Targeting of Social Spending in Brazil Page 8 2. Background: Urban Poverty and Social Expenditures in Brazil Overview of Urban Poverty present in the analysis of the 1996 PNAD recently completed for Brazil Urban Poverty Urban versus Rural Poverty. Strategy Paper.7 An analysis of PPV data for approximately the same period (1996- Whether measured in terms of income, con- 97) and using the same poverty line and sumption, access to services, or social indicators (such as school enrollment Figure 1 Poverty Headcount Ratio for NE and SE Urban Areas rates, infant mortality rates, and aver- age life expectancy at birth), poverty is more common and more severe among Brazil's rural population. Average 70.0% living standards are particularly low in 60.0% the rural Northeast, in part due to re- curring droughts and the large share of 50.0% the population that remains dependent 40.0% Percentage of Population on agriculture for subsistence. How- who are Poor 30.0% ever, as Brazil's population has be- 20.0% come increasingly urbanized, the mag- nitude of the urban poverty problem 10.0% and the share of urban individuals in 0.0% poverty have grown. According to Other urban Salvador Recife Fortaleza Belo Other urban Rio Sao Paulo NE Horizonte SE recent estimates based on 1996 PNAD data, roughly half of the poor (52.5%) live in urban areas, mostly in Figure 2 Composition of the Urban Poor smaller cities and towns outside the metropolitan regions.5 Fortaleza Recife 4% 4% Non-metropolitan Salvador Spatial Dimension of Urban Poverty. urban SE 5% 26% There are at least two notable spatial aspects of urban poverty in Brazil. First, urban areas in the South, Cen- terwest, and Southeast tend to have a much lower prevalence of poverty compared to those in the North and Sao Paulo 9% Northeast. Second, within a given Non-metropolitan urban NE region, poverty tends to be more Rio de Janeiro 41% 7% prevalent and more severe in medium Belo Horizonte 4% and small cities than in large cities and metropolitan regions.6 This disparity methodology also produced results consis- between smaller and larger cities is clearly tent with these findings, although the analy- sis is less disaggregated and only covers the 5 See the complementary background paper prepared by region. In the North and Southeast, metropolitan periph- Ferreira, Lanjouw and Neri: The Urban Poor in Brazil in eries have worse poverty indices than all but the small 1996 -- A New Poverty Profile Using PPV, PNAD and urban and rural areas. In the Northeast and South the Census Data. peripheries have lower poverty rates than the medium 6 The poverty rankings of the peripheral areas of the sized cities. metropolitan regions vary depending on the macro- 7 Ibid. The Poverty Targeting of Social Spending in Brazil Page 9 Northeast and Southeast. Figure 1 graphi- of the population come from either the urban cally represents the extent of the poverty Southeast or the São Paulo metropolitan problems faced by different urban commu- region. Very few of the individuals in the nities in the Northeast and Southeast. wealthiest quintile live in the Northeast. Geographical Distribution of the Poor. From the geographical decomposition of the Figure 2 shows the share of all urban poor consumption quintiles, it is easy to analyze living in the different geographical areas. the implications of poverty targeting. Un- Although most of Brazil's urban poor live in less a social program reaches at least some non-metropolitan areas, there is still a large individuals in the Northeast, it is not reach- number of poor that are concentrated in the ing the majority of the poor. Even consid- metropolitan regions of the Northeast and ering just the urban poor and excluding the Southeast. The huge populations of the rural poor, the story remains the same: a metropolitan areas, particularly in São Paulo social program must include the Northeast in and Rio de Janeiro, dictate that a city can order to have an impact on the majority of have a sizable urban poverty problem even the urban poor. if its poverty headcount ratio is low (see Figure 1 and Figure 2). For example, São Of course, it does not follow that a program Paulo has the lowest incidence of poverty of is badly targeted if it only benefits residents of a wealthier region such as Figure 3 Composition of the National Consumption Quintiles São Paulo. For example, a program serving only those Number of individuals São Paulo residents in the 25,000,000 bottom 20% of the national consumption distribution is 20,000,000 arguably well-targeted since all Rural SE program beneficiaries fall be- Urban SE low the poverty line. On the 15,000,000 Sao Paulo other hand, such a program Rio Belo Horizonte would have a very limited Rural NE Urban NE 10,000,000 Salvador impact on the national poverty Recife Fortaleza problem since relatively few of 5,000,000 the nation's poor reside in the São Paulo area. 0 Poorest (1st) quintile 2nd quintile 3rd quintile 4th quintile 5th quintile National per capita consumption quintile A Federal Perspective on the six metropolitan regions, but at the same Social Spending time, it has a greater number of poor than any other metropolitan area. The distributional impact of most programs differs quite significantly between spatial Figure 3 shows the geographical composi- units. In particular, in the wealthier spatial tion of each of the national consumption units, the incidence appears much more quintiles. The majority of the poorest 20% regressive than in the poorer units. One of the population live either in the rural simple reason underlying this observation is Northeast or non-metropolitan urban North- that there are very few individuals in the east. Very few individuals belonging to the wealthier units that are in the bottom quin- poorest quintile reside in the Southeast. At tile of the national consumption distribution. the other end of the consumption distribu- On the other hand, there are very few indi- tion, the geographical make-up is very dif- viduals in the poorer units that are in the top ferent. The majority of the wealthiest 20% quintile of the national consumption distri- bution. In other words, there are simply The Poverty Targeting of Social Spending in Brazil Page 10 very few individuals in São Paulo that are revenues, the choice is not to spend in dif- poor by national comparison. In the rural ferent parts of the country but on different Northeast, there are very few individuals programs within the same region. From this that are wealthy by national comparison, so perspective, it is instructive to compare the every program is by definition well-targeted. incidence of spending across spatial units based on the distribution of that respective spatial unit. In other words, Figure 4 Incidence of Public Primary Education, National Quintiles incidence analysis based on regional consumption quintiles 60% may provide a helpful gauge for adjusting regional poverty 50% alleviation measures. Education 40% Figure 4 and Figure 5 compare Primary 30% the incidence of primary edu- Public cation spending using both of 20% regional and national quintiles. 10% Figure 4 clearly shows how Incidence spending on primary education 0% in the Northeast is well fo- 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) cused on the bottom quintile of Consumption Quintiles (National) the national consumption dis- Fortaleza MR Recife MR Salvador MR Belo Horizonte MR tribution while spending in the Rio MR Sao Paulo MR Source: IBGE-PPV, 1996/7 Nonmetropolitan urban NE Nonmetropolitan urban SE Southeast is focused on the Figure 5 Incidence of Public Primary Education, Local Quintiles middle of the national con- sumption distribution. Figure 5 shows a very different pic- 45% ture with amazing similarity 40% across all spatial units. Figure 35% Education 5 shows that in all units, a 30% roughly equal share of primary Primary 25% school students comes from 20% the bottom quintile of the local Public of 15% income distribution. Even for 10% other types of social services, 5% it is apparent that locally con- Incidence structed quintiles (representing 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) different consumption levels) Consumption Quintiles (Regional) show similar behavior for Fortaleza MR Recife MR Salvador MR Belo Horizonte MR different spatial units. Rio MR Sao Paulo MR Source: IBGE-PPV, 1996/7 Nonmetropolitan urban NE Nonmetropolitan urban SE Incidence analysis on the basis of the na- Figure 6 illustrates the impor- tional distribution is useful for national pol- tance of the distinction between regional and icy making. Since there are so many more national programs in the Brazilian context, poor individuals in the Northeast, targeting where inter-regional income inequality is of social spending would indeed improve if large. Nearly 40% of the individuals living resources were shifted from the wealthier to in the urban Northeast have incomes that the poorer parts of the country. There is, place them in the poorest 20% of the na- however, another equally valid point of tional income distribution. By contrast, less view. From the perspective of a local policy than 4% of the individuals in Rio de Janeiro maker who decides on the allocation of local have incomes placing them in the poorest The Poverty Targeting of Social Spending in Brazil Page 11 20% of the national distribution. This inter- Overview of Public Social Spending in regional income inequality affects the Brazil evaluation of regional and national program targeting. Government Interventions and Urban Pov- erty. Government efforts to reduce poverty A national program cannot be considered and assist the poor have tended to focus on well-targeted if 100% of its benefits accrue the urban poor in Brazil. This finding is to the poorest 40% of the population of Rio hardly surprising, given the visibility and sheer number of Figure 6 Composition of the National Consumption Quintiles Brazil's urban poor, not to mention the 120.0% social problems as- sociated with urban poverty. The greater 100.0% density of the poor population in urban 80.0% areas also facilitates 5th (richest) quintile 4th quintile effective government 60.0% 3rd quintile 2nd quintile intervention. 1st (poorest) quintile Moreover, the con- 40.0% centration of wealth and administrative 20.0% capacity in urban areas endows urban governments with 0.0% All NE and Fortaleza Recife Salvador Urban NE Rural NE Belo Rio Sao Paulo Urban SE Rural SE more resources to SE Horizonte implement social programs for the de Janeiro. The poorest 40% of Rio de Ja- poor residing in their jurisdiction. Over the neiro's population are primarily members of last 10 years, tremendous strides have been the second and third national income quin- made in improving the access of poor urban tiles. Thus, such a program would not be households to health care, basic education, well-targeted toward Brazil's poorest indi- and clean water.8 Access to these services viduals in the bottom 20% of the national has also improved greatly among the rural income distribution. poor. However, the rural poor still tend to lag behind their urban counterparts in their From a local perspective, however, the access to services because of their initially evaluation of poverty targeting changes lower "starting point" and in some cases significantly. If 100% of the benefits of a because of a greater rate of improvement in program funded by Rio de Janeiro accrue to urban areas. the poorest 40% of Rio de Janeiro's popula- tion, then the program is arguably well- targeted. Since the local government con- cerns itself with the local population, a pro- gram that affects the relatively poor in the locality can be well-targeted even if such a program does not affect many individuals 8 See, among others, "Melhoria em Indicadores de that are poor in the national context. Saúde Associados a Pobreza no Brasil dos Anos 90" (October 1997) prepared by researchers affiliated with the University of São Paulo (Monteiro, D'Aquino Benicio, and Martins). This paper documents the changes which occurred between the 1986 and 1996 PNDS (Demo- graphic and Health Survey). The Poverty Targeting of Social Spending in Brazil Page 12 Social Expenditures in Brazil: Level and tion benefits are targeted to the poor drives Composition. the degree of poverty targeting achieved by When considered in aggregate, Brazil's the entire set of social programs. In other social expenditures do not seem inadequate. words, given current spending patterns, ef- Public social spending in Brazil among mu- forts to better target benefits from housing, nicipal, state, and federal governments labor, social assistance, sanitation, public amounted to 20.9% of GDP in 1995.9 transport, and nutrition programs will have a relatively small impact on the total benefit In Latin America, only Costa Rica, incidence compared to the better targeting of Panama, and Argentina spent similarly high health and education. Directing health and but slightly lower shares of GDP on social education subsidies to the poor is the key to achieving redistribution through Figure 7 Public Expenditures in 1995 by Program Area public social spending in Brazil. Intergovernmental transfers 40,000 comprise a significant share of 35,161 social expenditures. Thus, there 35,000 can be a large gap between the 27,968 amount of resources that a level 30,000 Social Security Education 25,000 Health of government administers and millions 21,738 the amount of resources it di- R$ Housing and Urban Labor 1995 20,000 rectly finances. Figure 8 and in Social Assistance 15,000 Urban Mass Transit Sanitation Figure 9 present the relative im- Expenditure Food and Nutrition 10,000 7,183 portance of federal, state, and 3,022 5,000 2,863 2,620 local governments in social 1,374 822 spending for different program 0 areas. programs. As Figure 7 shows, however, The information regarding the social security accounts for the majority of share of resources originating and adminis- social spending in Brazil. This finding does tered by the different levels of government not imply that Brazil spends an inadequate is important for developing an urban poverty share of its GDP on non-pension social pro- strategy for at least three reasons. grams such as health, education, and hous- ing. Brazil's expenditure levels on health, First, the dominant level of government in education, and housing are 3.4%, 4.3%, and the administration of program resources can 1.1% of GDP, respectively. These levels are affect the practicality of finely targeting comparable to those in other Latin American program benefits. For example, means countries. testing requires a certain level of adminis- trative capacity. Poorer municipalities can be ill equipped for designing and imple- In Brazil, education and health are the most significant categories of social spending for menting means tests. Usually, state and which benefits are not explicitly linked to federal governments have a greater admin- contributions. Together, these two areas istrative capacity and a greater ability to account for more expenditures than all other design and implement means tests. social programs combined (excluding social security). The significance of this finding is Second, the relative importance of each that the degree to which health and educa- level of government in disbursing program funds provides a guideline for determining where the primary decisions affecting a 9 See IPEA, 1998. "Gastos Sociais das Três Esferas de program are made. For example, housing Governo ­ 1995" The Poverty Targeting of Social Spending in Brazil Page 13 Third, the source of program Figure 8 Share of Resources Originating from Federal, State, and Mu- nicipal Governments, by Program Area, 1995 financing may change the way targeting issues are viewed. In program areas where the munic i- pal or state governments provide 100% the majority of resources, such as 90% urban transport, housing, and 80% sanitation, the more relevant 70% question is probably how the 60% benefits are distributed across the Municipal 50% State local income or consumption Federal 40% distribution. If local resources 30% are financing the programs, pol- 20% icy makers are likely to be inter- 10% ested in how well-targeted these 0% expenditures are to the relatively Social Education Health Housing Labor Social Urban Sani tation Food Security and Urban Assistance Mass and poor within their own jurisdic- Transport Nutrition tion. On the other hand, in pro- gram areas where federal money is dominant, such as nutrition Figure 9 Share of Resources Administered by Federal, State, and Mu- and health, perhaps it is more nicipal Governments, by Program Area, 1995 useful to adopt a national per- spective and ask how well re- sources are targeted to the poorer 100% groups within the national population.10 90% 80% In considering specific areas of 70% social spending, such as educa- 60% Municipal State tion, health, and urban services, 50% it is important to remember the Federal source and administration of 40% 30% funds. The source and admini- 20% stration of funds will serve as a 10% guide, indicating the levels of government where effort should 0% be exerted to produce the most Social Education Health Housing and Labor Social Urban Mass Sanitation Food and Security Urban Assistance Transport Nutrition efficacious changes in poverty reduction strategies. and urban programs derive most of their funds from municipal treasuries. Thus, an attempt to alter the targeting of housing and urban programs would require an appeal to municipal authorities. On the other hand, labor programs derive the majority of their funding from the federal government. Thus, an attempt to alter the targeting of labor 10 A problem common to nearly all benefit incidence programs requires an appeal to federal analyses is that the distribution of program benefits is authorities. considered in isolation from the incidence of the taxes used to finance the program. Therefore, the picture that a benefit incidence analysis can provide of distributive aspects of public spending is highly imperfect and partial. The Poverty Targeting of Social Spending in Brazil Page 14 3. Specific Areas of Social Spending This section analyzes selected social pro- health care in poor areas is typically lower gram on two dimensions. The coverage by and water services to poor areas are often consumption quintile shows the share of the intermittent. This difference in service be- population (or a subgroup of the population) tween poorer and wealthier classes intro- in each quintile that receives a given service. duces a systematic bias in the estimates that The share of the uncovered poor population follow. The incidence of services to the (for the purpose of this paper assumed to be poor should therefore be interpreted as a equal to the first quintile) has sometimes lower bound. been referred to as the error of exclusion where poor individuals are excluded from Again, it is important to remember that sig- the program. However, incomplete cover- nificant policy changes have occurred after age over a specific population can only be the date of the PPV survey, especially in the interpreted as exclusion if the entire popula- areas of health and education funding. This tion is supposed to receive the benefit.11 analysis obviously reflects none of the changes that have occurred after 1996-97, The targeting ratio refers to the share of many of which are likely to have been posi- program participants from the first quintile. tive in terms of their impact on the distribu- The share of participants from the other four tion of program incidence. quintiles has been referred to as the error of inclusion where non-poor individuals are included in the program. Each program has particular characteristics that complicate the analysis of both cover- age and targeting. The extent to which these complications have or have not been appro- priately addressed through the chosen meth- odology is briefly discussed in the context of each program. The applied methodology has some limita- tions that apply across most programs. In particular, the methodology assumes that the quality of the service received is the same for individuals from all quintiles (if conclu- sions are drawn in terms of benefit inci- dence), or that spending on beneficiaries from all quintiles is the same (if conclusions are drawn in terms of spending incidence). Almost universally, these assumptions are violated in that the poor receive less valu- able or less costly services. For example, the spending and quality of schools and 11In this paper coverage often refers to the entire popula- tion even though the target group of the program is much smaller. The target group for unemployment insurance, for example, is the group of all unemployed rather than the entire population, and low coverage among the popu- lation does not necessarily indicate exclusion. The Poverty Targeting of Social Spending in Brazil Page 15 Education Figure 12 Share of the Poor in Public Student Population, Urban Brazil, by Type of Facility Education System Overview Brazil spent R$27.9 billion on public edu- cation in 1995, an amount equivalent to 4.3% of GDP. State governments, which Higher education have responsibilities for the provision of secondary education under the Constitu- Adult (continuing) education tion, both provide and manage a larger Secondary education share of the country's educational re- sources than either the federal or municipal Primary education governments (see Figure 10 and Figure Kindergarten Figure 10 Share of Public Education Resources Managed at the Federal, State, and Local Levels Creche 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% Percentage of students attending public institutions who are Federal poor poor students enrolled in public basic edu- 20% Municpal 31% cation (crèches, kindergartens, and primary schools) is above 50%. The share of the poor in more advanced public education declines significantly. The share of the poor in public secondary schools is approxi- mately 25% and the share of the poor in university education is 0%. State 49% Figure 13 summarizes the incidence findings Figure 11 Share of Public Education Resources for different education levels in urban areas, Originating at the Federal, State, and Local Levels highlighting the progressive nature of basic education levels (daycare, kindergarten and primary education). The benefit incidence Federal Municpal 25% of secondary and adult education is concen- 28% trated in the third, fourth, and fifth quintiles. Spending on higher education is extremely regressive, with the vast majority of students Figure 13 Incidence of Public Education by Level, Urban State 80% 47% Urban 70% 60% Level, by 11). State governments bear the primary 50% responsibility for funding and managing 40% Education education resources, followed by municipal 30% governments and the federal government. Public of 20% The share of the poor in the public student 10% population varies greatly depending on the Incidence 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) type of facility and the level of education Consumption Quintile provided. As Figure 12 shows, the share of Creche Kindergarten education Primary education Secondary education Source: IBGE-PPV, 1996/7 Adult education Higher education The Poverty Targeting of Social Spending in Brazil Page 16 Figure 14 Reach of Public Education Among the Urban Poor The reach of public primary education is also high for poor individuals with 15-19 years of age, with 45% of the poor indi- Higher (20-24 years) viduals in this age group still attending public primary Adult Ed (15+ years) school. Secondary school en- rollment among the poor drops Secondary (15-19 years) significantly and is only 10% for poor individuals with 15-19 Primary (15-19 years) years of age. Likewise, very few of the poor are enrolled in Primary (7-14 years) public adult and university Kindergarten (4-6 years) education. Creche (0-3 years) Crèches 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% Coverage with daycare centers Percentage of all urban poor in the relevant age group attending a public educational facility (crèches) for very young chil- dren (0-3 years) is very low and increases with consump- enrolled in public university education tion. Coverage reaches more than five per- coming from the wealthiest quintile. The cent only for the top quintile. Public serv- PPV survey did not find any individuals from the first and Figure 15 Crèche Attendance, 0 to 3 Years, Urban second quintiles that were en- rolled in public universities. 100% The reach of public education 90% among the poor also varies sig- Old, 80% nificantly according to the type Years 70% of facility and level of education 3 to offered. As Figure 14 shows, a 60% 0 relatively small percentage poor 50% Urban children with 0-3 years of age 40% attend public crèches. Only 31% Attendance, 30% of poor children with 4-6 years 20% Creche of age attend public kindergar- 10% 0% tens. The reach of public pri- 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) mary education for poor children Consumption Quintile with 7-14 years of age is signifi- In public creche In private creche Not attending any cantly higher, with approxi- Source: IBGE-PPV, 1996/7 mately 81% of poor children in this age ices dominate for the first four quintiles but group attending public primary schools. are negligible for the top quintile. Public Table 2 Table 1 Distribution of Public Creche Attendees Creche Attendance, 0 to 3 Year Olds, by Consumption Quintile All NE and SE All urban NE and SE Per capita household consumption quintile per capita household per capita household 1st 2nd 3rd 4th 5th Consumption quintile consumption consumption All NE and SE 1 23.8% 35.1% In public creche 0.8% 2.3% 1.5% 2.7% 0.0% 2 32.6% 17.2% In private creche 0.6% 1.4% 0.5% 1.4% 19.7% Not attending any 99.2% 96.4% 98.0% 95.9% 80.3% 3 14.3% 28.3% All urban 4 22.6% 19.0% In public creche 2.1% 0.6% 3.1% 1.5% 0.0% 5 6.8% 0.4% In private creche 0.1% 1.9% 0.3% 2.7% 23.9% Total 100.0% 100.0% Not attending any 97.8% 97.5% 96.6% 95.9% 76.1% The Poverty Targeting of Social Spending in Brazil Page 17 attendance and the distribution Figure 16 Incidence of Children in Urban Public Crèche of public kindergarten atten- dance across the different con- sumption groups. Figure 17 40% and Figure 18 offer this same 35% information in graphical form. Creche 30% Primary Education Public 25% Primary school coverage (for Urban in 20% grades 1 through 8) among 15% rural and urban children ages 7- Children of 14 is 69% for the lowest quin- 10% tile and reaches 93% for the top Incidence 5% quintile. Urban primary edu- cation coverage is higher for all 0% 1 2 3 4 5 quintiles. Private primary Consumption Quintile Source: IBGE-PPV, 1996/7 schools are negligible for the poorest quintile but reach over daycare centers are progressive with almost one-quarter of all enrolled children (in urban 50% for the wealthiest quintile. As a result, and rural areas) coming from the bottom quintile. Table 2 and Table 1 Figure 17 Kindergarten Attendance, Urban 4-6 Year Olds provide more specific information regarding overall crèche attendance 100% and the distribution of public crèche 90% attendance across the different con- Old, 29.4% 80% sumption quintiles. Figure 15 and 43.4% Years 51.7% 54.4% 70% Figure 16 offer this same information 4-6 64.5% 60% in graphical form. 50% Urban 24.6% 40% 57.3% 19.0% Kindergarten Attendance, 4.3% 24.3% 30% Kindergarten coverage for children 20% 32.0% with 4-6 years of age extends from 31.2% 29.2% Kindergarten 10% 21.4% 13.3% 30% for the first quintile to 70% for 0% the top quintile. Public services 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile cover a share of 26-30% of children in the first four quintiles, while pri- In public kindergarten In private kindergarten Not attending any Source: IBGE-PPV, 1996/7 vate services increase with con- public primary school services are progres- sumption from 3% in the lowest to 49% in sive with 26% of the enrollment coming the highest quintile. Public kindergartens from the bottom quintile. Table 5 and Table are highly progressive with 42% of the en- 6 provide more specific information re- rolled coming from the poorest quintile. Table 3 and Table 4 provide more specific information regarding overall kindergarten Table 4 Distribution of Public Kindergarten Students All NE and SE All urban NE and SE Table 3 per capita household per capita household Consumption decile consumption consumption Kindergarten Attendance, 4 to 6 Year Olds, by Income Quintile 1 27.7% 23.3% Per capita household consumption quintile 2 14.2% 15.2% 1st 2nd 3rd 4th 5th 3 12.5% 12.7% All NE and SE 4 11.1% 10.0% In public kindergarten 26.5% 29.8% 28.7% 27.2% 20.6% 5 7.6% 9.9% In private kindergarten 3.0% 12.3% 18.6% 25.6% 49.1% 6 8.6% 7.6% Not attending any 70.5% 58.0% 52.7% 47.2% 30.3% 7 6.1% 8.9% All urban In public kindergarten 31.2% 29.2% 21.4% 32.0% 13.3% 8 5.6% 7.4% In private kindergarten 4.3% 19.0% 24.3% 24.6% 57.3% 9 5.6% 4.9% Not attending any 64.5% 51.7% 54.4% 43.4% 29.4% 10 0.8% 0.1% Total 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 18 Figure 18 Incidence of Children in Public Kindergarten in graphical form. Among the reasons for not attending 70% primary school, the urban poor most 60% often cite financial difficulties. Lack Public 50% of interest and lack of vacancy in in schools were other dominant reasons 40% for non-attendance. Among the ur- Children 30% of ban non-poor, financial difficulties Kindergarten 20% rarely present a reason for non- attendance. Lack of interest and lack Incidence 10% of vacancy in schools were cited 0% roughly as much by the non-poor and 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Per capita household consumption quintile poor alike as reasons for non- Fortaleza MR Recife MR attendance. Figure 20 and Figure 21 Salvador MR Belo Horizonte MR Rio MR Sao Paulo MR provide more information regarding Source: IBGE-PPV, 1996/7 Nonmetropolitan urban NE Nonmetropolitan urban SE the reasons for not attending primary education. garding overall primary school attendance and the distribution of public primary school As previously mentioned, the benefit inci- attendance across the different consumption dence varies between national and local quintiles. Figure 19 offers this information consumption quintiles. Figure 4 and Figure 5 show the distribution among public Figure 19 Primary School Attendance, Urban 7-14 Year Olds primary school students across the national and local consumption quintiles (see Page 10). From the 100% national perspective, public primary Old, 90% enrollment appears highly regressive for Southeastern urban areas and Years 80% highly progressive for Northeastern 7-14 70% 60% urban areas. When analyzed from 50% the regional perspective, public pri- Urban mary education appears highly pro- Attendance, 40% gressive for both Southeastern and 30% Achool Northeastern urban areas. The rea- 20% son for this difference is that there 10% Primary are relatively few poor (in the na- 0% tional context) in the Southeast and 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile there are relatively many poor (in the national context) in the Northeast. In public primary school In private primary school Not attending any Source: IBGE-PPV, 1996/7 Consequently, a program such as education will tend to benefit more Table 5 Table 6 Primary School Attendance, 7 to 14 Year Olds, by Consumption Quintile Distribution of Public Primary School Students Analysis is based on per capita household consumption All NE and SE All urban NE and SE Per capita household consumption quintile per capita household per capita household 1st 2nd 3rd 4th 5th All NE and SE Consumption decile consumption consumption In public primary school 68.3% 81.6% 84.6% 71.6% 41.5% 1 13.2% 15.0% In private primary school 0.5% 5.5% 6.2% 21.5% 51.1% 2 12.8% 14.6% Not attending any 31.2% 12.9% 9.3% 6.9% 7.4% 3 13.3% 13.1% All urban 4 13.3% 13.4% In public primary school 78.2% 85.3% 81.6% 64.0% 38.7% 5 13.5% 12.9% In private primary school 3.7% 6.1% 9.1% 29.7% 54.5% 6 9.9% 8.9% Not attending any 18.1% 8.6% 9.3% 6.3% 6.8% 7 9.6% 9.2% 8 6.9% 6.1% 9 5.5% 5.1% 10 2.1% 1.6% Total 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 19 Figure 20 Reasons for Not Attending Primary School, Urban Figure 23 Percentage of 15-19 Year Olds Attending Primary Poor, 7-14 Years Old School, Urban Brazil 50% 47% Work 45% 2% No vacancy Urban 44% 12% 40% School, Other No facility nearby 35% 30% 6% 33% Primary 33% 30% Attending 25% Completed the Olds desired level of 20% education Year 0% 15-19 15% of 12% 10% Financial Percentage 5% Not interested difficulties 17% 33% 0% 1 2 3 4 5 Source: IBGE-PPV, 1996/7 Per capita household consumption quintile school. Primary school attendance among Figure 21 Reasons for Not Attending Primary School, Urban Non-Poor, 7-14 Years Old 15-19 year-olds declines with increasing consumption, and is only 12% for the wealthiest quintile. Work 9% Secondary Education No vacancy 14% No facility nearby 0% Attendance at the secondary level (for Completed the desired grades 9 through 12) of children with 15-19 level of education 1% Financial difficulties Other 4% years of age drops drastically compared to 54% primary school attendance. Attendance is only 5% for the bottom quintile but reaches Not interested 50% for the top quintile. Almost all services 18% in the first three quintiles are public. Private services dominate for the wealthiest quintile. Source: IBGE-PPV, 1996/7 The benefit incidence of public secondary education is highly concentrated in the third of the poor in the Northeast simply because and fourth quintiles. The first quintile re- there is a higher concentration of poor in the ceives only 7.5% of the service. Table 7 and Northeast. Table 8 provide more specific information regarding overall secondary school atten- Primary school attendance is also high for dance and the distribution of public secon- individuals with 15-19 years of age from the dary school attendance across the different first and second quintiles. As shown in consumption quintiles. Figure 24 offers this Figure 23, 47% of poor individuals with 15- 19 years of age still attend primary school. Table 8 From the second quintile, 44% of individuals in this age group attend primary Distribution of Public Secondary School Students All NE and SE All urban NE and SE Table 7 per capita household per capita household Consumption decile consumption consumption Secondary School Attendance, 15 to 19 Year Olds, by Consumption Quintile 1 1.5% 2.1% Per capita household consumption quintile 2 5.9% 5.6% 1st 2nd 3rd 4th 5th 3 4.3% 7.8% 4 7.8% 11.2% All NE and SE 5 11.3% 13.6% In public secondary school 5.1% 9.8% 20.6% 27.7% 22.4% In private secondary school 0.2% 0.0% 3.1% 10.0% 28.0% 6 16.7% 14.2% Not attending any 94.7% 90.2% 76.3% 62.3% 49.6% 7 17.8% 18.2% All urban 8 15.5% 11.4% In public secondary school 5.8% 15.8% 26.5% 30.0% 21.2% 9 13.8% 12.8% In private secondary school 0.0% 2.6% 4.6% 11.0% 31.5% 10 5.4% 3.2% Not attending any 94.2% 81.6% 68.9% 59.0% 47.3% Total 99.9% 100.0% The Poverty Targeting of Social Spending in Brazil Page 20 Lack of interest is the primary reason for not Figure 24 Secondary School Attendance, Urban 15-19 Year Olds attending school among the urban poor with 15-17 years of age. Work and financial 100% difficulties are other dominant reasons for Old, 90% non-attendance. Among the urban non- Years 80% poor, lack of interest is also the primary 70% 15-19 reason for non-attendance. While work is 60% also a dominant reason, it is less of an in- 50% Urban hibitor for the urban non-poor than for the Attendence, 40% urban poor. Financial difficulties is cited as 30% School much by the urban non-poor as the urban 20% poor as a reason for non-attendance. Figure 10% Secondary 25 and Figure 26 provide more information 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) on non-attendance among the urban poor Consumption Quintile and non-poor with 15-17 years of age. Source: IBGE-PPV, 1996/7 In public secondary school In private secondary school Not attending any Figure 27 and Figure 28 show the share of Figure 25 Reason for Not Attending School, Urban Poor 15-17 Years Figure 27 Share of Public Secondary Students Belonging to Each Old Consumption Group (Regional Quintiles) 50.0% 45.0% Other 17% Work 40.0% 23% 35.0% Fortaleza MR 30.0% Recife MR Salvador MR Belo Horizonte MR 25.0% Rio MR Sao Paulo MR No vacancy 20.0% Nonmetropolitan urban NE 6% Nonmetropolitan urban SE No facility nearby 15.0% 0% Completed the desired level 10.0% of education 0% 5.0% Not interested Financial difficulties 42% 12% 0.0% 1 2 3 4 5 Per capita household consumption quintile (regional) Figure 26 Reasons for Not Attending School, Urban Non-Poor 15-17 Figure 28 Share of Public Secondary School Students Belonging to Years Old Each Consumption Group (National Quintiles) 50.0% 45.0% Novacancy 40.0% Work 1% 15% No facility nearby 35.0% 0% Fortaleza MR Other Completed the desired level 30.0% Recife MR 33% of education 0% Salvador MR 25.0% Belo Horizonte MR Financial difficulties Rio MR 12% Sao Paulo MR 20.0% Nonmetropolitan urban NE Nonmetropolitan urban SE 15.0% 10.0% 5.0% 0.0% 1 2 3 4 5 Not interested Per capita household consumption quintile (national) 39% information in graphical form. public secondary school students belonging to each consumption group in terms of both The Poverty Targeting of Social Spending in Brazil Page 21 individuals in the area. Figure 29 Higher Education Attendance, Urban 20-24 Year Olds Higher Education 100% Higher education attendance Old, 90% among the 20-24 age group is Years 80% negligible for the first three 20-24 70% quintiles. In fact, the PPV 60% sample does not include a sin- 50% gle student of higher education Urban Attendance, 40% from the first quintile. Cover- 30% age is 5% in the fourth and Education 20% 33% in the fifth quintile. The 10% incidence of university service Higher 0% is extremely regressive with 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) 22% of the services provided Consumption Quintile to the fourth and 76% to the In public higher education In private higher education Not attending any Source: IBGE-PPV, 1996/7 fifth quintile. Table 9 and Table 10 provide more specific regional and national quintiles. It is appar- information regarding overall university ent from these figures that public secondary education is Figure 30 Incidence of Higher Education, Urban not well-targeted to the poorest 20% of the population neither in terms of national nor re- 80% gional quintiles. In terms of 70% national quintiles, the poorest Urban 60% 20% of the population receives less than 10% of the public 50% Education, secondary education benefit 40% for all urban areas except the Higher 30% of non-metropolitan urban North- 20% east. Even in terms of regional 10% Incidence quintiles, the poorest 20% of 0% the population still receives 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) less than 15% of the public Consumption Quintile secondary education benefit, Source: IBGE-PPV, 1996/7 again for all urban areas except the non-metropolitan urban Northeast. In fact, the non-metropolitan attendance and the distribution of public urban Northeast is the only region where the university attendance across the different poorest quintile receives more than 20% of consumption quintiles. Figure 29 and the secondary education benefit, but this is Figure 30 offer this same information in primarily due to the concentration of poor graphical form. Table 9 Table 10 Higher Education Attendance, 20 to 24 Year Olds, by Consumption Quintile Distribution of Students in Public Higher Education Per capita consumption quintile All NE and SE All urban 1st 2nd 3rd 4th 5th All NE and SE per capita household per capita household In public higher education 0.0% 0.0% 0.5% 3.2% 13.6% Consumption quintile consumption consumption In private higher education 0.0% 0.1% 0.0% 2.1% 19.3% 1 0.0% 0.0% Not attending any 100.0% 99.9% 99.5% 94.6% 67.1% 2 0.0% 0.0% All urban 3 2.5% 6.9% In public higher education 0.0% 0.0% 1.7% 2.9% 17.4% 4 21.8% 20.3% In private higher education 0.0% 0.1% 0.6% 3.3% 24.8% 5 75.7% 72.9% Not attending any 100.0% 99.9% 97.8% 93.8% 57.8% Total 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 22 is less than 1% for all Figure 31 Incidence of Adult Education, Urban quintiles with public services dominating throughout. Given the 40% low frequency of adult 35% education in the sample, Urban results on distributional 30% incidence should be 25% viewed with caution. Education, 20% The analysis shows only Adult 15% 5% incidence in the bot- of tom quintile and a con- 10% centration of incidence in Incidence 5% the third to fifth quin- 0% tiles. Table 11 and Table 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) 12 provide more specific Consumption Quintile information regarding overall adult education Source: IBGE-PPV, 1996/7 attendance and the dis- tribution of public adult Adult and Professional Education education attendance across the different The coverage of adult and professional edu- consumption quintiles. Figure 31 offers this cation among individuals 15 years and older information in graphical form. Table 11 Table 12 Adult Education Attendance, Ages 15 and Older, by Consumption Quintile Distribution of Public Adult Education Students Per capita household consumption quintile All NE and SE All urban 1st 2nd 3rd 4th 5th per capita household per capita household All NE and SE Consumption quintile consumption consumption In public adult ed classes 0.1% 0.3% 0.6% 0.4% 0.5% 1 4.5% 7.4% In private adult ed classes 0.0% 0.0% 0.4% 0.1% 0.2% 2 15.0% 9.4% Not attending any 99.9% 99.7% 99.0% 99.5% 99.4% 3 30.1% 36.3% All urban 4 22.5% 18.3% In public adult ed classes 0.2% 0.2% 0.8% 0.4% 0.6% 5 27.9% 28.7% In private adult ed classes 0.0% 0.2% 0.3% 0.2% 0.2% Total 100.0% 100.0% Not attending any 99.8% 99.6% 98.9% 99.4% 99.2% The Poverty Targeting of Social Spending in Brazil Page 23 Health Care provided by those governments' public fa- Health Care System Overview12 cilities. Brazil spent roughly R$21.7 billion, corre- sponding to 3.4% of GDP on health care in The Poor's Access to Health Care 1995. Public outlays on health accounted The PPV provides new insight into the for 16% of all social spending. question of whether Brazil is achieving its goal of universal access to health care, espe- The current structure of Brazil's public cially for the poor. According to the PPV, health system has its roots in the Constitu- Brazil's poor are less likely than the non- tion of 1988, and the basic implementing poor to use health services. The gap in legislation for the health sector approved in health service utilization varies depending 1990. Known as the Sistema Unico de on the geographic area, but consistently Saude (SUS), the Brazilian health system favors the non-poor. If it were true that the has several features that make it unique in poor are less sensitive to health problems Latin America. than the rich, one explanation for the poor's lower usage of health care could be that they First, it is the only Latin American system are less likely to perceive the existence of a with a substantial separation of finance and health problem requiring medical attention. provision, where finance is largely public That is, the poor's lower usage of health while provision is predominantly private. services does not result from less access, but What is still more striking is the large share from their lower demand for care. The evi- of private provision that is for-profit and dence does not support this supposition. publicly financed. First, Brazil's poor are more likely than the non-poor to identify themselves as having a Second, Brazil is the only country in Latin health problem for which they did not obtain America to have completely eliminated the treatment. Second, while the majority in traditional separation between the Ministry both groups give "not necessary" as the of Health and the Social Security System. reason they did not get medical attention Coverage under one public system is univer- despite having a health problem, this ex- sal, and public money can be used to finance plains a much smaller share of the poor's any type of provider. failure to seek care. 80% of the non-poor who did not get treatment felt that it was not Third, along with the separation between necessary. Only 68% of the poor without finance and provision, there is also a separa- treatment felt that they did not need treat- tion by level of government, with financing ment. To summarize, lower use of health being handled mostly at the federal level but services by Brazil's poor cannot be ex- public provision being almost entirely a plained by differences in perceptions re- municipal responsibility. Both state and garding the need for medical attention. federal governments are responsible for offering the technical and financial support What limits access to health care? As that municipalities need to assure provision Figure 33 shows, poor individuals who did of health care services. Thus, a large share not obtain treatment even though they of funding occurs as inter-governmental thought it was needed most often cited a transfers, either directly to a state or munic i- lack of money for transport or treatment. pal government or as purchases of services They were five times more likely than the non-poor to give financial difficulties as the 12 The description of the Brazilian health system draws reason for not seeking medical care. As extensively from Brazil: Social Spending in Selected shown in Figure 31, the non-poor were more States, World Bank Report No. Br-17763, specifically, likely than the poor to state that treatment from chapter 3 on health, which was written by Philip Musgrove. takes too long as the reason for their not The Poverty Targeting of Social Spending in Brazil Page 24 seeking care; in fact, it was their second Less than 10% of the non-poor treated in most common explanation. They rarely private facilities paid anything for their care; mentioned lack of money as the constraint. even fewer patients in publicly operated Distance to the health facility was one of the health facilities incurred out-of-pocket ex- most frequently cited obstacles to obtaining penses. Because the government typically treatment for poor and non-poor Figure 33 Reasons for Not Seeking Medical Care, Urban Poor alike, although it was clearly a bigger problem for the poor. Facility didn't have Facility didn't have a When rural areas are excluded, the an appointment 2% specialist Other share of the poor with an untreated Treatment is too 1% 10% slow health problem is smaller; and the 4% proportion of the non-treated who say treatment was unnecessary is No time to obtain larger. The reasons "lack of treatment 2% money for transport or treatment" Lack of money for and "distance to the health facility" transport or are less of an obstacle for the urban treatment 7% poor than for those in rural areas, though these two obstacles are still Transport difficulties 0% the most common explanations for Not necessary 68% not obtaining care only after the Health facility is too reason that medical care was not far away necessary. 6% Source: IBGE/PPV 1996/7 An interesting question is whether the transport costs or the medical Figure 31 Reasons for Not Seeking Medical Care, Urban Non-Poor costs present more of an obstacle to the poor in their attempt to ob- Facility didn't have Facility didn't have a an appointment tain medical care. Unfortunately, Treatment is too specialist 0% Other 0% the PPV does not distinguish be- slow 5% 7% tween these two types of costs. No time to obtain However, the survey does provide treatment 2% some indirect evidence that travel Lack of money for transport or costs are the greater constraint. treatment First, the distance to the health 3% facility was one of the most fre- Transport difficulties quently given reasons for not 0% seeking medical attention. Second, Health facility is too the vast majority of patients, far away whether poor or non-poor, do not 3% pay for medical treatment in any of Not necessary the publicly subsidized health fa- 80% cilities (public hospitals and health Source: IBGE/PPV 1996/7 posts, and private hospitals and clinics with SUS agreements). For urban pays for health costs, it is reasonable to as- areas covered by the PPV, less than 2% of sume that transportation costs comprise the the poor who received medical care in a largest part of the financial constraint that public or privately operated SUS facility prevents the poor from seeking medical reported paying anything for their treatment. attention. The non-poor were only slightly more likely to pay for care in these health facilities. The Poverty Targeting of Social Spending in Brazil Page 25 Basic Health Care and Prenatal Care and hospitals are made by the non-poor. The PPV also provides information on sev- eral specific types of medical treatment of Public health facilities are critical to the particular relevance to the poor. Aside from urban poor's access to prenatal care. In measuring the poverty targeting of health care programs, it is impor- Figure 33 Where Health Care is Obtained, by Consumption Quintile, Urban tant to evaluate the types of serv- Brazil ices received by the poor. Early intervention with prenatal pro- 100% grams deserves special mention in Other 90% this analysis because of the tre- Pharmacy mendous impact such care can 80% Urban render on the health outcomes of 70% Private clinic both mother and child in impover- 60% SUS clinic Facility, ished areas. of 50% SUS hospital Type 40% Private hospital Prenatal Care by 30% Own house Subsidies to prenatal care are bet- 20% ter targeted to the urban poor than Coverage Public health post health care overall. According to 10% Public hospital the PPV, roughly 46% of patients 0% 1st (Poorest) 2nd 3rd 4th 5th receiving prenatal care in public (Wealthiest) health facilities are poor. As in the Consumption Quintile Source: IBGE-PPV, 1996/7 case of health care in the aggre- gate, the share of poor in the client popula- urban areas, the public hospital and health tion varies tremendously by type of facility. post system delivers nearly 95% of all pre- Nearly two-thirds of prenatal visits to public natal care received by poor women. The health posts and health centers are by poor non-poor are much less dependent on the women; at the other extreme, over three- public system, as they obtain approximately fourths of all visits to private SUS clinics one-third of their prenatal care in private (non-SUS) clinics. Nevertheless, the vast majority of all women, poor and non- Table 13 poor alike, are obtaining their prenatal Where Health Care is Obtained, by Consumption Quintile care through publicly subsidized heath All NE and SE Per capita consumption quintile services. Only the wealthiest 20% of Service Location 1st 2nd 3rd 4th 5th Public hospital 47.5% 47.8% 40.9% 32.8% 12.2% women currently receiving prenatal Public health post 34.6% 38.2% 35.6% 18.8% 6.7% care are more likely to go to a private Own house 1.5% 0.4% 0.1% 0.6% 1.5% Private hospital 0.0% 1.0% 2.5% 3.1% 8.7% clinic than one of the publicly subsi- SUS hospital 3.9% 3.6% 3.1% 8.2% 11.5% dized options. SUS clinic 1.7% 1.9% 6.1% 15.7% 23.4% Private clinic 1.8% 4.4% 4.6% 16.6% 33.9% Pharmacy 3.9% 1.9% 3.4% 2.6% 1.1% Targeting: The Incidence of Health Other 5.0% 0.8% 3.8% 1.6% 1.1% Total 100.0% 100.0% 100.0% 100.0% 100.0% Subsidies Data of health facility usage from the Table 14 PPV show that the poor almost exclu- Where Health Care is Obtained, by Consumption Quintile sively rely on public health care All urban Per capita consumption quintile whereas there is significant participa- Service Location 1st 2nd 3rd 4th 5th tion of private health care provision in Public hospital 52.0% 46.6% 30.1% 33.4% 9.6% Public health post 36.3% 34.1% 34.5% 13.4% 5.2% the higher quintiles (see Figure 33, Own house 0.0% 0.0% 0.0% 1.3% 1.3% Table 13, and Table 14). At the same Private hospital 0.0% 2.0% 2.8% 4.9% 8.2% SUS hospital 4.9% 4.9% 5.4% 7.9% 12.5% time, there is significant usage of pub- SUS clinic 1.7% 4.0% 10.3% 19.3% 23.3% Private clinic 2.1% 4.9% 9.6% 18.0% 37.3% Pharmacy 1.4% 1.0% 3.7% 0.9% 1.2% Other 1.5% 2.5% 3.5% 0.9% 1.4% Total 100.0% 100.0% 100.0% 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 26 Figure 34 Incidence of Overall Public Health Care, Urban privately run SUS hospitals and clinics on the other, differs greatly. Public hospitals and health posts 25% are rarely used by the top quintile, Urban and usage share by the poor ex- 20% ceeds their share of the population. Care, On the other hand, usage of SUS Health 15% hospitals and clinics (private fa- cilities publicly funded through Public 10% SUS) by the top quintile is over All of 40% of total usage but is almost 5% negligible at the bottom of the distribution.13 Incidence 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile This difference in utilization pat- terns has major implications for the degree of poverty targeting accom- Source: IBGE-PPV, 1996/7 plished by public subsidies to lic facilities, and in particular of private but health care in public versus privately run publicly funded facilities (SUS hospitals and facilities. Almost no one pays for health clinics) among the better-off. services received in hospitals and clinics belonging to the SUS system. It is reason- The incidence of health spending based on this simplified analysis needs to Table 15 be treated with great caution. Clearly, medical services and Percent of the Population Utilizing Health Services in the 30 Days prior to the Survey, by Consumption Quintile their costs differ greatly by fa- (regional distributions) cility, and presumably by con- Per capita consumption quintile Locale 1st 2nd 3rd 4th 5th sumption group of the patient. All NE and SE 9.4% 10.8% 13.1% 15.6% 17.6% Ignoring these complications, All urban 10.7% 12.8% 13.1% 14.9% 18.4% overall usage of publicly funded able to assume that the unit cost of serving health care appears to be almost flat across the poor is not higher than that for serving consumption groups. In other words, the the non-poor (and may well be lower). poor receive a share of public health serv- Therefore, conservatively, at least 80% of ices approximately proportional to their population share. Given the high income elasticity of health service demand (health 13 SUS hospitals and clinics, as discussed in this report, service demand typically rises more than refer to the items hospitais conveniados and clinicas proportional with income), this should not conveniados in the PPV questionnaire. The guide to the questionnaire explains that these terms refer to convênios be simplistically interpreted as a negative with the public health system. These facilities are thus finding without further analysis. Table 15 publicly funded. If the questionnaire was applied without and Figure 34 provide further information due reference to the guide, the question might have been misunderstood by respondents as referring to private on the incidence of overall public health convênios as well. In this case, the above analysis could care and health service demand. overstate the regressiveness of public spending for SUS hospitals and clinics. Piola and Nunes (2000) have un- A very diverse picture of the incidence of dertaken similar analysis of the incidence of health spending. In addition to some other methodological health spending emerges once data is disag- differences, they have excluded hospitais conveniados gregated by type of facility (see Figure 35, and clinicas conveniados from the publicly funded facili- Table 16, and Table 17). The share of poor ties. As a result, they find more progressive overall public health spending with 22.9%, 23.5%, 22.5%, among the patients served by public hospi- 18.7%, and 12.5% of federal public health spending tals and health posts on the one hand, and accruing to the first through fifth income quintile, respec- tively. The Poverty Targeting of Social Spending in Brazil Page 27 Figure 35 Incidence of Public Health Care by Type of Facility 84% of the health care received by the urban poor occurs in public hospitals 50% and health centers; privately run hos- 45% pitals and clinics with SUS agreements Facility, of 40% provided only 8% of the urban poor's 35% Type health services. On the other hand, by 30% private facilities do increase the non- Care 25% poor's access to publicly subsidized Urban 20% health care. According to the PPV Health 15% results, more than 25% of the medical Public treatment for the urban non-poor oc- of 10% curred in private SUS facilities. Usage 5% of publicly-funded private health care Incidence 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) facilities therefore appears to be highly Consumption Quintile regressive. In the case of hospitals, Public hospitals Public health posts/centers these facilities account for more than Source: IBGE-PPV, 1996/7 SUS hospitals SUS clinics half of SUS spending. the public subsidies to health services in privately run hospitals, and 90% of those in private clinics, are captured by the top three Piso de Assistência Básica (PAB) quintiles. By comparison, subsidies to pub- The new Piso de Assistência Básica (PAB) establishes a transfer of R$10-12 per person Table 16 per year for primary health care, which Percent of Patients in Different Facilities, by Consumption Quintile would be withdrawn from the most recent All NE and SE 25% increase in SUS tariffs. This program Per capita Public would leave slightly less federal funding consumption All Public Public SUS Health SUS quintile Care Hospitals Hospitals Post Clinic available for hospitals and more complex 1 16.3% 20.1% 8.3% 20.2% 2.1% ambulatory procedures. In 1998, the Minis- 2 19.5% 23.2% 8.8% 25.5% 2.7% try of Health further defined the basic pro- 3 22.2% 24.1% 9.0% 28.9% 10.3% cedures to be included in the health care 4 23.3% 23.0% 28.6% 18.2% 31.7% 5 18.6% 9.6% 45.3% 7.2% 53.2% package funded by PAB and established the All 100.0% 100.0% 100.0% 100.0% 100.0% values for each of the health procedures Table 17 covered by the PAB system. The purpose of PAB is to increase the extent and availabil- Percent of Patients in Different Facilities, by Consumption Quintile ity of basic public health care to the poor at All urban Per capita Public the expense of reducing provision of more consumption All Public Public SUS Health SUS complex and costly services (usually deliv- quintile Care Hospitals Hospitals Post Clinic ered in SUS hospitals and clinics). 1 19.4% 25.1% 9.9% 24.7% 2.0% 2 21.7% 26.7% 11.6% 27.6% 5.5% 3 20.2% 17.8% 13.4% 28.9% 14.8% Benefits for the Poor 4 21.0% 22.4% 22.1% 12.7% 31.2% There is no doubt that the Government's 5 17.7% 7.9% 42.9% 6.1% 46.5% All 100.0% 100.0% 100.0% 100.0% 100.0% efforts to provide free health care for the lic hospitals and health posts are rather well- poor has had a positive impact. As an ex- targeted, with approximately half of the ample, Table 18 provides evidence of the subsidies reaching the bottom two great reduction in infant mortality in Brazil. quintiles. Table 18 Moreover, private facilities in the SUS Infant Mortality 1980 1985 1990 1997 system play a minor role in improving Brazil 85.64 66.59 47.81 36.70 the poor's access to health care. Over North 83.61 63.30 44.59 35.60 Northeast 120.46 95.27 74.30 59.05 Southeast 64.44 47.96 33.57 25.23 South 57.70 41.18 27.36 22.55 Centerwest 66.44 44.15 31.19 25.39 Source: Simões (1997). The Poverty Targeting of Social Spending in Brazil Page 28 Nutrition Programs Programs for the distribution Figure 37 Coverage with Regular School Lunches of free milk achieve highest coverage among the second 80% quintile with almost 15% re- ceiving milk. The incidence is 70% heavily concentrated in the 60% first (29%) and second (33%) Lunches quintiles (see Table 19, Table 50% School 20, and Figure 36). In 1994, a 40% new milk distribution program Regular (Milk is Health) was created 30% with with a design intended to avoid 20% critical shortcomings in tar- Coverage geting, monitoring and evalua- 10% tion of previous programs. 0% The program is targeted at All students Kindergarten students Primary students Secondary students children (from six months to Poorest quintile Second quintile Third quintile Fourth quintile Richest quintile Source: IBGE-PPV, 1996/7 two years) and pregnant women under nutritional risk. The target program is decentralized, and federal re- population is selected in health centers. The sources from the Ministry of Health are transferred to each munic i- Figure 36 Incidence of Milk Distribution Programs pality on the basis of population size and esti- mates of child malnutri- 40% tion. The number of mu- 35% nicipalities participating in the program is increasing 30% Porgrams (587 in 1996; 999 in 25% 1997), however, fluctua- 20% tions in resources from Distribution year to year (R$ 139.6 Milk 15% of million in 1995, R$ 29.2 10% million in 1996, R$ 98.2 million in 1997) and inter- Incidence 5% ruptions in the transfer of 0% resources have been a 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) major problem. Per capita consumption quintile Source: IBGE-PPV, 1996/7 Free school meal programs Table 19 reach about 60% of students from the first Share of Population Receiving Donated Milk Belonging to Each Consumption Expenditure Table 20 Quintile Per capita Percentage in Quintile Living in Households that Received consumption Donated Milk quintile All NE and SE All urban Per capita consumption 1st 29.3% 34.5% quintile 1st 2nd 3rd 4th 5th 2nd 32.8% 26.8% All NE and SE 13.0% 14.5% 8.0% 5.9% 2.9% 3rd 18.2% 20.8% All urban 9.2% 7.2% 5.5% 2.0% 2.7% 4th 13.2% 7.7% 5th 6.5% 10.2% Total 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 29 Figure 38 Incidence of School Lunches three quintiles (see Table 21 and Figure 37). The incidence is closely related to the 40% income distribution of children in public 35% schools at different levels. The incidence 30% for free feeding at kindergartens is highly Lunches progressive, with more than 35% of the 25% benefit accruing to the first quintile. At School 20% primary schools, the incidence is also pro- of 15% gressive, with approximately 25% of the benefit accruing to each of the bottom three Incidence10% quintiles. At secondary schools, however, 5% only 5% of meals accrues to the first quin- 0% tile while 36% accrues to the fourth quin- 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) tile (see Table 22 and Figure 38). Consumption Quintile Kindergarten students Primary students Secondary students Source: IBGE-PPV, 1996/7 Table 21 Table 22 Percentage of students within consumption group with access to regular school Distribution of students in schools that provide regular school lunches, urban only lunches, urban only Per capita Per capita consumption consumption quintile All students Kindergarten Primary quintile 1 2 3 4 5 Secondary 1 25% 37% 25% 5% All students 58% 58% 60% 47% 22% 2 24% 23% 25% 19% Kindergarten 3 24% 18% 24% 25% students 64% 47% 42% 37% 13% 4 18% 16% 17% 36% Primary 58% 64% 72% 58% 39% 5 9% 5% 9% 15% Secondary 26% 35% 31% 37% 15% All 100% 100% 100% 100% The Poverty Targeting of Social Spending in Brazil Page 30 Urban Services first quintile. Whether these incidence fig- Housing ures are appropriate for the estimation of the The urban poor are typically concentrated in incidence of public spending depends on two types of informal areas: favelas and whether publicly funded programs are in illegal land subdivisions. Table 23 indicates fact directed at the neighborhoods that are the large number of favela homes in Brazil's characterized as favelas by the PPV respon- major cities. The share of favela homes dents. ranges from only 0.1% in Porto Alegre to over 40% in Recife. The Northern cities Many of the other urban poor live in the generally have a higher share of favela second type of informal area, illegal land homes than the Southern cities (however, subdivisions (loteamentos clandestinos). In the local definition of favela varies). While many cities, illegal subdivisions house the more recent and poorer migrants. Illegal Figure 39 Share of Quintile Living in Favela subdivisions result from illegal commer- cial operations in the periphery of a city 30% or a metropolitan region. Illegal land subdivisions are spatially more organized 25% Favela than favelas. Their physical layout is a in 20% somewhat similar to the formal areas of the city, often with clear street patterns Living 15% and lot arrangements but without planned open spaces for public facilities such as Quintile 10% of schools and health centers. However, 5% these lots are not serviced unless the Share Government later decides to provide 0% basic services. 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile (Regional) The types of housing tenure in urban Fortaleza Recife Salvador Brazil exhibit a peculiar pattern across Source:IBGE-PPV,1996/7 Belo Horizonte Rio de Janeiro Sao Paulo household incomes (see Figure 40). As favelas are often understood as low-income the PPV survey data indicate, those who neighborhoods, their residents are in fact own a house but not land (typically in in- quite heterogeneous. As Figure 39 and formal areas) and those who live in ceded Table 24 show, the non-poor compose a properties have the lowest average incomes. substantial share of favela residents ex- The majority of Brazil's urban households, ceeding 50% in some cities. The share of with the next higher level of mean income, the population that lives in favelas is 11% for the lowest quintile and declines to less Table 23 than one percent for Number of Favela Homes in Number of Homes (in Number of Favela Homes (in Favela as % of total Homes Brazilian City, 1990 thousands) thousands) the top quintile (see São Paulo 3,668 213 5.8 Table 25). Even in Rio de Janeiro 2,409 236 9.8 Belo Horizonte 517 52 10 urban areas, 86% of Salvador 470 17 3.7 Brasilia 422 0 0.1 the poor do not live in Porto Alegre 386 25 6.5 favelas. About 34% of Fortaleza 384 51 13.3 Curitiba 328 22 6.7 favela residents are Recife 311 131 42.2 Belem 257 39 15.1 from the first quintile. Goiania 251 4 1.6 Within the urban ar- Campinas 221 39 15.1 Manaus 218 10 4.7 eas, 44% of favela Santos 159 11 7.1 São Luis 151 6 3.9 residents are from the All above 15 cities 10,152 1,005 9.9 Source: Adapted from Gilbert (1996, Table 4.8). The Poverty Targeting of Social Spending in Brazil Page 31 Table 24 Table 25 Distribution of favela residents across the consumption quitiles Percentoftheconsumptionquintilelivinginfavelas,byregion Consumption quintile All NE and SE All urban Consumptionquintile 1 34.4% 44.4% 1st 2nd 3rd 4th 5th 2 26.7% 24.2% AllNEandSE 11.1% 6.7% 3.7% 3.5% 0.7% 3 16.8% 16.8% Allurban 13.6% 5.8% 3.5% 2.4% 0.4% 4 17.9% 12.2% 5 4.2% 2.3% All 100.0% 99.9% Transport own the house and the land on which the The share of commuters who use public house stands and have no outstanding loan transport declines with consumption level. for the house. The number of households Figure 41, Table 26, and Table 27 show the who rent accounts for only a small share incidence of different modes of transport. (12%) of all households, and their average Most of the poorest do not make regular incomes are relatively higher. The highest commuting trips because they do not have average income is found in the group of regular jobs. When they make trips, they are often unable to afford the cost of transpor- Figure 40 Housing Tenure and Poverty Rate, Urban Brazil tation requiring some form of payment and must instead walk to their destination. 70% 500 Where the poor have little choice of modes 450 60% other than walking, they have to spend time 400 (R$) and personal energy that could otherwise 50% 350 be used for more productive activities. 300 Capita 40% Per 250 As depicted in Figure 42, public transport 30% 200 Income is predominantly used by the second to 20% 150 fourth quintiles (population groups that 100 Monthly have a stable job but are not among the 10% 50 better-off and able to afford automobiles). 0% 0 In fact, the incidence of public transport for Own House still Rent Own House Ceded Own House Paying already Paid with already Paid the first quintile is only 9%, reaching a Own Land without Own Land high of 28% for the fourth quintile. Not Share of Pop Share of Poor Monthly Income Per Capita Poverty Rate surprisingly, the use of individual automo- Source: IBGE-PNAD 1996 biles is highly concentrated in the top two households who are still paying for their quintiles. While most public transport in- homes (or in other words, those who have been able to secure financing Figure 41 Incidence by Mode of Transport, Urban Commuters for their house). 70% Both the poor and non-poor house- holds in most cities have over 65% 60% home ownership (although without Commuters 50% title in many cases). Less than 25% Urban of the households live in rented 40% homes. The only significant differ- ence is that more poor households Transport, 30% of live in ceded and invaded property Mode 20% than non-poor households. by 10% Incidence 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile Source: IBGE-PPV, 1996/7 Public transport On foot Private Vehicle The Poverty Targeting of Social Spending in Brazil Page 32 Table 26 Table 27 Share of workers in each consumption quintile using various modes of Share of workers in each consumption quintile using various transportation to commute modes of transportation to commute All NE and SE All urban Transport Mode 1st 2nd 3rd 4th 5th Transport Mode 1st 2nd 3rd 4th 5th Public transport 17% 15% 15% 12% 10% Public transport 23% 20% 18% 15% 12% By foot 53% 48% 34% 32% 23% By foot 48% 39% 26% 23% 14% Private vehicle 2% 2% 7% 17% 35% Private vehicle 2% 3% 9% 21% 42% Other 6% 6% 4% 6% 3% Other 7% 6% 4% 5% 2% None (work where reside) 23% 29% 40% 33% 29% None (work where reside) 19% 32% 43% 36% 29% Total 100% 100% 100% 100% 100% Total 100% 100% 100% 100% 100% vestments in busways, suburban trains, and ments geared toward individual automobile metros are not as well-targeted as many users, almost all of which come from the top other social investments, their targeting is two quintiles. very favorable in comparison with invest- The transport voucher provided Figure 42 Incidence of Public Transport by Quintile by employers (Vale Transporte) is tied to formal employment. 30% Because the poor are typically employed only informally, trans- 25% port voucher coverage among the 20% first quintile is very low (7%) and Transport the benefit incidence for the first 15% Public quintile is even lower (6%). In of fact, 25% of the transport 10% voucher value accrues to the Incidence 5% wealthiest quintile while only 8% of the value accrues to the poor 0% (see Figure 43, Table 28, and 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile (Regional) Table 29). Fortaleza Recife Salvador Belo Horizonte Rio de Janeiro Sao Paulo Source: IBGE-PPV, 1996/7 Water and Sanitation The coverage of water and sani- Figure 43 Value Incidence of Vale Transporte, Urban Brazil tation services increases strongly with consumption levels. Cover- 30% age with safe water ranges from 36% for the first decile to 96% 25% for the wealthiest decile (see Table 30). Figure 44 shows the Transporte 20% Vale of 15% Table 29 Value of Distribution of the value of Vale Transporte received 10% by workers across the consumption quintiles. Per capita consumption quintile Incidence All NE and SE All urban 5% 1st 5.6% 8.1% 2nd 14.7% 19.7% 3rd 22.5% 21.7% 0% 4th 26.4% 25.7% 1 2 3 4 5 5th 30.8% 24.8% Source: IBGE-PPV, 1996/7 Per capita consumption quintile All 100.0% 100.0% Table 28 Workers' access to Vale Transporte by consumption quintile. Per capita consumption quintile 1st 2nd 3rd 4th 5th All NE and SE 7.2% 13.0% 18.0% 19.0% 15.8% All urban 13.1% 21.7% 19.6% 21.8% 14.4% The Poverty Targeting of Social Spending in Brazil Page 33 percentage of homes in each quintile that do not have access to safe water. Public sewer connections reach from 11% for the first quintile to 84% for the top quin- Table 30 tile. Figure 45 depicts the coverage of the Percentage living in households with access to safe water, by national consumption quintiles sewer network, and Table 31 shows the Per capita expenditure quintile percentage of households in each quintile 1st 2nd 3rd 4th 5th All NE and SE 36.1% 59.4% 77.9% 88.6% 95.8% that do not have access to any sanitation All urban 66.1% 81.6% 93.4% 95.1% 98.3% service. 45% of the poor live without any Figure 44 Share of Quintiles living Without Access to Safe Water kind of sanitation service; only 0.1% of the wealthiest quintile live in the same condi- tion. The incidence of services is regressive 80% with 12% of water services accruing to the Water 70% bottom quintile and 26% to the top quintile Safe to 60% (see Table 32). The differences are higher 50% for sewage network connection of which Access only 4% reaches the poor and 32% reaches 40% the wealthiest 20% of the population, in without 30% terms of national quintiles (see Table 33). 20% Quintile of 10% It is important to note that the incidence of service connection corresponds to the inci- Share 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) dence of current Government spending only Consumption Quintile if public spending covers a fixed amount of Fortaleza Salvador the current costs of services provided. If Recife Belo Horizonte Source: IBGE-PPV, 1996/7 Rio Sao Paulo public spending is mostly on the investment Nonmetropolitan urban NE Nonmetropolitan urban SE costs for new connections, the incidence of spending can be significantly different from Figure 45 Coverage with Sewage Network the incidence of current connections, which would then reflect the result of aggregate 30% spending over the past. An approach for determining the incidence of incremental 25% spending that would be appropriate in this case was recently proposed by Lanjouw and 20% Ravallion (1998). Network Sewage 15% Table 32 With Incidence of access to water network 10% Consumption Quitile All NE and SE 1 12.0% Coverage 2 17.0% 5% 3 21.3% 4 24.0% 0% 5 25.8% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) All 100.0% Consumption Quintile (Regional) Source: IBGE-PPV, 1996/7 Table 33 Table 31 NE AND SE COMBINED Incidence of the population by type of sanitation used Type of sanitation, by consumption group Dwelling Per capita connected to Concrete consumption quintile 1st 2nd 3rd 4th 5th public sewer cesspit Cesspool Ditch Other None Public sewer 10.5% 35.1% 57.2% 72.6% 84.1% Per capita Concrete cesspit 5.0% 14.1% 14.0% 10.6% 8.3% consumption quintile Cesspool 26.6% 27.3% 18.1% 11.6% 5.9% 1st 4.0% 9.7% 29.7% 35.2% 41.4% 69.3% Ditch 5.6% 4.1% 3.7% 2.5% 0.1% 2nd 13.5% 27.1% 30.4% 25.5% 29.9% 21.6% Other 7.8% 5.6% 2.6% 1.3% 1.5% 3rd 22.0% 26.8% 20.2% 23.0% 13.9% 7.0% None 44.5% 13.9% 4.5% 1.3% 0.1% 4th 28.0% 20.4% 13.0% 15.7% 7.1% 2.0% All 100.0% 100.0% 100.0% 100.0% 100.0% 5th 32.4% 16.0% 6.6% 0.5% 7.7% 0.1% Total 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 34 Transfer Programs of transfer programs is the treatment of The PPV allows for the analysis of two cash transfer payments in the construction of transfer programs: unemployment insurance consumption quintiles. Two alternative and pensions. Since transfer payments from approaches are followed in this paper. First, consumption quintiles are based on actual Figure 46 Incidence of Unemployment Insurance Benefits observed consumption (uncorrected quin- tiles). This approach does not consider that the consumption of beneficiaries would 45% likely have been less if they had not received Receipts 40% the transfer benefit. Incidence analysis fol- 35% lowing this approach may make a program 30% Insurance look regressive exactly because it succeeds 25% in lifting individuals out of poverty. To 20% overcome this problem, corrected consump- 15% Unemployment tion quintiles were constructed by replacing of 10% actual consumption of beneficiary house- 5% holds with what consumption would have Incidence 0% 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) been in the absence of the transfer program. Consumption Quintiles Consumption in the absence of a transfer All NE and SE, uncorrrected All NE and SE, corrrected program was calculated by estimating the All urban, uncorrrected All urban, corrrected Source: IBGE-PPV, 1996/7 coefficient that transforms income to con- sumption for different types of households Figure 47 Incidence of Pension Receipts and applying this coefficient to the reduction in income resulting from the elimination of 70% transfer payments 60% The coverage of unemployment insurance is 50% less than one percent for all quintiles with Receipts highest coverage in the second quintile (see 40% Table 34). The incidence of the benefit Pension of 30% amount is concentrated in the second and 20% fifth quintiles with the first quintile receiv- Incidence ing only 4% (uncorrected) or 13% (cor- 10% rected) of the benefit (see Table 35 and 0% Figure 46). While the concentration of 1st (Poorest) 2nd 3rd 4th 5th (Wealthiest) Consumption Quintile benefits in the second quintile appears plau- All NE and SE, uncorrrected All NE and SE, corrrected sible (few workers in the first quintile are All urban, uncorrrected All urban, corrrected formal and thus eligible), the concentration Source: IBGE-PPV, 1996/7 in the fifth quintile deserves more detailed these two programs can constitute a large analysis. Although this anomaly could indi- proportion of household income, one im- portant issue in the analysis of the incidence Table 35 Table 34 Distribution of Unemployment Insurance, by Per Capita Household Consumption Quintile Recipients Amounts Proportion of population in consumption quintile receiving unemployment insurance All NE and SE All Urban All NE and SE All Urban Consumption Expenditure Quintile Consumption 1st 2nd 3rd 4th 5th Expenditure Quintile Uncorr Corr Uncorr Corr Uncorr Corr Uncorr Corr 1st 6.9% 17.9% 28.4% 33.4% 4.1% 13.1% 20.3% 24.7% All NE and SE, uncorrected 0.1% 0.7% 0.4% 0.1% 0.4% 2nd 40.3% 33.3% 32.1% 27.1% 32.5% 30.1% 28.0% 23.6% All NE and SE, corrected 0.3% 0.6% 0.3% 0.1% 0.4% 3rd 21.8% 18.5% 9.6% 10.2% 19.5% 14.8% 9.6% 10.0% All urban, uncorrected 0.6% 0.7% 0.2% 0.1% 0.5% 4th 6.5% 5.9% 4.5% 3.9% 6.4% 4.5% 3.4% 3.1% All urban, corrected 0.7% 0.6% 0.2% 0.1% 0.5% 5th 24.4% 24.4% 25.4% 25.4% 37.5% 37.5% 38.6% 38.7% Total 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 35 cate problems in the benefit ad- Table 36 ministration, it could also be an artifact of the very low frequency Proportionofpopulationinconsumptionquintilereceivingpensions of unemployment insurance recipi- ConsumptionExpenditureQuintile ents within the PPV sample. 1st 2nd 3rd 4th 5th AllNEandSE,uncorrected 7.0% 10.5% 12.7% 13.9% 17.9% AllNEandSE,corrected 9.9% 11.0% 12.6% 13.4% 15.7% The coverage of the population Allurban,uncorrected 8.2% 10.4% 13.4% 15.0% 18.3% with pension benefits rises from Allurban,corrected 11.0% 11.4% 13.6% 13.8% 15.9% 8% in the first (corrected) quintile to 16% for the wealthiest quintile (see Table 36). As expected for a Table 37 program whose benefit levels are Distribution of Pensions, by Per Capita Household Consumption Quintile based on previous salaries, the Recipients Amounts All NE and SE All Urban All NE and SE All Urban incidence of the benefit is highly Consumption regressive with more than 50% of Expenditure Quintile Uncorr Corr Uncorr Corr Uncorr Corr Uncorr Corr 1st 9.7% 13.6% 11.7% 15.9% 3.6% 7.4% 4.3% 8.7% the benefit accruing to the wealthi- 2nd 16.0% 16.8% 14.9% 16.2% 6.7% 7.9% 6.2% 8.8% est quintile and only 7% reaching 3rd 19.8% 19.6% 19.8% 20.0% 9.5% 15.3% 10.8% 13.5% 4th 22.7% 22.4% 23.6% 21.8% 15.6% 18.8% 17.5% 20.7% the poorest quintile (see Table 37 5th 31.8% 27.7% 30.0% 26.1% 64.7% 50.5% 61.2% 48.3% and Figure 47). Total 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% The Poverty Targeting of Social Spending in Brazil Page 36 4. Poverty Incidence of Social Spending This section attempts to combine the inci- vide stimulation for further more detailed dence data from the PPV with actual social investigation along the lines proposed rather spending data for Brazil. While the program than be taken as a definitive judgment about incidence analysis and the classification of the incidence of social spending in Brazil. social spending are independently reliable pieces of analysis, the combination of the Social spending data is taken from a series two introduces a series of concerns and re- of studies undertaken by IPEA (Fernandes) quires several quite strong assumptions. As and refer to consolidated spending in 1995, a result, the following analysis should be with some updating to reflect more recent viewed as highly tentative. It should pro- information. Table 38: Summary of Consolidated Social Spending, 1995 with Updates Effective Benefit-Cost Total Budget Budget Budget Total benefits Total benefits Targeting to Ratio Spending Spending per Spending per to poor in cash to poor in kind bottom 20%* (R$bn/a) Total benefit to Current Benefit (R$bn/a) (R$bn/a) Poor to Poor Benchmarks General Investment 3% 1.0 33.3 208.3 Uniform Transfer Payment 20% 0.8 150.2 6.3 6.3 Education Creche 24% 1.0 4.2 26.3 0.00 Kindergarten 42% 1.0 1.1 2.4 14.9 0.45 Basic Education 26% 1.0 11.6 3.8 24.0 3.00 Secondary Education 7% 1.0 1.7 13.5 84.5 0.13 University Education 0% 1.0 5.1 #DIV/0! #DIV/0! 0.00 Adult Education/Training 5% 1.0 0.4 22.2 138.9 0.02 Health Care Universal Public Health Care 16% 1.0 21.8 6.1 38.3 3.55 Urban Investments Water Connection 12% 1.0 1.4 8.3 52.1 0.16 Sewer Connection 4% 1.0 25.0 156.3 0.00 Urban Public Transport 9% 1.0 2.6 11.1 69.4 0.24 Housing (Carta de Credito) 2% 1.0 7.2 50.0 312.5 0.14 Favela Upgrading 34% 1.0 2.9 18.2 0.00 Other Social Investments Microcredit Programs 20% 1.0 5.0 31.3 0.00 Land Reform 70% 1.0 2.0 1.4 8.9 1.40 Pension and Related Programs Pensions 7% 0.9 67.6 15.0 15.0 4.50 BPC (LOAS) 70% 0.9 1.2 1.6 1.6 0.76 Social Assistance Services Old Age Services 50% 0.8 0.0 2.5 2.5 0.01 Disabled Services 50% 0.8 0.1 2.5 2.5 0.02 Child Services (Kindergarten) 42% 0.8 0.2 3.0 3.0 0.07 Subnational Social Assistance Programs 70% 0.8 1.9 1.8 1.8 1.06 Other Transfer Programs Child Labor Eradication 80% 0.8 0.1 1.6 1.6 0.04 Minimum Income Programs (subnational) 70% 0.8 0.0 1.8 1.8 0.01 Nutrition Programs Food Baskets (PRODEA) 80% 0.8 0.2 1.6 1.6 0.10 School Lunches 25% 0.8 0.7 5.0 5.0 0.14 Maternal Nutrition (Milk Programs) 29% 0.8 0.1 4.3 4.3 0.03 Labor Programs Unemployment Insurance 13% 0.8 3.0 9.5 9.5 0.32 Severance Payments (FGTS)** 13% 0.8 9.6 9.6 Abono Salarial 13% 0.8 0.6 9.6 9.6 0.06 Others Drought Workfare (in drought years) 70% 0.9 1.2 1.6 1.6 0.76 Summary/Total 13% 131.8 6.44 10.53 * Targeting numbers in italics are staff estimates, not based on household surveys ** No budgetary spending Sources: IPEA, Fernandez et.al., World Bank staff estimates, includes updates to reflect more recent program changes. The Poverty Targeting of Social Spending in Brazil Page 37 as health care, water, and sanitation, and In order to arrive at crude estimates of bene- those that provide or lead to benefits pre- fit incidence, it is assumed that investment dominantly in cash, which includes transfer programs generally have a benefit-cost ratio programs but also in kind services such as of one, while different transfer programs education whose benefit is a stream of in- have a benefit-cost ratio of 0.8 or 0.9 re- creased cash income. Finally, immediate flecting different levels of administrative benefits (from transfer programs) are sepa- costs. In addition, programs are classified rated from those that accrue as a benefit into those that provide in-kind benefits such stream over an extended period of time Figure 48: Structure and Targeting of Federal Social Spending, 1997 Higher Education 100 80 Pensions and Public Servant Benefits 60 40 Others Secondary Education Urban Transport Labor Other Education 20 Health Housing and Urban Sanitation Nutrition Basic Education Land Reform Social Assistance 0 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Share of Spending Targeted to the First Quintile (Poorest) Shaded The Poverty Targeting of Social Spending in Brazil Page 38 (such as education and other investments). chart therefore represents the share of total federal social spending that accrues to the For social spending items that cannot be bottom quintile. directly related to the analysis of the PPV, assumptions on targeting are made based on Table 38 includes two indicators of targeting comparability with other programs. These effectiveness for each program. The first assumptions introduce additional uncertainty indicator is the budgetary cost per current in the aggregate estimates. benefit to the poor. The second indicator is the budgetary cost of total transfers to the Table 38 summarizes the main items of con- poor. The latter indicator is calculated as solidated social spending in Brazil in 1995. the inverse of the benefit cost ratio times the The overview shows social spending of targeting ratio. For the comparison of budg- approximately R$132 billion, of which R$ etary cost per current transfer, the benefit 68 billion refer to social security and public cost ratio is replaced by the ratio of current service benefits. Of this spending, R$17 benefits to costs (benefits accruing within billion or 13% accrue to the first quintile. one year). As a reference, Table 38 also These benefits to the poor can be divided includes two benchmarks that are hypotheti- into cash (R$6.5 billion) and in-kind bene- cal and do not refer to actual social pro- fits (R$10.5 billion). Benefits include those grams. The first benchmark is general pro- accruing immediately (transfer programs) or ductive investment in the economy generat- over the lifetime of beneficiaries (educa- ing a market rate of return and assuming a tion). Excluding social security and related distribution of returns proportional to Bra- items, social spending amounts to R$64 zil's income distribution. The second billion, of which about R$12.5 billion (or benchmark is a hypothetical universal trans- 19.5%) accrues to the first quintile. fer program that would distribute an equal cash amount to every Brazilian (poor or Figure 48 shows the structure and targeting non-poor) at an administrative cost of 20%. of Federal social spending in 1997. The chart excludes subnational spending, which Figure 49 and Figure 50 show a ranking of is significant especially for education and social programs by their cost effectiveness health care. The full box of the chart repre- in transferring either a current or a total sents the size of federal social spending amount of resources to the bottom quintile. (R$111 billion). Each horizontal slice rep- Theoretically, the ranking of Figure 50 resents a different category of spending, should guide resource allocation if the only ordered by the targeting ratio (share of objective of all these programs was long- spending accruing to the bottom quintile). term poverty reduction. If the objective was For each slice, the shaded area shows the immediate transfer to the poor (in case of an share of spending in a particular category emergency situation), the ranking of Figure that accrues to the bottom quintile of the 49 should guide resource allocation. population. The total shaded area in the The Poverty Targeting of Social Spending in Brazil Page 39 Figure 49 Budget Cost Per Current Benefit to the Poor 100 90 80 Poor 70 the to 60 50 Benefit 40 30 Current 20 per 10 Cost 0 Creche Lunches Reform Salarial Care years) BPCRuralPrograms (LOAS)Pensions ServicesServices Kindergarten Pensions UpgradingEducation Programs Connection TransportEducation Connection Budget Eradication PaymentLand Insurance Investment (PRODEA) Age Abono deCredito) Education (subnational) DisabledOld (Kindergarten)Programs)School GeneralFavela Basic (Milk Transfer HealthWater Public MicrocreditPublic SecondaryEducation/Training SewerGeneral(Carta University ChildFood LaborBaskets(inrelevant Urban Assistance Adult Programs Services Nutrition Uniform Unemployment Housing Workfare Universal SocialIncome Child Maternal Drought Subnational Minimum Source: WB staff estimates Figure 50 Budget Cost Per Total Benefit to the Poor 50 45 Poor 40 the 35 to 30 Benefit 25 20 Total 15 per 10 Cost 5 0 Budget Reform (LOAS) Creche Pensions LunchesCare Salarial Land Upgrading Education Eradication (PRODEA)BPC years)Rural KindergartenServicesServices Age Programs PaymentConnection Insurance TransportEducationPensions Connection InvestmentCredito)Education Basic School Health Abono de LaborBaskets (subnational) Programs DisabledOld Favela (Kindergarten) Programs) Water Public General Public Transfer Education/Training SewerGeneral(Carta University ChildFood (inrelevant (Milk Microcredit Urban Secondary ProgramsAssistance Services Nutrition Unemployment Adult Housing Workfare Child Universal Uniform Income Social Maternal Drought MinimumSubnational Source: WB staff estimates The Poverty Targeting of Social Spending in Brazil Page 40 Figure 51 graphically compares programs programs also have a growth objec- along three dimensions: each bubble repre- tive. sents one spending program; the size of each bubble is proportional to annual per house- b) For several programs, non-monetary hold spending (annualized in the case of benefits for the poor are difficult to investment programs) showing the relative measure. Therefore, the assumed importance of the program to beneficiaries; benefit-cost ratio may well underes- the horizontal position of the bubble shows timate benefits of several programs. the level of targeting of the program to the bottom quintile; the vertical position of the c) Targeting typically refers to average bubble shows the reach (coverage) of the spending in the recent past. New program among the bottom quintile. Pro- and additional spending may, how- grams in the lower left corner are poorly ever, have a different incidence. For targeted and do not reach many of the poor. example, the average targeting of The largest of these are pensions, unem- sewage investments in the past has ployment insurance, sewage provision, and been very regressive. However, as secondary education. Programs in the bot- coverage of the better-off population tom right corner are those well-targeted but increases, additional investment only reaching a small share of the poor may be better targeted. (typically social assistance programs). Pro- grams near the top left are universal pro- Despite the important limitations, the analy- grams, especially water and public health. sis suggests some interesting conjectures: Public pre-primary and primary education is better targeted but reach differs by level. a) As to be expected, social invest- For reference, the impact of distributionally- ments (education, urban services) neutral, annual growth of 4% is shown in the rank poorly in their effectiveness to top left corner. transfer resources in the short term. Counting all benefits through, some The analysis presented in Figure 49, Figure of these investments have benefits 50, and Figure 51 is instructive and permits that exceed those of some transfer the quantitative comparison of a wide range programs. Prioritization depends on of very diverse social programs. However, whether the objective is short-term several limitations need to be considered relief or medium to long term pov- before drawing simplistic and premature erty reduction. policy conclusions from this analysis. These limitations imply that the analysis cannot be b) Many programs are less cost- used as a direct guide to resource allocation effective in transferring total bene- but as a departure point for further in-depth fits to the poor than a hypothetical analysis. uniform transfer payment (many ur- ban investments, secondary, adult, a) The analysis ranks program by their and higher education, social secu- effectiveness to transfer resources to rity, and unemployment insurance, the poor. However, many of the among others). Since these pro- analyzed programs have additional grams do not withstand a simple objectives that need to be consid- cost-effectiveness test from a pure ered in a more comprehensive poverty reduction perspective, they evaluation. For example, programs should be justified on grounds not such as social security and unem- captured in this analysis. ployment insurance have an insur- ance function independent of their c) Figure 51 suggests a trade-off be- social objectives. Many investment tween targeting and reach among The Poverty Targeting of Social Spending in Brazil Page 41 the poor. The more complete the targeting to programs further away reach to the poor, the more difficult from the top left corner, or to re- it is to control leakage. This is the design existing programs such that challenge faced in up-scaling small they move toward the top right cor- and well-targeted social develop- ner, representing better targeting ment programs. The challenge is to and wider reach among the poor. either reallocate funds from pro- grams with inadequate reach and Figure 51 Reach and Targeting of Social Programs 1 General Investment/ Growth of 4% p.a. Public Health Primary Education 0.8 Poor 0.6 Water Merenda Escolar among 0.4 Reach Kindergarten Social Assistance Sewage 0.2 Pensions Unemployment Secondary Education LOAS/BPC Insurance Adult Education Creche 0 0 0.2 0.4 0.6 0.8 1 Targeting to Poor Source: WB staff estimates The Poverty Targeting of Social Spending in Brazil Page 42 5. Conclusions tive than a hypothetical uniform transfer This paper summarizes the coverage and program to all Brazilians (poor or non-poor) targeting of selected social spending pro- in allocating resources to the poor. grams in Brazil and compares the relative effectiveness of different programs in trans- In highlighting programs that are not well- ferring resources to the poor. targeted toward the poor, this paper attempts to provide suggestions regarding a possible The main conclusions of this paper are that reallocation of spending between and within only a relatively small share (13%) of social program areas that would improve social spending (including pensions) reaches the targeting and help alleviate poverty. poor and that many programs are less effec- The Poverty Targeting of Social Spending in Brazil Page 43 6. References Maria Alice da Cunha Fernandes, et al.; 1998a; Dimensionamento e Acompanhamento do Gasto Social Federal; Texto Para Discussão no. 547; IPEA: Brasília. Maria Alice da Cunha Fernandes, et al.; 1998b; Gasto Social Consolidado, IPEA: unpublished. Maria Alice da Cunha Fernandes, et al.; 1998c; Gasto Social do Governo Federal 1994/1997; Nota Técnica no. 001/98; IPEA: Brasília. Francisco H.G. Ferreira, Peter Lanjouw, and Marcelo Neri; 1998; The Urban Poor in Brazil in 1996: A New Poverty Profile Using PPV, PNAD and Census Data; unpublished. Peter Lanjouw and Martin Ravallion; 1998; Benefit Incidence and the Timing of Program Cap- ture; World Bank Policy Research Working Paper no. 1956; World Bank: Washington, D.C.