33340 Targeted Conditional Transfer Programs in Latin America: An Early Survey Guilherme Sedlacek, Nadeem Ilahi & Emily Gustafsson-Wright THE WORLD BANK JUNE 11, 2000 Paper prepared for the Regional Study: Securing our Future Office of the Chief Economist, Latin America and Caribbean Region. The World Bank EXECUTIVE SUMMARY This paper assesses the design and performance of Targeted Conditional Transfer (TCT) programs for human development in Latin America. These programs mainly came about in the early-mid 1990s in the form of Bolsa Escola in urban Brazil, and subsequently Progresa in Mexico and PETI in rural Brazil. Other programs such as the PRAF-BID II in Honduras, Red in Nicaragua and Beca Escolar in Ecuador are either in the pilot or design phase. A number of other countries are also contemplating putting in place their own version of TCT programs. The Bolsa Escola programs are administratively decentralized, i.e. they are installed and run by local states and municipalities in Brazil, whereas all the others (Progresa, PETI, PRAF-BID II, Beca Escolar and Red) are, or will be, run by the federal governments. The TCT programs give cash grants to poor families with young children on the condition that they visit health centers and/or keep their children in school. The underlying concept of the programs is similar--i.e. cash grants are given to the poor conditional on behavior which leads to improvements in their children's human capital. The good behavior varies from program to program. In the Brazilian programs it entails the families keeping their children in school. In Mexico and Honduras it also entails regular visits to health clinics. While in the PETI program, it requires that children regularly attend an after-school program. These programs have essentially five objectives. First, they hope to increase educational attainment and/or improve health outcomes and thus reduce future poverty. Second, by restricting the grants to the current poor and by improving their health, the programs aim to reduce current poverty as well. Third, by requiring children in beneficiary households to have minimum attendance in school, the programs aim to reduce child labor. Fourth, an implicit objective is that by providing income support to poor families, they act as a partial safety net--i.e., they prevent these families from falling into poverty in the event of an adverse shock. Last, some of these programs also provide supply side financial support to schools and health facilities. TCT programs are right in focusing on human development among the poor because the poor in Latin America fare much worse than non-poor on nutrition, health and education outcomes. The poor in Latin America have lower incomes than the non-poor, but more importantly, they also have significantly worse nutrition, health and education compared to the non poor. This is partly because they have worse access to services than the non- poor. In addition, poverty and child labor are closely associated and working children do much worse in school than their counterparts who do not work. These programs hope to decrease the likelihood of future poverty among the children of the currently poor by increasing their human capital. TCT programs are more likely to be successful in reducing short term and long term poverty than earlier approaches in the region. These programs explicitly target the poor. Thus, by design, they have an edge over earlier approaches to poverty reduction and 1 social protection (such as school feeding programs and other untargeted subsidies) that do not. At the same time, by focusing on the state of the future poor (children) they complement other programs for poverty reduction that focus more on current poverty (such as provision of basic services). The programs have been able to generate strong commitment within and outside the governments. First, federal and local governments have begun to see these programs as important element of poverty reduction strategy. There has been sizable government support, as seen by increasing amounts of financial resources allocated to the programs. Second, the local schools and health centers are important stakeholders. Various agents of civil society are actively involved in monitoring and enforcement of these programs. This promises better success for these programs than other social protection programs. The programs have sound criteria for selection and targeting. The federally administered (i.e. centralized) programs use geographical targeting at the national level. Progresa chooses the poorest localities in the country first while PETI chooses the localities with the highest incidence of worst forms of child labor. They then use means tests (or some variant of it) to select beneficiaries within these localities. The decentralized programs (such as Bolsa Escola in Brazil) are unable to target according to national priorities because they are local initiatives, however, within localities they do target the poorest neighborhoods first. They also use variants of means test to select beneficiaries. Preliminary evidence suggests that the programs have been reasonably successful in targeting, though some aspects need attention. Evidence from Brazil and Mexico on program targeting (based on comparisons of beneficiary and control populations) suggests these programs target well. Leakage rates (i.e., the non-deserving benefiting from the program) have been low. However, under-coverage rates (the deserving populations that is missed by the program) have been high. In decentralized programs such as Bolsa Escola, this is partly due to a lack of finances; some municipalities have had to ration coverage within the qualifying group to the poorest. The programs have improved educational indicators and outcomes. In the Bolsa Escola Programs in Brazil, a crude comparison of beneficiary and non-beneficiary families reveals school attendance is higher and dropout rates lower for the former. More beneficiary children enter school at the right age compared to their non-beneficiary counterparts. Mexico's Progresa has had a significant impact on education. A systematic comparison of program and control groups reveals that the program improved primary school attendance by 2.2% from a high base of 92%. The gains in secondary enrollment were even more significant--it increased by 8.4% from a base of 65%. There is evidence of improvements in access and usage of facilities in areas where programs address health and nutrition. The evidence on health outcomes is still not available, but crude data on access and usage of services suggests the programs have had a positive impact. The number of health visits increased by 20% in areas where Progresa was present, but only 5% in areas where it was not. For families with children under 5 it increased by 30%, and for those with pregnant women by 16%. On the nutrition end 2 there is evidence that beneficiary families increased their expenditure on some important food items by one-third more than non-beneficiary ones. The evidence on reduction of child labor is inconclusive with the exception of the PETI program in Brazil. The evidence on impacts of Progresa on child labor reveals mixed results. There was a 25% greater reduction in child labor in 12-13 year olds in the beneficiary compared to the non-beneficiary groups. However, except for the 14-15 age group, the difference in child labor reduction for all other age groups is not statistically significant. It has been difficult to ascertain the effects on child labor in the Bolsa Escola, partly because in Brasilia, where the evaluation was undertaken, there is little child labor. Target evaluations of the PETI program, which targets children in the worse- forms of child labor, reveals extremely low leakage rates and rate of excellent impact. Excessively stringent criteria for selection of families can lead to exclusion of the needy. The Brazilian programs use two conditions. First because some of the programs explicitly focus on poor families with children in school age, they exclude poor families with only pre-school and adolescent children. Second, they impose a minimum residency requirement on beneficiaries, ostensibly to discourage in-migration. By doing so, however, they can end up excluding needy groups, which raises concerns about equity. The view of this report is that for the sake of the latter group the residency requirement should be reduced, say, to one year. For the former groups, this report suggests two options. Either other programs, such as those for pre-school children and adolescents, could be run in tandem with the Bolsa Escola Programs. Or, the role of the Programs could be expanded along the lines of the Progresa program in Mexico to include poor families with only pre-school and adolescent children and catering to the health and nutrition needs of these groups. An issue that would require careful attention is that any expansion should ensure the programs still stay rooted in current administrative set up such as pre-schools, so that compliance with the program (minimum attendance) is monitored. The level of the cash transfer needs to be determined carefully. Three issues related to the cash grant are important. First, it is unclear from looking at the levels of the grants what the grant is meant to do--does it compensate beneficiary families for lost wages, or does it pull them up to the poverty line? Since the primary objective of the program are reduction in child labor and improvement in schooling outcomes, the recommend benefit level should compensate the families for the foregone earnings of each participating child. Further optimization would require rigorous field evaluations and a greater clarity of program objectives (improvement in minimum consumption or caloric intake levels). Second, the programs in Brazil currently use a flat grant that does not vary by the characteristics of the recipient. There may be gains from tailoring it to subgroups in the population. For instance, older children are likely to have a higher opportunity cost of going to school and it may be better to have the Bolsa increase with age (as it does in Progresa). Third, some of the Bolsa Escola Programs give one grant per family rather than per child. This is likely to reduce the incentive of families with many children from participating in the program. We suggest that the size of the grant should be based on the number of children in the family. While attempts to tailor the Bolsa will be efficient, 3 especially if they are based on empirical evidence, it is also likely they will add to the cost of administration, since it is expensive to identify and monitor different subgroups. Thus we recommend that an exercise to tailor the grant should be undertaken keeping in mind the administrative benefits and costs, as well as institutional capacity. Governments should not see the Programs as substitutes for investments in schools. The poor educational and health attainment are a consequence of both demand failure (poverty) and supply constraints (services' quality poor or lack of it). The quality of access to health and education is likely to be an important constraint, especially in poor regions. Thus governments should not see these programs as substitutes for other education-related interventions. This report recommends that programs without supply features (such as the Bolsa Escola) should follow the example of other programs in Brazil (PETI) and in the region (Progresa in Mexico) that provide explicit financial support to guarantee minimum service quality standards in local schools and health clinics. These programs are only partial safety nets--they protect the structural poor during crises and equip the next generation with risk reducing human capital. There are two ways in which these programs serve the purpose of a safety net. First, these programs directly transfer resources to poor families. Recent economic crises have seriously affected the health, nutrition and educational outcomes of the extreme poor. Thus by supporting the structural poor, these programs provide them with a safety net. Second, education and health investments reduce future vulnerability to economic shocks. Thus by encouraging human capital investments these programs allow a better brace against future shocks as well. Perhaps the biggest hindrance to the successful implementation and growth of the decentralized programs is their fiscal affordability. Not only are programs such as the Bolsa Escola implemented at the municipal level, but also their financing also has to come from local sources. This has created a particularly perverse problem in that the poorest municipalities are the ones that need these programs the most (since they have a high incidence of poverty) but are unable to afford them (since they also have low local revenues). The calculations in this report reveal that except for the richest urban municipalities, the program would amount to a fairly large burden on local finances. There is evidence even today that in some of the poorer municipalities where the programs are being implemented, shortages of funds have resulted in under-coverage of the deserving population. There are two ways of resolving the financing problem. One is to alter the selection criteria or the size of the grant to reduce the size of the beneficiaries or the amount of support respectively. The other is to top-off local deficit through financing from the federal government. This report recommends the latter because even though there may be some gains in adjusting the grant or the size of the beneficiary population, there is a limit to this strategy. In the extreme, such adjustments would start affecting the effectiveness of these programs in reaching their objectives. Federal transfers are perhaps the best option. However we strongly recommend that to preserve the incentives of local municipalities to efficiently run the programs there should be a local share that each municipality must pay. 4 1. INTRODUCTION The past three decades have seen a gradual shift in social policy in Latin America in two phases. First, many countries in the region have steadily moved from using general subsidies (especially for food and fuel) as the major instrument of support of the poor, to income transfer programs targeted to the poor. This shift came about largely for two reasons. First, countries faced administrative difficulties in keeping subsidies focused on the poor. Second, it also became clear that such subsidies were fiscally unaffordable. Thus income transfers directed to the poor only were seen as a cost-effective way of reducing poverty. The second shift came when countries moved towards targeted income transfer programs that require good behavior from beneficiaries. The fact that the poor in LAC have dismally low human development, as measured by education and health indicators became the rationale for the second shift. Making income transfers to the poor conditional on children's school attendance or pregnant mother's visits to health facilities was seen as an effective way to help the current poor while attacking the root causes of long term poverty. This paper provides an early survey of these targeted conditional transfer (TCT) programs for human development in LAC. Broadly speaking, these programs have five objectives. 1 First, by providing cash grants to families that are poor now, the programs hope to directly reduce the incidence of current poverty. Second, they aim to reduce long term poverty by tying these cash transfers to education and health attainment by children and pregnant mothers.2 Third, the reduction of child labor is an explicit objective in some of the programs (PETI, Beca Escolar) and implicit in others (Progresa, PRAF-BID II, Bolsa Escola).3 Fourth, these programs act as a safety net, by preventing the welfare of the current poor from falling below a survival threshold in the event of a negative shock.4 Last, most of the programs recognize explicitly that mere demand side support (i.e. giving cash grants for education and health) may not be sufficient. This is especially the case in poorer regions where there are a lot of potential beneficiaries but the supply of services (education, health) is deficient. Supply side support is an important part of Progresa in Mexico and PRAF- BID II in Honduras. For detail on objectives of specific programs see table 1.1. 1 It is important to note that these objectives are not listed here in order of priority; in fact different programs reviewed in this paper emphasize different objectives. 2 Not all programs address all aspects of human development. For instance the Bolsa Escola programs in Brazil focus largely on education. 3 In programs where the reduction of child labor is an explicit objective, the programs target regions where child labor is pervasive (PETI) or are installed in times when child labor is likely to be high such as in a crisis (Beca Escolar). In all other programs, the implicit objective is preventive in nature such that by creating incentives for families to put their children in school, the programs seek to create an indirect incentive against child labor. 4Because of their design, most of the programs do not cover the current non-poor who could fall in poverty after a negative shock. Only the Beca Escolar in Ecuador will be installed as a part of the coping strategy (i.e. after the shock) of the government in response to the macroeconomic downturn. 5 2. HISTORY OF THE PROGRAMS While TCT programs all share a common structure, they have evolved in different circumstances. Here we briefly describe the evolution of each of the programs surveyed in this paper. Since the early 1970's, Brazilian government and civil society have been discussing minimum income programs as a means to alleviate poverty. Bolsa Escola Programs In the mid 90s Bolsa Escola programs were first introduced in Campinas and Brasilia. Because Brazil is highly decentralized, various permutations of the original Bolsa Escola have been subsequently implemented in various municipalities and states. By 1998, there were more than 60 programs in operation covering a total of 200,000 families. Table 2.1 shows the extent of coverage of the programs in three metropolitan areas for which data is available. Even though the various Bolsa Escola programs have similar structure, they do differ on the margins as far as their objectives are concerned. For instance the original program in Campinas, was meant to be part of a larger social assistance network. On the other hand, the Brasilia program was largely seen as an educational program. The Bolsa Escola Programs are mainly urban programs. PETI In the past few years, the Brazilian government has pushed for the prohibition of dangerous and unhealthy forms of child labor (for children under age 18) and any type of labor for younger children (those below age 14)--except for professional training. The PETI program was introduced by the federal government with the aim of reducing the incidence of child labor in rural Brazil through education related grants. PETI was introduced in 1996, and by 1999 it was covering over 130,000 children. The program is expected to increase coverage in the next phase to 800,000 children (see Table 2.2). PETI is primarily a rural program. Progresa Part of the rationale and financing for Progresa in Mexico, came out of the phasing out of the tortilla subsidy. The Mexican government explicitly acknowledged that the subsidy had failed in reaching the poor and that a more targeted approach to poverty reduction was warranted. The federal government in 1997 initiated Progresa and funds used for the tortilla subsidy were redirected to Progresa. Progresa's expansion has been quite spectacular. By 1999, the program was operating in 31 states and covering 2.3 million families. See Table 2.3 for details. Progresa operates largely in rural areas in Mexico. 6 PRAF-BID The original PRAF program in Honduras was a poorly targeted and implemented. With the assistance of IADB the Government of Honduras has restructured and redesigned it in line with the lessons learned from the Progresa in Mexico. The new program incorporates excellent monitoring and evaluation components that aim to inform how to balance program design between investment in the supply of services versus incentives for service utilization (cash grants). The program is now targeted to the poorest 80 municipalities of Honduras and it is expected to become the cornerstone of poverty reduction and social protection in the country. Beca Escolar The Beca Escolar program in Ecuador was proposed to provide a coping response to the economic crisis facing Ecuador. It is expected to be an improvement on the earlier approaches to social protection in Ecuador, namely Bono de Solidaridad which was poorly targeted. It is yet to be implemented. Red de Protección Social The Red de Protección Social program in Nicaragua has been proposed as a means to attain a balance between immediate protection of nutrition and the long term accumulation of human capital through investment in primary education and preventative health. The Red is meant to fill the void that the other programs (Fondo Social Suplementario- FSS, Fondo e Inversión Social Emergencia ­FISE, MECD, and MINSA) leave in terms of health and education protection for the poorest.5 This program is also in advanced stages of development. 5Nicaragua: Red de Protección Social, IDB 7 3. RATIONALE There are three features of these programs that provide a rationale for their introduction and continued implementation. First, their focus on health and education of poor families is well justified given the current state of poverty in LAC. Second, the wedding of "supply" and "demand" features in some of the programs is particularly attractive. Third, because of their support to children in poor families these programs enjoy widespread support in countries, which makes them more sustainable than other programs. Human development and the poor The poor in LAC fare particularly badly when it comes to indicators of education, health and nutrition. Here we consider education and health indicators by poverty status in selected countries for which data are available. Table 3.1 shows grade completion by wealth status. Differences in grade completion between the poor (bottom 40%) and rich (top 20%) are much smaller at the lower grades than at the high grades. Except for Guatemala, more than 90% of poor and non-poor children have completed grade 1. However, the differences between rich and poor in primary school and grade 9 completion are stark. For instance in Brazil, only 8% of the poor and 38% of the non poor children of grade 9 age or above have completed grade 9. Part of the reason why older children in poor families have low completion of primary and higher grades is the high opportunity cost of time (see the discussion on child labor below). These data suggest that poverty-targeted interventions in education are likely to improve completion of grades 5 and higher. They also underscore the need for increasing the size of the cash transfer by age. In health and nutrition, poverty affects outcomes as well as access. Figures 3.1 and 3.2 show the distribution of infant mortality, child stunting by wealth quintiles. Annex 1, Figure 3.3 shows the distribution of children who are underweight by wealth quintiles. All three indicators reveal a clear association with wealth status. The poorest have the worst outcomes. The differential between rich and poor are two to four times for infant mortality, four to nine times for stunting and four to ten times for the proportion of children who are underweight. Part of the reason for low outcomes of the poor are their lack of access to services. Figures 3.4 and 3.5 show the distribution of antenatal care visits and child immunization by wealth quintiles. Again, the access of the poorest section of the population is far worse than the rest. The picture of access to health care by wealth status reveals that a demand side components that requires mandatory level of usage of health facilities is likely to be an effective way of improving the health and nutrition outcomes of the worst off in society. For its level of economic development, LAC has high incidence of child labor. According to the ILO, the overall incidence of child labor in LAC is about 10%.6 An in-depth four- 6 The numbers are sensitive to a) the age group of children and b) the nature of work considered. We use children in the 10-14 age group and take an inclusive definition of child labor--that where both work in income generating activities and in household chores is considered as child labor. 8 country study on child labor reveals poverty and child labor are intimately related.7 The proportion of children working falls monotonically with the quintile rank of their family (see table 3.2).8 The intensity of labor is high for working children. Table 3.3 reveals that almost one half of the working children in Brazil and Ecuador work more than five hours per day. On the surface, child labor does not seem to discourage school attendance. Of the children that work more than two-thirds are in school (Table 3.4). However, it does slow progression through school. Annex 1, Table 3.5 shows that compared to being in school only, school and work and work only increasingly reduce the grade attainment of children. These data suggest that programs that require children from poor families to have a minimum level of attendance during a month are likely to improve their educational attainment. Combining demand and supply Without the availability of adequate and reasonable quality services (schools, health clinics etc.) demand side interventions (i.e. conditional cash transfers to the poor) are not likely to succeed. This is partly because even if we are able to get children from poor families to attend schools (or get pregnant mothers to go to health clinics), the ultimate improvements in human capital will not take place if supply of health and education services is constrained or is of poor quality. At the same time, however, we should be aware that good quality supply is necessary but not sufficient for attaining positive human development outcomes. Thus the capacity to meet the increase in demand from these demand side programs matters. The existing structure of the TCT programs in LAC shows that program designers have done well to incorporate supply features where they are needed and to leave them out where they are not. On the one hand, Progresa, aware that it targets the most backward areas in Mexico where the quality of education and health services is likely to be poor, provides complementary support for schools and clinics in the form of additional funding. For PRAF-BID, this support is not tied, i.e. it does not dictate in which budget headings local authorities ought to spend the moneys. PETI pays a transfer to the local schools for every child in the program. Most of this transfer is intended to install the Jornada Ampliada, or the extended school day. On the other hand, where the quality of local services has been deemed adequate, programs have not provided financial support for improving supply. Most of the Bolsa Escola programs in Brazil do not provide help to the existing school systems because they are aware that the quality of local schools in urban areas (where these programs operate) is adequate enough to meet the increased demand for schooling that results from 7 The results here are based on Ilahi, Sedlacek and Sasaki (2000). 8 This is somewhat different from the "mixed" result that is obtained when child labor is regressed on income or consumption. In this case the evidence on the relationship between child labor and welfare is mixed. The endogenous nature of consumption or income may be responsible for biasing the estimated coefficients. It may be better (as reported in the main text) to look at the incidence of child labor by quintile rank. 9 the program. However, any future expansion of the Bolsa Escola Programs to poorer municipalities would require supply side support to schools also. Political goodwill augurs well for long term sustainability The existing programs are the outcome of a vibrant political and technical debate that took place in Mexico and Brazil during the 1990s. These programs count on the broad based political and technical support within the countries and are part of an emerging consensus in these countries on the need to prioritize support for poverty alleviation programs. As proof of this consensus all these programs have been financed entirely through local budget allocations (i.e. without foreign aid or donations) and the international financial institutions have been involved in only a supporting role. 10 4. EFFECTIVENESS Targeting the Poor The main objective of the programs is to support the current poor, Thus, how well these programs target the poor is critical in determining their success. Targeting depends both on design and implementation. Here we discuss separately the design features as well as the outcomes. The programs usually target in two phases. These are discussed below and specific details about each program's design are addressed in table 4.1. Targeting design: selection of geographical areas of intervention Since poverty targeting is an important criterion of these programs, most target the poor using country-level information. The federally administered programs (i.e. centralized programs) in our set follow this approach. Based on poverty maps constructed from household survey or census data, the first stage of targeting essentially entails picking the poorest regions in the country. Most of the programs reviewed here follow this procedure, though with some variation (see Table 4.1 for details).9 There are two exceptions to this targeting. First, because the Bolsa Escola programs are decentralized programs, initiated and administered by states and municipalities, they do not necessarily operate in the poorest urban regions of the country. Where in the country these programs are installed is largely a function of local desires and capacity (both financial and administrative) to have such a program rather than due to some nationwide poverty priorities. However, it is worth noting that even though the choice of the municipalities themselves is not based on any larger (Brazil-wide) notion of geographical targeting, the programs have good targeting within the chosen regions. The programs achieve this by starting with the poorest neighborhoods within a municipality. Second, since the primary objective of the PETI in Brazil is the eradication of rural child labor, the program chooses geographical regions by the incidence of child labor not poverty. It chooses localities in two steps. First it creates a profile of agricultural activities that have the worst forms of child labor. Then it selects the municipalities where these agricultural activities take place. Note that the municipalities ultimately chosen may or may not be the same as the poorest localities in Brazil. Targeting design: selection of beneficiaries The programs follow a wide variety of approaches in the selection of beneficiaries. Here we describe the approach in each one by one. In the earlier versions of the Bolsa Escola programs, an income-means test was used to identify the beneficiary group. A per capita income of one half of the minimum wage 9 See Sedlacek, Ilahi and Gustaffson-Wright (2000) for details. 11 was used as the cut-off point for the poverty line. The programs then further reduced the subset of the poor by two additional requirements--that families have children ages 7-14 and in some cases have a minimum amount of residency in the area. An income means test proved inadequate since a significant portion of family income is typically derived from informal occupations and unregistered sources. As an alternative, a Score system was developed. The scoring, which is demonstrated in Annex 2, Chart 1, considers various aspects of the potential beneficiaries' living standards--e.g. housing status, engagement in the labor market, educational attainment and number of children in the household. The weight of each characteristic differs by region. In the PETI program, a simple means test is used to select beneficiaries. Families qualify for the program according to the following criteria: a) the per-capita family income is less than ½ minimum wage, and (b) families have school age children (7-14).10 In the Progresa program in Mexico, there are three steps in the selection of household within the chosen localities. Progresa first uses income information to classify households into poor and non poor. The poverty line it uses is based on a per capita income of 320 pesos per month (or the cost of a standard food basket--canasta basic). Second, it performs discriminant analysis to incorporate household characteristics into a measure of poverty. An index (puntaje) is constructed based on the discriminant analysis and this index is then used to refine who is poor and non-poor. The underlying reason for combining traditional measures of poverty with those based on household characteristics is to incorporate multi-dimensional aspects of poverty into the selection process. In the current phase of the PRAF-BID II in Honduras, there is to be no second stage targeting. So essentially all the households in the chosen poorest municipalities will receive the cash transfers. This is partly based on the view that there is little variation in incomes within the poorest municipalities. The Beca Escolar (Ecuador) program intends to target poor families with children in the 6-15 age bracket. Those in the bottom 20% would be taken as the poor and a means test would be used to determine if a household qualifies. Targeting outcomes: success in targeting Only two of the programs are advanced enough to be judged for targeting--Progresa and Bolsa Escola Programs. Progresa has been evaluated formally, while the evidence on Bolsa Escola is from selected surveys in the program regions. In the absence of data on experiment and control groups, only indirect evidence on targeting is available for the Bolsa Escola Programs in Brasilia. The evidence suggests 10 The more recent program manuals suggest an additional requirement may be used in the future: that only families in which children are engaged in child labor be selected. If this condition for selection into the program is implemented then it is likely to create perverse incentives in that families that want to benefit from the program may put their children in work. 12 that the program has been well-targeted (Rocha and Saboia, 1998), meaning that the actual beneficiaries of the program were the intended beneficiaries based on the selection criteria established by the program. By comparing the characteristics of both beneficiaries and non-beneficiaries it is clear that the program did capture poor families with children of school-going age. A reason for this success is because the program uses a score system that allows the program to select beneficiaries by "scoring" their observed characteristics and then using a cut-off point to choose the most needy. Another reason is the manner in which the program has been implemented. The program was implemented in Brasilia gradually by first starting from the poorest neighborhoods and then expanding to less and less poor neighborhoods. This method of expansion uses a mix of geographical targeting (neighborhoods) and within a geographic area, targeting based on quasi-means test (household score). While many of the municipalities have been successful in reaching the potential beneficiaries, on average the programs have not been successful in reaching all potential beneficiaries. Although the program demonstrated appropriate targeting, as described above, the programs on average were not able to serve nearly 43% of the parameter population (the population established by the selection criteria). The reason why some programs had less that 100% coverage was financial--the program costs with 100% coverage may just have been too high. The evidence on how well Progresa has targeted the poor is also preliminary at this stage. Skoufias et al (1998)--the IFPRI team responsible for evaluating Progresa--find Progresa's targeting methodology sound. Using data from two of the regions served by Progresa, they also evaluate Progresa's targeting on the ground. They merge program data with household survey data from Mexico to evaluate targeting. They start with the household characteristics that Progresa used (based on its own discriminant analysis) to choose beneficiaries. This is essentially a proxy means test on permanent income. Skoufias et al (1998) get households with similar characteristics from a separate household survey (the ENCASEH) and check where these households fall in the income distribution (i.e. whether they are poor or not). Their finding is that the undercoverage rate (households that were poor but were not chosen) is quite low--about 14-16%. On the other hand, the leakage rate (i.e. households that were non-poor but were chosen) is rather high--in the range of 35%-38%. Table 4.2 shows the 2-by-2 targeting matrix for two regions where Progresa operates. Increasing Human Capital Educational attainment There are number of indications of successful outcomes in the Bolsa Escola program in Brasilia. First, a comparison of program data (experiment) and school census data (control group) reveals that drop out rates are much lower among the former (0.4% in 1996) than among the latter (5.6%; see table 4.3).11 Second, a larger proportion of the 11Note that the difference in dropout rate between beneficiaries and non-beneficiaries obtained in this way provides a upper bound on the actual difference since the former group consists of children from poor 13 children in beneficiary households enters the school system at the right age than do their non-beneficiary counterparts. Third, children in beneficiary households do exhibit a higher promotion rate (80%) than their counterparts in non-beneficiary households (72%) (Table 4.4). Finally, there was little difference between the beneficiary and non- beneficiary groups in learning outcomes (see table 4.5). It should, however, be pointed out that a systematic evaluation that controls for the fact that beneficiaries are not a random sub-group of the population may give better indications of program success. Two preliminary evaluations of Progresa have been performed at this point. Unlike in the case of Bolsa Escola, there is sufficient pre and post program data available to undertake a controlled assessment. Schultz (1999) conducts two tests to see if Progresa has made a significant difference in enrollment rates (see table 4.6 below). First is a crude regional comparison of pre and post program differences in enrollment rates between program and control regions. This comparison indicates that primary school enrollment in Progresa regions was 1.1% higher than in control regions (from a high base of 92% enrollment). Secondary enrollment (where a lot more slack existed) was 4.9% higher from a base of 65%. Second is a refined comparison based on individual data. It controls for community, household and school characteristics, but essentially tests the same differences. Here Progresa adds 2.2% more to the primary enrollment and 8.4% more to secondary enrollment compared to individuals who did not benefit. Schultz (1999) suggests that the refined estimates are more reliable than the crude ones because they control for other influences on enrollment rates that may not be captured adequately by the difference in outcomes between program and control regions/individuals. Regardless, there is clear indication that the program has had significant impacts in improving school enrollment, especially at the secondary level. Health and Nutrition attainment Bolsa Escola Programs do not have a health component. Preliminary results of the evaluation of the impact of Progresa on health by Gertler (2000) finds Progresa has a positive and significant effect on both access and outcomes. At the community level, visits to clinics in Progresa localities was 18% higher than to clinics in non-Progresa areas. At household level, utilization levels were 19% higher amongst poor individuals in treatment localities when compared to poor individuals in control localities. Utilization was higher among males aged 6-15 and 50 plus and females above the age of 15.12 Progresa also had a positive and significant effect on the number of prenatal care visits, increasing the number of daily visits by about 5 percent. The number of women making their first visit in their first trimester of pregnancy increased by about 8%, with concomitant reductions in the number of first visits at later stages of pregnancy. Participation in Progresa, lowered the probability of illness by 22% amongst children families who are more likely to drop out compared to a randomly chosen child. In other words the true effect of the program would be obtained when a randomly selected child is given the cash transfer and their drop out rate compared with that of the rest of the population. 12 Use of private health providers fell in treatment communities, suggesting that a good portion of the increase in utilization of public facilities reflects a substitution away from other providers. 14 aged 0-2. However, there was no evidence of impact on older children and no impact on duration of illness. The short-term nutrition impacts of Progresa were measured by a percentage change in the food expenditures of beneficiary families relative to non-beneficiary families (see Table 4.7). There was a substantial relative increase (by about one-third) among the beneficiaries in the consumption of bread, milk and cheese. Reducing Child Labor Since the reduction of child labor is not an explicit objective in the two programs that have already been evaluated, it is difficult to ascertain whether they have been successful in achieving this purpose. The indirect evidence from Bolsa Escola suggests that the programs may not have had a significant impact on child labor. This is partly due to the fact that the incidence of child labor was low in Brasilia before the start of the program. The evidence on the impacts of Progresa on child labor reveals mixed results (Progresa, 1999). A difference-in-difference comparison of 12-13 year olds in beneficiary and control (potential beneficiary) groups revealed a 25% reduction in the incidence of paid work. Except for the 14-15 age group, the differences for all the other age groups were not statistically significant. For the 14-15 year olds, Progresa seemed to have significantly increased the incidence of paid child labor. 15 5. OUTSTANDING ISSUES TCT programs as safety nets While the provision of safety nets is not an explicit objective of these programs, we consider how well they might serve this purpose. Education and health as risk-reducing assets During economic crises in Brazil, the least educated experience the greatest vulnerability (Thomas, 1999). Further, when a crisis hits, one of the coping strategies of the current poor is to pull their children out of school and put them in work to supplement family income (see Duryea, 1998; Ilahi 1999). Another could be reducing health expenditures (World Bank, 1999). While such strategies do reduce some of the short-term costs of a crisis, they create long term vulnerability. This is because they reduce the educational attainment and human development of children, who then are more likely to be vulnerable to economic risks in the future. Hence, by providing cash transfers to the poor conditional on school attendance, nutritional intake and health clinic attendance, these programs increase educational attainment among future generations and hence reduce future risks. Transitory versus structural poor Safety nets address vulnerability to risk. There are two types of vulnerable--those who are below the poverty line--the structural poor--and those who are above it, but can fall below it when a shock hits--the transitory poor. The TCT programs target the current poor (i.e. in normal times these would also be the structural poor). The current poor are particularly vulnerable to both economic and idiosyncratic shocks. Existing evidence from the region confirms this. In Ecuador, the recent economic crisis has severely affected health, nutrition and education outcomes of the extreme poor (World Bank, 1999). In Brazil, recent episodes of economic downturn and upswings have particularly exposed the poor to even worse outcomes (Thomas, 1999). Thus, the fact that these programs promise to continue to provide transfers to the current poor means they mitigate the risks the poor face. In their current design, these programs do not cover the transitory poor. This is because the transitory poor are the current non-poor and therefore theoretically, invisible to these programs. Is it possible to modify the programs to make them into better safety nets, or at least improve their "coping" qualities? There are a number of concerns that argue against tinkering with program design and parameters to make them into safety nets. First, the selection criteria for identifying the poor would have to be adjusted in order to track current income better. These changes could weaken transparency (and with it the governance of these programs because currently those who participate in the program are easily identified as poor through their permanent income). They would also add to complexity of managing these programs. Second, it is now clear that social sector 16 expenditures are pro-cyclical in LAC (see Hicks and Wodon, 2000, for details). This means that in the event of a macroeconomic downturn, there is a pressure on all social protection programs to cut budgets. In this sense, when expecting the programs reviewed in this paper to expand coverage to the "new" poor, care should be taken not to induce them to cut the transfers to the persistent (and in a sense more deserving) poor. Unless the total envelope allocated to these programs in increased during crises, expanding the coverage of the programs to the transitory poor could possibly result in a worse outcome in that the transfers to the persistent poor--who may have fallen further into poverty-- may have to be reduced. Finally, it is also important that policy makers who redesign these programs to make them better safety nets should also consider the "graduation" of families who experience improvements in welfare in up turns. When the economic cycle turns up, there will be many families that will experience an increase in income because of increases in employment and/or wages. It is imperative that such families are struck from program rolls. This will serve two purposes. First it will ensure that these programs stay trim and do not expand excessively in good times. Second it will keep them well targeted (by reducing leakages to the non-poor) and therefore maintain their political credibility. The "excluded" groups With the exception of Progresa and PRAF-BID II, which have health components, the education-only programs target poor families with children of school going age. They therefore overlook two types of poor families--those without children and those without children of school-going age (i.e. very young children and older children). Such criteria can create a potential problem since poor families without children in the 7-14 age group may also be deserving, especially those with very young (preschool) children. Further, poor families with adolescent children may also require support, as the adolescents may not yet be of the right age to work. In the case of Bolsa Escola, Table 5.1 below shows that the non-covered population is small in one dimension, but not in an another. First, families without children are not a significant portion of the poor, so the requirement that only poor families with children qualify does not exclude many. The second requirement, however, does. Of the poor families that have children in the 0-6 age bracket, as many as 57% do not have children in the 7-14 age group, implying they will not benefit at all from the program. The need for complementary programs for the excluded There is a need to expand the programs to poor families with adolescent children. Progresa already provides support for secondary education and new variants of the Bolsa Escola also plan to target the 15-17 year olds. For those with very young children, however, it may be best to leave these groups out of the current TCT programs that do not already include them. There are essentially two reasons for this. First, these programs are rooted in the school system and so the compliance of children of school going age is easy to monitor. Adding preschool children would require other ways of 17 monitoring the good behavior and this may either be unfeasible of too costly. Second, very young children have specific needs such as post-natal care and daycare which are best addressed by a separate program (see Pães de Barros and Mendonça, 1999). What may be best for this group is to have separate program fashioned along the lines of Bolsa Escola. An issue related to complementary programs for the "non-covered" is what to do with those who graduate from school. Since 14 year olds are unlikely to be ready to enter the labor force, there is a need for post school training programs to ease their entry into the work force. Some of the programs associated with PETI in Brazil are attempting to deal with this issue by having special training for these groups. The optimal size of the cash grant There are two issues related to the optimal level of the cash grant. First, is at what level it should be set, given program objectives. Given the fact that these programs have multiple objectives it may be difficult to come up with a precise level of the cash grant. In other words it is not obvious theoretically whether the program should pay a cash grant that brings the families up to the poverty line, or one that compensates the families for the opportunity cost of participating in the program (i.e. the opportunity cost of child time and the time cost of visiting the health center etc.). The answer is likely to depend on the priorities of program objectives. If the primary objective of the program is to reduce poverty then the grant should be adequate to bring the family up to the poverty line. If the primary objective is to have the children in school then compensating the families for the opportunity cost of child time as well as other indirect school costs may be best. Thus while it is not clear ex ante what the grant level ought to be, a program-by-program approach can yield the most appropriate level of the transfer. However, it should be noted that we might expect that households value the education and health services being offered, consequently the optimal level cash grants levels should necessarily be below the opportunity cost of the child times. Tailoring the grant to subgroups Two additional issues are whether the timing or the level of cash transfer should be changed according to the sub groups in the beneficiary population. This is because not everybody who participates in the program is identical and therefore, there may be gains in efficiency to be had if the program altered the timing or level of the grant to account for this heterogeneity of participants. There are two examples that illustrate this. First, Progresa, changes the level of the grant according to the school grade and gender of the child. This is based on the explicit understanding that families with older children face a higher opportunity cost than those with younger children so the grant for children in secondary school would have to be higher than for those in primary school and provides an incentive for families to keep their children in school. Also, the female-male differential in enrollment in Mexico widens at the secondary school level so a greater incentive is given to the girls to attend. Second, the Brasilia version of Bolsa Escola deposits the equivalent of one minimum wage in the savings account in the name of the beneficiary child each year. The child can withdraw the money once he/she completes 8th 18 grade. The idea here is to alter the timing of the grant to create an incentive to complete secondary school. While adjusting the cash grant level to tailor it to the various sub groups in the population makes economic sense (from an efficiency standpoint), it also adds to the cost of administering the program, since additional criteria have to be checked when making the cash transfer to the family or the banking system has to be involved. If these marginal administrative costs are not large then it would make sense to tailor the grant to the different subgroups in the program. However, if these costs are large then it may be better to leave the cash grant scheme simple. Financial constraints in decentralized programs In decentralized settings such as Brazil, the affordability of the Bolsa Escola programs has to considered alongside regional inequalities in poverty. The program is least affordable in the areas where they are needed the most. Table 5.2 shows that program costs are a much larger share of current receipts in Salvador, Fortaleza, Belém and Recife than in Brasilia and São Paulo. There are two reasons why. First poverty is much higher in the Northeast, so the pool of families that satisfy selection criteria is much larger than in the Center South. Second, the cities of Northeast are poorer than the ones in the Center South, so their fiscal revenues are much smaller than those of their counterparts in the Center South. At the cash grant level of one minimum wage the program is clearly unaffordable in almost all the cities except a few. Even after bringing the cash grant level to 1/5 of the minimum wage, program costs remain excessively high (4% in Salvador and 3.3% in Fortaleza). But cutting the size of the Bolsa runs the risk of jeopardizing the effectiveness of the program in inducing child enrollment and reducing poverty. The optimal size of the transfer should be set taking into account the parameters that maximize program success. Any gain from trimming the Bolsa has to be weighed against the cost of not meeting the goals of the program. The role for the federal government in decentralized programs If local municipalities tinker with the design of the program (i.e. changing the level of the Bolsa or the changing the poverty line) this would make the program more affordable, but it would also defeat its key purpose--i.e. poverty reduction and human development. Thus keeping the current targeting and transfer design of the program intact requires that the federal government "top off" with transfers to the poorest municipalities. The question is under what formula should these transfers from the federal government come? A feasible option is that the federal government can equalize the load of the program on the municipalities. So every municipality pays up to a maximum local share (say 2%), and the federal government takes up the remainder. This formula would mean that there would be no federal transfer to municipalities where program costs are less than the maximum local share. On the other hand, where program costs are very high (e.g. Salvador), the federal government would pay most of the costs. Since the poorer municipalities are also likely to need help with their school system, some of the transfer from the federal government can come in the form of "in kind" transfers for improving 19 the school system. Regardless, it is important to note that in order to maintain the incentives of the municipalities, the federal government should not pay the full costs. Program success and capacity First, Targeted Conditional Transfer programs require a fairly sophisticated understanding of what problem they are trying to address and therefore achieve. For instance, there is a need to clearly prioritize (as has been done in most of the programs reviewed here) the program objectives. This requires careful study of the existence of poverty maps, profiles, indicators of human development and the intersection of poverty and human development. Second, once the objectives have been clarified, the administrative capacity to implement and monitor a demand side program needs to be adequate. Prior experience in administering targeted programs may prove quite helpful in designing and implementing the targeting mechanism for a demand side program well. However, existing evidence on the quality of targeting in the region suggests that earlier programs (food basket programs, school breakfast programs etc.) have been poorly targeted. Therefore there is a need to pay special attention to this issue to ensure good targeting. The existing demand side programs reviewed in this paper have made a conscious attempt to target well. In Brazil, the existing household survey data (PNAD) have been used extensively to construct the right program parameters (proportion of minimum wages as the poverty line, the requirement of having school age children etc.). Current surveys are underway to set up a experiment-control groups so that not only targeting but also the effects of program on outcomes is evaluated in a systematic way. Progresa has a federal office that has been involved with international agencies (IFPRI and IADB) in setting up program design. They are also using a high quality team to evaluate the targeting and outcomes of the program. Even though PRAF-BID II is in a pilot phase right now, it is building on existing survey data from Honduras to construct a poverty map for first stage geographic targeting. It is also using the data to simulate the need for doing a means test to target. The program is designed to make the evaluation as accurate as possible. Hence, there are to be three types of "experiment" groups in addition to the control group (that did not receive any intervention). The three experiment groups are: 1) those that receive demand side intervention only; 2) those that receive a supply side intervention only and 3) those that receive both. Such a design will allow program evaluators to not only assess outcomes in a systematic way, but also answer the more important question of whether we need supply side interventions, demand-side interventions, both or neither. Integrating civil society The extent to which civil society and stakeholders participate and enforce good implementation is crucial for program success. In Brazil, the NGOs participate in the after-school program training teachers, developing instructional materials, etc. In the sate of Pernambuco, for example, there are four main NGOs which are involved in the PETI program: SERTA (Servicio de Tecnología Alternativa) works together with four municipalities in the training of monitors for the after-school program with the focus of 20 local sustainable development. Two other NGOs; CIELAS (Centro Universitario de Estudos da America Latina, Africa e Asia) and Pro-Cidadanía concentrate on the training of monitors with a focus on pedagogy, the reality of rural children, and training of teachers to compliment the work of the after-school monitors. Finally, the NGO Oficina do Saber, which is financed by the ILO (International Labor Organization), works together with families and educators. There are several other NGOs that provide professional courses for the graduates of PETI. In Mexico, Progresa puts heavy emphasis on the role of parents in program implementation. It promotes active participation and co-responsibility of parents. It also attempts to make its operations transparent by involving local communities through regular community meetings. A specific objective of these meetings is to select a voluntary community promoter who monitors the adequate use of cash grants by the beneficiary families (see Progresa, 1999). 21 POLICY CONCLUSIONS Our review of the programs gives rise to two types of policy concerns. One set relates only to the decentralized programs while the other relates to all types. The policy related issues that apply only to the decentralized programs are as follows. First, we caution that program effectiveness may be seriously affected if local municipalities adjust the size of the grant or the selection criteria to make the programs affordable. Second, the administering municipalities and the federal and state governments need to agree on cost- sharing for the programs. Third, there is a need to either expand the existing programs to include the "non-covered" populations or serve them with other complementary programs. A number of policy concerns are common to both decentralized and federal programs. First PETI and Progresa explicitly support schools, but Bolsa Escola Programs do not. This paper recommends that Bolsa Escola Programs should complement investments in school quality, especially when they are expanded to poorer regions. Second, the decision of whom to expand the programs to needs to be taken carefully. In the centralized programs, the programs should determine empirically, where the marginal peso should be spent--on the near poor family in existing areas or to poor families in near poor areas. Such a decision must also keep administrative costs of the latter in mind. In decentralized programs, the issue of nationwide targeting is crucial. The federal government needs to make its own calculations to prioritize geographical areas where interventions are needed most, so as to remain consistent with the objective of reaching the most needy in the country. Third, the level of the cash grant needs to be determined in a clear and systematic manner, keeping in mind the primary objective of the program in question. Fourth, it is better to base the means tests used in the programs on consumption rather than income measures. Fifth, there is a need to explore further how the role of these programs can be modified to make them better safety nets--i.e. improve their ability to protect from economic shocks. Sixth, there is a need to carry out a broad- based and systematic evaluation of these programs. A formal evaluation, such as the one for Progresa in Mexico should be conducted for the Brazilian programs also. In the absence of such an evaluation, it would be difficult to learn from the success and failures of such programs. Seventh, these programs are only a partial safety net. They cover the structural poor but not the transitory ones. There is thus a need to explore how their design could be altered to accommodate the latter group. Last, TCT programs require fairly sophisticated administrative capacity. Countries that contemplate installing these types of programs should pay attention to building their capacity to implement them. The conclusion of this paper is that the existing TCT programs are appropriately designed and well-administered. They have an important role to play in the larger country social assistance and poverty reduction strategies. They are likely to remain successful because they enjoy broad support in government as well as in civil society. However, the concerns raised in this paper should be kept in mind when replication or expansion of these programs is being considered. 22 References Duryea, Suzanne, 1998, "Children's Advancement Trough School in Brazil: The Role of Transitory Shocks to Household Income," Inter-American Development Bank, Washington. Filmer, D. and L. Pritchett, 1998, "The Effect of Household Wealth on Educational Attainment: Demographic and Health Survey Evidence," Policy Research Working Paper #1980, The World Bank. Gertler, Paul, 2000, "A Preliminary Evaluation of the Impact of Progresa on Health and Health Utilization," International Food Policy Research Institute, Washington. Preliminary Gwatkin, D., S. Rutstein, K. Johnson, R. Pande, and A. Wagstaff, 2000, "Socio-Economic Differences in Health, Nutrition, and Population," HNP/Poverty Thematic Group, The World Bank. Hicks N., and Q. Wodon, 2000, "Economic Shocks, Safety Nets, and Fiscal Constraints: Social Protection for the Poor in Latin America," in XII Seminario Regional de Politica Fiscal: Compendio de Documentos 2000, CEPAL, United Nations, Santiago, Chile, 381-407. Ilahi, Nadeem, 1999, "Children's Work and Schooling under Shocks: Does Gender Matter? Evidence from the Peru LSMS Panel Data," Background paper for Engendering Development, The World Bank, Washington. Ilahi, Nadeem, Guilherme Sedlacek and Masaru Sasaki, 2000, "Child labor, Poverty and Education: Empirical Evidence from some Latin American Countries," Human Development Network, Latin America and Caribbean Region, The World Bank, Washington., Mimeograph Paes de Barros, Ricardo and Rosane Mendonça, 1999, "Uma Avaliação dos Custos e dos Beneficios da Educação Pré-escolar," IPEA, Rio de Janeiro. Mimeograph. Pan American Health Organization, 1998, "Basic Indicators," PAHO-WHO. Progresa, 1999, "Evaluación de Resultados del Programa de Educación, Salud y Alimentación," Secretaría de Desarrollo Social, Government of Mexico. Rocha, Sonia, 1998, "Minimum Income Programs ­How Do They Apply to the Brazilian Nuclei?" Human Development Network, Latin America and Caribbean Region, The World Bank, Washington. Mimeograph. Rocha, Sonia and João Sabóia, 1999, "Programas de Renda Mínima Linhas Gerais de uma Metodologia de Avaliação a partir da Experiência Pioneira do Paranoá no Distrito Federal," IPEA, Rio de Janeiro. 23 Sant'Ana, Silvio Rocha, Moraes, Andréa, 1997, "Avaliação do Programa Bolsa Escola do GDF," Fundação Grupo Esquel, Brasil. Schultz, T. Paul, 1999. Preliminary Evidence of the Impact of Progresa on School Enrollments from 1997 and 1998. International Food Policy Research Institute Washington. Preliminary. Sedlacek, Guilherme, Nadeem Ilahi and Emily Gustafsson-Wright, 2000, "Brazil: An Assessment of the Bolsa Escola Programs," Human Development Department, Latin America and Caribbean Region, The World bank, Washington. Skoufias, Emmanuel, Benjamin. Davis, and Jere Behrman, 1999, "An Evaluation of the Selection of Beneficiary Households in the Education, Health and Nutrition Program (Progresa) of Mexico," International Food Policy Research Institute, Washington. Preliminary. Thomas, Mark, R., 1999, "Growth and Recessions: An Episodic Analysis of their Effect on Employment Income and Poverty in Metropolitan Brazil since 1982," Working Paper, Brazil Country Management Unit, The World Bank, Washington, Mimeograph. Waiselfisz, Julio J., Miriam Abramovay and Carla Andrade, 1998, "Bolsa Escola ­ Melhoria Educacional e Redução da Pobreza," Projeto Conjunto: UNESCO, UNICEF, POLIS. Brasilia. World Bank, 1999, "Ecuador: Human Capital Protection Project," Project Appraisal Document, Human Development Network, Latin America and Caribbean Region, The World Bank, Washington. 24 Table 1.1: Objectives of Programs Bolsa Escola PETI Progresa PRAF-BID II Beca Escolar Red de (Brazil) (Brazil) (Mexico) (Honduras) (Ecuador) Protección Social (Nicaragua) Poverty Reduction ü ü ü ü ü Human Development Education ü ü ü ü ü ü Health x x ü ü ? ü Nutrition Some1 x ü ü ? ü Child Labor Reduction ü (implicit)2 ü (explicit)1 x ü (implicit) ü (explicit) ü (implicit) Safety Net ü (implicit) ü (implicit) ? ? ü (explicit) ? Supply Side Support x ü ü ü ? ü 1Some of the Bolsa Escola Programs offer nutritional support also. 2The key activities identified as "worst forms of child labor" are differentiated by urban and rural areas. In urban areas, these are illicit activities such as drug trafficking, prostitution, or other harmful activities such as street vending. In rural areas, they include work in the collection or production of charcoal, agave, cotton, vegetable products, sugar cane, tobacco, horticultural products, citrus, salt, flour, fish, wood, textiles, tiles or ceramics and activities related to the extraction of stones and gems--mining. 25 Table 2.1: Bolsa Escola, Brazil: Coverage in three Metropolitan Areas Number of families Areas Below ? minimum Receiving benefit** wage* Belem 44,762 4,423 (June 1998) Belo Horizonte 51,730 3,300 (planned Dec./98) Brasilia 59,926 25,312 (Jan. 1999) * Estimates obtained from PNAD-1996 (Special Tabulations by Sonia Rocha) ** Information from local programs Table 2.2: PETI, Brazil: Evolution of Coverage Year Number of States Number of Municipalities Number of Children 1996 2 17 3,710 1997 3 48 37,025 1998 7 140 117,200 1999 12 166 130,963 Source: Ministry of Social Security and Social Assistance Table 2.3: Progresa Program: Evolution of Coverage Year Number of States Number of Municipalities Number of Families (thousands) 1997 13 466 404.2 1998 30 1,681 1,505.7 1999 31 1,984 2,301.2 Source: Progresa (1999) 26 Table 3.1: Education Grade Completed by Wealth Status in Selected Countries Brazil Colombia Peru Guatemala Dominican Republic (1996) (1995) (1996) (1995) (1996) Bottom Top Bottom Top Bottom Top Bottom Top Bottom Top 40% 20% 40% 20% 40% 20% 40% 20% 40% 20% Grade 1 completed (%) 92 99 94 99 95 99 68 96 87 99 Grade 5 completed (%) 46 90 63 94 75 96 24 87 57 92 Primary completed (%) 15 58 63 94 50 93 18 84 47 88 Grade 9 completed (%) 8 38 15 58 18 61 2 51 14 64 Source: Filmer and Pritchett (1998) Table 3.2 Incidence of Child Labor by Income/Consumption Quintile Brazil Ecuador Nicaragua Peru Bottom 20% 20% 49% 16% 42% 2nd quintile 13% 40% 14% 31% 3rd quintile 11% 31% 9% 28% 4th quintile 8% 25% 7% 18% Top 20% 5% 16% 7% 17% Source: Own estimates Table 3.3 The Distribution of Hours Worked by Children (10-14 years old) Brazil Ecuador Nicaragua Peru Less than 5 51% 52% 50% 85% 5 through 10 43% 40% 38% 13% More than 10 6% 8% 11% 2% Total 100% 100% 100% 100% Source: Own estimates 27 Table 3.4 School Attendance and Work by Children (10-14 years old) Brazil Ecuador Nicaragua Peru School only 80% 61% 78% 68% School and work 11% 26% 7% 27% Work only 3% 8% 4% 3% Neither 6% 5% 11% 2% Total 100% 100% 100% 100% Source: Own estimates Table 3.5 Grade Attainment of a Typical 12 Year Old Brazil Nicaragua Peru School only 3.5 3.9 4.3 School and work 2.8 3.3 3.8 Work only 1.8 1.3 3.4 Neither 1.7 1.4 4.0 Source: Own estimates Note: Based on the assumption that children start school at age 6 28 Table 4.1: Targeting Design and Target Population Bolsa Escola PETI Progresa PRAF-BID II Beca Escolar Red de (Brazil) (Brazil) (México) (Honduras) (Ecuador) Protección Social First stage (selection of geographical 6 4 4 4 4 areas) Second stage (selection of beneficiaries) 4£ 4£ 4£ 6 4£ Non poor families 6 6 6 4¢ 6 Poor families With pregnant 6 6 4 6 mothers ? With preschool- 6 6 4 children ? ? With school-age 4¥ 4¥ 4§ 4© children 4? With post school-age 4® 6 6 6 children ? Based on a national poverty map. Based on a national map of incidence of child labor. £ Means test/score. ¢ The non poor in the chosen municipalities ¥ 7-14 year olds § 6-12 year olds © 6-15 year olds ® A special program is being proposed for the 15-17 year old group. 29 Table 4.2: Progresa Targeting Outcomes; Regions 3 and 28 Region 3 Permanent income (per capita) criterion for selecting households Non poor Poor Total Household's status in 1174 101 (U=12%) 1275 Progresa: 1021 254 (U=14%) 1275 600 677 (U=22%) 1275 Excluded from Progresa 1626 (L=69%) 747 2373 839 (L=35%) 1534 2373 0 (L=0%) 2373 2373 Included in Progresa 2800 (75%) 848 (25%) 3648 (100%) 1860 (50%) 1788 (50%) 3648 (100%) 600 (16%) 3050 (84%) 3648 (100%) Region 28 Permanent income (per capita) criterion for selecting households Non poor Poor Total Household's status in 1521 156 (U=14%) 1677 Progresa: 1308 369 (U=16%) 1677 749 929 (U=22%) 1677 Excluded from Progresa 2262 (L=14%) 952 3214 1225 (L=16%) 1989 3214 0 (L=0%) 3214 3214 Included in Progresa 3783 (75%) 1108 (25%) 4891 (100%) 2533 (50%) 2358 (50%) 4891 (100%) 749 (15.3%) 4143 (85%) 4891 (100%) Notes: 1. The top # in each cell is obtained using the 25th percentile of log per capita permanent income as the poverty line. 2. The second # in each cell is obtained using the 50th percentile of log of per capita permanent income as the poverty line. 3. The third # in each cell is obtained using the log of the value of the standard food basket (canasta básica) as the poverty line Source: Skoufias et al (1998). Table 4.3 Dropout Rates Among Beneficiaries and Non-Beneficiaries (%) Bolsa Escola, Brazil Year Beneficiaries Non-Beneficiaries Total 1994 --- --- 6.2 1995 0.2 6.5 6.0 1996 0.4 5.6 5.2 Source: School Scholarship Program (1966) and School Census (1995) as cited by Sant'Ana and Moraes (1997) 30 Table 4.4 Promotion Rates Estimates 1995-1996 (%) Bolsa Escola, Brazil Year Beneficiaries Non-Beneficiaries Total (a) (b) (a) (b) (a) (b) 1995 79.8 67.0 79.5 70.9 79.6 70.8 1996 87.9 80.2 79.9 72.2 80.4 73.5 (a) Estimates by Sant'Ana and Moraes (1967) including the Literacy Cycle where promotion is automatic (b) Estimates which consider the 3rd. to the 8th grades only. Table 4.5 Learning Outcomes from the SAEB Sample of Fourth Graders (1997) Bolsa Escola, Brazil Subject Beneficiaries Non-Beneficiaries Number % correct Number % correct Portuguese 143 38.7 374 37.4 Mathematics 159 37.8 377 37.5 Science 162 42.4 377 43.5 Source: Waiselfisz et al (1998), p. 134 Table 4.6 The Impacts of Progresa on Primary and Secondary School Enrollment, Mexico Current enrollment Crude Refined rate difference-in- difference-in- difference§ difference£ Primary 92% 1.1% 2.2% (grades 3-6) Secondary 65% 4.9% 8.4% (grades 7-9) §Difference in post program enrollment rate between program and control regions minus the difference in pre program enrollment rates between program and control regions. £Based on estimated probit coefficient of program×eligible interaction term. The probit regressions also control for community, school and family characteristics. Source: Schultz (1999) Table 4.7 Percentage Change in Expenditures of Beneficiary Families Relative to Non-Beneficiary Poor Families Type of Food Percentage Increase Fruits and Vegetables 19.3 Milk and Cheese 33.8 Meat 24.2 Bread 32.0 Source: Progresa (1999) 31 Table 5.1: Number of Poor Families According to the Age of Children; Urban Brazil Nuclei Without With Children in 0-6 age group Children All Without children 7-14 year olds Belém 2,182 24,122 15,405 Fortaleza 6,404 69,227 35,467 Recife 7,278 36,384 21,204 Salvador 12,846 64,210 38,730 Belo Horizonte 3,297 31,447 13,949 Rio de Janeiro 17,862 47,450 24,561 São Paulo 25,066 34,913 25,064 Curitiba 1,062 8,483 4,243 Porto Alegre 4,357 18,270 10,589 Brasília 1,808 13,115 8,368 Total 82,162 347,621 197,580 Table 5.2 Estimated Program-Expenditures as a Proportion of Current Local Receipts Bolsa Escola, Brazil Urban area Number of % of Current Local Receipt Families (000) Cash grant of one Cash grant of ½ Cash grant of 1/5 minimum wage minimum wage minimum wage Belém 22.5 12.75 6.38 2.55 Fortaleza 61.7 16.71 8.35 3.34 Recife 40.3 12.27 6.13 2.45 Salvador 61.5 20.92 10.46 4.18 Belo Horizonte 29.9 4.10 2.05 0.82 Rio de Janeiro 54.1 3.04 1.52 0.61 São Paulo 59.0 1.44 0.72 0.29 Curitiba 7.2 1.74 0.87 0.35 Porto Alegre 13.3 3.21 1.61 0.64 Brasilia 32.3 1.23 0.62 0.25 Total 382.1 3.40 1.70 0.68 Source: IBGE/PNAD 1996, Special Tabulations by Sonia Rocha 32 Figure 3.1 Infant Mortality Rates by Quintiles; Selected Countries in Latin America, 1996 90.0 80.0 70.0 age) 60.0 months 12 births 50.0 (under live 1000 40.0 children per of 30.0 Deaths 20.0 10.0 0.0 Lowest Second Middle Fourth Highest Wealth Quintiles Brazil Peru Colombia Nicaragua Source: Gwatkin and others (2000) 33 Figure3.2 ChildStuntingbyQuintiles; SelectedCountriesinLatinAmerica,1996 50 45 40 are 35 who height-for-age age) 30 of average years5 below 25 (under 20 deviations children of % standard 15 two 10 5 0 Lowest Second Middle Fourth Highest WealthQuintiles Brazil Peru Colombia Nicaragua Source: Gwatkin and others (2000) 34 Figure 3.3 Children under 5 who are Undreweight; Selected Countries in Latin America, 1996 20 18 16 14 12 deviations weight-for-age 10 standard two average are who below 8 % 6 4 2 0 Lowest Second Middle Fourth Highest Wealth Quintiles Brazil Peru Colombia Nicaragua Source: Gwatkin and others (2000) 35 Figure 3.4 Antenatal Care Visits by Mothers by Quintiles; Selected Countries in Latin America, 1996 100.0 90.0 80.0 delivery to 70.0 prior visited 60.0 who 50.0 professional mothers of 40.0 % trained 30.0 medicallya 20.0 10.0 0.0 Lowest Second Middle Fourth Highest Wealth Quintiles Brazil Peru Colombia Nicaragua Source: Gwatkin and others (2000) 36 Figure 3.5 Immunization Coverage of Children, by Quintiles; Selected Countries in Latin America, 1996 90.0 80.0 70.0 60.0 immunized 1-2) Measles 50.0 and (ages 40.0 DPT3 for Children 30.0 of % 20.0 10.0 0.0 Lowest Second Middle Fourth Highest Wealth Quintiles Brazil Peru Colombia Nicaragua Source: Gwatkin and others (2000) 37