The Cost of Poverty Alleviation Transfer Programs: A Comparative Analysis of Three Programs in Latin America by Natàlia Caldés David Coady John A. Maluccio September 2005 MIDDLEBURY COLLEGE ECONOMICS DISCUSSION PAPER NO. 05-27 DEPARTMENT OF ECONOMICS MIDDLEBURY COLLEGE MIDDLEBURY, VERMONT 05753 http://www.middlebury.edu/~econ The Cost of Poverty Alleviation Transfer Programs: A Comparative Analysis of Three Programs in Latin America Natàlia Caldés,* David Coady,** and John A. Maluccio*** *Agricultural and Applied Economics Department ETSIA—Universidad Politécnica de Madrid [email protected] **Fiscal Affairs Department International Monetary Fund [email protected] ***Economics Department Middlebury College [email protected] Tel: (802) 443-5941 Fax: (802) 443-2084 September 2005 Forthcoming World Development ii Summary: A common criticism of antipoverty programs is that the high share of administrative (nontransfer) costs substantially reduces their effectiveness. Yet, there is surprisingly little rigorous empirical evidence on program costs. Improved information and a better understanding of the costs of such programs are crucial for effective policymaking. This study proposes and implements a replicable methodology for a comparative cost analysis of three similar poverty alleviation programs in Latin America, and assesses their cost efficiency. The findings underscore that any credible assessment of cost-efficiency requires a detailed analysis of program cost structures that goes well beyond simply providing aggregate cost information. Key Words: cost efficiency, poverty alleviation, human capital, Honduras, Mexico, Nicaragua iii Acknowledgments: This research is based in part on the IFPRI evaluations of PROGRESA (Mexico), PRAF (Honduras), and RPS (Nicaragua). We thank the staff of those programs, particularly Raul Perez, Hadid Vera-Llamas, Monica Orozco, and Daniel Hernandez (PROGRESA), Enrique Álvarez, Silvia Mejía, Manuel Enrique Vela, and Sonia Zuñiga (PRAF), and Leslie Castro, Norlin Sánchez, and Mireille Vijil (RPS) for inputs and support for this analysis. Seminar participants at IFPRI, Rafael Flores, Juan Manuel Medina, and Ferdinando Regalía provided helpful comments, as did two anonymous referees. We gratefully acknowledge funding for this research from the Inter-American Development Bank through the Norwegian Fund for Social Innovation. iv 1. INTRODUCTION It is widely accepted that social safety nets play crucial roles both in alleviating poverty and promoting social and economic development (World Bank 1997). Nevertheless, a common criticism of such programs is that a large proportion of their budget is absorbed by administrative costs and never reaches the intended beneficiaries.1 Depending on how such administrative resources are used, the poverty alleviation effect of the programs and, consequently, their overall cost-effectiveness, may be reduced. Proper assessment of the criticism that such programs are “expensive” is difficult, however, since there is little rigorous empirical evidence on their costs and cost structures.2 For example, in their review of targeted poverty alleviation programs in developing countries, Coady, Grosh, and Hoddinott (2004) find cost information of any sort for only 32 of the 111 programs examined, and most of these were from a single source (Grosh 1994). Moreover, the available cost information is rarely comparable between studies, even for similar programs. Some studies refer to administrative costs, while others consider costs only in terms of theft or other losses and leakages. When the focus is on administrative costs, it is often unclear whether the figures refer to the entire life of the program or only a specific period, such as the most recent year. For programs at different stages of maturity that have high fixed costs or undergo extensive learning-by-doing, analyses based on different time periods can lead to very different conclusions. Improved information and a better understanding of the costs of such programs are crucial for effective policymaking. This study proposes and implements a replicable methodology for a detailed, comparative analysis of the level and structure (the various activities being carried out) of costs for three similar poverty alleviation programs in Latin America. They are the Programa Nacional de 1 Educación, Salud y Alimentación (PROGRESA) in Mexico, the Programa de Asignación Familiar-Fase II (PRAF) in Honduras, and the pilot Red de Protección Social-Fase I (RPS) in Nicaragua. The primary objective of these programs is to generate a sustained decrease in poverty in some of the most disadvantaged regions in their respective countries. The programs’ underlying premise is that a major cause of the intergenerational transmission of poverty is the inability of poor households to invest in the human capital of their children. Supply-side interventions on their own, which increase the availability and quality of education and health services, are often ineffective in resolving this problem. These programs address this problem by targeting transfers to the poorest communities and households and by conditioning the transfers on attendance at school and health clinics. Since the total program budgets are the sum of administrative costs and total (cash and in-kind) transfers, we evaluate the cost-efficiency of each program by considering the cost of making a one-unit transfer to a beneficiary; we refer to this as the “cost-transfer ratio” or CTR (Coady, Perez, and Vera-Llamas 2005).3,4 How we use and interpret the CTR depends on how it is calculated and on program characteristics. Whether the fixed costs of setting up the program or only the variable costs of running it are included and whether the entire life of the program or a specific period is under consideration influence the CTR.5 Features of the program including targeting and monitoring, size, type, and delivery mechanism of the transfers (e.g., cash or in kind, demand- or supply-side), coverage, duration, and whether the program is expanding also matter. In this article, we propose strategies for how cost information can be used to assess the relative cost-efficiencies of the different programs, making clear that understanding design differences across programs is essential for making sensible comparisons, even for similar programs such as the three considered here. 2 While focusing on CTRs would be sufficient to evaluate a program whose sole objective was to disburse transfers, the programs considered in this article have more ambitious goals and specific design features aimed at achieving them. First, transfers are targeted to poor areas and to poor households within those areas. Second, transfers are conditioned on households investing in the nutrition, health, and education of their children. The combination of targeting and conditioning makes these programs operationally and administratively complex, and affects both the level and structure of program costs, as well as program performance. Hence, there is a potential trade-off: reducing the CTR may not be cost-effective if it comes at the expense of activities devoted to administrative tasks such as targeting the poor or monitoring compliance.6 For example, program expenditures arising from setting up and implementing program targeting rules will presumably have a return in terms of improved targeting effectiveness, but while the costs will be included in the CTR, the expanded benefits will not. Similarly, expenditures associated with setting up and implementing mechanisms for monitoring adherence to program requirements will presumably lead to greater effects on human capital, but will only be reflected as a cost in the CTR. Given these programs’ designs and multiple objectives, particularly improved human capital for children that is likely to yield returns over many years, we emphasize that it would be incorrect to interpret the CTR either as a measure of overall cost effectiveness or as a cost-benefit ratio. In Section 4, however, we discuss evidence on the relative targeting effectiveness and human capital impacts of the programs, facilitating a more comprehensive comparison of program costs. 2. DESIGN AND IMPLEMENTATION OF THE PROGRAMS To analyze the cost structures of these programs, it is necessary to understand how they operate and how they have evolved. Table 1 summarizes some basic features of each program. 3 [TABLE 1 ABOUT HERE] (a) Programa Nacional de Educación, Salud y Alimentación (PROGRESA) PROGRESA (Mexico) started in 1997 and was the prototype for the other two programs.7 Its cash transfers have two components. Children over age 7 (the starting age for grade 3) are eligible for education transfers. Transfers increase by grade and are higher for girls than for boys in middle school (grades 7–9). In 1999, monthly benefits were 80 pesos ($8) for grade 3.8 By grade 9, benefits rise to 265 ($27) and 305 ($30) pesos for boys and girls, respectively. In addition to enrollment, transfers are conditioned on an 85 percent attendance record, and children are allowed to repeat a grade, at most, twice. The second component of the transfer, for food security, health, and nutrition, is 125 pesos ($13) per month for each household, conditioned on household members making regular trips to health clinics for preventive health checks, and attending monthly nutrition and hygiene information sessions. The education and food security transfers are independent—beneficiaries can receive one and not the other, even if they are eligible for both. In addition to the cash transfers, beneficiary households with children under age 3 receive a monthly nutritional supplement. There is a ceiling of 750 pesos ($75) per month for education and food transfers combined. On average, the transfer to beneficiary households constitutes around 20 percent of preprogram annual household expenditures. The program design of PROGRESA (as well as of PRAF and RPS) calls for the money to be given to mothers. Transfer amounts are indexed to inflation and adjusted every six months, something not done in the other two programs. PROGRESA was targeted in two stages. The first stage identified the most marginal rural localities, using a “marginality index” constructed from the national census. The selected 4 localities were then visited to ensure they had access to the necessary infrastructure (schools and health clinics). The second stage targeted households within eligible localities, using census data specially collected in program areas to classify households as “poor” or “nonpoor,” based on a statistical analysis of income and other household characteristics. After beneficiary households were identified, a general assembly was held to explain the objectives of the program, incorporate households, and inform them of their responsibilities and rights. The expansion of the program throughout Mexico took place in several phases. The census for the first and second phases began in October 1996. In August 1997, Phase 1 began with incorporation of approximately 140,000 households in 3,369 localities. The first transfers took place in September 1997. Phase 2 began in November 1997, when a further 160,000 households in 2,988 localities were incorporated, with the first transfers taking place in January 1998. Expansion of the program has been determined largely by budget allocations, with the greatest expansion occurring in 1998, when nearly 1.63 million households in 43,485 localities were incorporated. By early 2000, the program had an annual budget of $1 billion and included nearly 2.6 million rural households in 72,345 localities in all 31 states. This constituted approximately 40 percent of all rural households in Mexico. (b) Programa de Asignación Familiar-Fase II (PRAF) PRAF (Phase II, Honduras) began in the second half of 2000 and includes both demand- and supply-side transfers.9 On the demand side, the education subsidy is 812 lempiras (L) ($54) per child per year, up to a maximum of three education transfers per household.10 This transfer is conditioned on the enrollment and regular attendance of all children who have not yet completed grade 4 of primary school. The food security, health, and nutrition transfer provided for pregnant women and children under age 3, is L644 ($43) per beneficiary per year, with a maximum of two 5 transfers per household. This transfer is conditional on pregnant women and children making monthly trips to health clinics for preventive checkups and growth monitoring. Transfers are distributed twice a year and, on average, comprise about 4 percent of preprogram total household annual expenditures (one-fifth of the equivalent percentage of PROGRESA). Unlike PROGRESA, where the supply side is left to the education and health ministries to manage, PRAF directly invests resources to improve supply-side services. For education, it makes grants to school parent associations. For health and nutrition, PRAF operates a community-based child growth and monitoring program that provides mothers with one-on-one counseling, as well as makes grants to local health service committees to improve the quality of health-care provided by the government health system. The program was geographically targeted to poor municipalities, which were chosen by ranking all municipalities according to the average rates of stunting observed in the 1997 National Census of the Height of First-Graders. Seventy municipalities with the highest rates of stunting were considered eligible (MNPTSG 2002). Of these, 50 were randomly selected, leaving the others as a control group for the program evaluation. In 40 of the chosen municipalities, all households with pregnant women, children under age 3, and/or children aged 6–12 who had not yet completed grade 4 of primary school were eligible for benefits (the remaining 10 municipalities selected received only the supply-side transfers described below). Transfers began in November 2000 and, by the end of 2002, PRAF had 47,800 beneficiaries and was operating in 50 rural municipalities (out of a total of 298) from seven departments. Eighty- seven percent of the households in these departments were classified as poor. 6