Report No. 42093-PE Peru Social Safety Nets in Peru September 25, 2007 Bolivia, Ecuador, Peru and Venezuela Country Management Unit Human Development Department Latin America and the Caribbean Region Document of the World Bank PROMARN Food and Nutrition Program for Abandoned Minors at Risk o f Malnutrition (Programa de Alimentaci6n y Nutrici6n del Menor en Estado de Abandon0 y Riesgo Nutricional) PRONAA National FoodAssistance Program (Programa Nacional de Asistencia Alimentaria) PRONA- MACHCS National Project for the Management o f Water Sheds (Proyecto Nacional de Mantenimiento de Cuencas Hidrogrhficas y Conservaci6n de Suelos) SIAF IntegratedFinancial Management System (Sistema Integrado de Administracibn Financiera) SIME IntegratedMonitoring and Evaluation System (Sistema Integrado de Monitoreo e Evaluaci6n) SIS-SMI Integral Mother and Child Maternal Insurances (Seguro Integral de Salud) SISVAN Information System for Food and Nutrition monitoring in Peru (Sistema de Informaci6n para la Vigilancia Alimentaria y Nutricional del Peni) SSN Social Safety Net UBN Unsatisfied Basic Needs UNDP UnitedNations Development Programme UNICEF UnitedNations Children's Fund WDR World Development Report UIT UnitTaxation based(UnidadImpositivaTributaria) VdL Vas0 de Leche, food assistance program Vice President: Pamela Cox CountryDirector: Carlos Felipe Jaramillo Sector Director: Evangeline Javier Sector Manager: Helena Ribe Country Sector Leader: Daniel Cotlear Task Team Leaders: Cornelia Tesliuc and IanWalker 1 CONTENTS SUMMARY ............................................................................................................................................................... VI 1 THE PATTERNOF EXCLUSIONAND VULNERABILITY INPERU .................................................... 1 1.1 WHYDO SOCIETIES NEED SOCIAL SAFETYNETS? ........................................................................................ 1 1.2 THERESILIENCE OF POVERTYANDINEQUALITY INPERU ............................................................................ 2 1.3 WHONEEDS PROGRAMS? SSN A PROFILE OFVULNERABLE GROUPSINPERU ............................................ 4 2 THE SOCIAL SAFETY NET INPERU ........................................................................................................ 10 2.1 MAINCOMPONENTS THE PERUVIAN SOCIAL OF SAFETY NET ................................................................... 10 2.2 PUBLIC SPENDING ON THE SOCIAL SAFETY NET ........................................................................................ 14 2.3 VASODELECHE ........................................................................................................................................ 20 2.4 WORKFAREPROGRAMS ............................................................................................................................. 22 2.5 FONCODES ............................................................................................................................................. 24 2.6 THEJUNTOSCCT PROGRAM ..................................................................................................................... 27 3 THE REDISTRIBUTIVE AND POVERTY IMPACTOF PERU'S FOODPROGRAMS ...................... 37 3.1 CRITERIA FORASSESSING REDISTRIBUTIVE THE ROLEOFFOOD PROGRAMS............................................ 37 3.2 FOOD PROGRAMS COVERAGE AND BENEFIT LEVELS .................................................................................. 38 3.3 FOODPROGRAMS'TARGETINGPERFORMANCE ........................................................................................ 40 3.4 FOODPROGRAMS'IMPACT MALNUTRITION ON POVERTY AND ................................................................. 42 4 THE WAY FORWARD .................................................................................................................................. 44 4.1 KEYISSUES FACING PERU'SSSNSYSTEM ....................................................................................................................... ................................................................................................. 44 4.2 RECOMMENDEDSTRATEGY 47 BIBLIOGRAPHY ..................................................................................................................................................... 51 ANNEX 1 PERU'S FOODTRANSFER PROGRAMSIN2006 . .......................................................................... 53 ANNEX 2: ESTIMATING THE INCIDENCEOFPUBLIC SPENDINGFORFOOD PROGRAMS ............54 ANNEX 3: ALGORITMO PARA EL CALCULO DE LA PROBABILIDAD DE POBREZA ......................... 56 ANNEX 4: FINDINGSAND LESSONSFROMTHE EVALUATIONS OFPERU'S FOODPROGRAMS ..59 ANNEX 5: IMPROVED PROXY MEANS TEST MODEL: AN ILLUSTRATION .......................................... 61 ... 111 FIGURES Figure 1-1: Malnutrition by expenditurequintile. 1996and 2001 3 Figure 1-2: SchoolEnrolment by Age and Quintile, 2003/04 ...................................................................... ............................................................... 6 Figure 1-3: Prevalenceof ChildLabor, Peruand LAC ................................................................................ -6 Figure 2-1: Recent trends in SSN spendingin realterms ........................................................................... 15 Figure 2-2: Public Spendingon Social Sectors inthe LAC Region ........................................................... 15 Figure 2-3 Peru's SSN Spendingas % of GDP, with regional and global comparators ............................. 16 Figure 2-4: Modeled impact on the poverty gap of increasedtargeting of SSN spendingtoward extreme poor Figure 2-5: Distribution of Vusode leche beneficiariesby quintile and area ofresidence...................... ............................................................................................................................................................. 17 Figure 2-6: Probability of receiving Vusode Leche by age and area of residence.................................... -21 Figure 2-7: Frequencyof delivery of Vusode Leche benefits inrural and urbanareas............................. 21 22 Figure 2-8: Trends inthe compositionof VdL beneficiaries, 2001-2003 .................................................. Figure 2-9: Distribution of beneficiariesofA Trubujur Urbunoby consumptionquintile, 2003 Figure 2-10: FONCODES's geographic targetingprogressiveness, comparedwith other social funds.....23 ...............22 Figure 2-11: The distribution of Central Governmenttransfers to sub-nationalgovernments,2006 Figure 2-12: FONCODES investment inreal terms ............................................................................................25 26 27 Figure 2-13:Juntos targetingprocess: geographic selection, householdassessment, and community validation .................................................................................................................................................... -28 Figure 2-15: Reduction inJuntos' Inclusion andExclusion Error with an ImprovedPredictionModel....29 Figure 2-14: Departmentalpoverty and extremepoverty rates, 2006 ......................................................... 33 Figure 2-16: Juntos' Simulated Impact onthe ExtremePoverty Gap ......................................................... 34 Figure 2-17: Juntos' Simulated Impact on ReducingChronic Malnutrition ............................................... 36 Figure 3-1: Coverage and benefit levels of Peru's SSN Programs comparedwith LAC ............................ 38 Figure 3-2: Targeting performanceof selected LatinAmerican food programs Figure 3-3: Targeting Performanceof selected Cash Transfer programsand Peru's food programs.........41 ......................................... 42 Figure 3-4: The SimulatedImpact of SSN Programson Poverty reduction ............................................... 43 TABLES Table 1-1:Poverty and extremepoverty inPeru. 1997-2006 ....................................................................... 2 Table 1-2: CircumstancesAggravating Povertyand Leading Indicators of Deprivationby MainAge Group and Quintile ........................................................................................................................................ 7 Table 1-3: CircumstancesAggravating Poverty andLeading Indicators of Deprivation by Main Age Group and Quintile ........................................................................................................................................ 8 Table 1-4: Incidenceof ExtremePoverty ..................................................................................................... 9 Table 1-5: Distribution of ExtremePoor by Type of Household ............................................................................................... .................................................................. 9 Table 2-1: SocialProtection Spending2006 12 Table 2-2: Public Spendingon Social Sectors, 1997-2007(% of GDP) .................................................... 14 Table 2-3: Fiscal Position and SocialAssistanceSpendinginLatin America, 2000-2001 16 Table 2-4: Juntos Conditionalities .............................................................................................................. ....................... 31 Table 2-5: Juntos PMT Algorithm: Inclusion andExclusion Errors ........................................................... 32 Table 2-6: Juntos' SimulatedImpact on PovertyHeadcount, Gap and Severity ........................................ 35 Table 2-7: Chronic malnutrition rate, children zero to five ........................................................................ 35 Table 3-1: Household-level coverageof Peru's food programs ................................................................. 39 Table 3-2: Foodprograms' coverageoftheir declaredtarget populations ................................................. 40 Table 3-3: Incidenceof Peru's food transfers on consumption ................................................................... 40 Table 3-4: Targeting accuracy: Share of public transfers reachingeach quintile ....................................... 41 iV Acknowledgements This report was prepared by Cornelia Mihaela Tesliuc and Ian Walker (LCSHS) with the help of a team o f consultants consisting of Lorena Alcazar and Rodrigo Lovat6n (decentralization of basic social infrastructure and community kitchen programs), Pedro Francke (decentralization of social programs), Anja Linder (review of social protection programs), and Juvenal Diaz (benefit incidence of social safety net programs). Pablo Lavado, Carmen Osorio and Andrea Portugal (JPAs, World Bank, Lima) provided invaluable assistance inconductingresearch and analysis. The Propuesta Ciudadana group undertook a qualitative data research that contributed in-depth information on empowerment, participation, and service delivery. The team would like to thank the representatives of community based organizations and local authorities in 18 provincial and district municipalities inthe departments of San Martin, Cusco, and Piura for their effort and contribution during interviews and focus group discussions. The work benefited from valuable comments, suggestions and peer review by Margaret Grosh, Laura Rawlings and Edmundo Murrugarra. Valuable comments were received from Daniel Cotlear, Helena Ribe, Kathy Lindert,JohnNewman, Carolina SanchezPhramo, and DavidWarren. The authors would like to thank the staff of MEF, Juntos and MIMDES (in particular PRONAA) for providing much of the data used inthis analysis. We would also liketo thank MEF for providing detailed comments on the draft final report. The work was greatly enriched by the suggestions and comments of participants in three workshops, in Lima, Of particular value were contributions from Cecilia Blondet (Instituto de Estudios Peruanos), Roxana Garcia Bedoya (Mesa de Concertaci6n para la Lucha contra la Pobreza), Juan Chacaltana (CEDEP), Marfil Francke(DFID), Martin Tanaka (Instituto de Estudios Peruanos), and Martin Valdivia (GRADE). In addition, we would like to thank Livia Benavides, Betty M. Alvarado and Erika Bazan (World Bank Office -Lima) for their collaboration inthe publication of this book. V SUMMARY This report - which forms part o f the World Bank's Programmatic Analytical and Advisory Activity, RECURS0 2 - appraises the scope and performance o f Peru's Social Safety Net (SSN) system and recommends a strategy for its future development. The report aims to contribute to the Government's plans to rationalize social programs, by proposing a strategic fi-amework, within which programs can be made more efficient and effective, with clear objectives and strengthened accountability frameworks, differentiatedbetween rural and urban sectors. Chapter 1provides an analysis ofthe pattern of exclusion and vulnerability in Peru, which highlightsthe following key issues: child malnutrition (affecting 26 percent of all under fives, or around 750,000 children); low education cover and high child labor among poor children (in the poorest quintile, 26 percent o f children aged 10-14 are out of school and 53 percent are working); and old age income insecurity (78 percent o f all elderly people have no pensions). Chapter 2 provides an overview of Peru's Social Safety Net (SSN) programs, their beneficiaries and spending trends. Peru has a plethora of programs - mainly food programs - which accounted for 0.68 percent of GDP in 2006, down from 0.96 percent in 1999. This i s a relatively low level of social assistance spending, relative to national needs or compared with regional benchmarks. Nearly half of the total (0.3 1percent o f GDP) goes on food programs, the biggest of which are: Vas0 de Leche (0.13 percent o f GDP); school breakfasts (0.06 percent), and community kitchens (0.04 percent). A further 0.07 percent of GDP is spent on the workfare program, A Trabajar Urbano; 0.24 percent on Social Funds; and 0.06 percent on the new CCT program, Juntos; the latter is set to roughly double in size in 2007. The chapter analyses in detail four major programs, which account for over 80 percent of SSN spending: the food program, Vas0 de Leche; the social fund, FONCODES; the workfare program, A Trabajar Urbano; and the rural conditional cashtransfer program, Juntos -identifying their main strengths and weaknesses. Chapter3 presents an in-depthbenefit incidence and poverty impact analysis for food transfer programs, which have historically been Peru's predominant form o f transfer to poor households. It shows that - although their poverty targeting is reasonably good -they contribute relatively little to reducing either poverty or malnutrition. The main reasons for the negligible poverty impact are the low overall level of spending; and the fact that the resources are spread too thinly. There were 27 different food based programs operating in Peru in 2004; and only one o f them had a monthly benefit above US$4. There are major problems with the effectiveness of large programs, such as Vas0 de Leche and Community Kitchens, which are made difficult to fix by their "ownership" by their existing beneficiaries, such as the Cornitis de Vas0 de Leche. Driven by such political economy dynamics, few of Peru's food programs reflect best practice in their technical design. Their low nutrition impact is due to not being focused on young children, and being too centered on food distribution instead of monitoring growth and showing mothers to how to feed and care for their babies. Chapter4 summarizes the main challenges and recommends a strategy to strengthen the SSN in Peru. It argues that the limited impact o f Peru's SSN programs to reduce poverty or malnutrition i s the result of a generalized "low quality equilibrium" inthe social assistance and nutrition sector, which i s rooted inthe lack of clear objectives and measurable goals. This has opened the way for the "capture" o f programs by interest groups, and for their political manipulation. One reflection of this i s the plethora o f programs; another is the generally derisory value of the benefits. Inan effort to satisfy more and more client groups, the available resources are spreadtoo thinly among small, low-benefit interventions, limitingtheir impact. The mediocre performance of Peru's SSN programs has reinforced the low overall level of social assistance spending. The Ministry o f Economy and Finance (MEF) has become skeptical of approving "more o f the same", setting up a vicious circle in which insufficient funding and poor program vi specification lead to weak impacts - which, in turn, undermines the case for more funding. To date, the decentralization process has done little to help resolve these problems. To break out of the low quality equilibrium, the report recommends establishing clear objectives and quality standards, and setting targets for improved outcomes. Program accountability should be strengthened through robust monitoring and evaluation systems. This line of action is now being pursued through the work o f the Results Based Budgeting (RBB) office in MEF and the social program optimization process (both supported by the World Bank's REACT DPL series). The latter has already managed to reduce the number o f social programs in Peru to a third o f the previous level. The report suggests that this process should now focus on improving the effectiveness the major programs such as Vas0 de Leche, Comedores Populares, School Breakfasts, FONCODES and Juntos, which account for almost three quarters ofthe resources used inthe SSN sector. The report argues that the SSN reform process needs to be anchored to a coherent national social safety net and poverty reduction strategy. The report suggests that the SSN interventions should be differentiated, as appropriate, between the urban and rural parts o f Peru. For instance, workfare programs to deal with cyclical unemployment only make sense in urban areas; and a nationally-led small-scale infrastructure program (such as FONCODES) only makes sense for rural areas. The Report also argues that the implementation arrangements for the SSN strategy should be differentiated for rural and urban areas, due the differences in capacity between the municipal administrations in the major cities and the rest o f the country. In the 30 biggest cities, SSN programs should be decentralized, to ensure better responsiveness to local needs and improved transparency. In the rest o f the country, for the foreseeable future, strong national agencies will beneededto ensurethat SSN programs are effective, but they should work with local governments, as appropriate. Inbothurbanand rural areas, the report arguesthat the central, strategic goal of SSN programs should be to tackle child malnutrition, which was identified inthe analysis o f vulnerability and exclusion (presented in chapter 1) as Peru's predominant Social Protection challenge. This point of view was emphatically the fight against chronic infant malnutrition - under the auspices of Plan CRECER - as "the central endorsed by President Garcia in his recent review of the Government's strategy, where he characterized objective of the Government"'. Chronic malnutrition affects over a quarter o f the under-five population nationwide, or around 750,000 children. This level o f incidence is extremely high for a middle income country such as Peru. It brings in its trail irreversible lifelong consequences, in the form o f physical and cognitive deficits in human capital accumulation, and engenders a vicious circle of inter-generational reproduction o f poverty, vulnerability and exclusion. A significant reduction in chronic malnutrition would have enormous benefits for the affected households and communities, reducing poverty and vulnerability and improving distributional equity. And Peru as a whole would benefit from improved quality labor supply and stronger growth potential. The lack of progress on child malnutrition in recent years reflects the failure of public policy to provide effective nutritional services and programs. For this reason, the report argues that the new SSN's standards and accountability system should place a heavy emphasis on nutrition outcomes. The relevant programs (including Juntos and all the feeding programs) should be focused on nutrition outcomes. They should be held accountable for what they achieve to improve the quality and effectiveness of growth monitoring, improve diets and reduce malnutrition among children under-five in their respective target populations. ' Discurso del Presidente Constitucional de la RepdblicaDr. Alan Garcia Pkrez, 28 dejulio 2007. vii I n rural areas (understood here to mean the whole country outside the 30 largest cities), the strategic focus o f the SSN should be on tackling chronic malnutrition among children under five. The rural strategy would need a strong national leadership, in coordination with the regional authorities. 0 Juntos should be re-focused on reducing chronic malnutrition, becoming the central program in the rural SSN. About a quarter of Peru's children are stunted. Changing this outcome will require increased parental awareness of the problem, improved understanding of good nutrition; and the economic wherewithal to turn this into improved feeding practices. Juntos could have a major impact on these issues. It offers substantial cash benefits ($30 per month) to beneficiary families, significantly improving their consumption capacity. It can require beneficiaries to participate in MINSA's nutrition monitoring and counseling sessions; and can support MINSA to improve those reach around 600 districts, covering 10 percent o f the national population - but almost 40 percent of services. And it is well targeted on communitieswith highmalnutrition problems: during2007, it will stuntedchildren. 0 FONCODES (or its successor entity) should continue with the role of supporting municipal development through the financing and construction o f small-scale infrastructure. Many rural municipalities will need such support for the foreseeable future.Inthe absenceof such an agency, the government's "investment shock" runs the risk o f sideliningareas with low implementation capacity. I n the 30 largest urban centers, where municipal capacity is greater, SSN programs should be decentralized. The financing of decentralized SSN programs should be supported by fiscal transfers, based on transparent, equitable criteria (poverty and population statistics). The government should encourage transparent decision-making by the municipalities, through requirements for budget consultation, participatory monitoring o f budgetexecution and the strengthening o f local poverty maps. A decentralized administration, coupled with a strong accountability framework, will open up spaces for the reform o f existing programs (such as Vuso de Leche) where political economy factors have made sweeping national-level reform difficult to achieve. Once some local governments have decided to improve the targeting o f their programs and can show better outcomes, the demonstration effects of benchmarking those results will generate increasing pressure on other municipal governments to do likewise. And at a local level, the authority and credibility o f municipal governments may offset oppositionto reform from the entrenched interest groups. At a programmatic level, the strategic foci ofthe urban SSN should be (a) re-distributional food transfers focused on poor children, to alleviate poverty and improve nutrition outcomes, and (b) an effective safety net mechanismto dealwith labor market shocks. 0 Urban food programs should be rationalized. The Registro Unico de Benejkiurios (RUB), being developed by MIMDES-PRONAA, should be usedto eliminate duplication of beneficiaries and clear the way for program fusion. Food programs should also strictly apply the rules on the definition o f eligible beneficiaries (e.g. limiting milk distribution to children under 6, as specified in the Vuso de Leche law). 0 A Trubujur Urbuno should be kept in place as a countercyclical program, to cushion households against temporary unemployment during cyclical downturns. It should be operated at a minimum level during normal times, but retain the capacity to scale up quickly during recessions or in the aftermath of disasters. These strategic orientations would clarify the objectives o f the SSN, as a basis for establishing quality standards and measurable goals for each program, and strengthening their accountability, through robust monitoring and evaluation systems. viii 1 THEPATTERNOFEXCLUSION VULNERABILITY INPERU AND 1. This chapter frames the analysis of the report by asking the questions: "why operate SSN programs?" and "for whom?" Section 1.1 discusses the general rationale for SSN programs. Section 1.2 shows how poverty and malnutrition have remained severe in spite of the recovery of economic growth, underliningthe need for programs to protect the most vulnerable to accelerate the reduction of poverty. Section 1.3 analyses the distribution of vulnerability and exclusion in Peru, by income quintile and by age. It identifies among the most vulnerable groups, children in extremely poor households, who are vulnerable to multiple sources o f deprivation, likely to endanger their future development. It also highlights the need to address exclusion from school and high unemployment among young adults; the lack o f social insurance in the adult populations; and income insecurity for old people with no pension rights. 1.1 Why do societies need social safety nets? 2. Social safety nets are public programs that redistribute resources to the poor or vulnerable groups, thus protecting people from the worst effects of poverty. They include cash transfers, food assistance programs, workfare, subsidies o f various types and programs that ensure the access of the poor to essential public services, such as school vouchers or scholarships; fee waivers for health care services; and social funds. Social safety nets are only one element o f social protection policy. The other element (not dealt with in this paper) is social insurance programs, such as contributory pensions or unemployment insurance, which need not include any re-distributional element. 3. There are two main reasons why societies need social safety nets. First, SSN programs are a short term solution to alleviate poverty and destitution in society. As is well known, a sustainable reduction in poverty over medium to long-term requires broad-based growth, together with human capital investment which reaches the poor (especially in education, health, and access to basic services). However, such outcomes may take years to achieve. In the meantime, social safety nets are often the only available policy instrumentto ameliorate the debilitating effects of extreme poverty. 4. Second, SSN programs can help households manage risksbetter. Risk i s part o f everyone's life, but its impact on the poor and vulnerable groups such as the elderly and disabled, i s often more severe than for others in society. Risk can materialize at household level (Le. illness, disability or death and unemployment), community or regional level (Le. floods, famine) or nationwide (Le. drought, global financial risks, shifts in terms o f trade). The adverse impact is often more damaging to the poor than to those with better endowments o f income, physical and mental well-being and o f long-term human development. Poor people who lose income may be forced to sell land, livestock or their tools, undermining future earnings potential, or to pull their children out of school and send them to work; or simply to eat less. Such reactions may help families survive from day-to-day, but will often make it harder for them to escape poverty inthe future. An effective social safety net can reduce the needto resort to such dysfunctional coping strategies. 5. Social safety nets can help to increase options for the poor. The knowledge that safety nets exist may open the way to take initiatives that incur risksbut bringpotentially highreturns. Crop insurance can encourage farmers to grow higher yield varieties of crops and using modem farming methods; employment guarantees made possible by workfare programs may encourage households to concentrate their labor on the highest return activities rather than working in diverse activities and earning less; and the knowledge that the government will provide assistance in bad times might encourage households to hold assets in more productive, but less liquid ways (such as investing in micro-enterprises rather than keepingcashunderthe mattress for a rainy day). 6. The existence o f credible and effective safety nets can also facilitate more effective economic policies by making it easier to generate consensusaround the need for reforms that generate losers as well 1 as winners. They can broaden support for sound fiscal and trade policy, and allow other social programs to concentrate on efficiency rather than equity goals. For example, if sound safety nets are in place, the pension program can focus on improving the efficiency o f providing benefits to contributing workers, rather than finding ways to provide cashtransfers to those who have not made adequate contributions. 7. The traditional role of SSN programs has been to redistribute income and resources to the needy, helpingthem to overcome short-term poverty. But a new generation o f SSN programs, called conditional cash transfer programs, includes mechanisms to help poor households escape from poverty. These subsidies are conditional on families investing in their children's human capital, through school attendance or regular use of preventivehealthcare services. Thus, they help reduce future poverty. 1.2 Theresilience of poverty and inequalityin Peru 8. Peru faces high levels of poverty and inequality. Based on the most recent available data (from the 2006 ENAHO survey), about 45 percent o f Peru's population is poor and 16 percent i s extremely poor (Table 1-1). Based on data from 2003, inequality, measured by the Gini coefficient of consumption per capita, stands at 0.42-below the Latin American average of 0.52 but much higher than the average for middle income countries o f0.30. 9. Despite economic growth, progress in combating poverty stalled between 2000 and 2005, but a strong reduction in poverty was registered in 2006. As a consequence o f the 1998 economic crisis, poverty increased by 7 percentage points between 1997 and 2001. The recovery o f growth after 2000 halted this deterioration, but only began to reverse it in 2006. However, the 44.5 percent poverty rate registered in 2006 still remains above that of 42.4 percent which was registered in 1998. Similarly, extreme poverty has only improved modestly in recent years, falling from 19.5 percent in 2001 to 16.1 percent in 2006. The problem of poverty i s particularly acute in the rural parts of the country, especially in the sierra, where almost 80 percent of the population is poor and more than half is living in extreme poverty (data not tabulated). -Ta 1997 1998 1999 2000 2001* 2002 2003 2004 2005 2006 Poverty headcount National 42.7 42.4 47.5 48.4 49.8 50.0 52.2 48.6 48.7 44.5 MetropolitanLima 25.4 24.1 31.4 38.9 28.3 34.3 33.7 30.9 32.6 24.2 Urban 33.0 34.0 37.3 35.3 42.6 39.0 39.5 37.1 36.8 31.2 Rural 66.3 65.9 71.8 70.0 76.9 74.3 75.7 69.8 70.9 69.3 Extreme poverty National 18.2 17.4 18.4 15.0 19.5 18.7 n.a. 17.1 17.4 16.1 MetropolitanLima 2.3 2.4 2.7 1.6 1.7 3.3 n.a* 1.3 2.0 0.9 Urban 7.6 7.5 6.3 6.1 10.0 8.2 n.a* 6.5 6.3 4.9 Source: INEI . *The methodologyfor estimating the poverty rate inPeruwas changedin2001; as aresult the Rural I 41.5 40.0 44.4 35.6 I 46.3 42.4 n.a. 36.8 37.9 37.1 estimates for 2001-2006 are not comparablewith earlier estimates; inaddition,povertyrates between2001-2003 are not comparableto 2004-2006. 10. Progress in combating malnutrition has also stalled. According to DHS data, a quarter of children under-five are malnourished on the height-for-age indicator (chronic malnutrition) and there i s a strong correlation between stuntingand income poverty (Figure 1-1). 11. The World Bank's Poverty Assessment (2006) identifies four main reasons for the limited progress inreducingpoverty inPeru: Since 2000, growth and investment have not been sufficiently strong and broad-based to bring the overall poverty rate down. They have beenbiasedtowards capital-intensive sectors, limiting 2 employment creation, and have had only a modest impact among the poorest households, especially inrural areas. 0 Inurban areas, low productivity among small firms inthe informal, labor-intensive sectors has limited employment and real wage growth. Recent improvements in employment and wage levels in urban areas were mostly concentrated among medium and large firms and among the highly skilled. In rural areas employment has increased, especially in the agricultural sector, although due to low productivity levels inthe sector, wage levels have remainedflat. Low asset endowments (human and physical capital) and poor access to financial and insurance markets and to effective safety nets increasethe exposure of poor people to shocks, limitingthe effectiveness of their coping strategies. As a result, they are more vulnerable to economic downturns, such as the 1998 crisis, and less able to benefit from the opportunities generated by periods of growth. Limited access to public services, infrastructure and public institutions, particularly in rural areas, has reinforced inequalities in social and economic opportunities between the poor and the non-poor. Figure 1-1:Malnutritionby expenditure quintile, 1996 and 2001 Share of stuntedchildren 50 % ofchildren under 5 whose heightfor age IS e-2 sd z-score _, 45 45 - r:40 40 - 35 0 30 9 25 20 0 % 15 10 5 0 Q1 Q2 Q3 Q1 Q2 Q3 Q5 Total Q4 Q5 Total Q4 Iea1996 2000 I 1 1996IB 20001 Source: Gwatkin Rutstein Johnson, Suliman, Wagstaff, and Amouzou, "Socioeconomic Diferences in Health, Nutrition, and Population in Peru", 2nd ed. Washington, D.C.: The World Bank, forthcoming. Note: Wealth quintiles basedon an index of householdendowment with assets. 12. Economic growth has now recovered strongly and looks set to stay high. Inthis setting, the critical challenge for Peru i s to ensure that this recovery o f economic growth leads to sustained poverty reduction and improved equity. Output in the economy has been growing for five consecutive years, reaching 6 percent in2005, the fastest rate since the mid-1980s and around 4.5 percent in2006 (INEI, 2006). Growth has been driven by investments and improved export performance in extractive industries, with Peru benefiting from high world prices for minerals. However, the links from growth to poverty reduction in Peru have been relatively weak, as growth and investment have been biased towards sectors with low labor intensity (such as mining). In addition, historically volatile growth rates have resulted into high levels o f uncertainty and low levels o f entrepreneurial confidence on the sustainability o f growth (Peru Poverty Assessment, 2005). 0 Urban economy: The main beneficiaries o f urban growth are those working in larger companies, where wages are higher and workers have access to social security benefits. For example, pay in companies with more than 50 workers rose by 6.3 percent in 2004, whereas pay in the smallest firms (with fewer than 10 workers) increased by 1.3 percent. However, three-quarters o f urban workers are employed in small and medium-sized firms and barely a quarter are infirms with 50 or more employees. 3 Rural economy: Some dynamic sectors of agriculture have taken advantage of risingdomestic consumption and strong overseas demand for asparagus, mangoes, lemons and other fruits and vegetables. However, the growth areas are concentrated mostly on the coast, and only account for a small proportion of the rural workforce. As a result, rural poverty levels remain stubbornly high(about 70 percent), and extreme poverty stands at 37.1 percent (Table 1-1). 13. The present report provides an in-depth assessment o f ways inwhich Peru's social safety net could be made better targeted and more effective. This complements the 2005 Poverty Assessment which estimated the potential impact on poverty levels o f a broad set of policies, including a more efficient social safety net, but also including adequate taxation policies and well-targeted, effective programs in health, education, basic infrastructure and utility services. 1.3 WhoNee& SSNPrograms?A profile of vulnerablegroups in Peru. 14. Consumption-poverty in Peru correlates strongly with other dimensions o f poverty, such as malnutrition, child labor, school drop-out, and illiteracy. The "life-cycle approach" developed by Hall and Arriagada2 i s commonly used to identify key vulnerable groups. Under this approach, the population is classified into age-groups: early childhood, school-age, youth, adulthood and old age. For each age-group, the circumstances that lead to or aggravate poverty and deprivation3are identified. Finally, the prevalence o f specific deprivations (such as malnutrition, child labor, illiteracy or unemployment) is presented by consumptionor wealth quintile, to assess to what degree it is concentrated among the extreme poor. Using this approach, Table 1-2 presents data on the incidence of different risks in Peru, by age group and consumption quintile; and Table 1-3 shows the number of individuals who are affected. The information was derived from the most recent surveys available, ENAHO 2003 for material poverty, and DHS 2000 for malnutrition or healthrisks, and classified into age groups and quintiles4. 15. The main risks identified inthis analysis are as follows: 0 For the 0 to 5 age range; malnutrition (deficient height for age, or stunting, affects 26 percent of all children under five, and 47 percent in the lowest quintile, Q1, compared with only 5 percent inthe richest quintile, Q5); child mortality (64 per thousand inQ1, compared to only 14 for Q5); lack of access to preschool education (only 39 percent of children 3 to 5 years old in Q1 are enrolled, compared to 75 percent inQ5). 0 For children aged 6 to 16:low school enrollment (11percent o f children aged 6 to 11 inQ1are not in school); high school dropout among older children (26 percent of children aged 12 to 16 in Q1 are out of school, compared to 8 percent in Q5); late entry to the secondary cycle (affecting 33 percent in Q1) and a high incidence of child labor (affecting 53 percent of those aged 10 to 14 in Ql). 0 For youth aged 17 to 24: low rates o f school enrollment, particularly among the poorest quintile (76 percent of youth in Q1 are not in school compared to 50 percent in Q5); relatively high youth unemployment rate (15 percent, compared to 8 percent for all adults) and high inactivity rates among the urbanyouth. For the adult population between the ages of 25 and 64, the lack o f social insurance protection is the main issue. Ninety nine percent of the extreme poor (Ql) do not have social insurance. Hall andArriagada(2001) inMexico: A comprehensivedevelopment agendafor the new era, World Bank, WashingtonDC. Poverty is definedina multidimensionalsense, encompassingsuch variables as consumption, nutrition, education, health, child labor and employment. 4 The indicators derived from ENAHO 2003 are classified using quintiles of real consumption per capita, while those derived from DHS 2000 are basedon wealth quintiles. For DHS information, see Davidson R. Gwatkin, Shea Rutstein, Kiersten Johnson, Eldaw Abdalla Suliman, Adam Wagstaff, and Agbessi Amouzou, "Socioeconomic Diferences in Health, Nutrition, and Population in Peru", 2nd ed. Washington, D.C.: The World Bank, forthcoming. 4 Employment in poor quality jobs (unremunerated household work or paid employment at very low wages) i s also a key issue. Illiteracy stands at 22 percent in Q1, compared to 12 percent national average). 0 Among the over 65 age group, income insecurity i s the main problem: overall, only 22 percent of the elderly have a pension and this falls to less than 5 percent in Q1. This may explain the high share of elderly who work (51 percent of the elderly in Q1 work, compared to only 27 percent o fthose inQ5). 16. The life-cycle analysis presented in Table 1-2 and Table 1-3 suggests that the group with the highest risk of poverty and other forms of material deprivation are children living in extreme poverty. Children are the age group with the highest incidence of poverty, and constitute the largest group of poor (children under 14 represent 47 percent of the extreme poor). A majority o f poor children are also affected by forms of deprivation other than consumption poverty, such as high rates o f infant mortality, malnutrition, school dropout, or low academic achievement. Infant mortality reaches 64 per 1,000 inthe poorest quintile, more than twice that inthe richest quintile (average of 28 per 1,000 inthe region). 17. Of particular note is the extremely high rate of chronic malnutrition (stunting) in Peru. Chronic malnutrition currently affects over a quarter of the under-five population nationwide (around 750,000 children) an extremely high level o f incidence i s extremely high for a middle income country such as Peru. The stunting rate rises to 47 percent among children from extreme poor households; such children accounted for 53 percent of the total cases in the country in 2000. Chronic malnutrition brings in its trail irreversible lifelong consequences, in the form o f physical and cognitive deficits in human capital accumulation, and engenders a vicious circle of inter-generational reproduction o f poverty, vulnerability and exclusion. A significant reduction inchronic malnutrition would therefore have enormous benefits for the affected households and communities, reducing poverty and vulnerability and improving distributional equity. And Peru as a whole would benefit from improved quality labor supply and stronger growth potential. 18. Access to education services, although improved substantially inrecent decades, also fails the poor. About one-third o f the students from the poorest quintile (in extreme poverty) do not complete primary education by age 14. There i s a marked difference between the school enrollment rate for children from the poorest quintile and other children (Figure 1-2). The poorest childrenjoin primary school later, repeat grades more often, and drop out earlier. This may be indicative that the efforts of the Government o f Peru on the supply side (better school infrastructure, and better supplies) are not enough - there is also a need to address the issue of demand-side constraints to the accumulation o f human capital in the poorest households. 5 Figure 1-2: School Enrolment by Age and Quintile, 2003/04 100% -z 90% 0 80% v1 p 70% I U c 5 60% c 3 50% E 40% s 30% 20% I 6 7 8 9 10 11 12 13 14 15 16 17 18 Age --eI-poorest -1)- 2-4 5-richest Source: Author's estimationsbasedonENAHO 2003/04. 19. Lower school participation of children from extreme poor households i s correlated with high prevalence o f child labor. According to the Inequality inL A C Flagship Report, child labor is substantially higher in Peru than the L A C region average: more than a quarter o f children work in Peru, compared to only 11 percent in LAC. Even more striking is the high incidence of child labor in extreme poor households, where 53 percent of children work, compared to the L A C average o f 17 percent (Figure 1-3). Figure 1-3: Prevalenceof Child Labor, Peru and LAC Child labor is widespread inPeru Peru LAC Average Source: "Inequality inLatinAmerica: Breakingwith History", World Bank(2004). 6 m a v w C $ . u m In 20. For rural girls, the phenomenon of child labor is often compounded by other forms o f exploitation. Many are sent at a very young age to work as domestic employees in the cities, where they work long hours, are severely underpaid and are often physically abused (Vega, 2001). In an effort to address this problem the Government has signed international agreements protecting children rights, introduced legislation that prohibits child labor and created a special department within MIMDES and an inter- sectoral commission to enforce the legislation. However, since the causes of child labor are mainly economic, legalprohibition, alone, is unlikely to end it. Table 1-4: Incidenceof Extreme Poverty (Total I 8.0 I 37.9 I 18.3 I 8.1 I 42.9I 20.8I 8.0I 40.2I 19.4I Source:Author's estimationbasedon ENAHO2003, annual sample. 21. The second largest group of extremelypoor people in Peru is the working poor (37 percent of the total). Among this group, the incidence of extreme poverty highest in rural areas, especially in the informal sector, such as independent workers and non-salaried household labor (see Table 1-4). The main circumstances leading to extreme poverty for this group are low-wage jobs coupled with high ratio o f dependents to working adults in the household. In contrast, extreme poverty i s virtually absent for non- manual workers, and i s also low for manual workers in urban areas. Rural and female wages are substantially lower compared to urban or male wages. 22. The two largest vulnerable groups -children and working-age adults in extreme poverty - overlap substantially in the same households, so programs to support children would also help their working parents, and vice-versa. About 94 percent of the extreme poor (Ql) live in households with children and where at least one adult i s working; and in 68 percent o f Q1 households, all adults in the household are working. However, 1 percent of the extreme poor live in children-only households which need Government assistancethrough special programs (Table 1-5). Table 1-5: Distribution of Extreme Poor by Type of Household Type ofHousehold YO Adults, all working, with children 68 Adults, not all working, with children 26 Adults only, all working 4 Adults only, not all working 2 Children-only households 1 Total 100 *)Childrenare definedas 14years oldor younger Source: Author's estimationbasedon ENAHO 2003 (INEI), annual sample. 9 2 THESOCIAL SAFETY NETINPERU 23. This chapter provides an overview of the main social safety net (SSN) programs in Peru. Section 2.1 describes the programs that constitute the social safety net in Peru. Section 2.2 shows that the SSN spending is low, compared to other countries in the LAC region. SSN spending is also small compared with the poverty gap o f the extremelypoor households inPeru. The last section (2.3) highlights some of the weaknesses inthe design of SSN programs. 2.1 Main Componentsof thePeruvian SocialSafety Net 24. Historically, the Peruvian SSN has been built around three types o f program: (a) a large and disparate set of feeding programs, with diverse target groups and varying objectives; (b) a social fund, FONCODES, which finances basic social infrastructure and income-generating projects in poor rural communities; and (c) workfare programs, which offer temporary employment. In 2005 a new program - Juntos was introduced inthe poorest districts in rural Sierra in four departments to provide conditional - cash transfers to poor families with children. Table 2-1 summarizes the main programs' expenditure, number of beneficiaries and benefit levels in 2006. In all, SSN spending totaled S 1,898 million Soles, amounting to 0.68 percent of GDP. Fourty six per cent o f the total (0.3 1 percent o f GDP) was for food programs, 10 per cent of the total (0.07 percent o f GDP) was for workfare, 35 per cent o f the total (0.24 percent of GDP) was spent on social funds and 9 percent of the total (0.06 percent o f GDP) was for the Juntos CCT program. 25. community entities - often without strong targeting or accountability frameworks - and the high level of Two striking characteristics o f the Peruvian SSN are the existence of programs controlled by programmatic disintegration: in a diagnostic undertaken in 2006, the Government identified 82 separate social programs. Box 2-1 provides a summary o f the historical roots of the Peruvian SSN model, which explains the origin of community-run programs as a responseto a vacuum left by the lack of government- led social programs in a context o f distrust engendered by dictatorial governments in the 1990s. This, in turn contributed to programmatic disintegration, as governments sought to meet the demands of as many interest groups as possible. Another factor in program disintegration in past years was donor funding: in the face of poor accountability frameworks, there was a tendency for donors to propitiate new delivery agencies, instead of funding established national programs. The following paragraphs summarize the main elements of the modern SSN'. 26. Food-based Programs. In 2004, 27 food-based programs were in operation in Peru (a full list is given in Annex 1). The largest programs in terms o f budget and number of beneficiaries are the municipally-run Vas0 de Leche (glass of milk) program for pre-school children; the school feeding programs (school breakfast and school lunch), and the community kitchen (Comedores Populares) for children and adults. There are several nutritional programs for infants and pregnant and lactating mothers but none is very big. These programs reach, altogether 6.7 million beneficiaries, or 25 percent o f the population, but the average subsidy per beneficiary is only US$3.3 per month6.About 70 percent of food programs' resources reach the poor (the bottom 55 percent of the population); and 30 percent goes to the extreme poor (the poorest consumption quintile). 27. The Social Fund. The National Compensation and Social Development Fund (FONCODES) was created in 1991 to finance small local investment projects. Its stated objectives were: to generate 'This study concentrateson the larger programs, whichaccount for the lion's share of SSN spending.The Peruvian SSN also includes important smaller programs, such as those for street children and disabled Peruvians, mainly operatedby MIMDES. The exception to the pattern of very low per capita food programbenefits, spread very thinly, is the Wawawasi program, which has much higher benefits (US$23 per month) but reaches relatively few children (40,000 nationwide). See Table 2-1. 10 employment, help alleviate poverty, and improve access to social services. The projects financed by FONCODES are demand driven and involve the community in execution and supervision. The most popular projects are the construction or rehabilitation of schools, health posts, water and sanitation systems, rural roads, secondary electrification schemes, and small-scale irrigation works. The social fund has been used to reach the poorest rural communities using poverty-map allocation methods and, in the past, it absorbed more resources than those channeled through programs that provide direct transfers to households (meals inthe case of food programs, or wages in the case o f workfare programs). At its peak, FONCODES' annual budget stood at around 0.3 percent o f GDP. Its relative importance has since declined somewhat, but it still amounted to 0.2 percent of GDP in2006. Itremains the single biggest SSN program (30 per cent of the total, S.567 million). However, its future i s now inquestion. 28. Worvare Programs. Two workfare programs have operated in Peru in recent years: A Trabajar Urbano and A Trabajar Rural. A Trabajar Urbano provides temporary employment opportunities for unskilledworkers inpoor areas, at relatively low wages. The program provides work for up to six months on temporary public projects undertaken to renovate social infrastructureand perform general community maintenance work. In 2006 it had a budget of S/.187 million, which accounted for 10 percent o f SSN expenditure, and about 0.07 percent of GDP. It reached 57,000 beneficiaries (about 0.2 percent of the population). Program participants inA Trabajar Urbano obtain a monthly wage of S/.300 (or US$91). This is set considerably below the minimumwage of S/.427 (US$129) per month in order to discourage the crowding out of other employment. Nevertheless, it is by far the largest benefit package of any SSN project. Incontrast, A Trabajar Rural is a budget umbrella used to bundle ad-hoc interventions. 29. Conditional Cash Transfers. The recently introduced conditional cash transfer program, Juntos, provides a monthly benefit of 100 Soles (about US$30) to poor families with children under 14 years old, conditioned on their take-up o f preventive health care visits, nutrition supplementation, school attendance, and obtaining birth registrationand identification documents. The program started inthe poorest 110 rural districts inthe departments of Huanuco, Huancavelica, Apurimac, and Ayacucho, benefitingabout 70,000 households. By the end o f 2006, it reached 210 more districts in five additional departments, reaching about 160,000 households in all. The program's cost was S/. 173 million in 2006, representing 9 percent o f SSN spending or 0.06 percent o f GDP. In2007 it will cover 638 districts and over 400,000 households. 11 Table 2-1: Social Protecl n Spending2006 Spending lumberofbeneficiarie tge benefivn mill. S.1 %of GDI thousands %ofpop. s.1 US$6' rota1Social Protection 10,522 3.76 Total Social Insurance, of which: 8,624 3.08 Pensions (SNP) 3,837 1.37 469 1.7 682 207 Other pensions 4,788 1.71 n.a Total Social Safety Net, ofwhich '': 1,898 0.68 Food-based programs 877 0.31 Glass of Milk (Vas0 de Leche) 363 0.13 2,959 10.9 10 3 Wawa Wasi (day care) 46 0.02 48 0.2 80 24 School lunch(Almuerzos escolares) 26 0.01 485 1.8 5 1 School Breakfast(Desayunosescolares) 182 0.06 1,926 7.1 8 2 CEIPRONOEI (food supplement for preschool children) 37 0.01 433 1.6 7 2 PACFO (food supplement for at risk children) 49 0.02 295 1.1 14 4 PANFAR(food supplement for at risk family) 13 0.00 79 0.3 13 4 CommunityKitchen, food for work and shelters ' 105 0.04 n.a Others 3' 56 0.02 n.a Total Workfare, of which 187 0.07 A Trabajar Urbano 187 0.07 57 0.2 300 91 Total SocialFund I Community Driven Development 661 0.24 FONCODES basic social infrastructureinvestments - 567 0.20 n.a PRONAMCHCS naturalresourcesmanagement - ' 94 0.03 n.a Conditional Cash Transfers program 173 0.06 JUNTOS 4' 173 0.06 270 1.o 100 30 'ro Memoria '' Average monthly wage, urbanarea 2006 2600 781 Average netwage unskilled worker, urbanarea 2006 - 32 1( Minimum monthly net wage (SI.)2006 500 15: GDP, mill. SI. 2006 " 280,2 05 Population, 2005 7,219,264 Source: Basedon data from INEI, SlAF MEF (Nationaland subnationalgovernment),DGAES MEF, - - I /Administrative expendituresare not included(exception:Glass of milk, and FONCODES programexecuted by sub nationalgovernments) 21 Expendituresexecuted by Nationaland Sub NationalGovernments. 3/ It includesthe following programs: Comedores infantiies, piloto subprograma infantil, equipospara el programainfantil y escolar, PEBAL, PROMANR,fundaci6npor 10s ninos, acta de compromiso, PANTBC,convenios, subsidios a comedores,situacionesde emergenciay programade cooperaci6n internacional. 4/ Expendituresexecuted by PCM. 5/ INEI. 6/ Exchage rate: S1.3.3 7/ Own estimates based on historic growth. 12 Box 2-1: The historical roots of Peru's SSN Model Peru's SSN model has strong roots inthe community-based development initiatives and political mobilizations of the 1960s and 1970s. Inthis period, community based organizations (CBOs) grew up to occupy a vacuum left by the weak public sector, in a setting o f centralized dictatorship and endemic economic crises. Later, in the 1980s, they were institutionalized and co-opted into the political process. Programs such as Vas0 de Leche and the community kitchens (comedores populares) started as informal initiatives o f poor communities, in the absence o f assistance from the government. The creation o f the comedores populares coincided with a social movement organized during the crisis o f 1976-1979. Following economic mismanagement by the Velasco Government, in 1976 a structural adjustment package was implemented, resulting in increased unemployment, inflation and reduced real wages. This reinforced popular opposition to the dictatorship and resulted in turmoil and strikes. In this context, hundreds o f community kitchens were set up inpoor neighborhoods to feedthe strikers. The community kitchens continued operating into the 1980s, when Peru returned to democracy, but economic revival remained elusive. In 1983, a new crisis led to a 10 percent fall in GDP and three-digit inflation. Against this backdrop, politicalparties competedfor popular support by offering funding for the CBOs. While FirstLady Violeta Correa offered support to the "cocinas populares", Alfonso Barrantes, the left-wing mayor o f Lima, promised a million o f glasses o f milk for the children o f Lima, to be delivered through an extended network o f "comitb de vas0 de leche". The administrations o f Fernando Belaunde (1980-1985) and Alan Garcia (1985- 1990) strengthened government support to the community kitchens. Under the Direct Support Program (PAD) started by the f r s t Garcia administration, the "clubes de madres" in Lima and Callao were given a cash subsidy, inaddition to the food transfers available to other community kitchens, a privilegedstatus still maintainedtoday. International donors were also supportive o f community-based initiatives. In 1992, under the Fujimori administration, the National Food Assistance Program (PRONAA) was created to provide food assistanceto rural and marginalized urban areas, and poverty alleviation was centralized under the Ministry of the Presidency (PRES), leading to considerable political discretion in support to the different programs. This, however, did not resolve the problems o f fragmentation and overlap o f social programs and the decade o fthe nineties saw growing corruption and clientelism. As a result o f this complex historical process, a movement that had originated in the poor neighborhoods, founded on principles o f reciprocity common among indigenous communities, came to depend heavily on assistance f?om the national and local Government or national and international ONGs. As a well-organized mass o f people that could be mobilized quickly, the CBOs become a powerful political instrument and governments have nurtured their relationship with them. At the same time, the CBOs themselves have sought to cultivate their relationships with the government o f the day. This has become an accepted strategy for CBOs in modern Peru. Groups that do not receive state support do not condemn those receiving it, rather, they commend and envy their ability to hook up with public funding. Inrecent years, the government has sought to reform the food programs and clarify their role within the social safety net. However, this has proved a daunting task, in the face o f resistance from long-time beneficiaries, mayors and the food producers who supply them. In2001, with support from the IDB and the World Bank, the transition government improved the transparency and accountability o f food programs, and began unifying them by transferring the FONCODES' infant and school feeding to the National Institute o f Health. In 2002, the Toledo government brought all the central government food programs under one organization - PRONAA - and established a multi-sectoral Board o f Directors to improve links with the education and health sector. The reform path set in 2002 included plans to further consolidate programs, improve their design by introducing health and nutritional components, to carry out impact evaluations and to develop a conditional cash transfers. However, this centrally-led trajectory was interrupted in 2003, when Peru decided to pursue the decentralization o f food programs and basic social infrastructure investments. Now, the Garcia administration has launched a new program consolidation and optimization process for the social sectors, led by the Technical Secretariat o f the Inter-ministerial Commission for Social Affairs (ST-CIAS), which aims to consolidate programs and promote results-based-budgeting, Adapted from Franke, P (2005) and A. Linder(2005). 13 2.2 Public spending on theSocialSafety Net 30. Total social sector spendinginPeruhas beenrelatively stable inthe last decade, averagingaround9 percentof GDP. In2004 Peru spent 3 percent of GDP on education, 1.5 percenton health, 3.9 percent on social protection (which includes pensions and the social safety net, SSN). Within Social Protection, pensions' spending predominates, accounting for 80 percent of the total, leaving only 20 percent for the SSN.7 As well as being small, the SSN has shown a worrying tendency to be pro-cyclical. The economic crisis that hit the country in 1998 and the recessionthat followed found Peru unprepared to protect SSN spending, and the structure of social spendingshiftedstrongly againstthe SSN. While spendingon health, education and social insurance all raised, SSN spending fell from 0.96 percent of GDP in 1999 to 0.76 percent in2004. (Table 2-2). This relative and absolute decline in SSN spendingafter 2002 is attributable to reduced spending on workfare and in the food programs operating in the rural sierra. The reduced financing programmed for the social funds in 2007 should be offset by higher municipal and regional investment, inthe contextof decentralization(Figure 2-1). Table2-2: PublicSpendingon SocialSectors, 1997-2007 (YOof GDP) 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006* 2007* Actual Actual Actual Actual Actual Actual Actual Actual Actual Actual Budgeted I.EDUCATION 2.53 2.65 2.96 2.87 2.88 3.04 3.11 3.00 3.06 3.07 3.06 11. HEALTH 1.27 1.34 1.39 1.35 1.37 1.45 1.51 1.50 1.52 1.67 1.58 111. SOCIAL PROTECTION 3.41 3.37 3.89 3.92 4.02 4.10 3.82 3.90 3.89 3.76 3.63 A. Pensions 2.52 2.63 2.93 3.16 3.25 3.35 3.17 3.10 3.18 3.08 2.97 B. Social SafetyNet 0.88 0.74 0.96 0.76 0.77 0.75 0.65 0.76 0.71 0.68 0.66 B1. Food-basedPrograms 0.40 0.36 0.41 0.40 0.48 0.43 0.39 0.40 0.37 0.31 0.31 B2. Workfare - 0.05 0.14 0.11 0.06 0.06 0.07 0.05 B3. Social Funds 0.48 0.38 0.55 0.36 0.24 0.18 0.15 0.30 0.24 0.24 0.17 B4. Conditional cashtransfers 0.04 0.06 0.13 TOTAL 7.21 7.36 8.24 8.14 8.27 8.59 8.44 8.40 8.47 8.50 8.27 - PAAG(MINSA), BCRP. * Uses estimatesof GDPbasedonhistoric growth. Source: Based on data fiom INEI; SIAF-DNPP (MEF); Pronamachs web; Perti en Ntimeros 1997 2004; PRONA; 3 1. SSN spendingin Peru is low comparedto other countries. Social spending is lower in Peruthan the average for Latin America in each of the three sectors: health, education and social protection. (Figure 2-2). The largest difference, though, is that for social safety net spending. SSN spending in Peru is amongthe lowest inLAC andwell below that of other regions. Peru spends barely half ofthe LAC region average of 1.3 percent of GDP and about one third of the 2.2 percent of GDP registered for OECD or Europe and Central Asia region (Figure 2-3). But, although it spends less, Peru's SSN programs cover a much larger fraction of the population than those of other countries in the region. This implies that the limited funds are dilutedacross a large number of beneficiaries, resulting inaverage transfers which cover only a fraction ofthe consumptiondeficit of poor households(Figure 2-4). 'Almost85percentof pension spending (about 2.6 percent of GDP) is subsidized with transfers fiom the central government. In2004 the government undertook a Constitutional reform that will reduce the hture liabilities of the civil servants pension regime, gradually reducing the amount of public subsidies required to fmance the public pensionsystemdeficit. 14 Figure 2-1: Recent trends in SSN spending in real terms Social safety netspending f997 2006 - (2004 constant prices) 80 I , 2000000 70 1800000 0Conditionalcash transfers 1600000 60 Social Funds 1400000 2 9 50 0 3 1200000 5 Workfare Ly 40 1000000 ; GK3Food-based 5 30 XI 800000 Programs v) 600000 zf 2 +Social safetynet 20 spending per 400000 capita 10 200000 0 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Source: Based on data .from INEI; SIAF- DNPP (MEF); Pronamachs web; Perd en Nzimeros 1997-2004; PRONA; PAAG (MINSA), BCRP. Figure 2-2: Public Spending on SocialSectors in the LAC Region Spending on health, education, social protection (social insurance and social assistance) as a share of GDP, most recent year PublicSocial Spendingas a share of GDP: LAC, OECD 35 0% 30 0% 25 0% 20 0% 150% 100% 5 0% 0 0% GTM-OO DOM02 PER-03 MEX-03 COL-04- CHL-00 ARG-03 BRA-04- LAC U S -01 OECD-01 Continental cons cons Average Europe41 Source: Lindert, Skoufias and Shapiro(2005). 15 Figure 2-3: Peru's SSN Spending as % of GDP, with regionaland globalcomparators Social Safety NetSpending, %of GDP most recentyear 2000-2006 2.5 ............................. 2.0 1.5 ................................................ 1.o ............................................. 0.5 OECD Europeand Middle East and LatinAmerica SouthAsia Peru (2006) CentralAsia North Africa and the Caribbean ~ Source: Ownestimates for Peru; Lindert, Skoufias and Shapiro (2005) for LAC; basedon Blank and others (forthcoming) for other regions. 32. Peru's low SSN spending compared with other countries reflects the low overall level of public spending, coupled with the low priority assignedto SSN programs. Low social sector spending inPeru is partly the result of low taxation and weak tax collection. In 2006, Peru's central government current revenues in Peru totaled only 17 percent of GDP. Although this reflects a noteworthy increase from the level o f 14 percent observed in 2002, Peru's tax effort remains inferior to that o f more developed countries and o f many less developed countries, too. As a result, fiscal space is limited. What i s more, within the available space, the priority assignedto SSN programs is also relatively low. Share of Total Public Spending going to SOCIAL ASSISTANCE (FiscalPriority) Less than 4.5% Between4.5% and 7% More than 7% L.. 8 - Morethan Colombia (0.9) Venezuela (1.O) Nicaragua (1.1) Brazil (1.4) Uruguay (4.7) * E * 24% = ? I 2 2 Between P, a Paraguay0.4) Costa Rica (1.1) LAC (1.3) Chile (1.7) Honduras (1.7) :$ $ 19% and Ecuador (1.O) a * 24% 'k a n & El Salvador (2.4) C P Peru (0.7) Argentina (1.2) Guatemala(1.1) k Under Mexico (1.1) m 19% Dominican Republic (1.O) 16 34. The vertical axis categorizes countries according to total spending as a share o f GDP; and the horizontal axis does the same for the share of total public spending devoted to social assistance (non- contributory programs). Brazil and Uruguay have high levels of social assistance spending as a share of GDP because they have high spending levels overall and devote large shares to social assistance programs. As a result, social spending accounts for more than 1.4 percent of national income. Other countries, such as Guatemala and Salvador, have low levels of government spending overall, but place a high priority on social assistance programs. But Peru is among the countries with the lowest level of public spending in Latin America and among the group o f countries who attach a very low priority to social assistance spending (less than half o fthe average for the region). 35. SSN spending i s low relative to the poverty gap but fairly large relative to the extreme poverty gap. In the absence of SSN spending, total poverty gap (the difference between the present consumption of poor households and the amount they need to consume to escape from consumption poverty) is estimated at about 7 percent of GDP. At 0.7 percent of GDP, total SSN spending could cover only 10 percent of that gap. However, SSN spending represents a much bigger proportion of the extreme poverty gap (the amount o f resources needed to take all households out of extreme poverty). In the absence of SSN transfers, the extreme poverty gap would be about 1.3 percent of GDP. 36. The reduction in poverty achieved by SSN spending is below its potential, because part o f the resources leaks to non-poor beneficiaries. About a third o f the SSN spending reaches the extreme poor and 70 percent reaches either the extreme or moderate poor; while 30 percent goes to non-poor households. As a result, it i s estimated that in practice, SSN spending in Peru reduces the poverty gap from 6.8 percent to 6.3 percent of GDP, and the extreme poverty gap, from 1.32 percent to 1.1 percent of GDP. Perfect targeting o f the existing SSN resources on the extreme poor would halve the extreme poverty gap. Under a more realistic scenario, where leakage is reduced by 50 percent (that is, that only 34 percent o f the funds go to households who are not extremely poor), the extreme poverty gap would fall by one third o f its pre-transfer level (Figure 2-4). Fig .e 2-4: Modeledimpacton the povertygap of increasedtargeting of SSNspending toward extremepoor PovertyGap vs SSN Spending SimulatedChange in ktrerne PovertyGap in 2003 I under2 scenanosoffmpmvedfargetmg 100 7.0 ; 90 6.0 80 5.0 70 3 B 60 ' 4.0 50 3.0 I 40 : 2.0 4 30 1.o 20 s 10 Re-transfer] ~ - t r a n ~ , e r ltransfer Inthe absenceof Current (leakageof lnprovedtargeting Perfecttargeting ~~ SSN Total poverty gap &trem poverty,gap SSN spending 58% of funds) (reduce leakageto (elimnateleakage) spending 34%) (Scen 1) (Scen 2) ~ Note: The left Dane1presentsthe total and extreme poverty gap beforeand after SSNtransfers. Note that the reductioninpoverty gap after SSN transfers is lowerthanthe SSN spending, because part ofthe resources does not reachthe poor. The rieht Dane1presentsthe simulated poverty gap for householdsinextreme povertyunder two scenarios. Scenario 1assumes 66 percent ofthe resources are receivedby the extremepoor (comparedto only 32 percent in2003). Scenario 2 assumesperfecttargeting, i.e. all resources reach only the extremepoor. Source: Author's estimationbasedon ENAHO 2003104 data. 17 37. The structure of social protection agencies within the Peruvian public sector is shown in (Figure 2-5). The following sections review the four largest SSN programs (which between them account for about 70 percent of all SSN spending) to suggest ways to improve their performance, covering: (i)Vas0 de Leche; (ii)workfare programs; (iii) social fund, FONCODES; and (iv) the Juntos CCT program. the 18 1r I g5 c -a,a, E c rn -0 3 5 3 C m $3 ,o L U 0 0 LL i c a, Y Y .-5 C I r ,r1 i 5 0 Lr 2.3 Vas0 de leche 38. Vuso de Zeche spends $110 million a year, which is over 40 percent of the overall budget for food assistance, and has almost 5 million beneficiaries. It has been criticized for having high leakages towards the non-poor or non-intendedbeneficiaries(Vazques 2005; Alcazar 2003). However, analysisundertaken for the presentreport shows that a large proportion ofthe resources are receivedby poor households: over 45 percent of the beneficiaries in urban area and 65 percent in rural areas are from the bottom two quintiles (Figure 2-5). The fact that the programis better targetedinrural areas is not surprising, given the higher incidenceof poverty there. 39. Given Vuso de Leche's considerable coverage of children in extreme poverty, it could become an effective instrument, given feasible improvements in its design and implementation. The program has a number of strengthsto buildupon: Beneficiaries': The programtargets children from zero to six years old, expectant and lactating mothers. All these groups have poverty rates above average, and constitute the key vulnerable group inPeru.As illustrated in Figure 2-6, children in the priority group have a higher probability of benefiting from the program. Coverage: The program has a large coverage among the poor population: 44 percent of poor households with children in urban area and 73 percent in rural area have one or more beneficiariesof Vusode Leche. Genderneutral: Survey data show no gender bias inthe program. Social capital: 23 percent of Peruvianhouseholdsparticipate incomitis de vas0 de Zeche; this is by far the largest CBO structure in the country. In contrast, the second largest association movement (usociucidn vecinul) attracts only 7 percent of households. The comitis de vuso de Zeche therefore constitute an important part of Peru's social capital and contribute to women's empowerment. With the right incentives, this network could play an important role in poverty reduction inPeru. Targeting: The program uses simple targeting mechanisms: demographic (age) and self- selection (mothers have to pick up the food rations inthe morning and have to contribute some oftheir time). Graduation: The program has an automatic exit rule - or would, if the age eligibility criteria were to be enforced. 40. The main issue with Vuso de Leche is that, over time, the program's implementation has deviated from the original idea of providing infants and pre-school children with a glass of milk seven days per week. Many beneficiaries(80 percent in rural areas) receivethe benefit in bulk once per month or even at larger intervals (Figure 2-7), commonly in the form of evaporated milk. This is more expensive and increases the likelihood that it will be shared with others in the household (as documented by the 2003 public expenditure tracking survey) or bartered for other goods or labor, as revealed by discussions in focus groups. Circumstantial evidence suggests that there may be a problem in some places with "phantom beneficiaries" who are manipulated to allow VdL committee members to benefit from their rations. The Comptroller General's Office reportedthat 47 percent of the VDL budget in 2004 was spent on evaporatedmilk, up from 34 percent in2001. The large milk companiesdedicate considerableeffort to securinga share ofthe Vas0de Leche market. 41. The second major issue is the crowding out, over time, of the priority group of children up to six years old by other age-groups.In2003, administrative data shows that only 58 percent of the beneficiaries are aged 0-6 years old, comparedto 69 percent in 2001 (Figure 2-8).Given the clear evidence that chronic malnutrition in Peru is often linked to inadequatecomplementary feeding in the age range from 6 to 24 The datasource is ENAHO2003/2004. 20 months, VdL could potentially have a much bigger nutritional impact if it were even more tightly focused onthat age range. Figure 2-5: Distribution of Vas0 de leche beneficiaries by quintile and area of residence Share o fbeneficiaries fiom a given quintile I Targeting performance of VdL by Area ..... . .. .. ... .. .... . . . .. Rural Urban Source: Author's estimation based on ENAHO 2003/04 Survey, annual sample, INEI. Figure 2-6: Probability of receiving Vas0 de Lecke by age and area of residence Probability of Receiving Vaso de Leche by welfare status, age and area of residence 1 8 .6 1year old, rural .4 -6- +1yearold,urban +6yearold,rural .2 +6year old, urban t13yearold,rural 0 -B- 13year old, urban 0 1 2 3 Welfare ratio: Per capita consumption I Poverty line Basedon ENAHO 2003/2004.INEl Note: Predictedprobabilities after controlling for household characteristics; on the x axis the population i s ranked according to the welfare ratio; 2 vertical lines correspond to total and extreme poverty lines. Source: Author's calculation based on ENAHO 2003/2004 annual sample. 21 Figure 2-7: Frequency of delivery of Vas0 deLeche benefits in rural and urban areas M3st VdL beneficiaries in ruralareas do not receivetheir glass of milkon a daily basis distributionofbeneficiaries bythe frequency of receivingthe Urban Rural Total Daily-Weekly Nbnthly Source: Author's estimationbasedonENAHO2003/2004. Figure 2-8: Trends in the compositionof VdL beneficiaries,2001-2003 j Distributionof beneficiaries by categoly I 2001 2002 jQ Childrenunder 6 Pregnant and nursingmothers Secondary target groupI Source: ComptrollerGeneral's Office, InformeMacro delProgramaVdL, 2003. Source: Author's estimationbasedonENAHO2003/04 Survey, annualsample, INEI. 42. To improve the poverty reduction impact of the program the government should consider the following measures: (i)amend the legislation to exclude children over 6 years old from eligibility; (ii) eliminate the earmarking o f one third of program resources to Lima and Callao; (iii)provide benefits sufficient to cover 7 days per week (in 2003 less than 50 percent of the districts were able to cover 26-30 days of supply per month); if necessary , to make this fiscally feasible, the age range of beneficiaries could be further restricted to those under 24 months o f age; (iv) integrate the program with health and nutrition services, making the glass o f milk conditional on human capital investments such as health visits, child growth monitoring, health and nutrition education. 2.4 Workfare programs 43. Peru has considerable experience with workfare programs that provide short term assistance and also build physical assets for the poor. Two workfare programs, A Trabajar Urbano and A Trabajar Rural, were initiated inthe early days o f the Toledo administration, inresponseto the economic recession 22 of 1998 to 2001. They provide temporary employment for unskilledworkers, at relatively low wages. In rural areas workers were hired for up to 6 months, to renovate social infrastructure and do maintenance work. However, today, A Trubajur Rural has no well-defined programmatic form and i s used as a convenient fiscal umbrella for ad-hoc interventions. 44. InA Trubujur Urbano, the period of assistance is up to 4 months. Program beneficiaries obtain a monthly wage o f S/.300 (or US$91), below the minimumwage o f S/.427 (US$129) per monthg.In2006 A Trubujar Urbuno spent Y.187 million, which is 10 percent o f SSN spending and 0.07 percent of GDP, and reached 57,000 beneficiaries (about 0.2 percent o f the population). The budget for the workfare programs has been decreasing over time". This reflects the fact that the Peruvian economy has now recovered from the economic crisis of 1998 and the recession that followed". However, it would be recommendable to maintain in place a minimumworkfare program capacity to facilitate a quick response duringfuture economic downturns or natural disasters. International experience suggests that it is hardto buildupwell-designedworkfare programs from scratch incrisis situations. Figure 2-9: Distribution of beneficiaries ofA Trabajar Urbano by consumptionquintile, 2003 r A Trabajar Urbano Share of beneficiaries from each quintile 30 . . ..... .. 25 ...... .. 20 15 10 5 I Q1 Q2 43 4 4 Q5 Source: Author's estimationbasedon ENAHO 200312004 45. A Trubujur Urbuno has many good design features, in line with international best practice. The principles for a good workfare program are that it should be counter-cyclical, should be self-targeted and should avoid distortingthe labor market (See Box 4.3). A Trubujur Urbuno does well on these tests, being characterized by: (i) reasonable labor intensity (share of the labor costs inthe total cost o f the projects i s about 60 percent); (ii) targeting mechanisms that combine poverty maps with self-selection, through good paying less than the minimumwage; in 2003 65 percent o f the beneficiary households belonged to Q1 or Q2; (iii) competitive selection o f projects at community level; (iv) emphasis on projects that improve the livelihoods o f the extreme poor, such as roads, schools and health posts; (v) community involvement in project supervision. 46. However, the wage level o f S/.300 (US$91) per month may be set a little too high. It is by far the largest benefit o f any SSN program in Peru. Although it i s 22 percent below the legal minimumwage, it i s more than what workers in the poorest quintile in Lima earn. This sets up incentives to try to remain in the program after the crisis has passed, and there is evidence that many workers on the scheme were already employed beforehand in informal sector activities. This minimumwage was inplace in2005; it was increasedto SI. 500 (USrS152) in2006. lo A Trabajar Urbano created 169,000jobs with duration of4 months from January 2002 to April 2004. A Trabajar Urbano covered0.5 percentofthe urbanpopulation in2003. 23 47. To improve the workfare program, some fine-tuning is suggested: (a) set the level of wage rate below the prevailing market wage for unskilled labor to promote self-selection; (b) align the workfare programs in the framework of decentralizedbasic social infrastructure investments to ensure the quality and maintenance of assets created; and (c) fund employment in social services (caregivers, health promoters) to increasethe participation ofwomen. Box 2-2: Workfare Programs Key Design and Implementation Issues - Public works programshave been demonstratedto work bothinmiddleincome (Chile, Argentina, South Africa) and low income countries (Senegal, Kenya, India and Bangladesh) and not to work in many others. This internationalexperienceoffers severallessons inthe designandimplementationofworkfare programs: WageRate: The key to ensuringself-targeting is settingthe wage rate low enough no higherthan the -- market wage for unskilled manual labor in agriculture or the informal sector during a normal year. While determining the precise level of the wage rate may not be easy, it is better to err on the side of startingwith a wage rate that is too low -ifthere is no demandat the offered wage rate, it can be raised. Settingthe wage rate at a low levelnotonly ensures that the workfare scheme will be well self-targeted, it also maintainsthe incentive to take up regular work when it is available, and it helps ensure that the programcanreach as many of those inneedas possible. Eligibility conditions and other means of rationing should be avoided; ideally the only requirement should be the willingness to work at the offered wage. Ifrationingis unavoidable, (for example if the demand for employment at the wage set exceeds the available budget), then explicit secondary criteria shouldbeused-the programmaytargetedto poor areas, work offered only inseasons of greatest need, the length of employment o f any individual limited, and/or additional targeting devices such as community-based selection of the neediest. Least desirable is rationing and entry determined by foremen or political figures. Women's participation can be enhanced through non-discriminatory wages, the provisionofon-site daycare andadequatelyprivate latrines. Labor intensity: The labor intensity, that is the share of the wage bill in total costs, should be higher than normal for similar projects inthe same setting. There is a trade off between immediate income gains through employment of the poor, and gains to the poor from the quality and durability of the assets created. In a crisis situation, where current transfers to the poor have high weight, high labor intensity is likely to be desired. Illustrativeaverage labor intensitiesrange from 0.5-0.65 percentin low income countries, and somewhat lower (0.4) in middle income countries, though labor intensities often vary significantlyby sub-projects Administration and Implementation: Administering and implementing an effective scheme is hard - involvingthe selection andmanagement of a plethoraof smallprojects over a wide geographicarea and many administrative entities. Ideally, public works schemes require a menu of works that is well- integrated into the local planning process yet elastic in size and timing. This can be difficult in low capacity settings because o f the forward planning and inter-agency coordination needed. In high capacity settings, fitting many small labor intensive projects into the sophisticated and often capital- intensive infrastructure plans of large and middle cities can be hard. Moreover, ensuring that the workfare program is poverty focused is often difficult because of conflictingpressures from alternate target groups(for example,the skilledunemployed). Source: Subbarao (2003). 2.5 FONCODES 48. The future ofthe Peruvian Fund of Social Development (FONCODES) is uncertain. Inthe context of the decentralization process and the social programs restructuring exercise, the Garcia administration has announcedthat it will be closed- but has also announcedthe need for a new entity to support small scale social sector infrastructure. 49. FONCODES was originally established in 1991with the aim of generatingemployment, alleviating poverty and improving access to social services. The government passed special legislation that allowed 24 FONCODES to circumvent the rigidities of planning and executing public investment through line ministries. The innovative modus operandi o f FONCODES included targeting usinga "poverty map" to assign resources to rural nzicleos ejecutores, demand-driven sub-project choice from a menu o f possible investments, and simplified procurement rules. This translated into over 40,000 projects for the construction or rehabilitation of schools, healthposts, water and sanitation systems, rural roads, secondary electrification schemes, and small - scale irrigation works for a total o f more than US$2 billion over 14 years of operation'*. 50. Evaluation studies show that FONCODES expanded access to basic social services and its investments reached the poor districts and poor households within those districts and have had positive effects on school attendance rates for young children. Schools that received FONCODES resources for school improvements experienced an increase in enrollment, which benefited poor children (Instituto Apoyo, 2000, Paxon and Schady 2002, and Schady 2005). FONCODES' geographic targeting is also more progressive than that o f social funds in Bolivia, Nicaragua, Honduras, Zambia, or Armenia (Figure 2-10), largely because of its predominantly rural focus and the use o f poverty maps in selecting the districts. Figure 2-10: FONCODES's geographic targetingprogressiveness,comparedwith other social funds 100 90 ui 80 2 70 'ft 60 50 c $5 v) 40 z 30 20 10 0 -+-Line ++Honduras of equality Arrrenia -w- Bolivia +Zanbia --eNcaragua -I- Peru ''In addition, FONCODES has executed a series of centrally-designed"special" projects which have included, amongothers, a school breakfast program, the distribution of uniforms for school children, shovels for farmers, and motorized canoes. 25 52. The flexibility, discretion and centralization which enabled FONCODES to reach the poor also made it vulnerable to political interference. Schady (2001) showed that between 1991 and 1995, FONCODES' spending was boosted before elections, and that projects were channeled to provinces where the political returns were expectedto be large. 53. Inthe last six years FONCODES underwent important changes to address some of these issues. In 2001, it was relocated within MIMDES (re-branded as the National Plan for Compensation and Social Development), and in 2003, it began decentralization to district municipalities. However, decentralization was conducted in an ad-hoc manner without developing first the fundamental elements of a new model o f decentralized provision of small-scale infrastructure. The accreditation of municipalities for decentralized administration of FONCODES resourceswas complicated and bureaucratic, but provided no guarantee o f adequate local capacity. Some questioned the rationale for any accreditation process, in a context in which many municipalities already administered much larger investment funds from FONCOMUN. FONCODES represented only 2 percent of transfers to municipalities from the Central Government in 2005 (Figure 2-11). The monitoring arrangements for the decentralized framework were also deficient: the management agreements(convenios de gestidn) betweenMIMDESand district municipalities did not include output indicators to monitor program performance. 54. The comparative advantage that allowed FONCODES to be an efficient manager o f small-scale infrastructure projects lay inthe different systems it developed to manage the project cycle: systems and instruments for designing, appraising, disbursing, and supervising projects, and for hiring project designers, evaluators, supervisors, trainers, and so forth. While the introduction of a decentralized project cycle gave district governments a few more responsibilities than before, most o f the know-how has remained in FONCODES zone offices; it has done relatively little to build the capacities of district governments to forge strong partnerships with communities and local service providers for effective small-scale infrastructureprovision Figure 2-11: The distributionof Central Government transfers to sub-nationalgovernments, 2006 Other social programs 11,50 Other transfers,432 anon, 2124 I/Socialprograms:PROVIAS,CommunityKitchen,Foodforwork,Shelters. Source: SIAF Local (MEF), millS/. (2006). 26 55. During the Toledo administration, funding for FONCODES was stabilized and reduced by around 50 percent, partly due to budgetary constraints and partly due to the transfer of the Desayuno Escolar programto PRONAA. The cut infinancing for the social fund was expectedto be compensatedby higher investment by municipalities and regions, in the context of decentralization. However, in 2005-2006 spendingrecovered. Figure2-12: FONCODESinvestmentin real terms FONCODESspending, 1991-2006 2004 constant prices 900 -................................................................................................................ ...... ............................................................................................. 100................................................................................................................. Source: Perli en Nlimeros 1997; Direccidn Nacional de Presupuesto Pliblico (MEF);SIAF-SP (MEF); INEI; BCRP. 2.6 TheJuntos CCTprogram 56. In 2005 the government introduced a conditional cash transfer (CCT) program-Juntos-that provides US$30 per month to families with children under the age of 14 or pregnant mothers living in extremely poor communities. Receipt of the transfers is subject to compliance with conditions regarding school attendance, obtaining birth certificates and identity cards, and attending nutrition education sessions and use of preventivehealth care. The program startedwith a pilot phase covering about 70,000 households in the poorest 110 districts of the departments of Apurimac, Ayacucho, Huancavelica, and Hubnuco.In2006 the program expandedinto five more departments, covering 200,000 households in320 rural districts and inner-city neighborhoods, in nine of Peru's 27 departments. By the end of 2007 the program will reach 400,000 households in over 638 districts across 14 departments, or about 2.5 million individuals representing9 percent of total population. The total cost of the program in 2006 was W.173 million, which is 9 percent of SSN spendingand 0.06 percent of GDP, making it the fifth largest program after FONCODES, Vas0 de Leche,A Trabujar Urbano andthe school feedingprogram. 27 rigure 2-13: Juntos targeting process:geographic selection, householdassessment, and community validatii n Household level Cornrnunity validation Select districts based on: Identify poor households with Validate lists. Exclude households 1. Extreme poverty 1. Children under 14 that have: incidence 2. Pregnantllactatingmothers 1. Professionalsearningmonthly 2. Povertygap wages 3. Unsatisfied Basic Needs 2. Pensioners 4. Chronic malnutrition 3. Shop/businessowners 5. PoliticalViolence 4. Alimony recipients 5. Vehicles 6. A certain number of livestock 7. Residents living in other cities 8. Relatives of local authorities 57. TargetingofJuntos, is done inthree steps, comparable to similar programs in Mexico and Colombia: (a) geographic targeting, (b) household assessments based on a proxy means test, and (c) community validation (Figure 2-13): (a) Geographic targeting. In the first stage all 1828 districts in the country were ranked using a composite poverty index13.The departments selected for the pilot phase (Phase 1) o f the program -Apurimac, Ayacucho, Huancavelica, and Huanuco-haveamongthe highest incidences of poverty in Peru (Figure 2-14). The five departments included in 2006 (Phase 2) present a more mixed picture. Puno and Cajamarca have a high incidence o f extreme poverty; while Ancash, Junin, and La Libertad are close to the national average. The program excludes departments that are sparsely populated or heterogeneous in terms of poverty, to avoid dispersion o f effort and economize on administrative costs. (b) Household assessment. Juntos is the first program in Peru to select beneficiaries using a proxy- means test14. Eligibility is determined using a household questionnaire (Ficha Socioecondmica l3 The index usedthe following component indicators: (i) extreme poverty rate (ii) poverty gap based on the total 2000 poverty map; (iii) percentage of households with more than two Unsatisfied Basic Needs (UBNj r u n n i n g water, electricity, schools, health services-- based on the 1993 Census; (iv) chronic malnutritionrate among children 6 to 9 using the weight and height census of 1999;and (v) the percentageof population centers affected by political violence based on a special census done through the Resettlement Support Program (Programa de Apoyo a1 Repoblamiento). Although some of these data sources are rather old, the ranking of districts using data fiom the ENAHO 2004 for total and extreme poverty yields similar results. It should be noted that the f r s t 70 districts in the pilot phase were selectedusing a composite index that excluded the extremepoverty rate. 14 The algorithm used by Juntos to identify the poverty status o f households i s detailed inAnnex 2. It is based on ENAHO 2001-2004 anduses a logistic regressionwith 8 predictors. Another program -Seguro Integral de Salud, a health fee waiver for the poor -plansto use a similar proxy-means test algorithm. 28 Unica FSU) that collects data on demographic composition (including the presence o f children - under 14 and pregnant mothers), education, dwelling characteristics, durable assets, livestock, and agricultural equipment. A score is calculated for each household and i s compared to the pre- established cut-off point. (c) Community validation. A community meeting i s held to validate the list o f beneficiaries. At this stage, the following households can be excluded: (i) with members employed in the formal those sector, pensioners, recipients o f alimony, business owners, money lenders, and relatives o f local authorities; (ii)owners o f livestock, cars or other large assets. The pilot phase showed some flaws inthis phase o f the process. Many are afraid to denounce others, while some believe the program representative; and promoters sometimes dominate - rather than facilitating - the discussions. should include everyone, to avoid dividing the community. The meetings are not always Figure 2-14: Departmentalpovertyand extreme poverty rates,2006 National poverty rate Madre de Bios 0 0 100 200 300 400 500 600 700 800 900 1001 Percentage of population .-- Poverty 8 E t r e m poverty Source: ENAHO 2006 (INEI). 58. Permanence in the program. A beneficiary remains in the program for four years; however this period is likely to be extended given that the exit rules are not yet defined. This could result in a progressive decline in targeting accuracy as households' poverty status changes, making it hard to accommodate those who have fallen into poverty or to exclude those who are no longer poor. Some CCT programs have systematic recertificationprocedures to deal with this issue. 29 the consumption o f recipient household^'^. While this is eight times greater than that o f the existing food 59. BeneJit level. The US$30 per month benefit given byJuntos is adequate - equal to about 25 percent o f programs (3.5 percent on average), the benefit level is comparable with that in Brazil's Bolsa Familia and Mexico's Oportunidades which give benefits o f about 20 percent o f mean household consumption. The value o f the Juntos benefit was not set (as in some other countries) based on the estimated costs o f complying with the program's conditionalities. Rather, the program opted for a benefit that was simple to administer and related to the extreme poverty line o f $1 a day. The transfer is the same, regardless o f household size or the number o f children. In this regard, Juntos differs from other CCT programs in Mexico, Brazil, Colombia, whose benefits vary according to the number o f children in the family and other criteria. 60. Conditionality. Juntos has adopted a relatively complex and ambitious set o f conditions for program eligibility, apparently in an effort to give political legitimacy to the program. Beneficiaries must sign an agreement called the Convenio de Compromiso Voluntario. Children under 14 must be enrolled in primary school with an 85 percent attendance rate and must visit the health center periodically. Mothers must attend pre-natal and post-natal clinics. Adults must have national identification cards and ensure that their children have birth certificates (Table 2-4). The program has a long way to go to improve arrangements for monitoring compliance. The evidence from the field suggests that beneficiaries are required to personally collect certifications from schools and health posts on a periodic basis, generating a rush to collect official stamps ("sellos") and interfering with the normal production o f services. This emphasis on filling forms rather then on providing the services can be detrimental for the impact o f program. A more sophisticated system would establish direct communication between Juntos and the service providing entities, in which institutional registers were used to verify compliance with program conditionality. 61. It would be recommendable after the pilot phase to consider simplifying the conditionalities and eliminate those which are not likely to have a major impact. For example, primary enrollment is already very high in Peru. According to the 2005 Census, 92 percent o f children aged 6 to 14 years in rural areas in the first selected 110 districts are already in school. It would make more sense to focus on increasing secondary enrollment, given that the ratio o f children o f secondary school age who are enrolled in school to the population o f the corresponding secondary school age is only 69 percent16. To encourage children in remote locations to attend secondary school a more substantial transfer could be envisaged, to cover commuting and/or boarding costs. At the same time it would make sense to strengthen the co- responsibilities related to nutrition, given that almost 35 percent children zero to five years old in the areas where the program operates suffer from chronic malnutrition. In July 2007, the Juntos Board decided to move in this direction, re-focusing Juntos on improving nutritional outcomes. This initiative was later endorsed by the national nutrition strategy, CRECER. 15 On average, Juntos benefit representsan increase of 25 percentof the consumption o f recipient households. The average consumption of extreme poor rural households in the 4 departments where Juntos operates is S/. 75, and their householdsize is 5.4 members. Adequacy was determinedas (100/(5.4*75) = 0.25. 16 According to World Development Indicators, World Bank, 2006. 30 Family co-responsibilities Education Enrollment inprimary school o f children 6 to 14 School Attendance: 85 percent o f the time Nutrition and identity Families with children 6 months to 3 years participate inthe nutritional program PACFO (papilla and education sessions) Health . Chlorine Note: These conditionalities were revised in July 207 to focus more on nutrition actions for pregnant women and childrenunder five. 62. Institutional arrangements. The institutional arrangements aim to involve the education and health sectors in decision making, to increase transparency, and isolate the program from political manipulation. The program is controlled by the Presidenciadel Consejode Ministros. An Executive Council chaired by a President and including the ministers of Education, Health, Women and Social Development, Finance and representatives of the trade unions and civil society approves the action plan, supervises and evaluates the program. The program has the status o f an executing agency with its own budget and is responsible for the main operations. A supervisory body including representatives o f the church, private sector, regional and local government, and the Mesa de Lucha Contra la Pobreza is responsible for social monitoring and transparency. 63. Supply side conditions. In the pilot phase, Juntos assigned a third o f its budget to strengthen the supply or relevant services and the program places a heavy emphasis on coordination at local level with the education and health sectors. This is a welcome recognition that a demand-side program to enhance human capital among the poor will only work ifthe supply-sideis inplace. However, to ensure that this i s effective it i s important to study the supply gap rather than giving the money to the health and education sector as fungible resources. However, to date, the line ministrieshave proved unable to use the additional funds suppliedbyJuntos effectively to improve supply conditions inJuntos areas (executing a very small proportion of the total). The government is presently analyzing options to improve this mechanism, as part o f the results basedbudgetingreforms. 64. Targeting accuracy. The program's targeting outcomes and poverty impact was modeled, using the database of the ENAHO 2004 survey." In the departments included in Phase 1, the algorithm correctly selects 66 percent of the extremely poor households as Juntos beneficiaries. However, given that 56 percent of the population of these four districts i s extremely poor, even a lottery system would select 56 percent. So the apparent success of the algorithm is largely due to the quality of the geographical targeting. This raises the question of what value is added by the costly and problematic process o f household-level data collection. It is also noteworthy that the precision of the algorithm diminisheswhen 17 For each household, a score was estimated usingthe officialproxy means test presentedinAnnex 2. Households with scores above the program thresholdwere "enrolled" inthe program, if located in one of the 110 districts fiom Apurimac, Ayacucho, Huancavelica, and Huhnuco where the program was implementedin 2005 (Phase 1). The same method was used to identify those households enrolled in 2006 fiom another 210 districts fiom the departmentsof Puno, Cajamarca, Ancash, Junin, and LaLibertad(Phase 2), as well as for the whole rural area. 31 the program expandsto districts with lesshouseholds inextreme poverty - ifexpanded to all rural areas, it would only correctly identify half of the extreme poor as beneficiaries (Table 2-5). IfJuntos i s to maintain such a system in place, it should consider developing a more sophisticated algorithm, using more indicators (see Box 2-3). Modeling results suggest that a more sophisticated algorithm could result in better coverage of the extreme poor, and less funds leaking to non-extreme poor beneficiaries." It should also be noted that inurban areas, the government is implementing SISFOH ("Sistema de Focalizaci6n de Hogares"), a proxy means test which aims to identify households in extreme poverty. Due to the greater variance inhousehold income, the case for investingin such a system is stronger inurban areas. Table 2-5: Juntos PMT Algorithm: Inclusionand ExclusionErrors '*For example, adding two predictors known for their high correlation with poverty status to the Juntos algorithm (household size and departmental dummies) would increase the share of extremely poor households covered by the program from 66 percent to 73 percent, reduces the fraction of extreme poor not covered by the program fiom 49 percent to 40 percent, and reduces the leakage o f funds toward non-extreme poor households by about Sl.20 million per annum (Figure 2-15). Another good predictor i s the rate of chronic malnutrition at the level o f the census unit. (The results ofthe regressions are presented inAnnex 4). 32 Box 2-3: Improving Juntos household-level targeting instrument The predictive power of the proxy-meanstest algorithm can be improved by: Using a larger set of predictors to predict poverty status. Juntos' algorithm uses only 8 variables to predict the poverty status of a household. In contrast, other CCT programs from Chile, Colombia, and Costa Rica reviewed by Castaneda and Lindert (2005) use between 13 and 16 variables. The FSU questionnaire includes many other factors strongly associated with poverty status, such as household size, location, the level of education of the adults, crowding, and house or livestock ownership. By not taking such variables into account, the official algorithm looses precision (see Annex 4). Calibrate the model on extremepoverty, not on totalpoverty. Using moreflexible prediction models differentiated across regions or departments. Avoid using data reduction techniques. Although Juntos collects a wide range of household indicators though the FSU questionnaire, it uses only a small fraction of them to predict poverty status. INEI arrived at this parsimonious model with 8 variables after a process of data reduction - instead of using the raw information from the questionnaire; it grouped that information into fewer categories which, in turn, are used as predictors. However, some precision is lost during the process of data reduction; using the original variables will generate abetter fit. Collect more relevant information in the FSU questionnaire. Juntos' FSU questionnaire collects a total o f 36 variables, lower than in the CCT programs from Chile, Colombia, Costa Rica or Mexico, which collect between 50 and 115 variables. Among the variables used to predict poverty in other LAC programs, which are not collected in the FSU, are the occupation ofthe householdhead or formal incomes. Note: Simulation based on ENAHO 2004 (INEI) and Juntos administrative data. Phase 1 correspondsto the 110 districts in the departments of Apurimac, Ayacucho, Huancavelica, and Huanuco; Phase 2 corresponds to another 210 districts in the departmentsof Puno, Cajamarca, Ancash, Junin, and La Libertad. Source: Authors' estimation based on ENAHO 2004. Figure 2-15: Reduction in Juntos' Inclusion and Exclusion Error with an Improved PredictionModel 100- I I ~ 70 60 Z 50 40 30 20 10 inclusion correctly classified as extreme poor exclusion extreme poor not cowred ~ L--_ _ _ _ _ ~ Juntos m ImprobedModel -, I Source: Authors' estimationsbased on ENAHO 2004 (INEI) 65. Juntos' impact on extreme poverty is likely to be large, due to the size of the benefit and good targeting accuracy. The modeling exercise suggests that in Phase 1 and 2 Juntos could reduce the extreme poverty gap by 19 percent in the departments where it intervenes; and this will reduce the national poverty gap by 2 percent in Phase 1 and by 5 percent in Phase 2. Ifthe program were expanded to all rural areas, the extreme poverty gap would fall by 14 percent. (Figure 2-16 and Table 2-6). These 33 impacts could be roughly doubled if the precision o f the targeting system could be perfected. They could also be increased by distributing benefits on a per capita basis rather than a household basis and redistributing program resources accordingly. Figure 2-16: Juntos' Simulated Impact on the Extreme Poverty Gap Phase 1 Phase 2 Phase 3 Before Juntos After Juntos I _ _ ~ ~ Note: Phase 1 covers 110 districts in Apurimac, Ayacucho, Huancavelica, and Huanuco; Phase 2 includes Phase 1 districts and another 210 districts in the departmentsof Puno, Cajamarca, Ancash, Junin, andLa Libertad. Source: Author's calculations basedon ENAHO data (200312004). 34 Poverty gap Povertyseverity Phase2: 320 Districts in Poverty headcount 0.523 0.470 Poverty gap 0.165 0.134 Povertyseverity 0.070 0.053 Phase3: All Rural Povertyheadcount 0.402 0.360 Povertygap 0.117 0.098 Povertyseverity 0.048 0.038 Note: Phase 1 corresponds to the 110 districts in the departments of Apurimac, Ayacucho, Huancavelica, and Huanuco; Phase 2 includes Phase 1 districts and another 210 districts in the departmentsof Puno, Cajamarca, Ancash, Junin, and LaLibertad. Source and methodology: Author's calculations based on ENAHO data (200312004). Beneficiary households(in the third column) are those which would qualify for Juntos participation based on their poverty status, as registeredinthe survey. 66. Juntos' potential to impact chronic malnutrition i s also very high. In Peru there are over 500,000 chronically malnourished children age zero to five. Of those, almost 40 percent live in districts where Juntos operates. The prevalence o f chronic malnutrition inthese areas is as highas 36 percent - compared with 13 percent in areas not covered by Juntos. Ifall chronic malnutrition were eliminated in the Juntos districts, the national stuntingrate would be reduced by 38 percent; even if half of it were eliminated, the national rate would be reduced by almost 20 percent. To achieve this effect, Juntos could strengthen incentives for families to participate in growth promotion programs, to seek micronutrients and other services offered by the health system, which are important for nutrition results. Engaging families into a growth measurement and counseling system would provide them with information about their child's nutritional status, and how to improve it. The income transfer will also help families to afford better nutrition for their children. (Figure 2-17). Table 2-7: Chronic malnutrition rate, children zero to five (Weighted statistics, weights =population of children less than 5years old) Juntos Areas Average Min P25 Median P75 Max Areas not in Juntos 13.2% 1.6% 6.2% 9.5% 18.7% 63.2% Juntos Phase 1 & 2 (320 districts) 36.4% 6.5% 28.4% 36.3% 44.3% 61.7% Expansion Juntos 2007 35.2% 8.2% 29.7% 35.4% 41.3% 60.1% Total 17.6% 1.6% 6.6% 14.7% 25.4% 63.2% Source: Authors' calculation basedon SIEN2005 and Juntos administrative database. 67. To achieve a strong nutrition impact, it is important for Juntos to continue focusing on rural areas with high malnutrition rates and to improve the design o f co-responsibilities to include a minimum o f important elements for diminishing malnutrition (for example, periodical growth monitoring visits, participation in counseling and health and nutrition education, pre-natal controls, institutional births, etc). At the same time, it is important to improve the quantity and quality o f primary health care services including guaranteeing a minimum protocol o f monitoring child growth until age two, providing 35 micronutrients and food supplements, improving the training of health personnel and community agents, improving the growth control and measurement systems. Figure 2-17: Juntos' SimulatedImpact on ReducingChronic Malnutrition :E60.35 -m 0.30 ._ 5 0.25 .... . gc 0.20 ,E 0 0.15 ._ 0 c 2 0.10 u a? 0.05 0.00 scenario 1: 100% of scenario 2: 80% of scenario 3: 50% of malnourished malnourished malnourished children in Juntos children in Juntos children in Juntos Source: Authors' calculationbasedon SIEN2005 andJuntos administrativedatabase 68. Monitoring and Evaluation. Juntos also needs to develop a coherent Monitoring and Evaluation system to ensure that good evidence is produced about the program's results and fed back into the decision process as the program expands. This should include a rigorous evaluation, to help decide whether the program should be confined to rural area only or widely adopted. The M&E system should include the following elements (a) outcome/impact evaluation (did the conditional cash transfer achieve the desired goals?); (b) implementation analysis (Are all parts of the program inplace? Are the health and education interventions implementedin all regions?); and (c) process analysis (How is the program being delivered? What benefits do participants receive? How do they get them?). The development o f a monitoring and evaluation system covering all these elements i s one of the priority challenges facing Juntos. The monitoring system should include the mothers DNI and the unique personal identification numbers (CUI) which are being included inthe birth certificates all o f children under new rules issuedby RENIEC . 36 3 THE REDISTRIBUTIVEAND POVERTYIMPACT OF PERU'S FOOD PROGRAMS 69. This chapter measures the effectiveness of food programs in redistributing income and reducing poverty inPerulg.It investigatescoverage and benefit adequacy, presents a benefit incidenceanalysis, and simulates their impact on poverty and inequality. The findings show that despite pro-poor distributional patterns, the small unit subsidies limit the redistributive, poverty, and inequality impacts of even the best targetedprograms. Their impact on reducingmalnutrition is also modest. 70. The chapter is organized in four sections. Section 3.1 presents the criteria used to assess program effectivenessand efficiency. Section 3.2 looks at coverage and the level of the benefits. Section 3.3 deals with targeting. The overall impacton poverty and malnutrition is summarizedin section3.4. 3.1 Criteriafor Assessing the Redistributive Role of FoodPrograms 71. In contrast to public spending on health and education-services with substantial externalities that mayjustify their universalprovision-SSN transfers are privately consumedgoods which should normally be targeted to the poor households or areas. The main rationale for safety net transfers is promoting equity, so they should be focused, especially, towards the extreme poor and it is important to assess how much they contribute to poverty reduction. This chapter focuses on the food programs, the largest component of the SSN sector in terms of budget and numbers of beneficiaries.Their impact on poverty reduction is assessedusingthree complementary criteria: Whatshare of the target group receives theprogram? To be effective, a program should cover all or a large share of its target group. To capture this dimension of programperformance, we estimate what proportion of individuals/householdsinthe target group ineach quintilereceivesthe program. How adequate is the benefit offered by the program? Substantial coverage of the target group is a necessary, but not sufficient condition to make a program effective. Using a public health analogy, to be effective, programs should provide adequate treatments. The subsidy channeled by these programs is comparedto the pre-transferconsumptionof the beneficiaries, especially the poor beneficiaries. How well are the program resources targeted to the poor and extreme poor? To assess how efficiently transfers are focalized, a benefit-incidence analysis i s undertaken*'. Estimates are presentedof the share of the subsidy that reachesthe extreme poor (Ql) and other quintiles. Previous studies have analyzed the incidence of food programs in Peru based on the distribution of program beneficiariesacross quintiles. However, this approachis only valid if similar benefits are receivedby everyone. Peru's food programs are complex, with variations in the number of rations consumedper month and the amount of subsidy given (some get free meals, while others pay a contribution, which differs from program to program). In addition, subsidies differ substantially across departments. The distribution of public spending across quintiles in this study is estimated by adjusting the program participation with the take-uprate, the level of subsidy from each region, and the subsidization rate. The analysis covers the seven largest SSN programs for which there are data in the ENAHO 2003 The chapter does not evaluate other possible impactsof food programs. Most programsinPeruhave a multitude of objectives. For example most food programsaim enhancingfood security, improvingnutrition, improvingschool attendance and academic performance, offering support to local farmers and domestic producers, developing social capital, or promoting community based organizations. However many times these objectives are not clearly specified and measurablegoals are not established. Given that most programs claim that their main objective is to helpthe poor children andtheir families, it seemsjustified to assess how they achievethis objective. *' Benefit incidence identifies who is benefitingfiom public services, and describesthe welfare impact on different groups of people or individual households of government spending (Demery 2004). It does this by combining informationabout the unit subsidytransferredby a program(obtainedfrom administrative data) with informationon the participationinthe programs(obtainedfrom householdsurvey data). 37 survey, which account for 80 percent of SSN beneficiaries and spending: (i)Vas0 de leche; (ii) Comedores populares; (iii) Desayuno escolar; (iv) Almuerzo escolar; (v) Comedor infantil; (vi) A Trabajar Rural; and (vii) A Trabajar Urbano. Full details are reported inAnnex 1. 3.2 Foodprogram coverage and benefit levels 72. Compared to the LAC region, Peruvian food programs have high coverage, but transfer extremely small benefits (Figure 3-1). Over 40 percent of Peruvian households benefits from some type o f food program. The coverage o f Peru's food programs, alone, surpasses the overall coverage of social assistance programs in Argentina, Brazil, Chile, and Columbia, all of which have relatively well-developed social assistance systems. However, Peru's spending on such programs i s lower than in other countries in the region, so that the average value of transfers per person i s between 5 times and 20 times smaller in Peru than in other countries (Figure 3-1; right panel). The value o f the food programs benefits i s less than two percent of the total consumption of beneficiary households. The most popular program, Vas0 de Leche, transfers, on average, less than US$2 per month per beneficiary, the equivalent of 5 percent o f the extreme poverty line of S/.115/month (about US$1.2 per day). Figure 3-1: Coverage and benefitlevels of Peru's SSN Programscompared with LAC Coverage A! Social Assistance - Average Value of Transfers (uS$pFp)- Social Assistance Share of Papulation N m g in Households Receiving at least one Benefit .......................................................... 40t............ .............................. ~~ ~~~~~ ..... g !. 30 40% aL 30% = 20 1 20% 10 10% 0% Peru LAC Argentina Brazil Chiis Columa Guatemla Peru LAC Argsnbna Erarii Chile C o b h a Guatemla Source: Based on Lindert, Skoufias, and Shapiro (2005). 73. Peru scores high on coverage of the poor with SSN programs, compared to other LAC countries, and this i s almost entirely due to two programs: Vas0 de Leche and Desayuno Escolar (Table 3-1). All other programs cover less than 5 percent of the population. Consolidating some of these tiny programs may economize on scarce human and financial resources, and help to capture economies o f scale. 74. Rural residents are better covered than urban residents. This i s a positive feature o f the Peruvian SSN system, given the higher incidence of poverty and extreme poverty in rural areas (Table 3-1). However, reaching the remaining 22 percent of extremely poor children in rural areas with no access to feedingprograms could prove a challenge, given the country's complicated geography. 75. The best way to analyze how effective the food programs are in reaching their beneficiaries is to analyze the coverage relative to the target group o f each program. For example, the primary target of Vas0 de Leche is children under age seven, Almuerzos and Desayunos Escolares targets school-aged children inpoor areas, and ComedoresPopulares targets all poor households. 76. With the exception o f Vas0de Leche and DesayunoEscolar, which cover 44 percent and 17percent o f their target groups, respectively, all other food-based or workfare programs are not effective in covering their target group (Table 3-2). To cover its target group o f "all poor households," the budget of Comedores Populares would have to increase 10-fold-clearly an unfeasible scenario. If programs are small compared to their target groups, they fail to accomplish their objectives, and the selection of beneficiaries is prone to political manipulation. 38 Table 3-1:Household-level coverage of Peru's food programs Note: The target group for Vas0 de Leche is childrenaged0 to 6 andpregnant and lactating mothers; for Comedor Popular it is poor families; for Desayuno andAlmuerzo Escolar it is childrenaged 7 to 13. The category "other" includes PACFO, PANFAR, and WAWAWASI which target younger children. Source: Author's estimatesbasedon ENAHO 2003/04 Survey, MEI. 77. The main food programs transfer very small amounts relative to household consumption. On average, the subsidy represents less than 2 percent o f the consumption o f a beneficiary (Annex 1 describes how the subsidy was estimated). This varies little with the type o f program. The most "generous" program is the comedorpopular for which, on average, the subsidy represents 2.2 percent of the consumptiono f a beneficiary. Inrelative terms, such transfers are more important for the poorest. For example, the subsidy from Vas0 de Leche represents 3.9 percent of the consumption o f those in the poorest quintile, and only 0.7 percent o fthose from the upper quintile.This reflects the fact that per capita consumption in the poorest quintile is one ninth of that o f the richest quintile. However, even for the extreme poor, subsidy from the four largest food programs is only a tiny fraction of their consumption deficit (the gap betweentheir consumption and the poverty line). 39 Programs Total Consumptionquintiles Povertystatus Area I I1 I11 IV V PX PNX NP Urban Rural Vas0 de Leche 2.0 3.9 2.2 1.8 1.3 0.7 4.0 1.8 1.3 1.5 2.7 Comedor Popular 2.2 4.5 2.4 2.6 1.3 1.4 4.3 2.5 1.4 2.0 2.7 DesayunoEscolar 1.7 3.3 1.6 1.1 0.6 0.5 3.2 1.5 0.8 1.o 2.2 Almuerzo Escolar 0.9 1.7 0.9 0.6 0.4 0.1 1.7 0.8 0.4 0.6 1.0 Total 1.8 3.5 1.9 1.6 1.2 0.7 3.5 1.7 1.1 1.4 2.3 3.3 Food Programs TargetingPerformance ' 78. Peru's food programs are better targetedthan is sometimes thought. Many previous analyses of the food programs in Peru have pointed to inadequate targeting and large leakages toward unintended beneficiaries (Vasquez 2005; Alcazar 2003). Using household-level data, the present study, however, finds that a large part of benefits go to the poor (either moderateor extreme). In 2003, 30 percent of total spending of W.816 million reached the poorest quintile and another 24 percent reached the 2ndpoorest quintile (Table 3-4). Peru's food programs also have better targeting performancethan other countries in LAC, outperforming similar programs inColombia and Guatemala(Figure 3-2). This comparatively good performance is due mostly to high-quality geographicaltargeting. However, targeting performance varies significantly from one program to another; desayuno escolar (with 71 percent of benefits going to the bottom two quintiles) and Vas0 de leche (with 55 percent) are better targeted than the comedores populares (which achieve only 41 percent, almost exactly pro-ratawith the population share). 40 Table 3-4: Targeting accuracy: Share of public transfers reachingeach quintile By Quintile Spending Total Q1 Q2 Q3 Q4 Q5 mil SI. Vaso de leche 100% 29% 26% 22% 17% 7% 356 Comedorpopular 100% 20% 21% 28% 22% 9% 98 Desayuno escolar 100% 44% 27% 18% 8% 3yo 117 Almuerzoescolar 100% 45% 33% 15% 6% 1Yo 24 Comedorinfantii 100% 43% 32% 14% 9% 3yo 26 Other 195 Allfoodprograms 100% 30% 24% 23% 16% 6% 816 Source: Basedon Lindert, Skoufias, and Shapiro (2005). Figure 3-2: Targeting Derformanceof selected Latin American food Drograms - -. . - I Benefit hcidence of SchoolFeeding Rograms Benefit Incidenceof Conplerentary Food Rograms % of Benefits Receivedby the PoorestTwo Cluintiles % of BenefnsReceivedby the FoorestTwo Quintiks 80% 180% .. ................~ ...............~ ....... . . ~ ~~.~ 70% .. . . . . 70% 60% . .. .. 60% 50% 50% 40% 40% 30% 30% j102 20% 20% 10% 10% 0% 1 I Colombia I Guatemala I R r u 1 mru Source: Basedon Lindert, Skoufias, and Shapiro (2005). 79. Nevertheless, Peru's feeding programs are less well-targeted than cash transfer (CT) programs in LAC, in Europe and Central Asia, and in OECD countries (Figure 3-3). There are two factors that might explain this. First, by design, the relative income threshold used to target benefits i s substantially higher in Peru's SSN programs than is the norm in cash transfer programs in OECD, ECA, and some LAC countries. Peru's programs target the poorest 50 percent o f the population. Incontrast, most CT programs in OECD and ECA countries target the poorest 10 percent of the population, while the other LAC comparators focus on the poorest 20 percent. Second, few programs in Peru use individual or household- level targeting methods, which are normally associated with better targeting performance, such as means or proxy means tests. 41 Figure 3-3: Targeting Performance of selected Cash Transfer programs and Peru's food programs - - - - - Share of ResourcesReaching the PoorestQuintile 1 100 ~.~ go ..~........... 80 70 60 50 40 30 20 10 0 I LAC Countries 1 ECA Countries 1 USA Peru Food-basedRogram - I I ~ Note: GMI=Guaranteed minimumincome, TANF= Temporary assistance for needy families, Source: Based on Lindert, Skoufias, and Shapiro (2005); Castaneda and others (2005) for L A C countries and US; and Tesliuc and others (2005, forthcoming) for the ECA region. 3.4 Food Programs Impact on Poverty andMalnutrition ' SO. Impact on reducingpoverty. To estimate the impact on poverty reduction, we simulate the level of poverty that would prevail if food-based programs were suddenly stopped. The results are presented in Figure 3-4.21Despite reasonable targeting and extensive coverage, food programs have only a modest impact on poverty and inequality. Inthe absence of food programs the poverty headcount would increase from 54.7 percent to 55.2 percent; this change is not statistically significant. The impact of food programs is slightly larger on those in extreme poverty; in the absence of the programs the extreme poverty headcount would increase by 4 percent. Food-based programs also have little impact on inequality (they reduce the consumption Gini by only 0.5 percent). The likely reason for this disappointing outcome is that the main programs distribute very small benefits which have little effect on the consumption deficit of their beneficiaries. As noted above, the biggest program, Vas0 de Leche, transfers less than US$ 2 per monthper beneficiary the equivalent of 5 percent o fthe extreme poverty line. - 21 The counterfactual consumption is estimated as observed consumption minus the value o f the transfer. This simulation does not take into account the behavioral response o f households to the reduction in public transfers. Confronted with a reduction inpublic transfers, households might be expected to react by working more, mobilizing other productive assets they have, and only as a last resort, by reducing their consumption. However, given that food-transfers are relatively small compared to the consumption o f the recipient households, the counterfactual consumption usedhere is probably close to the true one. 42 Figure 3-4: The Simulated Impact of SSN Programs on Poverty reduction Poverty Gap with and without Food Programs I I 40% I I I I 35% I 30 5% 30% 290% ~ 25% 20% 15% 10% 64% 70% 5% 0% I All population Beneficiaries All population Beneficiaries ~ I I P o ~ r t yGap Extreme Powrty Gap I __.__ ____ I Powrty gap in 2003 Simulated Powrty gap without Food programs 1 -I Source: Author's estimates basedon ENAHO 2003/04 Survey, INEI. 81. Impact on malnutrition. There is no evidence that Peru's food programs are contributing much to the reduction o f non-monetary dimensions o f poverty in Peru. Malnutrition remains high, especially among the poor. In 2000, 26 percent o f children still suffered from chronic malnutrition (deficient height- for-age), and this rose to 47 percent o f the children in the poorest quintile. 82. One reason why so much malnutrition can co-exist with so many food programs is that few of the programs target children under 2 years o f age, which i s the "window o f opportunity" where malnutrition can be combated effectively. School feeding programs intervene too late in the life-cycle. Moreover, Peru devotes a disproportionate amount of its food and nutrition program budget to purchasingand distributing food, instead o f supporting strategies that focus on prevention, promotion, and sustainable behavior-based solutions. It is well-established that food transfers are unlikely to have a significant nutritional impact, when little attention is given to feeding and health care practices, to micronutrient deficiencies, and to the rigorous monitoring o f individual children's growth curves, to identify at-risk individuals and devise adequate responses. 83. A handful o f nutritional programs offer more meaningful benefits but cover only a tiny fraction o f their target group hence do not make a dent in the national malnutrition rate.22 One program-the Supplementary Food Program for Groups at Risk o f Malnutrition (PACFO)---focuses on the under-two age group in rural areas. PACFO's impact evaluation revealed excellent targeting performance and relatively good nutrition outcomes (Maximize and Cuanto 2003). The program, however, has a small budget and operates only in eight departments, which implies that many poor children at risk o f malnutrition are not covered. In addition, it appears that the transfer o f the program from the National Institute o f Health to PRONAA has weakened the program's health education component. 22The rich literature on the performance of Peru's food programs is summarizedinAnnex 2. 43 4 THEWAY FORWARD 4.1 Key issuesfacing Peru's SSNsystem 84. Due to the lack of a clear accountability structure, and political economy factors, most of Peru's traditional SSNprograms contribute little to reducing poverty or malnutrition. There is a generalized "low quality equilibrium" in the social assistance and nutrition sector, rooted in the lack o f clear objectives and measurable goals. This has opened the way for the "capture" o f programs by interest groups, and for their political manipulation. As a result o f such political economy dynamics, there are major problems with the effectiveness o f decentralized programs such as Vas0 de Leche and Community Kitchens, and few o f Peru's SSN programs reflect best practice intheir technical design. For instance, the nutrition interventions generally lack a focus on children under two years old, where they would have the greatest effect. Many programs also have overlapping beneficiaries and the prevalence o f small programs leads to relatively high administrative costs. The Government has now begunto address this issue through the social program optimization process, which has already managed to reduce the number o f social programs in Peru to a third o f the previous level, and through plans to develop a Single Beneficiary Register for the social programs. 85. There is a vicious circle of poor performance and low funding. SSN spending is low compared to national needs and regional standards. Peru spends 0.7 percent o f GDP on the SSN, substantially below the region's average o f 1.3 percent o f GDP. The level o f spending i s also small compared to the pre- transfer poverty gap (6.8 percent o f GDP), but represents slightly more than half o f the pre-transfer extreme poverty gap (1.3 percent o f GDP). Without doubt, the mediocre performance o f Peru's social assistance programs has reinforced the low overall level o f social assistance spending. MEF has become skeptical o f approving "more o f the same", setting up a vicious circle o f poor quality and insufficient funding. 86. Although they are generally targeted on the poor, most SSNprograms attempt to cover too many beneficiaries, resulting in benefits that normally are too small to make a diference. Most SSN programs aim to target poor households, that is, the bottom 44.5 percent o f the income distribution in Peru23.But when this broad target group i s coupled with low overall spending, the inevitable result is very small subsidies for a large number o f people. This limits the SSN's impact on poverty and inequality. The food programs illustrate this well. In 2003, they covered an estimated nine million beneficiaries (a third o f the total population) with an average transfer o f only $2/month, equivalent to only 3 percent o f the per-capita income needed to surpass the poverty line or 6 percent o f that needed to surpass the extreme poverty line. 87. Peru 's spatialpoverty maps are used below their potential. The country has a long history o f using poverty maps to target expenditures to poor areas, and geographical or categorical targeting i s presently used to channel about 80 percent o f the SSN spending24.The first map was produced as early as 1977, based on an index o f unmet basic needs (UBN). Since 1991, FONCODES developed district-level poverty maps based on UBN techniques, and used them to redirect social expenditure toward the poor areas. After 1996, the Ministryo f Economy and Finance (through the statistics institute, INEI) developed improved poverty maps based on census and survey data to be used by to allocate pro-poor spending toward the poorest rural areas. However, although the legislation requires all public spending on SSN to be allocated using a pro-poor allocation index, in practice, the most spending is distributed based on .23 Following normal practice, income estimations used throughout this paper are based on ENAHO data for householdexpenditure,which is consideredthe best proxy for income insurvey data. 24 A recent international review of the accuracy of different targeting methods by Coady, Grosh and Hoddinott (2004) finds household-level targeting (means- or proxy-means test) and workfare have the highest targeting accuracy, judged by the degree of focalization of program resources toward the poor. Geographical targeting occupies the median place in terms of targeting accuracy. However, there are other factors to consider in chosing targeting mechanisms -including cost-efficiency in administering the mechanism, maximizing the coverage of the poor, and transparency to enhance credibility, perceptionsof fairness, and reducing fraud. 44 historical allocation. Moreover, the program-specific legislation on resource allocation sometimes conflicts with poverty map formulas, as in the case o f Vas0 de Leche, where a third o f the budget i s earmarked by law for the relatively richregions of Lima and Callao. 88. Reliance on community-based organizations leads to problems of exclusion. Many SSN programs in Peru rely on CBOs to deliver the benefits. Initially, implementation by CBOs brought substantial advantages: it avoided the creation of a large public apparatus, "crowded" in community resources, and contributed to social capital and to the empowerment of women. With time, however, this model reached its limits, and many poor people remain underserved, while some programs have been "captured" by the beneficiaries who gained access to the subsidies early on. 89. Reliance on in-kind transfers may impair the eflciency of the SSN. Apart from workfare schemes and the recently established CCT program, Juntos, all Peru's social assistance programs distribute food. Food provision has been politically more acceptable than cash transfers, due to the beliefthat men might use cash to buy alcohol. However, there i s no clear evidence that food distribution increases food consumption by more than an equivalent cash transfer. Food transfers also have disadvantages, from the point of view of economic efficiency, as they restrict consumer choice and involve hightransaction costs. Box 2-4: Municipalaccreditationand performanceagreements for the accountabilityof decentralizedSSN programs in Peru: lessons learned In June 2004, as a prior step for the decentralizationof food programs, Congress approved an Accreditation Law, under which the National DecentralizationCommission(now abolished) would certify the management capacity of local governments (provincial and district) and MIMDES negotiates performance agreements to regulate their managementof the programs. However, the criteria for the transfer ofresponsibilitiesare stated invery general terms, andthe law fails to specify the consequences of non-compliance with the targets includedinperformanceagreements. The fmdings of a qualitative study of the accreditationprocess in 18 provincial and district municipalities in the departments of Piura, Cusco and San Martin, undertakenas backgroundfor this paper, suggest that accreditationo f municipalities alone is no guarantee of better service delivery. It needs to be accompanied by municipal capacity building. The study highlighted several weaknesses: First, the process focuses on verifying the existence of documents and plans, and not on validating capacity to assume the new responsibilities. Second, the accreditationis not linkedto correctiveactions to strengthenthe identified weaknesses. Third,accreditationis perceivedas a "rubber - stamping" process: many municipalitieswhich felt they were unpreparedto assume new responsibilitieswere accredited anyway. As one municipality in the department of Piura said: "...there was no problem obtaining the accreditation- we were only askedtopresentpapers". Among the positive experiences arising from decentralizationreported in the study are: the adaptation of the food basket to local preferences, the introduction of procurement from local producers, the creation of municipal associationsto executeprocurement and distribution, and enhancedaccountabilitythroughthe comitb degestidn and comith de adquisicidn, and communityandcivil society oversight. (Based on qualitative data researchon participationand servicedelivery carried out by PropuestaCiudadanagroup). 90. Institutional disintegration undermines accountability and eflectiveness. Peru's social protection programs-pensions, labor market policies, and social safety net programs-are scattered across diverse ministries,agencies, and levels of government. The Ministryo f Economy and Finance (MEF)is incharge o f the pension system and the largest food program, Vas0 de Leche. The Ministry o f Labor and Employment is in charge of the urban workfare program and other labor market interventions. The Ministry of Women and Social Development (MIMDES) is responsible for all the food-based programs (other than Vas0de Leche), for FONCODES (the social fund for basic infrastructure investments), and for programs for children and youth at risk, internally displaced people and the elderly. MINSA i s responsible for nutrition but other agencies such as MIMDES,Juntos and the PCM (through the new Plan CRECER) also play an important role. Other ministries (housing, energy, and mining) operate subsidy schemes (housing, energy) that also, conceptually, belongto the social protection sector. 45 91. Decentralization has led to a rapid shgt in responsibility for program implementation to sub- national governments but has not yet established clear, workable accountability frameworks for Peru 's social protection programs. Decentralization o f SSN programs started in the `80s with Vas0 de Leche. More recently, three food programs -- Comedores populares, Albergues y hogares, Alimentos por Trabajo- were transferred to provincial municipalities. FONCODES' small scale infrastructure program i s being transferred to district municipalities; and PRONAA's feeding programs, to provincial government In2005, the lion's share of SSN spendingremained centrally controlled: in2005 only 15 percent of PRONAA's and 26 percent of FONCODES' budgets were decentralized. However, in 2006 a much larger share of funds i s budgeted for decentralization26.As the decentralized SSN model is further developed, it will be critical to establish equitable resource allocation rules at a national level: ceteris paribus, individuals should have equal opportunities to benefit from safety net programs, wherever they live. To ensure that decentralization leads to improved outcomes for beneficiaries, it will also be critical to establish a clear accountability framework, giving local government reasonable discretion in program implementation and clarifying the role o f central government in negotiating goals and monitoring the performance of the sub-national agencies which are to be charged with program implementation and service delivery. One instrument that has been applied to this end are municipal accreditation and performance agreements; however (as was detailed inBox 2.4), much work remains to be done to ensure that the decentralization process produces better outcomes for the beneficiaries. A differentiated strategy should be developed, which takes account o f the different levels o f capacity in larger and smaller municipalities. 25 In 2006 the government started a pilot in six provinces to consolidate school feeding and nutritional programs, prior to their transfer to provincial municipalities.The plan is to form two programs, one for children 6 months to three years old andpregnant and lactatingmothers; and one for school age children, to replacethe present set o f over 20 programs. 26 The budget for 2006 plannedto transfer 70 percent of the FONCODESfunds to 1149 district municipalities(out of 1647municipalitieseligible to receiveFONCODES funds). 46 Box 4-1: Overcoming the politicaleconomyconstraints to the reform of existing programs One option for improving the performance of the SSN sector is the reallocation of funds to better performing programs. But inPeru, about halfof SSN resources are implementedby community-basedorganizations (CBOs) which enjoy considerable political legitimacy and have a record of successfully opposing reallocations of funding. Thus, a major challenge for SSN policy in Peru is to fmd ways to improve the performance of such programsby improvingtheir designand implementationand strengtheningthe accountability framework As outlined inBox 2-1, the implementationof food-basedprogramsvia CBOs brought substantial advantages in the early stages. CBOs provideda service deliveryadministrationto reachthe poor ina givenarea, allowing the government to focus on the geographical allocation of resources usingpoverty maps, which could be monitored easily in Lima, without the needfor a large public administration.However, more recently some programs have beeneffectively "captured" by their beneficiaries. The governmenthas not enforcedexit policies, fearing loss o f political support, and has even legislatedloopholes, such as the Vuso de Leche norm that allows childrenaged 7 to 13 to continue benefiting, althoughthe programi s intendedfor those aged0 to 6. Since the programbudget is fixed, the low exit rate prevents others from entering. Such constraints mean that only the marginal increase in the programbudgetthat can be assignedto poorerareas, ashappenedduring2001-04. To improve the performance of the CBO-operated programs, the SSN system should promote transparent beneficiary selection procedures and should resist rule changes that permit individuals to stay in the program beyond the appropriate time. It should also promote administrative mechanisms which tend to "self-select" appropriate beneficiaries. For example, the daily delivery of benefits increases transactions costs and discourages the nonpoor - whose time has a higher opportunity cost from seeking the benefit. As appropriate, - the CBOs should also be used as a bridge between the primary health and education systems and the communities which they organize. To achievethese outcomes, clear accountabilityframeworks should be established, with goals for the number and characteristicsofbeneficiaries, for unit costs, and for outcomes (such as improvednutritional status), which shouldbe negotiatedwith eachprogram. Once suchaframework is inplace, the resource-allocationrules of the systemshouldtie future fundingto acceptable outcomes on these goals and indicators. 4.2 Recommended strategv 92. CZarijy the structure of the SSN system. Peru needs an articulating framework for its poverty reduction and SSN strategy, in which the goals of each program - and the system as a whole - are properly defined and understood by the relevant actors. This should include a clear assignment of responsibilities within the Central Government, and between Central Government and municipal authorities with regard to regulation, financing, implementation, and program monitoring and evaluation. The Garcia administration is starting to address this need through the process of fusion, integration and articulation o f social programs, which aims to clarify objectives and increase effectiveness. The Inter- ministerial Commission for Social Affairs (CIAS) has prepared an inventory of 81 programs in 12 broad programmatic areas, which have scope for fusion, rationalization or improved coordination - including 10 dedicated to food distribution and a further 9 linkedto nutrition. It has estimated that 80 percent of these programs have administrative costs greater than 10 percent of their budgets.27During 2007, it will develop detailed proposals to implement this strategy, to be incorporated inthe 2008 budget proposal. In each sub-sector a coordinating sub-committee of CIAS will be established to oversee the process. The goal is to improve coverage, reduce leakage of resources to unintended beneficiaries, reduce administrative costs, improve transparency and optimize impacts. This will be further reinforced by the development of a Single Beneficiary Register for social programs, underway inMIMDES. *'This i s an estimation made by MEF consultants. Although it i s being strengthened, the IntegratedFinancial Administration System (SIAF) still contains incomplete informationonthe breakdownbetweenadministrativecosts and direct service provisioncosts for some programs. 47 93. Anchor the program optimizationprocess to a coherent national social safety net (SSW and poverty reduction strategy, dwferentiated for rural and urban areas. The program optimization process will produce the bestresults ifit is informed by a clear vision o f the overall strategy of the government to tackle poverty and vulnerability in Peru. The strategy should be differentiated between the 30 largest cities and the rest of the country (small towns and rural areas), to reflect the very different nature of their needs and potentials. There are vast differences in capacity between the municipal administrations in the major cities and the rest of the country, so the implementation arrangements for the SSN strategy should be differentiated. In the 30 biggest cities, SSN programs should be decentralized, to ensure better responsiveness to local needs and improved transparency. In the rest of the country, for the foreseeable future, strong national agencies will be neededto ensure that SSN programs are effective, but they should work with local governments, as appropriate. 94. As well as differentiating implementationarrangements, SSN interventionsthemselves should also be differentiated betweenthe urban and rural parts of Peru. For instance, workfare programs to deal with cyclical unemployment only make sense in urban areas; and a nationally-led small-scale infrastructure program (such as FONCODES) only makes sense for rural areas. However, a central goal of SSN programs, in both urban and rural areas, should be to tackle child nutrition, working in close liaison with MINSA. The SSN's standards and accountability system should place a heavy emphasis on this theme. The relevant programs (including Juntos and all the feeding programs) should be focused on nutrition outcomes. They should be held accountable for what they achieve to improve the quality and effectiveness o f growth monitoring, improve diets and reduce malnutrition among children under-five in their respective target populations. 95. The rural SSN strategy shouldfocus on chronic malnutrition among children under five and needs a strong national leadership, anchoredon the Juntos CCTProgram Due to the limited capacity of municipal administrations in small towns and rural areas, there is a continuing need for a strong leadership from national programs, working in liaison with provincial governments. The central goal o f the rural strategy should be to reduce chronic malnutrition. At present, about a quarter of Peru's children are stunted. Changing this outcome will require changed feeding and health care practices; which inturn will depend on increased parental awareness of the problem; improved understanding o f good nutrition; and the economic wherewithal to turn this into improved feeding practices. Juntos could have a major impact on all these fronts and could thus play an important role intransforming nutritional outcomes in its target communities. In the first place, it offers substantial cash benefits (about $30 per month) to beneficiary families, significantly improving their consumption capacity. Juntos is also well targeted on communities with highmalnutrition rates. During2007, it will reach more than 600 districts nationwide. These cover some 10percent o fthe national population - but almost 40 percent of stunted children. Juntos i s working with the health system (MINSA) to ensure that regular monitoring of children's nutritional status and effective counseling are offered to all Juntos beneficiaries. This will also require the development o f nutrition standards which are comprehensible to parents28. 96. I n the 30 biggest cities, where municipal capacity is greater, SSN and poverty reduction programs should be decentralized.The financing of decentralized SSN programs should be supported by fiscal transfers, based on transparent, equitable criteria (poverty and population statistics). The government should encourage transparent decision-making by the municipalities, through requirements for budget consultation, participatory monitoring o f budget execution and the strengthening o f local poverty maps. A decentralized administration, coupled with a strong accountability framework, will open up spaces for the reform of existing programs (such as Vaso de Leche) where political economy factors have made sweeping national-level reform difficult to achieve. Once some local governments have 28 The World Bank's Recurso program of studies and technical assistance - of which this report forms a part - is supporting the reorientation of Juntos to concentrate on nutritionthrough a TA program, and is also supporting MINSA and CENAN inthe development of easy-to-understandnutrition standards, throughanew video. 48 decided to improve the targeting of their programs and can show better outcomes, the demonstration effects of benchmarking those results will generate increasing pressure on other municipal governments to do likewise. And at a local level, the authority and credibility o f municipal governments may offset opposition to reform from the entrenched interest groups. 97. At a programmatic level, the strategic foci of the urban SSN should be (a) re-distributional food transfers focused on poor children, to alleviate poverty and improve nutrition outcomes, and (b) an effective safety net mechanism to deal with labor market shocks. The urban SSN should aim to provide effective food transfer programs targeted on poor children, andto keep inplace a safety-net mechanism to tackle cyclical unemployment. SSN programs for the 30 largest cities should be operated by municipal governments, with financial support from fiscal transfers. This should be coupled with the strengthening o f transparency requirements, to promote community involvement in municipal budgeting and participatory monitoring of budget and program implementation - as mandated in the 2007 BudgetLaw. This would provide abasisfor benchmarkingand the promotion of best practices. 98. For both centralized and decentralized SSNprogram, strengthened accountabilityframeworks are the key to improved outcomes. To break out of the low quality equilibrium, the SSN system needs clear objectives and quality standards, and targets for improved outcomes, with a particular emphasis on chronic malnutrition. This line o f action is now being pursued by the Results Based Budgeting (RBB) office in MEF. The accountability and results management process should focus first on the major programs (such as Vas0 de Leche, Comedores Populares, School Breakfasts, FONCODES and Juntos, which account for over 70 percent of SSN resources). These programs have many strengths. FONCODES compares well on a global scale with similar programs on several indicators. A Trabajar Urbano is also in line with world-wide best standards. Vas0 de Leche has strong community roots that vitiate against crude political manipulation from the centre, and is better targeted than i s sometimes thought. Juntos has incorporated cutting-edge design features from Mexico's Oportunidades program, and i s well targeted. However, in each program, important gains could be achieved through pragmatic, feasible reforms in program design and implementationarrangements. 99. Juntos, as was argued above, should be the central program in the rural SSN. This CCT program couples large income subsidies with incentives to investment inhuman capital, and - working in liaison with MINSA - could have a major impact on chronic malnutrition. Juntos should give specialeffective priority to promoting attendance at growth monitoring and counseling sessions, supporting MINSA to offer improved nutritional and health services. Juntos' expansion should be limited to rural areas, giving continued priority to places with high rates o f chronic malnutrition. It should also continue to support RENIEC's efforts issue identity documents. However, to ensure the expansion is effective, the beneficiary identification system, system o f conditionality and the monitoring and evaluation system should all be strengthened. Among the issues to address are: a) whether the household-level component o f the beneficiary selection system i s adding enough value to justify its cost; b) the choice of appropriate conditions (little i s gained by the education conditions, given the level of enrollment) and establishment o f an efficient conditionality monitoring system; c) the adequacy o f the supply-side arrangements in nutrition and health; ad d) the need for a robust monitoring and impact evaluation system. Ifthese issues are successfully addressed, Juntos could become a key articulating element of Peru's social sector in low income rural areas. 100. Vas0 de leche is the key food program by size and importance. Its impact on reducing malnutrition inurban areas could be improved by (i)focusing resources on the legally defined group of beneficiaries (children 0-6 years old) - and preferably concentrating on children under 24 months; (ii)improving procurement and using fresh instead o f more expensive evaporated milk and ensuring on-site consumption; (iii)providing a service seven days a week instead of the actual practice o f giving milk on fewer days to cover more beneficiaries; and (iv) providing complementary health and nutrition education. The targeting performance could be improved by enforcing the oversight over the Vas0 de leche 49 committees and by eliminating legislation that confines one third o f the resources to Lima and Callao provinces. 101. Socialfind. Inrecent years, a process has beenunderway to decentralize FONCODES and lower its funding in the expectation that municipalities would increase investment to compensate. The present government has announced that FONCODES will be closed. However, many rural municipalities lack the capacity to develop and implement projects. In the face of such constraints, Peru still needs to reach closure on the definition of a system for small-scale rural infrastructure investments, which: (a) assigns resources usingpoverty targeting techniques; (b) provides adequate support to district municipalities; (c) establishes effective accountability mechanisms; and (d) ensures adequate coordination with the appropriate line agencies. FONCODES or its successor agency could make a vital contribution to building the capacities of district governments to forging strong partnerships with communities and local service providers for effective small-scale infrastructure provision. 102. A Trabajar Urbano has achieved effective self-targeting o f the poor and good impacts on household income earning and community assets. Although the economy i s presently buoyant, Peru should keep the program operating, maintaining a minimal program that can be scaled up during recessionsor natural disasters in a countercyclicalmanner. To improve implementation, some fine-tuning is required: (a) ensure that the wage level is set with reference to real market conditions inthe low-income informal sector; (b) consider the option of integratingthe workfare programs with municipal investments to ensure quality (in terms o f supplementary resourcesneeded to complement the labor production factor) and maintenance of assets created; and (c) consider as possible projects not only the construction o f physical assets but also the provision of social services (caregivers, health promoters) that might increase the participation ofwomen. 103. Move towards a high-quality equilibrium, increasing SSNfinancing towards the average in middle-income countries of 1.5 percent of GDP. Finally, for the reformed SSN programs to have a greater impact on reducing present and future poverty, more resources will be required. To move Peru's SSN system from the existing "low quality equilibrium" towards a "high quality equilibrium", as well as ensuring improved program quality and measurable outcomes on key goals, such as chronic malnutrition, it will also be necessary to reverse the erosion of fbndingwhich has flowed from the historic failure of management in Peru, the process of enhancing resources for the SSN should accompany - and provide existing programs to generate such outcomes. As part of the implementation of results-based budgetary incentives for - the reform process itself. Additional resources should be channeled towards those interventions that can demonstrate success in establishing robust accountability frameworks and documenting strong impacts. 50 BIBLIOGRAPHY Alcazar, Lorena. 2003. "Monitoring Social Outcomes and Policies in Peru: The Challenge of Decentralization." Background paper for the forthcoming Poverty Assessment o f the World Bank. Alderman, H., and David Stiefel. 2003. "'The "Glass of Milk' Subsidy Program and Malnutrition in Peru. Working Paper 3089. Washington, D.C.: World Bank. Blank, Lorraine, M. Grosh, Guillermo Hakim, and C. Weigand. (forthcomming 2006). "Social Protection SpendingDatabase." Washington, D.C.: World Bank. Blondet, Cecilia, and Carolina Trivelli. 2004. "Cucharas en alto." IEP. Bustamante Suirez, Miguel. 2003, "Caracterizacibn del Programa del Vas0 de Leche." Direccibn General de Asuntos Econ6micos y Sociales del Ministerio de Economia y Finanzas. Lima, February. Castaiieda, Tarsicio, and Kathy Lindert, with BBnBdicte de la Briere, Luisa Fernandez, Celia Hubert, Osvaldo Larraiiaga, M6nica Orozco, and Roxana Viquez. Forthcoming 2005. "Designing and Implementing Household Targeting Systems: Lessons from Latin America and the United States." Washington, D.C.: World Bank. Chacaltana, Juan. 2003. "El ImpactodelProgramaA Trabajar Urbano." Mimeo. Chacaltana, Juan. 2005. "La pobreza no es como lo imaginabamos." Mimeo. Cueto, Santiago, and Marjorie Chinen. 2001. "Impacto educativo de un programa de desayunos escolares en escuelasrurales del Perk" Lima. Francke, Pedro. 2004. "Reforma de programas de alimentaci6n escolar." Mimeo. Francke, Pedro. 2004. "Reforma de programas nutricionales infantiles." Mimeo. Francke, Pedro. 2005. "La descentralizaci6n de 10s Programas Sociales: En que Direcci6n y Cuanto se ha Avanzado." Mimeo. Francke, Pedro. 2005. "La institucionalidad de 10s programas alimentarios." Mimeo. Gajate, Giselle, and Marisol Inurritegui. 2002. "El impacto de 10s programas alimentarios sobre el nivel de nutrici6n infantil: una aproximacih a partir de la metodologia del `Propensity Score Matching'," Lima: Grupo de Andlisis para el Desarrollo, April. Instituto Apoyo. 2000. "Sexta Evaluaci6n ExPost del Foncodes, Evaluaci6n de Impacto y Sostenabilidad." Lima. Instituto Apoyo. 2002. "Public Expenditure Tracking Survey: The Education Sector in Peru, Appendix 1: Breakfast Program." Lima, Peru, September25. 51 LindertKathy, Emmanuel Skoufias, and Joseph Shapiro. 2005 forthcoming. "How Effectively do Public Tranfers inLatin America RedistributeIncome." Washington, D.C.: World Bank. L6pez-Chlix, J. R., L. Alcazar, and Erik Wachtenheim. 2002. "Peru: Public Expenditure Tracking Study." InPeru: Restoring Fiscal Disciplinefor Poverty Reduction, Public Expenditure Review. World Bank and Inter-American Development Bank, ReportNo. 24286. Washington, D.C. Maximize and Instituto Cuhnto. 2003. "Evaluacih de Impacto del Programa de Complementacih Alimentaria para Grupos de Mayor Riesgo" Lima, August. Paxson, Christina, and Norbert R. Schady. 2002. "The Allocation and Impact of Social Funds: Spending on School InfrastructureinPeru." The WorldBank Economic Review 16(2). Washington, D.C. Pollit, Ernesto, Enrique Jacoby, and Santiago Cueto. 1996. "Desayuno Escolar y rendimiento: a prop6sito del programa de desayunos escolares de Foncodes en el Peni." Background paper. Lima. Rawlings, Laura, L., Sherburne-Benz, and J., Van Domelen. 2004. "Evaluating Social Funds. World Bank, Regional and Sectoral Studies. Washington, D.C. Rawlings, Laura, and Gloria M. Rubio. 2005. "Evaluating the Impact o f Conditional Cash Transfer Programs." Tesliuc, E., and Lindert, K. 2002. "Social Protection, Private Transfers and Poverty in Guatemala." Backgroundpaper to the Guatemala Poverty Assessment. Washington, D.C.: World Bank. Tesliuc, Emil Daniel, M.Grosh, D. Coady, and L.Pop. (forthcomming 2005) "Program Implementation Matters for Targeting Performance: Evidence and Lessons from Eastern and Central Europe." Washington, D.C.: World Bank. Trivelli, Carolina. 2004. "Analizando la encuesta: 10s comedores de Lima metropolitana en 2003." In: Cucharas enAlto. Instituto de Estudios Peruanos. Lima. Valdivia, Martin. 2005. "Peru: I s Identifying the Poor the Main problem in Targeting Nutritional Programs." Vhsquez, Enrique. 2004. "Programas Alimentarios en el Perk LPor qui y c6mo reformarlos?" Study elaborated for the Instituto Peruano de Economia Social de Mercado, Lima, Peru, December 22. Vhsquez, Enrique. 2004. "Subsidios para 10s mas Pobres: s e r h beneficiados 10s niiios en extrema pobreza?", Los Niiios Primero, Observatorio para la Infanciay la Adolescencia. Vhsquez, Enrique. 2005. "Programas Alimentarios en el Peru" Por que y como reformarlo?" World Bank. 2003. Program Document for a proposed fourth programmatic social reform loan to the Republic of Peru. World Bank. 2004. "Inequality inLatinAmerica: Breaking with History." Washington, D.C. Yamada Gustavo, and Patricia Perez. 2005. "Evaluacih de Impacto de Proyectos de Desarrollo en el Peru." Universidaddel Pacifico. 52 ANNEX2: ESTIMATING INCIDENCE OFPUBLIC SPENDINGFOR FOODPROGRAMS THE Inassessingthe incidence of public subsidy distributedthrough food programs, we recognize that simply looking at the incidence of beneficiaries may be misleading. Program beneficiaries differ with respect to the monthly take-up rate (number of portions consumed per month) and the rate of subsidization (some get free meals, others pay). Inaddition, the level of subsidy differs substantially across departments. To estimate the distribution o f public spending across quintiles, we multiply the (monthly) distribution of beneficiaries with the take-up rate, the level of subsidy from each region, and the subsidization rate Many benefit-incidence studies use as a proxy for the distribution of the public subsidies the distribution of program beneficiaries across quintiles. The majority o f benefit-incidence studies of food programs in Peru follow the same approach. However, the distribution o f beneficiaries reveals the distribution of public subsidy only for programs that give the same "treatment" to all beneficiaries, i.e. they provide a uniform benefit per individual or per households at regular intervals. The food programs operating inPeru are more complex. Inthe case of food programs, the distributionofbeneficiaries is a poor proxy for the distribution of public subsidy, for three reasons: 0 First, for some programs, the utilization of food based programs varies across beneficiaries. For instance, 37 percent of the beneficiaries of the comedores populares receive meals daily, another 29 percent only few days per week, 16 percent only once a week and the rest at longer time intervals. To estimate the true incidence of the public subsidy, we need to correct for the frequency with which beneficiaries use these programs. 0 Second, food programs have are two trpes of beneficiaries: a first group does not pay for the good or services received (get the full subsidy), while others are charged a partial co-payment. For instance, only one in fifth beneficiaries of comedorespopulares receives meals free-of-charge. Food-based programs produce food and deliver food-rations for free, or at subsidized prices (Figure 1). The ENAHO survey does not collect information on the level of public subsidy enjoyed by those who consume free or subsidized food-rations. Instead, the survey asks if somebody was benefiting from the program, with what frequency, and ifthey hadto pay something for the food. From qualitativework, we know that the price of aration for paying beneficiaries is 50 percent subsidized. With this information, we can estimate the amount of public subsidy capturedby eachbeneficiary. Figure 1Shareof beneficiarieswho receivefood ration without co-payment, 2003 I I --1 I mso de leche cornedor desayuno club de almueno comedor popular escolar madres escolar infantil Source: Author's estimationbasedon ENAHO 2003/04 Survey, INEI. To estimate the level of public subsidy enjoyed by a beneficiary of a food program, we used the following algorithm: First, we determined how many rations per month were received by each beneficiary. Second, we estimated the gross consumption by multiplying the unit cost of a food ration (from administrativedata) with the 54 number o f rations consumed per month (reported in the ENAHO survey). For beneficiaries who enjoyed fiee rations, this is the level o f subsidy they received. Finally, for beneficiaries who made direct contributions (cash, labor, or food) for their food, the value of subsidy was assumed to be 50 percent o fthe gross consumption Third, the unit subsidy varies substantially across the 24 departments of Peru for many of the programs under consideration. When computing benefit incidence using an average unit subsidy, we assume that the benefit i s homogenous for all beneficiaries or departments of the country. In reality, in the case of desayuno escolar, the annual unit cost varied from 143 Soles in Lima and Callao to only 87 Soles in Tacna. To estimate the true incidence of the public subsidy, we need to correct for the variation inthe level ofaveragesubsidy distributedindifferent departments. 55 ANNEX3: ALGORITMOPARA ELCALCULO DELAPROBABILIDAD DEPOBREZA A partir de un pool de hogares entre 10s aiios 2001 y 2004 (utilizando la Encuesta Nacional de Hogares), se hizo la siguienteregresi6n logistica: Logit: Y = a + PX + p donde la variable dependiente, Y, se construye a partir de la lineade pobreza de la siguiente manera: Y = 1, si el hogar es considerado pobre; Y =0, si el hogar no es pobre Por otro lado, a es una constante, p representa el error y X esta constituida por la siguiente relaci6n de variables ex6genas (las cuales serandetalladas en pasos mas adelante): 1. analf-m: porcentaje de mujeres adultas analfabetas dentro del hogar. 2. edu-men: porcentaje de menores de edad que asisten a algun programa de enseiianza regular dentro del hogar. 3. combusto: acceso a fbentes industriales de combustible industriales para cocinar (gas, electricidad, kerosene). 4. no-equip: numero de artefactos ausentes en el hogar 5. serv3:tenencia de servicio de alumbrado, aguay servicios higihicos en el hogar. 6. tipom2: grupo de tipos de vivienda 2. 7. tipom3: grupo de tipos de vivienda 3. 8. tipom4: grupo de tipos de vivienda 4. Los resultados de laregresi6n son: Variable Coeficiente analf-m [12.661** 1.1832* edu-men 0.2276 [5.13]*** combust0 [12.841 ** -0.7624 * no-equip 0.4446 [27.40] *** serv3 -0.3769 [3.23]*** tipom2 -0.2593 [5.551 *** tipom3 [14,861* -0.8584 ** tipom4 -1.3172 [17.53]*** Constante ( 0 ) -1.3461 [12.48]*** Observaciones 17980 PseudoR2 0.182 Valor absoluto del estadistico Z en corchetes *significativo a1 10%; ** significativo a15%; *** significativoa1 56 Dentro de la base de Caracteristicas de la Poblaci6n(Cuestionario 100) se crean las siguientes variables: Objetivo: Toma el valor de 1 si cumple: "edad menor a 14afios" 6 "Mujer embarazada entre 12 y 49 afios ",y 0 en otro caso, a partir de la pregunta 115. Analfabetismo de Mujeres Adultas: Toma el valor de 1 si se trata de una persona de sex0 femenino, mayor de 18 aiios y que no sabe leer y escribir, segh la pregunta 108. Toma el valor de 0 en otro caso. Asistencia de Menores:Con el valor de 1 si es el cas0 de una persona menor de edad que asiste actualmente a algun centro o programa de enseiianza regular, dad la pregunta 110, y el valor de 0 en otro caso. Se agrega la base de datos por hogar, teniendo en cuenta la suma de adultos, de menores de edad, de mujeres analfabetas y de menores que asisten a algun programa escolar. Posteriormente, se filtra esta nueva base por aquellos hogares que cumplan con el Objetivo de contar con algun menor de 14 aiios o alguna mujer embarazada. Se construyen 10s ratios de: 9 Mujeresadultasanalfabetascomo porcentajedeltotal deadultos enelhogar, y 9 Menores que asistena algun centro o programa educativo como porcentaje del total de menores de edad en el hogar. A continuaci6n, se imputan 10s ratios anteriores, en el cas0 de encontrar vacios o valores perdidos, con el promedio de 10s hogares del distrito en cuestibn. D e la base de Caracteristicas de la Vivienda y del Hogar (Cuestionario 200), se obtienen las siguientes variables: N o equipo: Indica la cantidad de equipamiento con 10s cuales no cuenta un hogar. Toma valores entre el 1 y el 7, que se identifican con 10s siguientes artefactos: televisor a blanco y negro, televisor a color, refiigeradora, plancha elkctrica, cocina a gas, vehiculo motorizado y vehiculo a pedal. Servicios: Toma valores entre 1 y 3, dependiendo de si el hogar cuenta con el servicio de alumbrado conectado a red elkctrica, agua de red publica y servicios higidnicos con saneamiento. Combustible: Toma el valor de 1 si el combustible que mLs utiliza en el hogar para cocinar alimentos es de origen industrial, y 0 en otro caso. PAS0 6 Sejuntan esta base de vivienda resultante con la de poblaci6n que habia sido agregada por hogares. Luego se filtran aquellos hogares que cumplan con e l Objetivo y que cuenten con "encuestas completas". Posteriormente se procede a imputar las variables No Equipo, Servicios y Combustible con el promedio de 10s valores por distrito, en el cas0 que existan valores perdidos o vacios en labase resultante. 57 SE A~~~ADEN variables de tipo de vivienda obtenidas a partir de 10s materiales predominantes en las las paredes, techos y pisos (preguntas 203, 204 y 205, respectivamente). Despuis de realizar todas las 294 combinaciones posibles de dichos materiales, se escogieron 22 que engloban el 91,l percent, las cuales se agruparon en las siguientesvariables: Variable I TiDo Pared Techo Piso I 102 Adobe Teja Tierra Grupo de 126 Adobe Paja Tierra Tipos de 294 Estera Paja Tierra Vivienda 1 210 Piedra c/barro Paja Tierra 114 Adobe Caiia Tierra I 168 Quincha Paja Tierra 108 Adobe Calamina Tierra 150 Quincha Calamina Tierra Grupo de 252 Madera Paja Tierra Tipos de 276 Estera Calamina Tierra Vivienda 2 113 Adobe Caiia Cemento 101 Adobe Teja Cemento Piedra dbarro Calarnina Tierra Madera Calamina Tierra Grupo de 107 Adobe Calamina Cemento Tipos de 250 Madera Paja Entablado Vivienda 3 106 Adobe Calamina Entablado 24 Tierra Grupo de 232 Entablado Tipos de 23 Ladrillo Calamina Cemento Vivienda 4 5 Ladrillo Concreto Cemento 233 Cemento De encontrarse valores perdidos o vacios en estos Grupos de Tipos de Vivienda, se procede a imputarlos con el promedio del distrito. PAS0 8 A todas las variables anteriormente generadas (ver lista en Paso 1) se les multiplica por sus respectivos coeficientes encontrados en la regresidn del Paso 1, para despu6s hallar su distribucidn logistica, expresadacomo campo Y en labase, lo cual indicaria la probabilidad que tiene unhogar de ser pobre. A partir del resultado de la regresidndelPaso 1se realizaunaprediccidn de la probabilidadde pobreza de 10s hogares. Luego se 10s ordena en base a dicha probabilidad. Sabiendo que la pobreza en el area rural es de 65.9 por ciento en el pool de hogares 2001-2004, el umbral que se obtiene asociado a ese porcentaje es de 0.7567447. Aplicando el umbral, se obtiene el porcentaje de hogarespobres y no pobres dentro de cada distrito. END 58 0 . 0 0 0 0 . N 3 3 N 3 58 g& 5& 8 L & ANNEX5: IMPROVEDPROXY MEANS MODEL: ILLUSTRATION TEST AN Model (1) (2) (3) (4) (5) (6) (7) (8) DependentVariable lcpc lcpc lcpc lcpciof pobreza xpoor xpoor xpoor Share of illiterate female -0.308 -0.156 -0.148 -0.151 -0,151 0.779 0.559 0.394 0.418 (24.IO)** (13.34)" (12.72)" (13.43)" (14.01)"' (10.01)** (6.73)" (6.90)** Share of children attending school -0.065 -0.062 -0.068 -0.084 0.162 0.285 0.323 0.389 (4.62)". (4.96)" (5.44)" (7.19)" (3.02)'* (4.06)- (4.46)" (5.21)** Access to industrial fuel for cooking 0.218 0.222 0.152 0.165 -0.59 -1.381 -1.391 -0.934 ( I7.37)- (19.65)" (13.08)'' (12.10)" (11.66)"" (21.07)*' (20.42)" (10.85)" Number of missing appliances -0.119 -0.138 -0.131 -0.15 0.42 0.363 0.386 0.396 (33.71)'" (43.49)- (41.09)** (48.81)'* (29.46)" (17.21)" (17.84)" (17.60)" Access to water, sewage 0.169 0.16 0.135 0.145 -0.322 -0.383 -0.39 -0.397 (36.70)- (38.67)"" (31.80)" (30.53)" (17.17)" (16.59)". ( I6.22)'" (14.44)"" Housing conditions 2 -0.087 -0.073 -0.072 -0.063 0.031 -0.065 -0.08 -0.125 (7.71)" (7.22)** (7.10)" (6.12)" -0.64 -1.34 -1.59 (2.36)' Housing conditions 3 0.003 0.025 0.019 -0.029 -0.021 -0.236 -0.289 -0.447 -0.23 (1.96)* -1.5 (2.18)' -0.35 (3.32)** (3.94)" (5.89)" Housing conditions 4 0.025 0.049 0.01 -0.084 0.065 -0.549 -0.608 -0.613 (2.29)' (4.91)** -1.02 (9.12)" -1.59 (7.39)** (7.96)" (7.83)" Total number of HH members -0.117 -0.115 -0.094 0.239 0.258 (68.89)'" (67.50)" (62.31)" (27.29)" (28.35)" % malnourished children < 5 years old -0.95 (24.14)'" Area 0.039 0.103 (3.11)" -1.48 departamento==5 0.082 -0.513 (3.47)" (4.24)" departamento==6 -0.031 -1.81 departamento==7 0.183 -1.42 (7.71)" (4.69)" departamento==8 -0.159 -1.57 departamento==g -0.301 0.993 (11.45)" (8.41)'* departamento==lO -0.088 0.437 (4.14)" (4.39)" departamento==l 1 0.07 -2.161 (2.90)- (5.76)" departamento==l2 -0.056 (2.99)" departamento==l4 0.095 -1.04 (4.86)" (7.86)" departamento==l5 0.218 -0.622 (19.88)" (7.40)" departamento==l6 0.071 (3.44)" departamento==l7 0.379 -2.104 (5.77)- (2.91)** departamento==l9 0.047 -1.34 depaitamento==20 0.028 -0.637 -1.73 (7.21)** departamento==Zl -0.29 1.027 (15.97)" (11.92)" departamento==22 0.032 -1.36 departamento==24 0.322 -3.305 (7.72)" (3.86)" departamento==25 0.089 (2.95)" Constant 5.597 6.157 6.366 6.123 -0.805 -1.961 -3.417 -3.715 (243.42)" (277.67)" (266.71)*' (265.00)" (8.70)" (15.28)*' (23.69)" (24.15)" Observations 19590 19590 19248 19590 19590 19590 19590 19590 R-squared 0.36 0.49 0.5 0.56 0.19 0.25 0.29 0.32 Absolute value oft statistics in parentheses * Significant at 5%; ** significant at 1% 61