Report No. 28612-ME Mexico Poverty in Mexico: An Assessment of Conditions, Trends and Government Strategy June 2004 Colombia and Mexico Country Management Unit Latin America and the Caribbean Region Poverty Reduction and Economic Management Division FOR OFFICIAL USE ONLY Document of the World Bank This document has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization. This document has a restricted distributionand may be used by recipients only in the performanceof their Official duties. Its contentsmay not otherwisebe disclosed without World Bank authorization. MAIN ABBREVIATIONS & ACRONYMS ACERCA Programfor ComercializationSupport (Programade Apoyos para la Comercializacibn) APAVER Program for Drinking Water and Seweragein the Stateof Veracruz-Llave (Programade Agua Potabley Alcantarilladodel Estadode Veracruz-Llave) BANRURAL National Bank for Rural Credit (BancoNacionalde Crkdito Rural) CEC StrategicCommunity Centers (CentrosEstratkgicos Comunitarios) CENDIS Child DevelopmentCenters (Centrosde DesarrolloInfantil) CIDE Center for Economic Researchand Education (Centrode Investigacibny Docencia Econbmicas) CIESAS Center for SocialAnthropology Research and Studies (Centrode Investigacionesy EstudiosSuperioresen AntropologiaSocial) CONACYT National Council on Scienceand Technology (ConsejoNacional de Cienciay Tecnologia) CONADEPI Commissionfor the Development of IndigenousCommunities (Comisibnpara el Desarrollode 10s Pueblos Indigenas) CONAFE National Councilfor EducationalDevelopment (ConsejoNacional de FomentoEducativo) CONAFOVI HousingDevelopmentCommission (ComisibnNacional de Fomentoa la Vivienda) CONAPO National Councilon Population (ConsejoNacionalde Poblacibn) CONASUPO National Companyfor Popular Subsistence (CompaiiiaNacional de SubsistenciasPopulares) CONAZA National Commissionon Arid Zones (ComisibnNacional de Zonas hidas) DICONSA CONASUPOCommercialDistributor (DistribuidoraComercialCONASUPO) DIF IntegratedFamily DevelopmentSystem (Sistemapara el DesarrolloIntegralde la Familia) ECLAC Economic Commissionfor Latin America 11 ENCASEH HouseholdEconomic CharacteristicsSurvey (Encuesta de CaracteristicasSocioecon6micasde 10sHogares) ENESTYC National Survey on Employment, Salaries, Technology and Training in the Manufacturing Sector (Encuesta Nacional de Empleos, Salarios, Tecnologia y Capacitacibn en el Sector Manufacturero) ENEU National Urban Employment Survey (Encuesta Nacional de Empleo Urbano) ENIGH National Survey on Household Income andExpenditure (Encuesta Nacional de Ingreso-Gastode 10s Hogares) FAIS Social Infrastructure Contributions Trust Fund (Fondo de Aportacionespara la InfraestructuraSocial) FAPPA Fundfor the support of Productive Projects by Agrarian Organizations (Fondo Parael Apoyo a ProyectosProductivos de las OrganizacionesAgrarias) FIFONAFE Trust Fundfor the NationalEjido DevelopmentFund (Fideicomiso Fondo Nacional de FomentoEjidal) FIRA Agriculture Trust Funds (Fideicomisosconstituidos en Relaci6ncon la Agricultura) FIRCO Trust Fundfor SharedRisk (Fideicomiso de RiesgoCompartido) FONACOT National Fundfor Workers' ConsumptionDevelopmentand Guarantee (Fondo de Fomento y Garantia parael Consumo de 10sTrabajadores) FONAES National Fundfor Support to Social Enterprises (Fondo Nacional de Apoyo a EmpresasSociales) FOVI Fundfor Housing Banking Operation andFinancing (Fondo de Operaci6n y Financiamiento Bancario a la Vivienda) FOVISSSTE ISSSTEHousingFund (Fondo de la Vivienda del ISSSTE) IMSS MexicanSocial Security Institute (Instituto Mexican0 del Seguro Social) INDESOL National Institute of Social Development (Instituto Nacional de Desarrollo Social) INEGI National Statistics, Geography and Informatics Institute (Instituto Nacional de Geografia, Estadisticae Informhtica) INFONAVIT National Institute for HousingDevelopment (Instituto Nacional de Fomento a laVivienda del Trabajador) IN1 National Indigenous Institute (Instituto Nacional Indigenista) ... 111 INSP National Public Health Institute (Instituto Nacional de la SaludPiiblica) ISSSTE Social Security and Servicesfor the StateWorkers Institute (Instituto de Seguridady Servicios Sociales de 10s Trabajadoresdel Estado) LICONSA CONASUPOIndustrialMilk (Leche IndustrialCONASUPO) OECD Organization for Economic Cooperationand Development PAC CoverageExpansion Program (Programa de Ampliacicin de Cobertura) PAREIB Programto BringDownthe EducationalLag inPrimary Education (ProgramaparaAbatir el RezagoEducativo enEducaci6nInicial y BBsica) PEC Programfor Quality Schools (Programa de Escuelasde Calidad) PET Temporary Employment Program (Programade Empleo Temporal) PEMEX Mexican Petroleum (Petrbleos Mexicanos) PIASRE Comprehensive Program of Sustainable Agriculture and Productive Transformation in Areas of Loss Record (Programa Integral de Agricultura Sostenible y Reconversicin Productiva en Zonas de Alta Siniestralidad) PROBECAT Training ScholarshipsProgram for the Unemployed (Programa de Becasde Capacitacicinparael Desempleo) PROCAMPO Programfor Direct Support to Agriculture (Programa de Apoyos Directos alCampo) PROCEDE Program for Ejido Rights Certification (Programa de Certificacicin de DerechosEjidales y Titulacidnde Solares Urbanos) PRODEFOR Forestry Development Program (Programa de Desarrollo Forestal) PRODEI Programfor the Development of Pre-scholar Education (Programa parael Desarrollo de laEducacicinInicial) PRODEPLAN Program for Commercial Forestry Plantations (Programa de PlantacionesComercialesForestales) PROMUSAG Program for Women inthe Agricultural Sector (Programa de laMujer en el SectorAgrario) PRONABES National Scholarship Program for Higher Education (Programa Nacionalde BecasparalaEducacicinSuperior) PRONASOL National Solidarity Program iv (ProgramaNacional de Solidaridad) SAGARPA Ministry of Agriculture, Livestock, RuralDevelopment, Fisheriesand Food (Secretariade Agricultura, Ganaderia, Desarrollo Rural, Pescay Alimentaci6n) SFP Ministry of Public Function (Secretariade laFunci6nPitblica) SEDESOL Ministry of Social Development (Secretariade Desarrollo Social) SEMARNAT Ministry of Environment andNatural Resources (Secretariade Medio Ambiente y RecursosNaturales) SEP Ministry of Education (Secretaria de Educaci6nPliblica) SHCP Ministry of Finance (Secretariade Hacienday Crkdito Pitblico) SRA Ministry of Agrarian Reform (Secretariade laReformaAgraria) SSA Ministry of Health (Secretariade Salud) STPS Ministry of Labor and Social Prevision (Secretariadel Trabajo y Previsi6n Social) UDLA Universidad de Las Amdricas, Puebia UNDP UnitedNations Development Program WDI World Development Indicators WDR World Development Report V This report was preparedby a team ledby Gladys Ldpez Acevedo andMichaelWalton of the World Bank. Members of the team included: Alain de Janvry and Elisabeth Sadoulet (University of California at Berkeley); Marcela Rubio Siinchez, Jorge Moreno, William Maloney and Laura Rawlings (World Bank); Mdnica Tinajero, Maria Caridad Araujo, Jesds Torres and Sean Scott (consultants); Odracir Barquera (World Bank -production assistance). Material on the incidence of public spending was provided by John Scott (CIDE, Mexico) as part of parallel work on public spending managed by Steven Webb (World Bank). Specific inputs were provided by Ricardo Paes de Barros (PEA-Brazil); Jonathan Fox (University of California at Santa Cruz); Luis Felipe Upez Calva, Gloria Martha Rubio and Isidro Soloaga (Universidad de las AmCricas Puebla, Mexico); Lourdes Rodriguez Chamussy (UNDP-Mexico); Juan Martinez, Mauricio Santamaria, Cristian Baeza and Mario Torres (World Bank). Extensive comments on earlier drafts helped shaped the report. We would especially thank the peer reviewers, Rodrigo Garcia Verd6 (Banco de MCxico); Margaret Grosh. and Norbert Schady (World Bank). Valuable comments were also provided by Jose Mm'a Caballero, Mark Hagerstrom, Gillette Hall, Miguel Upez Bakovic, Ernest0 May. Gobind Nankani, Jaime Saavedra, Guillermo Perry and StevenWebb (World Bank). The work i s being undertaken under the general direction of David de Ferranti (Vice President, LatinAmerica andthe Caribbean, World Bank). The team received valuable feedback from Miguel SzCkely, Gonzalo Herniindez-Liconz and Carlos Maldonado (SEDESOL) and Rafael Freyre and Rodrigo Gallegos (Social Cabinet Secretariat) of the government of Mexico; the Poverty Measurement Committee and, at the design stage, from the joint government of Mexico-World Bank workshop or July 24th-25th,which involved representatives from all members of the Social Cabinet Thanks also to other members of the CONTIGO secretariat. During May, revisior meetings were held with the Social Cabinet and members from the ministries of Socia Development, Health, Agrarian Reform, Agriculture, Education, Finance, the Office foi Public Policy in the Presidency, and with president Vicente Fox. We thank thei comments andfeedback. 1This report exclusively representsthe views of the World Bank team. vi TABLEOF CONTENTS Summary and Key Messages xiv A. Poverty Conditions andTrends ' xv 3. Government Strategy xxvi C. Concluding comments: on action andfurther work xlix 1. IntroductionandFramework 1 A. Concepts: what is poverty andwhy does itmatter? 2 3. Frameworks for reducingpoverty 8 C. The Government's CONTIGOframework for poverty reduction 16 D. Summary 20 2. Well-being inMexico inInternationalPerspective 22 A. Patternsof progressinwell-being ininternationalperspective 22 3. An overview of the relationship between well-being and indicators of country conditions 29 C. Conclusion 38 3. The Structure and Trends inDeprivation 39 A. Trends ineducation, health, andbasic infrastructure services 39 B. The Structure, Correlates andTrends inIncomePoverty 48 C. Vulnerability 71 D. RegionalandEthnicdifferences inwell-being 80 Annex 99 4. Public Spending, Poverty and Inequality 108 A. Public spending, taxes and inequality 108 3. SelectedIssuesinthe Designof Socialprograms 130 C. Decentralization 151 D. Conclusions andissuesfor future work 160 Annex 162 5. Influences on growth, distribution and poverty reduction 165 A. Growth, inequality andfuture prospectsfor income povertyreduction 165 B. Structural andpolicy influences on growthanddistribution 167 C. Strengthening income growth for the moderate urbanpoor 179 D. Income growth for ruralhouseholdsinextremepoverty 189 E. Conclusions andquestions 202 6. The RoleandDesign of Evaluation 204 A. Mexico's pioneering role inimpactevaluation 207 3. RBM:The MexicanExperience within aGlobalTrend 209 C. International Lessonsfor Designof Result-based Management Systems 211 D. InternationalLessonsfor the Implementation of ResultBasedManagement 216 E. Challengesof ResultsBasedManagement 217 vii F. Conclusions andRecommendations 219 Annex 221 Appendix 228 References 233 LISTOFTABLES Table 1.Percentageof people livingbelow $2 per day xx Table 2. Mexico's social protection systemby poverty status xxxvii Table 3. Federal transfers xlv Table 1.1Official Poverty Lines inCurrent Pesos per month 8 Table 2.1 Incomeinequality inMexico ininternational perspective 25 Table 2.2 Enrollmentrates of young people inpoor andrichMexican households, 1992 and 2000 (inpercent) 27 Table 2.3 Years of education of males and females inMexico by age group 28 Table 2.4 The relationships between mortality, education andeconomic and social variables 37 Table 3.1 Steadyprogress butlarge gaps in schooling of Mexicanadults 40 Table 3.2 There hasbeen major progress ineducational indicators, butgaps and problems remain, especially at higher levels 41 Table 3.3 Enrollmentrates are muchlower for poorer groups at higher levels of education 42 Table 3.4 Comparative HealthOutcome Indicators: Chiapas, Oaxaca, and Guerrero versus National 46 Table 3.5 Many basic infrastructure services are reaching the bulk of the population 47 Table 3.6 There has been large progress inaccess to water andelectricity amongstthe ruralextreme poor, but not insanitation 48 Table 3.7 Marginalizedruralcommunities have low access to public services 48 Table 3.8 The composition of extreme poor in 1992 and 2002 by Household HeadCharacteristics 50 Table 3.9 The composition of the moderate poor by Household Head Characteristics (excluding the extreme poor) in 1992and 2002 51 Table 3.10 Decomposition of rural-urbandifferences inper capita income 54 Table 3.1 1Overview of Poverty Trends for Extremeand Moderate Poverty with Income andConsumption 56 Table 3.12 Inequality has declined at nationalbut risen inruralareas over the pastdecade, 1992-2002 60 Table 3.13 Patternsof income and spending growth, 2000-2002 62 Table 3.14 The evolution of real wages for different groups inthe urban labor force, 1994-2002 64 Table 3.15 MarginalValue of Education by Level along the Conditional EarningsDistribution, Mexico 1988-2002 66 ... Vlll Table 3.16 The evolutionof real wages for different groups inthe rurallabor force, 2000-2002 68 Table 3.17 Contributionof Transfer Income to the LevelinOverall Income, 70 2002 Table 3.18 The pattern of annual change inincomes inapanel of rural and urban 73 Mexicanhouseholds, 2000-2003 Table 3.19 Thepattern of annual change inincomes ina panelof Mexican households inpoor rural areas, 1998-2000 79 Table 3.20 Poverty by regionwith StandardDeviations 82 Table 3.21Municipalattributes, by 2000 classification of municipalities according to population 88 Table 3.22 Municipal attributes, by 2000-1990marginality reduction 89 Table 3.23 Poverty amongst indigenous groups 91 Table 3.24 Coverageof healthinsurance by region and indigenous status 92 Table 3.25 Municipal attributes, by presenceof indigenous 92 Table 3.26 Municipal attributes, by share of municipalmigrantsin2000 95 Table 3.27 Municipal attributes, by 1980-2000population growth 97 Table 4.1 FederalGovernment's overall pattern of revenuesandspending with respectto GDP, 1990-2002 110 Table 4.2 Growthinpublic spending categories, 1990-2002 113 Table 4.3 Public expenditure on 44redistributive"programs 116 Table 4.4 The distributionof spendingon all public programs analyzedwas roughly equal 2000-02 117 Table 4.5 The distributionof additional public education spending, by education level, 1992-2002 120 Table 4.6 The distributionof household expenditure before and after public transfers, 2002 122 Table 4.7 Indicators of deprivationover the life cycle 124 Table 4.8 Indicators of the coverage of selectedprogram inrelation to deprivation over the life cycle 125 Table 4.9 Numbers and spending on social programs, 2002 126 Table 4.10 A Stylized presentationof the Mexico's Social Protection Systemby Poverty Status (coverage rates refer to 2002) 133 Table 4.11Quality Educational Programs 150 Table 5.1 Both growth andinequality make a difference to the pace of poverty reduction 167 Table 5.2 Contributions to Mexico's past growth changes predicted by a global analysis of growthdeterminants 169 Table 5.3 Mexico's predictedfuture long-rungrowth, basedon a global analysis of determinants of growth 169 Table 5.4 Structural factors andmacro-stability are associatedwith changes in inequality 170 Table 5.5 There are large but declining differences betweenruralandurban wages 177 Table 5.6 The potential impactbetween selected structural reforms, inequality andpoverty 178 ix Table,5.7 Mexico has a highregulatory burdenininternational context 185 Table 5.8 The use of formal financial instruments increaseswith income 188 Table 5.9 School attendance andwork participation among 15 to 18 years old in marginal rural communities 190 Table 5.10 The composition of ruralincome in 1992 and2002 191 Table 5.11Programs insupport of incomes inrural areas 196 LISTOFFIGURES Figure 1.Long runtrends ininfantmortality andGDPper capita xvi Figure 2. Long runtrends insecondary school enrollment and GDP per capita xvi Figure 3. Overview of Poverty Trends for Extreme and Moderate Poverty xviii a. Extreme Poverty (food-based poverty line) b. ModeratePoverty (assets-based poverty line) Figure 4. Patterns of Growth for National Accounts, Survey Mean, and Bottom Quintile, 1994-2002 xxi a) Income b) Private Consumption Figure 5. Trends inInequality (Gini coefficient) xxii Figure 6. Unemployment and Informality inMexico xxiv Figure 7. Proportion of employed who earn less than two minimum wages at the state (left) andmunicipality (right) level xxv Figure 8. CONTIGO: Areas of public action over the life cycle xxvii Figure 9. The relationship between total tax revenues (in percent of GDP) and GDPper capita xxxi Figure 10. There i s huge variation in the inequality of different government programs xxxii Figure 11Public expenditure i s muchlarger inrelationto private spending for the poor xxxiii Figure 12. Closing the poverty gap: the impact on per capita household expenditure of the bottom five deciles of full reallocation of electricity subsidy andVAT subsidiesto Oportunidades xxxiv Figure 13. Investment inInfrastructure 1980-2002 xli Figure 1.1A simple conceptual framework for the changes inwell-being 10 Figure 1.2 The major actors and relationships of accountability for service delivery 12 Figure 1.3 CONTIGO conceptual organization of areas of public action 18 Figure 1.4 CONTIGO: Interactions Between Public Action Areas and the Life- Cycle 19 Figure 1.5 An application of the CONTIGO conceptual framework to rural development 21 Figure 2,l Trends inmeasuresof well-being 24 Figure 2.2a Difference in average years of education between top and bottom quintiles 26 Figure 2.2b Difference in average years of education between top and bottom X quintiles. B y age cohort, aroundyear 2000 27 Figure 2.3 Mexico has low average quality, but low dispersion, in the quality of education 28 Figure 2.4 The relationship between Income andMeasuresof Well-Being 30 Figure 2.5 The long-run relationships between incomes, infant mortality and secondary enrollments 32 Figure 2.6 Relationship between Income andMeasures of Governance 35 Figure 3.1 There have been steady gains intest scores across all school types 43 Figure 3.2 There have been large improvements in health indicators, including those relatedto diseases of the poor 44 a) Life Expectancy b) InfantMortality Figure 3.3 HouseholdInsuranceby Institution 45 Figure 3.4 Patterns of Growth for National Accounts, Survey Mean, and Bottom Quintile, 1994-2002 59 a) Income b) Private Consumption Figure3.5 Frequency Distributionsof per Capita Income 60 a) 2002 b) Rural, 1992,1996 and2002 c) Rural2000 and2002 d) Urban, 1992,1996 and2002 e) Urban, 2002 and 2002 Figure 3.6 The Proportion of Workers Earning Wages below Levels Correspondingto Different Poverty Lines; RuralandUrban2000-2002 64 Figure3.7 Yearly Ratesof Returnto Education Level 65 Figure3.8 Real UrbanWages by Informality 67 Figure 3.9 Wage GapbetweenInformalandFormalWorkers, 1997-2002 68 Figure 3.10 Remittances vs. Foreign Direct Investment 69 Figure 3.11 The Pattern of Annual Change in Incomes in a Panel of Rural and Urban MexicanHouseholds2000-03 73 Figure 3.12 The Pattern of Annual Change in Incomes in a Panel of Households inPoor RuralAreas, 1998-2000 78 Figure 3.13 Average grades of schooling at the state (left) and municipality (right) level (among those age 15 andolder) 83 Figure 3.14 Proportion of population ages 5 and older that speaks an indigenous language at the state (left) andmunicipality (right) level 84 Figure 3.15 Proportion of households who own no appliances at the state (left) and municipality (right) level 84 Figure 3.16 Proportion of employed who earn less than two minimum wages at the state (left) andmunicipality (right) level 85 Figure 3.17 Changes on socioeconomic variables over the 90s in Mexican municipalities 86 Figure3.18 Poverty and share of non-agricultural employment 90 Figure 3.19 2000 Share of population bornin a different municipality andlor state 94 Figure 3.20 1980-2000 population growth 96 xi Figure4.1 Mexico's tax collectioneffort i s low by international standards 111 Figure 4.2 Total and social spending has grown significantly in Mexico, apart from the crisis 112 Figure 4.3 Social spending was strongly pro-cyclical in the 1994-95 crisis but continued to grow inthe 2000-02 periodof stagnantgrowth 112 Figure4.4 Social spending is dominated by education, health and social security 114 Figure 4.5 Education and social security spending has grown most in absolute terms 114 Figure 4.6 There is huge variation in the inequality of different government programs 118 Figure 4.7 The inequality of rural and urban spending on programs in 2000 and 2002 119 Figure 4.8 Public expenditure is much larger in relation to private spending for the poor 121 Figure 4.9 Closing the poverty gap: the impact on per capita household expenditure of the bottom five deciles of full reallocation of electricity subsidy andVAT subsidies to Oportunidudes 130 Figure4.10 Number of beneficiary families of Opor~unid~de~ 136 Figure 4.1 1 Proportion of extreme poor, moderate poor and non-poor families receiving ~ p ~ r t u n i d transfers in2002 ~ e s 136 Figure4.12 School continuation rates inpoor rural communities intreatment and control villages under Progresa, late 1990s 137 Figure 4.13 Geographic patterns of public resource use in relation to state level GDP per capita: total resources, education spending, poverty-related spending andagricultural spending, 2002 153 Figure4.14 Geographic patterns of outcomes inrelation to federal transfers 2000 155 Figure 5.1 Investment in infrastructure in Mexico and Latin America fell steeply inthe 1990s 172 Figure 5.2 Average wages, growth andunemployment 174 Figure 5.3 The evolution of the relative demand for workers with tertiary education, Mexico UrbanAreas, 1988-2001 175 Figure 5.4 There are large differences in the pattern of wages amongst self- employed, informalandformal workers 181 Figure 5.5 Poorer workers experience greater informality across many dimensions 182 Figure5.6 Labor productivity inthe structured and non-structured sectors 183 Figure5.7 The policy framework of the Impulso strategy 184 Figure 5.8 The share of farm income intotal ruralincome has been falling, even for the extreme poor 191 Figure5.9 Returnsto education for childrenfrom marginal rural communities 193 Figure 5.10 Role of assets (human capital on horizontal axis) and economic context (vertical shift) on the incidence of poverty in 506 Mexican rural communities (Progresadata for 20,895 households) 194 Figure 5.11 Role of geographical location relative to a cluster for growth in manufacturing and services employment insemi-urban andrural municipalities 195 Figure 6.1 ResultsBased Management Framework 206 xii LIST OFBOXES Box 1. The Law on Social Development(Ley de Desarrollo Social) xxviii Box 2. Dimensionsof Decentralization xlv Box 3. Selectedexperiencewith ResultsBasedManagementinLatin America xlviii Box 1.1Mexico's Technical CommitteeonPoverty Measurementandthe Choice of IncomePovertyLines 6 Box 3.1 Usingthe Oaxaca-Blinder Methodfor Rural-UrbanDecompositions 53 Box 3.2: Poverty trends in2000-2002 55 Box 4.1 PRONASOL andthe political useof programs 131 Box 4.2. Veracruz 156 Box 5.1 Examplesof income generationamongstindigenous groups 201 Box 6.1 Building Results BasedManagementsystems: Lessons Learned 211 Box 6.2 Definitions: Monitoring andEvaluation 213 Box6.3 Engagingrelevant governmentandnon-governmentactorsinthedesign andimplementationof aRBMinMexico 217 ... Xlll Poverty inMexico An Assessmentof Conditions, Trends andGovernment Strategy Summary and Key Messages This report presents the results of an assessment of poverty conditions in Mexico and government strategies to reduce poverty. The report constitutes the first phase of a longer-term work program on poverty reduction in Mexico that the World Bank is undertaking in collaboration with the government of Mexico between 2003 and 2006. While the work has involved extensive collaboration with both the government and Mexican specialists working on poverty, the views expressed are those of the World Bank alone. This summary distills the report main conclusions and messages. The current situation is mixed with respect to poverty. In terms of well-being, Mexico has experienced major progress in some dimensions -notably related to basic service access- but much weaker progress on others -notably on the income of the poor. Despite the gains between 1996 and 2002, particularly for the extreme poor, poverty remains widespread andis only slightlybelow levels prevailingbeforethe 1994195crisis.' Related to government strategy, there are many strengths. The CONTIGO framework i s excellent as a conceptual framework, especially in its attempt to integrate the multiple dimensions of well-being and public action into a life-cycle approach. There are notable successes in specific programs -with ~ p o r t ~ n i d ~unusual in its combination of e s highly effective targeting and broad reach amongst the extreme poor. The emphasis on evaluation (especially in SEDESOL) i s commendable. The recent introductionof the Ley de Desarrollo Social is a potentially valuable attempt to institutionalize social development strategy and in particular provide more continuity across government administrations. But there are also many challenges. There i s some way to go to implement the principles of the CONTIGO conceptual framework across government programs. The quality of services is a major issue in many sectors. Social policies for the extreme poor are well developed, but not policies for their income growth. There is a broader pending agenda for the moderate poor, especially with respect to improving the productivity of the self- employed and informal enterprises. Most of the extreme and moderate poor fall outside the formal social protection system, and face significant risks, for example from health, unemployment or lack of income in old age. And there are a wide range of institutional issues to be tackled -from strengthening accountability, especially under decentralized structures, to social incorporationof excluded groups and linkingof rigorous evaluation to results-basedmanagement. 'Thisis defined in terms of those with incomes below the food-based poverty line. xiv Section A discusses poverty conditions and Section B reviews government strategy. Section C provides brief concluding comments on public action and future analysis. A. Poverty conditions andtrends Despite significant government efforts, poverty in Mexico remains widespread, and is closely linked to high levels of inequality. Poverty has many dimensions, including human capacities (especially education and health status), access to infrastructure, incomes, vulnerability and social status. Recognizing these multiple dimensions can be thought of as an imperfect approximation to the approach to well-being developed by . Amartya Sen, in terms of a person's freedom (or capability) to pursue a life of their choosing. This depends both on their human capacities, and their ability to convert these into "beings anddoings".2 This ability is powerfully shapedbothby the economic context -especially opportunities for productive work- and the political and social context in which people live. Inthis regard, Mexico has been going through two major transitions. Inthe economic sphere it has beenexperiencing a major deepeningof integration inthe international economy, while inthe political and social sphere ithasbeen going through a process of deepening of democratization. Yet both are complex and of a long-term nature. These transitions are central aspects of the environment for living conditions and government policy. Conditions and trends are reviewed for the various dimensions of poverty -that are matched to the dimensions laid out in the Government's CONTIGO strategy. We also review geographic aspectsof poverty. (I)Human c u p i t a l ~ c a p ~ ~ ~ e ~ Mexico has made important progress in terms of the humancapacities of the population, including health, nutrition and education. For both health and educational status and levels of service provision, Mexico is now reasonably close to where it mightbe expected to be for its income level. There have been large, long-term advances with respectto both infant mortality and secondary enrollments (Figures 1and 2). Inboth cases the progress was as good or better than the East Asia success^' case of Malaysia, one of the high- performingEast Asian countries that has the most similar structural and socio-economic characteristics to Mexico. However, despite this success on average, large differences exist bothbetween social groups, andbetweenregions within Mexico. See Sen, 1999. xv Figure 1.Long runtrends ininfant mortalityand GDPper capita 200 I 180 160 1M =444.04-78.39*1n(G DP)+3.49*ln(G DP)' R'= 0.6511 g 140 120 8 =P 100 0- & 8 0 6 0 40 2 0 0 0 2,000 4,000 6,000 8,000 10,000 12,000 G D P p e r capita (1995 U S D ) Source: WB staff calculations from World DevelopmentIndicators. Figure 2. Longruntrends insecondaryschool enrollmentand GDPper capita 100 80 6 0 40 G S E = -94.6-24.09*1n(G DP)-0.49*ln(G D P ) * R'= 0.6071 20 N=Z,OOS 0 00 -20 G D P p e r capita (1995 USD) Source: WB staff calculations fromWorld DevelopmentIndicators. Progress in human capacities was supported by a combination of expansion of programs that have aimed for universal service provision and have been increasingly pro-poor as they have expanded (especially basic education, public health programs, and water supply) and, in recent years, demand-side measures, that combine transfers with incentives for income-poor households to send their kids to school and attend health clinics. Oportunidades i s the flagship program for these transfers. These measures have been supported by a large expansion in social spending by the government -especially for education andpoverty reduction programs (see below). While this progress i s of a long-termnature, it has been consolidated anddeepenedunder current administration. Growth in social spending since 2000 i s especially commendable inlight of the stagnation ingrowth andrevenues inthis recent period. By contrast, public xvi spending on health has declined in the 2000-02 period, despite significant issues in the ~ector.~More generally, there remain major issues of low and uneven quality for both education andhealth services. (2)Assetspover~/P~t~monio With respectto infrastructure there have beenmajor advances in access to electricity and improved water sources4, and a more modest expansion in the coverage of improved sanitation. At the national level access rates were 9896, 90% and 80% respectively for electricity, water and sanitation in2002. While the poor have systematically lower access rates, they also experienced gains in the past decade. For example, for the rural extreme poor -the most disadvantaged group with respect to services- access to electricity increased from 63 to 90% between 1992 and 2002, while access to improved water increased from 38 to 58% in the same period. By contrast, access to improved sanitation for this group only rose from 22 to 26%. There has also been steady growth inthe ownership of household durables, that i s driven more by private savings and investment. However, there remain large inequalities in the ownership of housing and financial savings, with very little use of formal financial institutions. For example, while close to 70% of the moderatepoor own their house, only 5 percent reported that they were paying for their houses- an indicator of the predominantly self-financed processof house acquisition and construction. There i s evidence of Mexico lagging behindcomparators -especially East Asian ones- in infrastructure provision, and this could emerge as a major problem for income dynamics in the future. Mexico's investment fell in absolute terms in the early 19900's, and despite some recovery, still remains below 1990levels. This occurred ina periodin which the average shortfall inLatin American infrastructure relative to fast-growing East Asian countries was rising (Calderh andServtn, 2003). (3)Incomep o v e ~ / O p o r t u n ~ desingreso e Poverty changes in terms of incomes are driven by the interactions between growth and income inequality. Even with steady growth, poverty reduction tends to be slow, as a consequence of Mexico's high income inequality. And growth has not been good. One way of reading Figures 1 and 2 i s of large advances in social indicators, despite markedly slow progress along the income axis -with Malaysia leapfrogging over Mexico from an initially lower levelof mean income. Over the past decade, the pattern of overall poverty changes has closely followed the macroeconomic cycle andthe associatedrhythmof the labor market (Figure 3). The crisis of 1994-95 constituted a massive setback. Extreme poverty in 2002 was 16 percentage points below the 1996 level, but still only one percent below the pre-crisis 1992 level. Actual spending for 2003 is not yet available. Estimates indicate that health spending grew almost 24 gercent from 2002 to 2003 (SHCP, 2003). Access to potable water: runningwater inhouse, runningwater out of house, andby pipe-track. xvii While open unemployment i s not of great importance for the poor, low labor returns (in self-employment and wages) andunder-employment are, especially so for the poor. While poverty has only recently recoveredfrom pre-crisis levels, ,the measuredtrends in the 2000-02periodare encouraging, with a significant decline inextreme poverty, despite stagnation in average incomes. This was driven by a combination of substantial income growth inrural areas, combined with a decline in inequality inbothrural and urban areas inthisperiod, Figure3. Overview of Poverty Trends for Extreme and ModeratePoverty a) Extreme Poverty (food-based poverty line) 1992 1994 1996 1998 2000 2002 +National 1 Rural &Urban xviii b) ModeratePoverty (assets-based poverty line) 90 80 h 70 Ib aL 60 .*5 G -.-650 a a 40 c 2 30 r)i .I2 20 10 0 1992 1994 1996 1998 2000 2002 +National Rural +Urban Source: WB staff estimatesusingthe methodology of the Technical Committee for Poverty Measurement. There has been some debate within Mexico as to whether the measured changes in poverty reflect lack of comparability across surveys, since the 2002 ENIGH involved changes in both the questionnaire and the sample size. Recently, the Technical Committee on Poverty Measurement (CTMP) concluded that the 2000 and 2002 surveys are broadly 'comparable, at least for total incomes of households (CTMP, 2004). This report concurs with the CTMP's conclusions. It judges the 2002 surve to be superior, andcomparable with 2000 for total household (andper capita) income. This report also Y provides estimated confidence intervals for poverty comparisons, where feasible. This conclusion i s further supported by comparison with independent evidence on trends, especially from the series of labor force surveys (the ENET since 2000), for which the survey instrument didnot change. While the pattern of measured changes in income poverty i s consistent with the official estimates, not all the changes were sufficiently large to be statistically significant. The most important changes that were statistically significant were the notable reduction in extreme poverty in both rural areas and at the national level. When poverty i s measured in terms of consumption rather than incomes, declines in moderate poverty were also statistically significant at both rural and national levels; falls in extreme poverty at rural andnational levels continue to be significant. Measured declines inurbanpoverty are not significant whichever concept is used. The results both from the ENIGH and from the ENET (the labor force survey), are consistent with the view that there was growth in income of the poor despite economy- wide stagnation, driven by two factors: first, rapidgrowth inlabor incomes, especially for At the level of sub-categoriesof incomes-such as transfers-the surveys were not designedto be strictly comparable,andthe statistical assessment has to be undertakenon a case-by-case basis. xix rural unskilled wages (that increasedby over 20% between 2000 and 2002, according to the ENET), with some reduction in the premia to higher levels of education and formal urban work; and second, a substantial measured contribution from remittances and transfers (that include Oportunidades)for the ruralextremepoor. To put Mexico's poverty reduction in an international context, we can compare it with the most recent global poverty analysis that uses standardized poverty lines of approximately one and two dollars a day, at 1993 Purchasing Power Parities (Chen and Ravallion, 2004). While the methodology is different, the two dollars a day line i s quite close to Mexico's extreme, food-based poverty line, and the patterns over time are similar. As Figure 4 shows in relation to the Latin American average, Mexico has generally had slightly lower poverty incidence. This is because incomes are above average, and inequality close to the (high) Latin America average. Over the past two decades, and over the past decade, progress has been weak in both Mexico and Latin America, but Mexico's performance has been faster than the average since the 1994195 crisis. East Asia, by comparison, started very much poorer than Latin America two decades ago, but then made much faster progress. Only some five to ten percent of Mexico's population still live under $1 a day. This is a line that is close to those used in the some or the poorest countries in the world, and would represent a deep level of deprivation in the mexican context (as well as probably being more subject to measurementerror). Figure4. Tendenciesinpoverty incidence inMexico, LatinAmerica andEastAsia usinginternational poverty lines. (a) At "two dollars a day" (at 1993purchasingpower parities) - _---- = I - l o' E a s t Asia - L a t i n A m e r i c a a n d C a r i b b e a n- - -M e x i c o xx (b) At "one dollara day" (at 1993purchasingpowerparities) 60 3 50 E 4H ' 40 = .-*30 c$a20 2et10 c a s _e- - --u *--*-- 0 a 3 - East Asia ~- ~ S R Latin America and Caribbean- - - ~ ~ Mexico Note: The World Banks 1993consumption PPP exchangerate for Mexico is 2.102. The PPP one dollar per day poverty line is about 68.8195 pesosper personper month in 1993 prices. Source:Chen and Ravallion(1994) and WB staff calculations. The impact of the macroeconomic crisis and recovery i s abundantly clear from the poverty trends shown above. However, there have been notable changes in the relationship between macroeconomic performance and poverty reduction in the past decade. This i s illustrated by the significant divergence between growth in income and consumption from the national accounts, survey means, and the mean income of the bottom 20% of the distribution (Figure 5). Changes in measured average income and consumption from the surveys do not track National Accounts figures because of different concepts and measurements. Both suffer from measurement error, but survey- based evidence is generally considered a better measure of the welfare of households. In the 1994-95 crisis, income and spending from the surveys fell much more than in the national accounts, for both incomes at the mean and the bottom 20%. Survey-based income and spending rose fast inthe 1996-2000period, but less for the poorest. Finally in the 2000-02 period, survey and national accounts estimates both show little change, but there i s substantialgrowth for the poorest 20 percent. xxi Figure 5. Patterns of Growthfor NationalAccounts, Survey Mean, andBottom Quintile, 1994-2002 a) Income i b) PrivateConsumption 40% I 30% I 205L 10% 0% iI1 .lo% j i -20% -30% 1L -___- _-___ 199419% 19962ooo 2aX)-2002 1994-2002 Period HMmIeConsumptionpercapita MeanCurrentExpendlturspercaplta .Mean CurrentExpenditurespercapitalsf qufntiie Source: WB staff calculations usingENIGH, andNational Accounts. The observed changes in poverty reflect changes inboth mean income and in inequality. High levels of inequality are associated with higher poverty, and are also of interest in their own right. Inequality has tended to be counter cyclical (in contrast to characteristic patterns in Latin America in the 1980's), with the 1994-95 crisis mildly equalizing, the 1996-2000 recovery disequalizing, and the 2000-02 period of stagnation equalizing for incomepigwe 6). Mexico has movedfrom being around the (very high) Latin American average, to slightly below the average in 2002. While the pattern of inequality changes may in part reflect genuinely cyclical forces, there also appear to be longer-term structural factors at work. xxii Figure 6. Trends inInequality (Ginicoefficient basedon income) 60 55 50 45 4 0 . 35 30 1992 1994 1% 1998 UHW) 2002 LatinAmericaandthe Caribbean(withoutMexico) -CRural +Urban +National Source: WB staff calculations from ENIGH, various years, using the approved methodology by the Technical Committee for Poverty Measurementfor the calculation of income. Note: The Latin American averages are unweighted, there is little difference when using weighted averages. There are several features of the changes in inequality that may reflect structural forces, and are noteworthy. First, the returns to tertiary education of workers in the labor market rose significantly in the decade to 1997, but have fallen since. This i s generally attributed to the skill-biased technological changes induced by the large-scale opening of the Mexican economy to trade and foreign direct investment, which was consolidated under Chile -which also went through a profound opening to international markets- also experienced a large rise and modest fall in the wage premium to college graduates, suggesting that the effect of opening couldbe transitional, especially with vigorous expansion of education. Second, the last decade has witnessed a reduction in average wage differentials between ruraland urbanareas, probably reflectingrisingintegration between rural and urban labor markets, especially in the more developed parts of Mexico. This has been reinforced by relatively large remittance flows to ruralareas in the more recent past . Third, within rural areas, there has been a significant rise in inequality since the mid 1990's,with a sharp increase through 2000, followedby amore moderate fall See De Ferranti et a1 (2003) for discussion of overall patterns of wage inequality in Latin America and Lederman, Maloney and ServCn(2003) for a review of the effects of NAFTA. xxiii in2002 (Figure 5). The 2002distribution is now morelikethe 1992pattern.There has tended to be more rapidgrowth inhigher-productivity agriculture, and innon- farmruralincome growth inareas closer to manufacturing andservicecenters. (4) Vulnerabili~/ProteccidnSocial Vulnerability to a variety of adverse shocks is widespread, affecting ruraland urbanpoor and non-poor alike. Adverse shocks due to ill-health, employment conditions, meagre harvests and natural disasters tend to deepen poverty for the poor, and cause some of the non-poor to fall into poverty. Rural and urban workers appear to experiencequite similar patterns of income fluctuations, though the self-employed and informal sector workers experience higher income variability; this is particularly marked in rural areas, where farming is a major activity for the self-employed and informal workers. Large-scale crises, such as the 1994-95 macro shock, affected aEZ groups adversely, without markedly altering the pattern of changes across group^.^ The major gain from the perspective of vulnerability since 1995 has been the avoidance of a macroeconomic crisis. Yet there remain important covariate risks associated with natural disasters and the climate, and idiosyncratic health and employment risks. There is also the risk of having insufficient savings or support for old age. Some of these are systematically related to the characteristics of households -as with the greater fluctuations of incomes for the self-employed. While many individual workers move to the informal sector out of choice, there are structural reasons for the large size of the sector, associated in particular with the regulatory and tax burden of formalization and the weak links between social security costs and benefits. An important correlate of this i s the large fraction of the population that is not covered by formal mechanisms for provisioning against risks. The proportion of the work force in formal work experienced a large long-run decline from the late 1980's to the mid-l990's, followed by a partial recovery in the late 1990's, while the 1994-95 crisis and the 2000-02 period of stagnation tended to increase both unemployment andinformality (Figure 7). Chapter 3 providesthe analysis, findings, and literature review on vulnerability inMexico. xxiv Figure 7. Unemployment and Informality inMexico 8 1 . 4 In the recent period, weaker labor market conditions were reflected more in higher unemployment rather than lower wages. This may reflect a change in how the labor market is functioning, that could be associated with low inflation, that means that the mechanism of effecting broad real wage declines via price increaseshas to a large extent gone. It i s possible that future shocks could be associated with more long-term unemployment, as in countries such as Argentina and Colombia. This has potentially important implications for social protection policy, increasing the need for efficient mechanismsfor dealing with unemployment risk. (5) Socialpover~/Incorporaci~nsocial This important area is the hardest to assess quantitatively. A tentative qualitative assessment is that problems of both social exclusion and weak accountability of public institutions to poor groups remain of widespread importance, despite important progress in the recent democratic opening. There have been very valuable initiatives regarding greater transparency and accountability under the present government, but experience from other countries shows that for greater transparency to be eflective it i s crucial to have complementary institutional structures and processes in place. This can be challenging and can take time, especially where there are strong traditions of patron- client relationships and social exclusion. There is now a need to take stock of how much efforts inthis area are leading to better services andresponsive to poor groups. The position of indigenous groups is of particular importance because of historical patterns of social exclusion and deeper levels of poverty. The ENIGHdoes not include a question on ethnicity, but according to the 2000 Census, 4.4% of indigenous groups are in the bottom 20% of the overall distribution of income, and 80% inthe bottom 50%. Using the bottom quintile as a measure of extreme poverty (this i s close to the incidence under the food-based poverty line from the ENIGH), indigenous peoples account for about a fifthof the extreme poor, that is over twice their population share accordingto answersto the question on language at home in the 2000 Census. Indigenous groups typically suffer higher levels of deprivation interms of education andhealth status and access to services. (6)Spatial aspects of well-being xxv Cuttingacross the different dimensions of well-being is variation by geographic location. Living in a poor area can make a profound difference to life prospects. There are large differences in income poverty and other indicators of well-being across different regions, with a generalized gradient from north to south. The highlevel of deprivation inthe south i s confirmed by both CONAPO's index of marginality (based on access to basic infrastructure services, housing conditions, education attainment, and wage earnings), and the UNDP's Human Development Index (based on per capita GDP, educational achievement and enrollment, and life expectance). For both indexes, the southern states have the lowest rankingsamong all states (World Bank, 2002). These overall regional differences hide a great deal of spatial heterogeneity within states, both in terms of levels of well-being and patterns of change. In terms of numbers, there are large groups of the extreme poor living outside the poorest states (roughly a quarter of all the extreme poor inMexico live inurban areas inthe center states). Indeed, all partsof Mexico have steep gradients in conditions of living, from more developed urban areas, through peri-urban areas and smaller towns, to the more remote rural areas (as well as significant inequalities within any location). The heterogeneity across and within states is, illustrated by one indicator -the proportion of employed people earning less than two minimumwages inFigure8. A similar pattern of heterogeneity exists for other measures of well-being, such as schooling, access to various services and ownership of consumer durables. Figure 8. Proportion of employed who earn less than two minimum wages at the state (left) and municipality (right) level Source: WB staff calculations from 2000 Census. There i s an overlap between indigence and spatial indices of well-being, especially for rural indigenous groups. There are high concentrations of indigenous groups in many municipalities inthe south Pacific states and inthe Yucatan peninsula as well as in some municipalities of the west andnorthwest of Mexico. Among those who are self-identified xxvi as indigenous, or speak an indigenous language at home, a relatively highfraction live in relativelysmallrural villages, with low levels of services. Interms of trends, regional differences have long historical roots. There has been some long-term spatial convergence in most indicators of services and social conditions, but a tendency toward divergence in income and wage measures inthe 1990's, that appears to be associated with the differential effects of heightened international integration both before and after NAFTA. Areas closer to the border, or to urban centers, typically grew faster. B. GovernmentStrategy Inthis section we turn to the role of government strategy inreducing poverty. Itcovers the following: an overview on the administraion's CONTIGO framework for poverty reduction; issues of public finance and the distributional incidence of spending (this draws on work conductedjointly for a parallel public expenditure review and this report); and key issues of policy design. The third subsection on policy design discusses policies on social service provision, policies for income growth of the poor, and a set of cross- cuttinginstitutionalquestions. The Government's CONTIGO framework for poverty reduction and social development The Government has framed its strategy for reducing poverty in terms of an explicit framework on the nature of poverty andhow public action may improve the well-being of the poor. This framework is known as CONTIGO, which i s both a conceptual framework, and a set of programs that are particularly concerned with poverty reduction under the social cabinet. As a conceptualframework CONTIGO is excellent. Two features stand out (Figure 9). First, it explicitly recognizes both multiple dimensions of poverty and the complementary domains of public action, including human development, income generation, physical asset accumulation and social protection. Second, it embeds this within an understandingof well-being and poverty over the life cycle of individuals and households. It is unusual to have such a well-developed and coherent approach to assessingpoverty andframing policy as part of official policy. Within the CONTIGO framework, there are two areas that couldbe articulated further to underpin analysis and action. First, more attention could be given to the design of the supply side of social policy -articulating a framework for understanding the determinants of the coverage and quality of the full range of public services that affect the lives of the poor. A particular dimension of this concerns how the federal government can influence the coverage and quality of provisionof services at subnational levels; this i s of rising significance, in light of the relatively recent extensive decentralization in Mexico. Second, there is a case for giving more weight to issues of formal and informal institutional conditions, especially with respect to social incorporationand accountability. These are central determinants of both service performance and economic potential of poor groups. While issues of social inclusion and accountability are indeed treated as xxvii between social and economic development, and it is crucial that a poverty reduction strategy is not limited to a social development strategy. This assessment of CONTIGO'S operationalization in the initial years of the administration is preliminary becauseof the difficulty of analyzing intended impacts, and the small amount of readily available information on actual progress in terms of results. This is not a critique of the work of the social cabinet -who have made significant advances over the past three years- but rather a comment on the complexity of the challenge and the inheritance of programs from previous governments. Coordination between social and economic programs is made more difficult since the social cabinet is concerned primarily with social programs, while many complementary economic programs fa11under other organizational structures, includingthe economic cabinet. The government is now revisiting the operationalization of CONTIGO, placing the coordinating function of the social cabinet under the Presidency and exploring ways to focus on a more limited number of priority programs, rather than seeking to coordinate all programs that are important for poverty reduction. These appear to be sound moves. Inaddition to these developments within the executive branch, Congress recently passed a Law on Social Development (Box 1). This represents an important step to institutionalize a state policy for social development. It institutionalizes social policy in Mexico through a series of bodies to oversee social development policy and implementation, including a board for evaluation, a national commission on social development, an inter-ministerial commission and an advisory committee, which will include members from civil society. It also states that the principal (though not exclusive) executors of some social development programs will be municipalities, and includes a requirement that social development spending not be reduced in real terms in any year. The law has challenges in implementation, especially with respect to the workings of the various commissions created, and the primary emphasis given to municipalities. While the guaranteeon not reducing social spending as a share of GDP is desirable inprinciple, this introduces a significant inflexibility into the already inflexible budget (World Bank, 2004). As discussed below, not all social spending is effective and pro-poor, and other categories of spending (including parts of economic sector spending) are important for poverty reduction. Box 1. The Law on SocialDevelopment(Ley de Desarrollo Social) The Social Development Law, enacted in late 2003, establishes the principles and general guidelines to shape and institutionalize a state social policy that transcends the sexennial policies.' The law establishes the creation of a National Commission of Social Development, the creation of an Evaluation Council, and the budgetary mandates for social development programs that will permit the effective planning and coordination of social policy in Mexico. The law includes a chapter on priority areas; regions where inequalities are deeper and the conditions of poverty and marginality are broad and complex. It seeks to concentrate the energies and resources of public institutions and society so that Mexicans suffering high levels of deprivation in these areas reach, in the least possible time, acceptable levels of well-being. The * The Law will be fully implemented once the President decrees the specific operational rules and guidelines. xxix law stipulates that the principal executors of the programs, resources, and federal actions for social development will be the municipalities, except in those cases where actions or programs are explicitly assignedto the federal governmentor another institution. One of the contributions of the law is in the guidelines and criteria for the financing of the social policy. These are defined in such a way that the programs, funds, and resources aimed at social development are consideredpriority and of public interest. Spendingon these programscannot be lower, in real terms, than inthe previous fiscal year and must increase, at least at the rateof growth of GDP. Itempowers the federal government to establish and administer a fund of social contingency in light of economic and budgetary developments of unforeseenphenomena, of which the amount and minimum rules will be included in the FederalExpendituresBudget. The institutions createdby this Law are: 9 National Systemof Social Development.This systemwill group actorsand institutions concerned and involved with social development. I t will be established as a permanent mechanism of attendance, collaboration, coordination and consensus building of the three levels of government, as well as civil society andprivate sectors. P National Commissionof Social Development.This entity will be aninstrumentof coordination of the programs, actions, and investments for the achievement of the objectives, strategies, and priorities of the national policy of social development. The creation of this entity is fundamental for institutionalizing social programs. 9 National Evaluation Council. Thisentity maycarry out itsdutiesby itselfor throughone or more independent agencies. Its objective is to periodically review the goals, actions and results of social programs, in order to correct, modify, add, reorient or suspend programs. Institutions of higher education and of scientific research, and nonprofit organizations can all participate as evaluation agencies, chosen through a competitive biddingprocess. The law stipulates that for the mandated performance audit, the evaluations of social programs must include outcome, management, and service indicators in order to measure and report on their coverage, quality, and impact. The existing evaluation mandatenow acquires the status of a legal. requirement. This helps assures its continuation into the future. The Council also lays the foundations, guidelines and criteria for poverty definition, identification, and measurement. P Inter-secretarial Commission of Social Development. This entity will be an instrument of coordination of the actions of the Federal Executive in order to guarantee the comprehensive natureinthe design andexecution of the national policy of social development. P Advisory Committee of Social Development. This entity will analyze and propose programs and actionsthat affect the fulfillment of the national social developmentgoals. Source:Diario Oficial de laFederaci6n(2004). The roleof public finance insupporting programsfor poverty reduction There has been a substantial increase in social development spending since the 1994-95 crisis, including support for basic services andat least some well-targeted subsidies? This was important to some of the service coverage gains and reductions in extreme poverty. The Ministry of Finance's functional classification of spendingincludes the following three categories of spending: "economic sector", "government management", and "social development". "Economic sector" spending covers infrastructure, rural development, energy, transport, communication, and other services and economic activities, while social development spending includes education, health, social security, social assistance, regional and local development, and labor policy. Government management includes legislation, law and order, national security, election processes, governanceand environment. xxx While overall spending rose an average of 6.3% per year from 2000 to 2002, social development expenditure rose an average of 5.3% per year, and poverty reduction spending" rose 14.2% per year. Overall, the growth in social spending since the mid- 1990's was achieved despite fiscal astringency, especially through large reductions in "economic sector" spending. Expenditure on programs specifically targeted to the poor now represents 1.3% of GDP compared to 0.7 percent in 1990. Programs involving transfers to the poor, led by ~ ~ o r t u n i d sincesit was created, grew by an average of 8.4% per year duringthe ~O'S, ~ e and by even more -9.8% per year- after 2000. However, the most striking pattern i s the rapid pace of growth of social security spending - o f 35.2% per year since 1990- that is not targeted to the poor. The long-run reductions in economic sector spending are a possible cause for concern from a poverty reduction viewpoint because of the potential impact on both aggregate growth and the inclusionof the poor within the growth process. This is especially true in light of the different patterns of progressindifferent dimensions of poverty, notedabove -with much weaker gains in income poverty than in human capital. There has been some recovery in economic sector spending in the 2000-02 period, but spending levels remain below those prevailing in the early 1990s. It is, of course, crucial that economic sector spending be effectively oriented toward pro-poor growth. This is discussed further inthe section below, whenwe discusspolicies for income growthof the poor. In light of the likely need to increase economic sector spending, continued demands for social development spending, and the transitional costs of the pension reform, the fiscal position for spending affecting poverty reduction will be very tight inthe absence of a tax reform. A tax-increasing reformcan be one of the most poverty-reducing actions the state could take, provided increased resources are used eflectively. Along with many Latin American countries, Mexico has a Low level of tax revenue (Figure 10) and this greatly reduces the capacity of the state to take redistributive action. Evenaproportional increase in taxes can be highly progressive if resources are well used. The latter implies a combination of channeling of incremental resources into areas that are documented to be effective now, while linking incremental spending in other areas to institutional reforms to make spendingmore efficient and equitable. Incontrast to earlier periods there is now greater capacity to design transfers to offset lower benefits of a tax increase for poorer groups. loThis refersto programsspecifically targetedto the poor. xxxi Figure 10. The relationship between total tax revenues (in percent of GDP) and GDPper capita 0.45 S 0.40 I = I = 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 Log of GDP per capita (1995 USD) 1990-2000 Average Source: WB staff calculations, basedondatafromthe World DevelopmentIndicators. While increasing taxes is an important part of a poverty-reduction strategy, there is also great potential for more effective use of public spending." Within social spending plus subsidies, there is enormous variation in incidence, with some programs highly equalizing (mainly benefiting the poor) and others highly unequal (mainly benefiting the non-poor).12 On average, public spending appears to be much more equal than incomes, and at the margin, is becoming more equal and pro-poor, but there remain large areas of spending with very unequal incidence, and some with rising inequalities in incidence, for example ISSSTE pensions, tertiary education and IMSS benefits for active workers. This i s illustrated in Figure 11using a summary measure of the incidence of spending across different income groups in the population: the concentration coefficient (that is similar to the Gini coefficient) goes from -1 when all spending goes to the poorest, through 0, where everyone receives the same spending, to +1 when all spending goes to the richest. ''Spending analyzedhereaccountsfor over half of total programmablespendingand almost 1% of GDP in 2002. The coverage is primarily on social spending, plus subsidies-since few economic sector programs canbe linkedto specific households. The incidence analysis involves matching the use of programs from survey data to administrative costs from budget data. It makes no allowance for quality or for behavioral responses from households (Van de Walle, 2003). Despitethese limitations it is a useful measure of the orientation of governmentefforts across the population. Households are ranked in terms of "autonomous" spendingper capita, or spending before monetary transfers. This concept is directly comparable to autonomous income (or incomes pre-monetary transfers), with the assumptionthat the marginal propensity to consume from transfers is 50% (for details on the methodologysee World Bank, 2004a). xxxii Figure 11. There is huge variation in the inequality of different government programs (Concentration coefficients for 2000and 2002) Is 4.8 4.6 -0.4 -0.2 0.0 a2 a4 0.6 0.8 Note: The GINI refers to the distribution of private income. Procampo is not reported in 2000 becausethe sample of beneficiaries captured b the 2000 survey is not representative. VAT fiscal spending refers to exemptionsand the zero VAT rate.Y3 Source: Background work for World Bank (2004) and SHCP (2004) for the VAT fiscal spending. While the distribution of government social spending across different groups is only mildlyunequalon average, with wide variation, the size of spending in relation to private (pre-transfer) spending is much larger for poorer groups. This i s precisely because their incomes and spending are so low. Figure 12 shows the ratio of the value of measured public spending going directly to households to their private (pre-transfer) spending. The cost of such public programs is actually larger than private spending for the bottom three deciles of the rural population, and over 50 percent of the private spending of the bottom 20 percent of the urban population. This is only part of the story of their benefit to households, of course, since that also depends on the quality of the services, and how much they are valued by the recipients. Question of quality, which are discussed further below, are of major importance. l3 For details, see chapter4. xxxiii Figure 12 Public expenditure is much larger in relation to private spending for the poor (The ratio of program spendingto autonomoushouseholdexpenditure, by major program, national, 2002) 10 Primlymu. W SSA opoltunidades Lower Sec. mu. Preschool HIMSSHealth WProcampo UpperSec. Edu, ElectricitySubsidy PEMEXHealth IMSS Active Workers W ISSSTEHealth 2 W Tertiary mu. 1 0% 20% 40% 60% 80% 100% l20% 140% Source: WB staff calculations; see also World Bank 2004a. A further aspect of the story concerns the reach of different programs in relation to the "needs" of different groups from the poverty profile. The analysis for this report suggests a mixed picture. Some basic services -basic education, water, electricity- have near- universal coverage among the whole population. While not necessarily highly progressive, they reach most of the poor. Ingeneral, programs that are highly targeted to the poor actually have relatively modest reach -both in Mexico and elsewhere in the region (De Ferranti et al, 2004). Oportunidades is an important exception, with already 60% reach among the rural extreme poor from the household survey in 2002 and an estimated total 80% reach in 20O3l4. There remain gaps in basic service provision, especially for the extreme poor, and incomplete coverage of the social protection system -for health, old age, and unemployment. These are particularly marked for those amongst the moderate poor who have neither access to the formal social security nor to Oportu~idades. The analysis of incidence and reach suggests there i s considerable scope for redistribution within the budget for greater impact on the poor. To explore this, scenarios were developed which involved shifts either from the electricity subsidy or from the removal of exemptions under the VAT, with reallocations going to a hypothetically expanded l4Informationprovided by SEDESOL: xxxiv O~ortunid~des.These programs were selected for the scenario since they would in l5 principle be good for both efficiency and equity. The electricity subsidy and VAT exemptions are quite inequitably distributed now (with concentration coefficients of 0.24 and 0.41 respectively), implying more of the resources go to richer groups. Despite this, they do account for non-trivial amounts of spendingof the poor: electricity subsidies are equivalent to around 3% of private spending of the extreme and moderatepoor, and VAT exemptions are equivalent to some 4.5% of their private spending. However, if these programs were cut and resources saved were redistributed, for example, with a pattern similar to Op~rtunid~es, the net benefits to the poor would be substantial.l6Oportunidudes is equitable (with a concentration coefficient of -0.58) and probably has positive efficiency effects, since it provides incentives for investment in human capital. Two scenarios for redistribution are shown in Figure 13. The first redistributes the resources in line with existing distributions of programs, while the second assumes the 2002 gaps in coverage of households by O p o r ~ u n i would ~ ~ e ~be fully filled for the bottomdecile, 50% for the second decile and 25% for the third decile. Under these assumptions the bottom three deciles would experience large gains - making a substantial impact on extreme poverty- while the fourth and fifth deciles (as well as higher deciles) would experience small losses. While this i s only one among many illustrative scenarios, it shows the great potential for more effective use of the budget as an instrument of efficient redistribution. For example, for illustrative purposes on the tax side, we selected a tax change (exemption removal) that would have some negative effects on the poor which are then more than offset by the redistributive action. To the extent the same tax gains could be obtained by other tax reforms with less effect on the poor, the potential benefits from redistribution would be even greater -if potentially at some cost in terms of efficiency or administrative feasibility in the short run. '' '' These should be treated as illustrative only, since a further expansion of Oportunid~des,for example to roups not now covered, would raise new questionsof design. It is important to mention that this impact will not be achieved if another instrument is used instead of O ~ o r t ~ ~ i d a ~ e ~ . xxxv Figure 13. Closing the poverty gap: the impact on per capita householdexpenditure of the bottom five deciles of full reallocation of electricity subsidy and VAT subsidiesto Oportunid~es 14.000 m P r e - t r a n s f e r -Post-transfer ( C o r r e n t ) 12,000 W P o s t - t r a n s f e r (Reallocation) W P o s t - t r a n s f e r (Reallocation + T a r g e t i n g R e f o r m ) 10,000 8,000 6,000 4,000 2,000 0 1 2 3 4 5 Deeiles based on Expenditures N e t o f Transfers per capita Note:The targeting reformscenario assumes designchangesthat would close the coverage gap fully indecile 1,50% in decile 2, and 25% indecile 3. Source:WB staff calculations. Key issuesof policy designfor poverty reduction Of equal importance to potential shifts intax and spendingcomposition are issues of programandpolicydesign. The discussionis dividedinto three areas: the provisionof social services; policies influencingincome growth; and a set of crosscutting institutional issues. (1)Thedesignof public serviceprovision: socialservices and s o ~protection ~ l Human c~pitallCapacidades.Mexico, like many countries in the region and elsewhere, will have to run fast to keep up with the dynamics of returns to skills. Returns to primary and secondary education have been flat and are particularly low for rural areas. Returns to tertiary education are high, reflecting large rises in this premium in the early 1990s. While there is evidence of declines in the premium to tertiary education since 1997, tertiary enrollments remain highly concentrated in the children of richer households - who both appropriate the high subsidies and enjoy the highprivate returns. Inthe 1990's expansion of tertiary enrollment was actually disequalizing, with an increase in the gap between net enrollments from high and low income households. A continued push on both access and quality for secondary, and mechanisms for giving bright poor children access to tertiary education will be key to future progress. There are initiatives to provide funding or access for poorer children to tertiary education (including support to inter- cultural establishments, with a particular orientation to indigenous groups) but a broad- based student loan scheme will be challengingto implement inlight of weaknesses inthe tax collection system, which could make loan recovery difficult. A central issue, which is well documented for education services (and also appears to apply to health and other services) is quality; it is low on average and displays significant xxxvi variation, with the poor, and especially indigenous groups (at least those in indigenous schools), experiencing the lowest quality in schooling outcomes as measured by test scores. While socio-economic status is a major determinant of individual test scores, there i s an agenda of change on the supply side, where teaching reform to increase the accountability of teachers to outcomes will be central if the skills of the poor are to be upgraded. There are probably also opportunities for strengthening the resource base of schooling indeprived areas, with indigenous schools beingone example. The design of any additional support for education would, however, needto take account of the risk that expanded spending could be captured by increasedrents of public sector workers rather than better services for the poor. For both quantity, and especially, quality the primary focus should be on efficiency-improving reforms, International experience suggests that this is likely to involve complex institutional issues within the sector, between national and local governments, and between schools and the communities they serve. There are now an array of programs, including Carrera Magisteriul (Magisteria~ Career), Incentives for Higher Education Teachers, Escuelas de Calidad (Quality Schools), Compensatory Programs and Apoyo a Gestidn Escoluv (Supportfor School Exertion). Some of those which have been evaluated show specific gains, but also potential for design improvements. Others have not been evaluated. Consolidation of programs and systematic evaluation i s probably desirable; it is likely that deeper reforms to increase the accountability of teachers to outcomes will be central if the skills of the poor are to be upgraded.17 As an incentive program for human capital investment, O~ortunidadeshas had a significant impact on nutritional status and secondary school attendance amongst the poor, as shown by its impact evaluations. With the recent expansion into poor urban areas, it i s a program of national significance, which already reaches a high fraction of the rural extreme poor, as noted above. Since 2001, Oportunidudes extended education benefits to individuals 14 to 20 years old inpoor householdsenrolled inupper secondary education. After only one year of implementation, the program provided more than 400,000 scholarships for upper secondary, yielding a net impact of 38% increase in enrollment rates in rural areas and 6% in urban areas (Rubalcava and Teruel, 2003). A further extension is Jdvenes con Oportunidudes, which consists of a savings fund for lower secondary graduates wishing to continue their upper secondary education and finishing it before reaching 22 years old. Students can cash out their accumulated points and use them as collateral for micro credit, as down payment to buy a house or for home improvement, to buy health insurance, or finance higher education. This design i s promising, but the component has not yet been evaluated. Areas that could be strengthenedin Oportunidades,especially as a potentially integrating part of the CONTIGO strategy, include improving the connectivity with local service delivery and assessingits accountability mechanisms and proceduresfor graduation from the program. There is also a case for clarifying the balance between the program's dual objectives of increasing attendance at schools and clinics and transferring income to the l7Education quality is the topic of an ongoing work program inthe World Bank, and any statements here are necessarilypreliminary. xxxvii extreme poor. The extensive transfers to primary schooling do not have a large impact on attendance, since most children already attended school, but they can be justified as part of a comprehensive redistribution and social protection system, for which the objective of transferring resourcesto the extreme poor is key; this is taken up next. Managing vulnerabili~!/ProteccidnSocial. Social protection has two objectives: supporting efficient risk management of households, and redistributive transfers to the poor. These are sometimes mixed in the same program -and indeed this can bring political economy benefits, if the risk management component increases middle class support for programs that also redistribute to the poor. A sound strategy for risk management involves a balance between strategies to reduce risks, self-insurance and pooling of risks. The poor typically have relatively weak capacity for self-insurance, while risk pooling i s especially limited to extended families or other informal mechanisms, which tend to be costly, and to break down inthe context of covariate risks. Government action can becrucial inthese areas. Within such a framework, an important part of an overall risk management strategy for the poor and vulnerable involves buildingrural and urban asset bases and strengthening of financial systems (including micro-finance). It is thus closely linked to an asset- building strategy discussed below. There is a two-way relationship between assets and vulnerability: assets can be a means of managing vulnerability; conversely high levels of vulnerability can lead to reduced levels of investments, because of the risks these entail when credit andinsurance markets are incomplete. However, an assets-based strategy needs to be complemented by government social protection programs to deal with the risks that cannot be efficiently managed by individuals and groups themselves. A sketch of the current system is given in Table 2. The core of current structure of social protection i s oriented to the formal sector, including pensions, health and unemployment benefits. Beneficiaries of social security and health services from the two largest social security institutes, IMSS and ISSSTE, are primarily among richer and some middle groups in the population (Figure 10); the incidence of benefits of these institutes is roughly as unequal as the overall income distribution." This reflects the "truncated" welfare state that is so typical of Latin America (De Ferranti et al, 2004). Diagnosis of the formal system lies largely outside the scope of this study, though there are clearly areas for reform. With respect to the uncovered population, that is the primary concern of a poverty report, there i s a clear needto expand social protection to the poorer and informalhouseholds. l8 There are several other smaller federal and state social security institutes discussed inChapter 3. xxxviii Table 1. Mexico's social protection system by poverty status (an illustration) Extreme Poor ModeratePoor Non-poor OldAge No significant coverage IMSS (19% with rights) IMSS (27% with rights) ISSSTE(3% with rights) ISSSTE (12.4% with rights) Other social security institutes (federal and state) Major Health PAC (SSA) IMSS (34% with rights) IMSS (47% with rights) Risks SeguroPopular (SSA) ISSSTE (5% with rights) ISSSTE(15% with rights) IMSS-Oportunidades Private providers Privateproviders Segurode Saludpara Other social security lafamilia (IMSS) institutes(federal and state) Private providers Unemployment Very partial coverage Partial coveragewith formal Partialcoveragefrom Risks of PET severance pay and SICAT. formal severancepay Natural Disasters FONDEN FONDEN Largely inapplicable and Harvest Oportunidades Failure SocialAssistance Oport~nidades(42%) Partial coverageof Not applicable and Children DIF Oportunidades(16%) DIF Notes: Percentagesrefer to population covered ineach group. Fullprogram titles are: PAC-Program de Ampliacio'n de Cobertura; FONDEN-Fondo Nacional de Desastres Naturales; DIF-Sistema de DesarrolloIntegral de la Familia. Source: WB staff calculations fromENIGH2002. The main poverty-targeted transfer programs, especially ~ p o r ~ ~ n i d aandeProcampo, d ~ have good to very good targeting properties as redistributive social assistance (Figure lo), and also he1 reduce risk for recipients, since transfer amounts are not generally subject to shock^?^ However, these have incomplete coverage of vulnerable groups, and there is scope for broadening in two respects: (i) conditional transfers for groups in extreme poverty not covered by ~ p o r ~ ~ that~isdfocused~on children; this might n ~ e include adults incommunity service, and agedextreme poor; (ii) strengthening risk-based programs, such as public works, that can kick in automatically in response to covariate shocks (PET, the decentralized public works program has relatively small coverage and uncertain impact). This should be complemented by strengthening of information and accountability systems. With respectto health risks, Mexico has multiple providers: social security rights-holders have access to IMSSlISSSTE based curative care; there is a network of public curative hospitals and clinics provided by the health services of the states (SESA) and regulated by the Ministry of Health (SSA), as well as subsidized parts of IMSS; and an extensive private system serving all parts of the population. A strikingfeature of the health system i s the extent of use of private providers for major health risks, from the uninsured poor to l9 However, Procampo did decline significantly inthe wake of the 1994-95 crisis, underscoringthe cost of such macro-fiscal crises. xxxix the well off with social security rights.This is indicative of access or quality problems, at least for the poor. The Government has recently taken measures to extend health insurance for the uncovered population. The flagship program i s the SeguroPopular that was introduced in 2002, offering a package of specific services and medicines to those not covered by social security, with highly subsidized premia linked to ability to pay -with a zero premium for the bottom quintile of the population. This was complemented by a 2003 decree designed to progressively extend health insurance to all by 2010. The primary objective of Seguro Popular is to increase public resources going to curative care for poorer groups. This i s highly desirable in light of the generally inequitable pattern of overall spending in health. In 2000, the World Health Organization (WHO) ranked Mexico's health system 51 out of 191interms of overall performance, but 144out of 191 in terms of financial fairness (assessed in terms of distribution of potential health care costs interms of ability to pay rather than need). Moreover, Mexico spends less on health than international comparators. Seguro Popular is complementary to Oportunidades since it provides resources conditional on health risks. Like Oportunidades, it works on the demand side -albeit via finance of the in-kind provision of health services rather than a cash transfer. Seguro Popular also provides a potential instrument for changing incentives within the health system. The major challenges here are providing a means for Secretaria de Salud to provide genuine stewardship of the health system, through its "compact" relationship with the states, and for the state secretarias to fulfill a comparable stewardship role over public sector providers within their states. By end-2003, 625,000 families had beenaffiliatedto SeguroPopular in24 of nation's 32 states. This i s a notable achievement. However, since SeguroPopular has only just been introduced, there is not yet information on its impact. Indeed, its actual impacts will depend greatly on the behavior of households, on service providers, and on interactions between the federal secretaria and the states. While it has the potential to deliver major benefits, the extent of these remains uncertain. For example, there is uncertainty over the extent to which Seguro Popular will lead to households actually increasinguse of public health services, shifting from private to public providers (saving out of pocket expenses) or choosing not to enroll in formal social security. There are also unknowns on the behavior of health service providers. This uncertainty is comparable to the situation when Oportunidadeswas introduced (then as Progresa) inthe late 1990's. Getting information on these issues will depend on structured monitoring and evaluation both on household behavior and service provider performance. While current plans do indeed put considerable emphasis on monitoring and evaluation, there i s a need for particular attention to the design of impact evaluations, especially including mechanism to track the behavior of households and service providers over time (as was undertaken as an intrinsic partof Progresa's design with, inparticular, the useof longitudinalhousehold surveys). Seguro Popular potentially covers health risks for all uncovered families -with higher subsidies for the extreme poor. For other risk categories, a strategic issue for the future concerns middle groups inthe population-especially amongst the moderatepoor- who will never be reachedby ~ p o r t ~ n ~ ~ and s a d erelated programs that are primarily focused xl on the extreme poor, and will only participate in the formal social security system in the longer term. This constitutes a policy challenge for social protection that will be the focus of the next phase of work -it parallels the challenge of providing income opportunities for this group that i s the focus of the government's Zmpulso strategy, that seeks to improve the productivity and formalization of the informal sector and raise the potential productivity of workers (see below). Finally, a different type of vulnerability concerns risks of violence. Criminal violence is of particular importance in some urban neighborhoods, affecting both personal well- being and economic security. Rural violence continues to be an issue in some of the poorer areas of the country, where it is often associated with land conflicts, political or social divisions. Effective access to justice systems in Mexico, as elsewhere in Latin America, is biased against poorer groups. OECD (2003) found that there are serious and persistent problems in the judicial system in Mexico, as illustrated by long judicial processes, poor enforcement of judicial decisions especially at the local level, and high corruption. Actions to address these issues are required both at the federal and the state levels. While violence was not a focus of analysis in this report, it is flagged here as an issue of major concern for the poor. (2)Promoting income growth amongst thepoor As seen in Section A, there has been much weaker longer-run progress with respect to income poverty than for social dimensions of well-being. Future income gains amongst the poor will dependon both overall growth andthe pattern of growth, and especially the pattern of job creation. Scenarios undertaken for this report show that both growth and inequality matter for poverty dynamics. Slow growth, of 2% per annum, between 2002 and 2015, would have little impact on poverty if inequality does not decline, or reverts to 1998-2000 levels. But if inequality fell by lo%, extreme poverty would fall to 8% by 2015. Growth of 5% per annum would bring extreme poverty rates to 6% by 2015 if inequalit reverts to 1998-2000levels, butbringthemdown to 2% if inequality fell by 10 percent,2; The core of any job-creating strategy lies in the environment for private investment. In other words, a poverty reduction strategy i s inseparable from the design of a competitiveness and growth strategy. This relates to the current overall competitiveness agenda, which is currently under public scrutiny in light of Mexico's poor international performance relative to competitors such as China, especially after the initial gains from NAFTA were realized. This is a complex agenda, in which international evidence suggests that macro stability, infrastructure provision, tackling of protections of the private sector, and a range of measuresto improve logistics are all important. Economy-wide measures to strengthen competitiveness will need to be complemented by strategies specifically oriented toward income growth of the poor -both the extreme poor (mostly in rural areas) and the moderate poor (mostly in urban areas). Current government strategy for the extreme poor i s much more developed in the social arena 2o See scenarios in Chapter5. xli than the productive arena; this is a priority for the future. With respect to the moderate poor, the government has recently formulated a new strategy, termed Impulso, whose central thrust involves a variety of measures to increase the productivity, security and dynamism of the self-employed and small- enterprise sector, that is now largely informal. This is based on sound diagnosis, but there is an important agenda in policy design and execution. We sketch issues in asset-formation and the policy and regulatory framework for income growth inbothrural andurbanareas. In~rastructure and Assets/~atrim~~io. Throughout Latin America, infrastructure investment has fallen sharply in the 1990's -with rises in private investment not offsetting the fall in public investment, except in telecommunications. Recent work on cross-country growth finds that higher infrastructure stocks are associated with both higher growth and lower inequality (Calderrin and Chong, 2003, Calderrin and Serv6n 2003). In Mexico, during the ~O'S,investment levels have been at best flat and, after a short-lived boom in telecommunications in the late 1990's, private investment recently displayed a significant decline. Public investment has recovered since 2001 particularly in sectors such as electricity, roads and water. However, levels of investment remain low by international standards, and more effort will be needed to reactivate investment in infrastructure inboththe public and private sectors. Figure 14, Investment in New Infrastructure in Mexico 1990-2003 (excluding PIEMEX) 3.5% 3.0% 2.5% g P 2.0% 2e,1.5% m 1.0% 0.5% 0.0% 1990 1991 1992 1993 1994 1995 1994 1997 1998 1999 2000 2001 2002 2003 Note: Investment in infrastructure includes roads, electricity, railways, airports, ports, water, and telecommunications. Data for 2003 are estimates. Source: Anexo del Tercer Informede Gobierno 2003. While more spending on infrastructure is undoubtedly necessary, issues of policy and institutional changeare fundamental for bothefficiency and equity, whether these involve privatization, the regulatory framework for private production, or the possibly more challenging goal of reformingpublic provision. However, privatization in Mexico has led xlii to a mixed performance. In some sectors privatization has led to significant positive results -notably in ports, railroads and telecommunications. But in many areas privatization also created considerable market power for the incumbent, reducing the pressure for quality, low-price service, This may include telecommunications, airlines, toll roads, and commercial banks. Important parts of mexican infrastructure are either of low quality (electricity) or high cost (telecommunications, transport). Inaddition to these general concerns, specific issues for the poor include completing access to water and sanitation, improving service quality, and restructuring of electricity subsidies. Electricity subsidies are regressive and large, despite recent reductions. Those who consume more receive higher subsidies, except for the very top of the income distribution. The benefits of a hypothetical redistribution from electricity subsidies to ~ p o r t u n i d ~ ewere illustrated above. There is scope for price rationalization within the s residential subsidy, which could maintain lifeline subsidies for the poor, and release resources for other investments. A reduction in electricity subsidies could have large impacts on the poor, if combined with compensatory mechanisms both within the price schedule and from other programs. However, pressure groups are successfully blocking implementation of the 2002 price increase. Moreover, inApril 2003, the Senate approved an increase in subsidies for residential electricity prices in some regions inthe north and southeast ("with difficult climatic conditions"). This decision is clearly a step backward, and it can only further reduce the ability of the Comisidn Federal de Electricidad to invest in new capacity. There is also potential for savings in the electricity subsidy for irrigation, that mainly goes to large farmers. For the rural poor (and especially the extreme poor), there is an ongoing agenda of strengthening their productive asset base. Part of this involves the continuation of the complex process of land management and rights, notably PROCEDE (Programa Nacional de Certijkacidn de DerechosEjidales y Titulacidn de Solares Urbanos).This is a demand-driven program of landregularization which was createdin 1993. The program allows groups of producers to choose their property rights regime, delineates ejido boundaries, measures individual plots, and eventually issues certificates to individually owned plots (including house plots) as well as communally managed lands to each individual. PROCEDEhas led to positive results such as increasedlandaccess, improved governance and transparency at the grassroots level, increase in household welfare, and improved functioning of landmarkets. However, challenges remain, includingimproving investment and management skills, building capacity for conflict resolution, especially through alternative mechanisms and out-of court settlements, and seeking alternatives for reducing the costs of the program t (World Bank 2001b). Equally important is the provision of economic ~ n ~ r a ~ t r u cand e t ~ raccess tofinance. It is important that this be designed in ways that supports diversified production patterns. There appears to be potential for buildingon existing local financial intermediation, with a particular area of opportunity being the inclusion of the substantial flows of remittances that are being received by some 13% of rural households, and a fifth of the poorest. It should be mentioned that other work indicates that land reform remains an important issue for the poor in many parts of Mexico (World Bank 2001b, Finan, Sadoulet and De xliii Janvry 2002). In some areas the challenge is in facilitating the individualization of tenurial systems under the 1992 land reform. In others it is a case of supporting indigenous groups who prefer to keep communal systems. Issues of land conflicts are important in many areas. However, resolving land issues will only yield benefits if complemented by measures to strengthen the competitiveness of land beneficiaries with close links between titling, technical assistance to manage land independently, and structural measuresto raise productivity. For the urban poor -who account for the bulk of the moderate poor- the assets-based agenda i s centered on regularization of settlementlproperty rights, strengthening the quality of urban infrastructure, and broadening access to financial savings. Hdbitat is designed to face the challenges of the urban poverty, through a model of action that combines, among other aspects, the improvement of infrastructure and equipment in the urban-marginalized areas, and the implementation of social services and community development actions. With its different modalities, Hdbitat tries to support the population ina stateof assets poverty who live incities and metropolitan areas whose size, location, social and economic significance, converts them into strategic points, with particular care for the needs of women, people with different capabilities, children, teenagers, young adults, and the elderly. While of promising design, there are not yet impact evaluation results of this program. Policies and regulations. To complement an assets-based strategy, there is scope for strengthening the policy and regulatory environment for income growth of the poor. For the moderate poor, the Impulso strategy correctly focuses on the dynamics of growth in the informal sector, including self-employed and small-scale producers. A central issue is the costs of formalization. The costs, due to inflexibilitiesinthe labor market, taxes, costs of registering businesses, and other bureaucratic burdens, outweigh benefits for most self- employed and micro and smallfirms. One element of this involves the functioning of the labor market. The existing legislation probably has only minor effects on the overall pace ofjob creation, but it does constitute a constraint on the poor getting good jobs, since it increases the incentives for informal work (on the part of firms and workers). The debate on the modernization of labor legislation that started in 1997 (with the signing of "the agreement for a New Labor Culture") resulted in 2002 in a formal law proposal that is still under discussion in Congress. The reform proposal could represent a significant step forward in makingthe formal market more attractive, for example through the introduction of probation periods, initial steps in easing regulations concerning permanent contracts, efforts to modernize industrial relations, some simplification of bureaucratic requirements for small and medium size enterprises, measures to promote labor force training and measures to reduce corruption and legal uncertainty. However, the proposed law does not yet address the main issues affecting the labor market, especially those which are sources of inflexibility inhiringand firing. Beyondchanges to labor market legislation, other initiatives could be taken inthe short to medium-term to improve the balance between the formal and informal sectors. They xliv include enhancing the quality of services provided to formal workers and reducingthe tax wedge, inparticular employers social security contributions for low-skilledworkers. Within rural areas, there is a wide range of programs, whose coordination is unclear. An effort i s being made under the Programa Especial Concurrente mandated by the Ley de Desarrollo Rural Sustentable to coordinate different programs but only modest progress has been made. While Procampo continues to be progressive, most existing programs under Alianza para el Campo are probably regressive, and agricultural productivity growth of the recent past was weakest inthe poorest regions.21Since 2001, the following improvements were introduced in Procampo: the transfers are paid earlier in the crop cycle to facilitate the purchase of inputs and to encourage investments among producers by providing a more secure expected income; producers are allowed to use part of their payment as a collateral for loans; and the eligibility criteria was simplified and awareness of these criteria was promoted. Also since 2001, improvements were introduced in Alianza para el Campo, including: dissemination in the ejido sector so that a larger proportion of ejidatarios are aware of the program, understand its objectives, and are clear on how to access the resources; elimination of requirements for group participation (which may be difficult for the poorest) insome subprograms; and allowingejidatarios to purchase their own inputs directly from local distributors (rather than government- certifieddistributors). A broader consideration concerns the scope for a more structured and coordinated territorial strategy, that could buildon existing programs such as ~icrorregionesand the Ley de Desarrollo Rural Sustentable. These programs need to be institutionally linked and integrated into productive territorial strategies for agricultural and non-agricultural growth, and strategies for technological change. In shaping such strategies it will be important to focus on the role of rural non-farm sources of income. As a share of rural income this has risen sharply in the past decade, from 50% in 1992 to 70% in 2002?2 Even for the bottom quintile, that is more dependent on agriculture, this proportion rose from 30 to over 50%. In addition, the non-poor will typically be the key agents of economic dynamism, and it is of great importance to develop approaches to production strategies that support alliances and economic connectivity between the poor and the non- poor. Finally, there is a particular agenda around income growth of rural indigenous groups. This is both because these groups suffer high levels of poverty, and also because of distinct preferences and practices -for example in attitudes to the natural world and in collective modes of organization. Ongoing projects, for example in communal forestry and coffee cooperatives, show that it is possible to support dynamic productive products '' Thus, between 1990 and 2002 productivity in the production of grains in the north Pacific region (Baja California, Baja California Sur, Sonoraand Sinaloa) grew at 1.47% per annum, while inthe southernstates (Guerrero, OaxacaandChiapas) it grew at 0.63% (Work-in-Progress, Mexico RuralPoverty Study). 22 This i s the share of total household income from non-agricultural sources, from the ENIGH, in settlementsof less than 2,500 people. Non-agricultural income includes rural non-farm labor and enterprise income, remittances and transfers and rents. In 2002, the proportion of total income from rural non-farm labor andenterprisesincome was over 40% for all rural households. xlv through working with the organizations of indigenous groups. Where an external agent has directed or assisted the initial process, there are common characteristics, including: initial seed capital (not loans) by a grass-roots development agency; a large dose of technical assistance, but provided with a view of an exit strategy; andbuildingupon local models of intervention and a local socio-cultural vision of the goals and objectives of indigenous groups. (3) Crosscutting institutional issues: accounta~i~ityand decentral~zation?social inclusion, and evaluation Cutting across specific programs and social and economic strategies are broader institutional questions that will play a powerful role in the design and, especially, the implementation of poverty-reducing strategies. Cl~entel~s~~a~ronageand service delivery. As with much of Latin America, andindeed much of the developing world (World Bank, 2003), Mexico has beenstrugglingto reduce the importance of clientelistic and patronage-based relations with poor groups. There have been some notable successes, such as the shift from strongly patronage-driven programs associated with PRONASOL to the much less captured approach of ~ ~ o r t ~ ~However, ethe~legacy of history on institutional functioning is deep, and i d ~ . especially so in the complex area of service delivery. Both the use of services as sources of patronage (for jobs and service access), and the particular formation of interest groups amongst service providers, can make reforms for quality and equitable services a challenging and complex undertaking. Mexico i s in the midst of a major political, social andinstitutionaltransition, broadly associatedwith the deepening of democracy, but it is important not to underestimatethe difficulties andcomplexities inthis transition. Transparency and u c c o ~ n t u ~ i lGreater accountability is in principle a central part of i ~ . the answer to problems of clientelism and weaknesses in service delivery. An emphasis on transparency, participation and the buildingof the capacities to engage poorer groups are ways to both mitigate these risks, and strengthen use of local knowledge and capacities indesign and implementation. There has beenimportant progress inthe overall objective and recent laws, but this needs to be systematically introduced into specific local and sectoral settings -from municipalities to schools. However, there are even less blueprintsinthis area than inothers. Experience elsewhere suggeststhere are significant complementarities between measures that seek to improve incentives between policymakers and implementing agencies (including lower levels of government) and local mechanisms for accountability that seek to strengthenthe "voice" of poor and non-poor clients. There are also complementarities between processes of social movilization and reforms that increase local accountability, often working to reduce the capture by local elites -in what Fox (1992) has termed a "sandwich" strategy, that has occurred in some successful cases in Mexico (notably in parts of the DICONSA programs and some municipalities). There is again a strong case for experimentation linked to systematic monitoring and rigorous evaluation to give concrete informationon what does and does not work indifferent contexts. xlvi Managing ~ecen~ruz~zulion. Decentralization is a political fact of life, and is, inprinciple, desirable. Already a high proportion of poverty-related spending takes place at sub- national levels -with the new Law on Social Development- emphasizing municipalities as the core level for government action in the social arena, except in certain federal programs. In Mexico, there have been significant shifts of spending and program management to states and municipalities (Box 2). The implications of this for institutional interactions, program management, and distributional impacts are complex. Little is known about the impact of decentralized expenditure on poverty outcomes. It is unclear how the CONTIGO programmatic orientation can have a purchase under decentralized auspices. It is noteworthy that the flagship success -0porlunidades- i s a federal program. This report concurs with a World Bank study (2003) that concludes that relatively weak social outcomes in the poorest states of Mexico are the product of multiple influences, including lack of funding, inefficient use of funds, and difficult initial conditions (from high levels of income poverty, to highly dispersed populations andrelatively underdeveloped infrastructure networks). Box2 Dimensionsof decentralization A series of institutional reforms took place during the Zedillo administration to decentralize public spending. These included the creation in 1998of a new modality of federal transfers implemented through a budgetarybranch:Ramo 33 (R33): Fondo de Aportaciones Federalespara 10s Estadosy Municipios.R33 comprises seven funds: Fondo de Aportaciones para la Educacidn Brisica y Normal (FAEB); Fondo de Aportaciones para 10s Serviciosde Salud (FASSA); Fondo de Aportacionespara la Infraestructura Social (FAIS): Fondo para la Infraestructura Social Estatal (FISE) and Fondo para la Infraestructura Social ~unicipal(FISM); Fondo de Aportaciones para el Fortalecimiento Municipal (FORTAMUN); Fondo de Aportaciones ~ ~ l t i p l e(FAM); Fondo de Aportaciones para la Educacidn Tecnoligica y de Adultos s (FAETA); and Fondo de Aportaciones para la Seguridad Pliblica de 10s Estados y el Distrito Federal (FASP). Today, close to 65% of total educational spending and 75% of public health spending for the uninsured population are decentralized. The bulk of these resourcesare spent by state governments, with only 13% of R33 is spent directly by municipal governments. Table 2 Federal Transfers FUNDS OBJECTIVES BUDGET ALLOCATION CRITERIA (% R33) CRITERIA FAEB EstablishedInfrastructureandPersonnel, (62 4%) Basiceducation Previous Budget xlvii FUNDS OBJECTIVES BUDGET ALLOCATION CRITERIA (% R33) CRITERIA EstablishedInfrastructure andPersonnel, PreviousBudget FASSA23 (12.2 %) Healthfor uninsured population Formula: minimal per capita health Equity (1%) spending, non-coveredpopulation, mortality, poverty. Basic infrastructure: clean water, sewerage, 2.5 % Federal Equal state shares (transitional) drain, municipal urbanization, electricity for Revenue FAIS ruraland poor urbanareas, basic healthand (9.7 %) education infrastructure, housing, rural FISE (0.303 %) Formula: illiteracy, education,drain, roads, rural productiveinfrastructure. FISM(2.197 %) electricity, housing, income. Institutional Development Max 2 % FISM FAM Social Assistance 0.814 % Federal (3.1 %) EducationInfrastructure Revenue FAETA Technicaleducation EstablishedInfrastructure andPersonnel, 1 (1.7 %) Previous Budget Formula: Illiteracy, basic education, work trainin . FORTAMUN InstitutionalCapacity Buildingof the (9.9 %) Municipalities particularly inthe area of Population's size infrastructure FASP (1.4 %) Public Security Support Trial andconvicted population Source: World Bank Staff from the StatisticalAnnex of the President's State of the Union, 2003. Decentralization brings a classic tradeoff -with lower levels of government having advantages ingreater informationon local conditions and needs, but higher risks of local elite capture and weaker institutions. Community and municipal level participation has enormous potential. But both decentralization and "participation" carry risks of strengthening of local elite capture and patronage-based approaches, whether at state, municipal or community levels. This constitutes a major agenda for future experimentation, and diagnosis, in light of the central importance of subnational governments inservice provision. Excluded groups and the hard-to-reach poor. While there have been many initiatives over the years, indigenous groups (identified by language in the home, or self- identification) are on average much poorer than others in most dimensions of well-being, as noted above. This is true both unconditionally and conditionally with respect to characteristics such as education. As noted above, they account for some 10% of the overall population but 21% of the extreme poor in terms of income. There is a major outstanding agenda around indigenous groups, in terms of recognition, evaluating and strengthening experiments in bilingual education and conflict management. More work needs to be done to assess the importance of other social excluded groups, but a prior would be that socio-cultural processes are important for others -such as poor urban 23 After the 2003 reform to the health sector, the FASSA resources were split into two funds, FASSC and FASSP, which respond to the division of health goods into public goods, also known as community services, and personal services respectively. FASSP resources will be allocated to the states by a formula based on number of families to be covered, health needs, as well as effort and performance of the states. The reform, however, contemplates that the states should not receive less total resources (FASSP and FASSC together) as compared to the total amount of resources they received throughFASSA in2003. xlviii youth and slum-dwellers- as well. There has been a wide range of programs aimed at indigenous groups, and a careful stocktaking would be useful to assess their effectiveness. There can be other reasons why some categories of the poor will be difficult to include in development processesandwill require special efforts. Two issues are highlighted. Remote areas. There are a large number of very small rural settlements in Mexico, including many indigenous communities, which are typically very income poor and under-served by services. They raise special challengesbecause of the higher cost of service provision, and the fact that they can fall outside the net of those transfer programs that are linked to serviceprovision or attendanceat schools or health centers. This will require innovative action: for example strengthening the programof board-schooling for poor indigenous children. Shaping programs to the needs of very poor households. There will also be particular individuals, households, and groups within rural and urban communities, who have a weak capacity to respond to general initiatives -often because of histories of extreme vulnerability and exclusion. For these groups there i s also scope for experimentation with approaches that seek to have public agents or social workers mold responses to the multiple needs of particular groups andhouseholds, on the modelof the Puenteprogram inChile. Streamlining or redesigning of social programs. A different category of institutional issue concerns the large number of programs that directly or indirectly seek to reduce poverty. This i s a reflection of the long history of introducing programs -there has, if anything, been a decline in the number of activities under the present government, at least in SEDESOL. However, this report did not form a position on whether the numbersof programs shouldbe cut or not. During the present administration, as an example of the actions of one ministry, SEDESOL has made a significant effort to adjust social programs to better serve the needs of the poor. These programs fall under four categories. First, SEDESOL has created new programs such as Hdbitat, which contributes to overcoming urban poverty and improving housing conditions. Second, it has reinforced and extended existing programs: as noted above ~ p o r t u n i d ~ ewas significantly expanded and new s components introduced, notably Jo'venes con Oportunidades, which provides additional incentives for youth to complete high school. Third, some programs were transformed: Liconsa now distributes milk enhanced with iron and zinc, because the 1999 National Nutritional survey found a lack of these nutrients in urban children. Fourth, SEDESOL has eliminated a few programs that were not found to be effective, like the Tortilla subsidy. Other ministries such as Labor and Education have also re-structured their programs (Cravioto, 2004). The Ministry of Finance and a consulting firm, GEA, did a review of the re-structuring of social programs, xlix Perhaps of greatest importance is continued stocktaking concerning what does and does not work, building on successes, and pursuing opportunities for integration and complementary action. Further streamlining and coordination can possibly occur at the level of central ministries. Greater payoffs will probably come from designing experiments and greater coordination of programs at the local level. This takes us to a final issue. ~onitoringand evaluation. Last, but not least, monitoring and evaluation have a major role to play at both the technical and institutional level. There are some impressive aspects of evaluation in Mexico and SEDESOL in particular is a pioneer in the effort to deepen and extend this. Recent initiatives are in line with best practice, including Mexico's congressional mandate calling for annual evaluations, and its extension in the Ley de Desarrollo Social; the establishment and monitoring of presidential goals; SEDESOL`s registry of programs' beneficiaries; and the citizens' manual ~ ~ a n ~ a l Ciudadano), which provides information on SEDESOL's functions in addition to providinga mediumfor citizens' complaints and suggestions. There have been important efforts to improve the Padrdn de Beneficiarios (beneficiaries' registries) of the social programs. This is with the aim of further developing the areas of targeting and efficiency, and avoiding duplication of recipients. In this administration, SEDESOL has made significant progress in having the complete beneficiaries' registries for all its programs. It has also signed agreements with some state governments to complete the lists. A further effort will be madeto eventually include other ministries and programs. However, effective monitoring and evaluation is confined to only a few areas and Ministries, with the world-renowned evaluation of rural ~ ~ o r ~ n i d abeing d e s exceptional inthe Mexican context. Congress' mandaterequiringannual impact evaluation of all the one hundred and four programs providing public subsidy i s highly commendable inspirit, but brings a serious bias toward undertaking quick and superficial evaluations that bring few benefits. Moreover, the ~ ~ o r t u n i d ~model of impact evaluation, based on e s comparing the effects of treatment and control cases with randomized allocation of treatment, has limitations. It i s state of the art for a bottom-line assessment of whether a treatment is beneficial. I t is weaker for ongoing learning about the processes behind observed effects, and for understanding the role of local context. Both are key for the improvement of the design of interventions. What i s desirable is to combine the treatment and control approach and external evaluation approach, with systems of monitoring of progress across interventions structured both to maximize ongoing learning and improvement indesign, and to provide information in a user-friendly way to the public to increase social pressures for accountability (see Box 3 below on international experiences). SEDESOL is currently exploring a broadening of its evaluation agenda. To be effective in affecting government behavior, this needs to be integrated within an overall program of results based management. This is under development in both SEDESOLand government-wide though the Social Development Law. Box 3 Selectedexperience with Results Based ManagementinLatinAmerica 1 Chile: The Government of Chile initiated the design and implementation of a Results BasedManagement (RBM) systemin 1994 (Sistemade Controlde Gesticiny Presupuestospor Resultados)in order to improve public management. The Ministry of Finance launchedthe initiative, developing a number of instruments to foster a more transparent budgeting process, strengthen its analytical capacity, and improve program management. Although the chilean approach is decentralized (i.e., no single entity is responsible for developing and maintaining a national system as in Colombia), the processes are relatively well coordinated by a cabinet level inter-ministerial committee. The budget office in the Ministry of Finance plays a key role in coordinating the various decentralized initiatives, and uses performance data in formulating its budget on an annual basis. Performancei s evaluatedat the level of policies, expenditures, organizations, programs and individual public servants. The government has begun to link performance evaluationsto organizational incentives. Colombia: Colombia began its performanceevaluation system, SINERGIA, in the mid-1990's as part of the national development planning process. The exercise is led by the technically competent National Planning Department (DNP), which carries out the performance evaluations of ministries and programs. DNP presents annual reports on the government's strategic objectives and performance indicators to the Economic and Social Policy Council (CONPES) responsiblefor the oversight of socialpolicy inColombia. While the system i s well conceived from a technical point of view, it still lacks full coverage -as it is centered on evaluating public investmentprograms inthe National DevelopmentPlans and does not cover recurrent Government activities- and the performance information generated from the system is not systematically taken into account ingovernmentdecision-making. Bolivia: The 1990SAFCO (SistemaIntegradodeAdministracicinFinancieray Control Gubernamentales) Law required government entities to prepare annual operational plans with performance indicators. Most recently, the government attempted to develop a new performance evaluation system (Sistema de Seguimiento y Evaluacicin de la Gestibn Pu'blicapor Resultados, SISER) located in the Ministry of the Presidency and supported by the multi-donor Institutional Reform Project. In specific sectors, especially education and health, donor support for sectoralreforms has also resultedin better specification of sectoral goals and performance indicators. Overall, however, the system is far from fully operational. POAs (Annual Operational Plans) have always been treated in perfunctory fashions by both sectoral ministries and the Ministry of Finance, and the SISER project has beendiscontinued. A reform of the budget system that was supposed to accompany the development of the SISER never took off, in spite of the notable recentprogresswith a new financial managementinformationsystem. Brazil: Since the mid 1990s, the Ministry of Planning, Budget and Management has led an ambitious effort to turn the federal government's multi-year plan (PPA) into a strategicinstrument for results-oriented resource allocations and program management. Most of the PPA programs are assessed each year, and the report is submitted to Congress prior to the beginning of the annual budget cycle. Although the new PPA framework has enhanced government transparency and contributed to better strategic discussions about program goals and designs within the bureaucracy, there is little evidence that the annual assessment reports are seriously reviewed, let alone debated, by the Congress, and that evaluation results are systematically influencing resource allocations decisions. Highdegrees of rigidities in the budget structure and public management regulations limit the government's ability to manage for results. In spite of the multi-year character of these plans, they are still steps away from a robust medium-term expenditure framework as found inadvancedOECD countries. Source:WorldBank(1997), OspinaBozzi andOchoa (2003), Montes (2003), andWorld Bank(2002b). C. Concluding comments: on action andfurther work This overview of Mexico's situation with respect to both the poverty conditions and strategy has revealed a mixed picture. There has been considerable progress in the outcomes for the poor in both the long term and the recent past. But long-term progress has been much greater for health and education than for the incomes of the poor. And the li recent, and noteworthy, short-run gains in incomes of the poor, has only just reduced poverty below levels prevailing in the early 1990's. Moreover, becauseof Mexico's high inequality, poverty levels are highfor an upper middle-income country. Other dimensions of well-being -especially the insecurities associatedwith the range of risks faced by the poor and near-poor -are also of great concern. With respect to strategy, there has also been commendable progress, especially in provision of services, both over the long term, and in the actions of the present government. This applies both to the expansion of basic services, and to the extension of programs specifically targeted to the poor, with Oportunid~esparticularly impressive in both its targeting and its reach. But there remains a major agenda ahead, notably in sustained growth of incomes of both the extreme poor and the moderatepoor, in raising quality of all basic services, and indealing with risks. We conclude with some comments bothon areas for action and future work plans. Areasfor action in the short and ~ e d i term. This overview already provides a number u ~ of results with implications for action for poverty-reducing policy, inboththe short and mediumterm. Some examples are: 0 Pu~lic~nance.the short run, a tax reformthat increasestax collections can be In powerfully poverty-reducing. Eventhe removal of food and medicine exemptions under VAT could lead to substantial poverty declines if the resources are well- used, through channeling into an expanded version of Oportuni~ades,for example. To continue improving the progressivity of electricity subsidies is another prime target with large benefits for equity andefficiency. Social protection. In the short run, there is potential for further extension and consolidation of Oportunidades, looking to gaps in coverage for the extreme poor, backed by continued careful attention to monitoring, evaluation and accountability in light of the program's size. Inthe medium term, there is a need to reform the highly inequitable ISSSTE, and design options for the missing middle of the moderate poor. Seguro Popular shows great promise for helping the poor with health risks, but its impact is not yet established and rigorous evaluation will be important. Sewice quality. The service quality agenda is fundamentally a medium-term one of strengthening the accountability of providers -teachers, nurses, doctors- bothupwards interms of linkingincentives to results, and outwards to clients and communities. In the short-term there are many initiatives (e.g. Currera Magisterial for teachers; Cruzada Nacional por la Calidad de 10s Servicios de Salud for health workers) where specific design changes can increase effectiveness in improving service quality. The broader agenda of linking performance to results i s likely to require substantial strengthening both in accountability structures, especially at subnational levels, and in providing instruments whereby the federal government can influence reward the performance of service providers at the state and municipal level. lii Competitivenessand income growth of the poor. The competitiveness agenda is intimately linked to a poverty reduction agenda because of the centrality of employment creation. As one example, economy-wide measures both to expand infrastructure spending and deal with high costs and inefficiencies (whether in electricity, telecommunications, or roads) of infrastructure services, can be good for both growth and distribution. This type of measure will need to be complemented by targeted action for the poor inbothrural andurban areas: 2+ In rural areas, two areas for short-run action including genuine experimentation with territorial approaches -that will imply solving current coordination problems in some infrastructure programs, and continued facilitationof flows of remittances. 2+ In urban areas, there are short-term options for reducing the cost of formality -for example through increasedflexibility of hiringand firing and reducing tax and regulatory burdens- and a more complex longer- term agenda of broadening financial and infrastructure services, titling, andproviding technical assistanceto micro and small firms. Accountability. Getting effective accountability of government programs to citizens is a complex and medium-to-lon term endeavor, but one that is fundamental to effective government policy *$ .There have beennotable initiatives inthis, including the Law on Public Access to Information (2002) and the Law on Social Development (2003). However, international experience indicates that accountability also requires strong formal and informal institutions, including at local levels, to make such laws work effectively. Evaluation and results-based management. Monitoring and evaluation are fundamental for the review and ongoing redesign of poverty reduction programs. The government has good frameworks for evaluation (especially in SEDESOL) and a relatively new results-based management system (for example in Presidenciu) but these are weakly institutionalized. In the short run, there is scope for extending systematic evaluations of particular programs -especially extending the plans for systematic evaluation under development in SEDESOL to other ministries. Institutionalizing results-based management is intrinsically a mediumto long-run activity. Future work plans. There are many areas where deepening the analysis and exploring policy options are needed. Areas currently planned under World Bank's work programinclude the following: Rural and urban poverty. Parallel studies are under preparation on the particular characteristics of poverty in urban and rural areas. These will focus in particular on the structural and institutional influences on income growth -identified as a 24Examplesincludethe Citizens' ReportCardin Bangalore, Indiaand Cd'movamos? inBogotB. liii major issue in this overview- patterns of vulnerability and coping, and issues of social incorporation, as well as analyzing government programs inthese areas. Social protection. To complement the diagnostic on vulnerability in urban and rural areas, this study will focus on the effectiveness of current policies and institutions to support risk management. Itis expectedto focus bothon options for consolidating policies for the extreme poor, and the important middle groups not coveredby formal social security or likely to be reachedby ~ p o ~ t ~ n i d ~ e s . Decentralization and service delivery, Following the above studies, it is planned to deepen the understanding of how to get effective and quality service delivery under decentralized auspices. This i s of rising importance with the large and growing devolution of responsibilities to states and municipalities, and the complex issues of equity, institutional functioning and accountability that this raises. Both the current poverty assessment and future work are intended as a contribution both to government policy formulation and to broader debates, to strengthenefforts to reduce poverty in Mexico. It i s hoped that they can play a useful role in these important endeavors. liv Poverty inMexico An Assessmentof Conditions, Trends andGovernment Strategy Chapter 1. Introduction andFramework This report examines the structure andcauses of poverty inMexico, andexplores options for strengthening government strategiesto reducepoverty. Despite improvements in many aspects of well-being in Mexico, in both long and short term, concern over poverty is high on the agenda of societal concern and policy debate. Official estimates of poverty in terms of income (that is the proportion of people who have incomes below levels judged necessary for decent living) find that about half the population lives below the poverty line, with almost 20% living with insufficientincome to cover basic food needs. Despite significant reduction inincome poverty since the crisis of 1994-95-including inthe most recent 2000-2002period- poverty incidence rernains close to levels prevailinga decade ago. Lack of income i s only one measure of the well-being of disadvantagedgroups. Poverty has many dimensions including inadequate educational and health status, lack of access to basic services, vulnerability, low self-esteem and social exclusion. Insome of these- especially those relating to education, health and access to basic infrastructure- there has been substantial and sustained progress in Mexico. But even in these areas there remain important concerns over reaching the poorest and over low service quality for the poor (and the non-poor). Measures of vulnerability are less well established, but we present quantitative evidence that shows that risks of adverse shocks for the poor, and of falling into poverty for the non-poor, are widespread. There is even less quantitative evidence on questions of self-esteem and social exclusion but case study works indicates that these are important issuesfor boththe condition anddynamics of poverty. The causes of poverty are multiple. Government efforts to reduce poverty operates within a context in which both longer-term structural influences and short-term external forces play major roles. Structural factors include patterns of asset ownership, geography, besides formal and informal institutionalinfluences, which flow from political, social and cultural sources. Short-term external influences include the effects of international economic influences on the state of the economy, patterns of prices and employment opportunities, as well as weather-related shocks. Mexican governments have a long tradition of both economy-wide efforts to reduce poverty through job-creating growth and basic service provision, and a wide range of targeted policies for poverty combat. Some of the latter -notably ~ p ~ r t u n i d ~ e s (previously Progress)- arejustifiably world-renowned. The current administration took the valuable step of explicitly articulating an integrated social government policy in the form of the CONTIGO framework (Szkkely, 2000). This i s a conceptual framework for understanding the dynamics of well-being, that recognizes the manifold influences on 1 household living conditions, and also an organizing framework for a subset of the government's programs. This report aims to do the following. This introductory chapter outlines the overall framework for understanding well-being and analyzing public action to reduce poverty. Chapter 2 provides a brief overview of living conditions in Mexico in international context. Chapter 3 then goes to the core of the diagnostic of poverty conditions, examining patterns and trends for the different dimensions of poverty up to the most recent available information (generally 2002), and interpreting the casual influences on these trends. Chapters 4 and 5 then turn to government strategy: Chapter 4 reviews the overallmix, structure, andtrends inpublic spending andgovernment programs inrelation to the poverty and inequality with a particular focus on social services and transfers. Chapter 5 then discusses the influence of government policy on changes in income of the poor (andnon-poor) that often works through the more indirect, but potentially powerful, influences of economy and sectoral policies, rather than specific programs. Finally, Chapter 6 reviews issues in monitoring and evaluation in Mexico, within a global context, which is also an areainwhich there have been some important initiatives. This report is the first major product ina series of studies that will beundertakeninclose collaboration with the government of Mexico. These are expected to include an special in-depthfocus on issuesof ruralpoverty, urbanpoverty and social protection in2004 and onlinkswith decentralization in2005. Itshouldbetreated as acontribution to anongoing dialogue on the central issueof what poverty means andhow to reduceit inMexico. A. Concepts: what is poverty and why does itmatter? Discussions on poverty, in Mexico and elsewhere, have often been dominated by questions on the numbers of people livingbelow or above a poverty line defined interms of incomes or spending. This is an important part of the story, and this report will indeed devote space to the analysis of trends and causes of income poverty. But this i s only part of what is meant by poverty in a society. Both empirical work on what people think and conceptual considerations support the view that poverty has multipledimensions andi s as much about the future -in terms of risks and possibilities- as well as current conditions. One reason for taking a broader view of well-being and poverty i s that this reflects what people believe, not least those who live in poor communities. This i s reflected, for example, in the tradition of participatory poverty research, which seeks to systematically understand the nature of poverty from structured listening and interactions with groups and individuals living in poor areas. The international archive collected in Voices of the Poor was a worldwide undertaking of this type that included several Latin America countries.' SEDESOL recently undertook a somewhat similar exercise for Mexico, that relied on a questionnaire-based approach (SzCkely, 2003). While the methodological 1See referencesby Narayan and co-authors (1999,2000, 2001). The work covered some 60,000 people in 60 countries, including Argentina, Bolivia, Brazil, Jamaica and Ecuador in Latin America and the Caribbean. Follow up work has included other LatinAmerican countries, includingPeruand Colombia. 2 approaches differed, there were significant common findings. The Voices of the Poor studies found that ill-being (or poverty) did indeed mean lack of material meansbut also meant many other things, specially ill-health, social exclusion and isolation, insecurity and feelings of powerlessness, helplessness and frustration. By the opposite, a good life means having enough for basic material needs, being and appearing well, social well- being and inclusion, security and freedom of choice and action (Narayan et al. 2000 Chapter 2). In SEDESOL study for Mexico, 23% of respondents said that well-being (bienestar) meant having enough to eat, while 16% said it mean good-health, 13% basic services, 11% having work, 5% having peace. Mexico also has a rich tradition of ethnographic and other field-based work that documents the complexity of poverty.2 Participatory research is not a substitute for survey-based quantitative work, but does provide an invaluable source of understanding what well-being and poverty means for poor people andhow they advice government andother attempts to provide services. Taking a broader view of well-being and poverty also has strong conceptual roots. A major influence on development thinkinghas been the work of Amartya Sen. Three ideas from his work are of especialrelevance. The first concerns what constitutes well-being - and so poverty. Sen argues that it is what people can do or be, or their "functionings" that should be assessed as achievements in well-being. Such achievements can range from quite basic functionings, including "being well-nourished, avoiding escapable morbidity and premature mortality", etc. to more sophisticated ones, such as "having self-respect, being able to participate in the life of the community and so on". Material conditions, including income, might develop the realization of such achievements. Second, it is of great importance to distinguish between "achievement" and "the freedom to achieve" in assessing both the extent and normative significance of differences in well-being. The "capability set" represents the range of potential functionings that an individual can achieve, or the "overall freedom a person enjoys to pursue her ~ell-being".~ Third, questions of human agency are central to both normative and positive approaches - which we needto look at individuals as purposive agents rather than passiverecipients of their conditions andgovernment actions. Related to both the view of poor people and conceptual approaches such as Sen's i s the importance of the future, whether in terms of possibilities and opportunities, or the "freedom to achieve" (or its absence), or of the risk of falling into poverty, ill-health or other forms of deprivation. These approaches were at the core of the World Bank's 2000101 World Development Report on Poverty andDevelopment, as well as the series of global HumanDevelopment Reports of UNDP from the past ten years or so, including the recent country Human Development Report on Mexico (UNDP, 2003). And, of particular relevance to this report, they are consistent with the vision of social development expressed by the government of Mexico, as well as the work of the Technical Committee on Poverty Pioneering ethnographic work on the culture of poverty was by Oscar Lewis in Mexico. While the concept of a "culture of poverty" is now, correctly, controversial, the work forms part of a tradition that finds the complexity and socialembeddednessof poverty. Sen (1992) pp. 4-5 and p. 150 3 Measurement (see Box 1.1 at the end of this section). As an example, in 2002 the Secretary of SEDESOL, Josefina Vkquez Mota stated: "El desarrollo social es el proceso permanente mediante el cual se ampl~anla capacidades y las opciones de tas personas y comunidades para que puedan ejercer plenamente sus libertades y derechosy realizar todo su potencial productivo y creativo, de acuerdo con sus aspiraciones, elecciones, intereses, convicciones y necesidades ". (SEDESOL, 2002) ("Social development is the permanent process through which the capacities and opportunities of people and communities are expanded, so that they can fully exercise their freedoms and rights, and fulfill their productive and creative potential, according to their aspirations, choices, interests, convictions, andneeds".) Under this approachpoverty is the lack of what a society considers the basic minimumin terms of the range of dimensions that constitute well-being. As a general concept this is necessarily complex, and fundamentally political - e v e n if we can (and do) quantify relevant elements. As a recent report on poverty in Indonesia starts: "Poverty is an idea: a political and social idea that reflects a society's hopes andaspirations. Poverty is what we hope to eliminate". (World Bank, 2001a.) What does this mean in practice? We suggest explicitly recognizing various dimensions of poverty, which can provide the basis for both analysis of the condition of poverty and its causes. While there i s always an element of choice inconstructing such a list, here are a plausible set of categories that are consistent with international thinking and current work inM e ~ i c o . ~ Poverty of human capacities. This refers to lack of possession of minimal human capacities to lead a healthy and productive life. This can be framed in terms of a condition -as in freedom from "escapable morbidity or premature mortality" and possessingliteracy and other basic skills. It i s also often expressedinterms of access to a service of basic education and health services of adequatequality that support these ends. What i s considered "basic" depends on the level of development, notably in the area of education. Poverty of assets and infras~uc~ure~ Beyond human assets, lack of physical and financial assets i s an element of poverty. These include private goods such as housing, as well as access to publicly provided goods. In many societies access to a basic set of infrastructure services, such as water and sanitation, roads and electricity, i s considered essential to a minimum level of living standards. The nature of this list will again often depend on the level of development: basic water and sanitation services are almost universally judged necessities; in middle income countries electricity would typically be included. See World Bank (2001a) for a similar approach for the case of Indonesia. 4 Income or expenditurepoverty. This defines a minimal level of incomes or spendingthat allow the purchase and consumption of a minimum basket of private goods, including food, clothing, housing and privately purchasedelements for education, health and other services. It can be defined interms of either incomes or expenditures. While poverty lines tendto behigher for richer societies, they are usually heldconstant over time for at least the medium term, in order to analyze changes in "absolute" income poverty. InMexico the official poverty lines are absolute and basedon incomes, so we use the term "income poverty" as a convention. (See Box 1.1 for a discussion -as discussed there, there are reasons for preferringconsumption to incomes as a monetary measureof well-being, and we work with both inthis report). InMexico and elsewhere it is recognized that there are degrees of poverty defined by income (or consumption). This is reflected in the use of different lines, that range from the minimum needed to cover a basic food basket, to higher lines to cover spending on health and education, and to cover a more comprehensive set of consumption requirements. It is also reflected inthe use of poverty index that measure the distance poor people fall below the poverty line in question, and inequalities amongst the poor. Vulneru~ility.The issue of vulnerability, or the risk of falling into poverty, is emphasized in all participatory poverty work. There is also a tradition of economic analysis, though there are less well-developed measuresof vulnerability than there are for current poverty. This is in part because of data limitations, since special survey designs are required to capture the phenomena (employing panel surveys andlor retrospective questions on the experience of shocks). While panel surveys are much more limited in scope in Mexico than the core cross-sectional household survey instruments, the report presents some quantitative analysis on the pattern and trends of vulnerability, as well as discussing policy instruments for reducing vulnerability. Poverty of dignity and social inclusion. Socially based indignity and lack of social inclusion are further elements of poverty. Such social functionings (to use Sen's terminology) have a long intellectual history, going back in economics at least to Adam Smith's emphasis on the importance of "appearing in public without shame" and also includes "taking part in the life of the community" (Sen 1992, p. 115). However, a standardapproachto quantitative analysis of these areas as dimensions of poverty has not been established, and a variety of sociological and political analyses are likely to be necessaryto get a handle on them. This report will generally work with this list of areas of poverty. Indoing so three points are worth emphasizing: First, different dimensions of poverty will be- of varying degrees of importance for different individuals and groups. While multiple deprivations are of particular concern, there is no obvious metric for developing a unitary index. Nor i s this desirable, since it would obscure the structure of deprivation, which is an area for social and political debate. 5 Second, indebates and diagnoses on poverty, it i s common to treat poverty as a condition that someone either has or does not have -as with poverty lines or minimum service levels. This i s a misrepresentation of any sensible approach to valuation, which would give greatest weight to more severe degrees of deprivation. A common assumption, inthe language of economists, is that social welfare functions defined over a particular dimension are concave, or place continuously declining weight to increments to consumption or income. We are often interested inthe structure of a particular dimension of well-being across the whole distribution: there are important linkages between questions of poverty and inequality both for normative analysis of well-being and understanding causalrelations. Third, while all the areas listedabove areconstitutive ofpoverty, andso of social concern in their own right, they also have causal influences on other aspects of well-being in a society. For example, human investments and access to infrastructure have powerful instrumental influences on income poverty; social inclusion can influence the quality of service delivery, etc. Poverty, in its various dimensions, can also affect the broader pace of development, for example where lack of immunization or access to clean water increases risks of infectious diseases; where credit constraints of the poor lead to under- investment; or where social exclusion or polarization are associated with heightened distributive conflicts.' The latter effect is strictly about social inequalities, but there can be significant links with dimensions of poverty. Box 1.1. Mexico's Technical Committee on Poverty Measurement and the Choiceof Income Poverty Lines Untilrecently Mexico did not have an official position on what constitutes poverty. This is not unusual amongst countries, though a few, notably India and the United States, have a long tradition of maintaining an official income poverty line. Since the definition of poverty is necessarily a product of both technical analysis and political process, it is a major asset to have such an official position. This can help shift debate from how "many poor thereare" on to the really important questions of trends, causes and what to do. In 2000, the government of Mexico had the excellent initiative of convening a committee of experts to prepare an official position on poverty, the Technical Committee for Poverty Measurement (Comitk Tdcnicopara la Medicidn de la Pobreza or CTMP, by its acronym in Spanish). This comprised a range of academics, with observers from the government (The latter group does not have voting rights, in order to preservethe independence of the committee). The assessment work of the committee has been of high quality, and made full use of the international literature on poverty measurement (see CTMP, 2002, for a discussion). There is room for debate on a few of the choices made in methodology -and the committee itselfexplicitly recognizesthis. However, one of the most important contributions of the work is setting up a transparentand ongoing process. In terms of the overall conceptual framework the committee recognized the multi-dimensional and complex nature of poverty, and the need to employ multiple indicators in assessing the structure and trend -an approach broadly similar to that describedinthe text. In economics there has been a significant, relatively recent, literature on links between inequality and development; some of the theoretical arguments are essentially about poverty, others apply intrinsically to inequality. See for example Aghion et a1 (1999) and BBnabou (2000) for surveys, and De Ferranti et a1 (2004), Chapter 1for a brief discussionina LatinAmerican context. 6 A particular focus of the committee's work was the choice of poverty lines for income poverty. Three elements are needed to compute a poverty measure: (a) an indicator of well-being at the household level, such as per capita consumption or income; (b) a poverty line to which the indicator can be compared; and (c) a statistical tool (the poverty measure itself) used for reporting at the aggregate level the results of the comparisonfor each householdof its indicator with the poverty line. The key points are summarizedhere. Welfare measure: There are eight rounds of the Household Income and Expenditure survey (1984, 1989, 1992, 1994, 1996, 1998,2000, and 2002). The ENIGHis fielded on the third quarter of each one of these years. Issues have been raised regarding the comparability between the ENIGH surveys in the 80s and those in the 90s. Monetary income is comprised of income from different functional sources spanningthe 6-month period prior to the survey date. This feature makes possible the adjustmentby functional sources, Le., wages and salaries, profits, rental income and etc. Expendituresare obtained for different time periods according to the type of good purchased. The CTMP decided to use net per capita household current income as the welfare measure. The total current income of the householdsis the sumof monetary andnon- monetary income. The monetary income comprises earnings, income from businesses, capital income, transfers, income from cooperatives, and other income. The non-monetary income includes auto- consumption, payment-in-kind, gifts received in-kind, and an estimate of the housing rent. Before calculating the per capita monthly income of each householdtwo adjustments were made: 1) the category "other income" was subtracted from the total current income and 2) gifts granted was eliminated. The major questionhereis whether income or consumptionis a better monetary measureof well-being. There is in fact a strong case for usingconsumption as the indicator of material well-being, both because it will reflect the smoothing strategiesof householdsinresponse to income shocks and variations in savings over the life cycle, and is likely to be more accuratelymeasuredthan income.6 Inthis report we use both income and consumption. Normalized average monthly income and expenditure: The CTMP recommended calculating all income and expenditure in average monthly figures in August of the corresponding year's prices. Incomes are deflated usingthe generalprice index, while expendituresare deflated usinggood-specific price indexes. Poverty lines: Regarding the specification of the poverty lines, SEDESOL uses three lines, building on the CTMP's proposal. The first is the food-basedpoverty line, which is an estimate of the income required to purchase a food basketto satisfy minimumnutritional requirements. The Basic FoodsBundle (CBA, by its acronym inSpanish) was defined and calculatedby INEGIwith the collaboration of ECLAC. This bundle is calculatedseparately for rural and urban areas, and is based upon the food spendingpatternsof households who just satisfy minimumnutrient requirements, assuming all spending was on food. This is calculatedfor urbanandrural areas. Then, the food poverty line i s deflated usingfood specific consumerprice index from Banco de MBxico. To compute the second poverty line, food expenditures and food shares are ranked by per capita total expenditures based on the ENIGH Survey, By intersecting the estimated poverty line one with the food expenditure line and taking a range of roughly five percentile points to each side of this intersection, the average non food component is used in determining poverty line two. The inverse of the Engel coefficient i s the Orshasky coefficient that has to be multiplied by the corresponding poverty line one in order to obtain the poverty line two for urban and rural areas. This became known as the "assets" or patrimonio line. SEDESOL then added a third intermediate line, that i s intended to account for somewhat above the food-based line, Table 1.1 presents official poverty lines for the past decade at current prices, as used by SEDESOL. With respect to the calculation of lines, there are some questions over the constancy of the non-food ratio (official lines do not always use a constant EngellOrshasky coefficient) and over the choice of price deflator. This report works both with official poverty lines, and with re-calculated lines that use a constant 2000 Engel coefficient and Banco de MCxico price indexes for non-food items of the poor. These are reportedinthe annex to Chapter 3. As discussed there, the overall pattern of changes inpoverty is quite robust to different poverty lines, as well as choice of income or consumption. For discussionssee Deaton (1997) and Ravallion (1993). 7 Table 1.1. OfficialPoverty Lines inCurrent Pesosper month Zone 1992 1994 1996 1998 2000 2002 FoodBased Rural Urban 167,955.2 194.0 389.4 524.8 626.0 672.3 124,750.6 143.3 290.0 388.5 462.9 494.8 Capacities Urban 197,963.8 236.1 469.4 622.1 767.7 792.6 Rural 140,586.3 165.6 335.6 445.0 548.4 587.3 Assets Urban 333,122.7 414.5 804.2 1054.5 1,255.3 1367.4 Rural 218,794.4 278.7 527.2 690.9 842.6 946.5 Poverty measures: Measures of poverty need to take account of both the extent and depth of poverty - deprivation is much worse the further someone is below the poverty line. The most widely used poverty measures, endorsed by the committee, are the FGT indexes (Foster, Greer, and Thorbecke, 1984). The incidence of poverty, denoted by Po or H for the headcount index, is simply the share of the population living with income or consumptionbelow the poverty line. The depthof poverty, measuredby PI or PGfor the poverty gap, captures the distanceseparatingthe poor from the poverty line as a proportion of that line (the non-poor having a zero distance). The severity of poverty, measured by Pz or SPG for the squared poverty gap, takes into account not only the distance separatingthe poor from the poverty line, but also the inequality among the poor. Higher order poverty measures are rarely used and are harder to interpret. Denoting by Yi the indicator of well-being at the household level, (inthe absence of information on intra- household allocation), by N the population, by wi the household's weight, (household size times the expansionfactor, the sum of the weights being equal N), and by Z the poverty line, the FGT measures are usually applied with values of 8 equal to 0, 1, and 2 in: While these indexes have many desirableproperties, including decomposability across groups, they do not escape the essential arbitrariness of the poverty line itself. Poverty lines are of great use in focusing public attention, but there is not a sudden change in well-being when you cross the lines. Techniques for more robust assessment of whether poverty has changedare illustrated inChapter 3. inequality measures: The CTMP has not taken a position regarding how to compute inequality. This reportuses the CTMP per capita income and expenditure definitions for inequality measures. Alternative definitions of income and expenditure were also used and the yielding measures were consistent with the CTMP definitions. There are a number of areas for future work, which were noted by the committee. These include the choice between income or expenditure (noted above); questions on the use of equivalence scales for differing needs betweenindividuals (e.g. children) and economies of scale inconsumption for larger households; and price variation across regions. B. Frameworksfor reducingpoverty We now turn to frameworks for assessing causal mechanisms affecting poverty. This is, of necessity, a stylized sketch, since it would be impossible to summarize the vast literature on the topic. However, it is useful to have an organizing framework, or frameworks, for analysis of causes of poverty and the influence of policies. Moreover, there has beena significant evolution of thinkinginthe past decade. At the core of many approaches (including the one usedbyMexico's CONTIGO, that we outline below) is the interaction between assets and opportunities. A relatively simple 8 version of this was presented in the World Bank's 1990 World Development Report on poverty that saw the dynamics of poverty reduction as a product of the interaction between the expansion of human capital and jobs. Human capital formation was there conceived to be largely a product of government provisioning of education and health services; while jobs were determined by the level and pattern of growth. It then saw a complementary role for transfers and safety nets for those unable to participate in the growthprocess andat risk of shocks. This characterization remains valid but it is incomplete as a treatment of the process of expansion of well-being in a society and consequently of poverty reduction. There is a need for a broader account of capabilities in terms of the interaction between the economic and social assets of individuals and groups and the opportunities for their use, inboth market and non-market contexts. For this market, political and social institutions all matter. Of particular importance for the poor are the contextual determinants of opportunities for work (whether in wage-paying activities or self-employment) and the potential to directly or indirectly influence decision-making affecting service delivery. These are a function of the patterns of inequality, institutional structures anddevelopment strategy. To illustrate this general point we highlight three areas where a richer account of processesi s important: inthe determination of incomes; ininfluences on service delivery; andthe centrality of volatility. Income determinution and the i n ~ e r ~ t i obetween economic, politicul and social n mechunisms. Opportunities, assets and outcomes are related in complex ways. It would be impossible to give a comprehensive treatment of these complex interactions. The distributionof incomes is determined within the general equilibrium of the economy, and Mexico i s a complex modern economy, with market imperfections, incomplete information, strategic interactions andpolitical processesall takingplace simultaneously. We present a very simple general framework of the circular interaction between the distributions of assets (andopportunities), incomes (and other outcomes) andpower (and influence in society) -this i s discussed further in recent work on inequality in Latin America (de Ferranti et al, 2004). This i s schematically presentedinFigure 1.1, where the distribution of assets i s depicted at the top. These assets include not only physical and financial wealth, such as land or stocks, but also human capital endowments, such as education. Individuals may inherit some of these, but others are produced during a person's lifetime. People then make choices about the use of their assets in markets, where they are remunerated. Financial savings receive interests or dividends, inbonds or stock markets; land generates returns through rents or profits; human capital i s remuneratedinlabor markets. The combination of incomes arising from the remunerations of assets in these different marketscombines to form individual incomes. Primaryhouseholdincome dependon how individuals match to form households, and on their family size and composition decisions. Finally, secondary income are also affected by taxes and transfers, grouped together hereunder public redistribution. 9 Figure 1.1. A simple conceptual framework for changesinwell-being Assets &Opportunities Outcomes (& Incomes) These economic processes do not take place in a vacuum. At every step of the way, they are mediated by social and political institutions, very broadly understood, to encompass the rules and norms of behavior in society, but also more specific institutions which formally or informally regulate markets or affect government^.^ These include both the formal institutions in societies, and the socio-cultural processes that are products of the interactions between different groups in a society, notably those between dominant and subordinate groups, that affect the daily lives and practices of poor groups ininteractions with service providers, ineconomic exchange, inthe labor market and so on.' While formal and informal institutional arrangements are central to economic functioning, they are neither immutable nor indifferent to economics. Income and wealth - o r `economic power'- i s closely linked to political power, influence and voice and this isrepresentedherethrough the linkbetweenincome andpower, mediatedby political institutions and socio-cultural processes. Like the economic processes heroically summarized along the hypotenuse of this simple triangle, this side subsumes whole bodies of social science andpoliticaleconomy, inan extremely reduced form. Institutions shape economic realities in many ways. Despite the magic of the "invisible hand", markets themselves never really operate independently of regulatory institutions or the See North(1990) for a classic treatment of institutions as norms and `rules of the game'. As sociologists such as Charles Tilly and Pierre Bourdieu have emphasized, the latter are essentially relutionul, and are deeply intertwined with both the organization of economic production and the structure of power. See Heller and Mahoney (2003) for a review, Tilly (1999) andBourdieu (1990) for key examples of their work, and Rao and Walton (2004) discuss "inequality of agency" from a cultural perspective. A related strand of the economics literature on measurement of inequality concerns the development of the theory and measurementof polarization. This starts from the conceptualinsight that individuals evaluation of the salience of inequalities is linked to differences between groups with which they identify relative to other groups, from which they feel relatively alienated. SeeEstebanand Ray (1994). 10 purposeful activity of different groups, with more or less olitical and market power -a theme that Adam Smith himself forcefully emphasized. At the most basic level, the 1 enforcement of law and order is necessary to secure the property rights of those coming to sell, andthen of those who buy. And as the commodities themselvesandthe markets in which they are traded become increasingly complex -moving, say, from apples in a farmers' market to, say, micro-enterprise credit or financial derivatives -the enforcement of property rights and the need to correct for informational andother market failures also grows. Financial regulators, inspectors of working conditions in labor markets and anti-trust investigators are examples of governance institutions through which political systems provide a frame for the activities of markets, shaping economic processes and outcomes. Policy can also affect the distribution of assets, either through direct redistribution (such as via land reform or privatization processes) through taxation and subsidies and the influence on the options faced by individuals and groups (as in the direct financing and provision of education). The overall economic incentives with which agents are faced shapes differential rates of accumulation, which influence how asset distributions evolve. Finally, governments affect the distributionof disposablehouseholdincomes, through the need to finance their operations. Taxes serve to raise the revenues for a number of redistributive services and transfers, whether in kind (such as in the provision of free education or health care) or in cash (such as in unemployment insurance or various cash subsidies), as well as for the regulatory roles. Ineachof these steps initialconditions andpolicy choice affects the environment for the poor groups, and influences their choices over schooling, work, investment in physical assets, migration and a range of other behaviors affecting economic and social conditions. This `circular causality' between wealth, incomes and power, mediated through institutions, evolves through time and history. History has placed poor groups at a disadvantage in their initial economic conditions, in the shaping of policy and in its implementation. However, this does not imply we should be fatalistic. There i s scope for purposive action of policy-makers and a range of other agents, including the poor themselves, to break through vicious cycles of deprivation and the reproduction of inequality. Mexico's past and recent history provides both evidence of forces that have tended to perpetuate poverty and social exclusion and of such purposive action to reduce poverty and inequality. There is a wide range of initiatives under implementation now. However, in assessinghow to strengthen such initiatives we believe it i s useful to place them within this broader context of interactions between economics and institutions, sketchedinhighlyschematic terms inFigure 1.1. Service delivery. The delivery of services i s a key nexus for the interaction between institutions, economic behavior and well-being. Basic services such as education, health, electricity, water, sanitation and roads are both constitutive of well-being (as noted See Rothschild (2001) for a rich account of how Adam Smith and other enlightenment thinkers were greatly concerned with the abuse of unequal influence, and the role of purposive agency of different groups in shapingeconomic andsocial outcomes. 11 Most Latin American countries have long histories of relatively weak accountability to their citizens, associated with "weak and unequal" institutional structures, that granted disproportionate influence to elites." This has had a major influence both on the design and implementation of service delivery. The process of democratization at national and local levels, that restartedinmuch of the region inthe 1980's, has of course increasedthe demands of middle and popular sectors for better services. But the consolidation of democracy i s a long-term project, and the legacy of weak institutions, unequal influence and social excluded groups canbe strong. Take two areas of service delivery as examples: provision of infrastructure services and education. Infrastructure services were traditionally provided by public utilities in Latin America, inan approachthat was embedded inclientelistic structures. Utilities provided jobs for government supporters and subsidized electricity, phone, water or sanitation services to those lucky enough to have access to services. But pressures for efficiency were low, and rationing of services widespread. Fiscal difficulties exacerbated these problems. Privatization was a way of breaking through clientelistic patterns, providing incentives for greater efficiency, and a source of fiscal revenues. In many areas, privatizationdidyield efficiency and quality gains, with often positive pro-poor effects in terms of service expansion and modest overall distributional impacts (McKenzie and Mookerjee, 2002). Butthey typically brought layoffs, and insome cases increasedmarket power to domestic conglomerates and their foreign partners. Education provides a second example. Most countries in Latin America have a relatively longtradition of social demands for widespread provisioningof basic education, that has been reflected insupport from politicians and policymakers. There have been a variety of reforms in the "compact", especially in the decentralization of provision to the states or lower levels of government, and insome countries experiments ingreater client power of parents and communities at the school level. But education has also been a source of patronage in provision of jobs, and of extensive influence of teachers' unions. While unions have an important role to play indefending the interests of their members, and are sometimes forces for reform (e.g. in Minas Gerais inBrazil), they are more often source of resistanceto quality-enhancing reforms (Grindle, 2002). At the tertiary level, much of Latin America still has highly subsidized public universities, despite the fact that these are largely attendedby the children of higher income groups, who fiercely resist attempts to reducethese (inequitable) subsidies. This sketch is intended to illustratethe importance of assessingthe importance of how to strengthenservice delivery to the poor within abroader institutionalcontext. On volatility and vulner~bility. The third area we emphasize concerns the role of volatility in affecting economic conditions. This is both constitutive of well-being (see above), and has pervasive effects upon other aspects of economic and social behavior. loSee De Ferranti et a1 (2004) for a review of both historical roots and contemporary political and social institutional structures. 11 See Estache, Foster, and Wodon (2002), World Bank (2003b), and De Ferranti et a1 (2004) for discussion. 13 Volatility comes from many sources. This include influences that are specific to individuals or households --"idiosyncratic" shocks in the case of adverse effects- such as the ill-health, death or disability of a breadwinner, loss of a job in normal times, or theft, fire or damage of assets for the self-employed and small-scale entrepreneurs.There are also a range of common of "covariate" shocks that affect whole communities, regions of countries, including poor rainfall, natural disasters and economic recessions. All of these are importantinMexico. Work on volatility emphasizes that individuals, households and firms adapt to the prospect and occurrence of shocks.l2This can be through pursuing lower-risknower return strategies (e.g. drought resistant crops, pulling children out of school), through risk-diversification (as in household labor allocation strategies including migration), through a variety of approaches to savings -or self-insurance- including building up physical and financial assets that can be liquidized in times of needs, and through reciprocal social relations. Private mechanisms for managing risk are, however, often sub-optimal, notably because the lack of insurance and credit markets -and especially so in the face of covariate shocks. There i s a strong case for public action on social protection. But it i s desirable that this be structured around patterns of private behavior and specific problems of marketfailures andconcernsover distribution. This implies payingattentionbothto areas that can directly help suchprivate behaviors, most of all infacilitating asset accumulation and access to financial markets, as well as the more classic instruments of social insurance, involving transfers in response to the variety of adverse shocks, from natural disasters to unemployment. Both the capacity to deal with adverse shocks and the reach of public action has a powerful distributional dimension in most developing countries. The poor are typically more vulnerable due to their low asset base and weak access to financial markets to manage volatility. At the same time, the pattern of public provisioning that evolved throughout Latin America took the form of a truncated welfare state, that included relatively well-off households connected to formal work and institutions, and excluded the bulk of poorer and many middle groups. There i s a strong case for an expanded, and explicitly pro-poor approach to social protection throughout Latin America; this i s likely to involve both a mixture of reform of formal systems and special measures oriented to the poor. Mexico in transit". How does Mexico fit within these frameworks? It is in the midst of two major, long run transitions. In terms of economic institutions, this involved the radical opening up to external markets and capital, combined with internal privatization andliberalization; this was consolidated under NAFTA. Inthe political andsocial sphere, l2See Deaton (1997), Chapter 6 for a discussion of savings, consumption smoothing and social insurance from the perspective of household behavior and development policy. For recent practical applications to policy see World Bank (2001a) and for a framework and review of Latin American experience see De Ferranti et al (2000). World Bank (2001a) provides an application to Mexico, which is built upon in the work for this report. 14 it involved the deepening of democracy, moves to increase accountability of the state to its citizens, and political and administrative decentralization; this started in the 1990's and acceleratedunder the present administration. These transitions were from a systemof economic management and social provisioningthat was shaped inthe decades following the revolution and the later formation of the corporatist state, in particular in the 1930's under president L&aro Ciirdenas. The corporatist modelhadconsiderable success for severaldecades leading to highlevels of state and private investment and steady expansion of services. But it hadrunits course in economic terms by the 1970's with a legacy of inefficient patterns of production. Despite commendable goals of social inclusion, it failed to make a major difference to traditionally high levels of inequality. Implementation of basic services was historically far from universal. Giventhe limitationsof the state andthe inheritedextreme inequality, the result was typically one of partial, unequal and dualistic structuresof employment and social provisioning. This affected both patterns of access to services by citizens and the supply of servicesby public sector workers. With respect to service access, institutional inequalities took different forms in different realms of social provisioning. Insome realms dualistic principles of provision prevailed. This is most marked inthe social protection systems for pensions, healthandemployment -the "truncated" welfare state that i s so marked a feature of Latin America (De Ferranti et al, 2004). In others, universal principles took the form of a more gradual process of bringingpoorer andexcluded groups within the net of service provision-as inthe cases of basic education, water, sanitation and electricity (as discussed, for the recent period, in Chapter 3.). The ideology of universal provision of wage-goods also took the form of a costly, untargeted subsidy on tortillas prior to 1982. Despite minimum food prices for surplus producers the net effect was to tax agriculture in this period, because of an overvalued exchange rate. After the 1982-83 crisis and subsequent fiscal adjustment, the tortilla subsidywas gradually phasedout andeventually eliminated in 1998. On the supply side, the corporatist project was associatedwith provision of public sector jobs to government supporters, notably in the teaching service, in social security administrations and in utilities -a classical form of clientelism in Latin America and elsewhere. This part of the historical legacy remains relevant, in the challenges of achieving quality from service providers, and reducing rents associated with the leverage of organized workers, whether via their market power or connections with political parties. The two transitions can be viewed as seeking to take Mexico toward a modern, democratic, effective, state, with a dynamic job-creating private sector highly integrated into the global economy, and with broad-based provisioningof social services and social protection. But the transitions are necessarily long and complex, and the legacy of the past remains significant. Future chapters will be exploring this with respect to poverty conditions andgovernment policies. 15 C. The Government'sCONTIGOframeworkfor povertyreduction CONTIGO i s the mexican government's strategy to promote human and social development through a better coordination of programs and their responsibleinstitutions. There are two aspects of CONTIGO of interest here. The first is as conceptualframework for processes of poverty reduction and social development, and for the role of public action with this. The second concerns program management, which seeks to apply the framework in the complex organizational structure of the mexican government at central and subnational levels. In this chapter we focus on CONTIGO as a conceptual framework; we discuss government strategy inChapters4 and5. CONTIGO has two stated objectives: (i) provide basic social benefits to all mexicans, to including quality health care and education, adequate nutrition, housing, employment security and pensions; and (ii) to activate the levers that stimulate human development, that i s to provide opportunities for the productive and social integration of individuals through basic education, health, employment and training (Secretaria Tkcnica del Gubinete Social, 2003)13. Areas of potential public action are then organized into four areas (see Figure 1.3): a Human Development: ensuring the physical and intellectual development of the population through adequatenutritionand access to health services, education and training. a Income Generation: promoting employment and local development as well as providing credit andtechnical assistancefor productive projects. a Asset uccu~~lution: support for housing, savings and land titling programs that help households secure a base of physical assets, a Social Protection: reducinghousehold vulnerability to shocks andother risks such as low income among the elderly. These areas are complementary, reinforcing each other to create a virtuous circle for poverty reduction. Healthier and more educated individuals are better equipped to take advantage of income generating opportunities which in turn would facilitate asset accumulation. These, together with adequate social protectionpolicies can promote long- term investments that could break the intergenerational transmission of poverty. A particularly valuable element of the CONTIGO framework is the integration of these areas of public action with the life cycle of individuals and households. The target population ranges from pregnant women to the elderly. This potentially allows programs to bejoined around age groups and action areas so as to ensure that individuals needs of l3There is a slight variation in the stated objectives across documents. The CONTIGO website, for example, states as objectives: (i) to eradicatepoverty and (ii) the full developmentof the population ensure (see htt~:/fwww.conti~o.~ob.~~. 16 government services and assistance are provided at different times in their lives. Figure 1.4 illustrates. CONTIGO's framework also includes five principles that are intended to cut across all areas of public action14:- Equity to allocate more resourcesto disadvantagedgroups Transparencyto avoidthe discretional or politicaluse of resources Joint re~ponsibili~ betweenthe three levels of government and society Social cohesion to strengthen the social fabric and foster community development, governance anddemocracy Integrality to encourage efficiency in social spending by exploiting synergies, eliminatingduplication andensuring coherenceindifferent program objectives. 14Insomedocuments[environmental] sustainability is includedas asixthprinciple(see Secretaria TLcnica del Gabinete Social 2003). 17 Figure 1.3. CONTIGO's conceptual organizationof areas of public action AREAS SUB-ANAS 1. Education A. I.HumanDevelopment 2, Health and Nutrition 3. Training 6. Housing 7. Saving 8. PropertyRights 9. Insurance IV. SocialProtection 10. SocialSecurity 111.Protectionagainst Collective Risks CONTIGO i s an excellent overall framework for poverty reduction and social development. Intheory, it recognizes the ~ulti-dimension~ityof well-being andthe need for multiple categories of public action to raise living standards. The four core areas of public action Figure 1.3) map on to four of the dimensions of well-being discussedinthe preceding section. It also emphasizes the powerful complementarities between areas of public action. Embedding public action in the life-cycle is both useful for designing policies inrelation to different categories of needs, andpotentiallyfor the more ambitious target of shaping strategies that take account of interactions across different parts of a person's life-cycle and between generations. The emphasis on issues of transparency, joint responsibility and social cohesion i s consistent with the role of these factors in the quality and effectiveness of service provision for the poor (as well as in other areas of development process) discussed inthe preceding section. 18 Figure 1.4 CONTIGO:Interactions BetweenPublicAction Areas andthe LifeCj - Source: CONTIGO'Sweb page andpower point presentation, both available at: httD://www.conti~o.~ob.~ The CONTIGO framework i s probably one of most properly articulated in an international context. Perhaps ironically, it is in the low-income countries that there has been most extensive action in the past few years in articulating explicit frameworks for poverty reduction. This occurred because of the rising importance of nationally prepared poverty reduction strategies to underpin debt reduction and aid, especially from the international financial institutions. These frameworks typically also have a multi- dimensional concept of well-being, systematic categorization of areas for public action, participative engagement and aspire to place such action within a systematic results- based framework, with ongoing evaluation (though the last i s usually a long-term enterprise). In Latin America, Bolivia has a relatively thorough strategy, but implementation has been ill affected by social conflict. Even well crafted poverty reduction strategies cannot, at least in the short run, deal with profound social divisions andpoliticaldifficulties, especially inthe context of adverseeconomic shocks. Vietnam's strategy i s also of interest, since this i s not only providing a framework for national development strategy and underlyinganalysis, but the process i s now being replicated at provincial levels of the country. At the other endof the income scale, many richcountries havebeen puttingmore explicit focus on strategies for poverty reduction, and conceiving this as part of the broader project of combating social exclusion. Member states of the European Union, for example, developed national plans for social inclusion, in response to an EU-wide agreement to give this priority; this has been associated with work on assessinghow to interpret and analyze progress (Atkinson et a1 2000). In the UK this is one part of a 19 broader approach that is seeking to tackle historical problems around employment and poor service delivery, which is thoroughly integrated within a results-basedapproach. CONTIGO provides a sound general framework. Any specific area of public action will need a framework that is more detailed, taking account of the technical, economic and social issues for that area, whether it i s agricultural production or determinants of health. A variety of more specific frameworks will be referred to inthis work and inthe more in- depth follow up -as i s the practice within the government. At the level of a general framework, the one area that we will sometimes depart from in organizing thinking is in makingsocial inclusion a fifthareaof public action (inaddition to its presenceas across- cutting prin~iple).'~ This is because this has a direct relation with dimensions of social living that are constitutive of well-being (and poverty reduction) and because there are lines of public action that are specifically aimed at fostering social inclusion, for example inthe work of INDESOL. As an example of putting the considerations of this chapter together and taking the CONTIGO (and related) frameworks to one more level of structure, De Janvry and Sadoulet present a structure for linkingoutcomes and action for the poverty reduction and social development inthe ruralcontext, This is illustratedinFigure 1.5. D. Summary Poverty i s a major social concern inMexico. Poverty has many dimensions, including ill health and lack of basic skills, lack of access to basic services, insufficient income for a decent life, vulnerability and social exclusion and indignity. Recent thinking on the processes that determine the dynamics of poverty emphasize the complex interaction between economic assets and opportunities, and their embeddedness in, typically unequal, formal and informal institutional structures. This applies in particular to the central issue of service delivery to both the poor and the non-poor. There is also recognition of the importance of volatility, and associated insecurities, as an area for public action. Within this context, the government's CONTIGO approach is a good, comprehensive framework for organizing approachesto public action (but not for ranking them)I6. More emphasis could be given to determinants of service delivery and to social inclusion as a line of public action -but the framework provides a sound basis for both the design of overall strategy and the more specific approachesthat need to be developed in individual sectors. It is worth highlighting the valuable initiative to developing an independent committee charged with providing technical advice on official poverty concepts, including on the income poverty line. Finally, it is also laudable the approval of the Social Development Law (2003) which is an important step to institutionalize a state policy for social development. The law institutionalizes social policy inMexico through a series of bodies to oversee social development policy and implementation, including a board for evaluation, a national commission on social development, an inter-ministerial commission and an advisory committee, that will include members from civil society. It l5SeealsothediscussioninDeJanvryand Sadoulet(2003b) l6Paes de Barros(2003). 20 includes a requirement that social development funds not be reduced inreal terms in any year. 21 June 2004 Poverty inMexico An Assessment of Conditions, Trends and Government Strategy Chapter 2. Well-being inMexico inInternational Perspective Inthis chapter we presentanoverview of well-being inMexicoto provide the context for the more detailed analysis of posterior chapters. We first sketch medium to long-term trends in relation to other countries. Then we examine the extent to which Mexico's current position (around 2000) appears to be relatively good or bad given its income level and other structural conditions. The chapter uses indicators for three of the dimensions of well-being (and poverty) introduced inChapter 1: human capacities, assetslinfrastructure and incomes. Cross-country information on vulnerability and social incorporation are not available. However, attention i s paidto measures of governance, both as a correlate (and possible cause) of other dimensions of well-being, and as a roughproxy for the extent to which governmental structures reflect societal interests. To provide the international context, Mexico is, in some cases, placed in relation to the experience of all countries with data. Inothers Mexico i s compared with two regions of particular interest to debates on development inthe region: the rest of Latin America and with East Asia. Within the latter, Malaysia i s singled out to give a more concrete international comparator of an East Asia 44success"case. This country is probably the one in East Asia that is most like Mexico: it is now upper middle income, is relatively unequal, has been passing through a long transition from being natural resource-basedto primary reliance on industrialexports, it has significant socio-cultural divisions across ethnic groups, and has had a semi-authoritarian government for much of its recent history. There are also important differences of structure and history, and the comparison should not be over-stretched. Since this analysis uses information up to around 2000 or so, and covers slow-changing variables (such as child mortality), it can be treated as a snapshot of the current government's inheritance from the accumulated experience of previous decades. It might be considered a schematic baseline for the efforts of the present administration. A. Patterns of progressinwell-being ininternationalperspective Over the past few decades, Mexico has made substantial progress on many areas of well- being in terms of average indicators for the whole society. Selected variables are shown to illustrate this inFigure 2.1. In health services there has been a dramatic long-term improvement in both child mortality and life expectance, with no apparent breaks in the trends across decades. The number of children dyingbefore the age of five per 1000live births fell from over 140in the 1960's to less than 40 in the 1990's, while life expectance rose from 58 to over 72*. These large changes were broadly comparable with the Latin American and East Asian averages. For education, there i s a more complex pattern. Years of education amongst adults, that reflect the acumulative effects of earlier decades of education -grew rapidly from a relatively low level inthe 1960's, overtaking the Latin American average, anddisplaying some catch-up with the East Asian average. However, secondary enrollments first show a similar pattern of rapid gains between the 1970's and 1980's, significantly above the Latin American average, but then experienced slower absolute and relative gains in the 1990's. Average incomes experiences rapid growth until the 1980's, but slower long-run growth since, and somewhat below the Latin American average. The whole of the region experienced a severe recession in the 1980's, but Mexico's growth in the 1990's was worse than average-with the 1994-95crisis an important contributor to this. The major contrast with Malaysia is the sustained growth in incomes (despite, like Mexico, facing macroeconomic shocks in the 1980's and 1990's). Malaysia overtook Mexico in the 1990's on this dimension. There is less difference in progress on most human investment indicators, though Malaysia did not experience the slowdown of secondary enrollments inthe 1990'srelative to the 1980's. What do these patterns mean for the trends in deprivation? For any dimension of well- being, the pattern of deprivation i s a function of the average level of the indicator and its distribution across the population. Mexico, along with most of Latin America, is highly unequal with respectto income, and is, to varying degrees, unequal with respectto many other dimensions of inequality. For income, according to the most commonly used synthetic index of inequality, the Gini coefficient, Mexico was in 2000 more unequal than the (high) Latin American average, though less unequal than Brazil, Chile and Colombia (Table 2.1). In terms of the shares of different parts of the distribution, the bottom quintile of households (ranked by per capita household income) accounted for 3.1% of total income while the top decile accounted for 43.%. The ratio between the income of the top and bottom decile was 45 times.' An important part of the reason for highlevels of poverty inMexico is the highlevel of inequality. On average East Asian countries are much more equal, and so have lower levels of poverty for their mean income level. Malaysia has levels of income inequality significantly above the East Asian average, and, as shown in Table 2.1, i s only somewhat more equal than Mexico, and less equal than CostaRica andUruguay. 'This chapter uses the World Development Indicators for international comparisons. However, in 2001, national data from CONAPO reports that the child mortality rate was 27.4 per 1,000 born alive and life expectance was 74.3 years. All figures are from DeFerranti et a1(2004), which usedcomparablemethodsacross countries. 23 Y i \ \ f 0 asnMI Table 2.1. Income i :quality inMexico ininternationalperspective Cini coejjicient Share of top 10 Shareof bottom Ratio of percent in total 20 percent in incomes 0j1o"'~ income total income to I" decile Brazil (2001) 59.0 47.2% 2.6% 54.4 Guatemala(2000) 58.3 46.8% 2.4% 63.3 Colombia (1999) 57.6 46.5% 2.7% 57.8 Chile (2000) 57.1 47.0% 3.4% 40.6 Mexico (2000) 54.6 43.1% 3.1% 45.0 Argentina (2000) 52.2 38.9% 3.1% 39.1 Jamaica (1999) 52.0 40.1% 3.4% 36.5 Dominican Rep. (1997) 49.7 38.6% 4.0% 28.4 CostaRica (2000) 46.5 34.8% 4.2% 25.1 Uruguay (2000) 44.6 33.5% 4.8% 18.9 Malaysia (1997) 49.2 38.4% 4.4% 22.6 United States(1997) 40.8 30.5% 5.2% 16.9 Italy (1998) 36.0 27.4% 6.0% 14.4 Source: Authors' calculs Ins: for Latin America estimates from micro-data, see De Ferranti et al (2004), for othersWorld DevelopmentIndicators. Mexico has probably been highly unequal for a very long time, along with most of Latin America. Survey-based evidence for income differences i s only available since the early 1970s. Trends over time indicate that the Mexican income distribution has moved in different directions in the last three decades. There was an improvement in income distributioninthe ~ O ' Sa, substantial rise ininequality inthe 1980'~~ and relative stability in the 1990's, despite the major structural changes and economic shocks of the past decade. Together with the pattern of average growth, this implies relative rapid income poverty gains in the 1970'~~ a large reversal in the 1980'~~ little overall gain in the and 1990's, but with major fluctuations due to the Tequila crisis3. The 1994-95 crisis was actually associatedwith a small improvement ininequality (that was partially reversedby 20004), incontrast to the crisis experience of the 1980'~~the substantial worsening of or inequality in Argentina during its recent crisis. Finally, estimates for 2002 indicate a further, slightly larger improvement in distribution since 2000, which was an important source of estimated declines in poverty. We review the recent trends in more detail in Chapter 3. There are significant overlaps between incomes and other dimensions of well-being. The income poor tend also to suffer higher child mortality, have lower years of schooling and school enrollments of their children and lower access to services. Nevertheless,respectto service access, expansions typically have a specific distributional dynamic, providing for the rich and middle classes first, and then working down the distribution of income or wealth, reaching the poorer groups last. This implies that major service expansions start See de Ferranti et a1(2004), Morley (2001), Sztkely (2001). Differences between 1996and 2000 are not statistically significant. 25 off anti-poor and become more and more pro-poor the closer a country i s to universal pro~isioning.~ Detailed information on the structure and dynamic of educational inequality is available for Mexico in relation to other Latin American countries from recent analysis of comparable household surveys. For Mexico the product of the historical dynamics was a high level of inequality in adult educational attainment across income groups, In 2000 adults inhouseholds in the top quintilehad almost eight years more education than those in the bottom quintile. This was the largest difference of the set of countries with comparable data in Latin America, and had actually risen by half a year since 1992 Figure 2.2a). At the bottom of the distribution one dimension of this i s the level of illiteracy. Whereas self-reported illiteracy rates are less than five per cent in the top two quintiles, they were 30% inthe bottomquintile in2000, down from 34% in 1992 (for the second quintile illiteracy was declined from 17% in 1992 to 13% in2000). The evidence on educational dynamics is mixed. On one hand there was a modest reduction in the educational gap between the top and bottom quintiles of people in their 50's and 30's (Figure 2.2b). There was also a small decline inthe differences inenrollment from 13-17 year-olds between poor and rich households in the years spanning from 1992 to 2000. But there was a rise indifferences inenrollments for 18-23 year-olds (Table 2.2). Since the major divide in Mexico in terms of returns to education is now between those with and without tertiary education, risinginequalities at this level across income groups could be a source of further disequalizing pressuresinthe coming years. Figure2.2a. Difference inaverage years of education betweentop and bottom quintiles 0 - n A r o u n d 1990 m A r o u n d 2000 Source: De Ferranti et a1(2003) basedon micro data fromhouseholdsurveys. See Lanjouw and Ravallion for discussion of the issue in an Indian context, and de Ferranti et al (2004) for a summary of LatinAmerican experiences inthe 1990s. 26 Figure 2.2b. Difference inaverage years of education betweentop and bottom quintiles. By age cohort, around year 2000 4 0 (51-60) 0 (31-40) Source: DeFerrantiet al(2003) basedon micro datafrom householdsurveys. Table 2.2. Enrollment rates of young people inpoor and rich Mexicanhouseholds, 1992and2000 (innercent) I 13-17year-olds 18-23year-olds Bottomquintile Top quintile Bottomquintile Top quintile 1992 48 84 12 40 2000 57 90 16 52 Source: ENIGH1992and 2000. Educational quality matters as well as quantity. A recent international testing project (the PISA, Program for International Student Assessment) finds that Mexico educational quality falls substantially below that for OECD countries, and is also lower than that for Korea and Thailand. Education quality i s comparable to that of Argentina and somewhat better than in Indonesia and Peru, but that may be small comfort. As with all countries there i s significant variation in quality -that i s correlated with the socio-economic status of the households from which children come. Moreover, this increases differentials in quality-adjusted education across income groups. However, the variability of Mexico's quality is low by international standards, and much lower than Argentina's. Whether this i s due to government efforts to raise quality amongst the poorest, or the relatively consistent (and low) teaching standards that flow from countrywide features of educational supply, includingthe teachers' union, i s an issue for future work. 27 Figure 2.3 Mexico has low average quality, butlow dispersion,inthe quality of education (Test results for mathematicsfrom the PISA international study.) 600 HLD JPN HXO KOR FIN CA" NZL AUSGBR CiiE IS1 DNK FRA ^USW. LIE BEL IRL YO 500 DEU ESP RUS POL LUX ORC S THA 8UL ISR ROU C 0 400. a CUXO ALE ARG R IDH E BRA 300 * PER 200 !O DISPERSION Source:Authors calculationsfrom work oneducationalquality, basedon the Programfor International StudentAssessment (PISA). On a brighter note from a distributional perspective, Mexico has made substantial progress toward equalization of educational attainment between genders with respect to education, in common with many Latin American countries. This can be seen by the virtual elimination of gender differences that prevailed between 51-60 year olds compared with 21-30 and 10-20 year-olds, alongside the generalized progress in the 1990s.This is apatternsharedwith most LatinAmerican countries. Table 2.3. Years of education of males and females inMexicoby age group 10-20year-olds 21-30 year-olds 51-60year-olds Female Male Female Male Female Male 1992 6.5 6.3 7.6 8.3 3.9 4.6 2000 7.0 6.9 9.0 9.5 4.7 6.2 Source: ENIGH 1992and 2000. Lack of data precludes a more comprehensive review of long-run trends in the different dimensions of well-being in international perspective, especially for those dimensions such as vulnerability and social exclusion that do not have established comparable measures. However, the overall pattern i s clear. First, there has been major progress on average for health, education and some services, but halting progress on incomes. Second, well-being has been unequally distributed for all dimensions with information, with mixed patterns in the past decade or so -with poorer groups gaining relatively for 28 some services, losing ground (relatively) on tertiary education, and with a lack of clear trend on income differences, but with encouraging recent signs. B. An overview of the relationship betweenwell-being andindicatorsof country conditions Inthis section we examine the international patternof well-being, andMexico's position intermsof the relationship with other measuresof country conditions or performance.In other words, this turns to look at Mexico's performance conditional on its characteristics. The main conditioningvariable i s the mean level of national income, on the grounds that this is a measureof the level of aggregateresources a society has to support public action to reduce deprivation. For some variables we also use a multivariate analysis that takes a step toward a more complex analysis that allows for interactions between variables. The results should come with an important caveat over causality: while all are interesting on the associations, and many are suggestive of possible causal relations, in none of the results do we have the statistical basis to infer causality. This caution is intrinsic to such cross-sectional analysis. More in-depth, within-country analysis i s necessary to develop perspectiveson causal mechanisms, as i s planned for the work on Mexico. We first show the bivariate relationships between income and several measuresof human development and assets, using most recent data. These are shown in Figures 2.4a-g, which also presents a regression line of the relationship with mean incomes, and the values for Latin America, East Asia and Malaysia.6 For all variables there is a similar overallpattern of observations, that shows two results from this type of comparison: there i s indeed a positive association between average incomes and such measures of well- being; but there i s also wide variation for any given level of income. The latter illustrates how differences in structures, formal and informal institutions and policy choices may make large differences to health, education or service access for societies with comparable levels of aggregate resources. Mexico's position in relation to other countries i s generally close to, or somewhat worse then, what would be "expected" from this simple conditional relationship. For health, life expectance i s slightly better than the estimated relationships and child mortality slightly worse. For education, illiteracy i s slightly better (reflecting the historical expansion of basic education) but secondary enrollments are worse. Water coverage, and especially improved sanitation, i s below the estimated average relationship. However, it i s noteworthy that in all cases there are a large number of countries at either similar or lower income levels that perform better than Mexico: this illustrates the potential for better outcomes. This uses national income per capita adjusted for price differences across countries using information from the Purchasing Power Parity project. The regression line reported is for OLS with quadratic specification on log incomes: the actual specification is arbitrary but it fits the pattern of data reasonably well. We follow Todd and Hicks (2003). 29 \ a i * c( a cw B s 3 0 . 3 cl ** 0 8cl I * 1 . 0 .. 4 These figures show the cross-section relationship for the latest period. How has this relationship evolved over time? Figures 2.5a and b explore this. These figures pool cross- section and time series country numbers to estimate the relationship between two indicators of human development, infant mortality and gross secondary enrollments, and incomes. As for the above graphs, they use a quadratic expression for incomes to better fit the relationship in the data. This again shows that, on average, both human development indicators improve with higher income, but at a declining rate. There is of course a lot of variation around this average relationship -as has been extensively discussed in the literature. Some countries have achieved large gains in human development at low incomes, while others have had rapid income growth with relatively modest advances in human development (see Drkze and Sen, 1989, for a seminal discussion). In such a long-run perspective, Mexico experienced major catch-up from a position of large social deficits relative to its income level (the fact that this relationship shows Mexico being better than average, whereas Figures 2.4 shows modest deficits appears to be because the relationship has shifted over time: Figure 2.4 shows the contemporary relationship. Figure 2.5 the pooled relationship for the past 30-40 years). Malaysia also experienced major advances relative to its incomes. Again, the major difference is that Malaysiaovertook Mexico interms of income inthis period. Figure 2.5. The long-run relationships between incomes, infant mortality and secondaryenrollments (a) Infant Mortality 200 IM=444.04-78.39*ln(GDP)+3.49*ln(GDP)2 RZ=0.6511 N=3,388 0 0 2,000 4,000 6,000 8,000 10,000 12,000 GDP per capita (1995 USD) 32 (b) Gross Secondary Enrollment 100 ala 2o ysia 1970 GSE= -94.6-24.09*ln(GDP)-0.49*ln(GDP)* RZ=0.6071 Mexico 1970 N=2,008 0 ' 2,000 4,000 6,000 8,000 10,000 12 IO0 -20 GDP per capita (1995 USD) Source: WB staff calculationsbasedonWorld Development Indicators. We next look at indicators of governance. This i s a proxy for the institutional basis for capabilities in a society, both in terms of the formation of human capacities, and the opportunities to convert these into income and functionings. The measuresof governance come from survey results of subjective assessments various dimensions of governance covering 199 countries and territories. These indicators are based on several hundred individual variables measuring perceptions of governance, drawn from 25 separate data sourcesconstructed by 18 different organizations, developed by the World Bank Institute. These individual measures of governance are assigned to categories capturing key dimensions of governance and use an unobserved components model to construct six aggregate governance indicators. They are generally quite highly correlated with other country-level measures of governance, whether formed by experts (as in the Polity IV data base) or the views of a variety of intemationalinvestors (as inthe ICRGdata base). There are two reasons for showing this data. First, some of the measures, notably on voice and accountability, government effectiveness, the rule of law, regulatory capability and corruption, will be proxies, albeit imperfect, for important direct measures of well- being, including with respect to social povertylexclusion. As noted in Chapter 1, poor people across the world place great emphasis on poor government performance and adverse experiences with justice systems as dimensions directly affecting their well- being. Work by social scientists on the rule of law in Latin America also strongly underlines the positionthat a weak rule of law can exacerbateinequalities inliving, with poorer groups suffering greatest risk of abuse (Mendez et al., 1999). Second, issues of governance and accountability i s now recognized to be a central determinant of the effectiveness of service delivery, especially for the poor, as also discussed in Chapter 1 (World Bank, 2003b). Some illustrations of the associations from cross-country work are shown below. 33 Relative to both its income level and the LatinAmerican average, Mexico has low ratings for voice and accountability, the rule of law and corruption, but has average or slightly better ratings for its income (above the Latin American average) for government effectiveness and regulatory capabilities. Selected graphs of these results are shown in Figure 2.6. Malaysia's ratings are even worse than Mexico's on voice and accountability, but better on government effectiveness and rule of law. While we believe the international evidence supports the view that voice and accountability are important for better services for the poor and non-poor, this simple comparison provides a useful warning on a too simplistic perspective on these issues. Malaysia is a good example of an authoritarian developmental state, also typified by Korea and Taiwan, China in earlier stages of their development. For reasons that lie in social and political history, effective development states have been rare in Latin America. And with the (highly desirable) transition to democracy, this aspect of East Asian experienceis not relevant, even if other aspects are. The description of patterns has so far looked at trends and outcomes conditional on one country-level variable -mean incomes. We finally take a modest step toward a richer account of cross-country patterns by showing some results of multivariate regressions for health and education outcomes. These seek to calculate the partial associations between particular variables and the outcome, after controlling for the influence of other variables. The reader should be reminded that the patterns of influence in the real world are complex, and flow in many directions -coefficients and levels of significance on variables may be picking up such complex effects. The results are nevertheless suggestive. For health we explored the impact of a variety of country level characteristics on child and infant mortality, following a formulation in Filmer and Pritchett (1999), including average income, health spending, female education, measures of social difference (income inequality and ethno-linguistic fractionalization), provision of water, urbanization, dummy for Muslimcountry and dummy for tropical country. Inaddition we explored the impact of a variety of governance variables, including those discussed above. For education, a similar approach was followed, using measures of years of education and secondary enrollment. Table 2.4 presentsone illustrativeresult each for the child mortality and secondaryeducation, which are representativeof the results. 34 S t BL a .\. . *..*2, t 0 9 v 8d Several patterns of association, or its lack, are notable in the results on child mortality. Higher incomes, greater voice and accountability and greater urbanization have highly significant negative associations with child mortality. These are robust across choice of different measures for governance. By contrast, health spending is never significant nor is water access or distance to the equator. Female education is sometimes negatively associated with mortality, but not robustly so. The Gini coefficient is never significant, while higher ethno-linguistic fractionalization is associatedwith worse mortality in some formulations. Dummies for region are significant, indicating that there are patterns of regional conditions not caught by the other variables. These should be interpreted as effects relative to Africa, which is the excluded regional variable (and also has the worse outcomes). Thus the Latin America effect, while positive for health outcomes relative to Africa, i s worse than East Asia, Europe and CentralAsia and the MiddleEast. How should this be interpreted? This provides further support for the view that income interact with healthoutcomes, and also that there is stuff going on around the interactions with institutions, social arrangements or government effectiveness, captured in the governance variables. The positive influence of urbanization is likely to reflect both the greater density of service provision, and the large social and environmental changes associated with urbanliving. The absence of significance of health spendinginthe cross- section could be for a variety of reasons (Filmer and Pritchett, 1999). We highlight two. First, the factors that influence health are complex, including social behavior, environmental conditions, nutritionalstatus etc. Health spending i s only one of these. In micro studies within countries female education and water and sanitation are typically significant influences, but the patterns of international variation do not capture this. Second, even within those channels of influence potentially influenced by health spending, there are many "links in the chain" between government spending and actual effects, including the distribution of spending across groups, interaction effects with private spending on health, and major variability in the actual front-line interaction between health provider and citizen (Filmer and Pritchett, 1999). The fact that overall institutional arrangements seem to be strongly associated with health status, but health spending is not, provides at least suggestive support for the emphasis on the complexity of causes of outcomes and a central role for formal and informal institutions in the frameworks presentedinChapter 1. Mexico's position within this analysis i s illustrated by its residual, which is presented along with those of a few other countries of interest. This shows that Mexico's level of child mortality was somewhat high, after controllingfor all the other variables (including the "Latin America" effect). The same applies to Argentina and Brazil, while Chile, China, Korea, and Malaysia all have negative residuals, indicating "good" performance controlling for other factors. For education, the results were less robust, possibly due to a much more restricted data set for public spending. Incontrast to the health patterns, neither incomes nor indicators of governance have a significant association with secondary enrollments (nor with years of education of the adult population). By contrast, public spending on education is generally significant, as are regional dummies. A sample of one equation i s given in 36 Table 2.4. It is perhapsnoteworthy that now Mexico has a negative residual -indicating low levels of enrollment after controlling for all the other variables. While we should be careful not to over-interpret this result, it i s at least suggestive of poor performance at this level (in contrast with the almost universal enrollments at primary level). This will be examined inmore detail infuture chapters. Table 2.4. The relationshipsbetween mortality, educationand economic and social variables Resultsof an initial cross-country analysis log of child log of child net secondary DependentVariable mortality mortality enrollment Constraints on the Voiceand Voiceand GovernanceIndex Executive Accountubility A c c o u n t ~ ~ i l i ~ logof gdp per capita, PPP -0. 51* -0. 47" 7. 11 logof healthexpendituresas share of gdp -0. 03 -0. 03 logof educationexpendituresas share of gdp 12. 39* FemaleEducation (ages 25 and over) -0. 04* -0. 02 1.18 GiniIndex 0 0 -0. 08 MuslimCountry 0. 30* 0. 33" 14. 15 Ethno linguistic FractionalizationIndex 0. 33" 0. 36" 6. 08 Share of UrbanPopulation -0. 01* -0. 01" 0. 27 TropicalCountry 0. 05 0. 01 -10. 98 Share ofPopulation with Water Access 0. 002* GovernanceIndex -0. 06* -0. 21* 1.22 NorthAmerica -0. 55* -0. 59" 13. 37 WesternEurope -0. 79* -0. 82* 18. 45 SouthAsia -0. 57" -0. 62* 30. 19" EastAsia & Pacific -0. 78* -0. 79* 22. 83* EuropeandCentral Asia -0. 66* -0. 78" 20. 65* MiddleEast& North Africa -0. 64* -0. 76" 8. 43* LatinAmerica & Caribbean -0. 29* -0. 39* 19. 61* Constant 8. 98" 8. 30* -47. 63 R2 0. 9166 0. 913 0. 8997 Observations 137 154 43 Residuals: Actual minusfitted Argentina 0. 33 0. 31 3. 41 ' Brazil 0. 36 0. 44 3. 29 Chile -0. 4 -0. 4 2. 28 China 0. 02 -0. 06 Korea -0. 38 -0. 41 -1.35 Malaysia -0. 55 -0. 62 -2. 84 Mexico 0. 21 0. 18 -10. 09 Note: * denotes significant at the ten percent level. All results are OLS; further work will explore robustness and instrumenting for probably endogenousindependentvariables. 37 C. Conclusion Mexico has experienced considerable progress in many measurable dimensions of well- being over the past few decades. However, the progress i s mixed across indicators: large and consistent in health, large, but with signs of a deceleration, in education, and more faltering with respect to incomes, in the wake of the economic crises of the 1980's and mid-1990's. These patterns are likely to have been broadly reflected in patterns of deprivation for poorer groups, with large gains in some areas, but a decidedly bumpy ride with respect to income poverty. High levels of inequality remain a major adverse influence on deprivation, as in much of Latin America, and were notedto be salient with respect to both income and education. Despite the substantial progress on human indicators, Mexico's position in an international context was at best average, conditional on its income level -reflecting the adverse historical legacy of poor human conditions that the country has been steadily tackling. Finally there is evidence of the importance of factors associated with governance, that are probably correlated with the depth of social incorporation, and appear to be strongly associated with indicators of well-being. This further supports the attention to issues of institutions as a determinant of the quality of life for poor andnon-poor groups. 38 June 2004 Poverty inMexico An Assessment of Conditions, Trends andGovernment Strategy Chapter 3. The Structure and Trends inDeprivation Chapter 2 reviewed Mexico's progress as a nation in increasing its indicators of well- being incomparison to the rest of LatinAmerica andEastAsia. This Chapterexamines in more detail the current structure of poverty in Mexico, and recent trends. As with other parts of the report, the main organizing principle i s along the different dimensions of poverty or well-being. Section 1looks at dimensions of poverty associated with human investment and assets, analyzing trends in education, health, and basic infrastructure. Section 2 examines income poverty and discusses the structure, correlates and trends in poverty from labor and non-labor income usingofficial poverty lines. Section 3 presents some work on the nature of vulnerability in Mexico. Section 4 then examines Mexico's regional and ethnic differences in indicators of well-being across various dimensions, makinguseof disaggregatedspatial mapsof livingconditions. A. Trends ineducation, health, andbasic infrastructureservices Overview With respect to human investment -in education and health- and access to physical assets, there has been significant progress inMexico, both over the past decade andinthe period since 2000. It i s particularly noteworthy that the progress occurred despite the major adverse effects on the economy of the 1994-95 crisis, which constituted a massive setback to income poverty, as will be seeninnext section. Despite this progressinhuman andphysicalassets, major problems remain. These revolve inparticulararoundquestions of inequality and quality. Both education and health status, and access to social services, remain highly unequal. However, there has been a substantial improvement in the participation of the poorest income groups in lower-secondary education and health services for the uninsured, particularly inthe 1998-2002 period. Interms of quality, there are still major problems that apply with particular force to poorer groups in the society, exacerbating differences in access. These are most clearly documented for education, but also appear to apply to health. Comparable considerations of unequal access, and low and variable quality apply to access to physical assets, including housing, electricity, water, and sanitation. Problems of low and unequalquality of services appear to have as much to do with institutional functioningandpatterns of accountability as with resources. Education Education is central to incomes, citizenship, and the capacity to lead a fruitful and satisfying life. In 2000, mexicans between the ages of 25 and 65 had, on average, over 7 years of education. This average hides major variations. Younger cohorts were much better educated than older ones, with 21-30 year-olds having an average over 9 years of 39 education compared with just over 3 years for the over 60. There is also a steep gradient with respect to income, with 25-65 year-olds living inthe top 10% of households having 12 years of education, compared with 3.2 for those in the bottom 10% (Table 3.1). This gap has not been narrowing over the past decade or so. Adults in indigenous households only hadjust over 2 years of education. While differences ineducation betweenmen and women were historically significant, they have virtually disappeared for younger cohorts.' Parker and Pederzini (2002) find that "the overall gender gap in education in Mexico has diminishedto the extent that there are no apparent gender differences inyears of completed schooling for individuals under the age of 25" (p.24), although modest gaps remain inthe rural sector. Table 3.1. Steady progressbutlargegaps inschooling of Mexicanadults Years of education of 25-65 year-olds, by income decile Decile 1984 1992 2000 2002 1 2.1 2.2 2.8 3.2 2 2.3 3.0 3.7 4.2 3 3.0 3.6 5.O 5.0 4 3.0 4.2 5.6 5.6 5 3.6 4.8 6.2 6.5 6 4.1 5.3 6.9 6.8 7 5.0 6.0 7.4 7.4 8 6.2 6.8 8.3 8.2 9 7.2 8.2 9.5 9.8 10 8.6 10.8 12.1 12.0 Source: WB staff calculations from ENIGH(1984, 1992,2000, 2002). The tremendous expansion of the Mexican education system duringthe past decades can be seen from the growth of gross enrollments, which increased from 1.4 million to 31.5 million students from 1930 to 2003. This has been associated with a large increase in public effort inthe 1990's, with education expenditure rising from 4.9% of GDP in 1980 to 6.8% in 2003. The education system i s now so extensive that there are over 229 thousand schools (includingpreschools) staffed with about 1.5 million teachers. Mexico has achieved close to universal primary education and a rapidexpansion of lower secondary education. In 2003, there were 14.8 million students attending primary education, up from 14.6 million in 1980: slow growth in numbers i s because higher enrollment rates were offset by falling numbers of school-age children, reflecting Mexico's demographic transition. 92% of children are enrolled inpublic primary schools, with 5% of these in indigenous, bilingual schools and one percent in "community" primary schools, which are multi-grade schools in remote areas. Growth in gross enrollment has been particularly strong in indigenous schools, that increased from 589,000 to 854,000 students, and incomunity schools, that rose from 82,000 to 145,000 students. These patterns are also borne out by the 2000 Census that found a 93.5% primarynet enrollment rate for 6-11year-olds, upfrom around 88% in 1990. The number All numberscome from the 2000 ENIGH, except for that for indigenous years of education, that are from the 2000 Census. 40 of schools and teachers has increased faster than the number of students. While this is (potentially) good for quality improvements, it has also yielded higher costs. Rapid progress has been achieved in lower secondary schools. In 2003, 5.6 million students were enrolled in lower secondary, 92% of these in public lower-secondary education, compared to the 3 million in 1980. Progress in the last 13 years is illustrated by a rise of 38% in enrollment, a 42% increase in teachers, and a 60% increase in the number of schools. There have also been large increases in enrollments in upper secondary and tertiary education, albeit from a much lower base. From 1990 to 2003, upper secondary enrollment increased by 70% and tertiary enrollment by 79%. In upper secondary, there are 68% more teachers and 91% more schools. Intertiary education, there are 84% more teachers, andeducational establishments more than doubled. Table 3.2. There has been major progressineducational indicators, butgaps and problems remain, especially at higher levels Dropout Failure Terminal Gross Enrollment Efficiencv (Thousands) 1990-1991 Primary 4.6 10.1 70.1 14,401.6 Lower Secondary 8.8 26.5 73.9 4,190.2 TechnicalUpper Secondary 25.1 28.3 37.8 378.9 UpperSecondary 17.4 47.6 60.1 1,721.6 Tertiary 10.0 1,252.0 1995-1996 Primary 3.1 7.8 80.0 14,623.4 Lower Secondary 8.8 23.7 75.8 4,687.3 TechnicalUpper Secondary 26.7 29.5 45.5 388.0 Upper Secondary 17.0 44.5 58.1 2,050.7 Tertiary 7.1 1,532.8 2002-2003 Primary 1.5 5.4 88.0 14,857.2 Lower Secondary 6.9 18.9 78.8 5,660.1 TechnicalUpper Secondary 23.0 22.8 50.5 359.2 UpperSecondary 15.1 39.2 61.6 2,936.1 Tertiary 7.9 2,236.8 Source: Anexo del Tercer Informe de Gobierno2003. Growth in primary total enrollments has been accompanied by progress in terminal efficiency (the ratio of the number of children completing sixth grade to new enrollments infirst grade 6 years earlier) and repetitionrates, with a 20 percentage-point gain inthe last 13 years, from 70% in 1990 to 90% in 2003, due to reduction inboth repetition and dropout rates. However, terminal efficiency rates in lower secondary have not improved beyond 80% and the average estimated dropout rate is 6.9% in 2002-03 (Table 3.2). 41 Upper secondary terminal efficiency, dropout and failure rates2 have not significantly improved, especially for technical schooling. This later result i s particularly important since technicalltechnological schooling comprises 40% of the total upper secondary enrollment. Which groups have benefited from the expansion in enrollments? Table 3.3 shows net enrollment ratesby poverty status, educational level, andgeographic area. There are quite modest gradients between the extreme poor, moderate poor and non-poor for primary education, reflecting the high overall enrollment rate. However, the difference of net enrollments inrelationto income rises with higher levels of education and becomes very steep for upper secondary and tertiary education. Rural-urban differences also become more pronounced. For example, amongst 18-24 year-olds, almost 30% living in urban non-poor households attended university, compared to almost none amongst those living inruralextremepoorhouseholds. Table 3.3. Enrollment rates are muchlower for poorer groups at higher levelsof education Enrollmentratesbv Dovertv status. locationandeducation Level. 2002 ~~ Povertv Status Urban Rural National Urban Rural National Primary (6-1I) Lower Secondary (12-14) Extreme 87.8 88.1 88.0 54.3 55.O 54.6 Moderate 90.4 90.3 90.3 65.7 55.9 61.9 Non-poor 94.7 93.3 94.4 83.3 73.8 81.8 Total 92.2 90.9 91.8 73.8 59.4 69.6 Upper Secondary (15-17) University (18-24) Extreme 32.5 16.3 24.0 5.0 0.1 2.8 Moderate 40.8 22.0 34.0 10.2 1.1 7.4 Non-poor 63.8 37.8 59.3 28.5 6.0 25.6 Total 52.3 26.3 45.0 21.6 2.9 17.7 ~ ~~ Note: Poverty calculated usingSEDESOL's food- and assets-basedpoverty lines andcurrent income per capita usingtrimester income as reported by INEGI.Urbanareas are those localities with 2,500 inhabitantsor more, usingINEGI'sclassification. Source: WB staff calculations fromENIGH2002. Evidence on low and variable quality comes both from international and national testing. Chapter 2 discusses the Program for International Student Assessment (PISA) results. The Ministry of Public Education (SEP) also collects data based on national test, collected from all primary and lower secondary schools for the purpose of assessing the skills of their students over time. There i s only modest variation in average test scores by geographic region, despite much larger regional income differences. There are larger differences across category of schools, with urban private schools scoring significantly higher and indigenous schools scoring worst of all categories. Evidence from following a Following the Ministry of Education's definitions, terminal efficiency indicates the share of students that finish a given course on time, dropout is the proportionof students that do not continue into the next level of schooling, and failure the proportion of students that fail/ do not passthe school year. 42 cohort of average results from schools shows that there i s learning over time for all schools, with neither strong divergence nor convergence (Figure 3.1). Figure 3.1. There have beensteady gains intest scores across all schooltypes EstdndaresNacionales global test scores 550 530 510 490 470 450 430 410 390 370 350 1998 1999 2000 2001 2002 2Mgrade 3" grade 4" grade 5" grade 4thgrade --C-.CommunityCenter -I-IndigenousSchool Rural School +Urban Public School +Urban PrivateSchool Source: WB staff estimates usingEstdndaresNacionales, SEP. Using a data set from SEP, Upez-Acevedo (1999) analyzed inequality in learning achievement focusing on differences among individual children. This data set, which covers Mexico's five poorest states, was collected for the purpose of monitoring and evaluating a compensatory education program (PARE). Estimates of the standard deviations for the full sample and the various types of schools covered showed that the dispersion of test scores in Language and Mathematics i s large, which seems consistent with the PISA results in Chapter 2.3 Moreover, some groups of children were clearly lagging way behind others in learning achievement. As with school participation, these childrencome disproportionately from poor, rural, andindigenous families. Health Mexico's many achievements in the health sector over the past decades have led to significant improvements inthe health status of the population, a broadening of access to basic services, and growing public support for important public health measures. Rising prosperity has brought an increaseinlife expectancy, reductions ininfant mortality, and a decline in the death rate (Figure 3.2). Infant mortality decreased from 50.9 in 1980 to 20.5 in 20034, immunization among children has become nearly universal (97.7) and The PISAresultsfind significant variation, but less than inother countries such as Argentina. Salud: MBxico 2002. Infant mortality is the proportion of deaths under one age per thousandlive births. 43 gains have been achieved in maternal mortality -from 15.6 in 1980 to 7.3 in 2003. Life expectancehas increasedfrom 66.8 years in 1980to 74.9 years in2003. Figure 3.2. There have been large improvements in health indicators, including those related to diseasesof the poor a) LifeExpectance 78 66 64 62 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 zw(I 2301 2002 -)Male - C F m a l e InfantMortality Over the last decade, the mortality rate for children under age 5 fell around 20%, and mortality from diarrhea and pneumonia fell by more than 70% (Figure 3.2). These diseases are more prevalent among the poor rural population. Vaccine-preventable diseases have declined drastically, with no cases of polio since 1990 or diphtheria since 44 1980. There i s not data on health status in the ENIGH survey, so there i s not directly comparable information on how health varies with income category.' Despite substantial progress in basic health indicators, that appears to be partly due to centralized programs, the health system has many weaknesses from the perspective of poverty reduction. Coverage of the formal insurance system remains limited, especially for the poor. It appears to be particularly inequitable in the quality of provisioning for catastrophic health risks. Much of the population remains without access to social security. Households with workers engaged in the formal labor market benefit from health insurance through either Instituto Mexican0 del Seguro Social (MSS) or Instituto de Seguridady Servicios Socialesde 10s T r a b ~ j ~ o rdel e s Estado (ISSSTE). The shareof households with at least one rights holder member inthe social security system is almost 50% and has remained constant since 1992. Figure 3.3 shows household insurance by institution. According to the Ministry of Health (SSA, Salud: MBxico 2002)' the majority of households are uninsured. While in principle covered by the SSA's health services, only 15% of the uninsured population who used any services made use of the public system.6 Figure 3.3. HouseholdInsurance by Institution 60.0 50.0 42.6 10.0 0 3 03 15 0.0 m 0 5 7 __. _._ Nmc SocialSsudly Private %gumPapular VoluntsryMSS DoubleInsuranoe Source: Salud: MBxico 2002 For the poor living in rural areas, poor states, and marginal urban areas, access to basic care remains limited and of low quality. As a result the children of women living in extreme poverty are 2.5 times more likely to die before the age of one thanthe children of women who are not poor (World Bank, 2003~).Children mortality rates from infectious diseases are over three times the national average in Chiapas7, and 2.5 times in Oaxaca. ' The Encuesta Nacional de Salud has a much simpler, and less reliable income question: preliminary analysis finds implausibly small differences across income categories.They are not quoted here, since this will require further analysis. Private health servicesaccount for almost 60 of those householdsreporting any health expenditureinthe ENIGH2002. Section3.4 analyzes regional differences inbasic social indicators. 45 Health inequalities are also manifest in the adult population. For example, Chiapas, Oaxaca, and Guerrero (C0G)'s health outcome indicators continue to lag behind the national average. They are at the bottom of the list of states ranked by life expectancy at birth and disability-free life expectancy at birth, though differences seem quite small relative to income differences (Table 3.4). However, the gap in life expectance between the national average and each of the COG states has not been significantly narrowed down (World Bank, 2003~). Table3.4. ComparativeHealthOutcomeIndicators:Chiapas,Oaxaca,and Guerreroversus National Lifeexpectancy at birth: 2000 Disability-free life expectancyat (Years) birth: 1998(years) Male Female Male Female National 71.5 76.5 61.1 66.9 Chiapas 69.6 75.0 56.7 62.3 Oaxaca 69.9 75.2 56.6 63.3 Guerrero 69.9 75.3 55.2 65.7 Highest-Lowest {other than 75.1-7 1.7 79.4-76.3 63-1-58.7 66.8-63.3 COG states) Source: Mtxico PlanNacional de Salud2001-2006. Despite some recent improvements inmedical access and technology, the quality of basic health services continues to be low. A study by Bobadilla (1998) found that a baby born inan SSA hospital was three times more likely to die inits first sevendays of life than a baby of the same weight born in an IMSS hospital. While it is possible that the SSA- attended children are predisposed to higher levels of illness, it is clear that the overall quality of the two systems varies considerably with regards to obtaining satisfactory outcomes. Dissatisfaction i s shown through consumer behavior. In 2000, the National Medical Arbitration Commission (CONAMED) received nearly 774 official complaints regarding the quality of care. Many of these complaints dealt with ISSSTE, that could reflect the greater likelihood of complaining of better off households, rather than differences in quality. However, dissatisfaction is common among public institutions. World Bank (2001~)reports that in 1994 59% of mexicans thought that the current system "has some positive aspects but requires fundamental changes". Half of the users of SSA services were dissatisfied. Their principal reasonswere (a) poor service; (b) lack of resources such as drugs and well-trained personnel; (c) lack of access; and (d) highcosts. Inaddition, the survey revealed that quality issues were not limited to the lack of resources, personnel, and technical problems -40% of patients felt they were not treated adequately, 61% considered services too bureaucratic, and 8% did not receive medical treatment when needed. Surveys carried out in 1999 by SSA showed that only 18 out of the 36 essential drugs were available in surveyed urban and rural clinics. Basic drugs were lacking in more than 50% of the SSA units (Gomez-Dantes, 1999). According to the ENED (National Survey of Performance, 2002), there are substantial delays in getting medical attention in public institutions such as IMSS and ISSSTE. The ENSA (National Health Survey, 2000) reported that 82.8% of the population perceived that the service provided 46 was either good or very good. The RedCross and private hospitals hadthe highest rating, whereas IMSS and Federal District Department (DDF)hadthe lowest. Housingand Infrastructure Over the past decade, there have been major improvements in access to electricity, improved water source and sanitation, particularly inrural areas (Table 3.5). However, in 2002, 10% of all households inMexico did not have access to an improved water source and 19.1% didnot have any kindof sewerage. Table 3.5. Many basic infrastructure services are reaching the bulk of the National Urban Rural 1992 1994 2002 1992 1994 2002 1992 1994 2002 Electricity /Yes 92.4 95.0 98.0 98.5 99.3 99.5 76.2 83.5 93.5 No 7.6 5.0 2.0 16.5 6.5 Water Improved water source 80.7 84.7 89.9 92.2 95.6 96.2 50.3 55.8 70.6 Non-improved water source 19.3 15.3 10.1 1:; 1:: 0.5 3.8 23.8 44.2 29.4 49.7 ~ Floor Slab or finished floor 26.4 30.5 34.0 34.1 39.0 42.8 5.8 7.9 7.4 Cement 55.1 54.5 55.4 57.6 54.3 53.1 48.4 55.0 62.3 Dirtfloor 18.5 15.0 10.6 8.2 6.8 4.1 145.8 37.1 30.3 Sewerage Improved sewerage 65.4 69.8 77.0 81.6 86.0 91.1 122.5 26.2 34.2 Non-improved sewerage 6.0 2.4 3.9 4.0 1.6 1.9 11.3 4.3 9.8 No sewerage 28.6 27.9 19.1 14.4 12.4 7.0 66.2 69.4 56.0 The expansion of basic infrastructure services has included the poor. This is particularly notable for the gains inwater and electricity for the extreme poor inrural areas. Progress insanitationhasimprovedmuchlessfor this group, by contrast (Table 3.6). Despite this progress, there continue to be gaps in basic housing services for both the rural and urban extreme poor. In 2002, in rural areas, 53% of the extreme poor (those living below the food poverty line) lived in houses with a dirt floor and 11% lacked access to electricity. In2002, in urban areas, 18% of the extreme poor haddirt floor and 2.6 didnot have access to electricity. Another source of information, the Encuesta Nacional de Caracteristicas Socioeconcimicasde 10s Hogares (ENCASEH), also indicates that rural and marginalized communities inMexico continues to lack access to a wide range of basic services (Table 3.7). Inthe group made up of the smallest communities with less than 20 households, only 40% of the locations have electricity. Forty percent of the smallest villages still do not have a primary school. Sewerage, public phones, and post offices are virtually nonexistent. Mobile health units serve a majority of the smallest towns, but 30% are not served. Access to services improves with the size of the community, especially for schooling, but even inthe larger communities there are gaps, for example, inhealth care. 47 Table 3.6. There has been large progressinaccessto water andelectricity amongst the ruralextreme poor, but not insanitation Housingcharacteristics amongst the extreme poor and non-poor, Rural I Extreme Poor Non-Poor 1992 2002 I 1992 2002 Electricity Yes 63.4 89.5 88.1 97.7 No 36.6 10.5 11.9 2.3 Water Improvedwater source 38.4 57.5 64.7 86.0 Non-improvedwater source 61.6 42.5 35.3 14.0 Floor Slab or finishedfloor 2.4 1.6 11.9 18.4 Cement 33.7 45.5 67.4 72.4 DirtFloor 63.9 52.9 20.7 9.3 Sewerage Improvedsewerage 12.8 13.9 37.8 59.9 Non-improvedsewerage 9.0 11.4 15.3 8.9 No sewerage 78.2 74.7 47.0 31.2 Table 3.7. Marginalized ruralcommunities have low access to public services Percentof householdswithout access by service Community Size (Nun; tr of Households) upto20 21 to60 i l or more Electricity 59 40 20 Sewerage 90 87 84 Public phone 97 90 52 Postoffice 98 98 95 Preschool 68 28 6 Primary school 40 13 2 Telesecondaryschool 99 95 69 Secondaryschool 100 100 95 SSA clinic 98 93 76 I M SS-Soliduridud 100 98 90 Localhealth auxiliaries 72 47 41 Healthmobile unit 30 - 25 25 Source: ENCASEH. B. The Structure, CorrelatesandTrends inIncomePoverty Overview Incomepoverty remains both broad and deep in2002, according to official definitions of poverty, with a fifth of the population living in extreme poverty and half in moderate poverty. This reflects high levels of inequality, with the bottom 20% of the population 48 receiving just over 3% of total income, while the top 10% receives over 40 percenL8 While there is considerable heterogeneity amongst the poor, the extreme poor (those living below the food-based poverty line) are more likely to live in rural households in which the breadwinner's main occupation is agriculture (on small farms or as laborers), and have primary education or less. Indigenous groups suffer disproportionately from extreme poverty -an issue taken up in the last section of the chapter. The relatively small proportionof extreme poor who live inurban areas -especially inthe mediumor smaller cities and towns- also typically have low education and are most likely to depend on informal wage work, Those who live in moderate poverty (between the food- based poverty line and one that allows for acquisition of basic assets) are more likely to be urban, with primary education and depend on informal wage work, though a significant portionof formal workers inunskilled work also livebelow the poverty line. The persistence of high levels of income poverty reflects both the continued extreme income inequality in Mexico, and slow growth in average incomes over the past two decades. Inthe pastdecade this was largelydue to the massivesetbackfor poverty caused by the 1994-96crisis, which led to sharp increases in moderate and extreme poverty in bothrural and urban areas. Mexico has only recently brought poverty levels down below the levels prevailing in the early 1990s. The results of the 2002 household survey (ENIGH 2002) suggest a significant reduction in extreme poverty since 2000 despite stagnant overall growth. This is attributable to an apparent surge in income growth in rural areas, and equalizing patterns of income growth inurbanareas, where poorer groups experienced steady growth, while richer groups suffered income declines. While questionnaires changed in 2002, review of different sources and robustness tests undertaken for this report supports the view that the surveys are sufficiently comparable to draw conclusions, and that reductions inextreme poverty were statistically significant. This was associatedwith rapidwage increases for the low-paid and, for rural areas, large gains from remittances and transfers. This i s broadly in line with the conclusions of the Technical Committee on Poverty Measurement. TheStructure of Zncome Poverty As with non-monetary poverty indicators, monetary extreme poverty is especially severe inrural areas, where some 35% of individualsliveinhouseholdswith less than the food- basedpoverty line, comparedwith 11% inurban areas. Around 65% of the extreme poor live inruralareas'. It is disproportionately a feature of householdswhose heads mainjob i s inthe agricultural sector, whether as a self-employed farmer or rurallaborer, and has at most primary education. In urban areas, extreme poverty i s associated with household heads working in agriculture, manufacturing, construction and services; and also with low levels of education. The composition of the poor i s summarized in Tables 3.8 and 3.9. It is useful to compare the characteristics of the extreme poor with those of the * This is calculated from the ENIGH, the household income and spending survey. This is probably particularly weak at capturing capital incomes, which are typically more unequally distributed. The true distribution may be even more unequal. ' Table 1 in Box 1.1 shows the SEDESOL official poverty lines, which are also in Table A.3.4 in the annex. 49 moderate poor, treating people who have incomes between the food-based and assets- based lines as falling in the latter group. These are somewhat arbitrary categories, since there is a continuum of welfare levels across poverty lines, and heterogeneity within groups. However, the typical characteristics have a useful illustrative value. Table 3.8. The Compositionof the Extreme Poor in1992 and2002 by Household Head 'haracteristics 1992 2002 National Urban Rural National Urban Rural Proportion of extremepoor inthe population 22.4 13.3 35.6 20.3 11.4 34.8 Rural-urbancomposition10 100.0 35.0 65.0 100.0 35.0 65.0 Profileby education NoEducation-Primary Incomplete 74.3 58.2 82.9 65.3 51.3 72.9 Primary Complete 20.8 32.8 14.4 23.2 27.0 21.1 LSecondary Complete 4.2 7.1 2.7 9.5 17.6 5.2 USecondaryComplete 0.7 1.8 0.1 1.6 3.3 0.7 University Complete 0.0 0.0 0.0 0.4 0.8 0.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 Profileby employment Agricultural Laborer 22.7 6.5 31.2 24.3 10.1 31.0 Non-agricultural Laborer 35.3 64.3 20.1 29.6 55.9 17.0 Employer (under 5 employees) 4.3 2.8 5.0 2.5 1.5 3.0 Employer (5 or +employees) 0.8 0.2 1.1 0.3 0.0 0.4 Self-Employed 36.0 24.2 42.2 42.9 32.1 48.0 Non-remuneratedworkers 0.9 2.0 0.4 0.5 0.3 0.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 Profileby sector of activity Agriculture 49.6 11.7 69.1 52.7 12.7 71.7 Extraction 1.5 3.1 0.6 0.1 0.1 0.1 Manufacturing 11.8 24.0 5.4 8.6 17.4 4.4 Utilities 0.0 0.0 0.0 0.1 0.3 0.1 Construction 13.3 21.3 9.1 12.4 17.6 10.0 Commerce 7.9 11.3 6.2 8.8 17.7 4.6 Transportation 2.4 4.5 1.3 3.6 7.9 1.5 Financial Services 0.1 0.3 0.0 0.2 0.2 0.2 Services 13.6 23.9 8.2 13.5 26.1 7.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 Source:WB staff estimates usingENIGH1992,2 2. Following SEDESOL's L )adruraldefinition andthe official poverty measurement. lo 1992and2002figuresarenotequal,butarethesamewhenroundedtoonedecimalplace. The 50 Table3.9. The Compositionof theModeratePoor by HouseholdHead Characteristics(Excluding ieExtremePoor) in1 92 and2002 1992 2002 National Urban Rural ~ ~ t i o n aUrban Rural l Proportion of moderatepoor inthe population 30.1 30.6 29.4 31.4 30.6 32.7 Rural-urban composition 100.0 60.0 40.0 100.0 60.6 39.4 Profile by education No Education-Primary Incomplete 57.3 42.3 79.9 44.7 33.3 62.2 Primary Complete 27.5 35.3 15.8 27.8 30.1 24.1 LSecondary Complete 11.5 16.7 3.7 21.1 27.1 11.9 USecondary Complete 3.0 4.5 0.7 5.5 8.1 1.5 University Complete 0.8 1.3 0.0 1.o 1.4 0.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 Profile by employment Agricultural Laborer 8.5 1.4 18.8 9.3 1.8 21.0 Non-agricultural Laborer 59.5 75.5 36.5 57.3 69.7 37.8 Employer (under 5 employees) 4.5 3.4 6.2 3.4 3.2 3.8 Employer (5 or +employees) 0.9 0.1 2.2 0.1 0.1 0.1 Self-Employed 26.4 19.6 36.3 29.3 24.7 36.6 Non-remuneratedworkers 0.1 0.2 0.1 0.6 0.5 0.7 Total 100.0 100.0 100.0 100.0 100.0 100.0 Profile by sector of activity Agriculture . 23.2 2.4 52.5 20.4 3.3 47.2 Extraction 0.5 0.3 0.7 0.5 0.3 0.9 Manufacturing 17.0 20.8 11.6 15.9 18.7 11.6 Utilities 0.6 0.9 0.4 0.5 0.8 0.1 Construction 13.1 14.3 11.4 13.4 13.7 13.0 Commerce 14.8 18.8 9.1 16.5 22.6 6.9 Transportation 4.2 6.6 0.7 7.3 8.3 5.8 Financial Services 0.1 0.1 0.0 0.6 0.9 0.0 Services 26.5 35.7 13.5 24.8 31.4 14.5 Total 100.0 100.0 100.0 100.0I 100.0 100.0 Source: WB staff estimates usingENIGH 1992,2002. .Following SEDESOL's urbanlruraldefinition and the official poverty measurement. While the broad pattern of extreme poverty i s similar in 2002 to the situation ten years earlier, there has been a notable rise inthe educational status of household heads living in poverty and in extreme poverty -education failed to pay off in higher incomes because the lack of expansion in earnings opportunities. There was a slight urbanization of extreme poverty and an increase in the concentration of such poverty amongst the self- employed. With respect to sector of activity, the composition of extreme poverty has remained unchanged in the last decade. Table 3.9 shows that the extreme poor rural- urbancomposition has not changedinthe last ten years, 35% live inurban areas and65% inruralareas. The moderate poor are muchmore urbanized: 61% live inurban areas and 39% in rural areas. Compared with the extreme poor, the moderate poor are more 51 educated, more likely to be in non-agricultural work (notably services, manufacturing, construction, and commerce), and more likely to be urban workers. There has been an increase in the proportion of moderate poor in rural areas in the last decade, reflecting a more rapidtransition from extreme into moderatepoverty, than out of moderatepoverty. "Determinants" anal~sis2 ~ ~ 2 - d ~ ~ ~ g reducation, and sector u ~ h i c , Tabulations of the incidence of poverty among various household categories do not take into account the correlations between the household's characteristics. To assess the impact of various characteristics on the probability of being poor, probit models were estimated nationally (Table A.3.1). These results clearly show that education is associated with a lower probability of being poor after controlling for other factors. In terms of household head education, completed primary education is associated with a 31%lower probability of being poor, and lower secondary education a 34% lower probability, compared with households whose heads did not complete primary education. This i s a statistical analysis of conditional probabilities, that does notimplycausation: indeedthere are likely to be two-way directions of causation, with lower education reducing earnings and lower income reducing the capacity to invest ineducation. Ina similar spirit to the analysis of the probability of beingpoor, regressions were also estimated to analyze the marginalcontributionto income of household Characteristicsand location, The results of such regressions are provided in Table A.3.2, and give the marginal percentage gains in per capita income associated with various characteristics and location, after controlling for all other characteristics. The following summarizeskey results (all are significant at the 5% confidence level). Demographics: the larger the family, the lower the level of per capita income and consumption; however, the extent of this effect i s highly sensitive to assumptions over equivalence scales across the needs of different household members and economies of scale. There is also a nonlinear relationship between the age of the household head and per capita income with households where older heads enjoy higher income. Gender: there i s evidence that households with female heads have slightly lower per capita income levels than households with a male head (holding the number of adults constant). Access to social security: households with access to public social security have 22% higher per capita income levels than households without access to public social security, reflectingthe fact that this i s associatedwith higher-payingjobs. Education: the gains from education are large and have increasedover time for heads and for spouses. Households with a head who has completed secondary education have an income close to 43% higher than a household with a head who does not have any education -after controlling for other factors. The gain from some form of higher education i s twice as large. 52 Employment: having a head employed outside agriculture improves the household's living standards for most industrial sectors, relative to being employed inthe agricultural sector. This evidence, however, i s less clear for younger workers. Agricultural workers and self-employed individuals fare less well than do other workers and employees, while business owners fare better. Location: households in urban areas have an average 20% higher per capita income compared to those inruralareas (after controlling for all other observablefactors -this is explored further below). Zndigenous ~opulations:the regression results in Table A.3.2 are basedon data from the ENIGHsurveys, which do not have informationon indigenous peoples. Results obtained usingthe Census 2000 indicate that being indigenous reducesper capita income by 15% in rural areas and has no significant impact in urban areas after controlling for other factors. Explaining Rural-Urbanper capita differences In light of the salience of rural-urban per capita difference a particular focus was on analyzing the contribution of households characteristics to this difference. Such differences inincome may come from differences incharacteristics (say, a lower level of education for the household head and the spouse inrural areas) or from differences inthe returns to characteristic (say, a lower impact of education on earnings and thereby a lower per capita income in rural areas). In some cases, the differences in characteristics and in the returns to characteristics reinforce each other, but in other cases they partially offset each other. This was analyzed using the Oaxaca-Blinder decomposition of the income gap betweenassets andreturns (Blinder 1973 andOaxaca 1973). Box 3.1 Usingthe Oaxaca-BlinderMethod for Rural-Urban Decompositions Regressionsare estimatedfor both urbanandruralareas Where `R i s the income per capita in rural areas, PRrepresentsthe regressioncoefficients in rural areas, x~ represents the vectors of independent variables. is the error term. By simple algebra we come to equation (3): The f i s t term on the right hand side of the equation is interpreted as the share of the difference in income due to difference in "assets" or characteristics. The second term representsthe shares of the difference in income due to differences in the marginal influence of the assetslcharacteristicson household income. In some cases this can be interpreted as differences in a "return" in the market, for example in wage rates. In 53 other cases (such as the number of children), they may reflect more complex influences on household incomes. The shares of the difference in per capita income due to characteristics and returns are given inTable 3.10. Table 3.10. Decompositionof Rural-UrbanDifferencesinper CapitaIncome Share of differenceinper capita incomedue to: Assets Returns Total Difference Total infants and children characteristics 8.2 -4.7 3.5 Total adults demographiccharacteristics -0.8 14.7 13.9 Total educationof adults 33.4 1.4 34.7 Total employment characteristics 30.5 18.0 48.5 Total 100.0 Notes: Infants and children characteristics: are number and number squaredof children and infants. Adult characteristics: number of adults, the age of the householdhead, ifthe headis a female, if there is no spouse in the household and the employment and underemploymentconditions of the householdhead and spouse. Education: educationlevels of the householdheadand the spouse. Employment characteristics: sector of activity and position of employment in the firm by head and spouse. These results are obtainedfromdeterminantsof income regressionsinthe rural and urbanareas. Source: WB Staff estimates usingENIGH2002. Current incomeper capita as reportedby INEGI. Differences in numbers of children explain very little of the rural-urban difference, despite higher family sizes (on average rural families have 1.8 children while urban families have 1.3.). This would probably decline further if allowance were made for equivalence scales andeconomiesof scale (see above). Differences in the demographic characteristics of the adults and in their labor force participation account for 14% the difference in per capita income between rural and urban areas entirely due to differences in retums. This appears to be due to a number of effects: households without spouses have lower per capita income; and adult underemployment rates are higher in rural areas, rural household heads are older on average. The education of the household head and their spouse accounts for 35% rural-urban differences, almost entirely due to differences in educational endowments rather than differences in returns. Additionally, the returns to the education of the household head tend to be larger inurban areas, possibly becausethey reflect the relative scarcity of well- educatedlabor. Finally, almost 50% of the difference i s due to employment characteristics of the household head and their spouse, with important "asset" effects (a higher proportion in jobs of a worse category) and return effects (lower returns to jobs in the same sector.) This is due primarilyto the higher proportion of employment in agriculture and in small firms in rural areas as well as lower productivity in these activities. For example, a household with a heademployed inmanufacturing inurban areas has an expectedlevel of 54 per capita income 30% higher than an otherwise similar household with a head working in agriculture. This is consistent with the lower return to individual characteristics - especially education- inrural areas. Trendsin income and spendingpoverty As noted above, poverty trends suggest a pattern of slow long-run progress in the reduction of income poverty, overlaid with the dramatic impact of 1994-95 macroeconomic crisis. The evolution of poverty mirrors Mexico's macroeconomic performance well, except for the period 2000-2002. It is interesting to note that although the proportion of poor individuals fell between 1996 and 1998, the depth and severity of poverty increased, indicatingthat the poorestindividuals did not benefit initially from the economic expansion in this period (see Table A.3.3 for the informationon the series for the poverty gap and poverty gap squared). However, significant gains in terms of the depth and severity of poverty are observed since 1998. The ENIGH2002 results suggest a statistically significant reduction in extreme poverty since 2000 for the rural and national level, despite stagnant overall income growth. We now turn to examine these trends with particular attention to the recent period. We should note that there have been questions raised in public debate in Mexico as to whether the recent changes are (partly) due to measurement issues. The ENIGH survey was changedbetween 2000 and2002, with amongst other changes, more detailed income questions. This could theoretically have ledto changes inthe degree of under-reporting of incomes. The Technical Committee for Poverty Measurement has recently reported that they judge the surveys to be comparable with the improvements in 2002 providing greater precision to the estimates (Comunicado 3, Technical Committee for Poverty Measurement). However, for some specific categories of income (notably some transfers), there may notbe comparability. This report concurs with this assessment. Box 3.2: Poverty trends in2000-2002 From 2000 to 2002, real GDP per capita income decreased by 1.8%. In this context, many observers expectedpoverty to increase assumingno change in inequality. However, analysis of the householdsurvey by SEDESOL that largely followed the Poverty MeasurementCommittee methodology found that poverty decreased, particularly respectto extremepoverty. At the same time, the National Income and Expenditure Survey (ENIGH) 2002 experienced significant changes that could, theoretically, have affected the comparability with earlier rounds of the ENIGH, including that of 2000. The major changes of relevance were as follows. First, the ENIGH 2002 questionnaire included more questions on both incomes and expenditures compared to the ENIGH 2000. Second, the poor were over-representedinthe ENIGH2002 sampling design. In December 2003, after an initial statistical evaluation, the Poverty Measurement Committee (CTMP, 2004) concludedthat the 2002 survey is ingenerala superior and moreaccurate survey instrument, andthat l1Mexico's economyrecoveredquickly fromthepesocrisis andfrom 1996to 2000averagegrowth was higher than 5 percent per year. However, Mexico's economic performance faltered again in 2001 as a consequence of US economic recession. I n 2002, Mexico's economy was basically stagnated even though some sectors, like commercial agriculture, grew importantly from2000 to 2002. 55 it is in generalcomparablewith the 2000 survey at leastfor aggregate measures.They also concluded that there was a statistically significant fall in food-basedpoverty at both the national level and in rural areas12, plus a statistically significant decline in capacities poverty in urban areas. Other poverty changes are not statistically significant (measured declines may be indicative of a real decline but i s not sufficiently precisely measured for this to be statistically different from no change). However, they concluded that individual components of income would often not be statistically comparable. Work on comparing any effects of changes inthe survey on expenditures was not complete at the time of writing. The above conclusions are based on two categories of internal evidencefrom comparing the two ENIGH surveys (Rivera Vences, 2003; CortCs, 2003). First, Cortks shows that there is no statistical evidencefor an atypical increase in total income due to the increasein the number of questionsin the ENIGH. Categories or subcategories of income that did not change betweenthe surveys account for 99.1% of total income, in the majority of the cases the cells of the new categoriesof incomewere empty. However, in some cases the changed questionnaire appears to have led to implausible changes in the composition of income (for example betweencorporate and cooperative income) without making a statistical difference in the income aggregate for households and individuals. Second, with respect to the samplingframework, he concluded that the over-representation of the poor led to more accurate measurement but that this did not produce abnormal changes in the distribution of household characteristics amongst the poor (that would have occurred ifthe ENIGH2002 and2000 were effectively drawn fromdifferent samples). This report concurs with the broad conclusionsof the committee on the superiority of the 2002 survey and the comparability with 2000 for total household (and per capita) income. It also provides estimated confidence intervals for poverty comparisons where feasible. The conclusion is further supported by comparison with independent evidence on trends, especially from the series of labor force surveys, for which the survev instrument did not change. The intemretation of trends is discussedinthe text. Inthis report, we undertook a number of exercises to assess the robustness of population poverty trends. First, we estimated the statistical significance of the 2000-2002 changes inpoverty for bothincome andconsumption. Income poverty was estimated using the SEDESOL official poverty lines. As discussed inBox 1.1, we also re-calculate poverty lines in order to apply a constant 2000 Engel coefficient and Banco de Mtxico price indexes for non-food items of the poor - compared with SEDESOL's analysis that uses a variable Engel coefficient. (The Engel coefficient refers to the food share used). These are reported in the annex (Table A.3.5 and A.3.6). The overall pattern of changes in poverty in 2000-02 remains robust to different poverty lines. We found that poverty estimates for 2002 are slightly lower using a fixed Engel coefficient. Therefore, SEDESOL's approach of using a variable Engel coefficient in2002 is more conservative. We also estimated consumption poverty rates fixing these for one year to yield the same poverty rates as the official povertyline (2000 was chosen, but the choice i s arbitrar~).'~ 12Rivera Vences (2003). l3 2000, wecalculatedthe incomepovertyratesbasedonSEDESOLofficialpoverty lines.Wefixed For this poverty rateandcalculated the correspondingconsumptionpoverty line for the same year. The poverty line then projected to other years using a food and non-food CPI (constructed from Banco de MCxico's CPIs on goods and using consumption weights from ENIGH 2000), and consumption poverty rates were calculated. Consumption poverty lines are shown in Table A.3.7 and the constructed CPIs are shown in Table A.3.8 inthe annex. 56 Consumption i s a superior measure of well-being than income and is less likely to be subject to reporting errors ---orchanges inreporting errors on the income side. Of course, improvements in the survey of consumption may also have led to a reduction in under- reporting for consumption but it provides an initial crosscheck. Table 3.11 shows the income and consumption poverty trends. The overall consumption pattern is close to that for income poverty. For the recent period, national extreme poverty rates significantly decreased from 2000 to 2002. In rural areas there were significant extreme poverty declines for both expenditure and income-based trends whereas for urban areas the measured declines were not statistically significant.l4From 2000-02, consumption moderatepoverty declined significantly nationally andinrural areas. Table 3.11. Overview of Poverty Trends for Extreme and Moderate Poverty with Income and Consumption Significanceof 1992 1994 1996 1998 2000 2002 change2000- 2002 Income Nat~onal 22.4 21.4 31.1 34.0 24.2 20.3 *** Extreme Rural 35.6 31.0 52.4 52.4 42.4 34.8 *** Urban 13.3 10.1 26.5 21.2 12.6 11.4 National 52.5 55.8 69.6 63.1 53.1 51.1 Moderate Rural 65.0 12.3 80.8 75.1 69.3 67.5 Urban 43.8 43.1 61.9 55.1 43.1 42.0 Consumption Na~ional 24.8 22.0 36.1 33.4 24.2 21.0 ** Extreme Rural 41.8 38.1 52.9 52.0 42.4 31.2 ** Urban 13.0 9.8 24.5 20.5 12.5 11.1 National 51.2 50.4 64.0 60.6 53.1 51.3 * Moderate Rural 69.3 68.1 16.5 15.8 69.3 66.1 * Urban 38.1 31.0 55.3 50.1 43.1 42.2 Note:*** Significanceat 1percentlevel.** Significanceat 5 percentlevel. Significance at 10 * percentlevel. Source: WB Staff estimatesbasedonENIGH, usingSEDESOL's poverty lines for income measure andconsumption poverty lines inTable A.3.1. Second, we computed poverty trends for both income and consumption excluding transfers and imputed rent. We did not find a change in poverty trends when excluding either one of these two ~ateg0ries.l~ Third, we examined the changes in the distribution of total household expenditure per capita for several years (Figure A.3.1). This i s useful for assessing whether changes in poverty are sensitive to the choice of poverty line. We found dominance of first and second order from 2000 to 2002 up to the 50thpercentile indicating that the measured changes inpoverty were robust to the choice of poverty line. l4 resultsareconsistentwithRiveraVences(2003). These l5Resultsareavailableuponrequest. 57 Fourth, we looked at income trends from an independent survey source -the labor surveys. This has a more limited concept of income but to our knowledge there were not changes in survey design in this period. As reported below, measuredincomes from this source also increasedsignificantly for poorer households, These results support the view that the observed changes were mainly due to real developments. We now turn to the interpretation of these changes. We first examine the relationship between growth, inequality and poverty. We then look at the different sources of income, looking at the trends in labor income (that account for a significant share of total income), and for the most recent period, patterns of transfers and remittances. Growth, inequaZi~and poverty. An overview of the relationship between growth and poverty i s provided in Figure 3.4, for both income and private consumption. This includes information from national accounts sources, the survey mean and the average income of the bottom20% (as a proxy for income of the extreme poor) for three periods: the 1994-96 years of crisis; 1996-2000 years of recovery; and 2000-02 years of stagnation. Two patterns are evident. First,there are sharpdifferences betweenthe evolution of national accounts-basedgrowth and survey-based incomes in both the crisis and recovery. There is no difference for 2000-02. Although the differences are of opposite signs, the net effect for 1994-2002 is still notable: national accounts information suggest growth in real per capita incomes, while households surveys show a modest decline. These contrasts also exist, if in more muted form, for the series on consumption. While there are different views on the reliability of the different sources of information, for the purposes of assessing trends in household welfare we prefer the household survey information.l6 Second, there are differences between changes in average income and income of the bottom quintile. The crisis was mildly equalizing -with large falls at the mean and slightly smaller declines for the bottom 20%. This is consistent with other analyses that find relatively large declines at the top of the income distribution (Upez-Acevedo and Salinas, 2000). The recovery period was then disequalizing while the recent 2000-02 period quite strongly equalizing with respect to the parts of the distribution captured by these summary statistics. This i s consistent with the result of poverty decline despite stagnation. Figure 3.4. Patterns of Growth for National Accounts, Survey Mean, and Bottom Quintile, 1994-2002 l6 has beendiscussionof theseissuesindebates on global poverty trends, and especially on There interpretation of Indian data. See Ravallion (2001), Bhalla (2002), Deaton and Drkze (2002) and Deaton (2001) for discussion. 58 a) Income 40h1- -30% 1 199419% 19962OOO 2 4 " 19942002 PUiOd mGDPpereaplls(NaU-1 Aemunts) .Mean am'" lnmrwpreaplta .Mean current(neomepercapita Mqulntlle Source: WB Staff estimates usingENIGH, andNational Accounts. b) PrivateConsumption 40% 1 30%i zoB 10% ~ 0% -10% -20% -30%L 19!%19!M ~ 1 ~ Z o O o 2ooo;2w2 1994-2002 P d d Pr(vs1eComumpUO" percapita MesnCumt Expenditurepercapita 0MmCum1Expenditurespercapita1JtquinUle Source: WB staff calculations usingENIGHandNational Accounts. Broader measures of inequality are consistent with these patterns. In the past decade, there was a slight equalization at the national level after the crisis -more so for spending than income measures- then a widening back to 1994 levels between 1998 and 2000, and finally a significant equalization between 2000 and 2002. This is basedboth on the Ginicoefficient and also on the frequency distributions of per capita income. Behindthe trends there were quite divergent patterns between urban and ruralareas (Table 3.13). In particular, measuredrural inequality has experienced a very large increasebetween 1994 and 2000, especially when measured in terms of spending, followed by a fall in 2002. This widening included the period of the crisis which also corresponds to the implementation of NAFTA and the recent period of aggregate stagnation. Rising rural inequality is consistent with overall equalization in these periods because of equalizing 59 within-urban trends, and especially for 2000-02, a reduction inthe urban-ruraldifferential (Table 3.12). Annex Tables A.3.9 and 10 shows accumulated income per capita by population percentile, national, urban, andrural. Table3.12. Inequalityhasdeclinedat nationalbut riseninruralareas over the pastdecade, 1992-2002 Ginicoefficients Income Expenditure National Urban Rural National Urban Rural 1992 0.53 0.49 0.47 0.5 1 0.47 0.41 1994 0.54 0.50 0.44 0.52 0.48 0.40 1996 0.52 0.49 0.45 0.48 0.45 0.42 1998 0.54 0.49 0.49 0.49 0.45 0.44 2000 0.54 0.48 0.53 0.52 0.45 0.56 2002 0.51 0.46 0.49 0.50 0.45 0.48 Source: WB Staff estimates using the approved methodology of the Technical Committee for Poverty Measurementfor the calculation of per capita incomeand expenditure variables. The frequency distributions of per capita income show large disparities between urban and rural areas (Figure 3.5). The rural distributionis more concentrated around low per capita income while the urbandistributionis more evenly spread. From 1992to 1996, the ruralpercapita income distributionshifted to the left, meaningthat the per capitaincome distribution was more concentrated around the lower part of the per capita income distribution, From 1996 to 2002 this trend was reversed making the 2002 distribution more similar to the 1992 distribution. The same happened in urban areas but the shift from 1996 to 2002 was even larger. Figure 3.5. FrequencyDistributionsof per CapitaIncome a) 2002 i'". '"'\ i 1Mo Mnthly per capila3incom 2000 i O O 4 m 50Ml I- Rural -----Lkban b) Rural 1992,1996, and2002 60 6 Lo 9 6 9 ._ b 0"F 0 Lo 9 0 I I I I 0 500 1000 1,500 2M)O 2500 Monthlypercapita mcom c) Rural2000and 2002 Lo 6 9 v) I 1 I 0 500 Monthlypercapita1,500 1000 2000 2500 imam - 2 m -----2002 d) Urban 1992,1996, and2002 - - I 1 0 1 0 1Mx) -1992_____1996 Monthlyper capita3iocom 2ooo m 4033 5 m -__...-2002 e) Urban2000 and2002 61 0 3ooo 5000 lMX) percapitaincome Monthly 2ooo 4000 The pattern of changes in inequality is further illustrated in Table 3.13 which shows both income and spending growth by quintile for the 2000-02 period. The strongly equalizing national pattern was composed of high growth in rural incomes for the bottom four quintiles but declines for the top quintile and more modest and equalizing growth in urban incomes. Both measured income and spending significantly declined for the top urban and top ruralq~inti1es.l~is also noteworthy that changes in consumption were It often not significant, even when income changes were significant; this is consistent with consumption smoothing of households. Table 3.13. Patterns of income andspending growth, 2000-2002 Total percentageincreaseover the two-year periodinreal terms National Urban Rural Income Group Current Current Current Current Current Current Income Expenditure Income Expenditure Income Expenditure IstQuintile 9.7 *** 8.2 ** 5 5.1 ** 6.8 ** 10.5 * 2nd Quintile 7.6 *** 5.4 *** 3 *** 1.5 11.4 *** 5.9 3rdQuintile 3.7 *** 1.3 2.5 *** 0.9 8.9 *** 3.6 4th Quintile 2.1 ** 1.7 1.4 2 5 *** 1.8 5th Quintile -7.2 -3.4 -9.2** -1.5 -6.4 -12.9 Total -1.8 *** -0.1 -3.4*** 0.4 0.2 -4.4 *** Note: Current income and spending refers to all transactions on the current account of households; total spendingalso includes capital accounttransactions(sale and purchases of financial capital). *** Significant at 1percentlevel ** Significantat 5 percentlevel. Significant at 10percentlevel. * Source: WB staff calculations using ENIGH with income and expenditure calculated as recommendedby CTMP. Following Datt and Ravallion (1993), the changes in poverty rates are decomposed into growth and redistributive components between 2000 and 2002 (Table A.3.12 in the l7mightbenotedthatthereisnotanobviouschangeinmeasurementpracticeswouldleadtothispattern It of declines at the top and rises at the bottom of the distribution. Note also that the quintiles are calculated separately for each distribution: thus the bottom two to three rural quintiles are disproportionately representedinthe bottomnational quintile -thus the highgrowth for the latter. 62 annex). The results further confirm that in the period under consideration national extreme poverty, rural poverty, and urban poverty declined primarily due to the redistribution effect.l8 The economic slowdown disproportionately affected the top decile so income distributionbecamemore even.Ig Income sources. To explore further the interpretation of the trends we look at the composition of incomes for 2000 and 2002. As Table A.3.13 in the annex shows, labor earnings represents half of total income, when the latter includes the imputed value of non-monetary income. This Table includes the estimates and the corresponding standard errors. There are some classification changes (due to changes in the questionnaire) so we have aggregated "business" and "cooperative income", that otherwise show implausible shifts into cooperative incomes. There are also likely to be overlaps between these and labor income, both conceptually (since for self-employed activities part of the income i s a return to labor) and in the ways in which questions were answered and classified. The large shift from business and cooperative into labor income in rural areas between 2000 and 2002 probably reflects such a classification issue.20 Combined labor and business income accounted for 70% of urban and 63% of ruralincome. We turn to these next and then look at transfers and remittances, that were of greater importance to rural households. Labor market ~eve~o~ments. pattern of change in labor incomes will clearly have a The powerful influence on household welfare. Mexico is going through structural changes in terms of geographic composition, the skills of the labor force and the returns to those skills. Overlaid on these are the strong impacts of the macroeconomic cycle. For labor market developments we have the benefit of an independent survey of labor market characteristics which has covered urbanareas for some two decades and was extended to rural areas in 2000. This includes questions on labor incomes but not on spending or other categories of incomes. To draw the link with poverty we calculated the proportion of workers earning wages below the various income poverty lines b a s e d on labor income earnings of individuals inhouseholds inthe neighborhood of these poverty lines from the ENIGH.). As Figure 3.6 shows, the direction of changes in the proportion earning below such poverty-wages in urban and rural areas is the same as for the estimates of household poverty incidence though the pace of reduction i s somewhat smaller. Figure3.6. The Proportion of Workers Earning Wages below Levels Corresponding to Different Poverty Lines; RuralandUrban2000-2002 The resultsremain robust either usingincome or consumption. l9The changeinpoverty ratesdecompositionexercisefor the period 1992-2002alsoshows that national overty declined due to the redistribution effect f0Cortks(2004) also supports this as a classification issue. 63 1 Urban 1 ~ Rural Urban ~ Rural 2000 2001 1 Urban Rural 2002 1 BFood BasedLineUCapacitiesLine HAssets Line ] Note: Subsistence wage represents the average wage per hour of workers in the households in the line of poverty. This was calculatedusingthe ENIGH2002 and SEDESOL's poverty lines (2002). Source: WB staff calculations usingENET, severalyears. Urban and rural labor markets are considered separately. For the urban sector, the dramatic effects of the cycle are vividly shown in Table 3.14: labor earnings collapsed across the board in 1994-96 and recovered significantly (though remain below 1994 levels in 2002). Particularly noteworthy is continued strong growth in informal and formal wages in 2000-2002 while self-employed and employer's earnings declined on average. Table 3.14. The evolution of realwages for different groups inthe urbanlabor force, 1994-2002 Labor force structure Change inrealwages 2002 1994-96 1996-2000 2000-2002 Male Female Male Female Male Female Male Female Employer 5.5 1.9 -41.9 -52.9 24.5 18.0 -11.8 -12.2 Self Employed 27.2 25.7 -38.6 -48.0 25.6 24.0 -1.6 8.2 InformalSalaried 14.3 15.9 -32.1 -46.7 17.3 13.2 FormalSalaried 50.1 53.2 -21.4 -24.8 7.6 10.6 5.2 6.2 Source: WB staff calculations fromENET. The structural changes associated with this pattern are illustrated inthe Figure 3.7, which shows the partial effects of education along the income distribution. Returns for upper secondary rose sharply in the late 1980's and early 1990's and then fell after 1993. 64 Returns to tertiary education continued to rise until 1997, before falling to levels that remainedmuchhigher than inthe early 1990s(Figure 3.7).21 Figure 3.7. Yearly Rates of Returnto Education Level Mexico UrbanAreas, 1988-2002 12.0% +Primary Complete +Lower Secondary Complete 1 Upper SecondaryComplete *Tertiary Education Note: The yearly rate of return representsthe additionalcontribution to wages from an additional year of a certain levelof education. All the coefficients are statistically significant at the 5 percentlevelandconditioned onage, squaredage, gender, region(North, Center, South, andMexico City). Source: WB staff calculations usingthird quarter of ENEUfrom 1988to 2001and thirdquarter and urban sectionof ENET 2002. A more complex pattern of changes in rewards to education is illustrated by an analysis that looks at different parts of the distribution(usingquantile regression techniques -see Table 3.15). Returns are "convex" and become more so throughout the distribution until 1997-that i s they increased at a risingrate for higher levels of education. In 1988, when estimated at the median of the conditional earnings distribution tertiary education was associated with on average 52% more income compared to a person with upper secondary complete education. B y 1997, the premium to tertiary education had risen to 95%. However, when estimated'at the top of the distribution the premium to tertiary education "only" rose from 34 to 67% (implying higher relative as well as absolute, returns to upper secondary in the upper reaches of the income distribution.) Moreover, 21 The results took into account the change inthe ENEU94 questionnaire. These results are consistent with De Ferranti et. aL(2003) and World Bank (2000a) rate of return patterns, as well as with estimatesusingthe ENIGH. 65 while the premium to tertiary fell somewhat between 1997 and 2002 throughout most of the earnings distribution, they continued to rise at the top.22 1988 1992 QuantileO.l 0.25 0.5 0.75 0.9 0.1 0.25 0.5 0.75 0.9 Primary Complete 1.11 1.09 1.06 0.95 0.85 1.02 1.01 0.95 0.81 0.67 bwerSecodaVCOmPlete 1.21 1.18 1.19 1.24 1.27 1.15 1.18 1.21 1.25 1.3 UPPerseCO~UrycOmPlete 1.11 1.18 1.24 1.25 1.37 1.17 1.2 1.25 1.32 1.38 1997 2002 QuantilelO.1 0.25 0.5 0.75 0.9 10.1 0.25 0.5 0.75 0.9 Primary Complete 1.12 1.13 1.13 1.11 1.05 1.14 1.1 1.1 1.08 1.04 bwerSecodaQ'ComPlete 1.19 1.21 1.26 1.32 1.39 1.15 1.15 1.16 1.21 1.25 UPPerSecondary Complete 1.15 1.22 1.27 1.35 1.42 1.13 1.15 1.21 1.28 1.34 Tertiary 1.75 1.91 1.95 1.83 1.67 1.67 1.82 1.91 1.87 1.73 With respectto gender and geographic areathe results show that ratesof return to tertiary education are higher for bothurbanandruralmencompared to women, particularly inthe upper tail of the conditional earnings distribution. In sum, the returns to education increased in Mexico from 1988 to 1997 especially for higher levels of education and in the upper tail of the conditional earnings distribution. However, there was a reversal to this trend after 1997, especially for higher levels of education and in the middle and lower tail of the conditional earnings distribution. This may reflect a structural development, if expanding relative supplies of school leavers are offsettingthe secular tendency for risingrelative demandfor skills -especially at tertiary level {see De Ferrantiet a12003). Alternatively it may reflect a cyclical fall ineducation premia in times of recession which has also been observed in the data for Latin America {ibid). This issue i s taken up inChapter 5. But for the present, the labor force patternsby labor force status and education are fully consistent with the equalizing patterns of urban household income growth inthe 2000-02period. In terms of employment category {not shown), there was a significant widening in differentials for the self-employed, informal wage-workers and contract workers relative 22 To test the robustness of these trends we estimated the following models: the basic model only included age, squared age, and gender, the second model was the basic model plus region, the third model was the secondplus status inthe labor market, and the last model included all these variables plus sector of activity. 66 to formal wage-earners until 1999, followed by a closing of differences for the last two (lower income) categories. Finally, patterns of formal and informal wage growth for this recentperiodare shown in Figure 3.8.23 This shows increasesfor bothcategories between 2000 to 2002 followed by a flattening of informal and a fall informal wages inearly 2003. Along with the increase in unemployment (not shown here) this suggests a leveling off or partial reversal in income gains for some of the poor inthe periodafter the 2002 ENIGH. rigure3.8. RealUrbanWages byInformality Figure 4 Figure A. Real Wages by informality Figure B. Real wages by gender-informality 4,700 7 T 2,900 5,500 ,I......I...l_.._.....l.l...l..l... ...................................................... I B 2,750 -Formf.Men N 4,500 -Forma( 2,700 (!eft -m-hform!-Men 3,500 4,400 2,650 a b ) e -Form!-Wormn 7 2,600 ---hforml 3,000 -a-hform!-Womr 4,300 2,550 (rieM y,2,500 2,500 2,000 4,200 2,450 1,500 4,1004 . , , . , . , . 12,400 . ~ , , , , . i zooo-n 1,0001 ~ ~ , , , I 2001-I 2001-N 2002-1 2003-n zooo-11 2001-I ZOOI_N 2002-n 2003-n quarter quarter ource: WB staff calculations usingENEU. Controlling for relevant characteristics, the results in Figure 3.9 show that the wage gap between informal and formal workers increased along the conditional earnings distribution. However, since 1997 there has been a reduction inthis wage gap particularly inthe lower tail of the conditionaldistributionwhich is consistent with the results shown inTable 3.16. 23 I n this report, we use the term `informal' to refer to those unprotected by labor law and in turn this category is divided into two. First, owners of firms with fewer than 16 employees who do not have social security or medical benefits are identified as informal self-employed. In fact, less than 1percent of these firms have more than 5 employees. Second, employees in those small firms are identified as informal salariedworkers. 67 Figure 3.9. Wage Gap betweenInformal and FormalWorkers, 1997-2002 Informal Wage as a proportionof FormalWage 80% 79.4% 79% 78% na 76% 75% 74% 73% 72% 71% 70% 0.10 0.25 050 0.75 090 Qurnllle .1m 812002 Source: WB staff calculations using third quarterof ENEU1997to thirdquarter and urbansectionof ENET2002. BasedonWorld Bank 2000a. Turning to the rural sector, for which there is information for 2000-2002, the most strikingresult is the vigorous growth ininformal wages inthis period (Table 3.16). This too is consistent with the growth inhousehold incomes noted above. Measured growth in self-employed incomes i s low to negative. But this category -that largely covers peasant farmers- i s likely to be poorly measured in a labor force survey compared to a household income survey. The forces behind these changes will be explored in future work on ruralpoverty. Table 3.16. The evolutionof realwages for different groupsinthe rurallabor force, 2000-2002 Laborforce Realhourly wages (constant2002 Changeinrealwages (in structure pesos) percent) 2002 2000 2002 2000-2002 Male Female Male Female Male Female Male Female Employer 4.5 1.6 28.8 36.4 30.4 38.2 5.8 5.2 SelfEmployed 30.5 29.6 10.7 11.3 10.4 11.8 -2.6 4.2 Informal Salaried 26.6 20.3 9.5 7.9 11.5 10.1 Formal Salaried 20.9 24.5 18.6 18.1 20.7 20.6 11.3 14.1 Contract 5.3 4.4 15.0 10.4 16.3 11.3 8.7 9.4 Source: WB staff calculations from ENET. Transfers and remittances. The stream of remittances pouring into Mexico is growing year by year. In 2002, remittances amounted to some US$9,815 million - o r about 1.5 percent of Gross National Product and more than the US$4,224 million they comprised just a half dozen 68 years ago. As indicated in Figure 3.10, remittances increasedsteadily -growing by 50% from 2000 through 2002 while foreign direct investment decreased12%.24 Figure 3.10. Remittances overtook foreign direct investmentin2003 0 4fgggggg3gEgfggfgf +FDI +Remlttpnw Source: Banco de Mtxico, andWorld DevelopmentIndicatorsfor Remittances 1986-1995. Total transfers -remittances, Oport~nidades,and Procampo- account for an important share of total transfers for all income quintiles, but particularly for the bottom20% of the national income distribution, and even more so for rural areas (Table 3.17). Note that for this analysis households are ranked by their pre-transfer income since the effects of transfers are in many cases so large as to cause large re-rankings for the post-transfer income (that is usedinthe rest of the analysis inthis chapter). 24 While some of the increase may reflect better measurement, most observers consider there has been substantialreal growth. 69 Table 3.17. Contribution of Transfer Income to the LevelinOverall Income, 2002 National I IncomeGroup Remitrances Procampo Other Total Transfers Transfers 1stQuintile 11.2 5.9 2.4 24.4 38.0 2nd Quintile 2.5 1.8 0.8 9.5 12.8 3rd Quintile 1.o 0.5 0.2 6.8 8.0 4th Quintile 0.9 0.1 0.2 4.5 5.6 5thQuintile 0.1 0.0 0.2 2.5 2.8 Total 1.2 0.6 0.4 5.4 7.0 Income Group Remitrances Oportunidades Procampo Other Total Transfers Transfers 1stQuintile 4.0 0.2 0.0 19.6 23.6 2nd Quintile 0.8 0.1 0.0 8.4 9.2 3rd Quintile 0.7 0.0 0.0 5.8 6.5 4th Quintile 0.4 0.0 0.0 3.4 3.8 5th Quintile 0.0 0.0 0.0 2.6 2.7 Total 0.5 0.0 0.0 4.9 5.5 IncomeGroup Remittances Oportunidades Procampo Other Toral Transfers Transfers 1stQuintile 19.5 10.2 3.8 29.7 53.0 2nd Quintile 6.8 6.8 3.1 12.6 22.5 3rd Quintile 4.2 3.8 1.7 9.3 15.3 4th Quintile 1.7 2.2 0.9 5.5 8.1 5thQuintile 1.o 0.3 1.1 2.2 4.3 Total 3.5 2.5 1.5 6.8 11.8 Note: The other transferscategory includes pensions, severance payments, scholarships, and gifts from other households. Source: WB staff calculations usingENIGH2002. The most striking feature of the table i s the tremendous importance of remittances and other transfers (scholarships, pensions, transfers from other families) for the poorest rural households. In rural areas remittances accounted for 19.5% of average incomes of the poorest quintile and it i s judged to be an important source of income growth since 2002 (Inter-American Dialogue, 2004). We do not show results of the contribution to changes inincome by quintile andincome source becausethe ENIGHis not representative at that level of disaggregation and measurement errors are expected to be large. When all other transfers are added to remittances they account close to 36% of the income of the poorest quintile in rural areas (again ranked by pre-transfer income). ~portunidudesand remittances are the most important identifiedtransfer in2002. For urban areas, the contribution of remittances and transfers i s both smaller and more volatile. A sharp fall inremittances occurred for the bottomquintile. This may be due to greater measurement error given the much smaller number of households receiving remittances. 70 While both remittances and other transfers are of great importance to the level and change of incomes -especially for the rural poor- their coverage is very different. Remittances from abroad are received by 13% of all rural households and less than 4% of urban households. However, for the poorest quintile of rural households this proportion rises to over 20%. Average amounts received are much larger than other sources of income for these households. There is anecdotal evidence that some remittances are also used for community projects. The strongly progressive character of remittances is of particular note in light of the international experience that finds that migrants tend not to come from the poorest groups. By contrast, the reach of O ~ o r t u n i dwas ~ e ~extensive in ruralareas in2002 with over 40% of allruralhouseholdsand65% of the poorest uintile receiving transfers from this source -albeit with smaller amounts per household^!^ This i s an impressive coverage The urban coverage was much less -some 4% of households on average and 13% for the bottom quintile reflecting the only recent expansion of this program into urban areas. Administrative data suggests large further extensions in coverage since 2002. Inthis section we have looked at the structure and trends in incomes. The analysis showed the importance of a number of factors for incomes of the poor, includinglocation between rural and urban areas, education, labor force status, and remittances and transfers. The recent rapid growth in incomes of the poor (and all rural groups) found in ENIGHappears to be consistent with strong growth inwages inthe bottom parts of the distribution and the expansion of remittances and transfers, especially in rural areas. However, the observed growth in wages i s only the beginning of the story: future work will examine in greater detail the forces behind both urban and rural trends in labor earnings. C. Vulnerabilitf16 Overview The vulnerability of individuals andhouseholds to adverse shocks i s an intrinsic aspect of poverty (Chapter 1). Vulnerability can apply to many dimensions of living -to incomes, nutrition, health, social status and so on. It is important to distinguishbetween covariate (aggregate, economy-wide, common) and idiosyncratic (individual) shocks. The former affects everybody in a particularcommunity/regio~countrywhile the latter only affects a particular individual. In practice, even within well-defined areas, few shocks are purely idiosyncratic or covariate. Inthis section we present the results of a quantitative analysis of patterns of variation inincomes and consumption for the same households, drawing on two panel surveys (ENET and ENCEL27)-that followed the same households for a year or more. Ideally, we would like a map of mexican vulnerability to know not only who is poorer but also who is likely to become poor or become poorer and why. This can then form a basis for the policy question of how to shape public action to reduce the adverse '' 25 O~o~unidades increased its coverage of beneficiary families by 45 percent in the rural area and 232 ercent inurbanareas. This sectiondraws extensively on Maloney et a1(2003). 27For a description of this survey see Maloney et a1(2003). 71 costs of such vulnerability in a way that takes account of both sources and behavioral responses. Limitations imposed by data availability have constrained our efforts of constructing such a vulnerability map. In this section we present the results of some initial steps in that direction. There are always questions of measurement error for this kindof panel data, and for this reason an interpreting the results in this section needs to keep this inmind.28 First, we study how income shocks affect different groups of the urbanpopulation. We find that the less educated experience relatively less income variation than college educated, while informal self-employed and workers relatively more than their formal counterparts. Households with single men or women experience more variation but single mothers less than households with married couples and children. We then ask whether estimates of income variability innormaltimes are good predictors of income variability during aggregate shocks as the Tequila crisis in 1995, and find that relative patterns do not change significantly with the exception of disproportionately large falls amongst the college-ed~cated,~~Next, taking advantage of the introduction of rural households in the ENEU (renamed as ENET) from the 2nd quarter of 2000 we compare estimates of income variance for rural and urban households. We find quite similar patterns across these groups. Finally, we analyze the consumption variance of poor ruralhouseholds and its relationship with income variance using the ENCEL. This finds substantial variance for these households but significantly less for consumption than income -suggesting that even very poor householdsare able to smooth consumption, which i s consistent with McKenzie (2003) findings. Patterns of income change To illustrate the types of results obtained Figure 3.11 graphically portrays the overall pattern of changes in incomes for rural and urban households for 2000-03 with Table 3.18 providing the specific numbers. The quantities are changes in the log of incomes (that are close to percentage changes). Households were ranked according to this change in income from the largest negative change to largest positive change. Thus this shows that the median (middle) household experiences on average a very modest increase in income between one quarter and the same quarter the following year in both rural and urbanareas. But this hides a lot of variation. At the bottomof the shock distribution(20th percentile), households experienced an income fall of about 50%, while at the top of the shock distribution (80thpercentile) households experienced an income increaseof close to ~ 28 One source of error is due to attrition bias: recent work on US panel data by Neumark and Kawaguchi (2001) supports the view that the magnitude of attrition biases are not large. However, there remain questions of whether this result applies generally -including to the Mexican surveys- and there are other sources of measurementerror. 29 Several authors have examined the effect of the peso crisis on householdsincome, consumption, health, and education. For a comprehensive review of these results see McKenzie (2003) and Tolentino et al. (2003). McKenzie (2003) found that the peso crisis had an extremely widespread impact, lowering income and consumption for all age groups and education levels. However, the less-educated, ruraland agricultural workers experienced the smallest falls in income. In contrast, households living in metropolitan areas, highly educated household heads and workers in financial services and construction suffered the greatest declines. 72 50%.30Such changes reflect a combination of purely temporary shocks, long-term changes in incomes that happenedto occur inthe period and, indeed, measurementerror. While this cannot be sorted out from data this neverthelessprovides valuable information especially when combined with analysis of patterns of variation in incomes across householdswith different characteristics. For example, eventhis summary analysis shows that short-run income changes are significant in both rural and urban areas but slightly higher inruralareas inthis period. Figure3.11. The Pattern of Annual Change in Incomes in a Panel of Rural and UrbanMexicanHouseholds2000-03 (Size of changes for different percentiles of the distribution of changes inlog incomes) 1.5 f I - -.-8 1.0 e E 0.5 .-sP P S 0.0 Y 2 -0.5 II) 3 S -1.0 I -1.5 1 I 10 20 30 40 50 60 70 80 90 100 1 -c- Wral -A- Peereentit of Shock Urban 1 Source: WB staff calculations usingENET. Table 3.18. The PatternofAnnual ChangeinIncomesina PanelofRuraland UrbanMexicanHouseholds2000-03 Medianwithinshocks Average Std.Dev. 20" 3Uh 4Uh SOth 6Uh 7flh 8 8 9dh rodh ALL 0.078 0.678 -1.03 -0.48 -0.22 -0.01 0.05 0.13 0.25 0.41 0.68 1.17 RURAL 0.094 0.776 -1.23 -0.54 -0.26 0.00 0.05 0.14 0.29 0.47 0.76 1.39 URBAN 0.072 0.636 -0.98 -0.47 -0.22 -0.02 0.05 0.13 0.24 0.38 0.61 1.10 A Profile of V u l n e r a ~ iiniUrbanAreas in Normal and Crisis Periods ~ ~ Patterns of income shocks vary with the characteristics of households analyzed, Do younger or older, formal or informal, more or less educated groups experience higher variation in income? This is done by investigating how such variables influence the patterns of variation. It explores how income variations are influence by such characteristics relative to baseline patterns. A base category of households (or reference group) is (arbitrarily) chosen and then binary variables (dummies) introduced for 30Figure A.3.2 shows that in 2000-2002 poor households experienced positive shocks with a very low probability (high dispersion), while rich households experienced negative shocks with a very high probability (low dispersion). However, this result is not controlling for permanentincome. 73 characteristics that differ from these. In the analysis, the reference group selected was families headed by married, middle-age college-educated males working in the formal sector, with less than the mean number of children.31 Then binary variables were introduced for the education level of the head (primary incomplete, primary complete, secondary incomplete, secondary complete) for his or her age (one for those under 25 and one for those over 45); for more than the meannumber of children (1.3) inthe household, and for household structure (single mothers with children, single women without children, and single men without children). There are also binary variables for labor market status of the household head-informal self-employed, informal salaried- and a residual category that includes the head being out of the labor force, unemployed or otherwise not earning an income.32 Since we were interested in comparing normal and crisis conditions the analysis was undertaken for the pre-crisis period of the fourth quarter of 1992 to the third quarter of 1994 and compared with the crisis period between 1994 and 1995. Results are given in Table A.3.15 in the Annex. We use the convention of quoting results as "percentage" to give a clearer flavor of the results to the reader. The key findings innormal times (columns a-c inTable A.3.15) were as follows: First, less educated households -at least up to secondary incomplete- generally have lower dispersion in patterns of household income change. This may reflect lower possibilities for income growth. It may also reflect fewer opportunities to smooth income through credit markets or savings, so that families take measures to reduce or offset adverse income shocks, for example through putting additional workers in the labor market if the headloses hisherjob. Second, there i s significantly higher variation in income changes for the informal self- employed, the informal salaried and those households whose heads are not remunerated, on both the upside and downside. The self-employed also have a median income change below that of the reference group chosen (lower-than-base). Note that informal self- employed and workers also have lower average incomes than formal workers -as illustrated in Figure 3.14 in the previous section. Taken together, this seems consistent with the standard dynamics of a small firm sector: in any period some firms do exceptionally well and some less well than their salaried counterparts and, as i s the case inmany countries, firm mortality rates are very high.33This could be arguedto reflect a greater precariousness. But at least for the self-employed, other evidence finds that 31 Under 0.7 percentof the sample were single men with children whom we include in the basecategory. 32 We use the term `informal' to refer to those unprotectedby labor law and in turn this category is divided into two. First, ou'ners of firms with fewer than 16 employees who do not have social security or medical benefits are identified as informal self-employed -less than 1 percent of these firms have more than 5 employees. Second, employees in those small firms are identified as informal salaried workers. (Repetido; tainhiCnapareceen la p6gina 67). 33 See Levenson and Maloney (1998), Fajnzylber, Maloney and Montes (2003) find that the rate of micro firm owners returning to formal employment, one measure of mortality rates, is roughly equivalent in Mexico to that found in the US. 74 almost 70 percent of workers entering the sector from salaried work report do so voluntarily. Third, the same lower-than-base case growth in median incomes occurs for families headed by older workers. This may reflect a tendency towards declining income with retirement. The reverse logic may hold for families whose head shows no earnings at the beginning of the period. On average, many of those "unemployed" will findjobs so we would expect what we see, namely that the tendency of this group is to increase their incomes above the median of the referencegroup. Fourth,householdsheadedby single women do worse alongthe income distribution, and a similar pattern is visible for single men. This may suggest that jobs that young people hold ingeneral are those with fewer possibilities for big gains, greater possibilities for lay offs andmediocre performance at the median. Fifth, single mothers appear to have the same distribution of shocks as the reference group. While surprising at first sight, this finding i s consistent with anthropological studies of Mexicanfamilies. Selby, Murphy and Lorenzen (1990) find matrifocalfamilies have higher per capita incomes, proportionally more family members in the work force, lower dependency ratios, and that these families generally do as well as non-matrifocal household^.^^ Infact, Chant (1985) finds that "despite major structural constraints of the economic and social potential of matrifocal families, single parent units frequently fare better than male headedhouseholds" (p. 650). Let's now turn to the crisis period. There were, of course, large downward falls for all groups in the 1994-95 crisis with huge adverse effects on welfare -as reflected in the increasesinaverage levels of poverty discussedinthe preceding section. But here we are concerned with whether the pattern of relative income variability between different groups is different incrisis periods. This i s relevant to the design of safety nets designed to be usedduringcrisis. The key findings intimes of crisis (columns d-f inTable A.3.15) were as follows: Table A.3.15 (columns d-f) presents the estimates and significance levels of the same variables discussedabove (schooling, status inthe labor market, marriage status, age, and etc) but during the crisis period. The interpretation of these coefficients i s "compared to how a given group did relative to the base in `normal' periods, how did it do relative to how the given group didduringthe crisis?" 34Matrifocal householdsare not worse off because of the absence of a man, despite discrimination against women in the work force and the difficulties that women have getting well-paid employment. Although median household income is 14% lower than non-matrifocal households since they average one less member inthe householdtheir per capita incomes are 8.2% higher. They put almost as many membersinto the paid work force (1.38 vs. 1.4) and the ratio of dependentto members inthe work force is lower than the non-matrifocal. 75 The change inthe constantcaptures the shift for the referencegroup describedabove. Not surprisingly, what the findings reveal is a large shifting down of the distribution for this group with the median and the top tail (captured by the 80thquantile) following roughly equivalently, while the lower tail captured by the 20thquantile elongates significantly, meaning that some families suffered much larger declines probably reflecting the unusually large increaseinunemployment inMexico inthe crisis. There are no significant changes inthe pattern of changes for the self-employed (though they shared in the much larger average fall in income). In other words, the large shocks for the self-employed were in line with the reference group. This group experienced higherhighs, lower medians andmuchlower lows inbothgood times andbad. The same can be said for the informal salaried who work for them. Similarly, the distributions of the old and young,of those with above average sized families, single mothers and single men track the base case suggestingthat they also do not appear to have suffered any more or less ineither median or variance of income than the base case. The performance of single mothers is again somewhat surprising, yet consistent with Glewwe andHall's findings for Peru. The same cannot be said of those families with heads who earned no income. Their income fell sharply relative to the reference group inthe crisis period. Before the crisis, it couldbe that both median and the top tail of the conditional distributionwere previously higher than the reference group, perhaps reflectingthat the headoften got ajob acrossthe sample period. During the crisis this effect is reversed at the median and sharply attenuated at the top tail of the distribution perhaps reflecting the increased difficulty of getting ajob. Most striking are the results of less educated household heads comparedto the educated ones duringthe crisis.35For primary incomplete andprimary education, income falls only 23% compared to 30% for the base case. The less well-educated indeed appear to do better than they didinnormal periods relative to the college educated, experiencing lower lows, higher highs and, in the case of primary complete, a higher median. The premium at the median found for this group inTable A.3.15 is, infact, a feature of the crisis. There are two potential explanations for this. First, the premium to tertiary education tends to fall incrisis (Schady and Sgnchez-Phamo, 2003) and inMexico inparticular (McKenzie, 2003b and Chapter 5). There appears to have been a downward adjustment from the surge in high-end salaries before the crisis, for example in the financial sector. Second, there may be a greater propensity for poorer families to put extra workers in the labor force during crisis (Skoufias and Parker, 2003). The same effect appears although diminishedinmagnitude and statistical significance for secondaryeducatedworkers. 35 Several authors have examined the effect of the peso crisis on households income, consumption, health, and educationfor a comprehensivereview see McKenzie (2003a,b) and Tolentino et al. (2003). McKenzie (2003a) found that the peso crisis had an extremely widespread impact, lowering income and consumption for all age groups and education levels. However, the less-educated, rural and agricultural workers experienced the smallest falls in income. In contrast, households living in metropolitan areas, highly educatedhouseholdheads and workers in financial servicesandconstruction suffered the greatest declines. 76 Urban vs. Rural Income VuZnerabiZi~ Interms of designingsafety nets it is of great importance to know whether the shock patterns affecting urban areas are applicable to rural areas too. In order to explore this issue we take advantage of the modifications inthe ENEUfrom the second quarter 2000, which explicitly introduces rural areas36.Table A.3.16 presents a similar exercise to that undertaken in previous section, in this case, for both urban and rural households for the period 2000 to 2002 (which comprises a period of highgrowth in 2000 and stagnation in 2001 and 2002). This allows us to compare income variance inthe rural areas with what we know about the urban areas, The results for the urban population are very similar to thoseof the previous sections suggestingthat those findings are robust. The main message of Table A.3.16 is that results for rural areas are broadly similar to the urbanareas. Several findings merit attention. First, the overall shocks for the base group are strikingly similar. At the medianof the distributionthe base group shows increasesof income around 6-7%. Those householdsexperiencing the largest negative income shocks saw a fall in income of 31% and24% for urban and rural areas respectively. At the top of the distribution, both urban and rural areas presented an increase in incomes of around 42%.37 Second, informal and self-employed workers seem to show similar patterns. The lower lows of the self employed, and the higher highs of both are present in both samples. Hence the higher variance in the incomes of the self-employed and informal worker with respect to their formal counterpartsi s not unique to the urban world. Finally, we also found similar patterns, although with some differences, in the role that education plays on income shocks. In both urban and rural samples less educated household heads seem to present higher income increases in the median, which i s consistent with what we found above for the tequila crisis period. However, urban less well educated headed families seem to have lower income variance than the base group whereas the educational variables in rural areas behave pretty much like the self- employed and informal workers, indicating higher variance than the base group. Indeed when we run both samples together including interactive dummies for the urban households (not shown) we find the only significant differences are among the less well educated. And the message is that rural households with less educated householdheads have higher income variation with respectto the basegroup than their urban counterparts. Incomeand Consu~~tionVuZnerabiZi~In Poor RurhZAreas 36Although the introduction of rural households in the ENEU, renamed ENET, is an improvement important deficiencies are to be mentionedinthe collection of the data. The most important being that rural income does not include self-consumption which should be accounted as income too. This casts some doubts over the validity of rural incomeas well as its comparability with urban income. 37The magnitudesof the coefficients should be taken as suggestive as we have not yet applied Kennedy's correction for semi-logarithmic equations. Inferring conclusions about household vulnerability based on income data has major limitations for welfare analysis. If households are able to smooth consumption through borrowingor savings, income shocks do not necessarilyreflect changes inthe well-being of the household. In Mexico, the only panel survey with both income and consumption data i s the ENCEL survey of a subset of rural households. This was developed for the analysis and monitoring of PROGRESA in the 1990's, and so covers especially poor rural areas. This roughly corresponds to the lower parts of the rural distribution analyzed above. Where the EIWT measured changes across a five-quarter period, here we follow families from October 1998 to November 2000. Two sets of results are presented. The first uses the entire sample of both control and PROGRESA treated groups, putting dummies for those observations that change status either in number of children, structure of the household and usagelownership of land. An overview of the average patterns is provided inFigure 3.12, with the data provided in Table 3.19. There is again significant variation in experiences. Two features are noteworthy, First, the range of consumption changes is less than income changes (shocks) -especially for largest falls andrises. This i s consistent with households consumption smoothing especially for large negative and positive shocks. Second, the profile of income changes lies above that of consumption changes everywhere above around the 20fhpercentile of changes. This may be because incomes were systematically above expectations across the whole range so that households didnot spendall the increases. Figure3.12. The Patternof Annual ChangeinIncomesina Panelof Householdsin Poor RuralAreas, 1998-2000 (Size of changes for different percentiles of the distributionof changes inincomes) 3 2 I a 8 c v1 d2 0 r" 1 -2 -3 10 20 30 40 50 60 70 80 90 100 Dmileof Shock +Income +Consumption Source: WB staff calculations usingENCEL Survey (October 1998andNovember 2000). Table3.19. The Patternof AnnualChangeinIncomesina Panelof Mexican HouseholdsinPoorRuralAreas, 1998-2000 Medianwithin percentiles Average Std. Dev. IOth 20'" 3gh 40"' 50'" 70"' SO'" 90"" 100'" 78 INCOME 0.19 1.46 -2.35 -0.83 -0.37 -0.1 0.09 0.24 0.42 0.71 1.28 2.72 CONSUMPTION -0.16 0.64 -1.19 -0.75 -0.52 -0.36 -0.21 -0.07 0.08 0.24 0.43 0.8 Source: WB staff calculationsusing ENCELSurvey (October 1998andNovember 2000). The analysis of the effects of household characteristics finds for both models evidence that the less educated tend to suffer greater shocks in consumption across the board. At the median the least well educated appear to have experienced a fall of consumption of about 10percentagepoints. At the bottomof the distribution of changesthe base category experienced falls in consumption of 45% and those less well educated of the order of 55%. Similar magnitudes at the top of the distribution of changes were found for single men and single women. Households headedby older workers also experienced negative effects compared to the base group with about a 5-percentage point larger fall in consumption at the 20* percentile. These would be the groups that could be considered most exposed to consumption falls. Since they also broadly correspond to lower income groups they would, indeed, be the more vulnerable. This qualifies the results from analysis of incomes in the previous sections that found that lower educated groups were less vulnerable to income declines than others, and likely to experience larger increasesin the median and inthe top of the distribution. The same i s true for older household heads whose results did not differ from the base group. For the sample of householdsin poor rural areas, younger, better-educated workers seem to have better mechanisms to smooth consumption. Also strikingare the groups who showed differences inincome shocks with respect to the base group but do not present significant deviations in the c~nsumptionregression. This i s the case for informal salaried workers who appear to be slightly better at the lower tails than the base and the informal self-employed show exactly the same pattern of consumption shocks as the base or reference group. That is, the higher variance in incomes that wefind in both the rural and urban areas does not appear to translate into higher consu~ptionvariance. Again, consumption-smoothing mechanisms seem to be at play for informal workers. Finally, single mothers are not particularly adversely affected with respectto the basecase. This section provided quantitative information on patterns of income variation which is one input to assessinghow important vulnerability i s for the poor and for those at risk of falling into poverty. This represents a significant fleshing out of the profile of well-being, but is only the beginning of the work. Insum, the results suggests (i) the impact of that macroeconomic shocks -which by and large hit urban areas hardest- have been relativelyequally distributed across the urbanpopulation, and (ii) that inruralandeven in poor areas there i s at least some smoothing of consumption in the face of idiosyncratic income shocks. These results are consistent with McKenzie (2003). Several caveats are important to bear in mind. First, and perhaps most importantly, just because Mexico adjusted through evenly distributed income reductions in 1994 through broad wage reductions across all households and not via unemployment, this does not mean that future shocks will be transmitted the same way in Mexico. In fact, given today's macroeconomic circumstances and prevailing low inflation rates in Mexico which would preclude broad, real wage adjustment via price increases, Colombia and 79 Argentina may be good examples of how mexicans might manage shocks in the future, with very different implications for social protection policy. For instance, now that Mexico has single digit rates of inflation will the economy adjust more to shocks through unemployment than real wages as has been the case in the past? If so, will that unemployment be concentrated more in the poor or in the well off? Another problem is that, even though we find that income falls are not always larger for the poor, welfare falls are richer households can afford income declines better than poor households. For example, even if we found that in the crisis the poor lost 15% and the rich 2596, there is still an argument for aidingthe poor. Second, future work will need to include a mapping of the changes in consumption and income on to relative positions inthe consumption or income distribution. Evenif income shocks led to identical consumption losses, poor families may be less able to tolerate these than the better off particularly if they are close to subsistence. Then there is needto relate this to categories of shocks and how this relates to both existing and possible alternative categories of public action that can reduce adverse effects of vulnerability in an efficient fashion. D. Regionalandethnic differencesinwell-being Overview The distribution of well-being in a population i s not only about the economic characteristics of individuals and households. There i s extensive literature -in economics and other disciplines- concerning the relevance of group-based characteristics to the determination of incomes and other dimensions of well-being. Two salient aspects of this concern geography and ethnicity. Living in a poor area or membership of a different ethnic group can make a difference -sometimes a profound difference- to life prospects. History matters to this with current patterns of physical and human assets, as well as institutional structures, reflecting the past evolution of social interactions, political and economic developments. This is perhaps especially the case where current social difference flow from histories of domination and subordination as is the case of indigenous groups inLatinAmerica. Inthis sectionwe present the results of a spatial analysis bothof the pattern of well-being and changes in the past decade or so. The principal instruments of analysis for spatial conditions and developments are disaggregated ma s of conditions at the municipal level built primarily from the 1990 and 2000 censuses!' While geography and ethnicity are distinct dimensions, there are overlaps. The greatest concentration of indigenous groups identified by language spoken at home is in rural areas in the southern states of Mexico and the Yucatan peninsula with more specific other concentrations in other parts of the Mexico, for example inparts of the Northwest. This only provides a sketch of the issues that is intended to show that both geography and ethnicity matter for well-being and need to be taken account of in policy diagnosis and design. It also shows that this is only part 38These databases were preparedbyMariaCaridadAraujo. 80 of the story. Interms of numbers there are more of the extreme poor living outside the poorer southern states and who do not speak an indigenous language at home. For more in-depth analysis of conditions in the three poorest states of Mexico (Chiapas, Guerrero and Oaxaca) the reader is referred to the World Bank's report on the Southern States (World Bank, 2003~).There is a wealth of work on indigenous issues in Mexico; for a perspective from the World Bank see the Indigenous Peoples Profile (World Bank 1999a). Spatialheterogeneity and well-being Mexico i s characterized by great diversity within and across regions and states in its socioeconomic outcomes, assets andethnicity. There is a generalized gradient from North to South, but this is combined with considerable heterogeneity within any region or state. The highlevel of deprivation inthe Southis also confirmed with both CONAPO's index of marginality (based on access to basic infrastructure services, housing conditions, education attainment, and wage earnings) and the UNDP's Human Development Index (based on per capita GDP, educational achievement andenrollment, andlife expectance). For both indexes, the three southern states have the lowest rankings of among all states (World Bank, 2002a). Before turning to the more disaggregated analysis we illustrate the pattern at the level of regions following CONAPO's categorization, using measures of income poverty that are directly comparable with those discussedat the national level in Section 3.3, drawn from the ENIGH 2002. However, since the ENIGHseries is not designed to be representative at the regional level the numbers are approximate and standard errors are presented. The regional estimates can be interpreted as approximate measures of the true poverty incidence; state-level estimates are, by contrast, too unreliable to be meaningful (see LEipez Calva and Rodriguez Charnussy, 2003). The highest poverty incidence i s in the rural areas of the South Pacific states Xhiapas, Guerrero and Oaxaca- where still close to 50% of the population lives inextreme poverty, followed by the South Gulf and Caribbean region, where around 35% of the population are extremely poor (Table 3.20). This compares with 19% in the Center, 10% in the North and 4% in Mexico City. In terms of absolute numbers it i s important to note that there are large groups of the extreme poor living outside these south regions. For example, over a quarter of all the extreme poor inMexico lives inurbanareas inthe Center states. Note that poverty estimates include large standard deviations, particularly in the South- Pacific. This means that South-Pacific poverty estimates are around 41 to 52% with a 95% confidence. However, the standarderrors do not change the rankingof the regions in terms of povertyheadcounts. From 2000 to 2002, extreme poverty fell significantly in the South Pacific (6 percentage points), the Center (5 percentagepoints), and inthe South Gulf (5 percentagepoints). The north and Mexico City did not experience a significant change. Moderate poverty fell significantly only in the center (five percentage points) and rose slightly in Mexico City 81 (2 percentage points). However, it is premature to assess whether or not these poverty trends will continue. 2000 2002 2000-2002 Region Poverty Standard Poverty Standard Change in IHeadcount Deviation IHeadcount Deviation I Poverty South-Gulf 39.6 2.9 34.7 2.1 -4.942** South-Pacific 52.5 4.0 46.4 3.3 -6.050**" Center 23.8 1.7 18.6 2.3 -5.174*** North 10.5 1.1 9.7 1.7 -0.780 Mexico City 4.8 1.4 4.2 3.1 -0.616 South-Gulf 68.4 3.0 68.4 2.3 -0.005 South-Pacific 74.3 2.3 75.9 1.7 1.565 Center 58.5 1.o 53.0 1.6 -5.4 lo*** North 36.8 0.8 37.6 1.4 0.846 Mexico City 25.8 0.8 27.9 2.5 2.043" A much richer and more complex picture appears when we turn to lower levels of disaggregation. Figures 3.13-3.16 illustrate such diversity. They depict the spatial distribution of four socioeconomic variables across states and municipalities. Data comes from the 2000 population census. Colorslshading were defined based on quintiles of the distribution of the variable in question for municipality-level variables. The same color scale was kept for the state-level maps. Figures 3.13-3.16 show that these outcomes present more variation within states than between them. Moreover, they also illustrate the presence of clusters of contiguous municipalities with similar socioeconomic outcomes that do not correspond to the states administrative borders. Figure 3.13 depicts the distribution of schooling as measured by the average grade completed by adults across the Mexican territory. The state-level map shows a clear-cut divide betweenNorthern and Central Mexico -where schooling i s high- and the South of the country where it is low. However, as we move to a lower disaggregation, we find large heterogeneity throughout, While the municipalities that are on the US-Mexico border seem to exhibit, on average, higher schooling of between 7-12 years of schooling, there are also places with comparably high schooling outcomes to Central Mexico and the Yucat5n peninsula. The more disaggregated map also shows that there are many municipalities with very low average levels of education (below 4 years), especially in the South, but also inthe Center andpartsof the North. Figure 3.13. Average grades of schoolingat the state (left) and municipality (right) level (among those age 15 and older) I ' *,4487-599YI , 599-667 m-1209 Source: WB staff calculations from a municipal data base from Census and other sources. Figure 3 14 illustrates the proportion of indigenous population. Although the state-level figure suggests a relatively modest presence of indigenous peoples throughout the Northwest, South, and Center of the country, the municipal-level disaggregation shows the presenceof across-the-state-border-clusterswhere the presence of indigenous i s high. These are mainly inthe South, but also with clusters inChihuahua, Durango, andNayarit. 83 Figure3.14. Proportion of population ages 5 and older that speaks an indigenous languageat the state (left) and municipality (right) level Source: WB staff calculations from a municipal database from Censusand other sources. Figure 3.15, in turn, portrays the distribution of household durables ownership across states and municipalities. While the state-level map looks like the one for the distribution of schooling in Figure 3.18 in that it has a North-South divide, at the municipality level we see similarities with the map in Figure 3.19 suggesting that regions with a high presence of indigenous population are also places where households own less durable goods. Figure 3.15. Proportionof householdswho own noappliancesat the state (left) and municipality (right) level Note: Appliances include a radio, a TV, a VCR, a blender, a fridge, a washer, aphone, a water heater, a car or pickup truck and a computer. Source: WB staff calculations from a municipaldatabase from Censusand other sources. 84 Finally, Figure 3.16 representsthe percentageof workers earning less thantwo minimum wages both at the state and municipallevel. Inthis case, both maps show a clear regional pattern as the concentration of workers with low wages is increasingly denser in the South of the country as well as inthe YucatAnpeninsula. Figure3.16. Proportion of employed who earn less than two minimum wages at the state (left) and municipality (right) level 071 011 081 -059 089.099 Source: WB staff calculations from a municipal databasefrom Censusand other sources. Spatial heterogeneity over time We next turnto the question of change through time. Of particular interest is the question of whether poorer areas are catching up with or converging with richer areas. Inthe past decade the picture i s quite varied depending on the indicator. This i s illustrated by Figure 3.17 that uses municipal-level census data on each of the seven components of the CONAPO marginalityindex plus the percentage of adults (15 and older) who completed secondary school (Le. a total of nine years in Mexico). It plots changes on each of the variables between 12000 and 1990 (in the vertical axis) and their initial level in 1990 (in the horizontal axis). All of the variables (except the percentagewith secondaryeducation) can be thought of as dimensions of poverty and thus a decrease represents an improvement on average well-being ineach municipality. 85 Figure3.17. Changes on socioeconomic variables over the 90's in Mexican municipalities ADULTILLITERACY HOUSEHOLD WITHOUT PIPEDWATER i o - 1 lM) i, -30 , , , , - 1 0 0 ~ I 167055 80 !literacymi* 860711 100 90% hhwthoulpipedwater HOUSEHOLDWITHOUT ELECTRICITY HOUSEHOLDWflHOUT SEWERAGE 20- i , -i I 90% hhwiihouteieancity Ibo . 3Mna 90%hhwithout sewerage 5 0 ~ OCCUPANTS PER ROOM SHAREEARNINGLESS THAN 2 MINIMUMWAGES 3 1 2o i '1 0 Ibo 90aug mcupantslrmm 4 77101 32.9445 90%X eamrng<2 min.wages HOUSEHOLD DIRTFLOOR WITH ADULTS WITH SECONDARY EDUCATION 20J I B O i * 16- 9 - 12- p 8 - c b 8 4 - 0 - I 0 90%wisec. sduc. 28 5968 Source: WB staff calculations from a municipaldatabase from Census and other sources. 86 Figure 3.17 suggests a negative correlation between changes in illiteracy and the initial level. This means that improvements in well-being were relatively larger in places that were initially worse-off. A similar pattern is observed for lack of water and lack of electricity while there i s no clear pattern to changes in sewerage relative to initial levels. These patterns are consistent with convergence for those publicly provided services in which initial access was high (as with water and electricity) but an ambiguous pattern where coverage was relatively low (sewerage). This is consistent with other evidencethat public services typically work down the distribution, crowding in relatively disadvantagedgroups as access approaches universality (see De Ferranti et al, 2004). B y contrast, the pattern for completion of secondary schooling is clearly divergent. Inthis case, bigger increases occurred in places where the initial level of the variable was also larger. However, the pattern i s less clear for other variables such as the percentageof the employed population making less than two minimum wages or the average number of occupants per room in a dwelling. In these two cases no clear pattern of association between changes inthe variables andtheir initial level can be detectedinthe figures. One last variable that doesn't have an obvious pattern i s the percentage of dwellings with dirt floor. While a negative association is observed for localities with very low initial levels of prevalence of this type of dwellings, the relationship is attenuated and even reverses among the municipalities that are less well-off inthis dimension. Interestingly, of all the components of the CONAPO index, these three are variables represent assets which are, for the most part, not providedby the government. These findings can be complemented with those of authors who have explored the presence of convergence in different socioeconomic outcomes across Mexican states. Esquivel (1999) looks at per capita income across states and finds the presence of convergencebetween 1940and 1980followed by aperiodof stability duringthe 80's and divergence in the 90's. Esquivel's findings are consistent with those of Messmacher (2000). Garcia-Verd6 (2002a) in an analysis for 1940-2000 finds considerable persistence over time in the ranking of states with respect to the distribution of per capita income, much less so for the distributions of infant mortality and adult literacy. Of these three variables there i s convergenceacross Mexican states only interms of adult literacy. Finally, a series of studies have lookedat questions of whether the spatial distribution of wages has converged or diverged. These mostly find a pattern of divergence after the period of trade and foreign investment liberalization which started in the late 1980's and was consolidated under the NAFTA agreement. Inparticular, a recent analysis by Hansen (2003) finds patterns of divergence associatedwith skill levels, foreign direct investment andgeographic location. 87 Correlatesof changesin well-being To explore the correlates of changes in well-being we reproduced the CONAPO marginality index at the municipality level using only three of its components, the ones that can directly be associated to private -as opposed to public- provision: employed adults earning less than two minimumwages, households with a dirt floor and average number of occupants per room. We computed this index of marginality using principal- components for 1990 and 2000. The difference between the values of the index between the two years i s what we refer to as "changes inmarginality" or "changes inwell-being". We first compare a series of characteristics across municipalities of different sizes: urban (2000 population greater than 15,000); semi urban (2000 population between 2,501- 15,000); andrural(2000 population of 2,500 or less) inTable 3.21.39 Table3.21. Municipalattributes, by2000classificationof municipalities accordingto population Urban Semiurban Rural U S S=R U=R Marginality Decrease -7.48 -7.77 -7.81 *** * Population in2000 86,873 7,597 1,362 *** *** Percapitamunicipal expenditures3 677.80 795.60 1,306.00 *** *** *** Share of illiterate adults` 19.80 25.50 25.40 *** *** Share who speak indigenous language' 12.21 26.33 34.07 *** *** *** Shareof employment inmanufacturing' 14.79 11.68 10.35 ** ** Share of employment inservices' 37.84 24.14 19.96 *** *** Share of population served by state road2 57.12 44.11 25.03 *** *** *** Kmto closesteconomiccenter 93.95 127.85 188.22 *** *** *** Coastaldummy 0.12 0.06 0.03 *** *** *** Northern border dummy 0.02 0.01 0.01 ** ** Observations 1,018 998 381 1.In 1990 2. In 1995 3. In 1999, with respectto 2000 population Significant at: *** 1%, ** 5%, *10% Source: WB staff calculations from a municipal database from Census and other sources. Table 3.21 shows that basedon our index, the average change in marginality during the decade didn't diverge significantly across municipalitiesof different sizes. Basedon their size, Mexican municipalities differ in several aspects. The more rural a municipality, the 39 According to the municipal-level data used in this exercise in 1990 at the national level 22% of the population was indigenous and the 2000 reportsl8%. A difference between the CENSUS result and the dataprovidedhere is that inthe latter we are taking averages across municipalities (Le. each municipality is weighted equally, independently of population size) therefore this places a lot of weight in the southern states of Oaxaca and Chiapas where there a lot of small municipalities and the largest indigenous population). 88 more likely it i s to have larger per capita municipal expenditure^^^, more illiteracy, higher percentage of indigenous population, less employment inthe manufacturing and services sector, less access to roads. Rural municipalities are more isolated from centers of economic activity4*,from the coast and from the US-Mexico border. Also, while different from urban municipalities in some aspects, semi-urban and rural municipalities are similar among themselves in their illiteracy rates, their access to manufacturing employment and their proximity to the US-Mexico border. Based on the changes in marginality that occurred over the decade, municipalities were classified into two groups: municipalities with high decrease in marginality and municipalities with low decrease in marginality ----or even increase in it (the cutoff being the median decrease). Summary statistics for a series of municipal characteristics across these two groups are presentedin Table 3.22. The table separates urban, semi urban and rural municipalities. Table 3.22. Municipalattributes, by2000-1990marginalityreduction Urban Semi urban Rural Mg. Reduction Mg. Reduction Mg. Reduction High Low Diff. High Low Diff. High Low Diff. Marginality Decrease -10.10 -4.85 *** -10.71 -4.83 *** -11.49 -4.11 *** Population in2000 91,024 76,343 * 7,356 7,838 ** 1,330 1,395 Per capita municipal expenditures3 734.6 621.0 *** 890.8 700.3 *** 1,241.0 1,372.3 Share of illiterateadults' 18.80 20.70 ** 23.75 27.16 *** 27.05 23.70 ** Share who speak indigenouslanguage' 10.51 13.91 ** 24.19 28.47 * 36.28 31.84 Share of employment inmanufacturing' 14.94 14.65 11.05 12.31 8.44 12.27 Share of employment in services' 39.75 35.94 *** 25.64 22.65 *** 17.20 22.73 *** Share of population served by state road2 60.06 54.17 ** 46.82 41.41 ** 22.47 27.61 Kmto closesteconomiccenter 91.14 96.76 129.90 125.80 192.30 184.10 Coastaldummy 0.07 0.17 *** 0.06 0.06 0.03 0.04 Northern border dummy 0.03 0.02 0.01 0.01 0.02 0.01 Observations 509 509 499 499 191 191 1.In 1990 2. In 1995 3. In 1999, with respectto 2000 population. Significant at: ***I%, ** 5%, * 10% Source: WB staff calculations from a municipaldata basefromCensus and other sources. A first pattern illustrated in this table is that -unlike urban and semi-urban municipalities- there are few differences in the attributes of rural municipalities with high and low decreases in their indices of marginality. Specifically, high decreases in 40 The municipal expenditures data comes from the CEDEMUN data set. It is the total expenditure (administrative, public works, subsidies and other) for each of the municipalities in 1999. From the revenues side, it includes -among others- the following accounts: taxes, federal participations, rights, production and others. 41By center of economic activity we refer to those municipalities with a city of at least250,000 people and in the top 33% of the country's distribution in terms of their employment in manufacturing or in services. This definition suggests thereexist a tota1of 56 economic centersthroughout Mexico. 89 marginality are more frequent in places with high illiteracy and less employment in services. This suggests that the variables explored above are not the best choice of correlates of changes in poverty for rural municipalities. It would be important to include variables related to the local agricultural productivity, access to land and to other agricultural-related assets -this will be the subject of future work on ruralpoverty. An important pattern that is illustrated in Table 3.22 is that the urban and semi-urban municipalities with high decreases in marginality differ from the less successful ones in similar aspects. Specifically, urban and semi urban municipalities with high decrease in marginality have, on average, more investment (higher per capita municipal government expenditures) and better access to assets and infrastructure (less illiteracy, more roads, more employment in services). Interestingly, they also have a smaller indigenous population. A difference between urban and semi-urban municipalities that succeededin reducing marginality i s that, among the urban ones, it i s the largest ones interms of their 2000 population that achieved the biggest decrease in marginality, while the opposite is true for semi-urban municipalities. Note, however, that these cross tabs are only measuring association between the variables (in a multivariate regression some of the variables might not becameimportant) as causality could go inbothdirections. Finally, figure 3.18 shows maps of Mexican municipalities, distinguishing -among those where poverty42 i s high- the ones with high and with low employment in manufacture and services. We see that among the municipalities where poverty i s high, there are more that have a low share of non-agricultural employment than there are with a highshare. These maps also suggest that there exist large contiguous areas along Mexico that have similar employment andpoverty outcomes. Figure 3.18. Poverty and share of non-agricultural employment (Highrefers to top third) I I I Source: WB staff calculations froma municipal databasefrom Census and other sources. 42 As measuredby the CONAPO marginality index, includingall of its components. 90 Ethnic differences in well-being To explore the relationship between ethnicity and well-being, we first look at patterns of poverty for indigenous and non-indigenous individuals across the whole population and then turnto a spatial analysis, basedon concentrations of indigenousgroups. Since the ENIGH does not have a question on ethnicity it is not possible to estimate poverty statistics usingthe concepts and surveys that are the standard sources of poverty for Mexico andused in Section 3.2. However, the Census has a more limited question on labor incomes that can be used to generate indicative numbers. As Table 3.23 illustrates, 44% of indigenous groups are in the bottom 20% of the overall population distribution, and 80% in the bottom 50%. Usingthe bottom quintile as a measure of extreme poverty (this i s close to the incidence under the food-based poverty line from the ENIGH), indigenous peoples account for about a fifth of the extreme poor, that is over twice their population share according answers to the question on language at home in the 2000 Census. [ Proportion who are in: Category Headcount Poverty Gap SquaredPoverty Gap Bottom20% Indigenous 43.8 36.9 34.9 Non-indigenous 17.5 15.8 15.4 Total 20.0 Bottom50% Indigenous 79.3 58.3 49.9 Non-indigenous 46.9 27.8 22.0 Total 50.0 Top 10% Indigenous 2.3 Non-indigenous 10.8 Total 10.0 ~~ Note: Percentilescalculatedon the basisof per household wage income. Source: WB staffcalculations froma 2000Census. The 2000 Census indicates that close to 60% of the indigenous households did not have sewerage, 33.5% did not have access to potable water and 78.7% live in dwellings of only one room. With respect to education, net enrollment rates for educational levels beyond primary decrease dramatically for indigenous groups particularly inthe south part of the country. The lower secondary net enrollment rate is only 35% for the indigenous compared to 60% nationally. Moreover, 24% of the population that speaks an indigenous language in the south does not attend school in the age group 12 to 14 years old. The maincausefor dropping out of school is lack of financial resourcesto continue studying. More than 68.4% of the indigenous 12years and older work. The adverse gradient in access to services and education for indigenous groups also applies to health. Mortality levels are higher in the states with a high concentration of indigenous population (World Bank 2003~).More striking are differences in service levels. Coverage of formal health insurance i s substantially lower -a consequenceof the lower proportion of households attached to formal work. The M S S coverage rate in the 91 south (Pacific) is only 796, compared with 26% inMexico City and 31% in the north. In the south (Pacific), 89% of the indigenous population i s not covered by formal insurance systems (Table 3.24). Table3.24. Coverageof HealthInsuranceby Regionand IndigenousStatus South(Pacific Ocean) MexicoCity National Variable Speaks indigenous language Speaksindigenouslanguage Speaksindigenouslanguage Yes No Total Yes No Total Yes No Total IMSS 7.0 15.2 13.0 26.1 37.8 37.5 12.3 34.8 33.2 ISSSTE 2.7 7.4 6.1 8.1 13.5 13.4 3.1 6.4 6.2 PEMEX 0.3 1.1 0.9 1.2 1.3 1.3 0.4 1.1 1.1 Other 0.0 0.5 0.4 0.4 0.8 0.8 0.2 1.2 1.1 No insurance 88.7 74.7 78.4 63.0 45.7 46.0 82.5 55.5 57.4 Source: WB staffcalculations from XI1Censo de Pobhcidn y Viviendu 2000 (Survey). Turning to the municipal-level spatial analysis, regions with a higher presence of indigenous are still laggingbehindinmany socio-economic indicators. Moreover, data on their evolutionover time i s notconclusive on whether they are likely to converge towards the national average. Table 3.25 illustrates such differences. After ranking the municipalities according to the percentage of indigenous we compare the top quintile (i.e. those with 53% or moreindigenous) to the rest of the distribution. Table3.25. Municipalattributes,bypresence of indigenous Share who speak indigenous language' All5.85 others Top 20% Diff. 84.50 *** Population in2000 47,494 10,861 *** Per capita municipal expenditures3 888.5 580.1 *** Share of illiterate adults' 18.6 40.7 *** Share of employment inmanufacturing' 13.55 9.75 *** Share of employment inservices' 32.53 16.36 *** Share of population servedby state road' 51.39 27.43 *** Kmto closesteconomiccenter 113.79 160.15 *** Coastaldummy 0.08 0.07 Northern border dummy 0.02 0.00 *** 2000-1999changes Shareof illiterate adults -3.78 -7.91 *** Share indwellings without piped water -6.91 -9.46 *** Share indwellings without sewerage 3.93 14.06 *** Shareindwellings without electricity -11.31 -19.11 *** Average occupants per room 0.16 0.4 1 *** Shareindwellings with dirt floor -9.15 -12.43 *** Shareemployed earninglessthan two minimumwages -9.31 -4.90 *** Shareof adults with secondaryeducation 7.14 5.33 *** Observations 1,918 479 1. In 1990; 2. In 1995; 3. In 1999, with respectto 2000population. Significant at: *** 1%, ** 5%, * 10% Note: During the pastdecade, seweragecoveragegrowth was low with respectto population growth at the municipal level with mightexplain the positive signfound inthe table. Source: WB staff calculations from a municipaldatabasefrom Census and other sources. 92 We find that municipalities with a highpresenceof indigenous are, on average, smaller in population. Despite the national pattern towards higher per capita expenditure in smaller rural -as compared to urban and semi-urban- municipalities observed in Table 3.25, municipalities with a large indigenous population have a lower per capita municipal expenditure. They also have a higher percentage of illiterate adults, significantly less employment in manufacturing and services and they are more isolated, less served by roads, and less likely to be on the US-Mexico border. Table 3.25 also compares the changes over time of the components of the CONAPO marginality index, as well as of the percentageof adults with secondaryeducation across these two groups of municipalities. We find that municipalities with a highpercentageof indigenous improved faster than the rest in terms of illiteracy, piped water, electricity, andmaterial of the floor of the dwellings. Note however that the first threeof these refer to variables that tend to become of universal access over time. Furthermore, this table also shows that municipalities with a highpresence of indigenous did significantly worse than the rest in other aspects such as access to sewerage and crowdedness of the dwellings, earnings, and averageschooling of the adults. Net migrution a ~popu~tiongrowth d Mexico has experienced important demographic changes in the last twenty years. While domestic migration appears to be correlated with the availability of public and private assets and incomes we find that regions with a high proportion of incoming residents were relatively less successful inimproving their socioeconomic indicators over time. While in 1980, 35% of Mexico's 2,400 municipalities had more than 15,000 inhabitants (the cutoff population to be classified as urban according to CONAPO~~), in 2000, 42% of the municipalities were urban. Not only the average municipality became larger but in general there was increasing concentration of population in the urban areas. While in 1980, 87% of the population lived inurbanmunicipalities, in2000 this number increased to 92%.44 This, of course, came with a reduction in the percentage of people living in rural (from 0.96% to 0.54%) and semi urban (from 12.4% to 7.9%) municipalities in the same period. The number of localities increased in rural areas? I s there more population dispersion in ruralareas?We ask this because the GOM has insisted that a challenge is increasing populationdispersion that prevents them from effectively delivering programs and basic services. Figure 3.19 illustrates the distribution of the share of municipal residents non-native to the municipality or state where they lived in2000. This variable characterizes the relative "attractiveness" of different areas for domestic migrants. The map suggests that the regions that have a higher percentage of domestic migrants are located close to the US- Mexico border, to the larger cities (Mexico City, Guadalajara and Monterrey), and in 43 Note however that we are applying the CONAPO cutoff to municipalities rather than localities, which leads us to larger figures in terms of the total percentage of urban population in the country since rural localities inmunicipalities with more than 15,000 inhabitants are not accountedas such. 44 Note that inurban municipalities there might beruralindividuals. 93 tourist areas of the coast (both the Pacific and Yucatiin). Similarly, the places that seem less appealing to migrantsare clustered inthe poorer southern states of Guerrero, Oaxaca and Chiapas, as well as inthe dry Chihuahua. Figure 3.19. 2000 Share of populationborn ina different municipality andlor state 0Mexico %domesticmiarants Source: WB staff calculations from a municipal database from Census and other sources. Table 3.26 compares the top twenty percent of municipalities based on their percentage of municipal migrants in 2000 to the rest of municipalities in the country. By municipal migrants we refer to those who reside in a different municipality -but not a different state- than the one where they were born.45 Based on this definition, we find that municipalities with a high share of migrants are those that are overall better off in terms of levels (though not necessarily those that experienced the most rapid gains -see below). On average, they have bigger populations, higher per capita public expenditure, less indigenous population, less illiteracy, more employment inservices and manufacture, better access to roads and are closer to both economic centers and to the US-Mexico border. 45Although not reported here, a similar exercise that accounts for domestic migrants that traveled not only within the state but across the states reveals similar patterns in terms of the characteristicsof the locations with a higher percentageof non-native Mexican residents. 94 Table 3.26. Municipalattributes,by share of municipalmigrantsin2000 All others Top 20% Diff. Population in2000 33,260 67,825 *** Per capita municipal expenditures3 777.6 1,023.8 *** Shareof illiterate adults' 25.6 12.6 *** Share who speak indigenous language' 25.6 5.5 *** Share of employment in manufacturing' 11.3 18.7 *** Share of employment inservices' 25.96 42.54 *** Share of population servedby state road2 42.49 63.13 *** Kmto closesteconomic center 127.14 106.28 *** Coastaldummy 0.08 0.08 Northern border dummy 0.01 0.05 *** 2000-1990 changes Share of illiterate adults -5.03 -2.90 *** Share indwellings without pipedwater -8.26 -4.09 *** Share indwellings without sewerage 6.79 2.58 *** Share indwellings without electricity -14.04 -8.19 *** Average occupantsper room 0.24 0.10 *** Shareindwellings with dirt floor -10.49 -7.10 *** Share employed earning lessthan two minimumwages -7.32 -12.81 *** Shareof adults with secondaryeducation 6.56 7.67 *** Observations 1,917 479 1.In 1990 2. In 1995 3. In 1999, with respectto 2000population. Significant at: *** I%, 5%, * 10% ** Source: WB staff calculations from a municipal databasefrom Censusand other sources. If we look at how these two groups of municipalities performed over time we find patterns that are consistent with local inability to meet the needs of the incoming population. Specifically, municipalities with a highshare of migrants were less successful inreducing their gaps inilliteracy, inthe access to pipedwater or inelectricity. They also made less progress than the rest of the country interms of reduction of the percentage of the population living in homes with soil floors. Between 1990 and 2000, municipalities with high percentages of migrants experienced an increase in the percentage of people without sewerage as well as inthe averagecrowdednessof the dwellings. However-and this is probably relatedto the appealof these places andthe type of people they attract- municipalities with a high share of migrants also experienced larger increases in the average secondary schooling of adults and they had bigger reductions in the percentage of workers earning lessthan two minimumwages. An additional evidence of the changes in population distribution across mexican municipalities i s presented in Figure 320. It illustrates the population growth between 1980and 2000 includingboth natural growth as both as net migration and distinguishing between places that experienced a decrease in their population, as well as those where population growth was low and high. The figure shows that the municipalities where recent population growth has been the largest are spread across different regions of the 95 country: from the Yucath peninsula to the states of Baja California, Mexico, Guanajuato and Puebla. Places with negative population growth are also distributed throughout the country but there are a large number of them in the Northern-Central region of the country (inareas of Durango, Zacatecas, and Chihuahua). Figure3.20. 1980-2000populationgrowth growth Negatwe Positive](XI Posilivehigh Source: WB staff calculations from a municipaldatabase fromCensus and other sources. Table 3.27 compares characteristics of municipalities that experienced negative population growth between 1980 to those with positive high and low increases in their population over the same period. In summary, the table suggests that municipalities where populationgrew fast seemwell endowed interms of geographic characteristics and assets but they are not necessarily performing better in terms of the evolution of their socioeconomic indicators over time. More specifically, Table 3-27 illustrates that municipalities with a high increase in population differ significantly from the other two groups in several aspects. They are on average larger in terms of population and have lower per capita municipal expenditures. They have more employment inmanufacturing and in services, better access to roads and are closer to the main centers of economic activity of the country. When looking at the performance of socioeconomic variables 'over time these municipalities had higher decreases in adult illiteracy but experienced increases in the crowdedness of dwellings 96 and inthe proportion of population that lacks sewerage. On the other hand, municipalities with negative population growth differ from the other two groups in that they are less likely to be located on the coast or on the US-Mexico border. Table3.27. Municipalattributes,by 1980-2000populationgrowth Negative Low High N=L L=H N=H 2000-1980 populationgrowth -18.2 20.5 88.7 *** Population in2000 17,403 30,235 66,339 ***** *** *** *** Per capitamunicipal expenditures3 1,192.1 751.5 646.3 *** *** *** Share of illiterate adults1 20.5 23.6 24.3 *** *** Share who speak indigenouslanguage' 18.6 20.4 24.9 *** *** Share of employment inmanufacturing' 10.2 11.4 16.1 *** *** Share of employment inservices' 24.25 28.15 33.90 *** *** *** Share of population servedby state road' 35.07 45.84 55.52 *** *** *** Kmto closesteconomiccenter 146.13 123.92 105.11 *** *** *** Coastaldummy 0.04 0.09 0.10 *** *** Northern border dummy 0.01 0.02 0.02 ** * 2000-1999 changes Share of illiterate adults -3.46 -4.74 -5.30 *** *** *** Share indwellings without piped water -7.59 -7.93 -6.82 Share indwellings without sewerage 9.61 6.33 2.88 ** *** *** Share indwellings without electricity -12.52 -13.32 -12.67 Average occupantsper room 0.16 0.21 0.25 *** *** *** Share indwellings with dirt floor -10.49 -10.33 -8.83 *** *** Share employed earning lessthan two minimumwages -7.42 -7.74 -9.86 *** *** Share of adults with secondary education 6.50 6.68 7.08 *** *** Observations 624 880 879 1.In1990 2. In 1995 3. In 1999, with respectto 2000 population. Significant at: *** 1%, ** 5%, *10% Source: WB staff calculations from a municipal databasefrom Censusandother sources. Inthis section we have taken a brief look at patterns of well-being from a different angle to early sections - o f geography and ethnic grouping. Geographic factors cut across all dimensions of well-being. Living in a poor area can make a profound difference to life prospects. There are large differences inincome poverty between different regions with a generalized gradient from northto south, and Mexico City having the highest income and lower poverty rates. The highest poverty incidence is in the rural areas of the South Pacific states Ahiapas, Guerrero and Oaxaca- where still some46% of the population lives inextreme poverty, followed by the South Gulf and Caribbean, where over 35% of tlie population are extremely poor. This compares with about 19% in the Center, 10% in the North and4% inMexico City living inextreme poverty. From 2000 to 2002 extreme poverty fell significantly in the South Pacific (6 percentage points), the Center (5 percentagepoints) and inthe South Gulf (5 percentagepoints). The North and Mexico City did not experience a significant change. Moderate poverty fell 97 significantly only in the Center (five percentage points). However, it is premature to assess whether or not thesepoverty trends will continue. These overall regional differences hide a great deal of spatial heterogeneity within states both in terms of levels of well-being and patterns of change. In terms of numbers, there are large groups of the extreme poor living outside the poorest states (a quarter of all the extreme poor in Mexico live in urban areas in the Center states). Indeed, all parts of Mexico have steep gradients in conditions of living from more developed urban areas, through peri-urban areas and smaller towns, through to the more remote rural areas. There is an overlap between indigence and spatial indexes of well-being especially for rural indigenous groups. There are high concentrations of indigenous groups in many municipalities in the South-Pacific states, in the Yucatan peninsula as well as in some municipalities of the West and North-West of Mexico. Interms of trends, regional differences have long historical roots, There has been some long-term convergenceinmost indicators of services and social conditions but a tendency toward divergence in income and wage measures in the 1990's that appears to be associatedwith the differential effects of heightened international integration both before and after NAFTA. Areas closer to the border, or to urban centers, typically grew faster. The 2000-02 period appears to have witnessed statistically significant reductions in extreme poverty inthe South-Pacific, South-Gulf andCentral regions but not inthe North or Mexico City. For moderate poverty the only statistically significant decline was for the Central region. 98 ANNEX Robust Robust Variables dWdx Std.Err. P>z x-bar Variables dFldx Std,Err. Pzz x-bar urban* 0.019 0.005 0.000 0.753 adult squared -0.002 0.001 0.002 9.858 babies 0.039 0.006 0.000 0.483 age of HH -0.003 0.001 0.001 47,464 babiesquared 0.001 0.002 0.667 0.813 age squaredof HH 0.000 0.000 0.003 2490.800 children 0.043 0.005 0.000 0.860 female HH* 0.014 0.011 0.170 0.213 children squared -0.001 0.001 0.240 1.923 nospouse* 0.004 0.026 0.880 0.280 Household Head Characteristics Spouse Characteristics Education Primary Incomplete* -0.013 0.006 0.031 0.237 -0.019 0.006 0.004 0.177 PrimaryComplete* -0.015 0.006 0.028 0.204 -0.027 0.006 0.000 0.180 Lower Sec Incomplete* -0.031 0.005 0.000 0.037 -0.028 0.007 0.012 0.029 Lower Sec Complete* -0.031 0.006 O.Oo0 0.176 -0.030 0.006 0.000 0.161 UpperSec Incomplete* -0.034 0.005 0.001 0.032 -0.045 0.003 0.000 0.023 UpperSec Complete* -0.042 0.004 O.Oo0 0.061 -0.043 0.004 O.Oo0 0.048 University Incomplete* -0.033 0.007 0.005 0.035 Dropped (predictsfailure perfectly) University Complete* -0.041 0,006 0.006 0.056 Dropped (predictsfailure perfectly) NotWorking* -0.059 0.008 O.OO0 0.187 0.060 0.028 0.017 0.457 Searching* 0.432 0.122 o.Oo0 0.009 0.067 0.072 0.204 0.002 Formal* 0.016 0.018 0.348 0.263 0.003 0.024 0.887 0.128 Non-agriculturalLaborer* 0.012 0.011 0.251 0.441 0.009 0.022 0.661 0.115 Employer(under 5 employees)* -0.035 0.005 0.000 0.043 -0.041 0,003 0.000 0.005 Employer(5 or employees)* + -0.039 0.005 0.012 0.007 Dropped (predictsfailure perfectly) Self-Employed* -0.013 0.015 0.422 0.244 -0.014 0.020 0.520 0.103 Sectorof Activity Extraction* -0.043 0.003 0.002 0.004 Dropped (predictsfailure perfectly) Manufacturing* -0.030 0.006 O.OO0 0.131 0.004 0.015 0.802 0.047 Utilities* -0.037 0.005 0.003 0.005 Dropped (predictsfailure perfectly) Construction* -0.033 0.005 O.OO0 0.082 -0.035 0.010 0.133 0.002 Commerce* -0.045 0.004 O.OO0 0.136 -0.006 0.013 0.661 0.077 Transportation* -0.038 0.005 0.000 0.053 -0.022 0.022 0.463 0.002 FinanciafServices' -0.036 0.007 0.034 0.005 Dropped (predictsfailure perfectly) Services * -0.040 0.006 0.000 0.228 -0.020 0.012 0.156 0.095 Underemployment < 13brs.* 0.023 0.018 0.138 0.033 0.027 0.021 0.121 0,051 13hrs-19 hrs.* 0.023 0.018 0.137 0.020 0.065 0.031 0.004 0.020 19hs-39 h.* -0.007 0.006 0.257 0.133 0.028 0.017 0.051 0.070 FirmSize With c 5 workers" 0.019 0.008 0.017 0.449 0.026 0.026 0.247 0.184 With 5 to 9 workers* 0.013 0.012 0.244 0.072 0.047 0.041 0.137 0.017 With 10to 19workers* 0.023 0.018 0.141 0.048 0,004 0.028 0.892 0.009 99 Table A.3.2. Marginal contribution of householdcharacteristicsonpercapita income, 2002 Observations 17,167 R-squared 0.671 Prob>F 0 F(102, 17064) 145.18 RootMSE 0.533 Variables Urban 0.204 0.020 10.060 0.000 adult squared 0.020 0.002 8.540 0.000 Babies -0.302 0.017 -18.050 0.000 ageofHH 0.021 0.003 6.280 0.000 Babiessquared 0.032 0.005 5.920 0.000 agesquaredofHH 0.000 0.000 -4.100 0.000 Children -0.306 0.014 -22.050 0.OOO femaleHH -0.110 0.028 -3.860 0.000 children squared 0.028 0.004 7.780 0.000 nospouse 0.145 0.071 2.030 0.043 Adults -0.232 0.019 -12.020 0.000 socialsecurity 0.224 0.018 12.610 0.000 ~ o u s e h Head Characteristics o ~ Spouse Characteristics PrimaryIncomplete 0.179 0.023 7.870 0.000 0.117 0.026 4.480 0.000 PrimaryComplete 0.305 0.026 11.880 0,000 0.156 0.029 5.370 0.000 Lower Sec Incomplete 0.421 0.042 10.020 0.000 0.142 0.040 3.520 0.OOO Lower Sec Complete 0.428 0.030 14.500 0.000 0.137 0.031 4.490 0.000 UpperSec Incomplete 0.637 0.050 12.760 0.000 0.157 0.049 3.190 0.001 Upper See Complete 0.687 0.038 17.850 0.000 0.312 0.043 7.310 0.000 University Incomplete 0.970 0.054 18.080 0.000 0.493 0.061 8.090 0.000 University Complete 1.109 0.046 23.970 0.000 0.524 0.069 7.610 0.000 Not Working 0.707 0.056 12.720 0.000 -0.221 0.063 -3.490 0.000 Searching -0.916 0.100 -9.130 0.000 0.293 0.220 1.330 0.183 Formal -0.007 0.043 -0.150 0.879 0.004 0.063 0.060 0.950 Non-agriculturalLaborer -0.056 0.036 -1.570 0.116 0.003 0.051 0.060 0.948 Employer(under 5 employees) 0.479 0.048 1O.OOO 0.OOO 0.443 0.080 5.530 0.000 Employer(5 or +employees) 0.715 0.104 6.900 0.000 0.803 0.258 3.120 0.002 Self-Employed 0.039 0.051 0.770 0.443 0.033 0.068 0.490 0.627 Sectorof Activity Extraction 0.467 0.068 6.880 0.OOO 0.378 0.157 2.410 0.016 Manufacturing 0.168 0.033 5.130 O.OO0 0.002 0.050 0.050 0.961 Utilities 0.326 0.078 4.160 O.OO0 -0.551 0.233 -2.360 0.018 Construction 0.202 0.034 6.000 0.000 -0.002 0.135 -0.020 0.987 Commerce 0.245 0.034 7.280 O.OO0 0.091 0.046 1.980 0.047 Transportation 0.261 0.041 6.320 O.OO0 0.211 0.150 1.400 0.160 FinancialServices 0.350 0.187 1.870 0.062 0.347 0.138 2.520 0.012 Services 0.188 0.032 5.900 0.000 0.046 0.051 0.910 0.364 Underemployment Head< 13hrs. -0.133 0.046 -2.860 0.004 -0.076 0.036 -2.110 0,035 Head 13brs-19 hrs. -0.097 0.051 -1.910 0.056 -0.139 0.042 -3.300 0.001 H a d 19hrs-39 hrs. -0.003 0.021 -0.150 0.884 -0.071 0.026 -2.700 0.007 FirmSize With e 5 workers -0.106 0.025 -4.340 0.OOO -0.146 0.046 -3.190 0.001 With 5 to 9 workers -0.026 0.029 -0.890 0.373 -0.146 0.064 -2.280 0.023 With 10to 19workers -0.061 0.030 -2.030 0.043 -0.096 0.062 -1.550 0.122 100 Table A.3.3. Trends inHeadcount, Poverty Gap and Squared Poverty Gap 1992 1994 1996 1998 2000 2002 Headcount (Po) Extreme National 22.4 21.4 37.1 34.0 24.2 20.3 Rural 35.6 37.0 52.4 52.4 42.4 34.8 Urban 13.3 10.1 26.5 21.2 12.6 11.4 Moderate National 52.5 55.8 69.6 63.7 53.7 51.7 Rural 65.0 72.3 80.8 75.1 69.3 67.5 Urban 43.8 43.7 61.9 55.7 43.7 42.0 Poverty Gap (PI) Nat~onal 7.49 7.15 13.87 13.47 8.44 6.34 Rural 13.14 13.47 21.90 23.88 16.34 12.15 Extreme Urban 3.57 2.55 8.33 6.27 3.36 2.79 Nationa1 21.66 23.90 33.96 30.94 23.38 21.20 Rural 29.65 34.97 42.95 42.16 34.81 31.69 Moderate Urban 16.11 15.83 27.76 23.19 16.03 14.81 Poverty Gap Squared (P2) National 3.5 1 3.32 7.05 7.17 4.32 3.17 Rural 6.47 6.57 11.91 13.61 8.42 6.57 Extreme Urban 1.45 0.96 3.69 2.72 1.69 1.09 National 11.76 13.19 20.49 18.76 13.23 11.52 Rural 17.23 20.73 27.42 27.91 21.37 18.72 Moderate Urban 7.96 7.70 15.69 12.44 8.00 7.13 Source:WB Staff estimatesbasedonENIGH, usingSEDESOL's poverty lines for incomemeasure. Table A.3.4. Povertv lines (SEDESOL) ~~~~~ ~ 1992 1994 1996 1998 2000 2002 Foodpoverty urban 167,955.20 194.00 389.40 524.80 626.00 672.25 (PLI) Rural 124,750.60 143.30 290.00 388.50 463.00 494.77 Capabilities Urban 197,963.80 236.10 469.40 622.10 768.10 792.29 poverty (PL2) Rural 140,586.30 165.60 335.60 445.00 548.50 587.57 Asset-poverty Urban 333,122.70 414.50 804.20 1,054.50 1,254.50 1,366.85 W 3 ) Rural 218,794.40 278.70 527.20 690.90 843.30 946.94 Lines are presentedinmonthly per capitapesos of August of each year. Source: INEGI-CEPAL,SEDESOL. 101 Table A.3.5. Poverty Lines 2000deflated with food and non-food CPIs 1992 1994 1996 1998 2000 2002 Extreme Urban 167,955.20 194.00 389.40 524.80 626.00 672.25 Rural 124,750.60 143.30 290.00 388.50 463.00 494.77 Moderate Urban 326,157.38 384.69 739.81 1,008.50 1,254.50 1,364.22 Rural 220,477.39 258.68 502.03 681.18 843.30 913.47 Extreme poverty line is the same as Food Poverty line. The Moderate poverty line in 2000 i s the same as the Assets poverty line. The difference from Asset poverty and Food Basedpoverty lines in2000 corresponds to the non-food components of the bundle. This difference, in 2000, is deflated to other years with a non-food CPI constructed from Banxico's CPIs. Then, this deflated difference is added to the correspondingExtreme Poverty lines to construct the Moderate poverty lines. Table A.3.6. Poverty lineswith Fixed (Nominal) Engel Coefficient 1992 1994 1996 1998 2000 2002 Extreme Urban 167,955.20 194.00 389.40 524.80 626.00 672.25 Rural 124,750.60 143.30 290.00 388.50 463.00 494.77 Moderate Urban 336,581.15 388.77 780.36 1,051.70 1,254.50 1,347.18 Rural 227,218.53 261.00 528.20 707.61 843.30 901.17 Extreme poverty line is the same as Food Poverty line. Nominal Engel coefficients were calculated for 2000 poverty lines between the Food Based and Assets poverty lines. These coefficients were used to construct the moderatepoverty lines for the rest of the years. Table A.3.7. ExpenditurePoverty Lines 1992 1994 1996 1998 2000 2002 Extreme Urban 159,101.33 183.77 368.87 497.13 593.00 636.81 Rural 123,928.81 142.36 288.09 385.94 459.95 491.51 Moderate Urban 286,639.73 337.50 651.36 887.08 1,099.68 1,194.66 Rural 202,707.71 237.31 462.58 626.80 772.92 836.09 Setting 2000 as base year, extreme and moderate poverty lines were constructed so that they would yield the same poverty headcounts using expenditure as the headcount obtained in 2000 with the food based and assets poverty lines usingincome. The extreme poverty lines for the rest of the years were obtained deflating the 2000 line with the implicit food CPI from the official poverty lines. Difference from the Moderate poverty and Extreme poverty lines in 2000 correspondsto the non-food components of the bundle. This difference in2000 is deflated to other years with a non-food CPI constructed from Banxico's CPIs. Then, this deflated difference is addedto the correspondingExtreme Poverty lines to construct the Moderate poverty lines. 102 Table A.3.8. Consumer Price Indices 1992 1994 1996 1998 2000 2002 Food CPI: Implicitfrom Oficial Food Based Poverty Lines Urban 26.83 30.99 62.20 83.83 100.00 107.39 Rural 26.94 30.95 62.63 83.91 100.00 106.86 BANXICO CPIfor the Second Tier (between 1and 3 ~ ~ n ~ mwages)by Expenditure Category u m FoodCPI 25.50 29.14 59.35 80.7 1 100.00 109.21 Non-Food CPI 25.17 30.34 55.75 76.96 100.00 110.10 Estimated GeneralCPI 25.36 29.64 57.85 79.14 100.00 109.58 GeneralCPI 25.71 30.19 57.16 78.67 100.00 110.52 The non-food CPI and the estimated general CPI were calculated using expenditure shares from each category inthe 4th incomedecile according to ENIGH2000. Table A.3.9. Accumulated Income Share Percentiles, National. PopulationShare 1992 1994 1996 1998 2000 2002 10 1.2 1.2 1.3 1.0 1.1 1.3 20 3.5 3.3 3.6 3.1 3.3 3.7 30 6.6 6.2 6.8 6.0 6.3 6.9 40 10.6 10.0 10.9 10.0 10.3 11.2 50 15.6 14.9 16.1 14.9 15.3 16.5 60 21.9 21.0 22.6 21.2 21.6 23.1 70 30.0 28.9 30.8 29.2 29.5 31.4 80 40.5 39.5 41.7 39.8 40.1 42.5 90 56.4 55.4 57.7 55.4 55.7 58.7 92 60.8 59.9 62.2 59.7 60.1 63.1 94 66.0 65.2 67.4 64.6 65.1 68.3 96 72.2 71.8 73.8 70.7 71.5 74.9 98 80.7 81.2 82.5 79.1 80.5 83.8 100 100.0 100.0 100.0 100.0 100.0 100.0 Bottom20% 3.5 3.3 3.6 3.1 3.3 3.7 Middle40% 18.4 17.7 18.9 18.2 18.3 19.4 Middlehigh30% 34.5 34.4 35.2 34.2 34.2 35.6 Top 10% 43.6 44.6 42.3 44.6 44.3 41.3 Source: WB Staffcalculations basedonENIGH. 103 Table A.3.10. Accumulated Income Share Percentiles inRuralArea ~ ~ ~ ~opulat~onShare 1992 1994 1996 1998 2000 2002 10 1.6 1.9 1.8 1.5 1.4 1.6 20 4.3 4.9 4.9 4.1 3.7 4.3 30 7.9 8.9 9.0 7.6 6.8 7.7 40 12.4 13.9 14.2 12.1 10.7 12.1 50 18.0 20.0 20.5 18.0 15.6 17.4 60 24.8 27.2 28.1 25.3 21.8 24.0 70 33.0 36.3 37.5 34.6 29.4 32.3 80 43.6 47.9 49.2 46.5 39.4 42.8 90 58.0 64.6 65.5 62.9 54.1 59.1 92 61.6 68.8 69.8 67.2 57.7 62.6 94 65.5 73.7 74.6 72.2 62.2 67.6 96 70.1 79.5 80.3 78.1 68.2 73.3 98 76.2 86.5 87.4 85.8 77.2 81.5 100 100.0 100.0 100.0 100.0 100.0 100.0 Bottom20% 4.3 4.9 4.9 4.1 3.7 4.2 Middle40% 20.2 22.4 23.2 21.3 18.2 19.8 Middlehigh30% 33.2 37.3 37.4 37.6 32.3 35.1 Top 10% 42.0 35.4 34.5 37.1 45.9 40.9 Source: WE3 Staffcalculationsbasedon ENIGH. Table A.3.11. Accumulated Income Share PercentilesinUrbanArea PopulationShare I992 1994 1996 1998 2000 2002 10 1.7 1.7 1.7 1.6 1.7 1.8 20 4.4 4.3 4.3 4.2 4.5 4.9 30 7.9 7.7 7.9 7.7 8.1 8.6 40 12.3 12.0 12.3 11.9 12.5 13.3 50 17.7 17.1 17.7 17.1 17.9 18.9 60 24.4 23.4 24.4 23.5 24.4 25.7 70 32.7 31.4 32.7 31.5 32.6 34.2 80 43.9 42.1 43.5 42.1 43.2 45.3 90 60.2 58.2 59.7 57.1 58.6 61.2 92 64.6 62.6 64.1 61.2 62.9 65.7 94 69.7 68.0 69.2 66.1 67.6 70.8 96 75.8 74.5 75.6 71.9 74.0 77.2 98 83.7 83.1 83.9 80.2 82.1 85.8 100 100.0 100.0 100.0 100.0 100.0 100.0 Bottom20% 4.4 4.3 4.3 4.2 4.5 4.9 Middle40% 19.7 19.2 20.1 19.3 19.9 20.9 Middlehigh30% 35.8 34.7 35.3 33.5 34.2 35.5 Top 10% 39.8 41.8 40.3 42.9 41.4 38.8 Source: WB Staff calculationsbasedon ENIGH. 104 Figure A.3.1. Stochastic Dominance, 2000-2002 a) First Order Dominance Total Household Ex enditurePerCapita (Bottom 50 percent& -chf92 cdf96 cdfOO cdf02 529568 h YI r a -.- $ Y 5a U 0 pce Monthlyper capita expenditure Cumulative Distribution(First order dominance) b) Second Order Dominance Total Household Ex enditure Per Capita (Bottom 50 percent&)fxx92 fxxoo fxx96 fXXO2 1996 362.678 .3 5 Y2I F4 2a .e Y Y d8 2a z 4 0 Monthly per capita expenditure Deficit curve (Second order dominance) Source: WB Staff estimatesbasedonENIGH, 105 TableA.3.12. Growthand InequalityComponents2000-2002 Income Expenditure NationalUrbanRuralNationalUrbanRural 2000Extreme Poverty 24.24 12.58 42.38 25.41 14.35 42.62 2002Extreme Poverty 20.27 11.43 34.76 22.21 12.93 37.43 Growth 2.26 -0.12 5.24 0.70 -0.63 2.01 Distribution -3.95 -1.11 -7.61 -2.04 -0.50 -4.23 Residual -2.29 0.08 -5.25 -1.86 -0.29 -2.96 2000 ModeratePoverty 53.72 43.72 69.28 58.97 50.13 72.73 2002ModeratePoverty 51.66 42.04 67.45 58.20 49.61 72.30 Growth 1.42 0.56 2.03 -0.92 -1.65 -0.50 Distribution -2.01 0.06 -1.81 0.87 1.15 1.15 Residual -1.46 -0.62 -2.05 -0.71 -0.02 -1.08 Source: WB Staff calculations usingdata fromENIGH. Table A.3.13. Average IncomeShares, 2000 and2002 2000 2002 Estimate Std. Err. [95% Conf. Estimate Std. Err. [95% Conf. Interval] Interval] Labor Income 49.8% 1.4% 47.1% 52.5% 50.7% 1.1% 48.6% 52.8% BusinessI CooperativeIncome 17.9% 0.9% 16.2% 19.6% 17.6% 0.8% 16.1% 19.1% Rent Income 0.9% 0.2% 0.6% 1.3% 1.3% 0.3% 0.7% 1.8% TransfersLESSRemittancesIncome 5.3% 0.3% 4.7% 5.9% 5.8% 0.3% 5.2% 6.3% RemittancesIncome 1.5% 0.2% 1.2% 1.8% 1.2% 0.1% 1.0% 1.4% Financial Income 5.2% 0.6% 4.1% 6.3% 4.5% 0.8% 3.0% 6.1% Other Monetary Income 0.4% 0.8% 17.5% 20.5% 0.1% 0.1% 0.0% 0.3% Nonmonetary income 19.0% 0.1% 0.3% 0.6% 18.8% 0.3% 18.1% 19.5% Rural Labor Income 39.4% 3.0% 33.5% 45.4% 44.1% 2.7% 38.9% 49.4% BusinesslCooperativeIncome 22.3% 1.4% 19.4% 25.1% 18.4% 1.9% 14.6% 22.2% Rent Income 0.5% 0.1% 0.2% 0.7% 0.7% 0.1% 0.4% 1.0% Transfers Less RemittancesIncome 6.5% 0.5% 5.6% 7.4% 8.4% 0.8% 6.7% 10.0% Remittances Income 3.3% 0.6% 2.2% 4.4% 3.5% 0.5% 2.5% 4.5% Financial Income 7.2% 1.4% 4.4% 10.0% 4.6% 0.7% 3.2% 5.9% Other Imputed Income 0.3% 0.1% 0.1% 0.4% 0.4% 0.3% -0.2% 1.1% Non monetary income 20.7% 2.8% 15.1% 26.2% 19.9% 0.6% 18.8% 21.1% Urban Labor Income 53.0% 1.3% 50.5% 55.5% 52.7% 1.2% 50.3% 55.1% BusinesslCooperativeIncome 16.5% 1.0% 14.5% 18.5% 17.4% 0.8% 15.7% 19.0% RentIncome 1.1% 0.2% 0.6% 1.6% 1.4% 0.4% 0.7% 2.2% Transfers Less RemittancesIncome 4.9% 0.4% 4.2% 5.6% 5.0% 0.3% 4.4% 5.6% RemittancesIncome 0.9% 0.2% 0.6% 1.3% 0.5% 0.1% 0.4% 0.7% Financial Income 4.6% 0.6% 3.5% 5.7% 4.5% 1.0% 2.6% 6.5% Other Imputed Income 0.5% 0.1% 0.3% 0.6% 0.0% 0.0% 0.0% 0.0% Non monetary income 18.5% 0.4% -17.7% 19.3% 18.5% 0.4% 17.7% 19.3% Source: WB Staff estimatesusing ?NIGH. 106 Table A.3.14. MarginalValue of Educationby Quantile. Mexico1988-2002 1988 1992 Quantile Quantile 0.10 0.25 0.50 0.75 0.90 0.10 0.25 0.50 0.75 0.90 Primary Complete 1.08 1.08 1.05 0.99 0.90 Primary Complete 1.01 1.01 0.95 0.85 0.73 Lower Secondary Complete 1.14 1.13 1.15 1.18 1.22 LowerSecondaryComplete 1.13 1.15 1.17 1.19 1.23 Upper Secondary Complete 1.11 1.14 1.18 1.22 1.27 UpperSecondaryComplete 1.15 1.19 1.22 1.26 1.32 Tertiary 1.38 1.42 1.44 1.45 1.43 Tertiary 1.50 1.57 1.65 1.62 1.59 1997 2002 Quantile Quantile 0.10 0.25 0.50 0.75 0.90 0.10 0.25 0.50 0.75 0.90 Primary Complete 1.09 1.10 1.09 1.07 1.06 Primary Complete 1.09 1.07 1.07 1.06 1.05 Lower Secondary Complete 1.14 1.15 1.18 1.23 1.28 LowerSecondaryComplete 1.11 1.12 1.13 1.17 1.21 Upper Secondary Complete 1.15 1.19 1.22 1.29 1.36 Upper Secondary Complete 1.12 1.13 1.18 1.23 1.29 Tertiary 1.62 1.75 1.80 1.75 1.63 Tertiary 1.56 1.68 1.77 1.78 1.68 Note: The Model controls by: Gender, Age, SquaredAge, GeographicZone (North, Center, SouthandD.F.), Status inthe LaborMarket (Employer, Self-Employed, Formal Salaried, InformalSalaried, andOther SalariedWorker) andSectorof Activity (Primary sector, Manufacturingindustry, ConstructionandUtilities, Commerce,financial services andrent, TransportationandComunication), Social services(Education, Tourism, Health, Public Administration andEmbassy) andOther Services. FigureA.3.2. Shockson Labor Income and Labor IncomeDistribution (Averageand Standard Deviation by IncomeDecile) Mexico 2000-2003 - 0.7 1.Go 0.6 0.95 3E 0.90 0.5 5 -8 0.4 0.85 n 9 B 0.3 0.80 3 s 1 38 0.2 0.75 0.1 0.70 g I "s 0 0.65 3 -0.1 0.60 v1 -0.2 0.55 -0.3 0.50 1 2 3 4 5 6 7 8 9 10 IncomeDedle -E-Aver= t-StandardDeviation Source: WB staff calculations usingENET2000-2003. 107 Table A.3.15. Correlates of LogIncome Changes inUrbanHouseholdIncome Quantile Analysis: 1992-1995 Dependent variable Log Income Change PresamplePeriod 11 Crisis Period 0bservations 41676 41676 41676 11 41676 41676 41676 0.2 0.5 0.8 0.2 0.5 0.8 a b C d e f Primary Inc. 0.2324 *** 0.0085 -0.0649 *** 0.2320 *** 0.0358 0.0729 ** Primary 0.1832 *** 0.0047 -0.0552 *** 0.2241 *** 0.0597 *** 0.0662 ** Secondary Inc 0.1084 *** 0.0024 -0.0326 ** 0.1282 * 0.0118 0.0387 Secondary 0.0788 ** 0.0081 -0.0156 0.1665 * 0.0226 0.0278 Young 0.0411 0.0079 -0.0286 -0.1021 -0.0094 0.0195 Old -0.1918 *** -0,0467 *** 0.0142 -0.0456 0.0011 -0.0394 >1.3 Children 0.0278 0.0174 * 0.0039 0.0187 -0.0051 0.0124 SingleMothers -0.0466 0.0046 0.0078 -0.0011 0.0302 0.0306 Single Women -0.1627 *** -0.0550 *** -0.0793 *** 0.2190 ** 0.0615 * 0.1164 *** Single Men -0.1025 * -0.0305 -0.0014 0.0309 0.0264 0.0137 Informal Self-Employed -0.263 1 *** -0.0685 *** 0.1051 *** 0.0487 -0.0055 0.0181 Informal Salaried -0.0928 *** 0.0196 0.0771 *** 0.0712 -0,0081 -0.0001 No remuneration -0.1569 *** 0.0595 *** 0.2424 ***: -0.4475 *** -0.1244 *** -0.1064 *** Constant -0.4356 *** -0.0176 * 0.5289 ***11-0.4056 *** -0.2799 *** -0.3082 *** Analysis uses five quarter panels of the National Urban Employment Survey. The table above shows the coefficients of a regressions where five periods (1992:4 -1993:4, 1993:l-1994: 1,1993:2-1994:2, 1993:3 - 1994:3, 1994:3 -1995:3) of income growth havebeenregressedoncharacteristicsof Mexican householdsand their heads. We also included in the regressiona dummy interacted to one of each variables for the Crisis Period that goes from the 1994:3 to the 1995:3. Each group of columns a-d, b-e and c-f, correspond to the estimationsof the 20th, 50th andthe 80thquantile.The second column of eachpair retrievesthe coefficients of the interaction of each variable with the dummy for the crisis period. Standard errors have been caculated using bootstrapping techniques from Gould (1992, 1997). Number of bootstraps obtained by running the algorithm proposedby DavidsonandMackinnon (2000). Coefficients have beencorrectedfollowing Kennedy (1981). 108 Table A.3.16. Correlates of Log Income ChangesinUrbanand Rural Household Income Quantile Analysis: 2000-201 Urban Rural Variable Ouantile Quantile 0.2 0.5 0.8 0.2 0.5 0.8 Observations 28,786 28,786 28,786 10,905 10,905 10,905 Youth(Age<25) 0.117 ** 0.058 ** 0.050 0.041 0.031 0.050 Old (Age>45) -0.014 -0.001 0.048 ** -0.031 0.000 0.013 Change in labor hoursper week 0.002 ** 0.002 ** 0.002 ** 0.002 ** 0.001 ** 0.002 ** Children > 2 -0.041 ** -0.014 0.005 -0.018 -0.021 -0.018 No Education 0.026 0.037 ** 0.046 * -0.146 ** 0.054 ** 0.159 ** Primary Incomplete 0.054 ** 0.037 ** 0.010 -0.010 0.104 ** 0.158 ** Primary Complete 0.068 ** 0.038 ** 0.009 -0.014 0.060 ** 0.138 ** SecondaryIncomplete 0.018 0.043 ** 0.052 -0.119 **-0.024 -0.007 Secondary Complete 0.040 ** 0.023 ** 0.007 -0.050 0,025 0.065 High School 0.011 0.000 -0.007 -0.133 * 0.004 -0.047 Technical 0.046 * 0.022 0.009 -0.011 0.019 0.044 Single man -0.013 0.016 0.011 -0.053 -0.068 ** -0.120 ** Married man or woman Single woman 0.015 -0.008 -0.016 -0.038 -0.032 -0.042 Single woman w/ch -0.008 -0.009 -0.006 -0.038 -0.029 ** -0.055 * Employer -0.107 ** 0.018 0.113 ** -0.091 0.017 0.205 ** Self employed -0.137 ** -0.026** 0.103 ** -0.178 ** -0.045** 0.145 ** I n f o ~ salaried l -0.043 * 0.000 0.035 -0.036 0.010 0.128 ** Constant -0.315 ** 0.060 ** 0.419 ** -0.242 ** 0.068 ** 0.423 ** Source: WB staff calculations based on ENET Panel Data, 2000:2-2001:2, 2000:3-2001:3, 2000:4- 2001:4, 2001:2-2002:2,2001:2-2002:3,2001:4-2002:4. ** Significant at 5%,* Significantat 10% Note: All models for every sector were estimatedusingQuantile Analysis andrunningbootstrapping recursive iteration. 109 June 2004 Poverty inMexico AnAssessmentof Conditions, Trends and Government Strategy Chapter 4. Public Spending, Poverty andInequality This chapter turns to areas of direct government action, and in particular to the role of government programs financed by the Federal Government budget. The government budget.is potentially one of the most powerful instruments affecting patterns of deprivation, distribution and development. The chapter is divided into three parts. The first looks at the overall pattern of public spending and taxes, including the composition of spending over time, the incidence of spending across the population, and the extent of coverage of some of the major programs inrelation to the structure of needs identifiedin Chapter3. Itconcludes that the budgetis already significantly more equal inits impact on the population than private incomes. For the parts that can be measured the incidence i s roughly proportional but there i s very wide variation from highly regressive to highly progressive programs. There i s potential for budget shifts with substantial redistributive impact. The second part of the chapter examines selected issues in the design of social programs. Two issues are explored: the overall shape of the social protection system and the challenge of achieving quality in service delivery (especially for the poor). The last part discusses the implication of social spending and service delivery in the context of decentralization. The analysis of this last part enters complex terrain that will require considerable further work it i s intended to contribute to ongoing debates and to frame future work. The chapter draws extensively on parallel work being undertaken by the World Bank as input to its review of public spending in Mexico. Background work on spending was designed to feed into both poverty and public spending studies to ensure that these are fully complementary. A. Publicspending, taxes andinequality Action via the government budget can affect households and individuals via many channels that affect different dimensions of well-being. We focus on three channels in this chapter. First, provision of services such as roads, schools, health centers, water, sanitation and electricity can directly increase the quality of life and can expand human and physical productive assets. More productive assets can in turn lead to higher future incomesn -if after a lag of many years in the case of education. Second, incomes are also more directly affected by the taxes that have to be collected to finance government spending, and by direct transfers to households, whether these take the form of direct transfers, as in the case of pensions and Opor~un~dude~, or are hidden in subsidies to providers, as in the case of electricity and water inMexico. Third, the vulnerabilities that households and individuals face are substantially influenced by programs directly designed to provide protection against risks (including provisioning for health, unemployment or weather-related shocks) as well as indirectly, for example via the 110 influence of infrastructure on possibilities for household diversification and asset accumulation both of which affect risk-management possibilities.' Who benefits from spending is highly varied inall three categories andthis is a major theme of this section. It is also important to recognize that spending involves costs. This is most obviously true of the taxes required for its finance. Almost all taxes involve costs, including the administrative costs required to raise the taxes, the costs to individuals and firms in compliance andthe effect of taxes on patterns of work and production. Avoiding taxes - labor, income and VAT- i s one of the reasons why many small firms and individuals choose to operateinthe informaleconomy inMexico, potentially reducing their access to productive and social services. In a recent review of taxation in Latin America, Bird (2003) finds that typical estimates of costs, in relation to the total taxes collected, are in the range of 1percent for administrative costs, 4-5 percent for compliance costs, and 20- 30 percent for efficiency costs, with some estimateshigher than this. Onthe other side of the equation there is large variation in the effectiveness of spending, whether due to inefficiencies in performance of government administrations and service providers or outright corruption. An important theme of this report concerns the complex relationship between institutionalperformance andthe quality andeffectivenessof service delivery. This provides the context for the following discussion. To anticipate the conclusion, the report i s in general in favor of the expansion of government provisioning of some services to improve human development, incomes and reduce vulnerabilities. This is likely to involve increasedtaxes2But this only makes sense if it is combined with action to improve both the effectiveness and the equity of government spending. The analysis suggests there i s significant scope for doing both, building in part on initiatives such as Oportunidudesand a variety of measures to improve quality. Patternsof spendingand taxation-low tax efforts, but growingsocial spending We first review the overall pattern of spending and taxes. Mexico's overall tax effort i s low for its income level. In addition all spending was hard-hit by the 1994-95 crisis. Since then there has been substantial growth in social spending including in the recent period of stagnation of GDP. But this growth has been in large part financed by reduced economic sector spending. Mexico federal public sector spending was equivalent to 18% of GDP in2002 (Table 4.1) a relatively modest level for a country of Mexico's income level, a corollary of its unusually low tax collection effort. Of total federal spending, two-thirds, or 12% of GDP, were classified as "programmable" in the sense that they are allocated to a range of specific government programs designed to provide services or transfers to Mexican society. Within programmable spending a significant portion i s inthe form of earmarked transfers to subnational governments including teacher's salaries, part of the health `A further powerful influence on household risk-management is the condition of the financial sector (De Ferranti et a12000).Factors influencing the financial sector fall mainly outside the government budget and so are not discussed inthis chapter. World Bank (2004). 111 systemand social infrastructure. These are known as aportacionesthat are grouped under the Ramo 33 (R33) budget category and accounted for 23% of total programmable expenditure in2003. They fall inthe transfers category inTable 4.1, along with transfers to the pension systems and public enterprises. Outside programmable spending, sub national governments also account for a further 19% of total federal spending and over 3% of GDP through a tax-revenue sharing formula that is not earmarked to specific programs (known as participaciones.) While almost all analysis here covers programmable spending it i s the sum of this and participaciones that accounts for the bulk of action by the mexican state in provisioning of services and subsidies to the population. Federal resources dominate the resources of states and municipal governments: aportaciones account on average for almost 45% of state revenues and more of municipalities, while partici~acionesaccount for almost 40% for both levels of government. We discuss some of the issues for service provision raised by decentralization inthe last section of this chapter, Table 4.1: Federal Government's overallpatternof revenuesand spendingwith resnect to GDP. 1990-2002 1990 1995 2000 2002 Total Revenues 15.9% 15.2% 15.8% 15.8% TaxRevenues 10.7% 9.3% 10.6% 11.6% Non-Tax Revenues 5.2% 6.0% 5.2% 4.2% Total Expenditures 18.6% 16.1% 17.3% 18.0% Programmable Expenditures 7.3% 9.3% 10.7% 11.9% CurrentExpenditures 5.0% 7.4% 9.2% 9.7% Transfers 1.8% 4.8% 7.1% 7.5% Wage Bill 2.6% 1.8% 1.6% 1.6% Goods and Services 0.2% 0.2% 0.1% 0.1% GeneralServices andothers 0.4% 0.6% 0.3% 0.3% CapitalExpenditures 2.2% 1.8% 1.5% 2.2% N o n - P r ~ g r A m ~ bExpenditures l e 11.3% 6.8% 6.6% 6.1% FinancialCost 8.1% 3.8% 3.2% 2.5% Participaciones 2.8% 2.7% 3.2% 3.4% Others 0.4% 0.3% 0.1% 0.1% FiscalBalance -2.6 % -0.8% -1.5% -2.2% Source: SHCPandauthors calculations. The federal government's tax revenue at 11.6% of GDP is unusually low for Mexico's income level. The total tax collection effort i s illustrated in Figure 4.1. The low overall tax effort is due to a combination of relatively low taxes on trade and incomes and relatively hightaxes on goods and services relative to Mexico's income level. We return to the question of the role of tax policy when we look at budgetary options for a more equitable strategy, after reviewing spending patterns. 112 Figure 4.1 Mexico's tax collection effort is low by internationalstandards Average tax effort inrelation to logof GDPper capita, 1990-2000 0.45 0.40 m . 0.35 0.30 . m . 0.25 Nicaragua m 0.20 0.15 ' m unitedstales 0.10 2 a+ 0.05 0.00 4.5 5.5 6.5 7.5 8.5 95 10.5 115 Logof GDP per capita(1995 USD) 1990-2000Average Source:WB staff calculations from IMF,Government Finance Statistics. An initial overview of patterns of government spending is provided by analyzing the Ministry of Finance's functional classification of spending which includes the following three categories of spending: "economic sector", "government management", and "social development". "Economic sector" spending covers infrastructure, rural development, energy, transport, communication, and other services and economic activities, while social development spending includes education, health, social security, social assistance, regional and local development and labor policy. Government management includes legislation, law and order, national security, election processes, governance and environment. There i s often a special interest in "social" spending in relation to poverty reduction on the grounds that it i s of particular relevance to the social "needs" of the poor. Social spending can indeed be important. However, this can be quite misleading. As we show below, much social development spending is directed more to the non-poor than the poor, while social spending that does reach the poor can be important for income generation as well as social development. On the other hand, some "economic sector" spending, for example on basic infrastructure and some agriculture services, is also of particular relevance to income generation. With thesecaveats inmind, let's examine the patterns. Total public spending has experienced a significant increaseover the past decade, despite being hardhit by the 1994-95 crisis (Figure 4.2). Social development spending was also hard-hit by the crisis, but has risen faster than overall spending. Indeed, apart from the crisis, social spending has risen faster than GDP -and experienced significant real 113 growth in the period of stagnation from 2000-02 (Figure 4.3). The subcategory of targeted poverty programs also experienced rapid growth in 1998-2000 with a further acceleration in 2000-02. The substantial increase in social development spending since the crisis has been achieved despite fiscal astringency especially through reductions in "economic sector" spending which was only partially offset by growth since 2000 in the ruraldevelopment category. Figure4.2 Total and social spendinghasgrown significantlyinMexico, apart fromthe crisis. Programmable public spending, 1990-2002 1,200,000 200,000 * x :: x `. \I ,, ,.. '- " x- 0 " `1 t.Total +Social Development +-Economic Sector +Government Management Source: Anexo del Tercer Informede Gobierno 2003. Figure4.3 Socialspendingwas stronglypro-cyclical inthe 1994-95 crisisbutcontinued to grow inthe 2000-02periodof stagnantgrowth. (Growth of GDP andmajor public spending categories, 1990-2002) 25 -5 -20 3 3 N 3 a a zs 2s z s ; a sa ;w Q s o s s s a a -4-GDPRealGrowth &Total Spending +Social DevelopmentSpending -*-Poverty Reduction Note: Poverty reduction is part of social development spending but comprises only targeted programs. Source: Anexo delTercer Informe de Gobierno 2003. 114 At the level of broad program categories, Table 4.2 shows that long-term declines in "economic sector" spending have occurred because of reductions in energy, communications, and transport and "other" economic services spending. All categories grew in 2000-02 with very large increases in ruraldevelopment (agriculture and fishery) reversing the 1990-2000 decline. A recovery in spending on "economic sector" spending i s likely to be necessary for growth, competitiveness and poverty reduction. Some summingup results from cross-country experience are presentedinChapter 5 to support such a general position. However, it was beyond the scope of this report to analyse whether the particular rural development increase was beneficial to the poor. Indeed, most assessments of rural development spendingjudge that the bulk of such spending i s oriented to large producers in the north part of the country, with the exception of the Procamp program3(chapter 5; World Bank,2000; OECD, 2003). Table 4.2 Growth inpublicspendingcategories, 1990-2002 Percentper annum at 2002 prices 1990-2002 1990-2000 2000-2002 GDP 2.7% 3.3% 0.1% TotalPublic Spending 3.6% 3.1% 6.3% GovernmentManagement 5.0% 6.9% -3.8% Economic Sector Spending -1.3% -4.0% 13.4% Agriculture and Fishery Development 0.7% -8.2% 60.0% Energy -1.7% -2.6% 3.1% Communicationand Transport -2.0% -3.8% 7.1% Other economicServicesand Activities -4.8% -10.1% 26.5% Social Development 7.5% 8.0% 5.3% Education 7.4% 7.4% 7.5% Health 0.7% 1.1% -1.3% Social Security 35.2% 41.5% 7.7% Labor 5.8% 9.1% -9.1% Social Assistance 8.4% 8.1% 9.8% Urban and Regional Development 9.0% 8.6% 10.8% ...ofwhich poverty reduction 8.2% 7.0% 14.2% Source: WB Staff calculations from Anexo del Tercer Informede Gobierno 2003. With respect to social development spending, the major categories are education, health and social security (Figure 4.4). These have displayed quite varied patterns of growth in both the long term and in the 2000-02 period (Table 4.2 and Figure 4.5). Education has experienced steady and large long-run growth, with the exception of the crisis years, while health spending did not recover after the 1995 crisis (note that health spending includes health spending by the Ministry of Health (SSA), IMSS, and ISSSTE). Social 3 Procampo was introduced in 1993-94 to substitute for the previous guaranteed price support scheme (CONASWO) by the provision of direct payments to facilitate adjustment after NAETA for producers of basic crops. Although Procampo targets low-income farmers, the program's intended objectives of helping small farmers to switch to more productive crops or occupations have not been achieved (World Bank, 2000). 115 security spending rose dramatically from 2% of social development spending in 1990 to 25% in 2002 though this was primarily a result of bringingon to the federal budget the costs, both before and after the 1997 reform of IMSS, the social security system for private sector workers. As a result of the latter reform, transitional costs4havebeen 1.5% of GDP since 1998 (OECD, 2003). ISSSTE, the social security system for public sector workers, has yet to be reformed. There are no official estimates of the fiscal cost that a reform of ISSTE would entail but it would be only a fraction of what the IMSS reform is costing (OECD, 2003). Figure 4.4 Socialspendingis dominated by education,healthand social security Spending composition in2002 Social Assistance Regional& 2.3% 7 Urban Development 0.4% 7 11.4% 20.7% Source: Anexo del Tercer Informede Gobierno 2003. Figure 4.5 Educationand socialsecurity spendinghasgrown mostinabsoluteterms Levels of social spending by major categories 1990-2002 d m 300,000 j., & 250,000 4-r 0 200,000 m .I 35 150,000 R -3 100,000 3 50,000 0 m 0 -t-Education -k- Health +Social Security --+E- Labor -@-Food & SocialAssistance +Urban & RegionalDevelopment Source: Anexo delTercer Informede Gobierno 2003. Includes both reducedrevenues and higher expenditures. 116 While the major categories of education and health spending are of importance to the poor, as we explore inthe next section, there i s also a set of programsthat are specifically targeted to the poor that are included both inthe social development andeconomic sector pend ding.^ These are classified as "poverty reduction" programsby the government. They include most of the programs that now fall under food and social assistance, compensatory programs in education, as well as others which have poverty-targeting mechanisms. The largest antipoverty program i s Oportun~d~es(previously PROGRESA) which is also the largest item in the budget of SEDESOL, at 16 billion pesos (or almost 1.6 billion dollars). Other examples are Programa de Empleo Temporal (3.9 billion pesos); Programas Compensatorios del CONAFE (2.5 billion pesos); and Programa de Salud y Nutricidn de 10s Pueblos Indigenas (2.9 billion pesos). Poverty reduction programs have also experienced steady growth with a significant acceleration since 2000. In2002 they represented 1.3% of GDP comparedto 0.7% in 1990. Amongst such programs, social assistance expenditures, led by Oportunid~essince it was created, grew by an average of 8.4% per year during the ~O'S,and by even more -9.8% per year- after 2000. This initial overview of public spending patterns reveals impressive expansion of both social spending and poverty-targeted spending, There is apparent cause for concern over declininglevels of economic sector spendinginthe area of infrastructure and energy, and over fiscal outlays, which are significantly constrained by Mexico's unusually low tax effort. Butwho benefits from the observedpatterns? Thedist~bution~lincidence of spending We now turn to the question of the incidence of spending patterns and the potential impact of spending on the welfare of different groups inthe population. An assessment of these matters i s possible for those programs that can be linked to household characteristics and program use in the ENIGH. This applies to spending equivalent to over half of total programmable spending and almost 10% of GDP in 2002. There i s a strong bias in the analysis toward social spending plus subsidies -few income generating programs can be linked to specific households. We present the results of a comprehensive analysis that was undertaken for both the Public Expenditure Review (World Bank, 2004) and this report. This finds enormous variation in the distributional incidence of programs. More details will be available in World Bank 2004 (see also Scott, 2003, for a closely comparable earlier analysis limited to the 2000 ENIGH). The coverage of the analysis i s summarized in Table 4.3. Of spending covered in 2002, half is due to education, a quarter to health, 11% to pensions, 5% to two transfer programs ( O p o r t ~ n ~ dand ~ ~ eProcampo) and 6% to the residential component of the electricity subsidy. Note that the pension spending only includes the parts that are financed by the Federal Budget. For IMSS this includes payments to existing pensioners over the transition to the new system plus a social quota into the funds of current "Economic sector" spending, particularly rural development spending, also includes poverty-targeted programs. 117 contributing workers and a minimumpension guarantee. (Both the latter are intended to give the new system a progressive element across contributors). For ISSSTE it includes federal transfers to cover the deficit of the pay-as-you-go system that operates for public sector workers. The electricity subsidy does not include subsidies to agriculture which were almost 6 billionpesosin2002, about a sixth of residential subsidies. Table 4.3 Public expenditure on "redistributive" programs (inmillions of 2002pesos) Distribution 2000 % 2002 % Change of additional resources Total 539,456 100% 587,124 100% 8.8% 47,669 % Programmablespending 56.5% 54.4% % GDP 8.8% 9.5% Education 276,100 51.2% 298,404 50.8% 8.1% 46.8% Preschool 26,048 4.8% 29,076 5.0% 11.6% 6.4% Primary 101,445 18.8% 103,752 17.7% 2.3% 4.8% Lower secondary 56,131 10.4% 60,563 10.3% 7.9% 9.3% Upper secondary 37,964 7.0% 44,962 7.7% 18.4% 14.7% Tertiary 54,511 10.1% 60,051 10.2% 10.2% 11.6% Health 147,331 27.3% 155,946 26.6% 5.8% 18.1% Uninsured(SSA) 52,957 9.8% 55,435 9.4% 4.7% 5.2% Insured(IMSS) 75,437 14.0% 79,737 13.6% 5.7% 9.0% Insured(ISSSTE) 14,024 2.6% 15,101 2.6% 7.7% 2.3% Insured(PEMEX) 4,913 0.9% 5,672 1.0% 15.4% 1.6% Pensions 54,613 10.1% 66,714 11.4% 22.2% 25.4% IMSS active workers (1995 Law) 11,950 2.2% 12,380 2.1% 3.6% 0.9% I M S S pensioners(1973 Law) 32,476 6.0% 39,606 6.7% 22.0% 15.0% ISSSTEpensioners 10,188 1.9% 14,728 2.5% 44.6% 9.5% Directtransfers 22,306 4.1% 28,854 4.9% 27.9% 12.9% Oportunidades 10,265 2.0% 16,105 2.9% 56.9% 12.4% Procampo 11,595 2.1% 11,851 2.0% 2.2% 0.5% Electricityresidentialsubsidy 39,106 7.2% 37,206 6.3% -4.9% -4.0% Source: Backgroundwork for World Bank (2004). SEP (DGPPP); SSA (Boletinde Informacidn Estadjstica, CuentasNacionales y Estatales de Salud); IMSS (Coordinacirinde Presupuesto, Contabil~ad y Evaluacirin Financiera);Anew del Tercer Informe de Gobiemo (2003); Cuenta de la HaciendaPciblicaFederal (2002). To analyze the incidence of government spending administrative estimates of spending on programs are allocated across households in line with their use of the program. Households are ranked in terms of "autonomous" spending per capita or spending pre- monetary transfers. This concept is directly comparable to autonomous income (or incomes pre-monetary transfers) with the assumption that the marginal propensity to consume from transfers i s 50% (For details on the methodology see World Bank, 2004). One way to summarize the incidence of spending i s in terms of a "concentration curve" that are Lorenz curves of the share of a particular spending category ranked by the (autonomous) per capita household spending of individuals. A concentration coefficient can then be calculated that i s directly analogous to the Gini coefficient for incomes. 118 Concentration coefficients lie in the range of -1 to +l. values occur with Negative absolutely progressive {or pro-poor) spending; a value of zero is equivalent to exactly equal distribution of spending while positive values occur with absolutely regressive {or pro-rich) spending. Where spending is absolutely regressive but less so than the distribution of private spending or income it is typically described as relatively progressive. This can be illustrated for the aggregate effects of all spending categories analyzed. As Table 4.4 shows, the estimated total effect of spending programs analyzed is very close to being equal across the population in 2002. Programs used by the poorest decile of the population accounted for 9.5% of government spending while those by the richest decile accounted for 9.1%. The concentration coefficient for public spending was 0.03 compared with 0.48 for private spending, The public spendingprogramscovered are as a whole much more equal than private spending (and income) and experienced a modest gain inprogressivity in2002compared with 2000. Table 4.4 The distributionof spendingonallpublicprogramsanalyzedwas roughly equal2000-02* National Urban 3uraI Deciie 2002 2000 2002 2000 2002 2000 1 9.5% 8.9% 8.3% 8.6% 8.7% 9.3% 2 8.8% 8.9% 9.2% 9.0% 10.2% 9.7% 3 9.9% 9.1% 9.1% 8.4% 8.8% 9.4% 4 9.2% 8.6% 9.3% 9.5% 9.6% 9.5% 5 9.9% 9.4% 10.0% 10.1% 10.1% 10.3% 6 9.7% 10.1% 10.4% 11.4% 10.3% 9.9% 7 10.5% 11.6% 12.0% 10.6% 10.7% 10.3% 8 12.0% 11.2% 11.3% 12.1% 10.3% 9.3% 9 11.4% 12.2% 10.8% 11.6% 9.9% 11.2% 10 9.1% 10.0% 8.3% 8.6% 10.5% 11.2% Urban 72.5% 73.8% Rural 27.5% 26.1% Concentration 0.030 0.053 0.018 0.041 0.016 0.030 coefficient *Inthecaseof pensionsonly government transfers to activeworkers affiliated with IMSS are includedinthe total. Source: Backgroundwork for World Bank, 2004. This overall picture hides enormous variation in incidence with some programs highly equalizing and others highly disequalizing. This i s summarized in Figure 4.6, which ranks programs by their concentration coefficients. All ISSSTE programs analyzed (for pensioners, active workers, and health) are highly regressive, often even more unequal than the highly unequal autonomous spending. Tertiary education i s also highly regressive and IMSS pension-related programs only slightly less so. There are then a set of programs, including hospital health, IMSS health and electricity subsidy that are absolutely regressive, but muchless so than autonomous spending. Maternal andprimary health i s mildly regressive. Finally, there i s a set of programs that are either mildly progressive {Procamp, lower secondary and pre-schooling) or quite strongly pro-poor 119 (primary education, SSA health services) with Oportun~dadesdisplaying a very strongly pro-poor orientation. Figure 4.6 There is hugevariation inthe inequality of different government programs (Concentrationcoefficients for 2000 and2002) -0.8 4.6 4x4 -0.2 0.0 0.2 0.4 0.6 0.8 Note: The GINI refers to the distribution of private income. Procampo is not reported in 2000 becausethe sample of beneficiaries captured by the 2000 survey is not representative. VAT fiscal spending refers to exemptionsand the zero VAT rate. Source: Background work for World Bank (2004) and SHCP(2004) for the VAT fiscal spending. When we look at distributions within urban and rural areas the overall patterns are maintained, but almost all programs are relatively more progressive (or less regressive) within the urban population than within the rural population (Figure 4.7). This in large partreflects lower levels of all incomes and spending inrural areas discussed inChapter 3. For example, ~ p o r t u n i d ~ e ~ a relatively large fraction of the rural population covers (andso is only mildlyprogressive) but a small fraction of the poorest inurbanareas (and so is highly progressive). Similarly, only a small fraction of relatively better-off rural dwellers have rightsto ISSSTE and M S S or go to tertiary educational establishments. 120 Figure 4.7 The inequality of rural and urban spending on programs in 2002 (concentration coefficients) JSSSTEPensioners IIRural I3Urban -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 Source: Backgroundwork for World Bank, 2004. There were modest improvements inthe progressivity inmany (though not all) categories of spendingbetween 2000 and2002. The marginal incidence of public spending-that i s the distributional incidence of changes of spending- i s of particular interest to assess who benefited from expansions inprogram spending. To see this more clearly it i s useful to look at changes over the past decade. This is illustrated for the case of education in Table 4.5. Marginal incidence of spending on preschool, primary and lower secondary was strongly pro-poor (with the expansion of lower secondary to the extreme poor particularly notable) while spending on upper secondary was mildly regressive, and on tertiary education strongly pro-rich. Since private returns to tertiary education are also very high (albeit falling slightly inthe recent past) this amounts to rising subsidies to one of the principal mechanisms for the reproduction of inequality. Inhealth, the marginal incidence of SSA services has been strongly pro-poor in the 1996-2002 period while those of M S S services have been slightly regressive. Marginal incidence of public spendingon pensionshas been strongly regressive. 121 Table4.5 Thedistributionofadditionalpubliceducationspendingbyeducation level 1992-2002 Lower Upper Deciles Total Preschool Primary SecondarySecondary Tertiary 1 12.40% 18.70% 18.30% 16.70% 5.50% -0.90% 2 11.30% 11.60% 17.00% 16.10% 6.40% -0.20% 3 12.50% 14.30% 13.40% 16.10% 11.60% 5.90% 4 10.90% 13.50% 11.80% 9.70% 10.60% 9.50% 5 9.20% 10.50% 11.10% 6.30% 9.30% 8.40% 6 7.50% 8.20% 7.50% 6.40% 6.50% 9.30% 7 9.60% 9.20% 7.90% 8.80% 13.60% 9.60% 8 13.90% 6.20% 7.40% 9.30% 16.10% 35.30% 9 8.40% 4.80% 5.00% 9.40% 14.00% 9.90% 10 4.10% 2.90% 0.40% 1.30% 6.30% 13.10% Urban 73.70% 67.80% 64.40% 63.40% 84.10% 96.90% Rural 26.30% 32.20% 35.60% 36.60% 15.90% 3.10% Source: ENIGH 1992-2002, table 2.2.3. Population deciles orderedby per capita expenditure. The incidence of public spending can be considered a measure of government egorts across population groups. It is another matter to assess the benefits of such spending. A full analysis would need to take account of the value that different groups attributed to the government programs, net of any costs they occurred, as well as any second-round behavioral and external effects induced by the programs. There is not the information base to do this. As an alternative, albeit highly imperfect proxy for the benefits, the value of the programs is often equated to their cost in this type of analysis. With this assumption it is possible to assess the benefits in relation to the autonomous spending of different population groups. The benefits of public spending relative to autonomous spending is very much larger for poorer than richer households, As Figure 4.8 illustrates, the government programs analyzed were equivalent to over 140% of autonomous spending for the first decile and almost 80% for the second decile, compared with less than 20% for the top two deciles. It i s also interesting to see the sources of these benefits. For poorer groups, primary education i s the most important, followed by SSA health services, O ~ o r ~ and ~ ~ ~ ~ ~ e s secondary education. This rankingreflects both the levels and distribution of government spending: although primary education i s less progressively distributed than Oportunid~es(Figure 4.6) aggregate public spending was some seven times greater in 2002 (Table 4.3). These gains are much larger in rural than in urban areas with larger gainsfor the first through the fifthrural deciles than the first urbandecile, reflecting both greater ruralpoverty and the reachof government programs inrural areas. 122 Figure 4.8 Publicexpenditureis muchlarger in relationto private spendingfor the Poor The ratio of programspendingto autonomous household expenditureby major program, national, 2002 10 primuy mu. 9 IISSA oportunidades 8 LowerSec. Edu. IIPreschool 7 IMSSHealth .-3 "ampo CI 6 0 E3UpperSec. Edu. A 5 IIElectricitySubsidy 4 0PEh4EXHealth E4JMSS Active Workers 3 IIISSSTEHealth 2 IITertiaryEdu. 1 0% 20% 40% 60% 80% 100% 120% 140% Source: Backgroundwork for World Bank,2004. The net effect of government programs is to significantly reduce inequality once government programs are "added" to autonomous spending (on the assumption that individuals value government programs and their autonomous spending equally). In assessing this quantitatively it is important to note that measuredprivate spending in the ENIGH is substantially less than that from the national accounts. While this partly reflects different concepts some part i s likely to be due to under-recording of spending. By contrast, public spending i s almost certainly not under-recorded. Table 4.6 illustrates the overall impact on the distributionof spending under two assumptionsfor autonomous spending: using actual measures inthe ENIGH and adjusting upwards all spending equi- proportionately to bringit inline with the national accounts. Includingpublic transfers in cash and kind leads to large reductions in inequalities with the share of the poor rising significantly and the overall Gini falling from 0.48 for autonomous spending to between 0.39 or 0.43, dependingon the assumptions. 123 Table4.6 The distributionof householdexpenditurebeforeandafterpublic transfers,2002 After publictransfers* Autonomous Distribution Changeinshares expenditure National National Accounts Accounts 1 1.5% 3.1% 2.4% 101.6% 57.9% 2 2.7% 3.9% 3.4% 44.2% 25.2% 3 3.6% 4.9% 4.3% 33.2% 18.9% 4 4.7% 5.5% 5.2% 18.8% 10.7% 5 5.8% 6.6% 6.2% 13.9% 7.9% 6 7.0% 7.5% 7.3% 7.5% 4.3% 7 8.7% 9.0% 8.9% 4.2% 2.4% 8 11.2% 11.3% 11.3% 1.5% 0.8% 9 15.7% 14.9% 15.2% -5.3% -3.0% 10 39.2% 33.3% 35.9% -14.9% -8.5% Gini 0.481 0.393 0.431 -18.2% -10.4% *The distributionaftertransfers i s estimatedusingthe ENIGH (2002) for the distributionof public expenditureandusingtwo alternativesources for total householdexpenditures:(a) ENIGH(2002) and (b) NationalAccounts. Source: Backgroundwork for World Bank, 2004. This analysis of the distributional incidence of a subset of public programs has shown that this part of public spending i s much more equally distributed than private spending but with very wide variation, from highly disequalizing to highly equalizing programs. Some programs are having a large impact on the poor -assuming the spending is translated into real benefits and not offset by low service quality, for example. The progressivity of Oportunid~esi s striking, especially for a program that was already some 1.6 billion dollars and reaching 4.24 million families (about 21 million mexicans, three-quarters of them livinginrural areas). Inconcluding this section it is important to note that the analysis covered somewhatover half of programmable spending. It is not possible to undertake a comparable analysis of other spending categories. However, there are some reasons for supposing that they will tend to be regressive, at least more so than the average, proportional incidence of the programs analyzed here. To the extent that the benefits of economic sector spending are reflected in the distribution of income growth, we know that these have been distributed inahighly unequal fashion both across population groups (albeit with some equalization in 2000-2002) and to some extent geographically (Chapter 3.) Within economic sector spending, agriculture programs, with the exception of Procampo, generally have a bias toward large agricultural producers (World Bank, 2000). One significant item in non- programmable spending relates to the bailout for the financial sector in the wake of the 1994-95 crisis. Recent analysis suggests that this was highly regressive (De Ferranti et al, 2004; Halac and Schmukler, 2003). These factors need to borne inmindboth informing an assessmentof the overallimpact of the federal budget on distributionand the scope for more equitable public action. 124 Thereach of publicprovisioningin relation topatterns of deprivation The previous section looked at the distribution of public efforts across the population with a very rough estimate of a monetary equivalent of the benefits to different groups. This section takes a further step in relating programs to patterns of deprivation by comparing the reach of key programs to the needs of different groups. This takes as its point of departure the life-cycle approach of CONTIGO's conceptual framework. As shown in Figure 1.4, this framework seeks to respond to the changing needs of different groups over the life cycle with respect to objectives of social protection, expansion of human capacities, providingincome opportunities and the means to accumulate assets. A comprehensive analysis of this would involve undertaking a matching of the needs of different groups against both the potentiuZ for government action to make a difference and the detailed current efforts of the government. This was outside the scope of this report -indeed the information requirements for conducting such an exercise are not readily available. Instead a more aggregative and illustrative approach i s taken. We first revisit the major areas of deprivation andthen look at the survey evidence on the reachof selected programs. This is supplementedby a brief examination of the picture provided from administrative budget sources. Inthe spirit of CONTIGO's framework, Table 4.7 presents selected indicators for areas of deprivation over the life cycle (these are often termed "risk" indicators but we prefer to confine the use of the term "risk" to questions of shocks and fluctuations). Which aspects of deprivation are salient varies with age. For the very young, issues of protection against shocks and human capacity formation are closely intertwined, especially around good nutrition andhealth. For children and youth, capacity formation through formal schooling i s fundamental. For adults, income-earning opportunities, and the means to accumulate assets becomes the primary concern, as well as protection of themselves and their families against a variety of potential shocks -labor market-related, weather, natural disasters and health. For the aged it i s the capacity to finance their consumption, whether through a formal pension, savings or family support. Finally, some aspects of deprivation affect all age ranges -we include access to basic services and also the presence of absence of external remittances that i s a special factor for both income levels and managing shocks inMexico. The messages of Table 4.7 are clear. First, despite substantial progress in service provision and human indicators (see Chapters 2 and 3) there remain major gaps in deprivation. Second, there is a strong relationship between gaps and measuresof income poverty with the extreme poor systematically suffering greater deprivation (e.g. out of school, child labor, low skills and limited social security coverage). As elsewhere in this report, we use, for presentational purposes, the three categories of extreme poor (below the food-poverty line); moderate poor (between the food and "patrimony" poverty line) and non-poor but it should be remembered that this is only a way of organizing essentially continuous relationships as discussed inChapter 1. 125 Table 4.7 Indicators of deprivation over the lifecyclein2002 Population Area of deprivation Rural Urban group Extreme Moderate Non- Extreme Moderate Non-poor poor poor poor poor poor Neo-natallinfantmortality n.a n.a n.a n.a n.a n.a 0-4 Malnourished 18%of all childrenare stunted, probably concentratedamong 5 Does not attendpre-school 25.3 14.6 4.5 24.4 16.8 8.7 - Outof school(6-11) 3.9 3.1 0.9 5.2 2.0 1.1 14 Out of school(12-14) 16.2 14.9 6.7 11.5 12.8 3.6 Proportionnotincorrectgrade for age (12-14) 53.1 43.1 29.1 50.2 30.2 22.5 Childlabor(12-14) 14.0 18.3 15.5 10.1 9.9 4.8 Doesnot attendupper-secondary(15-17); (Out 78.8 64.6 44.7 66.0 57.5 33.5 of schoolor enrolledinless than upper- secondary) Doesnot attendupper-secondary(18-24); (Out 99.5 95.0 86.6 93.3 87.6 70.4 of schoolor enrolledinless thantertiary education) 15-24 Unemployed(15-24) 4.6 6.3 6.5 10.5 10.0 5.O -Male - Female 5.1 6.1 8.6 12.3 11.1 5.8 3.3 6.7 2.0 7.5 7.9 3.9 Inactive"(15-24) 36.0 29.5 20.5 32.5 26.7 13.6 - Male - Female 9.9 9.4 10.2 17.1 12.5 5.6 57.7 49.0 31.3 45.5 39.6 22.2 Low skillsb(25-40) - Male 55.4 34.8 18.0 29.0 14.6 3.9 50.4 31.0 14.8 26.2 14.4 3.7 - Female 59.4 38.1 20.7 31.2 14.7 4.2 Low skillsb(41-64) 85.4 77.3 53.2 62.7 42.2 21.1 - Male 79.9 73.6 50.8 61.4 37.4 18.1 - Female 90.7 80.6 55.1 63.8 46.3 23.8 25-64 Unemployed 0.4 1.3 1.1 4.2 2.4 2.1 - Male 0.5 1.8 1.3 5.7 2.2 2.4 - Female 0.1 0.3 0.7 1.7 2.7 1.6 Underemployed` 43.7 34.0 31.2 28.3 21.5 18.6 - Male 28.2 19.4 21.8 12.5 10.3 10.2 - Female 77.7 61.4 44.8 54.7 43.0 30.6 65 andover Does not receivepension 98.8 93.1 86.8 92.3 77.0 66.1 Had"catastrophic" healthexpensesd 46.8 36.5 26.3 47.2 35.0 21.7 Non-age Doesnot receiveremittancesfrom abroad 92.3 89.3 88.3 97.3 96.2 97.7 specific Nopipedwater 34.9 22.1 10.1 14.0 5.1 0.9 No sewerage 80.4 57.4 33.7 23.0 10.9 2.7 Noelectricity 9.3 4.1 1.1 0.6 0.2 0.0 a. Neitherat schoolnor inlabor force b. Noeducation or incompleteprimary c. Works lessthan35 hours (as proportion of labor force) d. Following SSA's definition which is 30percentof disposableconsumption, after spendingonfood, educationandhousing. Source:WB staff calculations from ENIGH 2002. We next look at the coverage of some of the major government programs following broadly the same headings as in Table 4.7. For those programs captured by the ENIGH we can effectively use the same categorization. Table 4.8 presents specific programs associated with areas of deprivation. Note that those areas of deprivation related to lack of access or use of a service inTable 4.7 -being out of school, lacking water supply and so on- are, to a large extent, the opposite of government programs of service provision. These are not repeated here to save space though such programs constitute an important part of government efforts. The messages are again quite clear. There are some programs that have significant reach amongst the poor population: ~ p o ~ (discussed ~ ~ ~ ~ ~ u ~ further inthe next section) and Procumpo inrural areas, SSA services, as well, of course, 126 as basic schooling, water and electricity (as discussed in Chapter 3). But for many the reach of services is inversely related to patterns of deprivation with greater coverage of the non-poor. This applies, for example to training, scholarships to higher education, access to formal health andemployment protection and pensions, and access to credit and savings programs. Table 4.8 i s mainly about specific programs. A whole domain of public action not included concerns policy influences on the expansion of income-opportunities many of which are not tightly linkedto particular population groups (though see an example from agriculture in the section on geographic equity below). These issues are discussed in Chapter 5. Table 4.8 Indicators of the coverage of selectedprograms inrelation to deprivation over the life cvcle Population Program related to areasof deprivation Rural Urban group Extreme Moderate Non- Extreme Moderate Nan- Poor poor poor poor poor poor o-4 Utilizes matemity care n.a. n.a. n.a. n.a. n.a. n.a. ParticipatesinOportunidades n.a. n.a. n.a. n.a. n.a. n.a. 5 Attends pre-school program 74.7 85.4 95.5 75.6 83.2 91.3 ReceiveOporfunidades grant for schooling (6- 27.4 21.7 7.8 1.8 0.9 0.1 11) 6-14 ReceiveOportunidadesgrant for schooling 63.2 44.6 15.3 6.3 2.9 0.3 (12-14) 15-24 Receivesscholarshipto attendschool" 21.7 24.0 13.5 6.3 2.2 4.3 Receivestraining 0.5 5.0 17.2 4.7 12.6 32.9 Receivesaywla alimentariaor despensa 0.6 2.0 8.7 2.6 7.9 16.7 ReceivesProcampo 10.6 6.9 4.4 0.1 0.2 0.1 Status at work: 25-64 Employer 2.2 2.7 6.4 1.2 2.1 6.7 Selfemployed 48.1 37.2 31.0 29.2 24.0 17.3 Informalsalaried 38.7 35.7 22.9 37.7 28.3 14.4 Formalsalaried 4.9 17.9 34.0 27.3 41.7 57.6 Other 6.1 6.5 5.7 4.6 3.8 4.0 65 andover Receivespension 1.2 6.9 13.2 7.7 23.0 33.9 Uses SSA services 29.3 25.8 11.8 16.2 10.0 3.7 Rightsto IMSSJISSSTE health services 2.9 12.6 27.6 19.5 36.3 53.2 Non-agespecific Access to housingcredit 0.5 3.9 14.4 6.4 17.6 33.4 Access to workers' credit fund (FONACOT) 0.0 0.8 4.1 1.8 5.7 10.6 Access to other savings fund 0.3 2.7 8.6 3.7 8.5 21.3 a. Includes alleducationallevels (not only highereducation).Crf!ditos educafivosweren't considered. b. Includes only working people. Source:WB staff calculations basedon ENIGH2002. This analysis of the reach of programsis biasedtoward programs that are includedinthe questionnaire of the ENIGH and so far gives a very incomplete picture of government programs to reduce poverty. For a broader picture it is necessary to go to administrative data. A matrix of programs by the Social Cabinet secretariat is available uponrequest and a summary by category of programs is inTable 4.9. According to the government's count there are 207 social programs -and note that this excludes many directly productive oriented activities, for example in the agriculture sector. There are, for example, some 48 programs in education (18 alone for education 127 quality andmanagement), 48 inhealth, 21on income generation, 35 concernedwith asset accumulation and 14 on social protection. Forming an overall assessment of these programs provedi s very difficult for this report. Very few have rigorous evaluations that allow assessment of effectiveness, despite important efforts by the government (notably SEDESOL) and congress to strengthen evaluation (see Chapter 6 for a discussion of strategies on evaluation). Table4.9. Numbersandspendingonsocialprograms,2002 TYPEOFPROGRAM Numberof % of budgetspenton 2002budget programs programsthat are allocationMxP poverty- targeted (millions) HUMAN DEVELOPMENT Education 48 29.2 116,799 Health 48 16.7 22,738 Food Supply and Nut~tion 8 50 2,223 Equity, Social Cohesion,Quality of .Life 14 35.7 158 and Rights TOTALHUMAN DEVELOPMENT 118 263 141,918 INCOME GENERATION 31 41.9 7,109 ASSETACCUMULATION 35 14.3 3,308 SOCIAL PROTECTION 14 14.3 84,868 CROSS-CU"G* 6 83.3 19,377 OTHERS 3 0 34 TOTAL 207 27.1 256,617 * Includes Oportunid~es,Estrategia de Microrregiones, Programs Estatalespor Demanda, Program Ha'bitat, Superacidn de la Pobreza Urbana, andAtencidn a JornalerosAgricolas. Source: Social Cabinet Technical Secretariat.( 1) Includes programs from the Ministry of Labor, the Ministry of Agriculture excluding Procampo which is not strictly considered a social program, and the Ministryof the AgrarianReform. If we bring together this general discussion of reach with that of incidence we can see that there is considerable variety of programs from the perspective of poverty and distribution. First, programs of basic service provision have extensive (and rising) reach throughout the population: some, such as basic education, have modestly favorable incidence; others, such as electricity, have extensive reach but are inequitable in incidence. Second, Oportunidades stands out amongst programs for which we have information as havingextensive reach and pro-poor incidence. Third, there appear to be a multiplicity of poverty-focused programs with only modest reach. These may be effectively reaching specific groups or this may be a sign of excessive proliferation of programs. And fourth, there are programs that have both low reach and are inequitable in incidence -from tertiary education to training. These programs are clearly not justified on poverty or distributional grounds; the economic case for them would have to lie on efficiency benefits. We have emphasized the difficulties of forming a comprehensive assessment of the pattern and design of programs. The government i s inthe process of re-orienting its own approach to assessing and managing programs. This will involve two features of relevance here. First, each sectoral ministry i s specifying a limited number of key goals and linking its own programs to pursuit of results toward these goals. Second, the Social 128 Cabinet secretariat will shift from seeking to track and coordinate all social programs, to focusing on a small subset of programs that the government judges to be flagship endeavors inthe fight against poverty that will receive particular attention with respect to design, evaluation, and outreach. Both of these moves are sensible and should facilitate future managementof social programs. In concluding this section, it should be emphasized that this analysis of reach and incidence says nothing about program effectiveness. We discuss this selectively in the second part of this chapter, in the subsections on design of social protection and the issues of service quality. Optionsf o r ~ n a n ~ i nthe expansionof poverty-orientedprograms g The section on the incidence of public spending showed that the government has experienced major achievements inthe use of the budget to reach the poor and to reduce inequality. However, the section on the reach of public programs showed that there are major uncovered needs. Moreover, there is at least suggestive evidence that economic sector spending needs to be substantially expanded to achieve equitable growth, that is crucial to reducing income poverty (see also Chapter 5 for cross-country evidence on infrastructure needs). This raises the question of financing. Mexico i s fortunate inhaving a sustainabledebt position thanks to prudent fiscal management since the 1994-95 crisis. But it faces expanding spending needs from the transitional spending for the IMSS reform and contingent liabilities under an ISSSTE reform that will have to take place sooner or later. This section illustrates two ways in which poverty-oriented programs could be financed -whether the spending i s inthe social sectors or ineconomic sector spending oriented to equitable growth. First, the government can undertake a tax reform to increase the aggregate tax effort. Second it can redistribute spending from more disequalizing to equalizing spending. Fromthe perspective of distributionalimpact, a good point of departure i s information of the kind presentedin Figure 4.6 that, together with budgetary information presented in Table 4.3, can be used to suggest possibilities for redistributive shifts in spending. But distributional incidence i s not the only criteria -it i s also necessary to take account of the effectiveness and efficiency impacts of programs, as well as the feasibility of making spending shifts. The preferred approach should be to shift spending from disequalizing spendingon programs that do not make sense on efficiency grounds to programs that are both equalizing and good for efficiency. Such shifts will not involve tradeoffs between equity and efficiency, though they will, by their very nature, involve shifts between groups. The feasibility of this will dependon political-economic factors. There is considerable scope for redistribution within the mexican federal budget that would either be efficiency-improving or have small adverseefficiency effects. Looking at disequalizing programs inFigure4.6 there is first a set of programs associatedwith IMSS and ISSSTE on pensions and health. For pensions, part of the reason for regressivity 129 reflects the workings of the labor market 4 M S S and ISSSTEpensioners are relatively well educated and intheir working lives were explicitly or implicitly setting aside part of their higher salaries for old age.6 This will become fully explicit for participants in the reformed IMSS pension system. However, it may also be the case that current IMSS and ISSSTE pensions reflect transfers unrelated to efficiency-related performance. It would, for example, be feasible and equitable to tax pensions as many countries do. This would have little or no adverse impacts on effi~iency.~Second, tertiary education i s highly unequalinits distributional incidence and, as notedabove, is associatedwith highprivate returns that provide strong incentives for household investment. It would be more equitable to increase cost recovery and reallocate the resources to bursaries or loan programs for poorer groups or to strengthening educational expansion at lower levels (upper secondary being the key dividing point between richer and poorer households now.) Third, the residential electricity subsidy i s now both disequalizing and clearly efficiency- reducing since it encourages higher consumption rather than energy saving. The most strikingly inappropriate element of the schedule are subsidies for hotter regions that also happen to be richer (e.g. Baja California). An efficient policy -that i s standard in electricity pricing regimes- i s to charge a premium, not subsidize, such electricity consumption since it can lead to increasedinvestment needs to expand peak capacity (for a detailed analysis see World Bank, 2004). To explore the potential gains from redistribution within the budget we choose two options for increasedresources, although there mightbe other possible options. First, to illustrate one possible tax reform we consider the removal of Value Added Tax (VAT) exemptions. While their removal makes sense on efficiency grounds, this would impose costs on the poor in the absence of action to direct the additional resources equitably. We select this not because it is the preferred reform -this report has no position on the best tax reform for Mexico. Designing a tax reform has to balance efficiency, equity, administrative and political considerations and this is a complex discussion that continues in Mexico. This particular reform was rather chosen becauseit i s relatively easy to do the quantitative analysis and i s one of the less advantageous reforms for the poor. If a more progressive tax reform were chosen, that would only increase the potential distributional benefits but possibly at a cost in efficiency or administrative feasibility. Second, we consider cutting or removing the residential electricity subsidy, whose reduction i s robustly justified on bothefficiency and equity grounds, and i s already being gradually pursued by the government (the gains to the government from reducing the electricity subsidy would come in various forms, both directly via immediate cash savings from current budgetary transfers and indirectly through a better profitlloss position of the Federal Electric Commission [CFTi]. The purpose of these scenarios is only illustrative anddoes not go into specific issues of design andpotential fundflows). This is true for the implicit labor contract within a pay-as-you go pension systemas well as for schemes that explicitly involvecontributions. Intheory increasedtaxation ofpensionscouldaffect labor market decisionsthough this effect is not likely to be large. 130 It is then crucial to reallocate resources to effective equalizing programs. We choose to use as a base for analysis Oportunidades, that i s both highly redistributive now and has positive efficiency effects through the incentives provided for schooling of poor children, as extensively discussed elsewhere. The potential shape of an expansion in design terms i s discussed in the next section -here we are concerned only with illustrating potential distributionalimpacts of spendingreallocation. Table A.4.1 in the annex presents the current (2002) distribution and incidence of Oportunidades transfers, the residential electricity subsidy and the VAT exemptions on food, medicines and other goods and services. In all cases we consider transfers net of administrative costs which are small in the case of Oportun~dades(though could potentiallybe larger inanexpandedprogram). For the total value of VAT exemptions we use the estimate from CIDE-ITAM (2003) which is more conservative than the estimate by SHCP (1.3% vs. 1.9% of GDP), For the calculation of RES see the appendix. Considering the joint redistributive impact of the three instruments we see that the residential electricity subsidy and the VAT reverse the highly progressive distribution of Oportunidadesand increase the resources available to the poorest only modestly despite their relative size (Oportunidudes currently distributes only 12% of the total transfers allocated through the three instruments). Table A.4.2 presents a variety of scenarios involving complete and partial reallocations fromthe residential electricity subsidy andVAT to Oportunidades,as well as reforms in the targeting mechanisms applied by Oportunidades and RES. In the case of Oportunid~deswe consider expansionswith the current distributionof benefits as well as expansions assuming full coverage to be achieved within the poorest 20% of the population. In the case of the electricity subsidy, we consider, as well as its total elimination, a reduction (and reallocation to Oportunidades) of 50% in two alternative scenarios: a) maintainingthe current distribution, andb) targeting only the poorest half of the population with a proportional (flat) distribution among beneficiaries. The redistributive opportunities suggestedby this analysis are not trivial, The impact of these resources on the poorest decile can be raised from a current 22% (of autonomous per capita household income) to a potential maximumof 177% with a full reallocation of the residential electricity subsidy andVAT expenditures to Oportunidudes,with the latter reformed as noted (Table A.4.2). The gain to be obtained from a total reallocation of RES to Oportunidades i s comparable to what could be obtained by eliminating VAT exemptions and allocating the proceeds as 20% to Op~rtunidadesand 80% as the overall (average) incidence of public spending estimated inWorld Bank, 2004. Figure 4.9 uses two scenarios to illustrate the redistribution impacts on Oportunidades. The first distributes the resources in line with existing distributions while the second assumes the 2002 gaps incoverage of households by Oportunidades would be fully filled for the bottomdecile, 50% for the seconddecile and 25% for the thirddecile. Under these assumptions the bottom three deciles would experience large gains -making a 131 substantial impact on extreme poverty- while the fourth and fifth deciles would experience small losses. While these are only illustrative scenarios they shows the great potential for more effective use of the budget as an instrument of efficient redistribution. Figure 4.9 Closingthe poverty gap: the impact onper capitahouseholdexpenditureof the bottomfive deciles of full reallocationof electricity subsidy and VAT subsidies to Oportunidades" 14,000 I P r e - t r a n s f e r =Post-transfer ( C u r r e n t ) 12,000 E P o s t - t r a n s f e r (Reallocation) 10,000 8,000 6,000 4?000 2,000 0 1 2 3 4 5 Deeiles based on Expenditures N e t o f Transfers per capita *Targeting reformrequires closing O p o ~ u n ~ d ~ dcoveragegap fully indecile 1,50% indecile 2, and 25% e s indecile 3. Source: Background work for World Bank, 2004. These simulations show that financing for poverty-oriented programs i s feasible. It i s of course crucial that expanded financing i s used effectively. That is why we used the example of ~ ~ o r t ~ ~ ~adprogram , ~ d e sthat appears to be highly effective inreaching the poor. However, in many areas where expansions are clearly needed, whether in social services or infrastructure, there are major questions over government efficiency and accountability. This raises the issue of the institutional basis for more effective and accountable service provision which forms one of the design issues taken up in the secondpart of this chapter. B. Selectedissuesinthe designof socialprograms The next two sections turn to a set of issues in design of policies and institutions. This section looks at social provisioning and especially at social protection and service quality. Section C turns to issues raised by decentralization. These are complex issues and the points raisedhere are intended as a contributionto debate and to future analysis. Mexico's system of socialprovisioningin transition InChapter 1it was noted that Mexico is going through two major longruntransitions - ineconomic institutions and policies and in political and social institutions. The second transition involves the deepening of democracy, increasing accountability of the state to its citizens anddecentralization to lower levels of governments. 132 Part of the heritage of the past was an important role for social development programs in the distribution of patronage, as a mean of sustaining political support. This can occur, for example, through the provision of public works or through public employment structures. Indeed under the administration of president Salinas, the major social project of PRONASOL has been documented to have been deeply embedded in a national clientelistic project shaped to create support for (a faction of) the PRI. (See Box 4.1). It is possible for programs partly motivated by patronage to bring real benefits but the accountability relationships are unlikely to be conducive to either high service quality or fully equitable participation inthe program.' Box4.1 PRONASOLand the politicaluse of programs The Program Nacional de Solidaridad (PRONASOL) was the principal poverty alleviation program of the Salinas administration and its budget grew rapidly in that period, from 0.2 percent to 0.6 percent of GDP between 1988 and 1994. It was essentially a program to provide local infrastructure. One of the program's most notorious elements was its "demand-driven" allocation mechanism using the organized participation of local communities, bypassing local government administrations and traditional PRI patronage structures, creating a parallel systemof Solidaridad committees. PRONASOL was discontinued in 1995, at leastunderthat label, following extensiveaccusations of its usefor politicalpurposes.Empirical studies support the view that it had relatively weak poverty targeting and was biased to specific political constituencies. The reaction to PRONASOL heightened concerns over transparency and public accountability in the Zedillo administration and was an important force behind the push toward more transparent forms of financing, and createdthe political space for the design and introduction of Progresa (now ~po~unidadesPRONASOL was also partially replaced by more formula-driven processes for ~ . allocation of resourcesto municipalities under the Fondo de Desarrollo Social Municipal (FDSM)in 1996, that was later replacedby the Fondopara la infraestructura Social Estatal (FAIS). Source: Estdvez, Magaloni, and Diaz-Cayeros(2002). Another part of the heritage of the past, that Mexico shares with muchof Latin America, was a ``truncated" welfare state, in which social security only reaches those parts of the population in formal work as already seen in the discussion of the reach of programs above. Both the overall institutional transitions and the specific reactions to programs such as PRONASOL ushered in a range of new directions to social provisioning that sought to increase efficiency and greater equity and in particular reach among the poor. Insocial protection, significant policy changes were the 1997 reform of IMSS pension system, the creation of PROGRESA and its subsequent expansion as Op~~tunidades, and the decline of food subsidies. Inbasic service provision there has been a range of policies that have sought to increase efficiency, quality and equity that range from privatization to a range of specific programs to raise educational standards. This has been accompanied by the extensive decentralization of education, health and basic infrastructure services to the states and municipalities financed by a mixture of unconditional tax sharing and earmarked transfers (see Section A.) All these areas have been complemented under the See Schady (2002) for an analysis of the political factors behind the allocation of the Peruviansocial fund inthe 1990'sthat is shownby Paxsonand Schady (2002) to bealso quite well targetedonpoverty grounds. See World Bank (2003b) for a general discussion on the relationship with accountability and service quality. 133 present administration with a major shift toward greater transparency as part of a broader concern with accountability mechanisms. This is clearly a major agenda and one that will involve substantial, long-term institutional (and behavioral) changes. The remainder of this section examines policy issues for the transition inthe areas of social protection and service quality. Thedesign of socialprotectionstrategy Social protection comprises the set of public interventions that are intended to support risk managementinthe population andprovide redistributive transfers to reduce chronic poverty. Such public interventions are only part of the picture of risk management and need to be understood within the context of the overall pattern of private and societal responses to risks and transfers. At a general level, risk management can be categorized into strategies that seek to reduce risk, provide for self-insurance and risk-pooling. Risks are commonly divided into idiosyncratic and covariate risks -and between small and large (or catastrophic) risks. Social protection instruments are typically about risk- pooling when private and community mechanisms are inadequate or costly. This i s more likely to be the case for large idiosyncratic risks (e.g. major health problems) and especially for large, covariate risks (e.g. natural disasters, crop failures or generalized unemployment increases.) While social protection is concerned with such risk-pooling and (transfers) public interventions can also help reduce risks (e.g. reducing agricultural production risk through irrigation in rain-fed areas or health risks through preventive health programs) and strengthen institutions for s e ~ - ~ ~ ~ ~(e.g. c e r a nthrough financial sector development).' In this section we look at the current design of the social protection system, its recent evolution and issues for design. As noted above, Mexico, like many Latin American countries, evolved a "truncated" welfare state with a sharply dualistic structure. A stylized account of the system is given inTable 4.10. This divides the population into the non-poor, the moderate poor and the extreme poor. This is somewhat arbitrary, since poverty lines do not mark dividing points inthe distributionof well-being (see Chapter 1) butit is ausefulorganizing principle that is usedby the government andelsewhereinthis report (ss in Chapter 3, we aggregate pobreza de capacidades and patrimonio into a single "moderate poverty" group, to simplifythe presentation). De Ferrantiet al. 2000. 134 Table 4.10 A Stylized presentationof the Mexico's Social ProtectionSystem by Poverty Status (coverage rates refer to 2002) Extreme Poor Moderate Poor Non-poor OldAge No significant coverage IMSS (19% with rights) I M S S (27% with rights) ISSSTE (3% with rights) ISSSTE(12.4% with rights) Other Federaland State social security institutes Major health Z~S~-Oportunidades IMSS (34% with rights) I M S S (47% with rights) risks Statehealth care ISSSTE (5% with rights) ISSSTE (15% with rights) systems Private providers Private providers SeguroPopular- Other Federaland State UniversalCoverage social security institutes Program (SSA) Segurode Sal& para a am ilia (IMSS) Private providers ~nem~loyment Very partialcoverage Partial coverage with formal Partial coverage from risks of PET severance pay and SICAT formal severancepay ~Oportunidades) Natural disasters Oportunidadesand FondoNacionalde Largely inapplicable and harvest FondoNacionalde DesastresNaturales failure DesastresNaturales Social Assistance Oportunidades(41.6%) Partial coverage of Not applicable and Children Sistemade Desarrollo Oportunid~es(16.1%) Integralde la Familia Sistemade Desarrollo W F I Integralde la Familia (DIF) The non-poor have extensive access to the formal security system (dominated by IMSS and ISSSTE) for old age and health risks. This is partly financed by contributions and partly by transfers from the budget, as we have seen above. There is no unemployment insurance but labor protection laws provide for severance pay -albeit in a manner judged by observers to be inefficient and costly for all concerned. The extreme poor have access to a set of services of programs that involve lower levels of transfer (~portun~dades), have low reach (Programa de Empleo Temporal, PET", for unemployment) or are generally judged to be of lower quality (for example SSA health provisioning)." Indeed, for health services, evidence on low quality is available from the opting out for private providers that occurs to varying degrees for all groups. B y way of illustration, 47% of the extreme poor suffer "catastrophic" health expenses (over 30% of lo is PET the largest public works program in Mexico. It was f i s t implemented as an emergency program during the crisis in 1995. The program's objective was to provide jobs in poor rural areas duringthe low agricultural season whenjob opportunities are scarce. This program offers jobs at two minimum wages or so, mainly to improve basic infrastructure, roads, highways, irrigation and reforesting projects. In 2000, over one milliontemporaryjobs were created. Since then, the number has been decreasingto about 900 000 in2002, Theprogram, which was createdas afederal programis now administratedby states. 11 Active Labor Market programs inMexico include SICAT, a training program for poor workers who lose their jobs, and CIMO (renamedPAC), that subsidizes on-the-job training in small firms. 135 spending on non-essentials) but only 29 percent make use of SSA health services and less than3% have rightsto IMMSor ISSSTEservices (Tables4.7 and4.8). It is also worth emphasizing that the moderate poor fall rather awkwardly in between protection systems for some risk categories -with relatively low participation in the formal system but also low access to Oportunid~es.With a few exceptions, neither the moderate nor the extreme poor have access to old-age pensions. The pension of the Distrito Federal (DF), which is financed by state revenues, is an exception. This appears to be reaching the bulk of the old inthe DF, and as a consequence, is not highly targeted to the poorest (whether that is judged to be good or bad depends on the goals of the scheme). Households respond to this structure inmany ways. Those who can, seek to place at least one member in the formal system in order to reap family benefits. But many households are excluded completely. This i s partly becauseof the costs of formalization -in paying taxes and inobserving labor and other regulations. While there is evidence that decisions of individuals to choose self-employment over formal work is often rational (World Bank, 1999c), this is conditional on the costs of formalization. Government strategy over the past 5-10 years has been dualistic: seeking to improve the efficiency of the formal social security system whilst extending or improving the parallel system for those excluded. The reform of IMSS is the most importance instance of efficiency-improving reforms in the formal system. The design and expansion of Op~rtunidadesis the most important program inthe parallel system. Inhealth there have been attempts to extend services to poorer, excluded groups, via IMSS SoZidaridad and the recentSeguroPopular (discussedbelow). The government i s keen to continue strengthening Oportunid~esto provide transfers andencourageinvestment inthe humancapital of the extreme poor. The government also wants to expand the healthcoverage for the moderate poor. What are the strategic questions for social protection in the future? Inthe long run, it would be expected that the formal system, appropriately reformed, would be extended to cover all groups inthe population. This is a feature of most OECD societies whose social protection systems typically combine a mixture of social insurance and social assistance elements -with different specific programs oriented to different groups. However, in the mediumterm, movingrapidly to an integrated, comprehensive systemwill be infeasible, given the current conditions. A more appropriate and achievable objective is to move toward a dualistic and comprehensive system with better efficiency and equity. We focus here on a set of design issues relevant to this, especially with respect to the extension of the parallel, poverty-oriented part of the system. The discussion is necessarily preliminaryandwill be the subject of more in-depthstudy inthe future. Within the formal social security system this is likely to involve a continuation of efficiency-improving reforms and reducing the costs of participation for firms and workers. High on the agenda are an ISSSTE reform, labor policy and other regulatory 136 reforms to reduce the costs of formality and inefficiencies of existing regulations. There i s probably also a set of measures to improve the institutional functioning within the formal system, for bothpension andhealth services, butthese have notbeen evaluatedfor this report. Such an approach should, over time, make the formal system both more efficient and equitable -especially through a shift to a contribution-based principle. It should also steadily increase the reach of the system that is likely to be of particular importance to the moderatepoor. Design issuesfor an extended, ~por~un~des-basedtransfer system Mexico's remarkable success with Oportunidades'* can provide the basis for filling in one part of a comprehensive system of social protection, that is primarily oriented to the extreme poor. The expansion and current reach of Oportun~~udes i s summarized in Figure 4.10. B y 2003 the program reached 3 million rural families and over a million urban families. This was equivalent to over 30% of all rural households and 55% of the extreme poor but still a much lower proportion inurbanareas -5% of the extreme poor. This is a truly impressive reach for a targeted program. If the pro-poor social protection system is to be expanded into a comprehensive safety net, with an extended U p ~ r t u ~ ~ ~asditsecore, this raises a set of design issues that we turn to next. Some of a s these could also increase its reach to the moderate poor. However, our understanding of the risk-management needs of the latter group and of policy instruments that could effectively respondto their needs is seriously incomplete. It will be one of the focuses of future work on social protection in 2005. Here we sketch some of the issues raised in a potential further expansion of Oportunidade~, Impact evaluations of this program are discussed in Chapter 6. ~ p ~ ~ ~ isna federally managed ~ ~ d e s program that involves a transfer to very poor households conditional on either their children attending school or the family attending health clinics. Orientation to poor households inrural areas was achievedby a two-level targeting process: first selecting localities with high levels of marginality according to CONAPO's marginality index; then selecting eligible households based on a points system, mainly using household assets. The effectiveness of targeting was already shown in the discussion of incidence using income-based poverty lines. The extensive evaluation work conducted by IFPRI (2000) has documented significant positive impacts on schooling attendance(especially'at secondary levels) and on health status of children, and negative impactson child labor. Since 2000, the programhas expandedsignificantly to cover most poor rural areas and into marginal urban areas with positive impacts especially for upper secondary attendance. (Rubacalvaand Teruel, 2003; Parker, 2003; Paqueo, Lopez-Acevedo, and Patrinos, 2003) 137 Figure 4.10 Number of beneficiaryfamiliesof O ~ o ~ n ~ ~ s 3,500 3,000 1,500 B% 1,500 1,000 300 1997 1998 1999 2000 2001 2002 2003 l R u r s l l U r b n n Source: Oportunidades. Figure 4.11 Proportionof extremepoor, moderatepoor andnon-poor families receiving~ ~ o r ~ transfers n ~ eins2002 60.0 R u r a l U r b a n - E x t r e m e P o o r &IM oderate P o o r "on-Poor - T o t a l Source: WB staff calculations usingENIGH2002. a) Balancing human capital formation and transfer objectives. A central feature of Oportunidudes,along with other conditional cash transfer programs, has been the use of one instrument to further two goals: transferring resources to poor households and providing monetary incentives for investment in the human development (or human capital) of children. While the analyses referred to above confirm the furtherance of both goals, design choices can make a large difference to the tradeoff between the goals. This can be illustrated for the education component of ~ p o r t ~ n i based on work by De d ~ e ~ Janvry and Sadoulet (2003). As currently designed, a high fraction of Oportunid~es resources are going to the children from poor families who would have attendedschool in any case. This i s illustrated by Figure 4.12, which provides the continuation rates from grade to grade for treatment and'control groups in the poor rural areas in what was then PROGRESA. This shows that the main impact of the intervention in this periodwas to 138 increase continuation rates inthe transition from primary to secondary schooling, from 67 to 76% -equalizing the continuation rate between the poor and non-poor in these communities. There was little impact on attendance and continuation either in primary (where enrollment rates are already high) or for those children who start attending secondary schooling. On the other hand, there is a large drop-off in attendance in the transition from lower to senior secondary schooling in both treatment and control villages. De Janvry and Sadoulet conclude that from the perspective of providing incentives for attending school amongst the out-of-school population of children, targeting could be improved. Almost all transfers provided at primary school levels have no effect on behavior. Even for those for children of secondary age amongst poor children some 65% would have attended in any case. On the other hand, in these poor villages, a quarter of children from poor families did not attend even with the offer of a transfer, while a quarter of children from non-poor families also didnot attend-but possibly would have ifoffered atransfer. Increasedeffects on attendancecouldbe achievedbytargetedlinked to the riskof not attending school (e.g. positioninthe family). Figure 4.12. Schoolcontinuationratesinpoor ruralcommunitiesintreatment and controlvillagesunder Progresu, late 19% (percent of cohort ina grade that proceedsto next levelof schooling) Continuationrate (?/o) Lower secondary school . . 100 Treatment villages 90 so 70 V 60 - 50 40 PROGRElSA INTERVENTION 45.6% I I I 1 I P2 I P4 P5 P6 S1 s2 I s4 Enteringgrade Source: De Janvry and Sadoulet(2003a and 2003b). The policy implications of this analysis depend on the importance attachedto the transfer andeducational incentive goals. This report places significant weight onthe transfer goal, in light of the extreme inequalities in Mexico and the fact that asset and economic opportunity-based strategies will take a long time to reduce extreme poverty. (See also De Ferranti et a1 (2004) for a Latin America-wide perspective on this). In this context, transfers linked to school and clinic attendance that are effectively reaching the extreme 139 poor can be desirable even when infra-marginal, since they are likely to have neutral to positive incentive effectsI3and are politically more acceptablethan unconditioned social assistance. At the same time it is desirable to shape part of the transfer to maximize behavioral effects on human development. This has been the direction that O~ortunid~es has been in practice moving. Since 2001, O~ortunidudesextended education benefits to individuals 14 to 20 years old inpoor householdsenrolled inupper secondary education. After only one year of implementation the program provided more than 450,000 scholarships for upper secondary, yielding a net impact of 38% increase in enrollment rates in rural areas and 6% in urban areas (Rubalcava and Teruel, 2003). A further extension is Jbvenes con Oportunid~esthat consists of a savings fund for lower secondary graduates wishing to continue their upper secondary education andfinishing it before reaching 22 years old. Students can cash out their accumulated points and use them as collateral for micro credit, as down payment to buy a house or for home improvement, to buy health insurance or finance higher education. This design is promisingbut the component has not been evaluated yet. b) Oportunidudes role in risk management. While not conceived as a risk-management instrument, Oportunid~esalready has safety net properties. For it provides a source of income for poor families that i s uncorrelated with either idiosyncratic risks (such as illness, death of a breadwinner or individualjob loss) or covariate risks (harvest failure, economic recession, natural disaster). There would also in principle be scope for temporarily increasing the transfer, at low administrative cost, in the event and in areas, where covariate shocks are associatedwith chronic poverty. This would, however, be an imperfect targeting instrument since the objective of risk-management is to reach those with who have adverse shocks and weak means of managing these. This could involve many households not covered by O~ortunid~es. For this there i s a needfor transfers that a contingent on households suffering adverse shocks: Seguro Popular (discussed below) i s precisely intended to strengthen such insurance for poor households subject to adverse health-relatedshocks and i s thus inprinciplecomplementary to Oportunidudes. c) Reaching un-covered groups over the life cycle. While O~ortunid~es has extensive and growing reach amongst the extreme poor with children there are many other uncovered groups that a comprehensive social protection strategy would be concerned with, It is again usefulto distinguishbetween providing transfers to the chronic extreme poor and providing risk-management instruments for those uncovered by the formal social security system(discussedbelow) though there will be overlaps between these, Amongst the chronic extreme poor it can be useful to divide the population into three categories, First, there are those adults who have low earning power because they are disabled or elderly. Most countries eventually move to provision of basic transfer for such groups including a state-funded pension. The feasibility of this depends crucially on ~ l3Even for the in-school population the transfers will provide incentives to increase attendance. There is the theoretical possibility of adverse effects on the work incentives for adults in the families, but this has not beendocumented. 140 the overall revenue position, and on the level at which the transfer is set. In Brazil, as a product of negotiations around the 1988 Constitution, a farmer's pension was introduced. This has proven to be effective inreaching the aged rural poor: for example inthe poor rural state of Cearti, which had extreme poverty of close to 50% in 1996, this scheme effectively eZi~inatedextreme poverty amongst the rural aged. However, this was associatedwith muchhigher transfers to the old than children amongst the extreme poor. Itis also relevant that Brazil's tax effort is of the order of double Mexico's. The specifics of design of such a pension for the aged and disabled would have to be carefully assessed for Mexico. Second, there are those families eligible for Oportunidades who do not participate. As noted above, in the poor rural Progresu areas in the late 1990's, about a quarter of poor households did not participate despite the offer. This requires further investigation but this may be a group of the extreme poor who may require more tailored treatment in relation to needs. Paes de Barros (2003) has proposed a form of social worker systemthat reaches out to the very poor and helps them make use of available social assistance and other services. Chile's Puente system works broadly along these lines. This could be explored in the context of small experiments, in rural and urban areas, to see if such a modeli s effective and affordable for Mexico. Third, there are the working poor without children, especially those in rural areas and with low education (see Chapter 3). These constitute some 7% of the extreme poor. Incomes of this group are essentially a function of the expansion of economic opportunities. Deepening of human capital amongst adults with low education is relatively difficult and costly. For this group, there are close links between chronic low labor incomes and the risks of fluctuations. The primary policy approach for this group should relate to affecting the rate, pattern and stability of growth, and opportunities for physical asset accumulation of poor (notably in agriculture, rural and urban non-farm self-employed activities and housing). These are discussed in Chapter 5. However, there can be a complementary role for work-related transfers in which publicly financed temporary work is offered at a low wage -a factor that i s critical for targeting effectiveness. This can complement labor income from other work where un- or under- employment i s high. It can serve as a mean of dealing with idiosyncratic risk in normal times and can be expanded to deal with covariate risks when harvests fail, natural disasters hit or there i s an aggregate economic shock. A well-documented example of an urban schemeinLatin America is Argentina's Trabajar (Jalan and Ravallion, 1999). The most famous case of a rural scheme i s the Employment Guarantee Scheme of Maharashtra (Ravallion, 1993). The potential relevance of works schemes at least in poor areas of rural Mexico is illustrated by the fact that in the ENCASEH survey of 1997, 52% of adults said they did not work (De Janvry and Sadoulet, 2003). As noted above, Mexico has a public works scheme inthe form of PET that is now managed by the states. Yet its reach seems to be small: in the same ENCASEH survey of marginalruralareas, only 1.5% of households had a member participating in PET. While it has not had a rigorous evaluation, the program has a reputation of being vulnerable to use for patronage purposes. Indeed 141 Argentina's Trabajar scheme also had such a reputation (as well as having excellent pro- poor targeting properties). This is an area inwhich design andinstitutionalcontext is key. While targeting effectiveness can be facilitated by a low wage other design features can increase community benefits and reduce risks of capture and use for patronage. These include linking workfare to local investment activities with community input and accountability mechanisms, and using transparent points mechanisms for ranking projects. Such measures will be weaker where local governments and communities are subject to extensive elite capture -an issue we discuss below in the context of decentralization. However, in light of the potential role of workfare within a social protection system it is recommendedthat there be further experimentation and evaluation of such schemes under alternative designs andindifferent contexts across Mexico. d)A continued rolefor food programs? While major food subsidies were phasedout over the 1980's and 1990's, Mexico still has some food-based programs, of which LICONSA and DICONSA are the oldest and most important. LICONSA started in 1944 and DICONSA was created in 1972. The LICONSA program (Leche Industrializada C O ~ A ~ U Pprovides milk powder and liquid milk at subsidized prices to low-income O ) families. A recent impact evaluation by ITESM (2003) found that poor households increased their weekly consumption of fortified milk by 3 liters per week. This in turn improved their nutritional status. DICONSA (Programa de Abasto Rural) is a public network of small stores providing basic goods to marginalized rural communities. An evaluation of the program(Grupo de Economistas y Asociados, 2003) shows that savings coming frombuying at DICONSA stores are equivalent to 10% of the total expenditures on basic foods by otherwise equivalent non-participant families. The long-term role of these food-based programs within an extended social protection system needs to be assessed. They may play a valuable complementary role to core instruments or there couldbe a case for phasing them out, for example, if Oportunidades were to become genuinely comprehensive. e) Managing exit. Oportunidades has so far been more concerned with entry than exit. I s there a need to pay more attention to graduation? There are two graduation principles. The first is demographic and behavioral and intrinsic to the conditionality of Oportunidades (and indeed to potential extensions sketched in this section). When children grow older or leave school, eligibility automatically stops. The second concerns economic improvement of households. For this it makes sense to institutionalize a periodic re-certification. This could involve the current once-off three re-certification after three years into regular such re-certifications. It makes sense to have a significant "cushion" inpart to ensure that households are not made to exit from the program when economic gains are temporary. There is also a case for ongoing re-evaluation of the points system. For example, a rural area may be experiencing a secular reduction in the proportion of families with dirt floors without significant increases in disposable incomes. Such a periodic re-certification i s a feature of universal means-testing schemes such as Chile's Fichu CAS. The latter scheme re-assesses household eligibility every three years -with the important difference from Oportunidades that this defines eligibility for a range of poverty-targeted benefits thus spreading the administrative costs 142 over many programs. Applying this in rural areas of Mexico would involve the full implementation of the Cbdula onica de Registro Poblac~o~al (CURP). For now, Op~r~unidades approach i s to resurvey households after 3 years to determine whether they remainpoor. Beginningin2003, householdswho are classified as poor will continue to receive the same packageof benefits. Other households will be transferred to receive a subset of actions (rural households after 6 years, urban households after 4 years) called EDA (Esquema Diferenciado de Apoyos), which includes education grants at the secondary and household level, health benefits and nutritional supplements. These households will receive EDA for three years, upon which point, according to the new rules they will no longer be eligible for Oportunid~es. f )Accountabili~.Inlight of the central and risingroleof Oportunid~eswithin Mexico's social protection andpoverty reduction strategy, assuring accountability i s of importance. O~or~u~idades`a recourse mechanism and reportedly has very low complaints rates. has It would be valuable to assess how well this is working and also to explore whether there may be unintended inter-institutional interaction effects, Researchon basic education and health services inMexico finds both very broad coverage and very uneven quality which raises the issue of how accountability mechanisms can work better to raise the lower end of quality. There may be aspects of Oportunidades operations that are relevant and interactive. Specifically, the following hypothesis could be tested inthe field, basedso far on limited ethnographic evidence: there may be Cross-ins~~tutio~aZincentives or disince~tivesfor Oportu~idadesprogram beneficiaries to push for accountability intheir education and health services. Positive incentives could occur if there were positive links between the Oportunidades program and teacher or health-worker feedback and performance. But negative incentives would occur if complaints put their status in Oportunidades at risk. For example, if parents complain about teacher or doctor absences, ineffective teaching or health services, then those teachers or doctors could seek to have families dropped from the list of beneficiaries. It is not clear if there are effective recourse once one is dropped from the list. If such cross-institutional disincentives for accountability do exist then that suggests the need to rethink and propose new and more effective accountability mechanisms. Expandinghealth coverage: managinghealth risks Major health problems can be sources of monetary poverty because of the cost of illness associated with both treatment costs and lost work. As already noted, close to half of the extreme poor in rural areas had "catastrophic" l4 health expenses. Within a social protection system there i s clearly a central role for mechanisms for providing insurance against such large or catastrophic costs. At the request of the executive branch of the federal government and after much debate, the mexican congress passed into law a significant reform of the system for healthcare provision and financing to the uninsured in April 2003, The implementation of Seguro Popular (Popular Health Insurance) is a center pillar of the reform. As stated repeatedly l4 definedbyWHO:morethan50%ofhouseholdincomenetofbasicfoodconsumption. As 143 by president Fox, the health reform in general, and the Seguro Popular in particular, is consideredby the current administration one of the most important political achievements inthe social sectors. Although SeguroPopular is the flagship program of the reform it i s only one of the three key changes introduced by the new legislation. Key challengesin theMexican Health Sector There are at least three key challenges for the mexican health sector today: equity, financial protection and quality of services. Interms of public policy inthe healthsector, the equity challenge is due to the existence of major inequalities in the allocation of public subsidies to and within the health care sector. The challenge of financial protection i s demonstrated by the fact that health shocks in Mexico significantly contribute to poverty and may lead to poverty traps. The quality challenge relates to the effectiveness of providers in the delivery of healthcare interventions in the country. Cutting across these i s a fourth challenge of an instrumental nature. This relates to the need to ensure a well functioning "compact" between federal and state levels and also between the state and the healthcare providers intheir jurisdiction (the "compact" is the set of relationships between policymakers and service providers that determine the incentive framework to achieve the equity, financial protection and quality goals in the sector, in this case also betweenthe federal level policy makers andthe state level-see Chapter 1 for discussion of a conceptual framework for service delivery). A weak "compact" between federal and state levels implies the potential for fragmentation of public health and healthcarepolicy in a context in which most healthcare delivery responsibilities are decentralized to the state level in Mexico. A weak compact between the state and the providers in their jurisdiction could lead to failure to set the appropriate incentive framework for such providers in the delivery of publicly subsidized healthcare services and, as a result, failure to achieve quality andfinancial protection. Health shocks are signi~cantzycontributing topoverty and thepoverty trap. Evidence (as good as we can get it today in the sector) suggests that both foreseeable (e.g. normal delivery) and unpredictable (e.g. diseaseof a newly bornchild) health events (shocks) are contributing significantly to poverty and the poverty trap. The poor expend a high proportion of their income as out-of-pocket payments for health with sometimes major adverse effects. More than 5 million citizen experience "catastrophic" health expenditures and at least 2 million of them fall into poverty as a result. The poor show much higher levels of such out-of-pocket payments than the rich in relation to their income suggesting that available risk-pooling mechanisms do not reach them andor do not protect them effectively. Although the current estimations of catastrophic events provide useful information as a preliminary approach to the problem, the methodological approachbasedon spending is only a poor proxy for a genuine measure of the impacts of health shocks on well-being -that should take into account permanent impacts on earning capacity, the ability to borrow to cover spending and a measure of impacts on human capital formation at the household level, in addition to actual spending as a share of income. Shifting from the current methodological a approach to one that captures the 144 long-term effects of health shocks on the household capacity to create humancapital and earn income i s an important priority for diagnosis, monitoringandevaluation. There are significant inequalities in the allocation of public sub~idiesto and within the health sector, First, there is a large per-capita public subsidy to formal workers via fiscal transfers to IMSS and ISSSTE, even larger than the per-capita subsidy to health services at state level that i s supposedly intended for the poorest segments of the population. These are regressive transfers (Figure 4.6). Second, there are large differences inthe per- capita transfer of federal public subsidies among states. The historical federal allocation mechanisms in the health system have favored the richest and most influential states in the federation. Third, there is an important capture of public health subsidies at state levels by the non-poor, largely as a consequenceof the existing incentive framework for public providers. There is a fragile compact between the SSA and the states and between the state and public providers in their jur~sdic~ions. Health services for the poor are decentralized to the state level inMexico. Although the SSA historicallyhad had a significant operational role, today, after significant decentralization to the state level of health service delivery for the uninsured, the SSA has very few instruments to leverage health policy and service delivery at state level. There are no doubts about the importance of decentralization to the state level in a country of the size and complexity of Mexico. However, the absence of a well functioning compact between the federal level and the state level can greatly contribute to fragmentation of health policy, increased inequality, and may reduce the accountability of states regarding the effective use of resources of the federation on services for the poor. The compact between the federal and the state level needs to be strengthenedfor the SSA to fulfill its functions of steward and regulator of health services. Financial transfers in the healthsector from the federal to the state level have been done on an historical basis, unrelated to the level of output or outcomes at the state level and focused on financing production factors rather than providing healthcare services to the uninsured. Reducing inequalities in public health financing requires a radical change in the way public subsidies are allocated in the health sector today. It requires an absolute (or at least a relative) reduction of public subsidies to IMSS and ISSSTE, an increaseinpublic funding to replace highlevels of out-of-pocket payments for the very poor, beginning at least with the poorest states, and a significant improvement of the federal-state compact to increase effectiveness andtargeting of public subsidieson the poor. The compact between states and state public providers face similar problems. States finance their own providers on a historical basis, financingproduction factors andwith no linkage between output or outcomes and the level of financing. Current provider financing mechanisms at state level set a perverse microeconomic incentive framework, that tends to favor delivery to the non-poor and low productivity. States need to develop strategic purchasinginthe health sector, linkingprovider financingto delivery of services to the poor (Strategic Purchasing i s defined as the process by which the state -the purchaser in this case- puts in place the set of contracting, payment and auditing 145 mechanismsthat set the right incentive to providers inthejurisdiction, particularly public providers). Themexican health sector reform At the request of the executive branch of the federal government the mexican congress passedinto law a significant reform of the system for healthcareprovision and financing to the uninsured inApril 2003. The new law focuses mainly on the financing of the public health sector. There are two strategic objectives of the reform: first, to improve access to basic health services and financial protection (reduction of out-of-pocket expenditures in health) for the extreme andmoderate poor; second, to change both the efficiency andequity of public financing (subsidies) and service provision within the public healthcaresector inthe country. The key features of the new legal framework include the following: (a) The separation of the funding andthe managementof public goods (andsome private goods with high positive externalities such as vaccinations) from medical health services. Public goods will be financed from general taxation under a new Fundfor Community Health Services. (b) The creation of Seguro Popular as a largely publicly financed (federal and state governments) basic health services insurance scheme, which provides funding for bothEssential Health Services (a packageof basic curative healthcare andpreventive services) and a Fund for Protection against Catastrophic Events. The latter is a centralized single risk-pool for high-complexity and high-cost health interventions, which will reduce the financial risk of all states and their beneficiaries regarding this intervention (known in the international health economics literature as truncation of the Pyramid of Risk). Note that while the Essential Health Services package is not a source of catastrophic risk management from the perspective of the complexity or cost of the intervention, it is intended as a risk-management instrument for the poor, given that relatively "simple" curative needs can have "catastrophic" financial consequencesfor the poor. (c) A significant increase in public financing for the public health system, expected to be fully implemented by 2010. However, to understand the full impact of the current reformon services to the poor, one needs to understand the impact of all the law's components in the health system in general, including SSA, states andindirectly potentially on IMSS andISSTE. 146 The role of Seguro Popular: protecting the poor and strengthening the federal-state compact Seguro Popular plays a central role in the newly approved reform. Although Seguro Popular has been designed with some features of a traditional health insurance arrangement (e.g. participants would pay premiums), the progressivity of the premium charge (the two lowest deciles do not pay any premium) and the heavy public subsidization for these deciles (by the state and the federal level) makes it basically a public scheme for the delivery of basic health services to the poor. Rather than a contributive health insurance, Seguro Popular has to be considered as a targeted transfer mechanism built around a defined package of health services with risk pooling mediated by the government. As such, traditional considerations of adverse selection and other insurance failures are much less relevant. Indeed concerns over adverse selection, that is the process whereby higher risk individuals or households enroll inan insurance scheme, would actually work in reverse. As a subsidized public scheme for provision of basic health services it i s the objective of the program to reach in particular those with the highest risk. The premium contribution constitutes only a marginal part of the overall financing and its role is more to strengthen the co-responsibility and citizenship dimensions of the scheme. Seguro Popular basically aims at addressing two of the key challenges of the mexican health sector: to improve the compact both between the federal and the state level, and between the state level and the state providers; and to provide a risk-management tool (with society level risk-pooling) for uncertain health shocks. As noted above, public goods and activities with high externalities (such as vaccinations and TB treatment, as well as interventions to overcome agency problems inhealth such as well child care) are to be addressed and funded by a separate fund, the Fundfor Community Health Services of SSA in synergic interaction with the health component of ~ p o r t u ~ ~ ~ aIn s . d eessence, ~ ~ o r ~ u n ~ d aprovides incentives to increase household demand from the poor for d e s effective basic health care services while Seguro Popular and the Fund finances the provision of services and potentially sets the right incentive framework for providers to effectively respond to such increaseindemand. Seguro Popular will allocate and transfer resources from the federal to the state level on the basis of capitation - o r the number of enrolled eligible families from the pool of households uncoveredby formal health insurance. Part of the capitation (25%) i s adjusted on the basis of relative health needs of the state based on mortality rates. There are also two additional adjustment factors, one aiming to stimulate state contributions, the other aiming to reward good state health system performance. However, these two adjustment factors account for a small fraction of the total allocation of resourcesby the SP. All three specific allocation factors will be adjusted on an annualbasis according to the assessment of program implementation. In this respect the capitation allocation of resources i s significantly biased (in comparison with the historical allocation) in favor of poor states, especially relative to the distortions on federal financing determined by historical allocations. 147 The impacts on the compact andrisk-management objectives are reviewed inturn. (a) The implementation of the Seguro Popular has a clear potential to improve the compact for both the federal-state and the state-provider relationship. It would do so as a result of: i)per-family transfers to participating states for eligible populations rather than the historical patterns on the supply-side; ii)transfers are intendedfor the provisionof the package of essential services rather than for an unspecified product; iii)given the linkage of financing to state health system performance, as well as the intended population and defined product, SSA would be able, for the first time, to collect and eventually use for management and financing purposes on a routinely basis, information on actual coverage for lowest quintiles, utilization andprovider productivity. Similarly, the same features of Seguro Popular provide the basis for the states to strengthen their compact with health providers. Monitoring enrollment of eligible population, ensuring that providers are actually delivering the package to the eligible enrollees andthat funds andresources are usedfor SeguroPopular andno other purposes will force states to set up monitoring and productioninformationsystems, quasi-contracts and other purchasing tools. At the core of the potential success of SeguroPopular will be the actual implementation of strategic purchasing at state level. The political and technical challenges for such implementation are significant. States will require both a close technical support and a close oversight by SSA to ensure such implementation. A successfulimplementation of the SeguroPopular would mostlikely trigger andrequire a transition towards a change of mission of the SSA and the state health secretaries: from one of perceiving themselves as a form of government holding company for public providers towards one of steward of the health system. While the traditional, holding company, missionfocuses them on the financial viability of their own providers, the new mission i s shapedaround ensuring maximumimpact of public subsidies on the poor. A lesson we have learned inthe last 15 years inLAC and even insome OECD countries i s that until the perceived policy function of ministries of health does indeed undergo the transition from managing and financing of their own public providers to maximizing impact on the population (and particularly on the poor), it is unlikely that the incentives and the constituency will exist for shifting to demand-side financing for services to the poor. At the core of such a transition lie two necessary changes. First, there i s a need to counterbalance the provider overseeing function (the traditional soul of a ministry of health) within the ministry with a separate financinglpurchasing function within or outside it. Seguro Popular will require this and will hopefully trigger its creation both at federal and state level. This is the implementation of a purchaser-provider split, functionally or even also organizationally. Only with this separation will there be the necessary tensions to re-focus the policy function of ministries of health to a new framework. This i s a complex but key transition: the two Latin American cases that have progressed most - C h i l e and Colombia- are just beginning to bear fruit. Second, the newly created function needs to increasingly shift financing from historical budgets to production-based financing. This can form the basis for a further transition to demand- side subsidies for services or insurance. Transitioning from historical supply side 148 financing to production-based and competition for clients at least at the margin (through arrangements inwhich money follows the patient) requires public providers to be able to dynamically adjust their production cost structures within the revenue constraints set by the purchaser. However, at the core of this process is the capacity to ensure that providers, particularly public providers, would be able to adjust their cost structures. The implementation of the Seguro Popular would most likely also trigger the policy discussion on restructuring public providers. In essence, the specific features of management and financing in the Seguro Popular would allow SSA to establish a strategic purchasing relationship with the participating states. The clear potential for improving the compact will be realized only when SSA fully implements and uses this strategic purchasing capacity. Establishing it is one of the most significant political and technical challenges the reformfaces to realize its potential. It has proven to be a significant challenge in most public health sector reforms in LAC and inOECD countries. However, no other strategy has proven to be effective inactually implementingthe required transition. (b) Seguro Popular has the potential to provide an improvedrisk-management tool (via increasing society level risk pooling for the poor) for uncertain health shocks. It aims at increasing risk pooling both through the Package of Essential Health Services and through the Fundfor ProtectionAgainst Catastrophic Events. The main difference between the current and the Seguro Popular scenario, besides its potential role in implementing strategic purchasing, is the reduction of out-of-pocket payments in the lowest quintiles. The very poor show high levels of out-of-pocket payments inMexico inrelationto their income that i s evidence of a non-functioning risk- pooling mechanism for the poor. Evidence on out-of-pocket payments and catastrophic events for the poor show that to actually reduce such spending Seguro Popular needs to: a) increase availability of free essential drugs for the very poor; andb) reduce direct out- of-pocket spending, at least for essential health services. Thus the focus of Seguro Popular on 131essentialdrugs and90 essential health interventions starting 2004. While, in theory, there are only modest structural differences between the current scenario and the Seguro Popular scenario, effective implementation under the new design could be of major significance: through reducing inequalities in the allocation of public financing (subsidies) for the health sector among states, IMSS and ISSSTE; increasing availability of essential drugs at public provider level; removing the "voluntary" and mandatory contribution the poor need to pay at public providers today; and reducing the crowding-out effect of the non-poor on utilization of health services. However, again, full realization of the benefits of the Seguru Popular reform will depend on effective implementation of strategic purchasing and on an absolute or relative shift of public subsidies away from the non-poor and towards the very poor. 149 Mexico can achieve the shift of public subsidies towards the poor in two ways, that are not mutuallyexclusive: reallocate current regressivesubsidiesfrom IMSS andISSSTEto the public health subsidized system andlor increase fiscal expenditures in health. Both strategies should be subject to implementation of strategic purchasing to ensure targeting and effectiveness of the additional financing on the very poor. Through its newly approved law, Mexico has already decided to significantly increase fiscal expenditures for the poor through the public health sector. Its possible that the discussions about the actual feasibility and sources of financing for the increase in fiscal contribution to the public health sector will triggedfacilitate a policy discussion about the reallocation of public subsidies currently allocated to the non-poor. Key challengesin the ~mplementationof SeguroPopular By October 2002, Seguro Popular had 584,722 beneficiaries (around 158,000 families), The states of Aguascalientes, Colima, Jalisco, Tabasco and Campechewere the first ones to participate in the new scheme. Soon after, Seguro Popular expanded its operations to 79 municipalities in 19 states. Services to Seguro Popular affiliates are offered in 46 hospitals and 277 health centers of the Health Service Systems of the States (SESA) which are regulated by the SSA (Operational Rules of Seguro Popular, SSA, 2003). By end-2003 625,000 families hadbeen affiliated in24 out of 32 states. These are important achievements. However, as discussed above, the key challenge for the success of the Seguro Popular is the actual implementation of an effective strengthening of the compact between the federal level and the state level and, particularly, between the state level and the health service providers. In the absence of such strengthening it i s unlikely that SeguroPopular beneficiaries would benefit from the full potential of the reformandincreasedpublic expending might result only inimproved conditions andrevenue for providers. A second challenge is the fiscal one. The new legal framework calls for a significant increase in fiscal transfers to the public health system (SSA and states). As discussed above, this can be achieved either by increased taxes or reallocations -whether from other sectors or from the currently regressive public subsidies within the health sector, namely ISSSTE and IMSS subsidies. This would configure a complex political scenario which takes us to the thirdchallenge. The third challenge is political. Strengthening the compact would necessarily affect the current balance of accountability inthe health sector for SSA, states and public providers. The experienceinLatin America shows that this i s apolitically painfuland longprocess. The fourth and key challenge i s the evaluation of the results of the reform. The core promise of the reform i s not the implementation of the Seguro Popular transference mechanisms, the Funds (Catastrophic and Essential Community Services) or even strategic purchasing. The core promise i s that the reform in general and the Seguro Popular in particular will actually improve utilization of services, financial protection and responsiveness (buen trato) for the extreme and moderate poor, resulting in better 150 health and less deprivation from health shocks in terms of incomes and long-term health status. Evaluating the impact of this reform is challenging methodologically and conceptually. The SSA has proven its capacity and focus on including evaluation and evidence as a key element for defining policy. The evaluation of Seguro Popular will be essentialto assess if and when this very promisingreformhas fulfilled its promise. Since Seguro Popular has only just been introduced there i s not yet information on its impact. Indeed actual impacts will depend greatly on the behavior of households, on service providers, on interactions between the SSA and the states. While it has the potential to deliver major benefits, the extent of these remain uncertain. For example, there is uncertainty over the extent to which Seguro Popular will lead to households actually increasing their use of health services shifting from private to public providers (saving out of pocket expenses) or choosing not to enroll in formal social security. There are also unknowns over the impact on the quality of health services. This uncertainty i s comparable to the situation when Oport~~~dadesintroduced (then as PROGRESA) was in the late 1990's. Getting information on these issues will depend on structured monitoring and evaluation bothon householdbehavior and service provider performance. While current plans do indeed put considerable emphasis on monitoring and evaluation there is a need for particular attention to the design of impact evaluations, especially those that can track the behavior of households and service providers over time (as was undertaken as an intrinsic part of Progresa's design with in particular the use of longitudinalhousehold surveys). Accountability and service quality As shown in Chapters 2 and 3, service quality in Mexico remains a major issue. The quality of educational services has not increased at the same pace as coverage and it remains low, Dropout and repetition rates are still highfor secondary education resulting inpoor terminal efficiency. Highdropout rates reflect partly financial problems and the high opportunity costs of enrolling children in schools for low-income families; but studies also suggeststhat they might also result from poor quality of educational services (L6pez-Acevedo, 2001). l5 National standardized students test scores show that approximately half the students have not reached the objectives of the educational programs. However, there have been some improvements since 1998 (Figure 3.1). The quality of education i s driven by the interaction between socioeconomic characteristics and service provision, particularly teachers' performance. The general thinking, discussed in Chapter 1 (Figure 1.2), emphasizes the centrality of patterns of accountability -the "compact" that defines the relationship between policymakers, service providers and front-line workers and the client power relationships between citizens and service providers, whether via choice of direct participation- these are often complementary. Mexico's educational system appears to have both a weak compact relationship, owing to the nature of bargaining and incentives for teachers, and weak l5According tothe 2000populationcensus, out ofthe7-29year oldpopulationwho left school, 27.5 per cent did it becausethey did not like or want to study, and 38.5 per cent for economic or familiar reasons. The second reason is more important for the older group (aged 20 to 29). 151 parents' participation. There have been a number of worthy interventions and some have had positive results, but these do not get to the heart of the determinants of provider and especially sanction poor teachers performance (Table 4.11). Thus the need to complement these with deeper reforms to increase accountability via both changes in the compact and client power routes. It is highly likely that analogousconsiderationsapply to other basic sectors e.g. health, roads, electricity, and water. Table 4.11 Programs for aualitv imwovements ineducation Q 1 " L PROGRAM SIZE KEYCHARACTERISTICS EVALUATIONRJBULTS Carrera 738,186 teachers a) Merit payment system: staff is L6pez-Acevedo(2001a) Magisterial in2002. voluntarily evaluated based on found that teachers 87 percentof experience, professional skill, enrollment intheprogram basic education educational level, and completion increasedstudents teachers are of accreditedcourses. learning achievement. 'enrolled in the b) Training of teachers. However, the higher the program. levelreachedby a teacher, the lower the studentstest scores. Incentives for Points system that is used to No formal evaluation. Higher Education calculated cash bonuses given in However, the points Teachers reward for good performance. system, betterdesigned Quality of teacher performance than CarreraMagisterial, accounts for the majority of points. providesincentivesfor Bonuses go from 1-14 minimum higher teachingquality. wages. Escuelasde 2.5 million Government grants to urban Qualitative evaluations Calidad(PEC) students and schools with students from low- have beencarried out. 9,827 schools in income households committed to However, there are no 2003. increasingquality. conclusivequantitative Schools participating have to results yet. create a project to improve school infrastructure andlor quality of the education. Compensatory 10million Improve equity and quality of World Bank 1999 and Programs students basic educationinpoor areas. 2003 showspositive Reduce dropout and absenteeism impact of the program on rates. students' learning achievementand reduction indropout rates. Apoyo a Gestibn Funds available to parents Noformal evaluation. Escolar (AGE) associations in preschool and primary schools under extreme poverty to buy school material and to finance projects chosen by each school. 5,000-7,000 pesosper year. Requires parents involved in school and teachers constant evaluation. Source: Government's website 152 While the government strategy to address quality of education seems to go in the right direction, some other actions should be considered or re-considered to complement this strategy. A law has been passed targeting public education spending at 8% of GDP by 2006 compared with close to 6.5% in recent years, a ratio that is broadly inline with the National Education Program (2000-06). However, the availability of these additional resources will clearly depend on progress on the fiscal reform and it is hard to see how the volume of resources could be raised so rapidly and be allocated efficiently. Whether these additional resources for education are made available or not, given the importance of the sector in public spending it is necessary that the GOM clearly define spending priorities and makes educational programs more cost-effective. Most resources allocated to education go to staff compensation (Mexico showing the highest share in OECD countries, at above 90%, along with Turkey and Portugal) while resourcesfor investment are scarce. It is therefore important to review and assess the impact of educational expenditure to identify the most cost-effective way to achieve quality in education. Improving quality of education requires a better evaluation of institutions, programs, teachers and a close monitoring of the implementation of reforms. Further efforts should be made in some specific areas. Inparticular, it i s important to accelerate work of the Instituto ~acionalpara la Evaluacidn de la Educacidn (INEE), the publication of evaluation results and to foster discussions around these results. To make schools accountable there is a need to go beyond particular programs such as PEC and, in particular, sanction poorly performingschools. This section covered social protection design andraisedthe issueof accountability for all services. It was concluded that moving rapidly towards an integrated and comprehensive system would be infeasible. A more achievable objective is to move towards a dualistic and comprehensive system strengthening ~ p o r t u n i d ~ e s andfindingoptions to reachthe moderate poor but without distorting the labor market. In doing so, achieving quality service delivery i s an essential area not only in health but also in the provision of other basic services. C. Decentralizationand geographic dimensionsof well-being Political and administrative decentralization has been a major feature of Latin America over the past decade or so. InMexico, there have been significant shifts of spending and program management to states and municipalities (Box 2, executive summary). The implications of this for institutionalinteractions, programmanagement and distributional impacts are complex. As we saw in Chapter 3, differences in levels and dynamics of well-being across space i s an important aspect of Mexico's overall inequality. In this section we introduce two themes: the geographic distribution of spending; and the relationship betweendecentralization andthe effectiveness and equity of service delivery. Thed i s ~ ~ ~ u tofi otransfers in relation to state needs n Figure 4.13 shows the geographic distribution of various public-spending items in relation to state GDP per capita in 2002. The first graph shows the distribution of total public spending (state own revenues, participaciones, aportaciones, municipalities own 153 revenues, and sectoral spending); the second shows the distribution of FAEB (adjusted per state school age population); the third graph is of the distribution of the two major poverty-oriented programs FAIS plus O ~ ~ r t u n ~ ~and e ;u ~thesfourth of all agricultural support programs, including PET. The following patterns are notable: Total spending, education transfers (adjusted for school age population) and (especially), and agricultural spending is positively related to GDP per capita. This means that relatively well-off states have greater spending overall and in these categories. With respect to education transfers there is wide variation around the broadly positive relationship, Mexico City obtains the largest transfer of 16 thousand pesos per school-age populationwhile a poor state like Guanajuato receives only 5 thousand pesos. The southern states receive an average of 8 thousand pesos and Veracruz (the fourth poorest state) gets only around 7 thousandpesos. Both FAIS and Oportunidudes'sgeographical targetingi s very good across states, with poorer states received the larger transfers. But there is considerable variation around this pattern. For example, Chiapas gets an average of 1,200 pesos per capita, Veracruz, that is slightly richer, gets half of this, while the national averageis around 400 pesosper capita. The Agricultural Support Programs benefit the northern and wealthy states of Sonora, Sinaloa and Baja California. Sonora gets an average of 4,500 pesos per capita while Oaxaca only gets 500 pesos per capita. FAIS and O~or~unidudes progressive targeting i s in part driven by application of relativelyclear formulae basedon needs. Inprinciple, this couldbe extended to other programs. However, political constraints are also relevant: in the case of the education and health transfers the bulk of spending goes to wages of teachers and medical personnel, and there i s significant inertia on the their location. Similarly most agricultural spending goes to support largeagricultural producers. The patterns reviewed related to spending in relation to average state incomes. It is also of interest to explore how the distribution of federal transfers relates to performance in terms of social outcomes. Figure 4.14 illustrates the simple relationship between sectoral transfers andthree areas of outcomes -in education, health and water and sanitation. For education, there i s essentially no relationship between drop-out rates and education transfers across states. For health, there is a somewhat negative relationship between health transfers per uninsured person and child mortality -implying states with worse health outcomes get less interms of transfers on average. This may inpart be becausethe share of uninsured is generally higher in poorer states and federal transfers do not effectively adjust for this. Finally for water and sanitation there is a somewhat positive relationship, implying states with a higher proportion unserved by improved water and sewerage services get higher transfers. 154 m u ; L g' e z . f 0 N ii 3 Interpretation of these patterns needs to be undertakenwith caution. Lower state transfers for an area with poor indicators (e.g. high levels of mortality) could reflect inequitable distribution of resources to areas of particular need. But it is important to recognize the multiple determinants of social outcomes. These include levels of private incomes, especially amongst the poor, and the effectiveness of government programs. The latter are in turn influenced by political and institutional factors such as the extent of elite capture and clientelism and bureaucratic capacities. Poorer states in terms of economic wealth may also be poorer in tems of such institutional variables. There can also be equitable or inequitable patterns of spending within states, mirroring the great diversity in incidence of spending at the national level discussed in part A of this chapter. A full analysis would need to look at the impact of government efforts (including transfers) on ~ ~ insocial outcomes-controlling for the rangeof other influences. ~ ~ ~ ~ e ~ e A recent review of the poorest three statesXhiapas, Guerrero andOaxaca(World Bank 2003)- concluded that relatively weak social outcomes were a product of multiple influences, including lack of funding, inefficient use of funds and difficult initial conditions (from high levels of income poverty to highly dispersed populations and relatively underdeveloped infrastructural networks). Inthis section we have provided some illustrations of differences inpublic effort across states that suggests that there are issues both with respect to resource allocation and with respect to how resources are used. Differences in the apparent "efficiency" of public action appears to be not only a matter of different initial physical conditions but also of institutionalconditions andreformefforts. There i s potential for action at state level, even inpoorer states. For example, Veracruz, the fourth poorest state interms of income per capita has introduced efficiency-increasing institutional reforms to improve service coverage and quality inboth their educational and health sectors (Box 4.2). It has begun to witness improvements in indicators such as infant and maternal mortality since 1999, but more time, andmore in-depth analysis will be neededto arrive at a full interpretation of the potential for action within poorer states. 156 Figure 4.14 Geographicpatterns of outcomesinrelationto federal transfers 2000 Education(drop-out rates), health(child mortality) andwater andsanitation(uncovered population) 4.0 3.5 0CHIS 3.0 0GRO 2.5 . 6 W l C H O V B R 4 O Z~A C D G O .CAMP YUC *OAX 0COL 2.0 GTO ~ I I A B $ PUE IAL SLP 0NAY El 1.5 SON BE 0MEX 0AGS DF MOR 1.0 NL OHGO COAH 0TAMPS 0BCS 0.5 0QROO 0TLAX 0.0 0 2000 4000 6000 8000 10000 12000 14000 16000 Basic Education Transfers per School Age Population, Pesos (FAEB) 8 7 O T L A X w e 6 MEX s^ ~c ; s * - O&JF mc%PRO .-3 & *OAX e3 4 1 *&iY * Y L C *SON 0QROO AOd 0SLP o Z A p~ ~N ~L ~ TAMPS -CAMP COL I- 5 5 3 0MlCH GRO 1 N%OAH SI& 1 DGO 0 0 200 400 600 800 1000 1200 1400 1600 1800 Health Transfers per Unisured Person,Pesos (FASSA) 40 35 +HGO ORO 30 TAB VER CHIS OAX J25 - c E20 20:15 +O R 0 CK 10 5 *DF * D f 0 0 50 100 150 200 250 300 350 400 450 500 Social Investment Transfers Der Caoita. Pesos(FISE and FISM) ~. ~ % Without Access to Potable Water +Share of pop without access to sewage Source:Webb andGonzalez (2003) andINEGI2000. 157 Box4.2. Veracruz-the potentialfor effective action at the state level Veracruz is the fourth pooreststate inMexico and receivestransfersper personbelow the national average. Ithastaken innovative measuresto improve the performanceof its educationaland healthsector. Inthe area of education, two state-ledinstitutional changes are notable since decentralizationbeganinthe early nineties: the integration of parallel federal and state educational systems and changes in the role of teacher unions in supervision. Previously, the federal and state systems operated almost independently, each subsystem with a full administrative superstructureand little coordination betweenthem. Veracruz is one of the few states that was able to successfully adjust to educational decentralization and integrated these two systems since 1994. With respect to the role of teacher unions, the federal Secretaria de Educacidn Ptiblica (SEP) negotiatesteacher wages with the national union leaving the state powerless in wage-settingdecision making. Inaddition, states also lack responsibility to monitor educationaloutcomes, even teacher attendances, because school supervisors are also union representatives. In Veracruz, while union supervisorsremain inthe federally run system, the state was able to replaceunion supervisorsinthe state system with state appointmentsby offeringthe union supervisorsretirement packages. Inthe area of healththere have beenalso some promising institutional reforms. These have included better integration and coordination between agencies in the state health sector that is characteristically fragmented. Inaddition, the creation of a state-payroll tax has helpedto increasethe state's contribution to health expenditure which increases the incentives of the state to ensure its effective use. The institutional changes have helped the state achieve its goal of reaching universal first-level coverage for its open population over the last three years. For this achievement the state was consequently awarded a certificate of universal coverage by the World Health Organization (WHO). This was mainly achieved through the leadership of Servicios de Salud de Veracruz (SESVER) which has the main responsibility of providing basic healthpackagesto the openpopulation. SESVER was createdby the government of Veracruz in 1997 andhas since increasedcooperationbetweenthe other agencies inthe fragmentedhealthsystem, including those run by the federal government, such as IMSS- Solidaridad, in order to help cover those not reached by these agencies. The state payroll tax, in 2001 contributed 32% of total health spending in relation to roughly 10% of state spending by other poor states. More analysis is needed in order to assess the full effects of these reforms. Source: World Bank, 2002a. ~ e c e n ~ a l i ~ a tandn communi^ driven approaches i o Accelerating poverty reduction and social development will require effective action across many levels -federal, state and local government, and communities. We finally look at issues at the level of communities and their interactions with government. Community Driven Development (CDD) programs might be one powerful instrument to achieve poverty reduction (De Janvry and Sadoulet, 2003). InCDD programs, organized community groups are given access to resources and power over their use. They manage investment funds and make decisions on planning, selection and supervision of projects using these funds. These programs have expanded rapidly in the World Bank's loan portfolio, for example, reaching US$Zbillion in mid-2003 and accounting for 11%of the total portfolio. The European Union's LEADER program follows a CDD approach, whereby Local Action Groups are charged with the responsibility of preparing local development plans, selecting investment projects and transferring EU funding toward those. In Mexico's programs, several programs relate to the CDD category, particularly programs run through the Consejos ~~crorreg~onuzes (SEDESOL) and Consejos 158 ~unicipales(SAGARPA). The case for territorial approaches to income generation in ruralareasis developedinChapter5. This approach is part of a "new" philosophy in ruraldevelopment strategies coming in contrast to state-led "integrated" programs and seeking instead an "integral" approach to rural development (see De Janvry and Sadoulet, 2003). The main dimensions of the approach include features that have allbeenpursued inMexico: 0 Decentralization to the municipallevel. 0 Devolution of management of common property resourcesto communities. 0 Participation of civil society organizations andcollective action. 0 Use of a territorial as opposedto a sectoral approach. 0 Implementation of coordination mechanisms between macro, sectoral and rural policies (e.g., Comisidn Intersecretarial para el Desarrollo Rural Susten~able, Gabinetede Desarrollo Socialy Humano inMexico; RuralLens inCanada). 0 Reconstructing a set of rural institutions following descaling of the role of the state andintroducing new institutions to overcome market failures. As an element of this approach, the modality of CDD offers much promise. It is, however, not free of risks and difficulties that largely remain to be resolved. We discuss both in what follows, drawing from experiences beyond Mexico. Solving the challenges offeredby the approachwill determine whether its promise will turninto a reality. Advantages of the CDD approachincludethe following: Access to and use of local info~ationand local social capital. With much private information locally public in agrarian communities but not available to outsider agencies, decision-making on the allocation of public goods and the choice of projects is better done at the local level. This allows projects and services to be adjusted to respond to the heterogeneity of poverty, to monitor project implementation through local information, and to use local social capital to enforce contracts. Empirical studies of decentralized targeting schemes by Galasso and Ravallion (2001) in Bangladesh; Alderman (2002) in Albania, and Faguet (2001) in Bolivia have all found that decentralization helped better target the poor or better identify localneeds. Real par~icip~tion. Community bargaining over resourcescan lead to a first best outcome for program definition and implementation if three idealized conditions prevail (see Binswanger andAiyar, 2003): 0 Equalinformation and foresight across community members. 0 All community members have equalbargainingpower. 0 There is a single overall budget constraint. While these conditions are typically not met, they set standards for what to achieve in order to improve efficiency and equity inCDD outcomes. 159 Contribution to empowerment. Giving to communities control over resources is an instrument for empowerment. Implementation of the CDD process and participatory evaluation are learning mechanismsabout collective action andlocal governance that can serve as catalyst for further local initiatives. Hence, through the processit puts into place, CDDs can be expected to strengthen participative culture, enhance local capacities in negotiation and collective action, increase self-esteem of local populations, improve access to information and create a sharedvision for local society on its possible future. But there are also risks with the CDD approach that need to be taken into account to secure success wit the approach. These include the following: Risks of capture and clientelism. This is the most frequently mentioned reservation about a CDD approach and about decentralization in general (see for example Bardhan and Mookerjee, 2001). The benefits of CDD's can be captured by local elites. They can also be used as populist and clientelistic tools by local politicians. Some localities may well have less corrupt and clientelistic administrations than state and national governments, but the opposite may also be true. Experience with LEADER Local Action Groups (LAG) shows that objectives differ according to the relative importance in the LAG'Sof civil groups versus democratically elected officials (Sumpsi, 2002). Municipal representatives tend to favor infrastructure and equipment projects while private sector representatives favor income generation projects. Hence, how decentralization affects resource allocation remains an empirical question. Certainly, heterogeneity of outcomes across localities i s to be expected and needsto be explained interms of forces inthe local politicaleconomies. Parallel ~ ~ i n i s t r a t i vstructure and lack of democratic representation. CDD's e sometimes establish a parallel structure to the elected local government and to deconcentratedbureaucratic institutions (e.g., Local Action Groups under the LEADER program in the EU; Consejos ~icrorregionaze~ Mexico). This may make project in implementation more efficient in the short runbut is not sustainableinthe long run. This approach, which has been extensively used for the management of social funds, has been criticized for this very reason (Tendler, 1998). The capacity of democratic local institutions and long-run accountability may also be weakened. It i s notable inparticular that local committees set up by CDD's tend to have weak by-laws and usually do not offer recourse mechanisms to local stakeholders. Hence, they tend to lack accountability mechanisms bothto their principal(EU, SEDESOL) andto their stakeholders. CDD's are not a substitute for institutional level development, both at the ministerial and sectoral level and at the municipal level. This suggests that CDD should be designed to complement democratic and sectoral administrative structures, not a substitute, and possibly folded into local democratic governance inthe longer run. CDD's should perform as part of a triangular power arrangement that includes effective local governance and strong local civil society organizations as the other two nodes. If these two other entities are inplace CDD's serve as a place for consultation, coordination andjoint priority setting for government and civil-society organizations (CSO's). "Real participation" can then happen. Without them, local committees acquire a life on their 160 own as they are not representative and can undermine the strength of both local governance and CSO's. Lack of local admin~strativecapacity. Local capacity to plan development, set priorities, managedecentralized funds andmonitor projects may be weaker than at the central level. Teams constituted to manage CDD activities may lack professionalism. If undertakings are relatively simple this may, however, not be a major problem as observed by Faguet (2001) in his study of municipal-level decentralization in Bolivia. What is important is that appropriate efforts be made to train a sufficiently large number of professionals in the management of local rural development to perform at the level of the consejos and municipalities. Local committees not specialized by types of projects. Local committees must decide equally on all projects submitted, whatever their content. Lack of specialization implies that there tends to be insufficient expertise in support of the more technical and sectoral undertakings. This suggests the importance of endowing local committees with budgets they can useto hire consultants to advise on specialized projects. Bottlenecks on dis~~rsement. Some CDD programs have suffered from low disbursement capacity with funds allocated to principals by donors or central governments but remaininglargelyidle. This is dueto severalcauses: 0 Low administrative capacity of committees. Complex administrative proceduresintroduced to check on moralhazards. 0 Many smallprojects with highfixed costs inselecting andmonitoring. 0 Lack of a sufficient number of intermediary facilitators such as NGO's or private sector consultants. 0 Complex mechanisms of participatoryappraisalwith uncertain methodologies. Lags intaking action due to low incentives for committee members. Scaling up problems. CDD's tend to work well at a small scale and with long time horizons. Scaling up the approach has proved difficult due to several factors (see Binswanger andAiyar, 2003): Lack of available budgets to secure the sustainability of projects and support their scaling up. Lack of a sufficient number of effective intermediary agencies (NGO's, private consultants). Public administration is not used to working following principles of subsidiarity andex-post rules of accountability. As a result coordination inthe co-production of services between central government, local government and the community is difficult andrestricts expansion of programs. Long learning process needed to reduce management costs before programs can expand inscale. Legalcontexts maynot allow direct disbursement to communities. There may be strong difference inobjectives between principal and communities, generating conflicts and limiting expansion. 161 Practices followed have not been adequately adapted to heterogeneous local contexts. There are diseconomiesof scale inproject management. Effectivedecentralization is not inplace, constraining community actions. An important instrument in support of scaling up is to organize horizontal cooperation across local committees. This requires formation of a network that can serve to exchange information and advice. Local committees should also have the legal capacity of forming associations when considering projects that have larger scale than the administrative unit (municipality) where they are located ( ~ u n c o ~ u n i d uin~Mexico, communities of e ~ communes inFrance). D. Conclusions andissuesfor future work The federal budget in Mexico is an important mean for redistributive action. It already has a large impact in terms of services and transfers for the poor. While a fully comprehensive analysis is not feasible the evidence suggests that the budget is already much more equal in its effects than private incomes. For the taxes and spending that can be analyzed it appears to raise incomes very roughly in line with incomes or spending, andspendinthe social arena, on average, broadlyequal amounts percitizen. But there is also large potential for further improvements. Mexico's overall tax effort is low for its income level and all spending was hard-hitby the 1994-95crisis. While there has been substantial subsequent growth insocial spending andtargetedpoverty reduction spending reached a peak in 2000-2002, this has in large part financed by reduced economic sector spending. Moreover within social spending there i s enormous variation in distributional impact ranging from highly regressive spending (ISSSTE programs, tertiary education, IMSS pension-related programs) to moderately regressive (electricity subsidies, probably much agriculture spending) to highly progressive (notably Oportunidudes).Moreover, despite substantial progress in service provision and human indicators there remain major gaps indeprivation. There i s a strong relationship between gaps and measures of income poverty with the extreme poor systematically suffering greater deprivation. There are some programs that have significant reach amongst the poor population: ~ p o r t u n i d ~and e ~Procumpo in rural areas, SSA services, as well, of course, as basic schooling, water and electricity. But for many the reach of services is inversely relatedto patterns of deprivation with greater coverageof the non-poor. The analysis of policy design suggests an important agenda for the future. This chapter focused on three major areas: (a) outstanding issuesinthe design of the social protection system, including gaps in coverage for the extreme poor, reaching the moderate poor (who largely miss out on both Oportun~~ude~IMSS) and the ongoing implementation and of policies to manage health risks for the extreme and moderate poor, especially under SeguroPopuZur; (b) the challenge of achieving quality in service provision and the need to complement specific measures (for example on teacher quality) with a much 162 broader approach to strengthening the accountability framework for service provision; (c) the major issues around service provision under decentralized auspices to the state, municipality and community levels that in particular needs shape policies that both reap the benefits of greater information and local responsiveness whilst managing the potential risks associated with weaker institutions, whether due to greater clientelism or exclusion or of lower bureaucratic capacities at local levels. These three areas all need further and will be the subject of future work drawing on analysis of Mexico's conditions and international experience, There i s both a major need and a huge opportunity to structure the wealth of natural experimentation going on inthe wide array of government programs and localexperiences to provide muchmore concrete andoftencontextualized results of what does anddoes not work. This will also involvea major deepening of the current evaluation agenda that SEDESOL has been pioneering. Some of the issues and techniques are discussedinChapter 6. 163 ANNEX Table11.4.1 RedistributiveInstruments Residential electricity VAT Fiscal * p ~ ~+~ nAll ~ ~ s *portccnidades subsidy Expenditure RES (RES) Public Expenditure 16,105 37,206 81,217 53,311 134,528 (MMx$) Deciles Distributional incidence 1 36.5% 3.4% 1.9% 13.4% 6.4% 2 23.4% 4.8% 3.4% 10.4% 6.2% 3 15.2% 6.3% 4.5% 8.9% 6.3% 4 10.0% 7.2% 5.7% 8.0% 6.6% 5 6.7% 9.8% 7.1% 8.8% 7.8% 6 3.4% 11.4% 8.2% 9.0% 8.5% 7 2.6% 12.1% 9.5% 9.3 % 9.4 % 8 1.8% 13.1% 11.5% 9.7% 10.8% 9 0.4% 16.4% 15.2% 11.6% 13.8% 10 0.1% 15.6% 32.9% 10.9% 24.2% cc -0.581 0.242 0.410 -0.006 0.245 Share of autonomouspc hhexpenditure) 1 17.5% 3.7% 4.5% 21.2% 25.7 % 2 6.3% 3.0% 4.7% 9.3% 14.0% 3 3.0% 2.9% 4.6% 5.9% 10.5% 4 1.6% 2.6% 4.5% 4.2% 8.7% 5 0.8% 2.9% 4.5% 3.7 % 8.2% 6 0.4% 2.7% 4.3% 3.1% 7.4% 7 0.2% 2.4% 4.0% 2.6 % 6.6% 8 0.1% 2.0% 3.8% 2.1% 5.9% 9 0.0% 1.8% 3.6% 1.8% 5.3% 10 0.0% 0.7% 3.1% 0.7% 3.8% ENIGH2002, 3er Informe de Gobierno, Fox. SHCP (2003). Population decilesordered by autonomousper capita expenditure. 164 TableA.4.2 Redistributiveimplicationsof expanding~ p o r t u n ~ financed e s throughreductionsinthe~ e s i d e n tElectricity Subsidy (RES) andFiscal ~ l ~xpenditureson VAT RES 100% to 50% to Opo~unidade~ loo%to ~ p ~ ~ u n i d a d e(fullcoveragepoorest20%) ~ , Current Oportunidades full coverage current RES targeted to coverage of poorest distribution current poorest50% 20% ortionatel Deci1es Distribution 1 13.4% 36.5% 44.1% 29.9% 38.8% 2 10.4% 23.4% 37.6% 26.1% 33.7% 3 8.9% 15.2% 9.0% 8.0% 11.4% 4 8.0% 10.0% 4.3% 5.3% 7.8% 5 8.8% 6.7% 2.3% 4.9% 6.2% 6 9.0 % 3.4% 1.3% 4.8% 1.0% 7 9.3% 2.6% 0.8% 4.7% 0.6% 8 9.7 % 1.8% 0.4% 4.9% 0.3% 9 11.6% 0.4% 0.1% 5.8% 0.1% 10 10.9% 0.1% 0.1% 5.5% 0.0% cc -0.006 -0.581 -0.714 -0.380 -0.667 Incidence (% autonomouspc hhexpenditure) 1 21.2% 58.0% 70.1% 47.5% 61.7% 2 9.3 % 20.9% 33.6% 23.4% 30.2% 3 5.9% 10.1% 6.0% 5.3% 7.6% 4 4.2 % 5.2% 2.2% 2.8% 4.0% 5 3.7% 2.8% 1.O% 2.1% 2.6% 6 3.1% 1.2% 0.5% 1.7% 0.4% 7 2.6% 0.7% 0.2% 1.3% 0.2% 8 2.1% 0.4% 0.1% 1.1% 0.1% 9 1.8% 0.1% 0.0% 0.9% 0.0% 10 0.7% 0.0% 0.0% 0.3% 0.0% 165 TableA.4.2 Continued VAT VAT RES + 20180 VAT 20180: l 100% 100% Oportunidades Current O p Public o ~ u n ~ ~ Current Pub. Spend. e s l O p o r t u n ~ ~ e sOportunidades,~ p o r t u n ~ e s , present full coverage Spending RES 50%: ~ p o r t u n ~ e s coverage poorest 20% Deciles Distribution 1 7.6% 18.4% 6.4 % 18.8% 36.5% 44.1% 2 6.7 % 13.7% 6.2 % 13.8% 23.4% 37.6% 3 6.3 % 11.6% 6.3 % 11.3% 15.2% 9.0% 4 6.4% 9.4% 6.6% 9.2% 10.0% 4.3% 5 7.0 % 8.8% 7.8 % 8.6% 6.7% 2.3% 6 7.4% 7.6% 8.5 % 7.5% 3.4% 1.3% 7 8.4% 7.9% 9.4% 7.7% 2.6% 0.8% 8 9.9% 8.6% 10.8% 8.3% 1.8% 0.4% 9 12.8 % 7.8% 13.8% 8.0% 0.4% 0.1% 10 27.5 % 6.2% 24.2% 6.7% 0.1% 0.1% cc 0.230 -0.193 0.245 -0.196 -0.581 -0.714 Incidence (% autonomous pc hhexpenditure) 1 22.0 % 53.5% 25.7 % 75.6% 146.4% 177% 2 11.0% 22.3% 14.0% 31.1% 52.8% 85% 3 7.6% 14.0% 10.5 % 19.0% 25.4% 15% 4 6.1% 8.9% 8.7 % 12.0% 13.1% 6% 5 5.4% 6.7% 8.2% 9.1% 7.1% 2% 6 4.7 % 4.8% 7.4% 6.6% 2.9% 1% 7 4.3 % 4.0% 6.6% 5.5% 1.9% 1% 8 3.9 % 3.4% 5.9% 4.5% 1.O% 0% 9 3.6 % 2.2% 5.3 % 3.1% 0.1% 0% 10 3.1% 0.7% 3.8% 1.O% 0.0% 0% 166 June 2004 Poverty inMexico An Assessmentof Conditions, Trends and Government Strategy Chapter 5. Influences ongrowth, distributionandpoverty reduction Growth in incomes of the poor has been particularly weak over the last two decades, especially when compared with gains inhuman and physical assets of the poor (Chapters 2 and 3). This chapter discussessome of the influences andpolicy options. While specific government programs are part of the story, economy-wide and sectoral influences on the level and pattern of growth are generally of greater importance. The environment for new investment i s fundamental to this, understanding this to include the range of market, infrastructure, regulatory and institutional influences on investment decisions. This matters both to the overall level of investment and to the patterns of investment across large and small firms, self-employed entrepreneurs and farmers, that jointly shape both the overall levelof growth and the extent to which growth is poverty-reducing. This chapter discusses the various influences on the investment environment, inequality and growth. It is divided into four sections. First, the discussion is motivated by illustrative quantitative scenarios that show that both growth and inequality matter to the changes inincome poverty. Second, the influence of a range of economy-wide influences on both growth and inequality are discussed. This relates past and proposed economy- ~ wide structural reforms to the poverty reduction challenge. The third and fourth sections then focus on the conditions of two key groups amongst the poor -urban households living in moderate poverty and rural households living in extreme poverty- and discusses the very different policy issues around improving the investment environment and developing a growth strategy for these groups. There are, of course, also urban households living in extreme poverty and rural households living in moderate poverty that may have distinct needs. However, we chose to concentrate on the moderate urban poor and extreme rural poor policy issues to give more focus to the questions. The conclusion both summarizes and notes questions and issues for debate and future work raisedby the preceding analysis. A. Growth, inequality andfuture prospectsfor incomepoverty reduction The pace of poverty reduction i s a product of interactions between changes in average incomes and income inequality. The higher the growth rate, the faster the pace of poverty reduction. However, the impact of growth on poverty i s generally lower when income inequality i s higher (that is the elasticity of changes in poverty rates to the changes in averageincomes i s smaller). A reduction ininequality that increases the share of the poor has both once-off effects and increasesthe responsiveness of poverty changes to growth. We show the quantitative importance of these factors by a simple projection of alternative scenarios for incomes and inequality in Mexico through 2015. It asks the question, what would happen to poverty if incomes and inequality followed certain paths? 167 To design possible scenarios we need assumptions on both growth prospects and plausible changes in inequality. The base used i s the level and distribution of incomes observed in the 2002 ENIGH. For growth a range was chosen from 5% annual growth from 2002 to 2015 in a "high" case to 2% annual growth in a "low" case with per capita income growth just under 1% lower .in each case owing to population growth. It is assumed that average incomes measuredin the survey grow at the same rate as aggregate growth. For changes in inequality, we assumed a range of scenarios from a modest worsening through to a "large" improvement. The distribution of income can change in many ways and not all matter for the poor (for example a shift from the very rich to the rich does not affect the poor). For convenience we assume that the Lorenz curve moved in or out by a given percent -5% for modest changes, 10% for large changes. A 5% worsening would take Mexico's level of inequality back to that prevailing in 1998-2000, a 5% improvement would take it to roughly that of Venezuela, and a 10% reduction to that of Argentina inthe early 1990's. The scenarios vividly show the influence of bothgrowth and inequality -selected results are shown inTable 5.1. Two percent growth with no inequality change leads to very little progress on extreme poverty by 2015. With inequality improvements there can be large reductions even with slow growth. Growth of five percent per annum leads to a dramatic reduction in extreme poverty with its virtual disappearance with inequality declines. There is also large variation for moderate poverty. For example, with five percent annual growth, moderatepoverty declines from 52 to 28% with no inequality change, but to 20% with a "large", 10%reduction ininequality. It is important to note that both growth and inequality improvement also help reduce the poverty gap and the squared poverty gap (these are measures of how far incomes of the poor lie below the poverty line. The squared poverty gap gives greater weight to the poorest with the largest gap). We should be more concerned with such measures of the depth of poverty than with the proportion of people who cross the poverty line. The squaredpoverty gap inparticulari s sensitive to changesindistribution. 168 Table 5.1 Both growth and inequality make a difference to the pace of poverty reduction Alternative scenarios for, food and assets-basedpoverty in2015 4 Average real GDP growth 2.0 5.0 Inequalitychange 5 0 -5 -10 5 0 -5 -10 Gini coefficient 50.77 53.32 50.78 48.24 45.70 53.32 50.78 48.24 45.70 Share ofthe bottomquintile 4.17 3.37 4.17 4.96 5.75 3.37 4.17 4.96 5.75 Extremepoverty headcount 21.22 18.25 15.30 9.85 8.32 6.36 5.51 3.69 1.73 Extreme Poverty Gap 5.31 4-94 3.90 2.53 1.68 1.08 1.20 0.54 0.14 SquaredExtreme PovertyGap 2.10 1.86 1.49 0.86 0.46 0.22 0.36 0.11 0.02 ModeratePovertyHeadcount 51.79 46.01 45.06 43.10 40.83 29.58 27.45 23.90 19.79 ModeratePovertyGap 21.61 18.81 17.33 15.02 12.81 9.86 8.39 6.26 4.48 SquaredModeratePovertyGap 11.24 9.84 8.65 6.88 5.38 4.35 3.53 2.35 1.51 These scenarios are intended only to motivate the discussion and illustrate the fact that both growth and inequality matters to poverty reduction. As noted, just how the income distribution shifts over time matters to poverty dynamics. Inequality could decline but leave out the extreme poor. Alternatively, the distribution could shift in a way that particularly benefited the extreme poor but helped the moderate poor less, as occurred between 2000 and 2002. Finally, these simulations have neither policy nor behavioral content. We now turn to a discussion of policy issues.' 3. Structural and policy influences on growth and distribution The hardest question to answer is the following: what i s the potential influence of external, structural and, especially, policy variables on both growth and inequality? This section examines this in terms of economy-wide factors, leaving to the next two sections the particular issuesrelevant to urbanandruralpoor. There are very large literatures on the determinants of income growth and inequality and, indeed, considerable work on Mexico. This section draws on selected results of recent work to explore the potential forces on growth and inequality. It should be stressed that the state of the art (indeed the state of the economic and social world) is far from a situation in which we can predict with any confidence the impacts of alternative structural changes or policies on incomes andinequality. However, we can say something on probable directions based on past experience. This i s done in three steps: first, we review what international cross-country analyses have to say on the relationships More sophisticated approaches can increase the behavioral content of such simulations. For example micro-simulation models can take account of decisions on education, labor force participation and fertility (Ferreira and Leite, 2003.) But there is no magic technique (yet?) for simulating with confidence the effects of policies on the full rangeof influences on growth anddistribution. 169 affecting Mexico's future growth and inequality; second, we look at evidence on patterns of change inthe labor market that i s particularly relevant to inequality and poverty; third, we discuss what this implies for the relationship between Mexico's structural reform agenda, inequality andpoverty. Lessonsfor Mexicofrom internution~levidenceon external, struc~urulandpolicy i n ~ ~ e n c eon growth and i n e ~ u u l i ~ s The first way we look at the relationships between factors associated with growth and inequality i s through use of some recent work in the tradition of cross-country analysis. This body of work compares the contrasting experiences across countries to draw conclusions on relationships between factors such as trade, external conditions, education and growth. This i s complemented by work from the smaller literature on relationships with changes inincome inequality. Care needs to be exercised indrawing causal lessons from this literature to the experience of a particular country -since, even with sophisticated econometric techniques, it can be difficult to assign causation from observation of patterns that appear to move together. But, with this caveat, the results provideusefulinformationon the likely relationships. While income levels and the income distribution are jointly determined, it is presentationally useful to start with results on determinants of growth. Here we look at the implications for Mexico of the results of a recent analysis of growth between 1960-99 from a global data base of 135 countries, 113 developing and 23 developed (see Loayza, Fajnzylber, and Calderrin; 2002, Tables III.2 and4). Usinga global database such as this makes possible muchmoreprecise andreliable estimationof the relationships than would otherwise be possible. Moreover, the results from the global analysis track well the experience of Latin American countries. The authors of the study statistically disentangled the influence of a wide range of factors on observed growth. A few aspects of this work are highlighted here. Growth determinants are divided between short-run cyclical effects (such as bouncing back after a recession), the influence of structural reforms, stabilization policies, and external conditions. Amongst structural measures, they used proxies for education, financial depth, trade openness, government burden, provision of public services and infrastructure, and governance.* For stabilization, measures of price stability, output volatility and systemic banking crisis were used. Finally terms of trade and period-specific dummies (to capture global conditions) were used as indicators of external influences. These various factors can be thought of as shaping both the overall investment environment and the efficiency of resource allocation. The application of the analysis to explainingMexico's past growth performance (to 1999) i s summarized in Table 5.2, where the model-predicted change in growth between various periods i s compared with actual changes. As can be seen, the results capture 2 The variables used were: secondary school enrollments for education; private financial credit over GDP for financial depth; total trade over GDP, adjusted for size and other country characteristics for trade openness; government consumption over GDP for government burden; and maintelephone lines per capita for infrastructure. 170 reasonably well the growth decline in the 1980's, the growth recovery in the 1990's, as well as the average change between the first and second half of the 1990's (the actual growth inthe latter periodwas actually higher than the model predicted). It i s noteworthy that the analysis indicates that there were growth-enhancing structural changes for all the periods with the fluctuations across periods explainedby changesin stabilization policies, cyclical effects and external conditions. Table 5.2 Contributions to Mexico's past growth changespredicted by a global analysisof growth determinants Contributionto growthrates inincome per capita Actual Projected Transitional Cyclical Structural Stabilization External change change convergence reversion reforms policies conditions 1980sVS. 1970s -3.87 -2.61 -0.49 -0.39 1.08 -0.98 -1.84 1990s vs. 1980s 1.72 1.80 0.05 0.19 1.51 0.24 -0.19 91-95 ~3.86-90 0.00 -0.17 0.03 0.45 1.02 -1.11 -0.56 96-99VS. 91-95 3.88 2.29 0.03 1.23 0.66 0.31 0.06 Source: Loayza, Fajnzylber, andCalderbn, 2002, Tables 11.4 and11.5. The model results can also be used to project Mexico's growth and we use this exercise to show the estimated influence of different structural policies. Two exercises are undertaken inthe study. The first i s basedon values of the explanatory variables basedon the projection of past trends -this can be in interpreted as projected growth for the first decade of 2000's relative to the 1990's. The second sets the explanatory variables to levels prevailing for countries in the 75thpercentile of the world distribution -that can be interpreted as the secular growth that would obtain if Mexico hadthese characteristics, ignoring transitional issuesof how it couldmove to such a position. Table 5.3 Mexico's predictedfuture long-rungrowth, basedon a global analysis of determinants of growth Contributionto growthrates inincome per capita At levelsofexplanatory At levelsof explanatory variables projectedfrom variablesof the 75th pasttrends percentileinthe world Growth in incomeper capita, 1991-1999 1.42 1.42 Predictedchange in growth 2.29 3.25 Projectedgrowth inthe 2000s 3.72 4.67 Contributionof structur~lfactors 1.98 2.03 Education 0.88 0.60 Financial depth 0.03 0.63 Trade openness 0.57 0.00 Governmentconsumption 0.02 0.00 Infrastructure 0.48 0.8 Contributionof macrofactors 0.74 1.22 Note:The total also includes contributions fromexternal, transitional convergenceandcyclical factors; these are not shown inorder to focus on policy-relevant factors. Source: Loayza, Fajnzylber, and Calderbn, 2002, Tables 111.2 and4. The results based on values projected from past trends suggests that Mexico could achieve an annual per capita income growth of 3.7% (and not far short of 5% interms of 171 total growth) with significant contributions from education, infrastructure, trade openness and macro stability. By contrast, if Mexico had the characteristics of a country at the 75th percentile of the global distribution for all variables, an even higher growth rate is predicted owing to larger contributions from financial depth and infrastructure (areas where Mexico has a relative deficit) as well as from macro stability (while it is assumed that a bankingcrisis would not occur, the history of macro instability i s a negative factor in the projection for Mexico relative to the 75th percentile). In this scenario, the contribution from education and trade openness is lower, indicating that Mexico i s above the 75thpercentile for these variables (however, the education variable used-secondary enrollment- does not take account of the quality issues discussed in Chapters 3 and 4. As emphasized there, additional efforts on education will be important to both growth andpoverty reduction). These projections are no substitute for an in-depth analysis of Mexico's short and long run growth potential. But they are indicative of what is achievable based on past international patterns and they underline the positive potential contribution of education, infrastructure, finance, trade and macro stability to growth. The pursuit of effective public action in these areas looks consistent with the achievement of the 5% medium term growth rate innational incomes used in the rapid growth projection shown inTable 5.1. So far the analysis has not taken account of distributional changes. We first look at some work from cross-country analyses that are close in spirit to that just presented. This uses results of a recent exercise by L6pez (2003), who used the same data set and approach as Loayza, Fajnzylber, and Calderh (2002), but with the addition of variables on income distribution from a global data set with information on changes over time for 133 cases. He finds considerable inertia in measures of income inequality -in line with other work- but also interesting patterns of association with the variables treated as growth "determinants". As with any multivariate analysis of this kind, the results for any variable refer to patterns of association after controlling for the influence of other variables. Outcomes for income inequality reflect the product of the interaction of competing influences. Key results are summarized inTable 5.4. Table 5.4 Structuralfactors and macro-stability are associatedwith changes in inequality Variable Influence oninequality (as measuredby changes inthelogofthe Gini) Education Decreases(not robustly significant) Financial depth Increases Trade openness Increases Governmentburden Decreases Infrastructure Decreases Governance No significant relationship Price instability Increases Cyclical volatility No significant relationship Bankingcrisis Decreases Note:resultsfrom GeneralizedMethodof Moments econometric analysis on across-countrydatabase. For variables usedfor structural conditions see note 2 above. Source:LEipez, 2003. 172 Education is associatedwith lower inequality -as well as higher growth- but the result i s not robustly significant across specifications. The lack of robustness probably reflects the fact that secondary enrollment i s a very poor indicator of determinants of returns to education that reflect the interaction between the relative supply and relative demand for skills amongst workers now in the labor force. Secondary education influences the relative supply of skills but only over time. Moreover, in Mexico, as in much of Latin America, the key divide has been between tertiary education and lower levels of schooling. This is discussedfurther inthe next subsection. Inequality-increasing associations were found with greater trade openness andfinancial depth -two areas of structural refoms that are associated with higher growth (Table 5.3). A negative influence of trade openness is consistent with studies on Latin America that interpret increasedtrade as mediating skill-biased technical change (see De Ferranti et al, 2003; and Shchez-Pkamo and Schady, 2003). The influence of greater financial depth could be interpreted as working through a lowering of the cost of capital, encouraging use of more capital-intensive technologies (that tend to be complementary with more skilled, better paid workers). These channels of influence on inequality are consistent with country-specific work on Mexico that has focused on impacts of trade- opening and capital investment on the labor market (see Hanson, 2003, and discussion of the labor market below). It i s also typically the case, and true in Mexico, that those working inthe financial services sector earnrelativelyhighsalaries. ~nfrastructure expansion is associated with reduced inequality as well as promoting growth. This is further supported by other recent country level analysis (Calderbn and Chong, 2003; Calderh and ServCn, 2003). The latter work uses similar data sets and approaches to the above studies but with a much more extensive set of information on infrastructural variables, including data on the stock of telecommunications, electricity and land transportation (roads and railways), as well as measures of service quality. These studies find that more andbetter quality infrastructure is good for both growth and income inequality. With respectto growtheffects, it i s the quantity, not the quality, that i s significant in the econometric analysis. After controlling for a range of other influences on growth, Calder6n and Servbn find that about a quarter of the variation in growth in 1996-2000, relative to 1981-85, i s explained by difference in infrastructure stocks. With respect to income inequality, both the quantity and quality of infrastructure has a significant negative influence: more and higher quality infrastructure is associated with lower inequality after controlling for other influences. Calderh and Chong estimate that an increase in infrastructure equivalent to moving up one quartile of the global distribution of infrastructure stocks is linked with a reduction in the Gini of 2.2 points in the following five year period and 12 points in the subsequent 35 years. Such a long-run effect is equivalent to a shift from the level of inequality prevailing in Mexico to below that inCosta Rica. The cross-country result on infrastructure resonates with the high degree of spatial heterogeneity that was discussedinChapter 3 -since better infrastructure canbe a major source of greater spatial integration In Mexico, during the 90's, investment levels have been at best flat and, after a short-lived boom in telecommunications in the late 1 9 9 0 ' ~ ~ 173 private investment recently displayed a significant decline. Public investment has recovered since 2001 particularly in sectors such as electricity, roads and water. However, levels of investment remain low by international standards, and more effort will be needed to reactivate investment in infrastructure in both the public and private sectors. Figure 5.1 Investment in infrastructure in Mexico and LatinAmerica fell steeply in the 1990's 3.5% 3.0 % 2.5 % & ; fi 2.0% 2 z 1.5% m 1.0 % 0.5% 0.0 % 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Note: Investment in infrastructure includes roads, electricity, railways, airports, ports, water and telecommunications. Data for 2003 are estimates. Source: Anexo del Tercer Informe de Gobierno 2003. We would finally note the more complex relationship between macro s ~ a b i l iand ~ inequality in this analysis. On one hand, greater price stability -that in Latin America means avoiding high inflation- i s associated with lower inequality. This i s consistent with country-specific analyses that find higher inflation goes with more inequality possibly because the rich have more financial and other instruments to cope with high inflation(See Ferreira and Litchfield, 1996, for Brazil). On the other hand, ban~ingcrisis are associated with lower inequality as measuredinthe household surveys used in this analysis. This i s consistent with the finding that the premia to tertiary workers tends to decline inthe wake of suchcrises, as noted inChapter 3 for Mexico, and as observed more generally for Latin America (Slinchez-Pdramo and Schady, 2003). However, this is likely to be only a smallpart of the distributional story of banking crisis. Recent work by Halac and Schmukler (2003)on Mexico and several other Latin American countries finds that banking crisis and government rescues are strongly regressivebut that the action takes place in the interactions with the financial sector and fiscal position. These are likely to be captured poorly, if at all, by household surveys. For Mexico, they estimate the total fiscal and quasi-fiscal costs of the 1994-95 crisis to be $135 billion, about a quarter of Mexico's 2000 GDP -or about two years worth of total 174 tax revenues. This primarily benefits those within the financial system -shareholders, depositors and borrowers- who are relatively wealthy within the Mexican income and wealth distribution. The transfers were financed by higher budgetary surpluses than would otherwise have been needed. On the basis of average incidence estimates, this was paidfor roughly proportionally by all the population -poor and rich alike. And the lost spending could have hurt the poor disproportionately at the margin, since the expansion of spending in many programs has been even more progressive (Chapter 4 and World Bank, 2004a). As banking crisis unfold, it is also usually the case that the larger and better-informed depositors move their money out of domestic assets first, potentially reaping capital gains when domestic assets prices fall (notably via exchange rate effects) and bigger and more influential equity-holders and borrowers get favored treatment in bailouts. The ENIGH, while only capturing a fraction of total capital incomes does register gains in financial income at the top of the distribution between 1994 and 1996 (De Ferranti et al, 2004, Chapter 8) inthe context of generalized declines intotal incomes from all groups. These considerations only sharpen the conclusion on the deep costs of macroeconomic instability in general, and the 1994-95 crisis in particular, for poverty and inequality. Chapter 3 showedhow severe was the impact of the crisis onpoverty reduction. Inequality dynamics in the labor market Labor incomes account for a large part of household incomes of the poor and a significant part of the action between policy and structural changes andinequality occurs via the labor market. The linkages'between the labor market and the household income distribution are mediated by decisions on household formation, labor force participation and choices over household size, so rises or falls in wage inequality are not always directly reflected in changes in household income ineq~ality.~ But it remains an area of interest for structural changes and potential policy which is of particular importance to the poor. There are several areas that affect wage dynamics. The economic cycle. The level of economic activity has a powerful influence on aggregate labor demand, the level of wages and unemployment. As Figure 5.2 shows, workers were hard-hit by the 1994-95 crisis in terms of high unemployment and falling wages. Unemployment increased sharply from about 3.5% in 1993-94 to over 6% in 1995, but then fell back to pre-crisis levels by 1998. The relatively quick recovery of employment i s partly due to highwage flexibility: manufacturing wages fell almost 2096, recovering to 1994 levels in 2000 and to 1993 levels only in 2003. As seen inChapter 3, these labor market dynamics was closely mirrored in poverty changes. Avoiding crisis and supporting stable growth i s clearly goodfor wage developments. See Bourguignon, Ferreira and Lustig (1998) for a discussion of the methodological approachof micro- simulations for analyzing the various channelsof influence on householdincome distribution. 175 Figure5.2 Average wages, growth, and unemployment 140 izn 100 80 60 40 i n 0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Msaufaeturlng industry wsgea (1993~100) l G D P(1993=100) +Unemployment rate Source: WB staff calculations basedon BE. The supply and demandfor skills. The structure of wages is as important for poverty as average wage levels. Mexico has a steep wage hierarchy, both with respect to schooling and spatial differences. This is partly because of the long-run inheritance of unequal schooling. Furthermore, it experienced rising differentials, especially for workers with college education, in the late 1980's and much of the 1990's, followed by declining differentials since about 1997. This can be interpreted as a product of the interactions between changes inthe relative supply and relative demand for different skill categories. Educational expansion of the pasttwo decadeshas ledto a gradual increase inthe relative supply of workers with secondary education and tertiary education -whose share in the workforce rose by some 50 and40% respectively between the late 1980's and late 1990's (De Ferranti et al. 2003, pp. 51 and 54). Since the premium to education (especially for tertiary education) rose, this implies shifts in the relative demand for skills more than offset these increases in relative supply. While observed wages are a product of the interaction between supply and demand changes it i s possible to derive the implicit shifts on the demand side by making assumptions on the extent to which different skill categories can substitute for each other in the production process (see Siinchez-Pkamo and Schady, 2003). The results of such an analysis is presentedinFigure 5.3: this shows that there were large increases inthe relative demandfor college graduates inthe 1990's, followed by amodest decline after 1997. 176 Figure 5.3 The evolution of the relative demand for workers with tertiary education, MexicoUrbanAreas, 1988-2001 1.2 0.0 I11IIIIVI11IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIIIVII1IIII\ 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 -c-a=l -4-(1=2 +a=3 Notes: The estimation of relative demandfollows Schady and SBnchez-Piramo(2003) in assuming three levels of elasticity of substitution (0)betweenupper secondary andtertiary educationworkers. The evolution of demandconsidersthree year moving averageestimations including 12 quarters per observation. Source:WB staff calculations usingthird quarter of ENEUfrom 1988to 2001. Mexico's pattern of rises in wage differentials for tertiary education i s shared by most middle-income Latin America countries (Behrman, Birdsall and Sztkely, 2001; De Ferranti et al., 2003). It i s generally associated with the radical opening to trade and investment that startedinthe late 1980's and, for Mexico was consolidated with NAFTA. This opening mediated a process of skill-biased technical change as firms adopted new techniques and modernized internal and external organization processes. The phenomenon of a levelling off, or partial reversal, of the relative demand changes also occurred in Chile in the late 1990's. Since Chile and Mexico are the two countries that have experienced the deepest international integration this supports the view that the boost to increased demands for high skills from opening up may be a once-off phen~menon.~ Just as the rise in skill differentials in the early 1990's tended to sustain or increase household income inequality, so the falling differentials between the late-1990's and 2003 will be an equalizing force for household incomes and was one of the influences behind the distributional improvement for 2000-02. The timing of the change around 1997 suggests that it was a structural rather than a cyclical phenomenon. While wage differentials may rise somewhat with the recovery, the overall past pattern provides some hope that further educational expansion will be associated with further declines in This is consistentwith theoriesthat skills matter moreduringperiodsof major change inproduction. 177 differentials and not be offset by further large increases in relative demand. This should be good news for both inequality andpoverty. InternationuZ migration and remittunces. An additional major influence on the structure of wages i s the large-scale migration to the United States. Over the past few decades this migration has been biased toward more skilled workers reducing their relative supply in the mexican economy. Chiquiar and Hanson (2002) show that in 1990 the education group of mexicans most represented in the United States was that with 12-15 years of schooling, corresponding to upper high-school and the leveljust above, which fall in the middle and upper segments of Mexico's wage distribution, Mishra (2003) estimates that this has been an additional force for rising skill differentials in M e ~ i c oMoreover, . ~ migration has historically been concentrated in some states, notably Zacatecas, Michoach, Guanajuato, Nayarit and Durango where mexican labor was sought by US employers and labor agents through temporal employment contracts. Between 1998 and 2001, 42% of the temporary migrants crossing the northern border to United States still came from the traditional sending states located in the northern part of Mexico (CONAPO, 2002). This was anadditional force for risingspatial differences inincomes. While migration has been inequality-increasing it has also reducedpoverty. As discussed inChapter 3, at least partof the flow hasbeenfrom very poor ruralareas andremittances have hada large (and rising) influence onthe reduction of extreme poverty of households involved. In terms of the spatial distribution, there has been a steady increase of emigration from some central and southern states such as Puebla and Oaxaca, and from the state of Mexico. Institutions and labor market functioning, Labor market institutions, both formal and informal, can modify market-determined patterns of wage differences. Three features of the mexican labor market are of particular relevance to inequality and poverty. First, the labor market has historically been highly flexible interms of wage adjustment -including downward real wage declines in crisis. This has been complemented by significant flows between formal and informal work, with a contracyclical, as well as a large structural component to informal employment. These factors have led to low open unemployment rates or only brief spells of high unemployment inresponseto crisis (see Figure 5.2), in sharp contrast to countries such as Argentina and Chile (Maloney, 2003). This may be changing: the achievement of low inflation rates appears to have reduced downward real wage flexibility and the labor market appears to be moving to a pattern of greater quantity adjustment to adverse demand developments -as inthe relativelyhigher unemployment in the recent mild recession (Garcia Verdd, 2002b). This increases the importance of developing social protection instruments to manageunemployment risks in an efficient fashion (see Chapter 4). 'Thisappears to apply less to the top of the skill distribution (graduates and post-graduates). Tremblay (2002) shows that relative to other developing countries, the number of such highly-skilled mexican migrantsinthe United States i s not high.For this partof the distribution the hightop salaries inMexico has tended to keep mexicansinthe country. 178 Second, the influence of unions on labor markets i s very different between most of the private sector and the public sector. Inthe private sector, the effect of unions appears not to increase wage-differentials with non-unionized sectors but rather to reflect bargaining for employment (Maloney 1999).This i s good for poverty reduction since it tends to raise formal employment, a factor that i s of especial value for the moderate poor. Unions of course have a rangeof other functions, including protection of worker rights and ensuring observation of health and safety standards. In the public sector unions are commonly associated with preservation of work-practices or firm structures that inhibit efficiency- improving reforms, notably in teaching, electricity and oil and gas (L6pez-Acevedo and Salinas, 2001, LarreandBonturi,2001, L6pez-Calva, 1998). Third, there remains an important division between the formal and informal sectors. While there are significant movements between formal and informal work, a significant partof which is voluntary (Maloney, 1999),highlevels of informality are associatedwith lower wage and the absence of formal social security. This is of particular relevance to the moderate poor in urban areas and i s discussed further in the section on this issue below. Spatial and r ~ r u l - ~ r developments. The 1990's rise inwage differentials had a spatial b u ~ dimension with areas closer to the border more integrated into international markets and with a higher concentration of foreign direct investmentexperiencing relative wage gains (Hansen, 2003). Inmany cases these were the initially better educated and higher income areas. This was an important factor behindrising spatial differentials inincomes noted in Chapter 3. Overlaid on this are large differentials between rural and urban wages. However, at an aggregate level there has been a reduction in rural-urban wage differentials in the past decade, apart from in 1994 (Table 5.5). This i s plausibly due to rising integration in the national labor market and, as noted in Chapter 3, has coincided with rising income inequality within the rural area. Rural wages were particularly buoyant between 2000 and 2002,but it is too early to tell ifthis will be sustained. Table 5.5 There are largebutdeclining differences between ruraland urbanwages Meanhourly wages by area 1992 1994 1996 1998 2000 2002 Rural 1,208 1.5 2.1 3.3 5.1 6.3 Urban 4,398 5.6 7.1 10.1 15.2 15.9 ~ u r a ~ U r WageRatio b ~ n 27.5% 27.0% 29.7% 32.5% 33.6% 39.6% Source: WB staffcalculations usingENIGH1992-2002. The structura~reform agenda andpoverty reduction We can now draw on the previous subsections to turn to the question of the relationship between Mexico's structural reform agenda and poverty reduction. Structural reforms have been a topic of considerable public debate inMexico. This has been heightened in the recent past by the China's symbolic overtaking of Mexico in terms of shares of the United States market. This has led to rising recognition of the need for a new round of structural reforms to increase competitiveness. Reliance on the gains from macro 179 stabilization and NAFTA will be insufficient to sustain rapid growth. The present government has sought to undertake reforms in a range of areas including tax, labor, electricity and competition reforms, but these have been resisted by various segments of mexican society, and there has been little or no passageof actual reforms. There are many considerations around the design of structural reforms. Here we provide a qualitative summary on what the experience reviewed above suggests for the relationship with poverty reduction andinequality (Table 5.6). Table 5.6 The potentialimpactbetween selectedstructuralreforms, inequality and poverty Structural reform Potential effects onpoverty andinequality inMexico area Macro stability Instability and crisis are pernicious for growth and regressive in their impacts. Macro stability is essentialfor poverty reduction. Trade andFDI Mostly consolidated in NAFTA. Opening increased trade and FDI but tended to opening increase wage and spatial inequality, at least until the late 1990's. Complementary measures are thus of central importance. Humancapital Despite past progress, there is a major educationagenda that is important to growth and essentialfor inequality andpoverty. However, effects oninequality andpoverty are gradual and felt in full over the long-term. Health policy matters for human capital of the very poor and to reduce income costs associated with major illnesslcatastrophicshocks. Infrastructure- Essential for growth, and has the potential to reduce inequality if used to integrate investmentand poorer ruralregionsand urbanneighborhoodsinto economic development. institutional reform Electricity Current structures are both inefficient and inequitable -reforms are necessary to growth and can be designed to be strongly pro-poor. In other countries use of private sector has led to equitable expansionas well as efficiency gains. There is a strong case for reduction of subsidies(see scenarios inChapter4). Labor reform Current labor regulations raise investment costs, are inefficient and encourage informalization. Labor reform can help the moderate poor and improve the investment environment; some (mainly non-poor) workers inthe formal sector may suffer reducedjob security. Competition policy Can reduce costs of market power (e.g. in telecommunications) and help growth; hurts relatively better-off beneficiaries of current concentration -potentially both owners and workers. Fiscalreform Fundamental input for poverty and inequality-reducing change through providing resources for education, infrastructure and transfers and compensation for any losers. These structural reforms can be divided into four categories. First, macro-stabilization, trade and investment integration are now firmly established, especially under NAFTA. These are central elements of a favorable investment environment, have been good for growth, but tended to increase inequality in the 1990's 180 -through the form of crisis resolution after 1994-95 and the (possibly transitional) effects on wage differentials of trade and foreign investment. Second, human capital development i s fundamental to reducing income inequality and poverty reduction but the effects of education, in particular, are a matter of decades, rather than years. There i s no controversy over the importance of human development, though there are major policy and institutional challenges of both design and implementation (see Chapter 4). Third, there are a set of reform measures that are important to current challenges of competitiveness and the investment environment -including infrastructure, energy, labor, competition- that have so far proven more difficult to implement. Our reading of the international evidence is that these can be inequality reducing, as well supporting competitiveness providedthey are well designed. Fourth, fiscal reform i s perhapsthe lynch-pin of all the reforms since increasedfinance i s central to many of the other reforms. Moreover, as argued in Chapter 4, a fiscal reform that increases the tax effort and allows higher social and economic sector spendingwill have an automatic tendency to be inequality reducing andcan be designedto be strongly pro-poor. These conclusions are no substitute for careful assessment of the effect of specific reform packages. However, the general message is clear. Competitiveness-oriented structural reforms are essentialfor income growth of the poor. Those who may lose inthe transition are more likely to be incurrently protected activities of the upper segments of the income distribution. However, economy-wide and sectoral reforms alone are not enough. To assure different groups amongst the poor participate fully in the economic benefits of reform, there will be a need to shape policies to respond specifically to their needs and potential. The next two sections illustrate this with respect to two groups -the moderate urbanpoor and the extreme ruralpoor. C. Strengtheningincomegrowth for the moderate urbanpoor In2002 there were some 19 millionpeople living inmoderatepoverty in urban areas (as well as a further 12 million in rural areas). The government has recently focused on the conditions of this group and has proposed a strategy, Impulso, to respond to their needs and potential (Presidenciu, 2003). This section draws on the latter diagnosis and discusses element of the strategy.6 In their diagnosis Presidencia uses a definition of moderate poverty of those in between the capacities poverty line and the assets-based line. This differs slightly from that usedin Chapter 3 where, to simplify the presentation, we used the term moderate poverty for those in betweenthe food-based and assets-based poverty line. Since the food- and capacities poverty lines are close, this does not make a significant difference to the analysis. 181 Diagnosisof the conditionsof the moderate urbanpoor. The moderate poor generally have enough resourcesto satisfy food and some other basic needs but not enough for a fuller array of housing, schooling, transportation and other requirements to pursuea decent life. Many heads of householdsinthis group have at least basic education. While a third have less then primary, some 30% have primary and 27% lower secondary schooling (in 2002). The moderate poor have some assets with close to 70% owning their house, although only 5% reported that they were paying for their house -an indicator of the predominantly self-financed process of house acquisition and construction. Their dwellings have on average 2.5 rooms and a bathroom and are usually built of long-lasting materials such as cement and bricks. Close to 90% of these households owned a television, almost three-quarters a refrigerator, and somewhat less than a half a washing machine. About 20% have a car. While average per capita income i s 1,063 pesos per month in 2002, ownership of such assets indicates some capacity for savings. It could also reflect the significant churning inthe income distribution, noted in Chapter 3, with those now classified as moderately poor having previously earned more andfallen on harder times. Interms of economic activity, aquarter of householdheads were self-employed and70% were workers. Most worked in services, commerce, manufacturing and construction (Table 3.10). Amongst all economically active members of households, about two out of three work in the informal sector. As Figure 5.4 illustrates, while there i s considerable variation and overlap across employment categories, most self-employed and informal workers have much lower labor incomes than formal worker^.^ The fortunes of the urban moderatepoor are closely linked to overall macroeconomic and labor market conditions in the economy since they work in activities that are typically integrated into the national economy. Their incomes are then especially influenced by wages for unskilledand semi-skilled work and to developments inthe informaleconomy. General influences on wages were discussed above: for much of the late-19SO's and 1 9 9 0 ' ~low-skilled wages were hit by the combination of poor overall wage growth ~ (especially due to the crisis) and the strongly skill-biased pattern of labor demand. As noted in the previous section, there is hope that the peak of this bias may have passed in the late 1990's. However, controlling for relevant individual characteristics there might be no differences inlabor income (Maloney, 1999). 182 Figure 5.4 There are large differences in the pattern of wages amongst self- employed, informalandformal workers Frequency distributions of wages Monthly labor income Urbanareas I I I I 0 2000 4000 6000 Pems 8000 10000 I --+-- Formal a I n f b d *... Selfempbyed salaried s h i e d S m e : ENE 2002,2nd quarter A primary focus of the diagnosis behind the lmpulso strategy concerns the nature and implications of informalemployment. A majority of the moderateurbanpoor work inthe informal sector and levels of incomes are lower than in formal activities. A higher fraction work without labor contract or social security (Figure 5.5). Most informal work i s inwhat is sometimes termed the "unstructured" sector in terms of production units. The International Labor Office defines this as including economic units of less than 15 employees in manufacturing or less than 6 employees in other activities, that are not registered by labor or fiscal authorities, often have no established premises andwhose workers do nothave laborcontracts or social security. 183 Figure 5.5 Poorer workers experience greater informality across many dimensions. 100 96.2 90 80 80.6 70 60 50 47.1 40 44.8 30 20 25.3 o ! I I i Extremeand Moderate Poverty Non Poor Intermediate Poverty -Percentage of employedages 18-65without contract -Percentage of owners and own-accountworkers with no account keeping -Wage employees under 1 minimumsalary, percentage of total wage employees Wage employeeswithout SocialSecurity, percentageof total wage employees Source: Presidencia (2003),fromENIGH. In2003, Mexico's non-structured sector included 4.9 millionestablishments,83.9 million of which were classified as self-employed individuals. Average labor productivity is significantly less than in structured activities (Figure 5.6) as a consequence of lower capitalization and the sectoral composition. Two-thirds of employment is in units with five or less workers. Within these micro-units, 37% are self-employed, 40% workers, one percent employers and 15% unpaid family workers. Almost 80% of employment i s in services and construction and the rest inmanufacturing. This partly reflects technological factors: in many manufacturing activities economies of scale make self-employed and very small-scale production uneconomic. Larger production units are more likely to be observed by the authorities and also more likely to be unionized. Both factors increase the probability of observanceof tax, labor and other regulations. ENET2003. Inmost national statistics, firms (businesses, enterprises) are those unitsthat makeupa legal or administrative entity whereas establishments are production units at a single location. Thus, there is at least one establishment per firm. For micro-enterprises and often also for small enterprises the two categoriescoincide. 184 Figure 5.6 Labor productivity inthe structured and non-structured sectors Co-ard -0s of 1993 Y 50*~n o o n I ~~ ~~ 60 45,000 40,oon 35,000 30,000 40 25,000 20,000 15,000 20 10,000 5.000 0 0 1996 1997 1998 1999 2000 2001 Average value added per employer aNon-StructuredSector I Sector Structured person (AVAPEP): Ratio of AVAPEP in the non-structured sector relative to the structured sector Source: Presidenciu (2003) from INEGI-STPS,EncuestaNacional de Empleo @NE)e INEGI, Sisterna de CuentasNacionalesde Mdxico, CuentaSatdlitedel Sector Informalde 10sHogares. In2002,nearly 73% of the micro firms obtained financing from friends andrelatives and only 0.2% from government programs. They very often rely on family ties and informal connections for labor contracting. About 70% of the micro firms did not pay taxes. Informal firms rarely comply with the various registration obligations. Finally, while most informal work is in the unstructured sector, it also exists in registered establishments inthe form of employment without a labor contract or social security. Does informality of employment and production have social and economic costs? In assessing this question it is important to distinguish between whether informality makes sense from the perspective of individual workers and firm owners, from the broader question of whether it has costs to households and society. Choosing informal work or enterprise status i s likely to be a rationalresponse to the costs of regulation inactivities in which self-employed or small-scale production makes economic sense, because of the low economies of scale. There is evidence from both economic andethnographic analysis that movements between formal and informal work are often voluntary (see Maloney, 2003 for a review). Individuals may prefer informal work because of more flexible working conditions -this applies particularly to women, who are more likely to work less than full-time and also to both younger and aged workers. Some workers work in a formaljob, save andthen move into higher-return andhigher-risk self-employed or micro enterprises. Households may place one of their members in formal work in order to acquire social security rights, andothers ininformal activities. Informality can, however, have costs. For the micro or small firm and self-employed, avoidance of regulations -and associated illegality- increases insecurity, reduces the probability of obtaining formal finance, can increase transaction costs in trading and investment decisions and vulnerability to extra-legal process (such as paying bribes or 185 protection money to policy, officials or informal enforcers). Workers typically enjoy poorer labor rights and security -though the absence of formal social security may be partially offset by informal mechanisms. Since firms and households are often joint activities, business insecurities over incomes and property rights translate directly into householdinsecurity. Strategiesfor growth of the moderate urbanpoor 3ased on an overall diagnosis along these lines, the government has proposed the Zmpulso strategy which involves a mixture of accumulation, formalization and institutional change. It has two broad thrusts: improving the productivity and formalization of the unstructured enterprise sector; and raising the potential productivity of workers. These are illustrated in the diagrammatic presentation of the strategy in Figure 5.7. The firm-oriented part involves a combination of changed regulations to reduce the costs of formalization, including a particular focus on labor market reform, extension of property rights, direct support to the micro and small-scale sector and finance. For workers, the policy orientation i s primarily to education and training. The remainder of this subsection looks at experience inthese areas. Figure5.7 The policy framework of the~mpulsostrategy Managerialadvlcing Source: Presidencia (2003). Regulation and the cost of fo~alization.Mexico's regulatory environment is typically much more burdensome than in OECD countries and in many areas is comparable to or more costly for businesses than the Latin American average (Table 5.7). This i s particularly marked for labor legislation. This hinders the entry of new innovative and potentially highly-productive firms or slow the exit of low-productivity ones, impedes 186 and creates incentives for informality. On-going regulatory costs in Mexico are thought to comprise 24% of total costs for micro firms. Table 5.7 Mexico hasa highregulatory burdenininternational context. Mexico Averagefor Latin OECDaverage America and the Caribbean GNIper capita (US$) 5,910 2,987 23,135 Informaleconomy as shareof 30 42 17 GDP Startina a business: - Duration indays 51 74 30 Minimumcapital (% GNIper 88 86 61 capita) Hiringandfiring workers Flexibility of hiringindex 81 56 49 Conditions of employment index 81 79 58 Flexibility of firing index 70 48 28 Enforcing contracts Durationindays 325 363 233 Proceduralcomplexity index 62 70 49 Gettingcredit Public credit registry index 0 35 18 Closinga business Cost as % of estate 18 15 7 Source: World Bank (2004b). A number of weaknesses continue to weigh on mexican enterprises and especially small and medium sized firms. Institutionalinfrastructure (for example bankruptcy procedures and the judiciary system) have to be strengthened and bureaucratic processes simplified. Well-designed bankruptcy rules contribute to increasing aggregate productivity by helping poorly performing firms to exit the market. It was only in 2000 that a new bankruptcy law in line with international best practices was enacted, as well as a law on guarantees. However, the new law has hardly been applied, suggesting that there were still shortcomings and that further reform i s needed. The law on guarantees has already been reformedin 2003. In March 2002, a new system to start a business within one day (SARE -Sisternu de Apertura Rdpida de Empresas) was created for most activities. The Federal Regulatory Improvement Commission (Comisio'n Federal de Mejora Regulutoriu, COFEMER) has signed SARE agreements with several state and municipal governments to ensure full cooperation and participation of these local levels, where an important part of the formalities related to the start-up of a business i s concentrated. In particular, COFEMER agrees with local authorities to assist them in implementing the SARE with the counterpart that local authorities are committedto report monthly on the results obtained. By easing firm creation, the new system is likely to make formal activities more attractive, but it is too early to assess the efficacy of the measures. 187 ...and the special case of labor policy. Amongst regulations, labor policies are of particular interest to the moderate poor. Strict employment legislation contributes to increase labor costs and creates incentives to opt for informality, especially for SME's. The absence of a probation period combined with relatively high severance payments increases the cost of hiringpeople on permanent contracts. At the same time, fixed-term and temporary contracts are strictly regulated, limiting the possibility of firms to circumvent hiringcosts in the formal market. Overtime hours are paiddoublethe normal time rates and therefore much more than in most developed countries. In addition, red tape and administrative burdens reinforce incentives to hire workers informally. One study found labor obligations to be those most often evaded partly because, when fully complied with, they account for 50-60% of annual profits (as opposed to 17-30% for Chile). Ingeneral, firms comply with some regulations and not other^.^ The labor reform proposal sitting on Congress could represent a step in making the formal market more attractive. Some issues covered in this law include: the introduction of probation periods; first steps in easing regulations concerning permanent contracts (with the introduction of a system of cumulative work hours); efforts to modernize industrial relations; emphasis on merit-based promotions; some simplification of bureaucratic requirements for SME's; measures to promote labor force training and measures to reduce corruption and legal uncertainty. However, the new law does not address the main issues affecting the labor market, inparticular those that could increase the flexibility of hiring and firing. Beyond changes to labor market legislation, other initiatives should also be taken into account to improve the balance between the formal and informal sectors. They include enhancing the quality of services provided to formal workers and reducing the tax wedge associated with formalization, in particular employers' social security contributions for low-skilled workers (Santamaria and Lopez- Acevedo, 2003). Property rights, land and slum upgrading. The supply of urban land and weak property rights represents a major problem for urbanhouseholds. While Mexico-specific evidence i s weak, De Soto (2000) has argued that the absence of property rights i s a major constraint on investment indeveloping countries, via effects on investment incentives and finance. An estimated 32,000 hectares of land are currently available for urban use-sufficient to meet only two years of demand. Myriad complex and costly building and subdivision regulations have contributed to making low and moderate-income land development an uneconomic venture for the private sector. The lack of agreement on basic legal and administrative principles, operational inefficiency and the antiquated and insecure information systems of state property rights registries has helped make real private property ownership difficult and expensive in Mexico. Widespread insecurity of tenure andinformalownership has followed as moderate poor households opt out of this costly, cumbersome system." The bulk of residential land development in urban areas occurs informally, at costs far higher than that of formal-sector development. 9 Levensonand Maloney (1998). loFreyre, 2002. 188 About two-thirds of the land on the periphery of medium and large towns consists of ejidos, while high-end private sector developers control most of the remainder. This form of communal landownership, dating from the Mexican Revolution, has hada complicated legal regimen that contributes to makingprivate freehold ownership assembly of parcels of significant size and rational land development difficult and expensive. Since 1992, ejido lands can be privatized but the process is cumbersome and centralized under the federal government. Municipalities must then get the explicit authorization of the federal authorities -which i s often slow incoming- to intervene, raising the cost of subsequent regularization. Many of Mexico's cities show discontinuous spatial patterns as formal development leapfrogs over ejido landsinsearchof lower-cost parcelsfar from cities and employment centers, creating much higher costs for infrastructure provision and congestion (extra traffic, air pollution). These problems have resulted in government provision of virtually all the parcels used in social housing projects, massive government purchase of informally-urbanized ejido land in order to subsequently regularize and transfer title of these lots to their de-facto occupants and the accumulation of land reserves by government -often, a problematic process- for these two other purposes (social housing projects andejido regularization). A demand-driven program of land regularization, PROCEDE, was created in 1993. World Bank (2001b) found that land markets in ejidos that have undergone PROCEDE function reasonably well, possibly even better than in the private sector. At the same time, there is strong evidence that producers remainseverely credit-constrained, implying that the award of certificates and titles did not increase credit supply. Further interventions should thus focus on improving investment and management skills, something where ejidos still appear to lag the private sector. World Bank (2001b) concludes that in order to successfully complete PROCEDE, the capacity for conflict resolution, especially throughalternative mechanisms and out-of court settlements, needs to be strengthened, andthat opportunities for reducing the costs of the program needto be sought. Conflict resolution will be of increasing importance in any effort to complete PROCEDE. Slum upgrading has occurred -if at all- in piecemeal fashion, without planning, neighborhood participation, sequencing of investments or integration of physical and social improvements. The fragmentation of these functions within different government agencies at all levels of government has impeded a comprehensive urban development approach. Only within the Ministry of Social Development (SEDESOL) there are three agencies that are involvedinurbandevelopment including Comisidnpara Regularizacidn de la Tenencia de la Tierra (CORETT); Fondo Nacional de ~abitacionesPopulares (FONHAPO); and the ComisidnNacional de Foment0 a la Vivienda(CONAFOVI). Within SEDESOL, another program, Hdbitat, was designed to face the challengesof the urban poverty through a model of action that combines, among other aspects, the improvement of infrastructure and the equipment in the urban-marginalized areas, the implementation of social services andcommunity development actions and regularization of tenure. With its different modalities, Hdbitat tries to support the urban population in 189 moderate poverty with particular attention to the needs of women, people with different capabilities, children, adolescents, young adults, andthe elderly. Financial services. As noted above, a large share of micro and small firm are self- employed individuals with little access to regulated financial institutions (unbanked). A recent World Bank study (2004b) found that the unbanked are far from being a homogeneousgroup. While they tend to be poorer and less educatedas a group compared to those who have access to formal financial institutions, they also include individuals from high-income groups. Around half of the top decile of income earners are unbanked and around ten percent of those with elementary school education have checking accounts. This study also found that more than half of the unbanked work inthe informal sector. This i s with formal bank requirement which emphasize steady employment references from employers (Table 5.8). Table 5.8 The use of formal financial instruments increases with income Type of Financial Instrument Incomegroup of medianuser (decile) ln~orma~banking instrum~nt~: Tandas* 5" Savingsat home 6`h Coop or "Caja Popular" 6" Overall Unbanked Sh Formal Banking instru~ents: Savings, debit card or "AFORE" account 7" Checking account gth Investmentaccount gth *Overallare Banked s" tandas informal rotating savingsclub. Source: World Bank, 2004b. The government is trying to improve the regulatory framework of all the different types of financial intermediaries, particularly those used by the unbanked. In this regard, Congress approved the Popular Savings and Credit Law (Ley del Ahorro y Crkdito Popular) in June 2001, which regulates all institutions that accept savings and deposits and gives credit to people with low incomes. Savings and loans institutions and credit unions operating under the previous legal framework have to change their status and become Popular Financial Societies (Sociedades Financieras Populares) by 2004. This law establishes a supervision system that includes the National Banking and Security Commission (Comisidn Nacional Bancaria y de Valores, CNBV) for this category of institution. A working group has been created to propose a new framework for the savings and loans sector, and the newly created Banco del Ahorro Nacional y Sewicios Financieros (BANSEFI) could become an important actor in the strengthening of this sector. However, BANSEFI and the newly created Financiera Rural are mainly rural. Therefore, much more needs to be done to reach the unbanked inurban areas. The recent entrance of Sanborns and Elektra (commercial establishments) into banking for low- income household seems to be an important step in that direction. Banco Azteca (Elektra's bank) is in a unique position at the moment because Elektra has a credit 190 history of its clients. However, interest rates in Banco Azteca are still very high and this bankfaces littlecompetition. Education and training. The evidence indicates that firms are more likely to train when they employ an educated and skilled workforce, invest in R&D and technology, and are relatively large and emphasize quality control methods or have foreign capital participation or and export to foreign markets." Thus there are strong complementarities between trainingand schooling and critical linksbetween firms' training, technology, and competition in product markets. Employers' decisions to train and the productivity outcomes of training depend on the stock of education and technical skills that individuals bring to the labor market. It is more cost-effective to impart training to workers who are adept at learning. Even public vocational training increases in effectiveness at higher education levels: as education has improved, vocational training institutes in many Latin American countries have reverted to their original mission of training for productive employment instead of implementing remedial adult education programs.l2 More rapid technical change necessitates a stronger education base. Given the complementarities between formal education and more technological dynamic firms, trainingis unlikelyto play acentralroleinupgrading workers inthe small-firmsector. Overview of strategies for income growth amongst the moderate urban poor. The government's new emphasis on the moderate poor represents an important complement to policies to the extreme poor and the diagnosis of the situation is sound. This will imply inparticular a firm-based approach, includingthe self-employed, and within this policies to reduce the costs of formalization while expanding services to the firms. There i s already an array of support policies and new initiatives, but it i s unclear that these are effectively tackling the costs of formalization. Both a deepening of the strategy and careful ongoing evaluation of ongoing initiatives are needed. D. Income growthfor ruralhouseholdsinextreme poverty The core of government strategy for the extreme rural poor has been on social development and there i s relatively weak coordination with productive programs. Yet the extreme poor needmore than basic human capacities, that are largely directed at the next generation under education and health programs, Oportun~dadesand welfare transfers, also provided by O p o ~ t ~ n i ~ ~ dTheye s . also need improved income generation opportunities. There are some programs specifically under CONTIGO that relate to productive activities, especially a range of credit programs and aspectsof ~icrorreg~ones under SEDESOL. However, while valuable, these are probably far from being the most important determinants of incomes of poor households. There i s a case for more extensive and articulated treatment of economic development strategies for the poor, An explicit income generation strategy i s also necessaryfor them andit needs to be specified and differentiated from a strategy for the moderate poor. This requires looking at what the extreme poor do and what options they couldhave. 11DeFerranti et al. (2002). Lopez-Acevedo and Salinas, 2001. 191 Thechallengesof rural income growthfor the extremepoor. Many of the extreme rural poor are unemp~oyed.In the marginal rural communities covered under the initial phase of Progresa the following question was asked to adults: "What did you do duringthe last week?" (ENCASEH 1997). Answers from 60,057 adults indicate that 51.8% of the adults in these communities did not work. l3 The answer is partially cultural as women tend to respond that they do not work, discounting work in the family's farm and underestimating effective work. Yet, it is strikingthat so many of the poor do not work. l4is It also strikingthat among school-ageyoung adults who do not go to school, many do not work. Inthe 15 to 18 age group, an average 13% of the males and56% of the females do not go to school anddo not work (Table 5.9). This is puzzling as cash transfers to those who go to school are meant to compensate for the opportunity cost of working. In fact, these young adults who qualified for Progresa transfers do not have paying occupations nor do they consider that they are working for a family enterprise. Table 5.9. School attendance and work participation among 15 to 18years old in marginal ruralcommunities Percentages by genderand age School, School Work, Nowork, no work andwork no school no school Males byage 15 40.4 8.8 34.8 13.4 16 25.8 6.9 49.1 13.6 17 14.9 4.7 63 12.3 18 10.9 3.4 67.2 11.4 Females byage 15 39.2 2.5 11.7 45 16 22.6 1.9 16.7 56.2 17 16.5 1.7 19.9 57.8 18 8.6 2.2 19.3 66.2 Source: Progresadata "WoMincludes anyamount of work, unpaidwork infamily production, andcases where the personholds a job eventhough she did not work this particularweek. Rural non-farm work is of rising importance. Amongst the extreme poor some 72% of heads of households say their primary occupation i s in agriculture, a proportionthat falls to 47% amongst the moderate poor (Table 3.9 and 3.10 in Chapter 3). However, this overstates the importance of agriculture as a source of employment and income. Most households -and many individuals- receive income from multiple sources, including agriculture, rural non-farm work and transfers plus, for the majority of households, the imputedrental income from living in their own home. As Figure 5.8 shows, the share of agricultural income fell sharply between 1992 and 2002. While the share of agriculture is larger for poorer households it has fallen for all groups. Even the bottom quintile has experienced a fall from about a 70 to below 50% shareof income from agriculture intotal income. l3Araujo, Maria Caridad, Mainde Janvry, and Elisabeth Sadoulet, 2001. l4 Notethat unemploymentas measuredby the labor force surveys is lower than this. 192 For the extreme poor, the fall in agricultural income has been concentrated in farm production -in crops, fishing, forestry and livestock, including consumption of own- produced goods (Table 5.10). The share of agricultural labor income has risen slightly to over 20%. Increases for the poor were spread over rural-non farm work and enterprise income, private remittances and transfers and a large increase in public transfers - dominated by O p o ~ ~ and Procumpo.~ The even more dramatic rise of rural non- ~ ~ ~ u e ~ farm income amongst the non-poor is primarily due to a surge in rural non-farm work, especially inrelatively well-payed activities. Figure 5.8 The share of farmincome intotal ruralincomehas been falling, even for the extreme poor Income share; by consumption quintile, 1992-2000 P i l Source:WB staffcalculations fromENIGH, 1992-2002 193 Table 5.10 The compositionof ruralincomein1992and2002 Poora Non-Doora 1992 2002 1992 2002 Agricul~ur~lincome Crop income 13.2 7.9 18.9 4.6 Other agricultureb 8.8 3.0 12.5 4.6 Self-consumption 16.1 5.9 7.2 2.6 Agricultural labor 19.6 21.9 10.8 9.6 Subtotal 57.7 38.7 49.4 21.4 N o ~ - ~ g r i c u ~ tincome u r ~ l Non-farmlabor 15.9 17.2 21.3 39.2 Enterpriseincome 4.8 6.8 8.8 5.5 Remittances 1.3 3.8 3.0 6.2 Other private transfers 3.9 5.5 4.2 4.2 Public transfers 0.4 16.1 0.2 4.5 Other 15.9 11.8 13.1 18.8 Subtotal 42.2 61.2 50.6 78.4 Notes: (a) interms of the food-basedpoverty line (or extreme poor). (b) other agriculture includes fishery, forestry and livestock. Source: WB staff calculations from ENIGH, 1992and 2002 This information from the national surveys is supported by results from poor areas covered inthe ENCEWProgresa surveys of the late 1990's. Inthe ejido, individuals with no secondary education work mainly on the agricultural labor market and inconstruction (De Janvry and Sadoulet, 2001). By contrast, those with secondary education work more on the non-agricultural labor market and in non-agricultural self-employment. The low skill, easy entry, labor market offered by agriculture and construction i s thus important as a source of income for poorest rural individuals. Agriculture remains an important source of employment for these individuals, as well as construction which is largely linked to agriculture and to remittances through income expenditures. Public work employment could potentially also be an important source of income for them. Yet, in the Progresa communities surveyed, only 1.5% of the households have a member participating to the PET (Temporary EmploymentProgram). Amongst non-agricultural activities some are linked to agriculture, particularly in services, and others are linked to regional manufacturing activity located in urban clusters. As a consequence, while individualhousehold members tend to specialize, rural households typically have highly diversified sources of income (so called pluriactivity). In the ejido sector, that is primarily composed of a farming population, 55% of total household income comes from off-farm activities (De Janvry and Sadoulet, 2001). Inthe ~ ~ o r ~ ucommunities,e 63% of the adults work as agricultural workers; 24% as ~ i d ~ ~ entrepreneurs in their own farms or micro enterprises, and 14% work as non-agricultural workers or employees. For a given availability of non-farm income opportunities the major determinants of who i s able to capture these opportunities are secondaryeducation, proximity to urbancenters andmembership inmigration networks. 194 Education has a low return in ~ a r g ~ n rural co~~unities. a l Calculation of expected lifetime earnings by educational levels in marginal rural communities shows the low returns to education offered in that context (Figure 5.9). In all three main occupations (agricultural wage employment, non-agricultural wage employment and self- employment), education buys nothing in these communities -at least in the late 1990's at the time of the survey. Migration yields substantially higher incomes for all education levels with significant returns to secondary education and very high returns to tertiary education -though the latter may also reflect selection biases or unobservedeffects for the relatively few in the rural community who really make it, whether through talent or connections to urban or international networks. Figure5.9 Returnsto educationfor childrenfrommarginalruralcommunities / 1Lifetimeearnings (pesodmonth) Migration 2000 1500 1000 500 Y . Primary Secondary 1 Secondary2 Secondary3 Higherthan secondary3 Source:Sadoulet, Finan and De Janvry 2001. There i s also significant heterogeneity across marginal rural communities. Some offer more opportunities than others. In Figure 5.10, the very poor communities covered by Progresa inthe late 1990's are classified into two groups with those with a more andless favorable economic context, especially in terms of infrastructure and proximity to markets. Human capital is measured in uneducated male adult equivalent using the marginal income effect of education as a weight inconstructing the index. The index can thus increase with the number of adults in the household and with their educational levels. The vertical axis measures the predicted poverty rate in the community from the econometric analysis of the patterns. This shows that human capital is important in reducing the incidence of poverty in a community but that the quality of the economic context is also very important in helping value human capital investments and reduce poverty. This is a vivid illustration of the complementarity between human investments and factors affecting economic context, such as infrastructure, and connectivity to markets. 195 Figure5.10. Roleof assets (human capital on horizontal axis) and economiccontext (vertical shift) on the incidenceof poverty in506 Mexicanrural communities Progresa data for 20,895 households) o Unfavorable context A Favorablecontext 0 0 0 # A A A 0 0I I I I 10 20 30 Humancapitalinuneducatedmaleequivalent mrcc: DeJanvry and Sadoulet (2003). Given the importance of off-farm incomes as an instrument for poverty reduction, it i s useful to analyze the role of geographical location. Looking at the growth inemployment insemi-urban andruralmunicipalities inthe 1990-2000period(Figure 5.11) shows that: (i) Proximity to a cluster of industrial employment is very important for growth in industrial employment in semi-urban and rural municipalities. Employment growth that is distant from a cluster (beyond 100km) i s rather low with a level of "autonomous" employment growth of some one percent per annum. (ii) Growth in services employment is also affected by proximity to an industrial or services cluster, but it i s less dependent on proximity to a cluster than industrial employment. 44Autonomous"employment remains relatively high at almost five percent per annum, further away than 100 km from a cluster. The main determinants of growth in servicesemployment are: e a highgrowth potentialinagriculture (cultivationof dynamic crops). 8 a highlevelof education. e highconnectednesswith surrounding municipalities (infrastructure). 196 Figure 5.11 Role of geographical location relative to a cluster for growth in manufacturingand servicesemployment in.semi-urban and ruralmunicipalities -~ ~" ". _." ". ~ ~ _____SWJ#C@S~ a ~ ~ ~ n ~ Source: Araujo, De Janvry and Sadoulet(2003). Current income-generatingprojects in rural areas. There in fact exists a multiplicity of programs to help residents of rural communities engage in productive projects. These include in particular programs under SAGARPA (notably Alianza para el Campo and Procampo), SRA and SEMARNAT. A list of some of the major programs is given in Table 5.11. Apart from Procampo, most of these programs are primarily oriented toward commercial agriculture and are thus of only minor direct significance for the poor. As noted above, SEDESOL has its own programs with a more explicit orientation to the poor. These include Crddito a la Palabra, Primer Paso Productivo (belongs to FONAES, Ministry of Economy since 200l), de Financiamiento, Acompa~amien~ola Fondos a Fomacicin Em~resarial(AFE), Capital de Riesgo (belongs to FONAES, Ministry of Economy since 2001), FONART (Fondo Nacional para el Foment0 de las Artesanias), Proyectos Productivos para Mujeres, and Microrregiones. INDESOL also supports productive projects for the poor through civil society organizations. Yet these programs are only weakly coordinated with the other and more important productive programs listed in Table 5.11. This i s in part because they depend on secretariats other that the three mainparticipants to the CONTIGO strategy. 197 Table 5.11. Programsinsupportof incomesinruralareas PROGRAM PROGRAMDESIGN AND BENEFICIARIES QUALITATIVE COMMENTON INCIDENCEAND EVALUATION Procampo Subsidy consisting of a lump sumper ha. paid Preliminaryanalysis shows to farmers according to the number of has significantimpactonthe income cultivated inthe baseperiod. The payment i s of subsistence farmers, favoring o f the order of 80 USDper ha. and per season. notonly increasedconsumption but also investment. The programprovides income support to 2.8 Largerfarmersbenefit million producers, of which 2.3 millionare proportionallymore. The subsidy benefitsownersas farmers on ejidos. In2002, ithad a planned well as tenants. expenditure of 12.4billionpesos (1.2 per cent The money transfers received o f public programmable spending). throughthe programhavereached up to 40 per cent of household incomefor low-incomefamilies. Alianzapara el Campo Farmmodernizationprogram. Collectionof The mediumand largefarmers subprogramswhose purposeis to promotethe benefitthe most fromAlianza, as technologicalimprovementof farms. reportedby FA0and SAGAR' s evaluation(OECD, 2003). In2002,itprovided8.7 billionpesosto 4.3 million Positiveimpactof the programon producersfor support investment. farmtechnology andproductivity Only 11per centof ejidatarios participatedinthe parameters. programin2000. ACERCA (Program de Contingentpricesupport subsidy to grain Favorsthe betteroff farmers since Apoyos a la producers. it targetsregions where agriculture Comercializacidn) To reducemarketdistortions, itcompensates i s highly commercial, but further producersfor the gap betweenapreviously analysis is required. establishedtarget priceandthe marketone. PET(Program de Incomesupport programthat provideswage Itbenefitspoorcommunities. Empleo TemporaE) employment inmarginalruralzones to buildor improveproductiveandsocialinfrastructure. In2000, over 1milliontemporaryjobs were created. Since then the number hasbeendecreasing to about900,000 in2002. PIASRE(Programa Programto prevent andcope with emergency Integral deAgricultura situations due to recurrentcrop failures. Itoperates Sostenibley inregionswhere agriculturalrisksare particularly Reconversidn high, promotingthe sustainableuse of natural Productiva en Zonas de resourcesinthose regions. Siniestralidad Recurrente) FAPPA(Fondo Para el Supportsproductiveprojectsfor dwellersof social Apoyo a Proyectos sector villages (ejidos andcommunities)who are Productivos de las not communityiejidomembersand ownnoland. Organizaciones Agrarias) PROMUSAG Supportsthe economic integrationof women in (Programa de la Mujer ejidos andcommunitiesthroughthe financingof en el Sector Agrario) productiveprojectsfor womengroups. PROGRAM PROGRAMDESIGN AND BENEFICIARIES QUALITATIVE COMMENTON INCIDENCEAND EVALUATION 198 PROCEDE(Programa Land-titling program for the social sector which has Ithasincreasedhouseholdwelfare de Certificacidn de the purposeof providing tenure security and by allowing participationinoff- Derechos Ejidales y operates ina voluntary basis. The purpose was to farmlabor markets. However, Titulacidnde Solares give farmers property titles that could beused as there has not yet beenmuch Urbanos) collateralfor loans, so as to facilitatetheir access to increaseincredit (OECD, 2003). credit. By the endof the 1990's this programhadissued certificates to more than 3 millionhouseholds. PRODEFOR(Programa Pilot programto improve forest management, deDesarrollo Forestal) promoteforest development anddiversify income sources of forest dependent populations by means of bothtechnical and financial assistance. PRODEPLAN Supports the development of commercial forest (Programde plantations with a subsidyof 65% of the Plantaciones installation costs for the initial 7 years. Comerciales Forestales) FIRCO (Fideicomisode Temporary employment programs, programs for Riesgo Compartido) the productivetransformation agriculture, renewableenergy programsandprogramsfor the management of small watersheds. FIFONAFE Trust fund that administersejidos andcommunities (FideicomisoFond0 funds generatedinparticularthrough indemnities Nacional de Foment0 from expropriationof ejido lands. The fund Ejidal) sponsors loan programsinsupport of female rural entrepreneurs and young 16 to 24 ejido farmers and training programs for landreconversion. FIRA(Fideicomisos Programsof access to financial services. OECD considersthat ithas constituidosen Relacidn preventedthe developmentof a con laAgricultura) well-functioning rural financial market. OpcionesProducrivas Grant supportsto the populationinpoverty inorder to promote development andthe formation of productive assets or working capital, promote sustainableproduction andthe diversification of productive activities andof self-employment. Source:OECD Economic Surveys: Mexico 2003; Mix, DeJanvry, and Sadoulet2003; SEDESOL. This preliminary review of conditions raises a number of issuesfor strategies for income- generation in rural areas. It i s clear that an education and transfer-oriented strategy - however important- i s not enough. In the more remote areas returns to education are very low. Creating the framework for income growth will be central. However, many existing government programs for income growth are oriented toward commercial agriculture. B y contrast,the rural poor will require attention both to the special problems of production in poor areas and amongst small farmers and attention to the growing importance of rural non-farm work. Taking such an integrated approach i s important in light of the non-viable character of most small farms because of their small size andor poor agroecological conditions. More generally there are problems of lack of private investment in rural areas because of low savings and a malfunctioning rural finance system. One promising approach involves geographic or territorially oriented strategies -these are discussed next. Finally, it should be mentioned that other work indicates that land reform remains an important issue for the poor in many parts of Mexico (World 199 Bank 2001b, Finan, Sadoulet and De Janvry 2002). In some areas the challenge is in facilitating the individualizationof tenurial systemsunder the 1992landreform. Inothers it is a case of supporting indigenous groups who prefer to keep communal systems (see Box 5.1 below). Issues of landconflicts are important inmany areas. However, resolving land issues will only yield benefits if complemented by measures to strengthen the competitiveness of landbeneficiaries with close linksbetweentitling, technical assistance to manage landindependently and structural measures to raiseproductivity. A geogru~hica~lyorientedapproach Territorial approaches to rural development are receiving growing attention as a component of "new" rural development strategies or Nueva R~ralidad.`~ Many countries have put inplace elements of a territorial approach. Within Mexico there are a numberof initiatives that would fall within this general approach including the approach under the Ley de Desarrollo Rural Sustentuble under SAGARPA and Microrregiones under SEDESOL (Caballero, 2004). We introduce the issues with a discussion of Microrregiones, by way of illustration, and then discuss issues in designing a more integratedapproach. Microrregiones program is an innovative response to the challenges of integrated rural development. The Microrregiones program focuses on marginal municipalities (numbering 1,334). Within these municipalities, localities with potential to serve as Strategic Community Centers (CEC) are identified (numbering263). These are localities which have the capacity of providing increased services in infrastructure, social services, and productive capacity to dispersed localities in their spheres of influence. It is consequently a local development strategy that seeks to both increase the provision of infrastructure, services and projects in poor municipalities and at the same time avoid dispersion of investments toward areas where population i s excessively dispersed and economic potential insufficient. In the long run, populations in the neighborhood localities may well move toward these regional centers. The Microrregiones program in principle tries to achieve a balance between the decentralization of infrastructure, social services and projects toward poor localities, and an effort to seek economies of scale around the CEC's with greatest potential in the municipality.'6 It thus tries to avoid a common critique of efforts at reducing poverty in a decentralized fashion, namely that dispersion of assistancecan be inefficient. Microrregiones program faces several challenges both conceptually and operationally. Microrregiones is not yet an integrated territorial program but a combination of programslinvestments from different ministries (Caballero, 2003). There is a particular need to coordinate the actions under the Microrregiones program and those under the SAGARPA'S territorial programs, including Ley de Desarrollo Rural Sustentable (see Dimension2 below). ''SchejtmanandBerdeguk. 2002. l6See CONTIGO2003. 200 With respectto the Microrregiones strategy, the following are a few initial observations. With 35% of national poverty and 53% of national extremepoverty located inrural areas and high levels of congestion in the major cities there i s a logic to pursuing a decentralized approach to poverty reduction that seeks to offer employment and investment opportunities in the regions (but not necessarily in the exact localities) where the rural poor are located. This is especially important for poverty associated with indigenous populations for whom territorial attachment is particularly high. As noted above, a territorial approach to ruraldevelopment offers the possibility of going beyond the traditional sectoral focus given to rural development, where agriculture was the main instrument focusing instead in an integral fashion on all potential sources of income (Caballero, 2004). We conclude from this that for a micro region to provide non-agricultural employment opportunities its belonging to a larger region with linkages to urban growth centers, favorable agriculture and connectednessto other municipalities are important. The micro- regional approach to poverty reduction could thus be seen as part of a two-pronged territorial strategy: Dimension 1: Regional development to generate local e~ploymentand investment opportunities. Regional development involves both an`increase inthe productive capacity of a region basedon its comparative advantages and institutional development insupport of expansion of its productive capacity. Regional development can be induced. Organizations for coordination, planning and promotion are essential. A region needs to define its comparative advantages, seek recognition for its specialties (quality, labeling, culture), seek economies of scale and linkages aroundits main economic activities (e.g., following CIRAD's Localized Agricultural Systems (SYAL`s) approach), promote its products, allocate public investments to attract private investment (e.g., inmaquilas), etc. Successful regional growth can thus generate local opportunities that can be seized for effective rural development (read poverty reduction). Under the new Ley de Desarrollo Rural Sustentable this regional effort could be achieved at the level of "districts" if functionally defined for this purpose(a definition which is still inprogress). Dimension 2: Rural development strategies to help the poor participate to the opportunities ofered by regional development. SEDESOL's microrregiones approach can usefullybe seen inthis perspective as well as the role of the Consejos~unicip~lesde Desarrollo Rural. The objective of these local efforts is to help the poor by increasing their asset endowments (ampliacidn de capacidades, formacidn de patrimonio), improving the quality of the context where they use their assets (generacidn de oportunidades)and providing them with protectionto allow them to take entrepreneurial risks (provisidn de proteccidn), so they can take advantage of the supply of employment and investment opportunities offeredby successful regional development. The microrregiones approach i s an innovative and promising concept but it i s unclear how it can be effective as currently designed. It should not be played in isolation, either in terms of links to other programs, or in terms of links between the poor and the non- 201 poor. The worst that can be done for income generation is to isolate the poor from the non-poor. The non-poor inevitably must be the mainagents inthe promotion of regional development. This requires well-orchestrated regional development with strong institutional development in support of the promotion of a regional dynamics. The Microrregiones and Ley de Desarrollo Rural ~ustentublestrategy, with channeling of resourcesto local areas, has the potential to be an effective instrument to attract regional employment and investment opportunities to these locations and their areas of influence. It is thus essential that these be fully integrated within broader strategies of regional development. The role of migration. Migration i s highly heterogeneous in terms of residence, return flows and migrants profile (UC San Diego Institute for Mexico-United States Studies, 2000). Migrants also differ in terms of their legal status. Although difficult to estimate precisely, the numbers of non-authorized Mexican migrants -who have either entered the UnitedStates without permission or have overstayed the term of their visa- are very large. Migrants profile i s changing from being men, with little schooling, working in agriculture and with low earnings, to being more balanced between genders and better educated (Chiquiar and Hanson, 2002). Over time however, although there is no single comprehensive and reliable source to analyze the profile of Mexican migrants, data from the various sources including CONAPO and the UC San Diego Institute for Mexico- United States Studies (2000) suggest that the characteristics of migrants are becoming increasinglydiverse. Remittances are a major reason for migration to the United States, especially for temporary migrants. As discussedinChapter 3, remittances sent by migrants working in the United States have increasedfrom US$9billion in 2001 to almost 10billion in2002, and over 13 billion in 2003. A recent study from the Inter-American Dialogue (2003) concluded that appropriate government policies could substantially increase the already considerable remittance flows -both for the people receiving them and for national development. Policies could further reduce the costs of sending remittances from the UnitedStates, particularly by expanding competition among the institutions that transfer the money. Inaddition, every effort should be made to encourage remittance senders and recipients to make use of banks and other financial institutions to exploit the full range of economic opportunities they offer. A recent study by the IADB (2004) concluded that remittances are almost entirely usedby migrants families for consumption expenditures. The recently created BANSEFIcould play a valuable role in improving the access and lowering the cost of remittance for the rural poor. Another initiative is the 3x1 program. This program supports citizen's initiatives in poor and high migrant localities. This program is funded with resources from the three government levels and from migrants (residents abroad). The projects are developed in the areas of basic social infrastructure, services, socio-productive projects, natural resources conservation, health and education, environmental restructuring, agricultural infrastructure, housing and urban improvement, highway and road infrastructure, and recreational projects. The program matches the municipal resources increasing the number of public works activities that can be performed and promotes citizen's participation. The program benefited over a million 202 people in 2003. Evaluations have indicated that even though the project does not benefit the poorest localities the investments made in this program have had positive externality effects on highpoor regions. Growth-related projects fur indigeno~sgroups. There is a particular agenda around income growth of rural indigenous groups (Box 5.1). This is both because these groups suffer high levels of poverty and also because of distinct preferences and practices -for example in attitudes to the natural world and a preference for collective modes of organization. There are promising models of indigenous development in the productive sector in Mexico. Where an external agent has directed or assisted the initial process there are common characteristics including(a) initialseed capital (not loans) by a grass- roots development agency a large dose of technical assistance, but provided with a view to maintain an exit strategy; and (c) building upon local models of intervention and a local socio-cultural vision of the goals andobjectives. Box 5.1 Examplesof income generation programs amongst indigenousgroups Communalforestry. The ejido and indigenouscommunity forestry enterprisesare an interesting example of programsthat have both suffered fromthe center-peripheryrelationship and, ina growing numberof cases, have resolved it in favor of local groups. Due to the particularities of the ejido and comunidad land tenure systems, between 70-85 percent of the mexican forests are within the boundaries of 7000 to 9000 ejidos and comunidades. Historically, from the past century into the early 1970's, the state allocated rights to timber extraction to capital-intensive initiatives, both Mexican and foreign, and ejiditarios and comuneros served as low-cost labor for extraction. Processing was concentrated in the urban centers preventing strengtheningof labor movementsbetweenextracting and processinglaborersand allowing extraction from different peripheral sites when old sites were no longer economic to exploit. The Union of Forest Ejidos of the MayanZone (OEFP) inYucatfin includes 18 ejidos with 2,791 ejidatario members. In Oaxaca, where 90 percent of the forest estate is within ejido and comunidadboundaries, and 45 percent of it is managed commercially, there are 54 forest ejidos and comuni~deswhich harvest their own timber and may have a processing mill; 30 are grouped in unions which have an industrial capacity and a formal enterprise. San Juan Nuevo, Parangaricutiro in Michoactin has an advancedforest industry and is actively training neighboring ejidos in technical and administrative skills. A movement has begunin Chiapas to create forest enterprises along the models of the mayan experience. A number of Tarahumara and Tepehuan-populatedejidos in Chihuahuaand Durango have forest enterprisesalthough these are more problematic due to the degraded state of their forests and the traditional domination of the marketing by non-indigenoustraders or ejido members. Coffee Cooperatives in Chiapas and Oaxaca. Over the past 15 years, thirty affiliated organizations in Oaxaca have managed an umbrella organization for marketing, credit, warehousing and processing of coffee. The organization Coordinadora Estatul de Productores de Cafe (CEPCO) was founded in 1989. This processhas fosteredparallel organizationssuch as UCIZONIand UCIRI(zapotecos and mixes in the Ixtmus of Tehuantepec) and the UCI (southern zapotecos). This cooperative has been successful in organizing market niches in the European coffee market capitalizing on its social identity. These organizations have benefited from NGO support and some financing to their members from coffee credit schemes and more recently Procampo and the Alianza para el Campo. To address the problem of fluctuating prices and reach higher-value markets, a growing amount of the coffee produced is organic. Organic coffee production is an increasingly important product in Chiapas as well. Unicin de 10s Ejidos de la Selva (LaSelva) is a tojolobal and tzotzil organization that involves 57 communities and 1,304 families in the area of Las Margaritas near the Montes Azules Reserve. About 1,000 hectares have been put into organic production with the objective of certification when the land has completed four years of strictly organic inputs. 203 Urban A ~ a p t a t i ~and Employme~tin the Yucatdn. The maya in the Yucathn peninsulahave adaptedto a n diverseeconomy. Traditional farmers maintaintraditional agricultural practicesinsmall communitiesinthe historical settlements and large numbers of farmers spend significant periods of time working in land preparation and harvesting activities in the cash crop areas of the peninsula and elsewhere in Central Mexico. Urbanization has createdjobs in the service sector where indigenous people find employment as domestics or as unskilled labor. Tourism has createdjobs in the resort areas as well as created a marketfor small-scalecommerceand sale of handicrafts. Becausethe process is so recent and the urbanregionshave not beenstudied, it is difficult to estimatethe population which considersitself indigenous-even when no longer speaks maya or their native language. What is known at the local level is that there are neighborhoods in the urban centers of the Yucatan that are recreating indigenous community lifestyles, adapted to the nature of urban employment and recreating religious and social events that maintain an indigenous affiliation and identity. This is a new frontier for indigenous peoples in Mexico which needs study to understandthe trends, potentials and problems. E. Conclusionsandissuesfor futurework There has been much weaker longer-run progress with respect to income poverty than social dimensions of well-being in Mexico. Future income gains amongst the poor will depend on both overall growth and the pattern of growth, and especially the pattern of job-creation. Scenarios show that both growth and inequality matter for poverty dynamics. Slow growth, of 2% per annum between 2002 and 2015, would have little impact on poverty if inequality does not fall or reverts to the higher levels prevailing in 1998-2000. But if inequality fell by lo%, extreme poverty would fall to 8% by 2015. Growth of 5% per annumwould bringextreme poverty rates to 6% by 2015 if inequality reverts to 1998-2000levels and to 2% if inequality fell by 10%.17 The core of any job-creating strategy lies in the environment for private investment. In other words, a poverty reduction strategy i s inseparable from the design of a competitiveness and growth strategy. This relates to the overall competitiveness agenda which i s currently under public scrutiny in light of the Mexico's poor international performance relative to competitors such as China, especially after the initial gains from NAFTA were realized. This is a complex agenda in which international evidence suggests that macro stability, infrastructure provision, tackling of protections of private sector andrangeof measures to improve logistics are all important. Economy-wide measures to strengthen competitiveness will need to be complementedby strategies specifically oriented toward income growth of the poor -both the extreme poor (mostly in rural areas) and the moderate poor (mostly in urban areas). Current government strategy to the extreme poor i s much more developed inthe social arena than the productive arena; this is a priority for the future. With respect to the moderate poor, the government has recently formulated a new strategy termed Impulso. This i s basedon sound diagnosis but there i s an important agenda in policy design and execution. With respect to the rural poor there have been large changes occurring with a relative fall in agriculture and rise in rural non-farm activities. Much current support i s for commercial farmers. A very different approach will be required for income growth of the rural poor. See scenarios in SectionA of this chapter. 204 A promising approach is that of the Nueva R u r a l i that involves territorial strategies, ~ ~ integrated approaches and local participation. Both Ley de Rural Desarrollo ~us~en~able and Microrregiones fall broadly within this approach but there are major questions over whether the institutionaldesign will be adequateto realize their objectives. The issues around income growth inurban and rural areas will need considerable further investigation on both conditions and current policy efforts. This will be a major focus of future analytical work. There i s also a broader agenda of learning about the processes behind observed effects and for understanding the role of local context. Both are key for the improvement of the design of interventions. Chapter 6 turns to questions of monitoring andevaluation to the Mexican case. 205 June 2004 Poverty inMexico An Assessmentof Conditions, Trends and Government Strategy Chapter 6. The Role and Design of MonitoringandEvaluation A key challenge for effective public administration is the establishmentof a reliable and effective monitoring and evaluation (M&E) system. The mexican state has a long tradition of implementing a variety of programs, including large-scale social programs that aim to relieve poverty through improving the coverage and quality of public education and health as well as the provision of basic services. However, as in most developing countries these programs have rarely been effectively monitored and evaluated. As a result, opportunities have been lost to reform individual programs and policies as well as to allocate resources across the portfolio of social programs basedon their relative performance. This lack of information regarding the impact of social programs and the extent to which they achieve their proposed objectives has posed dilemmas for past administrations aiming to manage such programs efficiently, and has limited the government's ability to channel resourcesinto the most effective programs. Inrecent years, the government of Mexico has taken some important steps forward to prioritize and improve both the monitoring and evaluation of its social programs. ~ u ~ i t o risi ~the continuous tracking inputs, outputs, and often outcomes, sometimes g relative to targets. Evaluations involve an assessment of performance. Impact evaluations address causality by measuring outcomes and attributing them to a specific intervention by estimating the counterfactual state of what would have occurred inthe absence of the intervention (see Box 6.2). In the mid 1990's, the Progress' conditional cash transfer program was developed incorporating the monitoring of input, output and outcome indicators, as well as a rigorous evaluation of the program's impact. Inthe mid-1990's the federal administration instituted a system of performance targets for all public sector agencies as the cornerstone of an effort to improve public sector performance, accountability and transparency. In 1999, Congress introduced a mandate calling for the external evaluation of all programs that use government funding. In,2003, the Social Development Law was enacted and calls for the establishment of an independent council to review and evaluate Mexico's social programs.2 These efforts have been part of a broader attempt to reform public sector institutions which has emerged region-wide as an In March 2002, Progresa changed its name to O p o ~ u n i ~ eand introduced several changes to its s objectives and operational features, including coverage expansion to urban areas. However, this chapter focuses on the designand evaluation of the original programandhenceuses its previous title. The Law establishes that the evaluation of the Social Development Policy will be the responsibility of Evaluation National Council. The Council can carry out its duties by itself or through one or more independent agencies. Its objective is to review periodically the achievement of the social objective of the programs, goals and actions of the Social Development Policy in order to correct, modify, add, reorient or to suspend them totally or partially. Institutions of higher education, of scientific research and nonprofit organizationscan all participate as evaluation agencies. When the evaluations are carried out by an agency other than the Council itself it is still the responsibility of the Council to select the outsideevaluation group through abidding process. 206 important element of the development agenda, This reformagenda includes efforts aimed at implementing civil service reform in federal government, laws and institutions for public access to information, introducing transparency and anti-corruption measures, and improvingmacroeconomic planning,financial managementandtax administration. Notwithstanding the merits of the initiatives describedabove, Mexico has yet to establish a systematic approach to M&E within any of its ministries, much less introduce an integrated government directive for results based management (RBM)3such as those implemented inthe United States or Chile. Even without a national directive, individual ministries and program operators would benefit from taking a more systematic approach to RBMand introducing rigorous impact evaluations of key social programs as was done in the Progresa program. A RBM approach based on the use of reliable, relevant monitoring data could benefit ministries responsible for managing a diverse set of social programs, often with uneven degrees of coverage, performance, targeting and costs. Additionally, the strategic application of impact evaluations could be carried out beyond Progresa to assess the effectiveness of Mexico's mainsocial development (and indeed other) programsthat have highdegrees of national coverage and absorb substantial resource allocations. Impact evaluations could also be used to pilot promising new social programs to inform decisions regarding their expansion basedon an assessment of their effectiveness. The president's Social Cabinet, which is responsible for coordinating development policy, is cornmitted to furthering the government of Mexico's more ambitious long-term goal of incorporating M&E practices throughout all its ministries as a cornerstone for effective management. SEDESOL4has launched an initiative to improve the M&E of the social programs under its administration. SEDESOL's goal is to create an M&E system that provides accountability through both performance monitoring and objective assessments of social policies and programs based on impact evaluations. This system i s intended to form a continuous feedback loop for refining the design and mix of social programs in Mexico.' It i s hoped that the SEDESOL system will become a model for other ministries in moving towards RBM. These efforts are central to the government's social development strategy CONTIGO, since they will provide a mechanism for improving the performance of poverty reduction programs prioritizing the allocation of public resources. The goal of SEDESOL's M&E system will be focused on principles of RBM.Figure 6.1 below representsthe general logical framework of RBM. RBMusesperformancemeasurement-the measurementon a regular basis of the results (outcomes) and efficiency of services or programs- to inform administrative decisions, particularly reforms to programs and budgetallocations. SEDESOL is charged with guiding and coordinating government efforts regarding social and human development. SEDESOL implements more than 20 social programs covering a range of interventions from workfare to food distribution to conditional cashtransfer programs. A World Bank financed InstitutionalDevelopment Fund(IDF) grant will support SEDESOL's attemptto combine both the monitoring of performance indicators and the impact evaluation of programs and in so doing, assist SEDESOL's effort of being more responsiveto the congressionalmandate. 207 Input Activities output -J, Outcome Impact Indicators j+ Indicators Indicators 4Indicators -b Indicators RBM measures progress toward short-term results (outputs) and longer-term results (outcomes) that can provide feedback to improve performance. A full RBM system also includes the strategic application of impact evaluations to assess whether the outcomes observed are indeed attributable to the program or policy that has been implemented. The integration of both monitoring and evaluation activities i s crucial. Reliable, timely monitoring systems are neededfor effective managementand are a key tool usedby high performance organizations. However, a monitoring systems cannot assess impacts since they provide no estimate of the counterfactual needed to address whether the program was indeed responsible for the observed outcomes. This i s what impact evaluations do. Furthermore, the credibility of impact evaluations i s enhanced when they are carried out by impartial, objective evaluators usingrigorous methodologies. The creation and implementation of a results-basedM&E system does not come without challenges. The system first requires the identificationof input, activity, output, outcome and impact indicators that are relevant to decision making. Reliable baseline values for output, outcome and impact indicators and realistic targets are also essential, increasing the need for foresight inthe planning of eachevaluation. As a corollary, the systemneeds a reliable information base to continually monitor these indicators and a framework to continually analyze the achievement of related targets. The continual monitoring and analysis of RBM allows for flexibility and provides the basis for needed corrections, increasing its effectiveness as a management strategy. However, these requirements are oftenhamperedby low capacity inM&Eactivities indeveloping countries. This chapter describes the evolution of M&E in Mexico with emphasis on the foundations towards building an RBM system, as reflected in the SEDESOL effort. The chapter reviews the experience with the Progresa impact evaluation and SEDESOL's designs for an RBMsystem inthe context of similar regionalefforts in sections A and B respectively. The international lessonsfor designing andimplementing an RBMsystemi s addressed in sections C and D of this chapter, highlightingissues of evaluation design, coverage and implementation. Section E addresses the challenges of an RBM system. The chapter concludes with some reflections on possible next steps and policy options for improving the M&E of public social programs as part of the operational implementation andrefinement of the CONTIGO strategy. 208 A. Mexico's pioneeringrole inimpact evaluation Mexico has been a pioneer in Latin America regarding the importance placed on the impact evaluation of social programs. As early as 1998, Mexico's influential conditional cash transfer program (CCT) Progresa, incorporated rigorous evaluation methodologies in its design as a key element for informing program design and expansion. Progresa (now Oportunidades) provides monetary transfers to poor families conditional upon investments in human capital, such as sending children to school or bringing them to health centers on a regular basis. The use of demand-side interventions to directly support beneficiaries represents a marked departure from traditional supply-side mechanisms, such as general subsidies and capital investments in schools and hospitals. The program introduced additional design innovations, including both a short-term focus on poverty alleviation through transfers and long-term focus through incentives for investing in human capital; an explicit creation of synergies in education, health and nutrition; and a focus on co-responsibility between the state and its citizens. The interest in assessing the effectiveness of these innovations prompted the incorporation of a robust impact evaluation designinto the initial stages of the program. Progresa went beyond collecting output indicators to focus on measures of impact such as malnutrition impact as measured through the height and weight of children. The main instrument of data collection used was program specific household surveys. Interms of methodology, in order to assess the net impact of the program, Progresa used randomized control designs. In the past, the central problem of impact evaluation had been assessingcausality by isolatingthe net impact of programs (changes in the value of outcome variables among participants compared to what they would have experienced had they not participated). This obstacle of evaluation arises due to the difficulty of , constructing control groups and observing such groups simultaneously with the groups participating in the program. Progresa was able to avoid such difficulties by carefully planning evaluation within the program's original design. The program utilized an experimental evaluation that involved the random assignment of beneficiaries into those who received support (treatment group) and those who met the criteria for support butdid not receive intervention (control group).6 This random categorization of control and treatment groups allows one to attribute differences to the program and not to chance. The early planning of data collection was also essential to Progresa's design, as it allowed for the compilation of baseline data, as well as repeated observations of treatment and control groups from which program impact could be assessed through difference-in-differences estimators. Without such planning and data collection the experimental nature of the evaluation design would not have beenpossible, The impact evaluation of Progresa provided the government with evidence of the program's success in reducing poverty, which aided the subsequent administration in deciding to continue and expand the program under the name Oport~~idades. only Not did its evaluation results influence the program's own future but also the creation of similar CCT programs inmany countries inLatin America and a growing number outside of the region, Many of these programs have included robust impact evaluations as part of Rawlings andRubio, 2003. 209 their program design, a testimony to the highstandards set by Progresu for the evaluation of social programs. Programs in Honduras and Nicaragua have experimental designs based on random assignment while the Colombian and brazilian programs have quasi- experimental designs but still rely on baseline and follow-up surveys of treatment and comparison group^.^ There can be resistance to the use of experimental evaluations with random assignment when governments feel they cannot exclude some groups from participation. When randomization i s not feasible, a second best is use of a quasi-experimental design that involves the generation of a comparison group through quasi-experimental approaches meant to approximate the effect of random control group selection, for example through matching methods or reflexive comparisons. Matching i s a statistical technique used to pair non-program participants with program participants who share characteristics. In a reflexive comparison the counterfactual i s constructed by assessing indicators of participants before participation inthe program. Thus, programparticipants are compared to themselves before and after the intervention and function as both treatment and comparison group.' However, such before and after approaches are much weaker in explanatory power since they cannot control for influences on outcomes over time that are unrelatedto the program. The Progresu impact evaluation has set high standards for other social program evaluations, both in Mexico and regionally. In addition, Progresu instituted other M&E innovations that, though less well known than the impact evaluation, are good examples of how a comprehensive RBM system can be designed. These elements include constructing a registry of beneficiaries (padrdn cinico de beneficiurios), a survey of beneficiaries perception of SEDESOL's social programs, constructing a ~unuuZ ciududuno(a booklet of information on how to access SEDESOL's social programs), and implementing a comprehensive monitoring system accessible to program managers throughout Mexico that regularly tracks key performance indicators. The objectives of these innovations are in line with the lessons learned from international experience implementingsimilar RBMapproaches. The following section discusses in greater detail lessonslearnt from the Mexicanexperiences with RBM. B. RBM:The MexicanExperiencewithina GlobalTrend In an attempt to manage public institutions and programs more efficiently, governments have begun to move away from traditional forms of management. Instead of solely focusing on the monitoring of inputs there i s a global trend to implement M&E systems, which are instead driven by results and impacts. These approaches are characterized as RBMorperformance measurement. Rawlings and Rubio, 2003. Such a methodology is currently being used in the World Bank`s current study of the evaluation of programsthat support Mexico's micro, small and mediumenterprises (SMEs) with information included in the EmSTYC panelsurvey (Tan andLopez-Acevedo, 2003). 210 Besides installing a performance-based culture and commitment to information management, RBM systems aim to establish the necessary bridges between evaluation and policy formation, political decision-making and resource allocation by linking evidence on the outcomes and impacts of public programs to management decisions, including on resource allocation. Furthermore, the RI3M approach aids institutional and capacity development and facilitates information exchange between beneficiaries, civil society andthe government through the dissemination of evaluation results. Ingeneral, the internationalliterature onRBMhasthe following results. First, according to reports from the OECD, improvedperformance of the public sector is a central factor in maintaining the social and economic welfare of individuals and the competitiveness of the economy. Consequently, performance management i s akey aspect of public sector reforms in many OECD member countries. In developing countries, performance management i s also a key aspect of poverty reduction strategies and social reforms. Second, there are three types of actions that form essential building blocks toward establishing performance-based M&E systems. These include creating an environment that expects andvalues highperformance, designing andexecuting a well-defined system for monitoring and evaluation, and aligning budgets with programs andgoals. Third, there appears to be no one right way to introduce performance management into the many institutions and policy-making activities of government. Often, depending on the presence (or absence) of certain elements, governments try one or more strategies: 1) comprehensive or whole-of-government; 2) sector-specific or 3) customer-focused approaches. In the whole-of-government approach, a number of countries have introduced government-wide strategic plans, performance indicators and annual performance plans and integrated them into annual budget documents (Australia, United States). Other related approaches include putting into annual financial reports program performance indicators that can be audited (Finland, Sweden, United States), or using performance agreements between ministers and heads of government agencies (New Zealand, United Kingdom). Romania at the national level and Argentina at the state level have initiated program-based budgeting strategies, while Malaysia has embraced the total quality management approach focusing on process reengineering and achieving strict quality standards. Total quality management has generally been introduced after reform processes are well underway, but the focus of quality management on customers i s relevant to reformefforts at all stages. Furthermore,total quality management focuses on monitoring results and using information for problem solving and decision making and, therefore, it provides elements for a sound monitoringand evaluation system. Other countries have chosen a sector-specific approach, such as the Philippines, which i s piloting performance M&E as part of their rural development sector performance information and management system. Still others have found it useful to focus on the 211 users or beneficiaries of government services or on one client group such as women or children (e.g., United Kingdom's Citizens' Charters, and in Egypt, the National Council of Women). This strategy includes developing key performance indicators that cut across line ministries with a specific focus on improving those government programs that support a particular population group. While some OECD countries have movedstraightto whole-of-government approachesto introduce performance management, other countries (such as the United States) began with pilot initiatives. By first piloting in a few programs and sectors, governments attempt to create favorable conditions for public sector learning and experimentation before mainstreaming the effort. Other countries have found that encouraging and supporting actors that are committed to results-based management in sectors where a clear reform effort i s underway (e.g. the health sector in Bangladesh, Ghana and the Kyrgyz Republic; the state health and education sectors in Santa Fe, Argentina) allows promising efforts to move forward regardless of whether national commitments have been made to implement a more comprehensive approach. This strategy can also help move forward a national agenda in a program area, rather than waiting for the entire government to embrace performance management. In developed countries, RBM systems have been applied to service delivery agencies such as health and education, to public services, to financial agencies such as treasuries andcentral banks.RBMsystems were adoptedby UKandNew Zealandinthe 198O's, by Canada inthe early 1990's and by USA and many OECD countries inthe late 1990's. A number of donor agencies, both at the multilateral and bilateral levels, are also making use of RBMapproachesinmanagingtheir own programs. The introduction of performance management strategies in developing countries i s still very new. As these governments move forward with their own strategies they are trying new and innovative approaches that are enriching the pool of data and experiences. Matsuda (2003) indicates that there have been efforts throughout Latin American countries since the late 80's to implement RBM's, notably in Costa Rica, Uruguay, Honduras, Chile, Colombia and Bolivia. Box 3 in the Summary and Messages summarizes these international experiences in terms of context, design and implementation basedon a report commissioned by the World Bank from IDEA (2003). Thus far, onethingis clear. There is noone particular strategyor approachthat is best for all countries. First, reform involves multi-year efforts and strategies inevitably evolve over time. Second, each country is unique in its socio-cultural and political context and its views of what i s feasible inperformance management. Inpractice, countries may use a combination of approaches from those sketched in Box 3 and adapt and test experiences from other countries to their own circumstances. International experiences provide lessons for the design (see Section C) and implementation (see Section 0)of an RBM system, in addition to indicating challenges specific to countries such as Mexico seeking to implement an RBMsystem (Section E). 212 C. Internationallessonsfor the designof RBMsystems International experience demonstrates that there is no blueprint for the design of an RBM system and it is essential to shape systems around the political and institutional context. Incentives, leadership and building capacity on the supply and demand side are all important (Box 6.1). Key elementsof design common to successfulcases include: 1. A clear articulation of goals and relatedmeasurableoutcomes, outputs, activities, andinputsbasedon a logicalframework approach. 2. The systematic monitoring of key performance indicators through a Management InformationSystem (MIS). 3. A series of external andinternal evaluations. 4. Institutionalcapacity-building activities. 5. Feedback mechanisms to link M&E results to policy decision-making an ultimatelybudget allocations. 6. The annualpublicationof performance results. Box 6.1 BuildingRBMsystems: lessonslearned The following lessons are base on World Bank work in supporting the establishment of performance management systems in Bangladesh, Egypt, the Kyrgyz Republic, the Philippines and Romania. I t is important to understand the political and institutional context in a given country before moving forward with the actual design of a results-based M&E system. There is no single blueprint for constructing these systems, andthe process is lengthy. Incentivesfor governments to produce information about their own performance are necessary. A sustained source of both internal and external (from the Parliament, civil society, the private sector and others) demand should be encouraged and supported for performance information that puts the government "on notice" that they will need to demonstrate that the policies and programs being implementedare meeting expectedresults. A successfulresults-basedM&Esystemneeds go well beyondhaving a capacityto managedata; it also must have sustained government leadership reviewing whether governmentis delivering effective services to its citizen. Without a strong champion who is willing to take on ownership and agree to the transparency and accessibility of performance information, the M&E system will not be built, andevenif it is, it will not beused. Just as there has to be demand so there has to be supply -capacity within the country to design, implement and usea results-basedsystem. Skills from social research, public management, statistics and data management must be mobilized to contribute to a systematic process of regularly assessing governmentperformance. Source: Jody Zall Kusek and Ray C. Rist, "Building Results-Based Monitoring and Evaluation Systems: AssessingCountry Readiness" World Bank PREM-Note 2004. Several mexican initiatives are in line with these best practice guidelines, including Mexico's Congressional mandate calling for annual performance data; the establishment and monitoringof presidential goals; SEDESOL's registry of programbeneficiaries, and SEDESOL's citizens' manual ( ~ a n ~ uCiududuno) which provides information on Z SEDESOL's functions in addition to providing a medium for citizens' complaints and suggestions. 213 Developing moni~o~ngsystems International experience with R3M reveals the need for a monitoring or information collection system in order to compile periodic measurements of performance indicators. Information on key indicators of inputs, activities and outputs should be frequent (e.g., every month or quarter) to allow for the rapid identification of problems in program implementation. Based on such information, managers are able to make timely adjustments to their programs. In Chile, one of the principal objectives of the public sector reforms titled Program Murco i s the implementation of a management information system within its institutions that includes easy-to-use databases. In New Zealand, departments in charge of government programs must provide a full set of financial statements to their minister andto the Treasury Department every month. These statementsare done on an accrual accounting basis. Unfortunately, information managementand collectionhasbeenespecially challenging to developing countries which are often faced with financial resource constraints and weak institutionalcapacity. This has translated into delays ininformationcollection, unreliable data and inaccessible information -problems that erode the validity of the link between results andpolicy decisions. Designingintegratedevaluations A comprehensive evaluation is defined inthe literature as containingprocessevaluations, cost-benefit studies and impact evaluations. Any single approach to evaluation contains its limitations. Hence, it i s increasingly recommendedthat an evaluation strategy contain a mix of types of evaluations, dependingon the needs of policymakers and that this mix draw on both quantitative and qualitative methods. Inparticular, the question of when it i s appropriate to conduct an impact evaluation i s critical since these often involve more resources, time and technical complexity than other types of evaluations. Impact evaluations that assess whether a program is directly responsible for observed outcomes should be strategically applied to projects that are innovative, replicable, involve substantialresource allocations andhavewell-defined interventions. Finally, the question of who conducts the evaluation i s important. External evaluations have the advantage of providing objective measures of program performance, enhancing their credibility. However, internal evaluations are important management tools that help buildinstitutionalcapacity and give those responsible for the design and implementation of programs greater controlover their work. Integrating monitoring and evaluation Both monitoring and evaluation are critical to effective public management. The challenge often consists of designing and applying the best combination of monitoring and evaluation approaches in support of the production of timely, reliable data useful to program managers, beneficiaries andpolicymakers. 214 Box6.2 Definitions: Monitoring and Evaluation Evaluation Evaluations involve an assessment of performance. Evaluation is the systematic and objective assessment of an on-going or completed project, program or policy, regarding its design, implementation andlor results. The aim is to determine the relevance and fulfillment of objectives, development efficiency, effectiveness, impact andlor sustainability. Evaluations are usually selective, discreet activities. Impact evaluations examine causality -whether a particular intervention is responsible for a particular outcome- through the construction of a counterfactual state estimating what would haveoccurred inthe absence of the intervention. Monitoring Monitoring is a tracking function. It is a continuous activity that uses the systematic collection of data on specified indicators to provide management and the main stakeholders of an ongoing development intervention with indications of the extent of progress inthe achievementof objectivesand use of allocated funds. Monitoring is ongoing, comprehensiveactivity. Good monitoring is essentialto effective management and often involves tracking indicators relative to targets. Both monitoring and evaluation should provide information that is credible and useful, enabling the incorporation of lessons learnedinto the decision-making process of bothrecipientsand donors. Source: Adapted fromthe Development Assistance Committee, OECD Working Party onAid Evaluation. Inthe case of Mexico, Sadoulet andDe Janvry (2003) propose an integrated monitoring and evaluation scheme with the following two goals: a) to serve as tool for improvingthe design and functioning of specific programs and; b) to ensurethe state's accountability to Congress. To achieve these goals, they proposed an integratedmonitoring andevaluation approach with two key components i)interactions between internal and external evaluators, stressingthe importance of beneficiary participation; and ii)a combination of short-term analysis (continual monitoring) with long-term impact evaluation. These two key components can be partially combined into a single processto reinforce the quality of each component as well as to maximize learning, accountability and transparency as part of a results-basedmanagementapproach.' SEDESOL is moving towards an integrated monitoring and evaluation strategy that includes a quantitative, qualitative and institutional foci." Based on a common, established set of social development goals, they are designing a strong performance monitoring system combined with selected evaluations of key programs, including impact evaluations. The quantitative component encompasses both short-term and long- term performance monitoring (external and internal), ex-ante simulations of program impact, internal program feasibility and sustainability studies, and social policy 9 For an application of this proposed monitoring and evaluation approach to the Microrregiones program and Indesol's set of projects, see Sadoulet and De Janvry (2003). loGonzalo Hernandez, SEDESOL "Evaluation and Monitoring SystemMexico" Powerpoint presentation. 215 evaluations. The qualitative focus involves beneficiaries in both monitoring and evaluation, and the institutional focus will concentrate on measuring performance in terms of institutionalcapacity and humandevelopment. The selection of indicators for monitoring andevaluation is an important aspect of system design, Emmanuel Jimenez, during the May conference on Social Policy Best Practices (2003), stressed the importance of selecting indicators that illustrate the program's performance both in the short and long term. The majority of indicators that measure socio-economic and human development outcomes are not revealed in the short term. Nevertheless, short-term variables such as the following often serve as good predictors of long-termeffects as well as aid managers inprogram administration: l1 1. Indicators of administrative efficiency andeffectiveness. 2. Indicators of the rate of useof services andsupportsprovided. 3. Perceptionsof beneficiaries. Furthermore, these indicators should be delivered and accessible not only to various levels of government, but also to members of civil society -those who have incentive to instigate reforms if these indicators are unsatisfactory. Sadoulet and De Janvry (2003) suggest that indicators chosen should capture the objectives and intended outcomes as well as unintended consequences. Beneficiaries participation can be useful in the selection of indicators with these characteristics. Buildingcapacity International experience suggests that the creation of a sustainable RBMsystem involves incorporating specific capacity building activities. In Chile, 180 professionals from different institutions and ministries were trained inprogram evaluations and a network of institutional experts give continual assistance to the system's operators. In Australia, a separatebranch within the Ministry of Finance was created with the specific objective of providing evaluation advice and support and training in accounting and auditing activities.12 The Mexican RBM plans to provide internal and external courses, lectures and seminars, in conjunction with the University of the Americas and CONACYT (Consejo Nacional de Cienciay Tecnologia), covering relevant M&Etopics. Capacity buildingactivities are especially important to developing countries, which often lack a culture of evaluation and whose personnel are unfamiliar with the tenets of RBM. These activities will aid developing nations in meeting the challenges of RBM and enhancing their ability to carry out strategic planning, data collection, monitoring, evaluation and communications. Although international consultants are useful in the training processby offering their perspectives, it has been shown that importing an RBM system without considering the country-specific context i s ineffective. Hence, most l1When monitoring data and impact evaluations share common indicators statistical analysis can be applied to identify variables that serve as good predictors of long-term program impact; these variablesare ideal candidates for key performance indicators. '*IDEAInternational. 2003. 216 training should be handled by national specialists and complemented by international experts inorder to builda sustainableRBMapproach. It is also usefulto create Technical Advisory Panelsfor each evaluation or for evaluations of similar programs with the objective of overseeing the process and incorporating the perspective of the various actors that have a stake in the evaluation process. The panels would be made up of evaluation experts, policy experts, members of Congress, beneficiaries and program operators. Evaluation experts would ensure the technical soundness of evaluation methodologies. Policy experts can advise evaluators on what indicators are most useful and relevant during policy decision-making in order to focus the evaluations and prevent excess spending on less useful information. Members of Congress should be involved as a way of building their understandingof the details and processes involved in carrying out the mission. Finally, it is often useful to include a representative of the programs' participants or operators. These individualshelp to ensure that the evaluation is practical, focusing primarily on policy-relevant indicators. In particular, it helps to ensure that the evaluation has credibility among these groups who are key to the long-termsuccess of policy change. Ensuringfeedback mec~ani~ms To be more than a bureaucratic exercise, RBM systems must include a feedback mechanism to link evaluation results to the budgetary process and policy decision- making. Without such mechanisms, efforts to build RBM systems are squandered. Linking the budgetary process to impact results creates incentives for managers to successfully runtheir programs. InAustralia, eachpiece of policy advice mustbelinkedto 11criteriatiedto the provision of strategic plans and previous impact evaluation of programs. InChile, program funds are allocated through a mechanism called fondo cursable -a Central Fund for Governmental Priorities.l3For new programs or the reformulation and extension of programs, ministries must make applications to this fund presenting in standardformat a budget proposal basedon its strategic plans andpast performance evaluations. InMexico, as stated by the 1999 mandate, yearly results of evaluations must be presented to Congress. Unfortunately, linking impact results to the budgetary process has been an especially challenging aspect of RBMfor developing nations. For impact information to be useful, the timing of evaluations must be sequenced to provide decision-makers with performance information necessaryto make resource allocation decisions. Such timing i s often difficult for developing countries due to data constraints and other factors. During the evaluation of Nicaragua's conditional cash transfer program, the collection of follow- up data was delayed due to coordination issues andhence impact results were postponed. Weak information systems have further consequences. In Mexico, Progresa payment records revealed that 27% of the total eligible populationinthe evaluation sample hadnot received benefits after two years of program operations. These deficiencies were l3IDEAInternational. ibid. 217 undetected by the information system.14 If evaluation results cannot be delivered in sequence, in a timelyfashion, they cannot be used as inputs in decision-making. Hence, such coordination is vital to the success of the RBMand special priority should be given to increasing organizational capacity and systems of data collection. Disseminatingresults International experience with RBM highlights the importance of disseminating performance results in order to promote commitment, participation and accountability. The dissemination of results is also often accompaniedby a public awareness campaign to educate the public in the new orientation in public service management. In Chile, performance results are made available to the Ministry of Works, Congress and the public. Australia publishes reports on the individual performance of federal and state programs, allowing for cross-program and cross-regional comparisons. Along these lines, the public provisionof annual evaluation results is included inthe mexican congressional mandate. However, accuracy, clarity and reliability of results should be ensured before disseminating information to general public. SEDESOL's Manual Ciudduno introduces the public to available public programs and hence acts as a tool of public awareness, which interms helps the public inunderstandingthe results once published. D. Internationallessonsfor the implementationofRBM Lessons from international experience indicate that the implementation of RBM should be an incremental process." It i s essential that the implementation begins with a strategic road map with specific milestones and remains flexible, as opposed to an approach with strict plans and regulations. The implementation should be carried out in a step-wise manner which includes first assessingpreconditions for the compatibility of the existing institutional framework with M&E, creating a plan and design of a simple system and implementing a pilot phase. Varying political, institutional and technical situations in each country present unique challenges for respective governments. The pilot exercise can postpone costs and allow agencies to learn the language and technology of measurement M&E. The pilot phase of implementation should be approached as an opportunity to resolve implementation headaches before the system is expanded. In addition, to increase the sustainability of the system, each step of the process should be participatory. The involvement of senior management, staff and other stakeholders gives them ownership of the process andincreasesits likelihood of success. l4 Rawlings and Rubioibid l5 IDEA International, ibid. 218 Box 6.3 Engaging relevant government and non-government actors in the design and implementation of a RBMinMesico16 Ministry of Public Function (Secretaria de la Funcicin Pliblica). This office is incharge of carrying out the financial, operational, performance and human and material resources of federal programs. For the audit exercise at the state level, there are coordination agreements with the state and municipal control units. In addition to the audit function, they have the responsibility for quality assurance of those programs where stipulated presidential goals have been established in collaboration with the Office of the President. The FederalAudit Office has suggestedcertain actions to improve evaluations inMexico including: improving the capacity of government officials to produce good terms of reference for evaluations; simplifying the rules of operation to reduce the number of administrative procedures for the evaluators to gain access to programdata; and improving ministries' incentive structuresfor contracting and supervisingevaluations. Presidencia. At the beginning of the current administration Presidencia started the construction of a presidential goal system (Sistema de Metas Presidenciales). This system was conceived as a tool that would allow movement from a norms-based government to one based on results. It is part of the strategic model for improving the quality and innovation of government. The first monitoring system developed to track the Presidential goals was based on a scorecard. However, it was considered that the scorecard was not producing the intended results and it was changedto a short-term indicator system. This new indicator systemis expected to be completed during the beginning of 2004. It is not clear at this point how soon the Sistema de Metas Presidenciales will be implemented. Ministry of Finance. Fromthe perspectiveof the Ministry of Financethere is not yet a useful and coherent M&Esystemof socialprograms.The current indicators do not go beyondtracking inputs andoutputs (e.g. expenditure, number of beneficiaries, people trained, etc.). Most of what is available to the Ministry of Finance are output indicators which do not help to detect deviations from the expectedprogram's goals. In linkingoutputs to decision making, expendituresare generally basedon past allocations and in only a few programs, such as Oportunidades, is the budget based on impact evaluation results. In the Ministry of Finance's view the implementationof a pilot to assess a program's cost-effectivenessshouldbe mandatory inorder to respondeffectively to the needof the population. They have indicated that they would be hard pressedto nominate any existing M&Esystems as bestpractice, which inturncould serve as an example to other programs wanting to buildaconsistentM&Esystem. Academia. In general, academicsinvolved in the area consider that recent legal changes incorporated in the Social Development Law (below) a good first step in promoting a sounddiscussionof a program's design and net impact. They also noted that there is more of an evaluation culture in the country at present in comparison to the recent past (2000). However, there is also a consensus on the existing difficulties in moving from measuring outputs to outcomes, given the lack of sufficient incentives to do so. Again, they expect that the new Social Development Law will be of great help inthis regard, given its emphasis in the measurement of outcomes. E. ChallengesofRBM In addition to recomendations for the design and implementation of a results-based M&E in Mexico, international experience also illustrates some associated challenges. Matsuda (2003) argues that Latin American countries face challenges at different levels intheir desire to movetowards RBMsystems. Low-capacitygovernments andeven some sectors within high capacity governments, like Brazil's federal administration, are not l6Theinformation isbasedonaforthcomingWorldBankM&EReadinessAssessment,2004.From December 8 through 12,2003 a World Bank mission had a series of interviews with mexican government officials involved inM&E. 219 equipped with a sufficient number of technically capable staff to manage the analytical tasks involved in program and organizational performance evaluations. Weak analytical capacity hinders eachphase inthe logic chain of RBM, ultimately breaking down the link between evaluation results andbudgetary and policy decision-making. Focusedcapacity- building efforts can alleviate this constraint, especially in those countries that already have a reasonable pool of civil service professionals and a developed network of academic institutions andpolicy analysis think tanks. InOECD countries, finance ministries have often beenthe driving force behind RBM initiatives. B y linking performance requirements with the budget process, finance ministries are in a strategic position to foster performance orientations within the public sector. Unlike cutting-edge reformers in OECD countries, however, in most of Latin America the finance ministries have not ledpublic sector reformefforts. More often than not, finance ministries activities have primarily been driven by concerns about expenditure control for fiscal management purposes, often neglecting other aspects of good public expenditure management, namely the effective allocation of budgetary resources to the government's policy priorities, and the fostering government-wide operational efficiency. A key challenge i s therefore to help ministries of finance gain the technical capacity to lead or at least fully take advantageof RBMefforts. The first implication of international lessons to SEDESOL i s that a unified M&E system cannot be designed all at once and cannot be implemented all at once. Incrementalism will have to come intwo ways -first, the design over time of the technical system to be more and more inclusive of the information needs of managers across the ministry and, second, the expansion over time of the organizational units who come under its umbrella. Since all of this seemingly cannot happen at once, the logic of systematically pilotinglexpanding the design and construction of an M&E system over time within SEDESOL is welljustified. A second implication is that the performance goals for SEDESOL will have to be clear and legitimate for all parts of the organization. Buildingaperformance basedsystem will require that the goals of the president of the country and the goals of the minister of SEDESOLare acceptedandthus frame the formulation of secondorder goals for eachof the units within the ministry. There will have to be an alignment within the ministry around goals and outcomes that at present i s not evident. The optimum situation would be one where the higher order goals andoutcomes cascadedown through the organization so that all parts of the ministry see their own mission and responsibilities in light of the overall mission. There needs to be a "line of sight" on the mission of SEDESOL from the top of the organization all the way to the bottom. Stated differently, the technical construction of the system (indicators, baselines, targets, etc.) cannot proceed if there i s not a clarity around and acceptance of the goals and outcomes to be achieved. Indicators can only be constructed and reporting can only begin when it i s clear what goals or outcomes are to be monitored andevaluated. The third implication i s the need for redefining the Congressional mandate for annual program evaluations. The mandate has been interpreted to be call for conducting annual 220 impact evaluations of all programs, an interpretation that has ledto frustrations in terms of fulfilling the mandate. First, as stipulated in the mandate, each program must be separately evaluated despite the overlap of several programs' objectives and geographic target populations. This raises the overall cost of evaluations, exacerbating financial constraints. Second, due to budget and time constraints, well-qualified external evaluation institutions, which are few inMexico, have little interest inparticipating inthe evaluations process. Third, the annual impact evaluation requirement does not allow the ministries to designmore efficient evaluation instruments. Lastly, the evaluation of social programs in Mexico often lacks a mechanism that provides feedback for improving the design and efficiency of current and future programs, Consequently, the lessons taken from the results of separate programevaluations obtained each year are not shared across programs, hampering the improvement of future program design. The congressional mandate would perhaps be better served if re-interpreted as a directive for performance measurement,based on monitoring data provided inannualreports, as well as occasional evaluations including longer-term impact evaluations, linked to an integrated feedback mechanism that leadsto better allocation of resources anddesignof programs. The fourth implication i s that with the approval in 2003 of the Social Development Law the evaluation mandate now acquires the status of a legal requirement and thus its continuation into the future. An evaluation requirement i s no longer explicitly neededin the Programa de Egresos de la Federacio'n (PEF) every year. Previously, the PEF yearly stipulated the main guidelines for evaluation and these guidelines had each year to be validated by the Congress. F. Conclusion.andrecommendations Mexico has come a long way in building better M&E systems and in fostering an evaluation culture. Influenced by the successful evaluation of ~ ~ o r ~ ~ n i d athee s , d government has moved to institutionalize evaluation through the Congressional mandate of 1999. The system of presidential performance targets and the recently-enacted Social Development Law of 2003 are also important reforms aimed at using M&E to support public effective sector management. More recently, through SEDESOL, Mexico i s attempting to broaden the scope and increase the effectiveness of program evaluation through the creation of an integrated RBMsystem. International experienceprovides severallessonsfor thistype of initiative: e The development and successful implementation of an RBM system i s an incremental process that requires strong leadership and a continual commitment from all of those involved. e The active participation of the Ministry of Finance should be encouragedinorder to eventually allow programperformance data to informbudgetary allocations. e Although the system has technical requirements to ensure the collectionof timely, reliable data, the primary focus of the RBMsystem should be on managementand on ensuring that the results are valued andusedas key inputsto decision making. 221 e The system should be crafted to the needs of managers and policymakers, accessible to all stakeholders -including civil society- and involve capacity- buildingactivities to ensureits effective implementation. e The RBM system should draw on a mix of internal and external evaluations, should provide regular performance data and should use impact evaluations strategically for programs that are innovative, replicable, involve substantial resource allocations, andhave well-defined interventions. Without a clear mandatefor transition to an RBMsystem and strong leadership, efforts at improving the performance of public sector institutions can fail. Strong leadership, commitment and involvement in RBM implementation are key to sensitizing different constituencies to the importance and benefits of RBMas well as increasing ownership of the process. The recently-enacted Social Development Law buildingon the congressional mandate of 1999,the system of presidential performance targets and a strong track record inthe impact evaluation of ~~~~~~~~~~e~ has positioned Mexico andSEDESOLto take a leadership position in implementing a successful RBM system. These efforts are all indications of the government of Mexico's commitment to performance management. SEDESOL's initiativeto implement an M&E systemthat i s result-basedis an opportunity to ensurethat the objectives of the congressional mandateare properly fulfilled. Apart from leadership, international experience indicates that further work is also needed inother areas inorder to successfully implementaRBMsystem. More work is neededon leadership and incentives and in resources, roles and responsibilities. For example, with respect to incentives, it is not yet clear why government agencies in Mexico would commit to supporting the development of a RBM. Fundamentally, the cultural shift that will be required to move to a performance orientation will need to be supported by a broad array of incentives to do so. These incentives will need to be carefully constructed to include both carrots and sticks -rewards and sanctions. But they do not exist at present. They will have to be constructed just as negative incentives will have to be removed. 222 m PI PI c f4 E i v3 2. Appendix A. Data Sets What follows is a brief description of each of the data sets at our disposal and the informationthat can beextractedfrom each. 1. Barro-Lee Data set on educational attainment at various levels for the male and female population. The data set includes estimates of educational attainment for the population by age -over age 15 and over age 25- for 126 countries in the world. (See Barro, Robert andJ.W. Lee, "International Measures of Schooling Years and Schooling Quality, AER, Papers and Proceedings, 86(2), pp. 218-223 and also see "International Data on Education", manuscipt). Dataare presentedquinquennially for the years 1960-1990. 2. Conteode Poblucio'ny Vivienda1995 Data providedby the Conteo comes from two sources: the Enumeracio'n,which counted all households in the country, and the Encuestu which took a sample of 80,000 households nationally (2,500 per state). The Enumeracio'n provides data on the space distributionof the populationand characteristics of households at the local levelwhile the Encuestaprovides information on families, social services, health services, migration and many other themes at the statelevel.' 3. ENCASEH The Survey of Socioeconomic Characteristics of the Households (Encuestu de Curacteristicas Socioecono'micas de los Hogures, ENCASEH) was the principal instrument for the collection of data of the "Program for Education, Health, and Food" (Progresa) for the identification of the beneficiary families. Sampling design: The universe of this study was constructed from householdslpeople belonging to the cities identifiedby PROGRESA as eligible by presenting a highdegree of poverty (according to 1995 "Conteo" and population census), located in the states of Guerrero, Hidalgo, Michoach, Puebla, Querktaro, San Luis Potosi, and Veracmz. Stratified and conglomerated sampling was employed. Information from 506 localities and 24,700 households was collected. As the main objective of the survey was to identify the results and impacts of PROGRESA, the sample was randomly divided into two groups: one which received the benefits of the program, 320 localities (treated) and one which did not, 186 localities (control group). The beneficiaries of the program were the poorest householdsin the treated localities. Note that there are some problems with the data from Chiapas; the Conteo underestimatesthe population due to the inabilityto executethe Enumeraci6n incertain areas. 230 This survey was carried out previous to the incorporation of the households into the program. An interview to all the households of the locality was carried out. The information obtained from the survey of the ENCASEH served as starting point for the evaluation of the conditions of poverty of the households, and permits the identification of the beneficiary householdsof the program. The survey shows information of diverse nature on the characteristics of the households that addresses the following subjects: identification of the household head and of the people responsible for the expenditure and children care; demographic features of the household and humancapital of its members; business activity of the members and their income; conditions of housing and availability of services; possession of goods, There i s also a locality questionnaire. 4. ENCEL The Encuesta de Evaluacidn de 10s Hogares (ENCEL) was designed as an instrument to evaluate the "Programa de Educacibn, Salud y Alimentacih (PROGRESA)" The Evaluation Survey consists of a follow-up survey every 6 months on these same households of ENCASEH. Since household information was collected every six months, a panelof households was created. 5. Encuesta Nucionul de lngresos y Gustos de 10s Hogares (ENIGH) 1992, 1994, 1996,1998,2000, and 2002 The ENIGHsurveys collect, periodically and systematically, socio-economic information of the households. This information is representative at the national, rural-urban, and marginality stratum (for 2002 only, according to CONAPO's classification) levels. The main objective of this survey is to generate information on current income and expenditure structure, financial income and expenditure structure, the value of the goods and services for self-consumption, the socio-economic characteristics of the household members, their labor conditions, and the household characteristics. For each year, the sampling process was stratified, multi-staged and by conglomerates. The final sampling unit is the household and all its members. In every stage, the selection probability is proportional to the size of the sampling unit, so the use of weighting factors i s necessary to obtain the appropriate estimates, 6. Sampleof the XI1Census2000 Two questionnaires were used for this Census, the basic one, which was applied to all mexican population, and the augmented (survey) which was applied to a sample. The sample design for the survey allows generating information at the municipal level for most indicators obtained from the augmented questionnaire. Inevery municipality, the sampling was by single-stage conglomerates, that is, complete geographical areas (primary sampling units) were selected. Such units vary across types of areas (rural- urban)andinthe sampledesign that is applied ineach. Thus, the sampling units mightbe AGEBs, blocks, or rurallocalities. 231 Given the main objectives of the sample, a sample size of approximately 700 households per municipality was estimated. However, in municipalities with the same or fewer number of households an adjustment for "finite population'' was done. Thus, for Chiapas, Hidalgo, Jalisco, and Veracruz the minimum sample size for municipality was fixed at 500 households; while in Oaxaca, Yucatiin, Puebla, and Tlaxcala the minimum ranged between 150 and 250 households. Inthese states, the municipalities with less than 1,000 inhabitants were surveyed completely. The former gave place to 2.2 million households interviewed nationally. Finally, given the sampling design, it is necessary to use weighting factors to obtain confident estimates. 7. Statistical Annex of the President's Addressto the Nation2001,2002 and2003 The Statistical Annex of the President provides a collection of data from the federal administration. The information i s obtained from the administrative records of various government ministries. It also collects several indicators at the state level. 8. Employment Surveys: ENE,ENET,and EMU During its evaluation, the Employment Surveys have had two sampling design stages. The first stage collected the information every year and was representative for rural and urban areas. The National Survey of Employment (Encuestu Nucionul de Empleo, or ENE),since 1996, andthe National Survey of UrbanEmployment (EncuestuNucionul de Empleo Urbuno, or ENEU), since 1987, collect information about the characteristics of the labor force 'inMexico at the national and urban level, respectively. These surveys are the basis for calculating official employment statistics. Both ENE and ENEU surveys share the same variables and structure. ENE and ENEU include context variables at the individual level such as: education, labor participation, hours worked, type of contract and benefit received, some questions about social protection coverage or labor force and other basic variables as gender, maritalstatus andage. The ENEUcollected the same sets of information that ENE, but ENEUonly focuses on the urban population; this means that the survey i s restricted to urban areas with more than 100,000 inhabitants. The geographical coverage on ENEUhas changed over time with the introduction on new urban areas and it currently covers the 45 largest urban areas in Mexico. The main advantage of ENEUi s that the survey i s based on a panel of individuals that can be followed for 5 quarters, The ENEU is structured as a rotating panel. In each quarterly sample there are five cohorts, each in a different state of completion in the interview cycle: one-fifth of the sample in its first interview; one-fifth in its last (fifth) interview; and three-fifths of the sample in intermediate stages. The ENEUconducts extensive quarterly household interviews in the 16 major metropolitan areas for from 1992 to 2000. The sample is selected to be geographically and socio- economically representative. The questionnaire is extensive in its coverage of participation in the labor market, wages, and hours worked that are traditionally found in 232 such employment surveys. Additionally, a household identification variable permits construction of household incomes. Inthe second stage since 2000, ENEUdisappears and ENEincorporates both ENEand EMUemployment databases, becoming the ENET and has apanel structure that allows following every individualfor 5 quarters. Specifically, since the second quarter of 2001, the ENE consists of a cross sectional database and it i s representative for the next mutually exclusive aggregatelevels: a) Nationallevel. b) Four levels of inhabitants (these levels allow dividingthe sample inrural and urbanareas). c) 48 mayor cities d) 32 Mexican states ENET is the acronym for ENE's new version, which starting 2000 has become a quarterly survey (Encuesta~ a c ~ o nde a lEmpleo Trimestral). 9. Polity IV The POLITY Project has evolved since the 1970s and has become a widely used source of cross-national, longitudinaldata on the authority characteristics of modern polities. It i s most widely used for its assessments of the degree of democracy and autocracy in the political structures of nation states. It collects coded annual information on regime and authority characteristics for all independent states (with greater than 500,000 total population) in the global state system and covers the years 1800-2002. The Polity fv Web site is housedat the Center for International Development and Conflict Management at the University of Maryland, College Park. 10.WBI Governance ResearchIndicators Dataset This dataset presents estimates of six dimensions of governance covering 199 countries and territories for four time periods: 1996, 1998, 2000, and 2002. These indicators are based on several hundred individual variables measuring perceptions of governance, drawn from 25 separate data sources constructed by 18 different organizations. These individual measures of governance are assigned to categories capturing key dimensions of governance, and the unobserved components model i s used to construct six aggregate governance indicators in each of the four periods. The point estimates of the dimensions of governance are presented as well as the margins of error for each country for the four periods. The governance indicators reported are an update and expansion of WBIs previous work, part of a research project on indicators initiated in 1998 (Kaufmann, Kraay and Zoido-Lobatbn 1999a,b and2002). 233 11.WorldDevelopmentIndicators2002 The World Bank draws on a variety of sources for the statistics published in the World Development Indicators. Dataon external debt are reportedto the World Bankdirectly by the developing member countries through the Debtor Reporting System. Other data are drawn mainly from the United Nations and its specialized agencies, from the International Monetary Fund (IMF), and from country reports to the World Bank. Bank staff estimates are also used to improve currentness or consistency. For most countries, national account estimates are obtained from member governments through World Bank economic missions. Most social data from national sources are drawn from regular administrative files, special surveys, or periodic censuses. 234 B. Estimationof Electricity Subsidy (RES) Coverage. The ENIGH from 1992 to 2002 report access to electricity (in 2002 it also identifies consumers with a formal contract), as well as household expenditure on electricity. In contrast to total income and expenditure, which are significantly underreported in ENIGH in comparison to the National Accounts, ENIGHover reports monthly electricity expenditure inrelation to administrative records by a magnitude by a factor close of almost 2. This may be explained by the fact that most households pay a bimonthly electricity bill but may forget to divide this when reporting monthly expenditures inENIGH. Public expenditure. We apply the residential tariff schedule covering the majority of householdsandclose to half of total residential electricity expenditure (tariff 1)to impute electricity consumption (kWh) to households on the basis of their reported expenditure (divided by two, for the reason mentioned in the previous figure). The great majority of the rest of households and expenditure corresponds is to tariffs 1A-lC, which are sufficiently similar to the latter to make this a reasonable estimate for them as well. To estimate costs of provisionwe assumedthe highesttariff rate for residential consumption to reflect true marginal cost, which implies a total subsidy for residential consumers close to the estimate reported by the federal government. For 1992-2000 this implies a single tariff schedule with three progressive marginal rates, but the 2002 tariff reform implies three different schedules, making direct imputation of electricity consumption levels more difficult in this year. For this reason, to estimate the distribution of the subsidy in the latter year we have applied the 2002 tariffs to the consumption structure estimated for 2000, assumingthe latter to have remained basically stablebetweenthe two years. 235 References Aghion, Phillippe, Eve Caroli, and Cecilia Garcia-Peiialosa. 1999. "Inequality and Economic Growth: The Perspective of the New Growth Theories." Journal of EconomicLiterature 37: 1615-60. Alderman, Harold, 2002. "Do Local Officials Know Something We Don't? Decentralization of targeted transfers in Albania." Journal of Public Economics 83: 375-404. Alix, Jennifer, Alain de Janvry, and Elisabeth Sadoulet. 2003. Payments for Environmental Services in Mexico: To whom, where, and how much? Berkeley, CA: University of California, Department of Agricultural and Resource Economics. Araujo, Maria Caridad, Alain de Janvry, and Elisabeth Sadoulet. 2003. Spatial Patterns of Non-Agricultural Rural Employment Growth in Mkxico in the 1990s. http://are.Berkelv.EDU/-alain and http://are.Berkeley.EDU/~sadoulet.Berkeley, CA: University of California, Department of Agricultural and Resource Economics. Araujo, Maria Caridad, Alain de Janvry, and Elisabeth Sadoulet. 2001. Supply and Demand-Side Dete~inantsof Of-Farm Employment in Rural Mexico. Berkeley, CA: University of California, Department of Agricultural and Resource Economics. Arias, Omar. 2001. Are Men Benefiting from the New Economy? Male Economic Marginalization in Argentina, Brazil and Costa Rica. Policy ResearchWorking Paper; 2740. Washington, DC: The World Bank. Arriagada, Ana Maria and Gillette Hall. 1999. Measurement of Inequali~.Washington, DC: The World Bank.Mimeo. Atkinson, Anthony and Francois Bourguignon (eds). 2000. Handbook of Zncome Distribution. Amsterdam: Elsevier. Atkinson, Tony, Bea Cantillon, Eric Marlier and Brian Nolan. 2002. Social Indicators. TheEU and Social Inclusion. Oxford: Oxford University Press. Attanasio, Orazio P. and Miguel SzCkely. 1998. Household Savings and Income Distri~~tionin Mexico. Inter-American DevelopmentBank Working Paper; 390. Washington, DC: Inter-American Development Bank. .2001. Wage Shocks and Consumption Variabili~in Mexico during the 1990s. Inter-American DevelopmentBank Working Paper; 451. Washington, DC: Inter- AmericanDevelopmentBank, Baffes, John, and Shah, Anwar. 1998. "Productivity of Public Spending, Sectoral Allocation Choices, and Economic Growth.'' Economic Development and Cultural Change46(2): 291-303. 236 Banks, J., R. Blundell and A. Lewbel. 1997. "Quadratic Engel Curves and Consumer Demand."` TheReview of Economics and Statistics74(4): 527-539. Bardhan, Pranab, and Dilip Mookherjee. Forthcoming. "Decentralizing Anti-Poverty ProgramDelivery inDevelopingCountries." Journal of Public Economics. Barrcin PCrez, Maria Antonieta. 2003. Evaluacidn del Programa de Atencidn a Jornaleros Agricolas: Impact0 en 2001, Comportamientoen 2002 y un Andlisis Comparativo entre 2001 y 2002. Cuarto informe. MCxico: UniversidadNacional Autdnomade MCxico (UNAM),Facultadde Econom'a. BarrdnPCrez, MariaAntonieta [et. al.] 2003. Programa deDesarrollo de Comunidadesy Pueblos Indigenas. Evaluacidn Externa. MCxico: Universidad Nacional Autdnomade MCxico (UNAMj,Facultadde Econom'a. Bautista, Sergio [et. al.] 2003. Evaluacidn del Efecto de ~portunidadessobre la utilizacidn de servicios de salud en el medio rural. Cuernavaca, Morelos, MCxico: Instituto Nacionalde SaludPliblica (INSP). Behrman, Jere, Nancy Birdsall, and Miguel SzCkely. 2001. Economic Policy and Wage Di~erential~in Latin America. Penn Institute for Economic Research (PIER) Working Paper; 01-048. htt~://ssrn.cornlabstract=294517 BCnabou, Roland. 2000. "Unequal Societies: Income Distribution and the Social Contract". American EconomicReview 90 (1j: 96-129. Bhalla, Surjit. 2002. Imagine There'sNo Country: Poverty, Inequality, and Growth in the Era of Globalization. Washington, DC: Institutefor InternationalEconomics. Binswanger, Hans, and Swaminathan Aiyar. 2003. Scaling Up Community-Driven Development: Theoretical Underpinnings and Program Design Impl~cations. World Bank Policy Research Working Paper; 3039. WashingtonDC: The World Bank. Bird, Richard M. 2003. Taxation in Latin America: The Balance Between Equity, Eficiency and ~~stainability. International Tax Program Paper; 0306. Toronto: University of Toronto. Blinder, Alan S. 1973. "Wage Discrimination: ReducedFormand Structural Estimates". Journal of HumanResources 8(4): 436-455. Bobadilla, J.I. 1998. Quality of Perinatal Medical Care in Mexico City. MCxico: Instituto Nacional de Salud Pliblica (INSP). Bourdieu. 1990. TheLogic of Practice. Stanford, CA: StanfordUniversity Press. Bourguignon, FranGois, Francisco H. G. Ferreira, and Nora Lustig. 1998. The Microeconomics of Income Dis~ributionDynamics in East Asia and Latin America. Washington DC: The World Bank Development Economics Research Advisory Staff. Mimeo. Browning, M. and T. Crossley. 2001. Shocks, stocks and socks: consumption smoothing and the replacement of durables during an unemployment spell. Toronto: York University. Mimeo. 237 Buchinsky, M. 1994. "Changes in the U S Wage Structure 1963-87: An application of quantileregression". Econometrica62(3): 405-458. . 1995. "Estimating the Asymptotic Covariance Matrix for Quantile Regression Models: a Monte Carlo Study". Journal of Econometrics68: 303-38. Caballero, JosC Maria. 2004. "Comentario al Programa Microrregiones." In Mejores Prdcticas de Politica Social. Conferencia Internacional, Mkxico 2003. MCxico: Secretariade Desarrollo Social (SEDESOL). Calderdn, CCsar andAlbert0 Chong. Forthcoming. "Volume and Quality of Infrastructure and the Distribution of Income: An Empirical Investigation." Review of Income and Wealth. and Luis ServCn. 2002. The Output Cost of Latin America's Infrastructure Gap. Working Papers CentralBankof Chile; 186.Chile: CentralBankof Chile. Calvo, Guillermo A. and Enrique G. Mendoza. 1996. "Mexico's Balance of Payments Crisis: A Chronical of a Death Foretold". Journal of International Economics 41(3-4): 235-64. Challa, Surjit. 2002. Imagine There'sNo Country: Posei-ty,Inequality and Growth in the Era oj` Globalization. Washington DC: Institute for International Economics. (Arriba hay unlibro del mismo nombreperoel autor aparece conioBhafla). Chant, Sylvia. 1985. "Single Pattern Families: Choice or Constraint?" Development and Change 16: 636-656. Chaudhuri, Shubham. 2002. Empirical Methodsfor Assessing Household Vulnerabili~to Poverty. New York: ColumbiaUniversity, Departmentof Economics.Mimeo. . 2003 "Assessing empowerment initiatives: a framework for empirically based inference." PowerPoint Presentation, Workshop on Measuring Empowerment, World Bank, February 2003. and Jyotsna Jalan and Asep Sryahadi. 2002. Assessing Household Vulnerabili~ to Poverty from Cross-Sectional Data: A Methodology and Estimates from Indonesia.New York: ColumbiaUniversity, Departmentof Economics.Mimeo. Chen, Shaohua, and MartinRavallion. 2004. How Have the World'sPoorest Fared since the early 1980s?Washington, DC: The World Bank. Mimeo. Chiquiar, Daniel and Gordon H. Hanson. 2002. International Migration, Self-Selection, and the Distribution of Wages: Evidence from Mexico and the United States. NationalBureauof Economic ResearchWorking Paper; 9242. Centro de Investigacidny Docencia Econ6micas, A. C. (CIDE) e Instituto Tecnol6gico Aut6nomo de Mtxico (ITAM). 2003. Andlisis de las Finanzas Pu'blicas en Mkxico. MCxico: CIDE e ITAM.Mimeo. Comitk TCcnico para la Medici6n de la Pobreza. 2002. Medicidn de la Pobreza: Variantes Metodoldgicas y Estimacidn Preliminar. Serie Documentos de Investigacibn.MCxico: Secretariade Desarrollo Social (SEDESOL). .2004. Comunicado3: Observacionessobre las cifras oficiales 2002. MCxico. 4 238 Conley, John P. and William F. Maloney. 1995. "Optimal Sequencing of Credible Reforms with Uncertain Outcomes". Journal of DevelopmentEconomics 48 (1): 151-167. CONTIGO. 2003. "Programas que Conforman CONTIGO'. http://www.contino,gob.mx/index.php?idseccion=12. Cortds, Fernando [et. al.] 2002. Evolucidn y Caracteristicasde la Pobreza en MJxico en la Ultima D6cada del Siglo XX. Serie Documentos de Investigacibn. Mdxico: Secretariade DesarrolloSocial (SEDESOL). .2003. iDisminuyd lapobreza?Mkxico 2000-2002.Mimeo Cravioto, Patricia. 2004. "Evaluacibn de Impacto. Programas que Otorgan Subsidios". Powerpoint Presentation at the conference organized by WBI and SEDESOL SeminarioAvanzado de Evaluacidnde Impacto. April 2004. Mexico. Cunningham, Wendy and William F. Maloney. 2000. Measuring Vulnerability: Who Sugered in the 1995Mexican Crisis.Washington, DC: TheWorld Bank.Mimeo. Datt, Gaurav and MartinRavallion. 1994.Income Gainsfor the Poorfrom Public Works Employment, Evidencefrom TwoIndian Villages. Living Standards Measurement Study (LSMS), Working Paper; 100. Washington, DC: The World Bank, Davidson, Russell and James G. MacKinnon. 2000. "Bootstrap Tests: How Many Bootstraps?"EconometricReviews 19(1): 55-68 De Ferranti, David, Guillermo Perry, IndermitGill and Luis Servdn. 2000. Securing our Future in a Global Economy.WashingtonDC: The World Bank. and Guillermo Perry, Daniel kderman, and William F. Maloney. 2002. From Natural Resources to the Knowledge Economy: Trade and Job Quality. WashingtonDC: The WorldBank. and Guillermo Perry, Indermit Gill, Josd Luis Guasch, William. F. Maloney, Carolina Shchez-PCiramo and Norbert Schady. 2003. Closing the Gap in Education and Technology.WashingtonDC: The World Bank. and Guillermo Perry, Francisco H.G. Ferreira and Michael Walton. 2004. Inequali~in Latin America. Breaking withHistory? WashingtonDC: The World Bank. De Janvry, Alain, and Elisabeth Sadoulet. 2001. "Income Strategies Among Rural Households in Mexico: The Role of Off-Farm Activities". World Development 29(3): 476-80. . 2003a. "Achieving Success in Rural Development." Paper presented at the ConferenceMejores Prdcticas de Politica Social, Mdxico, May 2003. 2003b. Toward Implementation of an Integral Approach. http://are.Berkelev.EDU/-alainl and http://are.Berkelev.EDU/-sadou1et/. Berkeley, CA: University of California, Department of Agricultural and Resource Economics. 239 . 2003c. Analysis of the CONTlGO Strategy with Emphasis on Local and Rural Development: Toward Greater Coordination and Eficiency. Berkeley, CA: University of California, Department of Agricultural and Resource Economics. Mimeo. De Janvry, Alain, Frederico Finan, and ElisabethSadoulet. 2002. Measuring the Poverty Reduction Potential of Land in Rural Mexico. Berkeley, CA: University of California, Departmentof Agricultural andResourceEconomics.Mimeo. . 2001. How Eflective are Educational Subsidies Programsfor the Rural Poor? Progresa in Mexico. http://are.Berkeley.EDU/-alai~ and http://are.Berkelev,EDU/-sadouletl. Berkeley, CA: University of California, Departmentof Agricultural andResourceEconomics. De la Torre, Augusto, Eduardo Levy Yeyati, and Sergio Schmukler. 2002. Financial Globalization: Unequal Blessings. World Bank Policy ResearchWorking Paper; 2903, WashingtonDC: The WorldBank. De Soto, Hernando. 2000. TheMystery of Capital. ~y Capitalism Triumphsin the West and Fails EverywhereElse. United States: BasicBooks. Deaton, Angus. 1997. The Analysis of Household Surveys. Microeconomic Analysis for DevelopmentPolicy. WashingtonDC: The World Bank. .2001. "Counting the World's Poor: Problems and Possible Solutions." World Bank ResearchObserver 16: 125-147. and J. Muellbauer. 1980, "An Almost Ideal Demand System." American Economic Review 70(3): 312-26. and Jean Drkze. 2002. "Poverty and Inequality in India: A Reexamination" Economicand Political WeeklySeptember7. Dercon, Stefan. 2002. Income Risk, Coping Strategiesand Safety Nets. DiscussionPaper; 2002122. Helsinki: World Institute for Development Economics Research (WIDER). Dreze, Jean and Amartya Sen. 1989. Hunger and Public Action. Oxford: Clarendon Press. (eds.). 1996.Indian Development: SelectedRegional Perspectives.Delhi: Oxford University Press, United Nations University (UNU)-World Institute for DevelopmentEconomics Research(WIDER). Durand Alchtara, Carlos H. 2003. Evaluacidn Externa del Programa de Identidad Juridica. Mtxico: Universidad Aut6noma Metropolitana. Edwards, Sebastian. 1998. "Two Crises: Inflationary Inertia and Credibility". Economic Journal lOS(448): 680-702. Escobar, Agustin andMercedes Gonzaez de la Rocha. 2002. Evaluacidn CuaZitativadel Programa de Desarrollo Humano Oportunidades. Seguimiento de Impact0 2001- 2002 comunidades de 2,500 a 50,000 ~abitantes. Mtxico: Centro de Investigacionesy EstudiosSuperiores en Antropologia Social (CIESAS). 240 EspinozaP., Leonidez. Evaluacidn Externadel programa Jdvenes por Mxico. Tlaxcala, MCxico: Universidad Autbnomade Tlaxcala (UAT). Esquivel, Gerardo. 1999. ConvergenciaRegional en Mkxico, 1940-1995. Documento de Trabajo; IX-99. MCxico: ElColegiode MCxico, Centrode EstudiosEconbmicos. Esquivel H., Gerardo, [et. al.]. 2002. Evaluacidn Extema del Programa para la Superacidn de la Pobreza Urbana. MCxico: El Colegio de MCxico, Centro de EstudiosEconbmicos. Estache, Antonio, Vivien Foster, andQuentinT. Wodon. 2002. Accountingfor Poverty in Infrastructure Reform: Learning from Lhttin America's Experience. World Bank Development Studies Series. Washington, DC: World Bank. Esteban, J. and D. Ray. 1994. "On the measurement of polarization." Econometrica 62 (4): 819-51. EstCvez, Federico, Beatriz Magaloni, and Albert0 Diaz-Cayeros. 2002. "The Erosion of One-party Rule: Clientelism, Portfolio Diversification, and Electoral Strategy". Paper presented at the 2002 Annual Meetings of the American Political Science Association Conference. Boston. August. Faguet, Jean-Paul. Forthcoming. "Does Decentralization Increase Government Responsiveness to Local Needs? Decentralization and Public Investment in Bolivia". Journal of Public Economics. Fajnzylber, Pablo, William F. Maloney and Gabriel Montes. 2003. LDC Micro Firm ~ynamics:How Similar are they to those in the Industrialized World?Evidence from Mexico. www.worldba~.or~laceconomist.Washington, DC: The World Bank, LatinAmerica and CaribbeanRegion.Mimeo. Ferreira, Francisco H. and Phillippe G. kite. 2003, Policy Options for Meeting the ~illenniumDevelopment Goals in Brazil: Can ~icro-simulationshelp? Policy ResearchWorking Paper; 2975. Washington, DC: The WorldBank. and Julie Litchfield. 1999, "Calm after the storms: income distribution and welfare inChile, 1987-94". The WorldBank EconomicReview 13(3): 509-38. Filmer, Deon, and Lant Pritchett. 1999. "The impact of public spending on health: Does moneymatter?" Social Science & Medicine 49: 1309-1323. and Jeffrey S. Hammer and Lant Pritchett. 1998. Health Policy in Poor Countries: Weak Links in the Chain. World Bank Policy Research Working Paper; 1874. Washington, DC: The WorldBank. . 2002. "Weak Links in the Chain I1A Prescription for Health Policy in Poor Countries". World Bank Research Observer 17(1): 47-66. Fonseca Sibaja, Asdnibal. 2001. El Sistema Nacional de Evaluacidn: un Instrumento para la Toma de Decisiones del Gobiemo de Costa Rica. Caracas: Centro Latinoamericanode Administracibn parael Desarrollo (CLAD). Foster, J., J. Greer, and E. Thorbecke. 1984, "A Class of Decomposable Poverty Measures.'' Econometrica 52: 761-766. 241 Fox, Jonathan. 1991. "Popular Participation and Access to Food: Mexico's Community Food Councils". In Scott Whiteford and Anne E. Ferguson (eds.) Harvest of Want: Hunger and Food Security in Central America and Mexico. Boulder: Westview Press. . 1993. The Politics of Food in Mexico. Ithaca and London: Cornel1University Press. .1996a. "HOWDoesCivil SocietyThicken?The Political Constructionof Social Capital inRuralMexico." WorldDevelopment24 (6): 1089-1103. . 1996b. "Los FondosMunicipales de Solidaridady laparticipacicinComunitaria en Oaxaca." RevistaMexicanade Sociologia 58(3): 145-168. Frankenberg E., D. Thomas and K. Beegle. 1999. The Real Costs of Zndonesia's Economic Crisis: Prelimina~Findings from the Indonesia Family Life Surveys. RAND Labor and Population Program Working Paper Series; 99-04. Santa Monica, CA: RAND. Freire, MariaEmilia. 2002. Urban Development in Mexico. Washington, DC: The World Bank. Mimeo. Galasso, Emanuela, and Martin Ravallion. 2001. Decentralized Targeting of an Anti- Poverty Program. Washington DC: The WorldBank. Gamanou, Gisele and Jonathan Morduch. 2002. Measuring Vulnerabili~to Poverty. Discussion Paper 20021'58. Helsinki: United Nations University-World Institute For DevelopmentEconomicsResearch(WIDER). Garcia-VerdG, Rodrigo. 2002a. Income, mortality, and literacy distribution dynamics across states in Mexico: 1940-2000.MexicoCity. Mimeo. .2002b. TheHumanDevelopmentZndex and its Application to States in Mexico. DireccicjndeEstudiosEconcimicos.MCxico: Bancode Mexico. Gil-Diaz, Francisco and Agustin Carstens. 1996. "One Year of Solitude: Some Pilgrim Tales about Mexico's 1994-95 Crisis." American Economic Review 86(2): 164- 69. Glewwe, Paul and Gillette Hall. 1998. "Are Some Groups More Vulnerable to Macroeconomic Shocks than Others?HypothesisTests Basedon PanelData from Peru". Journal of DevelopmentEconomics 56: 181-206. Gbmez-Dantes, Octavio. 1999.Prospects of the Mexican Health CareReform. Gould, W. W. 1992. "Quantile Regression with. Bootstrapped Standard Errors." Statu TechnicalBulletin 9: 19-21. , 1997. "Interquantile and Simultaneous-Quantile Regression.'' Statu Technical Bulletin 38: 14-22.ReprintedinStatu TechnicalBulletin Reprints 7: 167-176. Grindle, Merilee S. 2004. Despite the Odds: ContentiousPolitics and EducationReform. New Jersey: PrincetonUniverstiy Press. 242 Halac, Marina and Sergio Schmukler. 2003. Distributional Eflects of Crises: TheRole of ~inancialTransfers. Background paper for World Bank 2003 LAC Flagship: Znequali~in Latin America and the Caribbean: Breaking with History? Halvorsen, R. and Palmsquist, R. 1980. "The interpretation of Dummy Variables in Semilogarithmic Equations". American EconomicReview 70(3): 474-475. Hanson, Gordon. 2003. What has happened to wages in Mexico since NAFTA? Zmplicationsfor hemis~hericfree trade. National Bureau of Economic Research Working Paper; 9563. Heckman, James J. 1976. "Sample Selection Bias as a Specification Error", Econometrica,47(1): 153-62. and R. Robb. 1985. "Alternative Methods For Evaluating The Impact of Interventions." In J. Heckman and B. Singer (eds.), ~ n g i t ~ i nAnalysis of a l Labor Market Datu. Cambridge, MA: Cambridge UniversityPress. and Salvador Navarro-Lozano. 2003. "Using Matching, Instrumental Variables and Control Functions to Estimate Economic Choice Models", NationalBureau of Economic ResearchWorking Papers 9497. and Hidehiko Ichimura and Petra Todd. 1997, "Matching as an Econometric Evaluation Estimator: Evidence from Evaluating a Job TrainingProgramme", The Review of Economic Studies,64(4): 605-54. and Hidehiko Ichimura, Jeffrey Smith and Petra Todd. 1998. "Characterizing Selection Bias UsingExperimental Data". Econometrica,66(5): 1017-98. and Robert J. LaLonde and Jeffrey A. Smith. 1999. "The Economics and Econometrics of Active Labor Market Programs". InOrley Ashenfelter andDavid Card(eds). andb book of Labor Economics, VolumeZZZ. Elsevier SciencePub.Co. Heller, Patrick and James Mahoney. 2003. "The Resilience and T r a n ~ f o ~ a b iofl i ~ Social ~nequali~in Latin America". Background paper for World Bank 2003 LAC Flagship: ~`Znequali~in Latin America and the Caribbean: Breaking with History?' Institute for Democracy and Electoral Assistance (IDEA) International. 2003. Results- Bused Management System (RBM) for Social Programs: A Review of Znternational Experiences. Washington, DC: The World Bank. Instituto Tecnol6gico de Estudios Superiores de Monterrey (ITESM). 2003. Evaluacidn de Resultados del Progruma deAbasto Social de Leche, a cargo de Liconsa, S.A. de C.V. Resultudos de la Evaluacidn Enero a Diciembre de 2002. MCxico: ITESM. Instituto Nacional de Estadistica, Geografia e Informhtica (INEGI) Economic Commission for Latin America and the Caribbean (ECLAC). M a g n i t ~ eand Evolution of the Poverty in Mdxico 1984-1992.Aguascalientes, MCxico: ECLAC- INEGI. Instituto Nacional de Salud Pdblica (INSP). 2003. Evaluacidn Seguro Popular. Enero- Agosto. M6xico: INSP. 243 and Centro de Investigaciones y Estudios Superiores en Antropologia Social (CIESAS). 2003. Evaluacidn de Resultados de Impact0 del Programa de Desarrollo Humano ~portunid~es, Mexico: TNSP andCESAS. Jalan, Jyotsna and Martin Ravallion. 1999. Income Gains to the Poor from Worl$are: Estimates for Argentina's Trabajar Program. Working papers - Labor & Employment. Labor Market Policies and Institutions; 2149. Washington, DC: The World Bank. Jimdnez, Emmanuel. 2003. Comments on SEDESOL's Monitoring and Evaluation System. Paper presented at the conference Mejores Prdcticas de Politica Social, Mexico: May 2003. Mimeo. Kamin, Steven B. and John H. Rogers. 1996. "Monetary Policy in the End-Game to Eschange Rate Based Stabilizations: The Case of Mexico". Journal of Inte~~tionalEconomics 41: 3-4: 285-307. Katz and Murphy. 1992. "Changes in RelativeWages, 1963-1987: Supply and Demand Factors". Quarterly Journal of Economics 107(1): 35-78. Kennedy, Peter E. 1981. "Estimation with Corrrectly Interpreted Dummy Variables in SemilogarithmicEquations". American EconomicReview 71(4): 801. Koenker, Roger and G. Bassett Jr. 1978. "Regression Quantiles". Econometrica 46, 33- 50. . 1982. "Robust Test for Heteroskedasticity Based on Regression Quantiles". Econometrica50:43-61 and Kevin Hallock. 2001. "Quantile Regression". Journal of Economic Perspectives 15(4): 143-56. Krishna, Pravin, Tom Krebs and William F. Maloney. 2003. Tmde Policy, Income Risk and Welfare.Washington, DC: The World BankandBrown University. Mimeo. Kusek, Jody Zall and Ray C. Rist. 2004. Building Results-Based Monitoring and Evaluation Systems: Assessing Country Readiness. PREM Note. Washington, DC: The World Bank. Lanjouw, Peter and Martin Ravallion. 1999. "Benefit Incidence, Public Spending Reforms, and the Timing of Program Capture". World Bank Economic Review 13(2): 257-73. Larre, B. and M. Bonturi. 2001. Public Spending in Mexico: How to Enhance its Efectiveness. OECD EconomicsDepartmentWorking paper; 288. Paris: OECD. Lederman, Daniel, William F. Maloney, and Luis Servdn. 2003. Lessonsfrom NAFTAfor Latin America and the Caribbean Countries: A Summary of Findings. Washington, DC: The World Bank. Levenson, Alec and William F. Maloney. 1998. "The Informal Sector, Firm Dynamics and Institutional Participation." InWorld Bank (1999) Mexican Labor Markets: New Viewson Integration and Flexibility. Washington, DC: The World Bank. 244 Lewis, Oscar. 1959. Five Families: Mexican Case Studies in the Culture of Poverty. UnitedStates: Basic Books. Ligon, Ethan and Laura Schechter. 2002. Measuring Vulnerability. Discussion Paper; 2002186. Helsinki: UnitedNations University (World Institute for Development EconomicsResearch-WIDER). Loayza, Norman, Pablo Fajnzylber, and C6sar Calderbn. 2002. Economic Growth in Latin America and the Caribbean. Stylized Facts, Explanations, and Forecasts. Washington, DC: The World Bank.Mimeo. Lbpez, Humberto. 2003. Macroeconomics and Inequali~.Poverty Reduction Group. Washington, DC: The World Bank.Mimeo. L6pez-Acevedo, Gladys. 1999. Learning outcomes and school cost-effectiveness in Mexico: the PARE program. Policy ResearchWorking Paper; 2128. Washington, DC: The World Bank. .2004. Professional Development and Incentives for Teacher Peformance in Schools in Mexico. Policy Research Working Paper; 3236. Washington, DC: The WorldBank. and Angel Salinas. 2000. "Marginal Willingness to Pay for Education and the Determinantsof Enrollment inMexico." BackgroundPaper #5, completedas part of the Earnings Inequality after Mexico's Economicand EducationReforms study at the World Bank, ReportNo. 19945-ME. . 2001. Teacher Salaries and Professional Profile in Mexico. LCSHD Paper Series Vol. 63. Washington, DC: The World Bank. Lbpez-Calva, Luis Felipe. 1998. On Privatization Methods. Development Discussion Paper; 665. Boston, MA: HarvardInstitutefor International Development. and Lourdes Rodriguez Chamussy. 2003. The Evolution and Determinants of Poverty in Mexico, 1992-2002.Background paper commissioned for this work. Mimeo. Lustig, Nora C. 1998. Mexico: The Remaking of an Economy, 2nded. Washington DC: TheBrookings Institution. .2000. "Crises and the Poor: Socially Responsible Macroeconomics." Economia l(1): 1-30. Maloney, William F. 1999."Does Informality Imply Segmentation in Urban Labour Markets? Evidence from Sectoral Transitions in Mexico". The World Bank Economic Review 13(2): 275-302. andMariano Bosch, Jorge Moreno, MbnicaTinajero. 2003. Noteson Income and ConsumptionShocks. Backgroundpaper commissionedfor this work. Mimeo. and Wendy Cunningham and Mariano Bosch. 2003. Who Suffers Income Falls During Crises.WashingtonDC: The World Bank.Mimeo. 245 Martin, Robert. 2003. Consumption,Durable Goods, and Transaction Costs.Board of Governors of the Federal Reserve System International Finance Discussion Papers; 756. Martinez Valle, Adolfo. 2001. Social Znequality and Health in Mkxico 1990-1997: Zndividual and Policy Pathways - Former: Zmpact of Income Inequa~itieson Mexico`s Population Health. Washington, DC: The Johns Hopkins University and Pan American Health Organization. (PAHO). httr,://ris.bvsalud.or~/index.php?caten=rsch&action=ctrySearch&ctry=USA. Matsuda, Yasuhiko. 2003. Public Sector Reforms in Latin America and the Caribbean and their Zmpacts on Service Delivery. Washington, DC: The World Bank. Mimeo. McGraw, Barry 2002. TheKnowledgeEconomy and Lifelong Learning: Zmplicationsfor Developing Countries. PowerPoint presentation at the conference organized by the Baden-Wiirttemberg Foundation for Development Cooperation incooperation with the World Bank Group. October 9-10, 2002. The German State of Baden- Wiirttemberg. McKenzie, David J. 2002. "Are Tortillas a Giffen Good inMexico?'' Economics Bulletin 15(1): 1-7. 2003a. "How do Households Cope with Aggregate Shocks?Evidence from the MexicanPeso Crisis". WorldDevelopment, 31(7): 1179-1199. and Hillel Rapoport. 2003. Network efsects and the dynamics of migration and inequality: theory and evidence from Mexico. California: Stanford University. Mimeo. and Dilip Mookherjee. 2003. "The Distributive Impact of Privatization in Latin America: An Overview of Evidence from Four Countries." Economia 3 (2): 161- 233. Mtndez, Juan E., Guillermo O'Donnell and Paul0 SCrgio Pinheiro (eds.). 1999. The (Un)Rule of Law and the Underprivileged in Latin America. Notre Dame: Universityof Notre Dame Press. Messmacher Linartas, Miguel. 2000. Desigualdad regional en Mkxico. El efecto del TLCAN y otras reformas estructurales. Documento de Investigacibn; 2000-4. MCxico: Banco de MCxico. Mishra, Prachi. 2003. Emigration and Wages in Source Countries: Evidence from Mexico. New York: Department of Economics, ColumbiaUniversity. Mimeo. Montes, Carlos. 2003. Results-based Public Management in Bolivia. Overseas Development Institute Working Paper; 202, London: ODI. Morley, Samuel. 2001. The income distribution problem in Latin America and the Caribbean. Santiago, Chile: CEPAL. Nacional Financiera (various years). Economia Mexicana en Cifras. Aguascalientes: Nacional Financiera. 246 Narayan, D., R. Chambers, M. K. Shah, and P. Petesch. 2001a. Voices of the poor: Crying outfor change.New York: The WorldBank and Oxford University Press. Narayan, D., R. Patel, K.Schafft, A. Rademacher, and S. Koch-Schulte. 2001b. Voicesof the poor: Can anyone hear us? New York: The World Bank and Oxford University Press. Neumark, David, and Daiji Kawaguchi. 2001. Attrition Bias in Economic Relationships Estimated with Matched CPS Files. National Bureau of Economic Research Working Paper; 8663. North, Douglass C. 1990.Institutions, Institutional Change, and Economic Pegormance. Cambridge: CambridgeUniversity Press. Oaxaca, Ron. 1973. "Male-Female Differentials in Urban Labor Markets". International EconomicReview 14: 693-709. OECD. 2001. Knowledge and Skills for Life: First Resultsfrom the OECD Programme for Inte~ationalStudentAssessment (PISA)2000. Paris: OECD. ,2003. Economic OutlookMexico. Paris: OECD. Orozco, Manuel. 2003. "Changes in the Atmosphere? Increase of Remittances, Price Decline andNew Challenges". Inter-American Dialogue. Ospina Bozzi, Sonia and Doris Ochoa. 2003. "El Sistema Nacional de Evaluacidn de Resultados de la Gesti6n Piiblica (SINERGIA) de Colombia" In Nuria Cunill Grau and Sonia Ospina Bozzi, Evaluacidn de Resultados para una Gestidn Pu'blica Moderna y Democrdtica: Experiencias Latinoamericanas. Caracas: CentroLatinoamericanode Administracibn parael Desarrollo (CLAD). Paes de Barros, Ricardo. 2003. "Anaisis del DiseHo de CONTIGO'. Presentationat the July BrainstormMeetinginpreparationfor this work. Paqueo, Vicente B., Gladys Lopez-Acevedo, Harry A. Patrinos, andM.P. Savanti. 2003. "A Case Study of Mexico: Conditional Cash Transfers". InChanyong Park and Kye Woo Lee (eds.) Globalization and Social Safety Nets in Asia-Pacific Region. Seoul: KIHASA andMinistryof Health andWelfare. Parker, Susan and Carla Penderzini. 2002. "Diferencias de GCnero en la Educacidnen Mtxico". InElizabeth G. Katz and Maria C. Correia (Coord.) La Economia de Gkneroen Mkxico. Mexico: BancoMundialy Nacional Fianciera. Parker, Susan W. 2003. Evaluacidn del Impacto de Oportunid~essobre la Inscripcidn Escolar: Primaria, Secundaria y Media Superior. Mtxico: Instituto Nacional de SaludP6blica(IMP) y Centrode Investigacidny DocenciaEcon6micas(CIDE). Paxson, Christina and Norbert Schady. 2002. "The allocation and impact of socialfunds: spendingon school infrastructure inPeru". World Bank EconomicReview 16 (2): 297-310. Presidencia.2003. ImpuZso. PowerPointpresentation. Mexico City. Presidencia.Anexo 3er informe. http://www.informe.~residencia.~ob.~/index.php. 247 Rao, Vijayendra and Michael Walton, eds. 2004. Culture and Public Action. Stanford, CA: StanfordUniversity Press. Ravallion, Martin.2001. "Comment on `Counting the World's Poor,' by Angus Deaton." WorldBank Research Observer 16: 149-156. and Datt, Gaurav, and Chaudhuri, Shubham, 1993. "Does Maharashtra's Employment Scheme Guarantee Employment? Effects of the 1988 Wage Increase". Economic Development and Cultural Change, Vol. 41 (2): 251-75. University of ChicagoPress. Rawlings, Laura and Gloria Martha Rubio. 2003. Evaluation the Zmpact of Conditional Cash Transfer Programs: Lessonsfrom Latin America. Policy ResearchWorking Paper; 3119. Washington, DC: TheWorldBank. Rist, Ray. 2003. "Building aResultsBasedManagementSystem: Issues andChallenges". PowerPointpresentation.MCxico: The World Bank. Rivera Vences, 2003. "Estimaciones de pobreza, 2000-2002 prueba de hipdtesis". InstitutoNacional de Estadistica, Geografiae Informiitica (INEGI).Mimeo. Rogers, W. H. 1993. "Calculation of Quantile Regression Standard Errors." Statu Technical Bulletin 13:18-19. Reprinted in Statu Technical Bulletin Reprints, 2: 133-137. Roldiin JA, [et. al.] 2000. La desnutrici~na nivel municipal en Mkxico de acuerdo a un indicador mixto de estado nutricional. MCxico, DF: Instituto Nacional de la Nutricidn Salvador Zubirfin (INNSZ). Rothschild, Emma. 2001, Economic Sentiments: Adam Smith, Condorcet, and the Enlightenment. Cambridge, MA andLondon: HarvardUniversity Press. Rubalcava, Luis N. and Graciela Teruel. 2003. The Eflect of PROGRESA on Demographic Composition. Centro de Investigacicin y Docencia Econcimicas (CIDE) Working Paper; DTE-285. Mexico: CIDE. Salinas de Gotari, Carlos. 2002. Mkxico: the Policy and Politics of Modernization. Barcelona: PlazaandJanCs. Santamaria, Mauricio and Gladys Ldpez-Acevedo. 2003. Labor Reform in Mkxico: Analysis of a Proposal Drafted by Employers and Employees. Washington, DC: TheWorld Bank. Mimeo. Schady, Norbert. 2002. The (positive) efect of macroeconomic crises on the schooling and employment decisions of children in a middle-income country. Policy ResearchWorking Paper; 2762. Washington, DC: The World Bank. and Carolina SSmchez-Phamo. 2003. Ofl and running? Technology, trade and the rising demandfor skilled workers in Latin America. Policy ResearchWorking Paper; 3015. Washington, DC: The WorldBank. Schejtman, Alejandro, and Julio BerdeguC. 2002. Desarrollo Territorial Rural. Chile: Rimisp. 248 Scott, John. 1999. "Decentralization, Social Spending and Poverty in Mexico". BackgroundPaper for Government Programs and Poverty in Mexico. ReportNo. 19214-ME.Washington, DC: The World Bank. . Forthcoming. "Public Spending and Inequality of Opportunities in Mexico". Forthcoming in Quentin Wodon, ed. Public Spending and Inequality in Latin America. WashingtonDC: The WorldBank. Secretaria de Desarrollo Social (SEDESOL), Instituto Maya, Red para el Desarrollo Sostenible de MCxico, A.C. "Informe Final de la Evaluacih Externa del ProgramaMicrorregiones", Secretaria de Desarrollo Social (SEDESOL). 2003. Manual Ciudadano. http://www sedesol gob.~/pro~ramas/main . . .htm .2002. Discurso de la SecretariaJosefna VdzquezMota. Serie Documentosde Investigaci6n.MCxico: Secretariade Desarrollo Social(SEDESOL). ,2003. "Evaluation andMonitoring SystemMexico". PowerPointpresentation. 2003. "Estrategia de Microrregiones". http://www .sedesol.gob.~pro~r~as/~cro~egiones.htm. Secretariade Hacienda y CrCdito Mblico (SHCP). 2002. Cuentade la Hacienda Pdblica Federal. htt~://www.shc~.~ob.~docs/index a2.html .2003. Infomzes sobre la SituacidnEcondmica, las Finanzas Pdblicas y la Deuda Pdblica. MCxico: SHCP: .2004. Distribucidn del pago de impuestos y recepcidn del gasto ptiblico por decilesde hogaresy personas. Resultadospara el aiio de 2002. Mtxico: SHCP. SecretariaTCcnicadel GabineteSocial. 2003. Panfeto infomzativo. MCxico. Selby, Henry A,, Arthur D. Murphy and StephenA. Lorenzen. 1990, TheMexican Urban Household: Organizingfor Self-Defense. Austin: University of Texas Press. Sen, Amartya. 1973. On Economic Inequality. Oxford: Clarendon Press, reprinted in an expandededition in 1997. .1985a. Commoditiesand Capabilities.Amsterdam: Elsevier, . 1985b. "Well-being, Agency and Freedom. The Dewey Lectures 1984." The Journal of Philosophy 82(4): 169-221. .1992.Inequality Reexamined.Cambridge: HarvardUniversityPress. ,1999. DevelopmentAs Freedom.New York: Knopf Press, .2000. "Social Justice andDistribution of Income". InBourguignon andAtkinson (eds). Handbook of IncomeDistri~utio~ 1.Amsterdam: Elsevier. Vol. Skoufias, Emmanuel. 2001. PROGRESA and its Impacts on the Human Capital and Welfare of ~ouseholdsin Rural Mexico: A synthesis of the Results of an Evaluation by IFPRI. Washington, DC: International Food Policy Research Institute(IFPRI). 249 andB. McClafferty. 2001. Is PROGRESA working? summa^ of the resultsof an Evaluation by IFPRI. Washington, DC: International Food Policy Research Institute (IFPRI). and Susan Parker. 2003a. "Job Loss and Family Adjustments in Work and SchoolingDuringthe MexicanPeso Crisis". Mimeo. . 2003b. "The Added Worker Effect over the Business Cycle: Evidence from UrbanMexico". Mimeo. Secretaria de Salud (SSA). Various years. Boletin de Info~acidnEstadistica, Cuentas Nacionalesy Estatalesde Salud. MCxico: SSA. .2001. Programa Nacional de Salud2001-2006.Mexico: SSA. .Salud:Mkxico 2002.MCxico:SSA. .2003.Reglasde Operacidndel SeguroPopular.MCxico: SSA. .2004. Fair Financing and UniversalSocial Protection: TheStructural Reform of theMexican Health System. Mtxico: SSA. Sojo, Eduardo and Roberto Villareal. 2003. Estrategia para superar la pobreza moderada: Unapropuesta de accidn interinstitucional. MCxico: Oficina para las PoliticasPfiblicas, Presidenciade laRepfiblica. Sumpsi, JosC Maria. 2002. Desarrollo Local en Areas Rurales de la Unidn Europea: La Iniciativa LEADER, Madrid: UniversidadPolitecnicade Madrid. SzCkely, Miguel. 1998. TheEconomics of Poverty, Inequality and ~ e a l t hAccumulation in Mexico. New York: St. MartinsPress. .2000. "CONTIGO." PowerPointpresentation. Mexico City. .2001. TheI990s in Latin America: Another Decade of PersistentInequal~ty,but with Somewhat Lower Poverty. IADB Working Paper; 454. Washington, DC: IADB . .2003. Lo que dicen los pobres. Cuadernosde desarrollo humano, 13. Mexico: Secretaria de Desarrollo Social. Tan, Hong. 2001. Evaluating SME Programs: Some Lessons from Mexico's CIMO Program. Labor Market Policy Course, April 23-May 4, 2001, Washington DC: The World Bank. and Gladys L6pez-Acevedo. 2002. "Research Proposal: How well do SME programs work? Evaluating Mexico's SME Programs using Panel Firm Data". WashingtonDC: The WorldBank. .2003. Eval~atingSMEs' programs. World Bank Report. Washington DC: The World Bank. Mimeo. .2004. SMEsPrograms Evuluation. Washington DC: The World Bank.Mimeo. Task Force on Remittances. 2004. All in the Family: Latin America's Most Important Inte~ationalFlow. Washington, DC: Inter-American Dialogue 250 Tendler, Judith. 1998. TheRise of Social Funds: Whatare They a Model Of? Cambridge: MassachusettsInstituteof Technology. Tilly, Charles. 1999. Durable Znequality.Berkeley: University of California Press. Todd, andHicks. 2003.A Brief Exa~inationof the relationship between GDP/capita and Pe$ormance on Selected Social Zndicators in the Latin America and Caribbean Region.WashingtonDC: The WorldBank,Mimeo. Tolentino Mayo, MariaLizbeth [et. al.] 2003. Ali~entacidny Pobreza: Efecto de la crisis de 1994 en el consumo de alimentos en Mbxico. MCxico: Instituto Nacional de SaludPdblica. Tremblay, K. (2002),"Student Mobility between and towards OECD Countries: A Comparative Analysis". In OECD Znternational ~ o b i l ~oft ythe Highly Skilled. Paris: OECD. UNDP. 2003. HumanDevelopmentReport on Mexico. MCxico: UNDP. Universidad Aut6noma Metropolitana (UAM). 2002. "Sintesis temfitica de la primera entrega de resultados: Evaluacibn Externa del Programa de Empleo Temporal 2001-2001". Mtxico: UAM. Universidad Aut6noma de Tlaxcala (UAT). "Evaluacibn Externa del.Programa Mujeres Jefas de Familia". Tlaxcala, Mtxico: UAT. . "Programa Estatales por Demanda (Evaluacibn Externa)". Tlaxcala, Mkxico: UAT. Universidad Nacional Autbnoma de MCxico (UNAM). 2003. Programa Oport~nidades Productivas (EvaluacidnExterna 2002).Mkxico: UNAM,Facultadde Econom'a. Van de Walle, Dominique. 2003. "Behavioral IncidenceAnaisis of Public Spending and Social Programs". In Frangois Bourguignon and Luiz A. Pereira de Silva (eds.) TheZmpact of Economic Policies on Poverty and Zncome Distribution. Evaluation Techniquesand Tools.Washington, DC: The World Bank and Oxford University Press. Villagbmez, Alejandro and Fernando Solis-Sober6n. 1999. "Ahorro y Pensiones en Mexico: UnEstudioa1Nivelde las Fannilias". EconomiaMexicana,Nueva Epoca 13(2): 339-66. Webb, S. and Gonzalez, C., 2003 Bargaining for a New Fiscal Pact in Mexico. WashingtonDC: The World Bank.Mimeo. Welch, B. L. 1947. "The Generalization of Student's Problem when Several Different PopulationVariances are Involved". Biometrika 34: 28-35. World Bank. 1990. World Development Report 1990: Poverty. Washington, DC: The World Bank. . 1997. Colombia: Paving the Wayfor a Results-Oriented Public Sector. Report No. 15300-CO.WashingtonDC: The World Bank. . 1999a. Zndigenous Peoples Profile. Report No. 188899-ME. Washington, DC: The World Bank. 251 . 1999b. GovernmentPrograms and Poverty in Mexico. Report No. 19214-ME. Washington, DC: TheWorldBank. , 1999c. Mexican Labor Markets: New Views on Integration and Flexibility, Volume Two: Technical Papers. Report No. 19438-ME. Washington DC: The WorldBank. .2000a. Earnings Znequality aj?er Mexico's Economicand Educational Reforms. ReportNo. 19945-ME.Washington,DC: TheWorldBank. . 2000b. World Development Report 200001: Attacking Poverty. Washington, DC: The WorldBank. .2001a. Poverty Reduction in Indonesia: Constructinga New Strategy. Report No. 23028-IND. WashingtonDC: TheWorldBank. .2001b.Mexico: Land Policy: A Decade after the Ejido Reform. WashingtonDC: TheWorldBank. . 2001c. Mexico: A Comprehensive Development Agenda for the New Era. Washington, DC: The WorldBank. . 2002a. Mexico: Veracruz-Llave Poverty Diagnostic. Report No. 25161-ME. WashingtonDC: TheWorldBank. .2002b.Brazil Planningfor Pegormance in theFederal Government:Review of Pluriannual Planning. ReportNo. 22870-BR.WashingtonDC:The WorldBank. . 2003a. Universal Health Insurance Coverage in Mexico: In Search for Alternatives.WashingtonDC: TheWorldBank,Mimeo. . 2003b. World Development Report 2004: Making Services Work for Poor People.Washington, DC: TheWorld Bank. 2003~.Southern States Development Strategy. Washington DC: The World Bank. http://wb1n0018.worldbank.or~~AC~AC.nsf~CADocbyUni~54 lD538865DA8 25A85256DC5007A7BC8?0pendocument . 2004a. Public Expenditure Review. Report No, 27894-MX. Washington DC: TheWorldBank. .2004b. Broadening Access to Financial ServicesAmong the Urban Population: Mexico Cily's Unbanked.WashingtonDC:TheWorldBank.Mimeo. . 2004~.Doing Business in 2004: Understanding Regulation. Washington, DC: TheWorldBank andOxfordUniversityPress. 252