Report No. 35910-CR Costa Rica Poverty Assessment Recapturing Momentum for Poverty Reduction February 12, 2007 Poverty Reduction and Economic Management and Human Development Sector Management Units Latin America and the Caribbean Region Document of the World Bank LSMS Living Standards MeasurementStudy MCJD Ministry of Culture, Youth and Sports (Ministerio de Cultura,Juventudy Deporte) MIDEPLAN Ministry of NationalPlanningandEconomicPolicies (Ministerio de Planificaci6n Nacionaly PoliticaEconbmica) MPE Ministry of Public Education(Ministerio de Educaci6nPdblica) MTSS Ministry of Labor and Social Security (Ministerio de Trabajo y Seguridad Social) NA NationalAccounts NGO NongovernmentalOrganization OECD Organizationfor Economic Co-operationand Development OLS OrdinaryLeast Squares PANACI NationalInstitutefor the Protectionof the Blinds (PatronatoNacionalde Ciegos) PANARE NationalInstituteof Rehabilitation(PatronatoNacionalde Rehabilitacih) PAN1 NationalInstituteof Children(PatronatoNacionalde la Infancia) PPP PurchasePowerParity PREM Poverty ReductionandEconomicManagement PROCUMEN Economic Improvementand NourishingSecurity Program(Programade Mejoramiento Econ6mico y SeguridadAlimentaria) PROMECE Improvementof the Quality Educationof GeneralBasic andPreschoolProgram (Programade Mejoramientode la Calidadde la EducacidnGeneralBhsicay Preescolar) PROMESA "Remote" Learningfor Secondary Education PRONAMYPE NationalProgramfor the Micro and Small Enterprises(ProgramaNacionalde Micro y PequeiiaEmpresa) PSU Primary Sampling Unit R&D ResearchandDevelopment RIVM Old Age andDeathFund(Regimende Invalidez,Vejez y Muerte) SAB Attention for BeneficiariesSystem(Sistemade Atenci6n a Beneficiarios) SIMED NationalSystemof Improvementof Quality Education(SistemaNacionalde Mejoramientode la Calidadde laEducaci6nCostamcense) SINE NationalEvaluationSystem(SistemaNacionalde Evaluaci6n) SIP0 PopulationInformationSystem(Sistemade Identificacidnde la Poblaci6nObjetivo) UCR Universityof CostaRica(Universidadde CostaRica) UNDP UnitedNationsDevelopmentProgramme UNICEF UnitedNationsChildren'sFund WDI World DevelopmentIndicators Vice President: Pamela Cox Country Director: Jane Armitage Director PREM: Ernest0 May Director HD: Evangeline Javier LeadEconomist: David Gould Sector Manager PREM: Jaime Saavedra Sector Manager HD: Helena Ribe Sector Leader HD: Laura Rawlings Task Manager: Andrew D.Mason Co-Task Manager: Carlos E. Sobrado ... 111 TABLE OF CONTENTS EXECUTIVE SUMMARY ......................................................................................................................................................... xiv PART 1:THE NATUREAND EVOLUTION OF POVERTY INCOSTA RICA, 1989-2004 1. INTRODUCTION . ............... ..... ............................................................................................................... 1 2. RECENTPROGRESS, CURRENTCHALLENGES .................. ................................................................... ........4 3. GROWTH,INEQUALITY AND POVERTY ...... 36 4. POVERTYAND THE COSTA RICAN LABOR MARKET..................................................................... 66 PART 2: POVERTYAND SOCIAL SECTOR POLICY 5. INTRODUCTION .................... 101 6. EDUCATION.. ....................................................................................................................................... 103 7. HEALTH ................................................................................................................................................................. 130 8. SOCIAL PROTECTION .................... 160 PART3: RECAPTURLNGMOMENTUM FORPOVERTYREDUCTION 9. ELEMENTS OF A POVERTY REDUCTION STRATEGY FOR COSTA RICA................................................. 188 REFERENCES ........... ............................ . .................................................,.................................. ..............................................194 ANNEXES 1. Correlatesof Poverty in CostaRica 1989, 1994 and 2000-2004 .................... 201 2. EntropyInequalityMeasures.................... ........................................................................... 205 3. Poverty HeadcountandMethodologicalIssues: A SensitivityAnalysis ............................................................................... 206 4. Total GrowthElasticitiesof Poverty Reductionandthe Efficiency of Growth...................................................................... 212 5. HealthExpenditureEfficiency in CostaRicaandLatin America ........ 6. DetailedRiskProfiles for CostaRicansby Age Group and Poverty Level, 2004..................... TABLES 1.1 Key Socio-economicIndicatorsin CostaRica: Early 1990s-Early 2000s........................................................................ 1 1.2 Countries (latestavailable year, 1997-03) .............................................................................................................. Socio-economic Indicators- How CostaRicaCompares with Latin AmericaandUpper-MiddleIncome 2 2.1 Overallandextremepovertyby countryfor selected years ...... .................... 6 2.2 Povertyheadcountandcontributionto povertyby regionandarea....................................................... 8 2.3 RuralPoverty concentrationsby country........................................................................................................................... 9 2.4 Poverty,Average IncomePoverty Gap andSeverity by regionsand area ..................... 10 iv 2.5 HouseholdSize by Poverty Group in CentralAmerica .................... ....... ................................. 10 2.6 CostaRicaHousingConditionsandServicesChanges for All Poor And ExtremelyPoor ' 1989-2004 2.8 IndigenousPopulationandPoverty indicators,CostaRica, 2000........................................... 2.9 IndigenousPopulationandwell-beingindicators,CostaRica, 2000.............................................................. 2.10 Educationand Afro-CostaRicans..................................................... .......................................... 20 2.11 Afro-CostaRicaswell-beingindicators,CostaRica 2000...................................................... 2.12 CostaRica Years of Education'by Year andPovertyLevel ..................................................................... ............................................. 23 2.15 Employmentpercentages and rates (2004). ................................................................. 2.16 OpenUnemployment Rates Simulationsfor the Poor...................... ................................... 25 2.17 2004 Average Hoursper week worked, salary andtotal laborincome.................................... 2.18 Per Capita Income distributionby employment andpovertyclassification, 2004.......................................... 2.19 PerCapita non-laborincomeby povertygroup (2004) 2.20 CostaRica2004 Determinantsof poverty.............................................................................. 2.21 2004 Poverty Levels Usingthe 1989Income Distribution................................................................... 2.22 Sources of inequalityin CostaRica............ ........................... 3.1 Per capitagrowthperformance1989-2004(annualrates) 11............. .......................................... 38 3.2 Changesin povertyandgrowthelasticitiesof poverty 1989-20041/............................................ ..................... 42 3.5 Growthand inequalityinLatin American 1/ ............................................................. 3.6 Typology of Anti-Poverty Interventions,by Poverty Rate andPovertyDensity..... 3.7 Number of Cantonsin each of the FourPoverty Rate-PovertyDensity Categoriesby planningRegion(2004)....................... 3.8 SectoralDecompositionof Poverty Ch 3.9 Poverty Reductionand SectoralGrowth 1/........................................ ................................... 57 3.10 Decompositionof GrowthandDistributionalEffectson Changesin Poverty, by Sectorand UrbanlRural(1989-2004)andsub-periods) .................................................................... 3.11The Determinantsof GrowthandIncomeInequalityinLatinAmerica, 1960-2000 ............................. 61 3.12 ExplainingRecentChangesin GrowthandInequalityin CostaRica 1/ ................................... 4.1 Labor ForceParticipationRatesby Gender. 4.2 The Determinantsof Changesin Returnsto Education...................... ........................................... 82 4.3 Changesin the Variance of HoursWorked, by Gender and Sector........................................... 4.4 Proportionof Workers EarningAt, Bel UncoveredSectors (1989-1999 Average) ........................................... .......................... 89 4.5 EstimatedEffectof LegalMinimum Wageson Hourly Wages, Employment, Hoursand Monthly Earningsinthe CoveredandUncoveredSectors ........................... ..................................... 4.6 The ChangingShare of Female HeadedHouseholdsinCostaRica, AmongPoor Households V and Non-Poorhouseholds,1987-2004 ..................92 4.7 Education, Unemployment,Part-time Work and Self-employmentamong Male andFemale HouseholdHeads, 1987-2004............................................................................................................................................. 93 4.8 Structureof Femaleand Male-HeadedHouseholds 94 4.9 Earnings InequalityAmong Workers inC Immigrants(2000-2004).......................... ........................................ 96 6.1 CostaRica's Basic EducationIndicators, 104 6.2 CompletionRatesby Quintile andEducationLevel (%), CostaRica 1989-2004.............................................................. 113 6.3 Average Years of Schoolingby Age Groups and Immigration Status, 2004 ................................... 116 6.4 Reasonsfor 13 to 17Year-olds Not to Go to Schoolby QuintilesandGender in 2004...... 118 6.5 MPEPublic SchoolsSubsidies, CostaRica2004.............................................................................................................. 121 6.6 CostaRica's Number of Studentsper Teacher by Year 6.7 CostaRica's School InfrastructureIndicators,2002 6.8 Number of Studentsper Computer, 2002.......................................................................................................................... 124 SelectedSubjects, Levels and ................................................ 125 6.10 CycleIV PassRatesby SchoolType andRegion.............................................................................................................. 125 7.1 Key health Indicatorsin CostaRica: Early andMid 1990s-Early 2000s.......................................................................... 131 7.2 InternationalHealthIndicatorsfor Various RegionsandLatin AmericanCountries 7.3 RelativeDistributionof HealthExpensesby Level, CostaRica 1997-2004................................................... 7.4 Economic Impactof HealthReformProcess, CostaRica 2002......................................................................................... 139 7.5 CCSS andMinistry of HealthExpenditureStructure, CostaRica2004 ............................. 143 7.6 Source of HealthInsurance,CostaRica 1989-2004 143 7.7 Latin American CountriesRelationshipbetweenTotal He d HealthIndicators,and CostaRicanPerformance,2002........................................ ..................................................... 144 7.8 DoctorVisits from January to June 2006, CostaRica 146 7.9 UnsatisfiedDemandfor DoctorVisits, CostaRica2001......................................................................... 146 7.10 Hospitaluse, type anddays, CostaRica2001................................................................................................................... 147 7.11 VaccinationDeficits in ChildrenLess than Three Years Old, CostaRica 2002 7.12 Hedth andHousingServices by Quintile, Region, Zone andImmigration,Costa Rica2004......... 7.13 Summary Chart of MillenniumGoals for Health, CostaRica2004.................................................................................. 157 8.1 The Evolutionof the Public Spendingon Social Protectionin Costa Rica, 1999-2004 173 8.2 The Compositionof Public Spendingon Social Protectionin CostaRica, 1999-2004.......................... 174 8.3 CostaRica: Distributionof the Beneficiariesof the Main Programsof SocialProtection, 2003 ..................................... 177 8.4 Distributionof the Benefitsof Public SpendingSelectedSocialProtectionPrograms, CostaRica2003 ......................... 180 8.5 Costa Rica: Indicatorof Access to Main Programsof SocialProtection....................................... 182 FIGURES 2.1 Headcountpovertyin CostaRica........................................................................................................................................ 4 2.2 Latin America$2PPPJdayPoverty ... ................................................ 6 vi 2.3 CostaRica poverty evolutionby regions............................................................................................................................. 8 2.4 Costa Rica Housing, House type: Shack ....................... .................................................................... 11 2.5 CostaRica Housing, Tendency: "precario". 11 2.6 CostaRica Housing, Cooking: with firewood or coal......................................................................................................... 11 2.7 CostaRica Housing, Floor type: dirt 2.8 Costa Rica Housing, Roof: metal (zinc) ............................................................ 2.9 Costa Rica House Services, Water: no piped water............................................................................................................ 12 2.10 Costa Rica House Services, Water sources: "acueducto" 2.11CostaRica House Services, with sewer or septic tank .................................................................................... 2.12 CostaRica House Services, without electricity ................................................................................................................... 13 2.13 Costa Rica House Services, with garbage recollection .......13 2.14 Costa Rica House Services, with telephone line.................................................................................................................. 13 2.15 Povertyby birth place.......................................................................................................................................................... 14 2.16 Primary Net Enrollment Rateby Poverty Group, 2004 2.17 Primary Gross Enrollment Rate by Poverty Group, 2004 ................................. 2.18 Secondary Net Enrollment Rate by Poverty Group, 2004................................................................................................... 21 2.19 Secondary Gross Enrollment Rate by Poverty Group, 2004 ..... .................................................................. 21 2.20 PovertyLevel andLow Birth Weight 22 2.21 Headcount rates and InfantMortality, 2001............................................................................................................. 23 2.22 Gini coefficients inLAC..................................................... .................................................................................. 31 2.23 Gini by year and area .. 31 2.24 Income share by quintile, 2004 .......................................................... 32 2.25 Quintiles Income share ratios 2004 ........................................................................................................................ ............32 2.26 PovertyRates with All Four Adjustments .....,............ ...................................... 34 3.1 Poverty, growth, and changes ininequality. 37 3.2 The relationshipbetween the effectiveness of growth and inequality................................................................................. 38 B3.1 Survey basedincome growth vs. NA income based growth............................................................................................... 40 3.3 GrowthIncidenceCurves (1989-2004) 43 3.4 The impact of inequality on the efficiency of growth......................................................................................................... 44 3.5 Growth-IncidenceCurves for Urban and Rural Areas, 1989-1994, 1994-2000, and 2000-2004........................................ 46 3.6 Urban and ruralGICs (1994-2004) 47 3.7 Variation inPoverty Rates in Costa Rica, by PlanningRegion, 1989and 2004................................................................. 48 3.8 GICs, by Planning Region, by sub-period(1989-1994, 1994-2004)................................................................................... 49 3.9 Histogramof Canton-levelPer Capita Incomes, 1994and 2004 3.10 Distribution of Costa Rican Cantons, by PovertyRate and Pov 3.11 RelativeLabor Intensity, by Sector andPlanningRegion inCostaRica, (2004)............................................................... 56 3.12 Real Growth by Economic Sector, by sub-period 58 4.1 RealMonthly Labor Earnings(1999 colones).................................................................................................................... 67 4.2 National Unemployment Rate by Gender.................... .................................................................. 68 4.3 UnemploymentRates by Poverty Status, 1987-2004........................................ 68 4.4 UnemploymentRateby EducationLevel, 1987-2004........................................................................................................ 69 4.5 Percentof Workers in each Category who are Self-employed, 1987-2004 ................................. 71 4.6 Percentageof Poor Working Women Working Part-time, 1987-2004 72 4.7 Employmentby IndustrialSector, 1987-2004,All Workers............................................................................................... 72 4.8 MeanEarningsby IndustrialSectof in CostaRica, 1987-2004,All Workers (1999 colones) 4.9 Changesin Inequalityin Monthly Earnings, 1980-2004............................................................................ 4.10 Proportionof Workers at SelectedEducationLevels, 1980-2004..................................................................................... 77 4.11 Average Years of Educationof 20-year-oldsenteringthe Workforce 4.12 Returns to EducationinCostaRica, 1976-2004.................................................................................................. 4.13 Distribution of LegalMinimumWages Among Workers, 1988 and 1998........................................................................ 87 4.14 Distribution of LegalMinimum Wages andActual Wages, 1999 .............. 88 4.15 Mean(Log) Earnings inEachIndustrialSector, Adjustedfor Personaland Workplace Characteristics(relativeto commerce)............................................................................................................................... 98 5.1 Public SocialSector Spending(Education, Health, andSocialProtection)in Latin America ............... 101 5.2 SocialSecurity CoverageRatesfor the EconomicallyActive Populationin 15 Latin America Countries, 1990sand2000s.................................................................................................................................................. 102 6.1 Net PrimaryEnrollmentRatesinCostaRicaand SelectedCountries 2000-2004 .................... 104 6.2 Net Secondary EnrollmentRatesinCostaRicaand SelectedCountries 2000-2004 105 6.3 CostaRica's EducationalSystem.............................................................................. ......................................................... 106 6.4 GDP GrowthRatesin CostaRica 1979-2002 ................................... 108 6.5 Grade AttainmentLevelby Age in2004 109 6.6 Net andGross EnrollmentRates for Primaryand Secondary by Quintiles 1989-2004................................. 110 6.7 Gross andNet Enrollmentratesby level ............................................................................................................................ 112 6.8 Average CompletionRatesby EducationLevel, 1989-2004 .................... 112 6.9 SchoolAttendance for Different Age GroupsandQuintiles: 1989-2004........................................ 114 6.10 StudentsGoingto Their AppropriateSchool Gradeby Age Groups, 1989-2004.............................................................. 114 6.11 Underperformanceyears by Quintile........... ............................................................................................. 115 6.12 Average years of Schoolingfor 18 Year-olds and Older, 1989-2004 116 6.13 Poverty Level andyears of Educationfor 18 Year-olds andOlder, 1989-2004................................................................ 117 6.14 MPE ExpenditureLevels:Per Capita andas GDPPercent, 1978-2003............................................................................ 119 6.15 GovernmentExpenditurein Educationas Percentageof GDP, 2001-2005. ......119 6.16 Average MPE Budget Shares for the 1998-2001Period.................................................................................................... 120 6.17 MPE ExpenditureStructureby EducationLevel, CostaRica2004................................................................................... 120 6.18 PrimaryNet EnrollmentRate andEducational Public ExpenditureinLatin America, 2003 6.19 Secondary Net EnrollmentRate andEducationalPublicExpenditurein Latin America, 2003............................ 6.20 PrimaryCompletionRates with Projections, CostaRica 1994-2015................................................................................. 126 6.21 Secondary Net EnrollmentRateswith Projections,CostaRica 1999-2015 127 7.1 Direct CCSS Insurance ExpendituresandContributionsinCostaRica 1990-2003.......................................................... 136 7.2 Yearly Consultations by Age Group, Costa Rica 87, 92, 97 and 2002..................... .............................. 137 7.3 HealthInsuranceRealper Visit Cost, Costa Rica 1990-2004 137 ... V l l l 7.4 Share of the Economic Active PopulationContributingto System, CostaRica 1990-2004.............................................. 137 7.5 Number of EBAIS and Populationsper EBAIS, CostaRica2004......................................... 138 7.6 HealthExpendituresinLatin America, 2002 ................................................................................. 140 7.7 RealPublic Per Capitahealth Expenditure, CostaRica 1996-2004 . ........... .142 7.8 Compositionof Public Spendingon Health, CostaRica 1992and2004........................................................................... 142 7.9 Populationwith Insurance, CostaRica 1989-2004 ........................................... 145 7.10 PovertyandLow birth Weight at the CountryLevel, CostaRica2003 148 7.11 Supply Indicatorsand HouseholdPoverty at the RegionalLevel, CostaRica 2003.......................................................... 149 7.12 Distributionof Public Spendingon Healthby IncomeQuintile, CostaRica2001 ........................................... 150 7.13 Distributionof Public Spendingon Healthby IncomeQuintile andRegion, CostaRica2001 7.14 SocialExpense andIncomeLorenz Curves, CostaRica2000 ...... 7.15 MDG HealthGoal One: ReduceInfant Mortality andMeasles Un-Vaccinated. ........................................... 154 7.16 MDG HealthGoal Two: Reduce Maternal Healthand Increase B 7.17 AIDS Incidence(per 100,000) CostaRica, 1990-2004................. 7.18 Incidenceof Malaria, TB and Dengue, andMortality from TB, C 8.1 Organizationof the SocialProtectionSectorin CostaRica ............... BOXES 2.1 Componentsof income aggregate ............................................................................ 2.2 NicaraguansMigrantsin Costa Rica .................. 3.1 Poverty, growth, and the natureo ........................... ........................................... 40 3.2 Growth, Poverty,andthe Central AmericaFree Trade Agreement DR-CAFTA 4.1 Why do NicaraguansEarnLess than Others inCostaRica?................................................................................ 6.1 Costa RicanEducational System Composition: Non-teachinginstitutions................... ........................................... 106 7.1 ImmigrationandSocial Security ....................... 153 8.1 SocialProtection- ManagingSocialRisk, PromotingLong-termGrowth and Development 8.2 The Role of InstitutoMixto de Ayuda Social (IMAS) MAP 2.1 Costa Rica2004 Poverty Headcountby Regions.................................................................................................................. 7 ix Acknowledgements This report was preparedby a team ledby Andrew D.Mason (LCSHS) and Carlos Sobrado (LCSPP) under the general supervision of Helena Ribe (LCSHD) and Jaime Saavedra (LCSPP). The team comprised HumbertoLopez, Edwin Goiii Pacchioni (LCSCE), James Cercone, Jose Pacheco (Sanigest), Tim Gindling (University of Maryland Baltimore County); Juan Diego Trejos (University of Costa Rica), Catherine Marquette, and Kimie Tanabe (consultants). Logistical support, as well as assistance in processing the document, was provided by Claudia Isern, Martin Buchara (LCSHD), Marta Cervantes-Miguel, and Anne Pillay (LCSPP). The report has benefited enormously from comments, inputs, and other support provided by our governmental counterpartsand non-governmentalpartnersinCostaRica, duringaseries of meetings and discussions during the preparationof the report. In particular, we would like to thank President Oscar Arias, Guillermo Zliiiiga (Minister of Finance), Francisco de Paula Gutierrez (President of the Central Bank), Fernando Zumbado (Minister of Housing and Poverty Reduction), Leonard0 Garnier (Minister of Education), Ana Isabel Garcia (Vice Minister of Housing and Poverty Reduction), Diego Viquez (Executive Director,IMAS), and from the previous administration, Rocio Saenz (Minister of Health, Chair of the Social Cabinet) and Gilbert0 Barrantes (Minister of Economy, Industry, and Commerce) for their ongoing interest and support for this work, and for their comments and inputs at various stages of the Poverty Assessment process. We are gratefulto the many other CostaRicancounterparts and partners who helpedto organize and participated in the August 31,2006, conference in SanJose on the draft report. The feedback received during this event was very helpful in revising and finalizing the study. We would also like to thank Eduardo Lizano, Ronulfo JimCnez, and Roxana Viquez, along with the other organizers and participants of the Third Annual Conference of the Academia de Centroame'rica on Poverty in October, 2005, for the opportunity to present and discuss early findings of the study. The team benefited, as well, from the many other presentations and discussions that took place during the Conference. Miguel GutiCrrez Saxe from Estado de laNaci6nalso providedthe team usefulcomments at key stages of report's preparation. Maria Elena Gonzalez and Marita Begueri from the National Statistics and Census Institute (INEC) provided the team with valuable and timely support with Costa Rica's national household survey data and related information. Finally, we are grateful to our peer reviewers Edmundo Murrugarra,Louise Cord (PRMPR), and Dena Ringold (ECSHD) for their timely and insightful feedback at various stages of the work, and to Guillermo Perry (LCSCE), David Could (LCSPR), Ana Lucia Armijos (LCSPE), and JordanSchwartz (LCSFT) for their constructive inputs during the process. X I. IntroductionandExecutiveSummary Costa Rica is well-known for its socio-economic achievements Costa Rica has low levels of poverty and inequality by Latin American standards. It also performs well compared to other countries in Latin America and the Caribbean region, and to countries with similar income levels, in health and access to basic services. Its infant and child mortality rates are significantly lower, and its average life expectancy i s substantially higher. Some 97 percent of Costa Ricans have access to improved water supplies -a rate that also is high relative to comparable countries. Indeed, access to a range of basic services, including electricity and sanitation, i s generally high, as is access to suitable housing, both in absolute terms and by regional standards. Nonetheless, the countryfaces important challenges in the continuingfight against poverty Despite its considerable achievements, Costa Rica faces a number of important challenges: While the percentage of its population that is poor fell from 31.7 percent in 1989 to 22.9 in 1994, the poverty rate has not declined over the last decade; in 2004, 23.9 percent of the Costa Rican population was still poor. This stagnation i s surprising given that Costa Rica experienced relatively consistent economic growth duringthe period. Income inequality has been rising. While still relatively low by regional standards, inequality in Costa Rica, as measured by the gini coefficient, a widely used indicator of income inequality, rose from 0.44 in 1989 to 0.48 in 2004 - an increase that is economically significant. Despite considerable investment and some progress in education, the country still lags behind the Latin America and upper-middle income country averages for access and attainment at the secondary school level. This reflects, in part, the lasting effects of Costa Rica's economic crisis in the 1980s. Indeed, the poor still lag behind the non-poor in educational access and attainment, and this adversely affects their ability to participate in and benefit from economic growth. Even though Costa Rica has a relatively well developed set of social protection programs, many of its poor still fall outside the reach of the safety net. This study examines recent developments on the poverty front in Costa Rica, with particular emphasis on why poverty rates have not declined over the last 10 years despite consistent economic growth over the period. To accomplish this, the report develops a dynamic profile of poverty in Costa Rica to better understand the characteristics of the poor. It analyzes recent patterns of economic growth and the extent to which the poor have shared its benefits. The study also examines the dynamics of the labor market, including how recent labor market developments have affected the ability of the poor to generate higher incomes, particularly since 1994. It pays special attention to the impact that immigration from Nicaragua has had on poverty, and to the particular challenges faced by poor female workers. Finally, it examines the role and effectiveness of social sector spending and policies in improving the welfare of the poor and providingenabling them to escape poverty. xi The report was undertaken as part of an ongoing dialogue between the World Bank and the Government of Costa Rica on economic and social policy. Its contents reflect issues raised by Costa Rican officials in the previous and current administrations, as well as discussions with a number of non-governmental stakeholders. It draws on both recently published work as well as several newly commissioned studies. Several Costa Rican academicians and researchers collaborated closely in its preparation. The authors discussed background studies as well as the preliminary findings with a number of Costa Rican stakeholders, including members of the new administration. Progress in poverty reduction has stalled since 1994. .. The study identifies several reasons for the lack of progress in reducing poverty over the last decade. For one thing, growth in GDP - as well as household income - has slowed in recent years. Between 1994 and 2000, per capita household income grew at less than one-third of its 1989-1994 rate, according to Costa Rica's national household survey, the Encuesta de Hogares de Propdsitos Multiples (EHPM); and since 2000, average per capita income has barely changed. More importantly, while the benefits of growth were relatively evenly distributedbetween 1989 and 1994, an increasing share of the benefits has accrued to non-poor households since then. Indeed, poor and near-poor households actually experienced a decline in real income between 2000 and 2004. These shifts in how the fruits of growth are distributed reflect broader changes inthe Costa Rican and world economies that have led to a decline in the relative demand for less skilled workers at the same time that the relative supply of low-skilled workers has increased; among other things, the share of highschool graduatesinthe Costa Ricanlabor force declined duringthe 1990s, while the share of high drop-outs increased. This growing mismatch between the education and skill levels of the poor and the demand for labor has resulted in increased earnings inequality, significant increases in unemployment among the poor, and an increase in part-time (as opposed to full-time) work among low-skilled and poor workers - most strikingly among poor, single mothers. ...but with concerted effort Costa Rica can recapture its momentum in reducing poverty To ensure that the poor benefit from future economic progress, a multi-dimensional poverty reduction strategy is warranted. Many of the elements of such a strategy are consistent with the new administration's priorities, including an emphasis on increasing human capital among the poor and strengthening Costa Rica's social safety net. The proposed strategy would include efforts to: 0 Promote higher sustained levels of economic growth to increase economic opportunities for all Costa Ricans, includingfor the poor Strengthen the human capital of all Costa Ricans, with emphasis on improving secondary school education among the poor and bolstering skills formation among low-skilled workers so they can take full advantage of emerging economic opportunities Ensure that social protection (both social insurance and social assistance) i s available to the poorest, most vulnerable groups, thus protecting them against risks and shocks while improving their access to basic services xii Create an enabling environment for poor workers (particularly poor, working mothers) so that they can engage in more remunerative employment Define regionally differentiated investment strategies that respond to regional differences in patterns of poverty and thus maximize the effectiveness of anti-poverty measures at the local level, and Strengthen information systems for improvedpoverty monitoring, effective program targeting and monitoring and evaluation of poverty reduction programs to ensure that the Government's poverty-reduction efforts produce the desired results. Costa Rica's historical commitment to the social sectors has been important to the country's progress and accomplishments over the years. At the same time, the evidence makes clear that considerable gains could be realized by increasing the efficiency of current public social sector spending by strategically realigning resources and makinggreater use of targeted approaches to reach the poorest, most vulnerable Costa Ricans as a complement to existing universal programs. Utilizing existing fiscal resources more effectively in the fight against poverty i s particularly important given the fragility of Costa Rica's current fiscal situation. 11. What is the CurrentPovertySituationinCosta Rica? CostaRica has relatively low levels of poverty.. . According to official national poverty figures, Costa Rica's poverty rate is half that of the rest of Central America - although national poverty rates are not strictly comparable. But even by internationally comparable poverty lines, it has one of the lowest poverty headcounts in Latin America. Indeed, if poverty i s defined as households with income of $2 or less per person per day, 9 percent of Costa Rica's people live in poverty, a lower rate than for any Latin American country except Uruguay and less than half the regional average of 25 percent. Usinga $1-per-day poverty line, the poverty rate in Costa Rica i s only 2 percent, one-fifth of the Latin American average. EHPM data from 1989 to 2004 indicate that poverty declined significantly between 1989 and 1994, but has remained essentially constant since then. The proportion of the population that i s poor declined from 31.7 to 22.9 percent between 1989 and 1994 (Figure l), has hovered in and the 23-24 percent range since 1994. In 2004, it remained at 23.9 percent. Similarly, extreme poverty fell from 9.9 percent in 1989 to 6.8 percent in 1994, but remained at 6.6 percent in 2004. Although Costa Rica's poverty rates are generally low, this lack of progress over the last decade i s surprising considering that Costa Rica has experienced relatively consistent per capita GDP growth over the period, averaging 2.4 percent over the 1994-2004 period. Many studies show that economic growth i s normally associatedwith declining poverty. ...with distinct geographic patterns Although Costa Rica is relatively small, it has distinct geographic differences in the incidence of poverty. In 2004, the poverty rate was as high as 43.4 percent in the Brunca region and 38.6 percent in the Chorotega region, compared to the national average of 23.9 percent (Map 1). But it was only 18.4 percent in the Central region (which includes San Jose). These regional disparities have persisted over time; the Brunca and Chorotega regions had the highest poverty rates in 1989, and poverty rates in these regions have remained relatively high ever since. Differences between rural and urban areas, on the other hand, are relatively small. Poverty in rural areas was measured ... XI11 at 28.3 percent in 2004, compared to 20.8 percent in urban areas - a difference of less than 8 percentage points. Measured differences in rural and urban poverty are higher in most other Central American countries. Figure 1: Costa Rica House Services Water: no piped water I ca ~ 25% 0 2 20% 0% 1 I I I I I I I 1989 1994 2000 200tear2002 2003 2004 2005 To understand the geographic dimension of poverty in Costa Rica, it i s important to distinguish between areas with highpoverty rates and those with large concentrations of poor people. While the Central region has the lowest poverty rate, for instance, it is home to nearly half of Costa Rica's poor and about 40 percent of the country's extremely poor people. This is because roughly two-thirds of all Costa Ricans live there. This means that even though the other five planning regions have considerably higherpoverty rates, just over half of the poor and 60 percent of the extremely poor people live in these areas. These regional differences in the incidence and concentration of poverty have important implications for policy and investment approaches to reducing poverty at the local level. xiv Map 1: PovertyHeadcountby PlanningRegion,2004 COSTA RICA2004 POVERTYHEADCOUNT BY REGIONS NICARAGUA L ATLANTIC OCEAN PACIFIC OCEAN $& 43.4 I c kNAMA NATIONAL 23-9% Who are thepoor in CostaRica? Analysis indicates that people are more likely to be poor in Costa Rica if they come from: 0 Larger households-particularly those with more young membersunder the age of 18. 0 Households headedby women, especially those inrural areas. 0 Households in which the household head (or his or her companion) work in the informal sector or in agriculture. Rural areas in the two poorest planning regions, Brunca and Chorotega (this may reflect poorer infrastructure or weaker local economies inthose areas). Incontrast, education appearsto provide a way out of poverty: If a household head (or companion) has completed secondary education, the probability a household will be poor i s reduced by half, while having higher education reducesthe probability of being poor by 75 percent. Inruralareas, evenhavingsomeprimaryeducation (aproxy for literacy), reducesthe probability of beingpoor, controlling for other factors. Whilepoverty has stagnated, other socio-economicindicators have continued to improve Housing conditions (measured by the type and quality of housing, tenancy and the source of energy used for cooking) began the period high by regional standards, and continued to improve at least until 2000. Access to basic household services such as potable water, sanitation services, trash collection and fixed phone lines, also generally continued to improve through 2004 (again, xv from already highlevels at the start of the period). And while EHPMdata suggest that electricity coverage slipped somewhat between 1989 and 2000 - most notably among extremely poor households - access to electricity among the poor and extreme poor appears to have recovered duringthe 2000-2004 period. Health outcomesare generally good by regional standards.. . Costa Rica also performs well relative to other countries in Latin America and the Caribbean region, and to countries with similar income levels, in health outcomes (Table 1). Infant and child mortality rates are significantly lower than in comparable countries, while the average life expectancy i s substantially higher. While there are some measurable differences in heath status between the poor and non-poor, the differences are not as pronounced as in other countries in Central America. Data from Costa Rica's public health system show, for example, very little relationship between poverty and the percentageof low birth weight babies at the county (canton) level. Levels of health insurance coverage also are relatively high in Costa Rica. Indeed, social security coverage in Costa Rica i s among the highest in Latin America. Nonetheless, coverage rates remain significantly lower among the poor than the non-poor. In 2004, more than 30 percent of the extremely poor and 25 percent of all poor still lacked health insurance coverage, in contrast to only 16 percent of the non-poor. Moreover, recent declines in measles and poliomyelitis vaccination rates, as well as sharp increases in the incidence of malaria, dengue and tuberculosis, signal emerging challenges for the health sector. Indicator Costa Rica Latin America Upper-Middle and the Income Caribbean Countries Life Expectancy, 2003 (years at birth) 79 72 70 Infant Mortality Rate, 2003 (deaths per 1,000 live births) 8 25 21 Under-5 Mortality Rate, 2003 (deaths Der 1.000 live births) 10 31 27 I Net Primary Enrollment Rate, 2002 (%)' 1 90 1 93 1 91 I Net Secondary Enrollment Rate, 2002 (%)' 53 67 78 Access to Improved Water Source, 2002 (% of population) 97 91 91 While the education sector experienced setbacks in the early 1980s as a result of the country's financial crisis, outcomes have improved notably since then. Different data sources provide slightly different figures, but all indicate similar trends. The EHPM data indicate, for example, that primary-school net enrollment rates increased at the national level from 75 percent to 81 percent between 1989 and 2004, while secondary-school net enrollment rates increased from 32 percent to 47 percent. Achievement also has improved over time. From 1989 to 2004, primary school completion rates increased from 74 percent to 86 percent, while secondary school completionrates increasedfrom 28 percent to 37 percent. Average years of schooling inthe adult population rose from 5.5 years to 6.9 years. The poor and extremely poor have shared in the country's educational progress. As a result, a number of education gaps between the poor and the non-poor have narrowed over the period, with strong progress in school attendance, primary- school enrollment rates and completion, and average years of schooling. ...but the country continues toface important challenges in education Despite recent progress (and in contrast to such areas as health and access to potable water), Costa Rica continues to lagbehindLatin America and other upper-middle income countries innet secondary school enrollments (Table 1). Moreover, even though education gaps have narrowed, the poor still lag well behind the non-poor in a number of important areas. In 2004, primary- school net enrollment rates were still only 70 percent among children in the poorest income quintile, compared to 95 percent among those in the wealthiest quintile, while secondary-school net enrollment rates were only 33 percent, compared to over 70 percent for those in the highest quintile. Drop-out rates remain higher for poor students, and completion rates remain low, particularly at the secondary level. While primary-school completion rates were 71 percent for students in the poorest quintile in 2004, compared to 98 percent of those in the wealthiest quintile, secondary school completion rates were only 19 percent among students in the poorest quintile, compared to 59 percent for those in the wealthiest quintile. Evidence suggests, as well, that the quality of education in poor areas still lags considerably behind that in non-poor areas. As with poverty, income inequality in Costa Rica has been low by Latin American standards. Comparison of data from 15 countries inLatin America in 2000 suggests that only Uruguay had a lower gini coefficient than Costa Rica. Nonetheless, the data also show that income inequality has increased in Costa Rica since 1989; the gini coefficient rose from 0.44 in 1989 to 0.50 in 2001 before declining somewhat to 0.48 in 2004. Differences in family education levels play a critical role; in fact, differences in the education of the household head (or companion) can explain up to one-third of income inequality, depending on the measureused. An additional one- fourth can be explained by the proportion of household members who are worlung. Interestingly, several factors that appear to affect poverty, such as household size and gender of the household head, do not appear to be important sources of income inequality in Costa Rica. In sum, poverty and inequality in Costa Rica are low by Latin American standards, while most social indicators are high. Moreover, the country has experienced considerable socio-economic progress since the beginningof the 1990s. Nonetheless, the facts that the poverty headcount rate has stagnated at around 23-24 percent over the last decade despite relatively consistent economic growth, that income inequality has increased, and that the poor still lag behind the non-poor in educational access and attainment (with adversely consequences for their ability to participate in andbenefitfrom economic growth) all representsignificant challenges. To be able to confront these challenges effectively, it i s important to understand the factors that underlie Costa Rica's lack of progress in poverty reduction over the last 10 years, including why economic growth and government effort in the social sectors have not been effective forces for poverty reduction and what public policy can do to re-invigorate the country's fight against poverty. 111. Why Hasn't Poverty Declined inCosta Rica over the LastDecade? The absence of progress in poverty reduction in Costa Rica since 1994 presents something of a puzzle, at least at on the surface, as poverty has stagnated in the face of consistent economic growth during most of the period. Nonetheless, the evidence indicates that several factors combined to reduce the impact of growth on poverty over the last decade - factors related to the levels and patterns of growth, as well as to recent developments in Costa Rica's labor market. xvii Growth in per capita household income declined over time.. . While per capita GDP growth was generally positive since 1994, growth in both GDP and household income has slowed in recent years. Per capita GDP growth declined from 2.8 percent per year from 1989 to 1994, to 2.6 percent a year from 1994 to 2000 and then to 2.0 percent a year from 2000 to 2004. Household per capita income, as measured by the EHPM survey, slowed more dramatically. After rising by nearly 5 percent per year from 1989 to 1994, it grew only 1.5 percent per year from 1994 to 2000; between 2000 and 2004, average per capita household income barely changed (incomes grew by less than one-tenth of one percent per year). This slowdown, by itself, helps to explain the lack of progress inreducingpoverty. ...and thepoor benefited lessfrom the growth that did occur Perhaps more importantly, there has been a shift in how the benefits of growth have accrued to the poor and non-poor over time. From 1989 to 1994, the benefits of growth were.relatively equally distributed; according to EHPMdata, household income growth among the poor was 25.1 percent over that period, only slightly lower than the national average of 27.9 percent and slightly higher than growth among those in the middle 20 percent of the income distribution (Table 2). After 1994, however, income growth rates have been significantly lower for the poor than the non-poor. From 1994 to 2000, average per capita household income among poor households grew only 3 percent, roughly one-third of the national average (9.2 percent). And between 2000 and 2004, per capita income actually declined slightly among the poorest Costa Ricans (-1.3 percent among the poorest 25 percent of households). Indeed, since 2000, only households in the upper income quartile experienced positive per capita household income growth (a relatively small 2.2 percent between 2000 and 2004). Table 2: Cumulative ChangesinPer CapitaHouseholdIncomeinCostaRica (in percent),by quartile andsub-period from 1989-2004 1989-1994 1994-2000 2000-2004 Poorest Quartile 25.1% 3.0% -1.3% SecondQuartile 21.5% 5.0% -1.9% ThirdQuartile 23.1% 8.6% -2.6% Richest Quartile 31.9% 10.9% 2.2% National Average 27.9% 9.2% 0.3% This uneven distribution of the benefits of growth since 1994 can be seen not only at the household level, but across the different regions and economic sectors of Costa Rica. EHPMdata suggest that income growth has been faster in relatively wealthy counties (cantones) than in relatively poor ones. Moreover, growth has generally been higher in sectors that tend not to employ much poor or low-skilled labor - such as finance, commerce, and public administration - than in sectors where poor and near-poor workers tend to be concentrated - such as asculture, construction, manufacturing, and services. Changes in the relative demand and supply of unskilled labor... Several labor market developments since the early 1990s have combined to make it more difficult for the poor to raise their incomes, and thus escape from poverty: xviii Changes in the Costa Rican and global economies have led to increased demand for skilled workers relative to unskilledworkers. The increased demand for relatively skilled workers occurred at a time when Costa Rica experienced a decline inthe relative supply of such workers. Shifts in relative demand and supply for skilled and unskilled workers have resulted in increased earnings inequality, along with significant increases in unemployment among the poor and extremely poor since the early-to-mid 1990s. These shifts have also led to an increase in part-time work among low-skilled and poor workers, particularly among poor, female workers who are single parents. While changing patterns of investment and trade, along with technological changes, led to increased local demand for relatively skilled labor, setbacks in education resultingfrom the 1982 financial crisis were beginning to make themselves felt in the labor market. EHPM data show, for example, a decline in the proportion of secondary school graduates - and an increase in the share of secondary school drop-outs - in the labor force starting in the early-to-mid 1990s. An increase in female labor force participation as well as a surge in the supply of Nicaraguan immigrant workers in the mid-to-late 1990s, further increased the supply of low-skilled labor relativeto demand. ...ledto increased earnings inequality and higher unemploymentamong thepoor. .. Shifts in relative demand and supply for skilled and unskilled labor led to increasing earnings inequality starting in 1992 (although the data indicate that earnings inequality has started to decline again since 2002). This reflected higher market returns to those with higher levels of education. As importantly, shifts in relative demand and supply of skilled and unskilled labor have resulted in significant increases in unemployment among the poor and extreme poor since 1994 (Figure 2). Between 1994and 2003, unemployment rates rose from 8 percent to 17 percent among workers living in poor households and from 12 percent to 27 percent among those living in extremely poor households. Unemployment rates among non-poor workers, by contrast, remained steady at 5 percent or less. Unemployment among the poor appears to be largely structural (as opposed to cyclical), reflectinga skills mismatch between the supply of low-skilled, often poorly educated workers and firms' demand for such labor. xix Figure 2: Unemployment Rates by Poverty Status, 1987-2004 r 30 A 25 20 15 10 5 0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1998 1997 1998 1999 2000 2001 2002 2003 2004 /+'Extreme Poverty +All Poor t N o n - P o o r 1 Source: Author's calculationsfrom the HouseholdSurveysfor Multiple Purposes, 1987-2004 ...and a higher share of poor women working less thanfull-time The data also show an increaseinthe proportion of workers working less than the standard work- week (40 to 48 hours) since the end of the 1980s. The proportion of workers working full-time fell from 40 percent in 1988 to 31 percent in 2002, while the proportion working part-time increased. This increase in part-time work was driven by a significant increase in the proportion of poor women with part-timejobs - from just over 40 percent in 1987 to about 65 percent in 1995, where it has leveled off over the last decade (Figure 3). It is noteworthy that no other group (e.g., non-poor women, poor men or non-poor men) experienced an increase in part-time workers over the period. The increase appears to reflect the rapid increase in the share of female headed (largely single-parent) households in the population between 1987 and 2004 -from17.0 percent of all households to 26.4 percent - and an even bigger increase in the share of female headed households among poor households over the period - from 19.7 percent to 33.6 percent. Indeed, because part-time workers earn less than full-time workers for any given hourly wage, these trends have almost certainly contributed to the increase in poverty among female-headed householdsover the period. xx Figure3: Percentageof Poor WorkingWomen Working Part-time, 1987-2004 30 20 10 1987 1989 1991 1993 1995 1997 1999 2001 2003 l+Part-time +Full-time +Over-time 1 Source: Author's calculations from the Household Surveys for Multiple Purposes, 1987-2004 Certain labor marketpolicies also may have contributedto stagnatingpoverty rates Econometric analysis indicates that while Costa Rica's complex, multi-tiered system of minimum wages increasesthe hourly wage for many workers already employed in the formal sector, it also reducesformal sector employment among poor and near-poor workers. Specifically, a 10percent increase in legal minimumwages i s found to increase average wages of formal sector workers by about 1 percent, but to decrease formal sector employment by 1 percent (Table 3). The net benefits of the minimum wage system to formal sector workers are also unclear. While higher legal minimumwages raise formal sector workers' hourly wages, they are also found to reduce formal sector workers' hours worked per month. These two effects offset each other; as a result, monthly earnings for formal sector workers are no higher than they would have been in the absence of minimumwages. Together, these findings call into question the value of Costa Rica's minimumwage policy as atool for poverty reduction. Table 3: EstimatedEffects of Costa Rica's MinimumWages on Hourly Wages, Employment, Hours, andMonthlyEarningsinthe Coveredand UncoveredSectors (Percent change inresponse to a 10percent increase in real minimum wages) Covered Sector Uncovered Sector Hourly Wages 1.03% ns Employment -1.09% na Hours Worked -0.62% ns Monthly Earnings ns ns Note: ns = not significant, na = not available Other labor regulations may limit workers' - specifically female workers' - ability to earn their way out of poverty. Current law limits the ability to employ women at night. In the absence of suitable child-care arrangements, regulations that restrict women's ability to work non-standard hours may be contributing to the high observed levels of part-time work among poor female xxi workers and rising levels of poverty among predominantly single-parent, female headed households. Leveling the legal playing field between women and men by reducing legal barriers to women working non-standard work hours may make it easier for single mothers to obtain employment duringtimes when it i s easier to find others (whether extended family or friends) to care for their children. IV. Has immigrationfrom Nicaraguahinderedpoverty reductionsince 1994? Migration from Nicaragua into Costa Rica was extensive duringthe 1990s.Nicaraguans makeup, by far, the largest immigrant population in Costa Rica. At the height of Nicaraguan immigration in the late 1990s, roughly 20,000 Nicaraguans entered Costa Rica each year, although in-flows have slowed to about 9,000 per year since 2000. While the last official figure stems from the 2000 population census, it i s estimated that Nicaraguan immigrants made up roughly 7-8 percent of the populationof Costa Ricain 2005. Nicaraguan immigrants have less education, on average, than Costa Ricans (although they have higher average education levels than the Nicaraguan population as a whole). Nicaraguan workers in Costa Rica tend to be concentrated in three economic sectors: agriculture, domestic service and construction. Nicaraguan immigration does not appear to be a significant cause of stagnating poverty rates.. . Despite popular perceptions that the influx of poor Nicaraguan has hindered poverty reduction in Costa Rica, the data suggest that inflows of immigrants to Costa Rica have been too small to have had much of an impact on aggregate levels of poverty, Particularly since 2000. Indeed, simulation analysis using the EHPM data for 2004 indicate that in the absence of Nicaraguan households the poverty headcount would decline by less than 1 percent, a change that is statistically insignificant. Separate econometric analysis also indicates that Nicaraguan immigrant households are not significantly poorer than households of Costa Rican origin once education and other personal characteristics (e.g., household size, sector of work and location) are taken into account. This is not to say that Nicaraguan immigrants have no economic impact. The evidence suggests they do, but the impact i s small because different effects work in opposite directions and may largely offset each other. On one hand, by providing relatively low-wage labor, Nicaraguan immigrants almost certainly contribute to greater competitiveness of Costa Rica's agricultural sector. This likely has positive effects on economic growth (although the size of this impact i s difficult to measure). On the other hand, inflows of Nicaraguan immigrants have contributed to an increase in the relative supply of low-skilled labor at a time when the relative demand for it was declining. This has contributed to higher wage inequality in Costa Rica, although the size of this effect is not large. Indirectevidence also suggests that immigrant labor may have contributed to the observed rise in unemployment rates among low-skilled workers. Data limitations make it difficult to estimate precisely the size of these different effects. Nonetheless, the evidence suggests that the net impact of Nicaraguan immigrants on poverty has not been significant. Indeed, a much more important cause of stagnating poverty rates has been the growing mismatch between the low skills of all poor workers and the increasing demand for higher-skilled labor in the Costa Rican economy. Ensuringthat the poor acquire more education and greaterjob skills - particularly through improved access and achievement in secondary school - will thus be a key component of a poverty-reduction strategy for Costa Rica. xxii ...nor does immigration appear to affect adversely thepublic health system Concern also has been expressed that Nicaraguan immigration represents a financial drain on Costa Rica's public health services. The evidence suggests, however, that Nicaraguans have not had a significant negative impact on health services, especially once they have been in the country for several years. Data indicate that Nicaraguans account for 4 percent of total demand for health services in Costa Rica and 5 percent of the total health budget - less than their estimated 7-8 percent share of the population. At the same time, differences in insurance coverage between Costa Rican natives and Nicaraguan immigrants are not very large. Around 83 percent of Costa Ricans natives are covered by health insurance, either directly or indirectly (i.e., through a family member), while about 65 percent of all Nicaraguan immigrants are covered. But coverage rates among Nicaraguan immigrants increase significantly over time. For Nicaraguans who have been in the country for five years or more, the difference in coverage rates i s only about 5 percentagepoints. V. Canthe Social Sectors be a More EffectiveForce for Poverty Reduction? Strong social sectors - including education, health, and social protection - are critical to poverty reduction and to national development more broadly. Sound investments in education contribute to poverty reduction both through their positive effects on productivity and economic growth and, when the benefits are broadly shared by the population, through their effects on the ability of the poor to take advantage of new and emerging economic opportunities. Sound investments in health ensure a healthy, active and productive population. And sound investments in social protection serve to reduce people's vulnerability to poverty and increasethe economic mobility of the poor by ensuring that even the poorest families have access to basic services and the ability to invest in human capital. By facilitating human capital development among those who fall outside the reach of traditional (universal) education and health services, social protection programs can contribute directly to higher economic growth as well as to more effective poverty reduction. Public spending on the social sectors is considerable ... Costa Rica has had a long-standing commitment to social development and to universal access to core social programs in education and health. This commitment has been demonstrated, among other ways, by its levels of public spending on the social sectors. At 15.5 percent of GDP, Costa Rica's overall spending on the social sectors i s higher than the Latin American average of 12.5 percent (Figure 4). Only in the case of social assistance does Costa Rica spend below the regional average. This commitment to the social sectors has been an important to a number of Costa Rica's achievements. The country has high health indicators compared to other Latin American and upper-middle income countries. Its social security coverage i s also among the highest inthe region, rivaled only by levels in Chile. xxiii Figure 4: Public Social Sector Spending (Education, Health, and Social Protection) in Latin America (% of GDP; most recent estimate, 2000-2004) 20 w 1Latin America Average low 5 0 0 ow I - ~ 3 s . = . E-= 0 U With two important exceptions, public spending on the social sectors in Costa Rica i s either progressively distributedor "distribution neutral." The benefits of public spending on health and on social assistance are strongly "pro-poor" - that is, the poor receive a higher percentage of the benefits of public spending than their population share. Overall public spending on education is distributed more or less evenly across the income distribution, but there are important differences across levels of education. Public spending on pre-school and primary education i s strongly pro- poor, spending on secondary education i s essentially distribution neutral, and spending on advanced (or tertiary) education i s highly regressive, with the vast majority of the benefits accruing to the non-poor. Spending on social security (health insurance and pension benefits) also is highly regressive. To make the social sectors a more effective force for poverty reduction in Costa Rica, then, a key challenge will be to find ways to make spending on post-primary education and social security more pro-poor. ...but its impact could be improved.. . There is considerable scope for increasing the efficiency and impact of public social spending, particularly in education and social protection. For example, analysis of public spending on education in 16 Latin American countries shows that primary and secondary school outcomes in Costa Rica are significantly lower than would be expected given the level of spending (as a percentage of GDP). In the case of secondary education, cross-country regressions indicate that in 2003, net enrollment rates were about 11 percentage points below what would have been expectedgiven Costa Rica's level of public education spending(Figure 5). xxiv Figure5: Secondary NetEnrollmentRateandPublicEducationSpending inLatinAmerica, 2003 85 80 E 75 -E59i E EI'aaiv.. ^ ' 70 Brazil tloiiwa A I 65 I UIIUII 601F------------ z$' 55 50 Par. I C.R. 8 45 W Y I 40 Nic. I 35 Domi. R. I I 0 2 4 6 8 10 Public Expenditure in EducationasYOof GDP rdjusted R2= 0.37;t and Fvalue significant at p=0.5%, It is important to recognize that Costa Rica's current education outcomes are, in part, a function of recent history -in particular, the adverse impact of the 1982 financial crisis on education sector investment, spending and outcomes - and not only of current spending levels. Duringthe crisis, per-student spending by the Ministry of Education declined by more than 40 percent. While spendingbegan to increaseagain in 1983, increases were slow, and Costa Rica did not reach re,al 1980 per student spending levels again until 1998. While public spending on education in Costa Rica was 5.1 percent of GDP in 2003, it i s important to note that only four years earlier, it was only 3.5 percent of GDP. ...through more efficient public spending on education. .. Nevertheless, there i s considerable scope for improving the internal efficiency of public spending on education, improvements that would strengthen the sector's performance and enhance its impacts on growth and poverty-reducing impact. The Ministry of Education spends the vast majority of its budget on salaries -roughly 93 percent of its budget duringthe 1998-2001 (among the highest shares in Central American). This leaves only around 3 percent of spending for capital expenses and investments, and another 3-4 percent for non-salary, recurrent spending. This means that Costa Rica has very little to spend on expanding education access for the poor, training teachers, providing classroom and school materials, or on other basic needs. While fair and competitive teacher salaries are important, these non-salary expenditures are critical to ensuring educational access and achievement of children from poor households as well as to assuring education quality. There also i s scope for improving education outcomes by reallocating resources across levels of education. Specifically, there would be gains - particularly inpoverty reduction -from increasing relative spending on secondary education. In 2004, Costa Rica allocated roughly 23 percent of its education budget to secondary education - nearly 25 percent less than the average share for Latin American countries (30 percent). To be effective in reducing poverty, it would be important to dedicate a portion of increasedresource allocations to assuring that more poor children enroll in- and complete - secondary schooling of quality. xxv Focusing on improving efficiency of public spending in the social sectors is particularly important since Costa Rica needs to maintain fiscal discipline as a key element of sound macroeconomic management. The new World Bank Country Economic Memorandum for Costa Rica, "The Challenges for Sustained Growth" (2006), finds that building a solid foundation for higher growth and renewed poverty reduction through better public investment will require careful attention to fiscal constraints. In short, there are highreturns - to both growth and poverty reduction -of spending "better" rather than spending more. ...astronger strategicfocus in social protection.. . There also i s clear scope for improving the efficiency and effectiveness of social safety net programs, and making them a stronger force for poverty reduction. At present, the country operates a large number of generally small, largely uncoordinated, and sometimes overlapping safety net programs. A recent review of the country's safety net found that there are at least 47 different social insurance and assistance programs runby over 20 different government agencies. Yet despite the large number of programs, several key areas of risks faced by poor families do not receive sufficient emphasis - most notably programs dealing with early childhood risks and human capital development among poor children, which are critical to the fight against poverty. Consolidation and rationalization of Costa Rica's many social protection programs based on a more strategic focus - on areas such as early childhood and human capital development of the poor -should yield highreturns. Moreover, a significant proportion of the poor still fall outside the safety net. For example, despite the country's accomplishments with respect to social security coverage, analysis of the 2004 EHPM data indicates that 26 percent of all poor and 31 percent of the extremely poor still lack basic health insurance coverage. Similarly, benefit-incidence analysis of Costa Rica's main social assistance programs shows that 84 percent of the eligible poor still lack access to the country's child care centers (Centros Znfantiles), 57 percent of poor households have not benefited from the Family Housing Vouchers (BonoFamilia Vivienda), 46 percent of the eligible poor have not received social assistance pension benefits (Pensiones no Contributivas),and 32 percent of the eligible poor still lack access to Costa Rica's School Feeding Program (Comedor Escolar). ...and better targeting of CostaRica's poorest, most vulnerablefamilies While these (and similar) programs would make an important difference to the lives and livelihoods of poor families, eligibility currently i s not necessarily limited to them. Costa Rica's poverty reduction efforts would benefit from increased targeting to ensure greater access to the country's safety net programs by the poor and extremely poor, who for reasons of poor physical access or due to family financial constraints are either not covered or have not been successful in obtaining access to Costa Rica's universal programs. Well-targeted programs also can help conserve scarce budget resources, by focusing spending on the poorest, most vulnerable households. As in the case of education, the evidence suggests that focusing on areas with highest returns (programs to support human capital development among poor children, for instance) and by making more effective use of program targeting (as a complement to universal programs), Costa Rica's safety net could have a significant impact on poverty within existing budget envelops. While health outcomes have generally been good, even among the poor, and returns to public health-sector spending have been relatively high, there still remains scope for improving the effectiveness of health spending in the face of emerging risks. Specifically, recent, significant xxvj increases in the incidence of malaria, dengue, and tuberculosis, along with declines in measles and poliomyelitis vaccination rates, have highlightedimportant service gaps in the public health system. As in the cases of education and social protection, internal reallocation of the resources toward vaccinations and preventive health measures, along with public information campaigns will be important to avoid further deterioration of Costa Rican's health status and to ensure that the poor do not bear the brunt of the associatedhealthcosts. There are largepotential gains to spending "better" Insum, the fact that Costa Rica has a strong historical commitment to education, health, and social protection puts it on strong footing to use investment in the social sectors as a powerful tool for renewed poverty reduction. Nonetheless, the evidence suggests that there is considerable scope for increasing the impact of social sector policies on poverty through concerted efforts to increase access to quality services by the poor, particularly in education and social protection. The evidence also suggests that, given the levels of public spending, significant gains could be made even within current budget settings through strategic reallocations of resourcestoward areas of high impact (e.g., secondary education for all Costa Ricans), through greater internal efficiency of public spending, and through the strengthened use of targeted approaches to reaching the poorest, most vulnerable Costa Ricans as a complement to Costa Rica's universal programs. Given Costa Rica's current fiscal situation, spending existing resources better will be critical to its success in harnessing the social sectors for effective poverty reduction and in increasing economic growth. VI. Data, Poverty Measurement,andPoverty Reduction Strengthening data and poverty monitoring remain an important challengefor Costa Rica Good data and information systems play a critical role in the fight against poverty by making it possible to monitor the incidence of poverty and devise evidence-based policy responses. The main information source for poverty monitoring in Costa Rica, the EHPM, was originally designed as a labor force survey, and thus is a relatively weak instrument for poverty measurement. For example, it does not collect consumption data, generally considered the best measure of well-being for poverty analysis since it fluctuates less than income and tends to be measuredwith less error. In addition, the income aggregate computed from the EHPM is almost entirely composed of cash income (it asks only one question about non-cash income). Inaddition, cash income is computed using only eight questions - significantly fewer than is general considered adequate to generate a robust income aggregate. Such few questions could be expected to lead to relatively large reporting errors. Inan attempt to address this shortcoming, INEC, the Costa Rican statistical agency, has made several adjustmentsto the income aggregate - although it is not clear how well these adjustments compensate for the lack of complete information. Construction of the poverty lines also includes some questionable assumptions that may lead to inconsistencies in poverty measurement over time. Sensitivity analysis carried out as part of this study suggests that measurement issues are important. Specifically, measured poverty levels are found to be sensitive to adjustments in the income aggregate and to changes in assumptions regarding the setting of the poverty line over time. Changing severalcurrent assumptions can lead to as muchas a 2.6 percentagepoint drop in the poverty rate from currently measured levels. While this i s not a large number in absolute terms, it represents more than 10 percent of current poverty levels. Perhaps more importantly, changing the current assumptions leads to changes in relative poverty levels in rural and urban areas; combined adjustments lead to a fall in measured poverty in urban areas, but an increase in xxvii measured poverty in rural areas. Such simulations bring Costa Rica's measured rural-urban poverty differentials more in line with regional norms. More importantly, if correct, this change in the rural-urban poverty balance could have important implications for sub-regional perspectives on poverty reduction policy in Costa Rica. For these reasons, further work i s urgently needed to assess and, as necessary, strengthen both Costa Rica's household survey instrument and its poverty measurement methodology. The soon- to-be-released 2004 Income and Expenditure Survey (IES) for Costa Rica provides a much richer set of household survey data than i s available via the yearly EHPM surveys. It would be important, therefore, to analyze the IES survey closely to clarify pending empirical and methodological questions with respect to poverty measurement and poverty monitoring in Costa Rica. VII. Policy Options: A Strategy for Recapturing Momentum for Poverty Reduction The evidence arguesfor a multi-dimensional strategy to reducepoverty... Together, the evidence suggests that there i s considerable scope for Costa Rica to recapture its momentuminthe fight against poverty. Specifically, the evidence suggests that to ensure that the poor are able to participate in and benefit from future socio-economic progress, a multi- dimensional strategy for poverty reductionis warranted. Many of the priorities identifiedhere are consistent with those expressed by the new government administration, including emphasis on strengthening the human capital of the poor and improving the outreach and impact of Costa Rica's social safety net,. Elements of such a strategy would include: 0 Promotinghigher sustained levels of economic growth Strengthening the human capital of all Costa Ricans, with emphasis on the poor Ensuringsocial protection coveragefor the poorest, most vulnerable groups Creating an enabling environmentfor poor workers Defining regionally differentiated investment strategies that reflect regional differences in patterns of poverty 0 Strengthening information systems for improved poverty monitoring, effective program targeting and evaluation of program and policy impact Promoting economic growth. A large body of global evidence shows that assuring sustained economic growth i s critical poverty reduction in the long-term. Indeed, economic growth i s critical to providing expanded economic opportunities for all Costa Ricans, including for the poor. Efforts to promote sustained growth are critical to any country's poverty reduction efforts. But what will be the keys to faster growth in Costa Rica in the coming years? Recent empirical analysis has highlighted several key factors behind Costa Rica's growth performance in the 1990s. Among the most important were: investments in education, investments in infrastructure, strengthening of the country's financial sector, and increases in trade openness. The new World Bank Country Economic Memorandum for Costa Rica (2006) indicates that progress in these areas will continue to be important for growth in Costa Rica in the 2000s. In addition, the report highlights the importance of strong macro-economic and fiscal management, as well as efforts to xxviii strengthen the country's competitiveness through strengthening its system of research and innovation. As Costa Rica's recent experience indicates, however, growth alone will not engender poverty reduction. It will be critical to create conditions in which the poor are adequately prepared to take advantage of new and emerging economic opportunities. For that reason, efforts to strengthen the human capital of the poor, increase social protectioncoverage for the poorest and most vulnerable groups, and provide social supports such as affordable child care to enable poor working mothers to participate more fully inthe labor market would enhance any poverty reduction strategy. Strengthening human capital, with special emphasis on the poor. This is a key part of any country's development strategy, and should be at the core of any strategy to re-invigorate poverty reduction in Costa Rica. Key priorities include: Policies and investments to increase enrollment and graduation at the secondary school level, particularly among the poor. A fundamental element of any initiative to improve secondary school outcomes among the poor would be efforts to improve secondary school relevance and quality, particularly in poor, rural areas. This will require reallocation of education-sector resources toward secondary schooling, along with increases in the share of non-salary, recurrent spending in education, whether on quality- enhancing measures (both at the primary and secondary levels) or on efforts to improve secondary school access and achievement among the poor. Efforts to improve the capacity and employability of poor, low-skilled workers. Here the objectives would be to: bridge the skills gap between poor workers and the growing demand for higher skilled workers in the Costa Rican economy, and in doing so, to reduce structural unemployment, an important contributing factor to poverty. Findingthe right type of programto strengthen the labor market skills of poor, adult workers will be a challenge, however. International evidence suggests that only a small proportion of skills-training programs actually raise people's employability and earnings. It will be important, therefore, to identify effective approaches to strengthening people's job skills, drawing on international best practice. It also will be critical to monitor and evaluate Costa Rica's efforts to increase workers' skills to identify which specific program delivery models are most effective. Directing resources to address emerging health risks. Health sector resources should be aimed toward reversing the decline in vaccination rates for measles and poliomyelitis, and at preventive health measures and public information to deal with at malaria, dengue, and tuberculosis, in order to avoid deterioration of Costa Rican's health status, particularly among the poor. Ensuring social protectionfor the poorest, most vulnerable groups. Among the key functions of a social protection system are to reduce people's vulnerability to poverty, and increase the economic mobility of the poor by ensuring that even the poorest families have access to basic services and the ability to invest in human capital. To enhance the ability of Costa Rica's social protectionprograms to fulfill these functions, it will be important to: Strengthen the strategic focus of Costa Rica's social protection system on areas of particularly high returns, such as on early childhood interventions and programs that increase human capital development among poor children. xxix Rationalize and consolidate existing social-assistanceprograms to increase the impact of public spending, consistent with strategic priorities for poverty reduction. Expand social protection coverage (including support for human capital development, and health insurance and pension coverage) to the poor and extremely poor who currently fall outside the system. This can be accomplished through greater emphasis on targeting as a complement to universal programs and through strengthening program targeting mechanisms, such as the SIP0 database. With greater strategic focus and more effective targeting, Costa Rica should be able to increase social protectioncoverage significantly, even within the current real budget. It should be noted that there is a special role for "demand-side'' programs to help poor families overcome financial constraints to investing in their children. Demand-side programs, such as conditional cash transfers, provide .financialsupport to poor households provided they ensure that their children attend (and complete) secondary school. They have been effective in raising educational enrollment and attainment among the poor in such Latin American countries as Mexico, Colombia, and Brazil. Costa Rica has some experience with this type of program (for instance, CreciendoJuntos, Construyendo Oportunidades),but its programs have operated on a relatively small scale and, thus, their impact has been limited. Creating an enabling environmentfor poor, female workers. The evidence suggests that many poor women work part-time because they cannot find adequate or affordable child care arrangements. Current Costa Rican labor market legislation also limits women's flexibility to seek employment outside standard working hours. Creating an enabling environment for poor female workers would thus entail: 0 Providing greater social support, in the form of affordable child care options, to female workers, whether poor single mothers or female spouses, so that they can work full-time and generate higher earnings. Several policy options - alone or in combination - could serve to reduce women's child care-related constraints to employment, including: (i) expanding access to and participation in early childhood developmendpre-school education; (ii)expanding government subsidies to poor families for child care; (iii) providing before- and after-school child care programs in schools; and (iv) encouraging private firms to provide subsidized day care facilities at work. Reducing legal barriers to women working non-standard hours - by providing them the same working-hour flexibility as men - would make it easier for single, working mothers to seek employment during times when it may be easier to find alternative childcare arrangements (from extended family members, for instance). Defining regionally differentiated investment strategies. Most of the elements of a poverty reduction strategy - a commitment to strengthening the human capital of the poor, increasing availability of affordable childcare, and targeting support programs to the extreme poor - can appropriately be undertaken as national-level initiatives. Nonetheless, differences across regions in the levels of poverty and concentrations of poor people suggest that regionally differentiated policies and investments may be warranted in some cases. For example: Inareas like the Central Region, where the poverty rate is low but concentrations of poor people are high, investments in infrastructure and steps to improve the investment climate can be particularly effective. Through their impacts on business investment, such xxx approaches may be effective in generating employment among the poor by increasing demand for relatively low-skilled labor. Inplaces like the BruncaandChorotegaRegions, where the poverty rate is highbut the concentrations of the poor are low, investments or targeted support for education, training, and technical assistance, all of which raise people's economic mobility, may be more effective (and cost effective). Strengthening information systems. Costa Rica's poverty reduction efforts would benefit from strengthening and increasing the transparency of data, information and management systems, specifically by: Improving poverty measurement and monitoring. It will be important to strengthenCosta Rica's household survey, the EHPM, and to assess the strengths and weaknesses of the country's current poverty measurement methodologies to ensure that the poverty monitoring system can provide suitable empirical support to policy makers in developing poverty reduction strategy. The forthcoming availability of the 2004 Income and Expenditure Survey (IES) for Costa Rica provides a valuable opportunity to clarify a number of pending empirical and methodological questions regarding poverty measurementand poverty monitoring. Strengthening program targeting mechanisms. As noted above, there are potentially large gains to strengthening the information systems used to target programs to ensure that the poor and extremely poor have access to basic human services and adequate protection against risks. Several years ago, Costa Rica developed the SIPO database, which contains a registry of the poor based on a proxy means test. The SIPO database is not nationally representative, however, nor has it been updated to reflect households' movement out of (or into) poverty. Moreover, to date, it has not been adopted for use by many government agencies. Costa Rica's poverty-reduction efforts would benefit from efforts to strengthen and update the SIPO database or other instruments that would permit identification and targeting of the poor, along with efforts to make the information more widely available to the range of social programs focused on the poor. Developing institutional mechanisms for results monitoring and evaluation of interventions for more effective program management. Costa Rica does not yet have a system for monitoring the results of its social and poverty reduction programs. Developing such a system (and a supporting institutional structure) would go far in helping to re-invigorate the country's poverty-reduction efforts. In addition, developing the capacity to evaluate the impact of key social programs would help the Government and other stakeholders to identify effective (and cost effective) interventions. A strong monitoring and evaluation system would help policymakers ensure that interventions are producing the expected results and, if necessary, enable them to undertake mid-course corrections in programdesign or programtargeting to produce the desired outcomes. Improving transparency and public access to government information on poverty-related programs. This would help to create greater "democracy of information" in Costa Rica and increase accountability among government institutions involved in the fight against poverty. xxxj PART 1:THE NATUREAND EVOLUTION OFPOVERTY INCOSTA RICA, 1989-2004 1. INTRODUCTION 1.1 Costa Rica has made considerable progress in increasing income, reducing poverty and improving social indicators since the start of the 1990s (Table 1.1). For example, between 1990 and 2003, its average per capita GDP grew from US $3,151 to $4,410, an increase of roughly 40 percent. According to official estimates, the extent of poverty declined from 31.7 percent of the population in 1989 to 23.9 percent in 2004; extreme poverty declined from 9.9 percent to 6.6 percent. Primary and secondary school enrollment rates have increased over the period, as has life expectancy. Infant and child mortality, as well as child malnutrition, have declined from already low levels. Indicator Early 2000s Change 1 1990s GDP per capita, 1990-2003 (2000 Constant US $) 1 3,151 I4,410 1 40% increase I Poverty Headcount, 1989-2004 (Total, % of Persons) 31.7% 23.9% 25% decline Poverty Headcount, 1989-2004 (Extreme, % of Persons) 9.9% 6.6% 33% decline Net Primary Enrollment Rate, 1989-2004 (%) 75% 81% 9% increase Net Secondary Enrollment Rate, 1989-2004 (%) 32% 47% 48% increase Life Expectancy, 1990-2003 (years at birth) 77 79 7% increase Infant Mortality Rate (deaths per 1,000 live births) 1990-2003 15% 8% 47% decline Under-5 Mortality Rate (deaths per 1,000 live births) 1990-2003 17% 10% 41% decline Malnutrition prevalence, weight for age (% of children under 5) 6% 5% 17% decline 1982-1996 Malnutrition prevalence, height for age (% of children under 5) 11% 6% 45% decline 1985-1996 1.2 The country also performs well in health outcomes compared to other countries in Latin America and the Caribbean region and to countries with similar income levels (Table 1.2). Infant and child mortality rates are significantly lower than in comparable countries, while average life expectancy is substantially higher. Access to improved water sources i s nearly universal (97 percent) - also high relative to other countries. But Costa Rica does not compare favorably to other countries in its region in education outcomes (as measured by net enrollment rates), particularly at the secondary school level. In2002, for example, its net secondary enrollment rate was only 53 percent, compared to 67 percent on average in Latin American and Caribbean countries and 78 percent in upper-middleincome countries (Table 1.2). 1 Indicator Costa Latin America Upper-Middle Rica and the Income Caribbean Countries bet Primary Enrollment Rate, 2002 (%)* 90 93 91 bet Secondary Enrollment Rate, 2002 (%)2 53 67 78 11 Life Expectancy, 2003 (years at birth) 79 72 70 LnfantMortality Rate, 2003 (deaths per 1,000 live births) 1III 8 111I .25 111I 21 II Under-5 Mortality Rate, 2003 (deaths per 1,000 live births) 10 31 27 Access to an ImprovedWater Source, 2002 (% of population) 97 91 91 1.3 In spite of overall social progress and relatively low poverty numbers, Costa Rica faces several critical challenges. First and foremost, the incidence of poverty has essentially stagnated since 1994. Indeed, while the portion of the population living in poverty (the "headcount rate") declined steadily from 31.7 percent in 1989 to 22.9 percent in 1994, according to official estimates, the figure has hovered at or above 23 percent since that time (INEC, EHPMPrincipales Resultados, several years). This is all the more troubling since real per capita income' growth averaged 2.4 percent per year over the 1994-2004 period. A significant and growing empirical literaturefinds there is consistent and positive relationshipbetween positive economic growth and poverty reduction.* Moreover, Costa Rica's relatively low performance in secondary school enrollment and graduation rates -including continuing gaps in education access between the poor and the non-poor - have occurred even though public spending on education and on the social sectors more generally has long been above average for Latin America. How to re-invigorate the fight against poverty and make social sector spending more effective are thus important policy questions. 1.4 This Poverty Assessment addresses these and other pertinent policy questions in order to enhance the ability of the new Govemment of Costa Rica to reduce poverty through public policy. It will focus on understanding better how to strengthen the links between economic growth and poverty reduction in Costa Rica and how the Government of Costa Rica can more effectively use social policy to achieve poverty reduction goals. Specifically, the Poverty Assessment: presents a dynamic and multi-dimensional profile of poverty in Costa Rica, focusing on the evolution of poverty since 1989 and the key individual-, household- and community- level factors associatedwith being poor; analyzes the relationship between economic growth, inequality and poverty to better understand the macro-economic and sectoral factors underlying the lack of progress over the last`decade; I Measuredas the per capita Gross Domestic Product. See, for example, Bourguignon 2003, Dollar and Kraay 2002, Kraay 2004, Ravallion and Chen 1997, and Ravallion 2001. 2 0 analyzes how incomes are transmitted to the poor and non-poor via the labor market and how labor market policies and programs can be more effective in supporting poverty reduction; e analyzes the role and effectiveness of social policies and programs - particularly in education, health, and social protection-in improving the welfare of the poor; and e formulates recommendations for public policy by which the Government of Costa Rica can increase the effectiveness of economic and social sector policy in the fight against poverty. 1.5 Emphasis will be given throughout the report to the status and role of Nicaraguan immigrants-in the context of poverty and poverty reduction. In addition, because Costa Rica's household survey was not originally designed for the purpose of poverty monitoring - it was designed essentially as a labor market survey - special emphasis will be given to data and measurement issues to evaluate whether Costa Rica's statistical systems can be strengthened to better support policy-orientedpoverty analysis. 1.6 This Poverty Assessment is part of a continuing analytical and advisory program undertaken by the World Bank in coordination with subsequent governments of Costa Rica. The report is based on several newly commissioned background studies, along with other recent empirical analysis of poverty and social policy in Costa Rica. It also draws and builds upon several recent World Bank analyses on growth and economic and social policy, including Costa Rica: Social Spending and the Poor (World Bank 2003) which analyzed social sector spending and social outcomes in Costa Rica and provides recommendations for strengthening social policy; and the new Country Economic Memorandum (CEM) (World Bank 2006) and Investment Climate Assessment (ICA) (World Bank 2006) for Costa Rica, which focus on promoting sustainable growth and economic competitiveness through public policies and investments and a strengthenedinvestment climate. 1.7 The work on the Poverty Assessment also has benefited from a number of discussions with Costa Rican government officials, academicians, representatives of non-governmental think tanks and civil society groups. 1.8 The study is organized into three main parts. The first part of the report (chapters 1-4) focuses on the nature and evolution of poverty from 1989 to 2004. This includes presentation of a multi-dimensional profile of poverty (chapter 2), an analysis of the links between growth, income inequality and poverty reduction in Costa Rica (chapter 3), and an analysis of the role of the Costa Rican labor market in determining the earnings and opportunities faced by the poor (chapter 4). Part I1(chapters 5-8) examines the role of the social sectors and social sector policies on poverty reduction in Costa Rica, focusing on education (chapter 6), health (chapter 7), and social protection (chapter 8). The final part of the report focuses on recapturing Costa Rica's momentum in reducing poverty. This part focuses first on strengthening economic and social policies to reduce poverty (chapter 9) and then on improving data for poverty monitoring and results-based management of Costa Rica's poverty reduction efforts. 3 2. RECENT PROGRESS, CURRENT CHALLENGES 2.1 This chapter presents a multi-dimensional profile of poverty for Costa Rica, based largely on analysis of the country's household survey from 1989 to 2004. It discusses the evolution of income poverty at the national level, across rural and urban areas, and across Costa Rica's six planning regions; examines a number of non-income measures of well-being, including human capital, control of physical assets and access to basic services; analyzes the correlates of poverty in Costa Rica; and examines the evolution of income inequality over the period. Several main messages emerge from the chapter. First, the analysis shows that following significant progress in poverty reduction between 1989 and 1994, income poverty rates have essentially remained static through 2004. Second, a number of non-income measures oflwell-being have improved throughout the 1989-2004 period, although important gaps still remain between the poor and the non-poor in access to basic health and human services, including education and insurance coverage. Third, several factors are found to affect a household's probability of being poor. Controlling for other factors; households with more children and ones whose heads have low education levels, are female, or work in agriculture or the informal sector all have relatively high likelihoods of being poor. There also are instinct regional patterns of poverty. Finally, the analysis shows that income inequality has increasedin Costa Rica since 1989. Measuring Poverty in CostaRica 2.2 Poverty can be measured using different welfare indicators. Traditionally, income, consumption or basic needs indicators are usedto classify people as poor or not poor. Costa Rica uses its annual Multi-Purpose Household Survey (EHPM3)to measure poverty by computing the per capita household income and comparing it to two separate poverty lines. The EHPM is mainly a labor survey that includes some household and personal characteristics as well as income information. It was not originally designed to measure poverty and has some limitations for poverty analysis, which will be discussed later in the report. Nonetheless, to enable comparability with Costa Rica's official poverty statistics, the official income aggregates, poverty lines and poverty classifications generated by Costa Rica's National Statistics and Census Institute's, INEC, are used(see Box 2.1 for more details). Trendsand Patterns of Poverty in CostaRica 2.3 Analysis of the EHPM data show that poverty in Costa Rica declined significantly from 1989 to 1994, but that it has remained essentially constant since then. 35% Figure 2.1: Headcount Dovertv in Costa Rica *. +All Poor Specifically, the poverty 30% headcount (the proportion 2 25% - 7 of the population that is f 20% poor) declined from 32 percent to 23 percent over the 1989-1994 period. Since 1994, the incidence of poverty has basically 0% 1 I I I I I hovered around 23 percent 1989 1994 2000 200tea$002 2003 2004 2005 (Figure 2.1).4 Extreme Encuestade Hogaresde PropdsitosMbltiples. 4 2005 poverty figures publishedby INEC 4 poverty in Costa Rica has followed the same trend; the incidence of extreme poverty declined from 9.9 percent to 6.8 percent between 1989 and 1994, and has remained at or near 6.5 percent since then. Box 2.1: Measuring Poverty in Costa Rica The Income aggregate i s computed using eleven questions concerning an individual's first and second jobs, self employment income and non-lab(or income. Total income i s adiusted by adding a constant percentage to compensate for underreporting and other income not captured in the ~ u r v e y . ~The adjustment -Salary -Income corresponds to an increase of 17.4 percent -Social Security & bank Non Labor Income for urban households and of 35.5 percent Deductions -Pensions for rural households.6 Once the adjustment -Other deductions -Subsidies i s made, the total monthly income for each Self Employed -Scholarships household member is computed. Then, -Net Income -Other Cashtransfers household per capita income is obtained by -Productsusedat home I -Interest & rents adding up the income of each household Source: INEC, September 2003 member and dividing it by the household size. The extreme poverty line i s the value of a Basic Food Basket with 2,230 Kcal per day for urban areas and 2,316 Kcal for rural areas. The composition of the Basic Food Basket was drawn from the 1987-1988 National Household Income and Expenditure Survey, based on the average consumption patterns of the second, third and fourth deciles for the urban Basic Food Basket and deciles four, five and six for the rural Basic FoodB a ~ k e t .For ~ 2004 the monthly nominal value of the extreme poverty lines was C. 16,452 for urban households and C. 14,489 for rural households. Below this level, an individual can not meet the minimum required caloric intake, even if all his or her income were usedfor food consumption. The overall poverty line is computed by multiplying the Basic Food Basket value by the inverse of the ENGELS coefficient' computed with the same survey used for the Basic Food Basket estimation. For urban areas, the inverse of the coefficient is 1.97, and for rural areas 2.07. For 2004 the monthly nominal value of the overall poverty line was C. 35,866 for urban households and C. 28,543 for rural households. Below this level, an individual cannot meet his or her minimumcaloric need,plus basic non-food needs.' Construction of the poverty lines includes some questionable assumptions that may lead to inconsistencies in poverty measurement over time. Sensitivity analysis carried out as part of this study suggests that measurement issues are important. Specifically, measured poverty levels are sensitive to adjustments in the income aggregate and to changes in assumptions regarding construction of the poverty line over time. Changing several current assumptions can lead to as much as a 2.6 percentage point drop in poverty headcount or a reduction more than 10 percent from current poverty levels. Perhaps more importantly, changing the current assumptions leads to a fall in measured poverty in urban areas, but an increase in measured poverty in rural areas. Such simulations bring Costa Rica's measured rural-urban poverty differentials more in line with regional norms. More importantly, if correct, this change in the rural-urban poverty balance could have important implications for sub-regional perspectives on poverty reduction policy in Costa Rica. The objective of the adjustment i s to compensatefor information on income not asked for in the questionnaire (Le. own housing use value) and for under-reported income. The methodology to estimate the adjustment value used as a reference the Gross Domestic Product for the base year. The percentagesare computed over the income sources captured inthe survey. * The Engels INEC Principales Resultados Encuesta de Hogares de Prop6sitos Mliltiples, Julio 1996 Coefficient is the ratio of food expenditures to total expenditures. Poor households include extreme poor households. 5 2.4 In 2004, the total number of poor persons in Costa Rica was estimated at almost one million, with those living in extreme poverty estimated at just above a quarter of a million." A decline in poverty rates between 1989 and 2004 compensated for population increases duringthe same period. The net result was a small decrease in the total number of extremely poor persons from 294,000 to 280,000 (a 5 percent reduction) and a small increase in the total number of poor persons from 944,000 to 1,015,000 thousand (a 7 percent increase). 2.5 It is worth noting that Costa Rica has relatively low levels of poverty for its region - indeed, they are the lowest in Central America. This i s the case whether poverty i s measured using national poverty lines or internationally comparable dollar-a-day poverty lines. While national poverty lines are not strictly comparable across countries, the differences are striking. The existing data indicate that official national poverty rates in the rest of Central America are twice as high, on average, as in Costa Rica; official national extreme poverty levels in Central America are, on average, two-and-a-half times higher than in Costa Rica (Table 2.1). Country Costa Rica Guatemala Honduras Nicaragua Panama ElSalvador Year ++ (2004) (1998) (2004) (2001) (2003) (2002) Overall Poverty 23.9% 56.2% 50.7% 45.8% 37.2% 37.2% Extreme Poverty 6.6% 15.7% 23.7% 15.1% 16.7% 15.4% 2.6 Using internationally comparable $1and $2 "purchase power parity" (PPP) poverty lines (that is, defining as poor anyone in a household with income of $1 or $2 per day per person), Costa Rica has one of the lowest poverty headcounts in Latin America (see figure 2.2). For international comparisons, the use of the $1 and $2 (PPP) i s considered a best practice. Usinga $2 per day poverty line, Costa Rica has a poverty headcount of 9 percent - well below the rest of Central America and half the Latin America average of 25 percent. Usinga $1poverty line, the poverty headcount in Costa Rica is only 2 percent, one-fifth of the Latin American average (10 percent). Figure 2.2: LatinAmerica $2 PPP/day Poverty 90 I 1 loThe total number of poor and extremely poor was estimated by multiplying the headcount rates by census projections. 6 2.1 Costa Rim2004 Po nt by NICARAGElA ATLAiVTlC its extremely poor. This i s due to the high population concentration in the Central region - 63.5 percent of all Costa Ricans live there. As a result, the Central region has the highest contributions to poverty and extreme poverty; 48.9 percent and 38.9 percent of the poor and the extremely poor live in the Central Region, respectively (Table 2.2). This implies, in turn, that just over half of the poor and 60 percent of the extremely poor lives in the other five regions combined. These regional differences in incidence as opposed to the density of poverty may have implications for policy and investment approachesto reducingpoverty at the local level. % of Poverty Headcount Contribution to Poverty POPUlatiOn All Poor 1 Extreme Poor II All Poor I Extreme Poor AREA Urban 58.4% 20.8% 4.7% 50.7% 41.3% Rural 41.6% 28.3% 9.3% 49.3% 58.7% 'The central region includes Costa Rica Central Valley with the four biggest cities of the country: SanJose, Alajuela, Heredia and Cartago. Source: Costa Rica 2004 household survey (INEC) 2.10 Poverty rates in Costa Rica also vary between urban Figure 2.3: Costa Rica Poverty Evolutionby Regions and rural areas - although not as much as across planning 60% regions. As elsewhere in Latin 7-- America, poverty is higher in rural areas than in urban areas, 40% 14-Chorotega but the difference is not as large 30% -A- HuetarNorth -o- PacificCentra as might be expected. Rural .. ... 20% -. poverty was measured at 28.3 \e +HuetarAtlantic * " A n Y percent in 2004, compared to 10% - 20.8 percent in urban areas 0% 4 (Table 2.2). The ratio of 1989 1994 2000 2001 2002 2003 2004 poverty headcounts between the highest and lowest poverty planning regions i s 2.4, while the ratio between the rural and urban poverty rates is only 1.4.16 2.11 The rural poor account for only 49 percent of total poverty and 59 percent of extreme poverty. This is relatively low by Central American standards, but i s consistent with the relatively small differences in the incidence of poverty in rural and urban areas in Costa Rica. Rural poverty predominates in most of the rest of Central America; excluding El Salvador, the rural l6I t is recognized that part of these conclusions are driven by the methodology chosen to determine poverty status. A section in this chapter does a simulation to evaluate the importance of the methodology used. 8 poverty accounts for between 68 and 81 percent of total poverty, while rural extreme poverty represents 79 percent to 93 percent of total extreme poverty (See Table 2.3). The higher contribution of rural poverty to total poverty in Central America i s a function of both the higher proportion of rural populations in other countries (again, with the exception of El Salvador), and significantly higher incidences of poverty inrural areas compared to urban areas. Table 2.3: RuralPoverty concentrations by country I ICosta IGuate 1Hondur INicarag Panama - I IEl Rica mala as ua (1997) Salvador (2094) (1998) (2004) (2001) (2002) Rural Population (%) 41.6% 61.4% 51.7% 48.1% 44.4% 41.0% Contribution to overall poverty 49.3% 81.4% 73.7% 67.5% 77.3% 54.9% Contribution to extreme poverty 58.7% 93.1% 86.1% 78.8% 90.8% 65.2% Poverty concentration rate 1.19 1.33 1.43 1.40 1.74 1.34 Ext. poverty concentration rate 1.41 1.52 1.67 1.64 2.05 1.59 *' Ruralcontributionto extreme overallpoverty/ Ruralpopulation Ruralcontributionto povertyI Ruralpopulation 2.12 Poverty in Costa Rica also differs by region in depth and severity. In general, the depth and severity of poverty tend to be worse in areas with higher poverty headcounts. Indeed, the Brunca and Chorotega regions not only have more than double the national poverty rate and more than three times the national extreme poverty rate, but poverty in these regions i s deeper, with the highest poverty gap indexes17 for overall poverty (twice the national value) and for extreme poverty (more than triple than the national value) (Table 2.4). The Severity Indexes'* show very similar patterns to the Poverty Gap Indexes for extreme and overall poverty across regions. In sum, Bruncaand Chorotega not only have the highest ratesof poverty, but the averageincomes of poor households in these regions are further below the poverty line, on average, than in other regions of the country. 2.13 Average poverty figures also conceal differences in poverty levels across different types of households; for example, poverty is relatively high in female-headed households. Households headedby women make up 23.4 percent of the Costa Rican population, up 57 percent (from 14.9 percent) from just 15 years ago. Indeed, the poverty rate is 27.7 percent in female-headed households, compared with 19.6 percent in households headed by men. At 8.1 percent, the incidence of extreme poverty in female-headed households i s nearly 75 percent higher than in male-headedhouseholds, which have an extreme poverty incidence of 4.7 percent. The poverty Gap Index is the difference between income and the poverty line of the poor expressed as a roportion of the poverty line value and the population for each group. The poverty severity Index is a derivation of the Poverty Gap Index and takes into account the distribution of total consumption among the poor. 9 Table 2.4: Poverty, Average Income Poverty Ga? Index and Severity by regions and area Poverty # of poor persons Poor average income Poverty Gap Severity Headcount (thousands) (C./month) Index Index NATIONAL 6.6% 279.2 9,733 2.4 1.3 ICentral -........ 4.0% 109.9 9,972 1.5 0.8 Chorotega 13.6% 44.6 8,697 5.6 3.0 - - ._ ..- . _ _ _ 1.4 -. Rninri 15.3%I 50.0 1 9,895 I 5.2 I 2.6 I Huetar Atlantic 6.9% 28.8 9,621 2.5 1.3 Huetar North 10.9% 25.1 9,600 3.8 1.9 4.7% 115.4 10,874 1.6 0.8 AREA urban Rural 9.3% 163.8 8,930 3.6 1.9 2.14 As elsewhere in Central America, poor households in Costa Rica tend to be larger than non-poor households- although the difference is smaller. On average, households are smaller in Costa Rica than in other Central American countries. Indeed, the average Costa Rican household size is between 15 percent and 25 percent smaller than in other Central American countries (Table 2.5). The difference between household size of the poor and non poor i s also the smallest inthe region. Onaverage, extremely poor householdsinCosta Ricahave only 1.1more members than do non-poor households. Like the difference in family size, this figure i s decreasing over time. For other Central American countries, the difference in size between extreme poor and non-poor householdsranges from 1.9 membersto 4.4 members. Table 2.5: Household Size bv Povertv G r o w in Central America Country 3 Costa Rica' Panama Honduras Nicaragua Guatemala Year 3 2000 2004 20042 2004 2001 2000 ExtremelyPoor 5.9 5.6 8.7 7.3 7.4 8.4 All Poor 5.7 5.1 7.3 6.8 6.7 7.4 NonPoor 4.7 4.5 4.4 5.4 4.6 5.2 NATIONAL 4.9 4.6 5.4 6.1 5.3 6.4 * Source:LSMStype survey 10 Non-income measures of well-being 2.15 Poverty and well-being have a number of dimensions beyond the level of household per capita income. The sections that follow focus on poor people's ownership and control of assets such as housing, and their access to basic services like education, health, potable water, and collection of garbage by the poor. While most indicators of assets and basic services were already high for Costa Rica by regional standards at the beginning of the 1990s, the ccuntry has registered further improvements over time in most cases. Housing 2.16 Most of housing conditions in Costa Rica have improved over time, both for the overall population and for the poor. In addition, gaps between the extremely poor, the poor, and the national average have decreasedfor most indicators (see figures 2.4 to 2.8). With the exception of precarious housing tenancy (precario),the improvements are impressive. From 1989 to 2004, the proportion of households living in shacks, cooking with firewood, and having dirt floors in their houses declined by half or more. The percentage of houses with metal roofs increased from 90 percent to 98 percent; moreover, since 2000 there has been no difference in the proportion of extremely poor, all poor, and all households whose houses have metal roofs. I I Figure 2.4: Costa Rica Housing Figure 2.5: Costa Rica Housing Housetvae: Shack Housetenancv: "Drecario" 1 1989 1994 2000 2002 2004 1989 1994 2000 2001 2002 2003 2004 Year Year I-Ext. Poor -All Poor- 3- -National ,+,Ext. Poor -All Poor - 4- =National Figure 2.6: Costa Rica Housing Figure 2.7: Costa Rica Housing Cooking: with firewood or coal Floor tvDe: dirt 80 .3 28m 60 c) 40 rr 20 e0 o 1989 1994 2000 2002 2004 1989 1994 2000 2002 2004 Year Year -Ext.Poor -AllPoor --t- .National /-Ext.Poor -AllPoor - -t- .National 11 Figure 2.8: Costa Rica Housing Roof: Metal (zinc) x)O g 95 .e ;;J 90 sg 85 cr P, 80 0 75 1989 1994 2000 2002 2004 Year ~-Ext.Poor -All Poor- -+- .National Source: Nationalhouseholdsurvey, INEC CostaRica 2.17 Access to housing-related services such as potable water, waste removal, electricity, garbage collection and phone service also has improved over time for the entire population, including for the poor and for the extremely poor (Figures 2.9-2.14). Indeed, access to pipe water, treated water, sewers or septic tanks, and garbage collection improved dramatically from 1989 to 2004. The proportion of extremely poor households lacking piped water decreased from 28 percent in 1989 to 8 percent in 2004, for instance, while the portion of households with access to sewer systems or septic tanks increasedfrom 43 percent to 77 percent, and those with access to garbage-collection services increased from one-third to more than half. Moreover, the gap between the poor and the national average decreasedfor all these indicators with the exception of access to e~ectricity.'~ Figure 2.9: Costa Rica House Services Figure 2.10: Costa Rica House Services Water: no piped water Water source: bbacueducto" 1989 1994 2000 2002 2004 I989 1994 2000 2002 Year Year I 1 I ~~ "Ext Poor -All Poor- - National "Ext. Poor -All Poor- ~ I+ It- .Nation& l9There i s no difference between 1989 and 2004 in access to electricity. 12 Figure 2.11: Costa Rica House Services With sewer or septic tank Figure 2.12: Costa Rica House Services Without electricity g DO 90 .3 80 c & 70 60 % W I @ 40 1989 1994 2000 2002 2004 1989 1994 2000 2002 Year Year 2004 1-Ext.Poor -All Poor- I+- =National 1"Ext. Poor -All Poor- -+- .NalioMJ I I Figure 2.13: Costa Rica House Services Figure 2.14: Costa Rica House Services With garbage recollection With telephone line g 9U 80 .c 70 '5 60 m rc E o 30 @ 20 1989 1994 2000 2002 2004 I Year 1989 1994 2000 2002 2004 Year -Ext Poor -AllPoor - -+- .National /-Ek?.Poor -AllPoor - -t-.National Source: National household survey, INEC Costa Rica 2.18 Incontrast to the trend inincome poverty, housingconditions and access to basic services continued to improve well after 1994, although most improvements were registered prior to 2000. Changes in housing and basic service indicators are summarized in Table 2.6. As can be seen, after 2000 very little (or no) improvement was observed in the most of the selected indicators. Between 2000 and 2004, only access to telephone lines and electricity showed significant improvement, while access to water systems and to sewers or septic tanks has shown little or no improvement. 13 Table 2.6: Costa Rica Housing Conditions and Services Changesfor All Poor And Extremely Poor 1989-2004 Telephone line 'Ifchanges I 4 I 4 I 4 I for extremely poor andoverall poor were in the same direction, only one arrow was used, otherwisethe first arrow correspondsto all poorand the secondto the extremely poor. **Some indicatorswere renamedto reflectpositivecharacteristics. p <5 percent; 4-b :no statisticallysignificant change; all others p 20 010 born in migrated into Costa Rica during the 1990s, and it is estimated that by 2005, as much as 7 percent to 8 percent L c 20% born inNcaraaua of the Costa Rican population 8P 15% - was made up of Nicaraguan immigrants (Marquette 2006). Since 2002, poverty rates are Populationshare of Ncaraguans household between 8 and 11 percentage points higher in households 2000 2001 2002 2003 2004 where 20 percent or more of the members are Nicaragua- Source: World Bank Staff calculationsusingthe EHPM born compared with households with less than 20 percent Nicaraguan membership.21(The observed differences in the incidence of extreme poverty between immigrant and Costa Rican households generally ranged from zero to three percentage points, a figure that i s not statistically significant.22) ''Households with 20 percent or more members born in Nicaragua were classified as Nicaraguan households. "Differencesbefore2002arenotsignificantatp<=10percent. 22 With the exception of the six percentagepoint difference observed in extreme poverty rates in 2002. 14 2.20 Nicaraguan migrant households have no impact on nationalpoverty rates. Computing the poverty rate in Costa Rica excluding all Nicaraguan households from the sample does not significantly change poverty rates in the country. Simulations show that excluding all Nicaraguan households from the EHPMchanges poverty rates by less than one percentagepoint (Table2.7) - a difference that, again, is not significant. Table 2.7: Poverty Ratesin Costa Rica, 2000-2004 IncludingandExcludingNicaraguanMigrants I 2000 I 2001 I 2002 I 2003 I 2004 Poverty (including Nicaraguan migrants) I 23.1% I 22.9% I 23.5% I 21.4% I 23.9% Poverty (excluding Nicaraguan migrants) 22.8% 22.5% 22.6% 20.6% 23.1% Diference -0.3% -0.4% -0.9% -0.8% -0.8% 2.21 The migration rate increased in the 1990s and slowed after 2000. The unprecedented number of Nicaraguans who enteredCosta Rica duringthe 1990scame in search of employment. Duringthis period, the Nicaraguan migrant population grew from fewer than 90,000 to more than 200,000 persons, or from 2 percent to 6 percent of the total population. As many as 20,000 Nicaraguans entered Costa Rica in some years during the 1990s. This period of rapid in- migration by Nicaraguans to Costa Rica had ended by 2000. After 2000, better economic conditions in Nicaragua, among other factors, slowed Costa Rican migration rates, and the number of Nicaraguans coming to Costa Rica dropped to around 9,000 persons annually. By 2005, the number of Nicaraguans in the country had stabilized at around 300,000 persons or 7 percent of the national p0pulation.2~ 2.22 Permanent, seasonal, and irregular migrants. Most Nicaraguan migrants in Costa Rica, particularly those in San Jose, are permanent migrants. However, some Nicaraguan migrants may be seasonal or temporary migrants. As many as 100,000 seasonal migrants are in Costa Rica at peak harvest times, although the actual magnitude of seasonal migration is unknown. 2.23 Spatial concentration. Nicaraguans are concentrated in San Jose (40 percent) and in the northern border regions of Huetar North, Chortega, and Huetar Atlantic (more than 30 percent). 2.24 Mainly a younger working age population. Most Nicaraguan migrants in Costa Rica (70 percent) are young working age adults between age 20 and 39. Nicaraguans have almost twice as many adults as Costa Ricans in these age groups. Nicaraguan migrants are equally divided between men and women overall. However, there are proportionally more Nicaraguan women in the San Jose than there are men. This may be because San Jose has the biggest market for domestic workers. 2.25 Larger households. Nicaraguan households have approximately one more person (6 persons) than Cost Rican households (5 persons), and have lower educational attainment. While Nicaraguan migrants have significantly less educations than Costa Ricans, they have higher levels than Nicaraguans who remain in Nicaragua. Nicaraguan migrants on average have five years of education, compared to six years for the average Costa Rican. Most Nicaraguans have only incomplete primary education. 23Other reliable sources estimate the number of NicaraguanMigrants in Costa Rica at around 8 percent of the national population (see Marquette 2006). 15 2.26 High economic activity, concentration in low status and low paying occupations, and variation in occupation by region. Nicaraguan migrants are more economically active than Costa Ricans. There is a clear pattern of labor market segmentation in Costa Rica. Nicaraguan migrants are concentrated in lower status and lower paying occupations. InSan Jose, Nicaraguan men are concentrated in construction, and women in domestic service. In other regions of the country, Nicaraguans are concentrated in agricultural occupations. 2.27 Nicaragunas have lower relative pay, lower education, irregular status, poorer working conditions and greater job instability. Nicaraguans make up a significant proportion of the national labor force in agriculture (10 percent of the national labor force), construction (20 percent), and domestic service (30 percent). They work more hours and are paid less than Costa Ricans even within the same occupation, but there i s no evidence of formal wage discrimination against Nicaraguan migrants. Rather, lower educational levels relative to Costa Ricans are the main cause of lower wages among Nicaraguans. 2.28 Unemployment levels are higher in occupations in which Nicaraguans are concentrated and in San Jose. Unemployment levels overall are the same for Nicaraguans as Costa Ricans (6.5 percent). However, unemployment levels may be higher for Nicaraguan migrants than Costa Ricans within the lower skilled occupations in which Nicaraguans are concentrated (construction, domestic service and agriculture). 2.29 Proportionally fewer Nicaraguans have health insurance, but this may vary by time, poverty level, occupation, and region. Half of all Nicaraguan migrants have national health insurance and half do not. There are probably more uninsured Nicaraguans than insured ones if one takes irregular and seasonal migrants into account. The longer Nicaraguans are in the country the more likely they are to have insurance and most Nicaraguans in the country for periods of several years may eventually be insured. Poverty, occupation and region of the country affect insurance status. 2.30 Most Nicaraguans probably have access to health services, but their use of services may be lower than nationals. Nicaraguans account for 4 percent of the total demand for health services in the country and 5% of the total health budget, which is proportionately less than their overall share of the national population in 2000 (6 percent). Many Nicaraguans probably have access to public health facilities inCosta Rica regardless of their insurance or legal status. 2.31 There is no specific data on the overall health of Nicaraguan immigrants, but it may be similar to that of Costa Ricans. Health institutions in Costa Rica and vital statistics on illness or death do not generally record place o f birth or nationality. As a result, there i s no general health data on Nicaraguan migrants as a group. Representative survey data, however, do suggest that health indicators for Nicaraguan women and children are similar to those for Costa Ricans. This impliesthat levels of health may also be similar between Nicaraguan migrants and Costa Ricans in general. 2.32 School enrollment is generally lower among immigrants Education services may not reach Nicaraguan immigrants as effectively as health service, judging from official enrollment estimates. Some 79 percent of primary-school age Nicaraguan children are enrolled, compared to 95 percent of Costa Rican children. The difference in enrollment i s greater at the secondary- school level; 70 percent of Costa Rican secondary school age children are enrolled, compared to only 45 percent of Nicaraguan secondary school age children. Enrollment figures underestimate of actual enrollment among children from homes headed by Nicaraguans, however, because they 16 count only children born in Nicaragua as Nicaraguan immigrants. Children born to Nicaraguans in Costa Rica are recorded as Costa Ricans in school data. 2.33 Nicaraguan migrants face discrimination and stigmatization in Costa Rica, but data do not support many negative stereotypes about them. Negative perceptions of Nicaraguans often are reflected in the popular media and public opinion in Costa Rica. Discrimination in the workplace and in other settings may occur frequently. Nicaraguan migrants have been blamed for an array of negative trends in the country, including higher poverty levels, increased infant mortality, and stress on social services. These types of perceptions persist even though existing data do not support them. On the other hand, a recent representative national attitudinal survey suggests that many Costa Ricans also have positive perceptions of Nicaraguan migrants, particularly if they live or work alongside them. 2.34 Immigration has no direct impact on poverty in Costa Rica, but it has some indirect effects by contributing to wage inequalities. Poverty levels among Nicaraguans have not directly caused any significant increaseor change in overall poverty levels in the country because the size of the poor Nicaraguan migrant population i s small. But Nicaraguan migrants have indirectly affected overall poverty trends because he influx of lower educated Nicaraguan migrants in the 1990s addedto the already-increasing overall supply of less educated workers, and to the growing disparity between wages paid to less educated and more educated workers. These wage inequalities, in turn, were a major factor why poverty levels stopped declining after the mid 1990s. 2.35 Immigration policy focuses mainly on regulation and control. Costa Rica has no general strategy, integrated policy, or separate government agency to address issues related to Nicaraguan immigrants, their welfare or their integration. Specific government policies directed at Nicaraguan immigrants have focused mainly on regulating flows into the country or controlling illegal immigrants already in the country via amnesty measures. 2.36 Numerous NGOs provide specialized support and aid to Nicaraguan migrants. These NGOs are mainly active at the local level, and focus their activity in those areas of the country where Nicaraguans concentrate, such as San Jose, Huetar North, Chorotega, and Huetar Atlantic. NGOs working with Nicaraguans have tended to focus largely on providing legal support to migrants seeking residency and worker rights. 2.37 The level of irregular migration among Nicaraguans and the failed outcome of the seasonal worker program indicate that immigration policies by the Costa Rican government have not generally been effective incontrolling or regulatingflows from Nicaragua. 17 Box 2.2: NicaraguansMigrants in Costa Rica: Recommendations Marquette (2006) analyzes Nicaraguan immigration to Costa Rica and makes a number of recommendations regardingthe special needs of Nicaraguanmigrants with respect to poverty reduction inCosta Rica are: Government programs to help Nicaraguan migrants should target those in rural agricultural areas in Huetar North, as well as Chorotega and Brunca and in large slums such as La Carpi0 in San Jose were concentrations and poverty levels are higher. Efforts should be made to help areas that are most affected by seasonal migrants, and to develop strategies and resources for dealing with this type of migrant. Nicaraguan migrants seem to be less effectively reached by educational services in Costa Rica than health services. Their enrollment patterns vary by region; the need to increase primary school enrollment is particularly important in Huetar North and Brunca and for secondary enrollment in Huetar Atlantic and the Central Region outside San Jose. Seasonal migrants present schools with fluctuating enrollment and resource demands, which may particularly challenge schools in northern border areas. A single or standard registration mechanism for hospitals and health service providers should be developed to record nationality, allow comparability of data between service points and make possible a more valid assessment of the overall health profile of Nicaraguan migrants. Increasing health insurance levels among Nicaraguans i s essential to provide them with guaranteed access to the same quality of health care Costa Ricans receive. Public campaigns to raise social security affiliation rates among Nicaraguan migrants may be useful in this context. Nicaraguans need to be more fully integrated into existing primary care programs and perhaps be the target of specialized outreach programs in: general maternal and child health, family planning, prenatal care and child immunization. Nicaraguan migrants may achieve occupational mobility and its poverty benefits by improving their educational attainment and that of their children. This means increasing secondary school enrollment among the children of Nicaraguan migrants and providing training and educational opportunitiesfor older migrants. Efforts could be made to encourage training in higher skilled occupational areas such as tourism where the labor force is expanding rapidly, or in new areas such as the development of ethnic products (food, restaurants, entertainment) the demand for which i s generated by migrants themselves. Source: Marquette (2006). Poverty and the indigenouspopulation24in CostaRica 2.38 There are eight distinct ethnic or indigenous groups in Costa Rica - the Cabecar, Bribri, Ngobe, Terraba, Boruca (Brunca), Huetar, Malekus and Crorotega - and there are 24 indigenous t e r r i t o r i e ~According to the last census, there were almost 64,000 indigenous persons in Costa . ~ ~ RicaZ6 in 2000 (1.7 percent o f the population), o f which 42 percent live in the indigenous territories, 18 percent near the indigenous territories and 40 percent in the rest o f the country (Table 2.8). 2.39 Poverty is much more prevalent among indigenous people than in the rest of the population. Some 77 percent of indigenous people have at least one unmet basic need, according to the UBN indicator (UBN), compared to 39 percent o f the rest o f the population. The gap is even wider if one looks at the percentage of people with two or more UBNs: 56 percent of 24For this report a person is considered indigenous based on self identification. 25Different sourcesmention between 22 and 24 indigenous territories defined as specific geographic areas recognized by the government of Costa Rica. 26INEC, Census 2000. Costa Rican indigenous organizations estimate their population at 73,000, or 1.9 percent of the population. 18 indigenous people fit that description, compared to 14 percent for non indigenous-a ratio of four- to-one. 2.40 Place of residency i s more important than ethnicity in determining the probability of living in poverty. Living in indigenous territories is more determinative of poverty than being indigenous. Indeed, census data indicate that 85 percent of non-indigenous people with one or more UBN who live in the indigenous territories2' live in poverty, compared to 65 percent of those who live outside indigenous land (Table 2.8). Source: CostaRica2000 Census. INEC 2.41 Several other well-being indicators - including fertility rates28and a range of education indicators (illiteracy, average years of education, percentage of five to 15 year-olds going to primary school and percentageof persons with secondary or higher level of education) - produce similar findings: indigenous people living on indigenous land are the worst off, followed by non indigenous people living on indigenous land and then by indigenous people living outside the indigenous territories. Non indigenous people living outside indigenous lands are better off than any other group (Table 2.9). But there i s very little difference in labor net participation rates between any or the groups, and indigenous people living on indigenous land have the lowest open unemployment rates (less than halfthe national average). 21 About 6,000 non indigenous people live in indigenous territories. They represent 18% of the total number of people residing in indigenous lands. ** Low fertility rates are considered, for this study and poverty status, a positive characteristic. 19 Poverty and theAfro-Costa Rican population29 2.42 According to the last census, there were almost 73,000 Afro-Costa Ricans (2.1 percent of the population) in 2000, of which 74 percent live in the Limon province, 14 percent in San Jose, and the other 12 percent inthe rest of the country. Most (64 percent) live inurban areas.30(INEC, Census 2000). 2.43 InLimon province, Afro-Costa Ricans f'are better-than-average on well-being indicators, while the entire Afro-Costa Rican population has similar or better well-being indicators than Table2.10: EducationandAfro-Costa the Costa Rica average. This is true for Ricans Afro-C.R. housing conditions and services, education and insurance (Tables 2.10 and 2.1 1). The higher unemployment rate among Afro-Costa Rican's Secondar 31% does not seem to have an effect in any other Tertiar 18% characteristic. T O T A L 100.0% 100.0% Source: 2000 Census, INEC Do not Latrine go to or no No Bad Open No sanitary electricity ''literacy (13-17 Quality appliances No unemploy insurance service year- Housing ment' olds) Afro-C.R. 11% 2% 4% 26% 11% 6% 7% 16% Limon 20% 7% 8% 41% 15% 13% 6% Total 11% 3% 5% 32% 11% 7% 5% 18% Education and Health Access 2.44 EHPM data indicate that between 1989 and 2004 education outcomes (for instance, school enrollment, average years of schooling, attainment levels, and performance) improved, even for the poor and the extremely poor. But the indicators remain low, especially for secondary education. 2.45 Primary and secondary net3' and gross enrollment rates have increased for all poverty groups, and enrollment gaps between the poor and non-poor have decreased, since 1989.32 (Figures 2.16-2.19). Improvements in primary net enrollment rates are smaller but the absolute levels are high, reaching almost 67 percent for the extremely poor and 73 percent for all poor; improvements in secondary are much more impressive, although absolute values are low. Secondary enrollment rates increased 64 percent (net) and 79 percent (gross) for the extremely poor, and 48 percent (net) and 75 percent (gross) for all poor. On average, only around one third 29For this report people are considered Afro-Costa Ricans if they identify themselves that way. 30Urban areas include the urban periphery classification used in the 2000 census. 31Estimates are basedinmid year attendance. For the estimation of net enrollment rates using the EHPM, only the birthyear was available. Lack of a precise birthdate could have placed many students at the appropriate level or one year behind their level. Inthose cases the favorable assumption that they were at the appropriate level was made. As a result, this overestimates net enrollment rates. 32This is true for all gaps inprimary education and for gaps in relative terms for secondary net enrollment rates. 20 of poor students are performing at their grade level. Even the non-poor had low rates of secondary net enrollment in 2004 (52 percent). 2.46 Enrollment gaps between the poor and non poor have decreased.33 Educational gaps between the poor and non poor have decreased in relative terms, in all four indicators; only absolute differences at the secondary level have increased between 1989 and 2004 by an average of seven percentagepoints. ! Figure2.16: Primary NetEnrollmentRate Figure2.17: PrimaryGross Enrollment bv Povertv Group.2004 Ratebv PovertvGroua. 2004 100% 1 110% 80% 90% Ii 60% 70% ~ 40% I 20% 50% 0% 30% 1989 1994 2000 2004 I 1- 1989 1994 2000 2004 E x t . P o o r w A l l P o o r D N o n P o o r - A v e r a g e / l -Ex?.PoorWAllPoor-NonPoor-Average Figure2.18: Secondary Net Enrollment Figure2.19: SecondaryGross Enrollment Ratebv Povertv Groua.2004 Ratebv PovertvGroua, 2004 110% 7 1 1 60% 40% - - 20% I 1989 1994 2000 2004 t 1989 1994 2000 2004 1 1 Ex?.Po0r All P00 r Non Poor-Averagd 1 DExt.Poor5AllPoor mNonPoor-Averagi Net enrollment rates: Number students at the appropriate level (for their age) divided by the total number of students who should be at that level Gross enrollment rates: number of students at the level divided by the total number of students who should be at that level Source: World Bank staff calculations using the EHPM 2.47 Average years of schooling for the poor and extremely poor increasedfrom 1989 to 2004, although they decreased slightly for the extremely poor and were unchanged for all of the poor between 1989 and 1994, according to the EHPM (Table 2.12). Despite increases in years of education, the average years of education were still only 4.8 for all the poor and 4.2 for the extremely poor in 2004. That is less than the first two cycles of primary education. 33 This is true for all gaps inprimaryeducation, and for gaps inrelativeterms for secondarynet enrollment rates. 21 Source: Own calculations based on INEC 2004 Household survey 2.48 There is very little relationship between low birth weight and poverty at the county level. Counties with higher poverty levels34 do not have higher rates of low birth weights (Figure It is reasonable to assume that the high coverage levels of pre-natal and general medical attention decreasedthe incidence of low birthweight at all poverty levels. Figure 2.20: Poverty Level and Low BirthWeight * * lo 4 1I r Talamanca .. r .r Mina -** + n * + 8+ g % b I I c 1 -I- 4 2 4 30 4 5 3 40 * 50 *.I 10 6 4f Lo5 Chiles .r + * 4 - c Gofito 2 - - PovertyIncidence irce: Rosero. Luisand the CCSS 2.49 Poverty and infant mortality rates are not related at the regional level. The estimated linear regression between poverty headcount rate (independent variable) and infant mortality (dependent variable) is basically a straight line (Figure 2.21). In other words, the poverty rate of each region has no effect on infant mortality rates. 34 Poverty is measured as the presence of two or more unsatisfied basic needs. 35 Estimate linear regression R2value of 0.03 22 Figure 2.21: HeadcountRatesand Infant Mortality, 2001 I 50.0% g! 40.0% E 30.0% E S xg 8 20.0% 10.0% - 0.0% Central Huetar Huetar Pacific Chorotega Brunca Atlantic North Central Headcount Rate -Infant Mortality Rate +Fitted rce: EHPM and deathrecords. 2.50 The proportion of poor persons covered by health insurance i s relatively high in Costa Rica, although the data suggest that coverage has not increased since 1989. The percentage of poor persons covered by health insurance i s high in absolute terms - about 74 percent - but i s lower than among the non poor coverage rate (84 percent) (Table Table 2.13: Insured o ulation 2.13). . Insurance coverage for the extremely poor - 69 percent - is only National -0.6% ns slightly worse than for the poor. The NonPoor -1.6% * estimated coverage rates do appear All Poor 75% 74% -0.8% ns slightly lower in 2004 than in 1989, Extreme1 Poor 70% 69% -1.1% ns but the only statistically significant *significantat p<5%; ns: not significant at p<5% change since the 1.6 percent 1989 was Source: World Bank staffcalculations basedon INEC 1989 & decrease in coverage among the non- 2004 EHPM poor. 2.5 1 Health insurance financed by the government i s highly progressive, while pension-related ' health insurance is distribution neutral. According to the EHPM survey for 2004, a poor person i s four times more likely to receive state-financed health insurance than a non-poor person (4.7 percent, as opposed to 1.2 percent. Health insurance for retirees covers all socioeconomic groups at the same level: 5.8 percent. (Table 2.14) Nonfee healthinsurance providedby the government 23 Employment, income composition and salaries 2.52 While average unemployment rates are not high in Costa Rica, they are high among the poor and among (all) women. The national unemployment rate in Costa Rica was 6.3 percent in 2004, but it was 16.0 percent among the poor and 9.0 percent for all women. That means the unemployment rate is three-and-a-half times higher among the poor than the non-poor and about 84 percent higher for women than for men (Table 2.15). The probability of beingunemployed is highest among the extremely poor, especially those who live in urban households or are females. Inurban areas, more than one-fourth of the extremely poor, and one-third of extremely poor females, are unemployed. INECemdovment and unemdovment definitions Labor Force: 12 years and older Employedor Unemployed;Employed: 12 years and older personwho worked one hour or more last week; Visible underemployment: employedperson working less than 47 hoursper week, wants to work more and is capable of more work but can not find extra work; Invisibleunderemployment:employedperson working 47 or morehoursper week andwith salarybelow the lower minimumwage establishedby law. Unemployed: 12 years and older personwho did not work last week (onehour) and did look for work. Unemployedwith experience: unemployedpersonwho hadworkedinthe past. Unemployedwithout experience:unemployedpersonwho hadnever worked in the past. Inactive: 12 years and older person, who did not work last week (one hour), did not look for work in the last five weeks. Includesretired, students, stay at home persons, handicappednot ableto work andothers. Laborforce / total population= (Employed Unemployed)/ (Employed Unemployed Inactive). + Unemployed/ Labor Force= Unemployed/ (Employed +Unemployed). + + 2.53 Unemployment i s twice as prevalent among the poor as the non-poor, but the unemployment rate for the non-poor is four times higher.36Only 4.3 percent of all poor persons are unemployed, but the open unemployment rate for the poor is 16.0 percent. In contrast, 2.1 percent of the non-poor are unemployed, and their open unemployment rate is 4.5 percent. The reason for this apparent discrepancy i s that the open unemployment rate i s computed by dividing 36 Since not all households used in the official unemployment numbers had poverty classification, there are differences between the national values reported here and INEC official rates. In all cases INEC- reported rates are higher by 0.2 for open unemployment and 1.1percentage points for gross Unemployment. 24 the number of persons unemployed by the number of persons in the labor force. A much smaller share of the poor population i s in the labor force (26.7 percent), than the non-poor (45.7 percent). For the poor, in short, a lower participation rate in the labor force37i s an important cause of high open unemployment. 2.54 To assess the relative importance of participation rates, two simulations for the open unemployment rate were performed: first, usingthe Gross Participation Rate of the poor and the unemployed share of the non poor; and second, usingthe Gross Participation Rate of the non poor and the unemployed share of the poor (Table 2.16). The difference in the percentage of unemployed between the poor and the non poor explains 55 percent of the Open Unemployment Rate disparities, and different Gross ParticipationRates explain 45 percent. Table 2.16: ODenUnemDlovmentRatesSimulationsfor the Poor I I NonPoor I All Poor I Simulation 1 I Simulation2 I Gross ParticipationRate (GPR) 45.7% 26.7% 26.7% 45.7% Unemployed(UNEMP) 2.1% 4.3% 2.1% 4.3% Employed(GPR -UNEMP) 43.6% 22.5% 24.7% 41.4% Others ' 54.3% I 73.3% I 73.3% I 54.3% Total 100.0% 100.0% I 100.0% I 100.0% Open UnemploymentRate 4.5% II 16.0% 7.7% 9.4% Differencewith All Poor Open Unemployment Rate -8.3% -6.6% Note:GrossParticipation Rates = 100percent- Others 2.55 On average, each poor worker has to support four and a half household members.38One way to observe the combined effect of unemployment and gross participation rates is to estimate how many persons share the income of each worker. Each poor worker's income has to support twice as many household members (4.5) as each non-poor worker's income (2.3) (Table 2.15: Total household members over number of employed household member). The figure i s even higherfor extremely poor workers: 5.7 people. 2.56 The employed poor work fewer hours per week and have significantly lower salaries than the non-poor (Table 2.17). On average, a poor person works eight fewer hours per week than a non-poor person, and earns only 42 percent as much. Hours worked and earnings gaps are even larger between the extremely poor and the non-poor; the extremely poor work 18 fewer hours per week per week than the non-poor, on average, and earn one-quarter of the non poor hourly salary. Hours Der week 48 49 41 31 Colones per hour 945 993 415 238 Total monthly labor income 187,169 197,755 70,076 20,859 2.57 Labor income of the non poor i s almost six times that of the poor and fifteen times that of the extremely poor. The combined effects of lower participation rates in the labor market, higher 37Measuredas the size of the labor force dividedby total population(defined as the Gross Participation Rate inTable 2.10). 38Includinghim or herself. 25 unemployment, fewer hours worked and lower earnings per hour result in per capita monthly labor incomes among the poor (C. 13,214) that are 17 percent of those of the non-poor (C. 77,606); for the extremely poor the per capita monthly labor income (C. 5,167) is less than 7 percent of income earned by the non (Table 2.18). In other words, for each Colon a non- poor person earns from employment, a poor person earns 17 cents and an extremely poor person earns only 7 cents. 2.58 Non-poor households rely more on labor income than poor or extremely poor households. At least 71 percent of non-poor households' total income is from labor activities, compared to 53 percent for the extremely poor.40 This suggests that untargeted programs aimed at improving workers salaries would be regressivein nature. Table 2.18: 2004 Per Capita IncomeDistributionby EmploymentandPoverty Classification * Income ' Per Capitamonthly nominalincomein local currency (Colones). shares are computedas the percentageof the average values. 2.59 It should be noted that more than one-quarter of the income of the extremely poor comes from non-labor sources -asignificantlyhigher share than among the populationas a whole41 (Table 2.19). Indeed, cash transfers and subsidies alone represent almost 20 percent of the total income of the extremely poor. In general, subsides and scholarships are very progressive, and pensionsand returns to capitals are very regressive. 39 These issues -related to the labor market and poverty inCosta Rica - are examined in greater detail in chapter 4 of this report. 40 Since the adjustment values to the reported income (see Box 2.1) cannot be directly assigned to labor or non labor activities, the percentageslabor income (for the employed) are the minimum shares for the national average and the extremely poor. Actual values can be as high as 87 percent for the non-poor, 79 f;rcent for the poor and 67 percent for the extremely poor. As with labor income, these are minimum percentages. Ifadjustment values were to be allocated between labor and non labor income sources, the resulting percentages would be higher. 26 The Correlatesof Poverty 2.60 Analysis of the correlates of poverty helps deepen the understanding of how poverty and household characteristics are associated. By analyzing several variables at the same time in an econometric framework, the estimated effect of each variable can be isolated. Estimates obtained with this technique are closer to the true effect of the individual variables. The estimated parametershelp one to understand whether the variable i s positively or negatively associated with poverty and to assess the relative strength of association of the various factors to poverty outcomes.42 2.61 The analysis is limited by the variables used, and no direct causality effect should be assumed in the face of the statistical relationships uncovered. Since the number of variables in the EHMP survey is limited, the analysis does not take into consideration other factors as economic growth, violence, social and capital. Also, since causality can run in both directions between the variables and poverty, the relationships identified here should be interpreted as "statistical associations." 2.62 The variables used in this analysis include: household size; selected characteristics of the household head, including education; the household's geographic location; average access to services in the vicinity of the household (as defined by the Primary Sampling Unit (PSU)43;the household's labor characteristics; and nationality (see Table 2.20 for 2004 results). Household characteristics without enough variability were excluded from the analysis.* Variable selection was also determined by theoretical considerations. Individual regressions were estimated for 2004, Urban or Rural householdsfor overall poverty. No regressions were estimated for extreme poverty due to the low percentageof households with this character is ti^.^^ 2.63 "Probit" regressions were used to estimate the relationship between poverty and the characteristics of individual households.For each household characteristic an "Exp (B) parameter was estimated. Each estimated "Exp (B) value is the ratio of the probability of being poor when the variable is present to the probability of being poor without the variable. For example, an estimated value of 1.21 for the Exp (B) parameter associated to the variable "Female household head" means that the probability of being poor in a female headed household was 21 percent higher than that in a male headed household. Values above one represent an increase in the probability of being poor and below one a decrease in the probability of being poor%. The farthest away from one, the bigger effect the variable has. Exp (B) values of one (1) mean the variable has no impact on the probability of being poor. 42This type of analysis is often called "the determinants of poverty." It is important to note here, however, that while this type of analysis helps one to uncover statistical associations between specific variables and poverty outcomes, this analysis alone i s not sufficient to identify "casual" relationships between the s ecific factors and poverty. 4PAPSUconsists of a group of households within a census segment and sampled as a unit in the EHPM. 44For example, dummy variables with average values bellow 4 percent or above 96 percent were excluded. Exceptions were made for groups of variables like geographic location andjob industry becauseexcluding individual categories complicates the interpretationof estimated parameters. 45Even a continuous regression would base most of the parameters results in the observed values of the non extreme poor households. 46To transform Exp(B) values above one to percentage terms: subtract one and multiply by 100i.e. 1.5 is equivalent to (1.5 - 1.O)* 100= 50 percent increase in the probability of being poor. To transform Exp(B) values below one to percentageterms subtract Exp(B) from one and multiply by 100i.e. 0.75 is equivalent to (1 - 0.75)* 100= 25 percent reductionon the probability of being poor. 27 2.64 A strong association was found between overall poverty and household composition, education and labor characteristics. No strong statistical relationship was found between poverty and age of household head (or companion), nationality and access to most basic services. Geographic residence was found to be significant only for some rural regions. 2.65 Having young household members increases the probability of being poor, while households with more adults have lower levels of poverty. Without major differences between age groups (0 to 5, 5 to 11 or 12 to 17), having young household members i s associated with higher levels of poverty for urban and rural households. A larger number of working-age adults (18 to 59 years old) reduces the probability of overall poverty. 2.66 Female headed households have higher probability of beingpoor. This is especially true for rural areas, where female headed households have a 27 percent higher probability of being poor than male headed households. And it i s true even after controlling for household size, household head education level, region of residency and industry in which the head of household i s employed. The results suggest that other characteristics not included in the analysis - such as wage discrimination or limited social networks - may account for the estimated values.47 For urban households, the relationship i s only weak but also positive. No relationship was found between poverty and the household head age or the presence of a companion. 2.67 Education can be an avenue to avoid p~verty.~'The more education the household head or her or his companion has, the lower the probability the household will be poor. This i s true for all education levels in rural households and for households where the head (or companion) has completed primary or higher inthe urban households. 2.68 Since a person's education level is determined before his or her current economic status, a true causal relationship can be established. Completion of primary education reduces the probably of beingpoor by a third. If the household head or companion has completed secondary school, the probability household members will be poor i s almost half what it would be if he or she has no ed~cation.~'Having higher education i s the single most important characteristic of all: on average, higher education reduces the probability of being poor by 75 percent (compared to having no education). Even incomplete primary education, a proxy for literacy, reduces the probability of being poor inrural households by almost a fifth. 2.69 Rural households in the Chorotega and Brunca regions, respectively, have a 50 percent and 66 percent higher probability of being poor than rural households in the Central region. This i s true even after taking away the effect of all other variables included in the model. Possible reasons for these results are lack of infrastructure in these regions, their depressed economic conditions5' and low job mobility. These two rural regions also have the highest incidence of poverty (around 40 percent). No regional differences were found for urban households. 47 I t is important to note that single mothers living with their parentsor another family member are not included in this classification becausethey are not considered household heads. 48 The analysis uses the education attainment of the household head or her or his companion. 49 The differentials in the probability of being poor in percentage terms is given by the Exp(B) distance to one (1) multiplied by 100. 50 Withinruralhouseholds,The Choretega and Bruncaregions have the lowest percentage of their population employed; the Chortega region has the highest open unemployment rate, and the Brunca Region has the lowest Gross Participation Rate. 28 2.70 Basic services are not, related to the poverty status of the household^.^' With the exception of access to fixed phone line in rural areas, no strong relationship was found between housing services and poverty. Since most of the poor have access to electricity (96.9 percent), piped water (96.4 percent) and sanitation (85.5 percent) the results are probably a sign of the almost universal access of the services rather than their impact on well-being (just 45 percent of the poor have accessto fixed telephonelines). 2.71 Poverty and lack of insurance are strongly correlated, though the direction of the causality i s unclear. Poverty can prevent access to health insurance; likewise, lack of access to insurance and the expected negative effect on health can lower workers' productivity. 2.72 Job informality i s strongly correlated to poverty. If the head of household or companion works in the informal sector, the probability of being poor is increased by 31 percent for urban households and 42 percent for ruralhouseholds. This relationship occurs even after taking away the effect of the education level. Job informality is associated with lower salaries, job instability, restricted access to credit and lack of social services, mainly health services. With the exception of health services, the model does not consider such factors, which can explain the higher probability that people in the informal sector will be poor. 2.73 Household heads or companions working.in agriculture have a higher probability of being poor. If the household head or companion works in a non-agricultural industry,52the probability of being poor is, on average, 32 percent lower in urban households and 41 percent lower for rural households. The values range from 33 percent for urban households whose heads work in construction up to 53 percent for rural households whose heads work in transport and storage activities. 51Thevariable usedwas average values for all the households within a PSU. The effect of individual household access to the services has a minimumimpact in the variable value used. 52Low participation rates by the poor in the mining industry (< 0.5 percent), and utilities (<1.O percent) did not allow to determine statistically significance. These variables were not excluded from the analysis to have a complete classification group and easier parameter interpretation. 29 Table 2.20: Costa Rica2004 Determini its of poverty Urban Rural GrouDs Variables 0 to 5 1.48 1.44 + ** Number of HH 6 to 11 1.54 +++ **** 1.35 ++- ** members by age 12to 17 1.34 ** 1.26 ** ** 0.75 ** group 18 to 24 0.79 - 25 to 59 0.65 - ** 0.71 - ** 60 & older ns 0.85 - * Age ns ns HHHead Age * Age ns Female 1.21 + * I I ns 1.27 + ** With Companion Primary incomplete Education (none Primary complete 0.76 - 0.75 - excluded) Secondary incomplete 0.74 - 0.68 - Secondary complete 0.50 - ** 0.57 - ** With higher education 0.22 - ** 0.28 - ** Chorotega ns 1.49 + ** Regions (Central Pacific Central ns ns excluded) Brunca ns 1.66 ** Huetar Atlantic ns 0.84 +- ** I Huetar North Electricity n.a. ns Average value for Piped water ns accessinthe PSU Sewer or Sept. tank 0.47 - ns ns Phone line 0.74 - 0.58 - ** Labor No insurance ( Household average) 1.88 + ****II 1.38 + ** Informaljob 1.31 + ** 1.42 + ** Mining ns ns Industry 0.70 - ** 0.53 - ** Industry job Utilities ns ns (Agriculture Construction 0.77 * 0.67 - ** excluded) Commerce 0.69 ** 0.55 - ** Transport, Storage 0.60 ** 0.47 - ** Financial 0.65 ** 0.70 - Community, personalservices 0.70 ----- ** 0.59 - *** 'Odds Nationality 20% or more Nicaraguanborn HHmembers ns ns ratio value is the lationship between the probability of beingpoor w beingpoor with the variable = 0. Values above "1" meanthe variable increases the probability of beiAgpoor aid below one meansthe variable decreasesthe probability of beingpoor (the "+" and "-" signs are visual helpers of this * characteristic). ns: nonsignificant at p<= 10percent; *: significantat 5 percent <=p<=10 percent; **: significant at p<5 percent; nu: variable not used. Householdheador companion: the one with the highestincome 2.74 The correlates of poverty in Costa Rica have not significantly changed since 1989. Compared to 2004, the variables associated with poverty in urban and rural areas were very similar in 1989. Indeed, individual Probit regressions results sh0.w very few changes in the significance 'of the estimated parameters or their impact on poverty. (Annex ly3.Noteworthy differences in the 2004 results are: (i)the originally weak relationship between female headed households and poverty in urban areas is not present in any other year; (ii) incomplete primary education in rural areas is not significant in four of the other six years; and (iii)some urban 53Similar Probit regressions were performed for 1989, 1994, 2000,2001, 2002 and 2003 30 regions, mainly the Chorotega region before 2001, are significantly associated with poverty (compared to the Central region). Inequality in CostaRica 2.75 Like income poverty, income inequality in Costa Rica has tended to be low by Latin American ~tandards.5~As can be seen in Figure 2.22, out of 15 countries for which data are available from 2000, only Uruguay had a lower gini coefficient than Costa Rica. Figure2.22: GiniCoefficientsinLAC (Distributionof equivalizedhouseholdincome, 2000) uu 55 Source: Gasparini (2003). 2.76 Nonetheless, income Figure2.23: Ginibv vear andarea inequality in Costa Rica has increased 0.51 1 I since 1989. Measured by the gini 0.49 coefficient, inequality increased at the Q national level from 0.44 in 1989 to 0.5 z0 > 0.47 in 2001, and then decreased to 0.48 in 0.45 2004 (Figure 2.23). As can be seen 0.43 from the figure, measured income 0.41 inequality among urban and rural 1989 1994 2000 2001 2002 2003 2004 households follows very much the YEAR same pattern as measured inequality at the national level. 2.77 Urban and Rural households have similar levels of inequality. Urban and rural gini coefficients are not only similar (0.46 and 0.47 for 2004), but have occasionally changed rankings 54 Itmust be noted, however, that income inequality is higher on average in Latin America than in other regions of the world (see de Ferranti et a12004). 31 over the years. Rural households, which have traditionally experienced lower levels of inequality than urbanhouseholds, reported higher gini values from 2001 to 2003 (Figure 2.23). 2.78 More than half of the total income is captured by the wealthiest 20 percent of the population. The richest quintile of the population received 53 percent of all income in the country in 2004, as measured by the EHPM survey, while the poorest 20 percent of the population received only 4 percent of total income, an income ratio of almost 13 to 1. For each Col6n a person in the wealthiest quintile earns, a person in the poorest quintile earns less than 8 cents. While cross-country comparison of gini coefficients suggests that Costa Rica has relatively low levels of income inequality for Latin America, cross country comparisons of quintile income shares are less clear. Indeed, the upper and lower quintile income shares, and thus the corresponding quintile ratio, in Costa Rica i s very similar to those in other countries in Central America-for instance, Panama(2004), Guatemala (2000), and Nicaragua (2OOl)? Figure 2.24: Income Share by quintile, 2004 Figure 2.25: Quintiles Income share ratios 2004 ,o 20% 0 10% 8 0% Q1 Q2 Q3 Q4 Q2M1 03lQ2 WlQ3 051W Q5IQ1 QUlNTlLE QUlNTlLES 2.79 If inequality had remained the same as in 1989, poverty in 2004 would have been 25 percent lower than the actual values. One way to assess the importance of inequality changes is to estimate what poverty rates would be usingthe average income and poverty lines of 2004 and the income distribution of 1989. With the 1989 income distribution, the 2004 overall poverty rate would have been 18.3 percent or 5.6 percentage points lower than the actual 23.9 percent. The difference would have been bigger in urban areas (one third) than in rural areas (one fourth). Extremepoverty also would have decreased, but only 8.2 percent, or 0.6 percentagepoints. Extremely Poor All Poor 2004 2004 w. 89' Difference 2004 2004 w. 89' Difference National 6.6% 6.0% -8.2% 23.9% 18.3% -23.5% , Urban 4.7% 4.1% -12.0% 20.8% 14.1% -32.0% Rural 9.3% 7.5% -19.3% 28.3% 21.3% -24.4% 2.80 Decomposition analysis of the sources of income inequality indicates that education a1 attainment and the proportion of household members who work are the most important variables explaining income inequality. One category of income inequality measures, called entropy measures [See Annex 21, can be used to analyze the various sources of income inequality in an economy. Using the 2004 EHPM data, these measures show that the level of education of the household head or companion (whichever is the highest) can explain up to one-third of income inequality in Costa Rica, depending on the specific measure used (Table 2.22). In addition, the 55Figurestaken from recentWorld Bank LSMS-basedPoverty Assessments for Panama, Guatemala, and Nicaragua. 32 proportion of household members who work explains up to one-fourth of total income inequality. The proportion of inequality in Costa Rica explained by the either household size or access to basic services i s surprisingly low; these factors have one-third the impact they have in other Latin American countries. EntropyMeasurefor all CostaRica, IB decomposition for the rest of the values (see Annex 2 for a full explanation). Number of division per group in parenthesis. Educationis for the highest of the householdsheador companion. Other divisionsused andnot selected: Regions,Job Classification(high, middleand low pay), cooks with firewood or coal, sewer or septic tank (sanitationservice). How sensitive is observedpoverty to measurementand methodologicalissues? 2.81 All poverty measurement is affected by methodological choices that influence who is ultimately considered poor and who i s not. Technical and analytical decisions have to be madein computing an income (or consumption) aggregate and in defining and estimating the poverty lines used for the poverty classifications. Judging the quality of each decision i s not easy since one does not know the "real" underlying values with which to compare estimates (if there are any), and what is considered an appropriate measure in one country or in one moment in time within one country, may not be considered appropriate in another country or another time period. 2.82 This section examinesthe robustness of the current Costa Rican poverty measures to four different assumptions (or methodological considerations) used: (i) constant Engels ~oefficien?~ used over time; (ii) different Engels coefficients between urban and rural households; (iii)different extreme poverty lines between urban and rural households; and (iv) different adjustment for underreported income between urban and rural households (the impact of ' changing each individual assumption i s includedin Annex 3). 2.83 It is important to note that no claim is made here as to which is the best way to do the various adjustments or to compute poverty. The objective of this exercise i s to evaluate how sensitive the poverty measures are to alternative and reasonable assumptions. The resulting poverty estimates cannot be characterized as better or more accurate, or assessed to be a better reflection of poverty in Costa Rica. 2.84 Combining the effects of changing all the assumptions outlined above would result in changes in the poverty estimates for Costa Rica, with the most significant effects coming with 56 The Engels coefficient i s the relationship between the value of the food part of the poverty line and the total value of the poverty line. 33 respect to the urban and rural poverty headcounts. At the national level, the combined effect of all four simulations i s an initial increase of poverty until 1994, and a small decrease of 2.6 percentage points in the 2000 years. For urban households, the effect is a significant decrease in poverty for all years by an average of 6.7 percentagepoints, and for rural household the impact i s an increase in poverty in all years by an average of 4.6 percentage points (Figure 2.26). In the current (official) poverty estimates, rural poverty i s estimated to be only one-third higher than urban poverty (7.5 percentage points). With estimates revised on the basis of changes in the above assumptions, the results suggest that the incidence of rural poverty could be two-and-a-half times that of urbanpoverty (19.4 percentagepoints). Figure 2.26: Poverty Rates with All Four Adjustments 42% 38% c Q) 34% E c E 30% $ 26% U 8 22% I 18% 14% 10% 1989 1994 2000 2001 2002 2003 2004 -Urban Original &Urban All Changes 2.85 Policy recommendations would be very different under such revised scenarios. The revised poverty numbers suggest that policy recommendations should give more weight to the rural as opposed to urban dimensions of poverty - and particularly to higher ruralpoverty. In the current numbers (and assumptions), regional differences (between Central and Chortega regions, for instance) appear to be more important than the rural-urban dimensions of poverty. Conclusion 2.86 This chapter has presented a multi-dimensional profile of poverty for Costa Rica, based largely on analysis of the EHPM data from 1989 to 2004. It analyzes the evolution of income poverty at the national level, across rural and urban areas and across Costa Rica's six planning regions. The chapter also examines a number of non-income measures of well-being, including human capital, control of physical assets, and access to basic services, and it analyzes the correlates of poverty in Costa Rica. The chapter analyzes the evolution of income inequality over the period. Several main messages emerge from the chapter. First, the analysis shows that following significant progress in poverty reduction between 1989 and 1994, income poverty rates have essentially remained static through 2004. Second, a number of non-income measures of well-being have improved throughout the 1989-2004 period, even after progress on income poverty ended; 34 nonethelessimportant gaps remain between the poor and thenon-poor in access to basic health and human services, including in post-primary education, health services and insurance coverage. Third, several factors significantly affect a household's probability of being poor, controlling for other factors. Households with more children, whose heads have low education levels, are female, and work in agriculture or the informal sector all have relatively high probability of being poor. There also are distinct regional patterns of poverty. Finally, the analysis shows that income inequality has increased in Costa Rica since 1989. 2.87 Although poverty and social indicators in Costa Rica are good overall, the absence of progress in poverty reduction in Costa Rica since 1994 presents a puzzle, at least on the surface. Stagnating poverty levels have occurred in the face of fairly consistent economic growth. Why has poverty reduction been elusive to Costa Rica over the last decade? I s the problem related to the level of economic growth in recent years? Or i s it something about the pattern of growth? Income inequality did increase over the period. To what extent does rising inequality serve to dampen the salutary effect of growth on poverty? These questions are taken up in the next chapter. 35 3. GROWTH, INEQUALITY AND POVERTY 3.1 As discussed in Chapter 2, the incidence of poverty in Costa Rica fell from about 32 percent in 1989 to 23 percent in 1994, and has hovered at or around 23-24 percent since then. Stagnating poverty rates in Costa Rica since 1994 are somewhat surprising since Costa Rica experienced relatively consistent economic growth between 1994 and 2004 - per capita GDP growth averaged 2.4 percent per year over the period - and positive economic growth i s strongly associated with declines in poverty. In this context: What has been the relationship between growth and poverty reduction in Costa Rica? And why does growth appear to have become less effective in reducingpoverty over the last decade? 3.2 This chapter addresses these questions from the macro and sectoral perspectives. Several key messages emerge. First, while economic growth was generally positive since 1994, growth inboth GDP and household income has slowed over time. Per capita GDP growth declined from 2.8 percent per year from 1989 to 1994 to 2.6 percent a year from 1994 to 2000 to 2.0 percent a year from 2000 to 2004. Household per capita income growth, as measured by the EHPM surveys, slowed more dramatically. After growing nearly 5 percent a year from 1989 to 1994, average household income per capita grew just 1.5 percent per year from 1994 to 2000 and only 0.08 percent a year from 2000 to 2004. Such low levels of average household income growth may not be an effective force for poverty reduction. 3.3 There has also been an important shift in how the benefits of income growth have accrued to the poor and non-poor families over time. From 1989 to 1994, the benefits were distributed relatively equally. After 1994, however, income has grown significantly more slowly for the poor than the non-poor. From 1994 to 2000, average per capita household income of poor households grew at roughly one-third the national average. Moreover, between 2000 and 2004, per capita income actually declined for poor Costa Ricans. Indeed, since 2000, only households in the upper income quartile have seen their per capita household income grow. Inother words, since 1994the poor benefit less from whatever level of growth did occur. 3.4 This uneven distribution of the benefits of growth can be seen not only at the household level, but across different regions and economic sectors. EHPM data suggest that income growth has been faster in relatively wealthy counties (cantones) than in relatively poor ones. Moreover, growth has tended to be higher in sectors such as finance, commerce and public administration, which generally don't employ many poor or low-skilled workers sectors, not in sectors like agriculture, construction, manufacturing and services, where poor and near-poor workers tend to be concentrated. As a result, the sectoral and regional patterns of growth in Costa Rica have not been powerful forces for poverty reduction since 1994. 3.5 To ensure renewed poverty reduction in Costa Rica, the evidence highlights the importance of putting in place policies to promote higher sustained levels of economic (and household income) growth. But it also will be critical to ensure that the poor are better positioned to take advantage of growth and emerging economic opportunities. A key policy message, therefore, is the importance for Costa Rica of policies, programs, and investments that strengthen the human capital of the poor, that facilitate more productive participation of the poor in Costa Rica's labor market, and that protect and assist the poorest, vulnerable Costa Ricans in the face of changing global and domestic economic circumstances. 36 Poverty, growth and inequality: Stylizedfacts 3.6 There are three main reasons why growth can be more pro-poor (meaning that it leads to faster poverty reduction) in some cases than in others.57 First, growth can be more pro-poor simply because growth i s higher in some countries more than in others. Indeed, success stories of poverty reduction are typically associatedwith the achievement of high sustained growth rates. Figure 3.1: Poverty, growth, and changes ininequality Panel A Panel B 1 I 2 I n ? I ".d Per capita growth Change ininequality ISource: 1 Perry and others (2006) 3.7 For example, in the years 1981-2000 China's poverty rate fell from over 50 percent to 8 percent, owing to an impressive per capita growth rate of about 8.5 percent per year. Similarly, between 1993 and 2002 Vietnam cut in half its poverty rate, from 58 percent to 29 percent by growing at almost 6 percent per year in per capita terms. On the contrary, countries that have experienced economic stagnation or decline have witnessed dramatic increases in poverty: in 1993-2002 when Argentina's growth rate averaged 0.7 percent per year (a decline in per capita terms of .18 percent per year) poverty more than doubled from about 21 percent to slightly above 55 percent. 3.8 Usingcross-country data, Panel A of Figure 3.1 illustrates the importance of highgrowth levels for poverty reduction by presenting a scatter plot of changes in the $l/day poverty headcount against GDP per capita growth rates. The data reveal that, in general, higher growth rates are associated with faster poverty reduction. At the global level, a 1 percent increase in growth is associatedwith a decline of 1.25 percent inthe poverty headcount, on average. 3.9 A second reason why growth is more effective in reducingpoverty in some cases than in others relatesto the evolution of income inequality duringthe growth process. While it i s difficult to argue that poverty reduction can be sustained through redistributive policies in the absence of economic growth, it is clear that growth that i s equality-enhancing has a greater impact on poverty reduction than growth that is "distributionally neutral," (i.e., that leaves the income distribution unchanged). This point i s illustrated in Panel B of Figure3.1, which plots changes in poverty headcount against changes in the log of the gini coefficient. Here, the data suggest that every 1 percent increase in the gini coefficient i s associated with an increase in the poverty headcountof about 0.5 percent. 57 See Kraay (2006)for atechnicaldiscussion; see also Chapter IV of Perry et a1(2006). 37 Figure 3.2: The relationship between the effectivenessof growth and inequality I I 1I I * I Inequality (logged GiniIndex) Source: Perry andet al(2006) 3.10 A thirdfactor that influences the impact of growth on poverty reductionis the initial level ofinequality. There i s now ample evidence that even if inequality remains unchanged, poverty i s more responsive to growth when the initial income distribution i s more equal. This is illustrated in Figure 3.2, which plots the total elasticity of poverty reduction with respect to growth against the log of the gini coefficient for a number of countries.58 While it should be noted that there are limits to the use of the total growth elasticity of poverty as a measure of the effectiveness of growth (see Annex 4), the upward slope of the regression line in Figure 3.2 suggests that the growth elasticity of poverty reduction declines (i.e., becomes less negative) as income inequality increases. Growth and Doverty in Costa Rica 3.1 1 How does Costa Rica fare in each of these areas? Table 3.1 reports average per capita growth rates for the 1989-2004 period as well as for three sub-periods: 1989-1994, 1994-2000, and 2000-2004. For comparison purposes, the table also reports the average per capita growth rates of the other Central American countries and for the rest of Latin America. Table 3.1: Per capita growth performance 1989-2004(annual rates) 11 1989-1994 1994-2000 2000-2004 1989-2004 Costa Rica 2.81 , 2.60 2.00 2.5 1 Central America 2/ 1.27 1.64 0.10 0.91 Rest of Latin America 1.67 0.68 0.76 1.09 1/ This table reportsGDP mediangrowth figures for each group. 21 ExcludingCostaRica Source: WorldDevelopment Indicators. 3.12 The table highlights several interesting issues. First, it indicates that over the past 15 years Costa Rica's annual per capita growth rates at 2.5 percent was about 1.6 percentagespoints above those observed in Central America and 1.4 percentage points above the growth rates in the 58The growth elasticity of poverty reduction is defined as the percent (rather than percentage) change in poverty headcount associatedwith a 1percent rate of economic growth. 38 rest of Latin America. Inshort, the growth performance of Costa Rica has been relatively good by Latin American standards. Over the 1989-2004 period, there were only two Latin American countries that had higher average growth rates than Costa Rica: Chile and Panama.59Second, average growth rates for the 1989-2004 period have been quite stable; in the three sub-periods shown inTable 3.1, per capita GDP growth oscillatedbetween 2 and 3 percent per year. 3.13 This good performance (both in relative and absolute terms) over the 1994-2000 and 2000-2004periods seems to be at odds with the recent evolution of poverty. Table 3.2 reports the standard Foster Greer Thorbecke (FGT) poverty measures - the poverty headcount (PO), poverty gap (Pl), and squared poverty gap (P2) - over the same three sub-periods shown in Table 3.1. The data indicate that, on an annualized basis, headcount poverty in Costa Rica declined by about 1.7 percentage points in the first half of the 1990s, implying a growth elasticity of poverty reductionof -2.3, which is relatively high (inabsolute value) by international standards. Table 3.2: Changesinpoverty and growth elasticities of poverty 1989-200411 Changes inPoverty 1989- 1994 1994- 2000 2000 - 2004 1989- 2004 PO -1.75 0.02 0.21 -0.52 P I -0.74 0.01 0.00 -0.24 P2 -0.43 0.00 -0.01 -0.14 Elasticities 1989- 1994 1994- 2000 2000 - 2004 1989 - 2004 PO -2.30 0.03 0.45 -0.75 P1 -2.55 0.04 -0.01 -0.94 P2 -2.73 0.01 -0.11 -1.04 I!ThetabledisplaysthechangeinPO,PI,andP2inpercentagepointsatannualrates,andthe growth elasticity of poverty. Source: World Bank staff calculations. 3.14 Incontrast, at -0.75, the growth elasticity for the whole 1989-2004periodis on the low side (in absolute value) by international standards. This is the result of developments after 1994. Specifically, between 1994 and 2000 the growth elasticity of poverty was essentially zero (meaning that growth had no virtually effect on poverty). Moreover, between 2000 and 2004, the growth elasticity of poverty reduction was actually positive (Table 3.2) - that is, over the 2000- 2004 period growth was accompanied by a slight increase in poverty. A similar picture emerges when one looks at the P1 and P2 measures - although for them the poverty reduction-growth elasticities are essentially zero for boththe 1994-2000and the 2000-2004 sub-periods. 3.15 One challenge associated with assessing the poverty-reducing impact of growth i s related to the nature of the data being used. GDP growth rates are generally computed using National Accounts data, whereas poverty rates are estimated on the basis of household surveys of income, expenditure or consumption levels. If the growth rates implied in each of the different data sources were the same, there would be no problemcombining survey-basedpoverty estimates and national accounts-based growth rates to examine the relationship between growth and poverty over time. But it is well documented that household survey and national accounts data commonly 59Growthin Panamawas highly influencedby economic performancein the aftermathof the 1989 conflict with the UnitedStates. 39 generate different growth rates, with national accounts data tending to produce higher growth estimatesthan household survey data (Deaton 2005, Peny et a12006). 3.16 A key reason for this divergence is that national account-based GDP figures and household survey-basedincome figures measure different things: the former measure production, while the latter measurehousehold income. If this were the only difference, the problem could be addressed somewhat by using Gross National Income (GNI) rather than with GDP figures. In most countries, the differences between GDP and GNI growth rates tend to be small and, hence, working with one or another aggregate does not make much of a difference. Nonetheless, in countries like Costa Rica, where the flow of factor income to and from abroad - for example,. those related to INTEL-can be substantial, differences in GDP and GNIhave to be considered. 3.17 The differences between national accounts-based and household survey-based per capita growth estimates are not so much related to the distinction between production and income, but rather to the nature of the data sources themselves. Box 3.1 outlines several possible sources of discrepancies in growth estimates across national accounts and household survey data, as highlighted inrecent empirical studies. Box 3.1: Poverty, growth, and the nature of the data Deaton (2005) argues that there are a number of explanations for discrepancies between income estimates generated by household surveys as opposed to those derived from national accounts. First, not everybody who is asked to participate in a survey chooses to do so; if the better-off are less likely to respond (as evidence for OECD countries indicates) then survey-based growth rates may be biased downward. A second reason is that the way national accounts are computed, and the treatment of some specific items such as non-traded services, which may cause GDP growth rates to overestimate actual growth. Although in principle there is no particular reason to choose one measure over the other, comparing income and poverty trends based on a household survey with those based on the national accounts may create some problems. Figure B3.1 presents the scatter plot of survey-based growth rates against national accounts growth rates for a selection of Latin American Countries based on data in Gasparini et al. (2005). The regression line in the chart has an associated slope of 0.97 and an intercept of about -0.9. While the estimated slope suggests that there i s an almost one-to-one relationship between the growth rates corresponding to the two sources, the negative intercept indicates that national accounts growth rates tend to be much higher (almost 1 percent) than survey based estimates. I n particular, at low values of national accounts-based growth, the survey-based growth has a tendency to be negative. Figure B3.1: Survey-based Income Growth vs. National Accounts-based Income Growth -10 - -12 - I NA Based Source: Perry et a1 (2006) 40 3.18 To illustrate the differences across data sources, Table 3.3 reports per capita income growth rates using national accounts-based GDP and GNI measures, as well as measures based on Costa Rica's household survey, the EHPM. The data show that the GNIgrowth rates are more variable over time than the GDP growth rates, with rates during 1994-2000 just one-fourth of their 1989-1994 levels. Such deceleration in growth, usingthe GNImeasure, would be consistent with a deceleration of poverty reduction after 1994. Nonetheless, the GNIgrowth rate is still high enough - especially after 2000 - that one would have expected to see further declines in poverty during the last decade. 3.19 Comparing the household survey-based growth rates with the GNI-based growth rates one can see that they are quite similar for the 1989-1994 period (Table 3.3). Moreover, the implied growth elasticity for poverty reduction over this sub-period i s -1.4, a figure that i s consistent with international norms. After 1994, however, discrepancies appear between the two data sources. Household survey-based estimates produced higher growth rates in the late 1990s, while GNI-based estimates produced higher growth rates from 2000 to 2004 (2.33 percent per year using GNI compared to 0.8 percent per year using household survey). Indeed, household survey-based estimates of per capita income growth declined significantly from the pre-1994 to the post-1994 period, and were nearly zero from 2000 to 2004. This is consistent with the observed trend in poverty over these latter sub-periods. Nonetheless, average per capita income growth was still positive after 1994, suggesting that to understand fully poverty trends after 1994, one has to understand the evolution of income inequality - i.e., how the benefits of growth were distributed among the poor and non-poor over the period. Table 3.3: Alternative per capita income growth measures (annual rates) 1989-1994 1994-2000 2000-2004 1989-2004 GDP 2.81 2.60 2.00 2.5 1 GNI 4.04 0.97 2.33 2.36 HHSurvev 4.96 1.54 0.08 2.29 1/Onthebasis of $2000PPPper capitaGNI mediangrowth figures for eachgroup. 2/ ExcludesCostaRica. Source: WorldDevelopment Indicators. The evolution of inequality 3.20 The data suggest that increases in income inequality did serve to reduce the poverty- reducing impact of growth over the period, particularly after 1994. As shown in Figure 2.23 (Chapter 2), aggregate income inequality as measuredby the gini coefficient increasedfrom 0.44 in 1989 to 0.48 in 2004. Thus, depending on exactly how the income distribution changed over the period, one would expect that any level of growth would be less effective in reducing poverty than it would have been in the absence of an increase in inequality. Said another way, if inequality had remained constant over the period, growth would have led to a stronger positive effect on poverty reduction in Costa Rica. 3.21 To provide a sense of the impact of the observed increase in income inequality on poverty in Costa Rica, hypothetical poverty rates for 2004 were estimated for Costa Rica applying average per capita income growth to each household (as calculated usingEHPM data), underthe assumption that the income distribution remained the same between 1989 and 2004. In short, it was assumed that there was no deterioration in income inequality over the past 15 years. 41 The results of the simulation indicate that without any increase in income inequality over the period, the poverty headcount would have fallen to 18.2 percent by 2004, rather than stalling at the 23.9 percent. In other words, in 2004 poverty would have been nearly 6 percentage points lower than currently observed had income inequality not increased. 3.22 That growth was not effective in reaching the poor after 1994 can be seen clearly from more disaggregated analysis of Costa Rica's household survey, the EHPM. Table 3.4 shows average cumulative household income growth in Costa Rica across the three sub-periods under analysis, and for each of the four household income quartiles. As in Table 3.3, national averages were positive for each time period - although they decreasedover time. As can be seen from the data, however, although per capita household income growth appeared to be rather evenly distributed across quartiles between 1989 and 1994 (slightly higher among the top and bottom quartiles), the poor did much less well than the non-poor (and particularly than the wealthy) after 1994. Indeed, from 1994-2000, income growth among those in the poorest quartile was less than one-third the national average (and an even lower percentage of the income growth of the wealthiest quartile). From2000-2004, per capita household income of the poor actually declined. Indeed, over the four-year period, only households in the wealthiest quartile experienced positive cumulative income growth, according to the EHPMdata. 1989-1994 1994-2000 2000-2004 Poorest Quartile 25.1% 3.0% -1.3% Second Quartile 21.5% 5.0% -1.9% ThirdQuartile 23.1% 8.6% -2.6% Richest Quartile 31.9% 10.9% 2.2% National Average 27.9% 9.2% 0.3% 3.23 It is possible to develop an even richer characterization of the patterns of income growth across the income distribution by computing "growth incidence curves" (or GICs) for each of the sub-periods under analysis. These curves, introduced by Ravallion and Chen (2003), are simple and illustrative ways to analyze household per capita income growth across the income distribution. They show the proportional income change for each percentile of income (the poorest individuals in society would be close to 0 and the richest to 100). They have been used frequently in recent years to study the extent to which different sub-groups of the population participate in the growth process. GICs have two particularly interesting characteristics: (i) they can be used to compute the average per capita income growth rate experienced by different segments of the population;6' and (ii) they are able to capture much richer patterns of income inequality than those captured by gini coefficients or by the analysis quintile income shares (presentedinChapter 2). 6oThe integral from0 to a particular point inthe horizontal axis (sayp percent) would measurethe average growth rate of the p- poorest individuals. Clearly, the integral over the whole population equals the average growth rate (household survey-based). 42 3.24 Figure 3.3 presents national-levelGICs for Costa Rica over each of the three time periods under study. It highlights clearly the marked differences between the pre- and post-1994 periods - both in terms of average income growth and in the patterns of income growth across the income distribution. For example, over 1989-1994 the GIC had a U-shape, with a minimumat around 4 percent, indicating that every percentile in society enjoyed per capita income growth rates of 4 percent per year or higher. The two extremes of the GIC suggest that the very poor (the lowest decile) and the very rich (the top decile) benefited from particularly high growth rates, realizing growth in income of between 5 and 7 percent and between 7 and 8 percent, respectively. Given those levels of growth, it is not surprisingthat poverty declined significantly between 1989 and 1994 in spite of the fact that income inequality increased. 3.25 The post-1994 GICs paint quite a different picture, however. The GIC for 1994-2000 has an upward slope, an indication that income grew faster among those in the higher percentiles than among those in the lower percentiles. Not only did income inequality increase duringthe period, but income growth for most of those in the lowest two deciles61(those below the poverty line) hovered around 0.5 percent per year, indicating that the poor did not benefit much from growth over that period. True, some in the lowest decile experienced growth - around 2 percent per annum- but this was not enough to lift them out of poverty!* Thus, it should not be a surprise that even with positive aggregate growth, there was no much progress in poverty reduction. Figure 3.3: Growth Incidence Curves (1989-2004) 8 1 1 c s 3 5 2 1 0 -1 -2 0 10 20 30 40 50 60 70 80 90 100 Percentileof the population Source: World Bank staff calculations. 3.26 Between 2000 and 2004, the GIC once again takes a Ushape. But unlike the 1989-2004 period, the minimum level of per capita income growth i s around -1.0 percent (with an average of about -0.26 percent). As can be seen from the Figure, large segments of the population - including much of the population between the 10" and 80" percentiles - experienced negative 61In 1994,the poverty headcount was 19.96percent. Thus, the two lowest deciles would include all the poor. 62 Itshould be noted that GICs tend to have very large standarderrors at both tails of the distribution, which means that estimates at the extremes of the distribution are not very reliable. For that reason, one should be cautious about interpreting the growth rates enjoyed by either the extremely poor or extremely rich. 43 income growth. Thus, both the group of poor people between the 10 and 20" percentile and some of the initially non poor -those next to, butjust above, the 20" percentile - experienced declines in real income. This income decline prevented many poor from exiting poverty. Moreover, income declines among the near-poor had the effect of dropping some of them below the poverty line. The level of inequality 3.27 Even without a deterioration of the income distribution, high initial levels of income inequality (such as those typically found in Latin America and the Caribbean) may be an impediment to poverty reduction (Bourguignon 2002, Ravallion 1997, 2004). Indeed, Ravallion estimates econometrically reduced-form equations that analyze how poverty changes as a function of the interaction between growth and income inequality. Using a global database on growth and poverty from 62 developing countries, Ravallion finds the following empirical relationship: Change inPoverty (in%) = -9.33 (l-Gini)3x Per capita growth (%) 3.28 One implication of this expression i s that, depending on a country's initial gini coefficient, the growth elasticity of poverty could range from almost -5.0 (low inequality countries) to -0.5 (high inequality countries). In other words, societies with high levels of income inequality need to sustain much higher growth rates to attain the same level of poverty reduction as societies with low levels of income inequality (Figure 3.4). Figure 3.4: The impact of inequality on the efficiency of growth 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 GiniIndex Source: World Bank staffcalculations, basedon Ravallion (2004). 3.29 While Costa Rica's levels of income inequality are relatively low compared to other Latin American countries, recent increases may have adversely affected the relationship between growth and poverty reduction. Indeed, using the above empirical relationship, it is possible to make a rough assessment of the impact of recent increases in income inequality on the efficiency of growth in Costa Rica with respect to poverty reduction. Pluggingthe values of Costa Rica's gini coefficients in 1989 and 2004 into'theabove equation, one can see that risinginequality may have reduced the growth elasticity of poverty reduction by around 25 percent. In other words, given Costa Rica's level of income inequality in 2004, the country would need a growth rate of approximately 2.5 percent to achieve the same rate of poverty reduction it could have achieved with 2 percent growth in 1989. 44 3.30 For a given income level, poverty line, and distribution of income, it i s possible to "simulate" the potential trade-off between growth and inequality in the fight against poverty. Indeed, several recent studies of poverty in Latin America have addressed this issue in various ways (see, for example, Lopez and Serven 2005, Perry et a1 2006). In light of Costa Rica's income and current level of income inequality, where does the country stand then terms of this potentialtrade-off? And where does Costa Rica stand relative to others in Latin America? Table 3.5 provides some indicative estimates. The table report the maximum deterioration in income inequality that each country can tolerate without an increaseinpoverty, assuming that the country i s experiencing a 1percent growth rate. For purposes of internationalcomparability, inequality i s measuredby the gini coefficient, while poverty i s measuredusingthe $2/day PPPpoverty line. 3.31 The table suggests that progress in poverty reduction in Costa Rica i s relatively sensitive to changes in the income distribution, but the results are not out of line with most other middle- income countries in the region. Given a 1 percent growth rate, Costa Rica can tolerate an increase of nearly half a point in the gini coefficient, about the same level as Mexico, and Colombia, and more than Argentina, Chile and Brazil. Nonetheless, Costa Rica's progress in poverty reduction is about twice as sensitive to changes in inequality as such low-income countries in the region as Guyana, BoliviaandHonduras. Table 3.5: Growth and inequality inLatin American 1/ Country Country Argentina 0.40 Peru 0.64 Chile 0.42 St. Lucia 0.65 Brazil 0.43 Guatemala 0.67 Mexico 0.47 Paraguay 0.67 Costa Rica 0.47 ElSalvador 0.71 Colombia 0.48 Venezuela 0.82 Trinidad and Tobago 0.50 Ecuador 0.88 Dominican Republic 0.52 Nicaragua 0.88 Panama 0.53 Guyana 0.93 Belize 0.54 Bolivia 1.oo Uruguay 0.54 Honduras 1.25 Jamaica 0.60 1/ The table reportsthe maximumdeteriorationin inequalitythat each countrycan afford without increasesin poverty ($2 aday) when growthi s 1percent. Source: World Bank staffcalculations,adaptedfrom Perry et al(2006). Patterns of growth 3.32 A key message from the previous sections is that between 1994 and 2004 the pattern of growth in Costa Rica was not particularly "pro-poor.'' In this section, the growth-poverty relationship is examined across three different levels of disaggregation: (i) urbanand across rural areas; (ii) across geographic/administrativeregions; and (iii) sectors of the economy. across Urban and rural dimensions of growth and poverty 3.33 Figure 3.5 presentsGICs for the urban and rural population in Costa Rica. Several things can be seen from the graphs. While both urban and rural poor experienced income growth over the 1989 to 1994 period, the data suggest that the rural and urban poor fared quite differently. 45 The urban poor realized much better per capita income growth than the rest of the population, whether urban or rural. Indeed, high income growth among the urban poor seems to account for much of the relatively high income growth observed among the poor in the national-level GIC from 1989 to 1994. In rural areas, on the other hand, the wealthy experienced relatively high income growth over the period. These high levels of growth among the wealthy in rural areas appear to account, at least in part, for the relatively high growth experienced by at the highend of the national-level GIC from 1989to 1994.63 Panel A: Urban PanelB: Rural <-- 0 I O 20 30 p0 50 .60 70 80 90 IC0 0 I O 20 30 40 SO 60 70 80 90 100 Pereedikofthe ppdahon F'elcedile ofthe ppdahbn Source: World Bank staff calculations 3.34 From 1994-2000, the shapes of the urban and rural GICs, which slope upward,. very much mirror the shape found at the national level. The levels of growth are not similar to the national figures, however. Indeed, while part of the population - urban and rural - experienced positive income growth, a significant proportion of the both urban and rural populations - about one-third and slightly more in rural areas - experienced declines in per capita income levels. This is consistent with a slight rise in both rural and urban poverty during the 1994-2000 period.@ Between 2000 and 2004, the shapes of the urban and rural GICs differ once again. In sharp contrast to the 1989-1994 period, the urban poor experience the largest declines in income. In rural areas, the pattern i s more mixed, with the poorest and wealthiest households registering some income gains while the middle income percentiles experience declines in per capita income. 3.35 Patterns of urban and rural income growth are particularly striking when the post-1994 period is examined as a whole. For example, since 1994, 70 percent of urban dwellers - the poorest 70 percent - have experienced declining income (Figure 3.6). Losses have been particularly dramatic among the poorest percentiles, which have experienced average income declines in the 1 to 3 percent range over the 1994-2004 period. Again, this i s consistent with observed increases in urban poverty since 1994. In rural areas, there i s more variation, and the poorest people in rural areas appear to have experienced some gains. Nonetheless, those from the 10" to the 70" percentile in rural areas have experienced some decline in their incomes. Because the poverty headcountin rural areas was around 26 percent in 1994, this has meant some increase 63 Itshould be noted that since average income levels in urban and rural areas differ, the percentiles shown inthe rural and urban panels of Figure 3.8 are not completely comparable. World Bank staff calculations, using EHPMdata. 46 in rural poverty over the period as well. In contrast to in urban areas, however, the situation of the poorest of the rural poor has improved since 1994. Figure3.6: Urban and rural GICs (1994-2004) i 2~ I / -Urban -Rural 0 h IC ~5 -1 ' c? e -2 -3 i -4 0 10 20 30 40 50 60 70 80 90 100 Percentile of the population Source: World Bank staff calculations Suatial dimensions of growth and uoverty 3.36 Spatial variation in poverty rates within countries can be staggering. In 2000, the average per capita income in the poorest county in Brazil was roughly 10 percent that of the richest county; in Mexico, Chiapas had per capita income about 18 percent of that of the capital. Such regional differences can have important implications for poverty-reduction policy, even in geographically smaller countries like Costa Rica (Perry et a12006). Moreover, evidence indicates that there often are "convergence clubs" among rich and poor regions within c~untries.~~ This means that income in wealthy regions tends to grow relatively quickly while income in poorer regions tends to grow relatively slowly, so that incomes within rich (or poor) areas (or "clubs") tend to converge. But incomes across rich and poor areas tend to diverge, with poorer communities typically being left behind. 3.37 Costa Rica is no exception to the rule that spatial differences inpoverty can be important. Chapter 2 showed a large variation in poverty rates across planning regions. This section takes the analysis of spatial patterns of poverty - and of growth -deeper, to the canton Indeed, at the canton level, spatial variation in incomes in Costa Ricai s similar to the pattern found across counties in Brazil; in 2004, per capita income in the poorest Costa Rican canton was 11percent that of the richest canton. Large regional disparities in poverty can be seen in Figure 3.7, which shows variation inpoverty rates within and across planningregions in 1989 and in 2004. For each planning region, the figure reports the average poverty rate as well as the maximum and minimum poverty rates observed in different cantons. As can be seen from the figure, some Costa Rican cantons have poverty rates below 10 percent, while others have rates above 60 percent (80 percent in 1989). 65 See Perry et al. (2006) for the cases of Brazil, Chile and Mexico. 66 The canton-level analysis presentedhere is done usingEHPMdata. An important caveat in what follows is that the EHPM is not designedto make statisticallyunbiased inferences at the cantonlevel, butrather only at the planningregionlevel and acrossrural and urbanareas. For this reason, the results presented here should be consideredindicative. Nevertheless,the patternsuncoveredby the canton-levelanalysis are striking, andprovide some useful insightsfor investments to reducepoverty at the locallevel. 47 Figure 3.7: Variation in Poverty Rates in Costa Rica, by Planning Region, 1989 and 2004 ft 90 2004 80 .Min -Max 70 T T 60 50 40 30 20 10 0 , Central Atlantic Costa Central North Brunca Chorotega Central Costa Atlantic Central North Chorotega Brunca Region Huetar Rica Pacific Huetar Region Rica Huetar Pacific Huetar 3.38 In Costa Rica as elsewhere, regional differences in income and poverty are slow to change. This is consistent with the "convergence clubs" hypothesis that richer regions continue to be richer while poorer regions continue to be poorer. Indeed, as Figure 3.7 indicates, there was little change in the regional poverty rankings between 1989 and 2004, except that Brunca and Chortega basically switched places as the poorest regions. The Central Region and Atlantic Huetar continue to have the lowest average poverty rates, even though Central Region demonstrates large within-region variation. (Central Pacific and North Huetar maintain intermediate positions over time.) 3.39 Figure 3.8 presents the GICs from all six planning regions of Costa Rica for 1989-1994 and 1994-2004. Except for the poorest dwellers in Brunca, these GICs show that between 1989 and 1994, every percentile and region experienced income growth. While the U shape observed at national levels duringthat periodi s also present in Central Regionand Chortega, a clear pattern of pro-poor growth is seen in Central Pacific and North Huetar. The opposite i s true in Brunca andAtlantic Huetar. 3.40 More importantly, between 1994 and 2004, different income groups experienced different income growth rates, and in some cases, the poor were left behind. In the Central region, for example, while those in the upper percentiles of the income distribution experienced positive growth, those inthe percentiles below the poverty line suffereddeclines in their income levels. A similar pattern is observed in North Huetar, where the lowest 70 percent of the population experienced declines in income (while the upper 30 percent registered the highest income growth rates of all regions - an average of 2 percent per year). InPacific Central, the poorest percentiles also lost ground, with much of the middle percentiles experiencing essentially no income growth and those in the highest percentiles gaining some ground (an average of 1percent per year). In Chorotega, while no group experienced a decline in income, the rich fared better. 48 Figure 3.8: GICs, by PlanningRegion, sub-period(1989-1994,1994-2004) 12 CentralRegion 12 Chorotega -IC-89-94 " IC-94-04 --IC-89-94 IC-94-04 lo ]A -c 8 6 3 8 s 4 5 8 2 8$ 6 0 4 ~I 2 I -6 0 I O 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 I Percentile of the population Percentile of the populaaon Central Pacific BNnca I 16 I 24 -IC-89-94 IC-94-04 14 -IC-89-94 IC-94-04 22 ' 20 ;: I 4 .G12 1 2 1 0 I " ; 4 2 0 -2 -12' I I -4 d 10 20 '30 40 50 60 70 80 90 ~ 1; 0 10 20 30 40 50 60 70 80 90 100 Percentile of the population Percentile of the woulation 1 AtlanticHuetar North Huetar -IC-89-94 IC-94-04 I IC-94-04 6 4 ~ c- $ 2 8 0 ' -2 -4 1 -2 I I I J 0 10 20 30 40 50 60 70 80 90 100 0 I O 20 30 40 50 60 70 80 90 100 Percentile of the populaaon Percentile of the population iource: Own calculations 3.41 Incontrast,inBruncaandAtlantic Huetar, thepoorest populationexperiencednoticeable improvements in income between 1994 and 2004. So did a small cohort of the poorest population inChorotega. 3.42 Buildingon the earlier canton-level analysis, Figure 3.9 presents a frequency histogram of per capita income levels, by canton, in 1994 and 2004. Two main findings emerge from the analysis. First, the data suggest that the distribution of income across Costa Rican cantons i s bimodal (two peaks) rather than unimodal (one peak). This i s particularly clear in 2004. In 1994, 49 peaks can be identified would be around Colones 16,000 and Colones 30,000, re~pectively.6~ In 2004, the lower peak remains at about Colones 16,000, although the upper peak is now at about Colones 25,000. Figure 3.9: Histogramof Canton-level Per Capita Incomes, 1994 and 2004 Panel A. 1994 Panel B. 2004 0.35 7 0.3 I ' I 0.3 0.25 0.25 0.2 0.2 0.15 0.15 0.1 0.1 0.05 0.05 0 8.2 10.2 12.7 15.8 19.6 24.4 30.4 37.8 47 8.2 10.2 12.7 15.8 19.6 24.4 30.4 37.8 47 Per capitaincome(in 1000) Per capitaincome(in 1000) Source: Own calculations 3.43 Strong patterns of income growth can be seen among clusters of cantons (or "convergence clubs") over time. Four groups of cantons can be defined. One group includes cantons that had per capita income levels of less than Colones 20,000 in both 1994 and in 2004.68 This can be referred to as the low-low group. A second group includes cantons that enjoyed per capita incomes of above Colones 20,000 in both years (the high-highgroup). A low-high group can be defined as those cantons with average incomes below Colones 20,000 in 1994 and above Colones 20,000 in 2004, while the final, high-low group, includes those cantons with incomes above Colones 20,000 in 1994 and below Colones 20,000 in2004. 3.44 Analysis of these four groups indicates that there i s a tendency for cantons to remain above or below the Colones 20,000 threshold over time. Indeed, 65 percent of the cantons belong to the low-low group and 19 percent belong to the high-high group, while only 13 percent of cantons moved from the low to high group over time, and only 3 percent moved from the high to low group over time. Moreover, there were strong patterns of growth in these four groups of cantons. The average annual growth rate was 0.3 percent per year for the low-low group (almost stagnation), 1.1 percent for the high-high group, 5.0 percent for the low-high, and -5.8 for the high low. That is, cantons that were richer at the outset have tended to grow faster than cantons that startedpoor. Just a handful of cantons have been able to move up inthe ladder. 3.45 The existence of convergence clubs within the country has potentially important implications for public policy. Policy makers may have to face choices about whether to invest in areas with high expected rates of returns in the form of economic growth, or in poorer areas where investments may yield less aggregate growth, but would effectively reduce poverty. 3.46 Another policy issue involving the spatial dimension of poverty relates to whether a particular location has a highincidence (or rate) of poverty or a highconcentration (or density) of 67Real 1991Colones. 68The cut-off point of Colones 20,000 i s chosenbecauseit correspondsroughly to valley betweenthe twin peaks of 2004. 50 p0verty.6~Chomitz (2005) argues that a more refined use of spatial information on poverty can lead to more effective poverty-related investments. He proposes four spatial categories, each with distinct policy implications depending on the level and concentration of poverty (Table 3.6). TypeofArea I Low Poverty Density HighPoverty Density Type of Project HighPovertyRate 1 Type I1 Investments lack Typescale 1 - economies; 0 Rural roads therefore focus on interventions that 0 Other infrastructure. promote economic mobility, e.g.,: 0Education 0Cash transfers 0Agricultural R&D Low Poverty Rate Type I11 Type IV 0 Investments that boost labor demand I 0 Cash transfers I I Source: Adaptedfrom Chomitz (2005) 3.47 Inareas with highpovertyrates, but low povertydensities (Type Iareas), low population densities raise the per-person costs of infrastructure investments and prevent other types of productive investments from achieving economies of scale. As a result, these areas may not be able to develop substantial economic dynamism, so policy-makers should focus more on strengthening poor people's economic mobility (to more dynamic regions) or on providing direct poverty alleviation. Education initiatives, conditional cash transfer programs that provide incentives for human capital development, agricultural R&D or even payments for environmental services, may be most appropriate in this setting. In areas with low poverty rates and high poverty densities - for example, urban or relatively dense rural areas where economies of scale may exist - policies and investments that aim at fostering growth locally have good chances of reaching the poor and translating into effective poverty reduction. A major challenge i s such circumstances may be to ensure that wealthy groups do not "capture" the flow of resources intended for the poor. Where this i s a risk, targeted interventions - or even self-targeting mechanisms, such as those used in Argentina's and Colombia's workfare programs - may be particularly appropriate." Finally, areas with high poverty rates and high poverty densities have the potential to take advantage of projects with economies of scale, and because of their high poverty rates, are less likely to have problems of leakage of resources to the non-poor. Infrastructure investments, whether in (rural) roads or other types of economic infrastructure, may be appropriate projects in such setting. 69For a discussion on the incidence versus concentration of poverty at the level of planning regions, see Chapter 2. 70 The Argentine and Colombia workfare programs impose work requirements on welfare recipients at benefit levels that are lower than prevailing market wages for relatively unskilled labor. The idea is that by setting a sufficiently low benefit level, a workfare program will be self-targeted to those most in need; few of the non-poor inparticular will want to participate. 51 3.48 Figure 3.10 presents a scatter plot of poverty rates against poverty densities, building on the classifications outlined above and shown in Table 3.6.71 As can be seen from the figure, roughly one-third of all Costa Rican cantons fall into the high poverty rate-high poverty density category (Type 11), and another one-third fall into the low poverty rate-low poverty density category (Type 111). Incontrast, the high poverty rate-low poverty density category (Type I) and the low poverty rate-high poverty density category (Type IV) each account for about 15 percent of cantons. While cantons that fall into the low poverty rate-low poverty density category may not be top priorities for regionally differentiated poverty reduction interventions, those with both high poverty rates and high poverty densities would be. Moreover, at least under the 4-way categorization cited here, it may be that investments in infrastructure, such as roads or other types of economic infrastructure, may be appropriate complementary investments to national-level strategies (ineducation, for instance; see Chapter 6). Figure 3.10: Distribution of Costa Rican Cantons, by Poverty Rate and Poverty Density 80 70 60 50 W 23 40 $ 30 &b 20 lo ~ 0 0 0.2 0.4 0.6 0.8 1 1.2 I Densitv Note: Quadrantsare definedusingthe medianvaluesof the axes. Source: World Bank staffcalculations 3.49 Table 3.7 shows the number of cantons in each planning region that fall in each of the four poverty rate-poverty density categories. On the whole, the analysis suggests that different regions may warrant different types of poverty reduction interventions: Brunca, Atlantic Huetar, North Huetar and Chorotega have a large number of cantons with both high poverty rates and high poverty densities. These are cantons where investments in economic infrastructure may be cost effective. In contrast, the Central Region, which has a high concentration of cantons with low poverty rates but high poverty density, may be an appropriate setting for investments that boost labor demand or well-targeted conditional cash transfer programs that strengthenthe human capital and thus the productivity of the poor. 71 Poverty density is computed here as the number of people living under the poverty line in each canton divided by the highest value of this indicator among all cantons. 52 Table 3.7: Number of Cantons ineach of the Four Poverty Rate-Poverty Density Categories, by planning region(2004) . - - - Planning Region 2004 I I1 I11 IV Central Region 5 8 19 11 Chorotega 4 6 2 0 Central Pacific 2 1 5 0 Brunca 0 6 0 0 Atlantic Huetar 0 4 2 1 North Huetar 1 4 1 0 CostaRica 12 29 29 12 Source: World Bank staff calculations Sectoral decomposition of poverty 3.50 Which economic sectors have driven poverty reduction in Costa Rica? Table 3.8 summarizes analysis that decomposes changes in poverty (for each of the three FGT measures) into changes due to: (i)within-sector increases in welfare, (ii)increases in welfare due to population shifts across sectors, and (iii)an interaction effect.72 These components of the decomposition are labeled "intra-sectoral effects," "population-shift effects," and "interaction effects," respectively. The table presents the decomposition results for nine economic sectors at the national level, as well as disaggregatedby rural and urban areas. As can be seen in the table, within-sector improvements in welfare explain the vast majority of the poverty reduction seen in Costa Rica since 1989. At the national level, for example, it is estimated that over 70 percent of the progress registered between 1989 and 2004 came from within-sector improvements, while about 30 percent of the gains came from population shifts between sectors of the economy.73 The relative roles of within-sector gains and population shifts are similar in rural areas, but the patterns are different in urban areas where nearly all of the reduction in poverty can be explained by within-sector improvements inincome andwelfare. 72The approachfollows Ravallionand Huppi (1991). 73Gains fiom within-sector improvementsand populationshifts sum to slightly greater than 100percent. This is possiblearithmetically,becausethese gains were offset slightly by negative "interaction effects." 53 Table 3.8: Sectoral Decomposition of Poverty Changes (1989-2004) National PovertyMeasure Headcount Poverty Gap Poverty Gap Squared Poverty in period I 31.71 12.23 6.72 Poverty mperiod2 23.89 8.59 4.55 Sector Population share Absolute Percentage Absolute Percentage Absolute Percentage in period 1 change change change change change change Unknown 14.84 0.69 8.76 -0.62 17.03 -0.53 24.32 Agriculture and Fishing 26.19 -2.16 27.65 -0.98 26.96 -0.52 23.96 Mining 0.2 -0.02 0.29 0.01 -0.21 0.01 -064 Industry 13.25 -1.32 16.91 -0.49 13.38 -0.22 10.18 Electricity Gas and Water 1.72 -0.03 0.33 -0.02 0.64 -0.01 064 Construction 6.95 0.14 -1.76 0.04 -1.14 -0.02 0.73 Commerce 11.82 -0.92 11.79 -0.29 7 95 -0.1 4 6 TransponCommunicationandStorage 4.71 0.04 -0.49 0.04 -1.18 0.03 -1.35 FinanceInsuranceand RealEstate 2.77 0.14 .1.79 0.05 -1.34 0.04 -1.69 Services and PublicAdministration 17.54 -1.09 13.% 0.58 15.81 -0.41 18.59 Total Intra-sectoralefiect -5.92 75 66 -2.84 77.9 -1.73 79.34 Population-shifteffect -2.57 32.89 -1.07 29.47 -0.58 26.51 Interactioneffect 0.67 -8.55 0.27 -7.37 0.13 -5.85 Change in poverty -7.83 100 -3.64 ' 100 -2.18 100 Urban PovenyMeasure Headcount Poverty Gap Poverty Gap Squared Povertyin period1 26.16 9.65 5 20 Povertyin penod2 20.77 7.04 3.59 Sector Population share Absolute Percentage Absolute Percentage Absolute Percentage in period 1 change change change change change change Unknown 16.67 -0.7 12.97 -0.66 25.17 -0.57 35 65 AgnculNre and Fishing 3.92 -0.22 4.05 -0.05 1 9 0.01 -0.56 Mining 0.15 -0.05 0.93 0 0.08 0.01 -043 Industry 18.11 -1.84 34 I8 -0.73 28.03 0.34 21 31 ElectricityGas andWater 2.35 -0.19 3.57 -0 06 2.3 -003 1.63 Construction 6.15 -0.03 0.63 -0.15 5 83 -0.15 9.08 Commerce 16.74 -1.7 31.46 -071 27.13 -0.34 21.32 TransponCommunicationand Storage 6.51 0.27 -5.09 0.17 -6.52 0.12 -7.18 Finance Insurance and Real Estate 4.6 0.25 -4.65 0.04 -149 0.04 -2.21 Servicesand PublicAdministration 24.8 -1.25 23.15 -0.47 18.02 -0.34 21.37 Total Intra-sectoraleffect -5.46 101 19 -2.62 100.47 -1.61 99.98 Population-shifteffect -0.47 8.64 -0.07 2.81 -0.01 0.57 Interactioneffect 0.53 -9.83 0.09 -3.28 0.01 -0.55 Change inpoveny -5.39 100 -2.6 100 -1.61 100 Rural PovenyMeasure: Headcount Poverty Gap Poverty Gap Squared Povenyinperiod 1 35.83 14.15 7.86 Povertyinperiod2 28.25 10.76 5.89 Sector Population share Absolute Percentage Absolute Percentage Absolute Percentage in period 1 change change change change change change Unknown 13.49 -0.16 2.14 -0.33 9.85 -0.33 16.71 AgriCulNre and Fishing 42.7 -3.41 45.04 -1.62 47.76 -0.89 45.26 Mining 0.24 0 0 0.01 -0.44 0.02 1 Industry 9.66 -0.99 13.08 -0 33 9.63 -0.15 7.46 ElectricityGas and Water 1.25 0.1 -1.27 0 -0.11 0 0.22 Construction 7.54 0.18 -2.44 0.18 -5.3 009 -4.74 Commerce 8.17 -0.44 5.8 0 -0.08 0.07 -3.62 TransponCommunicationand Storage 3.37 -02 2.63 -0.07 2.16 -0.05 2.38 Finance Insuranceand RealEstate 1.41 0.07 -0.89 0.06 -1.89 0.04 -2.18 Servicesand PublicAdministration 12.17 -1.07 14.13 -0.7 2064 -0.47 23.87 Total Intra-sectoraleiiect -5.93 78.23 -2.78 82.23 -1.66 84.38 Population-shifteffect -2.17 28.63 -0.92 27.19 -0.49 24.89 Interactioneffect 0.52 -6.86 0.32 -9.42 0.18 -9.27 Change in poverty -7.58 100 -3.38 100 -1.97 100 54 3.51 Table 3.8 also highlights which sectors that have contributed the most to within-sector poverty gains. The largest intra-sectoral contributions to poverty reduction over the 1989-2004 period came from: agriculture and fishing (2.2 percentage points), industry (1.3 percentage points), services and public administration (1.1 percentage points), and commerce (0.9 percentage points). This is not very surprising, given that more than two-thirds of Costa Ricans were employed in these four sectors at the start of the period. For example, about 26 percent of Costa Ricans were employed in agriculture at the national level, and more than 40 percent of rural Costa Ricans were in agriculture. Similarly, more than 17 percent of Costa Ricans were in services and public administration in 1989. In contrast to agriculture, however, this activity i s concentrated in urban areas, where roughly 25 percent were in services and public admini~tration?~ 3.52 While the decomposition analysis presentedTable 3.8 can help clarify the role of specific economic sectors in recent poverty reduction (as well as the role of population movements from one sector to another), more information i s needed to attain a deeper understanding of the "pro- poorness" of each sector. One factor i s the size of the sector itself. It i s likely, for example, that growth in larger sectors (as measuredby their share of GDP) will have a larger impact on poverty reduction than will growth in smaller sectors. Indeed, if a sector accounts for a small share of economic activity, then it i s likely that relatively few people (either poor and non-poor) will benefit from growth in that sector. The labor intensity of a given sector also is important. This issue is highlighted by Loayza and Raddatz (2005), who argue that to the extent that different sectors have different levels of labor intensivity, they also are likely to have a different impact on poverty reduction (for a given level of economic activity). 3.53 Figure 3.11 presents the distribution of relative labor intensities of Costa Rica's main economic sectors, using the EHPM data for 2004. For data reasons, "labor intensity" is defined as the share of total income in a sector that i s allocated to unskilled workers. For each sector, the range of labor intensity values is computed using information at the level of the six different planning regions. As can be seen in the figure, agriculture and construction are clearly the most labor-intensive sectors, with a median labor intensity of nearly 60 percent. Transport, commerce, industry and services can be grouped in a middle category of labor intensity, with median intensities ranging between 20 and 45 percent, while utilities and financeheal estate are the least labor intensive sectors, with medians values below 20 per~ent.7~ 14It should be noted that sectoral population shares have shifted, sometimes dramatically, between 1989 and 2004. For example, the population share for agriculture and fishing declined dramatically from 26 to 15 percent over the period. Indeed, this movement out of agriculture into non-agricultural activities accounts for much of the positive effect of population shifts on poverty. Several sectors experienced increases in their population shares over time, including construction, commerce, transport, finance, and servicesand public administration. Commerce experienced the largest single increase in population share, increasing from roughly 12 percent to 16percent between 1989 and 2004. 75These labor intensities are to a large consistent with those in Loayza and Radatz (2005) who are basedon cross-country data. 55 Figure 3.11: RelativeLabor Intensity, by Sector andPlanning Regionin Costa Rica, (2004) 90 -Min -Max *CostaRica - Source: World Bank staff calculations 3.54 The notion that growth in different sectors has a different impact on poverty reduction is reinforced by the findings presented in Table 3.9. This table reports the results of regressing changes in headcount poverty on sectoral income growth, interacted with the share of each sector in the total income of each (observed) canton.76 The parameters in the table can thus be interpreted as sectoral elasticities of poverty reduction (i.e, the percent change in headcount poverty as the result of 1 percent change in production of the sector). Inspection of Table 3.9 indicates that four sectors that are likely to have significant poverty reducing effects: agriculture and construction (the two most labor intensive sectors), industry (although its average labor intensity occupies the middle a ground, its range i s actually quite similar to that of agriculture), and services. Among these sectors, construction can be interpreted as being the most pro-poor, in the sense of having a large sectoral growth elasticity of poverty reduction. Incontrast, growth in transport, commerce and finance (the least labor intensive sector) can be interpreted as not being particularly pro-poor. Among these sectors, only the coefficient on transport is statistically significant, and then only in the OLS regression; none of these sectors are significant inthe robust estimation. 76The results are presented using standard OLS as well as using a robust estimationtechnique that takes into account the possibility of outliers. 56 Table 3.9: Poverty ReductionandSectoral Growth 11 OLS Robust Agriculture -1.25 * -1.27 * -2.48 -3.I5 Construction -2.75 * -2.31 * -3.10 -3.24 Transport 2.42 * 0.84 2.24 0.97 Commerce -0.19 -0.20 -0.39 -0.51 Industry -2.42 * -1.58 * -4.00 -3.24 Services -1.36 * -1.37 * -2.46 -3.09 Finance -0.37 0.13 -0.41 0.I8 1/The dependent variable is the annualizedgrowthrate of headcountpoverty between2004-1989. The independent variables are per capitareal income growthindifferent sectors of cantons weightedby the share of this sector incomein total income over the same period. * Significant at the 5 percent level. Source: World Bank staff calculations 3.55 Another key factor affecting the poverty-reducing impact of sectoral growth is the level of growth itself. As noted above, a high growth elasticity of poverty in the absence of growth will not be very effective in reducing poverty. Figure 3.12 shows sectoral growth rates for Costa Rica, both per capita household income growth (using the EHPM) and per capita GDP growth (using the national accounts data). Reported growth rates are weighted by economic relevance of the sector, so the rates can be understood as the sectoral contributions to observed growth. According to the EHPM data, the three sectors that have grown the most over the past years are finance, commerce and services and public administration. This is true for the whole period 1989- 2004 and for most of the sub-periods under consideration. Note that of these three sectors, two appear to have no significant impact on poverty reduction. In other words, growth has not tended to occur in sectors that are particularly pro-poor. 57 Figure 3.12: Real growth by economic sector, by sub-period (1989-94,1994-2000,2000-04,1989-2004) 2 ) ..................................... . .. .... ........IWC2Om .............. .. ..... .. . .... ... P hmmc tHS1 mGDPINA) T''S&m (HS) RGDPiNA1 O 1 2wo 2004 I 1 1989.2004 2 0 IIlncornc iHS1 8GDPiNA) I 5 I O O S 00 O S e - a Y ) Ite: HS refers to estimates from the EHPMhouseholdsurvey, while NA refers to estimatesfrom NationalAccounts. Source: World Bank staff calculations 3.56 One final factor that affects the power of growth to reduce poverty at the sectoral level i s the inequality trend. If growth occurs in a sector without reducing poverty, then distributional changes may be a factor. Following Datt & Ravallion (1992), Table 3.10 reports the results of an analysis that decomposes changes in poverty by sector into their growth and distributional components. The results indicate that for the 1989-2004 period, at the national level, the only sector in which the distributional component has contributed to poverty reduction is mining- a sector where only 0.2 percent of the Costa Rican population i s employed. In all other sectors, the distributional effect worked against the growth effect, serving to dull the positive effect of growth on poverty. The adverse distributional effects are particularly marked in construction, finance and transport, where the distributional component more than offset the growth component.77 77 At least some of the adverse distributional effect inthese sectors likely stems from shifts of poor people out of agriculture and into construction, finance, and transport, as part of a broader structural change in the Costa Rican economy over the last 15 years. Incontrast to agriculture, whose population share declined dramatically over the 1989-2004 period (from 26 to 15 percent of the population), population shares in these sectors have increased over time. 58 y 1p: m f 9 f p! "! x I c.5 - 0 W ' 59 3.57 At the same time, it is important to note that the size of the growth component - and hence growth's impact on poverty - over the 1989-1994 dwarfs the distributional component (Table 3.10). Hence, growth is clearly the driver of poverty reduction duringthat period. This i s not the case after 1994, however. Between 1994 and 2000, the growth and distributional components basically offset one another, while from 2000-2004, the distributional component is larger than the growth component. Sources of growth: Sources of poverty reduction? 3.58 The previous section focused on patterns of growth and poverty reduction. Building on that analysis and on the recent empirical literature on growth, this section explores the determinants of growth and income inequality in Costa Rica and how recent developments in the key growth determinants have affected poverty trends. 3.59 Most of the factors that affect growth - whether domestic policies and investments or external factors - likely also have an impact on the income distribution. In some cases, the positive impact that a policy has on growth will be reinforced by its impact on the distribution of income (so-called "win-win" policies). In other cases, it is possible that pro-growth policies may be associated with higher income inequality. As discussed earlier, when growth i s accompanied by a widening of the income distribution, then the poverty-reducing impact of that growth i s less that it would be inthe presenceof distributionally neutral or equality-enhancing growth. It i s thus important to understand the growth and distributional impacts of growth-promoting policies as part of a country's poverty reduction strategy. Under certain circumstances, complementary policies, programs, and investments may be warranted to ensure that the broadest possible segment of the population-including the poor -can benefit from growth. 3.60 The analysis presentedhere buildson a recent study by Loayza, Fajnzylber, andCalderon (2002), Economic Growth in Latin America and the Caribbean: StylizedFacts, Explanations, and Forecasts. That study uses a multi-country database to relate per capita GDP growth to several determinants of economic growth that has received significant attention in the academic literature and in policy circles. These variables can be divided into four main groups: (i) structural policies (including education, financial depth, government burden, public services and international trade openness), (ii)institutions, (iii)macroeconomic stabilization (including inflation, output volatility, banking crisis), and (iv) external conditions (including terms of trade, and exchange rate over~aluation).'~ 3.61 Loayza, Fajnzylber, and Calderon estimate the determinants of per capita GDP growth over five sub-periods between 1960 and 2000. The parameterestimatesresulting from this model are shown in the first column of Table 3.1 1. The results indicate, in line with most of the empirical literature, that countries with higher educational attainment, better infrastructure, a more developed financial sector, and a more open trade regime tend to grow faster, controlling for other factors. The results also indicate that countries with smaller governments, better institutions, and better macroeconomic management tend to grow faster, all other things being equal. ''The econometric specification also includes variables that capture cyclical reversion and transitional convergence. For further discussion of the methodology and of technical details, see Loayza, Fajnzylber, and Calderon (2002). 60 3.62 The second column of Table 3.11 reports the results of estimating a similar model but with income inequality, as measured by the gini coefficient, as the dependent variable. These estimates indicate that for most of the policy areas considered, the inequality effect reinforces the growth effect. Indeed, the results indicate that such factors as education and infrastructure serve both to increase growth and to reduce income inequality (controlling for other factors). At the same time, there are some factors that increase growth that also appear to widen the income distribution. For example, while a more developed financial sector, an economy more open to international trade, and smaller government are all associated with faster growth, they also appear to be associatedwith higherlevels of income inequality. Table 3.11: The Determinantsof Growth and Income Inequality inLatinAmerica, 1960-2000 Variable Growth Change in logged gini lagged Inequality -0.242 (13.32) Initial GDPper capita -0.018 (3.80) Initial output gap -0.237 (8.52) Education 0.017 -0.022 (6.7) (2.77) Financial depth 0.006 0.014 (4.28) ,(2.83) Trade openness 0.01 0.024 (3.14) (3.04) Government burden -0.015 -0.018 (3.18) (2.71) Public Infrastructure 0.007 -0.016 (2.71) (3.32) Governance -0.001 0.005 (0.68) (1.74) Price Stability -0.005 0.008 (1.89) (2.16) Cyclical Volatility -0.277 0.112 (3.76) (1.41) External imbalances -0.006 -0.002 (3.90) (0.32) Banking crisis -0.029 -0.021 (7.42) (4.02) External conditions 0.072 0.051 (4.98) (1.87) Notes: t-stat inparentheses. Sources: Loayza, Fajnzylber, and Calderon (2002),Lopez (2004) 61 3.63 The results presented in Table 3.11 are based on analysis of countries across Latin America, but what can be said about the main drivers of growth in Costa Rica in recent years? Loayza, Fajnzylber, and Calderon (2002) simulate the effect of several key drivers of growth and conclude that, at least during the second half of the 1990s, the most important forces for growth in Costa Rica were progress on the trade and infrastructure fronts (Table 3.12).79 Incontrast, increased output volatility over the period worked to reduce aggregate growth.80 3.64 Drawing on the findings presentedin Table 3.11, Table 3.12 reports the simulated effects that progress on these seven drivers of growth had on inequality. The results of this analysis suggest that changes in several drivers of growth may also have contributed to increases in inequality. It i s important to note, however, that their contributions to inequality are much less than their contributions to growth. Infact, the table indicates that the combined effect of changes in these seven factors on the gini coefficient was only 0.13 percent (Table 3.12). At 0.28 percent, their combinedeffect on growth was more than twice as large. Table 3.12: ExplainingRecent Changesin Growth and Inequality in Costa Rica 1/ Area Growth Ineaualitv Education 0.08 -0.10 Financial sector 0.06 0.14 Trade opennes 0.19 0.46 Government burden 0.02 0.02 Infrastructure 0.20 -0.46 Inflation 0.03 -0.05 Volatility -0.30 0.12 Total 0.28 0.13 I / The table reports the contribution to growth and inequality (both in percent) as a function of recent progress made in key macro- and structural policy areas. The results are presented for the period 1966- 1999and are with respect to 1991-1995. Source: Loayza, Fajnzylber and Calderon (2002) and World Bank staff calculations 3.65 What does the above analysis say about the appropriate mix of pro-poor growth policies in Costa Rica?Recent evidence from the Latin America region suggests that attaining sustainable growth i s not only a function of implementing appropriate growth-promoting policies individually, but also establishing the right overall policy mix (Gallego and Loayza 2002). This suggests that a poverty reduction strategy that focuses solely on pro-growth policies that reduce inequality, such as education or infrastructure, and not on addressing bottlenecks in other areas such as strengthening the financial sector or improving the policy environment for increased international trade may well fall short (Box 3.2). Indeed, the evidence suggests that a development strategy that focuses on only a subset of a country's key policy challenges is likely to produce disappointing results with respect to growth and, ultimately, with respect to poverty 79 Loayza, Fajnzylber, and Calderon's econometric analysis is based on non-overlapping five-year averages between 1960 and 2000. Due to data limitations, the 2000-2004period could not be included in the current analysis. 8o Note that the figures in the Total row do not equal the observed growth rate over the period, but rather reflect the marginal contribution of the policies being analyzed (controlling for other factors). 62 reduction. A more productive approach would involve the implementation of complementary policies, programs, and investments designed to enhance the ability of the poor to take advantage of growth and the emerging economic opportunities that come with it. As i s discussed further in the coming chapters, this will involve efforts to strengthen the human capital of the poor (for example, by improving secondary school enrollment and attainment) along with initiatives to facilitate more productive participation of the poor in Costa Rica's labor market. It also will involve providing assistance to the most vulnerable groups to help them adjust to changing global and domestic economic conditions and to ensuretheir access to basic services. Box 3.2: Growth, Poverty, and the Central American Free Trade Agreement, DR-CAFTA A central factor in determining the future economic prospects of Central American countries is likely to be the ratification and implementation o f DR-CAFTA, the free trade agreement negotiated by Costa Rica, along with the Dominican Republic (DR), El Salvador, Guatemala, Honduras, and Nicaragua with the United States. DR-CAFTA is important, not only because the U.S. i s the major trading partner with the Central American countries, but becausethe treaty holds the potential of increasing trade and investment in the region. That, in turn, will be a key to increasing economic growth and improving the welfare of people inthe region, including those living in poverty. A recent World Bank study (2005) presents a preliminary assessment of DR-CAFTA. While quantifying the precise effects of any free trade agreement is challenging, the study draws on a number of different approaches and methodologies to reach its conclusions. The report finds that DR-CAFTA is likely to improve growth levels for the participating countries in Central America by increasing trade and investment levels. Greater trade levels are expected to result from the removal of virtually all remaining tariff and quota barriers to trade among all parties, consolidating - and in some cases expanding - the preferential market access that Central American countries have enjoyed in U.S. markets through the Caribbean Basin Initiative (CBI) program. DR-CAFTA is also expected to deepen regional trade integration (and increase trade levels) among the Central American nations themselves (and with the Dominican Republic). DR-CAFTA is expected, as well, to promote greater foreign and domestic investment by improving the certainty of these countries' market access with the US.,solidifying the broadeconomic reforms ofrecent years and spurringfurther reform efforts. Higher levels of investment and economic growth are, in turn, expected to contribute to long-term poverty reduction in Central America. Nonetheless, as has been found in the case of other recent trade agreements, the gains associated with DR-CAFTA will depend in part on the ability of the Central American countries both to adjust to the changes that the agreement will bring (including changes in relative prices) and to be proactive in handling the expected structural changes in their economies. Indeed, the study suggests that the magnitude of the benefits from DR-CAFTA will depend on the abilities of the Central American economies to pursue a "complementary p,olicy agenda," as the agreement's benefits are expected to be larger if accompanied by parallel efforts in areas like trade facilitation (e.g., ports, roads, and customs), institutional and regulatory reforms, investments in economic infrastructure, and investments in education and innovation. But what are the likely short-run impacts of DR-CAFTA on people's welfare? While EHPM survey data would not support in-depth analysis of the expected welfare impacts in Costa Rica, detailed analysis of the expected welfare impacts was undertaken for El Salvador, Guatemala, Honduras, and Nicaragua. Those findings can support inferences about the likely effects in Costa Rica (World Bank 2005, Marques 2005). The analysis indicates that the vast majority of the population in Central America is likely to experience welfare gains from implementation of DR-CAFTA, even in the short run. For example, 90 percent of Nicaraguan households, 89 percent of Honduran households, 84 percent of Guatemalan households, and 68 percent of Salvadoran households, respectively, were found to consume more than they produce of the basket of sensitive agricultural commodities to be liberalized under DR-CAFTA. This means that they can be expected to benefit from price changes induced by the treaty. In contrast, only about 8 percent of the Honduras households, 9 percent of Nicaraguan households, 16 percent of Guatemalan households, and 5 63 percent of Salvadoranhouseholdswere foundto produce more than they consumeof the basket of sensitive commodities and, thus, would be expected to experience welfare losses from DR-CAFTA related price changes. While the proportionof Central Americans who are likely to be adversely affectedby the treaty is relatively small, several provisions have been put in place to help protectthose who might be vulnerable to negative impacts. DR-CAFTA was negotiated to include a wide range of provisions to delay and soften the impact of liberalizing sensitive agricultural commodities. These provisions include grace periods for initiating liberalization, extended phase-out periods for tariffs, interim quotas and/or phase-downs of tariff-rate quotas, as well as special safeguardmeasures to protectlocal farmers from undue harm. Indeed, phase-out periods are as long as 20 years for some goods and, at least for a sub-set of the countries, white maize (an important staple crop produced by poor farm households) is exempt from liberalizing. Such provisions provide important protections for (net) producers of sensitive crops, giving them an extended timeframe over which to make necessary economic adjustments. The analysis suggests, as well, that complementary investments in education, rural infrastructure, rural finance and technical assistance, as well as targeted safety nets when necessary, can go far in ensuring that the poor, particularly in rural areas, have the means to take full advantage of the new opportunities arisingfrom DR-CAFTA. ISources: World Bank (2005), MasonandMarques (2006) Conclusion 3.66 This chapter has examined the links between economic growth, changes in the distribution of the benefits of growth, growth related policies, and Costa Rica's performance (or lack of it) in reducingpoverty between 1989and 2004. While GDP and householdincome growth rates were higher before 1994 than after, several other factors worked to mitigate the impact of growth (at any given level) on poverty over the last decade. Most importantly, growth in Costa Rica has been accompanied by increases in income inequality. These increases were not enough to offset reductions in poverty prior to 1994, when per capita income growth (and poverty) levels were relatively high, and when the benefits of growth were relatively evenly distributed. But that has not beentrue since 1994. Post-1994 growth has been higher among upper income households than among the poor; and since 2000, per capita income has actually declined for poor Costa Ricans. Indeed, between 2000 and 2004, only households in the upper income quartile experienced positive per capita income growth. 3.67 Uneven distribution of the benefits of growth can also be seen across the different regions of Costa Rica as well as in the sectoral patterns of growth over the period. EHPM data indicate that since 1994, growth has tended to be faster in relatively wealthy cantons than in relatively poor ones. Over the same period, growth also has tended to be higher in sectors such as finance, commerce, and public administration, which tend not to employ a lot of low-skilled workers, and slower in sectors like agriculture, construction, manufacturing, and services. For that reason, the sectoral pattern of growth has also not beenparticularly pro-poor. 3.68 Analysis of the links between macro- and sectoral policies and growth indicates that several types of policies and investments would promote both higher growth and greater income equality. The most important of these investments are in education and infrastructure. At the same time, there are several important drivers of growth that may serve to widen income distribution over time, particularly in the absence of a sufficiently well-educated workforce. For example, the analysis suggests that policies focused on financial sector deepening or on greater trade openness, while important engines of economic growth, may also contribute to higher income inequality. 64 3.69 A key policy message emerging from the analysis is not, however, to avoid pro-growth policies that may also widen the income distribution (at current education and infrastructure levels). Indeed, international evidence suggests that it is not only individual growth-promoting policies, but the overall policy mix, that i s important to effective and sustained growth. A more productive approach would thus be to design and implement policies, programs and investments that would better position the poor to take advantage of growth and the resulting economic opportunities that arise. This will involve efforts to strengthen the human capital of the poor, as well as to facilitate more productive participation of the poor in Costa Rica's labor market. IT will also involve the implementation of social protection policies that help the poorest, most vulnerable Costa Ricans groups adjust to changing global and domestic economic circumstances. 3.70 Chapter 4 examines the functioning of the labor market and how to enable the poor to participate in the labor market more effectively and remuneratively. Part I1 of this report (Chapters 5-8) then focuses on approachesto strengthening the human capital of the poor and on how social policy can enable the poor to participate more effectively in - and enjoy the fruits of - future economic growth. 65 4. POVERTY AND THE COSTA RICANLABOR MARKET 4.1 Labor earnings make up the vast majority of households' income in Costa Rica.81 So understanding the labor market is critical to understanding the incomes of the poor and the evolution of poverty in the country. This chapter examines the mechanisms by which earnings are transmitted to the poor, and explains how developments in the labor market have affected the ability of the poor to escape poverty. It focuses particularly on the years since 1994, when Costa Rica's poverty rate stopped declining. The chapter also analyzes the interaction between key labor market policies and poverty, and it discusses emerging demographic trends and their effects on labor and poverty. It pays special attention to the effects of Nicaraguan immigrant workers, as well as to recent increasesinthe portion of Costa Rican households that are headedby females. 4.2 A number of forces since the early 1990s have made it more difficult for Costa Rica's poor to escape poverty. Structural changes in the Costa Rican and global economy have increased demand for skilled workers relative to unskilled workers, and have raised relative earnings of people with secondary and higher education. At the same time, the proportion of secondary school graduates in the labor force has declined, while the number of secondary school drop-outs has increased. Shifts in relative demand for skilled and unskilled labor have resulted in significant increases in unemployment and part-time work among the poor and extremely poor. Indeed, between 1994 and 2003, unemployment rates rose from 8 percent to 17 percent among people living in poor households, and from 12 percent to 27 percent among people in extremely poor households, while unemployment rates among people in non-poor households remained at or below 5 percent. The evidence suggests that unemployment i s largely structural - that poor people, who often are poorly educated, lack the skills that firms are demanding. 4.3 These factors have contributed to a widening of earnings inequality in Costa Rica. Substantial increases in the number of female-headed households have also contributed to observed trends in employment, earnings and poverty; indeed, because of a lack social infrastructure to help single mothers hold higher paying, full-time jobs (early childhood development programs and daycare facilities, for instance), there has been a disproportionate rise in part-time work among poor female workers and in poverty among female-headed households. The evidence also suggests that some labor market policies - such as the country's minimum wage policy - may contribute to the situation by limiting productive employment opportunities among poor workers. 4.4 Inshort, analysis of the labor market bolsters the macro-levelevidence suggesting that policies and investments are needed - in general education, skills training, and better social services (to support single, working mothers) - to help the poor earn greater incomes in the face of a changing national and global economic environment. The evidence presented in this chapter is based on analysis of Costa Rican household survey (EHPM) between 1987 and 2004. The 2004 EHPM data indicate that labor earnings make up nearly 90 percent of total Costa Rican household income, and roughly three-quarters of the income o f the poor. While these are almost certainly overestimates due to lack of information or under-reporting of other components of household income, there is no doubt that labor income is the single-most important component of household income for most households. 66 Poverty and the Recent Evolution of the Labor Market in Costa Rica 4.5 This section examines the evolution of real earnings, unemployment, labor force participation and the structure of employment in Costa Rica. Real Earnings 4.6 Real earnings track the business cycle in Costa Rica: they fell from 1987 to the trough of the recession in 1991, then rose during the recovery from 1991 to 1994, fell from 1994 to the trough of another recession in 1995, rose from 1996 to 2002, and finally fell again from 2002 to 2004 (Figure 4.1). This pattern is the same for the earnings of salaried employees, self-employed workers, men or women. Poverty rates, meanwhile, tracked the change in real earnings prior to 1994; they rose during the recession of the early 1990s and then fell during the recovery from 1991 to 1994. After 1994, however, poverty rates stagnated, failing to fall even when real earnings increasedfrom 1996to 2002. Figure 4.1: Real Monthly Labor Earnings (1999 colones) Source: Gindling(2006), basedon EHPMdata from 1987-2004 Unemplovment 4.7 The puzzling combination of rising real earnings and stagnating poverty i s partly explained by a rise in unemployment rates from 1994 to 2002, especially among those most vulnerable to poverty. National unemployment rates were counter-cyclical prior to 1996; they rose with the recession of the early 1990s, fell with the recovery until 1994 (to 3.5 percent) and then rose again during the recessionfrom 1994 to 1996 (to above 6 percent in 1996; Figure 4.2). However, despite rising per capita GDP and rising average real earnings and incomes after 1996, unemployment rates remained high (6 percent to 6.5 percent) until 2004. After 1994, unemployment rates increasedfor both men and women. There are some differences by gender, however; in all years, the unemployment rate for women was higher than for men, and unemployment rates rose slightly more for women than men from 1994 to 2004. 67 Figure 4.2: National Unemployment Rate by Gender 9 8 7 6 5 4 3 2 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Zoo0 2001 2002 2003 2004 -Both Genders -I- Men Women Source: Gindling (2006), based on EHPM data from 1987-2004 4.8 The high and rising unemployment rates during the period of growth in the late 1990s and early 2000s were especially marked for poor households. Figure 4.3 shows that while unemployment rates for people living in non-poor households remained slightly below 5 percent for the entire 1996-2004 period, unemployment rates increased steadily and dramatically for people living in poor households. For all poor, unemployment rates increased from below 8.1 percent in 1996 to 16.7 percent in 2003. For the extremely poor, unemployment rates more than doubled during this period, from 12percent to over 27 percent. Figure 4.3: Unemployment Rates by Poverty Status, 1987-2004 0 I 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 [+'Extreme Poverty *All Poor -A- Non-Poor 1 Source: Gindling (2006), basedon EHPMdata from 1987-2004 68 4.9 The fact that unemployment rates remained high despite the recovery of the late 1990s suggests that there was an increase in structural, rather than cyclical, unemployment. Cyclical unemployment results from a fall in aggregate demand caused by an economic downturn such as a recession; it normally i s of short duration. Structural unemployment results when the skills of the unemployed are not the skills demanded by employers in a changing economy; it is likely to last a long time. Evidence of such skill-matching problems can be seen by comparing changes in unemployment rates for workers with different education levels (Gindling2006). 4.10 The data suggest that the rise in unemployment rates for the poor from 1996 to 2002 reflected rising unemployment for low-skilled laborers. As Figure 4.4 demonstrates, unemployment rates had followed the same pattern for all education groups prior to 1996. But from 1996 to 2002, unemployment rates rose for the less-educated, and fell for the more educated (i.e., those with a secondary school or university degree). By 2002, unemployment rates for secondary school drop-outs and those with just primary-school education were substantially higher than for those with secondary school degrees. This suggests that the economic growth in the late 1990s did not lead to increased employment opportunities for low-skilled, less-educated workers. At the same time, as will be seen below, relative earnings for such workers fell during this period. Along with the change in unemployment patterns, this suggests that the demand for low-skilled workers declined and the supply increasedduringthe period. Figure 4.4: Unemployment Rate by Education Level, 1987-2004 10 9 8 7 6 5 4 3 2 1 0 I t Primarialncompleta -+-PrimariaCmpletatSecundariaincornpieta/ I X SecundariaCompleta -X- Supenor Source: Gindling (2006), basedon EHPMdata from 1987-2004 Labor Force ParticiDation 4.11 From 1987to 2004, labor force participationrates increasedfor women and decreasedfor men (Table 4.1). These changes counteracted each other so that, on average, labor force participation rates changed very little over the 1987-2004 period. The increase in labor force participation rates was especially marked for women after 1996. Rates for women in non-poor families (40 percent in 2004) were greater than for women in poor families (21 percent), although the pattern of change was similar for both. 69 1987 1992 1994 1996 1999 2002 2004 Both Genders 53.9 51.5 53.1 52.2 54.8 55.4 54.4 Men 78.8 74.0 75.3 73.7 75.1 73.2 73.0 Women 29.4 30.0 31.6 31.1 35.5 38.2 36.8 4.12 Increasing labor force participation by women appears to have contributed to the high rates of unemployment from 1996 to 2002. Gindling (2006) broke changes in unemployment rates into three components: (1) changes in labor force participation rates, (2) changes in the probability of unemployment given non-employment (unemployment plus labor force non- participation), and (3) changes in the non-employment rate (unemployment plus labor force non- participation as a proportion of the population over 12 years old).** The first two components are related to increases in labor force participation rates, while the last i s related to changes in the demand for labor. Because labor force participation rates were increasing for women and falling for men, the decomposition was done separately for men and women. 4.13 The results suggest that the causes of the highunemployment in the late 1990sand early 2000s were different for men and women. For women, essentially all of the increase in the unemployment rate from 1994 to 2002 can be explained by rising labor force participation rates. Indeed, non-employment rates for women actually fell between 1994 and 2002. T h i s means that if there had been no increase in labor force participation rates, unemployment rates would have fallen for women from 1996to 2002. 4.14 For men, on the other hand, labor force participation rates fell, and non-employment rates increased from 1994 to 2002. The rise in unemployment rates for men during this period, then, cannot be explained by increases in labor force participation rates. Indeed, men already in the labor force were the ones who lost their jobs (Gindling2006).83 Salaried. Self-employed and UnpaidFamily Workers 4.15 The proportion of poor workers who were self-employed increased from less than 30 percent to more than 40 percent from 1987 to 2002 (it then fell slightly from 2002 to 2004). In poor families, this "informalization" of employment was especially pronounced among female workers; the proportion of poor female workers who were self-employed increased from less than 20 percent to over 40 percent between 1987 and 2004. The increase in the proportion of workers, especially women, who are self-employed, may have contributed to stagnating poverty rates in Costa Rica inrecent years. 82 The approachusedin Gindling(2006) follows Card and Riddell (1993). 83 These findings are reinforcedby analysis of employment as a proportionof the working age population (the inverse of the non-employmentratiosdescribedin the text). As aproportionof the working age population, employment fell for menbut increasedfor women throughoutthe 1987-2004 period. The increaseinthe employment of women was most rapidin the 1996-2002period. This suggests that high unemploymentrates for 1996to 2002 were caused by falling employment opportunities for men, and by the increaseinlabor force participationamong women (Gindling 2006). Similar analysis by povertygroup indicates that employment/populationratiosrise for the non-poorbut fall for the poor in the later 1990sand 2000s. This is further evidence that employmentopportunities for workers in families vulnerable to poverty declined inthe 199Os,while employment opportunities for workers innon-poor families increased. 70 Figure 4.5: Percent of Workers ineach Category who are Self-employed, 1987-2004 45 1 40 35 30 25 20 15 10 5 0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 t Female Poor -Male Poor t Non-Poor Source: Gindling (2006), basedon EHPMdata from 1987-2004 4.16 It i s worth noting that the proportion of non-poor workers in the low-paying, self- employed sector also increased from 1987 to 2004, although the increase was less dramatic than for poor workers: it went from 20 percent to 26 percent over the period. Part-time vs. Full-timeWork 4.17 From 1987 to 2002, the proportion of people working a standard work week (40 to 48 hours per week) decreased substantially - from 40 percent to 31 percent - while the proportion working part-time increased. The latter increaseswere driven by a dramatic rise in the proportion of poor women working part-time - from just over 40 percent in 1987 to around 65 percent in 1995, where it has stayed (more or less) over the last decade (Figure 4.6). The proportion of part- time workers increased for no other group (non-poor women, poor men or non-poor men). Because part-time workers earn less than full-time workers for any given hourly wage, these patterns suggest that the increase in part-time work among poor women may have contributed to the stagnation of poverty rates since 1994.84 84The share of workers working over-timealso increased during the period. This occurred because of an increase in the proportion of non-poor men working over-time - from 31 percent in 1987 to about 43 percent in 2004 (Gindling 2006). The proportion of over-time workers increasedfor no other group (e.g., poor men, poor women or non-poorwomen). 71 Figure4.6: Percentageof Poor Working Women Working Part-time,1987-2004 r I 70 - 60 50 40 / 30 20 10 1987 1989 1991 1993 1995 1997 1999 2001 2003 1 -+Part-time +Full-time -A- Over-time 1 Source: Gindling(2006),basedon EHPM data from 1987-2004 Industrial Structure of Employment 4.18 Two industries experienced significant increases in employment between 1987 and 2004: services and commerce (Figure4.7). That probably reflects the explosive growth of international tourism in Costa Rica in the 1990s(hotels and restaurantsare includedin "commerce," while tour guides and other tourist services are included in "services"). A smaller yet significant increase in employment also occurred in financial services and real estate. The only industrial sector to lose workers over this period was agriculture. Indeed, the share of workers employed in agriculture dropped significantly over the period, from about 28 percent in 1987 to 15 percent in 2004. Figure4.7: Employmentby IndustrialSector, 1987-2004, All Workers 500000 450000 - 400000 -r- 350000 300000 250000 200000 150000 100000 50000 0 1987 1989 1991 1993 1995 1997 1999 2001 2003 +Agriculture -+-Electricity, .-+e- Mining +Manufacturing gas and water -+-Construction -Commerce -Transport and communication 4 F i n a n c e and Raai Estate +Services jource: Gindling(2006), based on EHPMdata from 1987-2004 72 4.19 The evolution of mean real earnings by industrial sector from 1987 to 2004 is presented in Figure 4.8. It shows that the highest paying industrial sectors are financial services, transportation andutilities, while the lowest paying sector is agriculture. Interestingly, the sectors that experienced the biggest increases in employment financial services, commerce and services) did not experience very large wage gains. Over the 1987-2004 period, average real earnings actually declined in commerce, and they increased by only 0.6 percent and 3.3 percent in financial services and service, respectively. The biggest increase in average wages occurred in a sector that was losing employment relative to others: manufacturing. This suggests that although the share of employment in manufacturing declined over the period, there were substantial increasesinproductivity inmanufacturingover the period. Figure 4.8: Mean Earnings by IndustrialSector in Costa Rica, 1987-2004,All Workers (1999 colones) 35WOO 300000 25WOO 2wooo 150000 1WWO 5oooo 0 1987 1989 1991 1993 1995 1997 1999 2001 2003 --cAgrculture Manufacturing --i- Bectrictty gas andwater --ICConstructon --eComrce +Transport andcomrunicalion -finance and Real ktale --- Servces Source: Gindling (2006),based on EHPMdata from 1987-2004 4.20 As noted in Chapters 2 and 3, poor workers are disproportionately represented in agriculture. Indeed, more than one-third of all poor workers and nearly half of all extremely poor workers can be found in agriculture - although the share of poor workers in agriculture declined dramatically over the period (Gindling 2006). Another roughly 20 percent of poor workers were employed in commerce and services in 2004. This representeda doubling of the share of poor workers inthat industrialsector from 1987. Distribution of Incomefrom Salaried Employment and Self-Employment 4.21 The proportion of labor income from self-employment increased from 23 percent to 26 percent between 1987 and 2004. The increase was larger for women than men and for workers in poor families than for workers in non-poor families (Gindling 2006).85 For men, the proportion of labor income from self-employment increased as the share of men engaged in self-employed 85 The proportion of labor income from self-employment was relatively constant from 1987 to 1990, increased from 1990 to 2001, and then fell slightly from 2001-2004. 73 0 2 0.2 0.1 0.4 0.o -0 1 -0.1 73 From 1987 to 1992, earnings inequality continuedto fall, although at a slower rate than in the previous period. These changes occurred amid falling real earnings and hourly wages. From 1992 to 1999, real earnings rose substantially, with increases in real earnings being larger for each successively higher decile in the distribution. As a result, earnings inequality increased. From 1999 to 2002, inequality continued to increase, as earnings rose more at higher deciles than at lower deciles, in an environment of stagnating real earning^.^' From2002 to 2004, the real earnings of workers in the bottom three or four deciles of the distribution increased, while the real earnings of workers in the other deciles fell. This resulted ina narrowing inearnings inequality from 2002-2004.88 Factors Driving; RisingIneaualitv inMonthly Earnings89 4.26 Several techniques exist to break changes in inequality in monthly earnings down into components attributable to changes in the personal and workplace characteristics of workers. Gindling (2006) and Gindling and Trejos (2005) undertake two such decompositions: (i) the Fields decomposition, which enables one to estimate how changes in the levels of selected variables affect changes in observed earnings inequality, and (ii) the Yun decomposition, which enables one to estimate how changes in the levels and the variance of selected variables affect earnings ineq~ality.'~These analyses examine the factors affecting earnings inequality in five distinct periods: 1980-1985, 1987-1992, 1992-1999, 1999-2002 and 2002-2004. The decompositions examine the effects of several factors - including changes in education among the Costa Rican workforce, changes in the number and distribution of hours worked in the workforce, changes in the gender composition of the labor force, returns to work in different industrial sectors, employment in small as opposed to large firms, employment in the private versus the public sector. 4.27 These decomposition analyses indicate that three main factors drove changes in earning inequality (both inpositive and negative directions) between 1980and 2004: Changes ineducation among workers over time, Changes inreturns to education over time, and 87Accordingto the EHPMdata, averagereal earnings during this periodincreased, while accordingto Social Security data average real earningshave fallen (JuanDiegoTrejos, personalcommunication). Ifone looks at the distributionof all labor earningsinthe EHPM (as opposed to lookingonly at the earnings for paidemployees), one sees that averagereal earningsfor those inthe bottomfour deciles of the distributionfell while the average real earnings for those inthe top six decilesrose. 88Accordingto the EHMP, average real earnings fell duringthis period,while accordingto Social Security data averagereal earnings increased(Juan DiegoTrejos, personalcommunication). 89When CostaRicansrefer to earnings,they almost always refer to monthlyearnings. Yearly earnings for Costa Ricanpaid employeesinclude 12 months of pay plus a legally-required13" month bonus (aguinaldo),which is paidinDecember. Self-employed workers are obviously not paidthis bonus. This will create some non-comparability betweenthe reported monthlyearnings of paidemployees and self- 75 Changes among workers in hours worked. 4.28 These three factors had different effects over different periods. Gindling (2006) and Gindlingand Trejos (2005) find that a decline in returns to education from 1980 to 1985 was the most important factor associated with the decline in earnings inequality during that period. Moreover, they found that earnings inequality stopped declining after 1987 (until 2002) in part because returns to education stopped falling (and even increased somewhat between 1992 and 2002. From 1987 to 2002, the analysis suggests, an increase the variance of hours worked by Costa Ricans was the most important cause of rising earnings inequality. From 2002-2004, earnings inequality diminished for several reasons: returns to education began to fall again, the variance of hours worked among workers fell, and the distribution of education among workers becamemore equal. 4.29 Changes in the distribution of workers across large and small firms, across public- and private-sector jobs, and across industrial sectors had some, albeit smaller, impacts. For example, increases in the proportion of the work force employed in lower-paying small firms (ones with fewer than five workers) also contributed to increases in inequality from 1992 to 2002. The r) proportion of workers in small firms increased from 45 percent in 1992 to 50 percent in 2002. From 2002 to 2004 the proportion of workers in small firms fell from 50 percent to 46 percent, contributing to the reduction inearnings inequality in this period. 4.30 Declines in the wage premium paid to public sector workers, possibly due to the economic reforms undertaken since the early 1980s, appear to have helped equalize earnings over the 1980-2004 period. Changes related to the distribution of workers between industrial sectors were small over the entire period, suggesting that most of the change due to changes in industrial structure were attributable to within-industry shifts rather than to shifts in the distribution of workers across sectors; such changes were slightly equalizing in the 1987-1992 and 2002-2004 periods, but may have had a slightly disequalizing effect duringthe 1992-2002 period." 4.31 Insum, the Fieldsand Yun decompositions suggest that changes in inequality between 1980 and 2004 in Costa Rica were caused largely by three phenomena: increases in education inequality among workers, changing returns to education in the economy, and increases in the variance of hours worked among workers. These phenomena are examined in more detail in the following sections. IncreasesinEducational Inequality inthe Workforce 4.32 In the 1990s and early 2000s, the proportion of university-educated workers in the workforce increased, and the proportion of middle-income, secondary school graduates declined. These changes were driven by both domestic phenomena (a decline in real public spending on secondary and primary education and an increase in the number of private universities) and international phenomena(an influx of less-educatedNicaraguan migrants). 4.33 Patterns of educational inequality in the workforce changed over different sub-periods. Inequality in education among workers increased from 1980 to 1985 and from 1987 to 2002, and then fell from 2002 to 2004 (Gindling 2006), contributing to increased earnings inequality from ~ ~ ~~~ ~~ ~~ employed workers. Another source of non-comparabilityis that the reportedearnings of self-employed workers are likely to includereturnsto capitalas well as labor. For methodologicaland empirical details, see Gindling(2006). 91 Whether a disequalizing effect is observed depends on the specific measure of inequality used; see Gindling (2006),Table 3-2. 76 1980 to 2002 and to the decline in earnings inequality after 2002. Indeed, from 1987 to 2002 the proportion of workers who had graduated from secondary school actually fell, while the proportion of workers who were secondary school drop-outs (and who earn below average incomes) increased (Figure 4.10). The proportion of workers who graduated from secondary school (and whose earnings fall in the middle of the distribution of earnings) increased again from 2002 to 2004. Figure 4.10: Proportionof Workers at SelectedEducationLevels, 1980-2004 25 20 15 10 5 0 Secondary Drop-Out Secondary University Complete Source: Gindlingand Trejos (2005) Gindling(2006) 4.34 The slow growth in the proportion of secondary school graduates in 1987-2002 also contributed to a slowdown in the rate of growth of average education levels among Costa Rican workers (they declined despite continuing increases in the proportion of workers with university education). The average increase in years of education among workers was 0.2 years per annum from 1980to 1985, but less than half of that from 1987 to 2002. From2002 to 2004, the increase in averageeducation levels acceleratedagain, to 0.15 years per annum. This can be seen from an analysis of the average education levels for different age cohorts in the workforce, based on the 2004 EHPM data (Figure 4.11).92 As can be seen from the figure, education levels increased substantially for each birthcohort that entered the market from 1950 to 1979 (those born between the mid-1930s and the late-1950s). Incontrast, the increase in education levels stopped for those entering the labor market in the early 1980s through the mid-1990~'~ 92Cohorts are defined accordingto the year in which a 20-year old would be expected to enter the labor force (Gindling 2006). 93Funkhouser(1998)presents a similar graph of averageeducation levels by birth cohort. 77 Figure 4.11: Average Years of Education of 20-year-olds enteringthe Workforce 10 .-cm0Ea9 8 U w -0 7 $ 6 ` Q F 5 $ 4 3 195019541958 1962196619701974 1978198219861990 19941998 Year 20-year-olds entered the work force Source: GindlingandTrejos (2005) andGindling (2006) 4.35 These cohorts correspond to those born in a "baby boom" in Costa Rica, a period of elevated birth rates (CEPALICELADE, 2001). These baby boomers began to reach secondary school age in the early 1980s. Because of the size of this large cohort, it would have been difficult for the Costa Rican educational system to provide enough school resources. Unfortunately, its entry into primary and secondary schools also coincided with a reduction in public spending on education as a result of the economic downturn of the early 1980s, the debt crisis and subsequent economic adjustment programs.94 Duringthe late 1980s and early 1990s the supply of schools, textbooks, teachers and other school resources was not able to keep up with the increasing secondary-school-aged population. Indeed, Montiel, Ulate, Peralta and Trejos (1997) show that real spending per student in academic secondary schools in 1992 was only 73 percent of spendingin 1980, while spending per student in vocational secondary schools was only 53 percent of its 1980 4.36 Average education levels began to increase again for post baby-boom cohorts, beginning with those entering the labor force in the mid-1990s. This increase was driven by a rise in the rate of college graduates entering the work force, which inturn was the result of an explosion of enrollments in new private universities (most of which opened in the late 1980sand early 1990s). Before the mid-1980s, private universities enrolled a very small percentageof university students in Costa Rica; the increase in their enrollment occurred in response to a rise in demand by potential students who could not test into the more prestigious public universities. While average education levels increased in the late 1990s, the rate of increase was slower than it was in the 94In 1998the constitution was modified to require a minimumpercent of 6% of GDP be spent on public education public education (for primary, secondary and university education-no specific distribution by education level i s specified). 95Montiel, Nancy, Anabelle Ulate, LuisPeralta and Juan Diego Trejos (1997), La educacidn en Costa Rica: un solo sistema?, Serie Divulgacidn Econ6mica No. 28, Instituto de Investigaciones en Ciencias Economicas de la Universidadde Costa Rica, San JosC, Costa Rica and Ministro de Planificacidn Nacional (1998), Gobemado en tiempos de cambio, la administracidn Figquerez Olsen,San JosC, Costa Rica. For more detail on patterns of public spending on education over time, see also Chapter 6 of this report. 78 1950s, 60s and 70s. This i s because the substantial increase in the number of Costa Rican workers with university educations between 1995 and 1999 was partly counteracted by a continuing fall in the proportion of workers with a completed secondary education (Figure 4.10). From 2002 to 2004, when the proportion of workers with completed secondary education again beganto increase, the increase in average education levels accelerated. 4.37 As noted above, an influx of less-educatedNicaraguan migrants contributedto increased inequality of education levels in the Costa Rican workforce. Indeed, Nicaraguanmigrants are, on average, less educated than Costa Rican born workers (Gindling 2006). Some 63 percent of Nicaraguan migrants have a primary education or less, compared to 48 percent at most of Costa Rica-born workers. Similarly, just six percent of Nicaraguan migrants have college educations, compared to 20-22 percent of Costa Rica-born workers). And the proportion of Nicaraguan migrants who have completed secondary education is lower, while the proportion of Nicaraguan migrants who are secondary school drop ou,ts i s higher, than for Costa Rica-born workers. By contributing to the increase in the inequality of education levels among Costa Rican workers, the influx of migrants from Nicaragua added to the increase in earnings inequality. Changes inReturns to Education: Demand. Supply and InstitutionalFactors 4.38 Returns to education inCosta Rica have changed considerably over the last 30 years: they declined considerably between 1976 and 1983, stabilized until about 1993, and then rose again until around 2002 (Figure4.12). Increasesin returns to education, coming at a time of widening inequality in education levels of the workforce, are associated with greater earnings inequality since the early-to-mid 1990s. Indeed, these increases in returns to education are reflected in increases in absolute and relative earnings for university graduates. Real earnings of university graduates increased 18 percent between 1996 and 2002. Over the same period, real earnings for workers with a completed secondary education increased by 3 percent, while real earnings for workers with less than a completed secondary education were ~nchanged.~~ Using different methodologies and different measures of returns to education (or skill), other studies have found the same pattern of change: declines from 1976 to 1983 and stability (or small increases) thereafter (Funkhouser, 1998, Robbins and Gindling, 1997 and Sauma and Vargas, 2000). Similar changes have been identified as among the most important causes of increasing inequality in the United States, other industrial market economies (see Katz and Autor, 1999, for a review) and other Latin American economies (see Inter- American Development Bank, 1998, for a review). 79 Figure 4.12: Returns to Educationin Costa Rica, 1976-2004 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 19% 1998 2003 2002 2004 +AI workers +Paid Employees Source: GindlingandTrejos (2005) andGindling (2006) 4.39 Evidence suggests that three factors lie behind the increasing returns to education in Costa Rica from 1987 to 2002: (1) increases in the relative demand for educated workers (caused by investment in new importedcapital, a complement to skilled labor); (2) a slowdown in the rate of growth of the relative supply of educated workers (caused by a slowdown in the number of secondary school graduates among Costa Ricans and by the influx of less-educatedNicaraguan migrants); and (3) institutional changes (specifically changes inthe structure of minimumwages). 4.40 A couple of recent articles shed light on the role of each of these factors between 1980 and 1992. Robbins and Gindling (1998) find evidence that labor supply factors explain much of the fall in returns to education prior to 1987, and that both increases in relative demand and slowdown in the growth of relative supply of educated workers caused returns to education to increase in the post-1987 period. Funkhouser (1998) also finds that an increase in relative demand for more educated workers led to a halt in the decline in returns to education between 1983 and 1992. Funkhouser (1998) decomposes the increase in relative demand for educated labor into changes due to between-industry shifts and more general technological change common to all industries. He finds that technological change explains more of the increase in relative demand for educated labor than between-industry shifts do. Robbins and Gindling (1998) find that the increases in returns to education were not correlated with either exports or trade deficits, but that they were correlated with increased investment, which i s complementary to skilled-labor. They argue that the increase in demand for skilled workers, and the particular role played by increased investment, is evidence of skills-enhancing trade, "whereby trade liberalization induces an acceleration of physical capital imports, which through capital-skill complementarity raises relative demand" (Robbins and Gindling, 1998, p. 152). 4.41 Gindling (2006) updates the earlier analyses, using data that runs from 1980 to 1997 to examine the determinants of changes in returns to education. He does not attempt to measurethe effect of demand shifts directly, but rather examines several possible underlying causes of shifts inlabor demand (including several institutional factors). Ingeneral, the literature that attempts to explain shifts in relative demand for labor in the United States and Latin America has focused on 80 two phenomena: (i)trade-related phenomenon, and (ii)technological change-related phenomenon. In the U.S., skill-biased technological change (especially related to personal computers and industrial robotics) has been identifiedas an important cause of increasing demand for more-educated workers. For Costa Rica, skill-biased technological change i s likely to be introduced through the importation of newer capital that embodies new technologies. Gindling (2006) considers three relevant measures: (i)gross investment in machinery and equipment, (ii) imported capital, and (iii) direct investment. He also considers two of the most important foreign institutional factors affecting wages inCosta Rica: (i) legal minimumwages, and (ii) size and the structure of wages within the public sect0r.9~ Gindling also controls for exports and the real GDP. 4.42 A representative sample of Gindling's results, usingthree different specifications of the investment variable, i s presented in Table 4.2Y8 Due to the small sample size (only 16 years of data are available for the entire set of relevant variables), the findings should be interpreted with caution. Nonetheless, the results are suggestive. The most consistently significant findings are that labor supply i s negatively associated with returns to education and that the level of investment is positively associated with education returns. The ratio of the maximum to minimumlegal minimumwage (a measureof wage dispersion inducedby Costa Rica's minimum wage policy) i s positive and significant (at the 10 percent level) only in specification 1. Other variables, including foreign direct investment, exports, and the proportion of workers in the public sector, are not ~ignificant.9~These results are consistent with those presented in Funkhouser (1998) and Robbins and Gindling(1998). 4.43 Together, the regression results suggest that two factors contributed to the increase in returns to education in Costa Rica. First, relative demand for more educated workers rose due to increased investment in newer capital that most likely embodies higher technologies, inducing skill-biased technological change. The results suggest that it did not matter whether this investment was by domestically-owned firms or foreign-owned firms. Second, slower growth in the relative supply of more-educated workers beginning in the mid-1980s also contributed to the increasein returns to education. 97 The role of minimumwages inCosta Rica on wages, employment, hours worked, and monthly earnings is discussed in more detail later in the chapter. '*Thisresulti s robust to different specifications of the export variable and to usingreturns to education estimated from paid employees only. 99Alternative specifications of the export variables (e.g., non-traditional exports, balance of trade, current account balance) are also not significant. This suggests that trade-related phenomena are not important determinants of changes in returns to education in Costa Rica, although these results need to be interpreted with caution given the small sample size. 81 Table 4.2: The D erminants of Changesin Returns to Education Independent Variables (all in natural logarithms) investment foreign direct investment exports GDP index of supply maximum/minimumwage proportionin public sector constant number of observations R- squared 0.848 0.892 0.861 Standard errors in italics. * = significant at10% **= significant at 5% Note: Usingdatafrom the 1980to1997 Source: Gindling (2006) 4.44 Changes in investment spending are consistent with this interpretation. In the ten years prior to 1983, when returns to education began to increase, investment as a proportion of GDP was declining. For example, investment in machinery and equipment fell from 12 percent of GDP in 1972to 9 percent of GDP in 1982, and spending on imported capital equipment fell from 9 percent of GDP in 1972 to 5 percent of GDP in 1982. 'Incontrast, in the decade after 1983, investment in machinery and equipment as a proportion of GDP increased every year. Investment in machinery and equipment rose to 15 percent of GDP in 1993, and spending on importedcapital equipment rose to 10percent of GDP in 1993.'O0 4.45 Gindling(2006) simulates the relative magnitudes of the effects of changes in supply and demand on returns to education. Consistent with the earlier evidence, the simulations suggests that the decline in returns to education prior to 1983 can be entirely explained by the increase in the supply of more-educated workers. However, despite continuing increases in the relative supply of more educated workers after 1983, returns to education stopped falling, and even 'O0 Data on investmentare from the Banco Central de CostaRica. 82 increased slightly. This suggests that the halt inthe decline in returns to education after 1983 was due, at least in part, to increases in the relative demand for more-educated workers. Depending on the choice of the wage elasticity of labor supply, the simulations suggest that changes in the relative supply of educated workers accounted for less than half (between 20 percent and 50 percent) of the total increase in returns to education between 1987 and 2002. Thus, it i s likely that changes in the relative demand for educated labor were a more important cause of increase in returns to education during that periodthan were changes inrelative supply of labor. 4.46 As will be discussedfurther below, changes inthe structure of legal minimumwages also contributedto the increase in returns to education after 1992. Multiple legal minimumwages are set by the Costa Rican government, depending on the occupation and skill-level of the worker. Prior to 1992, no legal minimum wages were set (explicitly) for four-year university and technical secondary school graduates (although minimum wages were set for many workers according to their "profession," and a minimumwage was set for workers with a licenciado). In 1992, the government introduced minimum wages for university and technical secondary school graduates. Effectively, this increased the average legal minimum wages for these workers, leading to an increase in the minimumwage that applied to more-educated workers relative to less-educated workers. Gindling and Terrell (2004) present evidence that this change in the structure of legal minimum wages causedthe wage gap between less and more-educated workers (i.e., returns to education) to increase in the 1990s, thus contributingto greater earnings inequality in the Costa Ricanlabor market. Increasesin Inequality of Hours Worked 4.47 An increase in the variance of hours worked also contributed to greater earnings inequality between 1987 and 2002. As can be seen in first column of Table 4.3, in 1980, the variance of hours worked was greater among women than among men, and was greater among those working in private, small firms than in private, large firms or the public sector. From 1987 to 2004 the proportion of women in the work force increased from 29 percent to 37 percent (Table 4.1), while the proportion of workers in the small-firm sector increased from 47 percent to 50 percent. This suggests that at least part of the reason for the 1987-2002 increase in the variance of hours worked was the increase in the proportion of women in the work force, the increase in the proportion of workers in the private small firm sector, or both."' At the same time, the variance of hours worked also increased for both men and women in the private sector between 1987 and 2002, in large as well as small firms. T h i s suggests that the 1987-2002 increase in the variance in hours worked was a broad-based phenomenon that cut across gender and sectors. Over the period, dispersion in hours worked only declined(andjust slightly) within the public sector (Table 4.3). 4.48 For men, within-sector increasesin the variance in hours worked occurred because of an increase in the number of men working more than full-time (Le., overtime) in the private large firm and small firm sectors - from 32 percent (in large firms) and 36 percent (in small firms) in 1987 to 48 percent (in large firms) and 41 percent (in small firms) in 2002. At the same time, the proportion of men working full-time and part-time in both sectors fell.'02 These results are 101 Trejos (2000) presents evidence that these two phenomena were related becausemany of the new female entrants to the work force found work in the private small firm sector. 102 The standard work week in the private sector i s 48 hours, although many who work 40 hours a week consider this full-time also. Anyone working between 40 and 48 hours, inclusive, i s considered as working full-time. 83 consistent with evidence that an increase in the proportion of over-time workers was due to an increaseinthe proportion of men in non-poor families working o~er-time.''~ Table 4.3: Changesinthe Variance of Hours Worked, by Gender and Sector (expresse Logof Hours Worked) Level in Averaae Annual Chanqes Between 1980 1980-1985 I 1987-2002 2002-2004 All Workers 0.14 0.003 0.006 -0.018 Male 0.11 0.003 0.003 -0.009 Female 0.22 0.000 0.008 -0.029 - Male private small 0.15 0.006 0.006 -0.012 private large 0.07 0.002 0.002 0.000 public 0.06 -0.001 -0.001 -0.002 Female private small 0.45 -0.007 0.006 -0.027 private large 0.05 0.005 0.003 0.000 public 0.08 -0.001 0.001 -0.006 Source: Gindling and TI )s (2005) Ginc ng (2006) 4.49 For women, within-sector increases in the variance of hours worked occurred largely in the private small firm sector (the only sector where there was a substantial increase in the variance of hours worked). Unlike for men, the increase occurred because of an increase in the proportion of women working less than full-time (i.e., part-time). The proportion of women in private, small firms who work part-time increasedfrom 41 percent in 1987 to 53 percent in 2002, while the proportion working full-time and overtime fell. The increase in part-time work among women was especially pronounced among women working very few hours - for example, the proportion of women inthe small firm sector working less than 20 hours a week increased from 5 percent to 30 percent while the proportion of women working less than 10hours a week increased from 2 percent to 14 percent. These results are consistent with evidence (presentedabove) that the increase in the proportion of part-time workers was due to an increase in the proportion of women from poor families working part-time, predominantly inthe self-employed sector. 4.50 From 2002 to 2004 the variance of hours worked fell both among men and women. The reduction in the variance of hours worked between 2002 and 2004 occurred in part because of a shift of workers toward private, large firms (to 54 percent, up from 50 percent of all workers). The private, large-firm sector employs a larger proportion of full-time workers than does the small private form sector. The variance of hours worked fell within the private, small-firm sector and inthe public sector. 103See Gindling (2006); see also footnote 34. 84 4.51 In sum, the evidence suggests that one the most important causes of the increase in inequality in hours worked in Costa Rica between 1987 and 2002 was an increase in the dispersion of hours worked in the private sector; this was caused by an increase in the proportion of women working part-time in self-employment and small firms (i.e., in the informal sector), as well as by an increase in the proportion of men working more than full-time in large private sector firms (Le., inthe formal sector). LegalMinimum Wages, WageInequality and Employment in Costa Rica 4.52 Legal minimum wage laws are the most important labor market legislation affecting wages in Costa Rica. Duringthe late 1980s and the 1990s, important changes in the structure of the complicated legal minimumwage system in Costa Rica contributed to the increase in returns to education we have identified as a cause of increased earnings inequality in the 1990s. This section summarizes the key findings of three recent studies on the impacts of Costa Rica's minimumwages on wages, employment and wage inequality inthe country Costa Rica (Gindling and Terrell2004a, 2004b and 2005). Evolution of the Structure of Legal MinimumWages 4.53 Legal minimum wages in Costa Rica are set twice a year by negotiation within the tripartite National Salaries Council, which i s composed of representatives of workers, employers and the government. Only private sector employees are covered by this legislation; public sector employees (including those in state-owned enterprises) and the self-employed are not subject to minimumwages. Although public sector workers have their wages set by separate government decrees, interviews with officials at the Ministry of Labor indicate that changes in the legal minimumwages are oftenusedas a guide inthe settingof public sector wages. 4.54 In1987,individuals who worked asemployeesintheprivatesector wereassignedto one of 520 minimum wage categories. The vast majority were assigned to one of 506 minimum wages, defined by detailed industry and occupational classifications (For example, in the manufacturing sector there were 44 occupational categories). Thirteen of the remaining 14 minimum wage categories were defined by "profession" without an industry dimension (these included, for instance, librarians, nurses, accountants, laboratory technicians and drafters in architecture and engineering). Employees with five-year university degrees (licenciado) were subject to a separate minimumwage, which was typically the highest one. Employees with four- year university educations or technical high school degrees who were working as professionals in occupations other than the 13 specifically mentioned in the law were classified according to the 506 minimumwage categories defined by occupation and industry. 4.55 Beginning in 1988, following recommendations of an International Monetary Fund report, the Ministry of Labor began gradually reducing the 506 minimum wage categories for non-professionals by eliminating the variation in wages between industries. It did this by identifying broadly-defined occupational categories that were to be harmonized across industries, and then gradually increasing the lower minimumwages by a greater amount than the higher ones within each occupational category. By 1990, the manufacturing, mining, electricity and construction industries were consolidated into one category, with a total of approximately 65 minimum wages for workers without higher education. By 1995, there were only five industrial categories and less than 54 minimumwages. BeginningJune 1997, the industrial dimension was completely eliminated, so that there were only ten minimum wages for non-professionals: four based on skill categories (unskilled, semi-skilled, skilled, and specialized workers) and six for special categories(for instance, live-in domestics, stevedores, and day workers). 85 4.56 While the number of minimum wage categories for less-educatedemployees was falling, the categories for people with higher education were being consolidated and expanded. In 1993, new minimum wages were set for individuals with two to three years of university education (diplomados) and for graduates of five-year technical high schools (te'cnicos). In 1997, another new minimumwage category was added for workers with four-year university degrees. The 13 minimum wage categories for specific professions were largely eliminated, though. These changes resulted in a total of six minimum wage categories for workers with technical high school education or higher. The addition of the new minimum wage categories for diplomados, te'cnicos and university graduates increased the level of the minimum wage for these workers since prior to 1993 many of them would have had minimumwages that applied to less educated workers. Changes in the Distribution of MinimumWages 4.57 Gindling and Terrell (2004b) show that adding new minimum wages for workers with more education increased the minimum wage gap between more and less educated workers, and that this, in turn, significantly increasedthe actual wage gap between university-educated workers and others. Thus, this change in the structure of minimumwages contributed to the increase in returns to education which, as noted above, contributed to the increase in earnings inequality in the 1990s. 4.58 To examine the distributional impacts of legal minimum wages in Costa Rica, Gindling and Terrell (2004b) assigned each worker in the 1988-1999 EHPM the minimum wage that applied to his or her education, skill, industry and profession. This resulted in a distribution of legal minimumwages among workers for each year. Figure 4.13 presentsthe distribution of real minimumwages (in 1999 colones) among private sector workers who reported positive earnings in 1988 (at the beginning of the simplification process) and in 1998 (at the end). Spikes in the distribution of minimumwages represent legal minimumwages that applied to larger proportions of workers. For example, starting from the left (the lowest minimum wage) of the upper panel (the 1988 graph), the first spike is at the minimumwage for domestic servants, who represented approximately 7 percent of all workers and who received a legal minimumwage of 123 colones (in 1999 prices) or $0.43 (in 1999 U.S. dollars) per hour. There are no minimumwages over a large range of possible wages between the minimum wage for domestic servants and the next minimumwage, which is for unskilled workers ('peones and other production workers) in most industries.This second spike represents over 20 percent of all workers. Next there is a cluster of many minimum wages that surround two smaller spikes at the minimumwages for operators of machinery and specialized workers (supervisors) in most industries. Finally, at the very right of the distribution of minimumwages (after numerous very small spikes) i s a spike at the minimum wage of 578 colones or $2.00 per hour set for Zicenciados (five-year university graduates) who representapproximately 2 percent of all workers. 86 Figure4.13: Distributionof LegalMinimumWagesAmong Workers, 1988and 1998 _. - . -. .. . .. .-_.. ... ... . ... . ..- . . I 4 I I I Log of Hourly MinimumWage 8I Fraction of Workers to Which Each Minimum Wage Applies, 1988 .5 ._ -m c 0 U 0 Fractionof Workers to Which Each Minimum Wage Applies, 1998 Log of Hourly MinimumWage Source: Figure 1in GindlingandTerrell(2004b) 4.59 The lower panel of Figure 4.13 presents the distribution of (the log of) real minimum wages among workers who reported positive earnings for 1998. A comparison of the graphs the two years illustrates the changes in the structure of legal minimum wages. As in 1988, the spike at the far left of the 1998 distribution of wages is at the minimum wage for domestic servants (which again represents approximately 7 percent of workers) and the second spike occurs at the minimum wage for unskilled workers. It can be seen that the simplification and consolidation processcompressedthe distribution of minimumwages around the unskilled wage: while in 1988 the spike at the unskilled minimum wage represented 20 percent of workers, in 1998 the minimum wage for unskilled workers applied to more than 45 percent of workers. Moreover, there are three new spikes in the next range of minimum wages, all of which in 1988 were not significant: at the minimum wages for semi-skilled workers (12 percent of workers), skilled 87 workers (14 percent) and specialized workers (6 percent). The new minimumwage categories for workers with higher education resulted in several new spikes at higher wage levels, including a spike at the minimumwage for four-year university graduates (4 percent of workers). 4.60 Figure 4.14 overlays the distribution of the actual wages paid to workers on top of the distribution of legal minimumwages usingdata from 1999. It shows that most minimumwages affect workers in the middle of the distribution of wages. The lowest minimum wage (for domestic servants) falls in the 3rd decile of the wage distribution, while most minimum wages (for unskilled, semi-skilled and skilled workers) fall in the 4th and 5th deciles of the wage distribution. Minimum wages are also set for more highly paid workers; minimum wages for those with four year university degrees or for licenciados fall inthe 10thdecile of the distribution of wages. Figure 4.14: Distribution of Legal MinimumWages and Actual Wages, 1999 - proportionat each wage proportionat each rrinimmwage -.249041 a, Ps .-E 'E K 0 0 4 8 Density Functionsof Logof Wages, 1999 Source: Gindling(2006) Coverage of the MinimumWane 4.61 From Figure 4.14 it is clear that a large number of workers earn below the minimum wage in Costa Rica. In part, this i s because not all workers are covered by legal minimumwage legislation (for example, self-employed workers). But it is also true that a large number of workers in sectors covered by minimumwage legislation earn below the minimumwage. Table 4.4 shows that even among workers covered by minimumwage legislation, more than 30 percent earn less than 90 percent of the legal minimum wage. Among self-employed workers, more than 33 percent earn less than 90 percent of the legal minimumwage. Even if the analysis i s restricted to full-time workers, more than 30 percent of workers who should be covered by minimumwages earn less than 90 percent of the minimum wage applicable to their education, skill and profession. Further, Gindling and Terrell (1995) show that, on average, one fourth of full-time private sector 88 employees earn less than the lowest minimum wage applicable in each year that they studied (1976-1991). Workers earning less than the minimumwage are disproportionately female, very young (less than 19 years old), very old (more than 60 years old), have less education, live in rural areas and work in agriculture or personal services. Within i Below i Above I Sector 10%ofmw i 90%ofmw 110%ofmw I I ! I Covered (paid employees) 20.04 : I 30.64 :! I 49.32 Uncovered(self-employed) 10.32 I 33.37 I 56.3 Public I 9.7 ! I 10.29 !I 80.02 Source: Gindling(2006) The Impact of Legal Minimum Wages on Wages, Employment, Hours Worked, and Monthly Earnings 4.62 Gindlingand Terrell(2004b) estimate the impact of changes in legal minimumwages on actual hourly wages, hours worked, employment and monthly salaries in the covered and uncovered sector. They do so using individual-level pooled cross-sectionhime-series data. The explanatory variables include the (log of the) real minimumwage (in 1999 colones) that applies to a given worker in each year, human capital variables such as years of education, experience (cubed), gender of the worker, dummy variables for each occupation and skill category in the minimum wage decrees, as well as controls for endogenous changes in yearly average minimum wages and endogenous correlation of wages and minimum wages across occupation and skill categories. Equations are estimated separately for workers covered by minimumwage legislation (paid employees) and for workers not covered by minimum wage legislation (self-employed workers). 4.63 Their findings are summarized in Table 4.5. The table reports the percentage change in outcomes that would result from a 10 percent increase in real minimum wages. The results indicate that, on average, minimumwages have a significant positive impact on hourly wages in the covered sector and an insignificant impact on hourly wages in the uncovered sector. Specifically, it is found that, on average, a 10 percent increase in minimumwages leads to a 1 percent increasein averagehourly wages inthe covered sector. IO4For further details, see Gindlingand Terrell (2004b); the empiricalspecificationis also laidout in Gindling(2006). 89 Covered Sector Uncovered Sector Hourly Wages 1.03% ns Employment -1.09% na Hours Worked -0.62% ns Monthly Earnings ns ns 4.64 The estimated effect on employment i s negative and significant. The estimated coefficient indicates that a 10 percent increase in the real minimumwage reduces the probability of beingemployed in the covered sector by 0.0068. Evaluatingthis at the average probability of employment (0.625) in the data, Gindling and Terrell calculate that a 10 percent increase in the minimumwage reduces employment in the covered sector by 1.09 percent (Table 4.5). This can be interpreted as an elasticity of employment with respect to the minimumwage. 4.65 The estimated elasticity of average hours worked with respect to minimum wages, also reported in Table 4.5, indicates that a 10 percent increase in minimum wages would lower the average number of hours worked by 0.62 percent in the covered sector and would not have a significanteffect on hours worked in the uncovered sector. Hence, these results indicate that in Costa Rica employers respond to higher minimumwages by cutting back on the number of hours worked, as well as by reducingthe number of workers. 4.66 Finally, since Gindling and Terrell find minimum wages raise hourly wages of covered sector workers but lower the number of hours worked, they examine whether the overall effect of wages and hours translates into a positive or negative change in the monthly earnings that the remaining individuals in the covered sector receive. The results indicate that the increase in hourly wages is offset by the number of hours worked so that the overall impact on monthly earnings inthe covered sector is not significantly different from zero. 4.67 Insum, the evidence indicates that legal minimumwages have significant effects onthe covered sector labor market but do not have significant effects on the uncovered sector (self- employed). While the effect of higher minimumwages on actual hourly wages is positive, they have a negative effect on employment in the covered sector. They also have a negative effect on hours worked by those who are employed in the covered sector. Indeed, the positive effect of minimum wages on hourly wages is actually offset by the negative effect on hours worked, so that there is no net effect on monthly earnings for those who remain employed in the covered sector. The Impactof MinimumWages across the Distribution 4.68 The analysis above indicates that legal minimum wages are paid to workers in the 3rd through 10th deciles of the wage distribution. But this does not necessarily mean that legal minimum wages do not affect poor workers. If could be that, without legal minimum wages, workers currently earning minimum wages would earn less, landing them at the bottom of the wage distribution. It is useful, therefore, to estimate the impact of minimumwages on the actual wages and employment of workers who would likely be in the bottom of the earnings distribution if legal minimumwages didnot exist. To do this, GindlingandTerrell (2004a) adopt atechnique developed by Card (1996) in which they construct a distribution of predictedwages based on the 90 education, experience and gender of the worker. Workers in the lower deciles of this (predicted) distribution are those with lower skills who can be expected to earn lower wages, while workers in upper deciles of this (predicted) distribution are those with more skills who can be expectedto earn higher wages. 4.69 Gindling and Terrell (2004a) find that legal minimum wages have a significant positive impact on actual wages for workers in the 2nd through 5th deciles of the distribution of skills (most, but not all of the lower half of the wage distribution). They also find that minimumwages have a significant effect on the wages of workers in the 10th decile (most likely those workers with a university education). It i s noteworthy, however, that legal minimumwages do not have a significant impact on the wages of workers who are likely to be poorest - those in the lowest-skill decile. These findings are roughly consistent with the distribution of minimumwages and actual wages shown inFigure 4.14. 4.70 The employment effects of higher legal minimumwages, measuredboth in hours worked and number of workers, are negative and are significant inthe 2nd through 5th deciles of the skill distribution. Specifically, the effects of minimumwages on the number of workers in the covered sector are negative and significant for workers in the 3rd and 4th deciles, while the effects on hours worked are negative and significant inthe 2nd and 5th deciles. 4.71 The effect of minimumwages on monthly earnings is significant only for workers in the 3rd and 4th deciles. Thus, it i s only in these two deciles that those workers who remain in the covered sector experience an increase in earnings. Inthe 2nd, 5th and 10th deciles, higher hourly wages for those who remain in the covered sector are offset by fewer hours worked. Inall other deciles, minimumwages have an insignificant effect on monthly earnings. 4.72 Thus, legal minimum wages in Costa Rica have a significant effect on workers in the bottom half of the distribution of skills (although not at the very bottom - the first decile) and at the very top of the distribution (the 10th decile). At the same time, negative employment effects are significant for those in the 2nd through 5th skill deciles. Except for those in the 3rd and 4th skill deciles, higher minimum wages do not result in higher monthly earnings - even for those who remain employed in the formal sector. This i s because, in general, increases in legal minimum wages lead to higher hourly wages but also result in to fewer hours worked for those same workers. Changing Demographic Structures, the Labor Market and Poverty: Female Headed Households,Nicaraguan Immigrants 4.73 This section examines the link between changing demographic structures and labor market outcomes and poverty in Costa Rica. The section focuses on two issues in particular that merit special attention: the substantial increase in female headed households in Costa Rica in the 1990s, andthe impact of Nicaraguan immigrants on the labor market. Female HeadedHouseholds and the Changing Structure of Poor Families 4.74 The most notable change in the structure of Costa Rican households duringthe 1990sand early 2000s was an increase in the proportion of female-headed households. Between 1987 and 2004, the share of all households that were headed by women grew from 17.0 percent to 26.4 percent (Table 4.6). The proportion of poor households headed by women increased particularly rapidly - from 19.6 percent of poor households in 1987 to 33.6 percent in 2004. The share of female-headed households among the non-poor also increased over this period, although more 91 slowly, from 15.9 to 24.4 percent of non-poor households. These developments contributed to the stagnating poverty rates observed in Costa Rica since 1994. Table 4.6: The Changing Share of Female HeadedHouseholds in Costa Rica, Among Poor , MaleHeaded , Source: Adapted from Gindling (2006) 4.75 This rapid rise in poor female headed households appears to have been mediated in important ways through developments in the labor market - through increases in unemployment, part-time (as opposed to full-time) work, and increased self-employment, particularly among poor female workers. For example: Unemployment rates are higher for members of poor households than for members of non-poor households, and higher for poor female household heads than for poor male household heads. For instance, in 2004, unemployment rates were 15.9 percent for poor female household heads compared with 5.4 percent for poor male households (Table 4.7). Further, unemployment rates among members of poor households increased significantly from 1987 to 2004 (more than doubling from 1992 to 2004), with a larger increase for female household heads than for male household heads. Increasesinthe proportion of women working part-time in poor households i s a result of increases in part-time work among female household heads, employed (predominantly female) spouses and other employed household members. The proportion of poor female household heads working part-time increased from 46 percent in 1987 to 63 percent in 2004 (Table 4.7). The proportion of non-poor female household heads working part-time also increased, but by a smaller amount (from 31 to 42 percent). In contrast, over the same period, the proportion of part-time workers fell for all other members of non-poor households and for poor male household heads. 0 This evidence is consistent with the notion that the increasednumber of part-timefemale workers in self-employment and small private firms i s composed of the primary income earners (household heads) for their households. 92 4.76 This evidence is consistent with a situation in which many mothers with children at home have difficulty obtaining well-paid, full-time employment in the formal sector, and are forced to find low-paid, part-time employment in the informal sector. To the extent that the new female household heads are mothers with children at home, the increase in the number of female-headed households i s likely a contributing factor to the decline in the average earnings of households vulnerable to poverty and to an increase in poverty rates. 4.77 Who are these "new" female household heads of poor families? As we can see from Table 4.8, female-headed households are overwhelmingly single-parent household^.'^^ The typical female-headed household i s a single-parent household with children (the typical male- headedhousehold i s a two-parent household with children). Further, single-parent, female-headed households with children are more likely to be poor than either male-headed households or two- parent, female-headed households; 33.7 percent of single-parent, female-headed households with children were poor in 2004, compared to 21.2 percent of single-parent, male-headed households with children and 22.4 percent of two-parent, female-headed households with children (Gindling 2006). Some 59 percent of the increased share of female-headed households among the poor between 1987 and 2004 was due to an increase in female, single-parent households with children. A smaller proportion of the increase(25 percent) was due to an increase inthe number of female, single-parent households without children, while the remainder (16 percent) was due to an increase intwo-parent householdsheadedby women. 4.78 In sum, the evidence suggests that the rise in unemployment and part-time work in the self-employed sector has accompanied the rise of poor, female, single-parent households with children, and that this can help explain stagnating poverty rates and higher earnings inequality over the last decade in Costa Rica. Less is known about the underlying sociological causes of the increase in female headed households with children, however. For example, it i s impossible to tell from the EHPM surveys whether these are women who have never been mamed, were marriedbut have been divorced or widowed, or who have lived in union libres but no longer have another adult living inthe household. Additional investigationinto this issueis clearly warranted. 4.79 The share of female, single-parent households without children also increased from 1987 to 2004 (although at a much slower rate than the increase in such households with children). These women are older (more than 60 percent are older than 65) and not likely to be labor market participants. This suggests that these are older women who do not have access to the pensions of spouses. Unfortunately, again, the household surveys do not allow us to identify whether these `05A single-parent householdis one where, according to the householdsurveys, neither ahusband (esposo) nor companion(compar'iero)is present. 94 are women who were never married, who divorced, or whose husbands have died. Again, further analysis into this phenomenon i s warranted. The Labor Market Impacts of Nicaraguan Migrants 4.80 As discussedin Chapter 2, census figures reveal that the number of Nicaraguans in Costa Rica increased from 45,918 to 226,374 -from 2 percent to 6 percent of the population -between 1984 and 2000. This migration was caused largely by economic factors, and labor force participation rates for Nicaraguan migrants are higher than for native born Costa Ricans. As a result, the share of Nicaraguans among workers is slightly higher than in the population as a whole - a little over 7 percent according to the 2000 census. Nicaraguan immigrant workers are less educated, work more hours, and are paid less than Costa Rica-born workers (according to data from the 2000-2004 Household Surveys for Multiple Purposes, Nicaraguan workers are paid, on average, 65-75 percent as muchas the average Costa Rica-born worker). Further, Nicaraguans are generally "concentrated in lower status, lower paying occupations. In San JosC, Nicaraguan men are concentrated in construction and women in domestic service. In other regions of the country Nicaraguans are concentrated in agricultural occupations" (Marquette, 2005, p.63). 4.81 The large influx of Nicaraguan migrants to Costa Rica began in the early 199Os, at the same time earnings inequality began increasing. In the 199Os, about 20,000 Nicaraguan migrants entered Costa Rica each year. Migration rates slowed from 2000-2005, at about the same time as earnings inequality began decreasing. Did the influx of Nicaraguan migrants in the 1990s contribute to the increaseinearnings inequality duringthis period?'06 4.82 On average, Nicaraguans earn wages lower than Costa Rican natives. By increasing the number of low-wage workers in the Costa Rican labor market, the presence of Nicaraguans may have, by itself, caused the increase in inequality in Costa Rica. If the presence of Nicaraguan migrants i s causing the increase in inequality, then measures of inequality should decrease when Nicaraguan migrants are excluded from the sample. Table 4.9 presents two measures of earnings inequality - the gini coefficient and the log variance of earnings - for each year between 2000 and 2004, both including and excluding Nicaraguan workers. Interestingly, both measures of inequality are lower with Nicaraguans than without, indicating that Nicaraguan workers do not contribute directly to higher inequality. In general, differences in inequality measures with and without Nicaraguans are small. The gini coefficient is the same up to 2 digits whether Nicaraguans are included in the sample or not. Dependingon the year chosen, the difference in the variance of the log of earnings is a bit larger, but always in the same direction, suggestingthat the presence of Nicaraguans cannot explain directly the increaseinearning inequality. `06This section is based on analysis of Costa Rica's national household survey (EHPM), using data from 2000-2004. The 2000-2004 EHPM surveys includeda variable that indicates where the person was born, which is used to identify Nicaraguan migrants. In 2000 and 2001, the EHPM survey includes another variable that allows one to identify Nicaraguan migrants via self-reported nationality. Table 4-2 presents the distributionof Costa Ricanworkers in 2000 and 2001 by nationality. While these data are not perfect, comparing the 2000 census with the 2000 EHPM survey findings suggest that these data provide a reasonablepicture of Nicaraguanmigrantsin the CostaRicanlabor market. 95 2000 2001 2002 2003 2004 All Workers Gini coefficient 0.434 0.465 0.465 0.456 0.438 Loa Variance 0.748 0.870 0.878 0.832 0.760 ExcludingThose Born in Nicaragua Gini coefficient 0.438 0.469 0.467 0.461 0.441 Loa Variance 0.773 0.898 0.898 0.859 0.779 ExcludingThose With Nicaraguan Natiorality Gini coefficient 0.438 0.469 Loa Variance 0.769 0.895 4.83 To examine the impact of Nicaraguans on earnings inequality further, Gindling (2006) estimates the Fields decompositions (discussed above) including a dummy variable that indicates whether the worker i s a Nicaraguan immigrant. The results indicate that there i s no impact of Nicaraguan workers on earnings inequality in Costa Rica once various factors - including education, gender, region, hours worked, sector of employment, size of firm and work experience -arecontrolledfor (Gindling2006). 4.84 While Nicaraguan immigrants do not directly contribute to higher earnings inequality, they do affect key factors that influence earnings inequality-specifically, the distribution of education levels among workers in Costa Rica and (through that) returns to education.lo7 As noted above, Nicaraguan migrants are, on average, less educated than Costa Rcan born workers. The data thus indicates that the influx of Nicaraguan migrants into Costa Rica in the 1990s and 2000s contributed to the observed increase in the inequality of education levels among Costa Rican workers; and in this way, the presence of Nicaraguans has contributed to the increase in earnings inequality. 4.85 The same is true with respectto returns to education. As discussedabove and inChapter 2, Nicaraguan workers are considerably less educated, on average, than Costa Rica-born workers (in 2004 Costa Ricans workers had, on average, 8.6 years of education compared to only 5.9 years of education for Nicaraguan migrants). Indeed, including Nicaraguan migrants in the sample of workers decreases average education levels by 0.2 years. By lowering the relative supply of more-educated workers, Nicaraguan immigrant workers contributed to the increase in returns to education in Costa Rica during the 1992-2002 period. Gindling (2006) simulates the increasein returns to education causedby the influx of Nicaraguan immigrants. He finds that the influx of Nicaraguan immigrants into Costa Rica could account for between 4 percent and 40 percent of the increasein returns to education duringthis period, dependingon the estimate of the wage elasticity of labor supply that he uses. While the point estimate of their impact i s not very precise, the evidence does suggest that Nicaraguan immigrants were partly responsible for increasing returns to education, by decreasingthe relative wages of less-educatedworkers - both Costa Ricans andNicaraguans (Box 4.1). lo7Ginding (2006) find that the presence of Nicaraguan workers has on effect on inequalityof hours worked in the Costa Ricanlabor market, the third key factor driving recent changes in earningsinequality inthe country. 96 Box 4.1: Why do NicaraguansEarn Less than Others in Costa Rica? Analysis of the EHPM data from 2000-2004 indicate that Nicaraguan workers earn 65-75 percent the monthly earnings of Costa Ricans. Why is that? The main explanation appears to be the relatively low levels of education possessedby Nicaraguans compared to other workers in Costa Rica - although returns to Nicaraguan and Costa Rican education do appear to differ too. Gindling (2006) uses several earnings equation-based techniques to try to explain earning differences between Nicaraguan and other workers in the Costa Rican labor market, including the estimation of an earnings equation including a dummy variable identifying the Nicaraguan born and the OaxacdBlinder decomposition, the work-house in the economic literature for analyzing inter-group wage differentials.lo8 Using the dummy variable technique, Gindling finds that there are no significant differences in the pay of Nicaragua- and Costa Rica-born workers once factors such as education, gender, zone, hours worked, sector of employment, firm size, and experience are controlled for. For the most part, the results of the OaxacdBlinder decomposition are consistent with this. Using this second technique, Gindling finds that earnings differentials between Nicaragua-born and Costa Rican-born workers are due almost entirely to differences in their education levels. Gindling also finds that returns to Nicaraguans' education (that is, their years of schooling) are lower than returns to Costa Ricans' education - the coefficient in the earnings equation on education in the earnings equation is .08 for Costa Ricans and .05 for Nicaraguans). Thus, controlling for education levels (and other factors), Nicaraguan workers do appear to be paid less than Costa Ricans. Whether or not this latter effect reflects labor market discrimination is not completely clear. The evidence i s not strong. The results of the OaxacdBlinder decomposition suggest that the total labor market discrimination effect i s "zero," indicating no discrimination. There are plausible explanations about why that might be. First, it i s possible that the quality of Nicaraguan education is lower than the quality of Costa Rican education, and that these differential returns to education reflect real productivity differences. Second, it is possible that the reservation wage (the wage at which people are willing to enter the labor market) is lower for Nicaraguans than for Costa Ricans - that is, NicaragGan workers are willing to accept lower wages than similarly-educated Costa Rican workers. Yet labor market discrimination cannot be completely ruled out, at least in principle. If having a Nicaraguan education as opposed to a Costa Rican education is used by Costa Rican employers as a screening device for hiring Nicaraguan or Costa Rican workers, regardless of their actual productivity, then this would represent a form of ("statistical") discrimination. 4.86 The impact of Nicaraguan immigrants on wage premiums in specific sectors. Some observers have argued that the influx of Nicaraguan migrant workers to Costa Rica has had a significant depressing effect on the wages of Costa Rican workers with whom they compete. If this is the case, one would expect to see that mean wages in the industry sectors where Nicaraguans are concentrated fell in the 1990s, duringthe biggest surge in Nicaraguan migration. To examine this possibility, Gindling (2006), estimated an earnings equations across years of the EHPMsurvey (1990-2004) that excluded dummy variables for Nicaraguan identity and included dummy variables for different industrial sectors, including for domestic service. Changes in the estimated coefficients on these dummy variables measure changes in the relative average wages in each sector, controlling for changes in work place and personal characteristics (such as education). Figure 4.15 presents the coefficients on these industry sector dummy variables for 1990 to 2004, with the coefficients for agriculture, construction and domestic service (those sectors in which Nicaraguans are concentrated) presented in bold lines. The omitted industry dummy variable in the earnings equations is for the commerce sector, so what is reported in Figure4.15 are averageearnings of each industry relativeto average earningsin commerce. lo8For methodological details see Gindling (2006). 97 Figure 4.15: Mean (Log) Earnings inEach Industrial Sector, Adjusted for Personal and Workplace Characteristics (relative to commerce) 0 0.2 I1 .- 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 ' -c 'Agriculture +Construction Domestic Servants Manufacturing +Utilities +Transport and Communicatioris +Finance and Real Estate -Other Services Source: Gindling(2006) 4.87 As can be seen in the Figure, in construction, agriculture, and domestic service, the sectors where Nicaraguan migrants are concentrated, adjusted earnings grew relatively quickly in the 1990s and 2000s. In fact, the increase in adjusted earnings of those in domestic service was faster than that in any other sector. Incontrast, in most sectors with relatively few Nicaraguan immigrant workers (finance, utilities, transportation and communications, for instance), adjusted mean earnings remained relatively constant over the period. 4.88 Inshort, thereisnoevidencethat, after controllingfor changes inother characteristics of workers such as education, the influx of Nicaraguans had an adverse impact on the earnings wages paid to workers in different sectors of the Costa Rican economy. Rather, the evidence i s consistent with a situation in which Nicaraguan immigrants have been attracted to those industry sectors where earnings and wages (especially for low skilled workers) have been increasing. Wages in those industry sectors may have been increasing for one of several reasons - for example, because of an increase in demand for low skilled workers (such as in construction during the construction boom) or because Costa Rica-born workers have left those industries to work in other booming sectors that pay better for relatively high-quality workers. As an example of this latter phenomenon, low-skilled Costa Rican born women, who in the 1980s would have been domestic servants, may have found better paying work in new export industries (such as apparel, electronics or tourism), leading both to an increase in wages paid to domestic servants and to an increase in demand for female Nicaraguan migrants inthe domestic servant sector. Conclusion 4.89 This chapter's examination of the Costa Rican labor market suggests that a number of forces have come together since the early 1990s to make it more difficult for the poor to escape poverty. Structural changes in the Costa Rican and global economies led to increaseddemand for skilled workers relative to unskilled workers, and increased returns to secondary and higher 98 education. At the same time, Costa Rica experienced a decline in the proportion of secondary school graduates - and an increase in secondary school drop-outs - in the labor force. Shifts in relative demand for skilled and unskilled labor have resulted in significant increases in unemployment and part-time work among the poor andextremely poor since 1994. Between 1994 and 2003, unemployment rates rose from 8 percent to 17 percent among people who live in poor households and from 12 percent to 27 percent among those in extremely poor households, while unemployment rates among those in non-poor households remained at (or below) 5 percent. Moreover, the evidence suggests that unemployment i s largely structural, reflecting a skills mismatch between poor, often low-educated workers andfirms demanding labor. 4.90 Together, these factors have contributed to a widening of earnings inequality in Costa Rica's labor market. Moreover, recent increases in the number of female-headed households in Costa Rica, along with limited social infrastructure that would help single mothers find higher paying, full-time employment (early childhood development programs and daycare facilities, for instance), have contributed to a disproportionate rise in part-time work among poor female workers and inpoverty among female-headedhouseholds. The evidence suggests that some labor market regulations, such as Costa Rica's minimum wage policy, may also limit remunerative employment opportunities among the poor. 4.91 The evidence points toward several areas where public intervention to strengthen the ability of the poor to generateincomes from the labor market would be fruitful, including: Improving access to, and the quality ox education, especially at the secondary level. Increasing enrollment and graduation rates at the secondary level in Costa Rica will increase the education level and earnings of the average Costa Rican worker, reduce inequality in the distribution of education, and increase wages for less-educatedworkers (by decreasingthe relative supply of less-educatedworkers). Providing poor workers with the skills and other resources they need to escape unemployment. Unemployment is an important .cause of poverty in Costa Rica. The evidence suggests that it i s largely structural rather than cyclical, reflecting a mismatch between skills of low-skilled workers and emerging demand in the labor market. Therefore, in addition to strengthening formal education, increasing access to adult training programs outside of the traditional public education system would be beneficial. Internationally, the evidence on returns to skills training programs i s mixed. Identifying effective, market-responsive training approacheswould yield highreturns. Expanding access to afordable child care during standard working hours. Expanding the possibilities for child care for poor families during standard working hours would make it easier for poor single mothers (and female spouses in poor households) to obtain full-time work. Public policies to expand access to child care might include: expanding government subsidies to poor families for child care, providing before- and after-school child care programs in schools, and encouraging privatefirms to provide subsidized day carefacilities at work. Providing poor female household heads with the skills and other resources necessary to find and keep well-paid employment. Poor single female household heads have very low skills compared to other Costa Rican workers. Over 90 percent have not completed a secondary education. This suggests that programs designed to increase the skills of single mothers could contribute to reducing poverty in Costa Rica. One such set of policies would make it easier for women (particularly younger single mothers) to 99 complete more formal education. Another set of policies would provide training for adult single mothers. Reducing legal barriers to women who would like to work non-standard work hours. Current Costa Rican law limits firms' ability to employ women at night. Reducing legal barriers to women who would like to work non-standard work hours - essentially by leveling the regulatory environment for female workers so that they have the same working opportunities as men - might make it easier for single mothers to obtain employment duringtimes when it is easier to find others to care for their children. 4.92 It also i s important to recognize the trade-offs inherent in the legal minimum wage system - namely, that higher hourly wages bring lower formal employment among the poor and near-poor. A better understanding of this would ensure that future efforts to raise the minimum wages for low-skilled workers do not generate adverse and unexpected consequences for the employment of poor workers. IO0 PART2: POVERTYAND SOCIAL SECTORPOLICY 5. INTRODUCTION 5.1 Costa Rica has had a long-standing commitment to social development and universal access to core social services such as education and health. This commitment has been manifest (among other ways) in its levels of public spending on social sector programs. Indeed, at approximately 15.5 percent of GDP, Costa Rica's social spending i s above the Latin America average of 12.5 percent (Figure 5.1). In fact, social assistance is the only category of social spending on which Costa Rica spends below the regional average. Figure 5.1: Public SocialSector Spending (Education, Health, and SocialProtection) in Latin America (% of GDP; most recent estimate, 2000-2004) 'I a a" Source: World Bank staff estimates basedon World Development Indicators and selectedWorld Bank country reports. 5.2 This commitment has been an important factor underpinningCosta Rica's strong social performance. As noted in Chapter 1, the country's infant and child mortality rates are significantly lower than those in comparable countries, while its average life expectancy is substantially higher. Inaddition, social security coverage in Costa Rica i s among the highest in the Latin America region, rivaled only by Chile (Figure 5.2). 5.3 Yet despite levels of public spending in the social sectors, there i s considerable room to increase the efficiency and impact of public spending in some areas - particularly education and social protection. As noted in Chapter 1, in spite of progress, the country lags behind both Latin America and other upper-middle income countries in secondary school enrollment levels, and its poor people lag considerably behind the non-poor in educational access and attainment. Moreover, analysis of public spending on education indicates that Costa Rica achieves significantly worse primary and secondary school outcomes than would be expected given its level of spending. 101 Figure 5.2: Social Security Coverage Rates for the Economically Active Populationin 15 Latin American Countries, 1990s and 2000s 1 70 - 60 40 - Io: BO PY PE NI GU CO EC MX SA VE AR BR W CL CR Source: Rofmanand Lucchetti(2006) 5.4 Similarly, social protection could become a stronger force for poverty reduction that it is today. While Costa Rica's social security coverage rates are high by Latin American standards, for instance, coverage gaps continue to be larger among the poor than the non-poor; as many as one-third of the extremely poor still lack access to health insurance, for instance. Moreover, despite the country's accomplishments, a significant proportion of the poor still fall outside the reach of the safety net. These coverage gaps leave many of the poorest, most vulnerable Costa Ricans exposedto serious risks. 5.5 This following chapters focus on how Costa Rica can leverage policies and investments in education (Chapter 6), health (Chapter 7), and social protection (Chapter 8) to reducepoverty. 102 6. EDUCATION 6.1 This chapter is organized into six main sections. The first section discussesthe evolution of Costa Rica's educational indicators, and compares them with other countries in the region. Then the chapter describes the educational system, including the role of the public and private sectors, and the most recent government education policies. The third section describes in detail educational outcomes through time, with special attention to differences among income groups. Next, the chapter explores incidence, efficiency and quality indicators within the government education expenditure structure. The fifth part looks at the Millennium Development Goals (MDG) and Costa Rica's progress toward achieving them. The sixth and final part contains the conclusions and recommendations. 6.2 Education i s a key factor affecting inequality and poverty in Costa Rica. Education increases personal income, reduces inequality and decreases the probability of being poor. The EHPM shows that variation in educational attainment is the most important source of inequality, accounting up to a third of income disparities, and the educational attainment of heads of h~usehold''~ i s one of the major determinants of poverty (see chapter 2 for more details). 6.3 Costa Rica achieved important improvements in most education indicators from the late 1980s to 2004. Increased government expenditures between 1989 and 2004 improved the educational attainment of the population, and reduced inequalities between income groups. Education expenditure as a percentageof GDP increased65 percent during these years, secondary school net enrollment rates went up 48 percent, and the average years of education received by Costa Ricans increasedby 25 percent (Table 6.1). Improvements inprimary education were more modest, but the absolute levels are significantly higher. Nonetheless, the country faces a number of challenges. Despite significant levels of public spending on education outcomes - particularly at the secondary level - remain low compared to other Latin American countries and other upper middle income countries."' In addition, important education gaps persist between the poor and non-poor in access to education, persistence in schooling and quality of education. These gaps are particularly apparent beginning at the secondary level. 6.4 All-in-all, despite progress and advance in education since the early 1990s, Costa Rica has not been able to keep pace with changes inthe economy and inthe labor market. This has had an adverse impact on the ability of the poor to participate fully in and benefit from emerging economic opportunities, and that, in turn, has potentially important implications for the country's economic competitiveness. 109Or his or her companion, whoever has the higher labor income. While public spendingon education is both historicallyand currently highin Costa Rica, it should be notedthat the economic crisis of the early 1980s had a large impacton real education spending, including per student spendinglevels. Public spendingon educationdid not fully returnto per-crisislevels until the very end of the 1990s, and current educationoutcomes (includingcurrent educational attainmentof the workforce)appear to reflect,at least inpart, the post-crisis spendingtrajectory. 103 Table 6.1: Costa Rica's Basic Education Indicators, 1989-2004 Year I Change 1989-2004 Indicator 1989 2004 Value % Education expenditure as percentage of GDPa 3.1% 5.1% 2.0 65% a Ministry of education expenditures for 1898and 2003 Per capita over 5 to 24 year-olds. For the 1989-2002 period in 2003 constant dollars at $1 = 399 colones. `Students 13 to 17 year-olds in school but laggingone or more years according to their age. Source: World Bank staff calculations using the 1989 and 2004 EHPM 6.5 Costa Rica does perform well relative to other countries in Latin America and countries with similar income levels in terms of net primary enrollment rates"' (Figure 6.1). With higher primary enrollment rates than that of the Latin American average and other Central American countries"', the country i s better prepared than its neighbors to satisfy demands for workers with basic ed~cation."~ Figure 6.1: Net Primary Enrollment Rates in Costa Rica and SelectedCountries 2000-2004 100% 80% 60% 40% 20% 0% Note: 2002-2003 datafor Central American Countries, 2000 for all other countries. Source: DiGropello 2005 from Official Statistics of the Ministries of Education; The World BanklWDI (2003) For Honduras, 2004 LSMS 111 For international comparisons, official statistics from the ministriesof education were used. As discussedlater, there are significant differences between the official numbersand the EHPM estimates, although both show similar trends over time. `I' Panamaexcluded. 113 Other data sources, including the World Bank WDI, had differentnumbers not only for Costa Rica but for other countries as well. Depending on the data source, Costa Rica's net primary enrollment rates are just above or just below the Latin American average. Inall cases, Costa Rican values are very similar to the average for Latin American average and the upper-middle income countries, and above those for most Central American countries. 104 6.6 Costa Rica's favorable rating in primary schooling does not fully translate into secondary education, though. The country's secondary net enrollment rates are below other Latin American countries and the Latin American average, though still above all other Central American countries (Figure 6.2). Costa Rica's ability to attract employers who create higher-paying jobs will be limited by not having enough workers who have at least secondary school educations. Figure 6.2: Net Secondary EnrollmentRates in Costa Rica and Selected Countries 2000-2004 80% 70% 60% 50% 40% 30% 20% I Note: 2002-2003 data for Central American Countries, 2000 for all other countries. Source: Di Gropello 2005 from Official Statistics of the Ministriesof Education; The World Bank/WDI (2003) For Honduras, 2004 LSMS. 6.7 To strengthen the contribution of education to poverty reduction, the Government of Costa Rica faces several critical challenges. It has to expand secondary school coverage, improve quality, and increase educational opportunities. The new administration is committed to increasing the allocation of resources to education. This certainly will help. But the country also can increase the efficiency with which resources (existing as well as additional) are utilized. Currently, a very high percentage of recurrent expenditures are used to pay teachers' salaries, leaving relatively little funds to expand educational infrastructure in under-served areas (particularly in poorer, more remote rural areas), strengthen access to education among the poor, or improve educational quality. A fourth key challenge, therefore, is to increase the efficiency of public spending on education through stronger results-basedmanagement. 6.8 It is important to add that making significant progress against poverty will require not only the general strengthening of the education sector, but also specific poverty-focused actions, such as more and better-targeted support for education provided to children in poor households. In particular, the government should design targeted programs to reduce repetition and drop-out rates of poor children in primary and secondary education; increase public awareness of the benefits of secondary education at all economic levels; reduce the share of education budgets dedicated to salaries so that capital investment can be increased, and improve monitoring and evaluation of education programs. 105 6.tIf Costa Rica'rt levels of ed~icatici b.3j. 6.11 While the public sector traditionally has had the leading role in the education system in Costa Rica, the private sector is assuming an increasingly important role, particularly at the post- primary level, and it i s even taking the lead in higher education. Private enrollment is concentrated in the Central region of the country, which i s mostly urban. Around 80 percent of private students are in the central region, while less than 10 percent of students in private institutions are inrural areas.Il7 6.12 Although enrollment in private education i s still low (around 8 percent), the annual growth rate of private enrollment surpassed that of public institutions in 2003: 4 percent as opposed to 2.9 percent. A growing trend towards private high-school education is noticeable. Enrollment in private education centers doubled from 6 percent in 1980 to 12 percent in 2004. Today, almost one third of secondary schools are private (29.8 percent), which not only shows a greater demand for this type of education, but also indicates parents' reservations regarding the quality of secondary public education (State of the Nation, Reports X and XI). 6.13 For higher education, enrollment i s higher in private universities than in public ones. In 1978, the first private higher education institution, la Universidad Aut6noma de Centro Amkrica, was formed. Since then, the number of private universities has increased, reaching 49 in 2002. While 40.5 percent of college graduates came from private universities in 1994, that had increasedto 60.8 percent 10years later, in 2004. 6.14 The structure of the Costa Rican educational system did not change significantly between 1990 and 2005, but more than 30 individual programs were createdto address specific issues like quality (SIMED1I8), infrastructure (PROMECE"'), access (distance learning or "telese~undarias,"'~~),quality of preschool and general basic education (PROMESA'"), expansion of information technology, attention to special groups (indigenous, rural schools), equal access (PROCUMEN'22), school transport, scholarships, and enrollment rates. Most of the programs during this period were were not created as part of an overall plan or reform process (Estado de la Educaci6n 2005). The only effort to reform the entire system, the Edu-2005 project, did not have the support of the education sector or the backing of the National Congress (Barahona y Castro, 2003; Mora and Ramos, 2004). 6.15 Inthe lastfive years, educational policy hasfocused onhiringnew teachers andbuilding new educational facilities, but educational quality has not improved. A very comprehensive education policy framework by the MPE-Education Re-Launching - covers all education levels from day care to the fourth cycle, and addresses subjects such as subsidies, infrastructure, special groups, quality of education, teacher promotions, monitoring and evaluation, equity (access), organization and social aspects. But these guidelines do not constitute a major reorganization of the educational sector; they do not `provide for important financial changes, and the established command structures remain the same. Inshort, the framework can be seen as a set of middle-term measures that lack provisions to implement changes in an effective and efficient way. Some of the major shortcomings are that it lacks provisions to monitor and evaluate reforms and innovations, has little political support, and comes with few resources to implement and consolidate innovations (SANIGEST 2006). II7The rest are distributed in urban centers outside the Central region. 11*Sistema Nacional para el Mejoramiento de la Calidad de laEducaci6n Costarricense. `I9Programa de Mejoramiento de la Calidad de la Educaci6n Preescolar y General BBsica. 120"Remote" learning for secondary education. 121Programa de Mejoramiento Econ6mico y Seguridad Alimentaria 122Programa de Escuelasde Atenci6n Prioritaria. 107 6.16 In its continuing efforts to improve educational outcomes and create educational opportunities for those who, for various reasons, have not had access, Costa Rica recently developed the Educational Plan 2002-2006. The MPE has put forward four concrete goals for educational development: (i) to develop educational opportunities by allowing access, retentionof students through theirs school years, and encouraging scholastic success for all students (ii) to empower students with the knowledge, abilities, skills, attitudes and aptitudes to strengthen their integral development; (iii) strengthen integration and pertinence among diverse educational to approaches offered in technical instruction and respond to the needs and requirements for local and national economic development; and (iv) to improve MPE management increasing managerial efficiency, process transparency and resource allocation, rationality and process optimization. Some of the specific actions include quality improvement and universal access to pre-school, and improvement of primary education, especially for vulnerable groups (such as indigenous communities and children in schools with only one teacher). Educational outcomes 6.17 For this section of the report, education performance was evaluated using mainly two data sources: the Ministry of Public Education (MPE) and the yearly household survey (EHPM). Both sources have their strengths: the MPE has information for almost all students in Costa Rica, and the EHPM can be linked directly to other household characteristics such as poverty. Both have their weaknesses as well: the MPE computes enrollment rates using administrative records and census projections, and reports net enrollment rates above 100 percent,'23 and the EHPM does not have the month or day of people's births, which introduces some bias into the computations. HistoricalPerspective 6.18 The core theme that guided Costa Rican education until the end of the twentieth century was to promote education as the main and irreplaceable instrument for social mobility. The most significant state contribution to education came during 1956-1979, when public investment practically doubled to 6% of GDP. This happened, in part, becausethe mandatory period of basic general instruction was extended during the 70s to nine years. 6.19 The Costa Rican educational system was severely affected by a major economic crisis that began inthe 1980sand lasted Figure 6.4: GDP Growth Rates in Costa Rica 1979-2002 more than 15 years. From 1980 1 p10% to 1982, GPD growth was negative (Figure 6.4), average 5E 5% income decreased, and the B Ministry of I::g* Education's 0% expenditure per student decreased -5% by more than 40 percent.'24 After 1983 the per-student expenditure -10%4 slowly increased, but it did not I I , I , I , I , , , I , I , I , , I I , I ,j l reach the 1980 level until 1998. I Year 1 Source: SANIGEST 2003 lZ3 net Since enrollment rates are the number of children attending the appropriate school grade, divided by the total numberof children that should be at that school grade, values above 100% are not possible. '24From$352 down to $210 per capita (persons5 to 24 years old) inconstant 2003 dollars ($1=399 colones) Source: SANIGEST 2003 108 6.20 Lower personal income and the reduction of government expenditures on education had a major impact on education results. Attainment percentages peaked just before the economic crisis then decreased and did not reach the previous high levels until 14 years later. Figure 6.5 shows the attainment levels for primary (I and I1cycles), I11cycle (ninth grade), and secondary (eleven grade), usingthe 2004 EHPM.'25Attainments improved for each age group from those who were 85 in 2004 to those who were 40, then deteriorated and took almost 15 years to recover.I26 Figure 6.5: Grade Attainment Level by Age in2004 90% *6 80% 70% P- 6oyo 50% '' 5 40% ae 30% 20% ~~ 10% I 0% 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 Age in years Note:grade 6 markedperiod(---):38 to 24 year-olds; Grade 9 and 11markedperiod(-): 40 to 27 year-olds Note:individual valuesare three years movingaverages. Source: World Bank staff calculationswith the 2004 EHPM 6.21 To provide a long term perspective on educational system outcomes, the rest of the section will look at the period when the country had achieved attainment levels similar to those before the crisis. Due to data limitations, the analysis will focus on the years 1989, 1994 and 2000-2004. During those years, the EHPM also collected housing characteristics from the households, allowing a more complete poverty analysis in other sections of this Developments after 1989 6.22 Costa Rica has made significant progress in raising enrollment at the primary and secondary levels since 1989. The improvements have reached the poor as well as the non poor. Data show increases in the primary and secondary net and gross enrollment rates for all income quintiles and a reduction in enrollment gaps between the highest and lowest quintiles.'28 The A moving three year average was usedto smooth the presentation. 126A more detailed examination of the data shows improvement for sixth grade attainment up to 38 year- olds and for ninth and eleventh grade attainment up to 40-42 year-olds. Those years represent populations finishing the corresponding level around 1981to 1983. Also, the EHPMdoes not have comparable data before 1987. I28This is true for all the gaps in relative terms and for all the gaps in absolute terms with the exception of secondary net enrollment rates. 109 improvements came particularly in the 1989-1995 and 2000-2004 periods (Figures 6.6).'*' Improvements in primary net enrollment rates (Figure 6.6.a) were smaller, but the absolute levels were already high (almost three quarters for the lower quintile and 81 percent, on average by 2004). Improvements in secondary school enrollment were much more impressive in percentage terms, but were low in absolute terms. Secondary enrollment rates increased 77 percent (net) and 91 percent (gross) for the lower quintile, with average changes of 48 percent (net) and 69 percent (gross). Since the improvement in the lower quintileswas above the average, the gap between the lower and higher income quintiles decreased: in short, there was more education and less inequality! 6.23 But importantdifferences remain between income levels, and the absolute values are low, especially in secondary education. With barely a 70 percent net enrollment rate for primary education by 2004 and a dismal 33 percent for secondary education for the first (poorest) quintile (Figures 6.6.a and 6.6.c), the government of Costa Rica has a long way to go to provide the poor with the necessary basic skills to break the poverty cycle. Government policy should be designed or adjusted to target the poor, reduce the education gap and increase overall enrollment levels. Figures 6.6: Net and Gross Enrollment Rates for Primary and Secondary by Quintiles, 1989-2004 Figure 6.6.a: Primary Net enrollment rate Figure 6.6.b: Primary Grossenrollment rate I 100% 7 .. \I 120% 1 1 110% 100% 90% 80% 70% 60% 50% I 1989 1994 2000 2004 1989 1994 2000 2004 129Estimates are based in mid-year attendance. For the estimation of net enrollment rates using the EHPM, only the birthyear was available. For many students, the lack of a precise birth date could have placed them at the appropriate level or one year behindtheir level. I n those cases the favorable assumption of being at the appropriate level was made. The result of this assumption is net enrollment rates are overestimated. 110 ' Figure 6.6.c: Secondary Net enrollment rate Figure 6.6.d: Secondary Gross enrollment rate 120% loo%% 80% -I V I 100% 60% 80% 60% 40% 40% 20% 20% 0% 0% I 1989 1994 2000 2004 1989 1994 2000 2004 Net enrollment rates: # students at the appropriate level (for their age) dividedby the total # of students that should be at that level Grossenrollment rates:# students at the level divided by the total # of students that should be at that level Note: Nationalaverage values are very similar to the third quintile figures. Source: World Bank staff calculations using the EHPM 6.24 Since both net and gross rates are growing at the same time, one can only conclude that the loss in efficiency reflected in higher gross enrollment rates is due to late starting and re-entry to the school system (although, as presented later, the average number of repeated years has decreased), with no direct negative effects on net enrollment rates. This can be a consequence of a rapid expansion of the educational system but, even in such a case, attention should be given to improve efficiency within the educational system to meet existing or new problems posed by the higherenrollment rates. 6.25 Government school enrollment rates show very similar patterns as the EHPM data. Comparisons of official figures from the MPE (Figure 6.7) with values from the household survey show very similar patterns over the years. The fact that government enrollment rates are computed with population projection^'^' is probably the main reasonfor the discrepancies. Between 2000 and 2004, net secondary school enrollment rates increased nine percentage points in both cases (MEPand EHPM), and secondary gross enrollment rates increased 19(EHPM)and 21 percentage points (MPE). For primary gross enrollment, both data sources show very similar values: a half percentage point increase (EHPM) and a one percentage point decrease (MPE). Only primary net enrollment rates are slightly different: a five points increase (EHPM) compared with a one point decrease (MPE). 130Population projections are used in the denominator of all government enrollment rates. For the numerator, actual total number of students is used for gross rates and number of students at their appropriate grade (according to age) is usedfor net rates. 111 Figure 6.7: Gross and Net enrollment rates by level 1999-2004 I I 111.1% A v v A A v A v *109.9% 1 40% 1 I I I 1 I I I 1 1999 2000 2001 2002 2003 2004 Year '&Gross Primary - 0 -Net Primary-Gross Secondary - -Net Seconda Source: Minister of Public Education 6.26 Better enrollment rates produced higher completion rated3' at all Figure 6.8: Average CompletionRates by education levels between 1989 and Education Level, 1989-2004 _ _ - ~ _ _ _ _ _ 2004. In the long run, completion rates follow a very similar growth rate as net enrollment rates. The actual values are very similar too. For example, primary _ _ _ ~ ~ _ _ _ _ completion rates increased from 75 70% percent in 1989 to 86 percent in 2004 I (Figure 6.8), while average net enrollment rates increased from 75 percent to 81 percent. During the same period, secondary completion rates (I11 and IV cycles) grew from 31 percent to 45 percent, and average net enrollment rates increased from 32 percent to 47 percent. 1989 1994 2000 2001 2002 2003 2004 6.27 But, as with net enrollment, Finish6=primary; 9=III cycle; and 1l=lV cycle or secondary. disparities between income groups in Source: World Bank staff calculations with the EHPM completion rates have not changed, and absolute levels remain low, especially for secondary education. By 2004, primary completion rates of barely 70 percent for the poorest quintile and less than one in five for secondary graduates arejust too low (Table 6.2). Secondary completion rates for the second and third quintile (22 percent and 27 percent, respectively) are barely better. ~- ~ 131Completion rates are measured as the percentage of persons who finishedeach level. The reference groups are 14 year-olds for primary, 18 year-olds for I11cycle and 20 year-olds for IV cycle (secondary). 112 Average secondary completion rates are only slightly better for the non-poor: 40 percent. Clearly, secondary education has problems that go beyond the poverty status of the p~pulation.'~~ 6.28 Improvements in attainment levels have been poverty neutral at best and slightly regressive at worst. Moreover, government efforts to improve education levels in Costa Rica have not improved disparities between income groups. The differences between the fifth and first quintile have remained the same, with a tendency to increase between 1989 and 2004 for all levels of schooling: 1 point for primary, 4 points for I11cycle and 2 points for IV cycle. The government of Costa Rica should look at the low secondary completion rates as a national problem with higher incidence within the poor. 2000 78 64 80 95 31 45 22 45 61 39 35 14 19 66 51 2004 86 71 89 98 26 52 30 47 80 50 37 19 27 59 40 04-89 11 9 16 10 1 17 13 19 17 4 9 10 0 12 2 6.29 Regardless of the problems with enrollment and completion rates, the government of Costa Rica and the educational system in general have made considerable progress in increasing school access133and reducing inequalities between income groups. These gains have granted seven to 12 year-old poor children unprecedented access to school and 13-to-17 year-old poor children considerably better access as well. Indeed, data show attendance levels by seven to 12 year-olds in the poorest quintile are above 97 percent'34(up from 90 percent in 1989), reaching basically universal access (Figure 6.9.a). For the 13 to 17 years-old group the improvements are more impressive, with attendance levels for the first quintile children going from 45 percent in 1989 to 73 percent in 2004: a 62 percent increase! Attendance rate differences between the first andfifth quintilehave also decreasedsubstantially inrelative and absolute terms. 132Later in this chapter, reasonsfor not attending schools (among 13 to 17 year-olds) are analyzed. 133Access measured as the presence of children in the school system or attendance to school. 134The attendancelevels for the extreme poor (97.3 percent) and for rural children (97.5 percent) are basically the same. 113 Figures 6.9: School Attendance For Different Age Groups And Quintiles: 1989-2004 Figure6.9.a: 7 to 12Year-olds Figure6.9.b: 13to 17Year-olds 100% I V ~ 100% V 111, V 98% 90% 96% 80% 94% 70% 60% 92% I 50% 90% 40% ~ 88% 1989 1994 2000 2004 Linerepresentsthe averagevalues ~ ILine represent the average values 6.30 To increase our understanding of Costa Rica's educational system outcomes, impact and relevant factors, five more aspects will be reviewed: (i) students attending school at their appropriate level; (ii) number of underperformance years by students lagging on their school grade135;(iii) average years of schooling (attainment) for the 18 year-and-older population; (iv) relationship between poverty and attainment levels and (v) reasons for not going to school. 6.31 Improvements in enrollment and completion rates, and in attendance levels, have not Figure 6.10: Students Going to Their Appropriate increased the proportion of ISchool Grade By Age Groups: Costa Rica 1989-2004 students who are in school at the - - appropriate level for their age. 100% The increased levels of net school 1 90% enrollments observed between 80% 70% 1989 and 2004 were accompanied , 60% by proportional increases in 50% students lagging behind their 40% appropriate school grade.136 30% Since 1989 the proportion of 20% students at their appropriate level 0% has remained at around 80 percent 1989 1994 2000 2004 for seven-to-12 year-olds, with small variations, and at around 56 Source: World Bank staffcalculationsusingthe EHPM percent for 13-to-17 year-olds (Figure 6.10). 135That is, of the students laggingon their school grade, the number of years they are behind. 136Since gross enrollment rates includestudentsat any age, no directcomparisoncan be made with the results presentedhere. 114 6.32 For people lagging their grade level, the average number Figure 6.11: Underperformance Years by Quintile of underperformance years has 5.0 , 1 increased since 2000. After an 5 4.5 1 initial reduction from 1989 to % 4.0 1994, the average number of .I- o 3.5 lagging years rebounded in 2000-2004 by almost one year g! s 3.0 - (Figure 6.11). Average z 2.5 ! underperformance has 2.0 I decreased only 0.75 years for 1989 1994 2000 2004 the 15-year period (around 1 Source: World Bank staff calculations usingthe EHPM percent per year), and remains almost four years on average. 6.33 Average years of education have improved, but differences by income level and between urban and rural households have remained constant over time. Average years of education increased from five years to six and a half years from 1989 to 2004; but inequalities between quintiles and between urban and rural households have not changed in relative terms (Figures 6.12.a and 6.12.b). Gender differences are very small (0.2 years in favor of females) and have not changed since 1989 (Figure6.12.c). 6.34 Place of residency is a good proxy for education performance. Average years of education are very similar for rural households and the second income quintile, and for urban households and the fourth quintile. In the absence of any other indicator, a rural place of residency is a good proxy for persons with bigger educational needs. 115 Figures 6.12: Average Years of Schoolingfor 18Year-olds and Older: Costa Rica 1989-2004 Figure6.12.a: By Quintile Figure6.12.b: By Urbanand Rural 7 E Na ional r ~ 0 # ~ 5 E Rural ~ I 4 (D Q) * 1989 1994 2000 2004 c0 Figure6.12.c: By FemaleandMale I 7 1 I 1 0 1989 1994 2000 2004 4' 1989 1994 2000 2004 Source: World Bank staffcalculationsusingthe EHPM 6.35 Immigrant households (from Nicaragua) have significant lower levels of education, similar to households in the second income quintile. Persons living in households with 20 percent or more members born in Nicaragua have almost two years (a third) less education than the rest of the population (Table 6.3) - an amount very similar, on average, to people in the second quinitile. The differences are close to half a year for primary age children and to one year for secondary age children. Source: World Bank StaffCalculationsusingthe EHPM. 6.36 There i s a strong, negative relationship between years of education and poverty rates, with a premiumfor finishing I,I1and IV cycles. As the education attainment of people who are 18 years and older increases, the percentage of poverty and extreme poverty decreases in the population (Figure 6.13). The decrease in poverty i s bigger for a person finishing the first (third 116 grade), second (primary) or `forth cycle (secondary), suggesting a premium for completing either of the educational cycles'37. Figure 6.13: Poverty Level and Years of Education for 18 Year-olds and Older: Costa Rica 1989-2004 35% 30% 25% 20% 15% 10% 5yo I II Povertv IV 0% b- 0 2 4 6 8 10 12 14 16 Years of education Source: World Bank staff calculations usingthe EHPM 6.37 Economic incentives in the form of conditional cash transfer programs would improve attendance for 13-to-17 year-olds. Such programs should be accompanied by efforts to explain the benefits of secondary school. Regardless of income level or gender, lack of interest and economic reasons are the leading causes for 13-to-17 year-olds not to go to school. (Table 6.4). There are only some differences between income and gender groups. Most noticeably, there are higher "not interested" response rates from people in the higher quintiles and males; higher "economic reasons" rates from people in the lower quintile and male respondents (especially in the category "too expensive" for the first quintile), and higher "sickness, pregnancy or marriage" responses from females. 137Primiumsare representedin figure 6.13 by the sharper decreaseinpoverty rate for peoplefinishingthe I,andIVcycles. I1 117 Table 6.4: Reasonsfor 13 to 17 Year-olds Not to Go to School by Quintiles and Gender in2004 1 Low number of persons not going to school inthe higher quintilesdid not allow for individualquintile reporting. For the lowerquintilesthe numberof personsnot going to schoolwas higher and no groupingwas necessary. Source: World Bank staff calculations usingthe EHPM Government expenditures, incidence, efficiency and quality Expenditures 6.38 Public spending in education has increased since 1983 in absolute and relative terms; the rate of growth accelerated in 1999. After the 1980-1982 budget crisis, expenditures by the Ministry of Public Education (MPE) slowly recovered. By 1988, per capita expenditure^'^' had reached pre-crisis levels, but was not until 2003 (23 years later) that the it MPE expenditures reached 5.0 percent of GDP, the pre-crisis value. Several factors are responsible for the recovery. First,for the financing of the educational sector, decade cumulative GDP improvement values are the key determinant. During the 1990s, GDP growth had been erratic, and no constant expansion of national income was achieved. But by 2002, this value was 70 percent higher than in 1991. Second, total central government expenditures as a percentage of GPD also improved after 1993. Third, after 1996 the share of the MPEexpenditures in relation to total government expenditures increased from 20 percent to 27 percent in 2002. And finally, a constitutional amendment set the minimum share of educational expenditures at 6 percent of GDP. 13'Per capita over total number of 5 to 24 year-olds. 118 Figure 6.14: MPEExpenditureLevels: Per Capita and as GDP %, Costa Rica 1978-2003 $500-f In $400 ?n0 4- 3.0% C $300 8 I: #E 2.0% $200 I 8 1.O% capita I I ! I I $100 I 8 o.ooYo I $- Source: SANIGEST2006 6.39 With more than five percent of GDP allocated to public spending on education in 2003, Costa Rica ranks among the highest in Latin American'39. Not only that, but with $9,606 (2003 PPP dollars) GDP per capita, Costa Rica ranks third in Latin America and well above middle income countries ($6,104; 2003 PPP dollars).'@ Government proposals to increase that share to eight percent will certainly set a new standard for the region, not only in relative terms but also in absolute values. Figure 6.15: Government Expenditurein Education as Percentage of GDP, 2001-2005 I 7 ..................................... I , > , Source: World Bank Lindert, Skoufias y Shapiro (2005) - 139 Similar comparisonswere madeusingper capita expenditures,and the qualitativefindings remainthe same. Costa Rica's rank remainsamongthe highestin LatinAmerica I4O UNDPHumanDevelopmentReport2005. 119 Administrat Tslblcs 6.5: ~~P~ PublicSchools Subs ies, Casta Rim 2004 5.5.b: Shs budRct and quintitctot Efficiency 6.44 Costa Rica has low levels of efficiency for primary education. Higher rates of investment by the MPE have not fully paid off yet. Compared to other Latin American countries, Costa Rica's expenditure in education (as percentage of GPD) has not achieved the expected primary net enrollment rates. Even though there i s no direct relationship between the expenditure level and the enrollment rates,'@examination of Figure 6.18 places Costa Rica with expenditure levels above the average'45(vertical line), and with primary net enrollment rates lower than the average (horizontal line). In general, countries in quadrant Iare "high performers" and those in quadrant IVare "low performers." Figure 6.18: Primary Net Enrollment Rate and EducationalPublic Expenditure inLatin America, 2003 I 102 100 2 0 98 c -- E 96 94 c 92 90 `E 5 .- 88 86 84 0 2 4 6 8 10 Public Expenditure in Educationas %of GDP Adjusted R2=-0.07;t and Fvalue non significant at p=97%, Honduras 1999/2001; Ecuador, Brazil and Cuba= 2002; all others, 2003 Source: UNESCO, Institutefor Statistics, WEB page: http://www.uis.unesco.org/ev.php?URLJD=5 187&URL~DO=DO~TOPIC&URL~SECTION=201 6.45 Levels of efficiency in Costa Rica are not better at the secondary level than the primary level. Given Costa Rica's level of public expenditure on education, secondary net enrollment rates should be 11 percentage points above the 2003 value. With public expenditures in education at 5.1 percent of GDP, Costa Rica's secondary net enrollment rate should be closer to 64 percent (Figure 6.19). In general, countries on top of the fitted line in Figure 6.18 are performing above expectations and countries below it are underperformers. This is true for international comparisons using the selected countries. Possible reasonsare the low levels of enrollment rates variability, and country specific differences in the methodology used to compute expenditure and enrollment rates. Several functional forms were estimated, and none was significant; the results of the linear regression are included for illustrative purposes. 145Average within the sampled countries. 122 6.46 But one has to consider the evolution of public spending when evaluating primary and secondary net enrollment rates. While spending by the Ministry of Public Educationrepresented 5.1 percent of GDP in 2003, only four years early that value was 3.5 percent (28 percent lower!). The rapid rate of change in spending would dramatically change the efficiency conclusions reachedearlier. Education improvements take long pCriods of sustained support. And as shown earlier, it takes a long time to recuperatefrom shocks to the system such as the economic crisis at the beginningof the 80s. Figure6.19: Secondary NetEnrollmentRateandEducationalPublic , ExpenditureinLatinAmerica, 2003 90 85 Arg. 80 K 44 Chile * 75 -25 E El Salv., Bolivia 70 Peru* 65 60 z 55 F 0 ?' 50 Ecuador . -.. U5 45 Uruguay II C.R. $ 40 I cn I 35 1 4 Nic. Domi. R.4 I 0 2 4 6 8 10 Public Expenditure in Education as %of GDP ,djusted Fi2=0.37;t and Fvalue significant at p=0.5%, Honduras 1999/2001;Ecuador, Brazil and Cuba= 2002; all others, 2003 Source: UNESCO, Institute for Statistics, WEB page: http://www.uis.unesco.org/ev.php?URL_ID=5 187&URL-DO=DO-TOPIC&URL-SECTION=201 Oualitv of Education 6.47 To assess the quality of education, four indicators were analyzed: (i) student-to-teacher ratios; (ii) quality of infrastructure; (iii) number of computers per student; and (iv) standardized test scores. 6.48 The ratio of students per teacher has improved in the 2000-2004 period for all education levels. The number of students per teacher i s recognized as an important parameter for school quality: as the ratio decreases, the quality of education improves (ceteris paribus). Costa Rica has made important improvements - it reduced the ratio by 21 percent, or 5.4 students, for primary schools, for instance (Table 6.6). Such reduction in the student-to-teacher ratio is particulary impressive since it was achieved at a time that gross enrollment rates were increasing. 123 Stage 2000 2001 2002 2003 2004 Preschool 19.6 19.5 18.7 17.2 16.3 Primary (Cycles Iand 11) 25.3 24.7 23.5 21.5 19.9 Secondary (Cycles I11and IV) 18.9 19 16.6 16.6 16.1 6.49 The quality of public education infrastructure remains low - well below private school infrastructure. The probability that students will perform better increases with adequate physical resources (better buildings, bathrooms, classrooms, etc.). Two indicators were chosen to assess the quality of infrastructure: percentage of classrooms in "good condition"146 and number of desks per student. By 2002 less than 70 percent of primary or secondary classrooms were in good condition (Table 6.7). In other words, three out of ten classrooms did not fulfill the basic requirements for teaching. In private institutions, an average of 99 percent classrooms were in good condition. Indesks per students at the primary level, state institutions also are well behind private schools (they have 3.4 times more students per desk), Assuming two shifts a day, there are 3.2 primary students per desk, well above the optimum value of one. The gap between private and public institutions i s less for secondary school, but still substantial (2.6 times more students per desk). Table 6.7: Costa Rica's SchoolInfrastructure Indicators, 2002 I Satisfactory Classrooms Students per Desk EducationLevel Public Private Public Private Primary (Cycles Iand 11) 68% 100% 6.4 1.9 Secondary (Cycles I11and IV) 69% 98% 2.9 1.1 some cases up to three shifts per day (the third shift composedby night school students) Source: Based on MPEdata 6.50 Public schools' computer access lags far behind private and semi-private institutions. The ability to use computers and information technology is now considered a basic skill required inmanyjobs. Early exposureto the available tools, mainly computers and internet, is considered a must for a well- prepared labor force. The Government of Costa Education level Public Private Primary (Cycles Iand 11) 125 13.1 Rica has a long way to go to Secondary (Cycles I11and IV) 100 14.3 provide its students with the minimumtraining required to meet this need. By 2002, a student inthe public education system had to share a computer with one hundred other children. Private schools had many more computers per student: seven times what public schools had at the secondary level, and almost ten times what they had inprimary schools. 6.51 Standardized tests were re-introduced into the educational system in Costa Rica in 1988.'47 By 1994, a high failure rate led to pressures to changes the system. Also, the standardizedtests were perceived as exams exclusively based in memory, and thus an inadequate reflection of the objectives of the educational system. In 1999, the tests were modified to reflect the objectives instead of the contents of the educational curriculum. By 1999, standardized test were required for sixth, ninth and eleventh grade promotions. Promotion was based on a combination of test results and average school grades. The relative weight of grades and tests 146Evaluation of classroom conditions done by the MPE 147In 1973, standardized test were eliminated from the educational system 124 also has changed over the years. Since 2002, standardized tests make up 60 percent at all three levels (11, I11and IV cycles), and school grades represent40 percent.14 6.52 Influenced by the modifications introduced over time, test scores have fluctuated from year to year in all subjects. Table 6.9 shows a sample of such fluctuations for cycle I11and IV tests, with average test scores changing as much as 25 percentage points in one year. Given the highvariability of the test scores, very little can be said of their evolution over time. Table 6.9: Standardized Average Test Scores and Yearly Change for Selected Subjects, Levels and Years Source : MPE,20005 6.53 Private institutions prepare students better for the standardized tests, and no difference is Table 6.10: Cycle IV PassRates By School Type And found between urban and Region rural institutions. There are Year significant differences in 2000 2004 average pass rates between NATIONAL AVERAGE 62% II 62% public and private institutions School type for cycle IV (grade eleven) Public 56% 56% tests. On average, private Private 84% 83% institutions' pass rate i s 28 Subsidized 86% 87% percentage points higher than REGION that of public institutions Urban 62% 62% (Table 6.10). Contrary to Rural 59% 61% popular belief, no significant source: SANIGEST, 2006 difference i s found in cycle IVstandardizedpassratesbetweenurban andrural schools. Millenium DevelopmentGoals (MDG)for Education 6.54 There are two millennium development goals related to education: first, to achieve universal primary education (MDG #2); and second, to promote equity between sexes and the autonomy of women (MDG#3). 6.55 With 99 percent of 7-to-13 year-olds attending school, Costa Rica has achieved almost universal coverage for primary ed~cati0n.l~~Although the goal and objective of the MDG #2 148The reference period for school grades are the last two years (excluding the last trimester) for the I1and IV cycles, and the last three years for I11cycle. 14'Coverage defined as persons within the age range oing to school regardless of the grade they are attending to. The M D G refers to completion rates as latter stated in Costa Rica M D G goal one. 125 refers to primary education, the government of Costa Rica formulated specific goals other levels as well: 1) 100 percent primary completion rates by 2015; 2) 99 percent literacy for 15-24 year- olds by 2015; 3) for preschool education, reach 99 percent coverage for transition level by 2006 and 72.3 percent at interactive level I1by 2015;1504) increase I11and IV cycles coverage to 89 percent by 2015; and 5) eliminate inequality between the sexes at all levels by 2015.15' 6.56 To evaluate Costa Rica's progress, toward achieving the MDGs, past performance is considered the best predictor. Simple linear projections are made, and the assumptions behind them are qualified. The objective is to provide an idea how probable it is that the country will achieve the MDGs and to identify the necessaryconditions for it to do so. Of course, hocks to the systemor simple policies changes can alter the conclusions (in either direction). COSTA RICA'S GOAL, 1: 100PERCENT PRIMARY COMPLETION RATES BY 2015 6.57 As stated, Goal 1 is almost impossible to achieve. Even among high-income nations, some have not been able to achieve universal completion rates.I5* There will always be a small share of the population that i s very difficult, if not impossible, to reach. Among the reasons Figure 6.20: Primary CompletionRateswith Projections, hindering sixth grade Costa Rica 1994-2015 attainment are I difficult access to remote areas, cultural barriers, and development problems. Most of these problems will no be visible in the official statistics until attainment levels are closer to 100 percent. In general, reaching universal attainment 1 8 0 Y 0 1 , , , , , I only possible by I zz % %, , g, , g, , ,g, , p, , o, , , , rates in education is z z c u c u c u 8 8 c u c u a] making special Source: World Bank staff calculations considerations in the way the statistics are produced. Usingthe last 10 years as the reference period, a linear regression predicts completion rates of 96.3 to 99.4 percent by 2015 (Figure 6.20). The higher rate i s based on the observed rate of increase for the last 10 years; the lower prediction i s based on the observed growth rate for the last three years. The dotted line in Figure 6.20 represents the possible behavior with a different functional form, Outlook: di`cult to achieve, but signijkant progress in the right direction. COSTA RICA'S GOAL 2: 99 PERCENT LITERACY FOR 15-24 YEAR-OLDS BY 2015 6.58 According to the 1984 population census, the literacy rate among 15-to-24 year-olds was 97.1 percent.i53Sixteen years later, the 2000 census found that the literacy rate was 97.6 percent, I5OInteractive level I1refers to children between four and five years old. Transition level includes children between five and six and a half years old. 15'Since MDGs were defined by the government, official government figures are usedto evaluate them. For example, net primary enrollment rates by 2003: United States 92 percent, Ireland 96 percent, Australia 97 percent, Netherlands 99 percent, France 99 percent. Source: UNESCO 2005. 153Measured as the percentage of people who can read and write. 126 or one half percentage point higher. With the same rate of increase, literacy rates would be 98.2 percent by 2015. 6.59 Improvement in literacy rates are expected in future years. Nevertheless, reaching illiterate groups i s a difficult task, and it is hard to believe the proposed 99 percent rate can be completely achieved by 2015. Outlook: diflcult to achieve but signijkant progress in the right direction. COSTA RICA'S GOAL 3: a) 99 PERCENT COVERAGE FOR TRANSITION LEVELBY 2006, AND b) 72.3 PERCENT COVERAGEFOR INTERACTIVELEVEL I1BY 2015 6.60 The transition level coverage rate reached 90 percent in 2004. Before 2004, the average increase in coverage was 3 percentage points per year. Assuming similar growth rates, the coverage level would be 96 percent by 2006. Outlook: partially achievable. 6.61 Regarding interactive cycle 11, before 2000 only private establishments provided the service, and the coverage level was very low. As of 2001, public establishments began to provide the service, and coverage rates increased. In the first four years of public efforts at Transition Level 11, the coverage rate has gone from 6.6 percent in 2000 to 37 percent in 2004. Between 2005 and 2015, the goal of the Department of Public Education i s practically to duplicate the 2004 coverage, so that 72.3 percent of the net population can be reached. This means an annual increase of 3.3 percentage points (compared to the 7.5 percentage point growth in the last four years). At a first glance, sub-goal 3-b seems reachable. Nevertheless, considering the strong investments needed for infrastructure and re-adaptation of curricular programs, accomplishment of the goal is not as easy as projections suggest. Outlook: achievable with high investment levels by the government. COSTA RICA'S GOAL, 4: INCREASE I11AND IV CYCLES COVERAGE TO 89 PERCENT BY 2015 6.62 Given the dynamism shown by the coverage of high school education in the last six years, it i s possible to reach the proposed goal by the government if, and only if, the historical tendency continues to be at least the same during the next ten years, as the following graph shows. This implies increasing the coverage at an average yearly rate of 1.9 percent points between 2005 and 2015, when the coverage would reach 89.9 percent (Figure 6.21). Outlook: achievable but subject to keeping the same growth rate experienced during the last six years. 127 Figure 6.21: Secondary Net Enrollment Rates With Projections, Costa Rica 1999-2015 90 - - 85 2 80 - 75 - m cd 70 E 2 - rsl 65 - BO - 50 I I Year Note: values after 2004 are projections Source: SANIGEST2006 COSTA RICA'S GOAL 5: ELIMINATEINEQUALITY BETWEENSEXES AT ALL LEVELS BY 2015 6.63 The ratio of female-to-male primary net enrollment rates is 1.01. The same ratio for secondary schools i s 1.1. In both cases, females have higher enrollment rates. If goal 5 i s interpreted as improving the relative condition of women to at least equal that of men, the goal i s already achieved. If goal 5 is interpreted as making education gender neutral, Costa Rica has to make efforts to increase male participation in secondary schools. In order to reduce the present gender gap, government policies should be oriented to reducing dropout rates, a problem that affects men more than women. No specific government action i s specifically being taken to improve male secondary enrollment rates. Outlook: from achieved to uncertain depending in how the goalis interpreted. Conclusion 6.64 This section has presented the evolution of educational indicators, based mostly on the EHPM data from 1989 to 2004 and official figures for international comparisons. It describes Costa Rica's educational system, public and private. The chapter analyzes educational outcomes and inequalities between income groups in depth over time. It also examines the efficiency and incidence as well as the general quality of the public education sector. The main messages from the education section are: 6.65 The economic crisis at the beginning of the 1980s had on the level of government investment in education and education outcomes that lasted over 15 years. Educational indicators substantially improved from 1989 to 2004 and, compared to other countries, Costa Rica performs well above in primary education and average in secondary education. But such achivements have not come cheaply; by 2003, education expenditures in Costa Rica were above expected values for the enrollment ratesachieved. 6.66 Government educational policy has not changed over the last 15 years; several programs have been designed to address specific problems, but no integrated reform has taken place. With 128 more than 90 percent of the MPE budget paying labor costs, new initiatives, investments or programs are all but impossible to implement. Enrollment in private education institutions is still low but shows significant increases, especially in secondary and advanced education. 6.67 Almost 100 percent of seven-to-12 year-old children go to school, regardless of their income level, but net enrollment and completion rates are significantly lower for poor children. The main problem in primary education i s not how to attract children to school, but how to avoid repetition and eventual desertion. The problem gets worse as household income decreases. Government programs and policies targeting the poor would reduce desertion and significantly decrease repetition rates in primary school. Since higher income groups do not have major problems in primary education, new universal policies would be a waste of resources. Programs targeted programs to the poor, such as Conditional Cash Transfers, should be designed to reduce desertion and decrease repetition-not to promote attendanceto primary education. 6.68 For secondary education, Costa Rica has serious problems in several fronts. First,overall graduation levels are very low (less than four in ten children graduate from secondary school); second, disparities between income groups are not improving and in some instances are even getting worse; and third, children from poor households (the lower quintile), show dismal graduation rates (less than two in ten graduate from secondary school). Lack or interest was mentioned as frequently as economic reasons for not attending school. The government of Costa Rica should look at the low performance in secondary education as a national problem with higher incidence among the poor. Economic incentives, such as those provided by Conditional Cash Transfer programs would improve attendance for 13-to-17 year-olds. Such programs should be accompanied by efforts to promote the benefits of secondary school, with special attention to poor children and their families' needs. 6.69 The public education system has serious infrastructure problems, as indicated by the low quality of classrooms, the high number of students per desk and the lack of access to computers. The Government of Costa Rica has to increase its support for the improvement of infrastructure. Given the current MPE budget share allocated to salaries, this i s basically impossibleto achieve. 6.70 Lack of well-specified objectives and the absence of a monitoring and evaluation system make it difficult to identify and eventually correct problems. It is recognized that the recent high growth rate of public expenditures in education might have an impact on the indicators the efficiency of which only future evaluations would be able to assess. Costa Rica has to improve its education monitoring and evaluation system, including by making changes to its basic indicators (small modifications to the EHPM would make it possible to track basic educational outcomes more efficiently) and special programs (to track that benefits from which programs, for instance). It i s difficult to evaluate the impact of specific programs or initiatives; overall evaluations of the educational systems are very useful as a first step, but improvements in the systemwill require better analysis. 129 7. HEALTH 7.1 Empirical evidence from different countries shows that health improvements have multiple benefits. They increase people's productivity, and thus enhance economic growth and personal income.They also improve the quality of life. For these reasons, investment in health should be part of the overall country development strategy (Sanigest 2006). 7.2 Costa Rica has understood this argument. The health investments it has made since the 1950shave produced major, lasting gains for the population. Life expectancy grew by almost one year between 2000 and 2004, and the child mortality rate has dropped below 10 deaths for every 1,000 live births,'thethirdlowest in Latin America (UNDP, 2003) (Table 7.1). 7.3 Nevertheless, the country still faces major challenges. Some diseases, such as dengue, have become more prevalent recently. Others, such as AIDS, have improved in the last 15 years, but are still worse than in 1990-1991 (table 7.1). Vaccination rates for measles and poliomyelitis have increased, but also remain below their levels in 1990-1991. To achieve better results and long-lasting benefits in these areas, a series of reforms are still required, and issues such as financial sustainability, resourceallocation, and global system efficiency needto be considered. 7.4 This chapter provides a comprehensive view of Costa Rica's health system and its dynamics, and identifies key policy issues that would improve the impact of public health programs on poor and vulnerable groups. It i s organized into six sections. The first section describes health trends in Costa Rica in comparison to those in other countries in the region, and discusses the organization of the health sector in Costa Rica. Section 2 describes and analyzes health-sector reforms during the 1990s and the beginning of the new millennium. Section 3 examines health-sector spending (public and total), resource allocation and efficiency. The fourth section describes variation in health based on income, geographic location and other factors such as immigration status; it looks at indicators and outcomes, the redistributive nature of public health spending and other issues such as housing services and immigration from Nicaragua. Section 5 examines Costa Rica's progress towards the Millennium Development Goals (MDGs), and assesses the country's outlook for achieving them. Finally, Section 6 presents conclusions and policy recommendations. 130 Source: Stateof the NationProject 2005; MIDEPLAN web page; and Observatoriode DesarrolloUCR 7.5 Several messages emerge from the chapter:first, institutional reforms have improved the public health systembut serious financial vulnerabilities remain; second, the country's high levels of spending on health (compared to other Latin American countries) have produced good results with a high degree of efficiency; third, the public health system i s the most progressive social sector in the country, with only small differences in outcomes based on income, geographic location and immigration status (specific indicators point to gaps in coverage that have to be addressedcase by case);fourth, recent increases (in the last five to 10 years) in the incidence of several diseases (malaria, dengue and tuberculosis) and decreases in vaccination rates reveal important weaknesses in the Costa Rican health system; andfi'h, the lack of better monitoring and evaluation systems i s an obstacle to evaluating the efficiency or impact of specific programs on the poor. 7.6 By Latin American standards, Costa Rican health indicators are very good, if not excellent. Out of 15 indicators analyzed (Table 7.2), Costa Rica was better off in: (i) all indicators compared to the world average; (ii) nearly all of the indicators compared to the other five Central American c~untries'~~;and (iii) five out of nine indicators (55 percent) compared to the Latin American average (it ranked first, second or third in nine out of fifteen indicators, or 60 percent, compared to individual Latin American countrie~).'~~Even compared to Northern America, Costa Rica has similar or better values in five out of eight indicators (63 percent) for which datai s available. 154 With the exception of Nicaragua inHIV/AIDS prevalence. For six of the indicators, no data was available for the Latin American average. Caribbean countries are not included in the table. 131 Health Sector Organization 7.7 Costa Rica's health system i s organized around the Ministry of Health (MH), which formulates policy and i s currently undergoing reorganization to take on oversight and regulation, and the Costa Rican Social Security Institute (CCSS'56),which administers health insurance. The National Insurance Institute (INS)'57 administers workers' compensation insurance and traffic accident insurance, which complement health insurance. Two other institutions -the Costa Rican Water and Sewage Institute (AYA'58),and the University of Costa Rica (UCR) - also are considered part of the health sector. Caja Costarricense de Seguro Social 157Instituto Nacional de Seguros Acuaductos y Alcantarillados 132 7.8 The private health sector, which i s relatively has been growing in recent years. It concentratesprimarily on providing ambulatory care and marketingpharmaceutical products. 7.9 The CCSS was created as an autonomous state institution to provide care to workers. Its creation was consolidated in the 1949 Political Constitution. Health insurance has gradually evolvedover the last 50 years. Among the primary mileposts in its development are the 1961Law on Universalizationof Social Security,*60which established the CCSS's constitutional obligation to make social security coverage universal within a period of 10 years; the 1973 transfer of the hospitals to the CCSS Law;16' the 1975 extension of disability, old age, and death insurance to agricultural workers; the 1978 creation of the Voluntary Regimen for Protecting Independent Workers and their Families and the 1993 Law on Improving the Health of Costa Ricans,162which transferred to the CCSS care services that were primarily preventive in nature and had been the responsibility of the MH. 7.10 Beginningin 1995, the MHtransferred practically all responsibility for providingprimary care to the CCSS in order to focus on its regulatory functions and policy development. The MH stills manages the major nutrition and ir~tegral-care'~~program for children, which plays a major role in the fight against poverty by providing access to food to thousands of children, and an integral support program for minors. MHactivities are financed through the national budget. 7.11 The CCSS administers most of the financing and provisions of the country's universal health insurance system. National health insurance covers approximately 90 percent of the population, offering workers a broad set of services financed through payroll deductions, and covering the poor and unemployed through governmental contributions. Services are provided based on social security principles (solidarity, universality, unity, compulsoriness, equality and equity). In addition to health services, the CCSS provides economic and social protection in the form of disability, old age, and death (NM in Spanish) benefits to the insured, low-income population. Also, as part of the health insurance, it provides subsidies to people who are disabled because of illness and maternity. The CCSS administers a 29-hospital network that has a little more than 6,000 beds that provide first-level care. More than 90 percent of CCSS income i s made up of contributions by participants. 7.12 The National Insurance Institute (INS) administers the state's monopoly on private insurance. Its role in the health sector i s to administer workers' compensation and traffic accident insurance, which provide medical, surgical, hospital, pharmaceutical, rehabilitative-care and economic benefits incases of work-related and traffic accidents. 7.13 The Costa Rican Institute of Water and Sewer Systems (AYA) provides potable water- supply services, sewage and industrial waste liquid collection and treatment services. It establishes the norms on how the rain water systems work in the urban areas. To finance its activities, it charges the cost of services to users of its networks. Various estimates, each with a specific methodology, produce different results. Klevsen estimates that the relative weight of the private sector i s 20 percent, while SBenz and Le6n calculate it at 23 percent and Durin and Herrero at 31 percent. 160 Law No. 5349 161 Law No. 5349 Law No. 7374 163 Integral care look at all helth needs of the children as related aspectsto be considered at the same time. 134 7.14 The University of Costa Rica contributes to the health sector by: (i)educating and training professionals and technicians in health, (ii) hosting and participating in research and social action projects involving health and; (iii) providing some health services.'64It finances its health professional teaching and training actions through its general budget, which in turn i s funded by direct annual transfers from the national budget. Health Sector Reform 7.15 The economic crisis at the end of the 1970s and the beginning of the 1980scaused major deterioration to the country's public finances, causing stagnation in investments in the health sector and leaving the country unable to broaden health coverage. Then, as part of an effort to reform the Costa Rican state at the beginning of the 1990s, a sector analysis was done to identify the health sector's most relevant problems. That led to health sector reform, which addressed three large action areas: the financing model, institutional reorganization and modernization of management of the service provider network. FINANCINGMODEL 7.16 For almost 40 years, social security financing had a weakness because it was difficult to enforce mandatory membership. Health reform sought to reduce the problem by: (i) increasing contribution coverage by strengthening enforcement, establishing better systems to detect evasion, and extending coverage to independent-sector workers; and (ii)redesigning the financing model by adjusting contribution rates for independent workers and retirees, and modifying the sytem for calculating state contributions to the CCSS to cover the poor and the uninsured. 7.17 As a result of the reform, real contributions from the directly insured exceeded the real expenditures of the directly insured; but the difference (in effect, the subsidy to the non- contributors) started to decrease in 2000, and the positive balance disappeared in 2003 (Figure 7.1). Three explanations for the tightening fiscal situation are: (i) consulted with health- people care professionals more frequently; (ii) the cost per visit increased; and (iii); the share of and people contributingto the systemdecreased. In areas surrounding its main campusin San Jose. 135 Figure 7.1: DirectCCSSInsurance Expendituresand Contributions inCostaRica 1990- 2003 Source: CCSS Statistical Yearbook 7.18 The increasein consultation rates were for children less than one year old and for persons 65 and older (Figure 7.2). Since both age cohorts are considered vulnerable, it is difficult to argue against the increase; nevertheless, it has had a negative impact on the system's financing sustainability that the government has to address. The cost per visit increasedby an average of 2.4 percent annually from 1990 to 2004 (Figure 7.3). Finally, the share of salaried workers contributing to the system decreased steadily from 75 percent in 1990 to 62 percent in 2004 (Figure7.4).165The share of non salaried workers contributing to the system has fluctuated in both directions; methodology issues make it difficult to assessthe net impact. 166 With two very small recovery periods: 1991to 1993 and 2003 to 2004 166The behavior observed in coverage for non-salaried workers is the result of a problem with independent workers registering. Prior to 2001, the database overestimated non-salaried workers' coverage; when an independent worker went into the salaried worker category the database was not cleaned up. 136 Figure 7.2: Yearly Consultations by Figure 7.3: Health Insurance Real Per Visit Age Group, Costa Rica 87,92,97 and Cost, Costa Rica 1990-2004'67 2002 6 j 5 0 f I a 4 0 g 3 0 1 I a6io 1 0 U 0 0 Year 81987 11992 01997 02002 Source: CCSS Statistical Yearbook Source : CCSS Statistical Yearbook Figure 7.4 Share of the Economic Active Population Contributingto System, Costa Rica 1990-2004 90 - EO - g 6:.8 60- ---/ 2: 5 0 - 30 ~ Source: State of the Nation INSTITUTIONALREORGANIZATION 7.19 Beginning in 1996, the Ministry of health assumed responsibility for oversight, coordination, and regulation, and the CCSS took over the role of insurer and provider of preventive medicine (primary), care for medical problems (secondary) and hospitalizations "'Costa RicanColones with January 2005 base CPI. June ConsumerPriceIndex values were usedto standardizeall expenses. 137 Figure7.5: h'um 6,000 5,000 4,000 3,000 2,000 1,000 North ~ ~ Brunca ~ Char. r Atlantic ~ ~ North l South ~ o ~ ~ t ~ Huetar Paciiic # de EBAlS Redistribution of these resources among the different care levels, however, had a larger impact on the second-level care, since hospital expenses (tertiary care) stayed practically the same during the period studied (Table 7.3). Table 7.3: RelativeDistribution of Health Expensesby Level, Costa Rica 1997-2004 Source: CCSS 7.24 The reform began in the poorer regions, bringingbetter health services to less privileged populations. More health establishments (Le. EBAIS) were created during the reform in the relatively less developed geographic areas, health posts, community doctors' offices and small clinics predominate (Rosero, 2000). 7.25 The gains from investments made during the reform process were highly positive. Investments made by diverse international bodies had a significant economic impact (Cercone and Briceiio 2003). The most optimistic scenario, assuming the results observed were the sole product of World Bank investments plus the local contributionby the CCSS (a total investment of approximately $168 million), suggests an internal rate of return of 68 percent with a present benefit value of approximately $82 million by 2002. If the contributions from other bodies are in~luded,"~ internal rate of return would be 33 percent and the net discounted benefits would the be $33 million. The primary benefits were seen in the ,reduction of hospital infections, hospitalizations and infant mortality (Table 7.4). PresentValue in Source: SANIGEST2006 with CCSS data Public Spendingand Efficiency PUBLIC SPENDING 7.26 Health spending in Costa Rica is very high by regional standards. The country's health expenditures are among the highest in Latin America in relative terms (as percentageof GDP), as " OAddingcontributionsby the Inter-AmericanDevelopmentBank and the Governmentof Spain (for a total investmentclose to $217 million.) 139 well as in absolute terms (per capita PPP 17'). This is true for both public and private health expenditures. The UNDP Human Development Report (2005)'72reveals that Costa Rica spent 9.0 percent of its GDP on health in 2002 (6.0 percent in public and 3.0 percent in private spending), almost two percentage points above the Latin American average.'73 In 2002, total health spending per capita (PPP) in Costa Rica was almost 75 percent higher than the Latin American average (Figures 7.6). Figures 7.6: Health Expenditures inLatin America, 2002 7.6.a: Total as Percentage of GDP L 12% n $ 10% c f 8% 5 c 6% 0, P W x 4% -5I 2% I0% 7.6.b: Public as Percentage of GDP 2% 0% 8% 6% 4% 2% 0% ~ ~ ~~ 17'PPP standsfor Purchasing Power Parity units created by applying an especial exchange rate designed to make fair monetary comparisons between countries. $1 PPP i s equivalent to a US$1.06 from 1993. 17*Human Development Report (HDR)data was used because other indicators utilized for efficiency were present. The HDR data i s used for international comparisons. For the rest of the report, official government figures are employed. 173The average of the countries presentedin this comparison. 140 7.6.c: Total Per Capita inPPP LAC values are the average of the includedcountries Source: UNDP 2005 HumanDevelopmentReport, from World HealthOrganization,2005 (March healthexpenditure) 7.27 Public health spending in real 1996 colones grew from $376,812 billion in 1995 to $551,347 billion in 2000 and to $643,992 billion in 2004. During 2004, spending on health represented almost 30 percent of social investment in Costa Rica, making health the second largest sector after social assistance (32.6 percent). This does not include expenditures by the Institute of Water and Sewer Systems, the University of Costa Rica, nor the National Insurance Institute, though. An estimate by SANIGEST (2006) that included expenses from all entities showed that investment in the health sector grew to 40 percent of the total social investment, the highest for all social ~ect0rs.l~~Over the last five years, public spending on health has represented 5.3 percent of the GDP, and could be close to 7 percent of the GDP if the broader measurement i s used. Between 1990 and 2004, real public spending on health increased2.1 times. 7.28 Real per capita public spending on health increased more than a thirdfrom 1996to 2004. The Government of Costa Rica increased its real per capita health expenditure by an average of 3.9 percent per year from 1996 to 2004, for a total of 36 percent for the entire period. Official data from MIDEPLANshow per capita health expenditures of $44,662, up from $40,505 in 2000 and (Figure 7.7). 174The study use 2001 numbers, the last year when it was possibleto obtain complete data. 141 Figure 7.7: RealPublic Per Capita HealthExpenditure, Costa Rica 1996-2004 20,000 ~ 19% 1997 1998 1999 2000 2001 2002 2003 2004 Year Note: in constant 1996colones Source: SANIGEST2006 from MIDEPLAN web page 7.29 The CCSS accounts for the majority of health resources, and has increased its role over time. It spent more than 82 percent of the health budget in 2001, up from 74 percent in 1992 (Figure 7.8). The increase reflects both health sector reforms - specifically the fact that CCSS assumed almost all of responsibility for providing first-level care - and a faster growth rate than inother institutions. Figure 7.8: Composition of Public Spending on Health, Costa Rica 1992 and 2004 11 lOO~/O 1.4% 6.2% 9.3% 80% 9.1Yo 60% 40% 1 74% 20% 1 0% 7 1992 2001 CCSS mMOH AYA INS i U C R Source: Pan-American Health Organization 2003 7.30 Incontrast with the Ministry of Education, the CCSS' spending on wages and salary is small. CCSS allocates 53 percent of its budget to wages and salaries, compared to 93 percent by the Ministry of Education. In2004, the CCSS spent almost 30 percent of its budget on Goods and Services (Table 7.5). Also, capital expenses at the CCSS were 5 percent in 2004, five times the Ministry of Education's value. The capital expenses figure may be underestimated since 142 some entries, such as medical- and laboratory-instrument purchases, are counted as purchases of goods and services, not as asset purchases. In 2004, the Ministry of Health spent basically all of its budget on current expenses, of which almost two thirds were wages and salaries and one fourth were transfers. Wages and Salaries Purchaseof Goods and Services , Capital Formation Asset Purchase Source: CCSS 7.31 There were significant changes in the population, and thus in funding for insurance from 1989 to 2004 - with important policy implications. Costa Rica has experienced important population changes over the last 15 years. The fertility rate has decreasedfrom 3.4 in 1989 to 2.0 in 2004. And during the same period, life expectancy has increased by almost two years. The impact of these changes can be seen in the sources of revenue for health insurance: the proportion of salaried workers' families decreased 7.5 percentage points from 1989 to 2004, and the proportion of retired and "others" (which includes special forms of retiree) increased 2.1 percentage points each (Table 7.6). The declining weight of the insured retired will have a growing impact on the financial sustainability of the system.'75 The Government of Costa Rica has to prepare for population changes that will inevitably put a burden on the health system, includingthe retirement plan. Source of insurance Year Change 1989 2004 points % Salaried worker 21.6% 24.0% 2.4% 11% Family of salaried worker 55.3% 47.8% -7.5% -14% Self employed 6.5% 6.0% -0.5% -8% Bv the Government 6.3% 8.5% 2.2% 34% Nonfee minimum 3.2% 2.4% -0.8% -24% Retired 2.0% 4.1% 2.1% 106% Family of Retired 3.1% 3.1% 0.0% -1% Others 2.0% 4.1% 2.1% 102% 100% 100% 0% '100percentgovernment subsidized Source: World Bank staff calculations with the EHPM Costa Rica has a "pay as you go" pension system. 143 EFFICIENCY 7.32 To measureefficiency in the health sector, health indicators were compared with total per capita PPP health spending in the same 19 Latin American countries described in figure 7.6 above. Significant relationships between expenditure and the indicators were found in 15 out of 18 indicators for which data on Costa Rica were a~ailab1e.I~~ 7.33 Costa Rica consistently performs above the expected health indicator values for its total per capita PPP health spending. High levels of total health spending in Costa Rica are accompanied by very favorable health indicators. Indeed, in 14 out of 15 comparisons, Costa Rica performs above expectations, and in the only indicator where it does not, percentage of children under 5 years-old who are under weight, the expected value (4.5 percent) i s only slightly better than the observed value (5 percent) (Table 7.7 andAnnex 5). Compared to other individual LatinAmerican countries, only two -Chile and Cuba -performbetter than Costa R i ~ a . ' ~ ~ Table 7.7: Latin American Countries relationship Between Total Health Expenditures (PPP) and Health Indicators, and Costa RicanPerformance, 2002 S-curveY = e**(bO + (bl/x)) or ln(Y) =bO (bl/x). * Actual + observedandfitted values graphs are in Annex 5 CostaRicanvalue was 5.0 percentcomparedto an expectedvalue of 4.5 percent. Source: World Bank calculationswith the UNDP2005 HumanDevelopmentReport, from World Health Organization, 2005 (Marchhealthexpenditure) 176The same exercise was carried out tracking the 18 indicators and government spending as a percentage of GDP. None of the indicators was found to be significantly related to the level of spending (at p15 percent). For total health expenditures (public and private) as a percentage of GDP, only three of the 18 indicators were significantly related,to the level of spending. 177Chile and Cuba have better indicator values than expected (given their level of per capita PPPhealth expending) in all 15 indicators. 144 Health and Poverty INDICATORS AND OUTCOME 7.34 Affiliation to a health insurance plan in Costa Rica i s high and has not significantly changed over the years; nevertheless, important disparities across income groups can be found. The EHPM shows affiliation rates to health insurance (public and private) around 82 percent since 1989, with very little change over the years (Figure 7.9). But with affiliation rates for the poorest quintile below 75 percent (compared to 88 percent for the upper quintile), the Government of Costa Rica has a long way to go to reach truly equitable health insurance affiliation. The biggest disparity on health insurance affiliation involved Nicaraguan immigrant household^,'^^ where more than one third of individuals did not have healthinsurance by 2004. Figure 7.9: Populationwith Insurance, CostaRica 1989-2004 100% 8S2 90% - - - - a 4 - a fn S 0 0 `- 80% S '5 5 .-5 70% = cm \ / ________ /\ x P 60% n0 50% I I , 1989 1994 2000 2001 2002 2003 2004 ~-Awage +lst. Quintile -0-5th. Quintjle ++> 20% Nicaraguan Bod Source: World Bank staff calculations with the EHPM 7.35 Differences in affiliation rates seem to have little impact on access to health services. Since the public health system in Costa Rica is universal by law, no treatment can be refused to anybody, regardless of insurance status. In practice, affiliation increases access to information and to some types and quality of services. Nevertheless, doctor visits do not seem to be affected by affiliation status. Indeed, the 2001 EHPMshow very similar rates of doctor visits between the poorest quintile (41 percent) and the richest quintile (46 percent) (Table 7.8). Very little variation was found between regions, and urban and rural zones, and even less variation was found between nationals and immigrants from Ni~aragua.'~'Among people who visited doctors, the average number of visits was practically the same (2.8 visits) for all groups. Use of public establishments decreases as income increases, from 94 percent for the first quintile to 53 percent for the fifth quintile. Among public establishments, the poor favor EBAIS. ~~ 17*Defined as households with 20 percent or more members born in Nicaragua. 17'Important differences were found between males (37 percent visited a doctor) and females (50 percent visited a doctor), but there is no reason to believe poverty had any influence on the difference. 145 Table 7.8: loctor \ itsfrom January to June 2006, Costa Rica Number Public establishment 1 Private establishment Doctor of visits EBAIS CCSS CCSS INS TOTAL Clinic Other TOTAL Clinic Hospital Rx (a) (b) 1 41% 2.8 40% 29% 25% 0% 94% 4% 2% 6% QU 2 44% 2.8 37% 33% 21% 1% 92% 6% 2% 8% INT 3 42% 2.1 31% 34% 21% 0% 86% 11% 3% 14% ILE 45% 2.7 23% 32% 23% 1% 79% 17% 4% 21% 5 46% 2.1 12% 22% 19% 1% 53% 40% 6% 47% ZO Urban 43% 2.8 19% 34% 22% 1% 75% 21% 4% 25% NE Rural 44% 2.1 41% 25% 21% 1% 87% 10% 2% 13% R Central 44% 2.1 23% 33% 19% 1% 75% 21% 4% 25% E Chorotega 44% 3.0 42% 27% 20% 0% 90% 9% 1% 10% G CentralPacific 44% 2.8 30% 40% 21% 1% 91% 7% 2% 9% I Brunca 46% 2.9 27% 14% 50% 0% 91% 7% 2% 9% Atlantic Huetar 46% 2.6 48% 22% 17% 1% 88% 9% 2% 12% North Huetar 37% 2.6 34% 28% 25% 0% 87% 12% 1% 13% SE Male 37% 2.9 26% 29% 23% 1% 79% 16% 5% 21% Female 50% 2.7 30% 30% 21% 0% 82% 16% 2% 18% MI <20%Nic.born 44% 2.9 28% 30% 22% 1% 80% 17% 3% 20% GR* 2 2O%Nic. born 37% 2.8 37% 31% 20% 1% 88% 8% 4% 12% TOTAL 44% 2.8 28% 30% 22% 1% 80% 16% 3% 20% (a) Private clinic, hosDitalor doctc Source: World Bank staff calculationswith the 2001 EHPM 7.36 There is very little unsatisfied demand for doctor visits due to economic reasons with almost no difference among income groups. Among people who needed a doctor, 13 percent did not visit one; values were only slightly higher for lower-income groups and migrant households (Table 7.9). Strictly economic reasons (lack of money), were mentioned only in one percent of the population (ranging from 0 percent for the higher income groups to 2 percent for the lower income quintile. Even including "other" reasons as economic motivations for not going to a doctor only brings the unsatisfied demand to 9 percent for the poorest quintile. Went to a Didnot go to a doctor and neede one doctor Self Medicated Lack of Money Others2 1 TOTAL 84% 6% 2% 7% 16% 87% 5% 1% 6% 13% QUINTIL 86% 7% 1% 6% 14% 87% 6% 0% 7% 13% 89% 5% 0% 5% 11% ~ 2 0 %Nic. born 87% 6% 1% 6% 13% >20% Nic.born 84% 5% 3% 8% 16% TOTAL * Includes 'Only 87% I 6% 1% 6% I 13% people who neededa doctor are included in this table. lack of space, lost appointment, lack of time and others. Source: World Bank staff calculations with the 2001 EHPM 146 relationship was found between poverty and hospital access was found. There are only small differences among income or geographic groups in the percentage of QUI hospitalized persons and the number and lengh of hospitalizations. People in the lowest quintile, rural zone or from Nicaraguan migrant households have slightly higher percentage of hospitalized members, more hospital stays and higher use of public hospitals. As income increases, the use of 7.38 There i s also no clear relationship between rate of vaccinations for children under three years old and income level or for any other variable. Sometimes, lowe- income groups and children from Nicaraguan immigrant households actually show better vaccination rates (for tuberculosis first dose, and Hepatitis B third dose, for instance). Indeed, vaccination rates reported in the 2002 EHPM don't vary much on the basis of income groups, zone of residency, gender or migration status (Table 7.11). Germanmeasles. Source: World Bank staff calculations with the 2002 EHPM 147 7.39 There is no relationship between poverty at the county level and low birth weight. Poverty county levels (based on the existence of two unsatisfied basic needs) do not show any relationship with low birth weight (Figure 7.10). This pattern could be a result of prenatal care and other programs designed to improve maternal health that spread benefits among all the country's regions without regard to people's socioeconomic situation. . Figure 7.10: Poverty and Low Birth Weight at the County Level, Costa Rica 2003 La Cmz + + + 84 - lutrtina Talamanca m I 4 + t r , I I I i * * 10 6 & 30 +Osa 40+ 50 Q Los Chiles * ' 4 - G Q f h J PovertyIncidence ource: SANIGEST2006 7.40 Three supply indicators were used to relate poverty and health at the regional level: number of physicians, hospital beds and nurses per inhabitant. In all three cases, the health supply indicators improved as poverty increased (Figure 7.1l).'*' The primary conclusion is that the system tends to distribute more resources in regions with the highest incidence of poverty. Although exceptions exist for some indicators, the evidence suggests that there i s a favorable distribution of supplies for poor regions, showing the progressive nature of the health system in CostaRica. With the exceptionof hospitalbedsfor the North Huetarregionand nursesfor the Chorotegaregion. 148 Figure 7.11: Supply Indicatorsand Household Poverty at the RegionalLevel, Costa Rica 2003 7.11.a: PublicPhysicians per 1,000 Inhabitants and HouseholdPoverty Level 2 - 1.8 - + Central Paciflc 1.6 - 1.4 - + Chorotega 1.2 - + North Huetar + Brunca 1 -+ Atlantic Huetar 070 10% 0.8 -20% 30% 40% + Central 0.6 - 0.4 - 0.2 - 0 - 1.60 - 1.40 - - Central bcific - - 1.20 - ! 1.00 - 0.80 - Atlantic Huetar Chorotega , ,1 0% 15% ,a60 - 570 10% 20% - 25% 30% 35% 40% Central- 0.40 North Huetar Poverty Incidence 7.11.c: Public Nurses per 1,000 Inhabitants and Household Poverty Level - Central Pacific * North Huetar e Brunca 0.6 10% 30% 40% Central 0.2 O .0 l j Chorotega Poverty Incidence Source: SANIGEST 2006 149 REDISTRIBUTIVENATUREOFTHE HEALTHSYSTEM INCOSTA RICA 7.41 A guidingprinciplefor health policy inCosta Rica is to get healthsubsidies to the people who need them the most. An analysis of the distribution of public spending between different population groups suggests it has come to favor lower-income population groups. 7.42 Overall public health spending i s progressive. In 2001, the distribution of public spending on health by income quintile reflected a progressive redistribution.The poorest quintile, which receives 5 percent of the national income, received almost 30 percent of health public spending benefits, while the richest quintile, which receives 48 percent of the national income, received only 11percent of healthpublic spending benefits (Figure 7.12). Figure 7.12: Distributionof Public Spending on Health by Income Quintile, Costa Rica 2001 .-2 *5 30 a. 20 10 0 Quintile I Quintile I1 Quintile 111 Quintile IV Quintile V Quintk % Public Health Expenditures --t% National Income Source: SANIGEST2006 7.43 The progressive nature of public spending on health is stronger in the poorest regions of the country. The Central regionhas the lowest poverty headcountrate in 2001 (16.9 percent), and it is the least progressive in health expenditures: almost 20 percent of health expenditures are allocated on each quintile (Figure 7.13). On the other hand, the Brunca Region has the highest poverty headcount in 2001 (39.0 percent) and the more progressive health spending, with more than 40 percent of health expenditures allocated to the poorest quintile and aroundjust percent to the richest quintile. 150 r I National Central Chorotega Central Pacific Brunca Atlantic Huetar North Huetar 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Qulntlle 1 Quintlle 2 0 Qurntile 3 0 Qulntlle 4 r Qulntlle 5 Source: SANIGEST2006 7.44 Health is the most progressive social sector in the country. Compared to housing, education, social security culture and entertainment and income, health i s more progressive at all levels. Indeed, 2000 Lorenz curves show the progressive nature of public health spending for the entire population. In contrast, public education i s poverty neutral, and government housing program i s regressive for lower income households and progressive for upper income households (Figure 7.14). Figure 7.14: SocialExpense and Income Lorenz Curves, Costa Rica 2000 5 10 a a En EIC) 72! 1Eo; fa 100 Accumulated% o f famdies Source: Trejos 2002 151 OTHER HEALTHRELATEDISSUES 7.45 Health and Housing Services. Health related housing services such as piped water, water systems, sanitation (sewer or septic tank) and garbage collection have improved considerably from 1989 to 2004 (see chapter 2). By 2004, water access and sanitation were almost universal for all urban areas, andreached around 90 percent of rural dwellers (Table 7.12). Important gaps remained for some income groups (only 83 percent of lower quintile persons had adequate sanitation services, for instance) and regions (only three quarters of persons in the North Huetar region received water from water systems), though. And 96 percent of the urban wastewater collected in sanitary sewage systems was being disposed of in rivers without treatment. In addition, the Ministry of Health shows that in the last several years there has been an increase in the incidence of water-related diseases. Diarrhea, which i s associated with water resources, occupies second place as the cause of death in the group of diseases for which reporting i s mandatory. Table 7.12: Health and Housing Services by Quintile, Region, Zone and Immigration, Costa Rici201 4 I Piped IWater Sewer or GarbagePick GROUPS CLASSIFICATION Water System Septic Tank up 1 96% 86% 83% 59% 2 99% 93% 93% 76% QUINTIL 3 100% 94% 97% 85% 4 100% 97% 98% 90% 5 100% 98% 99% 95% Central 100% 98% 97% 94% Chorotega 93% 86% 82% 59% REGION Central Pacific 99% 95% 95% 74% Brunca 98% 84% 89% 50% Atlantic Huetar 97% 86% 91% 61% NorthHuetar 96% 76% 84% 41% ZONE I urban 100% 100% 98% 98% Rural 97% 85% 89% 57% 95% 82% 82% 74% OfficialAYA statistics 98% TOTAL 99% 94% 94% 81% Source: World Bank staffcalculations with the 2004 EHPM 152 Box 7.1 Immigration and Social Security According to the 2000 population census, around 225,000 Nicaraguanimmigrants live in Costa Rica, by far the largest group of foreigners in the country. These immigrants are mostly between 15 and 44 years of age, and tend to live in the urban areas. One-third of the Nicaraguans work in the service sector (maids, guards, sales) in both the rural and urban areas, and almost 8 percent work in construction (State of the Nation web page). The immigrant population's impact on the health sector has been of interest to several researchers (Rosero; 2002 and 2003, Castillo; 2002, PHO; 2003; Leal, et al., 2004). The studies agree that Nicaraguans do not have a negative impact on the services (State of the Nation, 2004). Barquero (2005) says that around 76 percent of the foreigners living in Costa Rica have some sort of insurance that gives them the right to use these services. Nicaraguans use health services in a lesser proportion than the percentage they represent in the country (they account for 6 percent hospitalizations and 4 percent external and emergency consultations). Foreigners have gone from accounting for 1.3 percent of external consultations in 1992 to 4.3 percent in 1997. There i s evidence o f greater use of medical services at the first attention level. The health sector reform that created the EBAIS also covers the Nicaraguan population. Immigrants use the EBAIS at a rate almost 30 percent higher than Costa Ricans. Some 42 percent of Nicaraguan men and 38 percent of Nicaraguan women are members of the Social Security system. In comparison, 78 percent of the Costa Rican men and 85 percent of the women participate in the system. Insurance by the state covers 9 percent of Costa Ricans and 7 percent of Nicaraguans. Some 75 percent of the Costa Ricans and 53 percent of Nicaraguans have access to the insurance system as direct contributors or as dependents (of a direct contributor). Castillo (2002) states that Nicaraguan immigrants generally have insurance, but it takes time to acquire it after they enter the country. Just one-third of the immigrants who have been in the country less than one year are insured, while 70 percent of those who have been in the country 5 to ten years are approximately insured. Health and the Millelnium Development Goals 7.46 There are three millenniumdevelopment goals (MDGs) related to health: to reduce infant mortality by two thirds; to improve maternal health; and to combat HIV/AIDS, malaria and other diseases. A recent study by a health policy consulting firm, Sanigest (2006), estimated the likelihood of CostaRica achieving the health MDGs. The results presentedhere are basedmainly on the Sanigestreport. 7.47 To evaluate the Costa Rican MDGs, past performance i s used as the best predictor. Simple linear projections are estimated, and the assumptions behind them are qualified. The objective is to provide an idea of how probable it i s that the country will achieve the MDGs, and to determine what the necessary conditions are for it to do so. Shocks to the system or simple policy changes can alter the conclusions (ineither direction). GOAL 1: REDUCE CHILD MORTALITY AND INCREASE MEASLES VACCINATION 7.48 Goal one has three indicators: reduce the death rate for children under five years old to 5.9 per thousand, to reduce the rate for children under one to 9.0 per thousand and to achieve measles vaccination rates of 95 percent for one year olds. 7.49 Regressionmodels predict that the goal for under five mortality i s achievable by 2015 (or at least very nearly) (Figure 7.15.a). The main determinant of under-five mortality rates i s under- one mortality rates. The difference between death rates for children under five and those under 153 one has been decreasing from around three per thousand at the beginning of the 1990s to around one per thousand during the last five years. Therefore, achieving the under-five mortality rate objective i s conditional to further reductions in the under-one mortality rate. The reduction for under-one year-olds death rate stated in the MDGs has been already achieved. A measles vaccination rate has fluctuated between 85 and 100 percent between 1995 and 2003. The challenge is to keep it at the 95 percent level (or higher), and it i s uncertain whether that can be achieved. Figure 7.15: MDGHealth Goal One: Reduce Infant Mortality and MeaslesUn-Vaccinated Figure 7.15.a: Under FiveYear-old Figure 7.15.b: Under One Year-old Mortality Mortality (000) (000) 20%, 1 15% - 15% 10% - 10% MDG2015 = 5.9 per thousand 5% 5% - 0% J , , , ,, , , ,, , ,, , , , , , , , , ,, , ,,1 -I 1 -0bsemd -Adjusted -I 1 -Observed -Adjusted i Figure 7.15.c: Percentageof One Year-olds Vaccinated against Measles M f f i 2015 = 95% 85% 80% - 1995 1996 1997 1998 1999 2000 2001 2002 2003 Source: Sanigest 2006 GOAL2: IMPROVEMATERNALHEALTH 7.50 Goal two has two indicators: reduction of maternal mortality by three-fourths, and have 97 percent of birthsassistedby qualified personnel. 7.51 The maternal mortality goal (to a rate of 4.8 deaths per 100,000 live births) will be very difficult, if not impossible, to achieve for Costa Rica. Maternal mortality recording substantially improved in 1998 with the creation of the National Maternal Mortality Evaluation System. Underreporting prior to 1997 has been estimated as high as 40 percent (UNDP, 2004). Data show an unexplained increase of 53 percent in maternal mortality from 1998 to 2000, right at the time 154 the new recording system was implemented (Figure 7.16.a). Adjustments to correct for the change in the reporting system require too many assumptions that would render the exercise invalid. As to births assistedby qualified personnel, the MDGgoal was achieved in 1993 and has kept improving since then (Figure7.16.b). Figure 7.16: MDGHealth Goal Two: Reduce Maternal Health and Increase BirthAttention Figure 7.16.a: MaternalMortality(100,000 Figure 7.16.b: BirthsAttended by Qualified live births) Personal 30 28 25 20 2015 = 97% 15 95% 10 1 5 MDG2015 = 4.8 per 100,000live births 90% Source: Sanigest 2006 GOAL 3: COMBAT HIV/AIDS. MALARIA AND OTHER DISEASE 7.52 Goal three has six indicators: stop and beginto reverse the spread of AIDS, condom use (no specific goal), and stop or reduce the incidence of malaria, tuberculosis, dengue and mortality due to tuberculosis. 7.53 The AIDS incidence rate in Costa Rica has fluctuated substantially between 1990 and 2004, reaching as high as 7.1 per 100,000 inhabitants and as low as 2.1 per 100,000 inhabitants. In one year (1996 to 1997) it doubled, while in another (2000 to 2001) it fell by half (Figure 7.17). Such fluctuations raise questions about the accuracy of the Figure 7.17: AIDS Incidence (per 100,000), Costa data. Still, the outlook for meeting the Rica 1990-2004 AIDS goal has improved as a result of recent progress achieved in several areas: the foundation of a national HIV/AIDS care assessment council, implementation of responsible sex campaigns that include encouraging the use of condoms, a decision to focus efforts on groups identified as being vulnerable, the creation and implementationof the General Law on AIDS, and the fact that it i s mandatory to report cases of the disease. As for condom use, the data available are relatively scarce and fundamentally Source: Sanigest 2006 based on the results of fertility surveys. Among Central American countries, Costa Rica has the highest percentage of people 155 using condoms (11 percent). Although above the regional percentage, i s still low, and suggests the needto step up efforts to encouragecouples to use condoms. 7.54 Out of the other four indicators, only tuberculosis mortality rate has declined steadily over time (Figure 7.18.c). Government reports of significant increases in malaria in 2005 are not official yet, and are not reflected in Figure 7.18.a. Tuberculosis and dengue incidence have significantly increased, and no sign of reductioncan be identified (Figures 7.18.b and 7.18.d). Figure 7.18: Incidence of Malaria, TB and Dengue, and Mortality from TB, Costa Rica, 1990-2004 Figure7.18.a: Malaria Incidence Figure7.18.b: TuberculosisIncidence (per 100,000) (per 100,000) 2500i 2202 19901991 199219931 994199519961 9971 998199920002001 200220032004 Year YFX Figure7.18.c: TuberculosisMortality Rate (1nn.nom Figure7.18.d: DengueIncidence (100,000) 3 1l\ tD0 t 4)2 6 .. ' * E L 'OD t :50 t 2 200 t a 2SQ.C >DO t I 1 5 0 L 100 t 50 0 1991 1592 1993 1994 1995 1996 1997 1998 1999 ZOaO2501 2DOz Z 3 3 2004 0 0 Year "Ear -MJ "*""Femsls *Total Source: Sanigest 2006 7.55 In summary, of the 11 indicators considered, the country has: (i)positive prospects in four (mortality of under-one year olds, mortality of under-five year olds, the percentage of births assisted by qualified personnel and mortality rates from tuberculosis; (ii) moderate and uncertain prospects in two others (the measles vaccination rate for under-one year olds and the AIDS incidence rate; (iii) poor prospects in four indicators (maternal mortality, and incidence of tuberculosis, malaria and dengue); and (iv) no evaluation for on (condom use). Table 7.13 summarizes the results for all the indicators previously discussed. 156 Maternal mortality recording substantially improved in 1998 with the creation o f the National Maternal Mortality Evaluation System. Underreporting prior to 1997 has been estimated as high as 40 percent (UNDP, 2004). During a press conference, the Ministry of Healthreported significant increases for 2005 not officially recorded yet. Source: SANIGEST 2006 Conclusion 7.56 The government of Costa Rica's historically strong commitment to the health system has achieved excellent results in the well-being of its people. Life expectancy levels are not only above all Latin American countries but also similar to other developed nations. Resources dedicated to the health system also are high for the region. And when outcome indicators are compared to spendinglevels, Costa Rica consistently performs above expectations. 7.57 The health sector is the most progressive of all social sectors in the country. The universal nature of public health services has reached most of the poor in the country. Public health supply indicators such as the number of physicians, hospital beds and nurses per inhabitants are higher in the poorest regions of the country. Also, some vaccination rates (Diphtheria Pertussis, tetanus and tuberculosis) are the same or even better for the poor. 7.58 Basic sanitary conditions and preventive care have improved the effectiveness of the health system. Average access to basic sanitary conditions i s not only high but also covers most of the poor: 96 percent of the lowest quintile population has access to piped water, 93 percent of the second income quintile population obtains water through a water system and has a sewer or septic tank. As for preventive care, use of primary service providers, the Integral Basic Health Care Team (EBAIS), i s significantly higher inthe lower income quintiles. 7.59 Contrary to popular wisdom, the presence of Nicaraguans does not have a significant negative impact on health services, especially with a few years after they enter the country. For example, 75 percent of Costa Ricans have insurance coverage by direct affiliation (that is, by contributing to the system) or through their family, while immigrants from Nicaragua who have been in the country for more than five years are 70 percent insured. Insurance by the state (non contributing system) i s the same for Costa Rica-born residences as for Nicaraguan immigrants (7 percent). 157 7.60 The health reform process started inthe 1990shas had a positive effect on the system and has contributed to good outcomes. The creation of small local clinics providing primary health services increased access by the poor to the health system and promoted preventive medicine. The gains from investments made during the reform process were highly positive, with a minimumestimated rate of returnof 33 percent. 7.61 And yet, despite the positive results attained by the government, meaningful deficiencies can also be found. 7.62 There are still structural problems in the Costa Rican health system, for instance. By 2003, the decline ininsurance affiliation rates, the increaseduse of the health services by children under one year-old (for increased preventive care) and the elderly (due to population dynamics, including longer life expectancy), and the increase in the cost of services had reduced operational surpluses (the difference between expenses and contributions) to almost zero, leaving few resources to finance non contributing - that is, the poorest - members. Given population dynamics and the increaseddemand for more expensive services, the financial vulnerability of the system will increaseunless actions are taken to solve the problem. The government of Costa Rica needs to address these issues as soon as possible before conditions worsen and radical measures become necessary. Actuarial studies of health insurance and appropriate policies to address the problem also are necessary to improve the financial outlook of the Costa Rican Social Security Institute (CCSS). Other areas for potential improvement are to create more transparency in insurance provided by the state, to reduce evasion due to under-contributingand under-reporting and to increasecontribution coverage of non-salaried workers 7.63 There still are differences in the system by poverty level and immigration status. Affiliation rates to the CCSS and doctor visits are significantly lower among lower-income groups and Nicaraguan immigrants, vaccination rates (for diphtheria and tetanus; measles, rubella and mumps) are lower in rural areas and for Nicaraguan immigrants (diphtheria and tetanus), and there are significantly fewer public nurses per habitants in the Chorotega region (the second poorest region in the country). Increasing affiliation rates to the CCSS will reduce several inequalities. For Nicaraguan immigrants, reduction of requirements (such as having legal immigration status) will increase affiliation rates. Information and promotional campaigns also can improve overall affiliation to the CCSS. 7.64 Important gaps still remain in access to basic sanitary conditions based on income, area of residency and immigration status. Access to a water system remains lower for the lowest quintile (86 percent) rural households (85 percent) and the Huetar region (76 percent). Only 83 percent of the poorest quintile households and 82 percent of immigrants from Nicaragua have sewer or septic tanks in their homes. Also, fewer low-income households have garbage collection. The government of Costa Rica should improve access to these basic services, especially by the poor. 7.65 Significant increases in the incidence of malaria, dengue (reaching more than 31,000 persons in 2005 compared to 7,000 in 2004) and tuberculosis (more than double the 1990's rate), as well as decreases in measles and poliomyelitis vaccination rates are evidence of some of the weaknesses of the health system. Even though no proof has been found that the incidence of these diseases is higher among the poor, it is known that the ability to deal with health problems decreases with income level. In other words, even if this problem presents itself equally for the whole population, the impact will be greater on the poor. Increasing the Ministry of Health's (MH) budget allocation for preventive measures as well as public vaccination campaigns is necessaryto avoid spikes in the incidence of these diseases. 158 7.66 The CCSS and the MH currently use two different regional systems. That decreases efficiency in coordination of sector policies and leads to unnecessarily duplication of staff, financial resources and materials. In addition, matching different information sources that include poverty conditions and indicators from INEC i s not always possible. 7.67 Finally, there i s also a fundamental issue regarding the quality of information on health. Information to evaluate specific health programs i s rarely available, especially with respect to the impact of government programs in general and on the poor. This hinders evidence-based policy analysis and the government's ability to assess the strengths and weaknesses of specific programs. Better assessments would allow for adjustments or new designs to improve the impact, efficiency and targeting of the public health programs. A monitoring and evaluation systemneeds to be put in place. Individual programs should have their own set of indicators and the tools to collect and analyzethem. 159 8. SOCIAL PROTECTION 8.1 Social protection i s traditionally defined as a range of measures adopted by governments to provide basic income security to their people, to protect them from unanticipated shocks, and to protect and develop the human capital of society's poorest, most vulnerable members of society - by strengthening their ability to thrive in the labor market, and assuring basic service access to those outside the reach of traditional government and privately provided services. Social protection is typically delivered via a variety of social insurance and social assistance programs (Box 8.1). Box 8.1: Social Protection- ManagingSocial Risk, Promoting Long-term Growth and Development People in developing countries face a range of risks. Some risks, such as economic recessions, harvest losses, natural disasters and wars, affect whole societies or large groups. Others, such as the illness of family members, loss of household breadwinners' jobs, and crime, may only affect individual households. Social protection consists of a range of public interventions that support society's most poor and vulnerable members and help individuals, families, and communities manage risks better. Social protection is delivered by a variety of informal, market-based, and public mechanisms - including regulation, government financing, and the direct provision of services. Intended to augment, not replace, family, community, and market-based risk management mechanisms, such interventions complement national economic policies and support strategies for poverty reduction and human development. Public social protection measures often are categorized into two main groups: social insurance and social assistance. Social insurance includes an array of insurance-type mechanisms, such as pensions and unemployment and health insurance that are intended to cushion the impact of shocks affecting income, employment or health, and thereby prevent families from falling into poverty. Such programs sometimes are categorized under the broad heading of "social security." Social assistance includes a variety of safety- net programs, such as workfare, assistanceto the disabled and indigent and cash transfers, all of which help individuals and families deal with temporary or chronic poverty, strengthen their capacities and achieve higher standards of living. Labor market policies and institutions often play a central role in social protection by influencing the nature and extent of the risks workers face, by providing a framework for social protection programs such as pensions, unemployment insurance and workfare, and by providing opportunities for skills or technical training so that people can find more remunerative employment. Well-designed social-protection programs can compensate for missing insurance markets or other private risk-mitigation instruments. In addition, certain types o f safety net tools-such as "conditional transfers," where payments are made contingent upon family investments in children's health or schooling-both provide short-term income support and strengthen longer-term investments in children's education, health, well-being and productivity. Source: "Volatility, Risk, and Innovation: Social Protection in Latin America and the Caribbean, Spectrum, World Bank, Fall2003. 8.2 Costa Rica has a relatively well developed set of social protection programs, in both the areas of social insurance and social assistance. As noted in Chapter 5, it has among the highest social security (health insurance and pension) coverage in the Latin America region (Figure 5.2), Moreover, the Government of Costa Rica operates a wide array of social assistance programs that provide support to poor and vulnerable groups and strengthen their human capital.'" 181Trejos (2006) argues that because of the high level of Government of Costa Rica involvement in the country's safety net, non-governmental organizations (NGOs) play only a minor role, mostly with the aim 160 8.3 This chapter summarizesthe key risks faced by the poor inCosta Rica. It reviews the main programs that make up Costa Rica's social safety net, analyzes public social protection spending, examines who benefits from social protection programs, and assesses whether existing programs adequately address the key risks faced by the poor. The chapter outlines several challenges for making social protection a more effective force for poverty reduction in Costa Rica, and concludes by highlighting key areas for public action to strengthen social protection's role in reducing poverty. 8.4 Several key messages emerge from the chapter. While Costa Rica's social protection system i s relatively well developed, the country still faces important challenges in supporting and protectingthe poor. Although social security coverage i s high by regional standards, a significant proportion of the poor still fall outside the reach of the safety net, and thus fail to benefit from the range of existing programs. In addition, several risk areas require greater programmatic focus - particularly in the area of assuring human capital formation among young children from poor families. The analysis also indicates that there is considerable scope for improving the effectiveness of social protection spending; indeed, over 70 percent of public spending on social protection goes to Costa Rica's contributory pension schemes that largely benefit the non-poor. Strengthening the poverty-reducingimpact of social protection would benefit from measures that would: (i) strengthenthe strategic focus of the system to address better the key risks faced by the poor; (ii) rationalize and consolidate social assistance programs, consistent with the enhanced strategic focus; and (iii) improve targeting of programs and, thus, coverage of the poor. The analysis shows such measures could contribute to important gains, even within existing budget envelopes. Principalrisksfaced by thepoor in CostaRica 8.5 As discussed in previous chapters, the poor in Costa Rica face a number of significant risks related to low human-capital development, low or precarious earnings, unemployment, and loss of income in old age, among others. People face different risks depending on their stage on the lifecycle. The probability of being poor i s also linked to the lifecycle of the household. When a household is being formed the risk of being poor increases upon the arrival of children. The risk of poverty declines once younger household members are old enough to enter the workforce and contribute to family earnings. When children beginto leave the household, the risk of poverty begins to rise again (Barquero & Trejos, 2004). This means that, on average, children and the elderly face relatively high level of vulnerability (all other thingsbeingequal). 8.6 The key risks faced by poor Costa Rican, according to the different stages of their lifecycle can be summarized as follows.'82 Childrenunder age 6 8.7 Children under 6 years of age are over-represented among the poor, as 32 percent of them lived in poor households in 2004. Thanks to the strength of the Costa Rican health services, 95 percent of births occur in hospital facilities, and vaccination coverage i s around 90 percent of complementing government actions rather than trying to substitute or fill service the gaps. Indeed, NGOsusually act underthe direction and financing of government institutions inCosta Rica. For this reason, this chapter focuses mostly on government social protection programs. 182This section draws heavily on the analysis presented in earlier chapters of this report as well as on new analysis of the 2004 EHPM survey by Trejos (2006). For details on Trejos's household-level analysis, see Annex 6 of this report, as well as Trejos (2006). 161 country wide (Programa Estado de la Nacidn, 2004). But there i s some evidence that monitoring of growth and development in the first year of life is weak, at least among the poor. Only about half of children living in poor households report visits to outpatient centers (e.g., the primary health care centers, EBAIS), even though it is considered essential for young children to have at least one medical check-up per year. And about 10 percent of poor children do not visit a health center for a check up despite reporting a medical need (Trejos 2006). 8.8 While data suggest that nutrition is not a high risk among poor children in Costa Rica (UNICEF2005), children under six in poor households face an elevated risk of stuntedphysical and intellectualdevelopment due to low levels of access to early childhood development and care services. Indeed, only about 25 percent of children from poor households attend Ministry of Health childcare centers or Ministry of Education preschool education prograrns.lg3Given the importance of early childhood development to cognitive achievement, educational attainment, and productivity later in life, this represents a risk with potentially long-term adverse impacts on poor children. Children age 6 to 12 (primary school age) 8.9 As with children under age 6, about one thirdof primary school age children belonged to poor households in 2004. They have ample access to primary education, however: 96 percent of the extremely poor and 97 percent of all poor children attend school. But only about one-third of the children from poor householdsfinishprimary school at the customary age, compared to the 54 percent of children from non-poor households. And roughly 30 percent of poor children never complete primary school. Adolescents age 13 to 18 (secondary school age) 8.10 Adolescents are also over-represented in poor households, although not to as great an extent as younger children: about 28 percent of adolescents live in poor households. About 70 percent of poor children in this age category are still in school, but many are not in the proper grade for their age. Indeed, around 27 percent of children age 13 to 18 in extreme poverty are still working to complete primary schooling (Trejos 2006). Moreover, secondary school enrollment is low. Only one-third of those in the lowest quintile attend secondary school. And of poor adolescents who attend, very few graduate: only 12 percent of those in extreme poverty and 19 percent of the total poor in2004 pass this milestone. 8.11 Data indicate a high level of youth inactivity among the poor as well. Approximately 17 percent of extremely poor youth neither study nor work. Among poor youth who are out of school and in the labor market, the unemployment rate is high, around 30 percent. Moreover, those who do work tend to be concentrated in low income, low productivity sectors; nearly half (47 percent) work in agriculture, while another 39 percent work in informal non-agricultural activities. Among the extremely poor, 77 percent work in the agricultural sector; while another 19 percent work in the informal sector. This suggests very direct links between failure to persist inthe education systemandlow income, precarious employment thereafter. Young adults, age 19to 24 8.12 Young adults often are part of "young" households that are just being formed (that is, they are couples who don't yet have children) or older households in which the young adults are working. Both types of household are associated with relatively lower levels of poverty. In 2004, lg3EHPMdata do notcapturethose who are takencare of by the CommunityHouseholdsProgram (Hogares Comunitarios)runby IMAS,the lnstituto Mixto de Ayuda Social. Nonetheless,it i s estimates that not more than 5 percentof poor childrenparticipatein this program (Trejos2006). 162 the average risk of poverty for this age group was 16 percent. Few young adults who are poor are still in the education system - only 19 percent of the extremely poor and 27 percent of the total poor - and when they are, the majority are trying to complete secondary education (53 percent of young adults attend high school, and 30 percent attend open education programs). Nevertheless, there i s a large group of poor, young adults who neither study nor work - about 28 percent of poor and 31 percent of extremely poor young adults. This group does face an elevated risk of poverty (27 percent). It also faces an elevated risk of encountering trouble with the judicial system (Trejos 2006). 8.13 Young adults in poverty are at high risk of having low levels of human capital, and although they have higher levels of labor market participation than adolescents, they also experience high levels of unemployment (30 percent, on average). When poor young adults are employed, they tend to work in low-productivity jobs in agriculture or the informal sector, or in the formal sector but under precarious job conditions. As a result, only 24 percent of poor adults have direct access to health insurance, and only 22 percent contributeto pension plans. Prime working-age adults, age 25 to 64 8.14 Prime working age adults face a below-average risk of poverty (19 percent compared to the national average of 24 percent) -although the risk is several points higher among women than among men (Trejos 2006). The data indicate a strong relationship between low human capital and poverty risk among prime working-age adults. Secondary school graduation as a key threshold; those who have not completed secondary education have a higher than average risk of poverty (for their age cohort), while those who have completed secondary education have a significantly lower-than-average riskof poverty. 8.15 Risk of poverty is also strongly associated with status in the labor market. Among the employed, the incidence of poverty is only 14 percent, while the risk of poverty jumps to 44 percent for the unemployed. The risk of unemployment i s considerably lower for prime-working age adults than for adolescents or young adults. Nonetheless, among prime working age adults who are poor, the unemployment rate i s 12 percent; the rate i s 18 percent among extremely poor prime working age adults. As with younger workers, poor prime working age adults who work tend to be concentrated in agriculture and in sectors without social security coverage, although their coverageratesareroughly twice those found among poor adult workers who are young. Older adults. age 65 and older 8.16 When time a person reaches age 65, the risk of being poor increases again. Indeed, about 30 percent of the elderly population count themselves among the poor (Trejos 2006). At the same time, this age group generally has high levels of health insurance; fewer than 10 percent of poor elderly people lack formal insurance. The greatest risk faced by the elderly stems from loss of income when they are unable to work and have no family members to support them. About 41 percent of the extremely poor and 31 percent of all poor among the elderly are not part of a pension plan (either a contributory or non-contributory, social assistance plan). A much higher proportion of the elderly would be uncovered was it not for Costa Rica's non-contributory pension program: 43-45 percent of the elderly poor report receiving social assistance pensions. At the same time, data suggest that the level of benefits from non-contributory pensions is not sufficient to lift these recipients out of poverty. Together the data on pension coverage and poverty suggest there is scope to increase both the coverage and the benefit levels of social assistance pensions to combat poverty among the elderly more effectively. 163 Risks associatedwith headinghouseholds 8.17 In general, the risks associatedwith heading households are similar to those outlined for prime-working age adults. But a clear gender dimension can be seen. Consistent with evidence presented in earlier chapters, females who head households face a relatively higher risk of poverty than their male counterparts (28 percent as opposed 20 percent). Poor females who head households also face greater barriers to remunerative employment in the labor market: they tend to have less education, higher levels of unemployment, a greater chance of holding part-time work or working inthe informal sector, and greater child-care responsibilities. Social Protection Programs 8.18 As noted above, social protection programs are public interventions aimed at helping people, families and communities to deal with risk more effectively as well as to assure a basic level of well-being to those in poverty or extreme poverty (Box 8.1). Generically, social protection interventions are categorized into to main groups: social insurance and social assistance. InCosta Rica these generic categories have very specific meanings. 8.19 Social insurance programs generally refer to Social Security (Seguro Social), which covers risks of illness, maternity, accidents, disability, old age and death via a "tripartite contribution" from workers, employers and the Costa Rican government. The state-run Social Security Agency (Caja Costarricense de Seguro Social, CCSS) manages this service. Because health insurance was discussed in the previous chapter, this chapter focuses primarily on old age disability and death insurance offered through the CCSS, as well as on smaller programs offered to different workers' associations (e.g., the Rkgimen de Pensiones y Jubilaciones del Magisterio Nacional or pension program for professional educators), referred to below as the contributory pension (or definedcontribution pension) programs. 8.20 Social assistance encompasses a host of programs aimed at supporting particularly vulnerable groups-including people who are not able to access mainstream (universal) social programs, those who havenot benefited from economic growth, and those affected by unexpected economic downturns or shocks. These programs - there are more than 45 of them - are executed by a number of government institutions, including CCSS, and they all make up what has come to be known in Costa Rica as Social Promotion and Assistance Programs. Within Costa Rica, social assistance programs are differentiated into two main categories: (i) promotion programs, which strengthen people's human capital, and (ii) assistance programs, organized under the "Social Protection Network," which supports people's basic levels of consumption. These programs are financed with revenue that comes directly from income and other taxes, as well as from other resourcesraised directly by the executing agencies. A large proportion of the funds are channeled to these programs through the National Development and Family Allocations Fund (Fondo de Desarrollo Nacional y Asignaciones Familiares) or FODESAF (Figure 8.1). 164 Figure 8.1: Organization of the Social Protection Sector in Costa Rica National Payroll Budget Sources -FODESAF Social Protection Programs t Social Security I Illnessand - Maternity 8.21 In the sections that follow, contributory pension programs and promotion and assistance programs are discussed inturn. Contributory pension urograms 8.22 Costa Rica's main contributory program aimed at addressing risks associated with old age i s the Disability, Old Age and Death Fund (Rkgimen de Znvalidez, Vejez y Muerte, RIVM), managedby the CCSS. There are also at least 14 other systems financed through national budget. The largest of these is the National Teachers Pension and Retirement Fund (Rkgimen de Pensiones y Jubilaciones del Mugisterio Nacional), which accounts for close to 70 percent of all expenses for the additional programs.184 Although reforms adopted during the 1990's closed access to most of the pension systems paid out of the national budget, an accumulated contingent liability will continue to impingeon public finances over the coming decades.'85The reforms also created a "second pillar" comprising a program of individual contributions (capitalization) with centralized collection by the CCSS and decentralized administration, either by private or public institutions; this second pillar i s restricted to the wage-earning workers. The reforms also aimed at strengthening a "third pillar" comprised of voluntary saving systems and not associated with one's status inthe labor market. 8.23 Although the country has a relatively modem labor market in which wage earnings predominate (close to 70 percent of all employed people earn wages), about 40 percent of the currently active workers fall outside the systems because they do not contribute. As noted above, lS4Some of these systems are actually non-contributory, like grace and war pensions, such as the Mag611 Prizes, or pensions given to ex-presidents of the Republic. However, these are only marginal within the total scheme off pension programs, so they are considered collectively here. There is also an additional system for the employees of the Judicial System, which is administered by thejudiciary and is managed through the national budget. See World Bank 2003. 165 the non-contributors include many workers from poor household-a problem that poses a great challenge for the future, particularly as the size of Costa Rica's elderly population increases. 8.24 The Disability, Old Age, and Death Fund, RIVM, i s financed inpart via direct "tripartite" contributions from workers, employers, and the Government of Costa Rica. The other programs are financed through a combination of workers' contributions and general revenues of Costa Rica's central government. The RIVM, together with a pension program for workers in the Judicial System, are the only programs that have a reserve fund that gives them a measure of financial sustainability . Several factors are creating challenges to the RIVM system's financial sustainability, however, including demographic changes (aging of the Costa Rican population), labor market changes (the growth of informal sector employment), program parametersthat have not been updated to reflect new economic realities, efficiency problems associated with investing the reserve fund, and a lack of proper controls on evasion. It i s estimated that, in the absence of recent reforms, contributions to the RIVM system would have been insufficient to cover payments by the year 2011. Counting interest on the reserve fund, contributionswould have been unable to cover payments by 2022, and it would have been necessary to start drawing from the reserve fund, which would have been depleted by 2028 (Martinez, 2005). Incontrast, the pension systems financed under the national budget were not created with reserve funds, so that the difference between workers' contributions and pay-outs to pensioners has to be covered by the national budget. For the group of these (other) plans, in 1999, workers' contributions represented 19 percent of pension payments, and by 2004, the role of these contributions had decreasedto 13 percent of payments. This suggests these systems will represent a growing financial burden on the national budget over time. 8.25 Recent reforms to the RIVM have sought to strengthen the financial sustainability of the system, to improve the distribution of benefits across the Costa Rican population, and to decrease evasion.'86 These reforms, approved in 2005, increase minimum contributions and raise the contribution rate 0.5 percent every five years until it reaches 10.5 percent (Trejos 2006). To make the system more progressive and to reduce evasion, the methodfor calculatingthe reference salary was changed and it was decided to increase the rate in inverse proportion to a worker's salary level. Early retirement options also were created to reduce fraudulent disability claims. With the recent reforms to the RIVM, it is estimated that the first critical date for the system - when funds from contributions are insufficient - has been postponed until 2041 (Martinez, 2005).la'In contrast, however, recent reforms to National Teachers' Pension and Retirement Fund have actually added to financial stress on the system by reestablishing benefits that had be abolished in 1995 (Trejos 2006).'88 8.26 As will be discussed further below, because these contributory pension programs are associated with work in the formal sector, their benefits go disproportionately to people at the higher end of the income distribution. This is the case for all such programs, but even more so for those programs supported by the national budget. Moreover, Trejos (2004) finds that the 186As i s discussed below, incorporating low-income and informal sector workers into the system, as obliged by the the Workers' ProtectionLaw (Ley de Proteccidn a1 Trabajador) remains a challenge. 187To improve access to pensions among workers who do not currently contribute, efforts have also been made to ensure universal coverage, as legally established under the Workers' ProtectionLaw (Ley de Proteccidn a1 Trabajador). These universal coverage provisions have met with opposition, however, and further progress in this area is pending. A Constitutional Court ruling reestablished benefits prescribed under an earlier law for workers who had contributed to the system for fewer than 20 years at the time of the 1995 reform. A subsequentdirective from the Procuraduria Generalde la Repu'blicatried to turn back this ruling. This created a political impasse that required further legal action to resolve. 166 distribution of their benefits has become more unequal over time. Key challenges Costa Rica faces thus include extending coverage to low-income and non-wage workers, improving the equity of benefits across participants, and ensuring long-term financial sustainability of these programs. These latter two challenges are particularly important for systems financed out of the national budget. Social promotion and assistanceprograms 8.27 Social promotion and assistance programs are implemented by a number of Costa Rican institutions, either as self-standing programs, or as components of larger programs or sets of activities. That makes it difficult to establish an exhaustive inventory of programs or to determine total spending on promotion and assistance. It i s important to highlight the effort the Costa Rican government has made through the Technical Secretariat of the Social Council (Secretaria Te`cnicadel Consejo Social) in recent years to make an inventory of promotion and assistance programs. The initiative was supposed to generate quarterly information on benefiaries and resources of (most of) the existing programs, disaggregated by region (Consejo Social 2005). This inventory is still incomplete, however. Moreover, information on beneficiaries has not yet been harmonized (across individuals, households, and projects), and the data on financing is limited primarily to direct transfer programs. Despite existing data limitations, this section attempts to review Costa Rica's main protection and assistanceprograms. 8.28 Promotion programs in Costa Rica are designed to support the development of people's human capital, with special emphasis on children. They also aim at improving people's habitat, and in doing so, strengthening the environment for human capital development. To a lesser extent, they also support increasedeconomic productivity among prime working age adults. 8.29 Several programs focus on strengthening human capital for children under the age of six. The principal program, known as the Centers for Education and Nutrition and the Child Centers for Integrated Care (CEN-CINAI, Centros de Educacion y Nutricion y Centros Infantiles de Atencion Integral), is a system of childcare centers managed by the Ministry of Health.'89 For school-age children and adolescents, programs tend to concentrate on offering incentives, either in cash or kind, to promote school enrollment and retention. These programs focus on providing complementary meals, scholarships, transportation and support for the purchase of school uniforms and supplies. Most of these programs are implemented by the Ministry of Public Education. The largest, in terms of resources and beneficiaries, is the School Feeding Program. The Instituto Mixto de Ayuda Social, IMAS, also runs several small programs to promote human capital development. It has two programs that targets women -one focused on women inpoverty (Creciendo Juntas), the other focused on adolescent mothers (Contruyendo Oportunidades). Although these programs have significant social welfare components, they are included as part.of promotion programs because they have training components and because they are expected to provide the basis for a future conditional transfer system consolidated under IMAS to promote greater educational achievement among children in poor households (Trejos 2006). 8.30 Several programs address habitat improvement. The largest i s the Family Housing Subsidy (Bono Familiar de la Vivienda),which i s managedby Banco Hipotecario de la Vivienda (BANHVI). Several MAS programs seek to strengthen land access andtenure, improve housing quality, and develop community infrastructure. Related initiatives include programs for development of rural aqueducts, implemented by the Znstituto Costarricense de Acueductos y 18'There are also programs for prenatal care and for growth and for development assistanceto minors, which are executed by the CCSS through the primary care facilities, EBAIS. Because these are fundamentally part of the broader public health system, they are not analyzed in depth here. 167 Alcantarillados (ICAA) and programs for basic environment health, implementedby the Ministry of Health through the Technical Council for the Social Medical Care (Consejo Te'cnico de Atencidn Me'dico, CTAMS). 8.3 1 Finally, promotional programs include initiatives that help prime working-age adults increase their productivity and incomes. This program area i s relatively undeveloped and operates with limited resources. The National Program for the Micro and Small Enterprises (Programa Nacional de Micro y PequeAa Empresa, PRONAMYPE) i s run by the Ministry of Labor and Social Security (Ministerio de Trabajo y Seguridad Social, MTSS), but has been frozen for the past three years. IMAS runs training programs, provides subsidies for productive projects, and has established a trust fund with the state commercial bank to provide collateral for credit extended to micro-entrepreneurs. The Institute for Agrarian Development (Znstituto de Desarrollo Agrario, IDA) also works with the National Council for Production (Consejo Nacional de Produccidn, CNP) to develop farmer settlements and basic infrastructure, and to strengthenagricultural credit. 8.32 Social Protection Network focus on several types of interventions: (i) compensation for people who have lost their jobs or experienced temporary emergencies; (ii) assistance programs for people in extreme and chronic poverty; and (iii)programs to counter socio-economic exclusion by reaffirming the rights of the minorities, assisting vulnerable groups and protecting those who are the subject of discrimination. 8.33 As with productivity-oriented programs, compensatory programs are not very well developed; nor is there a mechanism for counter-cyclical financing and implementation, an important feature of programs that respond to aggregate economic shocks or downturns. Nonetheless, there are several programs, including a job-creation program under the Ministry of Labor, the subsidies for emergencies under MAS, and the National Emergency Commission (ComisidnNacional de Emergencias, CNE), which is an inter-institutional organization in charge of responding to natural disasters. 8.34 Assistance programs are relatively well-developed, and include a non-contributory (social assistance) pension program. The non-contributory pension program i s administered by the CCSS and is focused on providing support to elderly poor who are not covered by contributory pension systems. IMAS administers a program of family subsidies, and the Social Protection Board of San Jose (Junta de Proteccidn Social de Sun Jose', JPSSJ) as well as IMAS provide financial assistance to private (often non-governmental) social welfare institutions (Box 8.2). Box 8.2: The Role of Znstituto Mixto de Ayuda Social, IMAS IMAS was created in 1971 with the mission to fight poverty. While IMAS only i s responsible for about 11 percent of total spending on social promotion and assistance, its importance stems from its flexibility in handling resources and in its potential to be a catalyzing and organizing force for social promotion and assistanceprograms. IMAS undertakes social investment in several different program areas. One key program area involves direct transfers to families that are poor or have suffered as a result of natural disasters (emergency response). The most important such program offers unconditional transfers for up to six months to help disadvantaged families meet their needsfor food, health (by paying for specialized consultations and medicines, for instance), public transportation, clothing, domestic equipment, funerary costs, rent, debt for public services and housing, or care for disabled people. I I 168 A second key program area involves support for young and school-aged children. This includes a small :hild care program called community homes (hogares comunitarios) for the preschool population, as well as growing programs that help poor families send their children to primary and secondary school. This program focuses on providing a monthly aid (scholarships) conditioned on school attendance. IMAS also implements programs aimed at poor women with children - one focused on women in poverty (Creciendo Juntas),the other focused on adolescent mothers (Contruyendo Oportunidades). Although these programs have significant social welfare components, they are based on the conditional transfer concept, which zombines short-term income support with support for human capital development for children and mothers. Production support programs have also played a role - albeit a relatively minor one - in IMAS's portfolio. Suchprograms include the Productive Ideas program, which provides subsidies to help heads of households or organized poor groups to establish or strengthening small-scale productive activities. IMAS also has provided scholarships and other forms of financial aid for people who attend the courses or take internship positions offered by business enterprises. It recently has established an alliance with Banco Nacional de Costa Rica (BNCR) and its subsidiary, BICSA, to establish a trust fund in the state bank to support credit lines for micro-entrepreneurs. IMAS also implements programs aimed at improving family's habitats, and provides financial assistance to private social welfare institutions. An important feature of IMAS's institutional set-upis its program beneficiary identification instrument, the SIP0 database(Sistemade Informacidn sobre la Poblacidn Objetivo). This system was developed based on the Chilean experience of identifying and selecting poor households to participate in social protection interventions. The system uses a proxy means test that looks at specified family traits, such as their housing characteristics, their income and the characteristics of the members of the household. This system faces several challenges, including making it more representative of poverty conditions in Costa Rica. It also has to update its information periodically. Nonetheless, it has the potential to support improved poverty targeting not only of IMAS programs, but of the range of Costa Rica's social promotion and assistanceprograms. The wide range of IMAS's activities creates the potential for duplication of efforts across programs that have similar objectives and constituencies, but are run by different ministries or agencies (such as the Ministry of Health and the Ministry of Public Education). This challenge points to a need, inherent in a multi-sectoral social protection system, for strategic.vision and technical coordination across programs and agencies. Source: Treios (2006) 8.35 Finally, several programs combat exclusion or the vulnerability of specific groups. Among the institutions involved in such programs are: the Patronato Nacional de la Znfancia (PANI), which assists children considered high social risks; the National Women Institute (Instituto Nacional de las Mujeres, INAMU); the National Council for Youth Policy (Consejo Nacional de Poli'tica para la Persona Joven, CNPPJ); the National Commission for Indigenous Affairs (ComisidnNacional de Asuntos Indigenas, CONAI). Several institutions also provide care to the disabled, such as the National Council for Special Rehabilitation and Education (Consejo Nacional de Rehabilitacidn y Educacidn Especial, CNREE), support rehabilitation (Patronato Nacional de Rehabilitacidn, or PANARE; the blind (Patronato Nacional de Ciegos or PANACI; and senior citizens (the National Council for the Older Adult - Consejo Nacional para la Persona Adulta Mayor -or CONAPAM). Financing.of social promotion and assistance programs 8.36 Financing for Costa Rica's promotion and assistance programs comes primarily via the Social Development and Family Allocation Fund (Fondo de Desarrollo Social y Asignaciones 169 Familiares,), FODESAF. Indeed, between 1999 and 2004, 80 percent of financing for promotion and assistance programs passed through FODESAF, on average. Promotion and assistance programs are also financed to a lesser degree with their own resources, from contributions from the JPSSJ and other institutions, and from the nationalbudget. 8.37 FODESAFwas established in 1974 - about a decade before the social emergency or the social investment funds, which were created in the region to mitigate some of the effects caused by the stabilization and adjustment policies in the 1 9 8 0 ~ . ' ~Its mandate is to address the structural determinants of poverty. The Fundhas several key features that dictate its contribution to social protection. FODESAF resources come from dedicated tax revenues - an arrangement designed to prevent resource volatility and promote financial sustainability (so that FODESAF would not have to compete with line ministries for the resources of the universal programs). Its targeted taxes are a 5 percent payroll tax on business and institution (collected by the CCSS) and a 20 percent of the sales tax (collected by the Finance Mini~try).'~'Despite the goal of reducing volatility, this type of financing i s pro-cyclical, so that in recessions, when the compensatory programs should be activated, the Fundactually hasfewer resourcesat its disposal. 8.38 Analysis of available financial information on FODESAF indicates, moreover, that its real income has declined by about 20 percent since 1999 (Trejos 2006). In large part, this is because FODESAF'srevenues from the sales tax have been below legislatedthresholds. In2004, for example, the Finance Ministrytransferred only 0.6 percent of the sales tax. This reduction in revenues from the sales tax has meant that FODESAF has become increasingly dependent on resources coming from payroll taxes. These have remained constant in real terms. By 2004 they represented 82 percent of all the FODESAF's incoming resources. Because of its declining resources, the Fund's role has been declining over time as well. While FODESAF has financed, on average, 80 percent of Costa Rica's promotion and assistance programs, the proportion has declined from 93 percent in 1999 to 69 percent in 2004 (Trejos 2006). This, too, raises questions about the long-termstability of financing of social assistance in Costa Rica. 8.39 FODESAF helps to finance about 47 programs implemented by 20 state institutions. The resources dispersed by the Fund have generally followed the same trend as its revenues - an accumulated contraction of 20 percent since 1999. Between 1999 and 2004, four programs received approximately 72 percent of FODESAF's resources on average (Trejos 2004): the Family Housing Subsidies (Bono Familiar de la Vivienda), the Non-contributory (social assistance) Pension Program (Rkgimen de Pensiones no Contributivas), the School Feeding Program (Comedores Escolares) and the CEN-CINAI Childcare Centers (Centros Znfantiles). Of these programs, only the Childcare Centers Program does not have a specific allocation level established by law. In addition to these four programs, two institutions receive another 13 percent of FODESAF resources: the antipoverty agency MAS and PANI, which assists children at social risk. Like the Family Housing Subsidy, the Non-contributory Pension Program and the School Feeding Program, PANI's fundinglevels are legislated. 8.40 This fact that a large proportion of FODESAF'sresourceallocations are designatedunder law means that the Fund has very little flexibility to allocate across programs or adjust to The Fundi s administered by the General Office of Family Allocation and Social Development (Direccidn General de Asignaciones Familiares y Desarrollo Social, DESAF), an annexed entity of the Ministry of Labor. 19'The law that createdthe FODESAF increased the rate of the sales tax from 5% to 8% inorder to finance it with these three percentagepoints, that is, 37.5% of the collected tax. As the sales tax has continued to increase over time, it was later agreed to fix the income to be 20% of the collected tax. 170 emerging priorities. Moreover, its flexibility has decreased over time. The share of its resources prescribed by legislation increased from 74 percent in 1999 to 82 percent in 2004 (Trejos 2004). This creates a number of problems. First, allocations are "permanent" and do not allow for periodic revisions; legal reform would be required to modify specific allocations. Second, allocations are not based on the performance of the implementing agencies or program impact. Finally, although FODESAF is charged with addressing the structural determinants of poverty, this mode of financing allocations has led it to finance activities that do not benefit its priority target group, the poor ("poblacio'n de escasos recursos"19*). FODESAF financed the Costa Rican Sports Institute (Znstituto Costarricense del Deporte, ICODER), which i s not oriented towards serving the poor, for example. Financing of the Family Housing Subsidy also results in explicit support for non-poor households, since the law that supports this allocation defines the program's target group as families in thejirst eighth deciles of the income distribution. Public spending on socialprotectionprograms 8.41 Analyzing public spending on social protection is difficult, given the large number of programs being implemented, the multi-sectoral nature of the activities, and a lack of consistent spending data and accounting practices across the various agencies and programs. Despite the data challenges, Trejos (2006) assembled and analyzed a detailed database of spending on social protection - both on contributory pensions and social promotion and assistanceprograms - from 1999through 2004. The key findings of his analysis are presented below. The evolution of social protectionspending 8.42 Estimates of public spending on social protection in Costa Rica for the 1999 to 2004 period are presented in Table 8.1.193 Includingthe contributory pension payments, it averaged about 5.5 percent of the GDP in these years, the equivalent of 22.8 percent of total central government spending and 30.3 percent of the public social sector spending. Even though social protection spending in 2004 had increased relative to 1999, levels have been relative flat since 2000. Indeed, the apparent increase in 2004 reflects, in part, that several previous years, and in particular 1999, had been years of contracting social sector (and social protection) spending (Trejos 2004). Overall, social protection spending increased at an average annual rate of 2.6 percent between 1999 and 2004 (1.2 per year percent between 2000 and 2004); per capita spending grew at an annual rate of 0.5 percent per year from 1999 to 2004 (although it contracted 0.8 percent per year iftracked from 2000 to 2004). 8.43 Spending associated with the contributory pensions made up 72 percent of total social protection spending, on average, over the 1999-2004 period. It has tended to increase its share over time. Between 1999 and 2004, spending on these programs has grown at an average annual rate of 3 percent (2.7 percent when 1999 i s not included), while spending per person grew 1. percent per year (0.7 percent per year from 2000). While spending on contributory pension programs represented3.4 percent of GDP in 1999, its share has increased to 4.3 percent over the last three years. Nonetheless, its weight in total public spending has been more or less stable, at around 16 percent on average over the 1999-2004 period, the equivalent of 22 percent of the public social spending, on average. 192See Trejos (2006). 193Expenditurespresentedhere do not includethose relatedto healthprograms, such as the primary care (EBAIS)or the enrollment in healthinsurance(Seguro de Sulud). Analysis of healthsector spending is presentedinChapter 7. 171 8.44 Spending for social promotion and assistance shows greater volatility than spending on contributory pensions, and also displays a decreasing trend. While it represented 28 percent of total social protection spending over the period, its share declined from 30.7 percent in 2000 to 26.5 percent in 2004. Real spending grew at a rate of 1.5 percent per year, on average, between 1999 and 2004 (although it declined 2.5 percent per year in real terms if 1999 is excluded). Inper capita terms, the average annual contraction i s about 4.4 percent. Despite this contraction, spending on social promotion and assistance as a percent of GDP tended to be stable at around 1.5 percent starting in 2004. This translates to roughly 6.3 percent of the total public spending and 8.4 percent of the public social sector spending. The composition of social protection spending 8.45 Table 8.2 presents the composition of spending on the principal social protection programs in 1999 and 2004. In the case of contributory pensions, the portion of the programs paid directly under the national budget grew quickly, at a real annual rate of 4.3 percent, on average. In 2004, payments to pension programs out of the national budget represented nearly two-thirds of all spending on contributory pensions, and almost half of the all social protection spending. 8.46 Spending on social promotion and assistance programs i s more varied. Two-thirds of spendingon these programs goes to social promotion, especially those focused on human capital development and habitat improvement - and, in particular three of Costa Rica's principal programs in the field (Childcare Centers, School Feeding Programs, and the Family Housing Subsidy). The productive assistance programs receive much smaller allocation: around 10 percent of all spending on promotion and assistance programs and less than 3 percent of all social protection spending. The share going to programs directed to poor women with children i s even lower. 172 Table 8.1: The Evolutionof the Public Spending on Social Protectionin Costa Rica, 1999 2004 - 2000 2001 2002 2003 2004 Average Total Expenditureon Social Protection Total Expenditure ' 246,052.8 266,941.O 266,484.3 273,998.9 271,294.3 279,912.7 267,447.3 Index2000 = 100 92.2 100.0 99.8 102.6 101.6 104.9 100.2 8.5 -0.1 2.8 -1.o 3.2 2.6 Percentage of the GDP 4.8 5.4 5.6 5.9 5.7 5.8 5.5 Percentage of the public expenditure 21.3 23.5 22.4 22.8 22.7 23.9 22.8 Percentage of the public social expenditure 30.1 31.4 30.2 29.9 29.3 30.8 30.3 Expenditure for inhabitant* 64,115.1 68,004.7 66,483.7 66,998.8 65,062.8 65,885.4 661091.7 Index2000 = 100 94.3 100.0 97.8 98.5 95.7 96.9 97.2 Rateof annual change 6.1 -2.2 0.7 -2.9 1.3 0.5 Expenditureon ContributoryPensions Total Expenditure ' 177,214.0 185,020.1 191,829.8 198,218.9 200,808.8 205,731.7 193,137.2 Index 2000 = 100 95.8 100.0 103.7 107.1 108.5 111.2 104.4 Rate of annual change 4.4 3.7 3.3 1.3 2.5 3.0 Percentage of the GDP 3.4 3.8 4.1 4.3 4.2 4.3 4.0 Percentage of the public expenditure 15.3 16.3 16.2 16.5 16.8 17.6 16.4 Percentage of the public social expenditure 21.7 21.7 21.8 21.6 21.7 22.6 21.8 Percentage of the expenditure in Social Protection 72.0 69.3 72.0 72.3 74.0 73.5 72.2 Expenditure by inhabitant 46,177.4 47,134.9 47,858.6 48,468.9 48,158.7 48,424.8 47,703.9 Index 2000 = 100 98.0 100.0 101.5 102.8 102.2 102.7 101.2 Rate of annual change 2.1 1.5 1.3 -0.6 0.6 1.o Expenditureon Social Promotion and Assistance Total Expenditure ' 68,838.8 81,920.9 74,654.6 75,780.0 70,485.6 74,181.0 74,310.2 Index2000 = 100 84.0 100.0 91.1 92.5 86.0 90.6 90.7 Rate of annual change 19.0 -8.9 1.5 -7.0 5.2 1.5 Percentage of the GDP 1.3 1.7 1.6 1.6 1.5 1.5 1.5 Percentage of the public expenditure 5.9 7.2 6.3 6.3 5.9 6.3 6.3 Percentage of the public social expenditure 8.4 9.6 8.5 8.3 7.6 8.2 8.4 Percentage of the expenditure in Social Protection 28.0 30.7 28.0 27.7 26.0 26.5 27.8 Expenditure by inhabitant 17,937.6 20,869.8 18,625.2 18,529.9 16,904.1 17,460.6 18,387.9 Index2000 = 100 86.0 100.0 89.2 88.8 81.o 83.7 88.1 Rate of annual change 16.3 -10.8 -0.5 -8.8 3.3 -0.5 I/Inmillionsofcolonesof2000,withtheimplicitgeneralgovernmentconsumptionpriceindexasdeflator. 2/In colones of 2000, with the implicit generalgovernment consumption priceindex as deflator. Source: Trejos 2006. 173 Table 8.2: The Compositionof Public Spendingon Social E otectionin osta Rica, 1999 rd 2004 1 Total Expenditureon 2omDositionbv Public Spending in Annual Rate of Programsand institutions Social Protection Sub-category ' 2004 as percentof Change Per 1999 2004 1999 2004 GNP EP EPS Total capita rota1Expenditureon Social Protection ' 246,052.8 279,912.7 5.8 23.9 30.8 2.6 0.5 :ontributory Pensions 72.0 73.5 100.0 100.0 4.3 17.6 22.6 3.0 1.o Insurancefor Disability, OldAge and Death (CCSS) 28.6 26.4 39.7 35.9 1.5 6.3 8.1 I.o -1.o Pensionspaidby the National Budget 43.5 47.1 60.3 64.1 2.7 11.3 14.5 4.3 2.2 Social Promotionand Assistance 28.0 26.5 100.0 100.0 1.5 6.3 8.2 1.5 -0.5 PromotionPrograms 18.3 17.3 65.6 65.4 1.o 4.1 5.3 I.5 -0.6 Support to the Formationof Human Capital 8.2 7.6 29.3 28.5 0.4 1.8 2.3 I.o -1.l Careto pre-schoolchildrenand their mothers 3.1 2.4 11.1 9.0 0.1 0.6 0.7 -2.8 -4.7 Child Care Centers (Ministryof Health) 2.7 2.1 9.7 8.0 0.1 0.5 0.7 -2.3 -4.3 Opportunitiesfor providingchild care (IMAS) 0.4 0.2 1.4 0.9 0.0 0.1 0.1 -6.0 -7.9 incentivesto access and to stay in school 4.6 5.0 16.4 18.8 0.3 I.2 1.5 4.2 2.1 School food program (ComedoresEscolares,MEP) 3.0 2.5 10.8 9.6 0.1 0.6 0.8 -0.9 -2.9 School voucher (Bono escolar, MEP) 0.3 0.2 1.2 0.6 0.0 0.0 0.1 -11.4 -13.2 Scholarships (Becas,FONABE) 0.3 0.8 0.9 3.1 0.0 0.2 0.2 29.5 26.9 Transportationof students (MEP) 0.9 1.1 3.1 4.1 0.1 0.3 0.3 7.1 5.0 Superemonos: Access to education (IMAS) 0.1 0.4 0.4 1.4 0.0 0.1 0.1 32.1 29.4 Supportto poor women 0.5 0.2 1.7 0.8 0.0 0.0 0.1 -13.8 -15.6 CreciendoJuntas (AFT = AMSP: IMAS) 0.4 0.1 1.3 0.3 0.0 0.0 0.0 -26.2 -27.7 ConstruyendoOportunidades(IMAS) 0.1 0.1 0.4 0.5 0.0 0.0 0.0 5.8 3.7 Improvementof the Habitat 7.2 7.1 25.9 26.7 0.4 1.7 2.2 2.2 0.1 FamilyHousingVoucher (BANVHI) 3.8 6.4 13.4 24.3 0.4 1.5 2.0 14.3 12.0 Improvementof housing(IMAS) 0.4 0.3 1.5 1.o 0.0 0.1 0.1 -6.7 -8.5 Titled propertieswith Access to public services (IMAS) 0.4 0.0 1.4 0.0 0.0 0.0 0.0 -60.1 -60.9 Communalinfrastructure(IMAS) 0.3 0.0 1.2 0.2 0.0 0.0 0.0 -30.5 -31.9 RuralAqueducts (ICAA) 0.6 0.1 2.1 0.5 0.0 0.0 0.0 -24.7 -26.2 Basic hygieneservices (CTAMS-MS) 1.7 0.2 6.1 0.7 0.0 0.0 0 1 -34.0 -35.3 Support to Production 2.9 2.7 10.4 10.2 0.2 0.6 0.8 I.2 -0.9 PRONAMYPE (MTSS) 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ProductiveIdeas (IMAS) 0.1 0.2 0.2 0.7 0.0 0.0 0.1 25.0 22.5 Training(IMAS) 0.2 0.1 0.6 0.3 0.0 0.0 0.0 -10.5 -12.3 IMAS-BNCR-BICSATrust Fund 0.0 0.1 0.0 0.3 0.0 0.0 0.0 Farmer Settlements(IDA) 2.1 1.4 7.3 5.5 0.1 0.3 0.4 -4.3 -6.3 Agriculture Reestructuring(CNP) 0.6 0.9 2.2 3.4 0.1 0.2 0.3 10.5 8.3 SocialProtectionNetwork 9.6 9.2 34.4 34.6 0.5 2.2 2.8 1.6 -0.5 Compensatoryprograms 0.4 0.4 1.6 1.5 0.0 0.1 0.1 -0.1 -2.1 Job Creation Program(MTSS) 0.1 0.0 0.3 0.2 0.0 0.0 0.0 -12.2 -13.9 Assistance for Natural Disasters(CNE) 0.3 0.3 1.o 1.2 0.0 0.1 0.1 4.1 2.0 Emergency Subsidies (IMAS) 0.1 0.0 0.2 0.1 0.0 0.0 0.0 -7.6 -9.4 Assistance Programs 6.9 6.3 24.5 23.6 0.4 1.5 1.9 0.7 -1.3 Non-contributorypensions(CCSS) 5.0 4.5 17.8 17.1 0.3 1.1 1.4 0.8 -1.3 Aid for socially disadvantagedfamilies (IMAS) 0.7 1.2 2.5 4.4 0.1 0.3 0.4 13.9 11.6 Aid for SocialWelfare Institutions(JPSSJ) 0.7 0.5 2.4 1.9 0.0 0.1 0.2 -3.2 -5.2 Strengtheningof the SocialWelfare Services (IMAS) 0.5 0.1 1.9 0.2 0.0 0.0 0.0 -34.7 -36.0 Programs againstexclusion 2.3 2.5 8.3 9.5 0.1 0.6 0.8 4.2 2.1 Childrenunder social risk (PANI) 1.7 1.5 6.0 5.7 0.1 0.4 0.5 0.4 -1.6 Women (INAMU) 0.3 0.2 1.o 0.9 0.0 0.1 0.1 0.9 -1.1 Youth (CNPPJ) 0.1 0.1 0.3 0.2 0.0 0.0 0.0 -6.9 -8.7 Indigenous (CONAI) 0.1 0.0 0.2 0.1 0.0 0.0 0.0 -8.8 -10.6 Disabled(CNREE-PANARE-PANACI) 0.2 0.3 0.8 1.1 0.0 0.1 0.1 8.2 6.1 InstitutionalizedElderly (CONAPAM) 0.0 0.4 0.0 1.4 0.0 0.1 0.1 11Total in millions of colones of ZOOO, with the implicit general government consumptionprice index as deflator. Relative structure by program. Source: Trejos 2006. 174 8.47 Spending on promotional programs have been increasing at a real rate of 1.5 percent per year, similar to the average of the social promotion and assistance programs. However, just like other programs, their evolution has been very volatile, and reflects a contraction in real per capita terms starting in 1999. Among promotional programs, ones providing education incentives, such as scholarships (including the IMAS aid) and ones for housing increased in real terms. Programs directed at children under six contracted in real terms even though coverage was relatively low at the beginning of the period. Programs for poor women with children also contracted in real.terms. Finally, production assistance programs grew - although less than average. In real per capita terms, they actually contracted over the period. 8.48 Social Protection Network programs absorb the remaining third of the spending on social promotion and assistance. Within the Social Protection Network, assistance programs have the highest share of spending-nearly 70 percent. The largest assistance program, by far, i s the Non- contributory pension program. This program represents half of all Social Protection Network spending, although its share has contracted slightly over the period. Compensatory programs are small, and they contracted during the period. Programs confronting exclusion are somewhat larger, and collectively expanded, essentially due to increased spending on programs for the disabled. However, these programs are still relatively small in the context of social protection spending (2.5 percent of social protection spending and less than 10 percent of social promotion and assistance spending), although they represent more than 25 percent of Social Protection Network spending. 8.49 Given that spending on contributory pensions respondsto accumulated debt, and that part of the resources allocated to social promotion and assistanceare definedby law, there i s not much scope for reassigning resources based on evaluation of program impact or changing priorities in the fight against poverty. In fact, in 2004, only 8 percent of the resources spent on social protection - equivalent to 30 percent of those spent on social promotion and assistance or 0.5 percent of the GDP - could be feasibly redistributed without legal changes. This high degree of inflexibility highlights the need for reforms in the legal framework controlling resource allocations in social protection - reforms that would improve the government's ability to take actions to improve the efficiency and effectiveness of social protection spending. Characteristics of Social Protection Program Beneficiaries 8.50 Table 8.3 summarizes some basic information about the distribution of the beneficiaries of the primary Social Protection programs based on analysis of the 2003 EHPM data (which had some basic information on access to several key social protectionprograms). The table shows, for example, the distribution of the pension recipients (without, for the moment, taking into consideration the amount of pension benefits received). Consistent with the fact that these programs representbenefits largely associated with formal sector and better paidjobs, the receipt of benefits tends to be concentrated among those with higher incomes. For example, nearly 51 percent of the beneficiaries are found in the 40 percent of families with the highest income levels. The poor have considerably less access to these programs; indeed, only 1.4 percent of the extremely poor receive contributory pension benefits. Consistent with the patterns of formal sector employment, benefits are largely concentrated in urban areas (76 percent) and in the Central Region of the country (75 percent). Interestingly, close to 40 percent of the beneficiaries are people under 65 years of age; this reflects the fact that benefits from programs funded under the national budget can be received at earlier ages, and that loop-holes enable early retirement via 175 claims of disability (Trejos 2006).'94Although women participate less in the labor market, there i s heavy female representation among the beneficiaries (48 percent). This is due not only to receipt of widow pensions, but also, more importantly, to the fact that women make up the majority of beneficiaries under the National Teachers Pension and Retirement Fund. 8.51 The counterpart of these programs is the non-contributory pension scheme (rkgimen de pensiones no contributivas por monto bdsico, RNC), which is the primary assistanceprogram of the Social Protection Network. This program, created with the approval of FODESAF (1973, does a good job of reaching the poorest groups: 56 percent of the beneficiaries belong to the 20 percent poorest households (Table 8.3). Over half of the program beneficiaries live outside of the Central Region (52 percent) and in rural areas; and 60 percent of the beneficiaries are women.'95 In the case of non-contributory pensions, two-thirds of all beneficiaries are in the primary target group (people over 64 years of age), even though the program allows for some beneficiaries to be younger - specifically disabled people, abandonedwidows, orphan minors and indigents from 50 to 64 years of age who are not qualified to work. An important weakness of the non-contributory pension program, however, i s the size of the benefits, which are equivalent to approximately the extreme poverty line (Le., enough for a person to cover the costs of their basic food needs), which provides only a minimumlevel of protection against poverty in old age. '94While two thirds of people receiving pensionsthrough the RIVM plan are over 64 years of age, half of those receiving pensionsunder budget-financedschemes are between 50 and 64 years of age; only 42% are over 64 years old. The peripheryregions account for 36% of the country's population, andthe rural zones 41%. 176 Table 8.3: Distribution of the Beneficiaries of Costa Rica's Main Social Protection Programs, 2003 Non- Indicator Child School School Scholar- Transport. Housing pensions `Ontrib' pensions contrib. Centers Care program food voucher ships of students Vouchers IncomeGroup ' 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 20% Poorest 10.3 56.0 51.3 38.7 61.6 42.0 40.0 25.3 20% next 16.1 21.8 27.5 26.3 22.5 34.8 27.4 27.5 20% intermediate 22.9 15.8 14.1 18.6 8.4 14.1 18.3 23.4 20% next 23.1 4.6 5.6 11.8 6.7 8.9 11.1 18.2 20% Richest 27.6 1.9 1.5 4.6 0.8 0.2 3.2 5.5 PovertyCategory ' 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Extreme Poverty 1.4 15.3 15.2 11.9 25.1 12.7 9.3 5.8 Poor ' 9.0 39.1 28.6 22.7 31.3 24.8 21.9 16.4 No poor 89.6 45.6 56.2 65.4 43.6 62.4 68.7 77.8 Region 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Central 74.8 48.1 50.0 53.8 45.4 55.5 33.7 60.0 Rest of the regions 25.2 51.9 50.0 46.2 54.6 44.5 66.3 40.0 Area 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Urban 76.2 48.5 29.9 45.3 36.6 51.5 9.9 55.0 Rural 23.8 51.5 70.1 54.7 63.4 48.5 90.1 45.0 Age 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Under 6 years of Age 0.0 0.0 76.1 3.2 0.6 0.5 0.0 9.2 6 to 11 years old 0.0 0.3 11.8 71.7 72.5 22.3 0.5 14.4 12to 17years old 0.8 1.2 1.8 24.2 26.9 67.4 89.0 15.6 18to 49 years old 8.5 13.1 9.5 1.o 0.0 9.8 10.5 47.1 50 to 64 years old 30.3 17.6 0.3 0.0 0.0 0.0 0.0 9.4 65 years old and over 60.4 67.8 0.4 0.0 0.0 0.0 0.0 4.3 Gender 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Male 52.5 40.0 43.6 51.O 50.6 45.0 47.7 49.0 Female 47.5 60.0 56.4 49.0 49.4 55.0 52.3 51.O Progressive Distribution Poor 0.52 2.72 1.55 1.09 1.77 1.42 1.18 1.04 40% poorest 0.67 1.98 1.39 1.09 1.41 1.45 1.27 1.18 Geographic Region 0.75 1.55 1.23 1.13 1.34 1.14 1.70 1.11 Zone 0.67 1.45 1.49 1.17 1.36 1.08 2.01 1.10 Gender Women 0.91 1.15 1.14 1.oo 1.01 1.12 1.06 1.01 11Familiesorderedby family incomeper householdmember. Excludeshouseholdswith ignoredor zero income(13,5 percent). 2/Relationship betweenthe beneficiariesof the poorest, vulnerable groupsor of remotezones and the referencepopulations Inthosedomains. The referencepopulationsare for pensions(50 years andover), child care centers(under 6 years of age), schoolfood and school voucher (6 to 11years old), scholarshipsand transportation(12 to 17years old) andhousingvouchers (totalpopulation). The programi s progressiveif its value is above one. Source: Trejos 2006. 177 8.52 The rest of the programs for which information on beneficiaries exists are promotional programs. The data indicate that a highpercentage of the beneficiaries of Childcare Centers of the Ministry of Health (Ministeriode Sulud CEN-CINAI) are poor or near-poor: around 80 percent of the beneficiaries belong to the poorest 40 percent of households (Table 8.3). Half of the beneficiaries live outside the Central Region, and 70 percent live in rural areas. The creation of FODESAF gave this program an additional boost since it provides for additional resources and takes care of served foods, preschool care, food supplies or milk for children up to six years of age; as a result, roughly three-quarters (76 percent) of the beneficiaries are in the zero-to-five age range. The program also provides food support to women who are pregnant or breast-feeding, so close to 10percent of the beneficiaries are inthe 18-to-49 age range; this also explains the greater presenceof women (56 percent) among the users of these services. 8.53 CEN-CINAI Childcare Centers provide daily attention and nutritional support to children, but this service i s only available to a limited number of beneficiaries. Integrated care i s limited to children living within a one kilometer radius of a center, so the availability of benefits depends on the where a center has been built. To increase coverage of the program, this restriction should be loosened. The program has its own selection criteria, which emphasize not only whether a child is from a poor household, but also whether he or she i s malnourished or at risk of malnutrition, or has some level of social risk (for instance, the mother needs to work, but has no one to look after her children). Program coverage is limited, and it would be even lower were it not for the fact that some services - for example, handing out food and milk to families with children at nutritional risk - are performed outside the centers. These "additional" services also help to explain the program's wide outreach to poor groups and regions, since if the attention i s placed on just on the beneficiaries of integrated care, the reach would be lower due to the residence effect. 8.54 The second promotional program for which data are available i s the School Feeding Program (ComedoresEscolures) implemented by the Ministry of Public Education. This program was also given a strong push with the creation of FODESAF. Although the program was originally designed as a universal program, the emphasis is on reaching primary school children, so it has achieved good coverage of children from poor households (72 percent of the beneficiaries are inthe primary school-age range). About 39 percent of the beneficiaries belong to poor households (bottom income quintile) while 65 percent of the beneficiaries belong to the poorest 40 percent of households. Moreover, around 46 percent of beneficiaries live outside the Central Region, while more than half (55 percent) live in rural areas. The primary limitation of the program is the quality (the nutritional value) of the food served; in the attempt to universalize the service, resources per student were set relatively low - a problem accentuatedby the fact that real resource allocations have been declining. Because of this, it will be important to move toward better use of geographic targeting, taking advantage of the poverty maps that have been elaborated from the 2001 census or of the census of child height that has been carried out in many schools. This would allow for more effective use of existing resources from a poverty reduction perspective. 8.55 Table 8.3 also presents data on several smaller Ministry of Public Education programs: the school voucher program (bono escolar) which provides assistance for the purchase of school uniforms and supplies; the student scholarship program, and the transportation program for students in remote areas. The school voucher directed to primary school students is the most effectively targeted: 62 percent of its beneficiaries belongto the poorest 20 percent of households (Table 8.3). The other programs focus on secondary schooling, so they tend to be less targeted to the poor, although their targeting is still progressive, especially when looking at the poorest 40 percent of households. These programs also maintain a pro-rural bias - especially the 178 transportation program, which was designed to focus on rural areas. Interestingly, 55 percent of the beneficiaries of the scholarship program are young women. This probably i s because female students tend to show a greater persistencein secondary schooling. 8.56 The last program for which there i s data i s the Family Housing Subsidy (bonofamiliar para la viviendu).This programprovides subsidies for construction, purchase or repair of houses. Established in 1987 as a subsidized credit, it later became a direct transfer. Its target group includes up to the eighth decile of incomedistribution since beneficiaries were defined as families having income of up to the equivalent of four times the minimumwage of a construction worker. This wide target population, together with weak control of the processes for granting subsidies, makes the Family Housing Subsidy the least well targeted all the programs examined here, and the one that has the lowest proportion of beneficiaries in rural areas after the Contributory Pension schemes. Only 25 percent of its beneficiaries belong to poorest 20 percent of families, and only about 50 percent belong to poorest 40 percent of families.'96 Although the program i s financed primarily from the FODESAF, there are legal restrictions to channeling all of the program's resourcesto the poorest population(Trejos 2006). The distribution of social urotection spending 8.57 Building on Trejos (2004), Trejos (2006) estimates the distribution of the benefits of social protection spending for selected programs. In the case of pensions and scholarships, the distribution of expenditures is not estimated in proportion to the number of beneficiaries, but rather in relation to the resources received, as this better reflects the benefit-incidence of spending. Inthe case of the pensions, a distinction is made between the RIVMplan of the CCSS and the systems that are financed under the national budget (RPN), because there are important differences between the two systems. The results of these estimations are presented in Table 8.4. Itis clear that the overall distribution of social protection spendingis driven by the distribution of the Contributory Pension schemes, which accounted for 74 percent of the total social protection spending in2003. 8.58 In contributory pensions, the distribution of expenditures is more regressive than the distribution of beneficiaries since the amount received i s directly linked to wages earned in the labor market. The distribution of benefits is even more regressive inthe case of pensions financed from the national budget than in the RIVM program. In fact, while 58 percent of total spending on contributory pensions reaches to the 20 percent richest, this percentage is 70 percent inbudget- financed pensions, compared to 38 percent in the RIVM (Table 8.4). Moreover, the poor receive practically no budget-financed pension benefits. The data show that benefits are concentrated among people living in urban areas and in the Central region. They also show a relatively high proportion of women benefitingfrom the budget-financed pensions, which i s consistent with the gendered patterns of employment in the education field. At the same time, although nearly 93 percent of the spending in both programs goes to people over 50, there are important differences between the programs. Nearly three-quarters of the RIVM benefits go to people 64 years and 196Here the beneficiariesare consideredto be all the people who belongto householdsthat havereceived housingvouchers since the program was launchedin 1987.Because this is a capitaltransfer, it is appropriate to considerthe cumulative number of beneficiaries, notjust the annual number of recipients. It should be noted, as well, that the numberof annualbeneficiaries is low - about 10,000householdson average- makingit difficult to generate reliableestimatesof the distributionof annualbenefits. 179 older, while two-thirds of the benefits of budget-financed pensions go to people between the age of 50 and 64.'97 Table 8.4: Distribution of the Benefits of Public Spending Selected Social Protection Programs, Costa Rica, 2003 Total Contributory Pensions Social Promotion and Assistance Indicator Social Budget- Protection Expenditure Total RlVM financed Total Promotion Network IncomeLevel ' 100.0 100.0 100.0 100.0 100.0 100.0 100.0 20% Poorest 14.9 3.1 7.5 0.6 48.6 44.2 55.8 20% next 10.9 6.3 12.3 2.9 24.0 24.9 22.5 20% intermediate 13.7 13.0 19.3 9.4 15.8 17.0 13.8 20% next 16.8 19.6 23.3 17.5 8.8 10.7 5.6 20% Richest 43.7 58.1 37.5 69.6 2.9 3.2 2.3 Poverty Group ' 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Extreme Poverty 4.3 0.3 0.8 0.1 15.4 14.8 16.5 Poor ` 9.9 2.8 6.5 0.7 30.2 25.4 38.2 No poor 85.8 96.8 92.6 99.2 54.3 59.7 45.4 Region 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Central 70.6 77.4 76.9 77.7 50.9 53.4 46.9 Peripheral Regions 29.4 22.6 23.1 22.3 49.1 46.6 53.1 Area 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Urban 72.3 82.4 80.0 83.7 43.6 42.4 45.7 Rural 27.7 17.6 20.0 16.3 56.4 57.6 54.3 Age 100.0 100.0 100.0 100.0 100.0 99.9 100.0 Under 18years of age 12.3 18to 49 years old 10.9 50 years old and over 76.8 92.8 92.6 12.1 62.2 Gender 100.0 100.0 100.0 Male 53.2 55.7 63.2 47.6 44.0 Female 46.8 44.3 36.8 48.5 I 53.8 52.4 56.0 1, imilies orderedby family incomeper householdmember. Excludeshouseholds with zero or ignoredincome (13 percent). Source: Trejos 2006. 8.59 Incontrast, the benefits of social promotion andassistanceprograms are more equitably distributed. Indeed, these programs, including non-contributory pensions (under the Social Protection Network), appear more successful in reaching the poor, people living outside the Central Region, women and people over 50 years of age. Promotionalprograms, because of their focus on human-capital development for children and adolescents under age 18, also are distributedprogressively, although less so than the Social ProtectionNetwork programs (a greater share goes to those in rural areas, however). Together, 49 percent of the benefits of social promotion and assistance programs goes to the poorest 20 percent of households (15 percent to the extremely poor); at least half of the benefits go to people in rural areas (and outside the Central region); and 54 percent of the benefits go to women. Disaggregating by age group, the 19'I t is important to point out that due to data limitations, the analysis of the benefit-incidence of the contributory pension programs presentedhere only takes into account the spending side of the equation (and not the contributions side); as such, it i s not possible to identify the net subsidy to beneficiaries. 180 data show that 47 percent of spending on social promotion and assistance programs goes to benefit children, while 31 percent of spending goes to people over 50. Access to the social promotion and assistance programs 8.60 Trejos (2006) also analyzes access by the poor and near-poor to four selected social promotion and assistance programs for which there are sufficient data: Childcare centers, the School Feeding Program, Non-contributory pensions, and the Family Housing Subsidy. As part of his analysis, he examines not only actual coverage, but program leakage, and what program coverage might be if programs were perfectly targeted to the poor. To ensure that his estimations are robust, he examines access (coverage) not only among the poor (households in the poorest income quintile) but also among those in the bottom 40 percent of the income distribution. The analysis is done over three time periods -the early 1990s, 1999, and 2003 -to see if coverage has improved over time. 8.61 Table 8.5 present four different indicators related to access and program coverage. The first indicator captures potential coverage of the target population. This compares the total number of the beneficiaries of the program with the total of the target population to assess the extent to which the program has the resources to cover all the relevant poor. Resources do appear to be sufficient in the School Feeding Program, the Non-contributory Pension Program and the Family Housing Subsidy, at least to cover families in the poorest income quintile. Moreover, based on 2003 data, the School Feeding Program and the Family Housing Subsidy appear to be able to cover the first two quintiles completely. Only the Childcare Centers of the HealthMinistry had insufficient resources to cover all the poor in 2003, since even with perfect targeting it could cover about only one-third of children under 7 from the poorest quintile (and about one-fifth of those from the poorest two quintiles). Aside from these coverage differences across programs, it i s important to point out that all programs have the capacity to increase coverage of the poor by improving targeting. 181 Table 8.5: Indicators of Accc i to Costa Rica's M a Social Protection `rograms Non-contributorv Indicator Child care centers School Feedingprogram pensions FamilyHousingSubsidy 1992/93 1999 2003 1992193 1999 2003 1990 1999 2003 1994 1999 2003 PotentialCoverage' 20% poorest 17.6 33.4 36.9 160.9 160.1 177.6 93.2 161.5 134.9 102.7 148.8 181.3 40% poorest 9.7 16.4 20.2 89.4 84.0 100.1 61.6 100.1 86.0 61.9 84.2 100.4 EffectiveCoverage' 20% poorest 5.0 17.4 16.4 61.9 58.1 68.1 35.7 43.5 53.6 17.9 38.5 42.7 40% poorest 3.7 14.2 13.9 56.6 54.8 64.5 29.9 40.5 46.1 26.3 45.4 50.3 Exclusions3 20% poorest 95.0 82.6 63.6 36.1 41.9 31.9 64.3 56.5 46.4 62.1 61.5 57.3 40% poorest 96.3 85.8 86.1 43.2 45.2 35.5 70.1 59.5 53.9 73.7 54.6 49.7 Leakages4 20% poorest 46.6 44.4 47.9 60.9 63.7 61.3 44.9 48.6 42.0 82.5 77.1 76.5 40% poorest 28.3 17.5 19.2 35.5 34.7 35.0 30.3 22.8 21.7 57.5 52.3 49.9 I/TotalBeneficiariesaspercentageofthetargetpopulation.Thetargetpopulationsconsideredare: under7yearsofage (child care centers), students of 5 to 17 years in public centers (school food program), people who are not occupiedor pensionedover 60 years of age (definedbenefit pensions)and householdsof inadequatehouses by quality of materials, hacinarnienfo or public services(housingvouchers). 2Percentage of the target populationwho are beneficiaries 3Percentageof the target populationthat is not taken care of by the program. 8.62 The secondindicator, efective coverage of the target population, is the percentageof the defined target population that actually is covered by and benefits from the program. While the coverage trend i s increasing over time in all four programs, it i s clear that effective coverage varies considerably across programs (Table 8.5). The School Feeding Program had the highest effective coverage in 2003, serving close to two-thirds of poor school-age children (whether one looks at the first or first two quintiles). The non-contributory pension program and the Family Housing Subsidy effectively covered about half of their target populations in 2003 (in the case of the former, when the first quintile i s considered and in the case of the latter, when both the first and second quintiles are considered). In contrast, the child care centers only covered between 14 and 16 percent of the target population (depending upon whether one focuses on the first or the first two quintiles). This suggests that with the exception of the general public health services, coverage of children under six by social protection programs i s relatively low, and translates into low investment per child(Trejos, 2004). 8.63 The third indicator, exclusions, reflects the percentage of the target population not covered by the program - that is, 100 minus "effective coverage." As effective coverage has increased over time, exclusions (also known as "errors of exclusion") have declined. Nonetheless, they are still above 80 percent in the CEN-CINAI Childcare program, around 50 percent for the Non-contributory pensions and Family Housing Subsidy programs, and roughly one-third for the School Feeding Program (Table 8.5). In the latter program, it i s important to note that there are additional exclusions that cannot be quantified. This corresponds to the populationthat does not currently attend school at all. As discussedearlier in the chapter (and in Chapter 6), this group is quantitatively important at the secondary level. 182 8.64 The fourth indicator estimatesprogram leakage, or errors of inclusion. When the target population i s defined as either the poorest 20 percent or 40 percent of families, the indicator shows the percentage of program beneficiaries that are outside of these target groups. It i s clear that for each program leakages are greater as the target group i s defined in a more restrictive manner. For this reason, leakages are higher if one defines the target group solely as the first quintile. Focusing on the first quintile, the Family Housing Subsidy program has the highest leakage (77 percent in 2003), followed by the School Food Program (61 percent), the CEN- CINAI Childcare centers (48 percent) and the Non-contributory pension program (42 percent). In general, there have been few improvements in leakage since 1990. When the target population is defined to include the first two quintiles, leakage i s reduced to about,half previous levels, although the housing subsidy continues to have high levels of leakage, most probably related to its very broad targeting criterion (the first 8 deciles). 8.65 Insum, except for the CEN-CINAIChildcare Centers, it would be feasible for the other three programs analyzed to cover all of the relevant poor population with existing budget resources, if this were their objective (and if programs could be perfectly targeted). Overall, there has been some progress in covering the poor since the early 1990s. However, it is important to note that, at least to date, not all the programs have been designed with the objective of covering all the poor. This i s most obvious in the case of the Family Housing Subsidy, which not only appears to make benefits available to an overly wide population group, but also provides heavy subsidiesto the effective beneficiaries (Trejos 2006). Key Institutional Challengesfor Social Protection 8.66 Costa Rica has a relatively well developed set of social protection programs relative to most other countries in Latin America -both in social insurance and social assistance. Its social security coverage (both health insurance and pension) i s among the highest in the region; and the government operates a wide range of social assistance programs to support and build the human capital of poor and vulnerable groups. Nonetheless, the social protection system faces a number of important institutional challenges. To ensure that social protection i s an effective force for sustained poverty reduction, it will be important to (i)improve coordination and strategic focus across social protection programs; (ii) strengthen financing of priority programs; (iii) improve data and information systems, (iv) improve coverage and program impact through better targeting; and (v) strengthenprogramimpact through results monitoring and evaluation. Improvingthe coordinationand strategic focus 8.67 As noted, Costa Rica implements upward of 45 social promotion and assistance programs. Aside from the few large-scale programs mentioned above, the vast majority are small, largely uncoordinated efforts, implemented by a cross-section of different ministries and agencies. Insome cases, different agencies carry out similar programs for similar constituencies, duplicating each others' efforts. Moreover, with the exception of educational scholarships, most programs offer assistance over a short time horizon - often too short to address the intended risk factors (Trejos 2006). Together, these issues - lack of scale, lack of coordination, inadequate time horizons and occasional duplication of efforts - all serve to limit the poverty-reducing impact of Costa Rica's social protection efforts. Factors such as the lack of coordination and overlap of programs also signal room for considerable efficiency gains - gains that would be important in Costa Rica's current fiscally constrained environment. 8.68 In addition, the evidence suggests that there is room for Costa Rica to strengthen the strategic focus of its social protection programs - that is, to increase its focus on areas with 183 particularly highpotential returns to public investment. The most notable examples are programs that address risks to early childhood development of poor children and seek to build human capital of poor children and youth - areas where there currently i s a relative lack of emphasis and coverage. Another programmatic area that i s relatively underdeveloped are compensatory programs to provide income assistance to families in the event of external shocks (Trejos 2006).'98 The lack of coordination and strategic focus in these programs suggests there is considerable scope for improvingthe efficiency and impact of promotion and assistance programs through consolidation and rationalization of programs around a commitment and strategic vision for poverty reduction. Strengtheninp the financing of uriority programs to reduce uoverty 8.69 Although Costa Rica allocates relatively high amounts of public resources to social policies and programs,'99approximately three-quarters of the public resources dedicated to social protection go to contributory pension programs, which largely benefit the non-poor. This means that only one-fourth of the public resources allocated to social protection go to programs.oriented to help the poorest, most vulnerable Costa Ricans. Since a significant proportion of resources spent on promotion and assistance programs also benefit the non-poor, this means that an even smaller proportion is being effectively directed to the fight against poverty. 8.70 The high degree of inflexibility in resource allocations across social protection programs i s another problem. As noted, over 80 percent of the resources for promotion and assistance programs financed through FODESAF in 2004 were set by legislation, for example. While this was intended to assure a minimum level of resources for specific programs, minimum funding levels have often not been implemented in practice. Nonetheless, setting resource levels by legislation has served to limit the government's flexibility to address new strategic priorities social protection as they emerge. Moreover, in some cases, legislative mandates serve to protect spending on the non-poor - as in the case of FODESAF's funding of the Costa Rican Sports Institute and the significant proportion of the Family Housing Subsidy that goes to households in fourth to eighth deciles (Table 8.5). 8.71 There also remain important issues of financial sustainability of social protection spending, particularly in contributory pension programs. Resources allocated to the contributory pension programs have grown significantly in recent years, particularly programs such as teachers' ofjudicial employees' pensions that are financed largely by the national budget. While recent reforms of the RIVM system have strengthened the financial state of the Disability, Old Age, and Death Fund, such reforms did not extend to teachers' or judicial workers' pensions. Meanwhile, the needto allocate more funds for pensions out of the national budget has served to limit or "crowd out" greater spending on social promotion and assistance programs that could fight poverty. Improvingdata and information systems 8.72 The information base on which programmatic decisions are made for social protection remains weak, and when information exists, the evidence suggests it is not used to its best advantage (Trejos 2006). As noted above, information on social protection spending is fragmented and dispersed across institutions. As a result, for the majority of the programs it i s not possible to know how much i s being spent or how much programs actually cost. So, a full 19' This would include programsto help individualsor families make economic adjustmentsin the face of long-runterms-of-tradechanges resultingfrom trade reforms, such as those that would occur under DR- CAFTA (see Box 3.3 in this report; World Bank 2005). 19' This findingreaffirmsearlier findings presentedinWorld Bank (2003) and Marques (2004). 184 accounting of resource utilization - either inflows or outflows -i s impossible. This issuetends to be most important for the smaller programs. 8.73 Information about the target population (the poor) i s also weak. The main source of information about income poverty is the EHPMdata collected yearly by INEC, the Costa Rican statistically agency. As noted in Chapter 2 (and accompanying annexes), the EHPM surveys were not originally designed for poverty monitoring, and thus have some limitations in this regard. The most important limitation involves the limited number of income variables that are collected.2m Perhaps more importantly from a programmatic perspective, IMAS gathers information on the target population inits SIPO database. But the SIPO continues to have several important limitations; it does not represent a comprehensive "census" of the poor, it i s not updated on a regular basis (to reflect changes in family's welfare over time), and it does not contain information on the "control population" (people who are not poor). With improvements, the SIPO could provide the basis for a strong database for poverty targeting, but this would require strengthening IMAS's technical capacity to collect data, as well as making an ongoing institutional commitment (and financing) to collect, maintain, analyze, and share the database with other relevant agencies (Trejos 2006). 8.74 The information base on program beneficiaries also requires strengthening. As with the target population, EHPMdata on of the receipt of program benefits is limited. Data collected by implementing agencies also has been less than optimal. For example, for many years, the number of beneficiaries of the school feeding programwas estimated from school enrollments, rather than having a count of the students actually being served. Moreover, incomplete information both on beneficiaries and on program costs precludes analysis of the unit costs of programs and, thus, the cost effectiveness of different interventions. In recent years, IMAS has made some important improvements in the systematic collection of information about the beneficiary population through the SIPO and through a database called SAB (Sisternu de Atencion de Beneficiurios) that provides information that could be used to follow-up with program beneficiaries. Nonetheless, IMAS has not usedthis data to its fullest extent, and to date it has not made the data available for other stakeholders, limiting its value as a program implementation or monitoring tool (Trejos 2006). Improvingcoverage and programimpact throuph better targeting 8.75 Costa Rica's outreach to the poor, and thus its poverty reduction efforts, would benefit from increased use of targeted approaches and improved targeting mechanisms to ensure that poor families, who because of lack of physicalaccess or family constraints, are not covered by (or have not be successful in obtaining access to) Costa Rica's universal programs. Well-targeted programs also can help conserve scarce budget resources by focusing resources where they are most needed. Lack of effective targeting to date is indicated by a high degree of leakage of promotion and assistance program benefits to the non-poor (Table 8.5). While this leakage has diverse origins, it is clear that there is room to improve the extent to which program benefits reach the poor. 8.76 Some of the problems in current programs i s a function of program design elements, as in the cases of the School Feeding Program and the Family Housing Subsidy. From the perspective of reducing poverty, review of these design elements may be warranted. Perhaps more importantly, some of the leakage stems from weak information- and inadequate beneficiary- selection systems. Indeed, many programs still do not have consistent, objective, and unbiased This included bothearnedincome and income received by households from the main social protection programs, whether contributory pensions or social promotion and assistance. 185 ways of identifying and selecting program beneficiaries. Rather, they depend on the perceptions of staff (Trejos 2006). This i s the case, for example, with promotion and assistance programs associated with the Ministry of Education. Although, as noted above, the SIP0 has great potential to support more effective targeting and beneficiary selection, the database requires updating and strengthening. It needs to be made more generally available to staff implementing non-MAS programs, and, perhaps, adapted to address the different needs of different categories of beneficiaries - for example, those inthe social assistance pension program. Strengtheningimpact through results monitoring and evaluation 8.77 The lack of data on potential and actual program beneficiaries and on program costs has precluded adequate monitoring and evaluation of Costa Rica's social protection programs. Moreover, until recently, results monitoring and impact evaluation have not been a key priority for the Government of Costa Rica. This can be seen in the vast majority of programs that have been designed without attention to the development of impact indicators or to collection of baseline data that would facilitate effective monitoring or evaluation (Trejos 2006). Results monitoring and selective evaluations of program impact are critical to identifying effective (and cost effective) social protection delivery models for poverty reduction. Identifying such models will be essential if the government is to rationalize and consolidate the social protection sector effectively. Identifying effective delivery models i s also critical in the face of fiscal constraints. Systematic program monitoring and evaluation also facilitate "retro-fitting" of potentially effective programs and projects by identifying the types of (marginal) changes in design and implementation strategies that could make those interventions truly effective. Conclusions 8.78 This chapter has analyzed Costa Rica's social protection system. It has noted its considerable accomplishments in social protection and discussed some of the key challenges the country still faces as it seeks to regain momentumin the fight against poverty. In this context, the analysis has highlighted the considerable scope for increasing the efficiency and impact of Costa Rica's social protection spending and noted the types of measures that could help make social protection a more effective force for poverty reductionin Costa Rica. 8.79 Where should the country start? As part of a broader strategy for poverty reduction, severalpriority actions can be identified: 0 To strengthen the strategic focus of Costa Rica's social protection programs, placing greater emphasis on areas of interventions with particularly high potential returns - such as on early childhood development programs and education-related interventions - that increase development of human capital among poor children and reduce the risk of the intergenerational transmission of poverty. 0 To rationalize and consolidate existing social assistance programs to ensure that public social protection spending has greater impact. Such reforms should be consistent with the strategic priorities identified for renewed poverty reduction in Costa Rica. To ensure that public spending on social protection i s more effective, it will also be important for the government to identify effective service delivery models for each of the main categories of risk faced by the poor. This will require strengthened results monitoring as well as strategically chosen evaluations of programimpact. As a complement to universal programs, to put greater emphasis on targeting social protection programs (including support for human capital investments and extension of 186 health insurance and pension coverage) to the poor and extremely poor who currently fall outside the system. Greater and more effective use of targeted approaches will require development of stronger targeting mechanisms. Costa Rica could improve targeting efficiency by strengthening and updating the SIP0 database, and makingit more widely available for poverty-oriented programs throughout the social protection system. Other possible targeting instruments, including updates of recent poverty maps, also could be usedto support the targeting process in a complementary manner. A promising programmatic approach in this area involves expanded use of conditional cash transfers, which provide monetary transfers to poor households on the condition the families ensure that their children attend and complete schooling. Such "demand-side'' programs have been effective in raising educational enrollment and attainment among the poor in such Latin American countries as Mexico, Colombia and Brazil, and they can be important complements to universal programs that focus on ensuring the supply of basic services to the population. They hold considerable promise for improving school enrollment and achievement among the poor in Costa Rica. Indeed, ensuring that poor children enroll and complete secondary education i s a priority for poverty reduction. The country has some experience with conditional transfer programs -through such programs as Creciendo Juntas and Construyendo Oportunidades - but these programs have operated on a relatively small scale and, as such, their impact to date has been limited. 8.80 It is important to note that with greater strategic focus, program rationalization and consolidation (with emphasis on effective delivery models), and better program targeting, significant improvements in poverty reduction could be made even within the existing real budget envelope. 187 9. ELEMENTS OFA POVERTY REDUCTIONSTRATEGYFORCOSTA RICA 9.1 This study provides ample explanation why Costa Rica is well-known for its accomplishments reducing poverty and achieving social progress. Its poverty and inequality levels are generally low by Latin American standards. It also performs well relative to other countries inLatin America and the Caribbean region, and to countries with similar income levels, in health and access to basic services. Its infant- and child-mortality rates are significantly lower than those in comparator countries, while its average life expectancy i s substantially higher. Almost all Costa Ricans - 97 percent - have access to improved water sources, also high compared to comparator countries. Indeed, both in absolute terms and by regional standards, access to a range of basic services, includingelectricity and sanitation, is generally high in Costa Rica. 9.2 Notwithstanding these considerable achievements, though, Costa Rica faces a number of significant challenges. While the percentage of the population that is poor declined from 31.7 percent in 1989 to 22.9 in 1994, the poverty rate has not declined over the last decade; indeed, in 2004, 23.9 percent of the Costa Rican population was still poor. The fact that poverty rates have stagnated over the 10 years i s surprising, considering that Costa Rica experienced relatively consistent economic growth over the period. Income inequality has been rising. While still relatively low by regional standards, inequality inCosta Rica, as measuredby the ginicoefficient, rosefrom 0.44 in 1989to 0.48 in2004. Although the country has invested considerably in education and made some progress as a result, Costa Rican students still lag behind the Latin America and upper-middle income country averages in achievement at the secondary school level. This reflects, in part, the long-term impact of Costa Rica's economic crisis in the 1980s. But the consequences are real. The poor still lag behind the non-poor in educational access and attainment, and this adversely affects their ability to participate in and benefit from economic growth. Moreover, even though the country has a relatively well developed set of social protection programs, many of Costa Rica's poor still fall outside the reach of the safety net. 9.3 This study has examined recent developments on the poverty front in Costa Rica, placing particular emphasis on why poverty rates have not declined over the last 10 years, even in the face of consistent economic growth over the period. It has developed a dynamic profile of poverty in Costa Rica to help explain who the poor are. It has analyzed recent patterns of economic growth and the extent to which the benefits of growth have reached the poor. The study also has focused on the dynamics of the labor market, including how labor market developments have affected the ability of the poor to generate higher incomes since 1994. Special attention has been paid to the impact that Nicaraguan has had on poverty, and to the special challenges faced by poor (particularly female) workers. Finally, the report has examined the role and effectiveness of social sector spending and policies in improving the welfare of the poor and providing them with the capacity to escape poverty. 9.4 The analysis has highlighted several reasons why progress in poverty reduction has stalled since 1994. First, GDP, along with household income growth, has slowed during recent 188 years. EHPM data show that between 1994 and 2000, household per capita income grew at less than one-third its 1989-1994 rate; since 2000, average per capita household income has barely changed. More importantly, while the benefits of growth were relatively evenly distributed between 1989 and 1994, an increasing share of the benefits of growth has accrued to non-poor households since then. In the face of negligible overall income growth, poor and near-poor householdsactually saw their incomedecline from 2000 to 2004. 9.5 These changes in the distribution of the benefits of growth reflect broader changes in the Costa Rican and world economies that have led to increases in the relative demand for higher skilled workers at the same time Costa Rica experienced a rise in the relative supply of low- skilled workers; among other things, for instance, the share of high school graduates in the Costa Rican labor force declined and the share of high drop-outs increased. Since the mid-l990s, this growing mismatch between the education and skill levels of the poor and the demand for labor in the economy has resulted in increased earnings inequality, significant increases inunemployment among the poor, and an increase in part-time (as opposed to full-time) work among low-skilled and poor workers -most strikingly among poor, single mothers. Towards a Poverty Reduction Strategyfor Costa Rica 9.6 The collection of the evidence presentedinthis study suggests there i s considerable scope for Costa Ricato recapture its momentum inreducingpoverty. Indeed, the evidence suggests that a multi-dimensional strategy for poverty reduction is warranted to ensure that the poor are able to participate in and benefit from future socio-economic progress. Many of the priorities identified here are consistent with those expressed by the new administration, including an emphasis on strengthening the human capital of the poor and improving the outreach and poverty reducing impact of Costa Rica's social safety net. The proposed strategy would include efforts to: Promote higher sustained levels of economic growth. This would generate an expanding set of economic opportunities for all Costa Ricans, includingfor the poor Strengthen the human capital of all Costa Ricans, with emphasis on improving secondary school outcomes among the poor and strengthening skills formation among low-skilled workers. This would ensure that the poor can take full advantage of emerging economic opportunities Ensure social protection coverage (both social insurance and social assistance) to the poorest, most vulnerable groups. This would protect them against risks and shocks, and improve their access to basic services Create an enabling environment for poor worker. This would ensure that they can engage in more remunerative employment, and would particularly benefit working mothers, who often lack such opportunities today. Define regionally diferentiated investment strategies that reflect geographic differences in patterns of poverty. This would maximize the effectiveness of anti-poverty measures at the local level. Strengthen information systems in order to improve poverty monitoring, efective program targeting and monitoring and evaluation of poverty reduction programs. This would ensure that the government's poverty-reduction efforts are producingthe desired results. 189 9.7 While Costa Rica's commitment to the social sectors has been important to its progress and accomplishments over the years, the evidence makes clear that considerable gains could be realized by increasing the efJiciency of spending through strategic realignment of resources toward areas with the highest potential returns, andby making greater use of targeted approaches to reaching the poorest, most vulnerable Costa Ricans. Increasing the efficiency and impact of spending as a complement to existing universal programs i s particularly important given the fragility of Costa Rica's current fiscal situation. A closer look at key priorities follows. 9.8 Promoting economic growth. A large body of global evidence shows that sustained economic growth is critical to poverty reduction in the long-term. Indeed, it is essential to ensuring expandedeconomic opportunities for all Costa Ricans, notjust the poor. 9.9 What will be the keys to faster growth in Costa Rica in the coming years? Recent empirical analysis has highlighted several key factors to Costa Rica's growth performance in the 1990s. Among the most important: investments in education, investments in infrastructure, strengthening of the country's financial sector, and increases in trade openness. The new World Bank Country Economic Memorandum for Costa Rica (2006) indicates that progress in these areas will continue to be important for growth in Costa Rica in the 2000s. In addition, the report highlightsthe importance of strong macro-economic and fiscal management, as well as efforts to strengthenthe country's competitiveness by strengthening its system of researchand innovation. 9.10 As Costa Rica's recent experience indicates, however, growth alone will not be enough to engender poverty reduction. It also is essential to create conditions in which the poor can take advantage of new and emerging economic opportunities. Thus, concerted efforts to increase the capacities of the poor, along with their abilities to participate in and benefit from future economic progress, would enhance any poverty-reduction strategy. Such efforts would includeprograms to strengthen the human capital of the poor, increase social protection coverage for them, and provide social supports, such as affordable child care, to enable poor working mothers to participate more fully inthe labor market. 9.11 Strengthening Costa Rican's human capital. Strengthening human capital, with special emphasis on the poor, should be a key part of any country's development strategy, and thus should be at the core of any effort to re-invigorate poverty reduction programs in Costa Rica. Key priorities include: 0 Policies and investments to increase enrollment, persistence, and graduation at the secondary school level, particularly among the poor. A fundamental element of any initiative to improve secondary school outcomes among the poor should be efforts to improve secondary school relevance and quality, particularly in poor, rural areas. This will require reallocation of education-sector resources toward secondary schooling, along with increases in the share of non-salary, recurrent spending in education, whether on quality-enhancing measures (both at the primary and secondary levels) or on efforts to improve secondary school access and achievement among the poor. 0 Efforts to improve the capacity and employability of poor, low-skilled workers. Here the objectives would be to: (i) bridge the skills gap between poor workers and the growing demand for higher skilled workers in the Costa Rican economy, and in doing so, (ii), reduce structural unemployment, which results from the mismatch between poor workers' skills and the market demand for labor and is an important contributor to poverty among the poor. 190 Finding the right type of program to enhancejob skills of poor, adult workers will be a challenge, however. International evidence on the impact of skills-training programs suggests that only a small proportion of such programs actually raise people's employability and earnings. It will be important, therefore, to identify effective approachesto strengthening people's job skills, drawing on international best practice. It also will be critical to monitor and evaluate Costa Rica's efforts to increase workers' skills so as to identify which specific programdelivery models are most effective. 9.12 While health conditions in Costa Rica are generally quite good, even among the poor, recent increases in the incidence of malaria, dengue, and tuberculosis, along with declines in measles and poliomyelitis vaccination rates, signal a need for renewed vigilance in the public health system. This will require some redirection of health sector resources toward vaccinations and preventive health measures, along with public information campaigns aimed particularly at the poor. 9.13 Ensuring social protection for the poorest, most vulnerable groups. Among the key functions of a social protection systemare: (i) to reduce people's vulnerability to poverty, and (ii) to increase the economic mobility of the poor by ensuring that even the poorest families have access to basic services and the ability to invest in human capital. To enhance the ability of Costa Rica's social protection programs to fulfill these functions, it will be important to: 0 Strengthen the system's strategic focus on areas of high potential return, such as early childhood interventions and programs to increase human capital development among poor children. 0 Consolidate and rationalize existing social assistance programs, in line with strategic priorities, to strengthenthe impact of social protection spending poverty reduction. 0 Expand social protection coverage to poor and extremely poor people who currently fall outside the system - by putting greater emphasis on targeting as a complement to universal programs and through strengthening programtargeting mechanisms, such as the SIP0 database. 9.14 With greater strategic focus and more effective targeting, Costa Rica's social protection system should be able to increase coverage of the poor significantly, even within the current real budget envelope. 9.15 It should be noted that there is a special role for "demand-side'' programs to help poor families overcome financial constraints to investing in their children. Demand-side programs, such as conditional cash transfers, provide financial support to poor households in return for ensuring that their children attend (and complete) secondary education. Costa Rica has some experience with this type of program (CreciendoJuntos, Construyendo Oportunidades), but such programs have operated on a relatively small scale and, thus, their impact has been limited. Conditional cash transfer programs have been effective in raising educational enrollment and attainment among the poor in such Latin American countries as Mexico, Colombia, and Brazil, andhold considerable promise for improving school achievement among the poor in Costa Rica. 9.16 Creating an enabling environment for poor workers. The evidence suggests that the reason part-time work is particularly prevalent among poor women i s because many working mothers are not able to find adequate or affordable child care arrangements. Current Costa Rican labor market legislation also limits women's flexibility to seek employment outside standard 191 working hours (by restricting their ability to work at night). Creating an enabling environment for poor female workers would thus entail: (i) providing greater social support in the form of affordable child care options, and (ii) leveling the regulatory playing field to allow women the same working-hour flexibility as men. 0 Expanding child care options for poor families would make it easier for poor female workers, whether poor single mothers or female spouses, to work full-time and generate higher earnings. Several policy options - alone or in combination - could reduce women's child care-related constraints to employment, including: (i) expanding accessto and participation in early childhood development and pre-school education; (ii) expanding government subsidies to poor families for child care; (iii) providing before- and after-school child care programs in schools; and (iv) encouraging private firms to provide subsidized day-care facilities at work. Reducing legal barriers to women working non-standard hours - by providing them the same working-hour flexibility as men - would make it easier for single, working mothers to seek employment during times when it may be easier to find alternative childcare arrangements (e.g., extended family). 9.17 Defining regionally differentiated investment strategies. Most of the elements of a poverty-reduction strategy for Costa Rica would benefit from taking a national approach. Nonetheless, differences across regions in the levels of poverty and concentrations of poor people suggest that regionally differentiated policies and investments may be warranted in some cases. For example: 0 Inareas like the Central Region, where the incidence of poverty is low but concentrations of poor people are high, investments in infrastructure and improvements in the investment climate designed to take advantage of economies of scale could be particularly effective. By increasing the profitability of business investment, such approaches may increase demand for relatively low-skilled labor, and thus generate employment among the poor. In placeslike the Brunca and Chorotega Regions, where the incidence of poverty is high, but concentrations of the poor are low, no such economies of scale exist. As a result, investments or targeted support for education, training, or technical assistance - all of which raise people's economic mobility -may be more effective (and cost-effective). 9.18 Even though Costa Rica is a relatively small country geographically, taking local as well as national considerations into account in the design of interventions can contribute to more effective anti-poverty interventions at the local level. 9.19 Strengthening information systems. Costa Rica's poverty reduction efforts would benefit from strengthening and increasing the transparency of data, information, and management systems across several dimensions: 0 Strengthening poverty measurement and monitoring. It will be important to strengthen Costa Rica's household survey, the EHPM, and to assess the strengths and weaknesses of the country's current poverty measurement methodologies to ensure that the poverty monitoring system provides suitable empirical support to policy makers.in developing poverty reduction strategy. The forthcoming 2004 Income and Expenditure Survey (IES) for Costa Rica will provide a valuable opportunity to clarify a number of pending 192 empirical and methodological questions regarding poverty measurement and poverty monitoring. Strengthening program targeting mechanisms. The SIP0 database, which contains a registry of the poor, is not nationally representative, nor has it been updated to reflect households' movement out of (or into) poverty. It also has not been adopted to date for use by many government agencies. Costa Rica's poverty reduction efforts would benefit if this database - or other instruments that would permit identification and targeting of the poor - were strengthenedand updated, and if other efforts were launchedto make the information more widely available to various agencies that operate a wide range of social programs focused on the poor. Developing institutional mechanisms for results monitoring and evaluation of interventions for more effective program management. Developing a system for monitoring the results of social and poverty reduction programs, along with a support institutional structure, would help the government and other stakeholders identify effective (and cost effective) interventions. A strong monitoring and evaluation system would help policymakers ensure that interventions produce the expected results and, if necessary, enable them to undertake mid-course corrections in program design or programtargeting. Improving transparency and public access to government information on poverty-related programs. This would help to create greater "democracy of information" in Costa Rica and increase accountability among government institutions involved in the fight against poverty. 193 REFERENCES Autor, D.; Katz, L.; and Krueger, A. November 1998, "Computing Inequality: Have Computers Changedthe Labor Market?' Quarterly Journal of Economics, Vol. 113, No. 4, 1169-1214. Banco Mundial. 2003, "Costa Rica: El Gasto Social y La Pobreza", Washington DC, Estados Unidos de Amkrica: Banco Mundial. Barahona, M.and Castro C. 2003, "Reformas Educativas en Costa Rica (1986-2002), en las Transformaciones en la Estructura Social en Costa Rica en el Cambio de Siglo", Instituto de Investigaciones Sociales, Universidad de Costa Rica. Barquero, Jorge y Juan Diego Trejos. 2004, "Tipos de hogar, ciclo de vida familiar y pobreza en Costa Rica 1987 - 2002". Poblacidn y Desarrollo, Revista Electr6nica. Volumen 2, n6mero 1, articulo 4, julio - diciembre 2004. Publicadoel 1dejulio del 2004. http://ccp.ucr.ac.cr/revista/ Bortman, Marcelo. 2002, "Indicadores de Salud: iMejor6 la equidad? Costa Rica 1980-2000". Ministerio de Salud/OPS Oficina Regional, San JosC, Costa Rica Bourguignon, F. 2002. Caja Costarricense del Seguro Social: Anuarios Estadisticos. Aiios varios. Direccion Actuarial y de PlanificacidnEcon6mica. Card, D. 1996, "The Effect of Unions on the Structure of Wages: A Longitudinal Analysis", Econometrica, 64(4): 957-979. Card, David and Craig Riddell, W. 1993, "A Comparative Analysis of Unemployment in Canada and the United States", in Small Differences That Matter, edited by David Card and Richard Freeman. University of Chicago Press, pages 149-190. Castro, Victor. 2005. Evaluacidn de efectos del Programa Ideas Productivas en las familias financiadas por el IMAS en la Gerencia Regional de Cartago, Costa Rica, aiio 2000. Trabajo final presentado en la Universidad de Costa Rica como requisito parcial para optar a1 titulo de Magister en Evaluaci6nde Programasy Proyectos de Desarrollo. Mimeografiado. CEPAL, / CELADE. July 2001, Boletin Demogrdfico No. 68,America Lutina, Facundidad. Cespedes, Victor Hugo, and Jimenez, Ronulfo. 1994, Apertura Comercial y Mercado Laboral en Costa Rica. San Jose, Costa Rica: Academia de CentroamCrica y Centro Intemacional para el Desarrollo Econ6mico. Cespedes, Victor Hugo. 1979,Evolucidn de la Distribucidn del lngreso en Costa Rica, Instituto de lnvestigaciones en Ciencias Econdmicas, Serie de Divulgacidn # 18.University of Costa Rica, San JosC, Costa Rica. Chomitz, K. 2005, "Development Policy for a Heterogeneous Space", in Jog0 Paul0 dos Reis Velloso (coord.), Estratkgias para o Nordeste e a AmazBnia (Cadernos F6rum Nacional 2). Rio de Janeiro: Instituto Nacional de Altos Estudos. 194 Consejo Social. 2005. Catdlogo de Programas Sociales 2002 - 2005. San JosC, Costa Rica. Presidencia de la Repfiblica, Consejo Social. Mimeografiado. Contraloria General de la Repdblica (CGR). 2005, Memoria anual. San JosC, Costa Rica: ContraloriaGeneral de la Repdblica. Cowell, F.A. 1980, "On the Structure of additive Inequality Measures", Review of Economic Studies, 47, pp.521-3 1. Cowell, F.A.and S.P. Jenkins. 1995, "How Much Inequality Can We Explain? A methodology and an application to the USA", Economic Journal, 105, pp.421-430. Datt, G. and Ravallion M. 1992, "Growth and redistribution components of changes in poverty measures. A decomposition with applications to Brazil and India in 1980s", Journal of Development Economics 38:275-295. North Holland. De Ferranti, D.; Perry, G.; Ferreira, F.; and Walton, M.2004, Inequality in Latin America and the Caribbean: Breaking with History? The World Bank, Washington, D.C. De Ferranti, D.; Perry, G.; Lederman, D.; Valdes, A.; and Foster, W. 2005, Beyond the City: The Rural Contribution to Development. The World Bank, Washington, D.C. Deaton, A. February 2005, "Measuring poverty in a growing world (or measuring growth in a poor world"), TheReview of Economics and Statistics LXXXVII(l),1-25. pp. DiGropello, Emanuela. 2005, Central America Education Strategy: An Agenda for Action, The World Bank, Washington D.C. Estadode la Educacih. 2005, Primer Informe, Programa Estado de la Nacih, San Jos6, Costa Rica. Fields, Gary S. 1998, "Accounting for Income Inequality and Its Change", Cornel1 University, May. Mimeo. Fields, Gary S. 2003, "Accountingfor Income Inequality and Its Change: A New Method, with Applications to the Distribution of Earnings in the Untied States", in S. Polachek, ed., Research inLabor Economics, Volume 22: Worker Well-being and Public Policy. Elsevier Publishing. Fields, Gary S., and Yoo, Gyeongjoon. June 2000, "Falling Labor Income Inequality in Korea's Economic Growth: Patterns and Underlying Causes", Review of Income and Wealth, Series 46, NO.2, 139-159. Fondo de las Naciones Unidas para la Infancia (UNICEF). 2005. Estado Mundial de la Infancia 2005. La infancia amenazada. Nueva York, EEUU:UNICEF. Funkhouser, E. 1998, "Changes in the Returns to Education in Costa Rica", Journal of Development Economics, Vol. 57-2,289-3 17. Funkhouser, E. 1999, "Cyclical Conditions and School Attendance in Costa Rica", Economics of Education Review, Vol. 18-1, 31-50. 195 Gallego, F. and Loayza, N. 2002 "The Golden Period for Gorwth in Chile: Explanations and Forecasts", Central Bank of Chile, Series on Central Banking, Analysis and Economic Policies, 6. Gasparini, L.; Gutierrez, F.; and Tornarolli, L. 2005, "Growth and Income Poverty in Latin America and the Caribbean: Evidence from Household Surveys", The World Bank, Washington, D.C. Gindling, T.H. January 1993, "Women's Wages and Economic Crisis in Costa Rica", Economic Development and Cultural Change, Vol. 41, No. 2, pp. 277-298. Gindling, T.H., and Berry, Albert. November 1992, "The Performance of the Labor Market DuringRecessionand Structural Adjustment: Costa Ricainthe 1980s", WorldDevelopment, Vol. 20, NO.11,pp. 1599-1616. Gindling, T.H., and Berry, Albert. 1994, "Costa Rica", in Labor markets in an Era ofAdiustment, Volume 2: Case Studies, edited by Sue Horton, Ravi Kanbur and Dipak Mazumdar. The World Bank, Washington, D.C., pp. 217-257. Gindling, T.H., and Robbins, Donald. April 29, 2001, "Patterns and Sources of Changing Wage Inequality inChile and CostaRica DuringStructural Adjustment", World Development, 725-745. Gindling,T.H., and Terrell, Katherine. August 1995, "The Nature of MinimumWages and Their Effectiveness as a Wage Floor in Costa Rica", World Development, Vol. 23, No. 8, pp. 1439- 1458. Gindling, T.H., and Terrell, Katherine. Fall 2004b, Michigan Journal of International Law, Vol. 26, NO.1, pp. 245-270. Gindling, T.H., and Terrell, Katherine. May 2004a, `The Effects of Multiple MinimumWages Throughout the Labor Market", IZA WorkingPaper No. 1159. Gindling, T.H., and Terrell, Katherine. November 2005, "The Effect of Minimum Wages on Actual Wages in the Formal and Informal Sectors in Costa Rica", World Development, Vol. 33, NO.11,pp. 1905-1921. Gindling, T.H., and Trejos, Juan Diego. July 2005, "Accounting for Changing Inequality inCosta Rica: 1980-99", Journal of Development Studies, Vol. 41, No. 5, pp. 898-926. ' Goldberg, Koujianou; Pavcnik, Penelopi and Nina. November 2001, "Trade, Wages, and the Political Economy of Trade Protection: Evidence from the Colombian Trade Reforms", mimeo. Dartmouth College. Hoynes, Hilary; Page, Marianne; and Stevens, Ann. October 2005 "Poverty in America: Trends and Explanations", NBER Working PaperNo. 11681. Huppi, M., and Ravallion, M. 1990, "The Sectoral Structure of Poverty During an Adjustment Period. Evidence for Indonesia in the Mid-l980s", Working papeer 0529. The World Bank, Washington, D.C. INEC. 2002, "Estimaciones y Proyecciones de Poblaci6.n 1970-2050", SanJose, Costa Rica, 196 August 2002 p.45. INEC. December 2004, Encuestade Hogares de Propdsitos M6ltiples: Julio 2004 Principales Resultados, San JosC, Costa Rica. INEC. September 2003, Manual de Creaci6n de Variables EHPM, INEC Unidad TCcnicade Programacidn, San JosC, Costa Rica. Inter-American Development Bank. 1998, "Facing up to Inequality in Latin America: Economic and Social Progress in Latin America, 1988-1999Report", Washington. Juhn, C.; Murphy, K; and Pierce, B. 1993, "Wage Inequality and the Rise in Returns to Skill", Journal of Political Economy, 101-3,410-442. Katz, L., and Autor, D. 1999, "Changes in the Wage Structure and Earnings Inequality", in 0. Ashenfelter and D. Card (eds.), The Handbook of Labor Economics, Volume 3, North Holland, 1464-1555. Katz, L., and Murphy, K. 1992, "Changes in Relative Wages, 1963-1987: Supply and Demand Factors", Quarterly Journal of Economics, Vol. 107,35-78. Knight, J. B., and Sabot, R. 1983, "Educational Expansion and the Kuznets Effect", American Economic Review, Vol. 73, No. 5, 1132-1136. Krueger, A., "The Relationship Between Trade, Employment, and Development", in Gustav Ranis and T. Paul Shultz (eds.), The State of Development Economics: Progress and Perspectives,Cambridge, MA, Basil Blackwood, 357-385. Kuznets, S. March 1955, "Economic Growth and Income Inequality", American Economic Review, Vol. 45, 1-28. Loayza, N.; and Raddatz, C. 2002, "The Composition of Growth Matters for Poverty Alleviation", The World Bank, Washington, D.C. Loayza, N.; Fajnzylber, P., and Calderon, C. 2002, "Economic Growth in Latin America and the Caribbean: Stylized Facts, Explanations and Forecasts", The World Bank, mimeo. Lopez, H.and Serven, L.2006, "A Normal Relationship? Poverty, growth and inequality" Policy ResearchWorking Paper, 3814, The World Bank, Washington, D.C. Marenco, Leda; Ana M. Trejos, Juan D Trejos y Marianela Vargas. 1998. Del silencio a la palabra: un modelo de trabajo con las mujeres jejas de hogar. San JosC, Costa Rica: Segunda Vicepresidencia de la Rep6blica de Costa Rica con el auspicio del PNUD. Marques, JosC S. 2004. "Evaluaciones de las redes de seguridad social en CentroamCrica. Analisis de 10s principales hallazgos." En Shelton H. Davis, Estanislao Gacihja y Carlos Sojo (editores): Desafios del Desarrollo Social en Centroame`rica. San JosC, Costa Rica: Facultad Latinoamericana de Ciencias Sociales (FLACSO) - sede Costa Rica y Banco Mundial. Marquette, Katherine. November 2005, "Nicaraguan Migrants and Poverty in Costa Rica", Report to the World Bank Poverty Assessment Team. 197 Martinez, Juliana. 2005. Reformas recientes de las pensiones en Costa Rica: avance hacia una mayor sostenibilidadfinanciera, accesoy progresividad del primer pilar depensiones. Ponencia preparada para el UndCcimo Informe sobre el Estado de la Naci6n en Desarrollo Humano Sostenible. San JosC, Costa Rica: Programa Estado de la Naci6n en Desarrollo Humano Sostenible. Ministerio de Hacienda de Costa Rica, March 2005, Cuadros Resumende Liquidaci6n del period0 2004, San JosC, Costa Rica. Montiel Masis, Nancy. 2000, "Reformas Econ6micas, Mercado Laboral y Calidad de 10s Empleos", pages 423-472 in A. Ulate, editor, Empleo, Crecimiento y Equidad: Los Retos de las Reformas Econdmicas de Finales del Siglo XX en Costa Rica, Editorial de la Universidad de Costa Rica. Mora R. and RamosP, 2004, Educaci6n y Conocimiento en Costa Rica: Desafios para Avanzar haciaunaPoliticade Estado. Serie de Aportes para el Aniilisis del Desarrollo Humano Sostenible, No. 8, Programa Estado de la Nacibn, San JosC, Costa Rica. Perry, G.; Arias, 0.;Lopez, H.; Maloney, W.; and Serven, L. 2006; Poverty Reduction and Growth: Virtuous and Vicious Circles. The World Bank, Washington, D.C. Portes, A., and Rumbaout, R.G. 2001, Legacies: The Sotry of the Second Generation, Berkeley, University of California Press. Portes, Alejandro and Hao, Lingxin. 2004, "The Schooling of Children of Immigrants: Contextual Effects on the Educational Attainment of the Second Generation", Proceedings of the National Academy of Sciences, 101(3), pp. 11920-11927. Programa Estado de la Nacibn. 2004. De`cimo Informe Estado de la Nacidn en Desarrollo Humano Sostenible. San JosC, Costa Rica: Proyecto Estado de la Naci6n. Ravallion M,July 1998, Poverty Lines in Theory andPractice, LSMS Working Paper No. 133, The World Bank, Washington D.C. Ravallion, M. 1997, "Can High Inequality Development Countries Escape Absolute Poverty?' Economics Letters, 56,5 1-57. Ravallion, M. 2004, "Pro-Poor Growth: A Primer", The World Bank, Policy Research Working Paper No. 3242. Robbins, D., and Gindling, T.H. June 1999, "Trade Liberalization and the Relative Wages of More-Skilled Workers in Costa Rica", Review of Development Economics, Vol. 3, No. 2, pp. 140- 154. Robbins, Donald; and Gindling, T.H. 1997, Liberalizacidn Comercial, Expansidn de la Educacidn y Desigualdad en Costa Rica, Sene de Divulgaci6n Econ6mica No. 27. Instituto de Investigaciones en Ciencias Econdmicas de la Universidad de Costa Rica. Robertson, Raymond. 1999, "Inter-Industry Wage Differentials Across Time, Borders, and Trade Regimes: Evidence from the U.S. and Mexico", Macalester College, manuscript. 198 Robinson, S. 1976, "A Note on the U-Hypothesis Relating Income Inequality and Economic Development", American Economic Review, Vol. 66,437-440. Rofman, Rafael, and Leonard0 Lucchetti, 2006, "Pension Systems inLatin America: Concepts and Measurements of Coverage," processed, World Bank, Washington, D.C. Rosero Bixby, Luis,2004, "Evaluaci6n del Impact0 de la Reformadel Sector Salud en Costa Rica mediante un Estudio Cuasi experimental". Revista Panamericana de Salud Mblica. 2004: 15(2):94- 103. Rosero-Bixby, Luis. "Los Retos de la Inmigracidn Nicaragiiense a Costa Rica", Estado de la Nacidn, SanJos6, Costa Rica. SANIGEST, 2006, "Poverty and Health in Costa Rica: Progress, Pending, and Policy Options". March 10.2006 SANIGEST, October 2003, Fortalecimientodel Sector Educativoen Costa Rica: Situaci6n Actual y Opciones de Politica. Sauma, P. and Trejos, J.D. 1990,Evolucidn reciente de la distribucidndel ingreso en CostaRica. Documento de Trabajo No. 132. Instituto de Investigaciones en Ciencias Econ6micas de la Universidad de Costa Rica. Sauma, P. and Vargas, J.R. 2000. Liberalizacidn de la balanza de pagos en Costa Rica: efectos en el mercado de trabajo, la desigualdady la pobreza. Documento mimeografiado. Seligson, Mitchell, Juliana Martinez y Juan Diego Trejos. 1997. "Reducci6n de la pobreza en Costa Rica: el impacto de las politicas p6blicas". En JosC V. Zevallos (editor): Estrategiaspara reducir la pobreza en Ame`rica Latina y el Caribe. Quito, Ecuador: Programa de las Naciones Unidas para el Desarrollo. Shorrocks, A.F, 1980, "The Class of Additively Decomposable Inequality Measures", Econometric, 48, pp.613-625. Shorrocks, A.F, 1984, "Inequality Decompositionby PopulationSubgroup", Econometric, 52, pp.1369-1385. Shorrocks, Anthony. January 1982, "Inequality Decomposition by Factor Components", Econometirca. SzCkely, Miqueland Hilgert, Marianne. November 2000, "What Drives Differences in Inequality Across Countries?" Inter-American Development Bank, Research Department Working Paper No. 439. Trejos, J.D. 1999, Reformas econdmicasy distribucidndel ingreso en CostaRica, Serie Reformas Econ6micas 37, Comisi6n Ecocn6mica paraAmCrica Latina y el Caribe, Santiago, Chile. Trejos, J.D. 2000, "Cambios Distributivos durante las Reformas Econ6micas en Costa Rica", pp. 473-556 in A. Ulate, editor, Empleo, Crecimiento y Equidad: Los Retos de las Reformas Ecdnomicasde Finales del Siglo XX en CostaRica, Editorial de la Universidade Costa Rica. 199 Trejos, J.D. 2000, "Reformas Econ6micas y Formaci6n de Capital Humano en Costa Rica", pp. 131-198 en A. Ulate, editor, Empleo, Crecimiento y Equidad: LQs Retos de las Reformas Ecdnomicas de Finales del Siglo XX en Costa Rica, Editorial de la Universidade Costa Rica. Trejos, J.D. August 2000, La Mujer Microempresaria en Costa Rica: Aiios 90", International Labor Office, Central American Project to Aid Programs for Micro-enterprises and Costa Rican National Institute for Women, Working PaperNo. 5, San Jose, Costa Rica. Trejos, Juan Diego, 2004, "Evoluci6n de la Equidad de la Inversi6n Social mblica desde 10s aiios Noventa", San JosC: Programa Estado de la Naci6n. Trejos, Juan Diego, Adrih Rodriguez, Inks Sfienz y Xinia Picado. 1994. La lucha contra la pobreza en Costa Rica: instituciones, recursos y programas. Documento de Trabajo No. 181. San JosC, Costa Rica: Instituto de Investigaciones en Ciencias Econ6micas de la Universidad de Costa Rica. Trejos, Juan Diego. 2004. La equidad del Gasto Social desde 10s aiios noventa. Ponencia preparadapara el DCcimo Informe sobre el Estado de la Naci6n en Desarrollo Humano Sostenible San JosC, Costa Rica: Programa Estado de la Naci6n en Desarrollo Humano Sostenible. Villasuso, J.M. 2000, "Reformas Estructurales y Politica Econ6mica en Costa Rica", in anabelle Ulate Quiros (ed.), Empleo, Crecimiento y Equidad: h s Retos de las Reformas Ecdnomicas de Finales del Siglo XX en Costa Rica, Editorial de la Universida de Costa Rica and the U.N. Economic Commission for Latin America and the Caribbean, San Jose, pp. 75-130. Viquez, Roxana. 2005. Sistema de Identijcacidn de la Poblacidn Objetivo: SIP0 en Costa Rica. Serie de Documentos de Discusidn sobre Protecci6n Social, No. 530. Washington DC, Estados Unidos de Amkrica: Banco Mundial. Wong, Gina and Picto, Garnett. 2001, "Introduction", in Wong and Picot, eds., Working time in Comparative Perspective, Volume I,Patterns, Trends, and the Policy Implications of Earnings Inequality and Unemployment, W.E. Upjohn Institute for Employment Research, Kalamazoo, Michigan. Yun, Myeong-Su. February 2002, "Earnings Inequality in the USA, 1961-1999: Comparing Inequality UsingEarnings Equations", mimeo, University of Western Ontario. 200 Annex 1 Correlatesof Povertyin Costa Rica 1989,1994 and 2000-2004 To compare the evolution of the correlates of poverty in Costa Rica, the exact same PROBIT regression, for urban and rural households, was run for 1989, 1994, and 2000 to 2004. As in the 2004 regression, presented in Chapter 2, the variables used in this analysis include: household size; selected characteristics of the household head, including education; the household's geographic location; average access to services in the vicinity of the household (as definedby the Primary Sampling Unit (PSU)'O1; the household's labor characteristics; and nationality.202 Household characteristics without enough variability were excluded from the analysis.203 Variable selection was also determined by theoretical considerations. Individual regressionswere estimated Urban or Rural householdsfor overall poverty. "Probit" regressions were used to estimate the relationship between poverty and the characteristics of individual households. For each household characteristic an "Exp(B) parameter was estimated. Each estimated "Exp(B) value is the ratio of the probability of being poor when the variable i s present to the probability of being poor without the variable. For example, an estimated value of 1.21 for the Exp(B) parameter associated to the variable "Female household head" means that the probability of being poor in a female headed household was 21 percent higher than that in a male headed household. Values above one represent an increase in the probability of being poor and below one a decrease in the probability of being poor204. The farthest away from one, the bigger effect the variable has. Exp (B) values of one (1) mean the variable has no impact on the probability of beingpoor. For further detail on the interpretation of the parameterestimates, see Box Al.l below. There are very few changes over the years in the significance of the estimates, the direction of the relationship to poverty (positive or negative) or the estimated value of the coefficients. Indeed, for statistically significant estimates, the direction of the relationshipnever changes for any of the 35 variables for urban or rural households (an impressive total of 490 estimates) - a very robust result. Also, the general rankingof the "grouped" variables - number of household members by age group (six variables) and education level (five variables) - i s the same over time (Tables A6.1, and A6.2). Nevertheless, some differences in the 2004 results are important to mention: (i)the originally weak relationship between female headedhouseholds and poverty found in 2004 in urbanareas is not present in any other year; (ii) incomplete primary education in rural areas is not significant in four of the other six years; and (iii)some urban regions, mainly the Chorotega region before '01A PSUconsists of a group of households within a censussegment and sampled as a unit in the EHPM. 202The only data difference was the lack of the Nicaraguanmigrant's households variable in 1989 and 1994. Since migrant population (from Nicaragua) in those years is believed to be low, no real impact on the estimates is expected. 203For example, dummy variables with average values below 4 percent or above 96 percent were excluded. Exceptions were made for groups of variables like geographic location andjob industry becauseexcluding individual categories complicates the interpretation of estimated parameters. 2MTo transform Exp(B) values above one to percentageterms: subtract one and multiply by 100Le. 1.5 is equivalent to (1.5 - 1.O)* 100= 50 percent increase in the probability of being poor. To transform Exp(B) values below one to percentage terms subtract Exp(B) from one and multiply by 100 i.e. 0.75 is equivalent to (1 - 0.75)* 100 = 25 percent reduction on the probability of beingpoor. 201 2001, are significantly associated with poverty (compared to the Central region). It is important to note that these differences in results over time may not all reflect real changes in the correlates of poverty over time. Some differences may reflect of real changes, while others might be a product of data weaknesses. Some tendencies in the changes of the estimated values were identified (increase in the "completed secondary" estimate value over time, and a decrease in the female household head estimate value over time). Given the limitations imposed by the quality of the data, such tendencies should be viewed as weak indications of possible changes in the way household characteristics relate to poverty, and in and by themselves is not proof of real changes. Box Al.l Technical note for the interpretation of the estimated parameter Exp(B) The estimatedparameter from the Probit regressions, Exp(B), is defined as: Where: P( 1) is the probability of being poor when the variable has a value of one, and P(0) i s the probability of being poor when the variable has a value of cero For interpretationpurposesi s easier to think in the change in the probability of being poor each variables ( has (in percentages): Change intheprobability of being poor (%) = 100 * P(1)-P(0) P(0) ) A very easy way to obtain the change in the probability of being poor i s to subtractthe estimatedparameter Exp(B) value from one (1) and multiply by one hundred: Change intheprobability of being poor (%) = 100 * (1 -Exp(B)) 100 (Exp(B)-1)=100 * * =Change inthe probability of being poor(%) 202 ij = * * * , = L L kklncu 0O0 0 0 - N I $ * : : : L L * * * * * * ::::: v: c E + + n U l n c o b c o ?kcqLnc\! 3 0 0 0 0 Annex 2 Entropy Inequality Measures General Formula Logarithmof the meandeviation E(O)+ a=O TheilIndex E(l)+ a= 1 Usingthe 1'Hospital rule we obtain Halfthe squaredcoefficient of variation E(2) + a=2 Decomposition of the Entropy Measures Any entropy measure of inequality can be decomposed into inequality between groups (IB) and inequality across groups (Iw)If we divide population into kjsubgroups, where j i s the number '05. of subgroups and p(y)j i s the mean of subgroup j, i s the measure for the inequality of subgroupj, fj = nj /n is the share of the population in subgroup j, and v =nj P (Y)j i s the nP ( Y ) share of the consumption of subgroupj, we obtain: With these definitions, entropy measurescan be divided inE(@=IsIw IB the + where reflects inequality given by the subdivision "k",and Iw i s the remaining inequality. *OsCowell andJemkins (1995) who were basedonpreviousworks of Bourguignon(1979), Cowell (1980) and Shorrocks(1980and 1984). 205 Annex 3 Poverty Headcount and MethodologicalIssues: A Sensitivity Analysis All poverty measurement is affected by methodological choices that in their own way influence who is ultimately considered poor and who is not. Technical and analytical decisions have to be made in computing an income (or consumption) aggregate, and in defining and estimating the poverty lines used for the poverty classifications. Judging the quality of each decision made i s not an easy task because one does not know the "real" underlying values with which to compare estimates (if there are any), and what is considered an appropriate measure in one country or in one moment in time within one country may not be considered so in another country or another timeperiod. Notwithstanding the problems associated with evaluating the quality of various well-being measures, there is a growing consensus about what constitutes good practice in poverty measurement, and comparisons with best practices along with sensitivity analysis to test the robustness of the analysis can be performed. Comparisons to similar estimations usingdifferent data sources (perhaps better, more complete data) also can provide some indications of the relative strengths and weaknesses of the measures being used. For example, as noted, the EHPM survey was designed originally as a labor survey and not in particular for poverty monitoring. Nonetheless, in 2005, the Costa Rican National Statistics Institute implementedthe "income and expenditure survey," a very complete household survey collected, on average, every 10 years and usedto update the Consumption Price Index. The income and expenditure survey is an excellent resource and can be used to evaluate the quality of existing poverty measures. Indeed, there would be significant benefits to using it to that end, as it could be used to help strengthen, as necessary, the quality of the yearly EHPMsurvey used to estimate and monitor poverty.206 This section examines the robustness of the current Costa Rican poverty measures to four different assumptions (or methodological considerations) used: (i) constant Engels coefficient207 used over time; (ii) different Engels coefficients between urban and rural households; (iii) different extreme poverty lines between urban and rural households; and (iv) different adjustment for underreported income between urban and rural households. Also, the combined effect of all considerations i s evaluated. The size and way to adjust for underreported income (one of the major limitations of the household survey) i s not evaluated because of the lack of information to formulate a sound alternative. Other minor considerations (the definition of household members, treatment of non-responsesto the questionnaire, etc.) are not evaluated here because no major impact on poverty i s expected. These aspects should be includedin a more complete study, perhaps once the income and expenditure survey data become available. It i s important to note that no claim is made here as to which i s the best way to do the various adjustments or to compute poverty. The objective of this exercise i s to evaluate how sensitive the poverty measures are to alternative and reasonable assumptions. The resultingpoverty estimates can not be characterized as better or more accurate, or assessed to be a better reflectionof poverty inCosta Rica. 206 The income and expenditure survey dataand supporting documentation were not available intime to be used in this study. *07 The Engels coefficient is the relationship between the value of the food part of the poverty line and the total value of the poverty line. 206 The Engels Coefficient To classify people or households as poor or extremely poor, income or consumption aggregates are compared to poverty lines. People are classified as extremely poor if their per capita income i s below the extreme poverty line, and as poor if it i s below the overall poverty line. Traditionally, the extreme poverty line is the cost to acquire a basic food basket with the minimum recommended amount of calories. The overall poverty line i s the extreme poverty line value plus an amount representing the cost of a minimumconsumption of non-food items (housing, clothing, etc.). There are no basic parameters to determine the non-food part value of the overall poverty line, as there are for the minimumrecommended amount of calories used for the extreme poverty line. "A common practice to determine the non-food component i s to divide the food component by some estimate of the budget share devoted to food." (Ravallion 1998). For example, the widely used poverty line for the United States developed by Orshansky (1963) assumes a food share of one third, which was the average food share in the U.S. at the time. Since the Engels Coefficient is the relationship between the food part of the poverty line (extreme poverty line value) and the total value of the poverty line the Engels coefficient for the U.S. example was 0.333. Use of constant Engels coefficients over time (constant proportional change over time), using the same Figure A3.1: Poverty Rates Updating Non-Food Engels coefficient over time Part of the Povertv Line with CPI produces poverty lines that do 32% not represent the same level of 3 30% ~ell-being.2'~ An alternative E 28% way to update the non-food 2 part of the poverty line would 26yo be to update its value using 24% changes in prices of the CPI P 22% non-food items. As can be 20% seen in Figure A3.1, estimated 1989 1994 2000 2001 2002 2003 2004 poverty levels would have been lower the assumption of a +Original +Updates using CPI *08Different Engels coefficients were computed for urban and rural households, but their values are the same over time. *09For example, if food prices increase by 25% and non food prices remain the same, the resultingpoverty line would be assuming a 25% increase in food and non food prices overstating the poverty line value and overestimating poverty. 207 over time had been relaxed. Indeed, by updating non-food prices with the non-food CPI, poverty in 2004 would have beenthree percentagepoints lower than under current, official estimates. Different Engels Coefficients between urban and rural households Different Engels coefficients were originally estimated for urban and rural households without any clear justification for doing so. The estimated Engel coefficient was 0.5076 for rural households and 0.4587 for urban households. It i s implied by the set values that the overall poverty line for the rural Figure A3.2: Poverty Rates with Same Engels Coefficient households was 1.97 times fnr All Hniiwhnlrlc the value of the poverty line, -- 36% and for urban households it was 2.18 times the value of 32% the extreme poverty line.210In E other words, the minimum C 28% z8 S non-food consumption for 24% urban households is bigger = 0) 20% than that for rural households. By doing so, a penalty was 16% imposed in rural households 1989 1994 2000 2001 2002 2003 2004 via lower minimum non-food 1 -Urban Original +Urban requirements, implying that -Rural Original +Rural Same hgels Same hgels rural households' needs are lower than those in urban households. An alternative approach would be to assume equal Engels coefficients for urban and rural households. At the national level, the alternative approach has no impact on overall poverty (Figure A3.2), but it would change the relative urban and rural poverty levels. Specifically, by assuming equal Engels coeflcients for urban and rural areas, poverty among urban households would decrease by an average of 1.8percentage points while poverty among rural households would increase 2.1 percentagepoints. 2101/ 0.5076 = 1.97 (rural) and 1/ 0.4587 = 2.18 (urban) 208 Different Extreme Poverty Lines between Urban and Rural Households Using different poverty lines for urban and rural households i s a common and valid practice if different patterns of food consumption are present or if urban and rural figure S3 PovertyRates With Same Extreme household face different food prices. Figure A33: Poverty Rates with Same Extreme Poverty Costa Rica is a small country with a Line for All Households well-developed road system, and the 34% difference in food consumption patterns between urban and rural households is not bigger than the 22% difference between regions or between poverty groups. Prices might be 16% different in rural areas, but the 14% 1989 1994 2000 2001 2002 2003 2004 estimates are not based in that -Urban Original -cUrbanSameEt.Line difference.211 A reasonable alternative -Rural Original -c-RuralSameExt.Line would be to use the same extreme poverty line for everyone in the country, regardless of rural or urban residence.212At the national level the new assumption has no impact on poverty, but, as with usinga single Engels coefficient across regions, it does affect rural and urban poverty levels; poverty in urban households would decrease an average of 2.3 percentagepoints, while poverty in rural households would increase by 3.0percentagepoints.213 Different Adjustment (UrbadRural)for underreportedincome Total income is adjusted to compensate for underreporting and other income not captured in the survey. The adjustment i s made by adding a constant percentage to the reported income. A different percentage is used between urban (17.4 percent) and rural households (35.8 percent). Different adjustment might bring the average income computed with the household survey in line with national account values, but for individual households and poverty estimates important biases are introduced: first, it i s a well known fact that incomes from national accounts and household surveys do n'ot match,214and second, by making different adjustments based on the place of residency, internal income distribution i s altered in different ways. A better way to determine the size of income adjustment could be obtained by a deeper analysis, for example a comparison of the 2004 Income and Expenditure survey and the 2004 EHPM. Since the 2004 211 The difference inthe extreme poverty line values is aproduct of pricingdifferent food basket. The same rices are used to estimate both food baskets. '12 A better estimate would be based inthe prices faced in eachregion (urbadrural), but that information was not available. 213 This and the prior change would bring estimated rural-urban poverty differentials more in line with those estimated elsewhere in the region. 214 A deeper analysis of the differences between national account and household survey income estimates is included on the Growth, Inequality and Poverty section in Chapter I. 209 I 1 I Income and Expenditure Figure A3.4: Poverty Rates with Same Income Adjustment Survey was not available, a fnr All Hniiwhnlrlq unique adjustment of 25 38% percent215 was used for all households. Poverty levels c 0) e 34% would be slightly lower under 30% C $ this assumption. Specifically 26% in 2004, poverty would have U8 22% been 1.2 percentage points I 18% lower than estimated, with a I I decrease of 2.7 percentage 14% points for urban households 1989 1994 2000 2001 2002 2003 2004 and an increase of 1.3 -Urban Original -cUrbanSameAdjustment -Rural Original -a- RuralSameAdjustment percentage points for rural households. (Figure A3.4). Combined effect of all the changes Combining the effects of changing all the assumptions outlined about would result in changes in the poverty estimates for Costa Rica, with the most significant effects coming with respect to the urban and rural poverty headcounts. At the national level the combined effect of all four simulations is an initial increase of poverty until 1994, and a small decrease of 2.6 percentage points in the 2000 years (Figure A3.5). For urban households, the effect i s a significant decrease inpoverty for all years by an average of 6.7 percentagepoints, and for rural household the impact i s an increase in poverty in all years by an average of 4.6 percentagepoints (Figure A3.6). Figure A3.5: National Poverty with All Four Simulations 35% , +Original &All Changes ~ 25 percent was the estimatednational adjustment value in the original reportused to determinethe urban andruralindividual adjustments. Sainz, Pedro.Comisi6nEcon6micaparaAmCrica Latina y el Caribe (CEPAL).Evoluci6nde laPobrezaen AmCricaLatina en 10s Afios Ochenta. 1991 210 FigureA3.6: Poverty Rateswith All Four Adjustments 42% 38% cQ, 34% 2 c 3 30% C 26% '0 8 22% I I8% 14% 10% I -Urban Original +Urban All Changes -Rural Original +Rural All Changes Indeed, changing the above-mentioned assumption leads to results that show a very different picture of rural relative to urban poverty in 2004. Inthe current (official) poverty estimates, rural poverty i s estimated to be only one-third higher than urban poverty (7.5 percentagepoints). With estimates revised on the basis of changes in the above assumptions, the results suggest that the incidence of rural poverty is two-and-a-half times that of urban poverty (19.4 percentage points). Policy recommendations would be very different under such revised scenarios. With the revised poverty numbers, policy recommendations would give more weight to the rural versus urban dimensions than the current numbers - and particularly to higher rural poverty. In the current numbers (and assumptions), regional differences (between Central, Chortega, etc.) appear to be more important than the rural-urban dimensions of poverty in the distribution of welfare across geographic areas. 211 Annex 4 Total Growth Elasticitiesof Poverty Reduction and the Efficiency of Growth The total growth elasticity of poverty i s commonly reported in the development literature as a measure of the poverty efficiency of growth. This i s defined as the percentage change in poverty for a given growth rate. Formally, denoting this elasticity byq, growth by g, and the log of poverty by P, q could be expressedas: Thus a higher q would give an indication of more effective poverty reducing growth. Intuitively there would be two routes through which poverty reduction performance could be improved: by achieving high growth rates for a given elasticity; andor by achieving a higher value (in absolute value) of q for a given growth rate. However, one has to be careful interpreting these figures. In fact, if we assume that income follows a lognormal distribution we can express: Thus poverty changes will be determined by the growth component 77, g and by the distribution component qGAG. It then follows immediately that the gross growth elasticity of poverty q can be rewritten as a function of the partial growth and inequality elasticities of poverty and of the observedgrowth and observedchanges ininequality: This expression can now beused to analyzehow q changes with AG and g. Consider for example the case of two economies (countries, states or regions) that are identical (the countries have similar values of 77, and qG so that differences in q will result from differences in AG and g. Assume also that over a given period of time, inequality changes in the same fashion in both places but the two economies have different growth rates (gl>g2>O). It is important to note that if AG>O, the total growth elasticity q of the economy with the highest growth rate will be smaller (higher in absolute value). Thus one would be tempted to interpret this as one state being more pro-growth and more pro poor, when the only thing that i s different in these economies i s the growth rate. Similarly, if AG