Report No. 28068-BO Bolivia Poverty Assessment: Establishing the Basis for Pro-Poor Growth December 15, 2005 Poverty Reduction and Economic Management Sector Unit Latin America and the Caribbean Region With Contributions From: UDAPE­Unidad de Análisis de Políticas Sociales y Económicas INE­Instituto Nacional de Estadística Document of the World Bank TABLE OF CONTENTS BOLIVIA 2005 POVERTY ASSESSMENT:MAINFINDINGSAND POLICY RECOMMENDATIONS ...........i INTRODUCTION............................................................................................................................................................ 1. TRENDSINPOVERTYAND INEQUALITY ........................................................................................ 1 IncomePoverty andInequality Poverty andThe IndigenousPopulation ..................... ................. .............................. 4 The Geography of PovertyandInequality............................................................. PovertyBeyondIncome ............... Who Are The Poor?IncomePove 2. GROWTH, POVERTY AND INEQUALITY ....................................................................................... 13 Patternsof Growth............ Sources andDeterminants The Impactof GrowthonPoverty andInequality.............................. FactorsDriving ChangesInIncomeDistribution 3. CONSTRAINTSTO EMPLOYMENT CREATION-THE DEMAND FORLABOR .................... 23 EmploymentCreation:ChallengesandOpportunities............................................................................... The ManufacturingSector: Characteristicsand InvestmentClimate ..... 24 CreationandTurnoverof ManufacturingEmployment ........................ Productivity andTechnicalEfficiency .................................... The DemandFor Labor......................................................................... Micro Constraints To CapacityUtilization andFirmSize ...... Policy InterventionsTo IncreaseThe DemandFor Labor ......................................................................... 37 4. CONSTRAINTSTO HUMANCAPITAL ACCUMULATION-THE SUPPLY OF LABOR ........41 Employment andEarnings: Trends andDisparities............... ................... The InstitutionsandPerformanceof Bolivia's Labor Market.................................................................... 46 Informal EmploymentandEarnings................................................................. Rural-UrbanMigrantsinThe Labor Market.................................................................. WhichPolicy InterventionsCanEnhance The Labor Market'sRoleinReducingPove 5. PROSPECTSFORGROWTHAND POVERTY REDUCTION ........................................................ 61 ................................................................................. 61 ................................... ........... BIBLIOGRAPHY ....................................................................................................................................................... 117 MAP .......................................................................................................................................... insidebackcover ANNEXES Annex 1.l: DataSourcesFor Monitoring PovertyandLiving ConditionsinBolivia..................................................... 69 Annex 1.2: The Geography of PovertyandInequality.............. ............................................ Annex 1.3: IncomePovertyandSubjectivePerceptions........... ............................................ Annex 2.1: TechnicalAnnex onDeterminants of Growth.......................................................................... Annex 2.2: ExplainingChangesinIncomeDistribution......... Annex 3.1: ManufacturingFirmSurveys................................ Annex 3.2: EstimatingRatesof Creation, TurnoverandReal1 Annex 3.3: Productivity andTechnicalEfficiency in Bolivian Manufacturing............... Annex 3.4: The DemandFor Labor .......... ............................................................................................................ 99 Annex 3.5: ExplainingThe Constraints To Capacity Utilizatio Annex 4.1: Determinants Of Earnings Differentialsin Bolivia Annex 4.2: EstimatingInformal-FormalEarnings Gap................................................................................................. 111 Annex 4.3: ExplainingThe LaborMarketPerformanceof Migrants ................................................................... 116 TABLES Table 1.1: PovertyMeasures. 1993-2002 ...................... Table 1.2: InequalityHas IncreasedinBo Table 1.3:Poverty According to UnsatisfiedBasic Needs ........................................... 8 Table 1.4: Bolivia's ProgressTowardKey Millennium Development Goals................................................................... 9 Table 1.5:Enrollment andAttendance RatesOf Bolivian Childre ............................ 11 Table 2.1: GdpPer CapitaGrowthinBolivia, LAC andThe Wor ........................................................... 14 Table 2.2: Growth Rate of Trend, Volatility andPersistencyof OutputPerCapita, Bolivia, LAC andThe World, 1960-1999 s to GDP Per CapitaGrowth in BoliviaandSelectedLAC Countries, 1971-2002......... 17 ........................................................................................................................................ 17 Table 3.1: InvestmentClimateIndicatorsof BoliviaandSelected LAC Countries ........................................................ 28 Table 4.1: Employment andEarningsRatiosBy Sector, MetropolitanAreas ... Table 5.1: Impactof EducationalUpgradingandEmploymentonWages...................................................................... 67 Table A 1.1.1:PovertyLinesinBolivia, 2002, Bs$Per PersodMonth . ............ Table A.1.3.1: Determinantsof Self-RatedPovertyinBolivia (18 Years Or 0 Table A.1.3.2: Determinantsof IncomeandSelf-RatedPovertyinBolivia(Hh)........................................................... 83 Table A.2.1.1: Determinantsof Differences inPer CapitaGDP GrowthinThe 1990s,BoliviaWith RespectTo Top LAC Performers ...................................................................................................................................... Table A.2.1.2: Determinantsof GrowthBetween 10Year Periods,Bolivia, 1970s-1990s ....................... ......87 87 Table A.2.2.1: Decompositionof ChangesinIncomeDistributioninBolivia, 1993-2002............................................. 89 Table A.3.3.1: Estimatesof The ProductionFunctioninManufacturing .............. Table A.3.4.1: Estimatesof ManufacturingUnskilledLabor DemandinBolivia. ................................ 100 Table A.3.5.1: Determinants of ManufacturingCapacity Utilization andFirmSi .................................... 106 Table A.3.5.2: QuantileRegressionDeterminantsof Capacity Utilization ....................... ..................107 Table A.3.5.3: QuantileRegressionDeterminantsof FirmSize ................................................................................... 108 FIGURES Figure 1.1: Open Unemployment RatesinBolivian CapitalCities. 1999-2002................................................................ 3 Figure 1.2: Change inHouseholdIncomePer Capitaby IncomeDecile andArea, 1999-2002........................................ 3 Figure 1.3: Gini CoefficientsinLAC, Distribution of EquivalizedHouseholdIncomecirca2000.................................. 4 Figure1.4:ConsumptionExpenditureandPovertyinBolivia, 2001 ............................................................................... 7 Figure2.1: GDP GrowthinBolivia, 1984-2003............................................................................................................. 13 Figure2.2: GDPGrowthby Sector andSelectedSub-sectors................ ............................................. 15 ....................................... 15 ....................................... ....................................... 19 25 Figure 3.3: Evolution andDistribution of ManufacturingOutput, Bolivia, 1995-1999.................................................. 9..................................... 25 26 Figure 3.4: Average LaborProductivity inThe ManufacturingSector, Bolivia, 1995-1999.......................................... 26 Figure 3.5: GrossEmploymentFlowsinThe ManufacturingIndustry, Bolivia, 1996-1999.......................................... Figure3.6: Bolivia HasHigher Ratesof Job Creation,Destruction& Reallocationby InternationalStandards ............31 32 Figure4.1: LaborMarket IndicatorsinBoliviaandLatinAmerica................................ Figure3.7: Job CreationandDestructioninBolivia's ManufacturingIndustryI s LargerinSmallFirms...................... 32 Figure4.2: Trends inLabor ForceParticipationin Bolivia 1989-2002........................... Figure4.3: Bolivia's DemographicTransitionandHumanCapitalAccumulation.............................................. Figure4.4: LegislatedEmploymentConditionsinBoliviaandLatinAmerica ............... Figure4.5: MinimumWage LevelsinBoliviaandLatinAmerica................................................................................. 47 Figure4.6: The Cost of Job Security, BoliviaandLatinAmerica, 1987-1999............................................................... 48 Figure4.7: Performanceof The LaborMarketinBolivia Comparedto LatinAmerica.................................................. Figure4.8: MigrationFlows inBoliviaby Origin andDestination, 1997 and2002, Heads of HouseholdOnly ............49 Figure4.9: MigrationFlowsin Boliviaby Origin andDestination, 1997 and2002, Heads andNon-Heads..................55 55 Figure 5.1: GDP Per CapitaGrowthProjections,Bolivia andLAC Average, 2000-2010 ........................... 62 FigureA.1.2.1: Incidenceof Povertyby Departamento, 200 FigureA.1.2.2: UrbanAreas: Incidenceof Povertyby Depa FigureA.1.2.4: Distribution of Poverty IncidenceandPoverty Gaps inMunicipalities................................................. 74 FigureA.1.2.5: Inequality inDistributionof ConsumptionExpenditure:Theil Index.................................................... 75 FigureA.1.2.6: Inequality and ConsumptionExpenditure ................ FigureA.1.2.7: Inequality andPovertyMeasures........................................................................................................... FigureA.1.2.8: Incidenceof Povertyby UBN and (High) Poverty Line ................................. ...................76 77 FigureA.1.3.1: Self-RatedAnd IncomePovertyProfiles FigureA.1.3.2: Impact of Regionof Residenceon Incomeand Self-RatedPoverty. ......................... 84 FigureA.1.3.3: Effect of Socioeconomic Characteristics FigureA.3.3.1: FactorElasticities inThe Bolivian ManufacturingSector ................... FigureA.4.1.1: Hourly EarningsDifferentials inBolivia, 1993-2002.......................................................................... 109 FigureA.4.1.2: Marginal Returns to EducationinBolivia 1990s....... ...................... 110 FigureA.4.2.1: Density of LogHourly Earningsby Sector,Met ....................... 112 FigureA.4.2.2: EarningsGap of BeingInformalinBolivia ....... ....................... 114 BOXES Box 1.1: IndigenousPopulationandPoverty................................................................................... Box 1.2: Bolivian SocialProtectionPrograms ................................................................................................. 10 Box 3.1: Simplifying Business RegistrationinLaPaz.................................................................................................... 29 Box 3.2: Constraints for Businessesto BecomeandRemainFormalinBolivia .... Box A.1.3.1: Explaining Incomeand SubjectivePoverty...................................... Box A.3.3.1: MeasuringFirmProductivity andTechnicalEfficiency ....................................... Box A.3.4.1: EstimatingManufacturingLaborDemand........................................ Box A.4.2.1: EstimatingSelectivity CorrectedInformal-FormalEarningGaps.............. Acknowledgements This report was prepared by a team that comprises Omar Arias (LCSPP, Task Manager), Wilson Jimenez (UDAPE), Fernando Landa (UDAPE), Magdalena Bendini (LCSPP), Mabel Martinez (consultant), Jessica Todd (summer intern, LCSPP); with background papers led by Leonard0 Gasparini (Universidad Nacional de la Plata, Argentina), Roberto Muiioz (University of Maryland), Alejandra Palma (University of Maryland), Walter Sosa (Universidad de San Andres, Argentina), Donald Pianto (University of Brasilia, Brazil), and Maria Tannuri (University of Brasilia, Brazil); contributions from Javier Monterrey (INE), Gustavo Canavire (NE) and Viviana Car0 (UDAPE); and the excellent assistanceof Lucy Bravo (LCSPP) and FerozamHashimi inthe preparation of the report. Jaime Saavedra Chanduvi (Sector Manager, LCSPP), John Newman (former Country Manager, LCCBO), and Vicente Fretes Cibils (Lead Economist and Sector Leader, LCCGA) provided overall guidance throughout as well as Norman Hicks (former Sector Manager for Poverty in Latin America) in the early concept stage. The peer reviewers were George Gray Molina (UNDP, Bolivia), Stephan Klasen (University of Gottingen), William Maloney (LCRCE), and Sergei Soares (LCSHD). Christopher Humphrey (LCCGA), Bruce Fitzgerald (LCCGA), and Vicente Fretes Cibils did a superbjob instreamlining the contents and messages of the final report. Carmen Romero (LCCBO) and Michael Geller (LCSPE) put together with excellence the final report. The team i s grateful for the very useful comments of Mauricio Santamaria (LCSPP), Luke Haggarty (LCSFR), Sara Calvo (LCSPE), Connie Luff (Country Manager, LCCBO), Carlos Mollinedo (LCCBO) and Neile Quintero (Country Economist for Bolivia, IDB) in discussions of earlier drafts of the report. And to Armando Ortuiio Yaiiez and Natasha Loayza from the UNDPoffice inLa Paz for granting access to the 2000 National Survey on Human Development Potential and Aspirations and insightful initial discussions. BOLIVIA 2005POVERTY ASSESSMENT:' MAINFINDINGS POLICYRECOMMENDATIONS AND ES.1 Bolivia faces high levels of persistent poverty and inequality. In 2002, 65 percent of the population was living in poverty and, of that, nearly 40 percent in extreme poverty. There was a decline inpoverty inthe mid-l990s, however, the rate today remains close to the level of the early 1990s. In addition, income distribution in Bolivia i s among the most unequal in Latin America. This report suggests three main reasons for the continuing high levels of poverty and inequality: 0 First,growth during the 1990s was concentrated in natural resource-basedexports, which have a relatively low demand for labor services. Labor-intensive sectors and poorer regions grew at a lower rate. The employment and income gains were insufficient to make a lasting impact on poverty reduction. Since 1999, negative shocks-such as the reversal of capital inflows, declining terms of trade and reduced exports, and the coca eradication program- reduced growth, reversing the earlier progress towards poverty reduction. Moreover, the high returns to capital compared to low returns to labor (especially unskilled labor) accentuatedthe already-high income inequality. 0 Second, the low productivity of firms, particularly in the informal labor-intensive sector, has held back the growth of both employment and wages. This has resulted from burdensome business and labor market regulations that discourage innovation and smaller companies from participating fully inthe formal economy, scaling up and improving productivity. 0 Third, the poor have inadequate opportunities to improve their human capital (e.g., through quality education, particularly secondary and above), despite recent progress in access to basic education. This results in low labor productivity and restricted access to better-paying jobs. Moreover, faced with high opportunity costs and inadequatesocial protection, many of the poor leave school early and end up working in low-paying urban jobs or in the rural economy. Those who do invest in education are often unable to translate it into large earnings gains due to limited learning outcomes by the poor (because of low quality schooling in poor areas, as well as health and nutrition deficiencies limiting early-childhood development) and limited access to the better payingjobs for their skills. ES.2 The main overall policy lesson is that broad-based economic growth, sustained over the long term, i s a fundamental and necessary condition to reduce poverty and inequality. However, this needs to be supported by policies to improve labor productivity and job creation. This can be accomplished through (i)removing obstacles to firm modernization and growth, and furthering their integration into the world economy; (ii) modernizing business and labor regulations, and providing the framework and incentives for firms to participate and remain in the formal sector, especially for small and medium firms; and (iii) strengthening human capital and social protection for the poor to enhance their productivity and ability to market their labor. 1. TheWorld Bank prepared this report with Unidad deAndlisis de Politicas Socialesy Econdmicas (UDAPE) of the Bolivian Ministerio de Desarrollo Econdmico and with the collaboration of the lnstituto Nacional de Estadistica (INE). i Establishiinathe Basis for Pro-Poor Growth I.TrendsinPovertyandInequality During the 1990s, poverty and inequality declined, but since 1999 most of the gains have been lost. ES.3 Duringthe period of growth, 1993-99, urban poverty declined from 52 percent to 46 percent, together with its intensity and severity. Urban income inequality remained unchanged. While adequate country-wide data i s not available, there i s evidence that both national andrural poverty may have also declined inthe 1990s. ES.4 The economic shocks of the late 1990s reduced growth and reversed progress on the poverty front. Between 1999 and 2002, poverty increased from 62 percent to 65 percent and extreme poverty rose from 36 percent to 37 percent. Poor urban households were particularly affected. Urban unemployment increased from 6 percent to 9 percent, and urban poverty reverted to the levels of the early 1990s. ES.5 Income inequality increased significantly during 1997-2002, and Bolivia currently has a Gini coefficient o f about 3,making it one o f the countries in the region with the highest income inequality, along with Brazil and Chile. The high income inequality reflects significant disparities in assets (e.g., education and land), household size, and earnings gaps by gender, ethnicity, location, and employment type. Nine of 10 Bolivians-poor and affluent alike-consider the distribution of income "unfair" or "very unfair". Poverty and inequality are everywhere, but the percentage rates and numbers of poor vary with location. ES.6 Bolivia's high poverty and inequality transcends rural-urban and regional boundaries. Although an overwhelming portion of the rural populace lives in poverty, there are also large pockets of urban poverty. Poverty i s concentrated inthe valleys and the central highlands, especially in Potosi and Chuquisaca, followed by Beni, L a Paz and Oruro. Santa Cruz and Cochabamba have lower poverty rates, but due to their large populations they contain a large number o f poor people. About 40 percent of the population inthe department of Santa Cruz i s poor, but the poverty rate i s only 20 percent in the capital. Many intermediate cities and small municipalities also have low poverty rates. Indicators of non-income poverty show more improvement than income poverty, but challenges remain. ES.7 Social indicators related to Unsatisfied Basic Needs (UBNs) and the Millennium Development Goals (MDG) improved significantly during 1992-2001. For example, child and infant mortality rates declined by 30 percent, net enrollment in primary education i s approaching 100 percent, and households without safe water and adequate sanitation fell from 50 percent to 30 percent. However, Bolivia still ranks among the worst inthe region in malnutrition, maternal and infant mortality rates, and i s off track to meet the MDG of universal completion of basic education. .. 11 Main Findinas and Policv Recommendations Peoples' self-perception of their well-being is generally aligned with income poverty. ES.8 Peoples' self-perceptions of their poverty and measures o f income poverty are both largely determined by employment, education, access to assets and basic services, ethnicity and location. Bolivians tend to fall into income poverty-and also consider themselves poor-when they are young, uneducated, unemployed or underemployed, indigenous, rural dwellers, and lacking basic services. ES.9 There are, however, some differences in income and self poverty perception surrounding ethnicity and location. Bolivian Quechuas tend to self-rate poorer than suggested by income poverty profiles, while the converse i s true for Aymaras. With equal access to basic services, rural residents perceive themselves as less poor than urban inhabitants, although they are more likely to be income-poor. Thus, exclusion andor cultural factors (e.g., sense of empowerment or identity) as well as location-specific characteristics (e.g., inequality, social capital, or crime) may have meaningful effects on Bolivians' self- perceptions. 11.Growth, Poverty and Inequality The growth of the 1990sreduced poverty and improved social indicators. ES.10 Growth averaged 4.7 percent per year (2.2 percent per capita) during 1993-98, surpassing other Andean countries for the first time in 40 years. Exports diversified beyond minerals and hydrocarbons to soybeans, coffee, sugar, wood. Investment levels exceeded 18 percent of GDP, four percentage points above the 1980s average. Sound stabilization and structural policies help to explain most of the 1990s growth expansion, which was accompanied by increased labor participation (especially among females) and improved overall productivity (1.2-1.7 percent per year). ES.ll This growth raised incomes (per capita income grew 13 percent cumulatively between 1993 and 1997) and reduced poverty. Specifically, the incomes of the poor in capital cities grew at the same rate than average incomes, and poverty declined from 52 percent in 1993 to 46 percent in 1999. In addition, improved rural education and living conditions suggest that poverty inrural areas may have also declined. ES.12 In spite of these gains, the severity of poverty meant that growth was insufficient to lift many Bolivians out of poverty. In 1997, 37 percent of Bolivians still were half-way short o f the income required to escape poverty. This implies that a 1percent increase in per capita income lifted less than half a percent of Bolivians out o f poverty-i.e., a poverty- growth elasticity inthe range o f 0.3-0.5, comparedto an average elasticity of 1for the region. Poverty reduction would have been greater if the growth had been sustained, more broad- based, and more labor-intensive. ES.13 Capital-and skill-intensive sectors (e.g., hydrocarbons, telecommunications and financial services) grew faster, with limited spillovers to agriculture and manufacturing, which employ over 60 percent o f the labor force. Exports diversified, but this did not ignite export-oriented labor-intensive sectors, and the export share o f GDP remained unchanged. 111 ... EstablishinatheBasisfor Pro-Poor Growth The growth in modem resource sectors-in the lowlands around Santa Cruz-did not spread as vigorously to subsistence agriculture and low productivity artisans, particularly in the Altiplano. ES.14 A key factor limiting development of more labor-intensive economic activity- necessary for job creation and poverty reduction-is low productivity, particularly labor productivity. Factors holding back productivity include (i) the ability o f firms and producers to adopt new technologies and production processes, train workers, and actively develop new products and markets-i.e., the demand for labor; and (ii)the ability of the poor to accumulate human capital and effectively utilize it inlabor markets-i.e., the supply of labor. 111. Constraintsto EmploymentCreation-the Demandfor Labor ES.15 Bolivia's weak business environment hampers investment, productivity and job creation. Total productivity gains in the 1990s largely reflected the improved resource allocation from economic reforms, rather than technology upgradingor innovation. Physical capital accumulation (linked to the adoption of new technologies) contributed little to growth, and overall labor productivity (GDP per worker) rose barely 0.5 percent per year during the economic boom, reflecting limited gains in labor productivity in the most productive sectors (e.g., petroleum, food and textiles). ES.16 Few small and medium enterprises (SMEs) grow larger. Small firms (10 or fewer employees) account for 83 percent of employment-largely unskilled-and 25 percent of output, while a few large firms (50 or more workers) generate two-thirds of output and only 9 percent of employment, largely skilled. In addition, labor demand in manufacturing firms takes over a year to adjust to changes in economic conditions, in the upper L A C range and 2.5 times the OECD average. Furthermore, there i s a high wage-employment trade-off. For every 10 percent increase inreal wages, demand for unskilled labor in manufacturing falls by 6.4 percent (twice the international average andinLAC'Shighend). A weak business climate hinders investment and job creation; small firms are most constrainedby regulations,contractreliability,creditandthin markets. ES.17 A small output market, burdensome regulations, limited and costly credit, and transport infrastructure limit the capacity utilization rates, prospects for expansion, and thus employment creation of manufacturing firms. For smaller firms, regulatory constraints (e.g., registration and operating licenses), high collateral requirements to obtain credit, and skilled labor bottlenecks are the most binding factors. For larger firms, input costs, including credit and access to technology, and market size represent the most binding constraints. More specifically: A thin, localized market. Domestic trade is highly concentrated and there are a limited number of firms exporting (for example, about 50 percent of large firms export, and only 20 percent of small-medium firms export). Burdensome business regulations and weak institutions. o Business registrationi s costly and subject to long delays, despite recent improvements. iv Main FhdinosandPoky Recommendations o Property is difficult to register and enforcement of contracts or property rights is uncertain. o Transactions and information costs are high, particularly with regards to credit, technology and information on foreign and domestic markets, accreditation, and disputes of contracts. Limited access to credit. The costly and high collateral required for lending, especially for smaller firms (over twice the loan amount, and mostly through real estate guarantees), reflects thin credit markets, ineffective assets registries, and insecure and costly debt recovery. A high cost of logistics and risky input and output market conditions. Supply chains are weak due to expensive and slow transportation (on a unit basis, 20 times costlier than in Brazil), customs dispatching remains cumbersome and costly, the quality of domestic services and inputs i s poor (e.g., unreliable supply of power), and as a result, inventorieshtock on materials are high (36 to 50 days). Restrictive labor regulations. Labor legislation (dating from 1943) mandates favorable conditions for workers compared to other countries in the region and elsewhere. This legislation-intending to protect workers-ends up increasing the total labor cost, making firms less competitive and discouraging equitable hiringin the formal market. As a result, it encourages informality, hindering productivity, andemployment creation. More specifically: o Uncapped severance payments lead to dismissal costs 2 to 3 times above most Andean neighbors andpoor countries inthe region. o Non-wage benefits (e.g., pension, health) are about 50 percent of labor costs. o Regulations restrict layoffs (including those due to economic shocks), seasonal work, overtime, and women's length of the work week and night work. Pervasive informality is interlinked with low productivity, reflecting high costs and low benefits of becoming and remaining formal. ES.18 Faced with few incentives to comply with regulations to start and run a business, many firms-particularly micro and small ones-remain outside the formal sector and lack access to formal institutions (e.g., credit and external markets). This restricts their potential for expansion as they cannot capitalize on productivity gains from innovation and economies o f scale. More specifically, e Smallerfirmsface low benefits andperverse tax incentives to become and remainformal. o Regulatory costs for the incorporation of partnerships are high. o Firms can not deduct valued-added taxes on purchases from firms registered in the simplified tax regime (SII), comprising most small firms. e One-size-fits-all labor regulations are not conducive for improving productivity in smaller firms. o Mandatory non-wage benefits (primarilysocial security) are costly to small firms and can amount to up to 8 percent of sales. This creates incentives to remain in the low productivity informal sector and thus forego access to formal institutions of credit, V Establishingthe Basis for Pro-Poor Growth training and exports. Even small firms registered to operate and report taxes often do not abide by labor laws. IV. Constraints to Human CapitalAccumulation-the Supply of Labor Getting a high quality education is harder for the poor, and labor market returns to education are not equal. ES.19 The combination of high opportunity costs and low returns to education discourage children from poor families to stay in school. The public education system, especially at the secondary level and in rural areas offers low quality education, limiting the capacity o f the poor to accumulate human capital and improve their earnings opportunities. Further, poor families face highopportunity costs and are often unable to afford keepingtheir children in school, insteadneeding them to help the family, either through income-generating activities or domestic and agricultural chores. ES.20 Returns to education are low-six out of ten graduates from high school are at risk of poverty because of these low returns. Inrural areas, only a post-secondary education offers a significant boost to earnings. As well, education does not carry equal returns for all workers. Education returns range from 0 to 60 percent for the least to best paidworkers with primary education, 20-30 percent for those with a secondary education, and from 50 to 150 percent for the college-educated. Workers from poor families tend to receive lower returns to education, due to limited learning outcomes by the poor (because of low quality schooling in poor areas, as well as health and nutrition deficiencies limiting early-childhood development) and limited access to the better paying jobs for their skills. The employment gaps faced by women, young and low educated workers and the earnings disparities solely relatedto gender, ethnicity, location and employment sector are above regional averages. Low opportunity costs for self-employment and non-wage benefits of informality encourage large employment inthe informalsector. ES.21 Bolivia's urban informal sector i s large and heterogeneous. In 2002, more than 55 percent of the labor force was in the informal sector, either as self-employed (40 percent) or salaried workers (15 percent). An additional 10 percent of workers were unpaid, principally working infamily businesses or are apprentices. ES.22 Informal employment largely reflects the low opportunity costs and non-income benefits of informality. For many Bolivians it offers a competitive alternative to low- productivity formal sector jobs or no work at all. As well, self-employment may be more attractive to certain sectors of the population, such as women seeking flexible work hours to balance their work and family obligations, or the indigenous who may face less discrimination as an independent worker than they might as an employee in a company. In fact, the self-employed perceive themselves as less poor than salaried workers with similar characteristics, an indication of the importance o f non-monetary benefits of self-employment. ES.23 Due to the low productivity of workers in the informal sector, the informal salaried workers do appear to have a significant earnings disadvantage when compared to salaried formal sector workers with the same skills andjob characteristics, particularly those vi Main Findinasand Policy Recommendations at the bottom of the salary scale. In part this i s due to the lower access of informal firms to programs to promote worker training, technology adaptation, or other kinds of productivity- enhancing interventions. Migration has improved earning opportunities but has had limited impact on reducing poverty. ES.24 While there has been some migration from less developed to more developed regions, it has remained small among the rural poor. Over the mid-l990s, urban areas attracted migrants largely from rural areas. Migration to main urban centers, especially from small cities, accelerated since the late 1990s. There is also significant migration to rural areas experiencing economic booms, especially inthe region around Santa Cruz. ES.25 For the rural-to-urban migrants, earnings were improved by migrating, particularly for those at the bottom of the earnings scale. That is, despite a potential lack o f contacts and urban know-how, migrants got competitive urban jobs. Thus, rural-urban migration likely helped to reduce poverty directly and possibly indirectly through remittances. ES.26 However, small migration flows limit the potential of migration to be an escape valve for the rural poor. About 350,000 people migrated during the 1993-1997 boom, of which only 69,000 were rural-to-urban migrants. Individuals from the poorest locations and indigenous household heads are more prone to rural-to-rural migration. The young, more educated, women, and small families are more likely to migrate to urban areas. As a result, while urban and rural labor markets seem interlinked, the absolute magnitudes of cross flow remain small. This partly reflects the high costs and possibly non-pecuniary factors that affect settlement decisions. V. Selected Policy Recommendationsto ReducePoverty andInequality ES.27 Restoring sustained economic growth and facilitating the development of labor- intensive sectors are essential to reduce Bolivia's poverty and inequality. Bolivia can have higher growth in the medium and long term tied to the development of the gas sector. However, to have a significant impact on poverty reduction, this must be accompanied by a policy environment that promotes, among other things, broader investment, increased productivity andjob creation. ES.28 To achieve these, policy reforms should focus on (i) removing obstacles to firm modernization and growth, and furthering their integration into the world economy; (ii) modernizing business and labor regulations, and providing the framework and incentives for firms to participate and remain in the formal sector, especially for small and medium firms; and (iii) strengthening human capital and social protection for the poor to enhance their productivity and ability to market their labor. Specific policy recommendations are elaborated below.2 2. These recommendations are complemented by other recent or ongoing reports, including the Investment Climate Assessment (Report No. 24746-BO, October 18, 2001), the Public Expenditure Review (Report No. vii EstablishiflatheBasisfor Fro-Poor Growth Removing obstaclesto firm modernization and growth and furthering their integration into the world economy. Measures that would help inthis area include: Simplifying procedures and lowering the cost of business registration, especially in large municipalities, scaling up IFC-supported efforts in this direction in L a Paz to other local governments. Implementing incentives (for example, limited, tax credits) to adopt new technologies, including in manufacturing not just hardware and software but also management techniques andworker training, andinagriculture small-scale ruraltechnology andnew crop varieties. Promoting expanded access to prudent financing for SMEs, among other ways through the overhaul o f laws on collateral. Increasing participation in world markets, in particular through enacting free trade agreements that will deepen exports and promote investment andtechnology transfer. Encouraging the creation of producer/exporter associations to reduce the cost of information to take advantage of trade and other market opportunities. Modernizing business and labor regulations, and facilitating formal sector participation. Firms would greatly benefit from general improvements in the investment climate, including: Reducing the cost of registration and business expansion for micro and SMEs, particularly the cost o f incorporation of partnerships, registering in the General Tax Regime and export licensing. This could be accomplished by further rationalizing documentation requirements (e.g., notarization) and government fees, and streamlining on-line business portals for registering andlicensing inmunicipal government offices. Establishing pilot initiatives that provide small firms incentives to become formal, encouraging small firms and producers to bid for government contracts (the Presidential Decree No. 27328 "Compro Boliviano" lays the legal basis), extending partial credits of value added taxes for eligible firms, and offering business development services (access to market credit, judicial services, management and accountingpractices) with special emphasis on supporting innovation initiatives and export production. Bolivia could benefit from the success experience of small business promotion agencies such as those in Chile, Italy, and the US. Streamlining labor regulations that currently limit the ability o f firms to expand and contract along with the economic cycles to align them with international practices, and reducing the cost of mandated labor benefits, which currently total about 40 percent of labor costs. Simplifying, reducing the cost, and increasing the transparency o f government bureaucracy procedures required to access technology, quality certification, accreditation, and dispute resolution. Strengthening institutions and coordinating relevant public agencies to reduce duplication and transactions costs, particularly the Superintendency of Enterprises and the Labor Ministry. 28519-BO, November 18,2004), the forthcoming Country Economic Memorandum(analyzing other constraints to growth) and the forthcoming Education Sector Study. V l l l ... Main Findinasand Policv Recommendations Strengthening human capital and social protection for the poor to enhance their productivity and ability to market their labor. This can be achieved through measures such as: a Raising the quality of the education system, particularly for the poor. Implement an education sector strategy geared towards developing basic cognitive skills and improving the productivity o f the labor force. Key elements of this strategy, which will be addressed in greater detail in an upcoming Bank education sector study, might include: o Filling coverage gaps in universal basic education, improving secondary education transitions and access to private higher education for poor students, and addressing low quality and inequalities ineducation achievement at all levels with results-based management, especially in municipalities with weak education outcomes. o Implementing a conditional cash transfer program (similar to Bolsa Familia in Brazil or Oportunidades in Mexico) that provides incentives for very poor families with children-at-risk to keep them in school and use preventive health and nutrition interventions. Any such program should be developed with medium-term fiscal constraints inmind, such that it can be sustained. The cost of similar programs inthe region range from 0.5 to 1percent o f GDP. a Zmproving labor market equity and opportunities. This could be achievedthrough: o Reducing obstacles to employment by expanding pre-school facilities and child care centers to facilitate women's and migrants' labor force participation and fostering community-led crime prevention in marginal urban neighborhoods to allow workers, especially women, to take advantage o f available job opportunities. o Training in high schools and colleges on relevant skills demanded in the labor market and encouragement of privately provided labor market intermediation services. o Using the newly developed consumption poverty map to target interventions aimed at income generation of the poor. These interventions may include growth- enhancing investments, targeted programs to develop human capital, community assets and income generation, and investmentsthat promote gradual integration of communities through migration. o Strengtheningpro-poor community investments andworkfare programs. VI. Prospectsfor Growth andPoverty Reduction ES.29 Bolivia's growth prospects remain vulnerable to domestic instability and external circumstances. Nevertheless, the country can improve its future growth potential through deepened economic reform. Simulations, that include some of the recommendations suggested above, indicate that if the policy determinants o f growth were to improve significantly, GDP per capita growth could be sustained at, about 4 to 5 percent per year. Because of the depth and breadth of poverty in Bolivia, and the skewed income distribution, these rates are necessary for the medium and longer term if the country's poverty level i s to be significantly reduced. Indeed, the national MDG target of reducing the incidence of extreme poverty in half by 2015 could be achieved with growth rates in this range, along ix EstablshinatheBasisfor Pro-Poor Growth with other pro-poor policy interventions. However, an even higher annual per capita growth rate i s needed to meet the MDG of reducing poverty in half. Economic simulations indicate that individual policy reforms will have relatively small impacts on growth and poverty by themselves, but can have a much larger impact when implemented as part of a comprehensive strategy of mutually-reinforcing reforms that includes macroeconomic stability. X Main Findinas and Policv Recommendations xi INTRODUCTION 1. Bolivia experienced important economic, political and social changes in the 1990s. Macro-economic stabilization inthe late 1980swas followed by market reforms to deregulate the economy, liberalize trade, simplify taxes, reform the pension system, privatize non-performing public companies, and decentralize public resources to municipalities. As part of the Heavily Indebted Poor Countries (HIPC) initiative in 2000-2001, the country developed its national Poverty Reduction Strategy (BPRS)with broadparticipation of different sectors and donors. 2. Bolivia's reform efforts swiftly paid off, with high rates of investment and growth. The economy expanded at an average annual rate of 4.7 percent (2.2 percent per capita) during 1993- 98, Exports diversified, social spending increased substantially, and living conditions improved, particularly education, health and other Millennium Development Goals (MDGs) indicators. However, progress in many areas was limited and not sustained. After several external and internal shocks in 1999, growth decelerated to an average rate o f only 1.7 percent (0 percent per capita) during 1999-2002. Fiscal imbalances and financial sector difficulties weakened macroeconomic stability, reducingjob creation andpoverty reduction. Bolivia today remains one of the poorest and most unequal countries inthe region. 3. Inthis context, the Bank incollaboration with Social andEconomic Policy Unit (Unidad de Analisis de Politicas Sociales y Economicas,UDAPE) and the National Institute of Statistics (Instituto Nacional de Estadisticas, INE) started the preparation of the Bolivia Poverty Assessment. Following the Government's request, the report focuses on selected knowledge gaps, identified with UDAPE and INE in consultation with other partners (IDB, UNDP), of the factors preventing more pro-poor growth inBolivia, andpublic policies to address these factors. 4. Since late 2003 the country situation changed dramatically. Discontent and social unrest over the economic situation and the gas export policy forced former President Gonzalo Sanchez de Lozada to step down. Since President Carlos Mesa took office, there have been some positive economic signs. However, political stability remains fragile. The upcoming constitutional assembly to reassess the economic policy framework and the implementation of the recent natural gas referendum pose serious challenges. There are concerns that social and political polarization may lead to diverging priorities that might be difficult to reconcile in Bolivia's multi-cultural population. 5. While the new political situation imposes new challenges and short term priorities, the generation o f income opportunities for the poor remains a critical issue. The findings and policy recommendations of this report aim to strengthen the analytical basis for the formulation of sound public policy for poverty reduction. To this end, the report examines the links between the evolution and determinants o f poverty and inequality in the 1990s, the sources and patterns of growth, andpersisting micro constraints to employment creation inBolivia. 6. This report i s structured as follows: 0 Chapter 1 examines the historical trends, characteristics, sources and determinants o f economic growth, and the evolution of poverty, inequality and the MDGs inthe 1990s 0 Chapter 2 analyzes the determinants o f changes in the income distribution in the 1990s, the profiles/determinants of self-rated and income poverty, and a detailed geographic characterization of poverty and inequality xi EstablishimtheBasisfor Pro-Poor Growth 0 Chapter 3 discusses the characteristics of the labor market, the supply o f labor, employment and earnings outcomes, with emphasis on the informal sector and internal migration 0 Chapter 4 examines the micro constraints to employment creation, productivity and labor demand inthe manufacturing sector 0 Chapter 5 assesses the prospects for pro-poor growth in Bolivia, particularly policy lever simulations for achieving the MDGpoverty target. 7. The work builds on previous analyses of poverty issues in Bolivia and several strands of analytical work in and outside the Bank, including: (i) previous Bank reports on poverty in Bolivia (World Bank 2000, 1996) that analyze poverty profiles, human development, and access to social infrastructure, credit and land; (ii)the 2001 report on Microeconomic Constraints to Growth inBolivia that thoroughly analyzes the constraints that Bolivia's investment climate pose to competitiveness; (iii)the 2004 Public Expenditure Review, which analyzes the fiscal sustainability o f major reforms, including the pension, civil service and education and health reforms, as well as the efficiency and equity o f social programs; (iv) Bolivia's national Ruraland Agricultural Development Strategy focusing on farm and non-farm production and other rural development issues, with technical support of the Bank; (v) a recent Bank study on poverty and nutrition defining the nature and extent o f the malnutrition problem in the country, the limitations o f current interventions and actions to improve immediate and long-term nutritional results; (vi) recent economic and sector work on the reform of the health sector and education, including support to the development of Bolivia's new education strategy and recent technical analyses for the development of recent Bank operations inthe social sectors ;(vii) several studies by the Bank, other partners and researchers, including recent regional studies (growth, indigenous people and poverty) and flagship reports (inequality, education and technology), IDB's 2003 Economic and Social Progress Report focused on employment inLatin America and UNDP's country HumanDevelopment and MDGReports. xii 1. TRENDS POVERTY INEQUALITY IN AND During the 1993-99period of economic growth, urban poverty fell from 52 percent to 46 percent and its intensity and severity alsofeel moderately. Although hard data do not exist, there is also evidence that rural and national poverty may have also declined during this period. The economicshocks and growth stagnation afer 1999 reversed any progress. During 1999-2002, poverty increased from 62 percent to 65 percent, and extreme poverty rose slightly. Poor urban households were particularly afSected. Income inequality increased significantly, and Bolivia currently has a Gini coeficient of about 0.58, making it one of the most unequal countries in the region. Poverty rates are highest in the valleys and central highlands, although urban areas contain the largest numbers of poor due to their high populations. Unsatis$ed Basic Needs and Millennium Development Goals (MDGs) indicators improved significantly during 1992-2001. Child and infant mortality rates declined by over 30 percent, net enrollment in primary education is approaching IO0 percent, and households without safe water and adequate sanitationfell from 50percent to 30 percent. There are some digerences in income and self-rated poverty, particularly regarding ethnicity and location, which call attention to non-monetary determinants of well being. Bolivian Quechuas tend to self rate poorer than suggested by incomepoverty profiles, while the converse is truefor Aymaras. 1.1 Poverty in Bolivia i s extremely high and has proved stubbornly difficult to reduce. In 2002, an estimated 65 percent of the Bolivian population were poor (with incomes insufficient to cover the basic food and non-food expenditures), while 41 percent lived in extreme poverty (incomes too low to afford the food basket of minimumcaloric intake). Bolivia i s also one of the most unequal countries inthe world, with most of the population living on very low incomes and a small elite controlling much of the country's wealth. This chapter traces poverty andinequality through the growth in the 1990s and the recent economic slowdown, discussing both national and municipal-level trends. It examines non-income poverty measures and concludes with an analysis of the determinants of the Bolivian poor's attitudes toward poverty. INCOMEPOVERTY AND INEQUALITY 1.2 Poverty in the 1990s. Available data (for departmental capital cities, Table 1.1) show that the growth episode in the 1990s ledto a decline in income poverty from 52 percent in 1993 to 46 percent in 1999, while the fraction of the population in extreme poverty decreasedfrom 24 percent to 21 percent.' A recent study (Klasen et al., 2004), based on improved demographic, education and living condition indicators, finds that poverty levels may have also improved in rural and small urban areas. However, due to fast population growth, the number of poor inlarge cities increasedfrom 1.5 million in 1993 to 1.7 million in 1997 while the number of indigent rose from 700,000 to 730,000. 1.3 The poverty gap (PG), an indicator of the intensity of poverty, aIso improved very little. This represents the percentage by which the average poor person's income falls short of the poverty line. This gap i s two times higher in rural areas than in urban areas, and four times 1. Lack of consistent, nationally representative household surveys for the early 1990s precludes establishing a national trend inpoverty and inequality. Annex 1.1discusses methodological issues. 1 EstablkhinatheBasisfor Pro-Poor Growth higher in the case of the extreme poverty gap. On average, the incomes o f the rural poor cover less than half of the basic consumption basket, while the extreme poor fall 43 percent short of the cost of the minimumfood basket inrural areas. Table 1.1: PovertyMeasures, 1993-2002 OfficialRates Poverty ExtremePoverty Poverty H I PG H PG Ti-pG- 7ExtremePoverty T p E - National 1997 63.6 33.7 36.5 18.9 Jational 1999 62.0 30.7 35.8 15.0 1999 63.5 36.0 40.7 22.2 2000 65.5 33.7 39.2 17.2 2002 65.2 36.7 41.3 22.3 2001 64.4 31.8 37.3 19.6 Capital 1993 52.0 22.2 23.7 8.4 2002 64.6 31.2 36.8 14.4 cities* 1997 50.7 21.0 21.3 7.5 Jrban 1999 51.4 22.4 23.5 8.9 1999 46.4 18.8 20.7 7.02 2000 54.5 25.6 27.9 11.0 2002 51.0 22.1 23.5 8.8 2001 54.3 24.6 26.2 14.7 Urban 1997 54.5 23.8 23.8 9.1 1999 51.4 22.4 23.5 8.9 2002 53.9 23.8 25.7 9.4 2002 53.9 23.8 25.7 9.4 :ural 1999 80.1 44.8 56.7 25.4 Rural 1997 78.0 49.4 59.0 34.4 2000 84.5 47.7 58.7 28.0 1999 84.0 59.4 69.9 45.1 2001 81.1 43.6 55.6 21.8 2002 83.5 57.6 67.0 43.3 2002 82.2 43.4 54.8 22.6 Note: H: headcount (% of population), PG: poverty gap (% Note: Basedon householdper capita expendituresinrural gap between average poor person's income and the poverty areas and householdper capita incomes for urbanzones. line). *Capital cities include: Sucre, La Paz, Cobija, Cochabamba, Oruro, Potosi, Tarija, Santa Cruz, Trinidad and ElAlto Source: UDAPE, based on household surveys, EIH (1993); ENE(1997); andMECOVI (1999-2002). 1.4 Poverty since 1999. The earlier improving trends in poverty have reversed since 1999. By 2002 poverty levels in the departmental capital cities went back to the level of the early 1990s. Rural poverty, particularly extreme rural poverty, showed a more positive trend between 1999 and 2002. Official poverty rates that rely on household expenditures for rural areas showed a slight increase in total rural poverty, and a two percentage-point decline in extreme rural poverty (Table 1.1).This may partly reflect rural-urban movements and within-rural migration to areas with expanding agricultural production (e.g., soy beans, livestock). 1.5 Interms of absolute numbers, most poor people live inurban centers. Over half the poor (2.9 million) and 43 percent of the extreme poor (1.4 million) lived in urban areas in 2002, up from one third (1.8 million) and one fourth (800 thousands) in 1997, respectively. On a percentage basis, however, poverty remains significantly higher in rural areas. In 2002 about 84 percent of rural inhabitants were income-poor (2.7 million) and 67 percent (1.8 million) were in extreme poverty, compared to 78 percent (3.3 million) and 59 percent (2.3 million) in 1997. 1.6 The recent economic slowdown put further stress on the employment and incomes of the poorest families, particularly in urban areas. The open unemployment rate in capital cities went up from 6 percent in 1999 to 9 percent in 2002, with a larger increase among women (Figure 1.1).The rise in unemployment affects more urban workers from the poorest families, who are twice as likely to be unemployed andtake longer to find ajob than the average worker. In2002, 2 Figure 1.1: OpenUnemploymentRates 60 percent of workers from the poorest inBolivianCapitalCities, 1999-2002 20 percent of families took at least two months to find a job, compared to 40 percent in 2000. An increasing fraction of workers cite factors related to the 10 - economic downturn (e.g., being fired, lack of sales or providers) as the main reasons 8 - for being unemployed during 1999-2002. There is also evidence of discouraged k 6 - workers leaving the labor force (the participation rate declined from 68 4 - percent to 65 percent). Underemployment has also become more prevalent. 2 Informality comprises over 80 percent of jobs in the poorest quintiles, and over one 0 -- third of those employed in 2002 were willing and available to work longer hours Total Men Women -upfromonefourthin2000. Source: UDAPEand INE,basedon MECOVIsurvey data, 1.7 The negative impacts on household incomes have been uneven. percent of Bolivians consider the current 10% income distribution "unfair" or "very 8% unfair" (Latinobarbmetro, 2001). The 6% standard measure o f inequality i s the Gini 1 4% coefficient, which compares the actual 290 distribution o f income to a hypothetically 1 Fs 0% equal distribution. It ranges in value from -2% zero (complete equality of income) to one -4% -690 (when all income is concentrated by only 1 1 2 3Deciles 4 5 6 7 8 9 10 of percapitai n c o m the top of the range, only below Brazil and Source: Author's estimates based on householdsurvey data. Chile, countries that nevertheless have an 3 EstablishinQ theBasisfor Pro-Poor Growth income per capita four to five times higher. A recent Bank regional study o f inequality in L A C finds that this high income inequality reflects significant disparities in assets (education, land, housing), household size, and earnings differentials by gender, ethnicity, sector and type of employment (De Ferranti et al., 2003). Figure1.3: GinicoefficientsinLACDistributionof Table 1.2: InequalityhasIncreasedinBolivia, EquivalizedHouseholdIncome, Circa2000 as Shownby Ginicoefficients 60 Equivalized Earnings Hourly Income Wages Capital Cities* 1993 50.34 51.89 54.01 1996 49.05 52.18 54.28 1997 50.37 52.58 54.27 2002 53.94 55.40 57.00 Urban 1997 50.52 52.64 54.25 2002 52.59 54.48 56.13 I I National *Source: Gasparini et. a1.(2004). 1997 56.85 56.28 58.07 2002 58.80 58.40 59.24 Note: * Capital cities include the capitals of the nine departamentos and ElAlto City. *Source: Authors' estimates based on household surveys. 1.9 Income inequality increased since the late 1990s. While there are different bases and measures of inequality, this result holds regardless the indicator used. For example, Table 1.2 presents the Gini coefficient based on equivalized household income (which adjusts for differences in the composition and size of households), total earnings (labor sources) and for hourly wages. The Ginis remained constant between 1993 and 1997 inthe main urban areas then increased from 1997 to 2002.2 The increase in income inequality mainly results from a less equal distribution of earnings (labor income represents around 90 percent of total income - cf. Chapter 3). POVERTYAND THE INDIGENOUSPOPULATION 1.10 Poverty i s linked to the situation of the indigenous population. Despite progress in political representation and some social indicators, the Bolivian indigenous people are still severely affected by poverty and exclusion (see Box 1.1). In2002 there were 1.4poor indigenous for each non-indigenous poor-up from 1.3 in 1997. 1.11 Low education levels are a principal factor behind the high incidence of poverty among the indigenous. However, even with the same level of education, indigenous people are less likely to exit poverty, in part because their returns to education are lower than for the rest of the population. The indigenous tend to be overrepresented in low productivity jobs both in the formal and informal sector. They are also concentrated in the disadvantaged central highlands and the valleys. The slow pace of improvement in this situation has contributed to recent social tensions. 2. The differences are statistically significant. See the backgroundpaper by Gasparini et al. (2004) for details. 4 Box 1.1:IndigenousPopulations and Poverty [nthe last 50 years, the indigenouspopulation (defined by spoken language) has grownfrom 1.7 to 3.9 million, and IOW represents half of the country's population. Their economic and political influence increased, as shown in the ,ast general elections, when parties representing the indigenous obtained 26 percent o f the seats in the parliament. The reforms of the last decade, especially the Law of Popular Participation and the Administrative Decentralization, recognized the juridical representation of the indigenous community and made resources wailable for social investment in municipalities with majority of indigenous population. Despite changes favoring :hem, the findings of a recent study by Jimenez and Landa (2004), summarized below, show that indigenous people are still severely affected by poverty and exclusion. Despite recent improvements, educational opportunities are less favorable for the indigenous population. The average education o f Bolivians (over 15 years old) i s 7.5, but only 5.9 years for the indigenous population :ompared to 9.6 years of education for the non-indigenous. The rate of illiteracy for indigenous people between 15 and 69 years old is 8 percent in urban areas and 25 percent in rural areas. Non-indigenous participation in secondary and tertiary education i s double that of indigenous people. Interms o f health, the indigenous population has disproportionately higher mortality rates and endemic diseases. Low educational attainment is related to child work. Child labor is more common in poor indigenous households than in non-indigenous ones. In2002, 31percent of indigenous children aged 9-11were working, four times more than non-indigenous children. The probability o f deserting school and working is higher for indigenous children, but householdeducation (particularly, mother's education) and income levels tend to reduce this. Non-indigenous labor income in 2002 was 2.2 times higher than that o f indigenous workers (for women, this ratio was 2.4). Analysis of the determinants of these wage differentials can be largely explained (up to 70 percent) by the lower human capital of indigenous workers. Returns to education are also lower for indigenous people. Labor market participation rates in2002 were higher and unemployment lower among indigenous workers, probably due to their concentration in rural areas and traditional agriculture. About 84 percent o f indigenous workers have an informaljob compared to 67 percent for non-indigenous workers. About 80 percent of indigenous women were self-employed while self-employment among non-indigenous women is 53 percent. Being indigenous limits the chances o f obtaining paid work, particularly inentry-level jobs. Source: Based onJimenez and Landa (2004a). THEGEOGRAPHY POVERTYAND INEQUALITY3 OF 1.12 UDAPE and INE, with the support of the World Bank, recently developed a consumption-based poverty map that provides poverty and inequality measures for disaggregated geographic units, including municipalities. This allows identification o f the areas with the highest concentration of poverty and inequality, usingtechniques that combine information from the 2001 Census and MECOVI household surveys 1999-2001 to generate new data on poverty- based consumption expenditures and three alternative poverty lines (see Annex 1.1 for methodological details). 1.13 While poverty i s widespread, patterns differ. The complete results are described in Annex 1.2to this chapter, andthey may be summarized thus: 3. This section summarizes the results o f the report Pobreza y Desigualdad en Municipios de Bolivia: Estimacidn del Gasto de Consumo Combinando el Censo 2001 y las Encuestasde Hogares (2003) led by UDME and INE and from the World Bank by Quentin Wodon (AFTPM), Werner Hernani (consultant), and Peter Lanjouw (DECRG) . 5 EstablishinatheBasis for Pro-Poor Growth Potosi and Chuquisaca have the largest poverty incidence, followed by Beni, L a Paz and Oruro, regardless of the poverty line used. The departamentos of Santa Cruz, Pando, and Cochabamba are the least poor in percentage terms at any poverty line, although even in Santa Cruz 40 percent of the population i s poor. However, the departamentos with lower poverty incidence concentrate the largest number of poor in absolute terms owing to their higher population density. The urban conglomerates of Beni andL a Paz present the highest levels of poverty (over 50 percent for the low poverty line) while Cochabamba and Santa Cruz are among the least poor (just under a quarter o f the population). Urban development in these two cities has created broad-based income opportunities, Rural areas all over the country are overwhelmingly poor (with no significant difference between low and high poverty lines), and incidence o f extreme poverty almost as high as total poverty. Rural areas of Cochabamba and Santa Cruz have a poverty incidence as high as that o f rural Potosi and Chuquisaca. Poverty i s widespread in a large number o f municipalities, both interms of magnitude and intensity and regardless of the poverty line. Many municipalities exhibit poverty above the national levels: a significant fraction above 80 percent. Poverty i s particularly concentrated in the valleys and the central highlands. In at least 20 municipalities with dispersed populations (e.g. Morochala, San Pedro Buena Vista, Ravelo), most residents are unable to cover basic food needs. 0 Bolivia also exhibits high levels of inequality in consumption at the local level. While in the more egalitarian departments (Tarija, Pando, and Beni) inequality i s mainly due to rural-urban disparities, in the most unequal (Potosi, Cochabamba, and Chuquisaca) inequality i s also pervasive within urban and/or rural localities. 0 The current system of inter-municipal transfers based on UBNpoverty targeting tends to be less responsive to municipalities with entrenched pockets of income poverty. 1.14 The results highlight the core connection between increases in average consumption (such as would occur with growth), poverty and inequality at the local level: poverty declines as consumption increases, but the association i s mediated by inequality. Municipalities with an average consumption far below the national average register poverty above the national level (Figure 1.4). These are typically located in more economically dynamic urban areas. At high levels o f consumption, very few municipalities have poverty above the national average. However, just below the average per capita expenditure level, several municipalities perform below and above national poverty levels. This reflects differences in inequality and particularly inpoverty intensity among localities. Inmunicipalities with a smaller poverty gap an increase in per capita consumption leads to a larger reduction in poverty, while in those with high poverty intensity a significant rise in per capita consumption makes a smaller headway towards reducing poverty. 1.15 Many localities may be caught in poverty traps (Azariadis (2004). The largest group of localities comprises municipalities with very high poverty and low inequality which are mostly sparcely-populated, remote indigenous communities living in subsistence. The second largest group includes those with both high poverty and inequality where small wealthier groups coexist with large poverty pockets and comprise larger urban localities with better resource endowments 6 Trendsin Poverty andIneaualitv and small cities dedicated to mineral exploitation or border trade with Brazil and Argentina. In manyof these localities poor basic infrastructure, costly access to markets, harshnatural resource endowments, low returns to human capital, and ineffective protection from natural and idiosyncratic risks, can all prevent very poor households from engaging in higher yield economic activities and long term investments in human capital. Low incomes also limit the demand of local goods and services. Raising consumption levels through growth alone may be insufficient and targeted interventions may be needed to increase the incomes of the poor inthese areas. Figure1.4: ConsumptionExpenditureandPovertyinBolivia,2001 (a) Low poverty line PovertyAnd LevelsOf Consumption Poverty And LevelsOf Consumption Incidenceof Poverly(low) PovertyGap(Low) (b) Extremepoverty line Source: Basedondata fromthe 2001Census and HouseholdSurveys 1999-2001. 7 Estabfishinathe Basisfor Pro-Poor Growth POVERTYBEYOND INCOME 1-16 A complementary approach Table 1.3: Poverty Accordingto Unsatisfied Basic Needs to income poverty measures i s to (% individuals) with unsatisfied basic needs (UBN) 1992 2001 National Urban Rural National Urban Rural and progress in human development outcomes. In the past decade, Bolivia made substantial progress in 70.8 living conditions, access to basic 43.7 91.2 social services and Social indicators. As a withthe UBN fraction Of Note: The Unsatisfied Basic Needs index is computed as the average PoPulation fell of four indices for housing, sanitation, education, and health. significantly from 1992 to 2001 Source: INE,basedon 1992 and 2001Census data. (Table 1.3). However, large gaps in UBNremaininrural areas andmany municipalities. Table 1.4: Bolivia's Progress Toward Key MillenniumDevelopmentGoals Source: UDAPE,2003. 1.17 There was also significant improvement in education, health and other social indicators linked to the Millennium Development Goals (Table 1.4). National child and infant mortality rates declined by 30 percent, net enrollment rates in primary education reached almost 100 percent and households without safe water and improved sanitation fell from 50 percent to 30 percent over the decade. However, malnutrition, and maternal and infant mortality rates remain 8 Trendsin Povertv andIneauaitv among the highest in the region and there are high levels of malaria, chagas, and tuberculosis. Bolivia still ranks low incoverage of basic services such as water, sanitation, andelectricity. The country i s not making sufficient progress to meet the MDG of universal completion of basic education, and progress in some indicators has been uneven. For example, infant mortality rates fell by 27 percent for the richest households and only a 5 percent for the poorest (WDI, 2003). Access to health services across socio-economic groups and indigenous and non-indigenous populations i s very unequal. Box 1.2: Bolivian Social Protection Programs Social protection programs, designed to help the extremely poor cope with risks associated with a drop in income (such as disease, financial crises, or natural disasters), have grown in size and innumber inBolivia inrecent years. As a percentage of GDP, spending on social protection has tripled since 1999, rising from just under 1 percent o f GDP to 3 percent in 2003. A good portion o f this increase is attributable to the BONOSOL program of cash transfers to all Bolivians over 65. Not including BONOSOL, public social protection spending stands at 1.7 percent o f GDP in 2003. Among the many programs forming part o f the Bolivian social protection network are those aimed at literacy, childhood development, health insurance, preventable disease, vulnerable groups (particularly indigenous people and women), poor neighborhood improvement, and provision of drinking water. Evolutionof Social ProtectionExpenditure I _ _ _ . _ _ 1 3 5 7 - - 3.0% 2.5% 2.0% 1.5% 1 0% 0 5% 0 0 % 1999 2000 2001 2002 2003 (P) 0 SP BSP (Without BONOSOL) The public sector plays a key role inthe provision o f social protection programs for the vast majority o f the population. Approximately 60 to 70 percent of SP programs (without the BONOSOL program) are funded through external credits, which raise concerns on the sustainability of the Bolivian SP system. With the exception o f the BONOSOL and the temporary employment program PLANE, municipalities administer the expenditures o f most programs, meaning that municipalities have a critical role to play in program implementation and in coordinating intergovernmental transfers. See Bolivia-Public Expenditure Review (PER) for a catalogue o f Bolivia's social protection programs. Although the amount o f fiscal resources spent on social protection is not excessive ina regional context and considering the needs o f the Bolivian population, social protection programs are generally not well-managed or targeted to make the most of scarce funding. Furthermore, major risk groups are not covered, for example youth and adults without health insurance and the vulnerable agriculture producers. As well, social protection spending shows an imbalance between the high spending for certain groups and limited resources destined to bigger groups facing higher risks. For example, BONOSOL expenditures are very high compared to the low funding directed toward the very successful SUM1 mother-child health insurance program and the PAN childhood development program, Following the recommendations of the 2004 PER, the Government i s developing a social protection system that will embody the vast variety of current programs backed by different sectors of government and different funding sources. The aim is to improve coordination, to standardize information an uniform criteria to implement and evaluate programs, and most importantly to increase efficiency. Source: UDAPE,UPF, INE and VIPFE 9 Establisbina theBasisfor Pro-Poor Growtb 1.18 Improvements in social indicators in part result from increased social expenditures and improved service delivery linked to decentralization (see Box 1.2). Expenditure decentralization started in late 1994 with the transfer of primary'responsibility for planning and implementation of public investment to the prefectures and municipalities, and expanded with the 2001National Dialogue Law. Social spending reached 18.5 percent of GDP (second-highest in LAC) in 2001 from 2.5 percent in 1986. Decentralization, backed by growing tax revenues, led to greater human capital and social investments in localities with high illiteracy and malnutrition and low water and sewerage c~nnection.~ 1.19 However, the recent economic stagnation may compromise improvements in social indicators and the achievement of MDGs. While both school enrollment and attendance rates have risen overall, they fell for children o f the poorest families (Table 1.5). 1-20 The increaseinthe use of trained Table 1.5: Enrollment and Attendance Rates of personnel and health facilities for infant and Bolivian Children (% of those age 5-12) maternal care during the 1990s slowed down inrecent years, and remains low despite free Enrollment Attendance access to the poor given through the publicly Quintile 2000 2001 2002 2000 2001 2002 provided basic health insurance program Poorest 83.9 82.9 80.1 83.5 81.9 79.2 (Seguro Universal Matemo Znfantil- 2nd 84.6 86.1 88.7 83.6 85.5 87.9 SUMI). 3rd 86.6 88.4 88.3 85.6 87.4 87.6 WHOARE THE POOR?INCOME 4th 90.8 89.2 91.5 89.8 88.2 90.8 POVERTY AND SUBJECTIVE PERCEPTIONS Richest 94.1 95.6 95.4 93.3 94.6 95.4 Total 87.1 87.5 87.8 86.3 86.6 87.2 1.21 While measuring poverty through income and expenditure poverty lines i s widely accepted, a complementary approach is to examine people' own subjective perceptions of their poverty. Since monetary poverty indicators are imperfect proxies of welfare, the priorities o f the population with regards to improvements in their well-being may not match those arising from income/expenditure poverty profiles. The determinants of people' self-rated poverty are valid considerations for public policy, and understanding them may shed light on factors underlying recent social demonstrations and lead to a more widely accepted poverty reduction strategy. A background study for this report (Arias and Sosa, 2004) analyzes these issues going beyond income/expenditure poverty profiles to examine the determinants of Bolivians' subjective perceptions o f well-being. The results are discussed inAnnex 1.3 of this chapter and are summarizedbelow. 1.22 The study finds that subjective poverty perceptions inBolivia are consistent with income metrics and lead to similar conclusions on who the poor are and the main determinants of poverty. Employment, education, access to assets and basic services, ethnicity and location are core determinants of both income poverty and self-rated poverty. Bolivians tend to have a greater likelihood of falling into income poverty or to consider themselves poor when they are younger, have low education, are unemployed or underemployed, have an indigenous heritage, live in rural areas, and lack basic services. Education and employment carry a greater weight in the poverty perceptions of Bolivians who are more leaning towards self-rating poor, which suggests 4. See World Bank (2004), Faguet, J. (2004) andBossert (2000). 10 Trendsin Poverty andIneaualitv that idiosyncratic factors are less important in the poverty perceptions of the less educated and the unemployed. Indigenous populations weigh unemployment more heavily than the non- indigenous in their self-rated poverty. This implies that actions to increase incomes and employment go hand-in-hand with those that affect Bolivians' own perceptions of well-being. 1.23 Some differences in income and self-rated poverty rankings call attention to non- monetary determinants o f well-being. Bolivian Quechuas tend to self-rate poorer than suggested by income poverty profiles while the converse is true for Aymaras. The self-employed report themselves less poor than salaried workers with similar characteristics, while individuals out of the labor force self-rate as poorer than those employed, despite their similar income poverty. Rural residents no longer self-rate poorer than urban inhabitants if they have equal access to basic services and similar socio-economic conditions, but remain more likely to be income-poor. Although Chuquisaca is the second income-poorest region, its residents self-rate the least poor in the country. Thus, exclusion andor cultural factors (e.g., sense of empowerment or identity), job flexibility, as well as location-specific characteristics (e.g., inequality, social capital, crime) may have meaningful effects on Bolivians' poverty perceptions. 1.24 All in all, income emerges as a sensible proxy measure of welfare and its determinants. Bolivians of all ethnic backgrounds care about education, employment and livingconditions, and the self-rated poor place an even higher priority on improvements in these factors. Programs to train, educate, or generate employment among the poor not only bring important income gains but are also valued by these populations. Nonetheless, the differences in subjective welfare rankings for various segments of the Bolivian population should remind policy makers of the significance of non-monetary aspects of welfare for these groups. 11 2. GROWTH, POVERTY AND INEQUALITY The growth spell of the 1990s reducedpoverty and improved social indicators. During 1993-98, growth averaged 4.7 percent per year (2.2 percent in per capita terms), surpassing other Andean countriesfor thefirst time in 40 years. Sound stabilization and structural policies accountedfor much of this growth spurt, aided by an expansion of the labor force and increases in productivity. As noted in Chapter 1,this led to a reduction in poverty, but this reduction was insuficient to pull large numbers of thepoor above the poverty line because of the large initial poverty gap. As well, growth was greater in capital- and skill-intensive sectors, with limited spillovers to agriculture and manufacturing, which employ over 60 percent of the labor force. As such, the impact of growth on poverty was reduced. A key factor limiting development of more labor- intensive economic activity was low productivity, particularly labor productivity. Productivity, in turn, was held back byfactors impacting: (i) the ability offirms to adopt new technologies, train workers, and actively develop new products and markets-i.e., the demand for labor; and (ii) the ability of the poor to develop human capital and efectively utilize it in labor markets-i.e., the supply of labor. 2.1 Over the last 40 years growth in Bolivia has been slow and did not have a significant impact on poverty and inequality. Even during the boom of the 1990s, growth was faster in sectors with lower direct job generation and small spillovers to the rest of the economy. The weak improvement in growth-enabling factors prevented Bolivia from realizing greater benefits from policy reforms. With the growth deceleration of the early 2000s, poverty gains were reversed, inequality rose, and social gains slipped. 2.2 This chapter examines the trends and determinants of economic growth and its implications for the evolution of poverty and inequality. It characterizes the patterns o f growth inthe 1990s and early 2000s, disaggregated by sector and region. It performs a simple growth accounting to assess the contributions of capital accumulation, labor force expansion, and total factor productivity to growth inBolivia. PATTERNSOF GROWTH 2.3 After macroeconomic stabilization Figure2.1: GDP GrowthinBolivia, 1984-2003(%) in the mid-l980s, Bolivia adopted numerous market reforms to increase l 6 5 private sector participation, align prices 4 with market forces and increase 3 integration into the global economy. The 2 Bolivian economy recovered from the 1 contraction o f most of the 1980s, and 0 expanded at an average annual rate of 4.7 percent (2.2 percent per capita) during 1993-98 on the basis of far-reaching 1 `*a4885 1988 1S7 888 1989 1SO 891 1992 1983 894 895 19% 897 898 19% 2000 2001 ZQ22003 -1 reforms and abundant external capital Source: Based on data from INE. (Figure 2.1). 13 Establishina theBasis for Pro-Poor Growth 2.4 However, due to negative external and internal shocks, the economy decelerated to an average rate of only 1.9 percent (close to 0 percent in per capita terms) during 1999-2003. The Russian financial crisis of 1998 and the international capital markets' subsequent turmoil ledto a sudden drop in external financing to Bolivia and many o f its trading partners, with a consequent deceleration of regional growth. In addition, the Brazilian and Argentine currency devaluations and ensuing crises in 1999 and 2001, respectively, caused a decline in both the competitiveness and demand for Bolivian exports. Internally, the coca eradication reduced incomes, and it was only partly compensated by the increase in soy production and new gas reserves exploitation. These factors, combined with the 2002 elections and social unrest, resulted in production disruptions, fiscal imbalances, financial sector troubles, capital outflows, and growth deceleration. 2.5 Bolivia's growth has trailed the average of the rest of the world Table 2.1: GDP Per Capita GrowthinBolivia,LACand the World, 1960-2002(%) (Table 2.1). GDP growth per capita was below LAC'S regional average 1961-70 1971-80 1981-90 1991-99 2000-2002 during the 1960s and 1970s and -0.20 declined further during the 1980s. It Andean 2.09 -1.18 0.77 0.06 Region* followed the regional trend in LAC** 3.44 -0.74 2.05 0.26 recovery during the 1990s, but still World*** 4.15 2.68 2.29 1.72 2.70 remained 0.5 percentage points Note: *Simple average, own calculations; **weighted average, n = below the LAC average during the 26; ***weighted average, n = 109. GDP measured at $1995 decade. growth than the other Andean Source: Based onLoayza, Fajnzylber and Calderon, 2002 and data Bolivia has had lower purchasing power parity. 2002 preliminary estimates. countries on average, except during fromWDI (2003). the 1990swhen its economy-grew twice as fast. Figure 2.2: GDP Growth by Sector and Selected Sub-sectors, Figure2.3: Bolivia's Sector ( Impositionof GDP, 1990-2002 Bolivia, 1993-2002(%) 1990 2003 11 10 12 .I 10 2 8 3 6 4 2 9 3 ' 0 IAGR BMIN BMAN OF,GS CON PCM I T & C MFlN -2 OS&PS IR&H NPA -4 Nore: See definitionsinFigure2.2 Source: Based ondata fromINE. * Preliminary. Note: AGR: agriculture; MIN: mining; MAN: manufacture; E.G.W: electricity, gas, and water; CON constmction; C M commerce; T&C: transport & communication; FIN:finance; S&PS: social and personal services; R&H restaurants &hotels; PA:public administrative services. Source: Basedondatafrom INE. 1. See World Bank (2004b) for a detailed discussion of Bolivia's macroeconomic evolution. 14 2.6 Capital and skill intensive sectors grew faster, with limited spillovers to agriculture and manufacturing (which employ over 60 percent of labor). The fastest growing sectors were construction, utilities, minerals (including hydrocarbons and natural gas), financial services, transport and communication, which expanded over 6 percent annually. Agriculture and manufacturing grew below the pace of the economy (Figure 2.2). In particular, the coca eradication program reduced production by 80 percent, hurting many Bolivian farmers. Agricultural expansion in other crops-led by soybeans-was insufficient to absorb the displaced agricultural laborers. 2.7 The composition of the economy shifted moderately towards services (Figure 2.3). Financial, business and real estate services increased their GDP share from 10 to 13 percent between 1990 and 2003. Agriculture remained at 15 percent of GDP. Industrial agricultural products grew over 10 percent per year but account for only 18 percent o f agricultural GDP. Hydrocarbons and minerals remained at around 10 percent o f GDP: exports of natural gas to Brazil rose, but that only served to offset production declines in other minerals. The share of manufacturing fell slightly from 18 percent in 1990 to 17 percent in2003, ledby food processing industries which account for more than half of manufacturingoutput. 2.8 Exports, as a percentage of GDP, had little change, although their composition shifted. The share of minerals and hydrocarbons in total exports dropped from about 95 percent in 1985 to about 50 percent in 2002, while non-traditional products (e.g., soybeans, coffee, sugar, wood) rose to 46 percent, largely due to the significant increase in soybeans. This structural shift was facilitated by intra-regional trade pacts (e.g., the Andean Pact and Mercosur) during the mid- 1990s. Thus, exports diversification has not ignited a dynamic export oriented model that can add value to natural resource sectors. 2.9 Economic growth did not affect all Figure 2.4: Average GrowthbyDepartamento, regions equally. The departamentos that Bolivia 1990-98 (percent) began with the highest per capita GDP 8 generally grew faster (Figure 2.4). An 7 I exception i s Pando, which began as number five in GDP per capita, but grew the fastest during 1993-98. The poorest departments had below-average growth and remain below the national income per capita. 2.10 In sum, the growth spur of the 1990s was unbalanced and not sustained. The growth in natural resource sectors-in the lowlands around Santa Cruz-did not spread sufficiently to subsistence Note: Numbers in parentheses refer to each department's agriculture-in the Altiplan-and low GDPper capitarankingduring the 1990-1993period. productivity artisans. A dynamic export Source: BasedOn data from INE. model that adds value to natural resource sectors has not yet ignited. 15 EstablihinnstheBasis for Pro-Poor Growth SOURCESAND DETERMINANTSGROWTH OF 2.11 Bolivia was a leader in economic reforms among the Andean countries during the 1980- 90s. Reforms took place in two stages.2 The first began in 1985 and aimed to stabilize the economy and develop the private sector. The public sector was reduced and monetary expansion was eliminated as a source of government financing, paving the way for the greater market- determination of prices. The second stage o f reforms took place in the early 1990s and emphasized policies to attract foreign direct investment (FDI), improve the efficiency and efficacy of social programs and public services, strengthen the financial sector and promote decentralization to strengthen the country's governance and institutions. These reforms were accompanied by a large amount of external financing. 2.12 The evidence from a recent study (Loayza, Fajnzylber and Calderon, 2002) suggests that the growth spur of the 1990s was mostly driven by structural rather than by cyclical factors. First, Bolivia registered an acceleration of the long-term (trend) component of GDP per capita, coupled with relatively smaller short-term fluctuations around this trend (that is, lower volatility and persistence of shocks). Second, total factor productivity increased significantly. And third, regression results from the study indicate that improved structural and stabilization policies explain a significant fraction of the growth during the 1990s. 2.13 For the entire 1960-90 Table 2.2: GrowthRate of Trend, Volatility' and Persistency2 period, Bolivia's growth rate of of Outputper Capita, inBolivia, LAC and the World long-term (trend) per capita 1960-99 output i s close to actual growth 1960-99 1960-70 1971-80 1981-90 1991-99 rates, and also below that of the Growth of Trend world, L A C and the Andean Bolivia 0.27% 0.45% 1.20% -1.97% 1.54% region. But during the 1990s it Andean Region" 0.75% 1.50% 1.85% -1.04% 0.76% matches LAC'S average and LAC** 1.42% 2.61% 1.62% 0.07% 1.51% World*** 2.35% 3.31% 2.51% 1.64% 2.00% doubles that o f the Andean Volatility countries. The volatility of Bolivia 0.0220 0.0384 0.0127 0.0147 0.0106 Bolivian per capita output during LAC** 0.0290 0.0208 0.0356 0.0314 0.0167 the past two decades was below World*** 0.0224 0.0195 0.0263 0.0213 0.0149 the L A C region and the world. Persistence While economic shocks were Bolivia 0.1911 -0.0406 0.78523 0.33924 0.0256 common during the 1970s and LAC 0.19853 0.0354 0.1641 0.23163 0.27253 1980s, the persistency of growth World**** 0.16113 0.18003 0.135g3 0.17603 0.14933 fluctuations was lower than that Note: * Simpleaverage, own calculations; **weighted average, n= 26; of L A C and the world during the *** weighted average, n= 106; **** weighted average, n=136. 1990s (Table 2.2). Standard deviation of cyclical component of log output per capita. 'AR(1) coefficient of persistency; 3significant at 5%; 4significantat 10%. 2.14 The significant increase Source: Based on Loayza, Fajnzylber and Calderon, 2002. intotal factor productivity (TFP) also corroborates that growth inthe 1990scannot be attributed to cyclical factors. The growth rate of TFP in Bolivia surpassed LAC'S average in the 1990s (Table 2.3). 2. World Bank (2004b). 16 Table 2.3: Factor Contributionsto GDPPer Capita Growth, inBolivia and selected LACcountries. 1971-2000 ~ Method1 I 1 1 1 1 Method2 Method3 Period1 Country I ,s;,",'th ILabor Capital TFPl I 1Labor Capital TFP2 Labor Capital TFP3 ~ Note: * Countriesshown had the decade medianannualTFP growth rate in the LAC region. TFPl assumes that capital growth is investment minus depreciation and the growth rate of the labor force. TFP2 adjusts for the educational attainment of the labor force. TFP3 adjusts for capital and labor utilization (employment rate, population employed and total hours). NA: Not Available. Source: Based on Loayza, Fajnzylber and Calderon, 2002. 2.15 Results of cross-country growth regressions indicate that Bolivia's growth acceleration in the 1990s i s largely attributed to better policies (Loayza et al. 2002). Improved structural and stabilization policies account for over 80 percent of growth in Bolivia in the past decade (see Annex 2.1 for details). Rebounds from past negative shocks (cyclical reversion) and external conditions had a smaller role. 2.16 Remaining weaknesses inthe policy framework as well as Bolivia's distinctive barriers to growth in good measure explain why Bolivia has been a growth under-performer. When comparing the growth drivers of Bolivia and high-growth LAC countries, Bolivia was outperformed in per capita growth on the basis of better policy performance o f the other countries (see Annex 2.1 for details). In fact, being a much poorer country Bolivia should have grew over 1 percent faster than most of these countries, according to transitional convergence theory. Country-specific characteristics explain from 0.7 to 2.4 percentage points of Bolivia's growth underperformance. These relate to factors such as natural resources, geography, the quality of the investment climate and political institutions, all of which limit the Bolivian economy's ability to reap greater benefits from structural and macro policy reforms. While these traits are challenging they can be circumvented with good p ~ l i c i e s .In ~ particular, despite recent progress, the institutions governing the investment climate in Bolivia fare behind most L A C countries. Bolivia's idiosyncratic characteristics cause it to grow 0.7-0.9 percentage points slower than Colombia and Peru, countries with their share of natural hurdles and social conflict butwith better andimprovingbusiness regulations. 2.17 The international evidence concurs that sound business regulations, including those for the entry-exit and expansion of firms and labor laws, are essential for productivity growth. They set the incentives for firms' learning and innovation, and enable the reallocation of resources to more productive uses. In fact, Bolivia's productivity gains o f the 1990s largely reflect the 3. See Sachs (2003),Rodrik and Hausman (2003). 17 EstablihinatheBasisfor Pro-Poor Growth improved allocation of resources ensuing from economic reforms but little technological upgrading or innovation. Capital accumulation (that often accompanies adoption o f new technologies) contributed little to growth, while labor productivity (GDP per worker) rose barely 0.5 percent per year during the growth boom. Chapters 3-4 discuss how a difficult investment climate, particularly weak business and labor regulations, constrain growth inBolivia. THEIMPACT OFGROWTH POVERTYAND INEQUALITY ON 2.18 Several factors dimmed the impact of growth on poverty in the 1990s. Bolivia's large income gap separating the poor from the non-poor (partly a result of past low growth) i s primarily responsible for the limited poverty reduction. This results in a low growth-poverty elasticity such that growth does not have a large, immediate impact on poverty. Moreover, the slower growth in the labor-intensive sectors and in the worse-off regions, as well as scant labor productivity gains, preventedlarger increases inemployment and earnings. The improvements in education and other social indicators, though certainly positive trends, take longer to translate into better income generation in an environment that i s not conducive to productivity growth. Altogether this ledto modest absolute income gains for the poor. 2.19 Growth in the largest urban centers was not biased against the poor. Per capita incomes of the metropolitan poor increased at a similar rate than average incomes (around 3 percent), although the richest 20 percent saw their incomes rise slightly faster. Figure 2.5 shows curves depicting the average annual growth rate of each income percentile during 1993-97 for capital cities and for national, urban and rural areas during 1999-2002. The fact that the poor benefited equally in relative terms from average growth i s reflected in the curve lying over the horizontal line in the top figure. While lack o f comparable household survey data for rural areas precludes definite assessments, the study by Klasen et al. (2004) suggests that growth did not bypass the poor inrural and small urban localities. 2.20 Note, however, that the resulting income gains for the poor are much smaller in absolute terms since they start from very low incomes. As a result, 1percent growth in per capita income lifts less than half a percent of Bolivians out of poverty (poverty-growth elasticities are 0.3-0.5, compared to an average elasticity o f 1for LAC). With the large number o f poor people and the severity o f poverty, the 13 percent cumulative per capita income growth of 1993-97 improved incomes, but was unable to pull a larger number of people in major urban areas below or even closer to the poverty line. About 37 percent of the population in 1997 still fell half way short of the income required to escape poverty, and 59 percent in rural areas could not even afford the basic food basket. 2.21 The results also corroborate that the economic slowdown led to large and relatively constant income declines (except for the richest 10 percent) in urban areas and slight relative income gains for the poorest rural households. Accounting for the recent real income declines, the incomes of the urban poor fell -0.8 percent per year during 1993-2002 18 Growth, Povem, andIneaualifv Figure 2.5: Growth Incidence Curves, 1993-2002 (% changeinhouseholdper capita income, by percentile) CaDital Cities 12 9 6 3 0 -3 -6 -9 +1993-1997 1999-2002 -12 ' 1 1 11 21 31 41 51 61 71 81 91 Percentile of the Household Income P e r Capita National, Urbanand Rural, 1999-2002 12 10 8 6 4 2 0 -2 -4 -6 -Rural -Urban -National -10 1 11 21 31 41 51 61 71 81 91 Percentile of the Household Income Per Capita Note: Based on householdper capita expenditures inrural areas and per capita incomes in the main occupation for urban zones to enhance the comparability across surveys of the 1990s-2000s. Source: Authors' estimates based on household survey data. FACTORS DRIVINGCHANGESININCOMEDISTRIBUTION 2.22 Several demographic and micro forces can be behind the increase in income inequality during the 1990s: the restructuring of employment across economic activities and occupations, changes inthe level o f education, the size and composition (gender, age, ethnic) of the workforce and households, and variations in the earnings premium attached to these characteristics. A companion study for this report (Gasparini et al. 2004) assesses the relative importance of these factors for changes in the distribution of earnings in Bolivia (which represent 90 percent o f total 19 EstablibinotheBasisfor Pro-Poor Growtb income, see Annex 2.1 for the methodology). Specific results are detailed in Chapter 3, but the main findings point to the following forces as drivers of distributional change duringthe 1990s: 0 A sizeable increase in the dispersion (Le,, inequality) of unmeasured or unobserved wage determinants (such as education quality, labor market connections, or unmeasured personal skills) i s the mainfactor behindthe increase inearnings inequality. These factors have played a very significant unequalizing role over the last ten years, especially inurbanareas. Changes inthe returns to education were an equalizing factor during the early 1990sinurban areas and an unequalizing factor thereafter. 0 Changes in the education structure of the work force have been mildly unequalizing, mostly inurbanareas. 0 Changes in regional earnings gaps have played a moderately equalizing role, particularly in rural areas. 0 Changes in the gender and ethnicity earnings gaps had little effect on earnings inequality, while the widening gap in hours of work between skilled and unskilled workers was mildly unequalizing. 2.23 Although some of the dispersion in unmeasured earnings determinants likely reflects measurement errors, the magnitude o f the results and short time span suggests that unmeasured factors like school quality, labor market connections, andor unmeasured skills (both cognitive and non-cognitive) have become more important determinants of earnings performance in Bolivia over the last decade, for instance through their impact on returns to ed~cation.~These factors alone account for an increase o f more than three points in the Gini coefficient of inequality in the distribution of wages between 1993 and 1997 in capital cities, three points between 1997 and 2002 in urban areas, and two points in rural areas during that period. The impact on the equivalized householdincome distribution i s smaller but still significant. 2.24 Contrary to the findings for other countries, there i s no evidence that rising returns to higher education increased earnings inequality between 1993 and 1997. On the contrary, wages grew faster for workers with primary education and reduced the Gini coefficient by over one point. In contrast, unskilled earnings lagged behind with the deceleration of growth in the last five years. The change in average returns between 1997 and 2002 was unequalizing, although o f smaller magnitudethan inthe earlier period. 2.25 The effect of education returns on earnings inequality is magnified by the uneven change inthe returns among workers (Gasparini et al. 2004). The estimated fall inincome inequality in capital cities during 1993-97 i s 2.5 Gini points when we account for the relatively faster decline inthe unskilledwage gap for workers at the bottom of the adjusted earnings scale whom tend to come from poor families. The contribution to the rising wage and household income inequality during 1997-2002 is also higher, since the increase in the skilled wage premiumbenefited only workers in the best-paid jobs, while returns declined for most workers in the bottom o f the 4. The evidence from the background studies (Gasparini et al. (2004), Tannuri-Pianto, Pianto and Arias (2004a, 2004b) is consistent with this interpretation of the residual wage dispersion to the extent that there are important differences in the returns to observed characteristics (cf. the returns to education) along the earnings distribution. See Chapter 3 and Lemieux (2004). 20 Growth, Poverty, andIneaualilv adjusted earnings scale. These results highlight the increasing importance of unmeasured worker skills that affect the productivity of measured human capital (years of education) for labor market performance in Bolivia. It could be that workers from better quality schools and/or with less favorable family background have borne the earnings erosion of the sluggish economy and rising unemployment. Furtherresearch is needed to ascertain the significance of these factors to guide the design of public policies. 2.26 The unequalizing effect of the moderate educational upgrading o f the workforce i s becoming visible in recent years. The impact of a better-educated work force was modest in capital cities between 1993 and 1997, large in urban areas between 1997 and 2002, and negligible inrural areas duringthat period. 2.27 The results point to reductions in the earnings differentials between several departamentos and Santa Cruz, particularly between 1997 and 2002 in rural areas. These changes are reflected in a relatively large equalizing effect on the income distribution in rural areas, since Santa Cruz i s one o f the richest regions. This convergence largely reflects the performance o f Cochabamba, L a Paz, and Oruro and it bypassed some poor regions such as Beni and Chuquisaca. Nevertheless, given that the latter are less densely populated, the overall effect has been equalizing. 2.28 Changes in labor supply and the gender wage gap have played a smaller role in the distributional changes. The gap in hours of work per week between skilled and unskilled workers widened over the last decade. While college graduates worked an average of three more hours per week in2002 than in 1997, those who completed primary school worked an average of two hours less in 2002 than in 1997. This had an unequalizing effect on the distribution. Meanwhile, the gender wage gap shrunk between 1993 and 1997 and increased thereafter, resulting in an equalizing change in the first period and an unequalizing force in the second. However, all of these impacts seem to be modest: the Giniincreases are less than half a point. 2.29 Summing up, changes in income inequality in capital cities over the last 10 years result from several forces, some of which offset each other. Much of the change cannot be explained because it i s attributed to unobserved worker characteristics, and further research would be required to pinpoint specific factors. During 1993-97 the unobserved and educational upgrading acted to increase inequality, but these factors were overcome by the equalizing impacts of the decline in the earnings gaps between more and less educated workers and the better-off and relatively worse-off regions. In contrast, several factors acted to increase inequality in the last five years, mainly the continuing rise in the returns to unobserved characteristics and the educational upgrading of the labor force, coupled with a moderate increase inthe education wage premium. Changes in regional wage gaps were equalizing, especially in rural areas, while changes in the gender wage gap have not affected income inequality significantly. Some of the associated policy levers of many o f these factors are examined inchapters 4 and 5. 21 EstabkhinatheBasisfor Pro-Poor Growth 22 3. CONSTRAINTSTO EMPLOYMENT CREATION-THE DEMAND LABOR OF Bolivia's weak business environment limits investment, reduces productivity, and hampersjob creation. Capital accumulation (tied to the adoption of new technologies) contributed little to growth, while labor productivity (GDP per worker) declined 2 percent per year over the 1990s. Total productivity gains of the 1990s largely reflected the improved resource allocation from economic reforms, rather than technology upgrading or innovation. This led to low job creation. Firms cite a small sales market, burdensome labor and business regulations, limited and costly credit and transport infrastructure as binding constraints to their operation. Thesefactors affect capacity utilization rates and firm size, and thus employment creation. Regulatory constraints such as registration and operating licenses, high collateral, and skilled labor bottlenecks are the most bindingfor smallfirms. Input costs and credit constraints bind the mostfor largerfirms. Faced withfew incentives to comply with regulations to start and run a business, many micro and small firms remain in the informal sector and cannot grow through capitalizing onproductivity gainsfrom innovation and economies of scale. To increase productivity, and thereby raise employment and income levels and reduce poverty, the Government should consider a number of policy measures to improve the business environment. Two broad types of actions are required: measures to (i) encourage productivity increases through greater participation in world markets, changing rules to ease access to credit and promoting worker training and technology adoption; and (ii) modernize labor regulations and promoteformal-sector participation by simplifying business registration, easing over-restrictive labor regulations, providing incentives for smallerfirms to join the formal sector, and working to create business associations to achieve economiesof scale. 3.1 This chapter adopts a labor demand perspective to analyze the firms' obstacles to expansion andjob creation, complementing the next chapter's focus on labor regulations and the characteristics of labor supply, employment and earnings. The analysis focuses on the main constraints to job creation and productivity improvement in manufacturing firms, with particular attention to small and medium enterprises (SMEs). The specific focus i s on the dynamics o f job creation and destruction, the characteristics o f labor demand, sectoral patterns o f manufacturing productivity, and the microeconomic factors affecting manufacturing capacity utilization and firm size. The results are usedto assess the prospects for job creation, particularly for unskilled workers, and to identify policy actions that can alleviate these constraints. 3.2 The focus on manufacturing i s due both to data limitation and the importance of this sector as a source of labor intensive growth. The analysis i s based on a 2000 investment climate survey (FACS) of formal manufacturing firms and a newly developed panel data from INE's surveys of manufacturing activities (See Annex 3.1). The FACS's and INE's manufacturing surveys are useful for analyzing firms' potential for job creation, but there are no similar data for other sectors (e.g., services) or microenterprises (less than five employees). While this limits the analysis, manufacturing and non-manufacturing firms of all sizes are likely to share similarities inthe business environment constraints. The chapter also draws from a field study, companion to the FACS survey that identifies compliance costs and legal requirements small firms face in startingand runninga business. 23 EstablihinatheBas&for Pro-Poor Growth 3.3 Currently existing preferential trade opportunities to the U.S. market through the Andean Trade Promotion and Drug Eradication Act (ATPDEA), and potentially even more far-reaching access if Bolivia joins in negotiations on the FTA with the US., offer increased potential for export-led growth to boost employment and earnings. The Bolivian Government can take policy actions that could leverage improved export potential to boost job creation. A variety of economic incentives have been considered to increase exports of textiles and wood products under the ATPDEA initiative, including tax incentives, the creation of manufacturing conglomerates linking large, small, and medium manufacturers (maquicentros), working capital financing and training. SMEs can play a crucial role in promoting more labor-intensive growth. Further reforms to business and labor regulations could reduce obstacles and increase incentives for small firms to participate informal sector institutions and to scale up. The following analysis aims to inform these public policy options. EMPLOYMENT CREATION: CHALLENGESAND OPPORTUNITIES 3.4 The limited employment opportunities to escape poverty rank at the top of Bolivians' concerns, especially the poor, as the demand for labor in the formal sector has not absorbed the increasing numbers-between 80,000 and 100,000-joining the work force each year. The job displacement caused by the coca eradication program, the predominance of subsistence agriculture, weak investments by formal domestic firms, and the low labor demand in hydrocarbons and telecommunications have all played a role in limitingjob creation. Few small firms grow into medium-size firms, with the result that a few large firms produce roughly 65 percent of GDP and only 9 percent of employment, while numerous small firms (of 10 or fewer employees) account for 83 percent of employment and only 25 percent of GDP. 3.5 Manufacturing i s labor-intensive and i s a major source of demand for agricultural products and small artisans. It generated 18 percent o f urban employment during the 1990s (thirdbehind commerce and services, which generated 31percent and 27 percent respectively) and contributed roughly 17 percent of GDP (the largest single sector). Over 300,000 people work in manufacturing in urban areas, of whom 60 percent work in family businesses or micro enterprises, 18 percent in small firms (five to 14 employees), 8 percent in medium-size firms (15-49 employees) and 13 percent in firms with over 50 employees. 3.6 Bolivian manufacturing has barely kept up with the overall growth o f the economy, and its 17 percent o f GDP share i s relatively small compared to regional standards (e.g., 21 and 24 percent in Ecuador and Peru, respectively). The sector has been particularly hit by the recent economic slowdown. Output growth i s close to 2 percent, only half the rate of the 1990s. 3.7 The manufacturing sector, in particular SMEs, faces significant structural limitations. The combination of inadequate infrastructure, limited supply o f skilled labor, restricted export growth, low capacity utilization, a small domestic market, high start up costs and burdensome red tape, highindebtedness and restricted access to credit markets all seriously hinder the sector, despite some recent improvements. 24 Constraintsto Emdovment Creation-The Demandoflabor THEMANUFACTURING SECTOR: CHARACTERISTICS AND INVESTMENT CLIMATE 3.8 Overall, manufacturing firms did Figure 3.1: The Majority of Bolivian not increase their employment during Manufacturing FirmsEmploy Only 5-14 workers 1995-1999 (Figure 3.1). The composition o f the sector in terms of firm size 100% 1650 remained relatively stable, except for 1998 80% 1600 when the number of SMEs fell. The 155011115-49 largest number o f firms was in the non- 60% 1500 metallic minerals, food, and wood 40% 1450 products sectors inthis period. 1400 20% +Total # of firms 1350 3.9 Total employment changed little 0% 1300 over the period, but there was a 1995199619971998 1999 reallocation towards larger firms and the petroleum and food sectors (Figure 3.2). Source: Authors' estimatesbasedonEEAMsurvey data. Approximately 50,000 employees work in the over 1,500 firms covered by the surveys. Employment inbigger firms expanded at an average annual rate o f 3 percent, whereas it fell 0.5 percent and 2.5 percent in small and medium enterprises, respectively. Food and petroleum products showed steady increases in employment between 1995 and 1999, while employment in wood-related activities contracted. About 70 percent of formal manufacturing employment was concentrated in firms with 50 or more workers, and over half of it was generated in the food, textiles, and miningsectors. Figure 3.2: Evolutionand Distributionof Employment inManufacturing, Bolivia, 1995-99 By Activity Type c 0 W Note: *AAG: Average Annual Growth. Source: Authors' estimatesbasedon EEAM survey data. 3.10 Manufacturing output expanded faster and was mainly generated in larger firms and in capital-intensive sectors (Figure 3.3). Value added grew at an average annual rate of 5.5 percent for larger enterprises, 3.8 percent for small firms and stagnated for medium-sized firms. Petroleumrefinery activities generated over 50 percent of manufacturingoutput. 25 Figure 3.3: Evolutionand Distribution of Manufacturing Output,Bolivia, 1995-99, $Bs 1995 Bv FirmSize By Activity Type 0% 1W% 5% ao% 4% 00% 3%1 (3 40% 2% 20% 1% 0% 0% 1935 1990 1997 1998 1999 0 Food 0 Bw &lobrn Textiles0 Wwden la Petroleum 0 Mneral Clher Note: *AAG: Average Annual Growth. Source: Authors' estimates basedon EEAMsurvey data. 3.11 Overall, there were limited gains in average labor productivity (measured as the ratio o f value added to total employment) over the period, and the gains were concentrated in the most productive subsectors (Figure 3.4). Labor productivity increased relatively evenly among firms, with average annual growth rates between 2.5 percent (large) and 3 percent (medium-size). On average, large firm productivity i s two to six times higher than that o f small and medium-size firms in the panel, respectively. These differentials are larger than those in the FACS survey. Labor productivity increased in the petroleum, food, beverages and tobacco industries (sectors where labor was already more productive), while textiles, wood and mining production did not gain. These comparisons are only partial indicators of overall productivity differences among sectors and they ignore the intensity of use of both capital and labor (analyzed further below). Figure 3.4: Average Labor Productivity inthe Manufacturing Sector, Bolivia, 1995-99 (Value Added per worker, $Bs 1995) y Firm Size By Sector ~ 5 . 1 4 ~ 1 5 4p50amlme 9 1 800 > 3% 5 2%g 0 1% " I 1995 1996 1997 1998 1999 0% 1995 19% 1997 1998 1999 AAG' Note: *AAG: AverageAnnual Growth. Source: Authors' estimates basedon EEAMsurvey data. 26 Constrain& to Em,ulovment Creation-The Demand of Labor 3.12 The weak and uneven performance of the manufacturing sector i s symptomatic of Bolivia's economy and poor investment climate. A recent Bank study concluded that a small sales market, institutional limitations, restricted and costly credit, high operational costs, and poor and costly transport infrastructure constrain competitiveness (World Bank, 2001). Despite recent progress, a comparison of Bolivia with other countries reveals many opportunities for improvement (Table 3.1). The study identifies the following key limitations faced by the manufacturing sector: A thin, localized market faced with stiff informal competition. The majority of domestic trade i s intra-regional, and inter-regional trade i s highly concentrated among L a Paz, Cochabamba and Santa Cruz, which account for close to 90 percent o f domestic sales. In many sectors at least one thirdof the market i s suppliedby lower cost informal producers and contraband. Less than 10 percent o f firms (almost half of large firms and 20 percent of medium enterprises) overcome domestic markets constraints by exporting. Existing export promotion programs can be improved to help overcome market and information barriers. Limited capital finance and stringent credit access. Lendingi s highly concentrated in a few local banks, very costly, and carries stringent collateral requirements. The institutional underpinnings for leverage-transparency, auditing, judiciary security, an asset registry, and a secondary market for capital goods-are lacking. An expensive and inflexible environment for business. The structure of supply chains, high inventory levels, high labor costs, scarcity of skilled labor, and poor infrastructure makes Bolivia a high cost and inflexible environment for business. Despite trade liberalization and improvements in customs administration, many firms report long delays and high costs of clearing customs. Costly and unreliable transportation services are one o f the most severe constraints. Although wages are low, the share o f non-wage benefits (e.g., bonuses, subsidies, social security and pension) in total labor costs are among the highest in the L A C region, ranging from 42 percent for small firms to 52 percent for large firms. Firms have difficulties filling highly skilled vacancies. Burdensome and costly licensing requirements and weak institutions. In 2004 a prospective entrepreneur needed to complete 15 procedures at a cost of 1.7 times Bolivia's average per capita income to start a business (Table 3.1). This i s an improvement from 1997 when the country had the largest number of procedures in the world and the cost stood at 2.6 times GDP per capita (Djankov et al, 2000), but still fares behind neighbor and similarly poor countries. Fixed costs remain prohibitive for smaller firms, and the ensuing barriers to entry hamper the economy's productivity, investments and growth. President Carlos Mesa has made simplification o f procedures a national priority. The IFC has supported simplification inLaPaz and is extending its work to other municipalities (Box 3.1). But fixed costs remain prohibitive for smaller firms and are likely to continue to contribute to the high rate of informality (Box 3.2). Property rights and contracts enforcement i s weak, with lengthy court cases and lack of other effective channels for dispute resolution. 27 Table 3.1: Investment Climate Indicators of Bolivia and SelectedLAC Countries Country Bolivia Chile Peru Colombia Ecuador Honduras Nicaragua Number of 15 Procedures 9 10 14 14 13 9 Starting a Business Time (days) 59 27 98 43 92 62 45 c Hiring and Firing Workers Registering Property Getting Credit Protecting Number of Procedures 47 28 35 37 41 36 18 Enforcing Contracts Closing a Business 19.3 31.1 Note: aO-lOO,O=least difficult; 100=most difficult; bsimple average o f the Difficulty of Hiring, Rigidity of Hours and Difficulty o f Firing indices; O=least rigid; 100=most rigid. 0-10, O=less legal rights; lO=more legal rights. d0-6, O=least credit information; 6=most credit information. e0-7, O=least disclosure requirements; 7=most disclosure requirements. Source: World Bank (2004a) and Djankov et a1(2002). 28 Box3.1: Simplifying BusinessRegistrationinLa Paz With the support of IFC and FUNDES International, Bolivia is simplifying business regulations to make it easier for SMEs to do business. Bolivia is simplifying municipal business regulations for the primary reason that the municipality i s often the starting point for new businesses. SMEs deal more with municipal authorities rather than national state agencies. Decentralization policies will give further power to the subnational level. A business registrationsimplification project was successfully implementedinthe municipality of L a Paz. This municipality was chosen because 33 percent of Bolivia's private sector i s located inL a Paz; it has a high degree of bureaucracy; business simplification was included in the Municipal Development Plan; and more important, there was support from both the local General Secretariat and the national Ministry of Economic Development. Success in the project in La Paz was due to close collaboration with municipal authorities and private businesses. Duringimplementation, the project (i) identified and built up the client commitment to reforms; (ii) diagnostics of the current situation; made (iii)designed the simplification program; (iv) developed needed regulations and procedures; (v) trained municipal officials; (vi) conducted information campaign for private sector; (vii) launched a new simplified business regulations; and (viii) adjusted the system based on the feedback from businesses and public officials. Proiect challenges The challenges that were encountered were mainly legal in nature. The legal foundation was so weak that the Municipal Resolution was the only resolution approved to support the procedures manual. Other challenges included personnel issues, such as high turnover and no clear definition of assigned duties or qualifications, the inability to comply with the time frames forecast in the design; Iand difficulties developinga MunicipalToolkit for business simplification. Results (i)91percentdecreaseinthenumberofdaystoobtainaoperatinglicense;(ii) percent A 32 decrease in the cost of compliance for business; (iii) 68 percent decrease in the number of required steps; (iv) creation of a "one-stop shop" for the operating license; (v) 20 percent increase inthe number of registered businesses; (vi) 70 percent increase of income generated by the business simplification process done in the Municipality; (vii) elimination of unnecessary requirements of about 50 percent; (viii) public disclosure of requirements and procedures; training and sensitization of municipal employees; (ix) new legal foundation; (x) entrepreneurs visits reduced from six to two; (xi) other red tape reduction applied by the municipality in another 30 procedures; and (xii) design of support software for business registration and authorization procedures developed by the municipality itself (www.ci-1auaz.gov.bo). Next stem The next steps to provide a sound investment climate in other municipalities are: to identify a municipality with minimum conditions for intervention, such as economic activity, bureaucratic barriers, political will, and institutional capacity; to reach consensus for the design and strategy implementation; to develop joint ventures with municipalities to generate positive experiences; tc replicate results in other important municipalities, such as red tape reduction; and finally to ensure the sustainability of reform efforts insulating them from political changes. 29 EstablshinatheBasisfor Pro-Poor Growth Box 3.2: Constraints for Businessesto Becomeand Remain FormalinBolivia A companion field study to the 2001 FACS report identified the excessively high costs of complying with most legal requirements to start and run a business among the main causes o f informality in Bolivia (World Bank, 2001). Some of these are being addressed by ongoing reforms but the following are remaining barriers to formality inBolivia: High costs to registration and business expansion. Despite recent improvements, Bolivian entrepreneurs incur prohibitively high costs through government fees (1.7 times income per capita) to acquire the necessary permits to operate. Registration requirements are higher for the incorporation of partnerships. Since most business projects require a minimum scale of operations for capital ventures, this situation hampers the realization of economies of scale and imposes fixed costs to business expansion. Small firms can not afford trained lawyers to deal with these procedures. Moreover, in order to export, firms have to register in the Registro Unico de Exportadores (National Register o f Exporters, RUE), which entails a down payment o f US$200 and US$20 monthly payments. Centralization and burdensome requirements. Excessive concentration of agencies running registers and licenses in L a Paz and archaic documentation requirements (e.g., notarization) also increase the red tape cost for micro and SMEs. The operating license and the patente municipal are more difficult to obtain than the tax registration for the smaller enterprises. This underscores the importance o f ongoing efforts to strengthen the capacity of municipalities to enforce registrationrather than on centralized entities. Negative tax incentives and low benefits. The Simplified Regime o f Bolivia's Internal Revenue Service does not allow extending invoices granting tax credits on purchases. This is a major handicap for a small business trying to sell its products to a formal business because the latter needs a formal receipt in order to deduct its taxes. The costs of registering in the General Regime for a small establishment are 20 times higher than for the Simplified Regime. Moreover, firms see little benefits o f being formal. Smaller firms often have limited access to procurements o f government purchases and subcontracting activities. Given the poor performance o f the judiciary and public services, they lack property rights on their assets, have no legal protection incase of theft or expropriation, cannot used them as collateral for a loan or easily transfer them. Many entrepreneurs do decide to register in some entities and not in others, which reveals their willingness to pay for some degree of formality if there are positive net benefits. For instance, while tax registration of small firms declined in relative and absolute terms between 1988 and 1997 compliance with municipal registration remained high. High costs of labor regulations. Mandate labor benefits imply a 42 percent increase in labor costs (8 percent of a small firm's annual sales) which are fully the responsibility o f firms. The compliance costs of labor regulations are far higher than the recurrent costs of compliance with municipal licenses or tax registration. Insofar as workers are unwilling to accept lower nominal wages to indirectly pay for these benefits, current labor regulations also discourage firms from becoming formal. CREATION AND TURNOVER MANUFACTURING EMPLOYMENT' OF 3.13 Changes in employment levels take place either through changes in the size of existing firms, or the entry or exit of firms from the market. Migration of formal jobs or firms to the informal economy i s another important source of job reallocation. Understanding these job flows i s important for more pro-poor growth for two reasons. 3.14 First, job reallocation is an important source of productivity growth. The movement towards a more productive economy i s intrinsically a trial and error process where new 1. Job creation and turnover was investigated using the EEAM manufacturing panel, given that changes in employment measured by household survey data fail to reveal the actual levels o f labor mobility or intra- sectoral employment reallocation. The methodology used i s described indetail inAnnex 3.2 30 ConstraintstoEmRlovment Creation- TheDemandofLabor technologies displace obsolete ones and employment destruction gives way to higher wage jobs. Productivity gains come from existing firms becoming more productive, the exit o f less productive firms, and the reshuffling of employment from less productive to more productive firms. Recent studies find that the job reallocation process contributes with 15 to 50 percent of the growth in aggregate productivity (IDB, 2003). 3.15 Second, excessive job reallocation can be inefficient where, for example, informal firms or SMEs could destroy jobs if underdeveloped financial markets prevent them from adjusting to negative shocks. Excessive bureaucratic burdens and rigid labor laws increase start up and adjustments costs for firms and push many to informality, and thus can lead to inefficient job reallocation and highfirm mortality. 3.16 Overall, manufacturing added Figure3.5: Gross Employment Flowsinthe few jobs directly during the second ManufacturingIndustry, Bolivia, 1996-1999 (% Jobs) half of the 1990s (Figure 3.5). Remunerated jobs were created at a rate of 20 percent, but the turnover of Unskilledlabor existing jobs was 18 percent, so that there were only 2.2 percent net additions. Employment creation and turnover was rather volatile: the rates of gross destruction fluctuated between 10 percent and 18 percent, -50, -97 97-98 98-99 96-99 96-97 97-98 98-9Q(96-99 while gross creation rates oscillated f between 10 percent and 20 percent. ,"," There i s no evidence of a bias against unskilled labor injobs reallocation. 3.17 About four out o f 10 jobs Source: Authors' estimates based onEEAMsurvey data. were reallocated among manufacturing workers over the period, a high rate compared with other developing countries. Figure 3.6 compares gross employment flows among selected countries. Although reference periods are different this offer useful benchmarks, and illustrates the high level of job reallocationinBolivia according to international standards. 3.18 These levels of reallocation likely reflect the utilization of temporary employment and unpaid family labor, as well as migration of workers andor firms to the informal sector. Permanent employees tend to enjoy more stability and represent larger labor costs for employers, whereas temporary and family help can be substituted more cheaply. 3.19 This is consistent with the finding that both the creation and destruction of jobs is larger insmaller enterprises (Figure 3.7). This mightreflect the flow ofjobs from SMEs to the informal sector. However, evidence from other countries clearly shows that employment reallocation among medium and large enterprises typically reflect changes in payrolls while in smaller enterprises it occurs mainly through the exit and entry o f firms. The high level of attrition of small firms inthe EEAMpanel i s consistent with this. Unfortunately, we cannot directly confirm this hypothesis since it i s not possibleto determine whether firms that drop from the sample also exit the market. The only sectors that added new jobs were food, petroleum refinery and mining. 31 Figure 3.6: Bolivia has High Rates of Job Creation, Destruction, and Reallocation by International Standards 45% 40% I A 35% 30% 25% 20% 15% 10% 5% 0% 1973-1988 1984-1992 1984-1992 1 11985-1991 1996-1995 1 United Bolivia ~ France Italy United States Kingdom DestruUCtion -A-Re-allocatioil Source: Based on Camhi, Micco and Engel (1997). Figure 3.7: Job Creationand DestructioninBolivia's Manufacturing Industry is Larger inSmallFirms(1996-99, % Jobs) BYFirmSize y Sector 50%yfA 2: W% I16% 50% 112% 40% 15% ._ 2c ,z 40% 0 10% gm? - - 30% -2 30% 0 5% gmE 20% I c G% z" ;20% 10% -5% 10% 0% -10% 0% < 15errp. 15-49 50&t ' 1 1 1 1 1 ' I Destruction A Netcreation 1 Source: Authors' estimatesbasedon EEAMsurvey data. 3.20 The results presented above suggest that an inefficiently high level of job reallocation i s behind Bolivia's sluggishnet job creation. The combination of a poor investment climate, high levels o f informality and constrained credit markets mean that: (i) firms adjust to negative many shocks through significant job churning; (ii)there i s likely an inefficiently high mortality of firms that negatively impacts productivity growth, as SMEs operating in the informal sector are prevented from achieving economies of scale, learning and adaptation of new technologies and productive processes; and (iii) Bolivian manufacturing workers face a volatile labor market, with limitedinsurance to facilitate efficiently matching their skills with availablejobs. 32 Constraintsto EmdOVmentCreation-The DemandofLabor PRODUCTIVITYAND TECHNICAL EFFICIENCY 3.21 Employment i s linked to firms' economic performance. The rate of utilization and mix o f inputs depends on their cost and contribution to the generation of output. The latter i s captured by the output-labor elasticity: the percentage increase in output relative to a percentage increase in labor inputs. The background study by Jimenez and Landa (2004~)estimated output elasticities with respect to capital, skilled and unskilled labor in Bolivia's manufacturing sector using the manufacturing panel data for 1995 to 1999. Their results can be used to assess: (i) whether existing technologies or economic fundamentals favor the use o f unskilled labor; (ii) which subsectors are more likely to generate labor-intensive growth; and (iii) efficiency (or the total productivity) with which firms operate. 3.22 Their results are presented inAnnex 3.3, and can be summarized as follows: 0 On average, expansion of capital leads to greater increases in output compared to labor. This reflects the relative scarcity andhighcost of capital inBolivia. 0 Manufacturing growth in labor-intensive sub-sectors would tend to favor the creation o f unskilledjobs. 0 Growth would favor the use of unskilled labor in the beverages and textiles sectors and would favor capital inminingandwood production. 0 There i s suggestive evidence of a high degree o f substitution between capital and unskilled labor, moderate substitution between labor types, and little complementarity of skilled labor to capital.2 0 There i s significant room for expanding manufacturing output by a more intensive use of the installedcapacity o f firms. 0 Firms in the lowlands, in the petroleum, wood products andor other industries sectors, show the highest levels of productivity. 0 There were no significant overall productivity gains duringthe period. 3.23 Note that these results refer to the relation between marginal changes in output and input utilization within manufacturing. H o w many jobs would be created in the economy depends-on the capital/labor ratio in each subsector and on indirect employment effects on other sectors of the economy. 3.24 The results also point to low levels efficiency in manufacturing firms. Many firms produce less than would be expected given their inputs and capacity utilization, age, location and sector. Inefficiency seems pervasive across firms, although beverages and tobacco and petroleum are relatively more efficient. 3.25 Altogether the results suggest that innovation i s stifled in Bolivian manufacturing. The mixtures in the use of skilled and unskilled labor are symptomatic of production processes that do not make intensive use of information technologies or modern management practices. Firms do not achieve sufficient productivity gains as they mature, reflect weak incentives to innovate, a highcost environment, limited sales andcredit markets, andbarriers to entry. 2. The proper assessment of the degree of complementarity between inputs should rely on cross-elasticities of input use, which were not estimated by Jimenez and Landa (2004~). 33 EstablishhinutheBasisfor Pro-Poor Growth 3.26 The large variability in firm performance hampers the ability of targeted industrial policy (such as subsidized credit or tax exemptions) to promote the expansion of businesses and employment. The manufacturing sector i s composed of many efficient and inefficient firms, malung it difficult to predict the final impacts of policies on any specific firm or subsector. There are likely to be higher returns to policies that improve the overall investment climate, including addressing infrastructure bottlenecks and barriers to market entry and exit, deepening of financial markets, and support to research and innovation. THEDEMAND LABOR FOR 3.27 The study by Jimenez and Landa (2004~)assesses manufacturing employment's response to changing economic conditions taking into account direct labor costs (real wages) through estimation o f demand equations for unskilled labor. The main findings are discussed below; Annex 3.4 presents the methodology and results inmore detail. 3.28 Firms and workers face a relatively high long run trade-off between higher wages and increasing unskilled employment. The long run demand wage elasticity for unskilled labor i s - 0.64, meaning that for every 10 percent increase in real wages labor demand declines in 6.4 percent (keeping output constant). This i s twice as high as the average international estimate of -0.3 (Hamermesh, 2003) and on the high end of estimates for other L A C countries, for example, Brazil (-0.4), Chile (-0.37), Colombia (-0.49), Mexico (-0.2), Peru (-0.2), and Uruguay (-0.69). Moreover, although the estimated elasticity does not account for indirect labor costs (e.g., mandated benefits) evidence for other countries shows that the elasticity of these labor cost components can be as high as the own wage employment elasticities (Saavedra and Torero, 2000; Mondino and Montoya, 2000). 3.29 Employment tends to adjust slowly in response to cost or output shocks. The estimated (half-life) speed of adjustment of labor demand i s 1.3 years, which again i s above international estimates (half a year) but aligned with those for Latin America (1 year in Brazil, 1.2 years in Chile, 0.4 years in Colombia, 0.8 years inMexico, 1.3 years inPeru, 1.5 years inUruguay). This suggests that it takes a relatively long time before fluctuations in the demand for labor adjust to changes ineconomic conditions. 3.30 Labor demand i s dominated by an inertial employment component. Firms rely solely on past employment decisions to adjust their payrolls (not on past wage and output changes). This i s consistent with idle capacity in which case employers can utilize the existing labor force more (less) intensely when faced with higher output demand (wages). 3.31 The long-term response of employment to changes in output seems low, although it i s not inconsistent with constant economies o f scale. The long-run employment-output elasticity is 0.47, implyingthat a 10percent increase in long-term manufacturing output results inan increase of 4.7 percent inunskilledemployment. 3.32 These results suggest that Bolivia's manufacturing employment i s quite flexible with respect to changes in wages (and very likely changes in non-wage costs also), but adjusts slowly in response to shocks. This, in turn, implies that policies which reduce the cost of labor by increasing labor productivity or reducing non-wage costs could be very effective in increasing employment levels. Moreover, the slow speed of adjustment o f employment to shocks i s likely related to several factors affecting firms' adjustment costs (underdeveloped financial markets, 34 Constrainbto EmDfOVmentCreation-The DemandofLabor firindhiring regulations, supply bottlenecks), macroeconomic volatility (real exchange depreciation, external shocks) and the degree of competition (entry-exit barriers). MICRO CONSTRAINTS TO CAPACITY UTILIZATION AND FIRMSIZE 3.33 Labor utilization can take the form of a more intensive use of the existing pool of workers or new hires. These relate to a heavier utilization of idle capacity and to the expansion o f firms. Capacity utilization (CU) i s a short-term measure of the intensity of labor use. Low CU levels are symptomatic o f production inefficiencies, which can result from economic downturns, tight credit, obsolete technologies and/or capital investments far exceeding the optimal plant size (e.g., due to stiff collateral requirements). Meanwhile, the number o f workers can be used to track the scaling up of firms in so far as many firms do take on larger payrolls over the long run. In fact, two thirds of the growth in the developed world industries during the 1980s comes from the growth in size of existing firms rather than from new entry (Rajan and Zingales, 1998). 3.34 L o w capacity utilization and the distribution of firm size are o f particular interest in Bolivia. The average capacity use for manufacturing firm was only about 56 hours per week in 1999, which i s consistent with 1.5 shifts of operation, far below other developing countries (World Bank, 2001).3 Smaller and non-exporting firms hadthe lowest capacity utilization. Firms cited lack o f demand as the predominant cause, followed by maintenance. Most manufacturing firms in Bolivia remain small: almost half the firms in the sample are small (4-15 workers), only 15 percent are relatively large (over 100 workers) and the average number of workers per firm i s 18. 3.35 In a companion study, Muiioz, Palma and Arias (2004) use the FACS data to analyze how investment climate factors affect the distribution of firm size and capacity utilization in Bolivia. They address three main issues: (i) constraints to the growth o f manufacturing main firms; (ii) principal bottlenecks affecting capacity utilization; (iii) o f policy changes that types can potentially improve capacity utilization and firm sizes. The main findings are highlighted below; Annex 3.5 describes the methodology and regression results. 3.36 The results show that capacity utilization rates are higher, on average, among firms that are younger, capital-intensive, pay lower wages, use higher quality inputs, and in the petroleum and food and tobacco sectors. Capacity utilization i s restricted by financial and credit constraints, and government requirements affecting the production process (although relevant to only 10 percent o f firms). Membership in a business association (gremio) has a positive and significant effect on capacity utilization. Proxies o f sales markets have a positive effect on C U but are not statistically significant, which reflects the pervasiveness of demand constraints for Bolivian firms. The following effects are worth emphasizing: 0 The age and capital intensity effects suggest that firms with more modern technologies have higher capacity utilization. Firms with a capital intensive production process are likely to enjoy economies of scale and better management efficiency and face a higher cost o f leaving the capital stock idle. 3. Alternatively, firms reported that they were using 57 percent of their installed capacity (down from 61 percent in 1998).This alternative measureof CUfor the most partyielded qualitatively similar results. 35 EstablshinatheBasisfor Pro-Poor Growth 0 Firms that use more costly inputs have higher capacity utilization. This is consistent with firms' reports o f input quality as an important factor affecting their performance. 0 Textiles and wood products had low capacity utilization, while petroleum production had capacity utilization twice as high as most other sectors. 0 High collateral to obtain credit constitutes a major obstacle for increasing capacity utilization. 0 Debt defaults show a significant negative correlation with capacity utilization, suggesting that debt default can become a lasting constraint to secure credit. 0 Firms mentioned several benefits of being in a gremio: access to credit, key inputs, and technical information; help with resolving disputes; information on foreign and domestic markets; and accreditation. 3.37 While there i s not clear cut prescription of what optimal firm size should be, the results indicate that Bolivian firms are constrained to grow by a weak business ~ l i m a t e .Bigger firms ~ are older, have access to larger markets, including external markets, are less capital-intensive, and are in the petroleum, and food and tobacco sectors. Difficulty in hiringskilled workers and the inability to sign contracts with suppliers also tend to reduce firm sizes. These signal the following points: 0 The effect of a firm's age highlights the importance of the learning and adaptation process that allows firms to exploit economies of scale as they age. 0 Larger markets go hand in hand with firms' capacity to grow. In particular, export orientation i s correlated with larger firm size. This i s consistent with two causal hypotheses: that larger firms tend to get access to external markets andor that exporting facilitates firm growth. 0 The lack of skilled labor prevents firms from reaching larger scales of operation through the adoption of new technologies and management practices. The existence o f contracts between the firm andits providers helps reduce transaction costs of expanding businesses. 0 The petroleum, chemical and glasskeramics, and food and tobacco sectors contain Bolivia's largest firms, which suggest that particular conditions in these sectors have allowed firms to exploit economies of scale. 3.38 The results underscore that micro factors bind firms differently depending on their scale of operation. For firms operating at comparatively lower scales regulatory constraints are most binding. Regulatory constraints such as registration and operating licenses, high collateral, and skilled labor bottlenecks are most binding for firms operating at scales (CU or size) lower than expected for their characteristics (the underperformers). Input costs and credit constraints bind the most for firms operating at larger than expected scales (the overperformers). Belonging to a business association impacts CU positively only for firms with outperforming CU rates. Thus, policies to improve institutional and credit constraints are of first order for underperformer enterprises relative to those relatedto firm characteristics andtheir operational costs. 3.39 The models are usedto runpolicy simulations, with some noteworthy results: 4. Financial variables effects on firm size were not robust due to the reversed causality between size and credit constraints (e.g., small firms are riskier clients and thus face more costly, limited financing). 36 An expansion inmarket size increases firm size, with a more pronounced effect inrelatively larger firms. The implication is that, as the economy grows firms that were already relatively bigwill grow much more than smaller ones. Policy interventions to encourage greater use and enforcement of contracts between firms can help all firms to grow. The simulations indicated that improved contract use and enforcement could lead to an increase inthe average number of workers per firm by 20 to 50 percent. Increasing access to international markets could also have the largest impact on firm size, especially for smaller firms. For example, tripling the fraction of exporters (from 17 to 55 percent) could double the number of workers inrelatively smaller firms. POLICY INTERVENTIONSTO INCREASETHE DEMAND FORLABOR 3.40 The results emphasize that an integrated public policy to improve labor markets cannot be restricted to traditional labor market policies (see the next chapter), and should include interventions to improve the investment climate in which firms operate. As agents of job creation, firms are key players for a well functioning labor market. 3.41 The Bank report Bolivia: Microeconomic Constraints and Higher Opportunities for Higher Growth provides comprehensive recommendations to improve the investment climate. These include actions to reduce red tape; to strengthen property and creditor rights, contract enforcement, registering o f property, and collateral law; to put in place mechanisms to increase access to prudent financing for SMEs; and to improve judiciary and public services, trade logistics and supply chains. These recommendations would also increase the demand for labor andthereby reduce the level o f poverty and improve the lives o f Bolivia's poor. 3.42 Here we focus on selected policy options that can improve the productivity o f the Bolivian economy, particularly of smaller firms. The essence o f development i s getting more output from the country's resources. This i s particularly important for Bolivia to overcome its low growth inertia and forge closer ties of natural resource based sectors to production and employment in non-tradable sectors. They would strengthen Bolivia's integration into world markets andnew investment opportunities. 3.43 Productivity growth requires an environment where resources (capital, labor) can move to their more efficient and firms are able to find more efficient ways to operate. This rests on the creation of new firms and the restructuring (expansion, exit) and innovation o f existing firms. In turn this requires focusing on sustained improvements in business environment regulations as well as specific policies to foster innovation and give smaller firms more incentives to participate informal sector institutions. 3.44 Accelerate the reform of business regulations. It i s critical to accelerate and expand the ongoing revamping of business regulations particularly those related to reduce the red tape and costs to start and close a bu~iness.~The Government has already taken steps in this direction. IFC-supported initiatives have helped to simplify business registration in L a Paz and other municipalities. Steps are also being taken with Bank support to the Government's 5. This is one of the top actions to accelerate economic development in developing countries highlighted by a recent panel of Nobel Laureates. See Copenhagen Consensus (2004), www.coDenhaeenconsensus.com. 37 EstablishinatheBasisfor Pro-Poor Growtb Regulatory and Corporate RestructuringProgram, learning from international reform experiences such as Colombia's and Peru's. Inthe short term, the following actions merit priority: Cutting by at least one half the cost of registration (to 85percent of GDPper capita)and business expansion for micro and SMEs, particularly the cost of incorporation of partnerships, registration in the General Tax Regime and export licensing, by further lowering government fees, eliminating unessential requirements (notarization), and streamlining one-stop business on-line registration and licensing in municipal government offices. The Peruvian and Colombian reforms offer valuable lessons. e Streamline labor regulations in line with international standards, which are currently a limitation on the ability of firms to expand and contract along with the economic cycles and their own competitiveness, and thus inthe long run hurt employment and wage levels (see Chapter 4 for details). 3.45 Foster increases in the efficiency of firms by supporting innovation. Productivity growth should be the cornerstone of policies for the productive sector. This means taking full advantage of information technologies, modern management practices, new production and quality control processes, and piloting small scale rural technology and crop varieties. A focus on innovation would greatly enhance the dividends of government support to firms and rural producers through the National Dialogue "productive pacts" embedded in municipal "Estrategias Productivas Integrales". Short-to-midterm actions may include incentives for technology adaptation, piloting technology service centers for SMEs and innovation networks (linking the productive sector, national and international universities and research centers inR&D activities), and gradually strengthening the national training system. The longer term calls for a gradual consolidation of a national innovation system as the country exhausts first order innovation and learning gains. Donor support can help develop specific policy options in a national innovation strategy drawing lessons from relevant experiences in the region (e.g., Brazil, Mexico, and Central America) and East Asia. 3.46 Implement policies to increase the benefits of formality for smaller firms. Besides general improvements in business regulations and the investment climate, smaller firms would benefit from complementary interventions that enable them to achieve economies o f scale and productivity gains. These may include: Establishing pilot initiatives that provide small firms incentives to become formal, encouraging small firms and producers to bid for government contracts (the Presidential Decree No. 27328 "Compro Boliviano" lays the legal basis), extend partial credits o f value added taxes for eligible firms, and offer business development services (access to market credit, judicial services, management and accounting practices) with special emphasis on supporting innovation initiatives and export production. The country could try to tap Millennium Account resources and learn from the successes of the U.S. Small Business Program and others like Italy and Chile. These incentives should be phased out gradually as firms grow in size and articulated, for example, requiring compliance with 38 Constraints to EmdovmentCreation-The Demandof Labor minimum tax and labor regulations to obtain bidding rights and access to business development programs. Facilitating small businesses and producers to achieve economies of scale, for example through trade or professional associations, in production andor contracting of services, reforming government policies or other interventions to overcome constraints in access to credit, technology, inputs, quality certification, accreditation, disputes resolution, or purchasing of health plan benefits. Institutional strengthening (stafing, training, technical assistance) of the Superintendence of Enterprises and coordination with relevant public agencies so that these can assume their increasingly more complex role.6 6. A recent workshop to learn from the ongoing reform experience of Colombia in corporate restructuring should help identify priorities and capacity building needs. 39 EstabfishinatheBasisfor Pro-Poor Growth 40 4. CONSTRAINTSTO HUMAN CAPITALACCUMULATION-THE SUPPLY OFLABOR Although improving, the public education system-especially in rural areas where poverty is most extreme-offers low quality education. Further, poor families face dificult choices and are often unable to afford to keep their children in school long enough, instead needing them to help the family, either through income-generating activities or domestic and agricultural chores. Returns to education are low-six out of ten workers graduating from high school are at risk of poverty. In rural areas, only a post-secondary education offers a significant boost to earnings. As well, education tends to have lower returns for workersfrom poor families. The employment gap faced by women, young and poorly education workers, and the earnings disparities solely related to gender, ethnicity and location are above regional averages. While workers in the informal sector tend to have lower productivity, the selfemployed may have no alternative options for work. Informal salaried workers do face an earnings disadvantage when compared to salaried formal sector workers, which could be improved if formal participation increased. Migration-in particular, rural to urban migration-is a useful tool for many poor people to improve their earnings. However, migration has been relatively small, limiting its scope to reducepoverty among the rural poor. To improve the ability of the labor market to eficiently match workers and jobs, and hence, improve incomes and reducepoverty, policy actions are required to (i) strengthen the education system, particularly for the poor in rural areas; and (ii) improve labor market equity. Education improvements shouldfocus onfilling gaps in universal basic education; improving secondary education; addressing low education quality and inequalities in achievement; improving private higher education accessfor low income students; and implementing a conditional cash transfer program, similar to Bolsa Familia in Brazil or Oportunidadesin Mexico. Measures to increase labor market equity could include expandingpre-schoolfacilities and child care centers tofacilitate women's and migrants' labor force participation, and offering training in high schools and collegesonjob search techniques. 4.1 Labor i s the source of most of the income of the poor. Thus, the labor market and the institutions that govern it can play a central role in reducing poverty and inequality. The results of Chapter 2 point to disparities in labor market performance. This chapter examines the characteristics of Bolivia's labor market, the supply of labor, employment and earnings outcomes. It characterizes the profiles of workers entering the labor market; takingjobs in the growing sectors, and lagging behind in the economy's slow down. A comparison of Bolivia's labor indicators and labor regulations with other countries concludes that these regulations are excessively restrictive and work against increased earnings and employment for both workers and employers. 4.2 The chapter has a particular focus on informal employment and internal migration given their central role inbalancing the labor surplus. The informal sector i s often seen as a haven for less-advantage workers who are denied superior formal sector jobs, mainly due to stiff labor regulations. However, the chapter finds that labor market segmentation i s a second-order source of low earnings relative to overall low labor productivity, and highlightsdifferences among types of informal employment. Moreover, rural migrants coming to urban areas during the 1990s did better than expected had they stayed in rural areas, particularly those at relatively low pay jobs 41 EstablisshinQthe Bask for Pro-Poor Growth for their skills set. However, migration has remained relatively small, limiting its scope to reduce poverty among the rural poor. 4.3 The chapter identifies interventions to improve labor markets and to better integrate informal and migrant workers into the economic mainstream. The creation of better-quality jobs requires making formality more competitive, with regulations that better balance the conditions for productivity growth and the protection of workers against job loss and their basic rights. Also needed are income generation investments for lagging regions and actions to ease internal migration costs so that the poor are better able to benefit from locations and productive endeavors with better economic prospects. EMPLOYMENT EARNINGS: AND TRENDSANDDISPARITIES 4.4 This section provides a comparative snapshot of Bolivia's labor market outcomes and examines recent trends in labor supply, employment, and earnings. It identifies the profile of workers that benefited during the 1993-97 boom, and those hurt the most with the recent slowdown. Bolivia's labor markets share many characteristics with other poor countries in the region, but Bolivia stands out for its low overall labor force participation rates, high informal employment, low returns to education (particularly secondary), and large disparities in labor market outcomes. After improvements in the 1990s, employment and earnings dxparities widened duringthe economic slowdown. LaborForceParticipation,Unemployment,andInformalEmployment 4.5 Bolivia has one o f the highest female labor force participation rates in the region (over 55 percent) but one of the lowest among men (below 80 percent) which results in one of the lowest gender gaps (Figure 4.1). Overall labor force participation has remained around 65 percent since the mid 1990s, increasing only slightly during 1997-2002 (Figure 4.2). The number of Bolivians in the labor market rose from 3.6 million to 4 million. Participation increased in urban areas especially in the main cities among women, and shows an apparent recent decline in rural areas. Participation rates remain much lower in urban areas (around 60 percent, 2.3 million workers) than inrural areas (76 percent, 1.7 million workers), especially among men. 4.6 After remaining steady duringmost of the 1990s, open urban unemployment rose almost 3 percentage points with the economic slowdown (see Chapter l), puttingthe number of jobless Bolivians in 2002 in 222,000.' Unemployment duration and involuntary underemployment are above the regional average2 and have increased since 1999. The overall unemployment rate in 2002 was similar among the poor and non-poor (about 5.5 percent). The poor represent 61 percent of the unemployed. This reflects the fact that they cannot afford labor idleness. Urban workers with primary education, particularly the young and women, were more affected by increasing unemployment (4-6 percentage points increase) than those with secondary and college educations ( 3 4 percentage points increase). In2002, 16 percent o f urban workers 15-24 years old were unemployed compared to 7 percent among prime-age workers. The high 1. It should be noted that these numbers do not consider the many Bolivians who are working only a few hours a week or who are discouragedworkers. 2. IDB (2003). 42 - Constraintsto Human CauitalAccumulation-The Suuulyof Labor unemployment among women, young, and low-educated workers stresses the importance of stronger safety nets, particularly the emergency employment program (PLANE). Figure 4.1: Labor Market IndicatorsinBolivia and Latin America Labor Force ParticipationRatesinLAC, Circa2000" Shareof InformalUrbanEmployment,Circa 2000 -- Uruguay wru Guatemala muador Paraguay eolivla Solbia Erazll Nicaragua muador Colon-bla I 1 - 8 - Argentlna , 1 kninican Rep. Paraguay r I Honduras Dnlnlcen Rep. 1 I Argentina - I - 0 Salvador I 1 chile mnduras I I Uruguay GuaterraIa 1 I Sawador Venezuela I I I Costa %a Nicaragua I I hbxko Chile r I Panama MeXlCO I I Brazil _I_ Panam 100 80 60 40 20 0 20 40 80 80 COsta moa 3 0 Self -toyed ISmallErma MarginalReturnsto Education,UrbanAreas 1990s** GenderWage Gap inUrbanAreas, Circa2000 Unexplalned Percentage Wag- Premium 101 Males __ I Brazil Honduras I C o l o m b i a Brazil I Guate-h I C h i l e Bolivia E l S a l v a d o r Dninlcan Rep. 0 o livia Paraguay Ezuador V e n e z u e l a Panam H o n d u r a s Nicaragua C o s t a R i c e chile ESalvador P a n a l l l a Uruguay N i c a r a g u a Costa Rice M e x i c o Venezuela Argentina P B , " mru U r u g u a y nnBXS0 Colorrbla i o 20 30 40 *National data except for Argentina, Bolivia, Mexico, and Uruguay. **Percentage annual return assuming degree completion. Source: Based on IDB 2003. 4.7 Informal employment is pervasive, accounting for the bulk of employment created in the 1990s, over two thirds of total employment in 2002 and 80 percent of jobs in the poorest quintiles. This comprises a mix of incipient entrepreneurs (self-employed, small firm owners), salaried workers in small firms and unpaid family labor. Bolivia has the highest rate of urban informal employment in the region, the highest for men (over 60 percent) and only below Guatemala and Paraguay for women (over 75 percent). Informal salaried employment i s less attractive for Bolivian women than in other LAC countries. 43 EstablishhinutheBasis for Pro-Poor Growth Figure 4.2: Trends inLabor ForceParticipationinBolivia, 1989-2002 Urban -- 90 T Oil 1996 70 0 1997 1997 60 0 1999 60 0 1999 50 02000 50 02000 40 02001 40 I32001 30 30 20 20 All Men Women A l l Men Women I -- MainCities I I nurai "....-I 90 I - 1 3 - 80 IIM 1996 I 70 0 1997 60 0 1999 50 [32000 40 0 2001 k sep Ebv wbr 1dL)rsldkc Ivn Jun wbv NN Nw Ebv Nw Nw 1969 1995 1531 1592 1m 1994 159 15% 15% 1537Irn am an1 zap 30 El20024 - - C A I 1 -+-Men -+-Womn 20 All M e n Women Note: P: Preliminary. Source: INEbasedon householdsurvey data for 1989-2002. EarningsDifferentials and Returns to Education3 4.8 Bolivia's labor market i s characterized by wide differences in earnings. The earnings gaps solely related to gender, ethnicity, sector, location and type o f employment are larger than L A C averages. Men earn on average over 30 percent more than women; indigenous salaried workers on average earn 25 percent less than non-indigenous; and average urban earnings differentials range from 20 to 55 percent across economic sectors, firm size and region. 4.9 Earnings growth was unbalanced. During the economic boom earnings grew relatively faster for women, self-employed and heads of households living in poor main urban centers (except for the poorest Beni and Chuquisaca) and working in utilities (including hydrocarbons), transport, construction, cornmerce and financial services. Earnings o f indigenous heads and the informal salaried and workers in agriculture, mining and the manufacturing sector lagged. Regional earnings differentials narrowed further during the slowdown except for Cochabamba and urban Chuquisaca, while agricultural and manufacturing earnings continued sluggish. 4.10 Education does not yield equally high returns for all workers. The urban earnings premiums range from 0 to 60 percent for the workers with primary education, around 20 to 30 percent for workers with a secondary education, and from 50 to 150 percent for college educated workers. Average returns to secondary education for urban workers remain among the lowest (around 6 percent per year of education) in the region. The average returns to higher education 3. See Annex 4.1 for supporting information on this section. 44 Constraintsto HumanCaDitalAccumulation-The SUDD~V ofLabor are close to the regional average (over 15 percent per year). The returns to schooling are lower in rural areas, and only a college education offers significantly higher rural earnings. Less educated urban workers accrued the largest wage gains during the growth of the 1990s together with the college educated, while those with secondary laggedbehind. The primary educated and rural workers with some high school were the hardest hit in the slowdown, while the college educatedposted minor averageearnings gains. 4.11 Despite the significant earnings gaps, the lion's share of low paidjobs and thus poverty in Bolivia largely reflects low productivity levels in the economy as a whole. Simulations show that reducing labor market disparities would have a larger direct impact on inequality than on poverty (Chapter 5 and Gasparini et al., 2004). This again reflects the pervasive low base earningsof most workers. About 60 percent of workers with complete secondary education are at risk of poverty (with wages yielding below $2/day PPP).4 Only a college education ensures a sufficient level of earnings to escape poverty. Thus, although education i s a major source of earningsdifferentials, it does not fully account for low-payingjobs. 4.12 Yet the low and skewedreturns to education can leadto "poverty traps." Facedwith low returns or dimprospects to complete a full course of secondary or higher education childrenfrom poor families drop out of school. Addressing the sources of low returns to secondary education and labor market inequities can contribute to accelerate primary to secondary transitions and unleash Bolivia's potential for productivity growth. The DemographicOpportunity 4.13 Demographic forces offer Bolivia an opportunity to reduce poverty. Apart from Haiti, Bolivia i s the only LAC country entering the stage of demographic transition where the "dependency ratio" (the fraction of the population who i s too young or too old to work) will decline for the next two decades (Figure 4.3). As Bolivia goes through this transition, labor force participation i s expected to rise. This can reduce poverty if the economy generates sufficient qualityjobs on the basis of higher labor productivity. 4.14 The demographic transition also allows the translation of the human capital accumulation of young cohorts into a more productive labor force. Bolivia's labor force has uneven educational accomplishments (Figure 4.3) with the main gap at the secondary level. Bolivia has a relatively high fraction of workers with some college education (only below Argentina, Panama, Chile and Costa Rica) but a large majority with primary or less (only lower than Nicaragua and Honduras). As the share of younger cohorts in the working age population will rise faster, older and poorly educated workers are replaced with younger workers who could have basic and secondary education. This i s a gradual transition and it would take more than a decade for skill investments to translate into a more productive labor force and improvements in national and family incomes. 4. IDB (2003). 45 EstablshinatheBasisfor Fro-Poor Growth Figure 4.3: Bolivia's Demographic Transition and Human Capital Accumulation Window of Opportunity (dependency ratio) Roliviaand 1,atin America l5 Note: Dependency ratio=(population age 65 and older or 15 and under)/population age 15 to 64. Source: IDB (2003). THEINSTITUTIONSANDPERFORMANCEOFBOLIVIA'S LABORMARKET 4.15 This section describes Bolivian labor regulations and compares the performance of the labor market with other countries. The analysis draws heavily on the recent IDB publication on regional labor markets (IDB 2003).5 4.16 Bolivia has overly protective labor regulations. Figure 4.4 shows the results of a recent study of labor legislation worldwide. The index captures the strictnessinregulations for hours of work, night shifts, leaves, and maternity benefits.6 On a scale ranging from 0 (least restrictive) to 1(most restrictive) Bolivia scores around 0.9, placingBolivia at the top of LAC and even above some countries inEasternEurope, the region with the most restrictive labor laws. 4.17 Bolivia's labor legislation (dating from 1943) restricts firms' abilities to adjust to economic conditions, both prosperity and downturns. Bolivia i s the only Latin American country with no cap on the severance payment for dismissal (one month salary for every year of service), resulting in firing costs two or three times higher than most Andean neighbors and poor LAC countries. Dismissal regulations do not provide partial indemnities under justifiable voluntary resignation. Lay-offs due to deteriorating economic conditions for the firm and seasonal work are not allowed, and overtime has to be approved by the labor authority. The intendedprotection from dismissal discourages firms from hiring new employees. Bolivia fares well with regardsto minimumwage regulations. The minimumwage i s relatively low compared to average wages and has a relativelyhighlevel of compliance (Figure 4.5). 5. We thank Carmen PagBs for facilitating the figures from the IDB report reproduced here. 6. See Djankov et a1(2003) for more details on the methodology. 46 Figure 4.4: LegislatedEmployment Conditions inBolivia and LatinAmerica (Index, 0-1) EasternEuropeMd CentralA8 Lam Amsnca anu Caribbean 0 0 1 0 2 0 3 0 4 0 5 0 5 0 7 0 8 0 9 1 Source: Basedon Djankov et al. (2003), taken from IDB (2003). Figure 4.5: MinimumWage Levels inBolivia and LatinAmerica (as proportion of wages of the median worker, 1996-2001) Nicaragua Colombia Casta Rica Peru Chile Salwdor Brazil Argentina Mexico Honduras Bolida Paraguay Uruguay o?/o 10% 20% 30% 40% 50% 80% 70% 80% EK Note: The wage usedfor comparison is the median wage for workers beween26 and 40 years oldthat work for more than 30hoursinthereferenceperiod of the surveys. Source: IDBbasedon officialcountry data. 4.18 Labor laws include provisions that discourage employers from hiring women and hinder effective enforcement. Women face a shorter work week, night work prohibition, weak domestic work regulations (e.g., irregular work hours) and employer-paid maternity benefits that can not be shared by husbands. Provisions against discrimination in remuneration and merit promotion are weakly enforced. Enforcement and labor conflicts resolution are too centralized in the underfundedand stretched labor ministry (e.g., conflict settlement councils are to be presidedby the labor authority inLa Paz). Incentives to evade stiff regulations make enforcement a daunting 41 EstablkhinatheBasisfor Pro-Poor Growth task. In terms of the degree of cooperation in labor relations, Bolivia ranks low according to employers' perceptions (below only Venezuela, Uruguay and Paraguay) and around the L A C average according to the opinions of the urban p~pulation.~ 4.19 Because so few workers are hiredinthe formal sector where labor laws potentially apply, most Bolivians do not enjoy their protections. About 64 percent of workers in main urban centers were self-employed or worked in micro-enterprises in 2002, up from 58 percent in 1993. A small fraction of the poor and extreme poor hold formal employment. Despite its heavy fiscal burden, social security coverage rates (aside from the Bonosol minimum pension) are only 11 percent for blue collar employees, 45 percent for white collar workers, and less than 10 percent among micro-firms workers. The fraction with the full benefits o f regulated contracts i s likely lower given the laxity and highcosts of law enforcement (via courts or settlement). 4.20 The one-size-fits-all approach to labor regulations i s misaligned with productivity levels of micro-firms and many SMEs. Figure 4.6 shows that dismissal costs are among the highest in LAC. Mandatory non-wage benefits (primarily social security) can amount up to 8 percent of a small firm's sales. Firms with a small scale of operations and low productivity cannot afford a wage plus these additional benefits at a manageable cost, and this drives them towards unregulated employment. Some small firms register to operate and report taxes, but do not comply with labor laws. The international evidence shows that non-salary costs are largely borne by employees through lower wages, and they can reduce total employment levels. Given low income levels and the poor quality o f public services, many workers opt for more suitable alternatives to regulated benefits, including universal support systems (such as the infant- maternal health care and minimumBonosol pension), that grant a minimumlevel of protection. Figure4.6: The Cost of Job Security, Bolivia and LatinAmerica, 1987 and 1999 EC& PslY CdWrba Bra211 V-Ua BdlYla cea18RIcB HCPLWBP P m Chile MBXlW El S v a W Average, 199 I\rgenlna Dan"lcanReplbl1C Ni.Xagra urQ ~ Y p-eglay 0 2 4 6 8 10 12 14 16 18 Cost of benefits measured in multiples of wages Source: HeckmanandPagBs (2003). 7. Basedon the results of cross-country surveys of employers' perceptions by the World EconomicForum (2001) andthe Latinobardmetro (1997) opinionpolls inLAC as reportedinIDB (2003). 48 Constraints to Human CaDitalAccumulation-The SUDUIY labor of 4.21 Informality can trap labor and other resources in low productivity activities.8 Many micro and small enterprises lack access to financing which would allow them to invest and innovate and thus increase labor productivity and wages. At larger scales of operation productivity is higher, there are benefits from participation in formal institutions (e.g., access to credit and external markets), and regulatory costs become more manageable. Many small firms may be trapped in a bad equilibrium: while low productivity and hence low earnings mean they cannot afford onerous business and labor regulations, informality limits the potential for productivity growth. Figure 4.7: Performance of the Labor MarketinBolivia Compared to LatinAmerica (circa 2000) Summary Efficiency Allocation of Incomes Summary Index of Equity inAllocation of Incomes L e s s efficient Summary Ranking of Insurance Summary Measure of Overall Efficiency in Labor Market Performance 4.22 A look at labor market outcomes confirms that Bolivia's labor regulations are hampering economic growth and, at the end of the day, the ability of workers to earn a better living and have job security. Figure 4.7 provides some regional comparison^.^ The income efficiency index captures the earnings differentials between workers of seemingly equal productivity (i.e., across gender, firm size, and sector of activity) and the equity index adds the differentials due to endowments (i.e., education). The insurance index captures the statutory benefits for job loss and social security and the percentage of workers actually covered by those benefits. Finally, the 8. De Soto (2000). 9. See IDB (2003) for a more detaileddiscussionof the methodology andresults. 49 EstablishinutheBasisfor Pro-Poor Growth index of overall efficiency factors in unemployment indicators and the cooperation in labor relations. Except for the insurance index, higher values imply a better functioning labor market. By these measures, Boliviaranks among the LAC countries with the least efficient and equitable labor markets. 4.23 In sum, Bolivia's labor regulations are among the least effective inprotectingworkers in LAC. The legislation imposes excessive regulatory costs that encourage informality. As a result, many viable small firms have a limited ability to scale up productivity and employment through investment and innovation. Thus, not only are Bolivia's labor regulations ineffective in protecting workers, but they also hinder the country's long-term productivity growth. INFORMALEMPLOYMENT AND EARNINGS 4.24 This section analyzes the ability of informal employment to create quality jobs and thus alleviate poverty in Bolivia. The informal sector has often been portrayed as a source ofjobs for workers who cannot obtain employment in the formal sector. It i s seen as an obstacle to greater productivity inthe economy and to increased per capita incomes for workers. Mountingresearch has questioned this monolithic view (for example, Maloney, 2004). Many informal jobs may reflect the voluntary choice o f workers given their preferences, skills, and competing earnings prospects. In Bolivia, about 44 percent of workers in the top earnings quintile are informal salaried or self-employed, many of whom conform to the characteristics o f an entrepreneurial sector. 4.25 Given our focus on job creation, we distinguish three groups o f workers:" informal salaried workers (workers in establishments with one-to-four employees plus domestic employees), formal workers (workers in establishments with five or more employees), and the self-employed (those who self classify as cuenta propia, employers o f micro enterprises and cooperative workers). With these definitions, 41 percent of employed individuals 15-65 years old in urban Bolivia were self-employed in 2002 and 50 percent were salaried workers (the rest includes unpaidfamily workers), o f which 14percent were informal. 4.26 The results indicate that to a great extent informal employment reflects the low opportunity costs and non-wage benefits of informality. For many Bolivians it offers a competitive alternative to low-productivity formal sector jobs or no work at all, Self- employment i s particularly attractive for certain sectors of the poorer population, such as women seeking flexible work hours to balance their work and family obligations, or the indigenous who may face less discrimination as independent workers than they might as salaried employees. The fact that the self-employed subjectively rate themselves as less poor than salaried workers with similar characteristics i s consistent with non-monetary benefits o f self-employment. However, most informal salaried workers do appear to face an earnings disadvantage when compared to formal sector workers with similar characteristics, reflecting the comparatively low productivity of smaller firms. 10. The definition of the informal sector depends on the aspect of the labor market that is the main focus of analysis (Lay 2001). A focus in social protection uses an alternative definition based on legal protections and benefits that informal workers lack. This i s strongly correlated with firm size inBolivia. 50 Constraintsto Human CWtaIAccumulation-The SUDDIVof Labor Profilingthe informalsalariedandself-employed 4.27 As in most of LAC, there are many similarities but also some important differences betweenthe informal salaried and self-employed. These are summarized thus: 0 The Bolivian self-employed are heterogeneous-from street vendors to small artisans under subcontracting production arrangements. They tend to have higher potential experience, the longest tenure in occupation, and a higher fraction wanting to work more hours. 0 The informal salaried sector mostly comprises workers in small shops (mainly in commerce or transport) with no contractual arrangement or benefits such as pensions and health insurance. On average, they are four to nine years younger than formal workers and the self-employed, have eight years of schooling, shorter occupation tenures, and are more likely to live with their parents. 0 Both groups work more hours than the formal sector employees and are disproportionately female (about half) and indigenous (44 percent o f the informal and 61 percent of the self-employed in 2002). 0 Despite similar average education, educational achievements are more diverse among the self-employed than among the informal salaried. While a higher fraction of the self- employed have not completed basic education (23 percent versus 16 percent in2002) the share with university education i s twice as high as among the informal salaried (8 percent and 3 percent in2002). I s thereinformal-formallabormarketsegmentation? 4.28 Medianhourly earnings in 2002 were 70 to 90 percent higher inthe formal sector than for the self-employed and informal salaried. Total earnings are also lower for the informal (despite working an average of 5 hours more per week). During the boom (1993 to 1997) the self- employed did relatively better than the informal salaried, but this pattern reverted during the recession (1998 to 2002). In 1997, the formal salaried earned on average 2.2 times above the income poverty threshold while this ratio was 1.6 for the self-employed and 1.2 for the informal salaried, an improvement relative to 1993 ratios of 2.6, 2.1 and 1.3. With the growth deceleration these earnings ratios became 3.3,2.0 and 1.9, respectively. 4.29 Yet the gaps in average hourly earnings among these three groups cannot be taken as evidence o f the superiority o f formal sector jobs. While the informal are relatively poorer this may have little to do with informality per se. Lower informal earnings may reflect the lower skills of workers (both observed and unobserved) rather than the characteristics of informal jobs. Most importantly, monetary earnings gaps do not account for quality differences in job characteristics such as flexible work schedules or benefits such as health insurance or training. Moreover, gaps in average earnings may mask the differential situation faced by workers whose unobserved characteristics place them at jobs with below or above average earnings for their skills sets. 4.30 We address these issues in two ways. First, we examine whether workers are "pulled in" or "pushed out" of the formal sector by examining the evolution o f the relative size and earnings of the three sectors. Second, we conduct a multivariate analysis to compute earnings gaps at different levels of pay adjusting by differences in sector participation and productivity-related 51 EstablihinnatheBasisfor Pro-Poor Growth characteristics o f workers. Below we discuss the key findings (see Annex 4.2 and Tannuri- Pianto et al. 2004 for more details). 4.31 The trend data do not support the existence of segmentation for the informal salaried, but suggest that some workers may have been pushed out to self-employment with the crisis. Table 4.1 compares employment shares of the three sectors over 1993-2002 in metropolitan areas with relative earnings at different points of the earnings distribution, that i s at low (lothpercentile), moderate (50") and high paying (90*) jobs. For example, in 2002 median earnings of the informal salaried were 59 percent those of the formal salaried. The view o f informal employment as inferior implies that the relative size of informal employment should move anti- cyclically with earnings (Maloney, 2004). The noteworthy findings are: 0 Informal jobs offered competitive earnings during the 1993-97 boom, but faltered with the slow down. Duringthe 1990s earnings grew faster for the self-employed, particularly at the lower pay jobs, and almost as fast for the informal salaried. Meanwhile the share of formal employment increased from 40 percent to 44 percent, self-employment rose slightly, and informal salaried employment fell from 22 percent to 16percent. 0 With the growth deceleration during 1997-2002, however, both the informal salaried and self-employed sectors expanded (to 21 and 47 percent, respectively) at the expense of formal employment (32 percent). Meanwhile, the informal salaried performed relatively better throughout the earnings distribution than formal workers. However, self-employed earnings fell at the lower payjobs as the sector expanded. Table 4.1: Employment and Earnings Ratiosby Sector, Metropolitan areas SectorNear 1993 1997 2002 Formal (inpercent) 40.03 44.23 32.13 Informal (inpercent) 21.45 15.85 21.27 Self-Employed (inpercent) 38.52 39.92 46.60 QuantileISec tors InformaWormal 10percent 0.43 0.44 0.64 50 percent 0.40 0.37 0.59 90 percent 0.33 0.30 0.41 QuantileISectors Self-EmployedlFormal 10percent 0.45 0.64 0.37 50 percent 0.60 0.72 0.64 90 percent 0.75 0.72 0.74 Source: Authors' estimates basedon household survey data. 4.32 The results suggest that during the economic boom, businesses expanded and absorbed many unskilled workers. This resulted in relative gains in low-paid, self-employed earnings while the size of the sector remained constant. As the economy decelerated many small businesses may have failed or laid off least productive workers, pushing many to self- employment as wages at low payjobs lagged. 4.33 The earnings regressions results suggest that informal employment to a great extent reflects the low opportunity costs and non-wage benefits of informality for many Bolivians. The main findings are: 52 Constrain& to Human CapitalAccumulation-The SUDDIYof Labor Lower slulls (e.g., 3 fewer years of schooling on average) explain almost the entire earnings disadvantage of the urban self-employed and about two-thirds of the gap for the informal salaried. The urban self-employed enjoyed returns to skills similar to the formal sector. Despite being overrepresentedin the informal sector, the indigenous and many women find in self-employment more mobility and flexibility in working conditions than in the patronage-tenured-basedformal sector. This i s consistent with the result from the subjective poverty analysis that the self-employed report themselves less poor than salaried workers with similar characteristics. However, informal salaried workers, particularly those at the bottom of the salary scale for their skills set, face quite large earnings penalties and may appear to be rationed out of formal jobs. Workers at the best payjobs for their skills set can voluntarily become informal without significant wage penalties. During the economic boom workers seemed capable of moving freely between the formal and self-employed sectors without any wage penalties. By 2002 the earnings opportunity cost of informality increased for workers in lowest pay jobs, including the low pay self- employed, who endured the brunt of the slowdown and rising unemployment. The lowest productivity workers mighthave been pushedout of the formal sector. The largest earnings penalties for poor householdsfrom non-formalwork are associatedwith being female. Penalties for ethnicity are quite small, though correlated with poverty. Low educated (male) workers do not receive large penalties from participating in the informal sector, while for those few poor workers with higher education the penalty of non-formal work i s quite high. 4.34 In sum, as inthe rest of LAC, the informal urban sector inBolivia is very heterogeneous, comprising an upwardtier of largely voluntarily informal workers and a lower tier for which the sector may function as a safety net. For many workers, especially the self-employed indigenous and women, informality offers more flexibility and mobility opportunities than the rigid formal sector. However, the majority of informal salaried workers seem to be at a substantial earnings disadvantage, reflecting the comparatively lower productivity of smaller enterprises. This low productivity i s directly related to their condition as informal, which limits their access to credit, to associations with other producers intheir sector, worker training, innovation, or other kinds of productivity-enhancing interventions. 4.35 The findings highlight that policy makers should not simply seek to reduce informal employment. The sizeable informal sector i s largely a reflection of voluntary choice given the overall low productivity of the Bolivian economy and inflexible and costly regulations. Broad productivity gains are needed to improve the quality of both formal and informal jobs through appropriate incentives and better business and labor regulations (addressed in the policy recommendations section of this chapter andthe previous chapter). RURAL-URBAN MIGRANTSINTHE LABORMARKET 4.36 Internal migrationserves to balance labor markets. Migrants leave the regions with fewer economic opportunities and move to the more dynamic regions. Individual migration decisions respond to economic opportunities and are affected by numerous factors: standards of living, unemployment rates, distances between origin and destination, family size, and the age and 53 Establishina theBasis for Pro-Poor Growth education of individuals. Migration flows have consequences for labor markets in recipient regions, demand for public goods, public expenditure, investment, poverty and inequality. 4.37 In light of the variation of poverty and inequality across localities, particularly the marked urban-rural poverty gap, internal migration could be important for poverty reduction in Bolivia. Yet little i s known about the relative socio-economic performance o f migrants and the characteristics o f those who integrate better to the urban economy. These issues are examined in the background work by Tannuri-Pianto et al. (2004) (Annex 4.3 summarizes the methodology). The analysis focuses on rural-urban migration since these workers are typically less skilled than the urban-metro ones, and thus may have a harder time integrating into the more developed urban labor markets. The focus i s mostly on householdheads that migrate since families usually either migrate together or leave the migration decision to the head of household. 4.38 Overall, the findings point at significant cross flow between urban and rural labor markets in Bolivia. Rural, urban and metropolitan areas are both sources and destinations, but metropolitan areas have recently become the principal destination of migrants. Rural-to-urban migrants did better than expected had they not migrated, particularly those getting jobs at the bottom of the earnings scale for their skills set. Migration Flows and the Profile of Migrants 4.39 There are important differences in the dynamics o f migration flows during the economic boom and recent slowdown. Figures 4.8 and 4.9 show the source and destination areas of migrants who migrated during 1993-97 and those who migrated during 1998-2002. Each bar depicts the fraction of migrants coming (top graphs) and going (bottom) to rural areas, small urban centers and large cities. The main findings are: 0 Upto 1997, the rural areas were the largest source of migration and the main attraction poles were urban areas (Over 60 percent of migrants went to small cities or metropolitan destinations). However, there was also significant migration to rural areas on the order o f rural-metropolitan migration. Migrants coming from rural areas preferred to go to other rural areas while those leaving small urban and metro areas were mainly attracted to other urban areas. 0 Migration to main urban centers has accentuated since the late 1990s, especially from small cities. Urban migration flows increased by about half from 1997 to 2002. The metro areas were the main attraction pole for migrants coming from all three regions. Yet there remains significant migration to rural areas, including an increase inurban-to-rural migration. 4.40 Overall, this suggests that urban and rural labor markets are interlinked with significant cross flow. The significant migration to rural areas may be linked to family networks that help reduce migration costs or provide a safety net for migrants that do not exist in urban areas, Also, these migrants maybe filling niches in the labor market at destination. A relatively low-skilled migrant in the metro areas could be relatively highly skilled in rural areas, compensating his migration. 54 Figure 4.8: Migration Flows inBolivia by Figure 4.9: Migration FlowsinBolivia by Origin and Destination, 1997 and Origin and Destination, 1997 and 2002 (Heads of Householdonly) 2002 (Heads andNon-HeadsofHousehold) Main source areas of migrationand destinations Mainsource areas of migrationand destinations 1991 2002 1997 2002 30 30 20 20 10 10 0 0 Rural Urban Metro Rural Urban Rural Urban Metro Rural Urban Metro Metro ~ ~ Mainattractionpolesand originof migrants Main attractionpolesand origin of migrants 2 3 -60 `60 - 1997 2032 50 50 - I I HI 30 30 20 20 10 10 I 0 0 Rural Urban Metro Rural Urban Rural Urban Metro Rural Urban Metro 1 Source: Own estimatesbased on householdsurvey data Source: Own estimates basedon household survey data. 4.41 However, absolute migration flows remain small. About 350,000 people had migrated in the five years previous to the 1997 household survey sample and only 69,000 were rural-to-urban migrants. The figures are similar for 2002. This partly reflects the high costs o f migration and, as noted before, non-pecuniary factors that affect settlement decisions. 4.42 In terms of characteristics, a striking difference between migrant and non-migrant household heads i s the overrepresentation of women: 36 percent of the rural-urban migrants were female in 2002. Migrants are younger by about six years compared to non-migrants, and typically have smaller families. In 2002, rural migrants had on average 7.3 years o f education, while rural non-migrants had only 4.5 years, and urban non-migrants 9.2 years. Participation in the formal, informal, and self-employed sectors in 2002 was similar for migrants and non- migrants. Younger people, whose gains are increased by the longer expected payoff period, are more likely to migrate. The more skilled and better educated workers are also more likely to migrate. 4.43 The results indicate that rural to urban migration in 1993-97 was led by younger and better educated heads of household with smaller families seeking work in more developed areas who probably face less credit constraints and higher expected earnings gains. Rural migrants do not come from the poorest or relatively least developed areas (as measured by the HDI-income and the infant mortality rates) and communities with higher levels of education have less migration, maybe because of the positive externalities generated by education. Individuals from the poorest locations and indigenous household heads are more prone to rural-to-rural migration. 55 EstabkhinutheBasisfor Pro-Poor Growth Earningsof migrants 4.44 Rural migrants to urban areas duringthe 1990s didbetter than would have been expected had they stayed in rural areas, particularly those gettingjobs at the bottom of the earnings scale for their skills. The returns to education are higher for migrant heads than for non-migrants (except at the college level) in 2002. There i s relatively large heterogeneity in rural areas, allowing the less skilled (for the urban labor market) migrants to benefit (in absolute terms) from migration. Despite potential obstacles such as lack of contacts, "urban know how" or skills that are demanded in the urban economy, migrants readily found competitive economic opportunities. 4.45 However, migration i s advantageous only for those with migrant-like characteristics. If a person has observable characteristics that do not conform to the typical migrant profile (described above) but still migrates, she will earn less than expected at destination. Thus, when individualswith large families coming from very poor localities migrate they tend to earn less in the urbadmetro sector than what their skills would predict. 4.46 Migrant earnings lagged with the economic slowdown. Inparticular, the performance of female heads o f household and indigenous individuals in the urban labor markets changed considerably from 1993 to 2002. Female migrants fared worse than their urban counterparts in 1993 and 1997, but there was a premium for female migrants at jobs with lower pay for their skill sets in 2002. On the other hand, indigenous migrant heads gettingjobs at the bottom of the earnings scale lost ground relative to non-migrant indigenous heads in2002. 4.47 The results are consistent with the change in the direction o f migration flows over the 1990s to early 2000s. There is little evidence of crowding out of low performing migrants in main urban centers during the growth boom. But with the growth stagnation, there was some reversion of migration flows towards rural areas. 4.48 Although migrants do not come from the poorest rural areas, rural-urban migration likely had a direct (and, through remittances, indirect) contribution to reduce rural poverty. However, the potential for rural residents to escape poverty through migration i s limited by the high costs of migration, the fact that the poorest people do not migrate to the urban areas, and by non- pecuniary aspects of resettlement decisions. WHICHPOLICYINTERVENTIONSCANENHANCE THE LABOR MARKET'S ROLEINREDUCINGPOVERTY INBOLIVIA? 4.49 The poverty and inequality impacts of growth can be enhanced by removing a number of obstacles faced by the poor to improve their productivity and market their labor. This includes steady support to their accumulation o f human capital in future years, and actions to improve equity in the labor market and income generation investments for lagging regions (including easing internal migration costs). Below are selected short-to-medium-term actions. 4.50 Strengthening human capital and social protection of the poor. To help address inequities that limit the ability of many poor Bolivians to access and benefit from entry into the labor market, the Government should continue and improve efforts to raise human capital, particularly in education and health. Given the fact that Bolivia's social spending i s already relatively high as a percentage o f GDP, social spending policy actions should focus on targeting 56 Constraints to Human CaRitalAccumulation-The SUDRIVof Labor worse-off localities, improving expenditure accountability and efficiency through close monitoring and results management, andfinding creative ways to tap private sector participation. Two specific areas of attention should be: 0 Implementing a "smart" conditional cash transfer program targeted to the poorest families with at-risk children, to offer minimumincome support to the poorest families made conditional on their participation in health and nutrition interventions (pregnant women and children age 0-6) and on school attendance (children age 7-12). Bolivia can learn from the very successful Mexico's Oportunidades, Brazil's Bolsa Familia, and similar programs in Honduras, Nicaragua, Colombia and Ecuador. The program can be piloted on high priority groups and grow gradually over time. Financing from natural gas tax proceeds would channel the benefits of this enclave sector to improve the welfare and human capital o f the Bolivian poor (regional cost norms range around 0.5-1 percent of GDP). 0 The program would complement ongoing supply-side efforts to advance the health and nutrition related MDGs and social protection (with Bank support). These include extending coverage o f primary health care in underserved areas, strengthening the municipal health service network, sustainable water and sanitation provision, and strengthening pro-poor community investments and the workfare program (PLANE). 0 Developing the Bolivian Educational Strategy geared towards the development of basic cognitive skills and the productivity o f the labor force, for which a Bank education sector study is to provide specific recommendations. Key issues for the mid-term implementation of the strategy include filling coverage gaps for universal basic education, improving secondary education transitions, addressing low quality and inequalities in achievement tapping on results-based management, improving private higher education access for low- income students, and developing labor market skills through computer and English-language instruction. 4.5 1 Reform labor market regulations to promote formality and productivity increases. The high rate o f informal employment reflects the high costs and small benefits o f employment inthe formal sector. Labor regulations should allow for growth to generate more andbetterjobs for the poor, balancing workers' effective protection with productivity growth. There are many policy options which could encourage increasing the productivity of both formal and informal jobs. These include: Establishing a cap for severance pay to bringexisting negative incentives for hiringbelow the regional average, while allowing workers to collect partial indemnities in some cases of voluntary resignation; Extendingthe use o f term contracts tied to special provisions (minimumrequired training investments) to make high-rotation hiringless attractive; More flexible rules concerning dismissals for economic reasons and seasonal work, at the same time that effective anti-cyclical safety nets are strengthened; Incentives for worker training such as similar tax treatment to investment inhuman capital as that of capital investments, co-investments of workers, and mutually agreed post-training job assurance by workers and firms; Provisions to accommodate different non-wage costs for smaller firms, under mutual agreement between employers and employees, such as simplified health/pension plans; 57 Establishinu the Basis for Pro-Poor Growth 0 Leveling contracting conditions across gender and ethnicity (including domestic work), strengthening anti-discrimination provisions, equalizing the length of the work week/day and co-sharing maternity leave benefits; and 0 Employingfair and timely out-of-court mechanisms to resolve labor disputes, possibly includingneutral arbitrationor mediation councils. 0 Institutional strengthening (staffing, training, technical assistance) of the Labor Ministryand coordination of relevant public agencies i s needed so that they can assume their important and increasingly more complex role. 4.52 As in much of the region, the labor market reform agenda has been stalled. There is a need for effective communication to workers, firms o f all sizes, and central and municipal governments o f the benefits from reforms, includinghigher productivity, more jobs with viable social benefits to more workers, potential higher fiscal revenues, and a more fluid labor market where rewards and mobility are based on skills and merit. The greater recognition of the importance of sound labor market regulations for productivity growth has recently motivated some important reforms in countries like Colombia, Chile and Peru and ongoing discussions in others like Ecuador. 4.53 Improve labor market equity. In order to increase labor market equity the government could also consider interventions to reduce earnings gaps that are not related to productivity. These would benefit from civil society and private sector sponsorship or support (learning from relevant experiences inthe region, e.g., Peru, Chile), andthey could include: 0 Encouragement for expansion of pre-school facilities and child care centers to facilitate women's and migrants' participation, learning from successful experiences in the region (e.g., Peru, Chile); 0 Training in high schools and colleges on job search techniques and strategies and promotion of privately-provided labor market intermediation services; 0 Fostering community-led crime prevention and control inmarginal urban neighborhoods that may prevent workers, especially women, from taking advantage o f available (night) job opportunities; 0 Considering moderate, productivity-driven increases in the minimumwage, as the economy resumes growth, closely monitoring for rising unemployment among the young and unskilled to avoid setting the level too high. 4.54 Increase economic opportunities in worse-off regions. Also needed are income generation investments for lagging regions and actions to ease internal migration costs so that the poor are better able to benefit from locations and productive endeavors with better economic prospects. Policy actions include: 0 Improving urban and inter-regional transportation to reduce commuting costs that hinder job access; 0 Using the newly developed consumption poverty map (UDAPE-INE)to guide the allocation of resources towards poor municipalities, particularly to target interventions aimed at income generation of the poor. This map could be considered for implementing the National Dialogue Law formula of inter-municipal transfers. Bolivia's geographic poverty profile calls for a wide range of interventions depending on local poverty and inequality levels, 58 Constraintsto Human CaDitalAccumulation-The SUDD~VLabor of location and resource endowments. These may include growth-enhancing investments, targeted programs to develop human capital, community assets and income generation, and investments that promote gradual integration o f communities through migration. U.S. Millennium Account resources could be tapped to finance entrepreneurial and social activities for lagging localities. 4.55 It is important to emphasize that the actions noted above should be accompanied by complementary reforms to achieve general improvements in the investment climate and bundled interventions to align the incentives of micro and small firms to participate in formal institutions, as outlined in the previous chapter. This will allow capitalizing on the synergies of advancing multiple reforms. 59 Establisbina theBasisfor Pro-Poor Growth 60 5. PROSPECTSFOR GROWTH AND POVERTY REDUCTION Although Bolivia's growth prospects remain vulnerable to external circumstances, the country can strongly improve its future growth potential through continued reforms. Simulations suggest that ifthe policy determinants of growth were to improve to that of the 7 f hpercentilefor the LAC region, GDP per capita growth could increase to over 4 percent per year, or over 5 percent a year if policy variables were to improve to the 7Sh percentile level of the world. Because of the large number of poor people in Bolivia and the skewed income distribution, high growth rates are necessary if the country's poverty level is to be signijicantly reduced. Considering the existing growth-poverty relation, annual per capita growth rates of 6.5 percent are required to meet the MDG goal of reducing poverty to 34 percent by 2015. The annual per capita growth rate required to achieve the alternative MDG target of reducing the incidence of extremepoverty by half by 2015 is about 1 percent lower. Simulations indicate that individual policy reforms will have relatively small impacts on growth and poverty by themselves, but can have a much larger impact when implemented as part of a more comprehensive strategy of mutually-reinforcing reforms. 5.1 This chapter builds on the results inprevious chaptersto assessthe prospectsfor pro-poor growth. It discusses scenarios that would put the country back on a fast-growth path and simulates policy levers of factors driving the income distribution that can reduce poverty and income inequality. GROWTH PROSPECTS 5.2 Bolivia's medium-run growth i s tied to the natural gas sector development, including the ongoing hydrocarbon projects with Brazil and Argentina andpossible exports of liquefiednatural gas. Recent studies estimate that these projects could lead to at least a 6 percent growth rate if they reach full capacity by the end of this decade (IMF2004, Klasen et al. 2004). As an enclave sector, the impact on the rest of the economy would be mainly through increased tax revenues, current account effects, and exchange rate appreciation. Klasen et al. estimate that growth would be somewhat larger ifthe government refrained from spending all the additional revenues in consumption, instead increasing public savings and avoiding the contraction of export-oriented sectors. Dueto limited output and employment spillovers, the direct benefits for the poor would bemodest. 5.3 Further progress in structural reforms can help Bolivia ignite shorter term growth and sustain it over the long run. The growth results discussedin Chapter 1can be usedto simulate orders of magnitudes (Figure 5.1). With growth determinants following the trends of the 199Os, Bolivia's average GDP per capita growth can be expected to be about 0.5 percentage points higher than its average rate of the 1990s (reaching 2 percent annual growth) during the next decade, compared with a 1.1 percent per capita growth gain for L A C (total of 2.6 percent per year). If the policy determinants of growth were to improve to that of the 75" percentile for the L A C region, Bolivia's GDP per capita growth could increase 2.7 percentage points to over 4 percent per year (similarto other LAC countries). Finally, if policy variables were to improve to the 75th percentile level of the world, then growth in Bolivia could reach an average of 5.6 percent per year. 61 5.4 The simulations suggest that improvements Figure 5.1: GDPPer Capita Growth in education, infrastructure, and trade integration Projections, Bolivia and L A C Average, could contribute the most to growth followed by 2000-2010(%) improved governance and financial deepening. Larger spillovers to production and employment I t5 6 % . - - might occur if the gas projects are accompanied g 5% 4% with policies and reforms in these areas. This -g9 3% requires a combination of economic policies and 2% 3 E reforms to reduce macroeconomic vulnerabilities, 1% 0% boost private investment, and exploit natural .-T'wg n Continuous Variables Variables endowments more efficiently. Trend are at Top are at Top 25% of LAC 25% of WorM 5.5 It is beyond the scope of this report to formulate a pro-growth reform agenda. In Chapter Source: BasedonLoayza, Fajnzylber and Calderon, 2002. 3 we offer some policy recommendations to address key micro constraints to growth faced by Bolivian firms, which are discussed in more detail in the Bank's 2001 Bolivia Investment Climate Assessment. Growth constraints and trade issues will be further analyzed in the Bank's forthcoming Country Economic Memorandum, Sustainable mechanisms for funding public infrastructure should be pursued using standard benefit-cost criteria and where donor funding i s not available the Government can seek viable domestic or external financing options (including public-private partnerships). A forthcoming Bank education sector study will provide specific recommendations to develop a new Bolivian Educational Strategy, but some recommendations are discussed in Chapter 4. The Bank's 2004 Public Expenditure Review offers recommendations to improve governance. The strengthening o f the financial sector i s being supported through the Bank's Programmatic and Corporate Restructuring adjustment loan to improve access to financing (particularly for small and medium enterprises) including collateral law reform and strengthening creditor rights. 5.6 Bolivia's growth in the 1990s in the midst of a favorable external environment (high growth of its main trading partners, significant capital inflows, the expansion o f natural resource exports), and the subsequent decline demonstrate the country's vulnerability to external shocks and dependency on external capital. The growth projections above, based on deeper structural reforms, do not factor in possible shocks-either negative or positive. A large portion of government revenue comes from natural gas exports, and shocks in its price level could affect fiscal accounts. Andersen and Faris (2002a and 2002b) suggest a stabilization fund for the natural gas revenues in order to mitigate price volatility, although these may not be as effective when price shocks are non-permanent (Nina and Brooks de Alborta, 2001). 5.7 Bolivia relies on external resources for one-third of all public expenditure. As noted in the 2003 Bank CAS, forthcoming levels of foreign assistance, particularly from bilateral donors, are uncertain, and the availability o f domestic and foreign private investment under the country's current circumstances may not be high enough to sustain growth. The Government should continue to pursue short-term aid conditional on good policy performance and advances in the aid harmonization agenda. 5.8 Investments in human capital, transportation and communications infrastructure, combined with an adequate institutional framework, can improve the prospects for and increase the impact of local and foreign direct investment. If local firms have access to technical assistance, credit, human resources, and information, spillovers from FDI are more likely. The 62 ProsDectsfor GrowthandPovertv Reduction creation of distribution networks, adoption of new technologies and access to foreign markets could enable the poorer regions of the country to better enjoy the benefits o f international trade.' 5.9 In summary, the results of growth regressions indicate that the structural reforms that Bolivia underwent during the 1980s and 1990s contributed positively to growth, especially during the 1990s. Future growth will be limited without further reforms. Moreover, lessons from past performance urge further actions to increase the impact of growth on poverty reduction. PROSPECTSFORPOVERTY REDUCTION: ACHIEVING THE POVERTYMDG 5.10 Bolivia requires sustained growth to achieve substantial reductions in poverty. The income distributions are heavily skewed, particularly in rural areas, making it more difficult for distribution-neutral growth to bring about significant, immediate declines in poverty. This i s illustrated by Figure 5.2, which shows estimates of the urbadrural per capita household income distributions for 2002 and the corresponding poverty lines. These show the fraction o f individuals at each income level, the area under the curve to the left of the poverty line being the fraction of poor individuals. Figure 5.2: Distributionof Log Household Per Capita Income, 2002, Urbanand Rural 0.5 - 0 4 045- 0.35 0.4 - 0 3 035- 0 25 0.2 0 2 - 0.15 0 15 - 0.1 - 0.1 0.05 - 0.05 0 1 2 3 4 5 7 8 9 1 0 - 2 . 1 0 1 2 3 4 5 5 7 8 9 1 C - Note: the vertical lines ineach graph represent an "average" poverty line for each area. Source: Authors' estimates based on household survey data. 1. For a more detailed discussion on the role of FDI in Bolivia, see the Institute for Socio-Economic Research (IISEC) and Overseas Development Institute (ODI) "Foreign Direct Investment and Development: The Case of Bolivia." 63 5.11 We conducted simulations to assess the impact of growth on poverty, in particular to assess the per capita growth rate required to achieve the Millennium Development Goal of reducing poverty to 34 percent by 2015 (Figure 5.3). Under alternative growth rates, assuming the income distribution does not change, the goal could be met if incomes grew at an annual rate of over 6.5 percent per capita for the next 11years, which means more than doubling per capita income by 2015. The annual per capita growth rate required to achieve the Figure: 5.3: Impact of Per Capita Growth on alternative MDG target of reducing the , Poverty inBolivia incidence o f extreme poverty by half by Changes inPoverty Levels 2015 i s about 1percent lower. These growth rates are significantly higher than the growth 35 rates in the 1990s. Poverty would decrease 30 25 to 29 percent with an annual per capita 20 income growth rate of 8 percent, but even in 15 this unrealistic scenario, poverty in rural 10 5 areas would remain very high (54 percent), 0 not reaching the MDG of halving the a e g a e w, 7 e g g " ` a ? $ a a i -a&, e x -& , % IMPACT OFPOLICYREFORMSONPOVERTY AND INEQUALITY 5.12 We have made policy simulations that assess the impact of reforms to several of the principal causes o f poverty and inequality identified in this report. The simulations: (i) extrapolate trends in improvements of education o f the 1990s to 2015 and accelerating the 90s trends; (ii) returns to education at different levels and for workers in different points of increase the earnings distributions; (iii)continue declines in fertility; (iv) reduce informal employment; (v) shift employment from primary activities to the manufacturing industry and/or skilled-labor intensive services; and (vi) eliminate gender, race and informality earnings gaps. These simulations are limited and are not meant to be predictive, but merely to offer a sense of the potential orders of magnitude o f alternative policies. Some results are illustrated in Figures 5.4 and 5.5 (see Gasparini et a1(2004) for details). 5.13 The mainfindings of the policy simulations are: 0 Because of the low level o f total national income and the high number of poor people, poverty cannot be reduced much by purely redistributive policies. 0 The distributional impact of a productivity-driven earnings increase would be greater ifit took place in primary economic activities (e.g., agriculture) in rural areas. Generalized earnings growth i s about neutral, but pro-poor if it took place inthe sectors employing the poor (e.g., the unskilled labor sectors), and less pro-poor if it happens in the skilled- intensive sectors (Figure 5.5). 64 Figure: 5.4: Impact of Different Policies onPoverty and Income Inequality inBolivia Tax Redistribution Change inPoverty Levels Distribution of Per CapitaHousehold Income Share @attornDec. 0ShareToo Dec +-Gm 1 40 - 70 35 1 30 - 60 25 - 50 20 - 40 .e 15 - 30 10 - 5 20 0 - 10 t40% t S % t=10% tSo% t=lO% t S O % 7 Base2W2 140% t=20% t-30% Alternative Policy Scenarios Changes inTotal and Extreme Poverty Incidence Changes inthe GiniCoefficient Note: Educational Structure: S1 extrapolatestrends in the 90s to 2015. S4 simulates the simulation of illiteracy, increases in primary school completion, high school and college attendance. Returns to Education: S1 assumes increases of 20 percent in the returns to education. S4 equalizes the returns for workers with conditionally low wages to the returns of the median workers. Employment Restructure: S1 extrapolates trends in the 90s into the next decade. S4 simulates massive movementsout of agriculture to manufacturing, commerce, and skilled-labor intensive services. Znfonnality: S1extrapolates trends in the late 90s to 2015. S4 assumes a 20 percentage points increase in the share of formal employment from reductions in informal salaried and self-employment. Wage Gaps: S1 eliminates the unexplained gender wage gap. S4 eliminates the informal salaried and self-employed unexplained earnings gaps. Number of Children 4 2 S1 extrapolates trends inthe 90s to 2015. S4 assumes a decline inthe proportionof households with more than three children. Source: Authors' estimatesbasedon simulations with 2002 MECOVI (Gasparini et al. (2004)). A 5 percent annual growth in earnings in urban areas has a considerable impact on poverty reduction (from 53 percent to 29 percent) while a similar earnings growth in rural areas would reduce poverty from 85 percent to 71 percent (cf. growth incidence curves in Figure 5.5). Considerably increasing the educational level of the working population does not change substantially the distribution o f income directly, and leads modest reductions in poverty. The same i s true, all else equal, for increases in the returns to education. Sectoral employment reallocations to manufacturing, commerce and the services sector yield modest impacts on income distribution and small effects on poverty. 65 EstablishinatheBasisfor Pro-Poor Growth 0 The impact of an increase in formal employment caused by reductions in informal salaried and self-employed workers would bring a reduction in the national poverty headcount ratio of less than 2 points. 0 A decrease in fertility generates equalizing and poverty-reducing effects that are relatively larger than the above scenarios. 0 Closing the unexplained wage gaps between non-indigenous and indigenous, and especially between formal and informal workers, would benefit the poor but again poverty and inequality fall little. Closing the gender earnings gap has a negligible impact on household inequality, largely because working women are more likely to be in the upper part of the householdincome distribution. Figure 5.5: Growth-incidence Curves from Different Policy Scenarios Annual 5% Growth inEarnings, By Sector Annual 5% Growth inEarnings, By Area 4.5 30 2 5 2 0 ..... 1 2 3 4 5 S 7 8 9 10 -All-wwban... ..rural Scenarios s l and s4 for the Educational Closing the Wage Gap Due to Gender, Race and Structure Informality 3 20.0{ \ 2.0 ......... -- .- ......-............-.. 0.0 loo 0 0 1 2 3 4 5 6 7 8 9 1C 1 2 3 4 5 0 7 8 9 10 .......01+s4 , gender &race +informalny Source: Authors' estimations based on simulations with 2002 MECOVI. See Figure 5.4 for definitions of the simulation scenarios. See Gasparini et a1(2003) 5.14 From these scenarios we conclude that, although all will help reduce poverty, the impacts of any single factor are small. Purely redistributive measures attained by fiscal policy seem to achieve higher short-term gains in poverty reduction but, unless these are moderate, they may have a negative impact on growth and this hinder poverty reduction in the medium and long term. While Bolivia i s very unequal, it i s foremost very poor. The low overall productivity 66 ProsDectsfor GrowthandPovertvReduction means a very low income base for the poor. Even after a sizeable change in the educational structure, or employment composition of the economy, the average wage just grows 8 percent and 2 percent respectively (Table 5.1). The impact i s muted by the fact that the wage premium of finishingprimaryschool i s rather small for workers at the bottom of the earnings distribution. Table 5.1: Impactof EducationalUpgradingandEmploymentonWages (WagesandNumberof Workers, UrbanAreas) Numberof workers Numberof workers Average wage Base 2002 Simulation4 Average wage Base 2002 Simulation 4 (0 (ii) (iii) (0 (ii) (iii) No education 3.1 90 0 Primary activities 4.6 216 23 Primaryincomplete 4.2 634 113 Manufacturing 4.6 325 451 Primarycomplete 4.9 137 677 Utilitiesand transportatii 6.8 215 161 Secondaryincomplete 5.3 331 226 Construction 4.7 213 135 Secondarycomplete 6.3 314 677 Commerce 5.6 639 790 Superior 13.7 551 564 Skilled-services 10.9 647 677 Total 2257 2257 Total 2257 2257 Average wage Average wage Real 7.0 Real 7.0 Simulated (s4) 7.5 Simulated (s4) 7.1 Change (%) 6.0 Change ( O h ) 1.9 Source: Author's estimates based on household surveys. 5.15 The main lesson from the analysis i s that sustainable growth, based on policies and reforms to enable the conditions for increasing productivity poor households, i s essential to reduce poverty substantially in Bolivia. Given current productivity levels, pro-poor social policies alone can only achieve modest poverty declines. There i s no magic bullet in a poor country like Bolivia. Actions on multiple fronts, as those discussed in Chapters 3 and 4 of this report, are needed to maximize the pro-poor characteristics of growth. These policies can also contribute to promoting growth in the long term and thus enable a virtuous cycle of growth and poverty reduction. 67 EstablishhatheBasisfor Pro-Poor Growth 68 Annex 1.1 Data Sourcesfor MonitoringPoverty and Living Conditions inBolivia There are two main data sources for monitoring recent trends in poverty in Bolivia: two national censuses and various household surveys (although of incomplete coverage and limited comparability). CensusData The censuses (Censo Nacional de Poblacidn y Vivienda) provide data on non-monetary indicators of well-being. The most recent are for 1992 and 2001. They are used for building poverty maps and targeting government interventions using indicators of Unsatisfied Basic Needs (Necesidades Biisicas Insatisfechas). An updated NBI-based poverty map was recently preparedby INE-UDAPEwith the 2001 Census. Household Survey Nationally representative household surveys are unavailable for the early 1990s. Until 1996, INE's multi-purpose household surveys only covered large urban areas. This report relies on the following household surveys: Integrated Household Survey (implemented in capital cities up to 1995), National Employment Survey (national coverage; June & November 1996, and November 1997, the richest in terms of contents), Continuous Household Survey of Living Conditions (1999 through 2003). The coverage i s national. These surveys have benefited from the support of MECOVI. The data i s more comprehensive, including modules on health, education, occupation, income and expenditures. These surveys present serious limitations in Table A.l.l.l: Poverty LinesinBolivia, terms of comparability: changes over time in the 2002 Bs$per persodmonth questionnaires make comparison of even basic Extreme socioeconomic variables difficult (incomes, Poverty Poverty education), and surveys before 1999 did not capture household expenditures. We rely primarily Rural 233.39 133.03 on the 1993, 1997, and 1999-2002 surveys, paying Urban 321.8 170.9 meticulous attention to consistency and Sucre 335.6 169.5 comparability issues. We use the official poverty La Paz 327.0 181.8 lines derived from household surveys by INE for Cochabamba 351.3 177.4 each department (Table A.1.1.1). Official poverty oruro 297.4 165.3 Potosi 273.5 152.1 rates are based on household per capita Tarija 351.3 177.4 expenditures in rural areas and household per Santa Cmz 343.9 174.7 capita incomes for urban zones given the high Trinidad 343.9 174.7 volatility and unreliability of rural income data. ElAlto 272.2 165.2 Since 1999, the Bank, through the Source: INE MECOVI program, has supported INE onthe development of nationally representative annual household surveys that have significantly improved the comparability of data for poverty monitoring and analysis. As the MECOVI program comes to an end, more funding from the government and donors i s needed to continue strengthening the national statistical capacity and improving the use of surveys for poverty targeting and program evaluation. Currently, only 0.5 percent to 1 percent of the national poverty reduction budget from donors i s allocated to improving the information system. The medium-term challenges for the overall institutional framework and governance of the statistical system remain considerable. The main objectives 69 are: (i) continue to produce nationally representative surveys implemented on an annual basis, covering social indicators and living conditions, access to basic infrastructure and other welfare indicators; (ii) improve survey questionnaires, fieldwork, and quality control in a range of survey activities; (iii) encourage wide access to and use of the survey data for policy analysis, and feed the design o f public policy; (iv) build an integrated statistical system to ensure coherence among different statistical activities in the country and promote stronger interaction between data producers and users; and (v) improve the quality and coverage o f administrative data and capacity to collect data at the sub-national levels. National Survey onHuman Development Potential and Aspirations In 2000, the UNDP sponsored a nationally representative (a sample of 10,000 persons) household survey, also conducted by INE, which captures the values and aspirations o f the Bolivian people with respect to themselves, their communities and the country's institutions. This survey includes individuals' poverty self-classifications, as well as many of the variables included in the MECOVI surveys such as the household structure, employment condition, income, assets and access to basic services. We use this to construct self-rated (subjective) poverty profiles. The main limitation of this UNDP survey i s that it does not capture incomes o f all household members, or household expenditures. This precludes straightforward comparisons of objective and subjective poverty determinants relying only on data from this survey. However, it i s possible to compare the results with those obtained with the MECOVI survey, since they share a similar sampling framework and common survey instruments (e.g., questionnaires). Developing a Consumption Poverty Map inBolivia Household surveys rarely allow for reliable estimation of income/consumption poverty in small geographic areas (e.g., municipalities). While census data often allows considerable geographic dissagregation (e.g., at the municipal level), it does not capture household income/expenditures. A recently developed methodology derives consumption poverty maps by combining data from both censuses and household surveys to fill in the missing income data in the census (Elbers, Lanjouw and Lanjouw, 2003). This methodology i s applied to Bolivia to derive income/consumption poverty maps at the municipal level with adequate statistical confidence levels usinghedonic regression models. The main sources o f data are National Census o f Population and Housing of 2001 and the MECOVI household surveys o f 1999, 2000, and 2001, which were combined to obtain a larger sample and then disaggregated by departamento and area. The methodology linked household consumption expenditure with variables measured in both household surveys and the census to impute the missingincome data (Elbers, Lanjouw and Lanjouw, 2003), and i s developed in two stages. First, an econometric model of estimated household expenditure was generated as a function o f variables from the surveys related to household structure, durable goods and housing equipment, basic services, and socio- demographic characteristics o f the members. Second, the parameters of this model were applied to the census data to obtain conditional estimates o f per capita consumption expenditure for each geographic unit. These can be grouped into four categories: householdfamily characteristics (type o f household, number of members, number of children), dwelling characteristics (construction 70 materials, access to water, sewage, electricity, durable goods), characteristics of the head of household (age, sex, education level, employment status, occupation), characteristics o f the community or residence area (education level, natal care, materials and space used in the household, services and energy consumption). For more robust estimations the survey sample i s divided into more homogeneous geographic units, namely regions and departments. The estimation errors o f the model are composed o f two parts: one attributed to geographic or location effects, and the other related to individual or idiosyncratic errors. The first indicates the presence o f non-observable characteristics that affect household consumption in a certain area or community; to address this, explanatory variables were generated for each community or area (cluster), which predicted the location error with fixed effects. The idiosyncratic error i s then estimated from a theoretical distribution (normal, t-student) with sampling procedures with at least 100iterations. Once the distribution of the consumption expenditure for each geographic unit was estimated, it was possible to calculate the usual indicators of inequality and poverty together with empirical confidence intervals. The poverty lines used to measure poverty were estimated by taking into consideration the basket of basic products comprised of the types and quantities of food typically consumed within and outside the household, along with the minimum caloric intake recommended by the World Health Organization. The household reference group was composed by families whose food expenditure levels are able to cover the minimum nutritional requirements within an interval 1 percent, 5 percent, and 10 percent above and below the adequate level of nutrition. The poverty lines were obtained by averaging the values o f expenditure inthe interval. In practice, few households cover the totality of the basic caloric and protein requirements, and generally, they are not fully represented. These factors, combined with expenditure measurement errors, explain the use o f interval rather than ordinal estimation. Two poverty lines were used for poverty estimations: a higher poverty line, which considers the upper limit of the value of non-food products, and a lower poverty line, determining an inferior limit (Wodon 1997). 71 ANNEX 1.2 The Geography of Poverty and Inequality' This section provides more detail results of the recent work of UDAPEandNE,with the support of the Bank to develop a consumption-based poverty map for disaggregated geographic units. The analysis presents new data on poverty in Bolivia measured by consumption expenditure and alternative poverty lines combining information from the 2001 Census and MECOVI household surveys 1999-2001. Besides the extreme poverty line two other lines were used: a high poverty line which considers the upper limit of the value of non-food products, and a lower poverty line determining an inferior limit (see Annex 1.1 for details). The results pinpoint the more unequal areas and localities with the highest concentration of poverty and the impliedrelativerankings. While poverty i s widespread in Figure A.1.2.1: Incidence of Poverty by Departamento, Bolivia, there are important 2001 differences across locations. The departments o f Potosi and Chuquisaca have the largest poverty incidence, followed by Beni, L a Paz and Oruro, regardless of the poverty line used (Figure A.1.2.1). Over 70 percent of the population in these departments i s poor. The departments of Santa Cruz, Pando, and Cochabamba are the least poor at any poverty line, but even in Santa Cruz 40 percent of the I ,population i s poor. These rankings are Source: Based on data from the 2001 Census and Household largely similar for extreme poverty. Surveys 1999-2001. Over 60 percent of the population in Potosi and Chuquisaca i s indigent, compared to 25 percent in Santa Cruz. Due to population density the departments with the highest poverty rates concentrate the largest number of poor: Cochabamba and Santa Cruz concentrate over 42 percent of Bolivia's population while Potosi and Chuquisaca have less than 15 percent. The urban conglomerates of Beni and L a Paz present the highest poverty incidence (over 50 percent) while Cochabamba and Santa Cruz are among the least poor (almost a quarter of the population) (Figure A.1.2.2). The patterns o f urban extreme poverty are very similar. Urban extreme poverty in Beni reaches 37 percent, three times higher than Santa Cruz and Cochabamba, the two most vibrant local economies of Bolivia. Indeed, the process of urban development inthese two large cities has created broad-based income opportunities. Rural areas all over the country are overwhelmingly poor regardless of the poverty lines used (Figure A.1.2.3). Over 40 percent o f the rural population has levels o f consumption insufficient to meet the basic food needs. Even rural areas of Cochabamba and Santa Cruz have levels of poverty as high as those of rural Potosi and Chuquisaca. Extreme rural poverty is 1. Based on the report Pobreza y Desigualdad en Municipios de Bolivia: Estimacidn del Gasto de Consumo Combinando el Censo2001 y las Encuestas de Hogares (2003) ledby UDAPEand INEand onthe World Bank sideby QuentinWodon(AFTPM), Werner Hernani(consultant), andPeterLanjouw (DECRG). 72 particularly concentrated in Potosi and Chuquisaca (close to 90 percent). Rural Pando and Beni present the lowest levels of extreme poverty (half of their population). Figure A.1.2.2: UrbanAreas: Incidence of Figure A.1.2.3: RuralAreas: Incidence of Poverty by Departamento, 2001 Poverty byDepartamento, 2001 __ m Source: Based on data from the 2001 Census and Source: Based on data from the 2001 Census and HouseholdSurveys 1999-2001. HouseholdSurveys 1999-2001 The municipal disaggregation shows that, regardless of the poverty line, many localities show both low per capita consumption and severe poverty, while in others moderate pockets of poverty coexist with high inequality in per capita consumption. Figure A.1.2.4 shows the fraction of municipalities for the range o f poverty incidence and poverty gaps observed in Bolivia using alternative poverty lines, the vertical dotted lines signifying the corresponding national values. A large share of municipalities exhibit poverty incidence above the national levels (most o f the curves in the left-hand graphs (a-b) fall to the right of the dotted lines). In fact, a significant fraction of municipalities have poverty incidences above 80 percent (see the peaks of the curves). The curves depicting the distribution of poverty gaps (right-handgraphs) also suggest that most municipalities have poverty intensity levels above the national average, although the distribution of municipalities i s more symmetric at lower poverty lines. Poverty is particularly concentrated in municipalities located in the valleys and the central highlands. Municipalities endowed with resources and larger urban populations have developed at a relatively higher rate, tend to concentrate larger levels of economic activity and have higher levels of consumption. This i s typically the case o f capitals of departamentos and relatively larger cities. The municipalities o f Cochabamba and Santa Cruz present indeed the lowest incidence and intensity of total and extreme poverty (about 20 percent of their population i s poor). However, some municipalities outside department capitals (e.g., Montero, Colcapirhua, Puerto Quijarro) have also been successful in reducing poverty levels. On the contrary, in at least 20 municipalities with dispersed populations (Morochala, San Pedro Buena Vista, Ravelo), virtually all residents live with insufficient consumption levels to cover basic food needs. 73 Figure A.1.2.4: Distributionof Poverty Incidence and Poverty Gaps inMunicipalities, 2001 (a) Highpoverty line (b) Lowpovertyline Source: Basedon data from the 2001Census andHouseholdSurveys 1999-2001. 74 The results also illustrate the important local interaction between poverty and inequality. Figure A.1.2.5: Inequality inDistribution Bolivia also exhibits high levels of inequality in of Consumption Expenditure: Theil Index consumption both between and within departments. Figure A.1.2.5 shows the Theil inequality index which similar to the Gini coefficient ranges from 0 (equal) to 1 (most unequal). Potosi, Cochabamba, and Chuquisaca have the highest inequality in consumption expenditures. In contrast, Tarija, Pando, and Beni show a more equitable distribution of consumption expenditure, partially reflecting greater economic opportunities by virtue o f being located along the borders with Brazil and Argentina, having a lower proportion of indigenous population, among other factors. About half of the inequality level in the most unequal regions i s explained by the large Source: Based on data from the 2001 Census and HouseholdSurveys 1999-2001. rural-urban income gaps, while in the more egalitarianinequality mainly reflects disparities within both urban and rural areas. The municipal data does not Figure A.1.2.6: Inequality and Consumption suggest a clear cut correlation between average consumption and inequality levels. Figure A.1.2.6 shows the inequality indexes and per capita consumption of municipalities with the correspondingnational levels (vertical and horizontal lines). While most municipalities with low consumption levels also have low inequality, among those with higher average consumption we find both low and high inequality levels. That is, for most municipalities (the mass of points in the bottom of the graph), there i s little evidence that higher Source: Based on data from the 2001 Census and Household average consumption entails higher Surveys 1999-2001. inequality. However, in municipalities that start with high inequality at a given level of average consumption (the more dispersed points at the top), a rise in average consumption first lead to a more unequal distribution before greater equality i s achieved. These patterns should be taken as tentative since they reflect cross sectional correlations. The municipal data does illustrate how the high levels of inequality in Bolivia are a key factor behind the country's high poverty levels. Figure A.1.2.7 shows the relationship between inequality (total and extreme) and the levels o f poverty incidence and intensity for alternative poverty lines. Again the horizontal and vertical lines depict national levels for each variable that divideeach of the graphs into four quadrants. 75 Figure A.1.2.7: Inequality andPoverty Measures (a) Low poverty line (b)Extremepoverty line Source: Based on data from the 2001Census andHouseholdSurveys 1999-2001. High inequality and low incomes are pervasive in many localities. Three groups of municipalities are worth distinguishing. The largest group comprises municipalities with very high poverty and low inequality (bottom-right quadrants in each graph), which are mostly low populated, remote indigenous communities living in subsistence. The second largest includes those with both high poverty and inequality (upper-right quadrants in each graph) and comprises larger urban localities with better resource endowments and small cities dedicated to mineral exploitation or border trade with Brazil and Argentina. Finally, municipalities with low poverty and low inequality (bottom-left quadrants in each graph) are typically found in more economically dynamic urban areas. Bolivia's geographic poverty profile call for a wide range of interventions depending on local poverty and inequality levels, location and resource endowments. Finally, the analysis finds that the relationship between consumption and Unsatisfied Basic Needs (UBN) measurements becomes weak among municipalities with lower levels of poverty. Figure A.1.2.8 shows the relationship between UBNpoverty and the levels of poverty incidence and intensity at high poverty lines with the corresponding national levels (horizontal and vertical lines). Severe levels of poverty, evidenced by high proportions of population with UBNandlow consumption levels, characterize a large number of municipalities inBolivia (note 76 the sizeable concentration in the upper-right quadrants of the graphs). However, the relationship i s less clear among the municipalities with moderate poverty levels, precisely those where the needs of going beyond geographic targeting are greater. This suggests that the current system of inter-municipal transfers based on UNB poverty (formula of the National Dialogue Law) targeting tends to be less responsive to municipalities with entrenched pockets of income poverty. Figure A.1.2.8: Incidence of Poverty by UBNand (High) Poverty Line Source: Basedon data from the 2001Censusand Household Surveys 1999-2001. 77 ANNEX 1.3 IncomePoverty and SubjectivePerceptions The background study by Arias and Sosa (2004) analyzes the determinants of subjective and income poverty estimate using binary regression models (see Box A.1.3.1).2 The study employs a nationally representative household survey (Encuestu Nucionul de Aspiraciones y Prioridudes de Desurrollo Humuno, ENAPD) conducted by the UNDP for its 2000 Human Development Report in Bolivia. The survey captures the values and aspirations of Bolivians, including their self-ratedpoverty perceptions, in addition to most socioeconomic variables inthe MECOVIsurveys (see Annex 1.1). Box A.1.3.1: Explaining Income and Subjective Poverty Arias and Sosa (2004) estimate binary regression models of the determinants of subjective and income poverty. For subjective poverty, the dependent variable i s based on individuals' responses to the question "Do you consider yourself poor?' in the 1999 ENAPD survey (on individuals 18 years or older). A major limitation o f the ENAPDsurvey is that it captures only a limited fraction of household incomes. The 1999 h4ECOVI survey is used to estimate poverty regressions based on per capita household incomes (urban) and expenditures (rural) and official poverty lines for a sample similar to the ENAPDsample. The study examines the role o f a multitude of socioeconomic characteristics on poverty classifications: Demographics (gender, age, household size, marital status), Human Capital (own, mother and father's education, work experience, public school attendance), Employment Status (unemployed, underemployed, or employed, temporary job), Type of Employment (employer, independent worker or employee), Occupation (white collar, blue collar), Place of Residence (urbadrural, departamento, migration status), Ethnicity (Spanish, Aymara, Quechua, mother tongue), Wealth (income, consumption expenditures, assets), and Social Capital (participation in social, economic, political or other organizations). As instandard inprobitllogit analysis the models relate poverty binary indicators (1if poor, 0 otherwise) to socioeconomic characteristics and can be thought as arising from the following process. Individuals rank themselves as poor by comparing her self-assessed welfare level wi with a personal critical value wi* that she deems adequateto avoid poverty: yi = 1if [wi* - wi= xi /I ui > 01, and 0 otherwise + where xi are observed characteristics, uicaptures unobserved determinants o f welfare and measurement errors, the p are the weights placed on the xi. The term wi* - wi i s an unobserved measure of net welfare capturing the propensity to consider oneself poor. In the case of income poverty wi is a measure o f income or consumption and wi* is a poverty line based on a minimum cost basic food basket. Probit yields estimates o f the conditional probability o f poverty for individuals with different xi. In this formulation unobserved (unmeasured) factors (ui) do not interact with observed determinants of welfare (poverty). That is, individual characteristics (xi) affect welfare (poverty) inthe same way for all individuals (p i s common to everyone, Le., independent of xi and ui). However, idiosyncratic welfare and poverty perceptions are unlikely to be formed so restrictively. For example, the impact o f education on welfare and self-rated poverty may differ according to the rank that unobserved characteristics grant an individual inthe latent welfare distribution, that is, depending on her propensity to self-rate as poor. In order to examine whether individuals who differ in their unobserved welfare determinants weigh socioeconomic characteristics differently in their poverty self-ratings, Arias and Sosa (2004) estimate binary quantile regression models (Manksi, 1985; Kordas, 2002). The results offer richer information to policy makers for discerning the priorities of different segments o f the population, specifically of individuals more likely to self-rate poor given their socioeconomic characteristics. 2. SeeRavallion andLokshin (2002,2001) for a recent similar analysis for Russia. 78 Self-rated poverty profiles are compared with conventional profiles based on income/expenditures. The results point to a few important differences in the effect of socioeconomic characteristics (such as education, employment, ethnicity) on the probability that individuals perceive themselves as poor and are classified as poor on the basis o f their family income. The findings also reveal the relative importance of poverty determinants (i.e., the priorities implicit in poverty self-ratings) according to individuals' ranks o f subjective welfare (i.e., their propensity to consider themselves poor). SELF-RATEDAND INCOMEPOVERTY PROFILES A first step is to compare simple profiles of income/expenditure poverty and self-rated assessments. Figure A.1.3.1 depicts both the fraction of individuals that are classified as poor according to their family per capita income/expenditure, as well as the fraction of individuals that self-rate as poor according to several socioeconomic characteristics. This i s restricted to the sample of household heads over 18 years old. The vertical lines indicate the total fraction o f income and self-rated poor for the full sample. The following are the mainresults: There i s a striking similarity between the patterns of income and self-rated poverty incidence. Whenever the income poverty rate of a group (characteristic) exceeds (is below) the overall rate o f income poverty, that same group is more (less) likely to self-rate poor than householdheads. With few exceptions, self-rated poverty rates are lower than (or similar to) income poverty rates. Poverty incidence i s 7 percentage points higher according to traditional income measures (55 percent) than with individuals' own direct assessments (48 percent). That is, individuals are less likely to consider themselves poor than expected given their level of family per capita income. Education significantly reduces both income and self-rated poverty. Income poverty incidence falls from 77 percent among household heads with no education to around 10 percent for those with completed college, while the parallel self-rated poverty rates decline from 74 to 12 percent. The largest declines in poverty incidence, by either measure, occur with the completion o f an education degree (basic, secondary or college). The profiles o f income and self-rated poverty are also very similar when considering employment, demographic and several living conditions indicators. The unemployed, the underemployed, blue collar workers, the self-employed, the indigenous population (proxied by mother tongue) and those with low access to assets and basic services have higher income and self-rated poverty rates. However, for a few important characteristics, income poverty headcounts depart significantly from subjective poverty rates, suggesting that the former may fail to adequately reflect non-pecuniary aspects of well-being. The self-employed and the Aymara population have subjective poverty rates roughly 15 percentage points lower than their income poverty rates. Quechuas and Aymaras have equal income poverty rates but the latter have lower self-rated poverty. The self-rated incidence of heads outside the labor force i s 10 points above their income poverty rate. 79 Figure A.1.3.1: Self Rated and Income Poverty Profiles for Bolivia (Head of Household, 18 Years old and Over) Education Employment 7 6 Seif tirployed 5 Erployee 4 3 tirployer 2 Under enployed 1 0 Outof LF 9 Unenployed 8 I tirployed 6 Private 5 4 Public 3 White Collar 2 1 Blue Collar 0 0% 20% 40% 60% 80% 0% 20% 40% 60% 80% alncorne Poor .Self Rated Poor Demographics Living Conditions Women Oectriciky -48% Self Rattec Men Toilet Poor Married Conputer %YO Divorced Radio 1-lrcome Poor Television Single Refrigerator Spanish Telephone Quechua Good Q. Roof Aymra Avg. Q. Roof Other Poor Q. Roof I I 0% 20% 40% 60% 80% 0% 20% 40% 60% 80% Note: Income poverty measuresare based on householdincome per capita for urban areas and rural per capita expenditures. Source: Authors' estimatesbased on 1999 UNDP survey and MECOVI 1999. 80 Probit regressions are used to isolate the independent effects of these and other variables on both income poverty and subjective poverty perceptions. Table A.1.3.1 presents results, also including as explanatory variables the levels of individual's income and family expenditures, indicators of assets, living conditions, temporary jobs, attendance to public schools, parental education, proxies of social capital, and whether the effects of key variables (e.g., education and employment conditions) vary by ethnicity (columns 3 and 5). Table A.1.3.2 compares results for income and subjective poverty only for variables common to the ENAPD and MECOVI surveys. For comparison purposes, each set of coefficients i s normalized (columns 3-4) so that they measure the relative importance of each variable with respect to all explanatory variables (the sign still indicates whether they lower (negative) or increase (positive) poverty). For example, the education coefficients reveal whether education is more important in reducing income poverty than self-ratedpoverty relative to other explanatory variables. The results are consistent with a multidimensional notion o f poverty in which incomes/consumption represent one particular dimension. They contradict the assertion that self- rated poverty mostly reflects idiosyncratic opinions that bear little systematic relationship to socioeconomic characteristics. Overall, income poverty measures provide a reasonable, albeit imperfect, characterization of welfare (poverty) rankings inBolivia. Specific results that confirm the findings from simple poverty profiles are worth noting: The probabilities of being income-poor and of self-rating poor are higher among heads with lower education, of younger age, unemployed or underemployed, in blue-collar occupations, with an indigenous (Quechua or Aymara) heritage, low access to assets and basic services andor living in rural areas. In many cases the relative importance of these characteristics in explaining income and self-rated poverty is strikingly identical (such as in the case of education) despite the fact that potentially different processes could be behind these poverty classifications. The small effects o f income and expenditures indicate that their impact on subjective welfare mainly operates through factors like education or employment. Significant income compensation is needed to increase the probability that an individual self-rates poor if her socioeconomic characteristics remain unmodified. Income poverty rankings fail to accurately reveal the relative subjective welfare status of the self-employed, workers outside the labor force, migrants, the indigenous population and region of residence effects. Salaried workers are more likely to self-rate as poor than the self- employed while household heads outside the labor force tend to consider themselves poorer than the employed with similar characteristics, but these rankings differ for income poverty. Migrants are equally likely to self-rate poor but less likely to be income poor than non-migrants. The discrepancies for ethnicity are distinctively relevant. Bolivian Quechuas tend to self- rate poorer than suggested by income poverty profiles (adjusting for observed characteristics) while the converse i s true for Aymaras. The relative importance o f education for self-rated poverty i s similar along ethnic lines, but the weight o f unemployment i s double for the indigenous' self-rated poverty. 81 Table A.1.3.1: Determinants of Self-ratedPoverty inBolivia (Probit estimates, individuals over 18 years of age) Empirical Models (1) (2) (3) (4) (5) Intercept 0.043 0.093 0.048 0.050 -0.064 Male 0.022 0.031 0.021 0.031 0.032 Age -0.004" -0.003* -0.004" -0.004" -0.001 Household Size 0.015"" 0.018* 0.015"* 0.015** 0.018" Married -0.078** -0.065 ** -0.077** -0.076"" -0.045 Education -0.059" -0.053" -0.059* -0.058" -0.044" Education x non-indigenous -0.000 0.001 Mother education -0.020* -0.018" -0.021* -0.021" -0.014* Father education -0.019" -0.017" -0.019" -0.019* -0.013* Public education 0.019 0.000 0.018 0.019 -0.011 Unemployed 0.558" 0.5 16* 0.902* 0.549* 0.826* Unemployed x non-indigenous -0.444** -.0478"* Out of the Labor Force 0.309* 0.264* 0.308* 0.298* 0.217* Underemployed 0.138" 0.136" 0.138" 0.144* 0.128* Employer -0.274** -0.179"" -0.275** -0.275"" -0.167 Employee 0.118* 0.115" 0.121*** 0.115* 0.117**# Employee x non-indigenous -0.006 -0.041 Temporary Employment 0.126" 0.117" 0.117*** 0.126" 0.092 Temporary x non-indigenous 0.011 0.012 Blue Collar 0.218" 0.20" 0.22" 0.214* 0.147" Rural 0.141" 0.095"" 0.141" 0.159* 0.004 Non-migrant 0.046 0.041 0.047 0.048 0.044 Quechua 0.218" 0.213* 0.210" 0.217" 0.154** Aymara 0.134** 0.122** 0.119 0.141** 0.045 Other Indigenous 0.345"" 0.344** 0.337** 0.352"" 0.280 Income (Bs$1000) -0.004" -0.003" Consumption (Bs$1000) -0.007* -0.003*" Community Org. -0.010 -0.047 Community Org. x non-indigenous 0.012 Political Org. -0.059 -0.054 Political Org. x non-indigenous 0.001 Economic org. -0.041 -0.045 Other Org. -0.042 0.024 Other Org. x non-indigenous -0.070 Regional Dummies Yes Yes Yes Yes Yes Observations 9018 9018 9018 9018 9018 * Significantat 1%; ** significantat 5%;*** significant at 10%. Source: Authors' estimates based onENAPD andMECOVI 1999householdsurvey data. 82 Table A.1.3.2: Determinantsof Income and Self-RatedPoverty inBolivia (Probit Estimates, Head of Household) Original Probit Normalized coefficients coefficients ENAPD MECOVI ENAPD MECOVI Self-rated Income Self-rated Income (1) (2) (3) (4) Male -0.022 0.021 -0.020 0.014 Age -0.004 ** -0.013** -0.004 -0.009 Household Size 0.007 0.217* 0.006 0.144 Married -0.050 -0.053 -0.046 -0.035 Education -0.061" -0.085* -0.055 -0.056 Unemployed 0.530" 0.628** 0.481 0.416 Out of the Labor Force 0.142*** -0.136 0.129 -0.090 Underemployed 0.075 0.104 0.068 0.069 Employer -0.303** -0.526" -0.275 -0.348 Employee 0.194" -0.107 0.176 -0.071 Blue Collar 0.263* 0.494* 0.239 0.327 Rural 0.170" 0.210* 0.154 0.139 Non-migrant 0.043 0.117** 0.039 0.077 Quechua 0.227" 0.087 0.205 0.058 Aymara 0.180** 0.399* 0.163 0.264 Other Indigenous -0.244 0.180 0.221 0.119 Intercept 0.109 -0.158 0.099 -0.105 Regional Dummies Yes Yes Yes Yes Observations 3491 3035 "Significant at 1%; ** Significant at 5%; ***Significant at 10% Coefficients in columns 3 and 4 are normalized for comparative purposes, their statistical significance i s similar to the original estimates. Source: Authors' estimates on ENAPDand MECOVI 1999 household survey data. The non-monetary benefits associated to settlement decisions are misrepresented by geographic income poverty rankings. In Figure A.1.3.2 the departments are ranked from the income poorest to the least income-poor. Although Chuquisaca i s the second income-poorest region its residents self-rate the least poor in the country. The residents o f Santa Cruz are equally likely to be income or self-rated poor than those o f L a Paz controlling for differences in individual characteristics. Residents o f rural areas are more likely to be poor than urban inhabitants measured by either income or self-rated poverty. However, rural residents no longer self-rate poorer than urban inhabitants ifthey have equal socio-economic and living conditions butremainmore likely to be income-poor. The discrepancies in income and subjective poverty rankings may be traced to non- monetary traits that affect the well-being o f these groups. These include exclusion and/or cultural factors (e.g., sense of empowerment or identity) as well as location-specific characteristics (e.g., inequality, social capital, crime) all o f which may have meaningful effects on Bolivians' poverty perceptions. 83 RELATIVEPRIORITIES OF THE SELF-RATED POOR This section sheds light on whether different segments of the population place different priorities on observed characteristics as a means out of poverty. Determinants of income and welfare not captured in household surveys (such as unobserved skills, education quality, labor market connections, personal attitudes) lead to differences in individuals' intrinsic propensity to self-rate poor. This in turn may affect the weight attached to observed characteristics in self- ratings of poverty. In fact the findings point to Figure A.1.3.2: Impact of Region of Residence important differences in the way measured on Income and Self-Rated Poverty (Probit socioeconomic characteristics are weighted Estimates) by individuals in their self-ratings of poverty. The results are shown in Figure POfO.2 A.1.3.3 which shows the relative weights (vertical axis) of the various poverty determinants for individuals with high (e.g., quantiles 0.3 to 0.5) to low (e.g., quantiles 0.7 to 0.8) conditional probabilities of self- rating poor (horizontal axis). The horizontal line represents the average relative weight. For many characteristics the weights increase or decline with the conditional probabilities o f self-rated ' poverty. Thus, the relative of importance Note: L a Paz i s the omitted region. Coefficients poverty determinants varies significantly measure the relative importance o f regional among individuals depending on howlikely effects with respect to all explanatory variables. they are to self-rate as poor. Source: Authors' estimates based on household survey data. Specifically, factors that help people move out of poverty (such as own and parental education or getting ajob) are more effective (the weights decline in absolute value with higher probabilities of self-rating as poor) precisely among Bolivians who are more likely to self-rate as poor given their observed characteristics. Chiefly, education i s less effective in reducing self-rated poverty of individuals for whom non- observed factors play a bigger role in their poverty perceptions. Meanwhile the effect of unemployment i s stronger at higher conditional probabilities of self-rated poverty, that is, it matters relatively more for Bolivians more prone to self-classify as poor. Among other results, the relative effect of rural residence increases systematically, becoming less noticeable among individuals who are more likely to consider themselves poor given their other observed characteristics. Meanwhile, having a Quechua heritage increases self- rated poverty (relative to non-indigenous) the same way regardless of non-observed determinants of welfare. These differences in the formation o f poverty perceptions are important considerations for public policy. Despite not being directly tangible, many of the unobserved factors driving them are amenable to policy interventions. For example, early slulls development, education quality and labor market connections are key for income and welfare more generally. Other 84 cultural and idiosyncratic factors are difficult to quantify let alone influence by policy. Yet ascertaining their role i s important to attune public policy priorities with the revealedpriorities of its target population. Figure A.1.3.3: Effect of SocioeconomicCharacteristics on Self-Rated Poverty Binary Quantile Estimates (relative weights) Household Sire Annual Income ($1000) o o s 2 , L- - - - - - - - - - - -- - ---I I I I I I I Employer Employee Temporary Employment Blue Collar Worker Non-migrant Querhus I Aymara I Note: Coefficients have been normalized (norm 1) so that they measure the relative importance of each variable w.r.t. all the explanatory variables. Source: Authors' estimates based on microdata from household surveys. 85 86 Annex 2.1 TechnicalAnnex onDeterminantsof Growth Actual Projected Transitional Cyclical Structural Stabilization External Specific Country. Difference Difference Convergence Reversion Policies Policies Conditions Effect Chile -3.46 -2.82 1.82 -0.18 -2.83 0.01 -0.11 -1.53 Argentina -2.18 -1.45 1.92 -1.10 -0.81 1.19 -0.25 -2.4 Costa Rica -1.95 -1.12 1.28 -0.06 -0.7 0.1 -0.24 -1.51 Panama -1.26 -1.21 1.05 -0.16 -1.85 -0.17 -0.22 0.14 Colombia 0.82 -0.83 1.17 0.04 -1.03 -0.16 -0.15 -0.7 Peru -0.79 -0.28 0.39 -0.46 -0.49 1.15 0.03 -0.9 I Table A.2.1.2: Determinants of Growth between 10year Periods, Bolivia, 1970s-1990s 1990sversus 1980sversus 1980s 1970s Transitional 0.11 0.02 Convergence Cyclical Reversion -0.02 -0.56 Structural Policies 1.34 0.38 Stabilization Policies 1.70 -1.53 External Conditions -0.59 -1.09 Actual Change 3.49 -3.62 Projected Change 2.54 -2.77 Source: Based on Loayza, Fajnzylber and Calderon, 2002. 87 ANNEX 2.2 ExplainingChangesinIncomeDistribution Microeconometric simulations of counterfactual distributions are helpful to characterize past distributional changes and to simulate the distributional impact o f changes in economic factors and public policies. The idea i s to simulate the distribution of labor income at time t as a function of individual observable characteristics affecting wages and employment, the parameters that determine the effect of these characteristics on market hourly wages and employment outcomes (participation and hours of work), and unobservable characteristics. A counterfactual distribution in time t l i s generated taking some of its determinants (parameters or distribution of characteristics) as if they were those of time t2 and this counterfactual distribution i s then compared to the actual distribution observed in tl. The difference between the two distributions can be attributedto the change in the selected determinants between t l and t2. This allows to isolate the contribution of changes in: (i)observed household characteristics (endowments), (ii) the returns to those endowments, and (iii) unobserved heterogeneity in the returns. Gasparini et al. (2004) estimate regressions for a reduced form o f a labor supply model with two equations, one for the number of hours of work and one for wages. The explanatory variables include the typical measures o f workers' human capital (education and experience, proxied by age and its square), demographic characteristics such as gender and ethnicity, job characteristics (sector of activity and labor-informality indicators), and geographical location. The earnings equations are estimated separately for householdheads and non-heads, both inrural and urban areas. The simulations are carried out for the periods 1993-97 and 1997-2002 and focus on the effects o f changes in the educational structure, the returns to education, the gender and regional earnings gaps, and unobserved earnings determinants. The decomposition analysis i s enriched with estimates of quantile earnings equations, (see below) which are used to generate counterfactual distributions when the whole family of returns to education (varying across quantiles) change or for changes in each of the return quantile coefficients. This procedure may provide a richer characterization o f past and predicted changes in the income distribution generated by economic and social changes or policy interventions. Particularly, when simulating changes in the educational structure, we can simulate the new individual wage from upgrading education according to the wage-education profile of the particular percentile to which the individual belongs. See Gasparini et al. (2004) for details. 88 Table A.2.2.1: Decomposition of Changes inIncome Distribution inBolivia, 1993-2002 Independent Contributionsto Changes inthe Ginicoefficient Model 1 Model 2 Model 3 A. 1993-1997 (main cities) Hourly Equivalized Hourly Equivalized Hourly Equivalize earnings income earnings income earnings income Observed1993-1997 -0.3 0.8 -0.3 0.8 -0.3 0.8 Returns Education- wages -1.0 -1.0 -1.8 -1.6 -1.5 -1.3 Education- hours 0.5 0.4 0.5 Gender -0.2 0.0 -0.6 0.0 -0.5 0.1 Unobservables 3.9 2.7 3.6 2.7 3.7 2.9 Regions 0.2 -0.6 0.0 -0.2 -0.1 -0.3 Educational structure 0.4 0.4 -0.7 -0.3 -0.1 0.3 Model 1 Model 2 Model 3 B. 1997-2002(urban areas) Hourly Equivalized Hourly Equivalized Hourly Equivalize earnings income earnings income earnings income Observed 1997-2002 2.0 1.5 2.0 1.5 2.0 1.5 Returns Education- wages 0.2 0.3 0.9 0.9 0.7 0.7 Education- hours 0.3 0.2 0.4 Gender 0.2 0.0 0.3 0.1 0.3 0.1 Unobservables 3.O 2.2 2.9 2.3 3.0 2.3 Regions -0.2 -0.6 -0.2 -1.0 -0.2 -1.0 Educational structure 1.o 1.5 0.5 0.9 0.3 0.9 Model 1 Model2 Model3 C. 1997-2002 (rural areas) Hourly Equivalized Hourly Equivalized Hourly Equivalize, earnings income earnings income earnings income 3bserved1997-2002 -5.8 -3.7 -5.8 -3.7 -5.8 -3.7 Returns Education- wages 0.0 -0.1 0.0 -0.2 -0.1 -0.1 Education- hours 0.4 0.3 0.4 Gender 0.0 0.1 -0.2 0.2 -0.1 0.2 Unobservables 1.9 1.1 1.3 0.7 1.4 0.8 Regions -1.1 -2.1 -1.6 -2.3 -1.6 -2.3 Educational structure 0.3 0.0 -1.4 -0.3 -0.9 -0.4 A Primer on Ouantile Regressions The technique of quantile regression (Koenker and Basset (1978)) is used extensively in the background studies for this report since it provides a richer characterization o f the effect of the explanatory variables (X) on the conditional distribution o f the dependent variable (e.g., the distribution of earnings, of firm size). When there i s sizeable unobserved heterogeneity in the data, mean linear regression models provide only a limited characterization of this distribution and of the role of explanatory factors. This i s probably the case in a heterogeneous country like Bolivia specially given the limitations of existing surveys. For example, we can estimate regression lines for various percentiles o f the adjusted (conditional) wage distribution (the distributionof earnings that results ifall workers had the same observable characteristics). For 89 instance, median regression (the 50th quantile) splits the sample in half (half of the residuals above and below the regression line) and gives the same results as Ordinary Least Squares (OLS, mean regression) when the wage distribution i s symmetric. This allows unobserved wage determinants to interact with measures of observed skills. This interaction is captured by regression coefficients that vary across percentiles of the adjusted wage distribution. This way we can recover different impacts of the explanatory variables throughout the entire distribution without imposing any prior assumptions, such as normality or constant variance of regression errors. Results are also robust to outliers inthe wage data. Suppose that X i s a dummy variable for gender (women=l), the quantile regression coefficient measures the gender wage gap between woman and an otherwise similar man (e.g., same education and experience) at the same conditional quantile of the wage distribution. For example, the coefficient in the 90th percentile would yield the wage disadvantage faced by women in the top10 percent best paid jobs o f any given level o f observed skills while the 10th percentile coefficient yields the gap for women in the bottom 10 percent of jobs of the earnings scale, Now suppose that X consists of years o f formal education. OLS provides a single estimate of the returns to education, the average for the whole population. Individual returns to education, however, may depend on some unobservable factors, like education quality, unmeasured skills, or labor market connections, and hence may differ across workers (Figure A.2.2.1A). Infact, recent studies for several countries suggest that returns are higher for workers at the top of the distribution. Moreover, it i s possible for the returns to education to increase for workers in the upper quantiles o f the wage distribution and decline for those in the bottom quantiles, leaving the average return unchanged (Figure A.2.2.1B). Quantile regressions allow assessingthese important potential differences. Figure A.2.2.1A: Differences inReturns to Figure A.2.2.1.B: Different Changes in Education Returns Over Time " OR ORU+, ,./' I W X.," OR'tr El E2 E 90 Annex 3.1 ManufacturingFirmSurveys FACS The FirmAnalysis and Competitiveness Survey (FACS) 2000, conductedby INEand the World Bank, analyzes the micro constraints to the growth of manufacturing firms in Bolivia. It relies principally on objective, quantitative measures of a broad range o f firm characteristics and problems affecting firms, such as capacity utilization (domestic markets, exports), sector, location, investment, sources of finance (credit collateral, debt exchange risk), labor productivity, operational costs (supply chains, inventory levels, cost of labor, infrastructure), and relations with government (taxation, licenses and permits, trade regulations, judiciary). A companion study identifies the compliance costs and legal requirements for informal firms and micro enterprises (less than five employees) to start and runa business. The survey covers a random sample of 659 formal manufacturing firms in the departamentos of L a Paz, Cochabamba and Santa Cruz that represent about 40 percent of the number of manufacturing firms in the country (85 percent weighted by employment). Almost half of the f sample are small (4 to 14employees) and only 14percent have over 100employees. Informal firms and micro enterprises are not included, since they rarely keep accounts and are difficult to locate for sampling. The Annual Survey o f Manufacturing; Activity Since the late 1980s, INE conducts an annual survey to measure basic characteristics of manufacturing activities (Encuesta de Establecimientos en la Actividad Manufacturera, EEAM). The survey consists of establishment data related to the gross value of production, value added, intermediate consumption, sales, primary good inputs and electricity, level of employment, salaries, taxes, and fixed assets. It has national coverage with the exception o f the departamento of Pando. The data i s collected as part of the firms' accounting exercise (from April 1to March 31 of the next year). Manufacturing firms are classified according to seven types of activities: food, beverages and tobacco, textiles and clothing, wood and wooden products, petroleum derived products, non-metallic mineral products and other industries. Firms in this sector are also categorized by size into three groups (excluding family and micro enterprises): firms with less than 14, between 15 and 49, and over 50 employees. The survey i s conducted in all firms with five or more employees. All firms are classified into Forced Inclusion (IF),that is, firms with more than 15 employees; and Sample Inclusion (SI), firms with more than five and less than 14employees. The survey covers between 1,500 and 1,600 firms annually from both the IF and S I groups, of which approximately two thirds are small enterprises. It is possible to construct a panel data set since firms are surveyed every year. The panel i s constructed using five surveys for the period (the only ones comparable due to changes in survey design) with firms for which data was gathered for two or more years. For firms with incomplete data (such as capital, value added, and employment), estimates were made based on other available data for the same firm. Since controls were not firmly implemented, it i s difficult to differentiate firms that have exited the market from those that dropped out of the sample, a problem most evident among S I firms. The panel i s comprised of between 650 and 850 firms, half of which are small enterprises. The number of medium enterprises i s slightly greater than that of larger firms each year. Around 50 percent of firms remained stable inthe sample over the period. Larger firms exhibit a lower rate of attrition than SMEs, and are thus slightly over 91 represented in the sample. Overall, the data i s believed to offer a good representation of the economic activity andemployment dynamics of the formal manufacturing sector inBolivia. 92 ANNEX 3.2 Estimating Rates of Creation, Turnover and Reallocation of Employment The methodology decomposes net changes in employment into the absolute creation and destruction that occurs between two points in time, resulting inthe reallocation of workers across different industrial activities. The creation of employment of firm i i s positive if the level of employment grows between t-1 and t, while the rate of destruction i s positive if employment falls. The rate of net growth of employment between t-1 andt for firm i i s then: where Ei s the level of employment of the firm in time t. The gross creation of employment (sum of all the new jobs generated across firms between t-1 and t) i s measured as: C, = 4; max(Net, ,0) The gross destruction of employment (sumof all the job losses betweent-1and t) is: D, =z& min(Net, ,0) The difference between Ct and Dt yields the net creation rate in the sector over the period. The weight each firm receives i s the average o f its employment levels during a given period. The rate of employment reallocation, which measures the total reallocation of employment among workers, i s the sum of the rates of creation and destruction. Landa and Jimenez (2004~)estimated a standard production function (Cobb-Douglas) relating firm's output (VAT) with the quantities of unskilled (Lu) and skilled (Ls) labor, and capital (K) factors, a set of characteristics related to the firm (Z), time effects (T) common to all firms, and an error component (E): lnVAT = a+ InK PI +PzIn +P3In +T +E Ls Lu where the coefficients p are the output elasticities o f each production factor. To measure firm's efficiency they estimate a production frontier (the maximum output obtained given the observed levels of input utilization) modeling the error term with a symmetric component (u) that captures non-observable aspects as well as measurements errors, and a non-symmetric component (v) that captures technical inefficiency, Le., the distance between a firm's actual production and its production frontier. Technical inefficiency occurs when the output of the firm i s below its production frontier, and i s measured comparing each firm's actual output level with that predicted by the model. Another common inefficiency measure is the ratio o f the standard deviation of v and u, that is: A = 0,/ fJu,which compares the variation inthe inefficiency under the control of the firm with external sources of output variation. The sources of inefficiency include obsolete technology, low quality o f inputs, and poor management and organizational skills. 93 Capital K i s measured as the value of the stock of fixed capital in the current period, including the value of machinery and other firm's fixed assets. Skilled labor i s the number of firm's managers and white-collar workers. Unskilled labor is the number o f workers in blue-collar occupations and low wage temporary workers, excluding non-remunerated family members. The model controls for the firm's age, the use of intermediate inputs (electricity and water) as a proxy of capacity utilization, subsector, time and regional effects. The models are estimated through maximum likelihood with random effects, and invariable time. 94 ANNEX 3.3 Productivity and Technical Efficiency inBolivian Manufacturing Employment creation is ultimately linked to firms' economic performance. Firms choose the quantities of inputsthey use (including labor) to minimize the cost of producing a given level of output. Thus, the rate of utilization and mix of inputs depends on their contribution to the generation of output (measured by output elasticities) and their costs. In a competitive market (i.e., with no distortions on price determination) a firm would utilize more those inputs with the highest output elasticity andor lower cost. These relationships are summarized by firms' production and cost functions. Ina background study for this report Landa and Jimenez (2004~) obtained estimates of the output elasticities with respect to capital, skilled and unskilledlabor in Bolivia's manufacturing sector using the 1995-99 EEAMpanel data (Box A.3.3.1). Box A.3.3.1: Measuring FirmProductivity and TechnicalEfficiency Landa and Jimenez (2004~)estimated a standard production function (Cobb-Douglas) relating firm's output (VAT) with the quantities of unskilled (Lu) and skilled (Ls) labor, and capital (K) factors, a set of characteristicsrelated to the firm(Z), time effects (T) common to all firms, and an error component(E): lnVAT = a t p , In K+pz1nLs+8, In Lu T+ +E where the coefficients p are the output elasticities of each production factor. To measure firm's efficiency they estimate a production frontier (the maximum output obtained given the observed levels of input utilization) modeling the error term with a symmetric component (u) that captures non-observable aspects as well as measurements errors, and a non-symmetric component (v) that capturestechnical inefficiency, i.e., the distance between a firm's actual production and its production frontier. Technical inefficiency occurs when the output of the firm is below its production frontier, and is measuredcomparing each firm's actual output level with that predicted by the model. Another common inefficiency measure is the ratio of the standard deviation of v and u, that is: = avlauwhich compares the variation inthe inefficiency underthe control , of the firm with external sources of output variation. The sources of inefficiency include obsolete technology, low quality of inputs,and poor management andorganizational skills. Capital K is measuredas the value of the stock of fixed capital in the current period, including the value of machinery and other firm's fixed assets. Skilled labor is the number of firm's managers and white-collar workers. Unskilled labor is the number of workers in blue-collar occupations and low wage temporary workers, excluding non-remunerated family members. The model controls for the firm's age, the use of intermediate inputs (electricity and water) as a proxy of capacity utilization, subsector, time and regional effects. The models are estimatedthrough maximumlikelihood with random effects, and invariable time. The results can be used to: i)assess the role of economic fundamentals, such as large differences in factor marginal productivity, in the utilization rates and degree of substitution of unskilled labor with respect to skilled labor and capital. That is, the extent to which manufacturing production processes (technologies) in Bolivia are biased against or in favor of the generation of unskilled jobs; ii)which subsectors are more likely to generate more labor- intensive growth; and, iii)estimate the levels of efficiency (or total productivity) with which manufacturingfirms operate. The overall results are reported in Table A.3.3.1. The empirical models control for the age of the firm, a proxy for its capacity utilization (use of energy and water), subsector and year effects. For Figure A.3.3.1 the models are also estimated by subsector. The coefficients for 95 capital, and unskilled and skilled labor are the factor output elasticities that measure the increase in output that could be derived from a 1percent increase in each of these inputs. For example, the 0.359 overall capital elasticity (in the preferred specification, col. 3) indicates that on average a 10 percent increase in the quantity of capital would raise manufacturing output by 3.4 percent. The results can be summarized as follows: The output elasticity of capital i s higher than for labor, so that, on average, its expansion leads to higher increases in manufacturing output compared to labor. This reflects the relative scarcity andhighcost o f capital inBolivia. Manufacturing growth in labor intensive sectors would tend to favor the creation of unskilled jobs. This follows from the result that the output elasticities are essentially similar for unskilled (0.24) and skilled labor (0.28). Since wages are higher for skilled workers this means that manufacturing expansion would tend to use unskilled workers more intensively. Growth would tend to be more unskilled labor intensive in the beverages and textiles sectors (elasticities of 0.65 and 0.45) and more capital intensive in mining and wood production (capital elasticities of 0.63 and 0.46). The skilled labor elasticity i s higher than the unskdled only in food (0.39) and mining (0.26) but only in the latter i s the marginal productivity difference likely to result inrelatively more skilled growth. The sectoral differences in factor elasticities i s consistent with the expected high degree of substitution between capital and unskilled labor, moderate substitution between labor types, and little complementarity of skilled labor to capital.' The correlation coefficient of the factor elasticities across sectors are -0.65 for capital and unskilled labor, -0.45 for unskilled and skilled labor, and -0.28 for capital and skilled labor. In fact, younger firms tend to reach higher levels of output for the same level of input utilization than older firms. Eventually older firms appear to be more productive but seemingly after a longtime (45 years according to the quadratic estimates). There i s significant room for expanding manufacturing output by a more intensive use of the installed capacity o f firms. A 10 percent increase in the capacity utilization rate, proxied by the use o f the intermediate inputs energy and water, would expand output by 1.9 percent. Firms locatedinthe lowlands, inthe petroleum, wooden products and/or other industries sectors, reach a higher level of output for the same level of input utilization. This maybe reflecting differences in market conditions and aspects of the business environment that favor the operations of these firms. There were no significant overall productivity gains in manufacturing during the period. The 1999 crisis hit the sector hard with a 10percent drop inthe level of output o f firms in the sample, keepingthe level o f input and capacity utilization constant. 1. The proper assessment of the degree of complementarity between inputs should rely on cross-elasticities of input use, which were not estimated here. 96 Table A.3.3.1: Estimates of the ProductionFunction in Manufacturing (Maximum Likelihood and Random Effects Estimates) (1) (2) (3) Dependent Variable lnValue Added Log Capital 0.420** 0.357"" 0.359** Log SkilledLabor 0.289"" 0.244"" 0.243** Log Unskilled Labor 0.321** 0.267** 0.281"" Age -0.012"" -0.009* Age Squared o.ooo* 0.000" LogWater and Energy 0.189"" 0.190** Beverage and Tobacco 0.203 Textiles -0.069 Wood Products 0.335** Petroleum 0.562** Mineral -0.142 Other 0.322"" Valleys -0.017 Lowlands 0.143" 1996 -0.008 1997 0.023 1998 -0.006 1999 -0.102"" Constant 6.040** 5.440** 5.122"" (3, 0.514"" 0.470** 0.468** (3" 0.742"" 0.705** 0.660"" Observations 3106 2745 2743 Number of Firms 969 787 787 Note: Omitted categories are food, highlands, and 1995. * Significant 5%; ** at significant at 1%. Source: Authors' estimates based on EEAM. 97 FigureA.3.3.1: Factor Elasticitiesinthe BolivianManufacturing Sector 0.6 0.5 0.4 0.3 0.2 0.1 - 1 +Capital -SkilledIabor dUnskilledbbor 1 0 Bev&Tob Textiles Food Other Petroleum Wood Mineral Note: Estimates from the Frontier model with fixed effects. Source: Authors' estimates basedon EEAMsurvey data. It is important to emphasize that the above results are not an assessment of the overall capacity for job creation (in absolute number) o f manufacturing compared to other sectors of the economy or between different subsectors in manufacturing. The results refer to the relationship between marginal changes in output and input utilization within manufacturing or each subsector. How many jobs are created in the economy depends on the capital/labor ratio in manufacturing (each subsector) and on the spillover effects on the rest o f the economy that is on the indirect generation o f jobs. For example, the expansion of mining and wood production would tend to generate fewer direct jobs per unit of additional output (or new capital investments) inthese subsectors. The results o f Landa and Jimenez (2004~)also point to low levels of efficiency of manufacturing firms in Bolivia, with considerable variation across firms. A large fraction of firms operate below the level of output expected given their level of input and capacity utilization, age, location and sector of operation (ie., below the optimal production frontier as defined by best performers). Inefficiency seems pervasive and bears little relationship with firm size, age and subsector. Beverages and tobacco and petroleum production present the lower levels and within-sector variation inefficiency. Altogether the results suggest that Bolivian manufacturing firms have not been adapting modem technologies and that, in fact, many older ones may have become obsolete. The relative indifference inthe use of skilled or unskilled labor (aside from their relative cost) i s symptomatic of production processes that do not make intensive use of information technologies and very little innovation. The apparent inability of firms to capitalize productivity gains as they mature may reflect their little incentives to innovate faced with a high cost environment, limited sales and credit markets, and institutional constraints to the entry-exit process. The failure to adopt new technology or management changes hampers their ability o f reaching the optimal production frontier. 98 ANNEX 3.4 The Demandfor Labor This section expands the previous results with a consideration of the (direct) cost of labor (real wages) in assessing the response of manufacturing employment to changes in economic conditions. This analysis is based on the estimates of demand equations for unskilled labor in the Bolivian manufacturing sector (Box A.3.4.1). The latter summarize the main parameters characterizing the dynamics of employment, including the response to changes in real wages (given by the own wage-employment elasticity) and the path of employment in response to shocks (captured by the lag employment term) and output expansions (given by the output- employment coefficients). These relationships are important to assess the impact of a variety of policies on employment. The wage elasticity i s of course key to assess the likely employment impact of measures that increase or reduce the cost of labor such as changes in minimumwages and non- labor costs (payroll taxes, pension benefits). Meanwhile, faster employment adjustments (to, say, an increase in the final demand for products, inputs scarcity or an external shock) imply fewer frictions for Bolivian workers and firms while they transition to the new equilibrium. The speed of adjustment of employment to shocks i s affectedby several factors including adjustment costs (constrained capital markets, firinglhiringregulations, supply bottlenecks), macroeconomic volatility (real exchange depreciation, external shocks) and the degree of competition in the industry(the less competition the slower the adjustment). Box A.3.4.1: Estimating ManufacturingLabor Demand The study by Landa and Jimenez (2004~)estimates a standard labor demand equation for unskilled labor (derived from the equilibriumequalization o f the marginal productivity o f labor with the real wage): where (li,t) i s the number of unskilled workers at time t (excluding managers and white collars) i s specified as a function o f employment in the previous period, the current and past levels o f real wages and output, time varying and firms fixed effects, and an error term. The lagged values capture the fact that firms do not react immediately to wage and output shocks due to adjustment costs, while the time effects proxy unexpected fluctuations in demand common to all firms, and firm specific fixed effects capture factors affecting labor demand that are specific to each firm (e.g., its managerialcapacity). Following the recent literature three estimation methods are used, including the preferred Arellano and Bond (1991)'s Generalized Method of Moments (GMM) instrumental variables (IV), based on the orthogonality between the lag dependent variable and the error term to estimate the differenced equation: Where A is the difference operator and the new error term i s not correlated with the lag dependent variable. All variables are assumed exogenous except for lag employment which is instrumented using its second difference lags. The output variable could not be instrumented with the data available inthe short panel. Table A.3.4.1 presents the main results, including the long run own wage elasticity of unskilled employment, which takes into account the dynamic adjustment in employment as a result of a change in wages, and the speed of adjustment inemployment measuredas the number 99 of years traversed before employment reaches half the level of its new equilibrium value. Column 3 shows the results which according to statistical tests represent the labor demand function more adequately (Landa and Jimenez, 2004~).The main results can be summarized as follows: e Firms and workers face a relatively high long run trade-off between higher wages and increasing unskilled employment. The long run demand wage elasticity for unskilled labor i s -0.64 so that for every 10 percent increase in real wages labor demand declines in 6.4 percent (keeping output constant). This is twice as high as the average international cross-country estimate of -0.3 (Hamermesh, 2003) and on the high end of estimates for other Latin American countries, for example, Brazil (-0.4), Chile (-0.37), Colombia (-0.49), Mexico (-0.2), Peru (-0.2), and Uruguay (-0.69). The short run elasticity (sum of 2nd-3rd rows in Table 3.3) i s half as big and lies within the middle range of recent estimates for some of these countries. Moreover, although the estimated elasticity does not account for the indirect cost of labor (e.g., mandatedbenefits) evidence for other countries show that the elasticity of these labor cost components can be as high as the own wage employment elasticities (Saavedra and Torero, 2000; Mondino and Montoya, 2000). Table A.3.4.1: Estimatesof Manufacturing UnskilledLabor Demand inBolivia (1) (2) (3) (Preferred) LogEmployment OLS Within Groups GMMdifferences 0.806 0.040 0.584 (-27.128) (-0.527) (-2.565) -0.223 -0.191 -0.307 (3.566) (3.359) (2.378) 0.112 -0.011 0.043 (-2.414) (0.307) (-0.793) 0.234 0.238 0.196 (-7.024) (-7.613) (-3.926) -0.107 0.074 -0.022 (3.497) (-2.578) (0.415) Year dummies Yes Yes Yes Longrun wage elasticity -0.572 -0.210 -0.635 Employment Adjustment (half-life) 3.21 0.22 1.29 Observations 1232 1232 924 Note: The long run own wage elasticity of employment is obtained as the sum of the short term elasticity coefficients (2nd and 3rd rows), divided by one minus the coefficient on lag employment (A, 1st row). The coefficient of "half life" adjustment in the level of employment i s computed as ln(.5)/ln(A). T-statistics in parentheses are robust to the presence of heteroscedasticity.Specification tests (2nd order auto-correlation, J- Hansen) support the results of Column 3 as the correct specification Source: Authors' estimatesbasedon manufacturing survey data 100 Employment tends to adjust slowly in response to cost or output shocks. The estimated (half-life) speed of adjustment of labor demand i s 1.3 years which again i s above international estimates (half a year) but aligned with those for Latin America (1 year in Brazil, 1.2 years in Chile, 0.4 years in Colombia, 0.8 years in Mexico, 1.3 years in Peru, 1.5 years in Uruguay). This suggests that deviations from firms' desired employment levels are relatively persistent so that it takes a relatively long time before fluctuations in the demand for labor adjust to changes ineconomic conditions. The long-term responseof employment to changes in output seems low although it i s not inconsistent with constant economies of scale. The long run employment-output elasticity i s 0.47, implying that a 10 percent increase in long term manufacturing output results in an increase of 4.7 percent in unskilled employment. However, the estimated long term elasticity coefficient is not statistically different from 1so that constant returns to scale could be describing the characteristics of manufacturing production. This would be more consistent with the previous findings of significant inefficiencies of firms in the sector. Firmsrely on past employment decisions to adjust their labor demand. The dynamics of labor demand i s dominated by the inertial employment component so that past changes in wages and output do not have an independent effect on current employment levels. This behavior i s consistent with the presence of idle capacity in which case employers can utilize the existing labor force more (less) intensely (each hour) inresponseto an increase in output demand (wages). Moreover, the defacto flexibility inthe labor market (given the high level of informality) and weak labor unions make past wage changes less relevant to the determination of present labor demand beyond their effect on past employment levels. Altogether the results suggest that Bolivia's manufacturing employment i s quite flexible with respect to changes in wages but adjusts slowly in response to shocks. One key implication i s that policies that increase the cost of labor (e.g., increase in the minimum wage or a rise in non-labor costs) beyond its marginal productivity could have a high cost in terms of reduced overall employment. Policy actions that aim to reduce excessive non-labor costs (payroll taxes), described in the last section of this chapter, could have a positive impact on long run employment levels. With respect to the slow speed of employment adjustment several of the explanatory factors noted above are relevant for Bolivia. Underdeveloped financial markets, volatile terms of trade and external capital, stringent labor regulations, and barriers to entry in the formal manufacturing sector, all likely conspire to lower the grease in the wheels of Bolivian's manufacturing labor demand. In particular, the deterioration of terms of trade, stop of capital flows and liability dollarization since the late 1990shas likely severely constrained the ability of manufacturing firms to undertake productive endeavors and expand employment in fuller speed. This contributes to the low levels of capacity utilization which in turn leads firms to first adjust hours of work before taking on new hires. Barriers to entry coupled with the disincentives of high non-labor costs increase the costs of employment adjustments for formal manufacturing firms although, as noted before, also lead many firms to join the informal sector. The latter means that the defacto regulations are in effect less binding than the de jure regulations so that 101 the speed o f adjustment of total manufacturing employment (including informal microenterprises) and the labor market overall i s likely higher. However, as we saw in chapter 3, inmany cases this means that unemployment is masked as underemployment inthe lower tier of the informal sector. The EEAMpanel data does not allow examining some of these hypotheses directly since it contains limited data on firm characteristics and lacks variables on the investment climate. In the final section of this chapter we use the FACS survey to examine more systematically the effect o f micro constraints faced by manufacturing firms on two aspects of job creation: capacity utilization rates and the distribution of firm size. 102 ANNEX 3.5 ExplainingThe Constraintsto CapacityUtilizationandFirmSize The companion study by Muiioz, Palma and Arias (2004) characterizes factors affecting the distributions o f firm size and C U of distinct groups of firms with special attention to firms located at the bottom quantiles of the conditional distributions of size and CU, firms with sizes or CU that are lower than predicted by their firm-specific characteristics, and measured business environment. The standard regression's focus on "average" firms may obscure the particular conditions that affect these firms. Key determinants and key micro constraints to CU and firm size growth may differ for firms at different points of the conditional distributions of C U and size. The study estimates simultaneously regressions for the 10 percent, 25 percent, 50 percent, 75 percent and 90 percent quantiles of the distributions of size and CU using the cross-sectional data from the 2000 FACS survey. It explores 50 different explanatory variables that the theory suggests could be important determinants of firm size andor CU. The variables considered include firm characteristics (age, sector, FDI, capital intensity); operational costs (costs of materials, labor, capital); limited demand and level of competition (undeveloped markets, access to exports, type o f competition); limited investment and sources o f finance (research and development, sources of credit, defaults, collateral requirements); shortage of production factors (access to skilled labor, rigidity of labor, technology); the business and legal environment (relations with suppliers/clients/workers,existence of contracts); basic services and infrastructure (electricity, water), as well as governmentalrequirements regarding licenses and permits. The dependent variable in the regressions for C U ("capac-hr") corresponds to the firm's number of actual working hours per week (the scaling factor transformation by the constant (1/168) hrs does not affect the regression results). The dependent variable in the firm size regressions ("firmsize") i s the (log) number o f employees reported by the firms. We first estimate a GLS robust regression for the basic models o f CU and firm size. Inthis basic model, the "minimalist model", we include as explanatory variables only those that different theories concur should be relevant. The group o f minimalist explanatory variables for CU includes: age of the firm (lnage), sector dummies, cost of materials (Incostinputcap), average wage (lnwage2n), interest rate (lnltinterest), and capital intensity (lnKin1). Then we include the other explanatory variables listed above. The GLS robust regression analysis uses blocks of variables introduced sequentially. Only variables that remain robust and significant are kept to arrive to a parsimonious extended model, which includes the age o f the firm (lnage), the size of its domestic market (lnmktsizepop, lnmktsizeinc), a dummy for exporters (exporter), capital intensity (InKinl), expenditures in R&D (InRyDspentl), and its interaction with capital intensity (RyDbyKin), the exporter dummy (RyDbyexporter) and a proxy for dishonest competition (RyDbydishonest). Inaddition, we examine variables related to financial and regulatory aspects, business relationships, contracts, infrastructure and human capital. One important caveat, which i s more relevant for the analysis o f firm size than for CU, refers to the possibility that some o f the independent variables could be thought as endogenous. Unfortunately, with cross sectional data and lack of good instrumental variables it i s not possible to formally establish the causality in some of the relations established, and care should be take in drawing policy implications. Thus, the analysis should be viewed as a first step to examine these questions through further research using datasets (e.g., panel data) suitable for addressing these methodological concerns. 103 Tables A.3.5.1-A.3.5.3 present the results for the variables that show more robust results. The variables considered explain 22-33 percent and 3 7 4 9 percent of the variation in C U and firm size among Bolivian manufacturing firms, respectively. We first summarize the main findings for the determinants of CU. Among the "fundamental" variables, capacity utilization rates are higher, on average, among firms that are younger, capital intensive, pay lower wages, use higher quality inputs, and inthe petroleum and food and tobacco sectors. The proxies for demand constrains (market size proxied by the population or per capita income of the firm's internal sales market and whether it exports) are positive and significantly correlated with C U but surprisingly turned insignificant even inthe minimalist model. This may reflect that the pervasiveness of demand constraints lead to insufficient variability in these variables in the sample. The proxy for the cost of capital (interest rate) reduces C U but becomes insignificant once other financial characteristics are added, particularly previous default. Among the "non-fundamental" variables (extended model), financial and credit constraints impose the most pressing restrictions on CU. Belonging to a business association (gremio) has a positive and significant effect on CU. Governmental requirements affecting the firm's production process (govreq) has a negative and sizable impact on CU but it i s measured imprecisely (10 percent of firms responded being affected). Firms reporting legal firing restrictions (hardtofire) tend to have higher CU. The results for firm size show the relevance of fundamental factors in explaining the distribution of firm size. Bigger firms are older, face larger markets including access to external markets, are less capital intensive, and again are in the petroleum, and food and tobacco sectors. Meanwhile, among capital intensive firms or firms with an exports orientation those that invest inR&D tend to be smaller. These signalthe followingeffects: Among other investment climate variables, only the difficulty to hire skilled workers and the availability o f contracts with suppliers has significant negative effects on firm size inBolivia. Although unmeasured determinants of firm size may correlate with the capacity to shop for scarce skills, the lack of skilled labor prevents firms from reaching larger scales of operation by adopting more efficient technologies and management practices. The existence of contracts between the firm and its providers helps firms to grow, as contracts reduce transaction and other hidden costs of expanding businesses. The availability o f contracts also facilitates the externalization o f activities leading firms to reduce size, but this effect i s seemingly unimportant in Bolivia. Thus, the difficulties for writing contracts and the shortage of skilled labor place constraints on firms' growth inBolivia. Note that results of Muiioz, Palma and Arias (2004) shows that the effect of financial variables on firm size i s not robust to the introduction of other control variables. This cannot be taken as evidenced that credit access i s not a serious constraint to the growth of firms in Bolivia given the reversed causality between size and credit constraints (e.g., small firms are riskier clients and thus face higher interest costs). However, the fact that financing does prove to be a crucial constraint for CU, especially collateral, corroborates that firm size and C U measure different aspects of firm growth. Liquidity constraints are less likely to be more bindingto firm size than C U inthe short runrelative to other demandandinstitutional factors. Firms that report legal restrictions to laying off employees have higher capacity utilization The effect of firing restrictions may suggest that labor rigidity leads firms to use their 104 payroll more intensively, getting more work from existing workers before hiring new workers. Alternatively, firms facing higher restrictions to fire skilled workers may have a more skilled workforce. Meanwhile, among capital-intensive firms or firms with an exports orientation those that invest inresearch and development tend to be smaller. 105 Table A.3.5.1 Determinants of Manufacturing Capacity Utilization and FirmSize inBolivia Capacity Utilization FirmSize I I Variable Minimalist Extended Minimalist I Modelfor Modelfor Variable Model ExtendedModel ~ ~....- capac-hr capac-hr lnage 0.2406 0.2435 Lnage -7.563 -7.911 (2.74) ** (2.64) ** (-2.30) * (-2.14)* lnmktsizepop 0.3016 0.2148 lncostinputcap 4.851 5.183 (4.92)** (3.28)** (2.88)** (2.73)** lnKinl -0.2506 -0.2023 lnwage2n -7.039 -5.701 (-2.94)** (-2.22)* (-3.36) ** (-2.40)* lnRyDspent1 0.0207 0.0110 (0.75) (0.37) lnltinterest -11.187 -4.376 1.4213 1.9706 (-1.86)' (-0.68) exporter (4.85)** (5.17)** lnKinl 4.392 4.614 -0.0112 -0.0203 (1.69)' (1.67)' RyDbyKin (-1.29) (-2.29)* Dsectorl 17.175 16.89 -0.5839 -1.2888 (2.33)" (2.OO)* RyDbyexporter (-1.65)" (-2.95)** Dsector2 6.61 7.267 hardtofire -0.1714 (0.87) (0.85) (-1.10) Dsector3 7.281 6.987 lnerrandsl (1.62) (1.26) Dsector4 24.177 27.01 gremio II (3.55)** (3.31)"" 0.7649 COll -7.109 Note: The dependent variable is capac-h = number of Note: The dependent variable is firm size = In working hours per week. Robust t statistics in (employment). parentheses. dsector1= food and tobacco, dsector2= textiles, garments, and leather; dsector3= wood, furniture, paper, and publishing; dsector4= petroleum, chemical and glass/ceramics; and omitted sector= metals, machinery, car parts, and other transport. Significant at 5%; ** significant at 1%; 'significant at 10%. Source: Authors' estimatesbasedon Bolivia's FACS survey. 106 Table A.3.5.2: Quantile RegressionDeterminants of Capacity Utilization Variable Q (10%) capac-hr lnage -0.145 (-0.1) lncostinputcap -0.068 (-0.09) lnwage2n -1.611 (-1.57) lnltinterest 2.121 (0.48) lnKinl 1.409 (1.24) dsectorl 6.376 (2.02)* dsector2 5.463 (1.49) dsector3 3.851 (1.43) dsector4 2.685 (0.8) coll -3.596 (-2.39)" default -2.921 (-0.83) otherlend 2.38 (0.88) gremio -2.272 (-0.81) govreq -12.575 (-2.86)** hardtofire 0.868 (0.35) Constant 40.241 (2.60)' Observations = 150. t-statistics inparentheses. 'a Significant at 10%; significant at 5%; * ** significant 1%. at Source: Authors' estimates based on Bolivia's FACS survey. 107 Table A.3.5.3: Quantile RegressionDeterminantsof FirmSize Q (10%) Q (25%) Q (50%) Q (75%) Q (90%) h(emp1oyrnent) h(emp1oyrnent) ln(ernp1oyrnent) h(ernp1oyment) h(ernp1oyment) 0.1480 0.2264 0.3222 0.3576 0.4532 Lnage (1.18) (1.90)" (2.60)** (2.25)* (2.67)** 0.0472 0.0825 0.2108 0.3683 0.4200 lnmktsizepop (0.52) (1.02) (2.24)* (3.28)** (3.14)** 0.0019 -0.1603 -0.1839 -0.3335 -0.2732 lnKinl (0.01) (-1.20) (-1.45) (-2.78) (-1.78)" 0.0231 0.064 0.0354 -0.0180 -0.0134 1nRyDspentl (0.64) (0.17) (0.69) (-0.32) (-0.22) 2.3517 2.0586 2.3197 1.4263 1.7078 exporter (3.50)** (3.33)** (3.59)** (2.12)* (2.33)* -0.0168 -0.0061 -0.0197 0.0049 0.0021 RyDbyKin (-0.58) (-0.27) (-0.83) (0.15) (0.05) -1.6201 -1.7009 -1.8655 -0.8295 -1.1606 RyDbyexporter (-2.20)* (-2.50)* (-2.59)** (-1.13) (-1.45) -0.1895 -0.0207 -0.0870 -0.4002 -0.4122 hardtofire (-0.71) (-0.10) (-0.42) (-1.46) (-1.21) 0.3571 0.4312 0.2825 0.1159 -0.0212 lnerrands1 (1.53) (1.89)' (1.25) (0.51) (-0.08) Cons (0.51) (0.46) (-0.23) (-0.58) (-0.83) No. Obs. 162 162 162 162 162 IR souared 0.2516 0.2809 0.3013 0.3715 0.3975 Observations = 162. The dependent variable i s firm size = ln(emp1oyment). The t-statistics are inparentheses. Significant at 10%; * significantat 5%; ** significant at 1%. Source: Authors' estimates based on Bolivia's FACS survey. 108 ANNEX 4.1 Determinantsof EarningsDifferentials inBoliva Figure A.4.1.1: Hourly Earnings DifferentialsInBolivia 1993-2002 (Heads of Household) Gender,Ethnicity, andEmploymentType Regions MainCities Main Cities Male Beni Non-indig. Chquisaca Self Employ. Inform sal. Tarija Gov. &Bus. Potosi Commerce om0 Constmct. Util.&Transp. La Paz Manufact. Cochabamba -0.6 -0.4 -0.2 0 0.2 0.4 -0.800 -0.600 -0.400 -0.200 O.Oo0 Urban Urban Male Beni Non-indig. Self Employ. Inform. sal. Go". & Bus. Commerce Tarija Construct. Util.6tTransp. Potosi Manufact. Cochatamba -0.6 -0.4 -0.2 0 0.2 0.4 0.6 -0.6 -0.4 -0.2 0 0.2 0.4 Rural Rural Male Potosi Non-indig. Chuqisaca Self Employ. Cochabamba Inform. sal. Cav. &Bus. Commerce LaPaz Construct. Tarija Util. & Transp. Manufact. -1.5 -1 -0.5 0 0.5 -1 -0.5 0 0.5 1 1.5 Note: Resultsfrom Mincer regressions(see background paper of Gasparini et al. for details). Pando and Beniwere excluded from the main cities (1993) and rural (1997) regressionsdue to insufficient observations. Omitted categories are female, indigenous, formal workers, and primary sectors (agriculture and mining) and Santa Cruz. Source: Authors' estimates basedon householdsurvey data. 109 Figure A.4.1.2: Marginal Returnsto EducationinBolivia, 1990s More efficient Urban Rural ...- 60 60I . 40 4 - - . 4.. .-..* 40 Primary 20 0 q=0.1 OLS q=o.9 ...e.. .P93 -1997 ...e.. ,1997 -2002 1 I 1 8 0 1 I Secondary 0'-, ...*..,1993-c 8 9 7 ...e.. ,1997 - - c 2 0 0 2 Technical College ::I 20 - 1 1 0- q=01 OLS q.09 q = o 1 OLS q = o 9 ..-4..,1997 xx)2 ..4..,7997 -2w2 200 200, 200 150 150 a0 KJO :. ......:.-* College 50 50 0 0 0- 9.01 OLS q=o.9 q l o l OLS q = o 9 ...*...1997-2002 ..-*.- ,897 -2w2 Source: Authors' estimates based on household survey data 110 ANNEX 4.2 EstimatingInformal-FormalEarning Gaps As stressed inthe recent literature, unconditional earnings comparisons cannot be used as evidence of the superiority of formal sector jobs. First, it cannot be claimed that the lower informal earnings are due to the characteristics of informal jobs rather than to the productive attributes of the workers (both observed and unobserved). One should compare earnings of Bolivian workers with similar characteristics, both observed and unobserved. Second, gaps in average earnings can hardly characterize the situation of workers at all points o f the earnings scale. Average earnings gaps may mask the differential situation faced by Bolivians whose unobserved characteristics place them at jobs with below or above average earnings for their skills sets. Third, and more importantly, monetary earnings gaps do not fully capture differences inthe quality of jobs across sectors in so far characteristics such as flexible work schedules, the degree of protection and non-monetary benefits (e.g., health insurance, training) are also valued (differently) by individuals. These need to be factored in as part o f the cost-benefit choice calculation o f workers. The background study by Tannuri-Pianto et al. (2004) attempt to address these issues by going beyond narrow average, cross-sectional earnings comparisons and a more careful multivariate analysis of earnings equations (Box A.4.2.1). They estimate differences in earnings across sectors for workers at different points of the sector-specific wage distributions conditioning on measured characteristics and the probability o f sector participation, that is, comparing the earnings distributions that would result if workers had the same set o f measured characteristics and similar sector self-selection propensities. It is useful to first compare the earnings distributions for formal, informal salaried workers and the self-employed (Figure A.4.2.1). It is clear that median earnings mask substantial disparities between workers at different points o f the earnings scale. The distribution for formal sector workers i s further to the right of the informal salaried reflecting their wage advantage at any wage level, and in fact, the distributions become further apart at the right tail-- earnings gaps are larger between workers at jobs with higher pay. Meanwhile the formal salaried and self-employed distributions converge at the right tail (the highest paid self-employed earn wages similar to the highest paid formal employees). A worker at the 0.10 quantile of the distribution for the formal salaried (whose wage places h i d h e r above 10 percent o f formal workers) earns about 122 percent more than a worker at the 0.10 quantile of the distribution of the self-employed in 1993. The earnings gap i s then reduced to 33 percent for the self-employed at the 0.90 quantile and goes to zero as we move to higher percentiles. 111 Figure A.4.2.1: Density of LogHourly EarningsBy Sector, MetropolitanAreas (Bs$, 2002) 1 1993 1993 2 - I I p 0" - -4 -2 0 2 4 6 -4 -2 0 2 4 6 Log hourly earnings Log hourly earnings 1997 1997 2 - I I E 0" C -4 -2 0 2 4 6 -4 -2 0 2 4 6 Log hourty earnings Loghourly earnings 2002 2002 .-h BE I 8 8 ................... >I I I I I I I I I -4 -2 0 2 4 6 -4 -2 0 2 4 6 Log hourty earnings Loghourly earnings Source: Authors' estimates basedon householdsurvey data. 112 Box A.4.2.1: Estimating Selectivity CorrectedInformal-Formal Earning Gaps The background study by Tannuri-Pianto, Pianto and Arias (2004) estimates differences in earnings in the informal, formal and self-employed sectors taking into account the fact that workers' decision to participate in a sector is affected by differences in earnings levels, which in turn are correlated with their characteristics (observable and unobservable). Average earnings do not capture the relative positioning o f workers in the conditional wage distribution, as the informal sector is very heterogeneous. The study uses quantile regressions to compare workers with the same rank in the sector specific earnings distributions. Quantile Mincer earnings equations are estimated together with multinomial choice models to correct for workers' self-selection into each sector using a two-stage procedure (Fitzenberger, 2003). Inthe first stage, the probability of participation in the formaYinformal/self-employedsectors is determined with a multinomial logit. This allows to consider simultaneity inthe choice of sector. The second-stagequantile Mincer equations include correction factors related to the probabilities of participating in each sector. This corrects sector-specific earnings equations for selection bias that cause participants to join a sector even when they have a low probability of being in that sector. Earnings include in-kindpayments. The results are then used to compute Oaxaca-type decompositions of the earnings gaps to measure the fraction explained by differences in worker characteristics and differences intheir returns across sectors. The study employs the EIH1993, ENE 1997, and MECOVI 2002 surveys for urban areas. LinkingPoverty and Conditional Ouantiles The quantile regression results allow to measure heterogeneity in the returns to workers' characteristics and determine ifpoorer households are more severely hurt by participating inthe informal or self-employed sectors and which characteristics are most damaging. The conditional returns from the first stage Mincer regressions are linked to workers' positions in the unconditional per capita household income distribution. This allows identifying the returns to characteristics received by each individual, as well as the returns they would receive in another sector, and therefore the penalty (or benefit) that they receive for their characteristics by not working inthe formal sector. To do this, the authors first identify the conditional quantile o f each worker. Quantile regressions are estimated at ten quantiles, and then the quantile to which each worker belongs is identified as the quantile with the smallest wage residual (inabsolute value). Next, household specific returns for each level of education are computed by averaging the return coefficients across workers that belong to the same household. The household specific returns for each education level on the (unconditional) household's centile are then regressed in the per capita household income distribution. The samples in these second stageregressions are composed of households with positive returns. Two specifications are considered, the first one including only dummies for the five quintiles and the second, the unconditional centile and its square. To properly gauge at the statistical significance o f the results, the second stage regressions are weighted to account for the standard errors of the estimates o f quantile return coefficients from the first stageMincerian equations. These earnings gaps in part arise from differences in productivity-related characteristics of workers throughout the distributions. The regression results offer further light on the issue of the relative performance of informal salaried and self-employed workers in the Bolivian labor market. The specific findings are discussed in detail in Tannuri-Pianto et al. (2004). Figure A.4.2.2 summarize the main results of the counterfactual decomposition of informal-formal earnings gaps at three quantiles of the earnings distributions (low, moderate and high pay jobs) for the period 1993-2002. Each bar shows the components o f the (log) earnings gaps attributed to differences in worker characteristics and differences in their returns. The left-hand side graphs show the computations using the characteristics o f a typical informal worker and the returns receivedinthe formal sector while the right-hand side graphs do the opposite. 113 Figure A.4.2.2: EarningsGap of BeingInformal inBolivia Worker with informal-like characteristicspaidinthe Worker with formal-like characteristicspaid inthe formal sector (usingformalsectorprices) informal sector (usinginformalsector prices) 100 1.m 0.80 0.80 tp P .I 0 6 0 2 0.w s" : 5 040 J 040 4u 0.20 0.20 I 1 1 "no 1993 1 1997 I2M2 1993 I 1997 12002 1993 1 1997 1 0.M 2002 1993 1 1997 I 2002 1993 1 1997 1 2002 loth -tile 501hgvMiilc 90th qmnfh j1993 I 1997 I 2002 IOthqvnlilr JOIhqvnlilc ~ 9 0 t h ~ n t i l e Irn hr to difference in work= sharaclcriatisa Duc to dffamcc in9clor pitu I Worker with self-employed-like characteristicspaid Worker withformal-like characteristics paid as self- intheformalsector (usingformalsectorDrices) emdoved usi in^ informal sector orices) 1.00 I Ion 0.80 0.80 t 0.60 0.60 s"5 .a5 0.40 0.40 0.20 4 0.20 -'""I 1 I -'--I I 1 I 1 n "" 0 nn 1993 1 1997 I2W2 1993 I 1997I 2002 1993 11997 1 2M2 1993 I 1997 1 2002 1993 1997 1 DO2 1993 I 1997 2032 lmh 50th 90th 10th 501h 90th mDus to dffcmcc in uorkcrsharactokiics DDuIodlfferencc insector prim mDuclodiff-cc UI wodcrcharaclcibriu. D h r lodiffaencsinrworpricea Note: Earnings regressions control for human capital (education and work experience), economic activity, gender, ethnicity, marital status, regional effects; and correct for self-selection into each sector. For the upper left figure, the gaps associated to characteristics and prices are measured as (Xf-Xi)*Bf and Xi*(Bf- Si), respectively. For the upper right hand side, the respective formulas are (Xf-Xi)*Bi and Xf(Bf-Bi); and similarly for the bottomfigures. Source: Authors' estimates based on household survey data. The results overall suggest that informal employment largely reflects the low opportunity costs and non-wage benefits of informality for many Bolivians. While urban workers in micro- firms and the self-employed had average hourly earnings 30 to 40 percent lower than formal workers, lower skills (e.g., 3 fewer years of schooling on average) explain almost the entire earnings disadvantage of the self-employed and about two thirds of the gap for the informal salaried. At the best paidjobs the returns to skills are not significantly different between the sectors and would seem to allow a choice for the worker. Segmentation might exist between the informal salaried and formal sectors for workers in low pay jobs for their skills set. Up until 1997 Bolivian workers seemed capable of moving freely between the formal and self-employed sectors without any wage penalties. However, by 2002 informal earnings penalties were back in force at the low payjobs as the lowest pioductivity workers (with any skill set) might have been cPumPntPd niit nf the formal sector. 114 A noteworthy finding is that the sector participation patternschanged over time. In 1993 and 1997, informal workers with formal-like characteristics were positively selected into the informal sector. In this time period, formal jobs were plentiful and workers with formal-like characteristics might have accepted those informal jobs which offered good opportunities. By 2002, this positive selection had eroded away. The reduction in the size of the formal sector from 1997 to 2002 may have been responsible by causing lower productivity workers from the formal sector to lose their jobs and move to the informal salaried sector. Again, this suggests that the lowest productivity formal workers with informal-like characteristics left (or were forced to leave) the sector. The significant positive earnings disadvantage for informal workers at the bottom of the earnings scale is likely a reflection of the lower productivity of micro firms and i s hardly compensatedby other non-wage benefits. Any unaccountedbenefits received by formal workers would only increase these differentials, and the tax burden on these low income workers (which would reduce the gaps) i s virtually non-existent. In kind payments of food, clothing, and transportation are included in the earnings estimates and should not bias informal earnings estimates. Training costs may be substantial, but formal workers enjoy wages from a minimum of 48 percent to a maximum of 210 percent more than their informal sector counterparts at the low payjobs for their skill sets, which seems quite large compared to any reasonable estimates of the costs incurredby training. Findings linking the earnings penalties of not being formal to the poverty status of workers further suggest that for low skill workers the opportunity cost of formality i s low. In 2002, the educational penalties are small for basic and primary education and increase for secondary and above, leading us to conclude that the low educated (male) informal workers do not face large penalties from not participating in the formal sector. For those with secondary education the penalties are larger, but uncorrelated with poverty. For those with higher education there are very few observations in poor households, although the penalty for informal work i s quite high. Meanwhile women in poor households face the worst penalties from informal work. Penalties for ethnicity are quite small though still correlated with poverty. 115 ANNEX 4.3 Explainingthe Labor Market Performanceof Migrants inBolivia Inthe background work by Tannuri-Pianto et al. (2004), a migrant is an individual who has moved from one cityAocality inBolivia to another in the previous five years. This i s allows comparability across surveys (EM 1993, ENE 1997, MECOVI 2002). Analyzing such recent migrants has the drawback of not fully allowing their assimilation at destination, but it restricts the sample to a more homogeneous group. The results apply to migrants at an early stage of integration to the urban economy. The earnings analysis focuses on rural-to-urban migrants. The approach for modeling migration i s based on Roy's model (1951) applied to an international migration context by Borjas (1999). In it, migration is assumed to be costly, and only occurs if net benefit (income differential minus cost) i s positive. Since the earnings of immigrants (who chose to leave) are only observed in the destination area, it i s impossible to know what their earnings would have been had they not migrated. Consequently, we must use the earnings o f non-migrants in the origin (chose to stay) as a proxy, which induces potential selection bias. This sample selection problem-inherent to the migration model (non-observable characteristics that influence the decision to migrate)-is resolved by a two-step method. In the first step, an individual's probability of migration i s estimated, and included in the second step, earnings equations for migrants at the destination, to correct for the fact that we only observe earnings of individuals who have migrated. This model i s appropriate to describe choices and earnings of average migrants, but presumes that gaps in average earnings fully characterize the situation of migrant workers at all points of the earnings scale. Also, returns to migration may vary across workers that come from and insert themselves in different points of the conditional wage distribution due to, for example, their endowment of unobserved skills or to differential opportunities (crowding effects) in the labor market. Moreover, selection biases may vary along the earnings distribution, e.g., motivations and opportunity costs differ between migrants at the bottom and top o f the distribution. The motivations and opportunities of individuals at the top o f the earnings distribution are likely different from those of the individuals at the bottom. Therefore, we estimate quantile earnings equations for migrants and non-migrants in urban regions correcting for selectivity bias using the methodology developed by Buchinsky (1998). The models are corrected for self- selectivity in 1997 and 2002, but not in 1993 because o f lack of data at origin, and include the Human Development Index (HDI) and its various components at the level of municipal sections (secci6n) as a variable correlated with the migration decision but not with earnings at destination. The probability models were estimated using Ichimura's (1993) semi-parametric least squares, butwere not very different from the results for the probit models 116 BIBLIOGRAPHY Andersen, L.,A. Brooks and A. Mercado. 2004. "Macroeconomic Policies to Increase Social Mobility and Growth in Bolivia." Institute for Socio-Economic Research, Catholic University of Bolivia, L a Paz, Bolivia. Andersen, L.,and R. Faris. 2002a. "Natural Gas and Income Distribution in Bolivia.'' Institute for Socio-Economic ResearchWorking Paper No. 1/2001, Catholic University of Bolivia, L a Paz, Bolivia. .2002b. "Reducing Volatility Due to Natural Gas Exports: I s the Answer a Stabilization Fund?" Institute for Socio-Economic Research Working Paper No. 11/2001, Catholic University of Bolivia, L a Paz, Bolivia. Arellano, M. and S. Bond. 1991. "Some Test of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations." Review of Economic Studies, 58: 277-297. Arias, 0 and W. Sosa. 2004. "Subjective and Objective Poverty in Bolivia." Background paper for the 2005 World Bank Bolivia Poverty Assessment. Azariadis, C. 2004. "Povety Traps." Forthcoming inthe Handbook of Economic Growth. Borjas, G. 1999. "Self Selection and the Earnings of Immigrants." Labor Economics, ed. by 0. Ashenfelter, 305-335, Princeton University. Bossert, T. 2000. "Decentralization of Health SystemsinLatin America: A Comparative Study of Chile, Colombia, andBolivia." Harvard School of Public Health. Boston, Massachusetts. Buchinsky,M. 1998. "The Dynamics of Changes inthe Female Wage Distributioninthe USA: A Quantile RegressionApproach." Journal of Applied Econometrics 13: 1-30. Camhi, A., E. Engel and A. Micco. 1997. "Employment and Productivity Dynamics in Chilean Manufacturing: Micro Evidence and Macro Consequences.'' In F. MorandC and R. Vergara (eds.), Empirical Analysis of Growth in Chile, Santiago: CEP andLADES, 197-225. CopenhagenConsensus. 2004. Available at http://copenhaganconsensus.com/ De Ferranti, D., G. Perry, F. Ferreira and M. Walton. 2003. Inequality in Latin America & the Caribbean: Breaking with History? The World Bank, Washington D.C. De Soto, Hernando. 2000. The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else (New York: Basic Books and London: BantamPressRandom House). Djankov, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer. 2003. "The Regulation of Labor." NBER Working Paper No. 9756. National Bureau of Economic Research, Cambridge, MA. .2000. "The Regulation of Entry." Quarterly Journal of Economics, 117, (1): 1-37. Doria Medina, S. 1990. "Caracten'sticas y Perspectivas de la Industria Boliviana." La industria: problemas y perspectivas. ILDIS.L a Paz, Bolivia. Elbers, C., Lanjouw and P. Lanjouw. 2003. "Micro--Level Estimation of Poverty and Inequality." Econometrica, Econometric Society, vol. 71(1):355-364. 117 Faguet, J. 2004. "Does Decentralization Increase GovernmentResponsivenessto Local Needs?: Evidence from Bolivia." Journal of Public Economics 88: 867-893. Ferreira, F. and P. Leite. 2003. "Policy Options for Meeting the MillenniumDevelopment Goals in Brazil: Can micro-simulations help?"World Bank Policy ResearchWorking Paper 2975. The World Bank, Washington, D.C. Fitzenberger, B. 2003. "Gender Wage Differences Across Quantiles, Accounting for Sample Selection." University of Mannsheim. Processed. Gasparini, L., M. Cicowiez, F. GutiCrrez and M. Marchionni. 2004. "Simulating Income Distribution Changes inBolivia: A Microeconometric Approach." Background paper for the 2005 World Bank Bolivia Poverty Assessment. Hamermesh, D.S. 1993. Labor Demand. New Jersey: Princeton University Press. Heckman, J. 2000. "Microdata, Heterogeneity and the Evaluation of Public Policy." Department of Economics, Universityof Chicago, Chicago, IL. Heckman, J. and C. PagCs. 2003. "Law and Employment: Lessons from Latin America and the Caribbean." NBER Working Paper No. 10129, National Bureau of Economic Research, Cambridge, MA. Ichimura, H. 1993. "Semiparametric Least Squares (SLS) and Weighted SLS Estimation of Single-Index Models." Journal of Econometrics 58:71-120. Inter-American Development Bank. 2003. Good Jobs Wanted: Labor Markets in Latin America. 2004 Economic and Social Progress Report. Baltimore: The Johns Hopkins UniversityPress. International Monetary Fund. 2004. Bolivia: Second Review Under the Stand-By Arrangement and Request for Waiver of Applicability and Modification of Performance Criteria. IMF Country Report 04/5. International Monetary Fund,Washington, D.C. JimCnez, W. and F. Landa.. 2004a. "Poverty and Indigenous Peoples in Bolivia, 1989-2002." Background paper for report: Indigenous People, Poverty and Human Development 1994- 2004. World Bank,Washington, D.C. . 2004b. "LBolivia Tuvo Crecimiento `Pro-Pobre' Entre Los Aiios 1993 y 2002?" UDAPE working paper for the World Bank BoliviaPoverty Assessment. .2004c. "Empleo y Productividad de laPequeiia y MedianaEmpresa de la Manufactura en Bolivia 1995-1999." UDAPE background paper for the World Bank Bolivia Poverty Assessment. Kakwani, N., S. Khandker and H. Son. 2003. "Poverty Equivalent Growth Rate: With Applications to Korea andThailand." World Bank Working paper, Washington, D.C. Kaufmann, D., M. Mastruzzi and D. Zavaleta. 2002. "Sustanined Macroeconomic Reforms, Tepid Growth: A Governance Puzzle in Bolivia?" The World Bank, Washington, D.C. Processed. Klasen, S, M.Grosse, R. Thiele, J. Lay, J. Spatz and M.Wiebelt. 2004. "Operationalizing Pro- Poor Growth Country Case Study: Boliva." University of Gottingen and Kiel Institute for World Economics. 118 Koenker, R. and G.Bassett. 1978."Regression Quantiles." Econometrica, Vol. 46 No.1:1-26. Kordas, G. 2002. "Smoothed Binary Regression Quantiles." Mimeo, University of Illinois at UrbanaChampaign, Department of Economics. Landa, F. 2003. "Las Dotaciones de la Poblacibn Ocupada Son la Unica Fuente que Explica la Desigualdad de Ingresos en Bolivia?: Una Aplicacibn de las Microsimulaciones." Draft, UDAPE. Lay, J. 2001. "Segmentation and Informality in Urban Labor Markets: Evidence From Bolivia andImplicationsfor Poverty Reduction." E e l Institutefor WorldEconomics. Loayza, N., P. Fajnzylber and C. Calderbn. 2002. Economic Growth in Latin America and the Caribbean: Stylized Facts, Explanations, and Forecasts. World Bank, Regional Study LAC. Machado, J., and J. Mata. 2003. "Counterfactual Decomposition of Changes in Wage Distributions UsingQuantile Regression." UniversidadeNova de Lisboa. Processed. Maloney, W., "Informality Revisited." WorldDevelopment 32, No. 7:1159-1178. Manski, C. 1985. "Semiparametric Analysis of Discrete Response: Asymptotic Properties of the Maximum ScoreEstimator." Journal of Econometrics, 32:65-108. Mondino G. and S. Montoya. 2000. "Effects of Labor Market Regulations on Employment Decisions by Firms: Empirical Evidence for Argentina." Research Network Working Paper R-391. Inter-AmericanDevelopment Bank, Washington, D.C. Muiioz, R., A. Palma and 0.Arias. 2004. "Determinants of FirmSize and Capacity Utilization in Bolivian Manufacturing Sector." Background paper for 2005 World Bank Bolivia Poverty Assessment. Muiioz, R., andA. Palma. 2003. "Determinants of FirmSize and Capacity Utilization inBolivian Manufacturing Sector." Background paper for the World Bank Bolivia Poverty Assessment. The World Bank, Washington, D.C. Mimeographed. Nina, O., and A. Brooks de Alborta. 2001. "Vulnerabilidad Macroecondmica Ante Shocks Externos: El Cas0 Boliviano". Institute for Socio-Economic Research Working Paper No. 10/2001, Catholic Universityof Bolivia, LaPaz, Bolivia. Rajan, R. and L.Zingales. 1998. "Financial Dependence and Growth." TheAmerican Economic Review 88: 559-586. Ravallion, M. 2004. "Pro-Poor Growth: A Primer". Policy Research Working Paper 3242, The World Bank, Washington D.C. Ravallion, M. and S. Chen. 2002. "Measuring Pro-poor Growth", Policy Research Working Paper 2666, The World Bank, Washington, D.C. Ravallion, M. and M. Lokshin. 2002. "Self-Rated Economic Welfare in Russia," European Economic Review. Processed. . 2001. "Identifying Welfare Effects Using Subjective Questions," Economica, 68: 335- 357. 119 Rodrick, D., and R. Hausman. 2003. "Economic Development as Self-Discovery", Journal of Development Economics (NETHERLANDS); 72, No. 2:603-33, December. Roy, A.D. 1951. "Some thoughts on the Distribution of Earnings." Oxford Economic Papers 3:135-146. Saavedra, J. and M.Torero. 2000. "Labor Market Reforms andtheir Impact Over Formal Labor DemandandJob MarketTurnover: The Case of Peru." ResearchNetwork Working PaperR- 394. Inter-American DevelopmentBank,Washington, D.C. Sachs, J. 2003. "Institutions Don't Rule: Direct Effects of Geography on Per Capita Income." NBERWorking PaperNo. 9490, NationalBureauof Economic Research,Cambridge, MA. Tannuri-Pianto, M.,D. Pianto and 0. Arias. 2004a. "Informal Employment in Bolivia: A Lost Proposition?" Backgroundpaper for the 2005 World BankBolivia PovertyAssessment. .2004b."Rural-Urban MigrationinBolivia: An EscapeBoat?" Backgroundpaperfor the 2005 World BankBolivia Poverty Assessment. Unidadde AnSllisis de Politicas Sociales y Econ6micas (UDAPE). 2003a. Estrategia Boliviana de Reduccibn de la Pobreza: Znforme deAvance y Perspectivas. L a Paz, Bolivia . 2003b. Pobreza y Deisgualdad en Municipios de Bolivia: Estimacibn del Gasto de Consumo Combinando el Censo2001 y las Encuestas de Hogares, L a Paz, Bolivia. United Nations Development Programme, (UNDP) 2000. Human Development Report 2000: Human Rights and Human Development, New York: Oxford University Press. . 1999.Human DevelopmentReport 1999: GlobalizationwithaHuman Face. New York: Oxford University Press. World Bank. 2004a. Doing Business in 2005: Removing Obstacles to Growth,Washington, D.C. .2004b. Public Expenditure Report. Washington, D.C. .2004c.Social SectorProgrammatic CreditDocument, Washington, D.C. . 2001. Bolivia Microeconomic Constraints and Opportunities for Higher Growth - Climate InvestmentAssessment, ReportNo. 24746-BO, Washington, D.C. .2000. Bolivia: Poverty Diagnostic. Washington, D.C. .1999.Consultando con 10sPobres: Reporte de SintesisNacional.Washington, D.C. . 1996. Bolivia; Poverty, Equity, and Income: Selected Policies for Expanding Earning Opportunities for the Poor. Report No. 15272-BO.Vol. 11: Backgroundpapers.Washington, D.C. 120 MAP SECTION IBRD 33374 70°W 65°W BOLIVIA SELECTED CITIES AND TOWNS B R A Z I L DEPARTMENT CAPITALS NATIONAL CAPITAL To RIVERS Porto Velho 10°S Abunã MAIN ROADS RAILROADS Miles PA N D O DEPARTMENT BOUNDARIES Riberalta Cobija Madre de Dios INTERNATIONAL BOUNDARIES Asunción Asunción 60°W This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank Puerto Heath Yata Guaporé Group, any judgment on the legal status of any territory, or any Lago Lago endorsement or acceptance of such boundaries. Madidi Huaitunas Rogaguado P E R U Beni Magdalena Lago B E N I Rogagua Santa Ana Lago de Yacuma San Luis Reyes Apere amoré San Apolo M Trinidad Paraguá San Borja S an 15°S L A PA Z Puerto Miguel Blanco Martín 15°S Acosta Lago Nevada Caranavi Ascención Ascención Titicaca Ichoa To Illampu Puno (6,362 m) Guaqui Quiquibey Concepción Concepción LA PAZ AZ Nevada Las Petas Viacha Illimani pay) (6,462 m) Chapare Gran d e San Ignacio C O C H A B A M B A Ichilo Yapacaani (Gu Cochabamba Montero S A N TA C R U Z Desaguadero San José de Chiquitos José Oruro Santa Cruz To Nevada Sajama Arica (6,542 m) Cordiller O R U R O Banados del Lago C o rdiller Aiquile Roboré Roboré Izozog Puerto Poopo Sucre Santa Suárez Suárez Salar de o Ana Coipasa Potosí Potosí To 20°S Iquique O Central Tarabuco C h a c Salar de C Pilcomayo G r a n To Campo Grande Carniri 20°S na Uyuni H Uyuni U Q U I S PA R A G U AY ce Pila y a A C A P O T O S Í To O C H I L E ccidental Villa Montes Mariscal Estigarribia Pacific To Calama G ra n d eeLd Yucuiba Tarija ípez TA R I J A Viljazón iljazón To Tartagal To BOLIVIA Abra Pampa 0 50 100 150 Kilometers To San Ramón de la Neueva Orán 0 50 100 Miles A R G E N T I N A 70°W 65°W 60°W SEPTEMBER 2004