Report No. 36568-LK Sri Lanka Poverty Assessment Engendering Growth with Equity: Opportunities and Challenges January 23, 2007 Poverty Reduction and Economic Management Sector Unit South Asia Region Document of the World Bank Table of Contents Preface vii Acknowledgements viii ExecutiveSummary ix xxi - 1. The Sri LankanEconomyin an InlernationalContext: Achievementsand Challenges I 2. Poverty, InequalityandVulnerability 9 3. A Profileof Poor Householdsand Lagging Regions 24 40 ... 111 List of Tables Table I: Healthoutcomes by wealth quintiles and sector xii Table 2: Selected provincialindicators, 2003-04 xv Table 3: Keyfindings, implicationsand knowledgegaps xx Table 1-1:Agricultural productivitygrowth, 1990-2000 4 Table 1-2: Selectedinfrastructureindicators,2000-03 5 Table 1-3: Grosstertiary enrollmentrate, 1997-2003 6 Table 1-4: Netforeign direct investment, 1980s-2000s 6 Table 1-5: Burdenof interest payments,2003 7 Table 2-1: Povertytrends in Sri Lanka 9 Table 2-2: Povertyand populationshare by province, 2002 11 Table 2-3: Povertyheadcountby districts 12 Table 2-4: Shareof each provinceinGDP 12 Table 2-5: Mean real per capita consumption 14 Table 2-6: Inequality:Gini coefficientof per capita expenditure 14 Table 2-7: Growthin mean per capita real consumption, 1996-97 to 2003-04 14 Table 2-8: Total change in real per capita income, 1996-97 to 2003-04 15 Table 2-9: Decompositionof Theil Inequalityindex 16 Table 2-10: Projectedpovertyheadcountin2015 17 Table 2-11:Averageannual percentgrowth per capita 18 Table 3-1: Distributionof employmentstatus of householdheads 25 Table 3-2: Povertyheadcountrates by employmentstatusof household heads 26 Table 3-3: Povertyheadcountby employmentstatusof youth (IO to 20 years of age) 26 Table 3-4: Povertyheadcountby industrywhere householdhead is employed, 2002 27 Table 3-5: Povertyheadcountratios by educationalattainmentof household heads 27 Table 3-6: Welfare indicatorsin Colombo City 30 Table 3-7: Povertyindicesand access to infrastructureby province 31 Table 3-8: Sectoralshares and inequalitymeasuresby province 33 Table 3-9: Povertyand educationalattainmentof householdheads by province 33 Table 3-10: Povertyand shareof paid employees inthe agriculturalsector by province 35 Table 4-1: Correlationof size of recent migrationinto Colombo City 43 Table 4-2: Estimatesof povertyheadcountratio in Colombo City 46 Table 4-3: Ratioof remittanceto consumptionexpenditure, 2002 47 Table 5-1: Selectedindicators,Sri Lanka and other countries 54 Table 5-2: Internationalcomparisonof health expenditures, 2002 56 Table 5-3: Child nutritionand healthstatus, by wealth quintiles and by sector, 2000 58 Table 5-4: Women's nutritionalstatus by wealth quintilesand by sector 58 Table 5-5: Net enrollment rates by income quintile, 1995-96 60 Table 5-6: Masteryskills at grade4, 2003 61 Table 5-7: Key educational indicators, 2002 63 Table 6-1: Householddistributionand poverty rates by sector and province, 2002 65 Table 6-2: Average annual growth rate of output and share intotal output value of the Agriculture, Forestryand Fisheries Sector, 1982-2004 66 Table 6-3: Average annual growth in wage earnings by sector, 1992-2002 66 Table 6-4: Average yields of selectedcrops, 2002-04 67 Table 6-5: Transportfacilities and rural access 73 Table 7-1:Maindata sources usedfor the analysis of the Northand East 77 Table 7-2: Labor Force Statistics(%) 2002 79 Table 7-3: Labor ForceStatistics(%) 2004 79 iv Table 7-4: Provincial GDP and populationshares 1990-2003 81 Table 7-5: Selected social indicators 85 Table 8-1: Literacy rates by sector 90 Table 8-2: Changing patternof housingstock in estates 91 Table 8-3: Selected attributesof estate householdsand asset index-poverty 92 Table 8-4: Profileof migrantsfrom estates (last 5 years beforethe survey) 95 Table 8-5:Welfare of householdswith migrants 96 Table 8-6: Asset index-poverty rates 98 List of Figures Figure 1: Poverty headcountin Sri Lanka ix Figure2: Povertyheadcountby district, 2002 ix Figure3: Distributionof real per capita monthly consumptionexpenditure(PCEXP) at 2002 prices X Figure4: Average annualgrowthof Gini coefficient xi Figure5: Poverty headcountby educationattainmentof householdheads, 2002 xii Figure 6a: Povertyestimatesfor DS divisions xiii Figure 6b: Accessibility indexat DS division level xiii Figure 7: Share of householdheadswith tertiary educationby origin district, 2001 xiv Figure 8: Provincialpovertyrateswithin the ruralsector, 2002 xviii Figure 1-1: Poverty headcounttrends in Sri Lanka, 1990-2002 1 Figure 1-2: GNI per capita, 1962-2002 2 Figure 1-3:The rank of Human DevetopmentIndexminusthe rank of GDP per capita, 2004 2 Figure 1-4:Annual rates of growth and povertyreduction, 1991-2002 2 Figure 1-5:Average annualgrowth rate of Gini coefficient, 1991-2002 2 Figure 1-6: Annual Growth Rateof Gini Coefficientand Poverty HeadcountRate 3 Figure2-1: Provinceand district boundaries 11 Figure2-2: Percentchange inheadcountbetween1990-91 and 2002 by district 12 Figure2-3: Growth-inequalitydecompositionbetween 1990-9t and 2002 13 Figure 2-4: Sectoraloutputs and contributionto GDP 17 Figure2-5: Distributionof real per capita monthly consumptionexpenditure(PCEXP)at 2002 prices 18 Figure2-6: Projected increasesin poverty headcountdueto economic shocks 19 Figure 3-1:Spatialconcentrationof poor population 29 Figure 3-2: Accessibility index and average driving distanceto Colombo correlatedwith district povertyheadcounts 32 Figure 3-3: Proportionof housing units usingelectricityor gas correlatedwith district poverty rates 32 Figure 3-4: Proportionof householdheadswith highereducationor no educationcorrelatedwith district poverty rates 34 Figure 3-5: Povertyestimatesfor DS divisions 37 Figure 3-6: Accessibitity indexfor DS divisions 37 Figure 3-7: Povertyheadcountand access to urban centersat DS division level 38 Figure4-1: Monthlywage in provinceas a percentageof wage in Colombo District 41 Figure4-2: Compositionof migrants by sectorof origin 41 Figure4-3: Shareof householdheadswith tertiaryeducation 43 Figure4-4: Educationalattainments and occupationfor migrants, recent migrants,and nonmigrantsin Colombo City 44 Figure4-5: Comparisonin housingconditionsby householdhead's migrationstatus 45 Figure 4-6: Homeownershipby householdhead's migrationstatus 45 Figure4-7: Percentof householdswith remittancesby consumptionquintifes 46 Figure4-8: Populationdensity 48 V Figure5-1: Sri Lanka's poor do relatively betterthan SouthAsian counterparts 55 Figure 5-2: Diarrheais more prevalentamongthe poor 59 Figure5-3: Privatetuition usageby quintile, 2003-04 61 Figure 5-4: Povertyand educational attainmentof householdhead, 2002 62 Figure 6-1: Distributionof population(2001)and the poor by sector, 2002 65 Figure6-2: Growth in incomes(95-96to 2002) by income groups (percentiles)for rural and agriculturalhouseholds 66 Figure6-3: Distributionof numberof owned agriculturalholdings and area by farms size, 1982 and 2002 68 Figure 6-4: Constraints Ratedas Majoror Severe Problemsby Rural Entrepreneurs 72 Figure 7-1:GDP Growth 1997-2003for Northand East 78 Figure 7-2: One Month Mean and Nominal Expenditures(2003-04) 80 Figure7-3: Foodexpendituresas a share (Yo)of per capita expenditures2003/04 81 Figure 7-4: EducationMastery Skillsof Primary SchoolStudents, 2003 85 Figure8-1:Nationalidentificationcard ownershipin estates is associatedwith better education and earnings 93 Figure 8-2: Incomediversificationis associatedwith lower poverty 94 Figure 8-3a: Primaryoccupation of estate population(age 15years and older) 95 Figure 8-3b: Monthlyearningsby primaryoccupationof estate residents 96 Figure8-4: Asset index-poverty in estates by road quality 99 Figure 8-5: Coverageof cash transfers in estates 105 List of Boxes Box 2-1: Establishingofficial povertylines 10 Box 2-2: Samurdhi benefitshave limited impacton households 20 Box 2-3: Howto improveSamurdhitargeting-aformula-basedapproach 21 Box 3-1: Underservedor poor settlements in Colombo City 30 Box4-1:Mitigatingexcessive urbanconcentration:theory and cross-countryexperiences 51 Box 5-1: Is public healthcarein Sri Lanka propoorand efficient? 57 Box 5-2: Impactof malnutritionon schooling and earnings 60 Box 5-3: Nutritionalstatus is responsiveto income 61 Box 7-1: Island-wideVulnerability Study by the World Food Program 83 Box 7-2: Vulnerabilityand PovertyProfiles (VPPs)of Villages inthe Northand East 84 Box 8-1: What factors determinepossessionof NlCs in estates? 92 Box 8-2: Externalemptoymentas a meansto achieve economic mobility 93 Box 8-3: Access to and quality of healthfacilities in the estates 100 Box 8-4: Improvementsin sanitationand healthinthe last 15years: Key informants'perspectives 100 Box 8-5: The negativeconsequencesof alcoholism inthe estates 101 Box 8-6: Perceptionson managementand trade unions in the estates 102 Box 8-7: Perceptionsabout changes in communitiesand householdsover last 15years 103 vi Preface Since 2003, the Department of Census and Statistics (DCS) in Sri Lanka and the World Bank have been working together to improve methodologies and analytical tools for measuring and analyzing poverty. This included developing an official poverty line and creating a poverty mapping system using small-area estimation techniques and Geographic Information Systems. This report uses these analytical tools as well as qualitative and quantitative research conducted in Sri Lanka and the broader development community to take a closer look at the trends, patterns and determinantsof poverty, with particularfocus on lagging regions and sectors. In many of these areas, the report extends and updates the analysis conductedfor the World Bank's PovertyAssessment for Sri Lanka2002. The report has been prepared in coNaborationwith DCS, particularlyon topics related to measurementof poverty and its correlates and poverty mapping. The analysis is primarily based on national data sets from the Household Income and Expenditure Survey (HIES), the Demographic and Health Survey and Labor Force Survey (LFS) for different years conducted by DCS, and the Consumer Finances and Socio Economic Survey (CFSES) for different years conducted by the Central Bank of Sri Lanka. These data were supplemented by a qualitative study and household survey of the estate sector conducted by the Centre for PovertyAnalysis and the Sri Lanka Business Development Centre, respectively. Notably, the poverty estimates in this report exclude the conflict-affecfed North and East, since HIES data necessary for measuring poverty is not available for this region. A chapter focusing on economic and social outcomes in the North and East provides an overview of the evidence available for this region from alternativedata sources, includingthe CFSES, LFS and smaller surveys by privateagencies. The report is intended to add to the debate and understanding of poverty and inequality in Sri Lanka. Acknowledgment of the slow and uneven pace of growth and poverty reduction across regions in Sri Lanka features prominently in the President's election manifesto, the Mahinda Chintana. The Chintana explicitly addresses regional inequalities in incomes, human development outcomes, and access to economic infrastructure. The 2007 Budget includes measures to strengthen links between emerging policy priorities articulated in the Chintanaand initial steps toward the creation of a medium-term budget framework. While the available data sources provide significant insights into poverty issues, the analysis is somewhat limited by the absence of a regular integrated householdsurvey, like the Living Standards Measurement Surveys conducted in many developing countries. In the absence of an integrated survey, information from different sources were combined and recent innovations like poverty maps were used to draw insights about the causes and determinants of poverty. The DCS is currently introducing changes to the HIES to cover a wider range of topics critical to assessing a household's well-being, which will greatly enhance poverty analysis and its ability to inform policy in the future. vii Acknowledgements This report was prepared by Ambar Narayan, Princess Ventura (Task Managers) and Nobuo Yoshida from the Poverty Reduction and Economic Management (PREM) unit of the World Bank's South Asia Region, drawing upon significant contributions from Nistha Sinha (health), Kirthisri Rajatha Wijeweera (education), Mona Sur and Dina Umali-Deininger(rural poverty), Muttukrishna Sarvananthan (North and East) and Aphichoke Kotikula (estates). The analysis of poverty using data from numerous surveys and the Population Census produced by the Department of Census and Statistics (DCS) was conducted in partnership with the DCS. In particular, the team is grateful to Wimal Nanayakkara, Suranjana Vidyarathne, G. Y. L. Fernando, and K. G. Tillakaratne (DCS) for valuable advice and contributions to the analysis. The report was edited by Hedy Sladovich (World Bank). Research assistancewas provided by Fei Gao and Gozde lsik (World Bank). Oxana Bricha, Prameela Namasivayam and Malathi Ratnayaka (World Bank) provided able assistance in handling logistical and contracting arrangements for the report, and Thelma Rutledge (World Bank) prepared the report for printing. The peer reviewers are Anila Bandaranaike (Director, Statistics, Central Bank of Sri Lanka), Gershon Feder and Linda Van Gelder (World Bank). The team is grateful to Tara Vishwanath (World Bank) for overall guidance on the report and design of the Estate household survey, as well as detailed comments at various stages. The team also gratefully acknowledges advice and comments received from Anila Bandaranaike and Rocio Castro (World Bank) during various stages of preparation; Harsha Aturupane and Kumari Vinodhani Navaratne (World Bank) for guidance on human development issues; and Uwe Deichmann and Piet Buys (World Bank) for contributions to the analysis using poverty and accessibility maps. The report benefited from extensive discussions with officials from government departments, notably the Plantation Human Development Trust and Ministries of Finance & Planning, Resettlement, Estate infrastructure and Livestock Development, Health, and Social Welfare. Comments on a previous draft received from Department of National Planning, Ministry of Finance & Planning benefited the final version. Acknowledgments are also due to participants at the Concept Note and Bankwide review meetings for comments and suggestions that shaped the report, and participants at the Regional workshop on Inequality (Delhi, December 2004) where preliminaryresults of the analysis were presented. The quantitative Estate household survey was conducted in collaboration with a team from Sri Lanka Business Development Centre (SLBDC) led by Rohanthi Perreira; a team from Centre for Poverty Analysis (CEPA) in Colombo led by Neranjana Gunetilleke and Sanjana Kuruppu conducted the qualitative study of the estates. Guidance on methodology, peer reviewing and partial funding for the qualitative study were provided by the World Bank's "Moving out of Poverty" (MOP) study team led by Deepa Narayan (World Bank). The questionnaire and sampling design of the Estate household survey benefited from the active participation of numerous individuals, notably, officials from DCS and the Statistics Department of Central Bank, the SLBDC and CEPA teams, and members of the World Bank Sri Lanka Country Team. The qualitative study benefited from suggestions received from government officials, researchers, and other industry related participantsat two stakeholder workshops organized by CEPA in Colombo. Financialsupport from the Department for InternationalDevelopment (DFID), United Kingdom for the estate and conflict-related studies greatly facilitated the new work undertaken for this report. The report benefited from oversight and comments provided by Kapil Kapoor (Sector Manager, South Asia PREM). Finally, the team thanks Sadiq Ahmed (Sector Director, South Asia PREM)for his valuable support; and Deborah Bateman (Country Coordinator), Peter Harrold (ex-Country Director for Sri Lanka) and Naoko lshii (Country Director for Sri Lanka) for their support and commitment to bring the poverty- and inequality-relatedchallenges to the forefront of the World Bank's engagement in Sri Lanka. ... VI11 ExecutiveSummary 1. The development story m Sn Lanka IS one of mixed success. The country is on par with middle- income countries and Millennium Development Goal timetables for universal primary school enrollment, gender parity inprimary and secondary school enrollment, and universalprovision of reproductivehealth services. At the same time, consumption income poverty persists and the poor continue to face basic welfare challenges such as malnutrition. 2. A number o f interrelated constraints prevent access by the poor to opportunities in more dynamic sectors o f the economy. In poor rural areas and the estates economic and geographic constraints include inadequate connectivity to markets and growth centers, lack o f electricity and transport facilities (infrastructure) and poor quality schools (public services). In poor urban areas constraints include inadequate access to clean water, electricity, sanitation and quality of housing. At the household level, the report assesses the cyclical nature o f poverty traps caused by low levels of education, poor nutrition and underemployment (mostly associated with the informal sector). Population in the estates, North and East, and the tsunami-affected coastal areas are more likely to fall into the poverty trap cycle due to historical disadvantages or recent events like civil conflict or natural disasters. I. Poverty,growth, and inequality trends in Sri Lanka 3. ConsumptionpoverQ reduction in Sri Lanka has been Figure 1: Poverty headcount in Sri Lanka modest-about 3 percentage points (from 26 to 23 percent) from 1990-91 to 2002)-and uneven across sectors. Urban poverty halved between 1990-91 and 2002, while rural poverty declined by less than 5 percentage points, and poverty in the estates increasedby about 50 percent-malung this sector the poorest in the country (Figure 1). The conflict-affected North and East are excluded fiom these estimates, since consumption data from HIES (the official source for poverty measurement) essential to measure poverty is not U90-91 95-96 I32002 Figure 2: Poverty headcount by district, 2002 o f Uva and Sabaragamuwa fell from 16 incidence of poverty fell to 11 percent for Western Province compared with 35 percent for Sabaragamuwa and Uva. 5. The largest poverty reduction in 2002 occurred in districts with low incidence of poverty in 1990-91 (Colombo and Gampaha in Western Province).2 In some districts-already I Source: DCS HIES 2002. 'Excludingthe North and East.Source: PeaceSecretariatusingCentralBankProvincialGDPnumbers. Poverty headcounts in Colombo and Gampaha were 16 and 15 percent, respectively, in 1990-91, and reduced by 10 and 4 percentagepoints, respectively,by 2002. ix 6. Poverty and vulnerability (the risk of Figure 3: Distribution of real per capita monthly consumption falling into or deeper into poverty) are closely expendit" linked, since the poor and those just above the poverty line are more susceptible to shocks. Figure 3 shows that the population is highly concentrated around the poverty line, implying that even small shocks can cause large increases inpoverty in~idence.~ 7. Current targeted welfare programs peljform well below potential. Despite a long historyofpublicly fundedwelfare programs the evidence suggests that current programs perform well below potential in protecting the consumption of the vulnerable and the poor. The largest cash transfer program, Samurdhi still misses a large proportion of the poor even Source: Staff estimatebased on HIES 199G91and 2002. Sources ofrisk at the household levelincludesickness and deathofa family member andunemploymentat the householdlevel. At the communitylevel risks includedrought, crop failure, andother naturaldisasters(World Bank 2006c, draft). 'Glinskaya(2000), usingdatafrom SLIS (1999-2000). CFSES2003-04. World Bankstaffcalculationsbased onDCSHIES 1990and 2002. X Sri Lanka (higher Gini indicates hgher inequality) Figure4: Average annuat growth of Gini coefficient increased at an annual rate of 2 percent, muchhigher than for East Asian comparator countnes with the __--____---_-._______----_I-- II exception of China (Figure 4). Thus, for every 1 percent annual growth in GDP per capita, the poverty headcount ratio declined by 0.4 percent in Sri Lanka, compared with 0.9, 1.4, and 2.6 percent inKorea, VietnamandThailand, respectively. 12. Rising inequality hinders poverty reduction. Had consumption distribubon remained unchanged from 1990-91, the 30 percent growth in average consumption by 2002 would have reduced poverty bymore than 15percentagepoints nationallyinstead Note: Gini coefficients of per capita consumption for China o f the observed 3 percentage points. At the average and Sri Lanka, ofper capita income for other countries. rate o f per capita consumption Gini increase o f the Source: China-Ravallion and Chen (2004); Korea-World Bank (2004a); Malaysia-Government of Malaysia (2001); past decade (2 percent annually) and continuing Thailand, Vietnam-PovCalNet at the World Bank; Sn GDP growth at the average rate over the past two years (5.7 percent annually), the poverty headcount in2015 is likely to be 15 percent, well short of the MillenniumDevelopment Goal of halving poverty between 2000 and 2015.' Sri Lanka will need to grow at 10 percent annually to achieve this target. Ifthe Gini coefficient remains unchanged at the 2002 level, however, a 5.7 percent rate o f annual GDP growth will be sufficient to reduce the povertyheadcount to 8 percent by2015.8 II. The correlates of poverty: householdcharacteristicsand spatialfactors 13. Poverty i s strongly associated with attributes of individualsihouseholds such as education attainment, employment status, and family size. After individual differences are accounted for (in a multivariate regression o f the probability o f being poor), the likelihood o f being poor also depends on a range of spatial factors, such as poor regional growth and employment opportunities, and the availability of infrastructure, such as roads and electricity. A. Householdand individualspecific factors 14. Larger households, especially those with children, are more likely to be Households with a member workmg abroad, however, have a significantly lower likelihood of being poor. 15. The human development challenges thatface Sri Lanka difler from those thatface most developing countries. Primary enrollment and literacy are near-universal in Sri Lanka. Similarly, basic indicators o f health-fertility, infant and child mortality, immunization rates, life expectancy, and maternalhealth-are uniformly highacross income groups. Goodbasic indicators inhealth correlate well with the highliteracy ofmothers inSri Lanka. 16. Low education attainment is strongly associated with poverty (Figure 5). In 2002, well over 30 percent of households with heads with schooling up to and including grade 5 fell under the poverty line, compared with less than 10percent for heads who completed at least grade 9. Regression analysis shows that a household is significantly less likely to be poor when the head has an education at the A-level and above. IncaseofSn Lanka, since HIESwasnot conductedin2000, thereferenceyear is 2002. Assumes population growth rate of 12 percent These projections are based on the approach proposed by Bourguignon (2003) to estimate growth elasticity ofpovertyunder certainrestrictive assumptionsabout the currentdistribution of consumption Please note the per capita consumption measure is unable to account for economies of scale in consumption and age of family members, andtherefore tends to underestimate the welfare of larger households xi 17. Disaggregated data on school enrollments show Figure 5: Povertyheadcount by education disparities among incomegroups. Educational attainment attainment Of 2o02 is significantly lower among children o f poor households. (percent) Net enrollment for grades 10-13 among the lowest income quintile (31 percent) is only one-half that of the Degree &above richest quintile (60 percent) and the net tertiary P f kade 9 &behwdegree enrollment rate for the lowest quintile (2 percent) 1s less than one-sixth that for the richest quintile (13 percent). Grade 6-8 I 18. The low quality of education acts as an additional Up to Grade 5 handicap for the poor in remote areas. Nationally, No schoohg students display a low skill level in first languages, English, and mathematics, and these indicators are even S O U ~ ~ ~ S Z O O Z0 10 20 30 40 SO lower for nonurban children. Absenteeism of teachers (about 20 percent nationally) is also higher in nonurban 1Source: Bank staff calculatlons schools. The poor quality o f education has led to a rapid 2002. increase inthe use of private tutors, and another urban-rural and rich-poor gap: CFSES (2003-04) shows that private tuition is used less by nonurban children, and twice as often by students inthe highest income quintile than those inthe lowest quintile. 19. Rich-poor gaps in health include low birthweight, malnutrition among preschool children, poor nutritional and sector status of adult women, and incidence of communicable diseases such as tuberculosis and diarrhea. Food availability, dietary awareness, and access to safe water and sanitation are often linked to income/consumption poverty (Table 1). 20. Inadequacies in nutrition and education affect lifelong earnings and trap households in a cycle of low capability and poverty. The higher incidence o f poverty among the less-educated and the fact that the poor are less likely to participate in higher levels of education perpetuates the vicious cycle o fpoverty. 21. The correlation between unemployment and poverty for thepopulation as a whole is weak, but the presence of an unemployedyouth is associated with a higher probability of the household being poor. More than 75 percent o f the unemployed are between 15 and 29 years o f age." Youth unemployment i s especially high among school dropouts, who in turn are more likely to belong to poor households. A household i s also more likely to be poor when at least one member i s employed in an elementary occupation (mostly in the informal sector)." Since underemployment i s often a feature in the informal sector, this finding supports the commonly held view that underemployment and poverty are closely linked. Individuals with lower education attainment tend to be employed in elementary occupations. Education i s thus an important underlying factor linlung labor market outcomes with poverty. Also, consistent with the pattern o f agncultural stagnation discussed later, employment as an agncultural worker increases the likelihood of poverty. loNanayakkara (2004). l1 Elementary occupations in the informal sector include wage labor in agnculture, fishing, mining and construction, street vendors, domestic helpers, and garbage collectors. xii B. Spatial or location-specific factors 22. Spatial characteristics at the district and DS division levels emerge as strong correlates of poverty. Households-identical inall characteristics-can have very different likelihood o f being poor depending on where they are located. A household located in a D S division with the average characteristics o f Colombo district i s 7 percent more likely to be non-poor than a similar household in a D S division with the average characteristics o f Monaragala district. This highly stylized example illustrates the uneven pattern o fpoverty incidence can be partly explained by the location of the household. 23. Poverty is concentrated in areas where connectivity to towns and markets, access to electricity and average educational attainment are relatively low, and agricultural labor is an important source of employment. Location attributes are highly correlated with each other, which indicate the many-sided nature of challenges faced by poor areas. Remote areas with poor connections to markets and cities are more likely to have lower access to electricity and lower average educational attainment. Figure 6: (a) Poverty estimates for DS divisions Ibl Accessibilitv index at DS division level low high Note: The accessibility index is calculated for every point as the sumo f the population totals o f surrounding cities and towns inversely weighted by the road networktravel time to each town. This map shows the mean of the access values for all points that fall into a given DS unit.The index is a measure o fpotential market integrationreflecting the quality and density of local transportation infrastructure, including 185 cities and towns inthe analysis. Source: Poverty map from DCS2004; accessibility map based on staff calculations. 24. Multivariate analysis finds that the probability of a household being poor falls by 2-3 percent with every unit increase in the accessibility index. Figure 6 vividly portrays the association between the accessibility index (potential for market integration using distance and availability o f roads from towns and markets) and poverty estimates at the DS division level, and i s based on the poverty maps produced by DCS. The probability of being poor is also greater when the household is located in a district with a higher proportion o fhousehold heads with only primary education and with l o w electricity usage. 111. Theimpact of internal migration-a consequenceof rising regional inequality 25. Migration offers upward economic mobility to those in economically marginal areas. Migration can affect cross-regional inequality by shrinking wage gaps between regions as people move in response to wage differences, and promote development in lagging regions through remittances sent back to the ... XI11 migrants' place o f origin. However, migration can also perpetuate regional imbalances, for example when the more educatedgravitate toward fast-growing cities. 26. Internal migration has almost doubled between 199697 and 2003-04 from 15 to 29per 1,000 households)." The primary migration destination i s Western Province, which includes the Colombo urban area. This trend is consistent with the expanding wage gap between Colombo and the rest o f the country, even in elementary occupations (informal sector). The overall economic benefits of migration from rural and remote areas to Colombo are large and have increased over the past decade. In2003-04, 81 percent o f internal migration was by job seeker^.'^ A substantial share o f household consumption expenditwe- around 26 percent4ame from remittances in2002 (usingHIES2002). 27. Poverty incidence in the origin district is Figure7: Share of householdheadswith tertiary strongly associated with recent migration to education by origin district, 2001 Colombo. Census (2001) indicates that a large (percent) number of migrants come from poorer districts and I xo, -- II districts in the North and East. Thus poverty seems to act as a "push" factor inducing households from economically disadvantaged areas to migrate. However, migration is more likely to be undertaken by the better-educated. Figure 7 shows that average education among migrants i s muchhigher than those in their district of origin. This indicates that the better availability o f jobs in Colombo acts as a "pull" factor for educated or skilled workers from laggingregions. 0migrantsby origindistrict residentsby district I 28. Migrants are also likely to be better-educated Note: Includes migrantsto Colombo city only. Source: wB staff estimates basedon census2oo,, Colombo City are almost twice more likely to have tertiary-level education than nonmigrants already livinginColombo City. Similarly, the proportion of migrantsworkmg inelementary occupations is much smaller than that of nonmigrants. When the household head i s a migrant, the household i s also more likely to have good housing conditions and amenities. Therefore, urban poverty i s unlikely to be a direct consequenceo fthe ruralpoor flooding urban centers. 29. Migration out of remote and lagging regions is more likely among the better-educated. It i s clear that improving education inremote areas can empower the poor with better choices to improve welfare. But over-agglomeration in Colombo created by the inflow of migrants i s leading to congestion and infrastructure bottlenecks and limitingthe potential for economic growth. Furthermore, strains on basic city services tend to be concentrated in already underserved settlements, thus reducingthe welfare o f the poor disproportionately. Cross-country analysis on the relationship between economic growth and urban concentration suggests the optimal "urban primacy" for Sri Lanka (the share o f the main urban center o f the country in total urban population) at 25 percent.14 Estimates for urban primacy for the greater Colombo area i s considerably above this level at 35-50 percent, putting the estimated "cost7' of over- l2 CFSES 1995-96 and 2003-04. These estimates are likely to understate migration, since migrants in these surveys exclude those who have migrated along with their entire household. l3CFSES 2003-04. Over the same period, external migration fell marginally from 63 to 60 per 1,000 households, but remittances from abroad grew at 11 per cent in 2003. This report focuses on internal migration because o f its close link with widening regional inequalities within Sri Lanka, as well as the current lack o f information on the impact o f external migration. The latter will be analyzed indetail inan upcoming trade study for Sri Lanka. l4See Henderson (2000). There are many country-specific characteristics that significantly affect the relationship between urban primacy and economic growth, but are not necessarily incorporated ina cross-country exercise such as this. xiv agglomeration at more than 1 percentage point in annual GDP growth." The development of alternate urbangrowthcenters wouldbetter utilizethe economic potential ofmigrationto urbanareas. similar to those for other provinces with the *Excludes Killmochchi, Mannar, and Mullaitivu. l5Given the limitations of a cross-country regression, this number is speculative. At the same time, it is a useful illustration of a broader point that Colombo urbanarea is overconcentrated,which inturn imposeseconomic losses on the country. l6CentralBank of Sri Lanka, Annual Report 1998. "DCS2004,PovertyStatisticsforSriLanka;CentralBankofSriLanka, CFSES2003-04. xv 2001) to post-cease-fire (2002-03) years.18 Unemployment fell from 13 to 9 percent in the North and from 16 percent to 10 percent in the East from 2002 to 2004, while the national unemployment rate dropped only marginally from 8.8 to 8.3 percent." 34. Signijkant constraints tosustaininghighgrowth in the NorthandEast remain. These include (i) poor availability and accessto financial services, (ii) poor access and quality o f economic infrastructure (roads, telecommunications, and water), (iii) time restrictions on the use of the A9 highway, (iv) fishing restrictions, (v) limits on mobility in certain areas such as Jafha, and (vi) outmigration of the better- educatedto the rest of the country or abroad. The tsunami o f 2004, as mentioned above, i s likely to have aggravatedthe poverty challenges inthe hardest-hit Eastern Province. 35. The cease--re since 2002 has presented the North and East region with the longest semblance of normalcy andpeace in recentyears. Initial studies have shown significant peace dividends for the North and East. The sustainability o f growth in the regon and prospects for significant poverty reduction will depend on whether durable peace i s achieved and the institutional constraints mentioned above are addressed. B. Poverty inthe estates 36. Thepoverty headcount in the estates increased between 1990-91 and 2002 and is now 7percentage points higher than the national average (Figure 1). The story inthe estates-home to only 5 percent o f the country's population-seems to be one o f stagnation, rather than a drastic fall in welfare. A high concentration o f consumption per capita within a narrow interval around the poverty line implies that small shifts can result in large swings in poverty, and the large increase in poverty headcount occurred due to a slight shift in the consumption distribution (Figure 3). A comprehensive Estate Household Survey and qualitative study were conducted to examine the determinants o f estate poverty in depth and discern patterns among types o f estates by size, location, type o f crop (tea or rubber), and management (regional plantation company, privately owned and state owned). An Asset Index (AI)score was used as a proxy for household wealth.20 37. Estate households with more earners andwith better-educated household heads tend to be better-of Possessiono fNational Identity Cards (NICs) among household headsis also associatedwith significantly lower AI-poverty rates, and those with NICs have better opportunities to earn outside the estates. Ownership of NICs i s particularly low among youth (age 16-19), which is probably relatedto the degree of isolation of households and estates, and, inturn, their increasedvulnerability to poverty. 38. More than 40percent of estate households rely solely on estate wagesfor earned income. The AI- poverty rate is highest among households with wage incomes from only one source, be that within the estate or outside it. Regression analysis shows that while wage employment outside the estates i s not associatedwith significantly higher welfare or earnings, households that receive income from enterprises tend to fare better.21Therefore, diversifiing to occupations beyond wage employment-particularly to income from enterprises-is associated with higher welfare, a pattern that also holds for the rural sector (see below). 39. The "ideal" diversi3ed livelihoodportfolio of an estate household would include remittancesfrom household members working overseas. Consistent with countrywide trends, estate households that receive remittances from overseas migrants tend to be better-off. Internal migration to urban areas, with or without remittances, does not seem to matter significantly for estate households' wealth or AI-poverty. Peace Secretariat using CBSL data. l 9DCSLabor Force Surveys (LFS) 2002 and2004. 2o AI-Poverty Rate is defined as the percentage ofhouseholds whose AI scores lie below the 30th percentile. The 30th percentile was used since the povertyheadcount for the estate sector i s 30percent. 21 Enterprise incomes come from nonagricultural businesses and from sales of crops, livestock, and livestock products. xvi This is partlybecause the rationale for internal urbanmigration is not limitedto short-term earnings, but rather includes the expectation of future opportunities, and access to networks and shlls development. 40. Higher poverty among estate hotrseholds is associated with the remoteness or lack of useableyear- round roads linking the estate to the nearest town. Nearly 42 percent o f estate households cannot use the road to the nearest town at all times o f year. Another social factor that emerged from the Estate Survey was alcoholism. About 80 percent of estate respondents mentioned alcoholism as a problem, and 75 percent o f community informants reported no improvement over the last 15 years. The increased availability of illicitly producedalcohol was perceived as aggravating the problem. 41. No clearpicture emerges as to why consumptionpoverty in the estates stagnated or worsened slightly over thepast decade. One trend in employment and diversification data for the period 1996-97 to 2003- 04, consistent with higher poverty, is a fall inthe average number o f income earners inestate households (from 2.3 to 1.7 per household). Ths may reflect a growing dislike for estate work by youth. The qualitative survey found youth avoid estate work in favor o f remaining unemployed until the right opportunity outside estates came their way. 42. A majority of households interviewed in the Estate Survey reported an improvement in their household conditions in the last 15years, despite the overall deterioration in the conditions of the estates. This difference inperceptions i s partly explained by the increasing role of nonestate employment among estate residents, which may partly de-link the condition o f the estate from that ofthe household. 43. Many aspects of health, education, and housing have improved in the estate; and this is also corroborated by the perceptions of respondents to the Estate Survey. Nevertheless, the estates lag well behind the rest o f the country on key indicatorsof health and education. For example, 37 percent o f estate children are stunted compared with 14 percent of rural children; and 48 percent o f estate mothers have low body mass index (BMI) compared with 23 percent o f rural mothers (DHS, 2000). Qualitative interviews indicated dissatisfaction with the quality o fhealth services. Access to hospital care i s a serious concern in remote estates; survey respondents reported an average one-way travel time o f more than an hour to the nearest hospital. Households perceived an overall improvement inaccess to education over the past 15 years, but the cost and quality of education and teacher shortages were frequently raised as concerns. 44. Nearly 30percent of thepopulation in the estatesector is poor but only I 3percent of the households interviewed reported receiving cash transfers from government welfare programs. The actual coverage rate in the estates contrasts sharply with the Samurdhi coverage rate for the rest o f the country (40 percent). Coverage by other social programs-with the sole exception o f sanitation-also appears to be low. Programs in childcare, early chldhood development services, housing, training and awareness, and microcreditreachless than 15 percent o fthe households sampled inthe Estate Survey. 45. Poverty traps in the estate sector and remote, rural areas of Sri Lanka share many characteristics, including a lack of connectivity and access to infrastructure to lack of education. Internal and external migration rates and participation in self-employment for the estates, although improving over time, are well below the rates for the rest o f the country (CFSES). A long history o f isolation o f these communities has contributed to inadequate networks and mobility beyond the estates. Even when estate residents are able to find work in outside jobs or migrate to urban areas, the opportunities are limited due to lower education attainment. 46. Qualitative analysis indicates that a root cause of persistence of poverty in the estates is the unique organizational structure of estates. Historically, the estates have employed resident workers who originally came from a foreign country; and even today, much o f the labor i s provided by a resident workforce. The relatively unchanged estate organizational structure is found to contribute to a sense of marginalization, leads to isolation, and adversely affects economic decisions o f households. The long- term future o f the sector appears to be in moving away from resident labor structure and toward a xvii standard employer-employee relationship. For this to work, however, the commercial viability o f estates has to improve to support higher labor productivity and wages. C. The challengeof ruralpoverty 47. Poverty reduction in the rural sector- home to 80percent of thepopulation and about 3.5million of the county 's poor-has been stymied by stagnation in the agricultural sector. Nearly 58 percent o f the rural population depends on agncultural, at least partially, for their livelihood.22Agriculture GDP growth slowed from 2.8 percent duringthe 1980sto 1.6 percent duringthe 1990sand to 0.9 percent during2002- 04, while national GDPhas been growing annually by 5 percent since the 1 9 9 0 ~ . ~ ~ 48. Households dependent on agriculture tend to be poorer (Figure 8). Nearly 24 percent o f rural agncultural households are poor, compared with only 16 percent o f nonagricultural households. In the poorest province o f Uva, the poverty rate for agncultural households (34.3 percent) i s double that for nonagriculture households (16.9 percent) (Figure 8), yet agriculture comprises 53 percent of GDP inUva. Even in Western Province, where agriculture makes up Figure 8: Provincialpoverty rateswithin the rural sector, 2002 only 3 percent o f provincial 3% - GDP, and where market access 30% I and economic infrastructure are substantially better than 25% elsewhere, 15 percent of aoY+ households engaged in agnculture are poor compared 15% with 9percent o fthose engaged 10% in other activities (Figure 8).24 According to CFSES (2003- 5% 04), the average monthly income o f those engaged in agnculture i s only 60 percent of that in industries and less than one-half o f that in services. 1 Source: Staff calculations based DCS HIES2002. 49. HIES data shows skewed and limited improvement in incomes in rural areas, especialb among agricultural households.The poorest 7 percent of rural and 10percent o f agricultural households suffered a decline in real income between 1995-96 and 2001-02. Only 19 percent o f agricultural households experienced an income increase greater than 10 percent. Raising agncultural productivity is therefore imperative for reducing rural poverty. Existing studies indicate that policies to ease farmer access to improved technologies and ensure sustainable use o f water and adjustments in existing trade and land policies can facilitate higher productivity." 50. Thegrowth of the rural nonfarm sector has signijicantly contributed to a reduction in rural poverty. The nonfarm sector has become increasingly important in rural areas, generating 67 percent o f rural employment in 2003. Nearly 52 percent o f the average rural household's income came from nonagncultural wages and self-employment in nonfarm enterprises (HIES, 2002).26 Even for households engaged in agriculture, the share of nonfarm income in their total income exceeded that o f agricultural income. The relative importance o f nonfarm income is higher for paddy farmers (probably because paddy 22DCSHIES2002 23Staff calculations based on data from Central Bank Annual reports, various years. 24Sectoral contnbution to provincial GDP from Central Bank data (2004). 25See, for example, World Bank (2003) and World Bank (2004~). 26Ofthe estimated 620,000 ruralnonfarmenterprises scattered throughout the country, most are inmanufacturing (41 percent) or trading (38 percent). The average rural enterpnse i s a mcroenterprise employing about 2 4 workers including hired labor, farmly members, and the relatively young, wth an age of about 9 years. xviii has lower value and its cultivation requires less labor days than most crops) and lower for farmers growing crops like tea, rubber, h i t , and vegetables. Among rural households owning and operating a nonfarm enterprise, the poverty rate i s 13 percent, compared with 23 percent for households without a nonfarm enterprise. The monthly income per capita for households operating a nonfarm enterprise is 20 percent higher thanhouseholds without such income. 51. Limited employment expansion in the agricultural sector means that poverty reduction will critically depend on the rate at which the rural nonfarm sector can absorb workersfrom the growing laborforce.27 Major obstacles for rural nonfarm enterprises include poor transportation, problems with accessing finance and the cost of finance, access and quality o f electricity, and marketing difficulties.28 The high correlation o f location-specific characteristics (like accessibility and electricity use) with the pattern o f regional poverty in Sri Lanka (section 11)reinforces this view. Addressing these constraints will improve opportunities for the estate population as well, since employment in enterprises i s associated with higher earnings inthe estates. V; Concludingremarks 52. Poverty reduction has been slow due to widening inequalities among income groups and across regions, and becausegrowth is concentrated in WesternProvince. More inclusive economic growth will require easing specific constraints affecting particular sectors, regons, and groups, but priorities critical for all include improving the quality o f education, access to infrastructure like electricity, connectivity to markets and urban centers, and access to finance for microenterprises. These improvements will expand opportunities for the poor and those inlagging regions interms o f moving to higher paying occupations, setting up or expanding microenterprises, or migrating to work in modern industries. Since many o f these opportunities are created in the urban sector, poverty reduction will require better and simultaneous coordination betweenrural development strategies and urbanplanning and development. 53. Policies to address poverty reduction must address multiple dimensions simultaneously. The President's ManifesteMahinda Chintana-with its focus on improving the road network, access to electricity and access to finance outside Colombo i s consistent with the needs o f lagging areas and sectors identified in this report as drivers o f change to accelerate growth and poverty reduction. Improving the connectivity of poorer and remote areas to markets will be particularly important, as illustrated by the resurgence o f economic growth inthe North and East following the opening o f the A9 highway to Jaffna. 54. Table 3 summarizes key cross-cutting and sectorhegion specific findings along with their implications for the way forward interms o fpolicy priorities and knowledge gaps for future analysis. The analytical underpinnings for these conclusions are found inthe chapters that follow. 27 In2003, 82 percent of the total labor force of roughly 7.2 million workers was in rural areas, and nearly 106,000 peoplejoin the labor force everyyear. 28ADB andWorld Bank(2005). xix Poverty reduction hasbeenslow during the past 0 Economic growth needs to be more inclusive of decade and unevenacross sectors-rapid inthe urban lagging regions and poor households. A better sector, but slow or stagnant inrural and estate sectors. understanding o f the patterns and causes of 0 Poverty reduction has beendampened byrising poverty requires representative household data for inequality between as well as within sectors and the entire country, including the Northand East. regions. 0 Current social welfare programs like Samurdhi 0 High concentrationo fpopulation around the poverty performwell below their potential, primarily due line suggests a sizeable vulnerable population at risk to targetingproblems. offalling into poverty inthe aftermatho f shocks. 0 A better targeting system (as currently being employed intwo Northern districts) will improve the impact o f Samurdhiprograms. Causesand impact of regionalhpatial inequality: Improving connectivity of remote areas to market Poor areas characterizedby low connectivity to centers and infrastructure availability. markets and towns, low availability o f electricity and International experiences suggest investing in human capital. Largest gaps exist betweenWestern Provlnce and the rest o f the country. transport inj-astructure is effective to develop regional growth centers, when complemented by Migration into Western Province has increased rapidly investments inkey areas (e.g., access to finance inrecent years dueto risinggaps ineconomic and electricity). Further analysis is neededto opportunities. identifythe comparative advantages of specific A high proportionofmigrants originate frompoorer investments and interventions to exploit those and conflict-affected districts. advantages The ability to migrate appearsto be linkedto Improving education and skills inremote areas can education. Migrants inColombo tendto bebetter- enhanceemployment choices, including the ability educated than nonmigrants at origin and destination. to migrate. Migration adds to the growthpotential inthe urban Better urbanplanning and quality o f urban center. However, overagglomeration inthe Colombo services canhelpreduce the costs o f urban area can limit growthpotential andreduce overconcentration andmitigate its impact on the welfare due to increased costs o fcongestion and strain poor. on services. Human development: 0 Reducing income poverty is likely to reduce Basic welfare indicators (e.g., child andmaternal malnutritioninthe mediumto long run.Inthe mortality, primary and secondary enrollments) are short term, nutritional outcomes will benefit from generally highfor boththe poor andnonpoor. But improving access to safe water and sanitation and rich-poor gaps exist along certain dimensions that can nutritional awareness. have lifelong effects on earnings, andperpetuate More analysis is neededto explain high dropouts poverty across generations. among poor children. Key question: are dropouts 0 Poverty is associatedwith higher rates o fmalnutrition explainedby supply or demand side factors, or a among women and children and lower education combination o f both? attainment; poor children are more likely to drop out andnot be enrolled at grade 10and above. The quality oflearning could be improved by emphasizing Englishand informationtechnology 0 Quality o f education, particularly outside urban areas, skills and ensuring teacher deployment to rural is an area o f concern. The poor are especially affected, areas. since they may not be able to compensatefully with private tuition. xx Rural sector: Improve farmer access to technologies, trade, land Poverty inrural areas is higher among agricultural and irrigationto helpraise agricultural households, due to stagnation in agricultural incomes productivity. during the past 10years. Obstacles to growth and the start-up o fnew Ruralnonfarmsector hasbecome increasingly enterprises include access to and quality o f energy important for employment and incomes; and income andtransport, as well as access to finance. from nonfarmenterprises appears to reduce poverty. Improve access by rural entrepreneurs to markets Givenlimited opportunities inagriculture, poverty and market information. reduction canbe improved through growth o fthe rural nonfarmsector to absorb workers from the growing labor force. Estate sector: Improving connectivity to towns, coverage of Lagsbehind national averages inpoverty rates and National Identity Cards and quality o f health and human development (e.g., malnutritionand education services canhelp improve economic educational attainment). Inthe estates the opportunities. concentration o f consumption around the poverty line Increasingthe coverage o f Samurdhi transfers in is much higher than for the country as a whole. estates cantake many above the poverty line. Poverty is higher inestates that lack an all-season road to the nearest town. Not possessinga NICi s linked Expansion o f coverage o f social program in withhugherpoverty amonghouseholds. microfinance and skills development can improve households' ability to diversify to other income Income from outside enterprises andremittances from activities. overseasmigrants reduces poverty. Butthe ability o f estatehouseholds to diversify their sources o f income More analysis is necessaryto explore how to is low compared wlththe rest o fthe country. improve the commercial viability o f estates, and move the structure toward a more standard labor Economic decisions and opportunities are adversely arrangement. affected by marginalization from the mainstream, and the current organizational structure o f estates. Conflict-afected areas: Removing constraints on the mobility o fpeople The conflict-affected Northand East lag behindthe and goods, suchas on fishing and onhours o f rest o fthe country ineconomic infrastructure and key operation o f the A9 highway, will yield significant human development outcomes. economic benefits. Remittances appear to have contributed to Fillingknowledge gaps inthe patterns and causes safeguarding income and consumption levels to a o fpoverty will only be possible withmore certain extent. complete and representative householddata. The cease-fue since 2002 has spurred enough Sustainable peace remains a necessary economic recovery inthe North and East to increase precondition for sustained economic growth and real GDP and reduce unemployment. poverty reduction inthis region. xxi 1. The Sri Lankan Economy in an InternationalContext: Achievements and Challenges 1. Human development indicators for health and education in Sri Lanka are now almost on par with developed countries. Sri Lanka has already attained or is close to attaining the Millennium Development Goals (MDGs) of universal primary school enrollment, gender parity in primary and secondary school enrollment, and universal provision ofreproductive health services well before the target year of 2015.l 2. Improvements in nonincome dimensions of welfare, however, are not matched by improvements in income or consumption poverty. In 2002, 23 percent o f the Sri Lankan population was poor, with per capita consumption expenditures below the national poverty line. Between 1990-91 and 2002, the poverty headcount rate declined by only 3 percentage points, well behind the pace o f poverty reduction necessary to attain the Millennium target o f halving poverty incidence by the target year of 2015. 1.1. Poverty trendsin SriLanka, 1990-2002 decline from 26.1 percent in 1990-91 to 22.7 percent in 2002 (Figure 1-1). During this decade Lanka,1990-2002 , national poverty increased by almost 3 percentage 40 points &om 1990-91 to 1995-96, and then fell by 2 more than 6 percentage points from 1995-96 to -g 30 2002.2 These estimates exclude the conflict- 20 73 affected North and East, since HIES data (the I 8 10 official source for poverty statistics in Sri Lanka) 0 suitable for estimating poverty is unavailable for National Urban Rural btate this region (see Chapters 2 and 7 for more details). 4. Economic growth duringperiods of poverty reduction was unevenly distributed, resulting in Source: HIES for different years, usingofficial poverty lines for Sn Lanka(data excludesNorth& East) 1.2. A cross-country comparisonof poverty, growth, and inequality 5. Why has poverty reductionbeen slow and unevenacross sectors and regions inSri Lanka? Some aggregate (macro level) studies suggest that Sri Lanka has fallen substantially short o f its income growth potential. When income growth for Sri Lanka i s compared with the Republic o f Korea, Malaysia, and Thailand t h s argument seems reasonable (Figure 1-2). In the 1960s, per capita income in Sri Lanka compared favorably with that o f these three counhles, but is currently less than one-tenth that of Korea, one-fourth that of Malaysia, and one-half that o f Thailand. And China, which lagged far behind Sri Lanka inper capita income inthe 1 9 6 0 recently overtook Sri ~ ~ Lanka after more than 10years of extraordinary growthperformance. 6. The UNDP's global Human Development Report 2005 shows that Sri Lanka's human development index, and its achievements in nonincome dimensions o f welfare is among the * Excludesthe ' SeeAnnex 1, Table A-1.l NorthernandEasternprovinces. 1 highest in developing countries, in sharp contrast to its rank in terms of GDP per capita. Sri Lanka's rank on GDP per capita minus that on Human Development Index (I-IDI) i s larger than most East and South Asian comparators (Figure 1-3).3 Since human development was an important prerequisite for East Asian countries' rapid growth during 1980s and 1990s, this indicator can be interpreted as further evidence that Sri Lanka has been unable to fully exploit its growthpotential. Figure 1-2: GNIper capita, 1962-2002 Figure 1-3: The rank of Human Development (current US$) Index minusthe rank of GDPper capita, 2003 20 ,, 1 I__-- moo0 16 15 MOO0 lzooo 10 DO00 5 -Si Lanka 8000 0 6000 -5 4000 -10 2000 0 N r - N r - N r - N r - N w m r - r - 0 3 0 3 o ) m o z z z a z z a m : : Source: World Bank, WDI (2005) Source: UNDP, HumanDevelopmentReport (2005) Figure 1-4: Annual rates of growth and poverty Figure 1-5: Average annual growth rate of Gini reduction in the 1990s(percent) coefficient inthe 1990s(percent) __- 7 9riLanka China (90- Korea Malaysia Thaiknd Vietnam (91-02) 01) Rep (90- (90-99) (90-02) (93-02) 01) 0Annualrate ofpovertyreduction Annual erowih rate Source: Realper capitaGDP data:- World Bank, WDI, (2005a). Note: Gini coefficientsofper capita consumphonfor China and Povertyheadcountratios:Sn Lanka:-DCS (2004); China:- Sn Lanka, of per capitaincome for the rest Ravallion and Chen (2004); Korea:-World Bank (2004a); Source: China-Ravalhon and Chen (2004), Korea-World Bank Malaysia:-World Bank (2004b), Thailand:-& Jitsuchon (2004); (2004), Malaysia-Govemmentof Malaysia(2001), Thailand, and Vietnam:-Glewwe et a1(2000) and Carolyn Turk (2005).4 Vietnam-PovCalNet at the World Bank (vanous years), Sn Lanka-HIES(2002) 7. Per capita GDP of Sri Lanka grew at an annual average rate of 3 percent between 1991and 2002 - healthy in absolute terms, but much lower than the rates at which the East Asian comparators grew in the 1990s in spite o f the East Asian economic crisis (Figure 1-4). Furthermore, the rate o f poverty reduction in Sri Lanka was slow even for the extent of growth The HDIi s a composite index of life expectancy, adult literacy, school enrollment, and incomeper capita. Annual rates of poverty reduction are based on national poverty lines using household surveys. For Sri Lanka, 1991- 92 and 2002 are used; for China and Thailand, 1990and2002; for the Republic of Korea, 1990and 2001; for Malaysia, 1990 and 1999; for Vietnam, 1993 and 2002. Consumption is used for Sri Lanka, Vietnam, and China, while income is used for Korea, Thailand, andMalaysia. Annual rates of growth are computed using real per capitaGDP. 2 that occurred. For every 1 percent annual growth in GDP per capita, poverty headcount ratio declined by 0.4 percent in Sri Lanka, compared with 0.9 percent in Korea, 1.4 percent in Vietnam, and 2.6 percent in Thailand (using Figure 1-4). These ratios must however be interpreted with caution since poverty headcounts are measured with reference to national poverty lines that are not comparable to each other. 8. The slow response o f poverty reductionto economic growth in Sri Lanka is linked to rising inequality. Figure 1-5 indicates how inequality (measured by Gini coefficients) increased at the annual rate o f 2 percent inSri Lanka, which is much higher than its East Asian comparators, with the sole exception of China. A s chapter 2 will lay out in detail, Sri Lanka witnessed widening intra- and inter-regonal inequality duringthis period. 9. A comparison with other South Asian Figure 1-6: Annual Growth Rate of Gini countries (see Figure 1-6) reveals that the average Coefficient and Poverty Headcount Rate annual rate o f poverty reduction in Sri Lanka was (percent) among the lowest inthe region over the last decade. This i s partly explained by the rate of increase in 2 0 inequality (as measured by the Gini of per capita 1 0 0.0 consumption) in Sri Lanka being among the highest -1.0 inthe region, exceeded onlybyNepal's5 -2 0 -3 0 -4 0 10. What are the obstacles to growth andpoverty I -5.0 ` I reduction in Sri Lanka? Major obstacles cited in existing studies include the two-decade-long civil war; inadequate infrastructure, particularly in rural areas; political instability; a large fiscal deficit; the stagnant agncultural sector; and labor market 61 C h i Coefficient Poverty Headcount Rate rigidities caused by some o f the existing labor regulations. 11. The next section uses cross-country Source. Vanous publications (see sources in Annex 1, TableA-1 2) See Annex 1, Table A-1.2 for detailed figures on Gini coefficients and poverty estimates. 3 1.3. Factors that lintit growth andpoverty reduction in Sri Lanka: insightsfrom a cross- country comparison 13. What does a cross-country comparison tell us about the prospects for growth and poverty reduction in Sri Lanka? A few specific factors stand out-including those identified in the DevelopmentPolicy Review of the World Bank (2004~). Lack of economic growth outside Western Province 14. The growing urban-rural gap (Figure 1-1) i s mirrored by widening gaps in income and poverty between Western Province and the rest of the country, since most of the country's urban population live in Western Province-where the country's largest city, Colombo, is located. Provincial GDP o f Western Province grew by an average of 6.2 percent annually during 1997- 2003, whereas that of the other provinces (excluding Northern and Eastern provinces) grew by 2.3 percent.6 In 2002, Western Province accounted for about one-third of the country's population but contnbuted one-half o f its GDP; as a result, poverty incidence in Western Province was 11percent compared to the national average of 23 percent. The Colombo district accounts for much o f Western Province's relative prosperity, with its large population, low poverty rate (6 percent), and muchofthe modern sector ofthe economy. 15. A higher trajectory for growth and poverty reduction in Sri Lanka would thus require improving growth prospects outside Western Province. While Westem Province fares better than the rest of the country in terms of human development outcomes, these differences are not as sharp as those ingrowth and income, suggestingthat factors other than the availability of human capital also play a role in constraining growth inthese areas. Subsequent chapters of this report will analyze some o fthese limitingfactors, whichinclude stagnation inagnculture, lack of access to and poor quality of infrastructure and services, inadequate connectivity to markets, and lack o f alternative markets or urbangrowth centers inlaggingregions. Stagnant agricultural sector 16. The lack of growth outside Western Province, areas that are predominantly rural, i s in a large part due to stagnation in Table 1-1:Agricultural the agncultural sector (Table 1-1). The rural areas are home to productivity growth, 1990-2000 88 percent o f the poor in the country, and 58 percent o f the rural population depend at least partially on agnculture for their livelihood (see chapter 6). Agncultural productivity grew at an annual average rate of 0.5 percent, which was much lower than other East Asian countries as well as some South Asian countries like India and Palustan. Thailand Vietnam 17. Despite slow growth in productivity and the shrinkmg contribution o f agnculture to national income over time, employment in the agricultural sector remained almost Source: World Bank ,WDI (2005) Peace Secretariat using Central Bank ProvincialGDP numbers. 4 Inadequate infrastructure 18. The economic infrastructure inthe country worsens the further one gets from Colombo, which is part of the Road Access to Telephone density reason for the lack o f regional markets or electricity mainlinesper (kmhq. km) alternative growth centers. In this Sri (2000) 1,000people (1997- Lanka is not alone: many other South Asian countries experience similar constraints to economic activity.' Chapter China 99 209 0.19 3 links access to infrastructure and the Korea, Rep. 100 538 0.88 spatial distribution o f poverty, but it i s Malaysia 97 182 0.20 also useful to look at where Sri Lanka Thailand 82 105 0.11 stands with respect to comparators Vietnam 36 54 0.29 countries in East Asia in terms of electricity and telephone service, and road networks (Table 1-2). 19. Access to electricity i s more o f a bottleneck in Sri Lanka than comparator countries in the region (World Bank, 2004~).The Iatest report from the Central Bank suggests that access to electricity has improved from 62 percent of the population in 2000 to 76 percent in 2003, which would still place it well behind all East Asian counties except for Vietnam. Sri Lanka also lags in terms of access to telephone mainlines, with only 49 telephone mainlines per 1,000 people. Since both industry and service sectors-critical fast growing sectors in Sri Lanka-rely on telecommunicationnetwork, the low penetration o ftelephones poses a challenge for the future. 20. The picture on road density is more complicated. Sri Lanka's road density of 1.5 kmper 1 km2ishighfor a developingcountry, buta longhistory o fpoorroadmaintenance meansaccessin remote areas i s not as good as the density statistics might suggest. A recent World Bank study finds that only 10 percent o f the road network is properly useable.8 Aggregate figures also mask inequality in access to infrastructure within Sri Lanka, which tends to reinforce spatial inequalitiesinincomes andpoverty, a theme that will be explored inchapter 3. Economic impact of the civil conflict and recent tsunami 21. More than 20 years of civil conflict has profoundly affected Sri Lanka's national economy, resulting in increasing economic and social disparities between the populations in the North and East and the rest o f the country. The accumulation o f public debt-at least partly a result of increased military expenditures inresponse to the conflict-has crowded out a range of propoor public services; although expenditures on health education were protected (at around 4 percent of GDP combined) as military expenditures expanded from 0.5 percent o f GDP in the 1970s to 5-6 percent in the 1990s. According to World Bank (2005a), Sri Lanka's military spending was among the highest inthe world at 5.3 percent of GDP, and more than 2 percentage points higher than that any o f its East Asian comparator^.^ After the ceasefire treaty in 2002 militaryexpenditures declined to a level comparable to China, Malaysia, andKorea. 22. The instability brought about by the civil conflict reduced investments andjob creation. At the macro level, the Central Bank attributes 2 to 3 percent o f annual GDP growth loss directly to the conflict." An intangible economic cost i s the energy and attention o f policymakers diverted from needed economic reforms and policies, especially during 1983-89 when the government E.g., ADB andWorld Bank, Sri LankaInvestmentClimateAssessment (2005). World Bank(2004~). See Annex 1, Table A-1.3 loCentralBankAnnual Report(1998). 5 leadership was preoccupied with the war inthe NorthandEast and an insurrectioninthe South. A similar distraction fi-om economic priorities is discernible during 1995-2001, when the civil conflict intensified. The conflict also prevented Sri Lanka fromreapingthe full economic benefits o f liberalization that took place since 1977. The economy grew at an annual average rate of 4.4 percent during the conflict years o f 1983-2001, compared to 4.6 percent during the preconflict years, 1964-82. Educationalattainment 23. Educational attainment, universally considered a pre- condition for growth, is inversely correlated with poverty in Sri Lanka (chapter 3). Thanks to its long-term commitment to education, Sri Lanka exhibits higheducational attainment at the primary and secondary levels. With gross secondary enrollment rates of 84 and 89 percent among males and females of secondary age, respectively, Sri Lanka actually outperforms many East Asian comparators (see Annex 1, Figure A-1.1). Primary school enrollment rates are near 100percent in Sri Lanka, similar to most o f the East Asian comparator countries. 24. Tertiary education outcomes in Sri Lanka have lagged behind East Asian comparator countries, suggesting constraints in the availability o f human capital. Gross tertiary enrollment rates for males and females were 6 and 4 percent, respectively, which is much lower than the rates inKorea, Malaysia, and Thailand (Table 1-3). The sharp rise in tertiary enrollment in East Asian comparator countries during the 1990s was not evident in Sri Lanka. In China and Vietnam, tertiary enrollment rates that were initially lower than Sri Lanka's in 1990 grew rapidly enough to surpass Sri Lanka's in 2002. Tertiary enrollments rates are the lowest for the poor in Sri Lanka; only 2 percent among the lowest income quintile. Conversely, the poverty incidence among graduates in Sri Lanka is only about 2 percent (estimated fromHIES 2002). Foreign direct investment and export industry 25. The inflow of foreign direct investment (FDI) and exports and the technology transfers that come with it can raise the productivity of domestic industries, enhance their competitiveness, and amplify overall economic growth. Foreign direct investment in Sri Lanka has been significantly lower than that of China, Malaysia, Thailand, and Vietnam since the 1990s (Table 1-4). 26. Studies on the East Asian "miracle" stress the role o f exporting industries in transferring technology to these countries (World Bank, 1993). Sri Lanka's export-to-GDP ratio i s comparable to China and Korea, and much higher than that of South Asian countries. However, it remains far behind Thailand, Malaysia, and Vietnam (see Annex 1, Figure A-1.2). Growth in exports as a share o f GDP since 1990 has also been lower in Sri Lanka than in almost all the East Asian countries. 6 Restrictive labor regulations 27. Some o f Sri Lanka's labor regulations make dismissing formal sector workers overly difficult and costly, and this has constrained private investments and inturnjob creation (World Bank, Doing Business in 2006). After Sierra Leone, Sri Lanka is the most expensive place in the world to dismiss permanent workers in terms o f weeks o f wages per employee, nearly double of that of any o f the East Asian countries. Alternatively Sri Lanka i s less rigdinother dimensions o f the labor market, with almost no restrictions in hiring a new worker and only some degree o f restrictions on changing workmg hours. The average of all three indices ranks Sri Lanka near the bottom in the comparator group, slightly less ngid than Vietnam, almost the same as Korea and muchmore rigidthan China, Malaysia, and Thailand (see Annex 1, Table A-1.4). 28. These riBdities in the Sri Lankan labor market are likely to impede growth and employment generation in the formal sector, which has direct implications for poverty and inequality. In developing countries, where insurance markets and publicly provided safety nets are imperfect, protection for workers by law can be an important social safety net. Overprotecting workers, however, as is the case in Sri Lanka, is likely to stifle job creation in the formal sector and push workers into the informal sector, where earnings as well asjob security are much lower. Such "informalization" of employment i s therefore associated with poverty; it drives a wedge between the relatively privileged and protected workers in the formal sector and a vast informal sector labor force, which contributes to higher inequality. Evidence presented inchapter 3 depicts the vast difference between these two groups: the presence o f a formal sector worker in a , household reduces the likelihood of being poor in Sri Lanka by more than 8 percent, while that o f an informal sector worker increases itby more than 6 percent (from Annex 3, Table A). Fiscal constraints 29. Sustainable fiscal performance i s a cornerstone for any viable growth strategy. In 2003, the fiscal deficit in Sri Lanka exceeded 8 Table 1-5: Burden of percent o f GDP and the public debt stood at over 105 percent o f GDP. interest payments, 2003 ercent of revenue) Large interest payments significantly restrict productive spending. The Sri Lanka 4 5 3 ratio of interest payments to total revenues i s around 45 percent in Sri Korea, Rep 5.1 Lanka, which is many times higher than that o f its East Asian Malaysia 10.5 comparators (Table 1-5). Thailand 5.8 30. An internationalcomparison o fnominal andreal lendinginterest rates for prime customers suggests the nsk o f a high fiscal deficit crowding-out private sector investment, which can hurt Sri Lanka's long-term economic development (Annex 1, Table A-1.5). On the positive side, the burden of large interest payments has had little impact on public expenditure for health and education as a percentage o f total public expenditure, which remained almost unchanged since 1990. The ratio of public education and health expenditures to GDP is, however, significantly lower than in Korea, Malaysia and Thailand (Annex 1, Table A-1.6); and the current fiscal situation is a deterrent to achieving any further increases inthese ratios. 1.4. Concludingremarks 31. This chapter has reviewed Sri Lanka's progress toward achieving poverty reduction, particularly in comparison with fast-growing countries in East Asia that have managed to reduce poverty rapidly. The evidence suggests that some o f the constraints Sri Lanka faces are shared by these countries, albeit invarying degrees. There also appears to be a combination o f constraining factors unique to Sri Lanka that limit growth and poverty reduction. 7 32. Economic growth in Sri Lanka has been stymiedby a combination o f factors, includingthe lack o f urban centers other than Colombo, inadequate infrastructure, the adverse impact o f the civil conflict on investment climate and size of public debt, a stagnant agncultural sector, lack of advanced shlls inthe labor force, and inefficiencies inthe labor market. At the same time, and in spite o f these obstacles, the Sri Lankan economy expanded by 45 percent in terms of per capita GDP and 30 percent in terms of per capita consumption during the past decade. The growth, however, has not translated into poverty reduction, primarily because it has been concentrated around Colombo and the neighboring districts, which has resulted inwidening inequality across regions and sectors. The uneven nature o f economic progress appears linked to stagnation in agriculture and differential access to critical infrastructure and markets. The report takes an in- depth look at the extent, nature, and causes o f wideninginequality. 33. Subsequent chapters present evidence on trends and patterns in intra- and inter-regional inequality, using household surveys and the population census, to identify interventions that will make economic development more inclusive. Chapter 2 analyzes the relationship between trends and patterns in poverty, inequality, and growth over the past decade. Chapter 3 focuses on understanding the nature and correlates of spatial distribution o f poverty, with the aid o f poverty maps and geo-referenced information on access to infrastructure, markets, and human capital accumulation. 34. Chapter 4 analyzes domestic migration and how people respond to the lack o f economic opportunities in remote, rural areas and estates, and the implications for future growth and poverty reduction. It also looks at the significant economic costs of unequal patterns of regional development, including the high concentration of population in Colombo urban area. Mitigating such over-concentration, while expandingeconomic opportunities to potential migrants from poor areas, will require the growth o f alternate economic hubs. Chapter 5 examines the human development challenges that limit the potential o f the poor, includingdeficiencies innutrition for children and women and educational attainment, found particularly among the poor inSri Lanka. 35. Revitalizing the agncultural and the rural nonfarm sector will help the poor inrural areas the most. Chapter 6 draws the linksbetween rural farm and nonfarm sectoral issues and poverty, and attempts to identify factors constraining productivity growth in the agricultural sector and limitthe growth ofthe ruralnonfarmsector. 36. Marginalized areas and groups must be drawn into the growth process to reduce poverty and close inter-regional growth gaps. Chapters 7 and 8 focus on two of these marginalized groups: the conflict affected areas and the estate sector. 37. While the civil conflict inSri Lanka has had economy-wide effects, the most direct impact has been felt by the North and East, which have been largely unable to participate in the economic progress inthe rest of the country. Chapter 7 attempts to understand how conflict and its aftermath--even though a ceasefire has held since 2002-1s associated with the extent and nature of poverty on the ground. The estate or plantation population, which has historically sufferedmarginalization i s especially vulnerable. Poverty inthe estates can only be understood in the context of the unique set o f economic, social, and political factors. Since this i s beyond the scope of existing national household surveys, Chapter 8 draws on quantitative and qualitative surveys o f estates, conducted for this report, to identifythe nexus of factors limitingthe economic potential o f estate households. 8 2. Poverty, Inequality,and Vulnerability 1. Poverty reduction in Sri Lanka has been slow during the decade o f 199&9 1 to 2002. This is partly because o f slow growth as described in chapter 1. Much of the growth bypassed the poor, and was geographically limited to Colombo and its neighboring districts-leaving 23 percent o f Sri Lanka's populationliving below the national poverty line. 2. A key challenge for Sri Lanka is both to increase growth and to ensure that growth is inclusive o f lagging regions and sectors o f the economy. This chapter presents past and present national and regonal trends in patterns o f poverty, drawing primarily from household surveys conducted over the last decade. 3. Given the large overlap between the poor and those vulnerable to the effects o f an economic shock, the chapter will also include a brief discussion on the extent and nature o f vulnerability in Sri Lanka and safety net programs to adequately protect them. Of course, the most significant aggregate shock to the country inrecent times has been the tsunami in December 2004 and the vulnerabilities associated with its impacts, 2.1. Consumptionpoverty: trendsandpatterns Poverty headcount Poverty gap (depth) Squaredpoverty gap (severity) 1990- 1995- 1990- 1995- 91 96 2002 91 96 2002 1990-91 1995-96 2002 National 26.1 28.8 22.7 0.056 0.066 0.051 0.018 0.022 0.016 Urbanb 16.3 14.0 7.9 0.037 0.029 0.017 0.013 0.009 0.005 Rural 29.4 30.9 24.7 0.063 0.072 0.056 0.020 0.025 0.018 Estatec 20.5 38.4 30.0 0.033 0.079 0.060 0.009 0.025 0.018 9 decade (all statistically significant changes).' Inthe first half o f the decade, the rise in national poverty was driven entirely by a rise in rural and estate poverty, while urban poverty actually declined. Inthe second half, the reduction in national poverty was driven by a decline in all 3 sectors. Thus while urban poverty has shown a falling trend all through the period, poverty in rural areas and estates spiked in the middle year, which contributed to a smaller net decline in poverty over the decade for these sectors. 6. The one caveat to interpreting the urban and rural poverty estimates i s the classification o f urban and rural areas, which changed between HIES 1990-91 and HIES 1995-96. The Town Council areas considered as urban in HIES 1990-91 were re-classified as rural in the later two rounds.* As a result, the estimated proportion o f urban population fell by around 7 percentage points from HIES 1990-91 to HIES 1995-96 and stayed almost constant from 1995-96 to 2001- 02. While this suggests caution in interpreting the magnitude o f changes in rural and urban poverty from 1990-91 to the later years, this change has minimal implications for the main conclusion that urban poverty has fallen dramatically while rural poverty has been relatively stagnant. The estimatesreportedbelow for provinces and regions, the definitions or boundaries o f which have not changed between surveys, also support this conclusion. Poverty in Western Province (WP) where Colombo i s situated (Figure 2-I), which constitutes a large part o f the urban sector-irrespective of whether pre or post-1995-96 definitions are used-more than ' Note that because the estate sector compnses of a relatively small part of the population, the HIES sample yelds poverty eshmates wth a higher degree of "error "However, thefall in national, urban and rural poverty headcount and the rise in estatepoverty headcount between 1990-91and 2002 were all statistically significant at 5 percent level (Annex 2, FigureA-2 1) T h i s implies that the definition of "urban" in 199Ck91 HIES includedUrban, Municipal and Town Councils, whereas that in 95-96 and 2002 HIES includedUrbanandMunicipal Councilsonly 10 halved between 1990-91 and 2002, while poverty reduction in predominantly rural districts has been minimaL3 7. The most recent national household survey for Sri Lanka, CFSES 2003-04, is an entirely different survey with differences in modules and field methodologies from the HIES, so that consumption estimates from the two surveys are not comparable. Moreover, since the official poverty line was estimated on HIES data, it cannot be applied to data from CFSES without careful adjustments. At the same time, comparisons o f consumption means and inequality measures between two rounds of CFSES-1996-97 and 2002-ffer some indications about the most recent economic status o fhouseholds (see Section 2.2 below). Regional differences inpoverty incidence and trends 8. Sharp differences in poverty incidence across sectors are mirrored by regional patterns (Table 2-2). Poverty incidence was only 11 percent in WP in 2002, and the poorest provinces o f Sabaragamuwa and Uva had headcount poverty rates of around 35 percent. WP accounts for about 33 percent o f the population but only 16percent o fthe poor. Uva and Sabaragamuwa together are home to 18 percent o f the population and 29 percent o f the 9. Most recent consumption data (CFSES 2003-04) indicates per capita consumption expenditure for WP to be more than double that for Figure 2-1: Province and district Uva and Sabaragamuwa, with the other provinces in boundaries between (Annex 2, Figure A-2.3). The differences between North-Central, Central, and Southern provinces are negligible, while Northwest Province has significantly higher consumption. These patterns track the HIES 2002 results and are broadly consistent withpoverty patterns seen inTable 2-2. 10. Poverty incidence varies even more between districts (Table 2-3). In 2002, while Colombo district had a poverty headcount o f 6 percent, 37 percent o f the population o fBadulla and Monaragala in Uva province lived below the poverty line. In 2002, poverty headcount rate in only 5 out o f the 17 districts for which data are available (covering 41 percent o f the population) was at or below the national headcount, indicating that the national poverty numbers mask much higher poverty rates in large parts ofthe country. Colombo district and WP accounted for 13 and 33 percent o f the country's urban population, respectively ,in 2002, and 13 and 30 percent in 199G91. The share of total population by province remained fairly stable over the decade, with the share of Western Province increasing by about 2 percentage points, while those for Southern, Northwest and Sabaragamuwa declined slightly (by 1percentagepoint or less) 11 least poor districts in 1990-91, were also in Figure 2-2: Percent change inheadcount between the top five in poverty reduction between 1990-91 and 2002 by district 1990-91 and 2002. Conversely, Hambantota, Monaragala, Ratnapura and Kegalle, which were among the six poorest districts in 1990-91, had an increase or no g o change in poverty. Kandy and Kalutara -20 I districts are the exceptions to this pattern being among the poorest districts in 1990- I 91 and the top five in terms of poverty reduction. Overall, gaps betweenpoorer and a o e , x , " p = " s 8 & $ $ z 5 9+ g ~ $ E~ $ 3 55 $ 4 less poor districts widened during 1990-91 0 9 2 2 q$ 5 3 5 w P Q 2 2 g r n & : to 2002, and the results in section 2.2 also 111 support this finding. 2 Source-HIES (1990-91 and 2002). 13. Growing regonal differences are also seen from Tab'e 2-4: Share Of each province in GDp consumption expenditures of provinces from CFSES Province 1990 1996 2002 data and provincial GDP trends. Per capita real Western 40.2 43.7 48.1 consumption grew at an average annual rate o f 3.6 North-Central 4.8 4.6 3.9 percent in WP between 1996-97 and 2003-04 Central 12.1 10.0 9.4 compared with 1.2 and -0.4 percent in the Uva and Northwest 11.1 11.3 10.1 Sabaragamuwa, respectively. Northwest Province Southern 9.5 9.0 9'7 was the only anomaly: in spite o f being one of the poorer provinces in 199697, per capita Sabaragamuwa 8.1 9.0 6.9 consumption grew at the highest average rate of 5.5 Uva 8.1 5.1 4.3 percent annually. As WP's share in national GDP Note: ExcludmgNorthem andEasternProvinces increased from 40 percent in 1990 to 48 percent in Source: Dept of National Planning. 2002, the shares o f all other provinces, with the 12 exception o f Southern, declined (Table 2-4). The share o f Uva, the poorest province, in GDP halved dung this period. Uva and Sabaragamuwa together accounted for only 11 percent of the GDP in2002, and 18percent ofthe population. 2.2. Poverty, growth, and inequality 14. The pattern o f poverty reduction during the last decade occurred during a period of reasonable growth in the national average consumption expenditures, particularly during the second halfo f the decade, along with a significant skewing o f the distributiono f cons~rnption.~ A clear trend o frising inequality over time occurred duringperiods of l o w as well as highgrowth. Linking growth and distributionalchangesto povertytrends 15. The relationship between poverty reduction, growth in consumption, and Figure 2-3: Growth-inequality decomposition changes in inequality can be quantified between 1990-91 and 2002 by a growth-inequality decomposition of *' change in poverty headcount. The so- called growth effect between selected years measures the simulated impact o f oo the increase in average per capita consumption on poverty headcount -100 (keeping the distribution unchanged fiom the initial year), while the redistribution effect measures the simulated impact on headcount o f the change in the distribution o f per capita consumption (keeping the Source World Bank staff calculations using HIES For comparable years (HIES 1995-96 and 2002 only, since incomes for 1990-91 are not comparable), the trends and patterns inper capita income inequality are similar to that for per capita expenditure (see Annex 2, Table A-2.2). See Datt and Ravallion (1992) for details. 13 at the full distribution o f per capita consumption for the two years (see Annex 2, Figure A-2.1). Moreover, the cumulative distribution of per capita consumption suggests that no matter where the poverty line i s drawn, the proportion of population below the poverty line was lower in2002 than in 1990-91. 19. Over this period, poverty declined slower in the Table 2-5: Mean per capita consumption rural sector than in the urban sector due to a greater increase ininequality as well as lower growth inrural Consn. areas. Poverty increased in the estate sector due to a quintiles 1990-91 1995-96 2002 rise in inequality as well as negative growth in mean Q1 1,045 991 1,068 per capita consumption. 42 1,499 1,445 1,596 20. The two interim periods-1990-91 to 1995-96 43 1,909 1,881 2,168 and 1995-96 to 2002-are quite different (Table 2-5 Q4 2,489 2,578 3,117 and Table 2-6). National mean per capita consumption Q5 4,871 5.274 7,325 grew by 3 percent fiom 1990-91 to 1995-96, with National 2,363 2,434 3,055 only the top 2 quintiles experiencing growth. Mean Urban 3,168 3,556 4,667 consumption increased by 12 and 6 percent for urban Rural 2,154 2,278 2,865 and rural areas, respectively, and fell for the estate Estate 2.103 1.685 1,985 sector. The Gini coefficient increased by 3 percent for Source: HIES 1990-91, 1995-96,2002 . - the urban sector and 14 percent for the rural sector, and fell for the estates. These numbers are consistent with the rise inrural and estate poverty and the small decline in urban poverty during this period. 21. Incontrast, from 1995-96 to 2002, national mean per capita consumption grew by 26 percent. Consumption o f the top and 4& 1 1990-91 1995-96 2002 quintiles grew by 39 and 21 percent, Urban' 0.37 0.38 0.42 respectively, and that o f the bottom two quintiles by 8-10 percent. Mean per capita Rural* 0.29 0.33 0.39 consumption increased by 31, 26, and 18 Estate 0.22 0.20 0.26 National 0.32 0.35 0.40 percent for urban, rural and estates, respectively. The Gini coefficient also 23. More recent trends from CFS'ES. The HIES Table 2-7: Growthinmean per capita trends from 1995-96 to 2002 are fairly consistent with real consumption, 1996-97 to 2003-04 those from the CFSES, although the two are different %annual %total surveys and the latter measures changesfrom 1996.97 Sector growth rate change to 2003-04 (see CBSL, 2003-04). Over this period, Urban 2.7 20.5 mean per capita expenditure increased by 18 percent Rural 2.4 18.1 for the country, and 21, 19, 4 percent for urban, rural, Estate 0.5 3.6 and estate sectors, respectively (Table 2-7). A sharp All 2.4 18.1 increase in inequality was also seen for all sectors, Source: CFSES Report (2003-04); Table 8.3. 14 using per capita income figures.' Aggregate growth for the bottom income quintile of every sector was 4 percent or less, compared to 25-26 percent for the top quintile (Table 2-8). The Gini for per capita income increased by 9, 8 and 26 percent for urban, rural and estate sectors respectively. 24. Thus, for rural and urban sectors, both surveys show rapid growth in average consumption, with the gains heavily skewed toward the upper end o f the distribution. For the estate sector, however, the CFSES rounds show negligible consumption growth and a large increase in income inequality; the HIES finds a sizeable increase in mean consumption and some increase in inequality, with a net result o f sizeable reduction inpoverty.* 25. Growth incidence czcwes. While trends in Gini coefficients and mean consumption levels by quintiles presented above hint at unequal distribution o f growth across the population, a more precise picture emerges from using growth-incidence curves (GICs) (Annex 2, Figure A-2.4). These confirm that while growth in consumption had a poverty- reducing effect between 1990-91 and 2002 in both rural and urban sectors, the benefits accrued disproportionately among the better-off. 26. Between 199&91 and 2002, growth in per capita consumption was negligble (below 1 percent) for the bottom 40 percent and sizeable for the top 20 percent of the rural population. A similar pattern is seen for the period between 1995-96 and 2002, although the gains in absolute terms were higher for all groups, consistent with a larger reduction in rural poverty during this period. The GIC for per capita income for the same period also shows similar results (Annex 2, Figure A-2.5). Between 1990-91 and 1995-96, consumptionfell for the lower 50 percent o f the rural consumption distribution, which explains the increase inrural poverty duringthis period. 27. The GICs for the urban sector show similar skewed growth in consumption. For the decade, the shape o f the urban GIC closely resembles that o f the rural GIC, albeit with somewhat higher levels o f per capita consumption growth, consistent with urban poverty declining more than rural poverty during t h s period. A notable difference between urban and rural areas is seen only for 1995-96 to 2002: the gains for those near the top o f the urban distribution appear to be especially large incomparison to the rest o f the distribution o f urban as well as rural population. 28. The GICs thus tell a story of highly skewed growth in per capita consumption over the decade, for urban and rural areas alike, a pattern that was even more pronouncedwhen one looks at the more recent subperiod o f 1995-96 to 2002. The decline in poverty incidence by 6 percentage points for urban and rural areas alike from 1995-96 to 2002 i s attributable entirely due to anupward shift inthe distribution, rather than any redistributiontowards the less well-off. Identifyingthe sourceof rise in inequa1it.y: inequality decompositions 29. To what extent is the rise in overall inequality explained by increase ininequalitybetween regions (districts or provinces) or sectors (among urban, rural and estate sectors), as opposed to Mean per capita consumption by quintiles o f consumption are not provided by the CFSES report This discrepancy may be explained bythe high sensitivity o f estatepoverty to small fluctuations (see Figure 2-5), the higher standard errors in all estate sector estimates in both surveys (due to the small size o f the estate sample), or the fact that the HIES estate sample for 1995-96 was not ex anb designed to be representative ofthe sector The GIC maps the average annual rate of growth o f real per capita consumption between the relevant years for all centiles (1 percent quantile) ofthe consumption distnbution (for details see Ravalhon and Chen, 2003) 15 an increase in inequality within regions and sectors? A Theil inequality index is used to decompose the national index into (a) interdistrict and withm-district indices; and (b) intersectoral and within-sector indices (Table 2-9). 30. The decomposition shows that while national and within-district indices rose by 57 and 52 percent, respectively, fiom 1990-91 to 2002, interdistrict inequality rose by as much as 112 percent. The intersectoral Percent 1990- increase index, on the other hand, increased less in 91 2002 between years percentage terms (52 percent) than did the National 16.8 26.4 57 ~ within-sector index (57 percent) and the Within-district 15.4 23.3 52 national index." The within-district and Inter-district 1.5 3.1 112 within-sector inequality indices are much Within-sector 15.5 24.4 57 larger than the interdistrict or intersectoral Inter-sector 1.3 2.0 52 indices for both years. This is expected, since the extent of variation within a district or a sector is much larger. 31. The important insights fiom this exercise are, first, the decompositions show substantial increases ininequality-by 50 percent or more-within andbetween districts (that are proxies for regions) as well as within and between sectors. Second, while inequality within districts or sectors continues to be much larger in magnitude than between sectors or districts, the percentage increase ininequality between districts was by far the highest. 32. These decompositions are aggregated at a geographical level (district or sectors) that are too large and heterogeneous to capture adequately the spatial dimensions o f poverty and inequality. These can be better analyzed using techniques that are able to estimate poverty and its correlates for smaller areas, like DS divisions, which chapter 3 will attempt to do. 2.3. Attaining theMDG of halvingpoverty: howgrowth and distributional changesmatter 33. What are Sri Lanka's prospects of attaining the MDG o f halving poverty by 2015 for reasonable assumptions o f growth and inequality changes? There i s no doubt that hgher economic growth will lead to greater poverty reduction, but what i s the responsiveness o f poverty reduction to growth? 34. Estimating growth elasticity and the impact o f inequality i s based on the approach proposed by Bourguignon (2003)." If inequality measured by consumption Gini remains constant at the 2002 level, elasticity of poverty reduction i s found to be 2.1, i.e., a 1-percent increase in per capita consumption expenditure reduces poverty headcount ratio by 2.1 percent. Since growth in GDP is different from that inconsumption, this elasticity (withrespect to growthinconsumption) implies that 1percent increase in GDPper capita will reduce poverty incidence by 1.6 percent." Butifthe consumptionGini grows at an average annual rateof2 percent, as it didbetween 1990- 91 and 2002, growthelasticity falls from 1.6to 0.9. loThe share o f the estate sector intotal population i s low, so most of intersectoral inequality is explained by differences between urbanand rural areas I'Among numerous methods for estimating growth elasticity, this approach was chosen because it has been empincally tested and allows for the impact o f Inequalityto be easily quantified (Annex 2, section I) This is calculated using a ratio o f GDP growth to Consumption growth for Sn Lanka obtained by comparing the growth Inconsumption fromHIESbetween 1990-91 to 2002 with the GDP growth from 1991 to 2002. 16 35. These elasticity estimates are used to generate Table 2-10: Projected poverty headcount in poverty projections for 2015 (Table 2-10). IfGDP 2015 fuercentl continues to grow at the rate o f the last two years, Given With increase the population growth rate stays at the current 1Gini 1 inGini level, and the consumption Gini i s unchanged at Assumed GDP the level of 2002, the poverty headcount will be growth rate: 5.7% more than halved to 8.2 percent by 2015. On the Assumed GDP other hand, if the consumption Gini increases at growth rate: 10% the average annual rate o f the last decade, the Note: (a) Populationgrowthrate is assumedto be 1.2 poverty headcount ratio in 2015 will be 14.8 percent. (b) Annual average growth of consumption percent, well above half the poverty headcount Gini is assumedto bethe ratebetween 199&9 1 and ratio in 2002. hthis scenario, Sri Lanka will need 2002. (c) 5.7 percentis the averageofGDP growth ratesin2003 and2004 (CBSL Annual Report2004). to grow at an annual average rate of around 10 Source: World Bank staff estimates basedon HIES percent to achieve the MDG target o f halving 2002, usingthe methodby Bourguignon(2003). poverty by 2015. 36. Like most such elasticity measures, these are calculated under certainrestrictive assumptions about the current distribution o f consumption (see Annex 2, section I). Although these simulations are illustrative, they still indicate that strong economic growth i s probably not enough to attain the MDG poverty target unless the growth process becomes much more inclusive o f lagging regions and populations. 2.4. Howpoverty trends relate tosectoralpatterns of growth 37. Macroeconomic data are broadly consistent with the growth inaverage levels of consumption observed in household survey data. Real per capita GDP increased by 41 percent cumulatively during 1991-2002, comparable to the growth inper capita real consumption o f 29 percent over the same period. For the interim periods, however, per capita GDP trends are at odds with consumption trends from micro-data. Real per capita GDP grew by 21 percent during 1991-96, and 16percent during 1996-2002, and this translates to an annual average growth o f 3.9 and 2.5 percent, respectively. Mean per capita consumption from survey data, on the other hand, grew by 3 percent fi-om 1990-91 to 1995-96 and 26 percent from 1995-96 to 2002, but such inconsistencies between national accounts and household survey estimates are common in many countries. Figure 2-4: Sectoral outputs and contribution to GDP I I Contributionto GDP(at constant1996prices)bySector Real per capita GDP and sectoral outputs BPgricuiture Q Industry EJSemces Pgricuiture industry -%- Semces Source CBSL, Annual Reports(multiple years up to 2005) 38. Irrespective o f whether one invests more faith in macroeconomic or household survey data, the important question is: what explains the rise in inequality that tempered the response o f poverty reduction to growth? Sectoral data is a useful starting point to see why predominantly 17 rural districtshegions lagged significantly behind urban regions. It shows that between 1991 and 2004, the share o f agriculture in GDP declined sharply from 28 to 19 percent, while that o f industry and services increased (Figure 2-4). A falling share o f agnculture in national output is a part o fthe common process of development, but in Sri Lanka agnculture is stagnant. Agricultural output per capita remained almost unchanged and in fact regstered negative growth during certain years, which i s consistent with the slow and uneven reduction inrural poverty. 39. GDP increased by almost 4 percent During 1991-1996, while per capita annually, agricultural output per capita actually shrunk by an annual average o f - 0.2 percent. T h s may have led to the spike 0.4 2.6 3.3 in m a l poverty observed in 1995-96. Between 1995-96 and 2002, agnculture did slightly better with output growing at an annual average rate of 0.4 percent per capita, and rural poverty reduced during this period (Table 2-11). Over the decade as well as during both subperiods, healthy output growth in industry and services (Table 2-11)i s consistent with the sustainedpoverty reduction experienced by urban areas. 40. In spite o f the dynamic growth o f other sectors, agnculture remains important for employment. Although agnculture's share intotal employment fell from 43 percent in 1991 to 34 percent in 2002, the number o f people employed by agriculture actually increased by 3.5 percent and agnculture employed more than one-third o f the total workforce (and likely much higher in rural areas) in2002. Stagnant agncultural output necessarily implies that a sizeable population in rural areas dependent on agriculture would have had minimal income growth over the past decade. 41. Chapter 6 will explore how lack o f growth in agriculture translates into rural wages and incomes, and how these impacts are distributed across the population. Unpachng the story of rural poverty will also require examining outcomes and challenges inthe rural nonfarm sector. 2.5 Nexus betweenpoverty and vulnerability toeconomicshocks disProPodionatelY Severe due to their Figure 2-5: Distribution of real per capita monthly limited access to insurance and safety nets. Moreover, vulnerability is also high for those just above the poverty line, since even a small shock can push them into poverty. 43. Consumption distribution and vulnerability to shocks. While the static poverty estimates highlighted in the chapter are unable to measure dynamic changes, they offer clues about the extent o f vulnerability that prevails in Sri Lanka. A small shock to consumption can cause large jumps in poverty incidence when a high concentration of the population hovers Source: World Bank staff estimation based on HIES 1990-91 around the poverty line, which is what and2002. 18 we see for Sri Lanka, although it fell somewhat from 1990-91 to 2002 (Figure 2-5). The concentration i s even higher inthe estate sector and increased over the decade. 44. Figure 2-6 shows that ifmonthly per capita : consumption were to fall by 10 percent o f the {Figure 2-6: Projected increasesin poverty poverty line (less than US$1.50) for every headcount due to economic shocks household, the poverty headcount will increase by 6 percentage points nationally and 10 percentage points for the estates. If monthly consumption declines by 20 percent, national and estate poverty headcount would increase by almost I O and 20 percentage points, respectively. Furthermore, the impact on poverty headcount rates could be even higher ifan adverse economic shock were to affect the lower quintiles 10% 20% 30% 40% 50% disproportionately, as is often the case. Reductionin monthly per capita consumption 45. According to a recent World Bank report expendiitureas %of the official poverty line (draft) the major individual risks faced by Sri ,Source: Staffestimationbasedon HIES2002. death of a family member and unemployment; and the main communitywide shocks include drought, crop failure, and other natural disasters, the most recent and disastrous being the ts~nami.'~Static poverty measures suggest significant increase inpoverty during 1995-96, a year o f severe and widespread drought. Certainpoor districts appear to be especially vulnerable. Large increases in poverty headcount rates from 1990-91 to 1995-96 occurred in distncts that experienced severe drought (Monaragala, Ratnapura, Matale and Puttalam-see Table 2-3). 46. Safety net or social welfare programs have a critical role to play inprotecting consumption inthe aftermathofa shock. While SriLankahas avariety of social safetynetprograms, many of these are inefficient interms o f achieving their stated objectives. The current landscape of socialwelfare programs 47. To be effective inreducing vulnerability among the poor, social welfare programs needto target assistance to the needy, ensuring a minimumlevel of consumption, particularly inresponse to income shocks. The typical clientele for such assistance would consist o f the poor and the most vulnerable among them.14 When measured against these broad objectives, the social welfare sector in Sri Lanka presents a decidedly mixed picture. A long history o f countrywide programs for the poor and vulnerable have created an enabling environment for such programs, but have performed far below their potential due to inefficient targeting and inadequate coordinationacross programs. 48. A multitude o f overlapping programs administered by different ministries constitutes the social welfare sector in Sri Lanka. Total expenditure for welfare programs, which amounts to almost 1 percent o f GDP in 2003, was distributed among programs like Samurdhi consumption grants to the poor; social security for disabled soldiers; social welfare for Internally Displaced People, Relief and Recovery programs; mid-day meals for children at schools; social assistance to vulnerable groups like elderly, women-headed households and disabled people; and Triposha (a nutritional supplement) for mothers and children. The Samurdhi transfers program is the most significant targeted welfare program intended for poor families, with an expenditure o f Rs. 9 billion ($ 90 million) inFY2005, which comes to about 0.4percent ofGDP.The effectiveness o f 13WorldBank (2006~). 14See WorldBank (2006~)for a detaileddiscussion on vulnerablegroups. 19 spending on social welfare by this program i s seriously undermined by poor targeting o f beneficiaries. 49. Samurdhi transfer program: A number o f studies, including an evaluation conducted by the World Bank point to large-scale errors in targeting of Samurdhi transfers. The program excludes about 40 percent of households in the poorest consumption quintile, while 44 percent of the total budget is spent on households from the top three quintiles. It covers close to 45 percent o f the population, with the result that the benefits are spread too thinly to have much impact on individual households (Box 2-2). Empirical evidence suggests targeting errors are systematic, with some groups (e.g., poor households inurban neighborhoods and estates) being less likely to receive Samurdhi than others (households intraditional villages). Qualitative results suggest that political factors, including party affiliation or voting preferences appear to influence allocation o f Samurdhi grants.' 50. Poor targeting is also evident from large discrepancies between the distribution of Samurdhi transfer budget with that o f poor population across districts (see Annex 2, Figure A- 2.6).16 A recent poverty mapping exercise has also revealed wide divergence between the D S division level poverty estimates and the distribution of Samurdhi beneficiaries across DS divisions. 51. The errors o f targeting relate to the way the program defines eligibility and selects households. The criteria for selection of beneficiaries are a combination o f family size and income applied by the program officers inthe field. Since income i s generally unobservable and difficult to corroborate, this results in a largely subjective selection process. The absence of any process for community validation, redressing grievance or monitoring of entry and exit also results ina process that is nontransparent and vulnerable to politicization." 52. The need to reform the targeting o f Samurdhi transfers has been a highpriority for various governments. A Welfare Benefit Act was enacted by the Parliament in 2002 to rationalize and improve the selection o f beneficiaries for all state-funded welfare programs. A Welfare Benefits Board (WBB) was set up in accordance with this Act and started developing a formula-based system for selection of Samurdhi beneficiaries that allows for greater objectivity and transparency, complemented by participation o f communities in validation and redressing o f grievances (see Box 2-3). Such a targeting system was implemented in two districts o f the conflict-affected North where Samurdhi was introduced for the first time ever in early 2005. lSSee, Glinskaya(2000) for the full analysis; also see World Bank(2002) andWorld Bank (2006c, draft). l6 The share of Samurdhi transfer budget received by each district in 2005 reveals that a few districts (Kurunegala, Gampaha, Anuradhapura and Colombo) receive much higher Samurdhi allocation than warranted by their shares in total number ofpoor, while others (such as Badulla, KegalleandNuwaraEliya) appear to receive significantly less (see Annex 2, FigureA-2.6 for details). l7World Bank(2006d) 20 However, extending this effort to the rest o f the country has proved difficult due to the challenges inherent in changing an entrenched system. In recent times, the government has publicly reiterated the need to target Samurdhi only to the poor; and the Ministry o f Samurdhi is considering various options, including ways to better involve communities in the targeting process, to achieve this objective over the next year. Box 2-3: An option to improve Samurdhitargeting- a formula-based approach 1 The Welfare Benefit Roard developed a proxy-means test fornmla (PMTF) to select Samurdhi beneficiaries based on an analysis conductedjointly by local statisticians and World Bank staff usinghousehold survey data. I l i e exercise correlated household or individual characteristics with welfare levels in an algorithm to proxy household income or welfare. The algorithm allows ranking of households using demographic characteristics, attributes of dwelling units, and ownership of durable assets that are more observable than direct measures of welfare like consumption or income. The YM'TF is to he compIemented by a strong community-based process for independent validation of bcnetlciasy lists and an appeals process to 53. Other key weZjizreprograms. A number o f programs other than Samurdhi target transfers to vulnerable groups. The largest among them i s the Public Assistance program administered by different provincial governments and targeted toward specific vulnerable populations: poor among the elderly and disabled, farmlies without breadwinners, destitute women, and orphans. In 2005, 365,000 families received an average monthly grant o f around Rs. 135, much smaller than the average Samurdhi grant. Inconflict-affectedareas, a separate social welfare programprovides dryration and cash assistance to internally displacedpeople. 54. Triposha, administered by the Ministry of Health, is the most significant targeted nutrition supplement program for pregnant and lactatingwomen and children (age 6-59 months) from poor families (580,000 beneficiaries in 2005). Other nutrition programs include a current pilot by the Ministries o f Samurdhi and Health that provides a basket of food items to pregnant and lactating mothers, an infant milk subsidy provided by the Samurdhi ministry, and a program by the Ministryof Healthproviding a nutrihonal supplement different from Triposha insome areas. In addition, a mid-day meal scheme for chldren IS being implemented in select schools, with plans for rapid and expanded coverage in2006; and there are special programs like food-for-educahon and nutrition for mothers and children operating in conflict-affectedareas. 55. For most programs other than Samurdhi, little is known about targeting effectiveness or impact on beneficiaries. A few studies, based on observations from surveys over the years, provide indirect evidence that the long-standing Triposha program has contributed substantially to the improvements in nutritional status among women and children, but much remains to be done to analyze the impact of Triposha as well as relatively new programs like mid-day meals, to inform any plans to improve or expand such programs. 21 56. Poor coordination among ministries and levels o f government often results in overlaps between the objectives and beneficianes o f different social welfare programs.I8 For example, the Provincial Public Assistance program, the smaller programs runby the Central Ministryo f Social Welfare and Samurdhi transfers are targeted to groups that are likely to overlap significantly. Rationalizingthe objectives and target groups o fprograms and integratingthis information across programs that serve near-identical objectives are needed to improve efficiency and reduce administrative and fixed-costs o f delivery. Even when integration is not possible, better coordination among programs is desirable to identzfi beneficiaries, reduce gaps and avoid duplication in coverage. Such coordination can also enhance the effectiveness of programs that are complementary. For example, when two programs serve different nutritional objectives, coordination o f the selection processes will better ensure that assistance is appropriately matched with specific nutritionneeds withinthe two target groups. 57. Effectiveness o f welfare programs can thus be enhanced by better coordination between programs, in terms o f their stated objectives, target groups, and selection o f beneficiaries. A recent effort by the government to take a sectoral approach inbudget preparation is an important step toward rationalizingsocial welfare objectives and allocations. 58. While Samurdhi and other social welfare programs are key elements in managmg ongoing risks o f the poor, large-scale disasters require more expanded assistance to cope with the immediate and longer-tern vulnerabilities. After the tsunami, the government and its development partners assembled important lessons for how social protection strategies can respond to future emergencies and facilitate the transitiontowardrecovery. Impact of tsunami and the role of social assistance 59. On December 26, 2004, Sri Lanka suffered was the worst natural disaster to confront the country inmodem times. The wave devastated two-thirds of the island's coastline spread over 13 district^.'^ Over a million people were affected: 35,000 dead, 20,000 injured, over one-half million residents displaced, and 150,000 workers without a livelihood. Almost 100,000 houses were at least partially destroyed and the total damage estimate was about $ 1billion.20 60. External shocks and natural disasters tend to affect the poor disproportionately. The poor typically have fewer assets, resources, and networks at their disposal to help them cope with initial shock and the transition to recovery afterwards. Loss o f assets and sources of income for the poor are also much harder to replace, and can leave them more vulnerable to future shocks, both at the household and the community level. 61. Geographicalpattern of damage. The tsunami affected coastal populations inthe Eastern, Southern, Western, Northern, and North Western provinces. Housing damage-a key indicator o f impact-affected Eastern, Southern, Northern and Western provinces in descending order o f damage suffered. Number of deaths and displaced persons, other key strong indicators o f the extent of impact also show similar patterns (Annex 2, Figure A-2.7).21 62. Note that both HIES and CFSESdata used inthis report arepre-tsunami, and therefore do not reflect changes that may have occurred due to the impact of the disaster. The tsunami i s particularly likely to have affected economic conditions in the Southern, Northern and Eastern provinces (for more discussion on the impact o f the tsunami on Northern and Eastern provinces, see Chapter 7). Although poverty data is not available for conflict-affected areas, all available These include Ministries o f Education, Health, Nation-building, Samurdhi and Social Welfare, in addition to Provincial departments. *' l 9Government o f Sri Lanka and Development Partners (December 2005). ADB, JBIC andWT3 (2005) DCS, Census o f Tsunami Affected Areas January 25,2005. 22 indicators indicate that Eastern Province, and the districts of Ampara and Batticaloa within the province, was among the lagging areas o f the countryprior to the tsunami. Inaddition, before the tsunami the three distncts affected in Southern Province had poverty headcount rates 3 to 9 percentage points higher than the country average (Table 2-3). Given the extent of damages in districts that were already lagging or vulnerable, it is likely poverty will increase in the affected areas at least in the short run. While no systematic post-tsunami household data is available to measure the short-term impact, the next round o f the HIES beginning in July 2006 is expected to provide information on the status of these areas and the extent o frecovery that has occurred. 63. Social assistance to reduce vulnerability. In the immediate aftermath o f disasters that disrupt local economies, a primary concern IS toprotect the consumption of families, until alternative means o flivelihood are reestablished. Within 3 months o f the disaster the government provided an unconditional grant of $50 per affected families for a total of four payments.'* An interim assessment conducted after the first payment indicated that the program was successful in covering a very high percentage o f the affected population, as well as people who had been minimally affected by the disaster (mis-targeting). Little more than one year after the disaster, there are encouraging signs o f longer-term recovery in sources of livelihood, including critical sectors like fishing and tourism. In keeping with its objective, the cash grant program also disbursed its last payment inDecember 2005. 64. The emergency cash assistance program highlights the value o f a flexible and timely approach to mitigate temporary vulnerabilities caused by a disaster. The post-tsunami program in Sri Lanka offers valuable lessons for similar (or smaller-scale) shocks at home and elsewhere. While a full stocktakmg of the program's impact and lessons must await a complete assessment, the interim assessment underscored the need to define clear eligibility criteria from the very beginning. The pre-existence o f beneficiay lists would minimize mis-targeting and help implementers inthe field. 65. A long-term recovery strategy to provide continued assistance to those inneed o f longer support i s best accomplished by integrating them into the country's welfare system. Inthe case o f Sri Lanka, large programs like Samurdhi are good candidates. And the challenge o f the social welfare system in Sri Lanka will be to become flexible enough to meet the needs o f those who have become vulnerable for the long-term because o fthe tsunami. 22This programwas supported by the World Bank's Tsunami EmergencyRecoveryProject. 23 3. A Profile of Poor Householdsand LaggingRegions 1. This chapter begns to unravel the complex characteristics associated with income and consumption poverty at the household level and the observed spatial patterns o f poverty. Rising inequality in Sri Lanka is manifested in widening consumption gaps between regions, provinces, and districts, and the uneven distribution of consumption within geographical areas. This unequal spatial development excludes large sections o f the population from the growthprocess. 2. Poverty incidence i s often strongly associated with attributes o f individuals andhouseholds such as demographics, education, ethnicity, land ownership, occupation and employment status. Beyond individual household attributes, variations in employment opportunities and infrastructure, such as roads, among districts and sectors also affect growth. The exact combination o f factors that keep a household below the poverty line is unknown, but partial analysis ofthe key correlates can suggest policy interventionsto reduce poverty. 3. The likelihood of poverty at the individual, household, and spatial level are explored using a multivariate regression o f the probability o f a household being poor. The regressions measure the effects of key correlates o f poverty when they are considered together, which can be quite different from simple correlations o fthe same vanables taken individually. The probit regressions (Table A, Annex 3), however, must be carefully interpreted since data limitations circumscribe the range of potential factors that can be taken into account, and the direction o f cause and effect i s sometimes impossible to determine. The regression results show that a variety of household- specific and spatial (district-specific) characteristics are significantly associated with the likelihood o f a household beingpoor. 3.1. Household-specificfactors associatedwithpoverty 4. This section focuses on the key household and individual characteristics associated with poverty. The probit regressions in columns 1and 2 o f Table A, Annex 3 provide a useful starting point for the analysis. The coefficients on household-specific factors reveal that a household is more likely to be poor when it i s located in the rural or estate sector, has at least one member working in the informal sector, includes an unemployed youth as a member, and where the household size is large. The presence o f at least one child is also associated with a higher probability o f being poor. Poverty is also more likely if the household head i s employed as an agricultural wage worker or i s inactive in the labor market or unemployed, although the correlation with unemployment is not significant. Households with a lower probability o f being poor have at least one member o f the household workmg inthe formal sector or a family member working abroad. Higher educational attainment of the household head lowers the probability o f beingpoor significantly. 5. Using these regression results as a framework, correlates o f poverty are considered independently as a first step in identifying how these factors constrain the ability o f poor households to increase their economic potential. These factors include labor market indicators like educational attainment o f household heads (human capital), underemployment and occupational status, and unemployment, While this makes the analysis tractable, it i s important to recognize that many of these factors are inter-related: educational attainment at least partly explains employment status; household demographics can play a role in determining education and employment opportunities; and land ownership may affect a household's access to assets critical for income generation. 6. In the analysis, poverty correlates identified as important by the multivariate probit are combined with relevant evidence from other sources to overcome some o f the data limitations imposed by the HIES. Land ownership, for example-an important correlate o f poverty in many 24 countries-is not included in the regression since HIES data does not provide household-level information that can be linked with poverty status. The same applies to information on access to safe water and sanitation and housing-other important correlates o f poverty-that i s available from the CFSES and the Population Census. Household demographicsand poverty 7. The probit regressions show clearly that larger households and especially those with children are more likely to be poorer than the average, while the elderly are likely to belong to households that are less poor than the average. A household size of six and above is associated with a 24 percent marginal increase inthe probability o f beingpoor, and the presence o f at least one child increases the probability by 5 percent. The presence of an elderly member reduces the probability o fbeing poor by 1percent. 8. The apparent strong correlations between household size and composition and poverty incidence must be considered inlight of the definition of poverty used in Sri Lanka, which does not take into account economies o f scale and equivalence scales inconsumption. Such effects are hard to quantify in a universally acceptable form, and are therefore excluded following the practiceinmany countries where consumption-based measures o f poverty are used. 9. Incorporating scale effects in consumption for Sri Lanka would not significantly impact poverty profiles as far as nondemographic correlates are concerned. But correlations with variables directly linked to household size and composition are a different matter. Ifeconomies of scale could be incorporated into the poverty measure, the effect o f household size and composition-namely the presence of children or the elderly-on the probability o f being poor would likely be different from what is described above. For example, larger households with more children would turnout to be less poor than what their per capita consumptionsuggests. 10. Finally, the presence o f a family member abroad has a significant marginal effect (8 percent) in reducing the likelihood of a household being poor. This indicates the critical role o f migration to foreign countries, and presumably remittances received from migrants, in determining the economic status o f a household. How employment status of householdmembers Linksto poverty 11. Poverty, underemployment, andunemployment. Evidence from many developing countries has shown that members o f poor households are much more likely to be underemployed; i.e., engaged in low-productivity activities in the form o f casual wage employment, than unemployed. Table 3-1: Distributionof employment status Indeed underemployment i s far more common of householdheads 1 among household heads in Sri Lanka than (percent) outright unemployment (Table 3-1). Infact Table 1990-91 1995-96 2002 3-2 indicates that there is no clear difference in Employed 75.7 75.4 '24: poverty incidence by employment status of the Unemployed 1.1 1.2 household head. The probit regression suggests Inactive 23.2 23.4 21.0 that if the household head is unemployed, 1 Source: HIES 1990/91,1995196,and2002. everything else being the same, the increase in the probability o f being poor is statistically insignificant. However, poorer households are more likely to have household heads who do not participate inthe labor market, and this coefficient i s significant. 25 12. While there is no direct evidence linhng Table 3-2: Poverty headcount rates by underemployment with poverty, the statistics on employment status of household heads underemployment in the CFSES report 2003/04--defined as employment with duration 1990-91 1995-96 of less than 35 hours per week-sheds light on Employed its prevalence and likely link to poverty.' Unemployed Underemployment i s usually found among the Inactive 25.6 less-educatedworkers in elementary occupations Source: HIES 1990/91, 1995196, and2002. (employment in informal sector, including population, unemployment among youth i s Table 3-3: Poverty headcount by employment clearly associated with poverty. The poverty status of youth (10 to 20 years of age) headcount rate for households with unemployed (Percent) youth (10-20 years of age) was 6 percentage 1990-91 1995-96 2002 points or more higher than households with Employed 28.4 30.8 25.0 employedyouth for the three surveys duringthe UnempIoyed 35.2 36.9 34.6 decade (Table 3-3). The presence o f an Inactive 29.4 33.4 27.7 unemployed youth in the household i s Source: HIES 1990191, I995/96,and2002. As noted in the CFSES report, the definition of underemployment does not take into account that even a person worhng for morethan 35 hoursper week can be underempfoyedifhe/sheis overqualifiedfor the job. Further analysis will benecessary to get abetter pictureofunderemployment and its associationwth poverty.A new roundof LFS wll includevariousmeasuresofunderemployment, which will provideopportunityfor such analysis. Rama(2003); HeltbergandVodopivec (2004) Nanayakkara(2004) 26 also associated with low-wage employment in the 3-4: headcount by agncultural sector. This i s expected, given this sector's industry where household head is stagnation co-existing with almost unchanged levels o f employed, 2002 ercent) employment (chapter 2). According to the probit regression, the marginal effect o f the household head fgculrureEishing 40.4 being employed as an agricultural wage worker on the ManufacturingJConstruction 24.7 probability o f being poor is positive and significant (4 Service 9.6 percent)' 3-4 suggeststhat the povertyrate among All households 22.7 household heads workmg inthe agncultural sector as paid . Note:Including paidemp]oyment only, source.HIES 2002, 20. Higher levels o f educational Table 3-5: Poverty headcount ratios by educational attainment expand economic attainment of household heads opportunities for households both in (Dercent) salaried employment and self- Education level 1990-91 1995-96 2002 empl~yment.~Even with high primary No schooling 38.1 45.3 45.5 and secondary enrollment rates for the Upto Grade 5 32.7 38.0 33.5 country as a whole, the relationship Grade 6-8 23.9 29.5 22.3 between poverty and the level of Grade 9-below degree 11.1 14.0 10.3 education attained by the household head Degree & above 1.2 1.1 1.9 remains high. The multivariate probit (Table A, Annex 3) shows that attainment of an A-level education and above by household heads i s associated with an 8 percent lower probability of being poor, and a 5' grade education and below i s associated with an 18 percent higher probability o f being poor. In 2002, the poverty headcount was more than 30 and 45 percent, respectively, if the household head had an education up to grade 5 only or had no schooling-significantly higher than the national average (Table 3-5). Poverty incidence in households whose heads have no schooling has also increased from 1990-91 to 2002, while the A number of papers in the economic literature have shown the links between incomes and education. Datt and Ravallion (2002) show the links between education and nonfarm economic growth, which results in economic diversity. Basu and others (2001) show that externalities of having an educated member on the household can result in higher earnings of other household members. 27 proportion o f such households in the total has declined from 16 to 13 percent. Even as school enrollments have gone up the disadvantages suffered by households lagging behind in education have expanded over time. Poverty and land ownership 21. Land ownership is closely associated with poverty in many developing countries, notjust because land provides the main source of income, but also because land ownership improves access to economic and social opportunities, especially in rural areas. Inthe case of Sri Lanka, a few caveats place limitations on the analysis o f the association between land ownership and poverty. First, since the HIES did not include detailed information on land ownership, the analysis draws on the CFSES, and the poverty rankings o f households from this survey are not strictly comparable with those based on the HIES. Second, land ownership in Sri Lanka i s quite complex and sometimes difficult to establish. The CFSES focused on defacto ownership o f the land without any reference to documentary evidence of legal ownership. This is a practical solution to the complex issue, but may also result in a certain degree o f misreporting. Moreover, the absence o f legal ownership may impose constraints as far as selling or mortgaging the landfor credit that further complicates the association betweenland ownership and poverty. 22. The CFSES 2003/04 report shows that land ownership rose with income level and that the difference between the rich and the poor was much smaller than in other countries in the South Asia region. For example, inthe poorest quintile 86 percent o f households own land, while inthe richest quintile 96 percent o f households own land. The size of the land parcel per household did not vary much among different income levels except for the nchest quintile. Households in the poorest quintile owned 117 perches on average, while the average ownership for the 4* and 5& quintiles were 134 and 197 perches, respectively. Gender, ethnicity, and religionof household heads 23. From the HIES, there i s no evidence to show that households headedby females are poorer than those headedby males (Table A-3.1,Annex 3). This does not necessarily imply, however, that female-headed households suffer no economic disadvantages. The lack of correlation, which i s often observed for developingcountries, may result from poorer households designating a male member as the household head due to social and cultural factors, even if the household lacks an adult male o f workmg age. Also, ethnicity and religion o f household heads are not correlated with poverty incidence (Table A-3.2, Annex 3). 24. Since the rural poor account for 88 percent o f the total poor o f Sri Lanka, the household characteristics associated with poverty identified above predominantly reflect these areas. The next section profiles the urbanpoor, a large proportion of who live inand around Colombo city. Characteristics of urbanpoverty 25. Why profile the urban poor separately, given that urban poverty rate is only around 8 percent? First, even a low poverty rate can translate into a significant number of poor people ina relatively small area with a large population, as the poverty map for divisional secretary (DS) divisions o f Sri Lanka shows (Figure 3-1).5 Large numbers o f poor people are found in Western Province including Colombo city due to its high population density. Second, as urban poverty rates have dropped consumption inequality has increased, widening the gap between the rich and the poor (chapter 2). A poverty map o f Colombo area shows that poverty headcount ratios vary widely within Colombo City, with pockets of deprivation amid relative affluence. This poverty Sri Lanka has four tiers of administration: 9 provinces, 25 districts, 324 divisional secretary's divisions (DS divisions), andaround 14,000 GramaNiladhari division (GNdivisions). 28 profile will identify factors that prevent some urban dwellers, even in Colombo, from availing themselves ofthe substantial economic opportunities in the country's most vibrant growth center. 26. A study o f poverty in Colombo is especially difficult due to the small size of the household Figure 3-1:Spatial Concentration of poor survey sample for the city. A few studies using less population formal data sources show a clear pattern endemic in the poor settlements/slums of Colombo City.6These areas are underserved in terms o f access to basic infrastructure facilities, and they lack stable income sources, access to infrastructure, decent housing, and clarity inlandtenure (Box 3.1). percent of them are engaged in ~ ~ s k i l l e ~ i r r e ~ u ~ a r employment activities. (2) Lack of access to basic mfrastructure and housing: 43 percent of residents in the underserved settlements picked availability of water for domestic use as their highestpriority int e r n of needs, 27 percent picked the availability and quality of sewer system; and 24 percent the availability of electricity. Manyresidents also expressed coiicern about poor quality of building materials used in housing construction. (c) Note: Dots are placedrandomlywithin aDS division; eachdot represents5pOpeople Source: DCS(2004). incentive o f residents to build proper houses or renovate their houses with proper materials; they also live in constant fear of sudden eviction by a legal entity. The studies also document the recognition among inany residents of` poor settlements that the key ingredients for moving out o f poverty are completing primary/secoritiary education, being able IO find work in a formal sector, or establishing their own businesses. .%xwc,c DFID and others (2002) and (iunrtilleke and others (2004) 27. Evidencefrom Census 2001. The findings reported in Box 3.1 are instructive, and can be validated to some extent using the Population Census o f 2001 that provides wider ~overage.~ Based on poverty map estimates and consultations with staff at the DCS, two GN divisions in Colombo City were chosen as "underserved" or poor GN divisions: Mudumpitiyu and Muhuwuttu. The estimates show poverty headcount ratios for these two GN divisions of 19 and 18 percent, respectively, more than three times of the poverty headcount ratio o f Colombo District. DFID and others(2002) andGunetillekeandothers(2004) The Census 2001 identifies areas only at the GN division level, which is larger than a typical urban underserved aredsettlement, a problemthat canbe minimized by careful selection of GN divisions. Underserved areas in Colombo were selectedfi-omthe Census (5 GN divisions) based on their estimatesof povertyheadcountratios, and out of them, 2 GN divisionswere chosenafter closeconsultationwith knowledgeableDCSstaff. 29 28. The results are consistent with Box 3.1. When compared to Table 3-6: Welfare indicators in Colombo City (Dercentl other households in Colombo Other Underserved District, the two poorhnderserved GNs GNs GN divisions have significantly less Light:electricity 86 56 access to clean water, electricity, and Toilet: exclusively for the gas, and a toilet exclusively for the household 61 37 household. These areas also have a Water: Tap withinuniupremises 53 31 much higher proportion of one-room Fuel: gas 58 30 hutsand shanties. Inthe underserved Wall: plank 9 35 areas, a much lower percentage o f Type ofhouse: hudshanty 4 13 individuals finish primary and House: single roomed house 19 33 secondary school and a much higher Individuals (age2 5 ) withhigher percentage o f working-age adults are than primary education 78 58 employed in the informal sector Individuals (age2 5 ) with higher (Table 3-6). than secondary education 43 16 Individuals (age 2 5 ) inelementary 29. Poverty inurban areas appears occupation 31 55 to be associated with l o w Proportion o fmigrants 27 12 educational attainment, employment - in elementary occupations in the Note: Underserved areas in Colombo city are GN 20 and 25 within DS code 3; "other GN divisions" in ColomboCity refer to remaining informal sector, and poor living GNs inDS3 andall GNdivisions in DS 27. conditions. These associations Source: Census ofPopulationandHousingCondition2001. explain why the urban poor are unlikely to increase their economic potential: low human capital and inadequate access to safe water and sanitation critical for health trap them inlow-paying and insecure employment. The proportion of migrants in underserved settlements i s also smaller than that in other areas o f the city, which is consistent with the findings inchapter 4 that migrants are, on average, better-educated and better-employed than long-term residents o f Colombo. 3.2. Thecharacteristicsof poorprovinces and districts 30. Regional or spatial characteristics can limit the economic potential even for individuals with favorable attributes. As shown inchapter 2, inSri Lanka the income and growth distribution pattern over the past decade shows steadily increasing spatial economic inequality. The decomposition in chapter 2 indicates that interdistrict inequality grew much faster than average inequality within a district. This section explores the regional (province or district-specific) characteristics that are strongly associated with poverty incidence. 31. The probit regression, described earlier that explains the probability o f a household being poor, shows the importance of district-specific characteristics (Table A, Annex 3, columns 1 and 2). L o w district-level averages for electricity usage and access to potential markets are strongly associated with the presence o f poor households. Even when the accessibility index i s omitted from the regression, the proportion o f household heads with education up to primary level or employment as apcultural workers in a district i s also significantly associated with the probability o f a household being poor.8 The next section explores how each o f these factors is associated with, and partly explains, the regional distribution o f income and poverty among provinces and districts. * Note that all spatial variables cannot be included together in a regression because of high multicollinearity. In particular, accessibility index has very high correlation with the other spatial variables taken together (see Annex 3, Table A-3.10), for which proportion of household heads with education of primary level or below and proportion of householdheads employedas agriculturalworkers are droppedwhen accessibilityis introducedinthe regression. 30 Access to markets and infrastructure 32. The association between poverty indices and measures o f access to business opportunities is evldent at different levels o f geographic disaggregation. The accessibility index, which is constructed as a measure of potential market integration and based on information about road network and location of major cities and towns, i s one measure. When t h s index measure is high the area has better than average access to markets. Another measure i s travel time to Colombo- which vanes between estimates based on the geographic information system on road networks, and actual travel times that depend on road and traffic conditions, which may be longer. Additional indicators at the province level that are useful for assessing access to infrastructure that improve the business environment include share of enterprises that use electricity, have a functioning landmobiletelephone line, and are locatednear a community with a bank. Table 3-7: Poverty indices and access to infrastructure by province Enterprises Enterprises Poverty Average Enterprises with a land located in a headcount Average travel time that use line/mobile community ratio accessibility to Colombo electricity phone with a bank Provinces (percent) index" (minJb (percent) (percent) (percent) Western 11 3.8 73 79 24 70 Central 25 3.1 200 80 7 47 Southern 28 3.1 229 68 18 62 NorthWestern 27 3.1 177 61 15 70 NorthCentral 21 2.9 304 61 8 75 Uva 37 2.8 295 62 23 78 Sabaragamuwa 35 3.3 152 76 15 1 70 Correlation with headcount -0.62 0.47 -0.32 0.2 0.14 Source` I HIESWB I ICs ICs ICs ICs ICs a. The accessibility index is calculated for every point as the sum of the population totals of surroundingcities and towns, inversely weighted by the road network travel time to each town. The numbers show the mean of the access values for all pointsthat fall into agivenprovince. b.The averagetravel time to Colombo city is estimatedtraveltime to each town basedon geographical informationof roadnetwork. The numbers showthe mean ofthe travel time for all pointsthat fall into agivenprovince. c. "HIES WB" denotes calculations by World Bank staff usingHIES 2002; "ICs' refers to World Bank "Sri Lanka: ImprovingtheRuralandUrbanInvestmentClimate" (2004). 33. The association between indicators of accessibility and poverty incidence comes through most clearly at the provincial level (Table 3-7). Western Province has the best access to business opportunities by any measure. Many of these indicators, however, do not necessarily explain the extent o f deprivationinother provinces. For example, Uva, the province with the highest poverty headcount ratio, has almost the same level of telephone connection as the Western province and even better access to a community with a bank, but it has the worst access to markets, a city, and electricity. Among these accessibility indicators, geographical isolation and high average travel time to Colombo seem to be most closely correlated with poverty. The correlation coefficient between the accessibility index and poverty headcount ratio i s -62 percent and that o fthe average travel time to Colombo i s as highas 47 percent. 34. The strength o f the correlations on factors that constrain investment and growth in poorer provinces is limited by the large variation likely to exist within a geographic area as large as a province. The relationships between poverty headcount rates and measures of accessibility at the district level are stronger than at province level (Figure 3-2). According to the probit regression (Table A, Annex 3) the probability o f a household beingpoor falls by almost 3 percent with a unit increase in the accessibility index of the district the household is located in, even after controlling 31 for other factors that affect the probability o f being poor. Moreover, the association is much reduced once districts in Western Province are excluded, suggesting that these indicators mostly explain differences among districts in Western Province than between Western Province and the rest o f the country. Figure 3-2: Accessibility index and average driving distance to Colombocorrelated with district overty headcounts E 300 - * 4 0 P=0.37 correlation=-0.7 3E, 200 - .-9E correlation=0.61 5 100- 2.5 1 I? = 0.49 f t 'E U * I 2 1 0 10 20 30 0 10 20 30 40 Poverty HCR(%) Povertv HCR(%\ Figure 3-3: Proportion of housing units using electricity or gas correlated with district poverty headcounts 1 100 I I s - 1 R* = 0.67 e 20i * *-\i \* %.ii 0 10 20 30 40 0 10 20 30 40 Poverty HCR (%) Poverty HCR (%) Source: Staff calculation using HIES 2002 and Census2001. 35. Data on the proportions o f enterprises with electricity, access to telephone lines, and proximity to a community with a bank are not available at the district level. Instead, two new indicators emerged as proxies for infi-astructure availability for economic activities at the district level from the 2001 Census: shares of housingunits that use electncity for lighting and gas for cooking fuel. Figure 3-3 shows that few households in poorer areas use electricity and gas. The probit regression (Table A, Annex 3) shows that the average usage of electricity in a district i s associated with the probability o f poor households in the district. The correlations for these indicators with poverty headcount are reduced once the districts in Western Province are excluded, but are still significant (above 30 percent). The size of rural and estate sectors, and inequality in a province 36. The probit regressions above show that the probability o f being poor is significantly higher when the household belongs to the rural or estate sector, even after allowing for the impact of other household-specific and geographic characteristics. Hence, provinces with a large share o f rural and estate population tend to be poorer. Table 3-8 shows that urban areas in all provinces tend to be significantly better-off than rural areas, with the exception o f Sabaragamuwa. Poverty in Sabaragamuwa is not just a rural feature: mean per capita consumption expenditure in the 32 urban areas of Sabaragamuwai s significantly lower than that o f other provinces, and i s in fact comparable to that of ruraE Westem Province. The two poorest provinces, Uva and Sabaragamuwa, also have relatively high shares ofpopulation inestates (Table 3-8). A high share o f estate population also explains why the poverty headcount ratio of Central Province is worse than that of North Central Province, in spite of Central Province having the richest urban sector and the secondrichest rural sector inthe country. Table 3-8: Sectoral shares and inequality measuresby province Meanper capita monthly consumption GINIofper Poverty Share ofpopulation expenditure capita headcount (percent) (Rs) consumption ratio expenditure (percent) Rural Estate Urban Rural Estate (percent) Western 11 69.7 0.9 5447 3800 1911 40 NorthCentral 21 95.2 1.o 5001 2574 4833 33 Central 25 70.2 21.o 5644 2630 2067 38 NorthWestern 27 95.7 0.5 5113 2582 1908 37 Southern 28 90.2 2.0 4918 2488 1979 37 Sabaragamuwa 35 87.4 8.8 3864 2317 1817 35 Uva 37 80.8 15.6 5282 2342 1765 39 Total 23 80.6 6.0 5285 2865 1985 40 Source: WorldBank staffcalculations usingHIES2002. 37. Table 3-8 also illustrates how inequality within a province explains its relative rank in terms o f poverty incidence. For example, access to infrastructure critical for business development in North Central Province ranks quite low (Table 3-7). Also, the majority o f its population resides in the rural sector (more than 95 per cent), and the mean expenditure for t h s sector i s not high in comparison with other provinces. The evidence so far suggests that the relatively low poverty incidence in North-Central Province i s attributable to two factors: (1) the lowest inequality among all provinces as measuredby Gini o f per capita consumption, and (2) a substantially better-off estate populationcompared to other provinces. Educational attainment and regional poverty 38. Average educational attainment at the province or district level is associated with higher poverty incidence. A higher share of household heads with no schooling in a province corresponds to a higher poverty headcount. Conversely, a higher share of household heads with at least GCE (OL) education in a province corresponds to a lower poverty headcount (Table 3-9). If Western Province i s excluded, both correlation coefficients are still significant at 25 percent in absolute terms. 33 0 10 20 30 40 0 10 20 30 40 Poverty HCR( O h ) Poverty HCR( O h ) 34 unemployed in Sn Lanka mainly because they can afford to wait for better job opportunities (see queuing hypothesis in section 3.1). Second, however low the wages are, the extreme poor must work because they cannot survive without income. Thus, in relatively poor areas, underemployment is likely to be associated more closely with poverty. Incontrast, in Colombo District where only 6 percent ofpopulation i s poor, unemployment inthe expectation ofbetterjob offers inthe future i s likely to be more common. Regionalpoverty and the agricultural sector 45. Employment inapculture i s associatedwith a higher likelihood o f a household beingpoor (section 3.1). Table 3-10 shows a highcorrelationbetween poverty headcount and the proportion o f agricultural paid significantly when excluded. The probit marginal effect on the probability of a household located in Note.The agricultural sector includes fishing andforest workers. thedistrictto bepoor. Source. World Bank staff calculations usingHIES2002. 46. Average per capita monthly income of households whose heads are workmg as an apcultural worker has a stronger association with poverty incidence in a province (Table 3-10) and at the district level (see Annex 3, Table A-3.8). This partly explains the high poverty headcounts in Southern Province and Sabaragamuwa in spite o f the relatively low share of agricultural wage workers: these provinces ranked at the bottom in per capita income when household headsworked as paid apcultural employees. 47. Therefore, when agricultural sector wage employment predominates in a province or district the poverty incidence i s higher, but much o f this association disappears when Western Province, which is primarily nonapcultural, is excluded fkom the analysis. What seems to matter more, and is more closely related to regional patterns of poverty, i s the average income o f households with agricultural employees ina particular area. This inturn suggests variation in the incomes o f agricultural households acrossregions. 48. The correlation between spatial patterns of poverty and variations in incomes o f agricultural households makes it important to understand what factors explain such variations- the subject o f chapter 6. Moreover, chapter 6 will look at the nonagricultural rural sector as critical income-diversification opportunities for households engaged inapculture. 49. Inreview, comparisons among provinces and districts show some evidence that poverty is concentrated inareas that are geographically isolated(interms of distance to markets and cities); where access to electricity and gas are limited; where the proportion o f highly educated household heads i s small; and where the proportion of agncultural wage workers in total employment i s high and the average incomes of households with such workers are low. Conversely, there is no clear evidence that inequality between provinces or districts is closely associatedwith differences inregionalunemployment rates. 35 50. Many o f the province and district characteristics are better for explaining differences in poverty incidence between Western Province and others rather than differences among provinces and districts outside o fWestern Province, which indicates the vast gap between Western Province and the rest o f the country. These differences suggest a story of two economies circumscribed by geography: one of Westem Province where there i s better access to markets and infrastructure, a higher concentration of educated people, and where nonagricultural sectors play a predominant role; and the rest o f the country where the converse o f these conditions holds. Although some provinces like North-Central fare much better than Uva and Sabaragamuwa, the differences within these provinces are muchless than the difference betweenthem and Westem Province. 51. Such broad district- or province-level profiles however obscure significant pockets of poverty that exist in even the richest districts. Even in the more prosperous Western Province, a vulnerable group like the estate populationhas one-half the per capita consumptionlevels of rural households. This i s also consistent with a result from chapter 2, that within-district inequality in consumption has increased by 52 percent over the decade-less than the 112percent increase in between-district inequality, but substantial all the same. 52. Analysis that i s based on more disaggregated geographical data can better reveal the spatial correlates o f poverty-particularly those that can be influenced by policy instruments, like access to markets and infrasmcture. The analysis below uses poverty maps and other geographical information at the level o fD S division. 3.3. DS divisionsas the unit of analysisfor characterizingpoor areas 53. The ability to conduct geographically disaggregated analysis below the district level i s usually hampered by lack of reliable data, since most sample surveys like HIES are designed to be representative at the district level. Recent efforts by the DCS and the World Bank have found ways around this problem by employing the statistical technique o f small-area estimation to combine Census and HIES data to estimate poverty incidence for subdistrict administrative areas (see Annex 3, section I), and by using information from Census and other sources to measure certainkey spatial characteristics below the district level. How spatial characteristics of DS divisionsrelate to poverty incidence 54. The probit regression of probability of a household being poor offers an opportunity to relate available D S division-level indicators with poverty incidence. The regressions presented in Table A, Annex 3 (columns 3 and 4) replace some o f the district-level indicators with D S division-level averages. The results are similar to those obtained with district-level characteristics. The marginal effects o f almost all the household-specific factors on the probability o f being poor remain unchanged. Spatial characteristics significant at the district level continue to be significant at the D S division level. A household is more likely to be poor if it belongs to a D S division with lower average access to markets, lower proportion o f households using electricity, and a higher proportion o f household heads with below-primary education (when the accessibility index is omitted from the regression). 55. These three spatial characteristics at the D S division level significantly affect the probability o f being poor even when the sample excludes households in Colombo District (see Annex 3, Table A-3.9). Thus they are important not only in explaining the variations in poverty incidence between the richest district and the rest o fthe country, but also the variations inpoverty outside this district. Poverty estimatesfor DS divisions 56. Poverty mapping using the small-area estimation technique provides new insights into the distribution o f poverty in the country, even for areas where more aggregated analysis suggests 36 low incidence o f poverty (see Annex 3, section I).map o f poverty headcount ratios at the D S A Division level indicates some interesting geographic patterns (Figure 3-5). As expected, poverty headcount ratios are substantially lower inColombo District and itsneighboring areas. Areas with highrates o fpoverty are especially prevalentinSouthernUva and SabaragamuwaProvinces. But pockets of extreme poverty exist even in districts with relatively low poverty rates - for example, some D S divisions inKalutara District of Western Province and inNorth-West and North-Central Provinces (e.g., parts o f Puttalam, Anuradhapura and Kurunegala Districts). Figure 3-5: Poverty estimatesfor DS divisions Figure 3-6: Accessibility index for DS divisions tow high Note: The accessibility index is calculated for every point as the sum o f the population totals of surrounding citie: and towns inversely weighted by the road network travel time to each town. The map shows the mean o f the access values for all points that fall into a given DS unit. The index is a measure o f potential market integration reflecting the quality and density of local transportation infrastructure, for 185 citiedtowns. Source: Poverty map from DCS (2004); accessibility map based on staff calculations. Relating characteristics of DS divisions with poverty estimates 57. Accessibility potential. As seen earlier, geographical isolation measured by the distance to the nearest market or town i s highly correlated with district poverty headcounts. This correlation is likely to be even more important at the D S division level, since accessibility should be more relevant for a smaller geographical area. Accessibility indices calculated for D S divisions shows that areas surrounding the Colombo District inWestern Province are well connected to towns and markets (Figure 3-6). These include the southwestern coastal areas surrounding Colombo City and the areas between Colombo and Kandy City. In general, the accessibility index appears to rapidly decline with distance from Colombo and its surrounding areas. 58. Comparing Figure 3-5 and Figure 3-6, the higher the accessibility index o f a D S division, the lower tends to be its poverty incidence. For example, the coastal areas surrounding Colombo District record a lugh accessibility index as well as a low poverty headcount ratio, whereas many D S divisions inMonaragalaDistrict o f Uva province are very poor and geographically isolated. A 37 scatter plot and a simple regression o f poverty headcount on accessibility index confirm a significant correlation (of -0.5 8) between these two indices (Figure 3-7). 59. Educational attuinment and Figure3-7: Poverty headcount and accessto urban electricity usage. DS division-level centers at D S division level indicators o f education attainment and usage o f electricity among households are also highly correlated with poverty estimates for D S divisions obtained fromthe poverty map. The correlation between poverty incidence and the proportion o f household heads with education o fprimary level or below is -0.62, and that between poverty and the share o f households using electricity is 0.79. 60. The spatial characteristics considered I here are also strongly correlated with each other. This is particularly true for the accessibility index vis-A-vis other spatial factors. A regression shows that D S divisions that are better connected to markets are also likely to have larger proportion of households with electricity and better education attainment among household heads, and belong to districts with lower unemployment rates and lower levels o f agricultural wage employment (Annex, Table A-3.10).9 3.4. Concludingremarks 61. This chapter identifies broad correlates and hints at some of the sector-specific issues critical to explainingpoverty outcomes and informing policy interventions. The analysis excludes the conflict-affected areas in the North and East, due to lack o f availability o f representative household data or Census information. Chapter 7 will piece together available information for these areas from a variety o f sources to draw a profile o f poverty and vulnerability in the context o f unique challenges brought on by the conflict. The correlates of poverty already identified are a usehl starting point for more in-depth analysis in subsequent chapters. Once the constraints that recur across poor households and areas have been identified, sector-specific analysis will address the specific reasons for stagnation in rural incomes, and why certain groups like the estate residents lagbehind the rest o fthe country inincome and nonincome dimensions o fpoverty. 62. A number o f household-specific factors are associated with a lower likelihood of being poor: the presence o f at least one formal sector employee in the household, a family member worlung abroad, andhigher educational attainment ofthe household head. Factors associated with a higher probability o f being poor include unemployment, particularly among youth, and underemployment-usually associated with employment in the informal sector. Consistent with the pattern of agncultural stagnation, employment as an agricultural worker increases the likelihood o f poverty. In urban areas, poor living conditions and a lack o f basic services are critical issues for the poor. 63. These correlations suggest a few avenues for poverty reduction: increasing productivity and incomes from agnculture, creating growth and employment in the formal sector, and investing in higher education. Since poverty i s strongly linked to employment in the informal This regression has an R2of 0.71, which i s very high given the short list o f explanatory variables and indicates high multicollinearity. This also explains why all the spatial variables cannot be included simultaneously in the probit regressionsofprobability o fbeingpoor. 38 sector, rapid growth of the formal sector will have a large impact on poverty-by drawing workers out o f the informal sector and raising informal sector wages. Removing labor market rigidities (see chapter 1) can stimulate growth and employment in the formal sector, particularly in urban areas. Better long-term economic opportunities for the poor require a secondary education and more marketable skills in the labor market. The high incidence of youth unemployment among school dropouts from poor households is likely to perpetuate poverty and sharpen inequality across generations. T o identify the types of interventionslikely to be effective in reducing dropouts and improving educational attainment, more information is needed to understand what factors constrain education among the poor. 64. While most o f these household-specific factors are quite intuitive, even after controlling for their effects, a number o f spatial characteristics emerge as strong correlates o f poverty. The stylized example using the probit regression in Table A, Annex 3 show that even when two households are identical inall household characteristics but differ in location-say one is located in a DS division with the average spatial characteristics of Colombo District and the other is located in an average D S division of Monaragala-the Colombo District household i s 7 percent less likely to be poor. Although this exercise i s a limited application of potential spatial and household-specific factors, it illustrates a message supported by all the available evidence: location matters in Sri Lanka, and location-specific characteristics are critical in explaining the uneven pattern of development. 65. A few spatial factors relevant to poverty emerge as critical constraints faced by lagging reBons in the country. Accessibility or the potential for market integration depends on distance from (and available road links with) towns and markets. Higher average accessibility of the D S division or district significantly reduces the probability of a household located there being poor, and i s therefore associated with a lower poverty rate for the area. Other important spatial factors include access to electricity and educational attainment, which partly capture the attributes relevant for growth and poverty reduction ina particular region-links to markets, the availability of infrastructure and human capital. These constraints are also highly correlated with each other, indicating the multiple challenges faced by poor areas. Remote areas with poor connections to markets and cities, for example, are also likely to have lower access to electricity and lower educational attainment. It seems intuitive then that policy interventions to promote regonal development would needto occur alongmultiple dimensions designed to complement each other. 66. Although urban poverty rates are low compared to the rest of the country, these rates translate to large numbers due to the high concentration o f population in Colombo and surrounding areas. Reducing urban poverty poses special challenges. Urban poverty i s highly concentrated within Colombo-and upto three times the poverty rate o fColombo District prevail in a few areas that are also severely underserved in basic services like water, sanitation, and housing. Inthe long run, improvingthese services will require more responsive urban municipal governments and better urbanplanning. 67. While transformations effected through medium-term strategies in regional development and urban planning will take time, targeted interventions can play a key role in protecting the poor and improvingtheir access to basic services in the interim. Targeted initiatives to provide clean water, sanitation, and assistance for housing to underserved settlements can improve welfare substantially among the urban poor; and safety net schemes can ensure a minimumlevel o f sustenance for the poor in rural and urban areas alike. The Samurdhi program, as the largest welfare program in the country, has the greatest potential in terms of outreach and capacity to reach the poor, but to achieve this objective, it must substantially improve its effectiveness in targeting the poor (see chapter 2). 39 4. Internal Migration, Remittances,and Urban Concentration 1. As earlier chapters have highlighted, poverty inthe lagging repons o f Sri Lanka-remote districts far away from Colombo, with a large rural and estate population-is a few orders o f magnitude higher than that in Western F'rovince, and this gap has been growing over the last decade. The growing regional imbalance contributes to certain dynamic processes, which can in turninfluence the future path of growth and poverty reduction. Perhapsthe most critical among these i s the movement o f people from backward regions to faster-growing areas in search o f better opportunities. Internal migration from ruralh-emote areas to the urban growth center doubled between 1996-97 and 2003-04, while the gap in economic performance between rural areas and the rest o f the country widened (CFSES, 2003-04). This chapter describes the trend and patternof internal migration inrecent years, and its connection to widening regional inequality. 2. The movement of people from rural areas to the urban growth center of Colombo has complex and profound effects on poverty and inequality depending on the causes, scale, and pattern of migration. Migration can reduce cross-regional inequality as people move inresponse to wage differences and reduce the wage gap. Remittances to the migrants' place o f origin in lagging economic regions can also reduce regional inequality. In fact, migration often represents the best available option to those fi-om lagging regions to better their economic status, and it can significantly improve the welfare and poverty levels of households who migrate. 3. Migration can also perpetuate regional imbalances by attracting the more endowed, in terms of slulls or wealth, to better employment opportunities in Colombo. Moreover, the large- scale movements of people into a single rapidly growing urban center can impose limits on growth due to overcrowding and worsen existing poverty. Thus, the scale and pattern of migration can influence the incidence and characteristics o f urban poverty, as well as the challenges faced byurbanregionsto keep the momentumo f economic growth going. 4. Trlls chapter analyzes the characteristics of migrants and the purpose o f migration that in turn affect the place of origin as well as the destination. A key element in understanding the impact of migration involves looking at the incidence, size, and use of remittances at the place o f origin. The effect of urbanagglomeration, which i s a consequenceof the pattern o fmigration, and the limits imposed by over-agglomeration on economic growth are also examined along with implications for policy. 4.1. Trendsin internalmigration over thepast decade 5. The combined proportion o f internal migrants and emigrants to other countries increased fi-om 78 per 1,000 households in 1996-97 to 89 in 2003-04, according to CFSES data. Of t h s total, internal migration increased from 15 to 29 per 1,000 households. In2003-04, 81 percent of internal migration was undertaken to seek employment.' The HIES data show similar increases in migration from 1990-91 to 2002. Both estimates probably understate migration, however, since migrants inthese surveys are defined as living away from other members of the same household. Inother words, these figures do not include the migration of entire households (see Annex 4, section Over the same time, external migration fell marginally &om 63 to 60 per 1,000 households. Remittances from abroad, however, still grew at 11 per cent in 2003. The focus o f this chapter i s however on internal migration, since it is a consequence o fregional inequality and inturn is likely to affect regional growth and inequality patterns. Internal migration rate is reported as higher (71 per 1000 households) than external migration in CFSES 2003-04 Report (part 2), ifthe reference period for internal migration is taken to be the same as that for external migration; this number is however not comparable with the internal migration rate from CFSES 1996-97. This is why the increase in internal migration is reported as 15 to 29 per 1000, which is for the sume reference period across the two surveys. 40 Figure 4-1: Monthly wage inprovince as a percentage of wage in Colombo District . I Occupatm m general r---- Elementary occupation __- - 4 0 / , m , , , , 1 ~ a I m 2 2 2mb 2mw m zm z z oom, oo oo oo wm mm o - ~ m N N N N -- -Westem (excl Colombo) ---Central Cobmb) ---Central Southern North-West 1 - --- Western(excl I - Southem -North-West North-Central ' ___ North-Central UVa -Sabaragamuwa 1-Sabaragamuwa Source: LFS (vanous years). 6. The sharp increase in internal migration measuredby CFSES is consistent with increasing economic inequality between regions through the past decade, which have largely resulted from the increased economic benefits o f migrating to Colombo and its surrounding areas. Figure 4-1 shows a widening gap in monthly wage earnings between Colombo District and all provinces other than Western Province. This gap does not necessarily translate into commensurate opportunities for potential migrants, however, since wages in the TEWA-protected formal sector-which i s relatively large in Colombo District-are set artificially high above market- clearing levels, and this labor market distortion hampers job creation in the formal ~ector.~ Elementary occupations, which are mostly unregulated and better reflect market forces, show a significant wage gap between Colombo District and other areas, and as such present a substantial economic incentive to migrate. 7. Incidence of internal out-migration increased for all provinces from 199&97 to Figure 4-2: Composition of migrants by sector of origin 2003-04, with the exception o f Western Province, the destination for most migrants. Most migrants came from the Northern and 80% Eastern provinces in 2003-04, a pattern 70% consistent 60% with lower economic 50% opportunities there due to the security 40% situation and the ability o f people to migrate 30% relatively safely within the country after the 20% ceasefire. 10% 0% 8. Indeed the HIES data show an 1990/91 2002 increase in the proportion o f migrants from Source: Staff estimates using HIES data, 1990-91 and rural and estate areas from 1990-91 to 2002, 2002. and a decline in migrants from urban areas ' The simplest economic theories posit that migration occurs when the expected economic benefit-determined by the earning differential and the probability o f finding a job at the destination-outstrips the cost o f migrating. The wages prevailing in TEWA-protected sectors-which are far more numerous in Colombo than outside and thus affect Colombo wages more-are high enough to impede new hiringand therefore reduce the probability of finding a job in this sector. Thus, the economic benefit from migrating can be overstated if onejust looks at the wage differential. 41 (Figure 4-2). Infact, the share o f migrants from the estate sector more than doubled during this period. For example, nearly 90 percent ofmigrants came fromrural and estate sectors in2002. 4.2. How migration affects theeconomy at the origin and destination 9. The economic literature suggests a wide range of likely welfare impacts depending on who migrates, where they migrate to, and why. Migration among highly slulled workers can reduce productivity and management skills in the source regons, which can in turn deepen regional inequality. At the same time, emigration of slulled workers can also generate benefits for lagging regions because remittances fi-om emigrants usually far exceed the incomes available in their place o f origin. The emigrants may also confer indirect benefits to their place of origin, such as inducing investment in education at the origin and the spillover of knowledge fi-om their destination to the place o f origin, some of which may raise the stock ofhuman ~ a p i t a l . ~ 10. The out-migration o f low-skilled workers on the other hand frequently reduces poverty at the place o f origm. Low-skilled workers are likely to earn l g h e r wages at their destinations and send remittances to household members in the location o f origin; also their emigration can raise wages or create new job opportunities for those left behind. In reality, however, the emigration of poor and low-slulled workers is often limited by their lack of financial resources, slulls required to obtain ajob, or access to social networks inurban areas.5 Similarly, the in-migration ofpoor or low-slulled workers at the target destination can raise unemployment rates, expand urban slums, and strain urban services. Although the link between migration and urban poverty in Colombo City attracts considerable policy interest, few studies have explored this issue with adequate data, and this chapter makes a beginning inthis direction. 11. Even though migration from ruralhemote areas to Colombo District and City widens spatial disparity, it may represent an efficient use o f the labor force. International experience suggests that internal migration to a few large cities i s an inevitable stage o f the development process.6 Cities provide a critical mass o f consumers, a network o f suppliers o f intermediate goods, and access to a large pool of shlled workers. These advantages attract workers from lagging areas, which can further widen the economic gap between cities and outlying areas. However, as agglomeration continues, congestion can hamper transactions and reduce productivity in the urban center itself. Hence, the externalities o f urban overconcentration may leadto a divergence inindividual and social benefits o f migration. Namely, migrants can continue to flow into an urban center long after the social cost o fcongestion exceeds the economic benefits o f agglomeration. 12. The net welfare impact o f migration trends depends on who migrates, where they end up, and how. Data for a comprehensive profile o f migrants within Sri Lanka is unavailable, but this obstacle i s partly overcome with data from the Census and household surveys and inferences about households who have migrated or those who have migrant members. The comprehensive use o f all three databases leads to a reasonable picture o f the economic and socially complex phenomenon o f migration (see Annex 4, section I). Characteristics of internal migrantsinto Colombo City 13. Migration is frequently triggered by a combination of "push-pull" factors associated with regional economic differences. Push factors that prevail in the place o f origin include lack of economic opportunities in a poor regon, which compel certain types of individuals to head for the urban center. Pull factors are prevail inurban centers like Colombo higher wages for highly See Lucas (2005) and Schiff (2006). See, for example, Mahmud(1989) for Bangladesh. See World Bank (2000), Worid Development Report 2000, chapter 6 42 15. Table 4-1 shows that the poverty incidence inthe district of origin i s strongly associated with recent Table 4-1: Correlation of size of recent migration into Colombo City migration into Colombo City. The correlation is With poverty headcount rate of stronger with the district poverty rates in 1995-96 than districts, 1995-96 0.56 with those for 2002, which i s consistent with the With poverty headcount rate of definition o f "recent" migration in 2001 Census. districts, 2002 0.39 Finally, although the aforementioned network effects Source: Census (2001), HIES (1995-96,2002) . - poor areas, this does not necessarily --- imply that the migrants themselves are j3igure 4-3: Share of householdheadswith tertiary poor. On the contrary, Figure 4-3 :ducation and residents by district of origin) suggests that skilled workers are more (migrants likely to emigrate from poor areas in Sri 80 -- i Lanka. The proportion of those with an 0-level (On) education or above i s much higher among household heads who emigrated into Colombo City than among those in the districts o f origin who remained. 18. The fact that this pattern i s true for all districts, and notjust the poorest, suggests the pull factors inColombo are strong enough to attract highly skilled workers from areas that are not iource: World Bank staffcalculationsusingCensus (2001) Sri LankanTamils constitute around30 percent of Colombo City's population, comparedwith only 4.5 percent of the total population,excludingthe North andEast (HIES, 2002). 43 economically disadvantaged. At the same time, the fact that these migrants are much better educated than other residents inthe districts of origm suggests that education level i s likely to be an important constraint to mobility for people inlagging areas. 19. Occupation of migrants. International experience suggests that well-educated migrants may not find jobs commensurate with their skills at their destination. If this were true in Sri Lanka, even though migrants appear to be among the better-educated in their districts of origin, their skills and talents may not be put to optimal use in Colombo City. Figure 4-4 finds some evidence infavor o f this kindof "underemployment" among migrants to the Colombo area. Even though tertiary educational attainment among migrants and recent migrants is almost twice that of nonmigrantresidents inColombo City, migrants seem to enjoy narrower employment advantages, as measured by the proportion o f (recent) migrants and nonmigrants employed in elementary occupations. A sizeable number of slulled migrants are worlung in jobs for which they are overqualified. At the same time, the proportion o f migrants worlung in elementary occupations is much smaller than that o f nonmigrants, and constitutes a minority (27 percent) of employment among migrants. Figure 4-4: Educational attainments and occupation for migrants, recent migrants, and nonmigrants in Colombo City 1 % of elementaly occupationamong workers YOof individuals withtertiary education 40 20 0 Source: World Bank staff calculations using Census (2001) 20. Migrants who did not migrate with the head of their households experience greater difficulties in finding good jobs; a larger proportion of such migrants work in elementary occupations (see Annex 4, Table A-4.3).8 A partial explanation may be that the educational attainment o f these migrants i s lower than that of other migrants, but still higher than that of nonmigrants. This is also consistent with the presence o f a network effect, although there may be other, equally plausible, explanations for t h s finding. Migration and urban poverty 21. Given that migration to the Colombo area appears to be a result o f both push and pull factors, the net impact on urban poverty i s uncertain. It does seem unlikely, given that migrants are, on average, better-off in educational attainment and occupation than nonmigrants, that migrantsthemselves are a significant proportion o f the poor inColombo. 8This finding resolves the apparent contradiction between the Census and CFSES report: the latter indicates that a majority o f internal migrants work as unskilled workers. This i s because CFSES can identify internal migrants only if the rest o f their household live in the district o f origin. The 2001 Census data, however, shows that those who have migrated with their household head actually constitute a dominant majority among all migrants. Among the migrant group closest to the one considered by CFSES, the Census also shows a large proportion (46 percent) employed in elementary or unskilledoccupations (see Yoshida and others 2006). 44 22. Housing conditionsand ownership: Figure 4-5 suggests mgrants into Colombo City have better housing conditions and are thus likely to be better-off than nonrnigrant~.~When the household head is a migrant, the household is more likely to have access to clean water, electricity, gas, and private toilets; and their houses are more likely to have brick walls and more than one room. Figure 4-5: Comparison in housingConditionsby household head's migration status (percent of households) t 80 20 I n Migrant nRecentmgrantI Source: World Bank staff calculations using Census (2001) 23. Migrants are more likely than nonmigrants to rent rather than own the house they live in (Figure 4-6). This Figure 4-6: Home ownership by seems somewhat inconsistent with the finding that household head's migration status migrants are likely to be better-off than nonmigrants, until (percent) we see that a high proportion o f renters migrated from iI 80 ------I conflict areas (see Annex 4, Figure A-4.4), and possibly 60 plan to return to their origin districts as soon as the conflict iI ends. The highproportion o f renters among migrants from I 40 ~ nonconflict districts may be due to a shortage o f housing and a recent hike in land prices in Colombo City that put 20 them at an economic disadvantage. 0 24. Poverty estimates using mapping methods. Typical Ownhouse Rent household surveys cannot provide statistically reliable poverty estimates for migrants to Colombo City due to the small sample size. However the poverty mapping method Source: World Bank staff calculation: described in chapter 3 which combines the Census data using Census (2001) with HIES data canbe usedto estimatepoverty for groups that arerelatively small. Becausethese estimates are based on simulations, they mustbe interpreted with caution. 25. Using this method, the poverty headcount among migrants and recent migrants (with residency of five years or less) to Colombo City is estimated to be well below that o f nonmigrant residents o f the city (Table 4-2)." These findings are consistent with the findings inchapter 3 that This analysis is conducted for households whose heads are (recent) migrants. This is reasonable because first, 80 percent of migrants belong to such households, and second, if household heads are not migrants, the characteristics of housing units, etc., may not reflect the result ofmigration. 10Furthermore, recent research shows that the poverty headcount rates for migrants and recent migrants estimated in this manner are likely to be overestimates, which suggeststhat the gap between migrants andnonmigrants may be even larger than what is suggested bythe poverty mapping estimates. (See Hoogeveen (2003) and Yoshida and others (2006) for details). 45 26. Thus Other countries, the evidence for Sri Lanka seems to suggest that urban Table 4-2: Estimates of poverty headcount poverty i s not a divect result o f the m a l poor ratio in CofomboCit ~igrationstatus of poverty Std. flooding the urban center. At the same time this does household head headcount error not preclude the indirect effects migration can have Nomigrant 10.9 1.o on urban poverty. The influx of better-educated and Migrant 5.6 0.6 shlled migrants from outer districts into the city can Recentmigrant 4.3 0.6 push the less skilled and educated long-term Source: Staff estmatesusingthe 2001 Census and residents out o f better types of employment, and theHIES2002. ' / ~ The percentage of households with remittance The percentage of households with remittance from abroad from within the country 8.0 ~ 6.0 4,0 i 4.0 - I 1 2.0 - m . . .......--....... m - - -....D.-......m , 2.0 - -..B--. 1990-91 I . - - B - - -1990-91I d 2 0 0 2 : 0.0 i I poorest second third fourth richest poorest second third fourLh richest 28. Incidence of remittances in migrants' places of origin. Links between internal and international remittance and poverty can be explored using HIES data. Figure 4-7 shows that incidence o f remittance from abroad is higher among higher-consumption quintiles. The highest share o f households with international remittance is also found among the richest consumption quintile. The incidence of remittances from within the country, however, i s less correlated with consumption level. For both types of remittance, the correlation between householdconsumption and incidence o f remittance i s stronger in 2002 than in 1990-91. These results however do not 46 necessarily imply that better-off households have higher incidence of remittance. Instead, it may be that a poor household becomes better-off due to remittances received. Unfortunately, there i s no easy way to distinguish between these possibilities using currently available household surveys in Sri Lanka. 29. What is clear from Figure 4-7 is that Table 4-3: Ratio of remittanceto remittances are received both by the poorest and consumption expenditure, 2002 richest groups. The proportion o f households in the (percent) poorest quintile that receivedremittance from abroad I Households or within a country (2 to 3 percent) are not that Consumption All with different fi-om the proportion o f households in the quintile households remittance richest quintile who receive remittances (4 to 7 Poorest 1.2 23.7 percent). Remttances constitute a substantial share o f 2nd 1.3 20.4 consumpbon expenditure for households who receive 3rd 2.3 27.1 them, around 20-28 percent for all consumption 4th 3.2 28.1 groups in 2002 (Table 4-3)." Remittances constitute Richest 3.3 27.3 only around 1-3 percent o f income for all Overall 2.3 26.0 households, reflecting the fact that most households I Source: Staff estimates using HIES 2002. I do not receive any remittances. 30. Do remittances reduce the likelihood of being poor? After controlling for a number o f household and spatial characteristics, receiving remittance from abroad has a significantly negative correlation with the probability o f a household being poor (chapter 3). However, this is not enough to say remittance reduces the probability o fbeing poor. To answer this question, it i s necessary to infer a migrant's countetfiuctual income, ie., the income level he or she would earn were they to work inthe area o f origin, which is not possible to obtain from HIES data. 31. A partial picture of the impact o f urban migration on poverty emerges from a special survey in Gampaha District that compared premigration income with postrnigration income.I2 Since premigration income is a reasonable proxy o f the counterfactual income, this comparison indicates whether migration actually improves the income level o f the migrant. The study finds that 73 percent o f in-migrants experienced income increases after their migration to an urban center in Gampaha District. The income increaseis not restrictedto the richest income group; and more than 80 percent o f even the poorest pre-migration income groups managed to increase their monthly income. This study is based on a relatively small sample from a specific area, so a clearer picture o f the impact of remittance on poverty must await analysis usingbetter data at the national level. 32. Impact of migration on education at the place of origin. Sometimes, as evidence from other countries shows, when the opportunity to emigrate increases the returns to education, more individuals are encouragedto invest more ineducation. However, it i s likely that only some o fthe educated people actually emigrate. So ifthe increase inhuman capital o f those who stay exceeds the loss from those who emigrate, the stock o f human capital in the place o f origin may actually increase because o fmigration, which i s inturn i s likely to benefit the laggingregion. 33. Ifthis hypothesis holds, educational attainment ofhousehold membersofmigrants should be higher than that o f households without migrants. The data show little or no variation in educational attainments of household heads or secondary enrollments o f children aged 14 and above based on the migration status o f household members (see Annex 4, Table A-4.1 and Figure A-4.5). This i s also largely true for enrollments o f all other age groups. Moreover, there i s no Per capita monthly remittance ranges from Rs. 270 (in 2002 Rs.) for the poorest income group to more than Rs. 1800 for the richest income group (see Annex 4, TableA-4.2). l2Ranabahu (2004). 47 significant difference in educational expenses between households with and without remittances. Educational expenses account for 11 percent of total household expenditure for households without remittances, but 13 percent for those with remittances. Therefore, there is no evidence that out-migration leads to greater investment or better outcomes in educabon. 4.3. Urban agglomerationand its effects ongrowth andpoverty 34. Migration i s mostly a private decision o f households in response to differences in economic opportunities between areas; and the ability to migrate, therefore, often represents a significant welfare-increasing opportunity for households. However, even as private benefits from migrating are high social benefits may go down, due to overcrowding and congestion in the country's major urban center, and result in a net welfare loss. Cross-country evidence also suggests that overconcentration ina single urban area can result in significant aggregate losses in growth, as compared to a country with a larger number of economically developed urban centers that can serve as alternative destinations for migrants. Indeed, it i s likely that Sri Lanka can obtain significant gains ingrowth and welfare by adopting appropriate policies to reduce agglomeration costs at Colombo. 35. Urbanization and overconcentration. Urbanization and economic growth in developing countries go hand-in-hand. The simple correlation coefficient between the share o f urban population in a country and GDP per capita (in logs) is as high as 0.85. Production is more efficient when concentrated in dense business-industrial districts in cities, due to factors like information spillovers amongst producers, more efficient labor markets, and savings in transportation costs. While such forces promote geographical concentration in a country's main urban area, there are also opposing forces. Immobile factors, such as land and natural resources, can become more expensive due to higher demand; and concentrations o f activity can generate pure external diseconomies such as congestion and pollution that can profoundly affect welfare and growth. The loss in economic growth due to over- or underconcentration in the main urban area can be substantial. According to recent cross-country empincal research, a 10-percentage point gap between a country's optimal and actual urban primacy - defined as the share of the main urban center of the country in the country's total urban population - can reduce its annual growth rate by more than one percentage point.13 I s Sri Lanka overconcentrated? 36. According to Census 2001, around one-half o f Sri Lanka's urban population (1.2 million) lives (per square kilometer) inColombo District, and one-quarter (0.6 million) in Colombo MC-live in an area of a little less 20.000 L than 40 square k~lometers.'~The population density of Colombo District i s many times higher 16,000 than that of the country, and the density in 12,000 Colombo MC is more than 4 times that of even 8,000 Colombo District (Figure 4-8). This leads to the 4,000 question: is Sri Lanka over-concentrated, and if so what are the consequences? Colombo MC Colombo SiLanka 37. There is much evidence to suggest that District national overcrowding i s a serious problem in Colombo, especially in the MC area. The strain on basic 13 Henderson(2000). j 4The total urbanpopulationof Sri Lankais 2.6 mlhon (Census2001) excludingcities intheNortheast. 48 services is evident in electricity, gas, clean water and sanitation, particularly in the poor and underserved areas of the city (see chapter 3). Residential land prices have skyrocketed in recent years, risingmore than 200 percent inreal terms in some areas and more than 100percent ineven some poor areas between 2000 and 2005.15 Census data from 2001 shows highnet out-migration from Colombo to its surrounding districts, especially among unskilled workers.I6 This may indicate costs o f overcrowding in terms o f rising housing costs and access to services that affect the poor more than the better-off. 38. Moreover, the rising trend o f workers living outside Colombo MC where housing is more affordable has also led to massive increases in traffic flowing into the city every day. The population o f Colombo MC more than triples during the day according to some estimates, when more than a quarter of the total population of WesternProvince flows in and out of the city each day (see Annex 4, section II). The economic and welfare implications of such congestion include increased vehicle running costs and lengthening commuting time, which negatively affecting productivity and welfare, andworsening air quality. 39. What is the aggregate cost of overconcentration? Cross-country empirical research mentioned earlier provides a way to estimate losses in economic growth. Using the parameters from the cross-country regression, the optimal urban primacy for Sri Lanka based on its per capita GDP i s estimated to be 24.5 percent (see Annex 4, section HI). Tahng all areas within a 10-kilometer radius from the center o f Colombo MC (a conservative estimate o f the size o f the Colombo urban area - see Annex 4, section IV), the actual urban primacy is as high as 35 percent. Instead, if the much larger Core Area o f the Colombo Metropolitan Regon (CMR) as defined by the Urban Development Authority (UDA) is taken to be the Colombo urban area, urbanprimacy o f Sri Lanka swells to well above 50 percent. 40. Simulations from the cross-country regression indicate that an urban primacy o f 0.35 implies a reduction o f at least 1.5 percentage points in annual average growth rate. Given the limitations o f cross-country analysis, these estimates should be considered speculative." Nevertheless, it i s a useful illustration o f a broader point made in chapter 1: that Colombo urban area is overconcentrated, and that in turn imposes economic losses on the country. And these losses are strictly in comparison to the counterfactual o f having more urban centers as viable destinations for migrants. Wow to mitigate excessive agglomeration in Colombo urban area 41. Section 4.2 shows that a highproportion o f migration into Colombo i s economically driven and originate from poor and conflict-affected areas. This i s likely to continue given the large gap in opportunities between Colombo and the rest o f the country. Given the strong incentive to migrate and the significant economic gains from migration to households, both urban development in Colombo and regonal development to promote alternate urban centers are essential to reduce the costs of agglomeration in Colombo and preserve the economic opportunities affordedby migration to those living inremote, laggingregons. 42. Lessons from economic theory and cross-county experience. Recent literature on Economic Geography suggests viable ways to mitigate the losses due to overprimacy. Urban agglomeration losses (such as congestion) are externalities; so unless there are institutional mechanisms internalizing these costs in migrants' private decisions, urban migration will surely exceed the optimal concentration level. Studies have identified a few factors as critical for this to l5According to Weeratne's independent landprice assessment in 2006. l6See Yoshida and others (2006). l7This analysis is just indicative of overconcentration in Colombo. Many country specific characteristics that may significantly affect the relationship between urban primacy and economic growth are not necessarily incorporated into a cross-country regressionmodel such as this. 49 happen: free-functioning land markets, and a strong role for proactive autonomous local governments andor competitive land developers in city development (Box 4-1).'* A number o f studies suggest inter-regional transport infrastructure as a way to reduce urban concentration by promoting regional development. Investment in hghway systems, inland waterways, and railways has all worked with some success in different countries to spread growth beyond the largest cities (Box 4-1). Investments intransport infrastructure alone, however, are not enough to improve the investment climate and exploit comparative advantages of specific areas. concentration: theory and cross-country experiences Restricting city size in the interests of its residents is one tn~chanisms/iiistitutionto incorporate the estrrnalities of. migration into the Inigrants' private decision to migrate. The collection of rents from land- holding through property tax can provide proper incentives to potential migrants through subsidies or taxes. The deconcentration o f Seoul city in the Republic o f Korea has been successful partly because the government provides proper incentives to migrants. To take such action, local governments mustbe able to control their ownbudgetsinorder to maximize the welfare o f local residents. Henderson and Recker (2000) suggest that decentralization of powers and democratimtioon of local governments is a powerful path to reducing excessive urban concentration. A sinlilar result can be achieved, at Ieast theoretically, through marketforces irt land markets, provided these forces can operate without monopolies or other kinds o f distortions. Henderson (2003) finds that investment in national roads and highway systems significantly reduces national urban primacy. SimiIar results are obtained by Gallup and others (1999), who suggest that historical investments in national navigable waterways induce inland habitation, and significantty reduce urban concentration. Rosen and Resnick (1980) find rail investment reduces national wban concentration. The decoaceatration of industryfrom the greater Sao Paulo region in Brazil to lower-wage hinterland cities followed major transport corridors first through SBo Paulo state and then into Minas Gerais, the interior state with the mainiron ore and other minera1and agricultural reserves. Transportation links may also have an opposing effect (Krugman 1991, 1999). Lowering transportation costs enables firm in large cities to compete with local producers in remote areas. The competitive advantages enjoyed by firms in the nlainurban area may then actually conspire to harm local business in hmterlmds and induce the perverse effect o f M e r concentration in larger cities. The net effect o f investing inbetter connectivity on urbanprimacy depends on which o f the two opposing effects dominates. For Sri Lanka, DO definitive answer is possible at this stage, but there is evidence to suggest that better connectivity will work infavor of spreading growth beyond the mainurban center. So;Jurce:Yoshidaandothers(2006). 43. The insights from cross-country evidence appear to be highly relevant for Sri Lanka. Lack o f adequate transport infrastructure outside Western Province (see chapter 3) indeed is a critical factor behind overconcentrated growth in Colombo and limits to regional development, but it not the only important factor. ADB and World Bank (2005) shows that access to finance, electricity, and transportation are all perceived as major constraints by both urban and rural firms and therefore act as bottlenecks to regional development. 44. The role o f institutionsand markets within cities i s also likely to be critical for reducing the costs o f overconcentration in Sri Lanka. Local governments, which can play an important role in providing incentives appropriate for achieving optimal city sizes, are hampered by lack of finances and capacity." Moreover, weak local governments are unable to provide adequate urban infrastructure and services, especially to poorer localities. Hence, improvements in the quality o f services across the board would mitigate some o f the worst impacts o f agglomeration on welfare and growth. This report suggests a number o f ways to rejuvenate urban functions o f Colombo Metropolitan areas and other urban areas through efforts to improve the functioning land markets and the private sector (see Annex 4, section V). See Krugman(1999); Hendersonand Becker (2000); Hendersonand Wang (2005). f 9See World Bank (2006e). 50 4.4. Summary and implicationsfor policy 45. This chapter profiles internal migration in Sri Lanka, which is a consequence o f increasing regional inequality, to understand its implications for poverty, inequality, and growth. In the absence o f pane1 data, the analysis uses "snap~hot~"o f migrants at different locations, cross- country analysis, and draws from cross-sectional databases. While these information sources are not substitutes for a properly designed panel survey, they reveal some o fthe underlyng dynamics ofmigrationpatterns. Obviously, the design of specific policy responses will greatly benefit from future surveys incorporatingsuch elements into their design. 46. The pattern o fmigration within Sri Lanka is from outlying areas to Western Provmce, with Colombo City as a major destination. The rising trend of intemal migration is consistent with an expanding wage gap between Colombo District and the rest o f the country, in both formal and informal or elementary occupations. Most migrants originate from poor and remote areas or the conflict-affected North and East, indicating that the quest for better economic opportunities is a powerful motive; the ceasefire of 2002 also allowed more free flow ofmigrants fromNortheast. 47. The characteristics o f migrants into Colombo City indicate that the average level of education is much higher among heads o f households who have migrated into Colombo City than residents at their district o f origin, which may cause lagging areas to fall further behind in terms o f their economic potential. At the same time, remittances sent by internal and international migrants to their family members at their places o f origin appear to improve living standards in these areas. 48. Migrants into Colombo City are likely to be better educated than long-term residents, and less likely to be poor or employed in low-paying elementary occupations than the latter by a wide margin. Consequently, migrants are also likely to enjoy better housing conditions and access to basic infrastructure. A large proportion of migrants into Colombo have migrated with their entire household, and these migrants tend to be more educated andbetter employed than those who have migrated by alone. This suggests that much o f internal migration is relatively long-term innature, and that this type o f migration contributes the most in terms o f transferring human capital and slulls to where they are valued. 49. Inthe urbancenter ofColombo, the inflow ofmigrantsadds to the stock ofhuman capital, which strengthens the city's economic prospects. At the same time, continued migration into Colombo i s likely to impose agglomeration costs that can reduce productivity and raise the cost o f doing business. There i s strong evidence to suggest this-skyrocketing land prices indicating scarcity o f land, increasing pressure on urban infrastructure, and a pattern of out-migration o f poor and unslulled workers to the periphery. Cross-country analysis suggests that the aggregate growth impact o f overagglomeration in Colombo urban area i s likely to be negative and substantial. Sri Lanka can therefore increase growth through regional development strategies that promote altemate growth centers-thereby reducing the agglomeration at Colombo and offering alternate economic incentives to potential migrants. 50. While the net welfare impact of migration i s ambiguous, migration does offer upward economic mobilityto those living in lagging regions. But such opportunities are also more likely to be utilized by those who are better endowed interms o f education or slulls, indicating that the absence of such attributes can act as powerful obstacles to the mobility o fhouseholds. 51. The findings therefore suggest a dynamic process at work in Sri Lanka, as a response of individuals to the sharp inter-regional differences in economic opportunities that exist. While being a consequence o f inequality, the pattern o f migration may also perpetuate regional differences, by increasing the endowment o f human capital inthe Colombo area at the expense of lagging regions. Moreover, as continuing migration leads to overconcentration at the country's 51 main growth center, there is an aggregate negative impact on growth and welfare as well. Ifthe trend of nsing regional inequality is not reversed, these processes are only likely to become stronger with time. Implications for a development strategy 52. Migration has the potential to efficiently allocate human resources spatially, and reduce poverty in remote lagging areas. Moreover, in case of remote areas with significant disadvantages, it may be more feasible to improve welfare by empowering people to move rather than movingjobs to people. MigrationinSri Lanka, however, appears to favor skilled or educated workers, even when the potential economic gains (as indicatedby wage differentials) are high.By investing in higher quality education in remote, lagging areas households will be better endowed withthe ability to migrate as a viable way to improve their welfare. 53. Raising the capacity of the rural poor to seek better economic opportunities in urban areas will improve welfare, but that must be complemented by regional development strategies to diversify opportunities for people living in lagging regions, which will in turn reduce overagglomeration in Colombo. Experience in a number o f countries suggest that investing in transport infrastructure can promote growth o f regional markets and urban centers; and the strong correlationbetween accessibility to markets and welfare (inchapter 3) suggest this i s likely to be the case in Sri Lanka as well. At the same time, improvingtransport infi-astructure by itself may not be enough to jump-start the process, and would need to be complemented by other key facilities necessary for market development. 54. Replicating the economic advantages o f Colombo in remote areas of the country is however not the answer. It would be enormously inefficient interms o f resource allocation to do so; and investing in regional markets that are not sustainable due to lack o f critical complementary factors may not benefit even these markets, as numerous experiences in other countries have shown. 55. A more sustainable approach would involve identzhing the comparative advantages of different geographic areas, improving connectivity of these areas to Colombo and other smaller markets, and providing the key complementary facilities needed to exploit these advantages. Potential regional growth centers must enjoy some favorable initial conditions in terms of existing markets, location, basic infrastructure, and institutions. International experience suggests that such initial conditions are often found in towns that are relatively close to (or enjoy better links with) the main urban center, rather than indistant, remote areas o f the country. A careful analysis to identify suchpotential sites would be invaluable to inform a regional growth strategy. 56. Reducing the losses from urban overconcentration also requires coordination between urban planning and regional development. A combination o f integrated urban planning by proactive autonomous local governments, competition among urban land developers, and well- functioning urban land markets i s likely to create interconnecting incentives that allow cities to achieve optimal sizes and improve infrastructureand services. Incase o f Colombo, improvements in infrastructure and services brought about by a better-functioning municipal government can significantly reduce costs of agglomeration arid improve the welfare o f the poor. Developing better institutionswill also likely help unlock the growth potential o f smaller cities. 52 5. Human DevelopmentChallengesand the Poverty Nexus 1. Sri Lanka's early achievements in human development are well known internationally. At the national level, the country is poised to achieve or i s well on its way to achieving most o f the MillenniumDevelopment Goals inhealth and education. An extensive network of schools and healthcare facilities has provided the entire population, including the poor, with universal access to a basic education and health services. As a result, the poor fare as well as the rest o f the country interms o f basic human development outcomes. 2. The challenges that face Sri Lanka inhuman development are quite different from those in most developing countries. Primary enrollment and literacy, which remain challenges for many developing countries, is near-universal in Sri Lanka. Similarly, basic indicators of health-like life expectancy, maternal health, fertility, infant and child mortality rates and immunization rates-are uniformly high in all parts of the country and across income groups. Good basic indicators in health are in no small measure due to high literacy rates among mothers in Sri Lanka. 3. The nationally aggregated social indicators, however, hide quality issues and disparities across income groups in terms of secondary outcomes. Some poor health outcomes, low educational attainment, and lower enrollment rates are more likely among low-income households, which inturn is likely to keep households inpoverty. Inhealth, some rich-poor gaps are prevalent, such as in low birth weight, malnutrition among preschool children, poor nutritional status o f adult women, and incidence o f communicable diseases such as tuberculosis, and diarrhea. Poor nutrition among the poor has important ramifications for health and earning potential, which in turn is likely to perpetuate the cycle o f poverty and low human capability. A healthy mother is an essential prerequisite for the birthof a health baby, and low birthweight and poor health among young children has been found to affect learning in school and also raise the risk o f chronic diseases in adulthood. Both are likely to have adverse impacts on lifetime earnings. 4. The challenge in education in Sri Lanka is low learning achievement. Despite significant investments in primary school enrollment and retention, students display only a weak grasp of first languages, English, and mathematics, and these weaknesses are even worse inthe non-urban sector. Chapters 3 and 4 have discussed the links betweenhigher levels of educational attainment and lower incidence o f poverty. This chapter will show that poverty i s also associated with lower net enrollment rates at secondary and tertiary levels, and the long-lasting impacts o f nutritional deficiencies. Inadequacies in education and nutrition can have lifelong effects on earnings, and trap households in a cycle of low capability and poverty. Private tutoring to supplement formal classroom lectures is more common among the rich, while school avoidance i s more common among the poor. The higher incidence o f poverty among the less-educated and the fact that the poor are less likely to be enrolled in higher levels o f education perpetuates the vicious cycle o f illiteracy and poverty. 5. T h s chapter examines the nexus between inequalities in economic status and nonincome dimensions of welfare. The analysis, drawing on existing data sources as well as a review o f the literature inthe country, will identify some o f the key challenges facing the health and education sectors today, to informthe way forward. Itreviews successes already achieved to ensure that the very same factors that have contributed to achievements in basic health and education in Sri Lanka are notunderminedby measures to address the key challenges facing these sectors today. 53 5.1. Health sector: achievements,outcomes, and challenges 6. Inadequate health care among the poor, who face the greatest burden o f ill-health, remains a pressing issue all over the developing world. Poor health outcomes are more likely among low- income households, which in turn is likely to keep households in poverty. There is little consensus on addressing the health-poverty nexus. Which poverty reduction efforts are most effective in improving the health o f the poor? Which health services should governments provide? One view argues that health services alone are not enough to eliminate the root causes of ill-health that are intimately linked to poverty. Another view argues that poverty reduction cannot be achieved when sickness and disease hamper poor people's ability to work and earn a living.' 7. Sri Lanka's health experience is a microcosm of this complex debate, in that the country has pursuedboth poverty reduction and direct health support. These efforts have worked well to improve basic health outcomes of the poor, but challenges remain along certain dimensions, such as nutrition outcomes among poor women and children, and on the curative side, for treating chronic conditions such as heart disease, mental illness, diabetes, and cancer. 8. The next section presents an internationalcomparisonof Sri Lanka's health outcomes. This is followed by an examination o f how the health system is accessed or used by poor households, and health outcomes, particularly nutritional status, among poor and rich households, and across sectors. Much of the data analysis is based on Sri Lanka's Demographic and Health Survey (DCS, 2002), defining socioeconomic status in terms o f assets or wealth, rather than income or consumptionthat are unavailable fiom the survey.' Profie of health outcomes in Sri Lankawith comparator countries 9. wha compared with other South Asian countries as well as a group o f lower- middle- income countries, Sri Lanka does better in most human development indicators (Table 5-1). Thefertility rate i s very l o w at about 2 births per woman and child; and the average mortality rates among infants, children and mothers, immunization rates and life expectancy at birth are much superior. The prevalence o f childmalnutririon~ rate(percentofpopulationages Adult lite& lower than the South Asian 15 and o\er) average, is more than twice that o f the average for lower- middle-income countries. Health expenditure,public (percent oftotal 10. Sri Lanka's achieves these results with relatively low total spending on health. Studies on the effects of health on productivity include Schultz (2003), Strauss and Thomas (1995), Strauss and Thomas (1998), BehrmanandDeolalikar(1988), HaddadandBois (1991). Households are classifiedinto quintilesbased on their asset mdex usingthe pnncipal components approachdescribed inFilmer andPntchett(2001). See Thalagala(2004) for anotherrecent analysisofSL DHS2000usingan asset index. 54 Expenditure per capita and total health expenditures (sum of public and pnvate) as a percentage o f GDP are less than both the South Asian average and the average for low-middle-income countries. The composition o f expenditures show the important role played by the public health system. Public expenditures make up one-half o f total health spending in Sri Lanka, which IS higher than the corresponding shares in South Asia and lower-middle-income countries. Public health expenditures as a percentage o f GDP i s higher than the South Asian average but lower than the average for lower-middle-income countries. 11. Health sector capacity, as measured by number o f hospital beds per 1,000 (3.1), is comparable to the average for lower-middle-income countries (3-8) and significantly higher than the South Asian average (0.7). However, number o f physicians per 1,000 people (0.5) i s only about a one-quarter o f the average in lower-middle-income countries (1.9). In summary, cross- country comparisons show that Sri Lanka's superior performance in health indicators occurs concurrently with relatively l o w spending on health and low i n ~ o m e . ~ 12. Thepoor in Sri Lanka also generally fare better in terms o f health outcomes Figure 5-1: Sri Lanka's poor do relatively better than South Asian counterparts compared to the poor in other parts of Stunting among children South Asia. Their relatively better health I status can at least partially be attributed to 1 SriLanka universal and free healthcare and a well laid 1 2000 29 out network o f preventive health services Bangladesh and hospitals. As noted earlier, Sri Lanka's /'! 2000 50.5 performance in malnutrition status is below India 1999 47 1 that o f middle-income countries and a f55.6 i -I comparison of stunting, an indicator o f I chronic malnutrition, among children Nepal2001 1 48.4 1 5 9 J reveals interesting contra~ts.~While the prevalence of stunting among children, is FOpulationAverage Poorest 20 Percent lower among the poor in Sri Lanka than Note: Refers to children under 5 years o f age. among the poor in Bangladesh, Nepal, and Source: Tabulations for Sri Lanka based on World Bank staff India,' the difference in the incidence of estimates using DHS (2000). Remainingtabulations taken from Gwatkin and others (2004). stunting within Sri Lanka between the poor and nonpoor i s starker than in any other South Asian country. Stunting among the poor i s double that o fthe average population inSri Lanka (Figure 5-1). 13. Access to safe water and education as-e two areas o f investment that typically improve health outcomes. Interms of access to safe water, Sri Lanka performs better than other South Asian countries but slightly worse than the average lower-middle-income counties. In adult literacy, Sri Lanka i s comparable to even middle-income countries and outperforms South Asia with less than 10 percent o f adult Sri Lankans being illiterate compared with more than 40 percent for South Asia (Table 5-1). The high literacy rate in Sri Lanka, especially among mothers, no doubt contributes to good basic health outcomes. Equityandefficiency ofpublichealthcare 14. Studies using time series data show that between 1952 and 1981, income growth alone could not have achieved the observed positive health outcomes and that public intervention had a HNPStats, World Bank; Annual HealthBulletin 2002 (Department ofHealthServices, Govt. of Sn Lanka, 2002) Stuntmg is defined as prolonged or severe nutrient depletion that eventually leads to retardation o f height or linear skeletal growth in children evident as unusually low height-for-age Compmsons are tabulated using recent rounds of DHS by Gwatkin and others (2004) combined with tabulations from Sri Lanka's DHS 2000 (DCS, 2002) 'CIassificationinto population quintiles based on asset index approach See footnote 2. 55 significant and positive effect.6 Dreze and Sen (1989) described Sri Lanka's approach to development as one of "support-led secunty" in which the government provided health and social sector services as a means o f promoting development without waiting for economic growth to do so. This approach was also taken by Chile, Costa Rica, Cuba, Jamaica, and the Indian state o f Kerala, and in Sri Lanka's case, led to the country being an outlier in terms o f its social achievements given its level o f income.' 15. Byregionaland internationalstandards, the health system in Sri Lanka displays highlevels o f technical eficiency and equity. Total health expenditures (as percent of GDP) in Sri Lanka are comparable to those of Thailand, Malaysia, and Korea; slightly below India, Russia, and the United Kingdom; but significantly lower than the United States (Table 5-2). Like Malaysia, Korea, Russia, and the United States, public spending on health in Sri Lanka make up about one- haIfo ftotal health expenditures. 16. At the same time, the share of spending onhealth intotal public expenditure is lower inSri Lanka than in all countries listed here except India. Moreover, total health expenditure per capita in Sri Lanka is far lower than other lower-middle-income and high-income countries, and comparable to India. The low unit costs suggest that health expenditures in Sri Lanka are comparatively more efficient than other countries given its health outcomes. Nonetheless, areas for improvement include allocative efficiency, quality of health care, and some poor and nonpoor dimensions (Box 5-1). 'Anandfor andKanbur (1991); Anand andRavallion (1993). See example Wang and others (1999) and Shiffman (2000). Sen (1981) showed Sri Lanka to be an outstanding performer insocial outcomesgven its income level and this generatedconsiderabledebate in literature, see Bhalla and Glewwe (1986), Bhalla (1988a and 1988b). 56 ,which results in overcrowding at a few large hospitals. `The delivery of preventive health care and outre~ch,particularly maternalandchildhealth services, is on par with international standards. However, to maintain its past achievementsthe declining expenditures on preventive care has to be reversed. Distribution and ,shortage of critical stag While there continues to be shortages in general surgeons, obstetricians, paramedics, pediatricians, medical teachers, nursing and paramedical tutors, nurses, there is an overproduction of genesdl doctors. This IS particularly problemtie given the comiitment by the government to absorb all medical graduates up to 2010 in the state sector. Although some form of health care is available within 1.4 kilometers of most homes, specialist staff in remote and conflict-affected areas is rare. Alternatively, districts, such as Colombo, Kandy, and Galle, have too many health personnel. According to the Annual Health. Bulletin2002, nearly 35 Dercent of snecialist doctors are in Colombo, - while Kilinochctii, Mullaitivu, and Mnnnar in the North haye none. sozIT(c':Slrllla (3006). 17. Equityin access topublic health services. Access to health facilities isnear-universal. Free public care services are available within a well-developed network of preventive services facilities and hospitals. For example, most people in rural areas live within 5-10 lulometers o f a peripheral health facility, which minimizes travel costs (Hsiao, 2000). In-patient care i s free and access is more or less equal across income groups. 18. Evidence from DHS 2000 further corroborates this. When asked about whom they will consult about a seriously illchild, mothers from rich and poor families are equally likely to report consulting a trained provider. Utilization o f maternal health services shows that women in both rich and poor households have a highuptake o f antenatal and postnatal care (Table A-5.I? Annex 5). Nearly 97 percent o f the births during the five years prior to the survey took place in government hospitals or maternity homes. Bothrich and poor households use preventive care and have contact with health workers (Table A-5.1, Annex 5). For example, on average, most children under age 5 had been weighed about 6 times, suggesting that most mothers comply with the recommended number o f visits (Table A-5.2, Annex 5). The starkest differences are across sectors: among urban, rural, and estate women, a smaller percentage of estate women report visits bymidwife for prenatal or postnatal care (Table A-5.1, Annex 5). Nutritional outcomesand poverty 19. Although there is little rich-poor variation in indicators such as childhood mortality and utilization o f health services, rich-poor gaps show up in l o w birth weight and low nutritional status o f children and women, and the incidence o f communicable diseases such as diarrhea and tuberculosis. It can be argued that these gaps are associated more with structural poverty issues like access to food, child-feeding practices, access to safe water and sanitation facilities, and less with "supply side" factors, like access to andutilization ofpublic healthsystem. 20. Low birth weight i s a proxy for intrauterine growth retardation and indicates that newborns have not attained their hllgrowth potential. L o w birthweight is more prevalent among poor and estate households. Nearly 25 percent o f babies born to mothers in the poorest income quintile were l o w inbirthweight. Mothers livinginthe estates had the highest percent of low birth-weight babies (Table 5-3). 57 21. Child malnutrition. d health status, by wealth Prolonged or severe nutrient depletion eventually leads to stunting, a retardation o f height or linear skeletal growth (unusually low height-for-age measures). Almost 30 percent of preschool children from the poorest income quintile and nearly 40 percent o f estate children are stunted, whereas only 3.5 percent o f children from the richest income quintile are stunted (Table 5-3). 22. Wasting, usually the result Estate 30.0 37.0 12.5 45.7 o f a short-term and acute food Population shortfall (low weight-for-height average 17.4 14.6 14.8 29.3 measures), i s more prevalent among children in poor households, but not any higher for estate chldren than for rural children. The percentage o f children who are underweight or have l o w weight for their age reflects a combination o f children suffering from bothchronic and acute nutritional depletion, and the prevalence o fbeingunderweight is highamongpoor and estate children. 23. Child malnutrition rates are high in North-Western, North-Central, Sabaragamuwa, Uva and Central provinces, while Western province, with the highest level of economic activity, has the lowest poverty rate and lowest prevalence o f child malnutrition. Hence, variation in child malnutrition rates across regions also reflects regional variation inpoverty rates. 24. Nutritional status of women. The intergenerational persistence o f poor nutritional status is evident in the poor nutritional status o f women residing inpoor and estate households (low body mass index). Nearly 40 percent o f poor women and almost one-half of estate mothers have low body mass. Mother's poor nutritional status is an important contributor to intrauterine growth, retardation, and low birthweight. Meanwhile, nearly 40 percent o f rich women and those residing in urban areas are obese (Table 5 4 , a pattern that is becoming increasingly common in many developing countries. 25. Long-term impact of malnutrition. Malnutrition among poor and estate children has a number of consequences for their education, adult health, and earnings. Fetal or childhood malnutrition increases the likelihood o f chronic noninfectious diseases in adulthood. Malnutrition can constrain a child's ability to learn. While it is difficult to discern the impact o f nutritional status on school 58 attainment, since households simultaneously invest in both, studies using longitudinal data show that malnourished children receive less education (see Box 5-2). This may be either because their parents invest less in education or because malnourished children have higher rates o f absenteeism from school due to higher rates o f illness. Poor nutritional status may also delay school entry, impair cognitive development, and potentially reduce lifetime earnings. 26. Malnutrition and exposure to disease. A recent analysis of Sri Lanka's progress toward the MDGs*found that the likelihood ofbecomingmalnourishedinchildhoodwas strongly associated with a household's access to sanitation facilitres (toilets) and safe drinkingwater (pipedwater).' This is consistent with evidence that diseases and infections that prevent absorption o f nutrients or increase dietary requirements are important immediate causes o f malnutrition and that children inpoorhouseholds without water and sanitahon services are at a greater risk ofbeingexposedto gastrointestinal infection and diarrhea (Figure 5-2). 27. Access to safe water and sanitation. Access to protected water supplies and sanitation systems among rural households in Sri Figure 5-2: Diarrhea is more prevalent among the poor I Lanka are better than other countries at a similar income level. Around 17 percent o f households still get dnnlung water from potentially unsafe sources like unprotected wells, rivers, streams, and tanks; and this proportion i s nearly 67 1 1 1 percent for estate sector households. 1 Although the needto boil drinkingwater has been widely communicated, most P5 11 2 0 1 2 0 1 percent percent I poor households appear not to follow 2G t By Wealth this practice." This exposes children in By Sector Quintiles poor households to significant risks o f diarrhea and gastrointestinal infection. Note: Percentageof childrenreportingdiarrheatwo weeks prior to the survey. 28. Poverty reduction measures are Source: World Bank staffcalculationsusingDBS2000. likely to reduce the incidence o f malnutrition in Sri Lanka, since much o f nutritional deprivation is concentrated among poor households somewhat, but measures are needed (Box 5-3). Additional measures to address childhood malnutrition must target food and health inputs as well as appropriate child-care practices, such as breast-feeding and nutritional supplementation o f very young children. In addition to these policies, improved access to clean water and sanitation are World Bank (2005~)Attaining theMillenniumDevelopment Goals inSri Lanka. Other significant correlatesincludemother's education, father's education, child's age, sex andbirthorder (see World Bank, 2005~). loSee Sinha (2006) for evidencefrom DHS 2000. 59 also important in preventing malnutntion because young children are particularly susceptible to infections. Empirical analysis consistently shows that reductions in poverty Iead to reductions in malnutrition. l'he Kenya, Kyrgyz Republic, Morocco, Mozambique, Nepal, Pakistan, Peru, Romania, South Africa and and cross-country regressions drawing on data from 61 developing countries. Household survey howed that a 2.5 percent per aniimi growth in income will lead to a 27-percent reduction in malnutrition by 2015. Only in Jamaica, Morocco, and Peru does such a growth in income lead to a 50- percent reduction in malnutrition by 2015 that is set out inthe MDG on nutrition. The elasticity bemeen income and underweight was estimated at about -0.54, hut this conceals vast vanations across countries.. In - Peru, the elasticity estimate was -1.13 \vhile in South Africa, it was --0.19. In Nepal and Pakistan the elasticity estiniste was 0.77 and -0.3,respectively (liaddad and ohm, 2003). Nutiitioiial is sensitive to income f'or many reasons. Increases in incomes enable laiiiilies to invest more in food and clean water and good hygiene, and access more effective chidcare. Higher Gross National 5.2. Education sector: achievements,outcomes, and challenges 29. Sri Lanka has hadnear universal primary school enrollment (96 percent) and gender parity inprimary and secondaryenrollments as far backasthe early 1990s.Furthermore, primary school enrollment rates are the same across income quintiles (Table Table 5-5: Net enrollment rates bv income auintile. 1995-96 5-5). Education in the country continues to be fi-ee with the government devoting 7-9 percent of its expenditure on education amounting to 3 percent of GDP. Government investment in basic education has also shown good equity, with underprivileged areas receiving higher per student allocation. Quality of education 30. Despite significant investments in the education sector since independence, students display only a weak grasp of first languages (37 percent)", English (IO percent), and mathematics (38 percent) (Table 5-6). These core skills are essential for higher education and securing 11i.e., either Sinhala or Tamil. ' 60 country contributes to the regional imbalances in 5-6: at grade 4, 2003 the quality o f education. The high demand for (percent) teachers in popular urban schools has led to an Mastery skill Sri Lanka Urban Rural oversupply o f teachers in urban areas and a First language 37 51 34 severe shortfall o f teachers inrural schools and, English 10 23 7 in particular, Mathematics 38 52 35 schools in economically Source. National Education Research and Evaluation disadvantaged rural areas where poverty Centre, University o f Colombo. incidence o f private tuition among 600 children according to their income 700 level. Clearly, students from 600 higher-income households have a 500 higher incidence o f private tuition 400 p than their counterparts in low- 300 income brackets. Less than one- 200 third o f children in the poorest 100 income quintile spend on private 0 tuition, while over two-thirds o f Bottom 20-40% 40-60% 60-60% Top 20% children in the richest quintile take private tuition (Figure 5-3)." 34. The above discussion and data underscores two things. First, private tuition has become an essential part o f Sri Lanka's educational process as it enhances the chances o f passing national competitive examinations. Second, since the poor are less able to afford private education this lowers their chances o f success at passing these exams. Indeed, provinces where a smaller percentage of students use private tuition have lower GCE O/L pass 12The only exception is inpreparation for the GCE AIL, when a larger percentage of estate students use private tuition than in urban or other rural areas. One reason for this as suggested by the CSFES is the high premium placed on educational achievement at the advancedlevel in the estate sector inrecent times. 61 rates (Table 5-7). For example, in the North Central province where only 40 percent o f the students used private tuition in2003/04, the GCE O/L pass rate was only 31percent. Incontrast, in the Western Province where 63 percent of the GCE OK. students took private tuition the corresponding pass rate was 48 percent. However, inthe case o f the GCE An examinations, the pass rates among provinces were very similar (around 50 percent), while the corresponding incidence o f private tuition was also high and show relatively little variation among provinces (around 60 percent). Poverty and educational attainment and enrollment 35. World Bank estimates based on HIES 2002 suggest that the poverty jFigure 5-4: Poverty and educational attainment of incidence i s higher among the less- 1householdhead, 2002 educated than the more-educated. Less -- I than 2 percent o f household heads with -s 40 tertiary education were poor in 2002, * I whereas almost one-half o f household 5 30 heads with no schooling were poor -0 0 t 20 I (Figure 5-4). It seems intuitive that a S> person with a higher level o f education i s IO less likely to experience poverty due to 0 greater chances of being employed in better-paying economic activities. 36. Although the country has near universal primary enrollment, there are ,burce:HIES2002. imbalances among the different income groups interms of net enrollment rates at secondary and tertiary levels (Table 5-5). Net enrollment o f children in grades 10 to 13 in the lowest-income quintile i s about one-half the rate for the richest quintile. At the tertiary level, the disparities widen even more: the net enrollment rate for the richest quintile at the tertiary level (13 percent) i s more than six times the rate for the poorest quintile (2 percent). 37. Moreover, 18 percent o f students who enter school fail to complete the compulsory formal education. Most students who drop out o f school come from (a) families living o f f the street, (b) economically disadvantaged areas, (c) conflict-affected areas, (d) the estate sector, or (e) are disabled. In addition, the poor are also more likely to stay away from school. While less than 1 percent of the children inthe richest quintile in the urban sector reported avoiding school in the CFSES 2003/04 survey, the corresponding figure for the poorest quintile in the estate sector was about 6 percent. Regional disparities in education outcomes 38. Like poverty incidence and growth, there are regional disparities in some educational outcomes (Table 5-7). There are wide disparities in the pass rates for GCE O/L and A/L examinations across provinces. While almost one-half of the students in the Western Province pass the GCE O/L exam, only around 30 percent o f students fiom the North Central, Uva, Central and North and East provinces do so. Outcomes in GCE O/L exam also appear to be linked to regional variations inhow much households spend monthly on private tuition. Households in the North and East spend about Rs. 222 monthly on private tuition fees, while households in the Westem Province spend more than double that amount at Rs. 571. 39. Another way o f lookmg at regional disparities ineducation outcomes i s to look at the sum of mean deviations (SMD) for completion o f grade 9, pass rates for GCE O/L and GCE A/L, and tertiary enrollment by province. The overall educational attainment score (i.e., SMD score) is 62 highest in the Western Province, which performs above the national average in almost all outcomes except GCE A/L (Table 5-7). In contrast, provinces with low SMD scores (such as Uva, North Central, and Central provinces) underperform on outcomes such as GCE O/L pass rates and tertiary enrollment. Table 5-7: Key educational indicators, 2002 (percent) Sabara- gamuwa Formal education completion (Grades 1-9) 85 GCE01L Pass 34 GCE A/LPass 57 Tertiary enrollment 9 Sumof mean deviations" (of above Indicators) -1 Monthly tuition expenditure 299 Poverty incidence (percent) 33 da: not available *SMDmeasuresthe overall attainment ofthe province inoutcomes It is calculated asC (X,-XsL)for all z Ito 4 and where XsLis the national average for the relevant outcome High positive score indicates a relatively higher performance relative to the national average while a low value indicates a weaker performance Source CFSES 2003104,Ministry o f Education and World Bank estimates based on Labor Force Survey (DCS) 5.3. Concluding remarks 40. Sustained public investments inhealth and education have clearly led to some o f the most impressive basic human development outcomes globally. Compared to countries with similar GNP andpoverty rates, Sri Lanka performs muchbetter on most healthand education indicators. Sri Lanka's poor tend to have better health and education outcomes than the poor in other South Asian countries. Yet, a few outstanding challenges remain that disproportionately affect the poor. These are important to address because deficiencies in human development have an impact on earning potential over the entire lifetime. 41. Inhealth, rich-poor gaps are prevalent inlowbirthweight, malnutrition amongpreschool children, nutritional status o f adult women, and incidence o f communicable diseases such as tuberculosis and diarrhea. These outcomes are closely associated with poverty because o f the direct link between poverty and the availability of food and water and sanitation. Poor nutrition has important ramifications for health and earning potential, w h c h in turn i s likely to perpetuate poverty and Iow human capability. 42. The relatively highprevalence of malnutrition in Sri Lanka, which i s out o f line with the other health indicators, i s a puzzle. Common correlates o f malnutrition are availability and utilization o f health facilities, female literacy, good hygiene practices and health knowledge, and insufficient access to food. Given that Sri Lanka fares quite well on the first two correlates, the best explanations for the relatively highprevalence o f malnutrition and communicable diseases among the poor are insuflcient access tofood and exposure to unsafe sanitary conditions. Given 63 that malnutrition contributes to the poverty trap by itself, interventions that seek to reduce malnutrition even at the current income levels needto be considered. 43. Current interventions to address malnutrition appear to have been successful in reducing chronic malnutrition among children, but not inreducing acute, short-term undernutrition, and or the prevalence o f low birth weight among infants. More needs to be done to convince mothers, especially in the estates and among poor households, to boil water properly to make it safe for drinking, and to improve access topiped drinking water, and closed wells and toilets with a view to reducingthe incidence o f gastrointestinal diseases. 44. Education is often considered a great "leveler," since it expands the economic opportunities available to poor individuals and households, especially in remote areas o f Sri Lanka. In spite o f near-universal enrollments at the primary level, rich-poor gaps in secondary and tertiary enrollments and rising private tuition usage suggest that poor children are at a disadvantage in acquiring the necessary skills to improve their economic prospects. Further analysis is therefore needed to understand what causes students to drop out even before completing grade 5, since poverty incidence is particularly high for this group. Since the current inequities in education are likely to worsen poverty traps at the household and regional level, Sri Lanka must focus on improving education for the underprivileged, particularly in remote areas and the estate sector where the challenges are most severe. 45. As discussed earlier, teacher deployment is one o f the weakest links in service delivery in education, which has a more pronounced impact on children inmore remote areas. More effective incentives for teachers to relocate to difficult areas have to be developed to address the regonal disparities in staffing in the education sector. Options for decentralizing teacher recruitment to provincial educational authorities need to be looked at to see if they can aid in mitigating the present problem o f weak teacher deployment and perhaps even mitigate teacher absenteeism in rural schools. Although Sri Lanka has done well inproviding universal access to basic education, qualify remains a challenge as evidenced by the poor mastery o f English,mathematics, and local languages o f students. 46. The health system must also gear up to deal with chronic noncommunicable diseases that go hand in hand with highlife expectancy, such as cancer and diabetes. Sri Lanka i s poised to become the third country with the oldest population inAsia, after Japan and Singapore: the share o f the population over 60 years o f age will increase from 9 percent in 2001 to 13 percent in2010 and 21percent in2025.I3 l3Annual Health Bulletin (2002) and Hsiao (2000). 64 6. The Rural Challenge: Raising Agricultural Productivityand Nonfarm Incomes 1. Rural areas are home to nearly 80 percent of the population and about 88 percent of the poor inthe country (Figure 6-1). Even though the rural poverty rate has declined from 29.4 percent to 24.7 percent during the period 1990191 and Figure6-1: Distribution of population(2001) and the poor bv sector, ZOO2 2002, this still translates to Distribution of Poor nearly 3.5 million people. Population Share 2. The slowdown in agncultural growth over the past decade has slowed poverty reduction among agnculture- dependent households. Therefore, measures to raise agncultural productivity and expand nonfarm income are needed increase employment 80% 88% and incomes and reduce Source: HIES (2002). poverty inrural areas. 6.I Agricultural households comprisethe majority of the ruralpoor 3. Agriculture i s an important source o f livelihood in rural Sri Lanka. According to HIES 2002, agriculturalhouseholds (defined as households engaged inagnculture, i.e. deriving any income from crop production, livestock raising and agncultural wage labor) comprise over 60 percent o f rural households in all provinces with the exception of Western Province (Table 6-1). InUva and North-Central provinces, 80-90 percent o frural households derive some income from agnculture. 4. The poverty rate among households engaged in agriculture (24.1 percent) i s significantly higher than nonagricultural households (16.4 percent) (Table 6-1). Conversely, the poorest rural households tend to be more dependent on agriculture as a source o f income. Agricultural farm and wage incomes account for 28 percent o f the incomes o f the poorest 10 percent o f rural households, compared to 7 percent for the wealthiest (Annex 6, Table A-6.1). Table 6-1: Household distributionand povertyrates by sector and province,2002 (percent) Percent ofpoor households by sector 1 Percent share of total households 1 Sector/Province Agriculture Nonagriculture All Agriculture Nonagriculture Rural 24.1 16.4 20.8 58.0 42.0 Region Western 15.0 9.0 9.2 32.9 67.1 Central 24.5 17.2 20.8 65.3 34.7 Southern 24.3 24.6 23.6 69.0 31.0 North-Westem 22.0 23.9 22.3 64.6 35.4 North-Central 19.0 17.6 18.1 80.3 19.7 Uva 34.3 16.9 31.8 89.6 10.4 Sabaragamuwa 30.5 28.4 28.9 I 62.0 38.0 65 Average annual growth Share in agriculture,forestry, and rate of output value fisheries output value Subsector 82-90 91-00 95-02 02-04 TE85 TE95 TE04 Agriculture, forestry and fisheries 1.7 1.9 1.4 0.4 33 28 22 Agriculture 2.8 1.6 1.6 0.9 24 19 16 Tea 2.9 3.4 4.8 -0.3 4 1 1 Rubber -1.2 4.3 -4.7 1.8 1 1 0 Coconut -3.8 2.5 1.3 3.1 3 2 1 Rice -0.3 -0.7 3.7 -4.5 6 4 3 Other 6.3 1.9 0.9 2.6 11 11 10 Forestry 6.1 3.3 1.1 1.3 2 2 2 Fisheries 0.7 6.4 4.1 -2.7 2 3 2 Agricultural wage worker Nonagricultural wage worker All wage workers Sector/Province 92-03 92-95 95-02 92-03 92-95 9 5 4 2 92-03 92-95 95-02 Rural 2.6 5.6 1.5 3.2 6.7 2.4 3.1 6.5 2.4 Provinces Western 1.9 2.5 0.1 3.5 5.2 3.1 3.5 5.2 3.1 Central 2.5 5.8 1.4 3.1 11.1 1.5 2.7 8.9 1.1 Southern 2.4 8.6 0.6 3.2 4.3 1.9 3.0 4.0 2.1 North-Westem -0.1 6.3 -2.5 1.7 6.2 0.1 1.3 5.7 0.3 North-Central 0.8 2.0 2.4 1.o 2.4 0.4 1.4 5.9 0.5 Uva 0.7 -0.4 2.1 2.1 10.6 0.3 2.1 6.7 1.9 Sabaragamuwa 1 0.8 1.3 0.3 I 2.2 5.0 1.9 I 2.0 4.7 1.5 Source: Staff calculations using Sri Lanka Annual Labor Force Survey data. I I 8. The average annual growth rate in wage earnings among agncultural wage workers was also below those of nonagncultural workers (Table 6-3). InUva and Sabaragamuwa, the provinces with the highest 66 poverty rates, the average annual growth rate in agncultural wage earnings were about one-half that of nonagriculturalwages. Challengesin raising agriculturalincomes 9. The increased agncultural productivity and competitiveness needed to raise incomes has been hampered, albeit unintentionally, by broad government interventions in agricultural commodity and factor markets intended to protect the interests of the farming population Policies on trade, marketing, agncultural technology, land and waters have unintentionally squeezed returns from agmultural production, limited productivity and income-enhancing investments, held back diversification to higher value activities, and "pushed" many out of agriculture into low-paying, insecure, casual nonagncultural wage labor. 10. Agricultural technology policy. Access to productivity-enhancing technologres by farmers has been constrained by restrictive seed and phyto-sanitary policies and the weakening o f the agricultural research systems and extension services. As the cross-country comparisons inTable 6-4 indicate, crop yields in Sri Lanka have considerable room for improvement, and excessive regulation may be serving more as a barrier to entry than as an environmental filter. Many requirements that are outdated and inadequate to meet rapid advances inresearch and technology occurring worldwide are also subject to costly permit and inspection procedures. Moreover, delays in releasing revised seed and phyto-sanitary regulations increases uncertainty about requirements for planting material imports and marketing. Table 6-4: Average yi ds of selecte Commodity Sri Lanka China I India I Indonesia Pakistan Thailand Vier Nam Rice 3,394 6,170 3,006 4,517 2,988 2,676 4,694 Maize 1,103 4,964 1,874 3,241 2,319 3,656 3,333 Groundnut 585 2,904 975 2,013 971 1,512 1,678 Soybean 1,049 1,781 865 1,270 1,250 1,356 1,317 Potatoes 16,543 15,462 18,555 18,555 15,051 12,054 13,159 Eggplant 6,882 18,63 1 16,146 7,034 10,556 5,826 Cabbage 13,889 18,743 21,330 20,334 14,856 11,011 17,972 Chilies 2,888 19,160 9,182 3,808 14,000 Tomatoes 7,593 26,121 14,789 12,678 9,964 26,095 Sugarcane 56,966 65,376 62,73 1 69,710 48,378 68,862 54,215 Pepper 623 1,530 237 718 3,093 1,733 Tea 1,450 863 1,690 1,405 295 1,011 Coconut 4,271 10,099 4,809 6,013 4,000 4,327 6,804 Rubber 798 1,292 1,596 896 1,623 876 Source: Food and Agricul -e Organization statistical database. 11. Agricultural research is performed by a large number o f government institutes,' which have been relatively successful inraising the productivity o fme, but less successful inraisingproductivity for other crops. According to past reviews, the public agricultural research system had been almost exclusively focused on rice, makes little use o f socioeconomic or financial analysis, and i s highly fragmented (Charles, 2002). Although the government established a number o f priority setting, planning, and competitive grant research funding schemes, the complicated procedures for accessing the grants deterred applicants from applying. Moreover, private sector investment in agricultural research in Sri Lanka has been hampered by the absence o f intellectual property rights protection, restrictive seed and phyto- sanitary regulations and procedures, and the subsidized sale o fplanting materials. 12. Further improving the effectiveness o f the agncultural research system inthe future as articulated in the National Agricultural Research Policy (2003) will require fostering a pluralistic national agricultural research system, including the government, private sector, NGOs and other agencies. Existing public agricultural research institutions need to strengthen demand orientation and improve the quality of 'These includethe DepartmentofAgriculture, plantationresearch institutes(tea, coconut, rubber), and severalnationalinstitutes reportingto the Councilfor Agricultural ResearchPolicy and universities. 67 research activities through greater participationby farmers and other stakeholders inprogram governance, priority setting, and the evaluation o fperformance. 13. Agricultural extensionservices have been severely weakened since they were devolved to Provincial Councils in the early 1990s under the 13* Amendment to the Constitution. Most field-level agncultural extension workers were reassigned as village facilitators, effectively eliminating their role as disseminators of agncultural information (Tabor et al., 2000). Analysis o f the SLLS data (1999/2000) found that only about 13 percent o f agricultural households report receiving technical assistance fiom a government extension agent (15 percent from all sources).* In 1999, the Department of Agriculture began piloting "fee-based" agricultural extension services as part of the Second Perennial Crops project. By design, however, this approach concentrates only on larger commercial farmers and enterprises. Its applicability to small holders may be more limited.3 14. The effectiveness o f the agricultural extension system could be improved by (i) expanding the supply of extension services through government sub-contracting to private firms, NGOs, and producer organizations; (ii)strengthening client orientation through adoption of participatory approaches in planning and implementation; (iii) providers to multiple sources o f innovation (research and linking others); and (iv) expanding the use o f new information and communications technologies to deliver a wider array o f information of value to farmers throughnew, innovative channels. 15. Land policy. Another constraint Figure 6-3: Distribution of number of owned agricultural holdings and to raising agncultural incomes is the area by farm size, 1982 and 2002. increasing proportion o f farmers dependent on very small landholdings `0 a 80% coupled with slow growth in 70% agricultural productivity. Analysis o f 58 60% the land ownership structures fiom .E E 50% the Census of Agriculture 1982 and 40% I0 2002 highlightthe significantjump in c0 30% the number o f farmers with an 5s0 20% 10% agncultural land holding o f less than `1 0% 1 acre (0.40 hectares). By 2002, 63 cL u) percent o f owned agricultural B holdingswere less than 1acre (Figure 6-3). Another 17 percent owned holdings less than 2 acres (0.8 0 4 acre 01- c2acres 2 c 20 acres - hectares). Such small farm sizes severely limit income-generating Source: DCS (1987) Census of Agriculture 1982, DCS (2004) Census of Agriculture- Sri Lanka 2002, Preliminary Release No.2. potential, especially if farmers are confined to only growing low value crops (such as rice). 16. Existing land legislation limits the efficient functioning o f land markets. A critical feature o f Sri Lanka's land ownership pattern i s government's ownership o f large shares of land. Agncultural land in Sri Lanka totals about 2.79 million ha (1.72 million ha owned by the state and 1.07 million ha owned privately) (World Bank, 1996). Lands transferred to farmers through various land settlement programs See Annex 6, Table A-6.3. 3Two private firms were contracted to provide technical assistanceinproject planning, design, and implementation o f commercial agricultural activities funded under the project (the fees charge is set at 5 percent o fthe project costs). 68 began in the 1930s but the most important of them was the Land Development Ordinance (LDO) of 1935.4 17. While these legislated government programs succeeded in promoting greater equity in land ownership, their highly restrictive nature hurt farmers in several ways. Land obtained through an LDO carries restrictions on mortgaging that preclude its use as collateral to access credit that households could use to finance both income-enhancing farm and nonfarm investments. For those wanting to remain in agnculture, the small landholdings, the lack o f secure property rights, and the legal restrictions on buying or leasing-in LDO land reduce incentives for productivity-enhancing investments. Those interested in shifting out o f agnculture into nonfarm activities, or merely moving to another location would have to leave without compensation for their land. In addition to fostering a large cadre of part-time farmers, these legislative provisions limit the ability of the land market to allocate land to its best use (World Bank, 2003). 18. The Land Development Ordinance (LDO) is inthe process o f being amended to address some of the land issues discussed above, and concerns about inefficiencies that can arise from the procedures laid out in the Bill and the absence of clear criteria for approving or disapproving applications.' The prompt processing o f the amendments to the LDO, including these concerns, would be essential to enhance agnculture productivity. 19. Wuterpolicy. Duringthe past five decades, irrigation development has served as a major pillar of the government's agncultural strategy to promote agricultural growth and rural development, and enhance food security. Most of these investments focused on the construction o f new dams for power generation and surface irrigation systems, which were also closely linked with the government's land resettlement program. The most important program was the Mahaweli Project initiated inthe late 1970s and managed bythe MahaweliAuthority o f Sri Lanka (MASL). 20. The longterm sustainability o fhuge investments such as expanding surface irrigation infrastructure is threatened on several fronts. Inadequate priority-setting and funding o f operations and maintenance (O&M) has led to the rapid deterioration o f canal systems and to poor quality o f services. This explains the need for repetitive and costly rehabilitation every five to six years. Institutional weaknesses in the water agencies combined with minimal involvement o f farmers impeded steady improvements in the quality o f and "user-orientation'' in service delivery. Poor reliability o f water delivery and the frequent lack o f access to water by tail-enders, combined with lack o f access to agncultural extension and improved technologies, contributed to low crop yields. Inadequate farmers' involvement in decisionmalung regarding water delivery, both in terms o f quantity and timing, has constrained their ability to switch away fiom rice or diversify to include higher value crops.6 Providing water for free also reducedthe incentive for farmers to save and use water efficiently. 21. Ensuringthe efficient and effective performance of irrigation systems will be important, not only to sustaining agncultural productivity growth over the longer term, but also inpreventing a large proportion of farmers reverting to rainfed agnculture and the associated increased production risks and income vulnerabilities associated with it. Recent estimates using SLIS 1999100 find that the landless, margmal and small farmers account for 59 percent o f total gross irrigated area (GVL).~ About 85 percent o f gross irrigated area in Sri Lanka is supplied with surface (major and minor) irrigation. Of these, landless, marginal and small farmers cultivate 26 and 42 percent of total GIA under major and minor schemes. Other key legislation included the Sale of State Lands (Special Provision) Law 1973, Land Development(Amendment Act) 1981, Agrarian Services Act 1979, Land Reform Law 1972 and 1975, Land Reform (Special Provisions) Act 1981, and the Agrarian ServicesDevelopmentAct No. 46 2000 5For example, people obtaining new landallocationsunder theLDOwill still berequiredto obtainapermitfirst. Currentwater delivery schedules are still designedfor paddycultivation. 7Marginal farmers are those who cultivate less than lacre; small farmers are those who cultivate 1to 2 acres and mediumfarmers are those who cultivate 2 to 4 acresofland. 69 (Annex 6, Tables A-6.4 and A-6.5). Including medium-sized farmers raises their share o f GIA to 85 and 81 percent, respectively. Effective action will necessitate (i)prioritizing expenditures toward rehabilitation and maintenance o f existing infrastructure; (ii)fostering greater user participation in managing systems and recovering costs; and (iii) re-orienting and restructuring water institutions to ensure efficient and client-oriented operations. 22. The Government is now faced with tightening intersectoral competition for water among various users (agnculture, drinlung water, industry, etc). Therefore, as it moves toward finalizing the National Water Policy and National Water Bill, priority would need to be gwen to: (i) promoting the shift from supply-driven goals to comprehensive planning, allocation, and management within a river-basin framework; (ii) formulating an appropriate regulatory framework and reprioritizing expenditures to support such a shift; and (iii)reforming institutional structures and procedures, building on increased participatory management of systems, to improve the management o fwater resources inSri Lanka. 23. Agncultural tariffs in Sri Lanka are remain highand are subject to frequent change. The government intermittently lowers the tariffs for major agncultural imports through duty waivers and controls import volumes through licensing during months when domestic prices rise. These frequent changes create considerable uncertainty, heightening price risks for farmers, consumers and local entrepreneurs, and greatly dampen incentives for private sector investments in storage. The high tariffs on agricultural commodities have also raised the cost o f these products for consumers, with associated impacts on consumer expenditures andpoverty Ievels (World Bank, 20060. 24. Phasing out tariff protection for various agncultural commodities gradually over the medium term would (i) the bias in favor of particular crops (e.g., rice, potatoes, chilies, onions) and thus allow reduce improved domestic resource allocation; and (ii) reduce the taxation o f consumers who pay above-world- market prices. The phased reduction in tariff protection will need to be accompanied by parallel policy changes, especially measures to lift the constraints on domestic, commodity and factor (land, seeds, technology, and water) markets and to improve rural infrastructure. These complementary actions will help ensure farmers the freedom and the capacity to alter their resource-use decisions to meet the changmg needs o f the market. 25. A strong commitment to removing policy and regulatory restrictions i s necessary so that those who choose to remain in agriculture can raise their productivity and incomes. In the short to medium term, policies are needed to ease farmer access to improved technologies, create a more transparent and stable trade policy regme, allow full and transferable ownership rightsto land, and ensure sustainable water use. Adopting policies to speed up currently laggingprivate sector participation and investments would also be critical to promoting growth in both the agncultural and nonfarm sectors. These include promoting a regionally equitable development strategy for rural infrastructure and services development with increased emphasis on operation and maintenance of physical assets to ensure their longer-term performance. 6.2 Increasing importance of nonfarm income 26. The nonfarm sector has increasingly gained importance as a source of income and employment in rural areas. The nonfarm sector generated 67 percent o f rural employment in 2003. One-quarter o f those employed in the sector are self-employed in nonfarm enterprises, and the remaining work in nonfarm salaried or wage work. Not only does the sector generate a large number o f rural jobs, but it also contributes substantially to household income. More than one-half (52 percent) o f the per capita income of the average rural household comes from nonfarm earnings. Agncultural households are also highly dependent on nonfarm incomes-the share o f their income from nonagricultural activities (41 percent) exceeds the contribution o f agricultural incomes (32 percent).8 ~~ See Annex 6, Table A-6.6 70 27. The relative importance o f agricultural incomes, however, differs considerably by type o f fanner. Agricultural incomes account for over 40 percent o f incomes for farmers growing higher value crops like tea, rubber, fruit, and vegetables. For paddy farmers, only 32 percent of income comes from agriculture,' so these households are more likely to rely on nonfarm activities to supplement income. 28. The growth of the rural nonfarm sector has significantly contributed to the reduction inrural poverty. The poorest rural households are heavily dependent on nonfarm incomes. Among the poorest 10 percent o f rural people, the average household derived about 44 percent of their per capita income from nonagriculturalwages, salaries and enterprise earnings (Table 6-1). A s mentioned earlier, the incidence o f poverty among rural nonagncultural households i s considerably lower than that among households engaged in apculture. Data from the HIES suggest that ownership o f and/or employment in a rural nonfarm enterprise are associated with significantly higher welfare. Among rural households owning and operating a nonfarm enterprise, the poverty rate i s 13 percent, compared to a poverty rate of 23 percent for households without a nonfarm enterprise. Households operating a nonfarm enterprise have monthly per capita incomes 20 percent higher than those who do not. 29. Incoming years, the ruralnonfarm sector will have to generate a highnumber o fjobs for the growing labor force. In2003 the size o f the labor force was roughly 7.2 million workers. About 12 percent of workers are employed in urban areas, 82 percent inrural areas, and the remaining 6 percent inthe estate sector. Eachyear nearly 106,000 people enter the labor force in Sri Lanka and many of them will continue to live in rural areas." Since the potential for employment expansion in the agncultural sector i s somewhat limited, expansion of employment and productivity in nonfarm activities will be essential to absorb this growing labor force. Profile of the rural nonfarm enterprise sector 30. According to recent estimates, approximately 620,000 rural nonfarm enterprises are scattered throughout the country. Most o f these rural enterprises are involved in productiodmanufachg (41 percent) or trading (38 percent), with a far smaller proportion in services (21 percent). Approximately 10 percent are engaged in manufacturing and the sale o f processed agricultural goods. Other rural manufacturing industries include garments, nonmetallic mineral products, furniture, and wood products. Most trading enterprises are engaged in selling processed (65 percent) and unprocessed apcultural products (57 percent), and only about 5 percent sell agncultural inputs. The main service-related enterprises include repair services (24 percent o f service-related enterprises), followed by personal services at 17percent (including barber shops, beauty salons, etc.) andhotelshestaurants (14 percent). 31. The typical rural enterprise employs about 2.4 workers, including hired workers and family members. Almost one-half the workforce inrural enterprises is comprised of family labor (60 percent o f workers in trading enterprises, 50 percent in services and 42 percent in production are family labor). Production- based enterprises tend to be larger with 3 workers on average while traders work alone or partner with one other person. Only 6 percent o f rural enterprises have more than 5 workers. The average rural enterprise has been in operation for 9 years. More than half (59 percent) o f these enterprises operate as stand-alone establishments, with the mainplace o fbusiness found outside the family homestead. Constraints to rural nonfarm enterprise growth 32. The potential for increasing incomes and reduce poverty in rural households through the non-farm sector is hampered by a number o f obstacles. A recent survey o f the investment climate inSri Lanka, rural entrepreneurs identified poor quality and availability o f transportation, poor access and high cost of finance, limited access and unreliable supply o f electricity, marketing difficulties, and poor coverage in telecommunications as the primary constraints to doing business (Figure 6-4). See Annex 6, Table A-6.6 10Between 1998-2003, 58 percent o f the new entrants lived inrural areas. 71 Figure 6-4: Constraints rated as major or severe problems by rural entrepreneurs investment Climate ConstraintsFaced by Rural Entrpreneurs Transportation Cost of Fimsnang Accessto Fimnclng Low demnd for goods and services Electriaty Lack of m k e t Informttron Telecom Skillsand Educationof Works5 B U S ~ ~bSaSi n g and Operattng Permts s Ecowmc end RegulatoryPolicyUncertainty I I Tax R a t s Access to land Labor Regulations Legalsyst&conflid resoIut1on Tax Admntatration corruptton Cwtorrsand Trade Regulations Crime theft and disorder 0% 5% 10% 15% 20% 25% 30% 35% Percent of firms citing constraintas major or severe Source:ADB & World Bank(2005). 33. While the ranking of constraints may differ by province, the top 7 constraints identified as major business obstacles are consistent across provinces (Annex 6, Table A-6.7). The relative importance of the obstacles faced also varies by type of enterprise. For example, electricity was identified as the most important obstacle by production and service enterprises (31 percent and 27 percent) as compared to trade-related enterprises (19 percent), which seem to be more affected by lack of market demand (16 percent) and financial infrastructure (14 percent). Separating firms by age suggests that financial infrastructure, lack of market demand and, to a lesser extent, road access are o f greater obstacles for start- upsthan for older and well-established enterprises. These factors, especially finance androad access, are also perceivedas more important constraints by small stand-alone enterprises as compared to larger ones. 34. The obstacles identified by rural firms have a negative impact on productivity, the level of investments made by existing firms, and the number o f new start-up enterprises. The biggest obstacles to productivity by firms in order o f priority include electricity, financial infrastructure, market demand and information, and road access. Evidence also suggests that larger rural enterprises are better able to deal with constraints imposedbypoor infrastructure than smaller ones (Jin et al, 2005). 35. Linkage between the business constraints andpoverty incidence. Since the nonfarm sector constitutes a significant part o f the rural economy, the constraints for nonfarm enterprises may be closely associated with poverty incidence. Infact, t h s is the case inSri Lanka for most o fthe business constraints identified above. As shown inchapter 3 (Table 3-71, better access to markets and banks, proximityto Colombo, and higher penetration o f electricity are all closely associated with lower poverty incidence, with the exception o f telecommunication coverage. Even though business communities inrural areas identify poor coverage in telecommunications as a constraint for rural nonfarm enterprises, there is no clear indication that poor coverage i s associated withhighpoverty incidence. 36. Transport. On average, enterprises are locatedwithin 10kdometers o f the nearest city or market, but because o f the poor road conditions travel is slow, so that the average travel time to the nearest commercial center i s more than 30 minutes. The average distances as well as travel time to the nearest commercial center are much higher for the poorer provinces -namely, Uva, Sabaragamuwa and Southern -comparedtothecountryaverages(Table 6-5). Roadconditions citedbyruralentrepreneurs asmajoror severe obstacles to doing business include road quality (36 percent), access to roads (33 percent), and 72 absence o f available transport (32 percent) (Table 6-5). In addition 29 percent o f rural enterprises are located in communities without public transportation to the main market and about 13 percent are in communities without public transportation to the nearest city. This greatly contributes to marketing difficulties as 47 ofrural entrepreneurs donot own their own vehicles. 37' Finance. The high Of andPoor to finance i s also a key constraint to rural Table 6-5: Transpod facilities 2nd rural Average distance Average travel time to enterprise performance and growth. Close to to nearest nearest commercial commercial center center by main means of Province (km) transport (minutes) Western 5 23 Central 5 22 Southern 13 41 North Western 6 20 North Central 7 29 Uva 16 70 Sabaragamuwa 18 59 NorthEastern 5 23 Total 8 31 access to equity capital as a major or significant constraint (as cited inWorld Bank, 2005b). 38. Most rural firms have limited access to financing from private commercial banks. Less than 12 percent o f rural firms apply for loans from these institutions compared to 41 percent from state commercial banks and Samurdhi. The smaller the enterpnse, the less likely it is to access financing from private commercial banks. Access to formal finance is especially restricted for investment purposes with internal sources (cash inhand) providing the biggest share o f investment (43 percent), followed by family and fiiends (35 percent). Trade finance appears to be an important source o f working capital with nearly 31 percent o f rural enterprises purchasing goods on credit. The share of financing from moneylenders i s relative limited, with rural enterprises primarily using this type o f financing for liquidity management purposes. 39. Collateral i s often critical to the availability o f finance. Almost one-quarter of loans to rural enterprises require collateral, and land is offered in 75 percent o f these transactions. When loan applications by nonfarm rural enterprises are rejected, lack o f collateral i s by far the most important reason (and the obstacles to using land as collateral was identified earlier in this chapter). While the government has taken steps to strengthen the enabling environment for financial institutions to expand financing to small enterprises, a number of regulatory and institutional constraints and market failures hurdles remain. The most critical regulatory weaknesses and institutional problems prkventing lending to small enterprises include: (i) deficiencies in debt-recovery legslation; (ii) incomplete land ownership and titling; (iii) absence o f efficient charge registries over movable collateral; (iv) limited availability o f credit information; and (v) a weak regulatory framework for asset-backed securities. Avenues through which access to financing opportunities for small rural businesses could be expanded are addressed in World Bank (2005d). 40. Lack of market information and isolation from supply chains are also significant barriers to the success o f small rural firms. About 27 percent o f rural enterprises complained about low market demand and 11 percent complained about lack of adequate market information. Inadequate market information can lead to businesses selling in local markets where prices may not be optimal or missing opportunities in markets where growth prospects are greater. The ADB survey on SMEs (2003) found that nearly 60 percent o f the owners viewed market opportunity as a serious barrier to their growth. Inthe investment climate survey less than 2 percent of all rural enterprises used any type of marketing assistance to sell their goods and services. A vast majority o f firms sell their goods directly to consumers or traders intheir own district and only a few sell to multinational parent companies or larger urban firms. Less than 10 73 percent o f all rural firms report selling under subcontracting arrangements that potentially would provide them with access to wider markets. Long and fragmented supply chains constitute another major challenge. Very few firms appear to be integrated into well-coordinated supply chains, limiting access to newtechnology, financing options, andwider markets. 41. Evidence fi-om other countries suggests that participation in business organizations and local chambers o f commerce could potentially strengthen marketing channels allowing businesses to share information on prices and quality standards, and obtain technical, financial and organizational services. In S n Lanka, however, participation in business associations i s limited among rural entrepreneurs. Only about 8 percent o f enterprises are members of any form o f business association, and 4 percent are members of a chamber o f commerce. 42. Electricity. Almost one-quarter o f rural entrepreneurs cited electricity as a major or severe problem and another one-quarter single it out as their most important constraint. Access to electricity remains a challenge in many communities, and even among enterprises with access, problems in reliability of supply and high prices limit growth. Electricity is heavily concentrated in urbanized areas such as Western Province, which has over 80 percent coverage. Rural areas such as Uva province are grossly underserved, with less than 40 percent coverage. Poor power supply exposes firms to frequent outages increasing production costs and tying up significant capital in back-up power production, resources that could be productively engaged in their core business. Just below 70 percent o f rural enterprises use electricity from the national gnd. The vast majority of these firms experienced power surges or outages; with one-half o ffirms reportingthe longest power outage ina typical monthas lastingmore than 2 hours. 43. Addressing the power problem will give rural businesses a strong boost-particularly the rural manufacturing sector, which i s the segment that generates more jobs than trade or service-oriented businesses. In the medium term, increasing electricity access in rural areas will require a number o f different measures that are outlined indetail inADB & World Bank (2005). 6.3. Concludingremarks 44. About 60 percent o f rural households in Sri Lanka are agncultural households. The poverty rate among rural agncultural households is significantly higher than that for nonagricultural households, and the poorest rural households are more dependent on agnculture as a source o f income. Poverty reduction inrural areas has been hampered by a slowdown inagncultural growth and slower growth inincomes in rural areas. 45. The nonfarm sector has become an increasingly important source of income and employment inrural Sri Lanka. The nonfarm sector generated 67 percent of rural employment in 2003. In the average rural household, more than one-half o f the income comes from nonagncultural wages and earnings from self- employment in nonfarm enterprises. Rural households that own and operate a nonfarm enterprise have a significantly lower poverty rate compared to those without a nonfarm enterprise. 46. Raising agricultural productivity and expanding nonfarm income opportunities are key to reducing rural poverty. To foster agricultural productivity growth, policy and regulatory restrictions need to be refashioned to ease farmers' access to improved technologies, create a more transparent and stable trade policy regime, allow full and transferable ownership rightsto land, and ensure sustainable water use. 47. The most serious obstacles to the development and growth o f rural nonfarm enterprises include poor transportation, problems accessing finance and the cost o f finance, poor access and quality of electricity supply, lack o f market information and isolation from supply chains. Given the importance o f nonfarm income for rural households, these constraints would explain why certain location-specific characteristics (like accessibility and electricity use) correlate well with the pattern o f regonal poverty in Sri Lanka (Chapter 3). As these constraints in rural areas are lifted, the productivity and growth of businesses are likely to improve and more start-up enterprises are likely to emerge. 74 48. Initiatives to strengthen and link producer organizations and business associations are likely to improve access to markets and marketing information for rural entrepreneurs. Finally, strengthening the regulatory framework for contract enforcement and making lower-cost dispute resolution easier and more available will help raise pnvate sector participation and investments inrural areas. 75 7. Social and Economic Situation inthe Conflict-AffectedNorthernand Eastern Provinces 1. The economic and social repercussions o f over two decades of conflict have affected people throughout the country. Over 65,000 people have died, nearly a million citizens have been displaced, private and public properties and economic infrastructure have been destroyed, local economies and community networks have been disrupted, and health and educational outcomes have deteriorated in districts in the North and East. The macroeconomic impact o f the conflict i s estimated at 2-3 percent o f GDP growth annually.' 2. The repercussions o f the conflict have fallen particularly heavily on Northern and Eastern provinces. These two regions have fallen behind the rest o f the country interms o f health and education outcomes and access to economic infrastructure and financial services. Moreover, the districts o fAmpara and Batticaloa inthe East sustainedthe most damage from the tsunami inDecember 2004. Measures to reduce regional inequalities must take into account the unique circumstances o f these two provinces. 3. The cessation o f hostilities, following the Cease-fire Agreement signed in February 2002, has stimulated economic recovery. Real GDP growthincreased fourfold to 12.6 percent inNorthern province and doubled to 10.1 percent in Eastern province.' Unemployment significantly dropped from 13 to 9 percent in the North and from 16 percent to I O percent in the East during 2002-04, while economic growth and unemployment inthe rest of the country remained relatively flat. 4. Despite the promising upturn in the local economies o f the North and East following the 2002 ceasefire, significant obstacles to sustained growth and poverty reduction in these areas remain. These include: (i) poor availability and access to financial services, (ii) access and quality of economic poor infrastructure (roads, telecommunications, and water), (iii) restrictions on the use of the A 9 highway, (iv) fishing restrictions, (v) limits on mobility in certain areas such as Jaffna, and (vii) out-migration o f the educated to the rest o f the country or abroad. 5. Peace dividends from the Cease-fire Agreement include the recent expansion o f the Samurdhi program to the Northem districts (starting with Jaffna and Mannar) and a national program that provides cash transfers to the poor (see Chapter 2). Additional analysis is needed to identify key constraints to raising incomes and improving living standards, but such analysis remains difficult due to the paucity o f household survey information with adequate coverage and representation for this region. 6. This chapter presents the economic and welfare situation in the North and East using available sources of survey data and smaller studies conducted by different organizations. Fouryears o f cease-fire has provided the first opportunity in decades for national surveys to include data fkom Northern and Eastern provinces. However, due to uncertainand volatile security circumstances inthese areas the HIES 2002 survey was not completed and the CFSES 2003-04 was completed but with uneven geographic coverage. The key data sources used in this chapter and their limitations are listed in Table 7-1. While the available data do not allow measurement of poverty incidence or identification of its proximate causes, they show how the region compares to the rest of the country along various dimensions o f welfare. The chapter also places these indicators o f welfare inthe broader context of the economy of the regon, and developments brought about by the cessation o f conflict since 2002, and major challenges ahead. * Central Bank o 1 Bank of Sri Lanka, Annual Report 1998. Central f Sri Lanka as it appears inPeace Secretariat Impact o fthe Cease-fire Agreement on Regional Economic Growth in Sri Lanka (2004). 76 Table 7-1: Main data sources i Indicators IYear(s) Data Source Key Limitation Provincial GDP growth 1997- Provincial GDP tabulated bythe Central Islandwide uniform 2002 Bank o f Sri Lanka (CBSL) as they deflator used appear inPeace Secretariat (2004) Provincial GDP and sectoral 1990, NationalPlanningDepartment as they Islandwide uniform growthrates 1995 appear inSarvananthan (2004) deflator used Labor force andunemployment 2002, Department o f Census and Statistics 2002 excludes Mullativu trends 2004 (DCS) Labor Force Survey and Killinochchi in Northern province Per capita incomes, 2003104 CBSL Consumer Finances and Socio Excludes Killinochchi, consumption, housing, Economic Survey (CFSES) Mannar, Mullativuin education outcomes, access to Northern province healthhanitation Provincial education mastery 2003 University o f Colombo National Outcomes cannot be skills Assessment o f Grade 4 Achievement reported separately betweenNorth and East Impact o fthe tsunami 2005 DCS Census o fTsunami-Affected Areas Availability, access, and 2003 World FoodProgram food insecurity Relies on secondary data utilizationo f food study Vulnerability rankings within 2004, Centre for InformationResources Cannotcompare results districts o fNorth and East 2005 Management Vulnerability Poverty across districts Profile (VPP) I II 7.1. Peacedividendsfollowing thecease-fire in 2002 7. The civil conflict reached a positive turning point in February 2002 when the Government of Sri Lanka (GoSL) and the Liberation Tigers of Tamil Eelam (LTTE) rebels signed a Memorandum of Understanding(MoU) resulting in a cease-fire. This has been the longest truce between the two sides to date. The economic embargo imposedby the government in 1990 on over 60 consumer goods including fuel, food, and medicines in rebel-held areas in the North and East was lifted in January 2002. The A9 highway, which links the districts in the North with the rest of the country, was also repaired and reopened in April 2002. The improved security and the mobility o f persons and goods fi-om the rest of the country to most o fthe North and East and vice versa brought noticeable economic and social benefits. 8. The cease-fire also permitted the collection o f household data fi-om conflict-affected areas for the first time in two decades. These efforts fill in some of the information gaps, but areas that remain insecure have yet to be surveyed. Inthe absence ofrepresentative data for all areas inthe Northand East, estimates of consumption, poverty, and other indicators of well-being cannot be computed. However, the information available from the data sources enumerated in Table 7-1, as well as other studies conducted by different agencies, is sufficient to draw preliminary conclusions about the economic and welfare situation inthe Northand East compared to the rest o fthe country. 9. Provincial GDPgrowth. Using Central Bank provincial GDP data the Peace Secretariat found that GDP of the Northernprovince quadrupled from an average o f 3.4 percent duringthe pre-cease-fire years (1997-2001) to 12.6 percent during the post-cease-fire years (2002-2003), and the GDP o f Eastern province more than doubled from 4.6 to 10.1 percent (Figure 7-1). Furthermore, GDP in the North- Central province, which shares borders with the North and the East, increased by 8.2 percent per annum in2002-03 comparedto a contraction of0.2percentper annumin 1997-200 1. 77 10. The qualitative study undertaken by CEPA, "Moving Figure7-1: GDP growth 1997-2003 for Northand East Out of Poverty in Conflict- Eastern Province I Affected Areas (2006)," also I reveals that the no-war no- 30 peace following the cease-fire 20 has allowed the people of the 10 North and East to re-establish 0 their livelihoods. 11. However, the rapid post- cease-fire expansion of economic activity in the wee: Impact ofthe Cease-fire Agreement on Regional Economic Growth Sri Lanka (2005). SCOPP Web Release. Northern, Eastern, and North- Central provinces was not enough to significantly affect overall national economic growth. During1996- 2003, the collective contnbution of these three provinces to national GDP was around 11-12 percent although they make up almost 20 percent of the total population, whereas Western province contributed 43-49 percent o fnational GDP and makes up only 28 percent o fthe population. 12. Agriculture andfishing. The engine o f growth during the post-cease-fire period in the North and East has beenthe agriculture and fishingsector. In2002-2004, this sector expanded by an average of 32 percent per year in Northern province and by 19 percent in Eastern province, compared to 4.3 and 4.9 percent, respectively, during 1997-2001. Paddy productionmore than doubled inNorthernprovince from an average of 65,000 tons during 1997-2001 to 138,000 tons during 2002-2003, and significantly increased in Eastern province from an average of 619,000 tons to 752,000 tons during the same period. The combined share o f the North and East in national paddy production increased from 27 percent in 1997-200 1to 31percentin2002-2003. 13. Given that the farming community inthe North and East consists largely o f small landholders, this growth is likely to be broad-based and propoor. Unfortunately, there is no poverty trend data available at this time to verify this hypothesis. 14. Other sectors of the economy in the North and East (industry and services) have also shown growth after the cease-fire. Growth inindustries, for example, has averaged 11.6 percent annually during 2002-03 compared to an average contraction of 2 percent per year during 1997-2001. 15. Since the cease-fire, investments in economic infrastructure by donors, social sectors by the government, and inprivate investmentshave risen. Unfortunately, comprehensive data on pre- and post- cease-fire investments inthe North and East are not available. Sectors such as bankng, retail trade, and communications show clear signs of increased investments inthe North and East beginning in 2002. In the communications sector, for example, the largest private cellular company, Dialog, did not operate in the areas under LTTE control prior to the cease-fire. Dialog has since invested over a million dollars in its infrastructure network inthe North and East. Of its total 1.5 million mobile customers, 250,000 or 17 percent are now from the North and East. This i s a significant increasing considering that only an estimated 13 percent ofthe population inSri Lanka currently resides inthe Northand East.3 Peace Secretariat Impact o f the Cease-fire Agreement on RegionalEconomic Growthin Sri Lanka (2004). 78 Impact of the cease-fire on labor force trends 16. The downward trend in the unemployment rate i s encouraging. Although the districts inthe North and East remain higher than the national average (8.4-1 1.5 Laborforce Unemployment percent in North and East districts compared to 8.3 Province participation rate percent nationally), unemployment fell from 13 to 9 Northern (part) 33.8 13.0 percent inthe North and from 16percent to 10percent Male 54.6 6.9 intheEastfiom200244 (Tables 7-2 and7-3). Female 15.7 31.5 17. Laborforce participation rates remain low. The Eastern 40.3 15.9 labor force participation rate was 34 percent in the Male 63.6 9.3 North and 40 percent in the East in 2002. The Female 18.0 38.0 corresponding figures for 2004 varied around 36 to 46 National excI. percent, respectively. In comparison, the North and East 50.3 8.8 corresponding national labor force participation rates Male 67.9 6.6 were 50 percent in 2002 and 49 percent in 2004 (LFS Female 33.6 12.9 2002 and 2004). I Census and Statistics. Table 7-3: Labor force statistics, 2004 18. Employment by sector. The employment by occupational (percent) group in the North and East mirrors the trend in the country. Laborforce Unemploy- About 27 percent of those employed in the North are skilled participation ment agricultural and fishery workers, 25 percent are employed in National 48.6 8.3 elementary occupations, and 14 percent are employed in craft North and related work. In the East the numbers are 30, 20 and 14 Jaffna 36.2 9.7 percent respectively; whereas nationally they are 20, 26 and 15 Mullaitivu 36.3 11.3 percent, respectively (LFS 2004). Killinochchi 40.3 8.5 Mannar 39.3 - 19. Employment and gender. The new jobs created recently Vavuniya 38.0 11.1 inthe Northand East tended to benefit men disproportionately. East The female unemployment rate in many districts in the North Batticaloa 41.4 10.6 and East remains high. About 44 percent of the women in Ampara 46.1 11.5 Mulaittivu, 30 percent inVavuniya, 24 percent inJaffna, and 23 Trincomalee 38.1 8.4 percent inBatticaloawere unemployed in2004 (Table 7-3). Source: Department of Censusand Statistics 20. Education and unemployment. The more educated who Labor Force Survev 2004. have recently moved back to the Northand East provinces have not been able to find jobs. As a result, unemployment is high for educated workers (completed GCE A.L./HNCE and above), at around 29 percent compared to the national average of 17 percent in 2004. Educated workers are now the most likely workers to be unemployed in the North and East. Unemployment among women who have completed GCE A.L. and above in 2004 i s even higher: 45 percent inthe North and 38 percent in the East (compared to 24 percent nationally). The composition o f the unemployed has shifted dramatically. In 2004 unemployment in the North and East was highest among the most educated group inthe labor force while in 2002 unemployment was highest for workers who had less than 10years o f schooling (45 percent inthe Northand 43 percent inthe East). Consumption, income and housing indicators 21. Both the Central Bank and the DCS attempted to cover the North and East in their household surveys since 2002, but large parts o f the North and some parts of the East are incomplete due to security constraints. Therefore, since the DCS was not comfortable with the sampling and the data collected from the North and East, this data was not analyzed. The CFSES 2003/04, on the other hand, provides data on the North and East. The CFSES data on Northern province has an urban-Jaffna bias since the survey was unable to capture the districts of Killinochchi, Mannar, and Mullativu, which make up one-quarter o f the 79 population inthe North. Inaddition, the CFSES did not cover families still livingincamps that are likely to be worse-off.` Thus, consumption and other estimates for the North from the CFSES are likely to be overestimates. The CFSES data i s also not comparable with the HIES (on whichthe official poverty lines are based), as a result of which comparable poverty estimates could not be calculated for the North and East. However, the CFSES does yieId measures that are roughproxies for welfare-such as income and consumption levels, savings and borrowing patterns, and housing conditions-that serve as points of comparison between the North and East and the rest o fthe country. 22. Comparisons between the per capita income/expenditures inthe North and East with those of other provinces may also be problematic because o f higher commodity prices in the North and East due to higher transport costs and monopolization o f trade in certain essential goods. Although there are no consumer price indices for the North and East at present, spot checks on commodity specific prices indicate that a basket o f minimumessential commodities would cost much more in certain parts o f the North and East (particularly the North) compared to the rest o f the country.' Thus, the per capita income and expenditure levels reported below for the North may be overestimates in terms o f the purchasing power they represent relative to the rest of the country. 23. Consumption expendituresfrom CFSES. Significant transfers to the populations of the North and East seem to have safeguarded income and consumption levels (Figure 7-2).6 Transfers constituted 37 percent of the total incomes o f households in the North and 24 percent in the East, while ,o,o Figure 7-2: One month mean and nominal expenditures, 2003-04 ~ the average for all provinces 35 was only 18 percent in 15.m 2003/04 (Figure 7-2). As a 30 result, the average monthly !O,oQO 25 income o f households in the 5,000 20 North in 2004 is similar to other provinces, with the 15 0,oOO exception of richer Western 10 province. The monthly 5 . 0 5 incomes in the East are much 0 0 lower and comparable to that o f the poorest provinces o f Sabaragamuwa and Uva (Figure 7-2). Soum CentralBankofSnLanka CFSES2003/04 1 -No ofhouseholds Ulncomepercapita 24. Thepattern of monthly 1 EZI Percapnaevendaure Transfers as % ofmcom per capita expenditures in 2003104 inthe North at Rs. 3,255 and the East at Rs. 3,136 are very similar to those of other provinces with the exception o f Western (Rs. 5,922) and North Western (Rs. 4,012) provinces (Figure 7-2). 25. The composition of household expenditures, however, shows that food makes up a far larger share of expenditures o f households in the North and East-about 42-43 percent similar to the poorest provinces of Uva and Sabaragamuwa (Figure 7-3). In the richest province (Western) the average household spends 27 percent of their monthly per capita expenditures on food. The average household 4See Table A-7.3, Annex 7 for statistics onpopulation living in refugee camps due to conflict andor the tsunami. 5 For instance, the price of LP gas (12.5 kg cylinder) was around 38 percent and 35 percent higher in Jaffna compared to Vavuniya during M a y and June 2005, respectively. The price of petrol was 12 percent higher and diesel 1Ipercent higher inJaffna compared to Vavuniya. InMannar district, duringM a y 2005, the price of a bag o f cement was Rs. 520 in govement-held territories whereas it was Rs. 630 inrebel-held territories. The CFSES 2003104 survey o f the NorthernProvinceexcludes Killinochchi, Mannar, and Mullativu. 80 expenditwe on housingandutl11ties, and the Figure 7-3: Food expenditures as a share of per capita thirdlargest on transport. expenditures 2003/04 (percent) 26. Housing conditions. A comparison o f housing conditions and facilities is also a useful gauge o f well-being. The CFSES suggests a mixed picture on this front although the results may again reflect an urban bias for the North (see Table A-7.4, Annex 7). Like many of the other districts, a large proportion of households in Eastern province (92 percent) own their own houses. Homeownership in the North (63 percent) i s in fact the lowest in the country. Houses in the North and East tend to be NorthCentral NorthWestern Note: Northern province excludesKilinochchi, Mannar, and Mullaitivu districts. Source: Central Bankof Sri Lanka (2005), Sri Lanka Socio Econormc Data2005, Colombo, pp35. Columns 1 & 2 -Department ofNational Planning cited in Sarvananthan (2004: 55). Sectoralcomposition 28. The predominance o f services and agriculture and fishing sectors (33 to 42 percent o f North and East GDP) has not changed during the past two decades. The industrial sector, on the other hand, has always been very small. In 1983, about 10,000 industrial units employed approximately 50,000 people and by 2003, 12,000 unitsemployed only about 45,000 people.' The services sector makes up one-half of 7DCS Census of Industries 2004. 81 the economy o fthe North and East, but since it is dominated bymicroenterprises informationon specific activities within this sector i s not available. 29. Although the agriculture and fishing sector continues to make up a significant part o f the economies o f the North and East, the importance o f the sector nationally has eroded during the conflict years. Before the conflict paddy, onions, chillies, tobacco, h i t s , dairy products, and fish were sold in significant quantities to other parts of the country. However, due to transportation difficulties and the trade embargo commercial agriculture inthe North and East was transformedinto subsistence agnculture. 30. Paddy i s the major agriculture product o f the North and East. Paddy production accounted for almost one-third o f total paddy production inthe country in 1980. However, this share had shrunk to less than one-quarter in 2000. Following the cease-fire, paddy production in 2002 more than doubled in the North and more than tripled inBatticoloa district inthe East (see Table A-7.1, Annex 7). 31. Fishinghas historically been a major economic activity inthe Northand East especially for Jaffna. In 1980, more than one-half of the total fish catch in the country was from the North and East (56 percent). However, during the past decades, mainly because o f restrictions imposed on fishing for security reasons, the region's share inthe total fish catch of the country dropped to 33 percent by 1990, and to just 16 percent in 2000. Encouragingly, once fishing restrictions were relaxed in 2002 following the cease-fire the share o f the North and East almost doubled to 31percent in2002 from 2000 (see Table A-7.2,Annex 7). Unfortunately, data i s not available on the impact o fthe tsunami on fishing inthe North and East. Poor economic infrastructure 32. Roads. Like some o f the other poorer areas in the country, the North and East suffer from a severe lack o f economic infi-astructure to support economic growth and reduce poverty. Although several `A' and `B' class roads have been repaired in the past four years, several rural roads are out o f service. The inadequacy o f the roadnetwork i s reflected inthe low accessibility potential inthe region (see Figure 3-6, Chapter 3). Information on the quality o f road is unavailable, but they are likely to be in poor condition due to years o f inadequate maintenance. 33. Even where roads are available, lack o f adequate transport facilities make it difficult for people in the interior to take their produce to market. For example, in the Vanni region o f the North a barebones bus service connects towns and villages. While there is frequent bus service between Omanthai (Vavuniya) and Muhamalai (Jaffna), n o bus serves interior routes. Also n o rail service i s available in the NorthbeyondVavuniya. 34. Similarly, minimalpower and telecommunicationsfacilities in certain areas o f the North and East deprive agricultural producers and fisherpersons access to information and markets. 35. According to the CFSES 2003/04 only 64 percent o f total households in the North and East have power, while the average for the island is 73 percent. This figure for the North is likely to be an overestimate because the CFSES could not cover the remote districts o f Kilinochchi, Mannar, and Mullaitivu where estimates suggest that up to 90 percent o f the households may not have power. Similarly, only about 20 percent of households inthe North and 14 percent inthe East have either a land or cellular phone while nationally about one-quarter o f households has one or the other. Access to telecom inthe Northand East i s better than inUva (9 percent) and Sabaragamuwa (13 percent), but worse than other provinces. 7.3. Vulnerability and human developmentoutcomes 36. Savings and borrowings as an indication of vulnerability. The North and East region has the lowest percentage o f households with positive savings. Only 34 percent o f households inthe East and 38 percent o f households in the North have positive savings, while the average for all provinces was 53 percent in 2003/04. Households inthe North and East also have higher borrowing rates than the rest o f 82 the country. On average, households in the East borrow as much as 44 percent of their total household income and the corresponding rate in the North is 39 percent, while the national average i s only 22 percent. About 65 percent o f households in the East and 58 percent of those in the North are indebted, compared to 49 percent of all households in S n Lanka. The number and average size of loans taken is also much higher in the North and East than in the rest o f the country and usually used for consumption (CFSES 2003/04). 37. Lower savings and higher incidence of borrowing among North and East households compared with the rest of Sri Lanka may be indicative o f greater vulnerability. Indebtedness and negative savings often result when households borrow or run down their savings to cope with income shocks. Available information is not enough to conclusively link these averages with the incidence o f such shocks, but is consistent with the intuition that households in conflict areas are likely to be exposed to more frequent economic shocks. 38. Studies using different methodologies also support the idea that vulnerability is pervasive in the North and East. A World Food Programme (WFP) study-on islandwide vulnerability focused onfood insecurity at the District Secretariat (DS) Division level using secondary data, Geographic Information Systems (GIs), and statistical tools-reported these two regions as the most vulnerable in terms of food access and utilization (Box 7-1). Of the 29 percent or 94 DS divisions in Sri Lanka that are most vulnerable to food insecurity, 43 percent were located in the North and East. In addition, all but 5 D S areas inthe Northwere identified among the most vulnerable to food insecurity. 39. Welfare and economic outcomes vary significantly among areas within the North and East as shown inthe Vulnerability and Poverty Profile (VPP) of the Northand East conducted by the Center for Information Resources Management (CIRM) o f the Northeast Provincial Council in 2004 and 2005 (see Box 7-2). The study attempts to identify the poverty status o f villages indifferent DS divisions within the Northand East along four dimensions: economic, health, education, and vulnerability. According to these criteria a large majority o f villages in all D S divisions o f Batticaloa, Trincomalee, Mannar, and Vavuniya are categorized as "extremely" or "highly" poor. 83 The aftermath ofthe tsunami 40. As the North and East were slowly emerging out o f the protracted civil conflict (since early 2002) the tsunami on December 26, 2004 devastated coastal populations. Coastal communities inthe North and East now face the double burden o f hardships from the conflict as well as the tsunami. The districts o f Ampara and Batticoloa inthe East, inparticular, suffered the most lost lives, displacement, and damage to infrastructure and housing. 41. Losses. According to the census of tsunami affected areas conducted by the Department o f Census and Statistics, out of the estimated 35,000 lives lost to the tsunami, 63 percent or 22,000 victims were from the North and East. About 347,000 persons were also displaced in the North and East by the tsunami. Of the 100,134 buildings either partially or completely damaged, around 89 percent were private dwellings. The East sustained the most building d a m a g e 4 8 percent of buildings destroyed islandwide were inAmpara and Batticoloa. 42. Fishing was the most severely affected livelihood sector in the North and fishing, micro and small enterprises, and tourism was severely affected in the East. Although the tsunami had a margmal impact on the country's overall macroeconomic performance, the regional economies of the affected areas were hard hit.8 Since the populations inthe North and East are mostly engaged in agnculture and fishing the tsunami is also likely to haveworsened livelihoods and the incidence o fpoverty inthese areas. Laggingeducationandhealthoutcomes 43. Lagging educational and health outcomes have a strong bearing on the likelihood o f future generations falling into poverty. In terms o f health and educational outcomes, the conflict has clearly caused the North and East to lag behind on these dimensions. Schools and health clinics have been destroyed, maintenance o f these assets has fallen short, andthere have been shortages innecessary human 'ADB/JRIC/JICA/WB(2005). 84 capital due to security fears. This is even more serious given that the child dependency ratio inthe North and East i s higher than in other provinces, 45 and 55 percent, respectively, while the national average is about 40 percent (CFSES, 2003-04). ocial indicators & E 3uao% s . - w 3 8 k 2 -u 0 Indicator s & G 3 s 7& 2a Low bn-thweight 18.0 25.7 22.7 24.3 <2.5 kg* Underweight 29.4 46.2 44.1 53.2 3-59 months* (2000) Access to safe 61.9 45.9 - - dnnkingwater** Nolatrine** 5.6 5.0 29.2 7.3 Source: Department of Census and Statistics (2004), Poverty Statistics / Indicators for Sri Lanka, Colombo, pp4 and42; CentralBank of S n Lanka (2005), CFSES2003/04. 44. Interms o f health outcomes,46 percent of children below five years of age inthe North and East are underweight, compared to 29 percent for the rest o f the country. The percentage rate o f babies born underweight in the country is 18 percent, but the proportion is 26 percent for the North and East. The figures are even worse in districts like Batticaloa and Vavuniya, where over one-half o f the children are underweight, and Trincomalee, Jafha, and Vavuniya where more than 30 percent o f babies born are underweight (Table 7-5). 45. Access to safe drinking water and safe sanitary facilities are also inadequate. Nationally 62 percent of households have access to safe dnnlungwater, but only 46 percent o f those in the North and East have such access. About 80 percent o f households in Sri L a n k have access to water seal latrines, while less than one-half o f households inthe North and East have such access (the lowest two among all provinces) (Table 7-5). 46. Serious access and quality issues in the education sector affect education outcomes. Eastern province has has the lowest literacy rate in the country (87 percent): female literacy is higher than male share o f population has secondary or tertiary education, compared with the country averages. Incontrast, literacy and education outcomes in Northern province I 40 are closer to those for the country as a 30 whole; tertiary education attainment is much higher (28 cent o f the Northern 20 population) than the average for the island I O (21 percent). 0 47. The North and East have higher than average student drop out rates and lower learning outcomes. According to the Census o f Schools undertaken in 2003 the repetition and drop out rates in the North and East were among the highest in I 85 the country after Sabaragamuwa and Uva provinces. At Grade One level, Sabaragamuwaprovince had the highest repetitionrate (2.6 percent) followed by Eastern (I.9 percent), and Uva and Northern (1.8 percent) provinces. 48. Primary school (Grade 4) children inthe North and East also have the poorest mastery o f skills in their mother tongue, the English language, and mathematics. The share o f Grade 4 children in2003 inthe North and East achieving mastery in their native language was only 23 percent compared to the national share at 37 percent. Likewise, the share of Grade 4 children in the North and East achieving mastery in the English language was only 5 percent which is one-half the national figure. In mathematics only 25 percent o f Grade 4 children achieved mastery compared to 38 percent o f the children nationally (Figure 7- 4). 49. The gaps in achievement, school retention, and repetition between the North and East and the rest of the country indicate a serious problem with the quality o f education in conflict-affected areas. Even more than lagging educational outcomes, the conflict may have affected the quality of basic services supporting the creation o f human capital, w h c h in turn may have exacerbated conditions leading to poverty. 7.4. Remaining impedimentstogrowth andpoverty reduction in theNorth andEast 50. The cessation of conflict in 2002 clearly benefited the local economies of the North and East. However, despite measures that improved security, and the slow reintegrationo f these areas with the rest o f the country, as well as higher economic growth, the Northand East's economic contributionto national GDP remains low. The sustainability o f future high growth in these regions also remains questionable. This section will explore institutional impediments to growth and poverty reduction affecting the North and East. 51. Constraints to mobiZity. A number of rigidities in the current institutional setup continue to impede the mobility o f populations between the North and East and the rest o f the country, imposing additional costs on economic activity and welfare. High security zones in the North and East still bar access to civilians and prevent the undertalung of productive activities. For example, one-third of the Jaffha peninsula is a highsecurity zone. Restrictions and occupation of prime commercial properties in Jaffna seriously hampers business development in the North and East since Jaffna has historically been the commercial hub o f these two provinces. 52. Fishing restrictions. Certain time and geographic restrictions on fishing still exist in the Northern and Eastern provinces, particularly inJaffha, despite some relaxation after the cease-fire. Fishing i s one o f the primary economic activities inthe Northand East, and prior to the conflict used to account for one- half of the total fish catch in the country. Security motivated fishing restrictions have also resulted in poaching within Sri Lankanmaritime boundaries by fishermen from India, Japan, Taiwan, and Thailand. 53. Restrictions on road use. Although the A 9 highway reopened in 2002, it was only open for vehicular and civilian traffic Monday through Saturday each week and 10 hours a day. Moreover, goods transported to and fiom the North and East had to go through four checkpoints and the quality of the A9 highway remains poor. At each checkpoint, the entire consignment was offloaded, checked, and reloaded. Entrepreneurs from Jaffha complain that it took them three days to transport goods to and from Colombo via the A 9 highway, which added to the cost o f doing business. Thereafter, the A 9 highway has been closed beginning the second halfof 2006 due to the worsening security situation. 54. Access to credit. Businesses based inthe Northand East have poor access to bank credit. Because they are unable to open a Letter of Credit (LC) at their local banks, traders in the province have to purchase goods fiom Colombo-based importers, which again adds to transaction costs and retail prices. Inaddition, there is anecdotal evidence ofcapitalflight from the NorthandEast to the rest ofthe country because o f fears that savings are not secure inbanks located inthese provinces. Since banks inthe North and East only Iendwhat they obtain indeposits from local clients, the issue o fcapital flight constrains the 86 availability o f credit in the area to create a vicious cycle, further exacerbated by more stringent in disbursement o f loans by local banks to farmers, fishermen, and traders, and collateral requirements that are higher than inother parts of the country. 55. Out-migration. Despite a large influx of internally displaced persons (IDPs) to Jaffna in recent times, the total number of people leaving the North and East may be higher due to obstacles inbanhng, fishing, and transport. As mentioned earlier, the unemployment rate among the educated is very highin the Northand East, and it appears that the bulk o f those migrating out are relatively better educated and shlled - a migrationpattern similar to other parts o f Sri Lanka (Chapter 4). Also, as indicated inChapter 4, the cease-fire o f 2002 has made it possible for people in the North and East to migrate in greater numbers to Colombo in search o f economic opportunities. Therefore, out-migration is further depleting the human capital urgently neededfor the rehabilitation and revival o fthese areas. 56. In conclusion, during the past four years of cease-fire several critical infrastructure facilities- roads and bridges, power supply and telecommunications, and schools and hospitals- have been rebuilt and restored, but to realize fully the growth potential in the North and East, the institutional constraints described above must be addressed. 7.5. Concludingremarks 57. Despite less than perfect data it is clear that the conflict has severely affected the welfare, social, and economic conditions o f people residing in the North and East. The tsunami aggravated existing poverty and welfare challenges. That is why the peace dividends from the cease-fire in 2002 are so encouraging. The worsening security situation in2006 would have added to the challenges in the region, although the impact o f the deterioration o f security on the economic welfare o f people inthe North and East is unclear at t h s point. 58. Undoubtedly, no amount o f investments by the government, various donors, and private investment can fully compensate for the shortfalls in economic infrastructure, incomes, health, education, and social challenges in the North and East region until lasting peace i s achieved. Without peace the institutional rigidities and obstacles to sustained growth in the North and East and prospects for significant poverty reduction will remain distant. 87 8. PovertyTraps in the Estatesand the Way Forward 1. The estate populationrepresents the most significant challenge to poverty reduction in S n Lanka.' Consumption poverty in the estates was higher in 2002 than in1990-91, contrary to the trend for the country as a whole. In2002 the estates were the poorest sector with a poverty rate 7 percentage points higher than the national average (chapter 2), and constituted about 5 percent of the country's total population but 8 percent of the poor. This chapter will examine the nature and determinants o fpersistent poverty inestates. 2. A wide range of factors contribute to poverty traps in Sri Lanka-from geographical isolahon and lack o f access to infrastructure to individual attributes like lack o f education and occupation (chapter 3). These factors contribute to a persistent cycle of l o w capability, minimal earning potential, and poverty among the estate population. Although many o f these factors operate in remote, rural areas of Sri Lanka as well (see chapters 3 and 6), the challenges in the estates are especially acute. 3. Many key human development indicators in the estates lag far behind even those in rural areas (chapter 5). Deficiencies ineducation are part o f the reason why the opportunities available through internal migration-an avenue out o f poverty for many households in remote areas (chapter 4)---are limited for estate households. The estate population has lower rates o f out- migration, self-employment, and participation in other sectors. In addition, even when estate residents are able to diversify their income sources by workmg in jobs outside the estate or migrating to urban areas, the economic benefits are often limited because o f their inability to find better-paying employment. 4. The persistence o f poverty traps in the estates is also a result o f long history o f social and economic isolation of estate residents over generations. The sector developed as an enclave, employing workers from outside the country who not offered citizenship rights until 1988. The unique employment structure of estates persists to this day, and much o f the workforce continues to be to be isolated to a large degree from the national mainstream. 5. Current national household surveys provide limited data to examine these issues more deeply. HIES and CFSES data do not distinguish among types o f estates-type o f crop or management, size, or location-to discern underlying patterns. They do not adequately cover issues such as location and connectivity o f estates, type o f work done on estates, or coverage of social programs and availability o f services critical for welfare of estate households. 6. A comprehensive householdsurvey, conducted especially for this study, is usedto provide data for the estate analysis. A complementary qualitative study explores underlying perceptions and attitudes o f estate residents and the relationships between management, trade unions, and households. The analysis asks two key questions: what explains the persistence o f poverty traps inthe estates, andwhat characteristics enable some poor people move out ofpoverty and stay out o f poverty while others fall into poverty or remain trapped in chronic poverty? Findings from quantitative data are triangulated with qualitative insights on perceptions o f estate residents on what they see as dnvers o f positive change. While the household survey was not designed to be representative of the estatepopulation, it was based on a large sample (of over 1,000 households in43 estates) with wide coverage interms o f location and characteristics of estatesto allow for rich analysis.* 'The definition of estates inthis report i s identical to that used by DCS: plantation areas, which are more than 20 acres in area and employing not less than 10residential laborers 2 Boththe studies covered only tea andrubber plantations. Since coconut plantations constitute only a small proportion o f the estate population they were excluded. The qualitative study was conducted as a 88 8.1. Settingthecontext: theestatesector 7. Large plantations growing tea, rubber, and coconut-started during the British colonial period-were self-sufficient enclaves apart from the national socioeconomy. Workers, primarily from South India, were confined to the plantation and were completely dependent on the management for all services related to daily life. Since independence the importance o fplantation crops in the national economy has declined, ownership has shifted from foreign to national, the labor force has become Sri Lankan citizens, and the estate structure has become less rigid. Origin and development of the estate sector 8. The British colonial administration introduced the cultivation o f tea and rubber as commercial crops inthe late 1800s. Changes inthe world commodity market as well as within the estate sector in Sri Lanka gradually reduced the importance of tea and rubber exports. By 2004, textile and garments had taken over as the leading foreign exchange earner, although tea remained an important export crop. Rubber diversified into a local manufacturing base in the 1950s and expanded rapidly after the introduction o f free trade policies and investment promotion zones inthe late 1970s. 9. Due to a range o f sociopolitical and commercial reasons, the Britishtransferred most of the plantation labor fiom South India. Today, estate residents usually earn most o f their income in wage employment on a plantation. Over the past two decades, the labor force in plantations has declined by more than 50 percent, from a peak of 542,000 workers in 1980 to an estimated current figure of less than 269,000.3 Over a similar period, the proportion o f smallholdings (not classified as estates) inthe tea sector has more than doubled (from 19 to 44 percent during 1982- 2002), and smallholdings for rubber are rapidly catching up with the estates despite a fall in the total area cultivated for rubber. The estate sector share o f agriculture has dropped from 27 percent in 1982 to 20 percent in 2002. These trends, along with the declining size of the estate labor force, indicate that the estates are indecline. 10. Policy changes affecting ownership and management. The Land Reform Law o f 1975 nationalized foreign-owned estates in two phases. Phase 1 in 1992 pnvatized management only by creating state-owned Regional Plantation Companies (RPCs) where each RPC entered into an agreement with a private company, the Management Agent (MA). Phase 2 in 1995 sold controlling interests in the RPCs to MASsubject to restrictions that allowed the government to exercise control over certain affairs (see Annex 8, section I).These policy shifts are likely to have hadprofound impact on estate management, workers, and the relationships between these actors. Poverty and human development inthe estates 11. Poverty incidence. The estate sector has the highest incidence o fpoverty, with a headcount rate of 30 percent in 2002. Unlike the trend in national poverty, estate poverty in 2002 was signijicantly above 1990-91 level^.^ Consumption in the estates has also become more highly concentrated in a narrow interval around the poverty line-again in contrast to the country, the concentration in estates has increased between 1990-91 and 2002 (see chapter 2). This suggests that even small changes or shocks to the estate economy can produce large shifts in poverty combination of focus group discussions(FGDs), key informant andhouseholdinterviews ina sample of 20 estates, 13 o f whichbelongedto the survey sample (see Annex 8, SectionI11for details). Statistical Abstract, DCS (2004) Even when the standard errors of poverty headcount are taken into account in computing 95 percent confidence intervals, the headcount inthe estate sector clearly increasedbetween 199&91 and 2002, and are also clearly higherthanthe national headcount in2002 (see Annex 2). 89 headcount rate, and overstate actual changes in welfare. Therefore, this poverty trend analysis takes into account the change inthe entire distribution, rather thanjust the headcount. 12. Bearing Inmindthe full density o f consumption, the story inthe estates seems to be one o f little movement over time, and relatively l o w variance within the sector. A significant majority o f the estate population consumes just enough to approach the poverty line; and over time, a slight worsening inthe situation appears to have affected almost the entire distributionrange. These two snapshots in time say little about movements in and out of poverty, but they do hint at the likelihood that poverty in the estates is endemic and linked tofactors that aflect the sector as a whole and to factors that have changed little over time. Insights from the qualitative study support t h s view. There i s much agreement in the characterization o f poverty between households across different types o f estates; and inmost cases, the perceptions and explanations o fpovertyby respondentswere restrictedto factors operating within the particular estate sector. 13. Poverty in the estates is multidimensional, a characterization that also comes through clearly inrespondents' qualitative interviews. While earnings and consumption are at the core o f the households' own understanding o f poverty, other elements such as health and education also figure strongly. At the same time, there i s some divergence between time trends o f consumption poverty and other indicators of welfare. Unlike consumption poverty, many aspects o f health, education, and housing have improved in the estates-especially in recent years thanks to initiatives undertaken by the government with various development partners. 14. Health and education. The Estate Sector Health Bulletin documents a gradual improvement Table 8-1: Literacy rates by sector in the health trends among RPC estates workmg (percent) with the Trust.' For example, the crude birth rate, All sectors crude death rate and infant mortality rates have 98687 declined steadily during the past 15 years. The gap Male in average literacy rates between the estate sector FemaIe and the urban and rural sectors has also narrowed 99697 76.9 Male over the past two decades (Table 8-1). Female 67.3 15. Inspite ofthese gains, the estates still lagwell 003-04 92.5 behind the rest of the country and the rural sector on key indicators o f health and education. Male and 90.6 female literacy rates in 2003-04 are 6 and 16 3 report (2003-04). percentage points lower than the rural averages, respectively (Table 8-1). A s reportedinchapter 5 (Table 5-2), 30 percent of estate children were low birthweight babies, compared with 17 percent o f rural children. Also 37 and 46 percent o f estate children are stunted and underweight, respectively, compared with 14 and 31percent o frural children; and 48 percent o f estate mothers have low body mass index (BMI), compared with 23 percent ofrural mothers. 16. Potential reasons for the relatively high incidence o f malnutrition in the estates include inadequate access to food and poor nutritional and dietary practices. The Estate Survey indicates that 13 percent o f sample households report food shortage. Moreover, food shortages appear poverty related-f those who reported experiencing a shortage, most attributed it to lack o f income, as opposed to nonavailability o f food. 'Plantation HumanDevelopment Trust (2005). This captures only RPC estates working with the Trust. 90 8-2). This reflects recent efforts Table 8-2: Changing pattern of housing stock in estates (percent) by the government and 'Type of housing unit Estate All sectors estate management to 199697 2003-04 199697 2003-04 allocate more resources Single house 10.2 28.1 88.5 91.2 in this area. However, Attached house/annex 0.0 7.9 9.7 2.8 the proportion o f line Line r m m h o w house 83.2 63.4 0.0 3.9 rooms is still much 0.2 0.5 0.3 0.8 higher than the country Other 6.4 0.2 1.5 1.4 average, and a Source. CFSES report, CBSL (2005). sustained effort is needed to close the gap between living conditions in the estates and other sectors. 8.2. Aprofile of poverty in theestatesector: householdand communityattributes 18. The Estate Survey conducted for this report explores a wide range o f household and community/estate characteristics. The insights from the survey data can be triangulated with results from the qualitative analysis o f factors influencing households' movements in and out o f poverty. Since the Estate Survey lacks consumption data, the poverty status of households is measured using an Asset Index (AI)score, which should be interpreted as a proxy for wealth rather than a flow o f income.6 To distinguish this measure from the consumption-based estimate o f poverty used elsewhere in this report, the welfare measure calculated from the AI will be referred to as AI-Poverty Rate, defined as the percentage of households whose AI scores lie below the 30th percentile. The 30* percentile i s a reasonable choice for the "AI-poverty line," given that the poverty headcount for the estate sector is 30 percent (HIES, 2002). 19. Demographic characteristics. The average dependency ratio--the ratio o f number of dependents (family members below the age of 15 and older than 64) to number o f worlung age family members (ages 15-64)-in estate households is 55 percent, compared to the national average o f 49 p e r ~ e n t .Multivariate analysis using the Estate Survey data indicates that ~ households with higher dependency ratios have lower AI, and conversely, households with more earners tend to be better o f f (see regression results in Annex, Table A-8.1). In the qualitative study, the demographic cycle o f a household-from the birth o f children into a family, through schooling, to reaching the age o f employment, marriage, and death-is identified by the households as having a strong impact on their current economic status and future prospects. 20. Education. According to HIES (2002), 14 percent o f estate household heads have no schooling, compared to 6 percent for Sri Lanka as a whole; 24 percent o f estate household heads attained grade 6-9 compared to 37 percent among all household heads. In the Estate Survey sample, 17 percent of household heads have no schooling, and only 7 percent have had education o f 0-level or above. AI-poverty rate among households whose heads have had no education i s 41 percent, compared with less than 11 percent among those with 0-level and above (Table 8-3). The multivariate model shows that household head's education beyond grade 5 has a significant The method employed to construct asset indices is similar to the approach adopted by DHS surveys around the world, including Sri Lanka, and inthe health section o f chapter 5 o f this report (for details, see Annex 8,Ssection Il). 7Poverty statisticshzdicatorsfor Sri Lanka, DCS (2004). 91 positive effect on the household's AI score and poverty (Annex 8, Table A-8.1). As seen later, education attainment i s related to better employment opportunities and higher earnings, especially inoccupations outside the estate. 21. The qualitative study reveals that estate households recognize education as a driver o f upward mobility. Households Poverty perceived improvements in access to Sample Asset rate distribution better quality education over the past 15 index (percent) .(percent) years-both within the estate and in the Headof household's education attainment larger towns, and saw that as a force for Noschooling -0.32 41 17 positive change. Doubts were expressed Grade 1-5 -0.19 35 44 about whether education actually creates Grade 6-9 0.11 23 31 employment opportunities for estate 0-level 0.85 13 5 residents, probably because there are so A-level and few examples in some o f the estates, as beyond I 1.68 I 4 2 well as a feeling of marginalization that Headof householdhasNational IDcard often clouds perceptions inthe estates. At NO -0.47 51 12 the same time, the role o f education in Yes 0.04 28 88 enabling greater participation in the .Total -0.02 30 100 national SOCiOeCOnOmy Was Strongly Note: The results are applicable to sample estates only, and are articulated, particularly by youth. notrepresentative of the estatesector 23. Why is possession of a NIC linked to welfare? It turns out that having a NIC has statistically significant links with key determinants of welfare, like education and economic opportunities, particularly outside the estates. Estate residents with a NIC are on average better 92 educated, work more days, and earn more from outside sources than those without a NIC (see Figure 8-1). Figure 8-1: NIC ownershiD inestates is associatedwith better education and earnings %with 0-level education or more #days worked last month Outside earnings last month M , ---_---___I 34 i M O x 20 iwoo XI 500 0 h8 2U24 25-34 3554 8 6 4 Totd 16-8 29-24 2534 3554 5 6 M Total W-EI 20-24 25-34 35-54 55-64 Total Aa* A#* AS* DNoNIC BWithNIC ON0 NK: BWithNK: QNoNIC mWthNffi Note: The resultsare applicableto sample estate only, and are not representative of the estate sector. Source: MOP Estatequanhtativesurvey (2005). I 24. The difference in education attainment i s particularly highamong the 16-19 and 20-24 age groups: around 30 percent o f youths with a NIC attain an 0-level education or higher, compared to 18 and 5 percent o f non-NIC youths in the two age groups, respectively. Given that the NIC ownership rate i s low among youths, this difference means that NIC ownership has critical socioeconomic implications for estate youth, especially. Ownership o f NIC also appears to be associated with more opportunities for employment outside estates. The estate residents o f age 16-34 who have NICs earn more from sources outside the estates, unlike those from their age group without NIC or the older generation, who rely heavily on estate wages (see Annex 8, Table A-8.2 for more details). Regressions show that, even after controlling for gender, education, ethnicity and location, having a NIC yields a significant premium to earnings. Diversification of the household livelihood portfolio 25. Qualitative findings indicate that the most successful way to move out of poverty i s by diversifying the household livelihood portfolio beyond estate employment-incorporating non- estate sources o f income including slulled work, and internal or external migration (see Box 8-2). Data from the Estate Survey clearly indicates that the ability to diversify income sources i s associated with higher welfare and lower poverty for households, particularly when the sources include income from enterprises, regular and slulled employment outside the estate, and overseas migration. Box 8-2: External employment as a means to achieve ecnnomic mobility Excerptsfroma life history(respondent: age 39, inale, married;rubber, RPC, Kalutaradistrict): ",...Until 1990mother was the onlyregisteredearnerofthe family sincefather was illand whatever she earnedwas not enough to sustain the family of six. Could not receive a satisfactory education: only up to the 5" gradc bccausc parentswere facingdrfi`icultiesin siipportingour studies.. .....Tried wage labor inthe estateuntil 1990hut it was nor payngenoughro cope with the gradually increasingexpenses af the family. From 1990 to 1994 throughthe help of a known mudalali (trader) I was i~Momtuwa (suburb of Colombo) working in a `tea kade' (roadside tea 93 26. Income diversification of estate residents. Estate households earn their income from three broad categories: estate wages, outside wages, and enterprise incomes-that includes income from nonagricultural household businesses and from sales o f crops, livestock, and livestock products. More than 40 percent o f estate households rely solely on estate wages for earned income, with "estate and outside wages" a distant second. The AI-poverty rates are highest among households that receive wage incomesfrom only one source; and lowest for households with incomes from all three sources (Figure 8-2). Figure 8-2: Income diversification is associatedwith lower poverty Distributionof income sources among AI-poverty incidence by income source households nterplse(uitNhithoutother 10% ax incomesource) I Estateand outsidewages OUtSldewageSonly Estatewgesonly B l None E4 Estatewagesonly None 0Outsidewagesonly 0Estateandoutsidewages O D 2 0 3 0 4 8Enterprise(withlwithoutotherincomesource) AI-poverty rate (%) Source: MOPEstate Survey (2005). . Multivariate regressions are usell to gauge the effect of diversification net of househo 1 attributes that may correlate with both the household's ability to diversify and AI. The regressions show that households earning both off-estate and estate wages are not significantly different in terns o f AI from those whose incomes are only from estates. Incontrast, households that receive incomefrom enterprises-whether singly or along with other income sources-fare significantly better (Annex 8, Table A-8.1). The same i s true of rural households-those owning a nonfarm enterprise have a significantly lower poverty rate comparedto those without one (chapter 6). 28. Why is diversification into any type of employment outside the estate not enough to increase household welfare? A disaggregated look at occupations is useful to understand why. Primary employment o f estate residents can be divided into 3 broad groups-estate, non-estate wage, and self-employed; and these can be further broken down by location or nature of employment (Figure 8-3).' Employees can also be classified as casual and regular (salaried) workers. Self-employed-those who derive their incomes from enterprises-account for about 6 percent o f workers. 29. Earnings from different primary occupations differ substantially (Figure 8-3). Workers in regular estate wage employment are better educated and earn almost a thousand rupees more per month than casual estate workers do, even though the number o f days worked per month i s almost the same (see Annex 8, Table A-8.3). Similar gaps in earnings and education also exist between casual and regular workers outside the estates. Working outside the estates thus does not necessarily imply higher earnings. Earnings from casual wage labor outside the estates are only slightly higher than that on estates, and even that may not translate into a welfare gap, since estates provide workers with numerous benefits. Regular employment inside or outside the estate and self-employment inside and outside the estate, on the other hand, are clearly associated with higher earnings. 8It i s sufficient to focus on primary employment exclusively because less than 7 percent o f workers living o n estates have secondary employment, and less than 5 percent are employed outside the estates. 94 Figure 8-3: (a) Primary occupation of estate (b) Monthly earnings by primary occupation of estate residents Total Earnings Self-employed 6000T 5000 4000 c Casual-Outside 8 3000 -r Casuai-Estate 2 2000 1000 0 Casual- Regular- Casual- Regular- Self- I Estate Estate Outside Outside employed I Source: MOP Estate Survey (2005). 30. Thus, certain types o f income diversification are associated with higher welfare o f estate households. Diversification of income sources among household members represents significant improvements in welfare onZy when the outside income sources include self-employment and/or regular (salaried) employment. Conversely, a reliance on casual employment-be that on or outside estates-seem to be associated with higher AI poverty. Notably, better education outcomes seem to lead to more remunerative employment, and especially so outside the estate. Successful households also often make a conscious decision to include estate work in their livelihood portfolio to accessbenefits such as housing. 31. Migration as a driver of economic mobility. Diversification appears to offer the most successful strategy out o f poverty, and qualitative findings indicate that migration is an important component o f the "ideal" diversified livelihood portfolio. Of the different types o f migration, which is more prevalent among the estate population? What are their welfare impacts? What are the characteristics that determine the pattern of migration and its effects? from the estate sector increased Table 8-4: Profile of migrants from estates considerably during the past (last 5 years before the survey) decade. The rate of external I fiverage " . I I migration from estates increased Distri- number of Remits Average from 42 to 49 per 1,000 bution months spent regularly age households between 1996-97 and (percent) in estate (percent) (years) 2003-04. During the same Gender period, internal migration also M a l e 60 2.3 77 29 grew, from 4 to 20 per 1,000 Female 40 2.5 58 28 households. The Estate Survey Type shows a much higher internal Permanent 6 n.a. 73 34 migration rate-257 per 1 , O O k Temporary 94 2.3 70 28 probably because it covers a five- Destination year window. Rural 6 3.4 80 28 33. Migrants from estate Urban 77 2.3 74 28 sectors are diverse in their Abroad 17 2.7 51 32 characteristics. A majority o f All migrants 100 2.4 70 28 migrants have gone to urban n.a. Not a ~ ~ l i c a l . 1 Note: The results are applicable to sample estates only, not representative areas of Sri Lanka to find work in ofthe estatesector. the past 5 years, and a sizeable Source: MOP Estate Survey (2005). 95 proportionto other countries (Table 8-4). Most are temporary migrants, returninghome for a little more than 2 months per year on average. Overseas workers are almost exclusively female, following the national pattern: women accounted for 64 and 53 percent o f all externalmigrants in 1996 and 2003, respectively (SLBFE, 2004). Most o f overseas employment is unskilled; domestic employment (primarily inthe Middle East) accounted for 82 percent of the jobs taken by female migrants overseas. 34. The pattern of migration is also linked to migrants' life cycle-particularly internal migration, which is highamong estate youth (Annex 8, Figure A-8.2). Migration for employment i s most common among those in their twenties-13 percent of this cohort lives outside estates. The qualitative study suggests that many residents work out of estate during their prime years, and then return. Male migrants start returning home around the age of 30, while female migrants mightretumhome sooner to marry. 35. Evidence fromthe Estate Survey makes it clear that the two groups of migrants-overseas and internal (mostly to urban areas)--are distinct in terms o f characteristics of the migrants themselves as we11 as their households of origin. The two groups are also almost mutually exclusive-at the time of the survey, less than 1 percent o f households had both internal and overseas migrants. Overall, 17 percent of estate households had some form o f migration activity within the past five years, with internal (predominantly migration to urban areas) and overseas migration accounting for 13 and 4 percent, respectively (Annex 8, Table A-8.4). 36. Foreign migration, remittances and welfare. Remittances from migrants living overseas seem to have highbenefits for households. Multivariate regressions show a higher AI associated with recipient households (controlling for a number of household, estate and location Urban Sri Lanka Abroad characteristics). But netting out Remit mig- the impact o f remittances, a Average family member workmg abroad has almost no effect (Annex 8, Note: Results applicable to sample estates, not representativeof the sector. Table A-8.5). Table 8-5 shows Source: MOP Estate Survey (2005). that households receiving overseas remittances have significantly higher average AI compared to all other households. 37. Who is most likely to migrate overseas?' Members o f larger households are more likely to migrate overseas, probably because larger families have a greater ability to diversify their income sources.1oOverseas migration also tends to be more common in larger estates, perhaps because of better access to networks and recruitment efforts. Overseas migrants also tend to come from households whose heads are less educated?probably reflecting the unskilled nature of most of thesejobs (Annex 8, Table A-8.4). 38. Interestingly, the average earnings o f family members (excluding migrants) of households with overseas migrants are higher than that o f other estate households. Their livelihoods are also more diversified and secure-a higher percentages o f members work off-estate inindustry, trade, and services, and hold regular employment in the estates-and even more so for households o f foreign migrantswho send remittances regularly (Annex 8, Table A-8.7)." For regressionresults pertainingto results inthis paragraph, see Annex 8, TableA-8.6. 10 Householdsize inthis particular analysis refers to the size before migration. Inthis context off-estate employment refers to employment while living on estate, not employment as migrants. 96 39. There are two possible explanations for these findings. First, qualitative evidence suggests that overseas remittances help develop other livelihoods among recipient households, by financing activities such as starting retail shops, hiringthree-wheeler vehicles, and migration o f other members to urban areas. Second, since the cost of overseas migration tends to be high, households receiving remittances could have had more financial resources to start with which enables them to invest in migration, thus weakening the observed link between overseas remittances and welfare. Both these explanations are likely to be true to some degree, but a definitive answer requires more information. The high cost of migration also helps explain the selective pattern o f overseas migration. Evidence from qualitative interviews of estate residents indicate that overseas migration i s seen more as a temporary diversification strategy than as a coping strategy by households, and the indication that households with more resources are likely to finance migration i s consistent with this evidence. 40. Internal (urban) migration and welfare. Multivariate regressions show that having a household member who has migrated to an urban area has no signiJcant association with the household's AI, irrespective o f whether the migrant sends remittances regularly or not (Annex 8, Table A-8.5). The average AI i s very similar for households with remitting and nonremitting urban migrants (Table 8-5). 41. What makes receiving remittances from internahban migrants uncowelated with higher AI inthe Estate Survey sample, even though most conventionalwisdom suggests that remittances are used to add to household assets, which is precisely what the AI indicator measures? The answer partly lies in the more heterogeneous nature o f internal migrants and their households. The evidence suggests that urban migrants who remit regularly come from relatively poor households, and their remittances are often insufficient, given their limited earning capacity, to lift their household members out o f poverty. This would explain why looking across households with remittances from internayurban migrants and those without, there is no discernable difference in average AI. Furthermore, migration to urban areas seems to be motivated by more than short-term monetary benefits, unlike most overseas migration. The qualitative study indicates that urban migrants hope to find opportunities for employment, develop slulls, or escape workmg in estates, which is deemed degrading by many, particularly youth. Such factors would weaken the link between having urban migrants from a household and the household's wealth or poverty status. 42. Who is most likely to migrate internaZZy?'2 Migrants to urban areas, unlike overseas migrants, belong to households with relatively higheducation levels, and a higher education most likely increases access to information, linkages, and networks necessary to find urban jobs. Larger household size and residency in estates for more than 20 years is associated with higher likelihood o f migration to urban areas, while the presence of young children reduces the likelihood. Unlike households with overseas migrants, the nonmigrant members o f households with urban migrants do not have significantly higher earnings or more diversified employment than those with no migrants. 43. Overall, the analysis of migration suggests that households consider overseas migration as an investment in a strategy to diversify income sources, with remittances being the "returns" to that investment. While remittances seem to improve welfare, better-offhouseholds are better able to afford the initial cost o f overseas migration. In comparison, internal (mostly urban) migration is more heterogeneous in terms o f the motivation to migrate and the characteristics o f migrants and their households. Inaddition to the motivation to diversify households' sources o f livelihood, urban migration occurs for reasons as diverse as seeking future opportunities, coping with poverty or seasonal unemployment in the estates, and escaping the much-stigmatized estate employment. 12Refer to Annex 8, Tables A-8.6 and A-8.7 for results to support this paragraph. 97 Qualitative insights also indicatethat longer-term internal migration is more successful as a driver o f positive change, as it allows the households to develop sustainable sources of income as well as social networks. 44. Nevertheless, significant obstacles to mobility or diversification remain, which is apparent from loolungat the simple statistics on migration. Inspite of the rapid increase inmigration from estates reported by CFSES in recent years, both internal and external migration rates are well below that for the rest o f the country, even though poverty rates in the estates are much higher and as such should provide greater incentives to migrate. Constraints to mobility of the estate population are likely to be linked to a long history o f geographic, social, and economic isolation. The Estate Survey finds that 81 percent o f households sampled have lived inthe estates for more than 20 years, indicating that a large share o f the population will probably have minimal links withnetworks and information sources inthe outside world, which are often critical to migrate or find outside employment. 45. The correlates of poverty described so far are consistent with much o f the qualitative evidence on the different economic groups and their characteristics. Focus group discussions (FGDs) revealed three categories o f households and that attributes o f the bottom group include high dependency, unstable or limited source o f income and ill-health, poor quality housing and facilities, and frequent mention of alcoholism. By contrast, attributes o f the top group include well-diversified and stable sources o f income, regular remittances from abroad, and access to productive networks.l3 Estate or community characteristics that matter for welfare Estate Survey is drawn from five (percent o fpopulation below 3 0 percentile o fasset ~ districts with significant estate populations. index) Table 8-6 shows that the AI-poverty rates are much higher for the estates in Ratnapura and Kegalle, and lowest for Kandy. This pattern persists even after controlling for other correlates o f poverty, while the correlations between crop or management type and AI poverty disappear (see Annex 8, Table A-8.1). Table 8-6 shows that the associations between type o f crop or management and AI-poverty incidence are ambiguous. This is also consistent with the findings o f the qualitative study. 47. Remotenessofestates. Although 77 percent o f estate households inthe sample live within 10 km. of a town, 42 percent o f households cannot use the road to the town all year round.This indicates a high degree o f isolation o f estates from markets and employment opportunities. Recognizing that the quality o f roads also depends on the geography of the locale, Figure 8-4 shows the AI-poverty rate by district and road quality. Inall districts, households inestates where roads to town are passable all year tend be better off. Moreover, after controlling for other estate, location, and household attributes, having an all-weather road connecting the estate to the nearest 13 The households assessedas being at the "bottom" reported long-term deprivation and felt that they were at the bottom o f the community group. The "middle" group reported that rather than face key deprivations they were constantly balancing the upward and downward pressures. The "top" level reported themselves as progressive "movers" and better-off than most inthe community. 98 town is associated with a 10-percent lower probability o f a household inthe estate being AI-poor 48. These results suggest that lack of road connectivity of an estate (throughout the iTigure 8-4: AI-poverty in estates by road quality year) with a town or market is an important Road to nearest town Dassable all year? constraint to the residents' upward mobility. The negative impact o f poor connectivity - 60 also came up in qualitative interviews- -s especially in the context o f factors that 40 > contribute to isolation fromjob opportunities P 20 and services outside, and the mainstream a economy in general. The results also suggest 0 that poverty incidence inthe estates is higher when they are located in poorer districts, such as Ratnapura and Kegalle (see chapter 2). t iource: MOPEstateSurvey (2005). What determines economic mobility of households 49. The correlates of AI poverty, supported in some cases by qualitative insights on what households identify as drivers o f change, point to factors likely to affect economic mobility o f estate households. Diverszfiing the household livelihood portfolio appears to be an effective path out o f poverty, particularly when incorporating income from enterprises and regular and slulled employment outside the estate. Overseas migration, and the remittances it generates for households, i s also an effective diversification strategy. Internal migration to urban areas may not yield significant remittances, but can provide other benefits in the form o f future opportunities, skill development, and as a strategy to cope with poverty or seasonal unemployment in the estates. A number o f other factors aIso play a role, primarily by affecting households' ability to diversify or improve income sources. Education improves the prospect o f more regular and remunerative employment, particularly outside the estate; and qualitative findings indicate that households perceive these benefits and look upon improvements in access to and quality o f education as a force for positive change. Not having a NIC appears to hinder education and employment prospects outside the estate, especially among youth. Lack of connectivity to towns by a road that is useable throughout the year -a relatively common problem inthe estates-limits opportunities for households to access markets and employment. 50. The qualitative study also suggests additional factors that communities and households themselves describe as important for households' upward or downward mobility. Wages, availability of work and cost-of-living appears to be one set o f factors: rising cost o f living unmatched by wage increases and lower availability o fwork were frequently mentioned as strong constraints to upward mobility. There were some differences between the tea and the rubber sector as the workers inthe rubber sector reported improved work availability due to changes in agncultural techniques and marketing. 51. Health shocks and access to healthcare. I11 health and death o f a family member were identified by households as important downward drivers, so much so that these can frequently override strong upward drivers such as a diversified livelihood portfolio and low number o f dependents. The risk o f chronic illnesses, especially among income earners i s a critical determining factor in the household's economic future. The type and quality o f healthcare available within the estate were also identified as influencingthe direction o fwelfare change. 52. While there are gaps between the estate sector and the rest o f the country, overall access to basic health services in estates seems fairly high. The Estate Survey shows that availability of a 99 doctor, nurse, or midwife is highbut far fromuniversal, and that it varies across different types of estates. Utilization o f maternal and child health services are hightoo, althoughlower than the rest o f the country. Ailments that require hospital visits pose special problems. The average travel time for a one-way trip to the hospital can often be over an hour, and more than 1.5 hours for those who live more than 5 kilometers from the nearest town, no surprise given the connectivity problems above. Furthermore, the qualitative study indicates a highdegree o f dissatisfaction with the quality of health services, and a perception that quality declined after privatization o f estates (see Box 8-3). percent have access to a nurse; and 77 percent have a heaith clinic located on the estate, Despite variations inavailability of medical personnel and health fac es across different types o f estates, women's use o f maternal and child health services is high, although lower than the rest of Sri Lanka. Most children have health cards and undergo growth monitoring, and have completed their vaccinations by the age o f one. In case o f aifments, estate households were most likely to consult a government-run faciliry (40 percent as conlpared to 14 percent who consulted an estate-run facility). Utilization of preventive health services is lower in estates than in other sectors (DIIS, 2000). Estate mothers are less likely to receive prenatal or postnatalvisits by a midwife or medical officer, or be advised on symptoms o f pregnancy complications. Although 80 percent of estate women give birthina hospital, this compares poorly with almost 100percent inother sectors. Ql,tnlitJi.The qualitative study also reveals widespread dissatisfaction witli quality of health services: many viewed the lack of trained staff, drugs and fimctioning equipment as a factor contributing to the deterioration o f health care in the estates. A direct link was often perceived between privatization and the deterioration of health facilities. The female focus groups were particiilarly critical of the elimination of free nutritional supplements to children and changes in maternity care in some estates, which they associated with productivity standards and rules introduced by h e privatized management in recent years. Privately owned and managed estates were most likely to be the most poorly served. The Estate Survey partly validates this obse~at~on-privately managed estates have lower availability of doctors and health clinics (but not midwifes) than Regional Plantation Companies. 53. Housing and sanitation. At the community level, the general condition o f estate housing (particularly line rooms) i s seen as a contributing factor to poverty. However, at the household level housing, together with related facilities o f toilets, water supply and electricity were frequently seen as improving and contributing to upward mobility. This i s consistent with the evidence presented earlier indicating improvements inestate housing conditions (Table 8-2). The improvements in housing stock were generally attributed to households' own efforts, although some acknowledge contributions from management. The Estate Survey indicates that housing programshavebenefited only 15 percent o f households inthe sample inthe past two years, which seems to be consistent with these perceptions. Sanitation programs turn out to be far more prevalent, with 24 percent o f households reporting benefits. The qualitative study indicates perceptions that sanitation conditions have improved in the estates over the last 15 years, which was also linkedto improvements in education and awareness (Box 8-4). Box X-4: fmprovenientsinsanitation and healthinthe last 15 years: Key informants' perspectives Key informants in RPC estates and in CoIonibo pointed out that health and sanitation conditions had improved considerably in the last 15 years. Inthe tea sector a significant minority also acknowledged the role of NGQs in improving living conditions, health, sanitation, and education in estates. In the FGDs, 100 the situation has improved, there IS a better interest, awareness and receptivity for family planning. births are spaced out and f`atnily planning practices have improved. linprovemenkscan be attribured to education (especially of youth) and the media; almost every house has a `I'V ." (C`Yf-+,Rubber, RPC, Ratnapura) `"Two new schools were started in 1993 and 1995. In 1998 we got etectncity In 2000 the government took over our haqxtal and upgraded it to a ruralhospital for the area...."(Female FGD, Tea, RPC, NuwaraEtiya). Success of welfare ~ n t e ~ e nwereomeasured by improved levels of awareness - especially inrelation ~ i ~ to health practices and placing high priont-y on the schooling of children. Programs that focused on clianging current attitudes andbehavior patterns were identified as important contributors to change, "In the past, general awareness levels were very poor, there was no knowledge of safe sanitation and related practices, children did not attendschool. This lack of knowledgewas identified as a key gap and we worked to fill this need. 'This year, a volunteer group of youth has been trarned to conduct awareness workshops. These groups organizednutntion interventions and competitions for home gardens." (CTL, Rubber, RPC, Kegalle) Source: CEPA (20051. 54. Alcoholism was widely seen as hindering upward mobility at the community and household level. Alcoholism adversely affects households' earning capacity, expenditure, and education o f children, creating intrahousehold conflict and disruptingcommunity life. Men-who are the primary consumers-tended to underplay alcohol consumption, abuse, and its effects; while women, youth, welfare officers and the estate management discussed it at length. About 80 percent of respondents in the Estate Survey report that alcoholism i s a problem in their estates, and 75 percent o f community informants report n o improvement over the last 15 years. The increasing availability or supply of alcohol was viewed as creating and increasing the problem; and a majority of community informants identified the sources to be illicit brews inside and outside the estates. Many better-managed RPCs are attempting community-level solutions to the problem, and frequently sought the participation ofyoung people insuch programs (see Box 8-5). - Box 8-5: The negative tonsequences of alcoholism in the estates Women, youth, management arid welthe officers \yere partiailarly critical of parznral and adult alcohol constimption (with the associated negative social consequences), and the increasing availability arid siipply o f alcohol as a direct reason for deterioration within a given community. Better-managed RPC's are attempting community-level solutions to the problein, includiiig frequent involvenient and participatioii 01' youth. "LVithin the estate we have two bars W e n men get their .;alary, the first thing they do is to go to the bar. `fie family conies atter that Peopleare addicted to 'kassippu' (moonshine)." (Female PGD, Tea, RtY', Nuwaru Eliya). "Most of the parenis are addicted to alcohol and children face lots of problems due to this. ....'This causes lot of problems in the night and children can`t even study propcrly. 'I'he problcrn is gctriny worse day by day. In some families both parentsdrink." (Youth RiD,rlccl, KI'C) "When talking of poverty, alcohol is a big issue......Kassippu (inooiishine) is brcwd and sold in the neighboring villages, now even `iced packs` are available. 1.as laws contributc to its conrinuance.The tine (tbr brewing and sale) is Rs 10,00(t-rhc dcalcr sends one of his assistants to prison and continues to sell. The vendors are seen hovering around on paydays to collect debts....." (C`TL. Rubber, RPC, Ratnclpurli) "Though both nlalcs and females drink, thc incidence is lower here than other estates. We worked with the people and took the initiative to arrange for police interventions. Children`s ducation and awareness has led to then] playing a lead role on educating their parents on the illeffects of alcohol. The police station is close to this di\.tsion this may have also contributed to the successof the programs." (C`TI., Ruhher, KPC`, Kegaller Sotirce: CEPA (2005). 55. Organizational structure of the estate sector. The empirical evidence on the nature o f poverty in the estates and the qualitative insights from households, management, and other key stakeholders seem to suggest that the organizational structure o f the sector i s a critical constraint to poverty reduction. The structure o f the "plantation system" marginalizes estate residents from the mainstream, as seen from the perceptions of estate residents themselves, as well as actual evidence on lack o f connectivity to towns, spotty coverage o f NICs, and inadequate services 101 including welfare programs (see section 8.3). The qualitative analysis finds that the sense o f marginalizationadversely affects economic decisions o f individuals; for example, leading them to reject estate work even when it makes sense economically, because it i s regarded as degrading. 56. The system also tends to create tensions arising out o f an adversarial but dependent relationship between workers and estate management, which explain fundamental issues of mistrust between workers and management. The qualitative interviews revealed little satisfaction among residents with regard to any form o f estate management-lack o f competence in production and lack o f care in human resource management were mentioned repeatedly as downward drivers in all but a few estates. Trade unions were acknowledged for their role as representatives of the workers to the management, but there was strong criticism o f what was perceived to be the self-serving nature of the unions and leaders, as well as the lack of representation (see Box 8-6). Box 8-6: Perceptionson managementand trade unions in the estates Little satispdction was expressed with regard to any form of estate maiqernt`nt. A considerable iiuinbers of -l focus group discussions reflected the view that the estate was deteriorating as a productive enterprise due to the management's overall lack of care and competence. There -ere a few individual instances of managers who successfully managed the workforce and were seen by the community as a positive influence ontheir lives. Inthese cases, the workers relatediothe specific manager rather than management at-large. "We cannot hope for a good future for the estate. The rnanagernent ISresponsible far that. They are not caring for the tea bushes (no pruning, fertilizing) and the crop is going down every year. Along with that, our income is going down too. The managementhas no knowledgeandthe bushes are not maintainedproperly. They havegrown tall and it is difficult to pluck......."(MaleEGD, Tea, RPC, Bactulfa). "The managementis not proper. 10 do work that can be done by 4. The estate is running at a toss. That's why we`re not given any facilities. The conipany estates are better. That's becausethcy maintainthem well, For 5 years we didn't get EPF (Employees' ProvidentFund) because it hadn't beenentered." (Male FGD, Tea, State, Kandy) "Compared to other factories this factory is better because the officers take care of us. Ifsomeone does not haveajob Evidence and perceptions on estate poverty trends 57. The discussion so far also has some implications for the question o f what may have contributed to the increase in consumption poverty in estates fiom 1990-91 to 2002 (as reported in the HIES). These "snapshot'' views of economic conditions cannot fully explain time trends, but they can provide useful clues. The perceptions o f estate residents themselves on the direction o f change bringa fuller understanding o f dynamic shifts within the sector. 58. As mentioned above, the full distributionof consumption is a far better indicator o fwelfare changes than the poverty headcount. And this suggests a small shift, rather than a drastic worsening o f the entire distribution from 1990-91 to 2002. The CFSES does not offer a clear story on why this occurred. On the one hand, the proportion of self-employed workers in the estate sector increased from 3 percent in 1996-97 to 10 percent in 2003-04; and more workers from estates now work in outside sectors like services and industry. On the other hand, regular employment in estates has shrunk from 68 to 49 percent, while casual employment has grown 102 from 29 to 41 percent. This trend also appears to be consistent with households' perceptions about lower availability of work (as reported above)--probably due to some estates preferring to employ more casual labor than registered labor, which allows them more flexibility during less profitable periods. Given the analysis in section 8.2, the first trend is consistent with a reduction inpoverty, andthe second with an increase inpoverty. 59. The CFSES reports that the number of income earners per household fell in the estates from 1996-97 to 2003-04 (from 2.3 to 1.7 per household), but remained unchanged for the country as a whole. This trend may have contributed to the rise in poverty in the estates during the 199Os, given that the analysis here indicates that number o f income earners in a household is associated with higher AI poverty. Interestingly, this has occurred even as the age dependency ratio has declined inthe estates (58 percent in 1996-97 to 55 percent in2003-04), as it has for the rest of the country. The fall in the number of income earners has meant that even as (age) dependency has declined, the dependencyper income earner has increased for the estates, while it declined for the rest o f the country. More analysis is necessary to understand why the number o f income earners declined, what effect this has had on household welfare, and whether it is a trend likely to continue. 60. Perceptions of changes within the estate sector. Overall, the residents perceived deterioration in living conditions on the estates over the last 15 years. In contrast to this, there was strong consensus among individual households that housing, sanitation, and access to education had improved. The difference in the perception at the community versus household level i s partly explained by the increasing role o f non-estate employment, as seen by the employment trends from CFSES. This may have served to de-link the fortunes o f the household from that o f the estate community to a certain extent-namely, when the estate i s not doing well, households can still increase their income from external sources. There is a clear pattern of positive perceptions about improvements in community and households for RPCs versus privately managed estates (see Box 8-7). The qualitative study often refers to poorer quality and availability o f housing and healthcare facilities inprivate estates, which are consistent with these perceptions. While the Estate Survey results do not suggest actual systematic differences in availability of facilities or outcomes by management type, these numbers also do not capture intangible issues, like poor quality o f services or management competence, that often dnve perceptions. 61. Attitudes of youth us an indicator of trends. An important insight that emerged from the qualitative analysis and the statistics on migration (section 8.2), is that educated estate youth are often not willing to work on the estate, primarily due to the stigma associated with it. At the same time, salaried employment in the non-estate sectors was not easily available to the estate youth, 103 relegating most to unskilled or semi-skdled jobs. This could explain why urban migration may not result in immediate welfare improvements for households. Outside employment may not lead to increase in earnings, although it still may be attractive to an estate youth as a longer-term investment to develop skills and links with outside markets. 62. The dislike for estate work was so strong that a number of youthFGDrespondents reported remaining voluntarily unemployed waiting for a job that matched their aspirations, or taking temporary employment. This phenomenon might explain the fall in number o f income earners reported in the CFSES, and i s consistent with the labor shortages reported by some estate managers. The general perception was also that youth mobility has improved, especially in the rubber sector where the estates are located close to rapidly developing townships. The empirical results in this section suggest that l o w educational attainment is a key obstacle to any such mobility. The youth respondents themselves cited marginalization due to their Indian Tamil heritage as "estate worker" as an obstacle to gaining opportunities, even in instances where they possessed the required qualifications and expertise. 63. Thus, perceptions o f estate residents about their own direction of change are somewhat at odds with the statistical poverty trends reported inthis chapter. This mayjust be due to the small sample in the qualitative study and the possibility that may not be representative o f the population. Another explanation could lie in the trend o f increasing opportunities for estate residents (including youth) to diversify their livelihoods and migrate outside the estates over the last 15 years. Although outside employment may not have resulted in an improvement in household's earnings, they tend to represent greater integration with the outside world as well as expectations for better employment in future, both o f which may translate to better perceptions among households about their current economic prospects. 64. At the same time, the perceptions were largely negative about conditions inthe estates, on past and future prospects. This must be understood inthe context o f factors discussed above: the adversarial relationshp with management, disillusionment with trade unions, and a deep sense o f marginalization caused by historical factors as well as the negative view o f estate work. The perceived way out o f poverty therefore seems to primarily rest inhouseholds' ability to diversify out o f estate work, particularly among the younger generation; and the empirical evidence in this chapter largely supports such perceptions. However, the evidence also indicates that there are a number o f constraints to households' ability to diversify effectively, which could be addressed by appropnate policy initiatives. 8.3. Social and welfareprograms in the estates 65. Given the poverty and human development challenges faced by the estate population, social and antipoverty programs can also serve as critical drivers of change. It is therefore important to examine the coverage and access to such programs in the estates-to identify patterns and potential gaps in coverage. This will also help us understand whether such programs can reduce the extent o fexclusion and marginalizationreportedby estate residents. 66. Cash transfers: Samurdhi and social welfare. Samurdhi cash transfers-as described in chapter 2--constitute the largest welfare program in the country. Coverage o f the estate population by Samurdhi and social welfare transfers (combined) appears to be low. Only 13 percent o f households in the Estate Survey sample report receiving any cash transfers fiom the government-less than one-half of the HIES-based poverty headcount rate of the sector, and in stark contrast to the 40-percent coverage o f Samurdhi for the entire country. The transfers do not appear to be well-targeted (Annex 8, Table A-8.8). While AI i s less accurate as a measure o f poverty status than consumptionexpenditure, one would still expect better incidence than what i s 104 seen here: 28 percent o f beneficiaries belong to the bottom two AI quintiles, compared to 23 percent inthe top two AI quintiles. 67. There i s wide disparity in coverage by district and management type (Annex 8, Table igure 8-5: Coverage of cash transfers inestates A-8.8). For example, 25 percent of sample percent) households in state-managed estates receive transfers, compared to 12-13 percent in RPCs 40 and private estates. The extent o f mistargeting 30 also varies widely by district and management 20 type: 42 percent o f state-managed estate 10 households in the top AI-quintile receive the 0 transfer, compared to only 7 percent o f RPC- managed estate households o f the same quintile; and 40 percent o f households in the j @Total (k)6RchestQuintile~ top quintile in Kandy receive transfers, compared with 11 percent in Kegalle and mrce: MOPEstate Survey (2005). Ratnapura (Figure 8-5). 68. Possession of a NIC appears to matter for coverage: 13 percent o f households whose head has a NIC receive transfers, compared to 9 percent of those who do not. Interestingly, the coverage o f the bottom three quintiles are noticeably higher for households whose heads have NICs than those who do not (Annex 8, Figure A-8.3). 69. Nutritional supplement through Triposha. Given the prevalence of malnutrition among estate women and children, the Triposha Program (see Chapter 2 for a description) i s highly relevant. The Estate Survey indicates Triposha coverage i s highinthe estates: about 80 percent o f new mothers and 66 percent o f young children reported receiving this supplement, although the proportion was lower in privately managed estates (62 and 47 percent, respectively). 70. Other social programs. The Estate Survey covered six broad types o f social program: housing, toilet and water supply (sanitation), training and awareness, microcredit, cr&che/child care facility, and early chldhood development services. Government programs are found to be the most prevalent. With the exception o f sanitation programs-which are relatively abundant and regarded positively by many households (as reportede a r l i e r h v e r y other type of program is available to less than 50 percent o f the sample households, and has a participation rate o f 15 percent or less.14 The least established programs are microcredit and childcare/cr&che, with participation by only 10 percent o f households (Annex 8, Tables A-8.9 and A-8.10). 54 percent o f households are found to notparticipate inor benefit from any program. 8.4. Implicationsfor poverty reductionpoliciesfor the estatesector 71. The long-term economic prospects o f the estate population are closely linked to the broader issues about future o f the industry, which would be the subject o f a separate study. Although structural changes in the sector can occur only in the medium to long term, more immediate welfare gains can come from addressing factors identified as critical dnvers of economic mobility. The analysis in this chapter suggests that these can be broadly characterized as facilitating mobility and migration, encouraging self-employment and alternative skills development, and expanding the provision of state welfare services and other socialprograms. A wide range o f interventions can support these objectives: connecting estates better with nearby l4A program is considered "available" to all households in the community if at least one household from the community participates inor benefits fromthe program. 105 towns; improving the coverage of NICs, particularly among the youth; increasing access to and quality o f health and education; developing programs to increase knowledge and sktlls in alternative economic activities; and tackling alcoholism by involving communities and women and youth inparticular. 72. Social and wegareprograms can also play an important role in e#ecting these changes. The coverage of public social assistance programs, with the exception o f Triposha, is inadequate. The coverage o f Samurdhi and other transfer programs are much below the poverty rate of the sector. The proposed revamping of the Samurdhi targeting system (see chapter 2) may increase coverage and improve targeting among the estate population. Other social programs-such as housing, livelihood generation, childcare and early childhood development-also seem to have wide gaps incoverage. Expansion of rnicroJinance inparticular, which currently benefits only 10 percent of households (in the Estate Survey sample), can yield sizeable benefits. This chapter suggests at least two areas where better access to finance can have significant impact: setting up microenterprises and financing overseas migration, both of which are likely to have initial fixed costs that preventpoor households from making optimal choices. 73. L o w rates of participation by households, even when the programs are available in their communities, raises more questions: the problem may involve inadequate scales of these programs or difficulties in mobilizing estate communities-especially for programs like microcredit that rely on community participation-or a combination of both. The success of such programs will depend on sorting out such issues of access, scale and participation, and design interventions accordingly. It will also be important to learn from success stories-identified by communities in the qualitative study as the programs that built awareness to change attitudes and behavior patterns relating to sanitation, nutition,and schooling. 74. Role of the estate management. The estate management's incentives to facilitate such changes may be limited, because o f perceptions that these may further reduce the availability o f labor for the estate. On the other hand, by actively encouraging the drivers o f change identified above, the estates can contribute to a positive image as an employer and improve the status o f estate work. This would be in the long-term interest o f the sector, which needs to be able to attract workers from the labor market. The qualitative study revealed encouraging signs that some well-managed estates are increasingly seeing the value intaking such steps. 75. The long-term solution to poverty in the estates clearly lies in mainstreaming the sector. Perhaps the most enduring link between the current system and the enclave plantation past i s labor living within the commercialproperty, which limits movement o f workers and marginalizes the population. Severing this link, perhaps by providing land rights to long-term residents, would relieve management o f welfare responsibility toward residents and the obligation of residents to provide labor to the estate. This would help mainstream the sector by changing the current parameters of the employer-employee relationship. 76. Such a solution will however not be easy to implement, and it i s likely that large increases inlaborproductivity will berequiredfor the industryto remainviable. To achieve that, itwillbe necessary to take a broad view o f the sector and identify the changes that will be required. Current shifts in tea and rubber production occurring naturally in Sri Lanka-seen inthe higher productivity and increasing share o f smallholdings in total production-hint at the kmd of restructuringthat may be necessary. The survival o f the industry, as well as the long-term welfare o fits labor force, will depend on its ability to re-invent itselfto achieve higher productivity. 106 Epilogue 1. Many studies show that Sri Lanka has been more successful at achieving human development and less successful at reducing income poverty. Certainly economic growth has been slower than East Asian countries with comparable levels o f per capita income in the 1960s and a similar trajectory of human development. Povertyreductionhas been troublingly slow even for the growth achieved (averaging around 3 percentper capita annually during 1991-2002) due to widening inequalities between households and across regons. Western Province experienced growth three times that o f the rest o f the country (excluding the North and East) during 1997- 2003. 2. A primary focus ofthis report has been to understand what explain the highand widening inequality between and within regons or sectors; and how can economic growth be made more inclusive of laggingregions and sectors. Individual or household-specific factors associatedwith the economic status of a household include education, employment or occupabon and family size. Taking the differences across individualshouseholds and combining this information with where a household i s located yielded the best insights. The spatial disparities in poverty mirror differences between districts (or at a more disaggregated level, D S divisions) in attributes like connectivity to towns, access to infrastructure such as electricity, and average educational attainment. Another study by the ADB and World Bank (2005) also identifies similar factors, like energy availability and transport, as constraints to business development - and therefore limiting economic growth-outside the Colombo area. 3. The poor suffer from specific disadvantages that limit their future productivity and incomes and thus contribute to poverty traps. High dropouts among children o f poor families at secondary and tertiary levels destine this population to low educational attainment, and low lifetime earning potential. Low birthweight and malnutrition among poor children and mothers often affect lifelong and cross-generational earnings. While the incidence o f malnutrition links to poverty, more in-depth analysis will yield useful information on its types, patterns and causes. For example, many poor children drop out of school before Grade 5. Further evidence i s also necessary to understand what factors cause many poor children to drop out o f school evenbefore Grade 5 -low quality of schools, particularly outside urban areas, and the inability of the poor to compensate for poor schooling with private tutoring can be one explanation. Improving the quality o f learning among the poor thus constitute a significant policy challenge, along with other factors that may contribute to low educational attainment andnutritional status. 4. Concentration of poverty inparticular regions or sectors stems from constraints specific to that group (chapters 6 to 8). At the same time, a few crosscutting issues affect all sectors and regions (chapters 3 to 5). Improvements in quality of education, connectivity to markets and urban centers, quality, and availability of infrastructure like electricity and financing for micro- enterprises all expand opportunities for the poor and lagging regions/sectors. Such Opportunities can occur in different forms - by finding work inhigher-payingjobs perhaps or by migrating to urban areas, or starting upmicroenterprises, or expanding enterprises to access to larger markets. 5. Many of these opportunities have cross-sectoral implications and require coordination across sectors. The issue of migration from lagging areas to Western Province underscores why. Migration offers a viable means for upward mobility to those living in laggng regons, and improving the quality of education inthese areas i s likely to improve the ability of people to take advantage o f this opportunity. But part of the welfare and growth potential from the migration process i s lost because o f "costs" associatedwith overconcentration inthe Colombo urban area, where almost all the in-migration occurs (chapter 4). Thus, the development o f alternate growth centers or secondary cities and a strategy for better urbanplanning mustbe coordinated with rural and estate development strategiesto bringlaggingregions into the path o f economic growth. 107 6. At the sector-specific level, limited and skewed growth inagncultural incomes has clearly limited poverty reduction and increased inequality sharply in rural areas. Improving agncultural productivity is thus a key issue, given that 88 percent of the poor live inrural areas and 58 percent o f rural households derive at least some income from agriculture. The rural nonfarm sector is a bright light on the horizon and has great potential for creating employment and raising incomes. Addressing the constraints that face rural entrepreneurs - which includes those mentioned in paragraph 2 above -IS thus critically importantfor poverty reduction(chapter 6). 7. The conflict-affected areas of North and East are characterized by various economic and social gaps between this region and the rest o f the country (chapter 7). The challenges are especially severe in the case of Eastern Province, which suffered extensive damage by the tsunami in 2004. The cease-fire o f 2002 brought enormous economic benefits to the regon, which underscores the importance of sustaining peace. Improvements in the security situation will also permit better data collection fromhouseholds inthe region, which will greatly aid future analysis andpolicymalung. 8. The highvulnerability and stagnant incomes o f the estate population are closely related to their social, geograpluc, and economic isolation. Structural changes inthe sector that will address these issues permanently, such as moving away from the resident labor structure, can occur inthe medium to long term, as constraints to productivity and profitability o f the industries are overcome. Inthe interim, welfare improvements inthe estates can be achieved through drivers of positive change that, broadly speaking, facilitate mobility and diversification of livelihoods among estate residents (chapter 8). 9. Many o f the issues highlightedinthe report are broadly consistent with key elements o f the government's strategic vision for development articulated in the President's manifesto Mahinda Chintana. The Chintana's vision for regional development--of improving road networks and access to electricity and finance in rural areas-is consistent with the needs o f lagging areas and sectors outlined in the report. Specific actions arising from this strategic plan include the government's efforts to improve the poverty impact of public expenditures, including its plans to improve the targeting of the Samurdhi transfers program countrywide. Any improvements in the effectiveness o f welfare expenditures will directly benefit the poor. 10. The upcoming surveys plannedby the government, such as the DHS2006-07 and the HIES 2006-07, will add a wealth o f new information to fill existing knowledge gaps and advance the necessary policy debates and strategic thirhng on economic development that will arise in the years ahead. 108 References Alderman, H.,J. Behrman, V. Lavy and R. Menon (2001). "Child Health and School Enrollment: A Longitudinal Analysis". Journal of Human Resources, 36(1), 185-205. Arunatilake, N., S. Jayasuriya and S. Kelegama (2000). The Economic Cost of the War in Sri Lanka. Colombo: Institute o f Policy Studies. Asian Development Bank (2005). Key Indicators 2005: Labor Markets in Asia: Promoting Full, Productive, andDecent Employment. Manila: Asian DevelopmentBank. Asian Development Bank and World Bank (2005). Sri Lanka: Improving the Rural and Urban Investment Climate. Colombo, Sri Lanka: World Bank. Asian Development Bank, Japan Bank for International Cooperation, Japan International Cooperation Agency and World Bank (2005). Sri Lanka 2005 Post-Tsunami Recovery Program: Preliminary Damage and Needs Assessment.Colombo. Barker, D.J.P. (1998). Mother, Babies and Health in Later Life (2nd ed.). Edinburgh, London, New York: Churchill Livingstone. Basu, K., A. Narayan, M.Ravallion (2001). "Is Literacy Shared Within the Household?: Theory and Evidence for Bangladesh". Labour Economics, 8,649-65. Behrman, J. R., and M.Rosenzweig (2004). "Returns to BirthWeight". Review of Economics and Statistics, 86(2), 586-601. Bhalla, S. S. (1988a). "Is Sri Lanka an Exception?A Comparative Study o f Living Standards", in Srinivasan, T.N. and P.K. Bardhan (Eds.), Rural Poverty in South Asia, 89-117. New York: ColumbiaUniversity Press. Bhalla, S. S. (1988b). "Sri Lanka's Achievements: Facts and Fancy", in Srinivasan, T.N. and P.K. Bardhan (Eds.), Rural Poverty in SouthAsia, 89-117.New York: Columbia University Press. Bhalla, S. S., and P. Glewwe (1986). "Growth and Equity in Developing Countries: A Reinterpretationof Sri Lanka's Experience". WorldBank EconomicReview, 1,35-63. Boissiere, M., Knight, J. B., and Sabot, R. H.(1985). "Earnings, Schooling, Ability, and Cognitive Skills". American EconomicReview, 75(5), 1016-30. Bourguignon, F. (2003). "The Growth Elasticity of Poverty Reduction: Explaining Heterogeneity across Countries and Time Periods", in T. S. Eicher and S. J. Turnovsky (Eds.), Inequality and Growth: Theory and Policy Implications. Cambridge: MIT press. Center for Poverty Analysis (2005). Moving out of poverty in the estate sector in Sri Lanka: Understanding growth and freedom from the bottom up. Background study for the Sri Lanka Poverty Assessment. Colombo: CEPA. Central Bank o f Sri Lanka (2005). The Consumer Finances and Socio Economic Survey Report 2003/04 Part I,Colombo: Central Bank o f Sri Lanka. Central Bank o f Sri LankaAnnual Reports (various years). Colombo: Central Bank o f Sri Lanka. Centre for Information Resources Management (2004). Vulnerability Poverty Profile, Various Districts. Trincomalee: NorthEastProvincial Council. Charles, S.H. (2002). Sri Lanka agricultural technology policy and its impact. Paper prepared for the Sri Lanka Agricultural Policy Review of the World Bank. Mimeo. 109 Datt G., and M. Ravallion (2002). "Is India's economic growth leaving the poor behind?" The Journal of Economic Perspectives, 16(3), 89-108. Datt, G. and M. Ravallion (1992). "Growth and Redistribution Components of Changes in PovertyMeasures". Journal of Development Economics, 38,275-292. Deaton, Angus and Jean Dreze (2002), "Poverty and Inequality in India: A Re-Examination". Economic and Political Weekly,September 7,3729-48. Department of Census and Statistics (1987). Census ofAgriculture 1982,Colombo. Department o f Census and Statistics (2002). Labor Force Survey in the North and East, Colombo. Department o f Census and Statistics (2003). Household Income and Expenditure Survey 2002, Colombo. Department o f Census and Statistics (2004a). Atlas on the Buildings Affected by the Tsunami 2004, Colombo. Department o f Census and Statistics (2004b). Census of Agriculture-Sri Lanka 2002, Preliminaly ReleaseN0.2, Colombo. Department o f Census and Stahstics (2004~).UfJicial Poverty Linefor Sri Lanka. Retrievedfrom http://7NWW.sfatistics.gov.lk/poverty/OffcialPovertyLineBuletin.pdf. Department o f Census and Statistics (2004d), Poverty Statistics/Indicators for Sri Lanka. Retrievedfrom http://www.statistics.gov.lk/poverty/PovertyStatistics.pdf. Department o f Censusand Statistics (2005). Census of Tsunami Affected Areas. Colombo. Department o f Census and Statistics (various years). Labor Force Survey in Sri Lanka. Colombo. Department o f Census and Statistics and World Bank (2005). A Poverty Mapfor Sri Lanka Policy Note, Colombo. DFID, UNDP, and UN-HABITAT/LTMP Urban Poverty Reduction Project (2002), Poverty Pro$le, City of Colombo: Urban Poverty Reduction through Community Empowerment, Colombo, Sri Lanka. Dreze, J. and A. Sen (1989). Hunger and Public Action, Oxford: Clarendon. Eelens, F., T. Schampers, and J. D. Speckmann (1992). Labour migration to the middle east: From Sri Lanka to the GulJ: London :KeganPaul International. Elbers, C., J.O. Lanjouw, and P. Lanjouw (2003). "Micro-level Estimation o f Poverty and Inequality". Econometrica, 71(1), 355-364. Filmer, D. and L. H. btchett (2001). "Estimating wealth effects without expenditure data--or Tears: An applicationto educational enrollments instates o f India". Demography, 38(1), 115-32. Gallup, J.L., J. Sachs, and A. Mellinger (1999). Climate, Navigability, and Economic Development. Unpublishedpaper, HarvardUniversity,MA. Ganepola, V andP. Thalaysingam (2004). Poverty and Conflict: A Review of Literature (Worlung Paper SeriesNo.1). Colombo: Centre for PovertyAnalysis. Gazder, H. (2003). A review of migration issues in Pakistan. Paper prepared for Regional Conference onMigration, Development and Pro-Poor Policy Choice inAsia. Glewwe, P., H. Jacoby and E. King (2001). "Early Childhood Nutrition and Academic Achievement: A Longitudinal Analysis". Journal of Public Economics, 81(3), 345-368. 110 Glewwe, P., M. Grangnolati, and H. Zaman (2000). who Gainedfrom Vietnam's Boom in the 199Os? (WorldBank Policy ResearchWorlung Paper 2275). Washington, DC: World Bank. Glinskaya, E. (2000). An empirical evaluation of Samurdhi Programme. Background Paper for Sri Lanka Poverty Assessment 2002, ReportN o 22-535-CE. Government of Malaysia (2001). Review of Second OutlinePerspective Plan, 1991-2000. Gunatilleke, G. (2000). "Sri Lanka's Social Achievements and Challenges", in D. Ghai (Ed.), Social development and public policy: A study of some Successful experiences, 139-189. Macmillan. Gunatilleke, Godfkey, et a1(2001). The Cost of the War: Economic, Social and Human Cost of the War in Sri Lanka. Colombo: National Peace Council. Gunatilleke, N. A. A. Cader, and M. Fernando (2004). Understanding the dimensions and dynamics of poverty in undersewed settlements in Colombo (Worlung Paper Series No.3). Colombo: CEPA. Gwatkin, D., S. Rutstein et al. (2004). Initial country level information about socioeconomic diffeYencesin health, nutrition andpopulation (2nded.). WashingtonDC: World Bank. Haddad, L. and H. Bouis (1991). The Impact o f nutritional status on agricultural productivity: Wage evidence from the Phlippines. Oxford Bulletin of Economics and Statistics, 53(l), 45-68. Haddad, L., H. Alderman, S. Appleton, L. Song, and Y. Yohannes (2003). "Reducing Child Undernutrition: How far does Income Growth Take Us?" World Bank Economic Review, 17(1), 107-31. Heltberg, R. and M. Vodopivec (2004). Sri Lanka: Unemployment, Job Security, and Labor Market Reform. Unpublishedmanuscript, World Bank. Henderson, V. (2000). "The Effects of Urban Concentration on Economic Growth". NBER WorkingPaper No. 7503. Henderson, V. and H.Wang (2005). "Aspects o f the Rural-Urban Transformation o f Countries". Journal of Economic Geography, 5,23-42. Henderson, V. andR. Becher (2000). "Political Economy o f City Sizes and Formation". Journal of Urban Economics, 48,453-484. Henderson, V., Z. Shalizi, A. J. Venables (2001). "Geography and Development". Journal of Economic Geography, 1,81-105. Herath A.(200 1). Cost of Compliance of Sanitary and Phytosanitary Requirements in Beverages and Spices in Sri Lanka. Colombo, Sri Lanka: Economics Research Unit, Department of Export Agriculture. Hoogeveen, J. (2003). Measuring Welfarefor Small Vulnerable Groups: Poverty and Disability in Uganda. Unpublishedpaper, World Bank. Indrasiri, L.H. (2006). Population Distribution & Density Levels of Colombo District by Divisional Secretariat. A table prepared for the Sri Lanka Poverty Assessment. Jin, S. Q., K. Deininger and M. Sur (2005). Sri Lanka's Rural Non-Farm Economy: Removing Constraints to Pro-Poor Growth. World Bank. Jitsuchon, S. (2004). Poverty Maps Construction in Thailand. Unpublished paper, Thailand DevelopmentResearchInstitute, Bangkok, Thailand. 111 Klasen, S. (2001). Malnourished and Surviving in South Asia, Better Nourished and Dying Young in Africa: What CanExplain this Puzzle? (Discussion Paper No. 214). University of Munich. Krugman, P. (1991). "Increasing returns and economic geography". Journal of Political Economy, 99,483-499. Krugman, P. (1999). "The role of geography in development". International Regional Science Review 22,2, 142-161. Kumar, Arun, (1999). TheBlack Economy in India. New Delhi:PenguinBooks India. Lucas, Robert, E.B. (2004). International migration to the high income countries: Some consequencesfor sending countries. Unpublishedpaper, BostonUniversity. Mahmud, W. (1989). "The impact of overseas labour migration on the Bangladesh economy: A macro-economic perspective", inR. Amjad (Ed.) To the Gulfand Back: Studies on the Economic Impact of Asian Labour Migration.New Delhi: EO-ARTEP. McNay, K., R. Keith and A. Penrose (2004). Bucking the Trend: How Sri Lanka has Achieved Good Health at Low Cost - Challenges and Policy Lessonsfor the 21StCentuly. London: Save the Children. Mendez, M. A., C. a. Monterio and B. M. Popkm (2005). "Overweight Exceeds Underweight Among Women inMost Developing Countries". American Journal of Clinical Nutrition, 81,714- 721. MinistryofFinance(2005). Medium TermExpenditureFramework,Colombo. Murray, C. J. L., K. Xu, J. Klavus, K. Kawabata, P. Hanvoravongchai, R. Zeramdini, A. M. Aguilar-Rivera and D. B. Evans (2003). InMurray, C.I.L., and B.D. Evans (Eds.), Health System Pe~ormance Assessment: Debates, Methods and Empiricism, Geneva: World Health Organization. Muthaliph, T.M.Z (2005). "Measuring the Regional Contribution to the Sri Lankan Economy: Estimationand Analysis of Provincial GDP 2003". News Survey, 26(1), 11-18. Colombo: Central Bank of Sri Lanka. Nanayakkara, A.G.W. (2004). Employment and Unemployment in Sri Lanka - Trends, Issues and Options.Colombo: Department of Censusand Statistics. Organization for Economic Co-operation and Development (OECD) (1998). Maintaining Prosperity in an Ageing Society. Paris: OECD. Osmani, S.R. (1994). "Economic Reform and Social Welfare: The Case of Nutrition in Sri Lanka". American EconomicReview, 84(2), 291-296. Pathmanathan, I., J. Liljestrand, J. M. Martins, L. C. Rajapaksa, C. Lissner, A. de Silva, S. Selvaraju, P. J. Singh (2003). Investing in Maternal Health: Learning from Malaysia and Sri Lanka (Health Nutritionand Population Series). Washington D.C: World Bank. Peace Secretariat (2004). Impact of the Ceasefire Agreement on Regional Economic Growth in Sri Lanka. Colombo. Piyasena C. (2004). "Case Studies of Successful Micronutrient Programs: The Sri Lankan Experience". Food and Nutrition Bulletin, Geneva: World Health Organization. Popkin, B.M., M.K. Richards, C. Monteiro (1996). "Stunting i s Associated with Overweight in Children of Four Nations that are Undergoing the Nutrition Transition". Journal of Nutrition, 126, 3009-16. 112 PresidentMahindaRajapakse (2005). Mahinda Chintana PresidentialManifesto, Sri Lanka. Rama, M. (1999). The Sri Lankan Unemployment Problem Revisited (World Bank Policy ResearchWorking Paperno. 2227). WashingtonD.C: World Bank. Ranabahu, R.A.S.P. (2004). "The Consequences of Urban In-migration in Sri Lanka: A Case Study ofGampahaDistrict". Sri Lanka Journal of Population Studies, 7,71-87. Ranna-Eliya, R. P., N. de Mel, D. Samarasinghe, H. Aturupane, and H. Wijeratne (1997). ResourceMobilization in Sri Lanka 's Health Sector. Colombo. Rannan-Eliya, R. P. (2001). Strategies for Improving the Health of the Poor - The Sri Lankan Experience. Colombo: Institute ofPolicy Studies HealthPolicy Program. Rannan-Eliya, R. P., B. Pande, J. Killingworth and A. Somanathan (2001). Equity in Financing and Delivery of Health Sewices in Bangladesh, Nepal and Sri Lanka: Results of a Tri-countly Study. Data International Limited, Nepal Health Economics Association and Institute of Policy Studies. Ravallion, M.and S. Chen(2003). "Measuring pro-poor growth". Economics Letters, 78,93-99. Ravallion, M. and S. Chen (2004). China's (uneven) progress against poverty (World Bank Policy ResearchWorking Paperno. 3408). WashingtonD.C: World Bank. Rosen, K.and M.Resnick (1980). "The Size Distribution of Cities: An Examination ofthe Pareto Law andPnmacy". Journal of Urban Economics, 8, 165-186. Sarvananthan, M. (2003). "Economic Revival in North and East Sri Lanka: What Are the Impediments". Economic and Political Weekly,38(19), 1844-1850. Mumbai. Sarvananthan, M. (2005), Poverty in the Conflict Affected Region of Sri Lanka: An Assessment, Colombo. Sarvananthan, M. (2005). "Post-Tsunami Sri Lanka: Swindlers Hold Sway". Economic and Political Weekly,40(17),1683-7. Mumbai. Sarvananthan, M., (2004). "Conflict Time Economy of the North & EastProvince of Sri Lanka: An Introduction". IndianJournal of Regional Science,36(l), Kolkata. 53-72. Schiff, M. (2006). "Brain Gain: Claims About Its Size and Impact on Welfare and Growth are Greatly Exaggerated", inOzden, C. and M. Schiff (Eds.) International Migration, Remittances h TheBrain Drain. Washngton, DC: World Bank. Sen, A. K. (1981). "Public Action and the Quality of Life in Developing Countries". Oxford Bulletin of Economics and Statistics, 43,287-3 19. Sen, A. K. (1988). "Sri Lanka's Achievements: When and How?" in Srinivasan, T.N. and P.K. Bardhan(Eds.) Rural Poverty in South Asia, 549-56. New York: Columbia University Press. Shiffman, J. (2000). "Can Poor Countries Surmount Maternal Mortality?" Studies in Family Planning, 31(4), 274-289. Sinha, N. (2006). Poverty and Health in Sri Lanka: Successes and Challenges. A background Paper for Sri Lanka Poverty Assessment, World Bank. Smith, L.C. and L. Haddad (2002). "How Potent is Economic Growth in Reducing Undernutrition? What are the Pathways of Impact?: New Cross-country Evidence." Economic Development and Cultural Change, 51(I), 55-76. Sri. Lanka Bureau of Foreign Employment (2004). Annual Statistical Report of Foreign Employment 2004. RetrievedMarch 21,2006 from hti~~,'/wwtv.slbC.l~/f~b/Ftat - maii~html. 113 Strauss, J. and D. Thomas (1998). "Health, Nutrition and Economic Development". Journal of EconomicLiterature, 36,766-817. Svedberg, P. (2000). Poverty and Undernutrition: Theory, Measurement, and Policy. Oxford: Oxford University Press. Svedberg, P. (2004). Has the Relationship between Undernutrition and Income Changed? Copenhagen ConsensusOpponentNote. Tabor, S. R., S. Abeyratne, and R. Epaarachchi (2000). An agro-industry strategy and policy reform actionplanfor Sri Lanka, Ministry of Industrial Development (with the assistance of the USAID).Colombo: The Ago-Enterprises Project. Thalagala, N. (2004). "An Index to Measure Socioeconomic Status: Asset Approach". The Journal of Collegeof Community Physicians in Sri Lanka, 9,32-39. Tilakaratna, K.G. and A. Satharasinghe (2005). Headcount Index and Population Below Poverty Line by DSDivision -Sri Lanka: 2002. http://www.statistics.gov.lWpoverty/index.htm. Turk, C. (2005). Vietnam: A Comprehensive Strategy for Growth and Poverty Reduction. Managing for Development Results Principles in Action: Sourcebook on Emerging Good Practice, 44-50. United Nations (2003). Assessment of Needs in the Conflict Affected Areas of the North East, Colombo. UnitedNations Development Programme (2005). Human Development Report 2005. New York: UNDP. Urban Development Authority (2001). City of Colombo. Retrieved from http://www.buildsrilanka.com/cdp/8~City%20ofo/o20Colombo.htm. Wang, J., D. T. Jamison, E. Bos, A. Preker and J. Peabody (1999). Measuring Country Pe~ormanceon Health: SelectedIndicatorsfor 115 Countries (Health, Nutrition and Population Series). Washington, DC: World Bank. World Bank (1993). The East Asian Miracle: Economic Growth and Public Policy. New York: Oxford University Press. World Bank (1996). Sri Lanka Non-plantation Crop Sector Policy Alternatives, Report No. 14564-CE.Washington, DC: Agriculture andNaturalResourcesDivision, Country Department 1, SouthAsia Region, World Bank. World Bank (2000). WorldDevelopment Report 1999/2000.New York Oxford University Press. World Bank (2002). Sri Lanka Poverty Assessment, Report No. 22535-CE. Washington, DC: World Bank. World Bank and Asian Development Bank (2002). Poverty in Bangladesh: Building on Progress. ReportNo. 24299-BD. Washington, DC: World Bank. World Bank (2003). Sri Lanka Promoting Agricultural and Rural Non-Farm Sector Growth. Colombo: RuralDevelopmentUnit,World Bank. World Bank (2004a). Republic of Korea: Four Decades of Equitable Growth. A case studyfrom Scaling up Poverty Reduction: A Global Learning Process and Conference. Shanghai. World Bank (2004b). Malaysia: 30 Years of Poverty Reduction, Growth and Racial Harmony. A case study from Scaling up Poverty Reduction: A Global Learning Process and Conference. Shanghai. 114 World Bank (2004~). Sri Lanka: Development Policy Review, Report No. 29396-LK. Washington, DC: WorldBank. WorldBank(2005a). WorldDevelopment Indicators 2005. Washington, DC: World Bank. World Bank (2005b). A Poverty Map for Sri Lanka-Findings and Lessons, (Policy note). Washington, DC: World Bank. World Bank(2005~).Attaining the Millennium Development Goals in Sri Lanka: How Likely and Wiat Will it Take to Reduce Poverty, Child Mortality and Malnutrition, and to Increase School Enrolment and Completion? Washington, DC: Human Development Unit, South Asia Region, World Bank. World Bank (2005d). Sri Lanka Improving Access to Financial Services Selected Issues. Washington, DC: World Bank. World Bank (2005e). Sri Lanka Development Forum: The Economy, The Tsunami, and Poverty Reduction. Washington, DC: World Bank. World Bank (20050. Treasures of the Education System in Sri Lanka: Restoring Pe$ormance, Expanding Opportunities and EnhancingProspects. Colombo:World Bank. WorldBank(2006a). GlobalEconomic Prospects 2006. Washington, DC: World Bank. WorldBank(2006b). DoingBusiness in 2006. Washington, DC: WorldBank. WorldBank(2006~).Sri Lanka: Underpinning Growth withEquity, Draft, WorldBank. World Bank (2006d). Poverty and Social Impact Analysis for Sri Lanka: A Case Study. Washington, DC: World Bank. World Bank (2006e), Decentralization and Service Deliveiy in Sri Lanka: Assessment and Options, draft, World Bank. World Bank (20060. Agriculture Import Liberalization and Household Welfare in Sri Lanka, Draft, World Bank. World Bank (2006g). Resilience amidst Conflict: an Assessment of Poverty in Nepal, 1995-96 and 2003-04.Washington, DC: World Bank. World FoodProgramme (2003). Identification of DSDivisions of Sri Lanka: Vulnerablefor food insecurity. Vulnerability Analysis andMappingUnit, Colombo: WFP. World Health Organization (2004). Sri Lanka: Reproductive Health profile. WHO Regional Office for South-EastAsia. World HealthOrganization(2005). The WorldHealth Report. Geneva: WHO. Yoshida, N.,K.G. Tilakaratna, and D. B. P. S. Vidyarathne (2006). Does migration provide an impetusfor pro-poor growth?, a backgroundpaper for the Sri LankaPovertyAssessment, World Bank. Food andAgriculture Organization statistical database. www.centralbanklanka.org WI..W.nioe.gov.lk I.x~~~~.statistlcs.pov.lk vnw.tafren.aov.1k 115 Annex 1 Figures and tables referredto in Chapter 3 Figure A-1.1. Gross Secondary School Figure A-1.2: Exports of goods and services Enrollment Rate (YO) (Yo of GDP) 0Femak 80 60 40 20 0 Source:ADB, Key Indicators 2005 Sortrce: World Bank ,WDI 2005 Enroll all children inprimary school by 2015 Net enrollment ratio inprimary education(for age 6-10 years) reached 96% in2002 Make progress towards gender equality and Ratios o f girls to boys inprimary, secondary, and empowering women by eliminating gender disparity tertiary educationreached 95%, 102% and 114%, inprimaryandsecondary schoolby2005 respectively Reduce infant and child mortality rates bytwo- Infant mortality rate fell down from 18 in 1991to thirdsbetween1990and201.5 in2o02; child mortality rate declined from 22 in1991to 14in2002 Reduce maternal mortality rates by three-quarters Maternal mortality ratio declined from 42 in 1991 between 1990 and 2015 1to 28 in2002 Provide access for all who needreproductive health Contraceptive prevalence rate i s high at 70 in services by 2015 2000; 96% o fbirths are attended byhealth staff 116 Periods Consumption inequality (Ginicoefficient) Poverty headcount rate Start -- End year Start I End Start I End Bangladesh (l) 91/92 --- 2000 25.9 30.6 58.8 49.8 India (*) 93/94 --- 99/00 29.0 32.0 29.2 22.7 Nepal (4) 95/96 --- 03/04 34.2 41.4 41.8 30.9 Pakistan (5) 01/02 --- 04/05 28.0 30.1 34.4 29.2 Sri Lanka (6) 90191--- 2002 32 40.0 26.1 22.7 Source:(1) World Bank andADB (2002); (2) StaffEstimationbasedon DeatonandDreze(2002); (3) World Bank (2006g); (4) World Bankstaffestimationbasedon PIHS2000-01 and2004-05; (5) HIESsurveys (DCS) Note: Povertylines are defineddifferently across counhies; so povertyheadcountratios are not comparable across countries. China Korea, Rep Malaysia Thailand Vietnam NA NA Source: WDI 2005 Note: The definition of militaryexpenditures i s slightly different from Annual Report2004 to makeinternational comparisonscomparable. Table A-1.4: Comparisons on rigidity of labor regulation in Januarv 2005 I index I index I index I waaes) Korea, Rep 44 60 30 90 Malaysia 0 20 10 65 Thailand 33 20 0 47 Vietnam 44 40 70 98 Source: Doingbusinessin2006 117 Tabfe A-1.5: Interest Rates in 2003 (%)' Lending Real Sri Lanka 5.1 China 3 Korea, Rep 3.9 Malaysia 2.7 Thailand 5.9 3.8 Vietnam 9.5 3.9 Health Country (2002/03) (2002) Sri Lanka 2.2 1.5 China NA 2.6 Korea, Rep 4.3 2.6 Malaysia 7.9 2 Thailand 5.2 3.1 Vietnam NA 1.5 Note: Current and capital expenditureexpressed as apercentage ofGDP 'Lending rates are collected by the IMF as representative interest rates offered by banks to residents. The terms and conditions attached to these rates differ by country, however, limiting their comparability. 118 Annex 2 I. Projectingtheratepovertyreduction:elasticityofpovertyreductiontogrowth Although economic growthis the most powerful force for reducing income poverty, the extent of poverty reduction would be limited if benefits of the growth were skewed in favor o f the rich. Therefore, projection of poverty reduction needs to estimate the impacts on poverty reduction o f economic growth (growth effect) as well as of a change in income/consumption distribution (distributional effect). Various methods are proposed to estimate the impacts: Datt and Ravallion (1992) propose a method based on an identity among poverty reduction, a change inmean income, and a change in its distribution. The method does not require any assumption on the income/consumption distribution but needs at least two household surveys for reference and comparison years. To estimate the growth effect, this method shifts the reference year's distribution so that its mean is equal to that of the comparison year. The difference in poverty rates between the reference distribution and the shifted one represents the sole impact o f growth of mean income/consumption on poverty reduction. To estimate the distributional effect, the comparison year's distribution is shifted so that its mean is equal to the reference year's one. Since the reference year's distribution does not differ from the shifted one in mean income/consumption but in distribution, the difference in poverty rates between the reference distribution and the shifted one represents the sole impact o f a distributional change on poverty reduction. The advantage o f this approach i s that no assumption i s needed for the income/consumption distnbution. Bourguignon (2003) also proposes a method based on the growth-poverty-inequality identity. By assuming the income/consumption distribution i s log normal, where the distribution i s simply defmed by its mean and standard deviation, the growth effect and the distributional effect can be represented as a simple function of a change in mean and standard error, respectively. This method has several advantages over others. First, it i s parsimonious in that it needs only two parameters to estimate the effects, namely the mean and standard error o f income/consumption distribution at the reference year. This means, in addition, that only one year of household survey is necessary for estimation. The extent o f poverty reduction can be estimated by a bunch o f poverty predictors; however, the more information a projection needs, the more likely the projection i s to contain errors. Second, the formulas for both growth and distributional impacts are simple, as shown in Bourguignon (2003). Third, Bourguignon (2003) shows this method outperforms other methods inpredictingthe extent ofpoverty reduction. H e found, usinga cross- country and time series database, that the predictedrates o f poverty reduction by his method fit to the actual rates better than those by other methods. This implies that the assumption on income/consumption distribution i s not harmful in predicting poverty reduction. The innocuousness o f the assumption i s also enhanced by the fact that the growth effect between 1990-91 and 2002 in Sri Lanka estimated by this method i s very similar to that o f Datt and Ravallion's approach. Quentin (2002) raises a question on independence between growth rates and distributional changes. For example, assuming a high growth rate and no change in income distribution could be unrealistic. He proposes to estimate the correlation between growth rates and changes in income distribution, and then compounds the direct impact fkom economic growth with the indirect impact through a distributional change caused by the growth. The downside of this approach is that the correlation needs to be estimated with very limited observations. In Sri Lanka, only three household surveys are currently comparable in estimating poverty headcount rates, implyingthe maximum two observations to estimate the correlation. Another concern is about whether the correlation from the previous surveys can predict future relationship between 119 growth rates and distnbutional changes. Structural changes in growth and inequality are often observed ina longtime span. 11. Figures and tables referred to inChapter 2 Figure A-2.1: 95 percent confidence intervals of headcount ratios 8-- c E R- i; 5 *- V s M- s -1 m m vi 90-91 95-96 2002 90-91 95-96 2002 year year -8 5 3 - I * A - I U i;3- B 5 V s 6%-, vi I I I Figure A-2.2: Density and cumulative distribution of per capita expenditure %ne=&. 1423 0 2OW 4WO 60W 8000 IWOO 0 2000 4WO 6000 8000 low0 rpcexp-2002 rpcexp-2002 I- dm-2002 -----der-1990 1 1- cum2002 -----cum1990 I 120 FigureA-2.3: Per capita nominal consumption expenditure(2003-04) 6,000 , 4,000 2,000 0 I Source: CFSESreport (2003-04) I FigureA-2.4: Growth incidence curves for per capita consumption 1-02) Natronal(95-96 to 0ldX) Source: WorldBank staff calculations using HIES (different years). 121 FigureA-2.5: GICs of per capita income(pcinc) and consumption(pcexp) between1995-96 and 2002 National (95-96 to 2002) National (95-96 to 2002) 1 10 20 30 40 50 60 70 80 90 100 1 I O 20 30 40 50 60 70 80 90 II % ofpopulationrankedby pcinc %ofpopulationrankedbypcexp FigureA-2.6: Shareof districts inpoor population and Samurdhibudget _____ I 12 -gm10 f a $ 6 4 2 0 0 % of poorpopulation(2002) B% oftotal Samurdhi allocation(2005) Source: DCS, HIES (2002); Ministryof Samurdhi (2005) No. of damaged houses No. of displaced No. of deaths E4 Eastern P Southern 0 Northern0 Western I L L Sources:DCS (January, 2005); Government of Sri Lanka and Development Partners(December 2005) 122 Table -4-2.1: Ranking of districts bv Dovertv headcount bv districts Province District Rank (1990-91) Rank (1995-96) Rank (2002) Colombo 2 1 1 Western Gampaha 1 2 2 Kafutara 14 6 4 Kandy 17 13 8 Central Matale 8 15 11 Nuwara Eliya 3 9 5 Galle 10 10 9 Southern Matara 9 11 10 Hambantota 15 8 13 North-West Kurunegala 7 4 7 Puttalam 4 7 12 North-Central Anuradhapura 5 5 3 Polonnaruwa 6 3 6 Uva Badulla 11 14 16 Monaragala 16 17 17 Sabaragamuwa Ratnapura 12 16 15 Kegalle 13 12 14 Source:HIES90-91,95-96, and 2002 (DCS) Note: Districtsinthe Northemand Easternprovinces are excluded since no data are available from HIES Table A-2.2: Comparison between per capita ncome Percapita income 'er Capita consumptior Quintile 95-96 2002 95-96 2002 1 703 766 991 1068 2 1215 1381 1445 1596 3 1698 1984 1881 2168 4 2472 2952 2578 3117 5 5966 7809 5274 7325 1 Total 2411 2978 2434 3055 GIN1 I 0.43 0.46 0.32 0.40 Source: IES95-96 and 2002 Note: Quintiles of per capita consumption expenditure 123 Annex 3 I. The role of poverty maps in analysis There are good reasons to believe that poverty rates much higher than the national average can be found inspecific areas in Sri Lanka, even within districts that on the aggregate show a relatively low incidence o f poverty. Thus there i s a long-standing demand for an understanding o f poverty and inequality at finer levels of spatial disaggregation than what is available as direct estimates using HIES data (district level). Reasonably accurate estimates o f poverty, at the D S Division or lower administrative level, can greatly facilitate monitoring and evaluation o fthe existing poverty alleviation programs and geographic targeting o f future government interventions. This i s possible through an exercise in Poverty Mapping - a technique developed in Elbers et a1 (2003) and since implemented in many countries around the world - with the objective to provide statistically reliable estimates of consumption-based welfare indicators. The DCS o f Sri Lanka has initiatedthis exercise -the first-ever attempted in SouthAsia -with technical assistance from the World Bank. The poverty mapping method takes advantage o f strengths o f both the HIES - that includes consumption aggregates but lacks enough sample size to estimate poverty at the geographical unit below district - and the Population CENSUS - that has enough sample size but lacks consumption aggregates. Using this method, members o f the DCS and World Bank staff produced a map of poverty headcount rahos at the D S division level, for the year 2002. The results of this exercise, some o f which are presented inthis report, suggest the important role that poverty maps can play inthe analysis o f spatial inequality and its correlates. Perhaps most importantly, such maps allows policymakers to draw visual and statistical links between poverty and spatial factors that are not apparent from more aggregated analysis. At the same time, it i s important to bear in mind that poverty mapping uses statistical techniques to circumvent the absence of statistically representative household data, which can introduce errors inthe poverty estimates. Giventhis, it is best to use the poverty maps inconjunction with other tools available to policymakers to get a handle on spatial distribution and correlates o f poverty, rather than rely exclusively on the precise ranking o f geographical areas indicated by the map to guide policy. 124 11. Tables referred to in Chapter 3 TableA: Results of multivariateprobit regression: probability of a household beingpoor Source: Staffestimation basedon HIES 2002 data Notes:**refers to YOsignificance, * 5%, 1 and + 10%. Accessibility index is omitted from the regressions with average educational attainment (for district or DS division) due to multicollinearity 125 Table A-3.1: Poverty Headcount Ratio by gender of household heads (YO) 1990-91 1995-96 2002 Male 26.0 29.1 23.0 Female 26.6 27.6 21.5 Table A-3.2: Poverty Headcount Ratio by ethnicity of household heads (YO) ETHNICITY 1990-91 1995-96 2002 Sinhala 26.3 28.3 22.3 Sri Lanka Tamil 26.2 30.8 26.8 IndianTamil 18.8 37.3 25.7 Sri Lanka Moors 27.7 30.4 24.6 Table A-3.3: Poverty Headcount Ratio by religionof household heads (YO) RELIGION 1990-91 1995-96 2002 Buddhist 27.5 29.7 23.3 Hindu 22.3 35.4 27.8 Islam 26.5 29.4 23.3 Christian 15.1 13.5 11.0 Table A-3.4: Poverty incidence and accessto infrastructureby district Poverty Average Average time % ofhousing % ofhousingunits District name Headcount accessibility to Colombo using electricity usinggas for Ratio (%) index for lighting cooking &el Colombo 6 3.9 43 86 48 Gampaha 11 3.8 58 83 24 Kalutara 20 4.0 99 72 17 Kandy 25 3.2 184 71 14 Matale 30 3.1 219 51 6 Nuwara El 23 3.0 195 54 7 Galle 26 3.3 152 73 11 Matara 27 3.1 210 72 9 Hambantot 32 2.9 286 49 4 Kurunegal 25 3.2 166 52 4 Puttalam 31 3.0 195 52 8 Anuradhapura 20 2.9 309 49 6 Polonnaru 24 2.9 292 46 4 Badulla 37 2.9 251 59 6 Moneragal 37 2.7 316 31 3 Ratnapura 34 3.1 168 45 5 Keealle 32 3.5 120 57 5 L -__- ._ Correlation with IKX (Yo) -70 61 -71 -8I 126 Source HIES 2002 WB ICs ICs CENSUS 2001 CENSUS 2001 Notes 1 The accessibility index i s calculated for every point as the sum of the population totals of surrounding cities and towns, inversely weighted by the roadnetwork travel time to each town The numbers show the mean of the access values for all points that fall into a given dismct 2 The average travel time to Colombo city IS estlmated travel time to each town basedon geographical information of road network The numbers show tlie meanof the travel time for all points that fall into a gwen dismct 3 "HIES 2002 WB" denotes that the world bankstaff calculated these figures usingHIES 2002, "ICs' refers to "Sn Lanka Improvingthe Rural and Urban InvestmentClimate (2004)" Table A-3.5: Poverty and Educational Attainment of household heads % ofhheadswith Yo of District Poverty tertiary education household Share of employed name Headcount (higher than G.C.E. heads with no population in the Ratio (%) (OK)) schooling agricultural sector Colombo 6 42 3 2 Gampaha 11 30 3 9 Kalutara 20 30 3 20 Kandy 25 25 7 28 Matale 30 18 7 43 Nuwara Eliya 23 10 11 76 Galle 26 20 8 32 Matara 27 19 9 44 Hambantota 32 15 9 48 Kurunegala 25 21 5 37 Puttalam 31 16 5 36 Anuradhapura 20 20 6 59 Polonnaruwa 24 12 7 53 Badulla 37 17 14 69 Monaragala 37 13 13 71 Ratnapura 34 16 10 47 Source: DCS (2004) Author's calclulation based onHIES2002 LFS2002 Table A-3.6: Poverty Headcount Ratio (%) Unemployment rate (%) Western 11 8.9 NorthCentral 21 8.4 Central 25 8.9 NorthWestern 27 7.8 Southern 28 10.6 Sabaragamuwa 35 9.8 Uva 37 6.0 Source HIES2002 WJ3 LFS 2002 NotestLFS 2002 refers to "Annual Report of Sri Lanka Labor Force Survey 2002". 127 Table A-3.7: Povertv and unemdovment rates bv district District name HCR Unemploymentrate Colombo 6 9.1 Gampaha 11 8.2 Kafutara 20 9.8 Kandy 25 11.6 Matale 30 7.6 Nuwara Eliya 23 5.4 Galle 26 8% Matara 27 11.0 Hambantota 32 13.3 Kurunegala 25 7.9 Puttalam 31 7.5 Anuradhapura 20 7.1 Polonnaruwa 24 11.7 Ratnapura 34 9.2 Kegalle 32 10.6 Badulla 37 6.3 Monaragala 37 5.2 Source: HIES 2002 WJ3 LFS2002 Table A-3.8: Poverty and paid employees in the agricuhral sector by district Per capita monthly household District Poverty Share of agricultural paid income for householdswith name Headcount employeesintotal heads working inthe Ratio(%) employment (%) agricultural sector as paid employees Colombo 6 2 2463 Gampaha 11 4 2873 Kalutara 20 6 1907 Kandy 25 10 2001 Matale 30 5 2435 Nuwara El 23 4% 1996 Galle 26 16 2017 Matara 27 15 1811 Hambantot 32 14 1920 Kurunegal 25 6 2212 Puttalam 31 12 2280 Anuradhapura 20 19 2262 Polonnaru 24 12 1960 Badulla 37 27 2003 Moneragal 37 14 2088 Ratnapura 34 12 1467 Kega11e 32 10 1468 Source: World Bank staff calculations using HIES 2002 Notes: The agricultural sector includes fishing and forest workers 128 TabIe A-3.9: Results of multivariate probit regression: probability of a household being poor (all coefficients represent a change inprobability for a marginal increase in a n explanatory variable) Sample excludes Colombodistrict With access variable Witheducation & shareof agrz emp Householdcharacteristics Presence o f elderly -0.016 -0.016 (f.89)+ (1.90)+ Family member abroad -0.091 -0.091 (8.41)** (8.40)** Presence o f a child or children 0.065 0.065 (7.83)** (7.78)** Unemployment o fthe youth 0.048 0.049 (5.OO)* * (5.14)** Household size o f4-6 members 0.105 0.I05 (12.10)** (12.07)** Household size more than 6 members 0.260 0.258 (20.74)** (20.61)** Located in the rural sector 0.092 0.089 (8.27)** (7.98)** Located in the estate sector 0.021 -0.002 (1.07) (0.10) At least one formal sector worker -0.095 -0.094 (9.87)** (9.74)** At least one informal sector worker 0.075 0.076 (9.74)** (9.90)** Characteristicsof the householdhead Male 0.014 0.015 (1.44) (1.59) Unemployed 0.033 0.035 (1.48) (1.54) Inactive 0.034 0.033 (3.22)* * (3.19)** Completed 5 grade o f less 0.186 0.183 (13.55)** (13.29)** Completed 6-9 grades 0.100 0.100 (7.31)** (7.28)** Completed A/L or above -0.087 -0.087 (3.98)** (3.98)** Working as agricultural wage worker 0.044 0.040 (3.85)* * (3.48)** Districtcharacteristics Share o f agricultural employment o fHHeads 0.000 (0.84) Unemployment Rate 0.004 0.004 (2.79)** (2.27) * Characteristicsof DSdivisions Accessibility index -0.012 (2.14)* Share o f householdsusing electricity -0.001 -0.001 (5.76)** (2.53)* Share o fHHeads withprimary education 0.002 (3.73)** Observations 14880 14880 Source Staffestimation based on HIES 2002 data Notes ** refers 1 % to significance,* 5%, and+ 10% Z-statistics are inparentheses 129 Table A-3.10: Regression of accessibility index of D S divisions DS division characteristics Share of using electricity 0.022 (7.01)** Share ofHHeads withprimary education -0.001 (0.1%) District characteristics Share of agricultural employment of HHeads -0.029 (7.87)** Unemployment Rate -0.058 (3.15)** Constant 8.2 11 (23.99)** Observations 249 R-squared 0.71 130 Annex 4 I. Comparison of data sources for information on migration and remittance Census ofPoplation and Housing Condition 2001: The Census i s supposed to be an authentic statutory record o f all people resident in Sri Lanka. It not only provides the usual information on individual characteristics, availability o f physical infrastructure, and housing conditions, etc., but also enquires about "leaving home". The questionnaire asks about the place o f birth, the place o f previous residence, and the length o f stay inthe current location. FromCensus data, migrants can be defined as someone who has previously resided somewhere other than the district of current residence. In principle, the Census covers all migration flows except for international migrants who went left Sri Lanka, and people who moved from one place to another within a district. The Census has been used in studies o f internal migration focusing on patterns o f inter-regonal and rural-urban migration. Another major limitation o f using the Census for migration analysis i s its lack o f information on remittances. The Census does ask whether there are any household members residing outside, butnothing is asked relatedto remittances receivedfrom them. HIES and CFSES: HDES provides information on remittance from internal and international migrations. Households are asked whether there are any household members residing outside, and if so, how much remittances they sent inthe last year. Such information is useful to measure the impact o f remittance and migration on rural development and economic disparities among distncts. Bothsurveys and especially CFSES provide information on demographic characteristics of internal and international migrants. Since the Census does not include this, the information from surveys i s valuable incomplementingthe Census. One potential issue i s that the household surveys likely underestimate the size o f migration. Migrants in these surveys are only those who live outside the current residence o f the other members o f the same household. Therefore, in contrast with the Census, if a household as a whole moved somewhere, these surveys would not treat them as migrants. In fact, the census indicates 80 percent of migrants into Colombo city moved with their household heads, suggesting a high probability that all household members moved into Colombo city. The HIES and the CFSES cannot elicit any information regarding such migrants, which may significantly bias profiles ofmigrants ifthey were exclusively relied uponfor this information. 11. Populationgrowth and congestioninColombo Congestion in Colombo MC i s a growing concern in Sri Lanka (see Figure A-4.6 below). Population inColombo M C has increased since 1871 and reached 0.6 million in 2001. Colombo MC's over-concentration issue i s evident when looking at its extremely highpopulation density. Colombo MC's population density grew from around 15 thousand per square km in 1981 to around 17 thousand in 2001, which are much higher than Sri Lanka and even Colombo District. The highpopulation density may explain why population growth in Colombo MC slowed down since 1970s and i s much lower than the national growth rate (Figure i). i s also the likely reason It for highpopulation growth in the broader metropolitan area - Colombo District - and Western Province: over-concentration in Colombo pushes many residents outwards, at the cost o f long daily commutes. In fact, congestion in Colombo MC is much worse than the population density figures suggest during daytime when population inthe M C area expands to more than 2 million. Severe traffic congestions are caused by the hightraffic volume, and aggravated by factors such as shortage of parking areas and poor public transport facilities. As a result, average vehicle speed is only around 10kilometers per hour within most parts o f the city duringthe day. 131 111. Cross-country evidence on the economic costs of over-concentration Henderson (2000) provides an estimate o f economic loss due to over-concentration inthe largest city or the largest metropolitan area in a country. H e examines the cost imposed by excessive concentration (interacted by per capita income and national scale) on economic growth using a panel o f 80-100 countries every 5 years fiom 1960 to 1995. Henderson confirms that there are "optimal" primacy points-urban concentration in the largest city-that depends on income and scale. Specifically, efficient urban primacy levels rise upto an income per capita o f $5,000 (1985 PPP), then peak, and decline. The rise is rapid, but the decline after the peak is modest. For example, for a country with national urban population o f 8 million, the optimal primacy rate changes from 15, 28, to 26 percent as the country's GDP per capita rises from $850, $3000, to $17200. Henderson (2000) shows the list o f countries with highly excessive urban primacy in 1990. The list includes the usual suspects, such as Argentina, Panama, Costa Rica, Chile, Korea and Thailand. On the other hand, 30 countries belong to a group of countries with satisfactory urban concentration, which includes USA,Canada, Australia, andNew Zealand. IV. What constitutes Colombo urban area: estimate usingpopulation density pattern In Figure A-4.7, each dot represents population density for a GN division in urban areas o f Colombo District. Larger dots refer to population density for Colombo MC, while smaller dots for other GN divisions. It is clear from that the distribution o f population density changes dramatically at around 10kmfrom the center of Colombo city. Inside 10 kmfiom Colombo MC, the variation o f population density i s large and the average population density i s high and declining with distance from the center. On the other hand, outside 10 km, the variation in population density is small and does not change much with distance. This suggests that there is a structural difference in habitation between inside and outside 10km from the center. A satellite image also indicates that there is a clear continuation o f habitation till 10 km fi-om the center o f Colombo city. This i s actually a conservative estimate o f the Colombo urban area, since it is smaller than the Core Area o f the Colombo Metropolitan Region (CMR), also known as Capital Territory, which consists of the municipal councils o f the City o f Colombo, Dehiwala, Mt. Lavinia, and Sri Jayawardena Pura-Kotte, and a few local authorities in the Western Province. CMR, on the other hand, consists o fthe entire Western Province.' V. Improvingurban services in Sri Lanka Based on analysis on data available in Sri Lanka and international experience, the Service Delivery Report (2006) made several recommendations to rejuvenate/enhance urban functions of Colombo Metropolitan areas as well as other urban areas. First, financing urban services by levying a tax on the annual rental value of land or any species o f immovable property will make urban services more demand dnven. This i s especially the case for urban areas outside the Colombo Metropolitan regions. This would not only enhance local governments' own source of revenues, but also foster more efficient pricing o f local services, aligns costs with benefits more closely, and make infrastructure services more demand-driven. Second, there will be likely benefits from expanding the role o f private sector inurban planning. This would involve the government reducing its role in making decisions about land use by auctioning off public lands, and focusing on planning, provision o f urban services and infrastructure, and levying an appropriate tax on land holdings. By doing this, the government can help markets to allocate resources efficiently to mitigate the risk o f market failures. See http://www.buildsrilanka.com/CDP/ 132 Third, in the Western Province, some form of integrated metropolitan urban planning and land management IS worth exploring. Urban development of Colombo M C is closeIy linked to that of neighboring areas. Integrated metropolltan urban planning will help mitigate the cost of over- concentration of Colombo M C as well as foster economic growth in Western Province as a whole. Indeed, the city of Colombo DevelopmentPlandevised by UDA appears to be consistent with this approach, althoughitis not clear when this planwill be translated into action. VI. Figures and tables referred to in Chapter 4 Figure A-4.1: The shares of domestic migrants and recent migrants (with lessthan 5 years of residence in Colombo city) by origin district Figure A-4.2: The shares of poor population and recent migrants by origin districts (except for Northern and Eastern Provinces) I -% ofpoorpop 95-96 - * '% ofpoor pop 2002 I I st YOofrecentmigrants 133 FigureA-4.3: Educational attainments and occupation for non-migrants and recent migrants %of elementary occupation amongworkers %of individuatswith tertiary education 60 'to 40 20 0 Figure A-4.4: Housing ownership by household heads' origin district Figure A-4.5: EnrolIment in secondary school for age 14 in 1990/91 and 2002 100% ~ . p t ~ z ~ t m z 80% Whde smple W rece~\edremittance 60% 40% 20% 0% Below G5 & None Education Above G5 Source: Staff estimationbasedon HIES 1990/91and2002 134 FigureA-4.6: Average annualpopulationgrowth between1981 and 2001 I 1 5 1 0 0 5 0 0 Colombo MC Colombo Western Sri Lanka Distnct Province national Source: Indrasiri(2006) and DCS (2004) Figure A-4.7: Population density for urban areas in ColomboDistrict 1990191 2001102 Remittance Remittance No Remittance Remittance No on2yfrom onlyfrom From remit- onbfrom onbfiom From Both remit- Abroad Domestic Both tance Abroad Domestic tance Noschooling 9% 13% 10% 11% 5% 7% 6% 6% Upto G5 37% 41% 70% 43% 30% 37% 26% 34% G6-8 24% 21% 3% 23% 23% 21% 28% 22% G9&below degree 28% 24% 17% 22% 41% 35% 37% 36% Degree&above 1% 1% NIA 1% 0% 0% 3% 2% 135 All households Householdswith remittance Consumptionquintile International Internal Total Jntemational Internal Total Poorest 6 7 13 131 140 271 2nd 14 8 22 205 127 332 3rd 32 18 50 378 211 589 4th 71 29 100 626 255 882 Richest 168 59 226 1377 482 1859 Overall 58 24 82 668 278 946 136 Annex 5 I. The HealthSystemin Sri Lanka The government system consists o f a network of hospitals providingbothinpatient and outpatient care. This network includes three broad tiers o f types o f curative care institutions. The primary care institutions include Peripheral Units, Maternity Homes, Central Dispensaries and Maternity Homes, District Hospitals and Rural Hospitals. These primary care health facilities have maternity wards and offer basic medical care. A network o f smaller facilities, called Central Dispensaries, provide mainly outpatient care such as treatment o f minor injuries. Base and Provincial Hospitals, located mainly in large towns, provide secondary level care. Teaching and Special Hospitals provide tertiary care including treatment of cancer, tuberculosis, leprosy and other chronic diseases. At the end o f 2002, there were 576 hospitals and 411 Central Dispensaries (Annual Health Bulletin, 2002). The primary, secondary and tertiary government health institutions together provided about 3 beds per 1000 individuals in 2002. This i s higher than the South Asian average but comparable to other countries with GNP per capita similar to Sri Lanka's. A breakdown of number o f hospitals by provinces and districts shows that government health facilities are widely available throughout the island (Annual Health Bulletin, 2002). According to Hsiao (2000), the expansion in hospitals took place not on the basis on some planning criteria by the Ministry of Health but in response to demands from legislators (Hsiao (2000), page 37). Despite the good availability o f facilities across districts, the availability of beds per IO00 population varies widely within provinces and across districts. Measured on a per thousand capita basis, Colombo has the highest availability of hospital beds at nearly 5 per 1000 capita while Vavuniya and Kilinochchi have the lowest availability at about 2 beds per 1000capita. The government health system does not mclude health facilities on estates. Health facilities on pnvately owned estates are managed by Regional Plantation Companies. Facilities on nationalized estates are managed by Janatha Estates Development Board and State Plantations Corporation. he estates have 52 hospitals, 192 Maternity Wards and 405 dispensaries. In 2001, the government took over 15 estate hospitals (Annual Health Bulletin, 2002). The private sector in Sri Lanka consists of medical clinics and private hospitals. Since the government provides free inpatient care, the private sector mainly provides outpatient care and higher cost care to those who can pay. In 1997, private hospitals comprised less than 5 percent o f total bed capacity (Hsiao, 2000).' There are an estimated 500 to 800 full-time private general practitionerswho provide outpatient care from private clinics (Hsiao, 2000). Some of the private clinics are staffed by government medical officers working part time outside duty hours (before 8 am and after 4 pm) (Hsiao, 2000). Public sector doctors were granted right to private practice by the medical department in the 19* century since it was difficult to raise official salaries. This i s similar to the practice inUK, Jamaica and Singapore. Public sector physicians have historically been paid wages below the market wage and senior government physicians obtain the bulk o f their income from private practice. Full time private doctors and medical practitioners are concentrated inWestern province, mainly inColombo. 1Ministry o f Health does not gather data on private health institutions. Hsiao (2000) reports statistics for the private sector based on a survey conducted in 1998. 137 11.Tables referredto inChapter 5 Notes: World Bank staff calculations using Sri Lanka Demographic and Health Survey (2000). Refers to maternal health services utilized during pregnancy by currently women aged 15-49 years with births within 5 years preceding the survey. TableA-5.2: ChildrenwithHealthRecord.bvWealth Ouintilesand bv Sector Ofthose withHealthRecord %with ChildHealth Number of times Percentage Never Develotment Record weighed weighed By Wealth Quintiles Poorest 96.0 5.7 4.9 Richest 99.6 6.2 1.5 By Sector Urban 99.3 5.9 1.8 Rural 99.2 6.4 1.2 Estate 90.5 6.1 9.9 Population Average 98.5 6.I 2.2 138 Annex 6 Tables referred to in Chapter 6 TabIe A-6.1: Shares of ,Monthly Per Capita Incomes by Rural ExpenditureDecites & Source (2001102) Note. Other income includes incomes from Samurdhi, food Stamps, other cash receipts and rental income Source Authors' calnilntionsfrom HIES 2001/02. Source: Authors calculations using Sn LankaAnnual Labor Force Survey data Source of Agnc Percentage o f Households With Access to Technical Assistance Extension North North North Assistance Western Central Southern Eastern Western Central Uva Sabaragamuwa Total All agencies (government, NGOs, others) 4 6 16 7 4 6 4 6 17 1 36 3 5 2 146 15 Government 3 7 15 7 4 6 3 8 13 3 32 9 4 8 12 7 13 2 139 Distribution of gross irrigated area (%) Distribution of major scheme imgated area ("A) Minor scheme Note: CuItivated and imgated lands are calculated from land used, not necessary owned byhouseholds.Note: Farmsize is based on land owned exclusively by households. Marginal: own less than 1acre, small: own between 1 to 2 acres, medium: own between 2 to 4 acres, and large: own more than 4 acres `ableA6.5: Access to Irrigation by Expenditure Quintile Quintile of Real Per Capita Expenditure - IS` Sh Total (poorest) znd 3d 4Ih (richest) Rural Sn'Lanka (distribution by quintile) Distribution of gross cultivated area ("h) 17.1 15.8 20.3 20.3 26.5 100 Distribution of gross irrigated area (%) 17.0 15.9 20.5 20.1 26.4 100 Distribution of major scheme irrigated area (%) 14.8 14.7 21.1 22.3 27.0 IO0 Distribution of minor scheme irrigated area ("A) 24.1 18.0 16.0 15.5 26.3 100 Average agricultural household Gross irrigated land(acres) 1.4 1.3 1.7 1.7 2.2 1.7 % irrigated to gross cultivated area 48.4 48.9 50.3 55.8 49.7 50.6 Distribution of gross imgated area ("A) Major scheme 34.0 44.9 46.9 50.1 43.3 44.1 Minor scheme 61.1 35.0 38.2 29.7 42.3 40.8 Lift imgation 3.2 12.6 9.5 7.4 3.8 7.3 Well irrigation 1.8 7.4 5.4 12.8 10.5 7.8 Total 100 100 100 100 100 100 140 xo m g j c vc Annex 7 Tables referred to in Chapter 7 Table A-7.1: Paddy Production 1980-2002 Productia (MT) % Share of Total 1980 1990 2000 ~ Sri Lanka 2,134,000 2,539,000 2,861,000 Northern Province 248,000 137,000 84,000 181,947 11.6 5.4 3.0 Jaffna 65,000 34,000 21,000 45,593 1.3 0.7 Kilinochchi 43,000 N.A. 36,165 3.0 1.7 Mannar 115,000 33,000 19,000 25,914 5.4 1.3 0.7 Mullaitivu 33,000 20,000 22,000 24,425 1.5 0.8 0.8 Vavuniya 36,000 7,000 22,000 49,850 1.7 0.3 0.8 Eastern Province 439,000 483,000 602,000 Ampara 251,000 302,000 436,000 Batticaloa 93,000 136,000 58,000 Trincomalee 95,000 45,000 107,000 Source: Department o f Census and Statistics (2003), unpublished data. NorthEastProvincialCouncil (2003:64) Note: Kilinochchi and Mullaitivu data for 2000 and Trincomalee data for 1990are estimates Metric Ton YOShare of Total 1980 1990 2000 2002 1980 1990 2000 2002 Sri Lanka 167,410 145,790 267,680 273,280 100 100 100 100 Northern Province 66,580 24,150 8,190 33,090 39.8 16.6 3.1 12.1 JafTna 41,310 14,450 - 8,340 24.7 9.9 N.A. 3.1 Kilinochchi Mannar 14,730 7,410 20,930 8.8 5.1 7.7 Mullaitivu 10,540 2,290 -- 3,820 6.3 1.6 1.4 Vavuniya EasternProvince 27,090 23,210 35,520 51,870 16.2 15.9 13.3 19.0 Ampara Batticaloa 11,780 13,630 20,980 32,890 7.0 9.3 7.8 12.0 Tnncomalee 15,310 9,580 14,540 18,980 9.1 6.6 5.4 6.9 Table A-7.3: NE POLilationLivingin Refugee Camps as a Result of Conflict and Tsunami District Conflict (end 2003) Tsunami (Jan 2005) Ampara 7,055 62.727 Batticaloa Kilinochchi 7,282 (end 2004) Mannar 8,361 (end2002) Mullaitivu Vavuniya Northern Province 143 Western Central Southern Northern(a) EasternNorthWestern NorthCentral UvaSabaragamuwaSri Lanka Typeof ownership-ownhouse 90.5 76.4 94.9 63.3 91.5 95.1 97.6 85.4 90.6 89.2 Floorarea per Person(Sq. Mt.) 20.2 15.5 16.4 15.1 11.0 17.8 16.9 13.6 16.1 16.8 Roomsper Person 1.2 1.0 1.0 0.8 0.8 1.2 1.1 1.0 1.0 1.1 Wall Type- Bricks/CementBlock 79.5 68.6 19.3 81.2 76.9 81.4 79.2 73.7 67.9 76.5 Floor Type Cement - 84.4 73.2 80.3 81.1 79.6 81.o 65.3 63.6 15.4 77.9 RoofType-Tile/Asbestor 86.2 59.3 91.1 75.9 76.5 78.0 76.0 71.0 77.7 78.7 Table A-7.5: Share of Selected Villages inNortheast by Level of Vulnerability 2004 (%) DSdivisions Extreme Very high High Poor: Lower Displaced poverty: poverty: poverty: poverty: Villages Code5 Code4 Code3 Code2 Codel Batticaloa District Koralaipattu South 60 20 8 10 2 KoralaipattuNorth 55 27 10 8 Trincomalee District Eachchilampattai 58 20 4 18 Kuchchaveli 43 20 10 7 11 10 Mannar District Madhu 40 8 14 13 7 18 Manthai West 32 30 20 13 1 4 Vavuniya District Vavuniya 38 30 21 10 1 Source: Centre for Information Resources Management (2004), Vulnerability Poverty Profile: Batticaloa District, December, NorthEast Provincial Council, Trincomalee. 1 WESTERN SOUTHERN SABARAGAMUWACENTRAL W A EASTERN NORTH NORTHNORTHERN SR- WESTERN CENTRAL LANK!. 1997 7.7% 4.4% -9.9% 11.1% 4.9% 9.0% 13.1% -7.6% 20.7% 6.3% 1998 7.1% 10.3% -8.4% -1.9% 3.6% 16.1% 4.2% -4.9% 9.1% 4.7% 1999 12.3% 7.3% 0.4% -2.7% -14.1% -4.8% -9.9% 20.2% -8.1% 4.3% 2000 7.9% 4.6% 10.4% 8.6% 1.9% -4.1% 6.5% -1.0% -9.1% 6.0% 2001 -4.2% 1.3% -6.3% -1.9% 15.6% 8.2% 0.9% -5.6% 1.6% -1.5% 2002 3.7% 2.2% 11.0% 5.5% -4.0% 1.9% 0.3% 11.5% 17.0% 4.0% 2003 8.8% 8.7% -6.0% -0.7% 9.2% 19.1% -3.1% 5.0% 8.3% 6.0% AVERAGE 1997-2001 6.0% 5.5% -3.0% 2.5% 1.9% 4.6% 2.7% -0.2% 3.4% 3.9% 2002-2003 6.2% 5.4% 2.1% 2.3% 2.4% 10.19'0 -1.4% 8.2% 12.6% 5.00/, 144 Annex 8 I. Theevolutionofownershipandmanagementofestates Duringthe colonial period and first two decades ofindependence (1830to 1972),the estates were owned by foreign ("sterling") companies registered under British law or local ("rupee") private companies registered under Sri Lankan Iaw. Land was obtained through a series of Acts and Ordinances, which most often acquired land with little or no compensation paid to current ownershsers. Labor was acquired through migrant labor from South India. Critics o f this structure, even during the pre-independence period, saw it as an enclave o f foreign capital, management and labor, which also deprived the rural population o f their land and seriously affected their sustainability as peasant farmers. Nationalization: These critiques culminated in the introduction o f policies to end foreign ownershp and unequal distribution o f land. The LandReform (Amendment) Law No.39 o f 1975 resulted in the nationalization o f all privately owned estates, with 417,957 acres o f estate land owned or possessed by public companies vested in the Land Reform Commission. At the time, this represented about 63 percent o f the country's tea acreage, 32 percent o f the rubber acreage and 11 percent of the coconut acreage. The management o f these entities was handed over to government organizations. In 1976, the government established the Janatha Estate Development Board and the State Plantations Corporation for plantation-style management to exploit economies of scale. Re-privatization phase 1(management): On the recommendations of a task force, the government decided to restructure the state owned plantations in 1992. The first phase o f the program transferred only the management aspects o f plantation, granted at a nominal annual rent of Rs.500. The government created 23 state-owned Regional Plantation Companies (RPCs) where each RPC entered into an agreement with a private company, the Management Agent (MA), chosen through an open bid procedure, where only S n Lankan bidders were allowed. The MAS were contracted for an initial period o f 5 ?4years with provision for extending by further periods of 5 years subject to certain levels o f profitability. However, due to the short-term nature o f the lease agreements, the newly formed MASfound it difficult to raise money to runthe plantations. Re-privatization phase 2 (selling controlling interest): In February 1995, the newly elected government decided in favor of a fuller privatization of State Plantations. A program for the sale o f controlling interests inthe RPCs was announced inJune 1995,with the following elements: (a) reduction of the lease period from 99 years to 50 years and the nominal lease rentals to be increased substantially fkom Rs.500 per year per estate and revised annually; (b) MASo f RPCs that had shown operational profits were eligible to purchase 51 per cent of the shares at the Colombo Stock Exchange market price on an all or nothing basis; (c) 20 per cent of the shares to be offered for sale to the general public; (d) 10 percent o f the shares to be distributed free of charge among the employees o f the RPCs; (e) the remaining to belong to the government for the time being; and (f) government to own a Golden Share in each RPC in order to exercise control over certain affairs. 145 If. Asset index Due to the lack o f consumption information inthe survey, Asset Index is usedto capture the well- being of estate households.' The principal component method is used to choose appropriate weights o f household assets for the index (Filmer and Pritchett, 2001). The basic idea of principal component i s to find a linear combination of the asset variables that contains the most information. The assets used are listed in the footnote.2 The Asset Index (AI) performs well in terms o f robustness, internal coherence and external validity. The asset index is proven to be robust to the choice of assets since the ranking o f households do not change much after excluding certain assets. The Spearman rank correlations o f the base case asset index (including all assets) and other indices (calculated with more limited number of assets) are high and significant in all cases. The internal coherence o f the Asset Index is shown by the fact that households in the higher AI quintile own more assets. External validity can be shown by comparing the Asset Index to household attributes that are conventionally correlated with poverty such as education. Households inhigher AI-quintiles have higher education attainment. 111. SurveyMethodology The quantitative survey was conducted between October and December, while the qualitative study filed work was carried out between March and M a y o f 2005. The surveys cover estate sector, definedby the DCS as plantation areas o f more than 20 acres inextent that have at least 10 residential labourers. The quantitative survey covered more estates and households, while the qualitative survey employed more instruments and hold in-dept interviews with various stake holders. Differences between methodologies o f the two surveys are summarized below. ' The asset index, or the f i s t principalcomponent, is expressedas K AssetIndex, = sk(akr- f sk ,where ak,is the value of asset k that household i has, Zk is the ak) k=l mean, and sk is the standarddeviationofthe asset k. 2 Own vehicle ; Own motorcycle; Own TV; Own radio or recorder; Own VCRNCD Player; Own refrigerator; Owntelephone (land or cellular); Own sewing machine; Number of water buffaloes; Number ofcows; Numberof goats ;Numberofpigs; Numberofpoultry; Own farm with title; Own farmwith grant or permit; Own farm without document; Own home plot with title; Own home plot with grant or permit; Own home plot without document; Own dwelling with title; Own dwelling with grant or permit; Own dwelling without document; Number of rooms; Dwelling type-Line room (DB); Dwelling type-line room (SB); Dwelling type-Twin cottage; Dwelling type-Separate house; Dwelling type-Temporary Shed; Dwelling type-Upstairs Barrack; Cookingfuel-Firewood;Cookmg fuel-Sawdusdpaddy husk; Cookingfuel- Kerosene; Cooking fuel-Gas; Cooking fuel-Electricity; Wall-Brick; Wall-Mud; Wall-Wood; Wall-Metal sheet; Roof-Concrete; Roof-Tile; Roof-Tin sheet; Roof-Asbestos; Roof-CadjadpalmyraWthatch; Floor- Terrazzo/ tile; Floor-Cement; Floor-Wood; Floor-Dungimud; Floor-Sand; Amenities-Electricity - Main grid; Amenities-Electricity - Non grid; Amenities-Own toilet; Amenities-Home garden plot; Source of drinking water-Inside well; Source of drinking water-Outside well; Source of drinking water-Unprotected well; Source of drinking water-Tube well; Source of drinking water-Public tab; Source of drinking water- Inside tap; Source of drinking water-Riverhank, Toilet-Water seal; Toilet-Pour flush; Toilet-Pit; Toilet- Bucketlatrine. Most asset variables take 0 and 1values, unless indicatedotherwise. 146 Oualitative Survev - Quantitative Survey Number of estates 20 estates* 50 estates divided into 100PSUs** Number o f 157 households 1030 households (1007responded) households Methodology 3-stage sampling Level 1:district 50 estates were selected purposively fiom (poverty and spread o f crops), Level 5 distncts with significant estate activity 2: estate (based on management based on management type, and crop and type, size, remoteness, and profile remoteness. The DCS then applied the oflabor), Level 3: household stratification and drew a sample of 100 (stratify HHinselected estates into PSUs from a total o f 668 census blocks in 3 categories: `bottom', `medium' the 50 estates. About 10households per and `top' based on FGD PSU were randomly drawn and assessment). mterviewed. Coverage: Crop 10estates for eachcrop, over- 35 tea and 15 rubber estates, reflecting sampled rubber estates. greater extent of tea estates inSri Lanka. Coverage: Nuwara Eliya, Badulla, Kandy, Nuwara Eliya, Badulla, Kandy, Districts Ratnapura, Kegalle and Kalutara Ratnapura, andKegalle Coverage: More private estates than inthe Only 4 pnvate estates inKandy and Management type quantitative survey Ratnapura districts, out of the initial 50 estates. Coverage o f FGDmaycover non-resident estate Onlyhouseholds located inthe estates resident workers workers Survey Office Based Information, Householdquestionnaire (by a respondent Instruments CommunityTime Line, Female withinthe households) and community FGD, Male FGD, YouthFGD, questionnaire (by key informants). Individual Life Story, and Certain topics are answeredby the female Additional K e y Person Interviews. householdheads or spouses of heads. Remark Thesample estates in the qualitative survey are a subset of the quantitative survey, with the exception of seven estates that do not overlap with the quantitative sampIe. Note: Number of estates. Initially, 50 estates were selectec however, the survey was conducted in fewer estates because o f the difficultyinidentifying small, privately owned estates. Thisproblem resulted inthe small sample size o fhouseholds inprivately managed estates. *For the community-level data collectioninlarge estates, an estate division was defined as a `community'; insmall and mediumestates, respondents viewed the entire estate as a `community'. **Inthe final data set, the householdsurvey was conducted inabout 106PSUs. 147 r - 1 80 I 60 40 I 20 I O Figure A-8.2: Percentage of migrants to estate population by age group I Migrationfor work I e15 15-19 20-24 25-29 30-34 35-44 4554 55+ I Figure A-8.3: Coverage of transfers by NIC I I I I / Source: MOP Estate quantitative survey (2005) 148 Table A-8.1: Poverty profile: Multivariate regression results (1) (2) (3) OLS Probit OLS Estate's major crop is rubber 0.081 0.099 0.093 (0.62) (1.32) (0.70) Estate's management: State 0.417 0.836 0.357 (1.06) (20.63)** (0.90) Estate's management: Private 0.391 -0.201 0.415 (2.03)* (2.65)** (2.1 1)* District dummy: Kandy -0.153 -0.363 -0.102 (0.40) (0.27) District dummy: Nuwara 0.280 -0.065 0.302 (4.19)** (I.67) (4.27)** District dummy: Ratnapura -0.352 0.252 -0.376 (3.10)** (3.63)** (3.16)** District dummy: Kagalle -0.395 0.134 -0.405 (2.62)** (1.52) (2.64)** Road passable all year 0.186 -0.098 0.185 (3.31)** (2.95)** (3.25)** Distance to the nearest town (Km) -0.004 (0.76) HHsize 0.093 -0.058 0.097 (6.32)** (6.00)** (6.04)** Head edu: grade 6-9 0.356 -0.128 0.350 (6.14)** (3.96)** (6.04)** Head edu: 0 level or above 1.251 -0.242 1.254 (1 1.98)** (5.19)** (11.95)** Head's age 0.006 -0.004 0.006 (2.76)** (2.63)** (2.41)* Head hasNational ID 0.306 -0.160 0.291 (3.71)** (3.20)** (3.52)** Prop. Offemale FM -0.178 -0.015 -0.171 (1.44) (0.21) (1.38) Head is IndianTamil -0.129 -0.010 -0.151 (2.03) * (0.26) (2.33)* Live in estate over 20 years -0.172 0.063 -0.180 (2.50)* (1.55) (2.61)** Prop. of employed FM 0.078 (0.60) Dependencyratio -0.135 0.073 -0.125 (3.16)** (2.98)** (2.73)** Dummy:HHhas migrant(s) working abroad 0.416 -0.066 0.399 (3. I3)** (0.95) (3.00)** HHreceive income from estatewages and salaries -0.140 (2.05)' HHreceive income from outsidewages andsalaries 0.018 (0.32) HHreceive income from entrepreneurial sources 0.308 -0.155 0.295 (3.63)** (3.16)** (3.44)** HHismember oftrade unions or politicalparties 0.063 149 (0.90) HHparticipated inhousingprogram(s) (dummy) 0.298 -0.157 0.236 (4.19)** (3.88)** (3.00)** HHparticipated inother social program(s) (dummy) 0.085 (1.44) Constant -1.050 -0.998 (5.85)** (4.60)** Observations 95 1 951 95 1 R-squared 0.30 0.3 1 Note: significant at 5%; * ** significantat 1%Absolute value oft statisticsinparentheses "Prop. ofemp. FM" refersto proportionof employedfamily members. OLS: Ordinary Least SquaredNote: Model (1) is the mainregressionwhere all the essential attributes are regressedagainst asset index, the dependentvariable, with OLS. Model (2) employs the same right hand side variables, but use povertystatus (0-1) as dependent variable. Reportedvalues in Model (2) are marginal effects or probability of gettingout of poverty. Badulla is the omitted district; therefore, the coefficients are changes in welfare comparedto Badulla. Outside earnings 469 520 645 663 618 559 NONIC Estate earnings 426 1,468 1,921 1,789 411 1,093 Total earnings 895 1,988 2,566 2,451 1,029 1,652 Outside earnings 907 1,140 1,583 1,108 835 1,178 WithNIC Estate earnings 358 896 1,666 1,882 1,007 1,465 Total earnings 1,265 2,036 3,249 2,989 1,842 2,643 Outside earnings 623 1,019 1,480 1,041 806 1,046 Total Estate earnings 402 1,008 1,694 1,868 929 1,386 Total earnings 1,025 2,027 3,174 2,909 1,736 2,433 150 Estate-casual 3,327 23 2.9 56.79 Estate-regular 4,210 24 21.6 9.45 Inside-casual 3,222 21 11.6 2.2 Outside-casual 3,693 19 10.2 15.07 Outside-regular 5,000 24 27.3 10.11 Inside-self 6,993 24 26.8 2.86 Outside-self 4,652 25 21.0 3.17 Employer 24,000 28 57.1 0.36 All workers 3,853 22 9.9 100 Factory 3,238 23 6.I 12.63 Field 3,262 23 3.0 75.52 Sundry 4,486 24 4.5 6.77 Supervisor 5,433 27 51.4 2.69 Other 5,817 26 35.5 2.39 All estateworkers 3,462 23 5.5 100 Agnculture 4,003 20 9.5 30.13 Industry 4,303 22 16.8 20.54 Trade 4,757 25 17.8 21.89 Service 5,711 23 26.8 27.44 All non-estate workers 4,698 22 I 17.5 I 100 Table A-8.4: Profile of estate householdswith migrants Destination Urban Abroad All* Size of estate Smaller than 150 16 I 17 151-250acres 16 3 20 25 1-500acres 13 5 18 Larger than 500 2 8 11 District Badulla 6 1 7 Kandy 12 10 24 Kegalle 6 12 19 Nuwara Eliya 26 3 29 Ratnapura 3 3 7 Management RPC 13 4 17 State 14 12 27 151 Private 3 0 3 Major crop Tea 14 3 1s Rubber 5 10 17 Ethnicity ofhead Sinhala 3 6 Sri Lankan Tamil 14 23 IndianTamil 14 1s Muslim 0 0 Headhas NationalIDcard N o 12 4 20 Yes 13 4 17 Head ofhousehold's education attainment No schooling 11 18 Grade 1-5 15 20 Grade 6 -9 11 13 0level 17 21 A level andbeyond 12 0 12 Total 13 4 17 Note: Migrants refer to individuals who used to live inhouseholds in the past 5 years, migrated for work only*: "All" includesa small groupofhouseholdswhose migrantswork inrural areas. UrbanMigration OverseasMigration (1) (2) (3) (4) Work inurban areas -0.121 0.021 (1.51) (0.14) Work inurban areas and remitregularly -0.192 (1.13) Work abroad 0.431 0.071 (3.20)** (0.37) Work abroad andremit regularly 0.684 (2.65)** Observations 970 970 970 970 R-squared 0.27 0.27 0.28 0.2s Table A-8.6: Probit regressionof likelihoodof an estate householdhaving a migrant UrbanMigration OverseasMigration (1) (2) Estate'smanagement:State 0.535 1.471 (2.08)* (4.03)** District dummy: NuwaraEliya 1.048 0.643 (6.74)* * (2.05)* 152 District dummy: Ratnapura -0.079 0.701 (0.31) (1.89) District dummy: Kagalle 0.155 1.598 (0.64) (4.93)** HHsize 0.277 0.I86 (8.08)** (3.81)** Head edu: grade 6-9 0.012 -0.185 (0.09) (0.92) Head edu: 0 level or above 0.465 -0.708 (2.02)* (1.50) Head's age -0.001 -0.015 (0.23) (1.88) Head has National ID -0.267 0.144 (1.42) (0.52) Head is Indian Tamil -0.113 -0.097 (0.77) (0.50) Live in estate over 20 years 0.356 0.157 (2.12)* (0.68) Prop. o f female FM 1.216 -0.424 (4.02)** (1.01) Presence o f children aged 6 or younger -0.650 -0.292 (4.34)** (1.48) Presence o f children aged 7-14 -0.560 -0.119 (4.07)** (0.62) Presence o f FM aged 65 or more -0.063 0.209 (0.40) (0.97) Constant -3.271 -2.606 (7.60)** (4.59)** Observations 1002 1002 Absolute value o fz statistics inparentheses;* significant at 5%;** significantat 1% Note: Householdsize before migration is the sum of current household size and number o fmigrants. Dependent variables: (I) migration: household has any migrant(s) working inurbanSri Lanka; urban (2) overseas migration: household has any migrant(s) working abroad. Table A-8.7: Characteristics of households by migration destination and remittance Destination Urban Sri Lanka Abroad No Remit regularly Yes No Total Yes No Total migration Asset Index -0.02 0.04 -0.01 0.56 -0.19 0.20 -0.02 AI-Poverty rate (%) 22 28 23 25 42 33 31 Earnings per worker 3560 3644 3581 4377 3842 4116 3469 Earnings per capita 1740 2215 1858 2087 1674 1886 1796 Total HHearnings 6597 10068 7458 8649 7973 8320 7220 %HHmembers who work: off-estate inagriculture 6 8 6 3 14 8 10 off-estate in industry 6 15 8 16 8 12 5 off-estate intrade 7 4 6 7 16 I 1 5 off-estate inservices 4 14 6 8 10 9 3 estate casual employees 59 36 53 47 15 32 51 153 estateregular employees 1 6 3 13 18 15 10 estate: total 60 42 56 60 33 47 62 % ofHHhead attained: grade 1-5 edu. 56 34 50 60 53 56 43 grade 6-9 edu. 20 44 26 25 21 23 33 O/L or better 9 9 9 0 5 3 7 %highest edu inHHi s OK or better 23 19 22 20 11 15 22 Table A-8.8: Percentages of householdsreceiving cash transfer (Samurdhi/sociaI welfare) Asset index quintile 1 (poorest) 2 3 4 5 (richest) Total Size of estate Smaller than 150 8 I 5 20 15 9 13 151-250 acres 20 10 12 9 4 11 251-500 acres 14 17 8 10 6 12 Larger than 500 8 13 21 17 9 13 District Kandy 13 0 13 - 36 40 23 Nuwara Eliya 0 0 5 0 0 1 Badulla 26 15 15 23 20 19 Ratnapura 29 39 32 17 11 28 Kegalle 4 13 8 9 11 8 Management RPC 13 14 13 12 7 12 State 13 0 18 40 42 25 Private 33 25 9 0 0 13 Major crop Tea 17 12 11 12 10 12 Rubber 7 19 30 19 13 15 Ethnicity of head Sinhala 20 25 30 20 11 17 Sri Lankan Tamil 15 27 14 19 27 20 Indian Tamil 12 9 12 10 7 10 Muslim 100 33 20 33 HeadhasNational IDcard N o 7 14 6 5 29 9 Yes 16 14 14 13 10 13 Head ofhousehold's education attainment N o schooling 13 18 14 18 22 16 Grade 1- 5 14 12 11 13 7 12 Grade 6 - 9 16 15 14 10 15 14 0level 0 0 29 17 3 8 A level and beyond 0 0 50 0 0 4 Total 14 14 13 13 10 13 154 Table A-8.9: Access to social programs (%) Housing Wateritoilet Training Microcredit Crech ECD None Size o f estate Smaller than 150 51 68 60 44 25 28 25 151-250 acres 59 74 50 35 38 36 8 251-500 acres 47 71 35 36 41 50 12 Larger than 500 54 86 31 21 61 61 0 District Kandy 55 71 15 45 41 30 0 Nuwara Eliya 37 55 57 30 27 30 17 Badulla 59 72 31 25 22 50 19 Ratnapura 56 94 57 86 69 56 0 Kegalle 58 100 42 25 68 34 0 Management W C 50 72 47 36 38 42 13 State 49 68 17 51 34 34 0 Private 74 100 26 51 49 29 0 Major crop Tea 49 68 42 35 31 40 14 Rubber 61 95 57 49 73 44 0 Total 51 73 44 38 38 41 12 Note: The table showspercentages o fhouseholds living incommunity where the social programs exist. ECD refers to early childhood development program. None refers to communities where none of the social program exists. Table A-8.10: Participation by households in social programs (%) Housing Wateritoilet Training Microcredit Crech ECD None Size of estate Smaller than 150 13 21 13 7 7 7 64 151-250 acres 18 26 12 8 11 9 55 251-500 acres 18 23 8 9 12 14 48 Larger than 500 13 28 6 10 19 14 51 District Kandy 12 26 2 11 11 3 50 NuwaraEltya 16 18 16 5 8 8 59 Badulla 19 26 6 3 8 12 57 Ratnapura 14 28 11 36 21 19 35 Kegalle 9 30 6 3 14 6 58 Management RPC 16 24 11 9 11 11 54 State 10 25 2 12 10 3 49 Private 12 21 9 24 9 9 51 Major crop Tea 16 23 11 8 10 11 54 Rubber 13 28 8 15 17 10 52 Ethnicity o f head Sinhala 13 17 7 13 11 10 61 Sri Lankan Tamil 18 24 3 5 10 9 56 IndianTam1 15 25 13 10 11 11 52 Muslim 10 10 0 20 0 10 60 155 HeadhasNationalIDcard No 18 25 7 9 9 9 55 Yes 15 24 11 I O 11 I 1 54 Headofhousehold'seducationattainment No schooling 17 25 11 8 9 11 56 Grade 1-5 15 24 10 8 10 11 54 Grade G -9 15 25 9 12 13 11 50 0level 12 19 10 13 12 4 60 A level and beyond 25 21 21 17 13 13 67 Total 15 24 10 10 11 11 54 Note: ECDrefers to earlychildhooddevelopmentprogram.Nonerefers to householdsthat did not participatedin any programs. 156