Poverty and Welfare in Sri Lanka 103281 SRI LANKA Poverty and Welfare: Recent Progress and Remaining Challenges* * Prepared by David Newhouse, Pablo Suarez-Becerra, and Dung Doan from the World Bank’s Poverty and Equity Global Practice. This analysis draws on background papers by Lidia Ceriani, Gabriela Inchauste, and Sergio Olivieri on the components of poverty reduction, by Nicola Amendola, Susan Razzaz, and Giovanni Vecchi on the robustness of Sri Lanka’s poverty measure, by Ana Abras on inequality in access to services, and by Dung Doan and Dilhanie Deepawansa on pockets of poverty. The team also gratefully acknowledges comments on earlier drafts from Lidia Ceriani, Gabriela Inchauste, Sergio Olivieri, and Charles Undeland. The analysis draws heavily from “Ending Poverty and Promoting Shared Prosperity: A Systematic Country Diagnostic”, led by Gabriela Inchauste and Charles Undeland, and benefited greatly from the comments of peer reviewers Dean Jolliffe, Ambar Narayan, Tara Vishwanath, and Thomas Walker. 1 TAbLe of CoNTeNTS List of Abbreviations ..........................................................................................................................................................................................................3 Country Context and Scope of Analysis .......................................................................................................................................................................5 Trends in Poverty, Welfare, and Inequality ..................................................................................................................................................................9 Sri Lanka Has Made Laudable Progress in Reducing Poverty.........................................................................................................................9 Living Standards and Access to Basic Services Have Also Improved........................................................................................................ 13 However, Progress in Reducing Poverty Has Not Been Universal ............................................................................................................... 16 Drivers of Poverty and Inequality Reduction ............................................................................................................................................................ 23 Broad-based Growth Benefited the Poor Particularly from 2006/07 to 2009/10 ................................................................................ 23 Higher Labor Earnings Were the Main Contributor to Poverty Reduction ................................................................................................ 26 Potential Causes of Increased Labor Income ......................................................................................................................................................... 30 The Gradual Structural Transformation.............................................................................................................................................................. 30 Urbanization and Agglomeration ......................................................................................................................................................................... 34 Rising International Food and Tea Prices during 2006-2009 ..................................................................................................................... 35 Growth in domestic consumption ........................................................................................................................................................................ 37 The Current State of Poverty and Inequality............................................................................................................................................................ 40 Despite the Encouraging Reduction in Poverty, Living Standards Remain Modest for Most of the Country ................................ 40 Geographic Pockets of Poverty Remain ............................................................................................................................................................. 40 The Poor and Near-Poor Tend to be Rural, Young, and Disconnected from Productive Earnings Opportunities......................... 42 The Living Standards of the Near-Poor are Closer to Those of the Poor Than of the Top 60 Percent. ........................................... 46 Poverty in the Estate Sector .................................................................................................................................................................................. 49 Inequality is Rising and Access to Services Remains Inequitable.............................................................................................................. 50 Social Protection and Poverty ...................................................................................................................................................................................... 54 Public Transfers are Fragmented and Small in Both Their Operational Scope and Financial Scale ................................................ 54 Demographic Changes are Imposing New Challenges to the Social Safety Net .................................................................................... 60 Conclusions ....................................................................................................................................................................................................................... 62 Appendices ........................................................................................................................................................................................................................ 66 Appendix 1: Household and Labor Force Surveys in Sri Lanka .................................................................................................................... 66 Appendix 2: Survey-to-Survey Imputation in Sri Lanka ................................................................................................................................ 68 Appendix 3: Key issues in Estimating Poverty in Sri Lanka........................................................................................................................... 70 Appendix 4: Multi -dimensional Inequality Among Sri Lankan Children .................................................................................................. 72 Appendix 5: Profile of the Poor and Near-Poor by Sector and in Former Conflict Provinces ............................................................. 74 References ......................................................................................................................................................................................................................... 76 2 Poverty and Welfare in Sri Lanka LIST of AbbRevIATIoNS CCPI Colombo Consumers’ Price Index CPH Census of Population and Housing DCS Department of Census and Statistics DS Divisional Secretariat FAOSTAT Statistics Division of Food and Agriculture Organization GDP Gross Domestic Product HIES Household Income and Expenditure Survey I2D2 International Income Distribution Database IPS Institute for Policy Studies of Sri Lanka LFS Labor Force Survey PPP Purchasing Power Parity WDI World Development Indicators 3 4 Poverty and Welfare in Sri Lanka I. CouNTRy CoNTexT ANd SCope of ANALySIS Analysis of Sri Lanka’s recent progress in reducing poverty and inequality is directly relevant to the new government’s development agenda. The newly sworn-in president ran for election on a platform that featured, among other goals, inclusive growth and support to the agricultural sector. The pursuit of these and other goals of the new administration can be informed by a fuller understanding of recent developments in household living standards across the country. Yet the World Bank’s most recent poverty assessment in Sri Lanka, covering the period from 1990 to 2002, was published a decade ago. Since then, domestic economic growth, the end of the civil conflict, and fluctuations in global markets have led to substantial changes in Sri Lanka’s economic environment. To inform the new government’s development policies, this report examines five topics related to recent developments in poverty and welfare. Sections II through V of the report focus on (i) trends in poverty, welfare, and inequality since 2002, (ii) labor market outcomes associated with the observed reduction in poverty, (iii) four potential causes of this poverty reduction, (iv) the state of poverty and inequality in 2012/13, and (v) the role of social protection in reducing poverty. Section VI concludes by pointing out future implications and remaining knowledge gaps to continue to reduce poverty and improve living standards. This analysis draws mainly on data from the 2002, 2006/07, 2009/10, and 2012/13 rounds of the Household Income and Expenditure Survey, supplemented by annual rounds of the Labor Force Survey from 2002 to 2012. Since the surveys could not be conducted in parts of the Northern and Eastern provinces before 2011 due to the civil conflict, their geographical coverage varies from year to year. To ensure comparability, all historical trends presented in this report correspond to the same geographic area. With the exception of figures that are based solely on 2012/13 data, the figures exclude Northern and Eastern provinces, which account for about 12.9 percent of the total population. A more detailed description of the data is provided in Appendix 1. 5 I. CouNTRy CoNTexT ANd SCope of ANALySIS Overall, Sri Lanka’s record of poverty reduction has been encouraging, mainly due to increased labor earnings as demand for labor rose across a broad range of sectors. Excluding Northern and Eastern provinces, the poverty headcount rate fell from about 22.7 percent in 2002 to 6.1 percent in 2012/13. Per capita incomes also rose for the poor, corroborated by increases in the share of consumption devoted to non-food items, ownership of durable goods, and school attendance. Wages and employment grew as well, particularly in the manufacturing, construction, commerce, transport, and communication sectors. Sizeable increases in agriculture minimum wages led to higher earnings for agricultural workers. While it is difficult to identify the underlying causes of this poverty reduction, this report focuses on four potential factors. They include (i) the economy’s gradual structural transformation out of agriculture into more productive sectors, (ii) urbanization and agglomeration around key urban areas, (iii) rising international prices for food and tea that raised earnings in agriculture, and (iv) strong domestic aggregate demand that has boosted economic growth. Of these potential factors, a more rapid structural transformation and increased agglomeration have the most potential to sustain poverty reduction in the future. Growth in the agricultural sector during this period largely reflected rising prices and an expansion in arable land, neither of which are likely to be sustained. Domestic aggregate demand, meanwhile, has been led by the construction and transport sectors, spurred in part by public investment in the aftermath of the conflict. These sectors, however, cannot be relied upon to produce sustained growth. This suggests that efforts to further improve living standards of the poor focus on promoting further structural transformation and urbanization. Despite this recent progress, living standards remain low and pockets of severe poverty persist. Around 40 percent of the population subsists on less than twice the poverty line, which was $2.75 per capita per day in 2005 PPP terms, or 225 rupees per day. Furthermore, living standards of the near-poor – those above the national poverty line but below the 40th percentile -- are similar to those of the poor. Moreover, low-income households’ ability to access basic services and public facilities has barely improved since 2002. The population in Northern and Eastern provinces is particularly disadvantaged in terms of consumption, labor market outcomes, educational attainment, and housing conditions. Finally, inequality increased sharply from 2009/10 to 2012/13. popuLATIoN LIvINg uNdeR $2.75 A dAy IN 2005 ppp 40% TeRmS, oR RS. 225 A dAy 6 Poverty and Welfare in Sri Lanka Spending on social assistance programs is low and has declined, and therefore contributed little to poverty reduction. In the short run, more generous social assistance programs, as well as multi- sectoral interventions targeted to the remaining pockets of poverty, can help support the existing poor. In the longer term, adjustments to Sri Lanka’s social protection system are necessary to address challenges related to a growing middle class and an aging population. Policies can also do more to speed the structural transformation out of rural and peri-urban agriculture into more productive work. Roughly 30 percent of the workforce, and about half of the working poor, toil in the agriculture sector. Many of the poor live in peri-urban areas – over half of the poor are estimated to live within 30 km of a main agglomeration area. Policies that help connect these workers to productive employment opportunities off the farm can contribute to sustainable poverty reduction. Important knowledge gaps remain, however, starting with the relative importance of the four potential factors described above in explaining the observed poverty reduction. Internal migration and more detailed information on access to infrastructure have yet to be thoroughly investigated. Nor does longitudinal data exist to examine how, or how frequently, households enter and exit poverty. Little is known about the underlying causes of the rise in labor demand – did workers’ skills improve or was technology used more effectively? How much do labor regulations constrain further growth in wage employment? Bridging these knowledge gaps and documenting how public policies affect the poor can inform much-needed efforts to further improve their living standards. 7 8 Poverty and Welfare in Sri Lanka II. TReNdS IN poveRTy, weLfARe, ANd INequALITy SRI LANKA hAS mAde LAudAbLe pRogReSS IN ReduCINg poveRTy Poverty reduction since 2002 has been solid by regional standards. The poverty rate is defined as the proportion of the population whose household’s per capita expenditure falls below the poverty line. The current poverty line was set at Rs. 1,423 per person per day in 2002 rupee, and has been updated since using the Colombo Consumer Price Index (CCPI). A new poverty line for 2012/13, which will more accurately reflect recent consumption patterns, is currently being developed. But under the current line, which allows for monitoring poverty trends, headcount poverty fell from 22.7 percent to 6.1 percent between 2002 and 2012/13 in districts outside Northern and Eastern provinces (Figure 1). Sri Lanka’s poverty is low by international standards. For cross-country comparisons, the World Bank currently uses an extreme poverty line of USD 1.25 per person per day (in 2005 PPP terms). By this measure, extreme poverty in Sri Lanka decreased from 13 percent in 2002 to less than 3 percent in 2012/13, and is lower than many of Sri Lanka’s neighbors, other post-conflict countries, and other comparable countries (see Figure 2). Figure 1: 10 Conflict's "Second "Third Ceasefire Final Peace 100% Sri Lanka has made early days Stage" Stage" Stage significant progress and India's 8 intervention 80% Headcount Poverty Rate in reducing poverty Real GDP Growth (%) incidence amidst robust 6 60% growth 4 28.8% 40% 26.1% 22.7% Source: WDI and PovcalNet 2 13.5% 20% 7.4% 6.1% for figures prior to 2002, World Bank staff calculations 0 0 based on HIES excluding Northern and Eastern -2 -20% Provinces from 2002 to 1984 1986 1988 1990 1992 1994 1996 2000 2002 2004 2006 2008 2010 2012 2014 1998 2012/13. GDP Growth Poverty with national line Notes: Poverty rates are obtained by inflating the 2002 poverty line to 2006/07 using the base 2002 CCPI and subsequently inflating the poverty line to 2009/10 and 2012/13 using the base 2006/07 CCPI. This, along with the exclusion of Northern and Eastern provinces, causes the results reported above to differ slightly from officially reported estimates. 9 II. TReNdS IN poveRTy, weLfARe, ANd INequALITy Figure 2: Sri Lanka’s extreme poverty rate (latest available) 50% poverty rate is low relative 43.3 PPP$1.25 Headcount to comparable countries 40% 32.6 32.8 30.3 30% 23.7 Source: WDI 20% 18 19 12.7 14.1 10% 5.6 6.5 8 8.5 5.3 1.6 2.4 2.4 2.8 2.9 0% Dominican Rep. Lao PDR Peru Pakistan Thailand Bhutan Vietnam Sri Lanka Colombia Tajikistan Bolivia Nicaragua Georgia Indonesia Philippines Nepal India Cambodia Bangladesh Other measures of consumption poverty have also fallen along with poverty incidence. The poverty gap index, which measures the average shortfall of the total population relative to the poverty line, narrowed from 5 percent to 1 percent of the national poverty line between 2002 and 2012/13. During the same period, the poverty severity index, which gives greater weight to those further below the poverty line, also fell rapidly from 1.6 percent to 0.3 percent. Figure 3: Poverty gap, poverty severity Poverty gap and poverty 10% 25% 22.7% severity have been Poverty Incidence 8% 20% decreasing 6% 5.1% 13.5% 15% Source: World Bank staff’s 4% 10% calculation from HIES 7.4% 2.6% 6.1% 2% 1.6% 1.4% 5% 0.8% 1.0% 0.4% 0.3% 0 0 2002 2006/07 2009/10 2012/13 Poverty gap Poverty severity Poverty incidence The recorded decrease in poverty is robust to a variety of methodological choices in constructing the poverty line. These include the varying geographic coverage of the surveys described above, the selection of the reference group from which the weights of the Consumer Price Index (CPI) are determined, the inclusion of savings and durable goods in the consumption aggregate, and the methods used to inflate the consumption aggregate over space and time. Robustness checks show that these have minor effects on the estimated poverty rate. The most important issue is the CPI, which is used to inflate the poverty line. Using unit values to measure food inflation instead of the food component of the CPI could raise the estimated headcount ratio by a mild but noticeable amount (see Appendix 3). 10 Poverty and Welfare in Sri Lanka how IS poveRTy meASuRed IN SRI LANKA? POvErTy IncIDEncE The official poverty figures in Sri Lanka refer to the share of individuals whose household per capita consumption falls below the official poverty line. This indicator is referred to as the poverty headcount index and is the standard measure of the incidence of poverty. The consumption aggregate used to calculate the headcount index is the sum of all food and non-food expenditures collected in the Household Income and Expenditure Survey. The consumption aggregate is spatially deflated, using unit values of expenditure items from household survey, to take into account differences in the cost of living across different districts. To obtain per capita consumption, the spatially deflated consumption aggregate is divided by the number of household members, excluding those who are members of the household but usually live elsewhere in the country or abroad. Once per capita household consumption is calculated, it is then compared to the national poverty line. This line was defined as the expenditure for a person to meet the daily calorie intake of 2,030 kcal based on the Cost of Basic Needs approach, and was set at Rs 1,423 in 2002. For more recent years, this line has been inflated using the Colombo Consumer Price Index (CCPI), to keep the national poverty line constant in real term since 2002. To facilitate comparisons across countries, the World Bank uses a common poverty line of USD 1.25 per person per day in 2005 PPP terms – approximately Rs 100 per day or Rs 3,100 per month in 2012/13 – to measure extreme poverty. This represents the mean of the poverty lines found in the poorest 15 countries ranked by per capita consumption among 88 surveyed countries over the period 1990-2005 (Ravallion, Chen, and Sangraula, 2009). Other commonly used poverty lines for international comparisons across middle-income countries are USD 2.50 a day and USD 4.00 a day (2005 PPP), which correspond to, respectively, Rs 6,058 and Rs 9,692 a month in 2012/13 prices. Sri Lanka’s national poverty line is equivalent to about USD 1.50 a day in 2005 PPP terms. This is moderate by regional standards, but below what one might expect from a country at Sri Lanka’s level of development (see Figure 4). The poverty line is based on consumption patterns in 2002, which is likely to have changed significantly in the last decade, making it important to update the current poverty line soon. 11 II. Trends In poverTy, welfare, and InequalITy (conTd.) Figure 4: Sri Lanka’s national 2.00 poverty line is mid-range Bhutan National poverty line (2005 PPP) by regional standard Maldives 1.50 Pakistan Sri Lanka Source: WDI, I2D2 Nepal India Bangladesh 1.00 0.50 0 2,000 4,000 6,000 8,000 10,000 12,000 GDP per capita (2011 PPP) Note: Sri Lanka’s and Bhutan’s poverty lines are for year 2012, India’s for 2011, Bangladesh’s, Pakistan’s, and Nepal’s for 2010, and Maldives’ for 2009. POvErTy gAP AnD SEvErITy OF POvErTy Although the headcount index is the most common measure of poverty, one shortcoming of this measure is that it does not take into account the intensity of poverty, i.e. how far from the poverty line the poor are. A measure of poverty that overcomes this problem is the poverty gap index. The poverty gap index is the average shortfall of the total population from the poverty line, measured as a percentage of the poverty line itself. A third common measure is the poverty severity index, which is calculated as the average over the total population of the squared shortfall from the poverty line. Since trends in the poverty gap and poverty severity indexes tend to track the headcount index, however, this report focuses on the latter as a measure of changes in poverty. THE BOTTOm 40 PErcEnT The analysis also focuses on the poorest 40 percent of the population to discuss the welfare of the less-well-off more broadly. The consumption cut-off line for the bottom 40 percent (the threshold below which the poorest 40 percent lays) in 2012 is Rs. 6,771, or approximately USD 2.75 a day in 2005 PPP terms. 12 Poverty and Welfare in Sri Lanka LIvINg STANdARdS ANd ACCeSS To bASIC SeRvICeS hAve ALSo ImpRoved. consumption not only increased, but also adjusted in line with rising living standards. Households in the bottom two quintiles were increasingly able to afford non-food items, as the non-food budget share for this group rose from 33.8 percent in 2002 to 45.1 percent in 2012/13. Shifts in the pattern of food consumption are also consistent with higher living standards, as households have been devoting a larger share of their budget to protein and less to cereals and starches. 32.1 Figure 5: 35 Share of food expenditure 27.6 27.6 Household foods 30 26.8 consumption patterns 21.8 25 reflect higher living 2002 20 17 standards 2012/13 11.7 15 12 10 8.2 7.1 Source: World Bank staff’s 4.4 3.7 5 calculation based on 2002 and 2012/13 HIES 0 Liquor and rich food Cereals and Vegetables Fats, sugar, Milk food and others and fruits Protein tobacco starches Per capita incomes increased, especially for the poor. Because of a change in the way income was calculated starting from 2006, changes in income are only comparable from 2006/07 to 2012/2013. During this period, average per capita incomes grew only slightly, about 2 percent overall. But for poorer households with lower per capita consumption, growth in real per capita income was over twice as fast, and was in line with similar growth rates in per capita consumption.2 Figure 6: Household income 0% 2% 4% 6% grew faster among the Poorest Decile bottom 40 percent during 2006/07-2012/13 2nd Decile Source: World Bank staff’s 3rd Decile calculation based on 2006/07 and 2012/13 HIES 4th Decile Top 60 percent Average 2 The growth in household per capita income reported above contrasts with the figures on total household income reported by DCS, which remained constant in real terms. The discrepancy arises mainly because average household size fell during this period. Be- cause larger households require more income than smaller households to maintain the same standard of living, per-capita income is preferred to total income as a measure of household welfare. 13 II. Trends In poverTy, welfare, and InequalITy (conTd.) Increased welfare was also reflected in higher ownership rates of durable assets among poorer households. Between 2006/7 and 2012/13, ownership rates of refrigerator, motorcycle, washing machine, computer, and telephone more than doubled among the bottom 40 percent. Most remarkably, the share of the bottom 40 percent that owned a cell phone rose from 13 to 76 percent (see Figure 7).3 This impressive improvement could have resulted from a combination of factors, including rising wages as well as the falling costs of motorcycles, refrigerators, and computers. Figure 7: 100 Ownership of durable 90 80.3 Asset ownership rate (%) 75.9 assets among the bottom 80 68.1 70 40 percent increased 60 significantly between 50 31.2 2006/07 and 2012/13 40 26.6 27.2 26.7 30 20.1 11.8 20 13 9.7 8.2 6.3 10 Machine 4.7 Source: World Bank staff Washing 1.2 Computer 0.8 Motor car, 0.3 van 1 calculation based on HIES. 0 Refrigerator scooter Machine (domestic) (mobile) Television Motor cycle, Sewing Telephone Telephone 2006/07 2012/13 Note: Unfortunately, data on assets were not collected in the 2002 HIES. Increase in ownership rates of Increase in share of cell- Increase in share of refrigerators for the bottom phone owners for the bottom population purchasing 40% of the population 40% of the population electricity 2006/7 - 2012/13 2006/7 - 2012/13 2002 - 2012/13 15 63 25 percentage points percentage points percentage points 3 Unfortunately, information on asset ownership was not asked in the 2002 HIES. 14 Poverty and Welfare in Sri Lanka greater ownership of durables coincided with a considerable expansion in access to electricity. Overall, the share of the population living in households that purchased electricity in the last month expanded from 69 percent in 2002 to 94 percent in 2012/13 (Figure 8)4. This expansion in electricity is reflected in brighter night-time lights (Figure 9) not only in key cities, including Colombo, Kandy, and Jaffna, but also on the south coast of Galle, Matara, and Hambantota. Figure 8: Proportion of population living in households Access to electricity 100 94% expanded 88% that purchased electricity 82% 80 69% Source: World Bank staff 60 calculation based on HIES. 40 20 0 2002 2006/07 2009/10 2012/13 Figure 9: Night-time lights became brighter in key cities and the North Source: Christopher Small, based on VIIRS and DMSP- OLS imagery. Note: White areas show stable areas of lighting over the period 1992-2012, while red areas show lights that existed in 2012 but not in 2002 or 1992 4 Since 2006, the HIES survey asks households to directly report access to electricity. Unsurprisingly, the share of households report- ing access is within a percentage point of the share of households reporting positive expenditure. We report the latter because it is also available for 2002. 15 II. Trends In poverTy, welfare, and InequalITy (conTd.) The reduction in poverty is also reflected in improved education and health outcomes. While little change was observed in primary school attendance and completion, secondary school attendance rose moderately from 2002 to 2012/13. Secondary school completion, measured by the share of 17 and 18 year olds who graduated from senior secondary school, saw a much larger increase. Health indicators also improved during this period. The rate of infant mortality fell from 13 to 8 children per thousand births, and the rate of under-nutrition, while still high, decreased from 30 to 25 percent (see Figure 10). Figure 10: Education and health outcomes have improved 89.2% 86.3% 100 85.1% Source: FAOSTAT, WDI, 90 DCS, and World Bank staff 80 60.7% 70 54.4% calculation based on HIES 50.7% 60 42.2% (%) 50 29.6% 29.1% 26.7% 24.9% 40 30 13.0% 10.6% 9.2% 20 8.2% 10 0 Secondary School Under Mortality School Attendance nutrition rate Completion lnfant rate (per 10 births) 2002 2006 2009 2012 howeveR, pRogReSS IN ReduCINg poveRTy hAS NoT beeN uNIveRSAL The pace of poverty reduction varied across regions. Figure 11 shows how poverty rates at the DS Divisions changed between 2002 and 2012. Eastern and Northern provinces, marked by gray, were not covered by the 2002 HIES, and are therefore excluded. Of those areas covered by both surveys, poverty reduction was greatest in Kalpitiya, Mundel, and Vanathawilluwa divisions in Puttalam district, as well as in Badulla and Hambantota districts. In contrast, most DS divisions in Monaragala, Colombo, and Gampaha made little progress in reducing poverty. 16 Poverty and Welfare in Sri Lanka Figure 11: Pov. reduction (% points) Jaffna Poverty reduction has 29.8 - 37.1 22.6 - 29.8 15.4 - 22.6 been uneven Kilinochchi 8.2 - 15.4 1.0 - 8.2 No data Mullaitivu Source: Sri Lanka poverty map 2012/13, estimated from a combination of the Mannar Vavuniya CPH 2011 and HIES 2012/13 Trincomalee Anuradhapura Polonnaruwa Puttalam Batticaloa Kurunegala Matale Kandy Ampara Gampaha Kegalle Nuwara EliyaBadulla Colombo Moneragala Kalutara Ratnapura Galle Hambantota Matara In the Estate sector, living standards remain low despite large declines in consumption poverty. Poverty incidence in the estate sector fell by more than 19 percentage points between 2002 and 2012/13, from 30 to 10.9 percent. Most of this impressive decrease occurred between 2006/07 and 2012/13 (Figure 12). As a result, the sectoral gaps in poverty headcount rates have narrowed significantly in recent years, even though the estate sector has traditionally lagged behind the rural and urban sectors. 17 II. Trends In poverTy, welfare, and InequalITy (conTd.) Poverty in the Estate Sector Poverty incidence Percentage of estate population in the bottom 40% of the 19 percentage points national population 2002-2012/13 60% Asset ownership Below national average Other measures of well-being continue to lag in the Estate sector. Asset ownership rates in the Estate sector, for example, remain well below the national average. Furthermore, more than 60 percent of the Estate population falls in the bottom 40 percent of the national per capita consumption distribution, making a large portion of the Estate sector vulnerable to poverty. Box 2 contains more details on the reduction in monetary poverty in the Estate sector. Section V below documents in detail how, despite a relatively small difference in headcount poverty, the Estate sector fares much worse than the rural sector along several other dimensions of welfare. Figure 12: 35 30 Poverty in the estate 30 27.7 24.7 sector decreased rapidly 25 Poverty rate(%) during 2006/07-2009/10 20 13.8 15 10.9 10.5 Source: World Bank staff 10 7.9 6.2 7.8 6.8 calculation based on HIES 5 4.1 1.8 0 2002 2006/07 2009/10 2012/13 Estate Rural Urban 18 Poverty and Welfare in Sri Lanka poveRTy ReduCTIoN IN The eSTATe SeCToR Although the estate sector accounts for only 4.4 percent of the total population – based on the 2011 Census of Housing and Population – its traditionally acute poverty situation, geographical remoteness, as well as unique socio-demographic conditions puts the sector high on the poverty reduction agenda. As shown in Figure 12, most of the decrease in estate poverty during the period of 2002- 2012/13 occurred between 2006 and 2009 (from 28 to 11 percent), coinciding with the period when poverty reduction had the highest elasticity with respect to economic growth (Figure 19). One potential explanation for this sharp decline in estate poverty is the substantial increase in the price of tea, which is the major output of the estate sector as well as Sri Lanka’s largest exported commodity by export value. The Colombo auction price for tea surged from Rs 199/kg in 2006 to Rs 361/kg in 2009 while the quantity of total tea exports remained stable. This resulted in much higher profit margins and revenue for the tea industry, as reflected by higher ratio of auction price and production cost (see Figure 13). The higher profit in the tea industry may have been passed onto estate workers in the form of increased earnings. As shown in Figure 14, between 2006 and 2009, only real wages of estate workers rose, while those of rural and urban workers stayed flat. Figure 13: Tea price, relative to 400 1.6 Exports (million kg) cost, rose substantially Price/Cost ratio 317 313 300 290 1.4 between 2006 and 2009 270 274 267 242 267 271 1.34 1.18 200 1.15 1.1 1.15 1.2 1.06 1.18 1.02 Source: Central Bank of 100 1 1.0 Sri Lanka: Economic and Social Statistics. The price to 0 0.8 2004 2005 2006 2007 2008 2009 2010 2011 2012 cost ratio is the ratio of the auction price to estimated Total tea exports Price/Cost ratio production cost. 19 II. Trends In poverTy, welfare, and InequalITy (conTd.) Figure 14: Real earnings for estate 150 wage workers outpaced that for rural and urban 138 Poverty rate(%) wage workers during 124 126 2006-2010 115 118 118 114 112 114 109 115 104 104 104 104 104 Source: IPS and World Bank 102 100 99 101 99 100 93 1.8 10.9 staff calculation based on 99 95 LFS. 90 2002 2003 2006 2007 2008 2009 2010 2011 2012 Urban Estate Rural Moreover, the real wages of estate workers continued to rise even more drastically between calendar years 2009 and 2010, promptly responding to the increase in minimum wages set by Wage Board (see Figure 32 and Figure 33). The reported poverty statistics for 2006/07, 2009/10, and 2012/13, on the other hand, are based on consumption data collected between July and Jun of two consecutive years. Thus, the poverty reduction during 2006/07 – 2009/10 might reflect the large wage increases for estate workers during the first half of 2010. The large reduction in estate poverty is corroborated by other welfare indicators, such as the multidimensional welfare index reported by UNDP (2012), which indicates a decline in multidimensional poverty from 21.1 percent to 11.4 percent between 2006 and 2009. Durable asset ownership in the estate sector also increased considerably, particularly in landline and mobile telephones, refrigerators, and televisions, signaling better living conditions and potential time savings, which could in turn further improve earnings capacity (Figure 15 below). Figure 15: Durable asset ownership increased in the estate 100% sector 75% 74%77% 65% 65% 50% Source: World Bank staff 41% 40% 40% calculation based on HIES 25% 15% 11% 11% 6% 8% 3% 4% 6% 1% 2% 4% 1% 1% 1% 0% Computer Refrigerator Motor cycle, Motor car, Telephone Telephone Television scooter van (domestic) (mobile) 2006/07 2009/10 2012/13 20 Poverty and Welfare in Sri Lanka However, more than 60 percent of estate residents belong to the poorest 40 percent of the national population, who lived on less than $2.75 a day in 2005 PPP term in 2012/13 (Figure 16). As documented in Section V below, the Estate sector also lags well behind the rural sector in a variety of non-monetary indicators. Figure 16: Expenditure in the estate 100 Cummulative population sector has increased, 80 but still lags behind the 60 national average 40 Source: World Bank staff 20 calculation based on HIES 0 5,000 10,000 15,000 20,000 Real expenditure per capita (2012/13 prices) Estate 2002 Estate 2012/13 National 2012/13 National poverty line 2012/13 Bottom 40% threshold 21 22 Poverty and Welfare in Sri Lanka III. dRIveRS of poveRTy ANd INequALITy ReduCTIoN Broad-Based growTh BenefITed The poor parTIcularly from 2006/07 To 2009/10. Sustained and pro-poor economic growth from 2002 to 2012/13 was at the root of Sri Lanka’s poverty reduction. Real GDP per capita grew by 5.6 percent per year between 2002 and 2012. More importantly, during this period, growth in real per capita consumption was higher for the bottom half of the distribution than for the top half (see Figure 17). The per capita consumption of the bottom 40 percent of households grew by 3.3 percent per year, a rate that within the South Asia region was only exceeded by Nepal and Bhutan. Figure 17: 5 Growth of household 4 expenditure per capita Percent change 3 was pro-poor during 2002-2012/13 2 1 Source: World Bank staff 0 calculation based on HIES -1 0 20 40 80 100 Percentile of per capita consumption (2006/07) Growth incidence 95% Confidence interval Mean growth rate Figure 18: Percent growth Income per capita grew 0% 1% 2% 3% 4% 5% 6% faster between 2006/07- Poorest Decile 2012/13 among the bottom 4 deciles 2nd Decile 3rd Decile Source: World Bank staff calculation based on HIES 4th Decile Top 60 percent Average 23 III. dRIveRS of poveRTy ANd INequALITy ReduCTIoN yet the extent to which growth has reduced poverty has fluctuated considerably. Figure 19 shows that the elasticity of poverty reduction with respect to GDP per capita growth was highest during the final stage of the civil conflict from 2006 to 2009/10. The elasticity during this period, in fact, was exceptionally high by international standards. In contrast, it was only moderate during the previous cease-fire period from 2003 and 2006, and much lower during the postwar period from 2009 to 2012/13. These fluctuations were likely influenced by the underlying drivers of growth, which differed considerably between 2002 and 2012/13. Figure 19: 9 Sri Lanka Growth had the largest 8 (2006-2009) impact on poverty during Poverty reduction-growth elasticity 7 2006-2009 6 Tajikistan Source: WDI, HIES 2002- 5 Vietnam Nicaragua Pakistan 2012/13. 4 Sri Lanka Nepal (2002-2012) Bolivia Bhutan 3 Peru Bangladesh Lao Colombia 2 Phillipines Indonesia Sri Lanka 1 Sri Lanka (2002-2006) (2009-2012) Georgia India 0 0% 5% 10% 15% 20% 25% Annualized percent decrease in PPP $1.25 headcount poverty rate Note: The reported countries are selected based on 3 criteria: having similar GDP per capita with Sri Lanka around 2002 (Bolivia, Georgia, Indonesia, Nicaragua, Philippines), having similar GDP growth during 2002-2012 (Tajikistan, Vietnam, Lao), countries in South Asia region (Bangladesh, Bhutan, Nepal, Pakistan, India), and countries with recent civil conflicts (Colombia, Peru). For each country we used the closest years to 2002 and 2012 for which poverty information was available. growth in the agricultural sector was particularly beneficial for poverty reduction during this period. Figure 20 illustrates this point by showing how much poverty fell among workers aged 15 and above in a particular sector in response to that sector’s growth. For example, poverty among agriculture workers decreased by one percentage point per one percent growth in per capita agriculture output between 2002 and 2006/07, as compared with only 0.2 percentage point per one percent growth in the service sector. 24 Poverty and Welfare in Sri Lanka Strong poverty reduction among agriculture workers 2002 – 2009/10 Headcount poverty rate Agricultural Sector Industrial Sector Service Sector 18 percentage points 13 percentage points 09 percentage points As a result, even modest growth in agriculture from 2002 to 2009/10 greatly reduced poverty for agriculture workers. Real growth in the Agricultural sector averaged less than 4 percent between 2002 and 2009/10. But during this time, the headcount poverty rate among agriculture workers declined by 18 percentage points, as compared to only 13 and 09 percentage points for workers in the industrial and service sectors (see Figure 21). As will be discussed in more detail in Section IV, this was partly due to the surge in the international prices of food and tea from 2006/07-2009/10 being passed on to agricultural workers. Despite robust growth, poverty reduction slowed from 2009/10 to 2012/13, and poverty among agriculture workers remained unchanged. Figure 20: 1.2 Growth-poverty semi-elasticity 1 Growth had stronger 1.0 poverty reduction impacts 0.8 0.6 in the agriculture sector 0.6 0.6 during 2002-2009/1 0.4 0.3 0.3 0.3 0.3 0.2 0.2 0.2 0.1 0.04 Source: World Bank staff 0.0 calculation based on HIES 2002-2006/07 2006/07- 2009/10- 2009/10 2012/13 Agriculture Industries Services Not employed Figure 21: 30% Poverty rate fell most 25% 26 Poverty Rate by Activity rapidly among agriculture 20 21 20% workers during 2002- 17 15% 14 2009/10 12 12 10% 8 8 8 7 7 6 5 5 5% Source: World Bank staff 2 0% calculation based on HIES Agriculture Industries Services Not employed 2002 2006/07 2009/10 2012/13 25 III. dRIveRS of poveRTy ANd INequALITy ReduCTIoN hIgheR LAboR eARNINgS weRe The mAIN CoNTRIbuToR To poveRTy ReduCTIoN Increased labor earnings were the key factor in reducing poverty. A simple framework for analyzing the decline in poverty is to decompose it into portions due to changes in labor income, different components of non-labor income, the employment rate of adults, and the dependency ratio. Between 2002 and 2012/13, increased labor income accounted for about 60 percent of the 16.5 percentage point reduction in headcount poverty. This is similar to decomposition results from Bangladesh and Nepal, but labor played a smaller role in Sri Lanka than in other countries in Latin America and Europe. “Non-labor income” accounted for another 27 percent of the reduction in Sri Lankan poverty during this time, with the decline in household size accounting for the remaining 13 percent. Smaller household size 13% Increased labour earnings 60% 27% Increased non-labour income *Increases in international remittances and higher returns to education were important secondary factors accounting for the reduction in poverty. *The decline in Samurdhi transfers, after adjusting for inflation, slowed the rate of poverty reduction by 10 percent. most of the increase in labor income was in turn due to increased returns to work both in and outside agriculture. A more sophisticated decomposition approach can be used to decompose the change in poverty into a portion explained by changes in characteristics on the one hand, and changes in the returns to those characteristics on the other hand. Box 3 describes this approach in greater details, and Figure 23 displays the results. Between 2002 and 2012/13, the role of increased returns to work stands out. Increased returns to agriculture accounted for about 31 percent of the decline, of which roughly two thirds were higher returns to self-employed farmers. The other main factor was an increase in the returns to paid non-farm work, which accounted for 28 percent of the poverty decline. Combined, the increases in the returns to work explain nearly 60 percent of the reported poverty reduction. 26 Poverty and Welfare in Sri Lanka These higher labor earnings have particularly benefited workers with less education since 2006. As illustrated in Figure 22, real wages of workers with General Certificate of Education (GCE) degrees and above either fell or remained relatively flat between 2006 and 2012. In contrast, the wages of workers with primary schooling or less rose robustly. Moreover, the wages of these workers have risen in line with the increase in minimum agriculture wages since 2010. Given that less educated workers tend to be low-income earners, their rising wages positively contributed to poverty reduction. Figure 22: 135 Workers with less 130 schooling benefited from Wage index (2002=100) 125 rising real wages 120 115 110 Source: IPS and World Bank 105 staff calculation based on LFS 100 95 90 2002 2003 2006 2007 2008 2009 2010 2011 2012 G.C.E (O/L) No Education Primary incomplete Primary G.C.E (A/L) & higher Figure 23: 35 27.9 25 Decomposition of the 20.4 MVR million 15 11.2 decline in poverty between 9.6 9.6 8.7 7.5 6.8 4.7 3.9 5 2.3 1.6 1.2 0.7 2002-2012/13 -0.7 -5 -2.2 -4.9 -9.6 -15 Source: World Bank staff -25 -20 Paid non-farm Paid farm Transfers in kind Self-employed farm Education self-employed non-farm ethnicity, region Dividends transfers Rents self-employed mix Occupation Disability and relief Other non labour income Samurdhi Savings rate Pensions Residuals Remittances calculation based on HIES International Age, gender, Domestic Returns Returns 27 III. dRIveRS of poveRTy ANd INequALITy ReduCTIoN deCompoSINg fACToRS CoRReLATed wITh poveRTy ReduCTIoN IN SRI LANKA Comprehensive household surveys such as the Household Income and Expenditure Survey enable a decomposition of changes in poverty according to how different factors changed and how returns to those factors changed. In order to quantify the contribution of each factor to poverty reduction, the decomposition is implemented in four stages. First, the determinants of education level, sectorial and activity choices are separately estimated for each year. Overall, the simulated distributions are close to the true distributions, indicating that the underlying specifications of these models can be used to simulate shifts in the composition and structure of the labor force. As a second step, labor income is separated into farm and non-farm income in order to estimate the earnings equation for each period, estimated separately for household heads and for other household members. Regressions are run for each of five groups of workers in each year: (1) salaried farm; (2) salaried non-farm; (3) self-employed farm; (4) self-employed non-farm; and (5) self-employed mix of farm and non-farm. Salaried employment is modeled at the individual level while self-employed income is at the household level. In the third step, coefficients from the previous step’s regressions are used to simulate counterfactual distributions by changing one variable at a time and by observing the effect of each change on the distribution. Fourth, the counterfactual poverty rates are compared to the observed poverty rates in order to quantify the impact of each element on poverty reduction. Given that changes in multiple factors could have interaction effects, the cumulative effect of these decompositions are computed. This effect is calculated following the methodology proposed by Bourguignon et al. (2008), which entails: (1) sequentially calculating the effects of changes in characteristics of the population like age, gender, region and area, followed by changes in sector, education and activity structure of the population; and (2) using the results from (1) to sequentially calculate changes in farm and non-farm earnings due to changes in the returns to these characteristics; followed by changes in non-labor incomes, and changes in the consumption-to income ratio. 28 Poverty and Welfare in Sri Lanka 29 Iv. poTeNTIAL CAuSeS of INCReASed LAboR INCome The observed increases in labor earnings could have been driven by several complementary factors. Higher labor demand, higher labor productivity, better public infrastructure that improved production efficiency, increasing urbanization, and movement out of agriculture all likely played a role. Although it is difficult to untangle and quantify the impacts of each factor, we focus on four major factors that have potentially contributed to the increases in labor earnings that reduced poverty. The gRAduAL STRuCTuRAL TRANSfoRmATIoN Sri Lanka’s economy has become a more industrialized economy since 2002. With the exception of 2008, the annual growth rate of the industrial sector has consistently and increasingly outpaced that of both the agricultural and service sectors (refer to Figure 25). As a result, the industrial sector is becoming a larger contributor to total GDP (Figure 24). At a more disaggregate level, the sectors contributing most to economic growth since 2009 include the manufacturing and construction industries, as well as transportation and communication services (see Figure 26). Sectoral contribution to GDP growth Figure 24: 100% The industrial sector is growing as a share of 80% national GDP 60% Source: DCS, World Bank 40% staff calculation 20% 0% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Agriculture Industries Services Figure 25: The agriculture sector 12% has consistently grown 10% Annual growth rate at lower rates than the 8% industrial and service 6% sectors 4% 2% Source: DCS, World Bank staff calculation 0% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014* Agriculture Industry Services 30 Poverty and Welfare in Sri Lanka Figure 26: 100% Contribution to GDP growth GDP growth has been 80% fueled by growth in (USD million) 60% the manufacturing, 40% construction, commerce, transportation and 20% communication sectors 0% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Agriculture, Forestry and Fishing Manufacturing Source: DCS, World Bank Construction Other industries staff calculation Wholesale and Retail Trade Transport and Communication Banking, Insurance and Real Estate Other services In contrast, the agriculture sector accounts for a declining share of both gDP and employment. As shown in Figure 27, in 2014 the sector accounted for 28.5 percent of total employed workers in 2014, down from 34.4 percent in 2002. Its share of GDP also dropped from 14.3 percent to 10.1 percent during the same period. Figure 27: 40% 34.4% The agriculture sector 33.8% 33.5% 35% 32.0% 31.2% 32.6% 32.4% 32.5% 32.8% 30.7% has been declining, albeit 30% slowly 29.8% 25% 28.5% Percent 20% Source: DCS, World Bank 14.3% 15% 13.7% 13% 12.5% 12.3% 12.1% 11.9% staff calculation 11.9% 12% 11.2% 11.1% 10.8% 10.1% 10% 5% 0% 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Agriculture’s share of GDP Agriculture’s share of employment (excl. Northern & Eastern provinces) Agriculture’s share of employment (National-wide) This structural transformation coincided with rising employment and wages in several sectors outside agriculture. Except for the period from 2006 to 2009, real wages in manufacturing, construction, commerce, and to a lesser extent, transport and communication, all increased amidst growing employment shares (Figure 30). This suggests that firms are willing to hire more workers, and pay them more, as the value of their contributions have increased. This is also reflected in increased productivity, as measured by value added per worker outside agriculture. Figure 28 shows that productivity growth in industry has outpaced that in agriculture since 2006, while productivity in the service sector grew rapidly since 2002. 31 Iv. poTeNTIAL CAuSeS of INCReASed LAboR INCome The shift out of agriculture has resumed since 2009/10, though at a slower pace for the poor. Figure 29 shows a fall in the share of workers in agriculture from 2009/10 to 2012/13, following a slight increase from 2006/07 to 2009/10. At least part of this increase was due to the surging international prices of food and tea during 2006-2008, which made the agriculture sector temporarily more profitable and attracted low-income workers back into the sector. Interventions that increase the transition of poor workers away from farm works, especially for the poor, will help further improve living standards by shifting workers into more productive and better paid jobs. Figure 28: Real value-added per worker index 180 Real value added per worker in industrial sector 160 has outpaced that in agriculture since 2006 140 120 WDI, World Bank staff calculation based on LFS 100 2002 2003 2006 2007 2008 2009 2010 2011 2012 Agriculture Industry Services Figure 29: Agriculture’s share of employment 50% More workers who 41.8% 41.6% 40% 37.9% belonged to the bottom 40 percent worked in the 28.8% 29.5% 30% 26.8% agriculture sector during 2006/07-2012/13 20% 10% Source: Word Bank staff calculation based on HIES 0% All workers Workers in bottom 40 percent 2006/07 2009/10 2012/13 32 Poverty and Welfare in Sri Lanka Figure 30: Changes in real wages Increased labor demand was reflected in higher 14% 12% wages and employment 10% in many sectors between 8% 6% 2009 and 2012 Percent 4% 2% 0% Source: IPS and World Bank -2% staff calculations based on -4% LFS Others Public utilities Manufacturing Mining Construction Agriculture Commerce Transports communications Financial & business- oriented services Community & family- oriented services Transports & 2002-2006 2006-2009 2009-2012 Changes in employment share 0.8 0.6 Percentage points 0.4 0.2 0.0 -0.2 -0.4 -0.6 Others Public utilities Manufacturing Mining Construction Agriculture Commerce Transports Transports & communications Financial & business- oriented services Community & family- oriented services 2002-2006 2006-2009 2009-2012 Notes: Employment is change in percentage point change in share of workers employed in each sector. Real wage increase is the increase in real wages, including only wage workers. Wages are deflated using the Colombo Consumer Price Index 33 Iv. poTeNTIAL CAuSeS of INCReASed LAboR INCome uRbANIzATIoN ANd AggLomeRATIoN Sri Lanka has experienced a dynamic process of urbanization leading to greater spatial connectivity. Both night-time lights (Figure 9) and data from human settlement layers (Figure 31) show increased density in the corridors from Colombo to Kandy and Galle/Matara corridors, as the area has become a multi-city urban agglomeration. The night-time light image also illustrates that smaller single-city agglomerations have emerged around Trincomalee, Batticaloa- Akkaraipattu, and Jaffna, although they are still poorly connected with the prime agglomeration of Kandy-Colombo-Galle. Figure 31: Urbanization along the Kandy-Colombo-Galle corridor Source: World Bank SAR urban team’s based on compilation from Global Human Settlement Layers data sets Peri-urban areas appear to be growing most quickly. Growth in night time lights between 2001 and 2011 was concentrated in a few DS divisions. The top 10 percent of the DS divisions accounted for 30 percent of the increase in night-time lights, while the top 20 percent contributed half of the increase. In general, these high-growth DS divisions were moderately built up peri-urban areas in 2014, where new settlement and economic activity created brighter and denser night-time light. The colombo-Kandy-galle agglomeration is at the forefront of growth and poverty reduction. The area around this agglomeration, including a 60 km buffer, contains a third of the county’s land mass, but houses 70 percent of the population and over 55 percent of the poor. This area includes all or part of 6 districts: Colombo, Gampaha, Kalutara, Kandy, Kegalle, Kurunegala, and Matara. Combined, these districts account for 83 percent of national value added, and 87 percent of manufacturing value added. 34 Poverty and Welfare in Sri Lanka rIsIng InTernaTIonal food and Tea prIces durIng 2006-2009 A surge in the international prices of rice and tea from 2006 to 2009 helped boost real wages in the agricultural sector. The price of rice rose by about 30 percent, and the price of tea – Sri Lanka’s largest export by value – increased by more than 50 percent, from $2.1 to $3.3 per kilogram (Table 1). Since there was little change in yields and production costs during this period, rice and tea production became more profitable. This led to slight increases in both the employment share and real wages in agriculture, which were balanced by declines in the employment share and wages in manufacturing and commerce sectors. The agriculture sector employs a disproportional number of poor workers, and as discussed above, this external price shock had a positive impact on poverty reduction. Table 1: rice and tea prices spiked between 2006 and 2009 2002 2006 2009 2012 Rice Market Price Guaranteed Price (2013 Rp) 32.1 28.9 37.7 30.9 Yield 3.9 4.1 4.3 4.4 Percent of Land n/a n/a n/a 63.8 Tea Colombo auctions, $/kg 2.1 2.1 3.3 2.8 Yield (Kg per Ha) n/a 1.4 1.3 1.5 Cost of Production (2013 Rp) 307.1 331.2 350.8 418.0 Percent of land n/a n/a n/a 13.3 Exports (‘000 MT) 292 327 290 320 Exports (bn 2012 Rs) 159 164 166 180.4 Exports (% of GDP) 3.9% 3.1% 2.8% 2.4% Source: Sri Lanka Central Bank, Economic and Social Statistics of Sri Lanka Higher rice and tea prices were eventually followed by sharp increases in the agriculture minimum wages. The minimum wages for agricultural workers set by the Wage Board rose dramatically in 2010 and again in 2012, while minimum wages in other sectors remained flat through the last decade (Figure 32). Real agricultural wages, unlike those for industrial and service sector wages, tracked these minimum wage increases almost exactly (Figure 33 below). 35 Iv. poTeNTIAL CAuSeS of INCReASed LAboR INCome Figure 32: 150 Real wage rate index (1978 base) 150 Real wage index (2002 = 100) Minimum wages for 120 agriculture workers 130 increased sharply 90 110 between 2009 and 2012 60 90 30 Figure 33: 0 70 Real agriculture wages 2004 2006 2008 2010 2012 2002 2004 2006 2008 2010 2012 tracked minimum Agriculture Services Industry and Commerce Agriculture Services Industry and Commerce agriculture wages closely Source: Central Bank of Sri Lanka, Economic and Social Statistics of Sri Lanka 2014 The increase in the agricultural minimum wage in 2010 is consistent with the reduction in poverty among agricultural workers between 2006/07 and 2009/10. While wage data were collected for each calendar year, the consumption data underlying the reported poverty statistics for 2006/07, 2009/10, and 2012/13 were collected between July and June of two consecutive years. Thus, the strong poverty reduction observed between 2006/07 and 2009/10 could partly capture the impacts of the higher agriculture wages that occurred during the first half of 2010. Agriculture households benefited during this period regardless of whether they owned land or worked for themselves. Table 2 shows the increase in average per capita consumption, according to whether household members work in agriculture and whether the household owns agricultural land. If increased minimum wages alone were the primary driving factor behind the increase in agricultural returns, one would expect to see a larger increase in consumption among households with agricultural wage employees than self-employed farm households. On the other hand, if the increase in food prices were not passed on to farm laborers, households that owned agricultural land should have seen larger increases in consumption. Between 2006/07 and 2009/10, the increase in consumption was markedly higher among households with members working in agriculture, either as employees, self-employed, or employers, than among households without members working in the sector. But farmers that owned their own land, if anything, saw slightly lower gains between 2006/07 and 2012/13 than landless farmers and households with a member working in a salaried agricultural wage job. This suggests that agricultural land and labor markets, perhaps assisted by the increase in the agricultural minimum wage, successfully passed on the gains from rising world food prices to agricultural laborers and landless farmers. 36 Poverty and Welfare in Sri Lanka Table 2: The large increase in agriculture minimum wages and international food prices benefited both farmers and agricultural wage workers Household owns Household members Annualized increase of median consumption land for cultivation working in agriculture per capita 2006/07-2009/10 2009/10-2012/13 No land None 3.1% 3.8% Employees only 7.2% 2.8% Own account and employers 6.4% 2.8% Agricultural Land None 2.6% 4.5% Employees only 4.9% 1.9% Own account and employers 4.3% 1.4% All 3.8% 3.2% Source: World Bank staff calculation based on HIES gRowTh IN domeSTIC CoNSumpTIoN GDP growth has been largely led by domestic private consumption. Private consumption made up 67 percent of GDP in 2013, having expanded by 6 percent annually during 2002-2013 (Figure 34). The second largest component of Sri Lanka’s GDP has been private investment, which accounted for 23 percent of 2013 GDP after an average annual growth of 9 percent during the same period. Net exports, government consumption, and change in stocks all made minor contributions to total GDP. Public infrastructure spending, while accounting for only 6 percent of GDP in 2013, expanded rapidly by 18 percent per year between 2002 and 2013. Figure 34: A. Composition of Aggregate Demand B. Growth in Demand Components (percent of GDP) (Index, 2002 = 100) Domestic aggregate demand has been 150 650 largely led by private 550 consumption 100 450 Source: DCS, World Bank 50 350 staff calculatio 250 0 100 0 -50 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Private consumption Government consumption Private consumption Government consumption Government investment Private investment Government investment Private investment Change in stocks Exports Change in stocks Exports Imports Imports Although evidence is lacking, expanded infrastructure investment and access to electricity may have boosted productivity by improving connectivity and access to markets. While still accounting for only a small proportion of aggregate demand, the dramatic increase in public 37 Iv. poTeNTIAL CAuSeS of INCReASed LAboR INCome infrastructure investment could have spillover effects in terms of improving production efficiency, reducing transportation costs, connecting remote areas, and ensuring a stable power supply. Moreover, studies from other countries demonstrate that access to electricity led to increased productivity among small-scale informal workers in South Africa and broad-based improvement in labor productivity in Brazil.5 The higher coverage of electricity and increased ownership of such assets as motorcycles and mobile phones that was documented in Section II, therefore, might also have increased labor productivity in Sri Lanka. Among the four factors discussed above, growth in domestic consumption and rises in international food prices are not likely to be sustainable in the years to come. Fiscal constraints preclude further large-scale infrastructure investments, and agriculture workers, especially low- income earners, may be vulnerable to international prices drops. In fact, tea price fell in 2014 in response to declining demand from Middle East countries. While it is unclear, given the available data, how this has affected agricultural and estate workers, policies to speed up the structural transformation can help reduce Sri Lanka’s vulnerability to adverse commodity price shocks from external markets. 5 See Dinkelman (2011) and Lipscomb, Mobarak, and Barham (2013) 38 Poverty and Welfare in Sri Lanka 39 v. The CuRReNT STATe of poveRTy ANd INequALITy deSpITe The eNCouRAgINg ReduCTIoN IN poveRTy, LIvINg STANdARdS RemAIN modeST foR moST of The CouNTRy. consumption in 2012/13 remained low for a large share of the population. Figure 35 shows, on the vertical axis, the share of Sri Lankans whose household per capita consumption falls below the expenditure level depicted on the horizontal axis. The intersection of the curve with the green line indicates that roughly a third of Sri Lankans live below the international poverty line of $2.50 per day in 2005 PP terms. In fact, the expenditure threshold for the poorest 40 percent of the population, in terms of the per capita consumption, is merely USD 2.75 per day (2005 PPP); and just over 60 percent of the population lives on less than USD 4.00 per day. In light of these low consumption levels, it is useful to identify the characteristics of both the poor and the near-poor. We define the near-poor as individuals living in households with per capita consumption above the national poverty line but below the 40th percentile threshold. The poor and the near-poor are in turn compared with the top 60 percent (henceforth, the non-poor). In particular, the following sub-sections will profile residential location, demographic characteristics, and asset ownership of poor and near-poor Sri Lankans. Special attention is given to those living in the estate sector, which has traditionally been poorer than the urban and rural sectors, as well as the conflict-affected districts in Northern and Eastern provinces. Demographic characteristics and important living standard indicators of different expenditure groups are displayed in Table 6 and Table 7 (Refer Appendix 5). Figure 35: 20 40 60 80 100 Cumulative population (%) Consumption remains low for a large share of the population Source: World Bank staff calculation based on HIES 0 0 10,000 20,000 30,000 40,000 50,000 Real expenditure per capita per month (Rs) Official poverty line Bottom 40% threshold $ 2.50 (2005 PPP) $ 4.00 (2005 PPP) geogRAphIC poCKeTS of poveRTy RemAIN Despite the low national poverty rate, geographical pockets of poverty remain a serious concern. Figure 36 shows estimates of poverty at the DS level, based on an updated poverty map that combines the 2011 Census of Population and Housing with the 2012/13 Household Income and Expenditure Survey. The map shows three main pockets of poverty. The first is the former conflict districts in Northern Province, Mullaitivu (28.8 percent), Mannar (20.1 percent), and, to a lesser extent, Kilinochchi district (12.7 percent). The second is Batticaloa district (19.4 percent) in Eastern province, and the last one is Monaragala district (20.8 percent) in Uva province. 40 Poverty and Welfare in Sri Lanka Figure 36: No. of poor people Jaffna Jaffna Poverty map 2012/13 6,500 - 18,500 4,500 - 6,500 3,500 - 4,500 Kilinochchi Kilinochchi 2,500 - 3,500 500 - 2,500 0 - 500 Figure 37: Mullaitivu Mullaitivu Number of poor people by Mannar Mannar DS Division Vavuniya Vavuniya Trincomalee Trincomalee Source: Sri Lanka poverty Anuradhapura Anuradhapura map 2012/13, estimated from a combination of the Polonnaruwa Polonnaruwa 2012/13 HIES and the 2011 Census of Population and Puttalam Batticaloa Puttalam Batticaloa Housing Kurunegala Matale Kurunegala Matale Kandy Ampara Ampara Kandy Gampaha Kegalle Kegalle Gampaha Nuwara EliyaBadulla Nuwara EliyaBadulla Colombo Colombo Moneragala Moneragala Kalutara Ratnapura Kalutara Ratnapura Poverty rate (%) 35.0 - 60.0 25.0 - 35.0 Galle Hambantota 15.0 - 25.0 Galle Hambantota 5.0 - 15.0 Matara Matara 0.0 - 5.0 The near-poor are prevalent in both poor and affluent districts. In general, areas with higher poverty rates tend to have a larger portion of the bottom 40 percent as well, especially in Monaragala, Mannar, and Mullativu. But there are also relatively affluent districts such as Ampara and Nuwara Eliya, where poverty rates are slightly over 8 percent yet over 45 percent of the population belongs to the national’s bottom two quintiles (see Figure 38 below). 41 v. The CuRReNT STATe of poveRTy ANd INequALITy Figure 38: 100% The near-poor are 80% prevalent in both poor and affluent districts Percent of population 60% Source: World Bank staff 40% calculation based on HIE 20% 0% Kalutara Kandy Kurunegala Hambantota Polonnaruwa Ratnapura Colombo Gampaha Vavuniya Puttalam Ampara Nuwara Eliya Kegalle Sri Lanka (All) Matara Anuradhapura Matale Jaffna Trincomalee Galle Badulla Kilinochchi Batiacaloa Mannar Monaragala Mulativu Poor Near-poor Non-poor High poverty incidence does not always coincide with a large concentration of poor people. Most of the poor live in populous and relatively affluent districts such as Kurunegala, Ratnapura, and Kandy. Kurunegala, for instance, is home to 7.7 percent of the country’s poor people even though only 6.5 percent of its population lives under the official poverty line. In contrast, Mullaitivu and Mannar, where estimated poverty rates are very high (28.8 and 20.1 percent, respectively), account for only 3.4 percent of poor people nationwide due to their small populations. The same is true at the DS division level, where large numbers of poor exist in Gampaha, Nuwara Eliya, and Ratnapura districts. The poor and near-poor Tend To Be rural, young, and dISCoNNeCTed fRom pRoduCTIve eARNINgS oppoRTuNITIeS. The poor are overwhelmingly rural. The rural sector accounted for over three quarters of the country’s population (see Figure 39) and over 85 percent of poor Sri Lankans nationwide in 2012/13.6 As a result, much of the poverty reduction observed during the period of 2002-2012/13 was driven by the reduction in rural poverty. Specifically, the decrease in rural poverty headcount ratio from 24.7 percent to 6.8 percent during this period is roughly equivalent to a decrease in the number of people living under the poverty line from 19.8 percent to 5.3 percent of the total population.7 6 The official definition of urban and rural is based on areas administered by urban and municipal councils, which underestimates the extent of urban areas. 7 Estimated based on sectoral population shares from the 2001 and 2011 Census of Housing and Population, and the estimated poverty rates reported in Figure 1. In the 2001 census, which does not cover 7 districts from Northern and Eastern provinces, the rural sector accounted for 80 percent of the population. 42 Poverty and Welfare in Sri Lanka Figure 39: 4.4% Most Sri Lankans live in rural areas 18.3% Source: World Bank staff calculation based on CPH 77.3% 2011 Estate Urban Rural The poor and near-poor are disproportionately young. This is partly because poorer households tend to have larger household sizes. As can be seen in Table 6 (Refer Appendix 5), 45 percent of the bottom 40 percent is below 25 years old, as compared to only 37.8 percent of the top 60 percent. The poor and near-poor have lower education attainment. Data from the 2012/13 HIES reveal that about 55.7 percent of adults (aged 21 and above) in the bottom 40 percent did not finish secondary education, as compared to only 46.2 percent of the non-poor. Since most of those adults had stopped attending school when they were interviewed, their reported education levels are a determinant, rather than a result, of low earnings. The poor and near-poor are less likely to be working (see Table 6). More critically, unemployment is exceptionally high among youths in the bottom 40 percent. Young Sri Lankans (15-24 years old) in general are much more disconnected from productive activities than those above 24 years old. Furthermore, the situation is even more acute among near-poor youths.8 If this pool of abundant, young, and cheap labor is not fully utilized, the combination of high youth unemployment and a young poor population has the potential to threaten social stability. Among those poor and near-poor that are employed, a large proportion is engaged in agriculture. In particular, about one in every two poor workers works in the agricultural sector, whereas one in every two non-poor workers in the service sector (Table 6). The agriculture sector also employs a majority of near-poor workers (42.9 percent). This disproportional concentration of low-income Sri Lankans in farm works, which typically have lower value added and wages than the other sectors, reinforces the case to accelerate the structural transformation and shift workers into more productive sectors. 8 This is based on analysis of the HIES data, which uses a different definition of unemployment than the LFS data. According to this definition, youth unemployment was 34.2% for the near poor in 2012/13, as opposed to 26.6% for the near-poor and 26.8% for the poor. 43 v. The CuRReNT STATe of poveRTy ANd INequALITy gender inequality is more apparent in labor outcomes than consumption or educational inequality. Gender inequality in Sri Lanka is in fact quite limited, as reflected in not only per capita consumption but also education outcomes. Sri Lankan girls are as likely as, if not even more likely than, boys to attend and complete primary and secondary school. However, women still face substantially worse labor market outcomes, with much lower labor participation rate and higher unemployment rate than men (see Figure 41). Figure 40: 40% Youth unemployment 32% 27.9% remains high 26.9% Unemployment Rate 24% 21.6% 21.2% 21.0% 18.5% 19.5% Source: IPS and World Bank 17.0% 16.9% staff calculation from LFS 16% 8% 3.7% 3.4% 3.4% 3.1% 2.8% 3.2% 2.7% 2.0% 2.0% 0% 2002 2003 2006 2007 2008 2009 2010 2011 2012 15-24 years old Above 24 years old moderate disparities between Tamils and Sinhalese persist. The poor, like the overall population, are largely Buddhist and Sinhalese. But poverty rates for Sri Lankan Tamils are almost twice that of Sinhalese overall, after including Northern and Eastern provinces (see Figure 42). These disparities remain even though nearly a quarter of Sri Lankan Tamils – more than the national average – live in urban areas. Figure 41: 90 18 Women continue to lag Labor force participation rate (%) 80 16 behind in labor market Unemployment rate (%) 70 14 outcomes 60 12 50 10 Source: IPS and World Bank 40 8 staff calculation based on LFS 30 6 20 4 10 2 0 0 2002 2003 2006 2007 2008 2009 2010 2011 2012 Male unemployment rate Female unemployment rate Male participation rate Female participation rate 44 Poverty and Welfare in Sri Lanka Ethnic disparities in poverty can be explained by differences in a few key characteristics. Sri Lankan Tamils tend to live in poorer districts, and after accounting for these geographic differences, the disparity in poverty rates falls from 6.1 to 4.7 percentage points. Further accounting for three key characteristics of the household head – age, education, and gender – reduces the poverty penalty faced by Sri Lankan Tamils to 2.8 percentage points. The penalty falls to merely 1.2 percentage points when differences in household size are taken into account. In other words, while Sri Lankan Tamils are more likely to be poor than Sinhalese, the vast majority of this penalty can be explained by differences in geographic location, the age, gender, and education of the household head, and average household size. Figure 42: 15% Poverty rates are higher Headcount poverty rate 12.0% among Tamils 10% 9.3% 9.4% 9.1% 5.9% 5.9% 6.0% Source: World Bank staff 5.0% 5% calculation based on HIES 0% Sinhala Sri Lanka Hindu Sri Lanka Tamil Tamil Moor 2012 2012 (including Northern and Eastern Provinces) Figure 43: 100 31% 39% 58% The poor and near-poor Share of consumption 80 devote a similar share of 60 consumption to food. 40 69% 61% 42% Source: World Bank staff 20 calculation based on HIES 0 Poor Near poor Upper 60% Food Non-food 45 v. The CuRReNT STATe of poveRTy ANd INequALITy The lIvIng sTandards of The near-poor are closer To Those of The pooR ThAN of The Top 60 peRCeNT. The near-poor fare only mildly better than the poor when it comes to consumption patterns and asset ownership. For example, the near-poor spend an average of 61 percent of their budget on food, which is only eight percentage points less than the poor do, as opposed to 42 percent for non-poor households (See Figure 43). Likewise, the near-poor are also far behind the non-poor in terms of ownership of many durable assets (see Figure 44). The difference between the poor and near-poor is slightly more pronounced for housing conditions. About 54.5 percent of the poor lives in a housing unit with permanent walls, floor, and roof. The figure is considerably higher for the near-poor (69.4 percent) and non-poor (84.3 percent). Figure 44: 100% 91% 95% 90% 86% Asset ownership among 79% 80% the near-poor tends to 67% 70% 65% be low 60% 55% 62% Percent 50% 48% 41% Source: World Bank staff 40% 32% 29% 27% 27% calculation based on HIES 30% 23% 25% 20% 15% 14% 10% 12% 10% 7% 8% 5% 2% 0% 1% 1% 0% Computer Refrigerator Motor Motor car, Sewing Washing Telephone Telephone Television cycle, van machine machine (domestic) (mobile) scooter Poor Near poor Upper 60% Figure 45: 50 Average travel time 40 to public facilities for the bottom 40 percent 30 hardly improved during 2006/07-2012/13 Minutes 20 10 Source: World Bank staff calculation based on HIES 0 Maternity Maternity Maternity Pre-school Pre-school Pre-school Primary Secondary/ Collegiate Hospital Dispensary Clinic Primary Secondary/ Collegiate Hospital Dispensary Clinic Primary Secondary/ Collegiate Hospital Dispensary Clinic Poor Near Poor Not Poor 2006/07 2009/10 2012/13 46 Poverty and Welfare in Sri Lanka Figure 46: 40% 34% A significant proportion of the near-poor borrow 30% 25% 25% from retail shops rather Percent 20% 22% than banks 21% 10% 11% Source: World Bank staff calculation based on HIES 0% Poor Near poor Upper 60% Borrowed from banks Borrowed from retail shops Both the poor and near-poor reported little improvement in their access to basic services since 2002. Sri Lanka has made considerable investments in road infrastructure in recent years, and as noted earlier, motorcycles have become more prevalent among the poor. Yet between 2006/07 and 2012/13, the bottom 40 percent of Sri Lankans, as did the top 60 percent, reported almost no improvement in travel times to public education and health care facilities, except for access to clinics (see Figure 45). Moreover, it takes the poorest 40 percent longer to reach these facilities than the non-poor. This suggests that infrastructure investment has been concentrated on main roads and highways, rather than the smaller feeder roads that tend to help low-income households reach public facilities. The poor and near-poor have limited access to formal credit markets. The richest 60 percent are far more likely to borrow from banks than retail shops, while the near-poor utilize retails shops nearly as much as the poor do (see Figure 46). Lacking access to formal credits not only leaves poor and near-poor households little room to smooth their consumption in the face of adverse income shocks, but also restricts their ability to invest in business and/or capital to improve their earnings. The children of the poor and near-poor are less likely to pursue secondary and collegiate education. Figure 47 and Figure 48 show attendance and completion rates among children at ages that correspond to each education level. There is little difference for primary and junior secondary schooling. Poor children are actually more likely to attend primary school than more affluent children, even though they are slightly less likely to complete their primary degree. But attendance 47 v. The CuRReNT STATe of poveRTy ANd INequALITy and completion rates decline significantly for senior secondary and collegiate education across all expenditure groups. Moreover, at these higher levels of education, the gaps between poor, near- poor, and non-poor students are substantial. For example, only 18 percent of poor children aged 17 to 18 attended college, as opposed to 60 percent of their non-poor counterparts. Figure 47: 100% Poor children are less likely to attend secondary 80% and collegiate schools Attendance rate (percent) 60% Source: World Bank staff 40% calculation based on HIES 20% 0% Primary Jr. Secondary Sr. Secondary Collegiate Primary Jr. Secondary Sr. Secondary Collegiate Primary Jr. Secondary Sr. Secondary Collegiate Poor Near Poor Not Poor 2006/07 2009/10 2012/13 Figure 48: 100 Poor children are less 80 likely to complete Completion rate (percent) 60 secondary and collegiate schools 40 20 Source: World Bank staff 0 calculation based on HIES Primary (12-13) Sr. Secondary (17-18) Primary Completion on time (12-13) Secondary Completion (12-13) Primary Completion on time (12-13) Secondary Completion (12-13) Poor Near Poor Non-Poor 2006/07 2009/10 2012/13 The gap between rich and poor in attending secondary and collegiate education is growing. The share of 17 to 18 year olds attending college increased by 13 percentage points for near-poor students, 9 percentage points for non-poor students, but only by one percentage point for poor students. The growing disparity between the poor and near-poor will have long-term consequences and make it more difficult for poor children to escape the cycle of poverty. 48 Poverty and Welfare in Sri Lanka poveRTy IN The eSTATe SeCToR Poverty in the estate sector reflects the unique characteristics of its residents. The poor and near-poor in the Estate sector are mainly young, Hindu Tamils, working in tea estates in the agriculture sector, and living in free housing units provided by estate owners. The dominance of plantation estates as the major employer in the estate sector explains the high employment rates observed in the Estate sector, whether poor, near-poor, and non-poor. Yet due to low wages, remoteness, and lack of government-sponsored basic services, these higher employment rates do not translate into better living standards for poor Estate residents. As displayed in Table 6 (refer Appendix 5), there are striking gaps between the estate poor and rural poor in housing conditions. The dependence of estate workers on their employers for accommodation and basic services means that their housing conditions, sanitation, and water quality can easily be compromised. For example, only 5.6 percent of the estate poor owned a house in 2012/13, and only 11.3 percent of the Estate near-poor do. In contrast, 90.4 percent of the rural poor owned a house. In 2012/13, slightly more than a quarter of the estate poor and also non-poor lived in a house with a permanent roof, while three quarters of the rural poor did. Moreover, only 55.1 percent of the estate poor and 47.9 percent of the estate near-poor reported to have access to clean water, substantially lower than 84.3 percent of the rural poor. HOuSIng cOnDITIOnS 2012/13 Percentage of estate poor that own a house 5.6% Percentage of estate near poor that own a house 11.3% Percentage of national poor that own a house 83.7% Percentage of rural poor that own a house 90.4% AccESS TO cLEAn WATEr Percentage of estate poor with access to clean water 55.1% Percentage of estate near poor with access to clean water 49.1% Percentage of rural poor with access to clean water 84.3% Percentage of national poor with access to lean water 82.8% The poor in the Estate sector have low education attainment. Only half of the poor adults living in estate areas had completed primary school, and only 2.3 percent had completed secondary school in 2012/13. This is significantly less than the corresponding figures in the rural sector, which are 65 percent and 8.8 percent, respectively. Low levels of education make it more difficult for workers in the rural and especially the estate sector to shift to more productive sectors that reward higher levels of education. 49 v. The CuRReNT STATe of poveRTy ANd INequALITy Estate households devote a lower share of consumption to protein than rural and urban households. Consumption of protein-rich foods improved only slightly from 3.5 to 4.9 percent of the food budget. Worse still, the share of expenditures devoted to alcohol and tobacco rose from 7.4 percent to 8.6 percent. Taken together, these indicators suggest that estate households have much lower living standards than their rural counterparts, despite the small difference in poverty rates. 100 100 Share of food expenditures Share of food expenditures Figure 49: 80 80 Estate households spent 60 60 less on high-protein food, 40 40 but more on alcohol and 20 20 tobacco than rural and urban households 0 0 URBAN (2002) URBAN (2012/13) RURAL (2002) RURAL (2012/13) ESTATE (2002) ESTATE (2012/13) URBAN URBAN (2002) (2012/13) RURAL (2002) RURAL (2012/13) ESTATE (2002) ESTATE (2012/13) Figure 50: Cereals and Starch Items Vegetables and Fruits Meat Eggs Estate households Protien Rich Food Milk Food Fish Pulses consume less protein Fats, Sugar, Confectionery Liquor and Tobacco and Others from meat and fish than rural and urban households INequALITy IS RISINg ANd ACCeSS To SeRvICeS RemAINS INequITAbLe Source: World Bank staff calculation based on HIES Inequality in both consumption and income has increased since 2009. Inequality in Sri Lanka has traditionally been moderate by the standards of the region and other countries at similar levels of development. Between 2002 and 2009, inequality in both consumption and income declined, as pro-poor growth led to sizeable increases in consumption for the bottom 40 percent of the consumption distribution.9 But between 2009 and 2012/13, inequality in both consumption and income increased. The increase in consumption inequality was particularly sharp between 2009/10 and 2012/13, erasing the decline in inequality observed since 2002. Sri Lanka is the only country in the region that experienced this rapid increase in consumption inequality in the last decade (Figure 52). Furthermore, household surveys often underestimate inequality, because the truly wealthy are a tiny share of the population and are less likely to participate in household surveys. While higher inequality may in part be a byproduct of growth, it can strengthen the case for more progressive fiscal policy. 9 The income inequality figures are based on a sample that omits the top 0.1 percent of households 50 Poverty and Welfare in Sri Lanka Inequality 2002-2009 2009-2012/13 Inequality in consumption Inequality in consumption Inequality in income Inequality in income Figure 51: 46.4 44.4 50 43.2 43.1 Inequality has increased 40.2 39.9 38.9 Pakistan (2007/08) 36.3 since 2009/10 40 Nepal (2003/10) Gini Coefficient 30 India (2004/09) Figure 52: 20 Bhutan (2003/10) Sri Lanka is the only 10 Bangladesh (2005/10) country in the region Sri Lanka (2009/10) 0 to experience a recent Consumption Income -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Inequality Inequality increase in inequality Annualized Change in Gini Coefficient 2002 2009/10 2006/07 2012/13 Source: World Bank staff calculation based on HIES most of the increase in inequality between 2009/10 and 2012/13 was driven by changes within population groups. The change in consumption inequality during this period can be decomposed into changes in inequality between and within a variety of groups. In this case, groups are defined based on residential location, education and ethnicity of the individual, and key characteristics of the household head. Among the characteristics considered, changes in inequality across sector of residence (urban/rural/estate) and employment sector of the household head (agriculture/ industry/services/not working) contributed most to the overall increase in consumption inequality. This means that changes in the average consumption of households in different residential sectors, and with heads working in different employment sectors, favored better-off households. In contrast, changes in average consumption across education levels, ethnicities, genders of household head, and between post-conflict and non-conflict areas contributed little to the overall increase in inequality. 51 v. The CuRReNT STATe of poveRTy ANd INequALITy Table 3: Decomposition of the increase in inequality during 2009/10-2012/13 Within Between inequality inequality Education 99% 1% Ethnicity 102% -2% Conflict-affected areas 100% 0% Residential sector 93% 7% Education of Head 96% 4% Gender of head 100% 0% Employment sector of head 92% 8% Source: World Bank staff calculation based on HIES consumption and income inequality is associated with unequal access to health and education services. Since 2006/07, the HIES has asked respondents to report travel times to public facilities. Figure 53 shows, in percentage terms, how much longer poor and near-poor households reported it took to travel to public facilities than non-poor households. Poor households in 2012/13 reported that it took about 20 to 30 percent more time on average to reach education and health facilities than non-poor households. The difference is largest for public dispensaries and secondary/ collegiate schools. Figure 53: 100% The poor have limited 80% access to health and 60% education facilities 40% Source: World Bank staff 20% calculation based on HIES 0% Maternity Maternity Pre-school Pre-school Primay Primary Secondary/ Collegiate Hospital Dispensary Clinic Secondary/ Collegiate Hospital Dispensary Poor 2006 2009 2012 Near Poor Clinic Lack of accessibility to school may hinder secondary school completion for poor students. Sri Lankan children attend school longer than their counterparts in neighboring countries. In Sri Lanka, almost all 17 and 18 year olds have attended at least one year of secondary school. But as noted above, poor children in Sri Lanka are much less likely than near-poor and non-poor children to complete secondary school. Inequality is also apparent in access to basic housing infrastructure as well as health and nutrition outcomes. The latter is critical to children’s development and future earnings potential. Appendix 4 discusses multi-dimensional inequality among Sri Lankan children in more details. 52 Poverty and Welfare in Sri Lanka 53 vI. SoCIAL pRoTeCTIoN ANd poveRTy pubLIC TRANSfeRS ARe fRAgmeNTed ANd SmALL IN boTh TheIR opeRATIoNAL SCope ANd fINANCIAL SCALe. Sri Lanka’s social protection system is fragmented and limited. The main social assistance program in Sri Lanka is Samurdhi subsidies. These consist of small monthly stamps worth between Rs. 200 to Rs. 1,500, given to families identified as poor by community offices. Besides Samurdhi subsidies, Sri Lankan households interviewed in the 2012/13 HIES also reported income received from five other social assistance programs. They are: fertilizer subsidies; school meals program, which provides lunch to students at poor schools; monthly assistance to the disabled; assistance for elders over 70 years old with no income; and the Thriposha food program, which provides nutritional supplements to women that are pregnant, breastfeeding, or whose children are undernourished. These programs are all implemented by different institutions, which may hamper administrative coordination. Table 4 displays the benefit level, number of beneficiaries, and the implementing institution of each program. School meals program Samurdhi subsidies Thriposha food programme Social Protection System Assistance for elders Monthly assistance to the disabled Fertilizer subsidies 54 Poverty and Welfare in Sri Lanka Table 4: Sri Lanka’s social assistance programs Program Implementing Beneficiary type Benefit level Beneficiaries Expenditure Institution 2011 2011 (Rs. million) Samurdhi DCGS Poor families Monthly 1,541,575 9,043.4 subsidy stamps worth (Families) of Rs.210, 750, 1200, 1500 per family Monthly NSPD Persons with Monthly 11,216 403.8 assistance for disabilities, of low assistance (Families) disabled income families worth of Rs.3,000 per family with a disabled New Elders NSE Elders over 70 years Monthly relief 229,892 (for 190 Assistance with no income (who of Rs.1,000 per 2012) (Allocation Program have been receiving elderly person for 2012) PAMA or Samurdhi benefit) School Meal ME Grade 1-5 students Mid-day meal 1,117,219 2,486 of primary and secondary schools in selected rural areas and students from special education National MH All pregnant mothers Two take- 873,509 980 Supplementary and lactating home packs of Food – mothers for first Thriposha once Thriposha 6 months and a month infants and children from 6-59 months deviating from the normal weight and those who are with growth faltering Source: Talakaratna et al. 2012, Safety nets in Sri Lanka: An overview By international standards, social assistance is less generous in Sri Lanka than in many other comparable countries. The World Bank has conducted a benchmarking exercise that examines the generosity and targeting of social assistance programs across 40 lower middle-income countries. One of the main indicators of generosity is the adequacy of benefits for the bottom quintile, which is defined as the total amount of social assistance transfers received by the bottom quintile, divided by their total consumption. In other words, this measure indicates the average contribution of social assistance programs to the budget of the bottom quintile. By this measure, Sri Lanka’s 55 vI. SoCIAL pRoTeCTIoN ANd poveRTy social assistance is tiny, at merely 6.6 percent, significantly below Pakistan, the Philippines, and Bolivia (Figure 54). not only are Sri Lanka’s social assistance programs small, but the budget devoted to social transfers has also fallen in recent years. Spending on social transfers, with the exception of fertilizer subsidies, has declined in real terms (see Figure 55). This accounts for the result, presented in Figure 23 in Section III above, that the fall in the real value of Samurdhi transfers slowed poverty reduction by 9.6 percent between 2002 and 2012/13. Because of the decline in spending, by 2012/13, the six programs listed above combined amounted to only over 3 percent of total household consumption for the bottom consumption quintile (shown in Figure 57 below). Figure 54: Adequacy of social assistance 40 for bottom quintile (Percent) Sri Lanka’s social 35 assistance programs are 30 25 small by international 20 standards 15 10 5 Source: World Bank Atlas of 0 Social Protection Indicators Pakistan Peru Tajikistan Bhutan Nepal Bangaladesh Sri Lanka Colombia Vietnam Philippines Bolivia for Resilience and Equity Figure 55: 200 4.5 Spending on social 4.0 transfers have 150 Percent of GDP 3.5 Rs billions declined 100 3.0 2.5 Source: World Bank staff 50 2.0 calculation based on HIES 0 1.5 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Samurdhi Pensions Fertilizer Others % of GDP (RHS) Figure 56: Share of household consumption 2.0% Transfer programs have 1.7 limited scale 1.5% 1.0% 0.8 0.8 Source: World Bank staff 0.6 calculation based on HIES 0.4 0.4 0.5% 0.3 0.3 0.3 0.2 0.2 0.2 0.0% Bottom 2nd Quintile Total Quintile Samurdhi School Feeding Fertilizer Others 56 Poverty and Welfare in Sri Lanka Samurdhi has, therefore, had a minor and decreasing impact on poverty reduction. Samurdhi transfers are too small to make a large impact on poor households’ budgets, as they contributed only 1.7 percent to household consumption of the poorest 20 percent of the population in 2012/13 (see Figure 56). Based on data from the HIES, subtracting the Samurdhi benefit from household consumption would increase the national poverty rate by 2.1 percentage points in 2002. But in 2012/13, the comparable figure had declined to merely 0.6 percentage points (Figure 57). Samurdhi contribution to poverty reduction Contribution to household consumption of the poorest 20% 1.7% 2012-13 Increase in poverty rate if Samurdhi benefit was removed 2.1 percentage points 0.6 percentage points 2002 2012/13 The poverty reduction impacts of other social assistance programs are even smaller. A recent DCS report examines how much the absence of nine programs other than Samurdhi would raise the national poverty rate in 2012/13 (DCS 2015, p. 4). These programs include the school food program, fertilizer subsidies, the Thriposha program, disability and relief, elderly payment, scholarship, health and medical aids, food and other commendation, and disaster relief assistance. Combined, their impact on poverty is almost negligible at 0.5 percentage points. Figure 57: Decrease in poverty rate due to value Samurdhi has negligible of Samurdhi transfer (% point) impact on poverty 2.5 2.1 2.0 Source: World Bank staff 1.5 1.2 calculation based on HIES 1.0 0.7 0.6 0.5 0.0 2002 2006/07 2009/10 2012/13 57 vI. SoCIAL pRoTeCTIoN ANd poveRTy Lack of generosity in Sri Lanka’s social protection programs is a larger problem than poor targeting. The country appears to be moderately well-targeted by international comparison. The international benchmarking analysis mentioned above also examines the share of social assistance programs that reaches the bottom quintile of households. Out of 40 lower middle-income countries included in the exercise, Sri Lanka in 2006 was estimated to transfer the 10th largest share to the bottom quintile, defined on the basis of pre-transfer income. Among countries in South Asia, Sri Lanka’s social assistance programs transferred the largest share to the bottom quintile (Figure 58). As an additional point of comparison, an older meta-analysis of 85 programs from the 1990’s and early 2000’s found that the median program transferred 25 percent more resources to the poor than a transfer of a fixed amount to all households.10 Samurdhi’s targeting performance by this measure is respectable, since it transfers 55 percent to the bottom two quintiles, which is 35 percent more than what would be allocated to the bottom two quintiles by an equal transfer. Figure 58: Sri Lanka’s social 35 Percent of social assistance accruing to bottom quintile assistance is better 30 25 targeted than in 20 neighboring countries 15 10 Source: World Bank Atlas of 5 Social Protection Indicators 0 BGD 2010 NPL 2010 PAK 2009 LKA 2006 for Resilience and Equity Among Sri Lanka’s social transfer programs, Samurdhi and the school lunch program are better targeted than fertilizer subsidies. Figure 59 shows the percentage of benefits that accrue to each quintile from Samurdhi, school lunch, fertilizer subsidies, and three other programs – old age transfers, disability, and Thriposha – combined. Samurdhi and school lunch programs transfer nearly the same amount to the bottom quintile, and in both cases roughly 55 percent of benefits go to the bottom 40 percent. The other three programs, when considered as a group, distribute 53 percent of their transfers to the bottom 40 percent. The fertilizer subsidies are the least well targeted transfer, with only 45 percent of the benefit being devoted to the bottom 40 percent. Figure 59: 100% 5.0% Despite considerable 9.6% 18.7% 8.4% 15.4% 12.2% leakage, Samurdhi and 80% 20.0% 13.5% 19.1% 15.6% school lunch program 60% 18.7% 26.6% 16.3% 25.0% are better targeted than 40% 22.4% fertilizer subsidies 25.8% 20% 38.8% 20.2% 38.8% 30.0% 0% Source: World Bank staff Samurdhi Fertilizer School Others calculation based on HIES Lunch Bottom quintile 2 3 4 Top quintile 10 See Coady, Grosh, and Hoddinott (2004) 58 Poverty and Welfare in Sri Lanka Samurdhi’s targeting performance has slightly worsened in recent years. Figure 60 shows how Samurdhi’s targeting performance has declined since 2002. In 2002, 42 percent of all transfers reached the bottom quintile and 70 percent reached the bottom 40 percent. But by 2012-13, this had fallen to 39 percent and 65 percent, respectively. Reforming the social assistance programs, including the Samurdhi, to reduce leakage and improve coverage of the disadvantaged can make it more helpful for the poor. Figure 60: 100 2.8% 4.0% 2.9% Percent of Samurdhi benefit 5.0% accruing to each quintile 9.4% 9.7% 8.9% Samurdhi’s targeting 80 9.6% 17.6% 18.3% 20.3% 20.0% performance has 60 worsened since 2009 28.1% 28.0% 25.2% 26.6% 40 Source: World Bank staff 20 42.1% 40.0% 42.6% 38.8% calculation based on HIES 0 2002 2006 2009 2012/13 Bottom quintile 2 3 4 Top quintile Private transfers, including international remittances, are not compensating for the reduction in public assistance. In general, few poor households benefit from international employment opportunities. International migration has increased in recent years, and for households that receive them, remittances from abroad are an important source of income. For these households, remittances account for roughly 30 percent of household consumption. But it is rare for the poor to benefit from remittances – only 4.5 percent of households in the bottom quintile receive them, as opposed to 10.7 percent for the top quintile. Therefore, only 9 percent of the total amount of remittances reaches the bottom quintile. And if remittance income is subtracted from consumption, the poverty headcount rises about only 1.4 percentage points. Figure 61: International remittances 35.1 40 33.9 33.5 32.6 31.3 tend not to reach the poor 28.8 30 22.4 18.5 Percent 20 14.5 Source: World Bank staff 10.7 calculation based on HIES 9.0 8.0 7.2 10 6.4 4.5 0 Percent that Average share Share of recieve of consumption remittances among recipients Bottom 20% 2 3 4 Top 20% 59 vI. SoCIAL pRoTeCTIoN ANd poveRTy demogRAphIC ChANgeS ARe ImpoSINg New ChALLeNgeS To The SoCIAL SAfeTy NeT Sri Lanka’s population is aging rapidly, which will substantially increase the demand for social support for the elderly in a near future. Elderly dependency rates are projected to rise fast, as the number of persons aged 60 years old and above is expected to double from 2011 to 2041. As can be seen in Figure 55 above, the recent bulk of spending on social assistance comes in the form of pensions. Furthermore, the share of employment in the public sector increased from 23 to 26 percent between 2002 and 2013, which will place further demands on the pension system going forward. But as of 2012, only about one third of the workforce reported that their employers contributed to a pension. The combination of low population growth and a rapidly aging population will create both financial and operational pressures to provide additional support for the elderly in the future. Figure 62: Less than one third of the 31.2% work force receive pension contribution by their employers 43.6% Source: World Bank staff 25.2% calculation based on HIES Wage Employee Wage Employee Non-wage with Pension without Pensions Employee Figure 63: 100 Sri Lanka’s population is Projected dependency ratios aging 80 60 Source: De Silva and Indralal (2012) 40 20 0 1981* 2006 2016 2026 2036 2046 2056 2066 2076 Old Age Dependency (60+yrs.) Child Dependency (<15 yrs.) Total Dependency Economic growth and an emerging middle class may also put further demands on the social protection system in the future. For example, as more workers enter wage jobs, the demand for unemployment insurance may rise. Active labor market programs, if designed and implemented well, can also help social assistance recipients graduate to productive employment. 60 Poverty and Welfare in Sri Lanka 61 VII. CONCLUSIONS Sri Lanka can build on a successful recent track record in boosting living standards for the poor. Poverty fell substantially from 2002 to 2009/10, and remained low in 2012/13. Household incomes also rose for the poor, as did other welfare indicators such as the share of non-food consumption, ownership of durable goods, and education attendance. In general, poverty reduction followed strong growth. But from 2006 to 2009, poverty fell despite moderate GDP growth, in part because of accelerated growth in the agricultural sector as food prices increased. In line with this, poverty fell slightly faster for agricultural workers than for non-agricultural workers, especially between 2006/07 and 2009/10. Increased labor earnings, due to a broad-based increase in labor demand, drove poverty reduction. When decomposing the drivers of poverty reduction, increased returns to agricultural and non-agricultural work each accounted for about a third of the decline. Across different non- agricultural sectors, employment and wages both increased, reflecting a broad-based rise in the demand for labor. This increased demand was particularly noticeable in construction, commerce, and transport and communication, and led to particularly strong earnings increases for less well- educated workers. While it is difficult to disentangle the underlying causes for the broad increase in the returns to work, we focus on four potential causes. First, a gradual structural transformation has led to a weakening emphasis on agriculture in the economy. The share of agriculture in GDP has declined, and employment in agricultural fell rapidly from 2002 to 2006 and again in 2012. The shift out of agricultural employment was particularly notable for youth. Second, Sri Lanka has experienced improved agglomeration and spatial connectivity, particularly in the main urban center that encompasses Colombo, Kandy, and Galle. Images of night-time lights show large increase on the outskirts of this center, as well as the secondary cities in Jaffna and Batticaloa-Akkaraipattu. Highly built-up areas along the Colombo-Kandy-Galle corridor contributed the most to national value added and manufacturing value added. Third, Sri Lanka benefited from increased aggregate demand. The period from 2002 and 2013 saw strong growth both in private consumption and private investment. Government investment, while still accounting for a small share of total aggregate demand, grew fivefold from 2002 to 2012/13 at a rate of 18 percent per year. Finally, agricultural workers profited from increases in food and commodity prices. Increases in the domestic price of rice and the world price of tea from 2006/07 to 2009/10 benefited agricultural workers and farmland owners alike, and were followed by large increases in minimum wages for agricultural workers in 2010. This was reflected in rapid wage growth and strong poverty reduction for agricultural workers. 62 Poverty and Welfare in Sri Lanka Not all of these potential drivers of growth and labor demand are sustainable in the future. In particular, fiscal constraints have already led to an announced reduction in government investment. More generally, the transport infrastructure and construction sectors are not likely to be strong sources of sustained and consistent growth. In addition, the prices for food and tea fell significantly from 2009 to 2012, which helped slow the reduction in poverty from 2009/10 to 2012/13. This raises the stakes for accelerating and managing the structural transformation and agglomeration in the coming years. Despite the fall in poverty, living standards generally remain low. The near-poor, defined as those living above the poverty line but below the 40th percentile, subsist on less than twice the poverty line and appear similar to the poor in terms of various welfare indicators. Furthermore, spatial inequality is substantial; poverty rates in the poorest DS divisions stayed as high as 30 to 40 percent in 2012/13. Monaragala district and the former conflict areas in Northern and Eastern provinces also stand out as severely poor. Poverty rates are also disproportionately high for vulnerable groups such as youth and ethnic minorities; and unemployment is high for youth and women. Finally, the Estate sector, despite relatively low levels of monetary poverty, is severely disadvantaged according to alternative welfare indicators such as education attainment, food consumption, and housing quality. To ensure continued improvement of living standards and shared prosperity, Sri Lanka can provide additional support to poor households. Devoting more generous resources to social assistance, while developing a broader social protection strategy to confront an aging population, will help poor households invest in the next generation. Support to poor household can take the form of multi-sectoral interventions to areas with particularly high poverty rates, such as providing job opportunities to youths through public works or job matching services. Such programs whose cost increases with the number of beneficiaries can be targeted to areas with high poverty rates. On the other hand, investment in infrastructure and other public goods may be targeted more efficiently to urban areas where most of the poor live. Although rural areas account for over 85 percent of the poor, over half the poor are estimated to live within 30 km of the two agglomeration areas of Colombo-Kandy-Galle and Jaffna-Trincomalee-Batticaloa. Further progress also depends on accelerating the structural transformation by connecting poor workers to more productive employment opportunities. To the extent that fiscal constraints allow, this could include prioritizing infrastructure investment to keep pace with increasing urbanization, which will place further demands on transportation infrastructure. Although 95 percent of the population can now access electricity, more cost-effective power generation will be required to support further productivity growth. Effective governance of cities will also be important to enhance the ability of workers to take advantage of more productive job opportunities in and around urban areas. Several remaining knowledge gaps make it difficult to identify the best approach to speed up the structural transformation in the Sri Lankan context. We propose six questions where further rigorous research could provide useful guidance for policy: 63 VII. CONCLUSIONS • How have recent infrastructure investments, including the expansion of access to electricity, affected labor productivity and poverty? Studies on the relationship between infrastructure and labor productivity can shed light on the importance of further spending to increase infrastructure in the future. • What are the status and determinants of internal migration? Surprisingly, given the importance of urbanization, there are little data on internal migration. Who is migrating where and why? How do households fare after migrating? Understanding these issues can inform interventions that provide either information or financial incentives to help households make informed choices about where their earnings opportunities are greatest.11 • To what extent and how do households enter and exit poverty? What adverse shocks are near- poor households most vulnerable to and what government interventions could help mitigate these shocks? • What determines labor market outcomes and skill acquisition? In particular, what explains the prevailing low rates of female labor force participation and high rates of youth unemployment? To what extent are labor market regulations enforced? And do they constrain the creation of salaries jobs? • How vulnerable are the poor and near-poor to a downturn in food and commodity prices? How important was the recent increase in prices in explaining the poverty reduction in the last decade? Above all, little is known about the beneficiaries and effects of public programs, investment and services. This analysis has drawn extensively on Sri Lanka’s impressive statistical infrastructure, which is well-equipped to monitor trends in poverty and labor market outcomes. It is invaluable for taking stock of the recent past, and providing an informed basis for speculating about the future. But with a few exceptions, this type of broad descriptive monitoring does not lend itself to generating specific policy recommendations and conclusions. Improvements in the statistical infrastructure can support more informed policy-making. This would partly involve a shift of emphasis from monitoring to evaluation. Practical steps could include adding new questions to existing surveys on the quality of services and programs, supporting one-time evaluations of selected programs in particular areas, and establishing a longitudinal household survey. The latter would be particularly valuable as a way to evaluate the benefits of public investments in human capital, as well as the dynamics of households’ participation in public programs. Efforts to fill knowledge gaps on these and other topics can help identify and prioritize the public policies and programs that best support the welfare of poor households. 11 For example, in Bangladesh an $8.50 financial incentive to migrate to urban areas during the lean season induced 20 percent of rural households to send a migrate to urban areas, which raised consumption of these poor households and made them 10 per- centage points more likely to re-migrate within 3 years (Bryan, Chaudhury, and Mobarak 2014). 64 Poverty and Welfare in Sri Lanka APPENDICES 65 APPENDIX 1: HOUSEHOLD AND LABOR FORCE SURVEYS IN SRI LANKA The analysis in this report is based mainly on Sri Lanka’s two large periodic surveys: the Household Income and Expenditure Survey (HIES) and the Labor Force Survey (LFS). On the one hand, these surveys provide extensive data that are representative at the district level and allow for poverty analysis over an extended period. On the other hand, this report needs to take into account several inconsistencies across survey rounds to ensure the reported statistics are chronically comparable. The most prominent issue is inconsistent geographical coverage over time because the surveys could not be implemented in various areas affected by the civil conflict. Table 5 below lists the areas excluded from the surveys’ samples in their most recent rounds. Specifically, the 2002 HIES does not cover Northern and Eastern provinces, which account for 12.9 percent of the national population (based on the 2011 Census of Housing and Population). The 2006/07 HIES excludes Northern province and Trincomalee district – equivalent to approximately 7.1 percent of the population. In the 2009/10 round, the survey excludes only 3 districts, namely Mannar, Kilinochchi, and Mulaitivu, which collectively house about 1.5 percent of the population. A special HIES was carried out in 2005 to record household economic welfare after the country was hit by the Tsunami disaster in December 2004. While this survey round covers all districts nationwide, it interviews only 5,000 households and provides data representative at only the national level. Not until 2012/13 does the HIES collect district representative data for all 25 districts, with 20,540 households being interviewed. Similarly, the LFS excluded conflict-affected districts in its earlier rounds. A special nationally representative round of the LFS was also conducted in 2005 to measure impacts of the Tsunami, with 5,350 households being interviewed. Since 2011, the LFS covers all districts and interviews approximately 19,400-20,200 households in each round. Table 5: Geographical coverage of the HIES and LFS Year LFS HIES 2002 Northern and Eastern provinces Northern and Eastern provinces 2003 Northern province n.a. 2004 Northern province n.a. 2005 All districts included, smaller sample size All districts included, smaller sample size 2006 Northern and Eastern provinces Northern province and Trincomalee district in Eastern province 2007 Northern and Eastern provinces 2008 Northern province n.a. 2009 Northern province Mannar, Kilinochchi, and Mulaitivu districts in Northern province 66 Poverty and Welfare in Sri Lanka APPENDIX 1: HOUSEHOLD AND LABOR FORCE SURVEYS IN SRI LANKA Year LFS HIES 2010 Northern province 2011 All districts included n.a. 2012 All districts included All districts included 2013 All districts included Due to the surveys’ inconsistent geographical coverage, Eastern and Northern provinces are excluded in our trend analysis, unless noted otherwise, to ensure the reported statistics are compatible across survey years. A separate analysis is devoted to these conflict-affected areas in Section III. This report does not use data from the 2005 surveys since they capture impacts of the weather shocks and are representative only at the national level. It is also important to note that the HIES and LFS contain similar but unidential questions on employment. In particular, occupation and industrial categories differ between the two surveys, leading to slightly different industrial categorization in the reported labor market outcomes. This report also accounts for the different definitions of working age across HIES rounds. The 2002-2009/10 HIES define working age as 10 years old and above, whereas the 2012/13 HIES selects 15 years old and above. The statistics on labor market in this report refer to only individuals aged 15 years old and above. Not all inconcistencies in the data can be harmonized, however. First, the analysis on asset ownership and debt cannot be extended to year 2002 because the 2002 HIES did not include questions on household durable assets and debts. Second, questions on usual employment status and activity in the 2002-2009/10 HIES did not specify any particular reference period. The 2012/13 HIES, in contrast, refers to the last one week for employment questions. These issues, fortunately, do not invalidate the report’s analysis. 67 appendIx 2: survey-To-survey ImpuTaTIon In srI lanka In Sri Lanka, new poverty estimates are typically released every three years after each new Household Income and Expenditure Survey (HIES) dataset has been collected and processed. In contrast, labor force statistics are based on the labor force survey (LFS), which is fielded every quarter. Since increased labor income is a main driver of poverty reduction in Sri Lanka, combining information from the LFS and HIES could be a logical and cost-effective way to generate more frequent estimates of poverty. This in turn would help make poverty estimates more useful as inputs into the development of economic policies. Generating more frequent poverty estimates by combining two or more surveys is called survey- to-survey imputation, and in the Sri Lankan context involves three steps. The first step is to identify a set of common variables in the HIES and LFS. The second is to generate a prediction model using data from the current round of the HIES. Finally, data from the most recent LFS are plugged into that prediction model to estimate the current levels of poverty. The underlying assumption in this method is that changes in the household characteristics used to predict consumption, such as the household head’s occupation, closely tracks changes in household consumption. But this may not be the case in Sri Lanka. To verify whether the survey-to-survey imputation method could be used to generate more frequent poverty statistics in Sri Lanka, we used data from the LFS and HIES, covering a common geographic area from June to December in both 2006 and 2009. One test is whether prediction models based on the 2006/07 HIES and imputed data from the 2009 LFS can accurately project poverty forward to 2009. Another important test is whether prediction models based on the 2009/10 HIES and imputed data from the 2006 LFS can project poverty backwards to 2006. Unfortunately, the LFS data do not accurately track poverty trends in Sri Lanka (Newhouse et al. 2014). This is especially true for the estate sector. According to the HIES, estimated poverty in the estate sector fell rapidly from 2006/07 to 2009/10, from about 34 to 10 percent. But predicted poverty based on imputed data from the LFS fell by only 2 percentage points. This is consistent with large numbers of households in the estate sector moving from just below to just above the poverty line, in ways that could not be captured by the explanatory variables in the LFS. 68 Poverty and Welfare in Sri Lanka appendIx 2: survey-To-survey ImpuTaTIon In srI lanka For the rural sector, where most poor people live, the forward predictions perform somewhat better. The predictions explain 70 percent of the 6.5 percentage point reduction in rural poverty between 2006/07 and 2009/10. But backward projection does not fare as well. A model from the 2009/10 HIES and imputed data from the 2007 LFS predict rural poverty rate in 2006/07 to be only 10.3 percent, much lower than the 15.4 percent rate directly estimated from the 2006/07 HIES. Finally, the method produces more accurate poverty predictions for urban areas. There are important differences in the wording of the questions between the LFS and HIES that made the predictions less accurate. This leaves two options to pursue frequent poverty estimation for Sri Lanka in a cost-effective way. The first is to include questions on assets and other welfare indicators to the LFS; but this may detract the survey from its main purpose of measuring labor market outcomes. The other option is to implement a new household survey that has a short questionnaire that includes variables necessary to impute household expenditures accurately. 69 AppeNdIx 3: Key ISSueS IN eSTImATINg poveRTy IN SRI LANKA geogRAphICAL CoveRAge of The hIeS SAmpLeS The Household Expenditure and Income Survey (HIES), from which official poverty figures are estimated, cover different populations over time because the survey could not be implemented in conflict-affect areas. The 2002 HIES excludes 8 districts in the Northern and Eastern provinces – about 13-14 percent of the national population – which are generally poorer than the rest of the country. The 2006 and 2009 HIES exclude 5 and 3 districts, respectively. Not until the 2012/13 HIES were all districts in Sri Lanka covered. National estimates of poverty typically include all available districts in the survey. Because of this, the 2002 HIES is likely to underestimate national poverty and, consequently, the pace of poverty reduction since 2002. However, limiting the sample to common districts only lowers the poverty rate by 0.6 percentage points. ALTeRNATIve defINITIoNS of CoNSumpTIoN AggRegATe In theory, the consumption aggregate used to calculate poverty statistics should include only items that are welfare-enhancing. Yet Sri Lanka’s official consumption aggregate includes savings, which have ambiguous correlation with welfare but often account for a significant proportion of rich households’ total expenditure. There are also some inconsistencies in the inclusion of durables goods and elder and child care services and their recall periods across different rounds of the HIES.12 We construct a new consumption aggregate from the 2002 and 2009 HIES to eliminate these inconsistencies and finds that the estimated poverty rates for 2002 and 2009 are both 0.3 percentage points lower than the official figures, the Gini index also decreases by 5 index points in both years. The estimated poverty reduction between 2002 and 2009, as a result, remains unaffected. INfLATINg The poveRTy LINe oveR TIme Sri Lanka’s poverty line deflator, the Colombo Consumer Price Index (CCPI), is based on retail prices in the capital district of Colombo. This characteristic limits the CCPI’s ability to capture changes in living costs of the poor. The prices actually paid by low-income households are often different from urban retail prices, partly because low-income households often consume commodities of lower quality, and partly because they are more likely to purchase in remote areas or through informal markets. Moreover, the CCPI reflects the consumption pattern of an average Colombo residents and assigns larger weights to richer households when aggregating budget shares across households. Data from the HIES show that poorer households spend a larger budget share on food, with smaller shares going to education, transport, and housing as compared with the shares used in the CCPI. Since food inflation was higher than non-food inflation during the period of 2002- 2009, the CCPI might underestimate the increase in living cost of low-income households and, consequently, underestimate the poverty incidence in more recent years. Constructing alternative price indexes to address these issues increases the estimated annual inflation and poverty rates between 2002 and 2009 by a mild but noticeable amount. 12 This sensitivity analysis was undertaken before the 2012/13 HIES data became publicly available. Therefore, we were unable to conduct robust check with respect to the official estimated poverty statistics for 2012/13. 70 Poverty and Welfare in Sri Lanka AppeNdIx 3: Key ISSueS IN eSTImATINg poveRTy IN SRI LANKA Specifically, if the bottom decile was selected as the reference group, instead of the Colombo population, annual inflation during 2002-2009/10 rises from the official estimate of 11.39 to 11.97, and the headcount index from 8.2 to 9.7 percent. If equal weights are assigned to households when aggregating household budget shares, annual inflation is estimated at 11.63 and poverty rate at 8.9. Using unit values of expenditure items from the HIES to construct survey-based price indexes, we also obtain higher inflation and higher poverty rates. Since our consumption bundle was restricted by items that are common between the 2002 and 2009/10 HIES, we cannot construct a survey-based index that has the same consumption bundle as the CCPI. Nevertheless, the official food-only price index is similar to our food-only, survey-based indexes. Also, using a weighted average of the official CPI and official food CPI to inflate the poverty line for the bottom 40 percent increases national poverty rate for 2012 by 0.8 percentage points. Therefore, the official poverty trend is robust to alternative sources of price data, at least for food items. SpATIAL pRICe AdjuSTmeNTS The spatial price index used by DCS to control for differences in living costs across regions is based on the HIES unit values and disaggregated at the district level. Our sensitivity analysis with regard to the index formula, the population reference group, and the level of disaggregation shows that alternative indexes only have minor impacts on the costs of living and ranking of poverty rate at the district level, but the declining trend in poverty incidence does not change considerably. Overall, although alternative methods of constructing the poverty line and calculating the consumption aggregate result in different poverty estimates, Sri Lanka’s impressive reduction in poverty remains robust. 71 appendIx 4: mulTI-dImensIonal InequalITy among srI lankan chIldren As pointed out in Section two above, income inequality is associated with unequal access to basic goods and services. Such inequality is particularly concerning among children because their access to such goods and services is often beyond their control. We consider inequality among Sri Lankan children in three critical areas: education, as indicated by secondary school attendance of children aged 12-17 years old and secondary completion by children aged 17-18 years old; health and nutrition, as indicated by the prevalence of diarrhea and stunting, and consumption of iron- rich foods, and household infrastructure (access to permanent housing structure, clean water, electricity, and toilet). Compared to other countries with similar income levels, Sri Lanka has performed relatively well in providing her children with almost universal coverage of primary school (97 percent) and electricity (98 percent), and reducing the prevalence of diarrhea to only 4 percent (Abras 2014). Yet by 2012/13 the country still falls short of universal coverage in several areas, specifically secondary school attendance (87 percent) and completion (53 percent), permanent housing structure (77 percent), and access to iron-rich foods (51 percent). Stunting rate also remains high (19 percent). These coverage rates, however, do not reveal how access to such goods and services and avoidance of the diseases are distributed across different socio-demographic groups. One measure to access multi-dimensional inequality is the Human Opportunity Index (HOI), in which opportunity is broadly defined as access to basic goods and services that are critical for the physical, mental and social development of children. The HOI is calculated as the difference between the overall coverage rate of an opportunity and a misallocation penalty that represents how much the opportunity is unequally distributed across circumstance groups.13 That means the higher inequality, the larger the misallocation penalty and the difference between HOI and coverage rate. Analysis based on data from the 2002-2012/13 HIES and 2007 DHS shows that among the selected indicators, inequality in opportunity, as reflected by the penalty, is larger in secondary school completion (9 percentage points), access to permanent housing (7 percentage points), and to a lesser extent, the probability of not being stunted (4 percentage points). As shown in Figure 65, household wealth, and education attainment of the household head are two major determinants of unequal opportunities in secondary school attendance and completion, access to electricity and permanent housing structure, and not being stunted. As well, living in conflict-affected areas significantly hinders children’s access to iron-rich foods and safe water, and avoidance of diarrhea. Being stunted is strongly affected by wealth and education of household head, as well as household size. Location plays an important role in determining access to electricity, permanent housing structure, safe water, and toilet. 13 Abras (2014) provides a useful discussion on the construction of the HOI. 72 Poverty and Welfare in Sri Lanka appendIx 4: mulTI-dImensIonal InequalITy among srI lankan chIldren Figure 65: 100% 100% 100% Determinants of multi- 80% 80% 80% dimensional inequality 60% 60% 60% among children 40% 40% 40% 20% 20% 20% 0% 0% 0% Source: World Bank staff Secondary Attendance Age Electricity Permanent Safe Toilet Diarrhea Not Mother Kid Iron Complete Age 12-17 Housing Water Stunted Iron calculation based on HIES 17-18 Gender Gender Head Gender Gender Head Gender Gender Head Age Head Educ Head Age Head Educ Head Age Head Educ Head Quintiles HH Size Quintiles HH Size Quintiles HH Size Location Conflict Location Conflict Location Conflict Changes in the HOI can be decomposed into (i) the composition effect, which is due to changes in the distribution of circumstances in the population, (ii) the scale effect, or the proportional change in the overall coverage across different circumstance groups, and (iii) the equalization effect, i.e. change in the coverage of under-covered groups while keeping the overall coverage rate constant. A major proportion of the changes in HOI in Sri Lanka are attributed to the scale effect, suggesting public investment in public facilities would benefit all (see Figure 66). The equalization effect is largest in access to electricity. Figure 66: Scale effect accounts for 20 Change in HOI (percentge point) most of the changes in 15 HOI during 2006-2012 10 Source: World Bank staff calculation based on HIES 5 0 -5 Primary on time (age 12-13) Secondary Complete Age 17-18) Attendance Age 12-17 Electricity Electricity Permanent Housing Safe Water Composition Scale Equalization 73 74 Table 6: Profile of the poor and near-poor All Districts Estate Rural Poor Near Upper Bottom All Poor Near Upper Bottom All Poor Near Upper Bottom All Poor 60% 40% Poor 60% 40% Poor 60% 40% Sex Male 47.8 47.7 46.7 47.7 47.1 43.4 48.8 48.3 47.8 48 48.2 47.5 46.5 47.7 47 Female 52.2 52.3 53.3 52.3 52.9 56.6 51.2 51.7 52.2 52 51.8 52.5 53.5 52.3 53 Education (aged 21 or more) Incomplete primary 36.5 27.9 17.4 29.3 22.1 47.9 45.8 37.2 46.2 42.8 35.5 27 17.7 28.5 22.2 Incomplete secondary 55.3 55.8 46.2 55.7 50 49.8 49.6 51.1 49.6 50.2 55.8 56.2 47.4 56.1 51.1 Secondary 6 11.3 18.9 10.4 15.5 1.9 3.6 7.5 3.3 4.9 6.4 11.6 19.1 10.7 15.5 Post-secondary 2.3 5.1 17.5 4.6 12.4 0.4 1.1 4.2 0.9 2.2 2.4 5.2 15.8 4.7 11.2 Age 0-14 33.4 28.8 23.4 29.5 25.9 39.3 34.1 24.7 35 31.1 32.7 28.4 23.6 29.2 26 15-24 16.4 15.3 14.4 15.5 14.8 14.4 12.9 13.6 13.2 13.3 16.8 15.3 14.3 15.5 14.9 25-49 30.7 32.9 34.8 32.5 33.9 28.1 31.3 34.7 30.7 32.2 30.9 33 34.8 32.6 33.9 50-64 12.3 14.9 18.3 14.5 16.8 10.3 13.8 20.4 13.2 15.9 12.5 15 18.3 14.6 16.7 65+ 7.2 8.2 9.1 8 8.7 7.9 8 6.7 8 7.5 7.2 8.3 8.9 8.1 8.6 Ethnicity Sinhala 64.9 67.6 78.9 67.2 74.2 4.9 7 12.4 6.6 8.8 71.4 75.7 85 75 80.7 ANd IN foRmeR CoNfLICT pRovINCeS Sri Lanka Tamil 19.9 13.7 8.8 14.7 11.2 23.3 17 12.5 18.1 16 19.6 12.3 7 13.6 9.8 Hindu Tamil 6.2 6.6 3 6.5 4.4 71.6 74.6 73.4 74.1 73.8 0.8 1.1 0.7 1.1 0.8 Sri Lanka Moor 8.7 11.9 8.7 11.4 9.8 0 1.4 1.6 1.2 1.3 8.1 10.7 7.1 10.2 8.4 Religion Buddhist 62.3 65 72.7 64.5 69.4 5.9 7.3 12.9 7 9.2 68.9 73.3 79.7 72.5 76.7 Hindu 21.9 16.9 9.5 17.8 12.8 81.1 81.5 76.1 81.4 79.4 17.3 11.3 6.1 12.3 8.7 Islam 8.8 12.2 9.2 11.6 10.2 0.2 3.2 2.7 2.7 2.7 8.1 10.8 7.5 10.3 8.7 Christian 6.9 5.9 8.6 6.1 7.6 12.8 8 8.3 8.9 8.7 5.7 4.7 6.7 4.9 5.9 Economic Activity (aged 15 or more) Employed 47.2 48.8 50.8 48.6 50 51.5 57.9 62 56.9 59 47.2 49 52.1 48.7 50.7 Unemployed 3.9 4.3 2.8 4.2 3.4 3.7 4.8 3.8 4.6 4.3 3.8 4 2.8 4 3.3 Out of labor force - Student 9.1 9.2 9.8 9.2 9.6 7.7 6.1 5.5 6.4 6 9.3 9.4 9.7 9.4 9.6 Out of labor force - Other 39.8 37.7 36.6 38 37.1 37.1 31.1 28.6 32.1 30.7 39.6 37.6 35.4 37.9 36.4 Economic Sector (employed aged 15 or more) appendIx 5: profIle of The poor and near-poor By secTor Agriculture 51.9 42.9 26.1 44.3 32.7 69.8 66.9 64.9 67.3 66.3 52.2 44.5 29.9 45.7 36 Industries 27.7 28.8 23.6 28.6 25.4 16.5 19.7 13.7 19.2 16.9 28.3 28.7 23.7 28.6 25.6 Services 20.5 28.3 50.3 27.1 41.9 13.8 13.4 21.3 13.4 16.8 19.5 26.8 46.4 25.6 38.4 Housing infrastructure Has Safe Water (protected well, tap) 82.8 85.4 92.3 85 89.4 55.1 47.9 42.2 49.1 46.5 84.3 87.1 92.4 86.6 90 Has Toilet 93 96.2 94.7 95.6 95.1 95 96.5 96.1 96.3 96.2 92.8 96.7 97.2 96 96.7 Has Permanent Walls 84.1 91.6 96.8 90.3 94.2 88.7 95.3 96.1 94.2 94.9 84.3 91 96.5 89.8 93.7 Has Permanent Roof 72.1 83.3 93.4 81.4 88.6 25.7 23.4 26.6 23.8 24.9 75.6 87.5 95.1 85.4 91 Has Permanent Floor 71.2 82.8 90.7 80.9 86.8 72.9 83.4 91.1 81.6 85.2 70.2 81.1 89 79.1 84.8 Has Permanent Walls, Roof and Floor 54.5 69.4 84.3 66.9 77.3 17.7 19.6 25.1 19.3 21.5 56.8 71.8 84.2 69.1 77.8 House Ownership Owns 83.7 85.1 88.3 84.9 87 5.6 11.3 13.1 10.3 11.4 90.4 92.2 93.5 91.9 92.8 Free Rent 8.7 7.8 3.8 7.9 5.4 76.4 70.9 71.6 71.9 71.8 3.1 2.8 1.6 2.9 2.2 Relief Housing 1.2 0.9 0.9 0.9 0.9 6 6.9 5.9 6.7 6.4 0.8 0.4 0.3 0.4 0.4 Rent/Other 6.5 6.2 7 6.3 6.7 12 10.9 9.4 11.1 10.5 5.7 4.7 4.6 4.8 4.7 Table 7: Profile of the poor and near-poor in post-conflict provinces Northern and Eastern Provinces Other Provinces Poor Near Poor Upper 60% Bottom All Poor Near Poor Upper 60% Bottom All 40% 40% Sex Male 46.7 47.2 46.1 47.1 46.7 48.2 47.8 46.7 47.9 47.2 Female 53.3 52.8 53.9 52.9 53.3 51.8 52.2 53.3 52.1 52.8 Education (aged 21 or more) Incomplete primary 38.6 28.6 21.1 30.5 26.4 35.9 27.7 17 29 21.4 Incomplete secondary 54.3 54.6 48 54.5 51.6 55.6 56 46 56 49.7 Secondary 5.5 11.7 17.2 10.5 13.5 6.1 11.2 19.1 10.4 15.8 Post-secondary 1.6 5.2 13.7 4.5 8.5 2.4 5.1 17.9 4.7 13 Age 0-14 38.5 31.2 25.3 32.6 29.4 32 28.2 23.2 28.9 25.3 15-24 20.3 18.7 16.1 19 17.7 15.4 14.5 14.2 14.7 14.4 25-49 29.6 32 34.1 31.6 32.7 31 33.1 34.8 32.7 34 50-64 7.7 12.6 16.6 11.6 13.8 13.5 15.4 18.5 15.1 17.2 65+ 4 5.4 7.9 5.1 6.3 8.1 8.8 9.2 8.6 9 Ethnicity Sinhala 5.9 13.4 14 11.9 12.8 80.8 79.3 85.7 79.5 83.4 ANd IN foRmeR CoNfLICT pRovINCeS Sri Lanka Tamil 74.9 60.6 61.4 63.4 62.5 5.2 3.6 3.3 3.8 3.5 Hindu Tamil 1 0.5 0.4 0.6 0.5 7.6 7.9 3.3 7.9 5 Sri Lanka Moor 17.7 25.4 24.1 23.9 24 6.3 9.1 7.1 8.6 7.7 Religion Buddhist 5.9 13.3 13.2 11.8 12.4 77.4 76.1 78.9 76.3 77.9 Hindu 63.8 51.8 51.1 54.2 52.8 10.7 9.4 5.1 9.6 6.8 Islam 17.8 25.4 23.9 23.9 23.9 6.4 9.4 7.7 8.9 8.1 Christian 12.5 9.5 11.7 10.1 10.8 5.4 5.1 8.3 5.2 7.1 Economic Activity (aged 15 or more) Employed 40.3 39.9 41.7 40 40.8 48.9 50.7 51.7 50.4 51.3 Unemployed 6.3 7.2 4.5 7 5.9 3.3 3.7 2.7 3.6 3 Out of labor force - Student 13.8 11.8 10.3 12.1 11.3 8 8.7 9.8 8.5 9.3 Out of labor force - Other 39.5 41.1 43.5 40.8 42.1 39.8 37 35.9 37.4 36.4 Economic Sector (employed aged 15 or more) Agriculture 46.6 37.8 26.4 39.4 33.2 53 43.7 26 45.1 32.6 appendIx 5: profIle of The poor and near-poor By secTor Industries 27.9 26.9 20.9 27.1 24.1 27.6 29.1 23.8 28.9 25.6 Services 25.5 35.3 52.7 33.5 42.7 19.4 27.2 50.1 26 41.8 Housing infrastructure Has Safe Water (protected well, tap) 92.5 93.2 95.3 93.1 94.1 80.2 83.7 92 83.2 88.7 Has Toilet 82 92.6 95.5 90.5 92.7 95.9 96.9 94.7 96.8 95.5 Has Permanent Walls 78.8 86.7 94.2 85.1 89.1 85.6 92.6 97 91.5 95 Has Permanent Roof 73.2 82.6 90.3 80.8 84.9 71.8 83.5 93.8 81.6 89.2 Has Permanent Floor 81.9 88.4 94.7 87.2 90.4 68.4 81.6 90.3 79.4 86.2 Has Permanent Walls, Roof and Floor 67.1 76.4 86.5 74.6 79.8 51.1 67.9 84 65.2 77 House Ownership Owns 84.4 87.4 86.3 86.8 86.6 83.5 84.6 88.6 84.4 87 Free Rent 8.5 6.2 6.6 6.7 6.6 8.7 8.1 3.5 8.2 5.3 Relief Housing 1.8 0.4 0.3 0.7 0.5 1 1 1 1 1 Rent/Other 5.2 6 6.9 5.9 6.3 6.8 6.3 7 6.4 6.8 75 Poverty and Welfare in Sri Lanka RefeReNCeS Abras, A 2014, ‘HOI in Sri Lanka’, unpublished manuscript. Amendola, N, Razzaz, S & Vecchi, G 2014, ‘Poverty trends in Sri Lanka: How robust?’, unpublished manuscript. Bryan, G, Chowaudhury, S & Mobarak, AM 2014, ‘Underinvestment in a profitable technology: the case of seasonal migration in Bangladesh’, Econometrica, vol. 82, no. 5, pp. 1671-1748. Ceriani, L, Inchauste, G, Olivieri, S & Shivakumaran, S 2014, ‘Understanding Poverty Reduction in Sri Lanka: evidence from 2002-2012/13’, unpublished manuscript. Coady, D, Grosh, ME & Hoddinott, J 2004, ‘Targeting of transfers in developing countries: Review of lessons and experience’, The World Bank, Washington, DC. DCS 2015, ‘Poverty Indicators, Household Income and Expenditure Survey 2012/13’, accessed on 29 June 2015, De Silva, WI & Indralal, W 2012, ‘The age structure transition and the demographic dividend’, Sri Lanka Journal of Advanced Social Studies, vol. 2, no. 1, pp. 3-46. Dinkelman, T 2011, ‘The effects of rural electrification on employment: New evidence from South Africa’, The American Economic Review, vol. 101, no. 7, pp. 3078-3108. Lipscomb, M, Mobarak, AM & Barham, T 2013, ‘Development effects of electrification: Evidence from the topographic placement of hydropower plants in Brazil’, American Economic Journal: Applied Economics, vol. 5, no. 2, pp. 200-231. Newhouse, D, Shivakumaran S, Takamatsu, S & Yoshida, N 2014, ‘How survey-to-survey computation can fail’, World Bank Policy Research Working Paper No. 6961, The World Bank, Washington, DC. Ravallion, M, Chen, S & Sangraula, P 2009, ‘Dollar a day revisited’, The World Bank Economic Review, vol. 23, no. 2, pp. 163-184. 76