Sri Lanka - Poverty assessment

Report No. 13431-CE Sri Lanka Poverty Assessment January 11, 1995 Country Management Unit-A Country Department I South Asia Region - -e .4~.ws; % *+5,-%- . .. . S S;-~teW,orld - :-,-tzF ~t.'=;"-4  SRI LANKA POVERTY ASSESSMENT Abbreviations AGA Assistant Government Agent CDF Cummulative Distribution Function CPI Consumer Price Index CWE Cooperative Wholesale Establishment DCS Department of Census and Statistics FCD Food Commissioner Department FGT Foster, Greer, Thorbecke (class of poverty measures) FOD First Order Dominance FSP Poor Relief Food Stamp Program GDP Gross Domestic Product GNP Gross National Product HIES 1990/91 Household Income and Expenditure Survey JSP Janasaviya Program JTF Janasaviya Trust Fund LFSS 1985/86 Labor Force and Socio-Economic Survey MASL Mahaweli Authority of Sri Lanka MDMP Mid-Day Meal Program NCHS U.S. National Center for Health Statistics NGOs Non-Governmental Organizations NWSDB National Water Supply and Development Board OER Official Exchange Rate PFP Policy Framework Paper POs Partner Organizations PPP Purchasing Power Parity SA3CI Country Operations, Industry and Finance Division, South Asia III Department, World Bank SOEs State-Owned Enterprises t/ha Metric tons per hectare TVET Technical and Vocational Education and Training VAT Value Added Tax WIHO World Health Organization  SRI LANKA POVERTY ASSESSMENT' Preface This is the first Poverty Assessment of Sri Lanka supported by the World Bank. It is perhaps of special interest because Sri Lanka is a country long known for its enlightened human resource policies and above-average health and education outcomes, but more controversial economic management. The main body of the report is divided into four chapters. Chapter I starts with an examination of trends in consumption poverty in the post-Independence period, and the sources of income growth for poor households. This is followed by a discussion of Sri Lanka's performance regarding indicators of educational achievement and health, fertility and nutrition outcomes. The chapter ends with an analysis of the characteristics of the poor as revealed by recent household survey data. Chapter II discusses Sri Lanka's growth performance, the ongoing economic reform program, and two special topics closely related to poverty reduction prospects--unemployment and the constraints to agricultural sector growth. Chapter III discusses current issues in human resource development. Chapter IV examines the array of public transfer programs and makes suggestions for reforms aimed at increasing the poverty alleviation impact of the programs. It then discusses the role of Non-Governmental Organizations (an important component of the country's safety net) and what can be done to create a more supportive environment for their activities. The Executive Summary provides a succinct account of the report's main conclusions and recommendations. This is an integrative task and as such it draws on a number of other Bank reports as well as studies conducted by others, which are referenced throughout the report. The main mission for the Assessment took place in February 1993 and was combined with a Public Expenditure Review mission led by Mr. Menahem Prywes. We gratefully acknowledge the valuable assistance provided by the Bank's Resident Mission in Colombo during and after the mission. We would also like to express our gratitude to the Department of Census and Statistics for making available to us the data from the two household surveys and assisting us in their use. Several Sri Lankan government officials and academics provided valuable comments and suggestions on an earlier version of the report, including Dr. (Mrs.) Patricia Alailima, Dr. R.M.K. Ratnayake, Prof. W.D. Laksham, Mr. Charitha Ratwatte, Mr. A.S. Jayawardena, Mr. A.G.W. Nanayakkara, Mr. Chandra Munasinghe, Dr. Leslie Gunaratne, Dr. P.B. Jayasundera, and Dr. Lal Jayawardena. We are grateful to all of them for their contributions to the report. This report was prepared by Hugo Diaz (task manager, SA3CI), with inputs from Gaurav Datt, Soniya Carvalho, Dileni Gunewardena, Tom Selden, Neville Edirisinghe, Ruben Suarez, Mohammed Allak, Paul Blay, Douglas Lister, Robert D. Hunt, Theresa Bradley, Pedro Alba, Menahem Prywes, Jytte Laursen, and Dilesh Jayanntha. Peer reviewers for the task were Helena Ribe and Roland K. Peters.   SRI LANKA POVERTY ASSESSMENT CONTENTS Pages Abstract Executive Summary ....................................................... i I. POVERTY TRENDS A. Consumption Poverty in the 1953-85 Period ............................. I B. Consumption Poverty .n the 1985-90 Period ............................. 6 C. Consumption Poverty in the 1990-93 Period ............................ 10 D. Sources of Income Growth .................. ...................... 12 E. Trends in Physical Indicators of Human Welfare ......... .. .............. 16 F. Characteristics of the Poor from Household Survey Data ....... .. ........... 23 II. ECONOMIC GROWTH AND POVERTY ALLEVIATION A. Sri Lanka's Growth Performance .................................... 30 B. The Post-1989 Economic Reform Program and Effects on Poverty .... ......... 32 C. Ensuring Sustained Growth ............ ............................ 37 D. Unemployment and the Labor Market ................................. 37 E. Constraints to Agriculture Sector Growth .............................. 43 III. HUMAN RESOURCE DEVELOPMENT A. Education and Training . ........................................... 51 B. Health ....................................................... 55 C. Water Supply and Sanitation ....................... 61 IV. TRANSFERS, SAFETY NETS AND NGOs A. Transfer Programs ....................... 66 B. Non-Governmental Organizations .76 ANNEXES: 1. Documents Cited in Main Text. 82 2. Analysis of Poverty From Household Survey Data ............ ............... 86 3. Tables and Figures from the Analysis of the 1985/86 and 1990/91 Household Surveys .95 4. NGOs in Sri Lanka .129 MAP   Sri Lanka Poverty Assessment Abstract Post-Independence Sri Lanka has made impressive progress in reducing the incidence of poverty, reflected in terms of improvements in human development indicators as well as in consumption terms. This progress was made possible by a combination of two factors: (i) a strong push on public provision of basic health and education together with income transfer programs that enabled higher consumption by the poor; and (ii) a relatively good long-term growth performance, with real per capita GDP growth of about 2.5 percent per year on average for the period 1950-93. While Sri Lanka has achieved human development indicators far superior to most other developing countries, it is still a low income economy. Though its long-term growth rate in per capita terms compares favorably with most of the developing world, it falls well short of the growth rate achieved in the last three decades by the high-performing East Asian economies, such as Korea, Malaysia, Thailand and Indonesia. A stronger growth performance along the lines of these other economies would have enabled a much deeper reduction in poverty. The main reason why Sri Lanka's economic growth lagged behind the high-performing East Asian economies is that policies were less supportive of economic growth. Problems with macroeconomic management have constrained domestic saving and investment and fueled inflation; exports growth was constrained by inward looking trade policies, interventions in the labor market, and, at times in the past, exchange rate overvaluation; and the agriculture sector's performance has been constrained by pervasive controls on land ownership and use, marketing and pricing. Starting in 1977 an effort was made to bring economic policies more in line with those of the most successful Asian economies. The new strategy sought to increase the role of markets and the private sector by reducing restrictions on pricing, investment and extemal trade and payments. Unfortunately, however, the reform program faltered in the 1980s. The program failed to address the restructuring of the overextended public sector, and since early in that decade the country has been immersed in a costly civil war. Unsustainable macroeconomic imbalances emerged in the second half of the 1980s, and growth was very low during those years. Renewed efforts at economic reform from 1989 onwards have shown the large potential for economic gains, with the economy expanding at around 4 percent per year in real per capita terms, a surge in private sector investment and rapid growth in labor-intensive manufactured exports. While the revitalized reform program has rightly aimed at reducing macroeconomic imbalances, social spending has been protected. Even though data on poverty indicators are not available for this recent period, the broad-based nature of growth, the increase in labor-intensive manufactured exports and the protection of social spending suggest that poverty probably declined further in the early 1990s. The above suggests that the chief focus of Sri Lanka's future poverty alleviation strategy would need to be the acceleration of broad-based economic growth, with emphasis on agriculture and labor-intensive manufacturing, while protecting public spending for human development. This will enable Sri Lanka to reap poverty reduction gains in consumption terms in line with its social indicators. These tasks constitute a major challenge, especially in view of the civil war and the persisting resource constraint in the public sector. The direct negative impact of the civil war on poverty reduction is obvious --resulting from the severe disruption of economic activity in the conflict areas and a large diversion of resources away from high-priority spending in order to finance military and police activities. In addition, the indirect negative effects in terms of the adverse impact on incentives for private investment and therefore growth could be substantial. Hence a peaceful resolution of the civil war is of critical importance for poverty reduction. Along with this resolution, Sri Lanka will need to continue the implementation of the economic refomn program that was resumed in 1989. The apparent disruption to the program implied by the short-term, politically motivated package of measures of April/May 1994 needs to be reversed quickly. Structural reforms need to be deepened. This includes completing the rationalization of the trade regime; tax reforms; rationalizing the structure of public expenditure; deepening the privatization of public enterprises and the deregulation of foreign direct investment; pursuing reforms leading to a more flexible labor market; and tackling land market rigidities and other high-priority areas of reform in agriculture. At the same time, essential reforms of basic social services programs --including education, health and water supply and sanitation-- are necessary to improve service quality and ensure that the truly poor benefit from these programs. Finally, income transfer programs need to be rationalized and better targeted to the poorest segments of society.   EXECUTIVE SUMMARY 1. This is the first Poverty Assessment of Sri Lanka conducted by the World Bank. It is perhaps of special interest because Sri Lanka is a country long known for its enlightened human resource policies and above-average health and education outcomes, but more controversial economic management. Perhaps because of these traits, the country has evoked a rich body of economic and social research on poverty and related matters from both local and foreign social scientists. The present report has greatly benefited from this literature, and a number of specific pieces are referenced throughout the main text and in the Bibliography in Annex 1. The country is also well endowed with household survey data going back for several decades, and the report has also drawn on some of these data. This Executive Summary first summarizes Sri Lanka's record in poverty alleviation and then outlines a recommended strategy for poverty alleviation in the future. A. Sri Lanka's Record Summarized Progress in Reducing Consumption Poverty 2. In the recent literature, the concept of "poverty" is generally understood as encompassing two dimensions. One dimension is the extent to which individuals or households are unable to achieve a certain minimum consumption standard. This dimension of poverty is known as "consumption poverty". The second dimension of poverty is defined by the extent of premature mortality, morbidity burden and illiteracy in a society. This dimension is usually referred to as the human development dimension of poverty. Both dimensions must be assessed in studying the extent of poverty in any given country. 3. The evidence reviewed in this report indicates that, in the post-Independence period, Sri Lanka has made good progress in reducing consumption poverty. National accounts data, available since 1950, show that the performance of the economy in terms of real per capita GDP growth --about 2.5 percent annual average in the 1950-93 period-- has been good, resulting in the tripling of private consumption per capita in real terms. Moreover, these consumption gains have accrued more or less proportionately to all income groups, because the distribution of household income by decile towards the end of the period was about the same as in the early 1950s (even though there were fluctuations in in the intervening years). Thus, it can be concluded that there has been a large decline in the incidence of consumption poverty in the post-Independence period, regardless of the particular "poverty line" that one might choose. The data also show that, as compared to many other developing countries, Sri Lanka had a relatively egalitarian distribution of consumption around 1990, with a Gini coefficient of 0.30. 4. A more detailed examination of the evidence indicates that there was little consumption poverty decline between 1950 and 1965. Between 1965 and 1985 there was rapid poverty decline, propelled by rapid growth of per capita income and consumption and a household income distribution that improved early in the period but then reverted to a degree of inequality similar to that at the beginning of the period. Between 1985 and 1990 there was very little growth of per capita consumption; however, poverty continued to decline on account of an improvement in the distribution of consumption. It should be noted, however, that our survey data for the 1985-90 period exclude the North and East (probably about 15 percent of the population of the country). Finally, in 1990-93, consumption per capita grew at 1.8 percent annually, despite substantial macroeconomic  - ii - adjustment. The broad-based nature of growth in this period, together with the fact that adjustment did not adversely affect government social spending, suggest that poverty may have declined further. 5. While progress in reducing consumption poverty has been impressive, the analysis in this report also shows that substantial poverty remains. When measured in relation to the reference poverty line estimated in this report, which is very close to the "one-dollar-a-day" (per person at 1985 purchasing power parity) poverty line suggested by the World Bank as a benchmark to measure poverty throughout the developing world, the proportion of individuals in Sri Lanka with consumption expenditure below the poverty line was about 22 percent in 1990. Progress in Human Development 6. In addition to progress in reducing consumption poverty, Sri Lanka has done very well in terms of literacy and other basic education indicators, and mortality and fertility have both declined to the point where the country has almost completed its demographic transition. These accomplishments are a consequence of the country's enlightened health and education policies in the past several decades. There has been a strong drive by the public sector to provide universal primary education to boys and girls alike, and country-wide coverage of basic health services. At present, the very good outcomes of these policies in terms of indicators of basic education enrollment, literacy, mortality and fertility clearly mark Sri Lanka as an outlier in relation to other countries at similar levels of per capita income. In 1990, Sri Lanka had a literacy rate of 88 percent overall (83 percent for women), as compared to 60 percent average for all low-income economies (48 percent for women). Its infant mortality rate was a very low 19 per thousand live births, as compared to 69 per thousand average for all low-income economies. Its total fertility rate was 2.4, as compared to 3.8 average for all low-income economies. Moreover, Sri Lanka has done very well by international standards in terms of the pace of change of education and health indicators in the 1960-90 period. A recent Bank study found that Sri Lanka ranked 19th out of a sample of 71 developing countries in terms of the decline in infant mortality in that period. In terms of increases in net primary enrollment, Sri Lanka ranked 8th out of a sample of 43 developing countries in the same period. 7. There are, however, some areas of Sri Lanka's human development performance that need attention. Data on morbidity suggest that the incidence of malaria has increased sharply. Evidence from a number of surveys indicates substantial undernutrition in young children by reference to international norms, although the latest (1993) data suggest a significant recent decline in the incidence of such undernutrition. The incidence of wasting (low weight-for-height, reflecting acute undernutrition) among children 6-36 months old has been estimated at 15.9 percent in 1979/82, 18.2 percent in 1988/89 and 14.3 percent in 1993. For stunting (low height-for-age, reflecting chronic undernutrition), the corresponding estimates are 30.5 percent, 33.2 percent and 22 percent respectively. Unlike other developing countries, however, undernutrition does not result in high levels of infant and child mortality, possibly because of a combination of good basic health services and education of the mothers. In basic education, there is much scope for improving equity and quality. 8. Sri Lanka's unemployment rate is high by international standards. It has exceeded 14 percent on average since 1980. Moreover, unemployment spells are long: in 1990, 85 percent of the unemployed had spent more than one year searching for a job. There is also a positive relationship between poverty and the incidence of unemployment, with the poorest group (that of individuals with consumption below 80 percent of the poverty line) having the highest unemployment rate at 25 percent in 1990. While the debate on the nature and causes of this high unemployment continues, it would  - ijj - appear that factors contributing to high unemployment include: excessively restrictive labor legislation that raises the cost of labor to firms; poor quality of education acquired by many workers coupled with high labor market aspirations; and the public sector's tendency to hire large numbers of unemployed graduates, thus increasing incentives for unemployment. In addition, the official definition of the unemployed in use in Sri Lanka is very broad. These factors result in widespread queuing for rationed jobs in the regulated sector of the economy, both public and private. Such queuing, in turn, is made possible by the generalized availability of private transfers to support the unemployed while queuing. This interpretation suggests that in order to reduce the unemployment rate significantly it would be necessary to create good jobs (i.e., such as those in the regulated sector) at a faster pace than that of labor force growth, improve the flexibility of the labor market by easing regulatory restrictions, lower aspirations of entrants to the labor market and improve the quality of education and training. Continuing to rely on the public sector to create jobs for the unemployed beyond the needs of an efficient service, however, would be fiscally unsustainable. The Persistence of Poverty and Missed Opportunities for Higher Growth 9. Sri Lanka remains a low-income economy. And, though its long-term growth rate in per capita terms compares favorably with most of the developing world, it falls well short of the growth rate achieved in the last three decades by the high-performing East Asian economies, such as Korea, Malaysia, Thailand and Indonesia. This can be seen in the following figures. In 1960, Sri Lanka's per capita GNP, estimated at US$141 in current (1960) dollars, was substantially higher than that of Thailand (US$96) and Indonesia (US$51), about the same as in Korea (US$156), and only 50 percent lower than in Malaysia (US$273). By 1992, Sri Lanka's per capita GNP in current (1992) dollars had risen to US$540, as compared with US$6,790 for Korea, US$2,790 for Malaysia, US$1,840 for Thailand and US$670 for Indonesia. A stronger growth performance along the lines of the high-performing East Asian economies would have allowed a much deeper reduction in Sri Lanka's poverty indicators. 10. The main reason why Sri Lanka's economic growth lagged behind the high-performing East Asian economies, even though its social indicators are comparable (and in some cases better) than those of these economies, is that policies were less supportive of economic growth. Problems with macroeconomic management have constrained domestic saving and investment and fueled inflation; exports growth was constrained by inward looking trade policies, interventions in the labor market, and, at times in the past, exchange rate overvaluation; and the agriculture's sector performance has been constrained by pervasive controls on land ownership and use, marketing and pricing. 11. Following a period of rapid growth in the second half of the 1960s, growth slowed down in the 1970-77 period as a result of a combination of external shocks and inadequate domestic policy responses. Starting in 1977 a new economic strategy was pursued which sought to increase the role of markets and the private sector by reducing restrictions on pricing, investment and external trade and payments. Although growth initially accelerated, difficulties developed as the reform effort slowed down in the early 1980s; in particular, the reform program did not address the restructuring of the overextended public sector. The civil war that began in 1983 and persists to this day also hurt economic performance. Unsustainable macroeconomic imbalances emerged in the second half of the 1980s, and growth was very low during those years. 12. In 1989, a revitalized economic reform program was launched, and growth performance was strong through 1993. The rebound was broad-based with strong growth in industry  - iv - and services, and a rapid expansion of domestic and foreign private investment. Real GDP rose by an average of 5 percent during 1990-92, with significant expansion of exports. In 1993, growth and external sector performance strengthened further. Real GDP is estimated to have grown by 7 percent, mainly as a result of continued strong growth in manufacturing as well as a recovery in agriculture from the 1992 drought. Much of the expansion in output came from relatively labor intensive activities such as garments and the services sector. The recent recovery in agricultural output is also expected to have raised rural employment and incomes. 13. In April/May 1994, however, the government announced a wide-ranging package of economic measures which, if fully implemented, would have a very high fiscal cost. The measures would, in the short-run, raise incomes of many low- and middle-income households, including some that may have been bypassed by the benefits of growth in recent years. At the same time, the package could worsen macroeconomic imbalances significantly, increase inflationary pressures, and severely damage private confidence both domestically and abroad. Thus the new measures could have a large cost in terms of lost GDP growth and a reduced pace of poverty reduction. Effects of the Post-1989 Economic Reform Program on Poverty 14. The post-1989 economic reform program, if it continues to be consistently pursued and deepened, can be expected to speed up the pace of growth and poverty reduction over the medium and long term. The question can be raised, however, as to the short-term effects that such a program may have on the poor. On balance, the post-1989 economic reform program does not appear to have had a negative short-term effect on a majority of the poor, although some specific groups may have been affected. It is useful to distinguish between the aggregate effects of the program and those of specific reforms. Concerning the aggregate effects of the program, it may be noted that per capita consumption in the 1989-93 period grew at an average annual rate of 1.8 percent in real terms, or a total increase of about 7 percent over the four-year period --a modest but significant increase. This implies that, unless there have been offsetting adverse changes in the distribution of consumption, poverty must have continued to decline. The broad-based nature of the economic recovery and the protection of social spending (see below), suggest that the distribution of consumption is unlikely to have been adversely affected during this period. 15. Concerning the effects of specific reforms in the post-1989 program, the major such reforms were those in the areas of fiscal adjustment, privatization and liberalization of the external sector. The fiscal adjustment (a decline in the budget deficit from 15.7 percent of GDP in 1988 to 7.4 percent of GDP in 1992, followed by an increase to 8.1 percent of GDP in 1993) mostly took the form of a reduction in capital expenditure, and the cuts especially affected capital transfers to public corporations and statutory bodies. Such cuts could have adverse effects on long-term growth, but is unlikely that they have had adverse short-term effects on the poor. Another way to look at possible adverse short-term effects on the poor of the fiscal adjustment is to examine the evolution of spending on social services, both current and capital, since 1988. This shows that aggregate spending on social services (including spending on education, health, welfare, housing and community services) did not decline after 1988 as a share of GDP. Such spending amounted to 8 percent of GDP on average for the three-year period 1986-88, and averaged about 8.8 percent of GDP in the five-year period 1989- 93. Hence the fiscal adjustment did not target social services, a particularly sensitive area for the poor. There are, however, some serious efficiency and equity shortcomings affecting spending in social services --not necessarily originating in the period after 1988-- which need to be addressed by the government.  - v - 16. Privatization of State Owned Enterprises (SOEs) has been another major thrust of the post-1989 economic reform program. Divestiture has concentrated on small to medium size enterprises in industry and services. In addition, the government has shifted management of its estates in the plantation sector to 22 private management companies on a profit-sharing basis. The privatization program would tend to raise the rate of growth of the economy by enhancing efficiency, and hence indirectly speed up poverty reduction. In Sri Lanka retrenchment has been voluntary in most cases, i.e., the retrenched employees self-selected after being offered a compensation package; involuntary retrenchment has only taken place in a few cases were SOEs were divested by asset liquidation. In all of the divested SOEs retrenched employees were compensated, though there were several different compensation packages providing uneven benefits. In the tree crop estates, there has been no retrenchment of workers except for a small number of white collar employees. However, there is some concern that health services provided to estate workers and their families may have deteriorated with the transition to private management. 17. The third major area of reform since 1989 has been the reform of the trade regime and adoption of other measures to increase openness of the economy. The pace of trade reform has been fast in the last four years, including the elimination of most non-tariff barriers, the lowering of ad valorem tariffs (with the maximum tariff falling from 100 percent in 1989 to 45 percent in 1994), bringing the minimum tariff up to 10 percent with few exceptions, and reducing discretion in the implementation of tariff rules. Other measures to increase openness have included reversing the policy of heavy regulation of foreign direct investment and liberalizing the external payments system. In the long-term, these reforms can be expected to increase efficiency in production and consumption, increase the flows of foreign direct investment (which is already happening), accelerate growth and speed up poverty reduction. Lowering protection to import-substituting manufacturing could, however, temporarily displace workers from affected firms, thus leading to short-term hardship. The severity of such hardship would partly depend on how quickly displaced workers are able to switch to those sectors where incentives would have improved as a result of the trade reforms, i.e., export industries. This is partly a function of the flexibility of the labor market, where --as already noted-- there is room for improvement. In agriculture, quotas on the importation of several important foodstuffs, i.e., potatoes, onions, chilies, black gram, lentils and maize, were removed with the 1994 budget (however, imports of onions were subsequently banned). While this could cause hardship for a segment of the farming population in the short-term, it should also lead to a decline in the real price of these foodstuffs for consumers. This would be especially beneficial for the poor, for whom food expenditure is a large share of total consumption. B. A Strategy for Future Poverty Alleviation Characteristics of the Poor 18. According to the analysis in this study, the incidence of poverty is higher in rural than in urban areas (24 and 18 percent respectively in 1990-91), and is about the same in all regions with the exception of the Western region (where the capital is located), where it is significantly lower. Thus poverty is not a localized phenomenon, but can be found throughout the country. The study also found that individuals are more likely to be poor if working members of the household are employed in agriculture than in other major sectors of activity; that households where poor individuals live are larger; that the poor are less educated than the non-poor, but the differences are not very pronounced; that the poor have somewhat lower labor force participation rates, which is partly explained by age  - vi - composition; that there is a positive correlation between poverty and unemployment; and that the poor devote a very large proportion of their consumption expenditure to food, with rice alone accounting for 25 percent of their total consumption. It was also found that the incidence of poverty for female- headed households is about the same as for male-headed households (a result also obtained in recent Bank studies on Viet Nam and Indonesia). However, the study did not go deeper into gender issues and the relationship between gender and poverty. This is an area where more research should be undertaken. Concerning the relationship between ethnic origin and poverty, the study found that the incidence of poverty for Sri Lankan Tamils was about the same as for the majority Sinhalese, but this result should be qualified by stressing that our sample for this analysis excluded the North and East. Poverty incidence was estimated to be lower for Indian Tamils than for the majority Sinhalese. However, this does not take into account the fact that calorie needs, and hence minimum required food expenditure, may be greater for Indian Tamils on account of their concentration in the plantations which involves hard physical work. Also, social indicators and social services for Indian Tamils are still poor relative to the rest of the population. 19. These characteristics of the poor, along with the concerns noted earlier, suggest the outlines of a strategy for further poverty reduction. Since poverty is not a phenomenon restricted to a small group in society, but rather it affects more than one-fifth of the total population of the country, broad-based growth at a high and sustained pace will be necessary in order to achieve a rapid decline in poverty. Along with broad-based rapid growth, Sri Lanka would need to address the inefficiencies in social sector spending, improve the quality of education and training and reform labor market regulations. These tasks constitute a major challenge, especially in view of the ongoing civil war and the persisting resource constraints in the public sector. The direct negative impact of the civil conflict on poverty alleviation is obvious--resulting from the severe disruption of economic activity in the conflict areas and a large diversion of resources away from high-priority spending to finance the conflict. The magnitude of this diversion can be seen in the following numbers: in 1973-77, military and police expenditures as a percentage of GDP were just 0.7 percent; by 1988-93, they had reached 3.7 percent of GDP. In addition, the indirect negative effects, in terms of the adverse impact on incentives for long-term private investment and therefore growth, can also be substantial. So, a peaceful resolution of the civil war is of critical importance for sustained poverty reduction. Along with this effort, Sri Lanka will need to continue the implementation of the adjustment program that was resumed in 1989. The strong positive impact of the reforms are already visible from the experience of the past four years. The apparent disruption to the reform program implied by the short- term, politically-motivated package of April/May 1994 needs to be quickly reversed. The reform program needs to be deepened in the areas of agriculture, land and labor markets, and financial sector. At the same time, essential reforms of the social programs--including education, health and the transfer programs-- are necessary to improve service quality, maintain financial sustainability and ensure that the truly poor benefit from these programs. Ensuring Sustained Growth 20. Growth has been high in the 1989-93 period, but as already noted there is a serious danger that it will be derailed by the recently announced package of government spending initiatives. Even before this package was approved, there were concerns about inflationary pressures. To address the problem of inflation, fiscal deficits needed to be reduced further through expenditure cutbacks. So, to ensure that unsustainable macroeconomic imbalances do not develop and inflationary pressures do not accelerate, the new government now in place should reconsider these initiatives and return to the fiscal adjustment course being followed before the new initiatives were announced.  - vii - 21. The new government should also continue to pursue the pending agenda of structural reforms vigorously. This includes completing the rationalization of the trade regime by eliminating remaining quantitative restrictions and moving towards a tariff range with a low average and dispersion; rationalizing the structure of public expenditure (e.g., as discussed in this report for health, education, water supply and transfer programs); deepening the privatization of public enterprises; deepening the deregulation of foreign direct investment; removing financial sector distortions; and improving the efficiency of land and labor markets. 22. These reforms will create the conditions for more rapid sustainable growth of the private sector of the economy, which is the only realistic avenue to poverty reduction for the great majority of the poor. Rapid growth of the private sector, together with the low growth of population expected in the coming years, should lead to progressive tightening of the labor market and lowering of unemployment rates. Given the positive correlation between unemployment and poverty, a tighter labor market should be especially beneficial for the poor. The Key Role of Agriculture 23. In 1992, agriculture accounted for about 26 percent of GDP and over 40 percent of total employment. In the long run, as the economy continues its structural transfornation, it is to be expected that the proportion (and eventually the absolute size) of employment in agriculture will decline while those of manufacturing and services increase. In the more immediate future, however, Sri Lanka will remain to a large extent an agrarian society, and the rate of growth of agriculture will be critical for the speed of poverty alleviation. This is especially so in view of the higher incidence of poverty among agricultural workers, and in rural areas as compared to urban areas. Hence, there is a need for a strong focus on policy and institutional reforms in agriculture. 24. Growth performance in agriculture in recent years has been lackluster due to numerous policy constraints. These constraints will need to be removed. Three areas may be highlighted in this context, namely: reforms towards a better functioning land market; deepening the privatization and deregulation of the tree crops sector; and the further liberalization of agricultural trade and domestic marketing. The latter is important for poverty alleviation not only because it would tend to increase efficiency of resource allocation in agriculture, but also because it would ensure that the poor as consumers can take full advantage of the opportunities afforded by the world food market. Also important is to achieve a better understanding of what public goods and services should be provided by the government to support the growth of private agriculture, and the modalities of such provision. A major organizing principle in confronting this issue is that for agriculture to achieve high sustainable rates of growth it would have to increase its international competitiveness in crops other than rice, since rice can no longer be the engine of agricultural growth (unless productivity in rice production increases so much that the country becomes an efficient exporter of rice). Hence provision of supporting public goods and services must be supportive of that change of emphasis. Human Resource Development 25. The strategy for poverty reduction should also include a strong focus on the education and health sectors, to ensure that the excellent historical performance in these sectors is sustained. This in turn would ensure that the poor have adequate endowments of human capital which would enable them to take advantage of job opportunities opened up by an expanding economy (and, indeed, would make that very expansion feasible).  - viii - 26. In education, the country has achieved high levels of enrollment in primary and secondary education, and this is reflected in the fact that differences in educational achievement between the poor and the non-poor as measured by years of schooling are not large. Thus, for males over ten years of age, the average number of years of schooling in 1990 was 5.9 years for those with consumption below the poverty line, 6.2 years for those with consumption between 100-120 percent of the poverty line, and 7.7 years for those with consumption above 120 percent of the poverty line. A similar pattern prevailed for females. However, there is ample room to improve the quality of the education provided to all children, as well as to improve the equity with which resources are allocated. Improving quality would require reducing expenditures on salaries by reducing teacher numnbers, many of whom are in the cities where they are in surplus, and providing more resources for essential school supplies and materials. Additional measures to improve quality of general education would include emphasis on science and language training, especially English as a second language, a greater effort in teacher training, and revision of the curriculum with an aim to make general education more relevant to the needs of the economy and society. Improving the equity of access to good quality general education, which would be strongly pro-poor, would require a reduction in the existing disparities in resources among schools, both within and across provinces. 27. In health, the government health services are faced with the challenge of a changing population structure and consequently changing disease pattern. There are also several critical problems concerning the composition of public health expenditure. These include insufficient attention to preventive services (reflected in the apparent rise in incidence of infectious and parasitic diseases), unbalanced functional distribution of current expenditures (with too much being absorbed by higher- level health facilities), declining ratio of current expenditure to capital stock, and insufficient expenditure on repairs and maintenance of health facilities. Confronting these problems will be key to ensure that the poor, and others using the public health facilities, have access to effective health services. The government should also consider taking over health services in the estates, in order to ensure that the estate population has access to the same quality of services as the rest of the population. 28. Addressing the above efficiency and equity issues should be the first priority in health and education. Once sufficient progress in this respect has been achieved, it may make sense for Sri Lanka to set a goal of increasing public expenditure in health and education as a percentage of GDP. Public expenditure in education as a percentage of GDP is low compared to a number of other Asian countries such as China, India, Indonesia, Korea, Thailand and Malaysia. Public expenditure in health is only about average for the region. In both cases, public expenditure as a percentage of GDP is lower now than in the 1970s. 29. A strategy to address undernutrition in young children would need to be formulated and it should be broad-based. Preventive health services should be strengthened since infectious and parasitic diseases are a major contributory factor to undernutrition in young children. This should be complemented by a strong program of nutrition and hygiene education, and by expanding access to safe water and adequate sanitation facilities. Nutrition supplementation, as in the Triposha program, may also have a role in a broad strategy, but it is probably less critical than the previously mentioned interventions. 30. Substantial progress was made in the 1980s towards the goal of achieving universal coverage of safe water and sanitation facilities, and by 1990 about two-thirds of the population had access to some type of formal water supply source, and close to 60 percent had access to satisfactory sanitation facilities. There are, however, significant disparities in coverage among provinces and between urban and rural areas, and the current investment program of the National Water Supply and  - ix - Drainage Board would reinforce these disparities rather than reduce them. Hence, the government needs to reassess its expenditure priorities in this area. Transfers, Safety Nets and NGOs 31. The final component of a poverty alleviation strategy for Sri Lanka concerns the strengthening of its safety net. This, in turn, has two components. The first is to ensure that public transfer programs intended to assist the poor do indeed help those who need it the most. The second is to ensure that a supportive environment exists for the activities of private non-profit organizations, which make an important contribution to the country's safety net. 32. Public Transfer Programs. The government appears to be meeting the needs of the population displaced by the civil conflict reasonably well, although some improvements may be possible in terms of targeting and the provision of water and sanitation facilities in the camps. The real solution to the problems of these households is, of course, to end the conflict and resettle them back in their places of origin. On the other hand, present programs addressing the needs of the destitute poor --defined in this report as poor households with household heads who are disabled, elderly or widows; destitute orphans; destitute elderly; and the destitute disabled-- appear to be severely underfunded and in disarray. The central government should conduct an in-depth review of these programs, jointly with the Provincial Councils, in order to formulate a plan of action to provide the programs with more adequate resources and management. 33. Meeting the basic needs of the destitute poor should be the number one priority of public transfer programs. After this priority is met, and to the extent that the availability of fiscal resources may permit, government assistance in the form of income transfers should be concentrated on the poorest 20-30 percent of households with able-bodied members (with periodic re-screening and flexible mechanisms for exit of households that are no longer eligible as well as for entry of newly eligible households). That such better targeting is feasible at low cost has been shown by the relative success of community-based targeting methods employed by the Janasaviya Program (JSP). The government should consider consolidating the JSP and Food Stamp/Mid-Day Meal programs into a restructured income transfer program using community-based targeting procedures. A package of reforms could be designed along these lines that would result in greater poverty alleviation without increasing public transfers as a percentage of GDP. This would require further work on targeting procedures, in order to refine the approaches followed under the JSP and ensure that benefits are concentrated on those most in need of assistance. 34. Non-Governmental Organizations. Although little hard data are available, NGOs are a key component of the safety net in Sri Lanka, along with public and private transfers. They could play a much larger role. They number in the tens of thousands, and are involved in a rich variety of activities with an emphasis on the provision of social welfare services. A large share of their resources originate in foreign donations, and they are constrained by the availability of resources and of managerial and technical skills. 35. Until recently, the government had followed a laissez faire policy with regard to regulation of NGO activities. In December 1993, however, new regulations were passed that impose burdensome requirements on NGOs and specify stiff penalties for non-compliance. The government should reconsider the new regulations, and in doing so it should carefully weigh the costs and benefits to society of the regulations. In addition, there are several positive steps the government could take to improve the supportiveness of the policy and institutional environment for NGO work. These include the development of an information base; the establishment of a central unit to be the focal point for  - x - coordination with NGOs; and the creation of NGO panels in selected ministries and government agencies. 36. A new institutional mechanism for government-NGO cooperation in helping the poor is the Janasaviya Trust Fund (JTF), established in 1991. Its basic concept is that of an apex agency which receives funds from the government and donors and disburses those funds for specified purposes to Partner Organizations (mostly NGOs), which implement programs that target the poor, with their active participation. Current programs include a Credit Fund, a Rural Works Fund, and a Nutrition Fund. The concept is similar to several programs of "social investment funds" that have been operating successfully in Latin America. In Sri Lanka, the JTF initially deviated from its original concept and became too involved in the direct administration of activities it funded. More recently, however, the JTF has reverted to its original concept of an apex agency working with a broad ensemble of NGOs and other Partner Organizations. This is a very positive development. Careful monitoring of JTF's operations will be needed to ensure it achieves its objectives as originally conceived. Directions of Future Work 37. As noted in the opening paragraph of this Executive Summary, there is a rich literature on poverty and related topics as they pertain to Sri Lanka. One topic that has not been explored in detail and deserves further attention is that of the relationship between gender and poverty. Another unexplored topic is that of private inter-household transfers, which have been found in other countries to be a very important part of society's safety net, and which may be substitutable with public transfers to a significant degree. More work is also needed in order to refine the targeting mechanisms for the income transfer programs. 38 It is also important that the government start monitoring poverty indicators on a regular basis. This could be done on an annual basis now, since starting in 1993 the Department of Census and Statistics has added a household income and expenditure module to its quarterly labor force survey for the last quarter of the year. Thus the government could now define an official poverty line and proceed to monitor poverty indicators (incidence and other indicators) on an annual basis. Such annual monitoring would be a good complement to the annual reporting of national income growth and other results of the national accounts. The poverty indicators would give an indication to the public as to how widely the annual gains from growth are being shared among the population.  CHAPTER I - POVERTY TRENDS 1.01 The first three sections of this chapter discuss trends in consumption poverty in Sri Lanka since the early 1950s'. First, the trends in the 1953-85 period are examined, using data on private consumption from the national accounts, population estimates, official price indices and data on household income distribution from several household surveys. Then, a more rigorous analysis is conducted of the 1985-90 period, using data from two comparable household surveys conducted in 1985/86 and 1990/91. This is followed by a discussion of the 1990-93 period. 1.02 The main results of the analysis of consumption poverty are: (i) There was a substantial decline in the incidence of consumption poverty over the period 1953-90. (ii) The decline was particularly rapid during the 1965-85 period, when a rapid growth in per capita income facilitated a rapid growth in per capita consumption. (iii) In 1990-93, given the broad-based, rapid growth in GDP per capita, and lower but still significant growth in per capita consumption (despite substantial macroeconomic adjustment), coupled with protection of social spending and increasing real wages, it is likely that poverty declined further. A. Consumption Poverty in the 1953-85 Period 1.03 To assess trends in consumption poverty rigorously, it would be necessary to use household data from a number of comparable surveys spread out over the period under analysis. Applying the methodologies that have been developed in recent years, with a poverty line held constant over time in terms of real purchasing power, it would be possible in theory to generate a series of points in a time-poverty incidence space, and then adjust a trend line using regression techniques. 1.04 Such an approach is however very demanding in terms of data and resources. Moreover, there could be problems of comparability among the various surveys. In the case of Sri Lanka, the Department of Census and Statistics (DCS) has undertaken four household surveys, i.e., the Socio-Economic Surveys of 1969/70, 1980/81, 1985/86 and 1990/91. The latter two surveys are used later in this chapter for the analysis of the 1985-90 period. There is also a series of household surveys conducted by the Central Bank; these are the Consumer Finances and Socio-Economic Surveys of 1953, 1963, 1973, 1978/79, 1981/82 and 1986/87. The surveys in the two series are very different in design and probably not comparable. As noted by Michael Lipton and Martin Ravallion in their paper Poverty and Policy (World Bank, Policy research Working Paper No. 1130, April 1993), a suitably comprehensive measure of current consumer spending on all goods and services is generally preferred to income as a measure of current living standards in developing countries. There are two reasons for this. First, current consumption is generally considered to be a better indicator than current income of current standard of living; it is assumed that instantaneous utility depends directly on consumption, not on income as such. Second, current consumption may also be a good indicator of long-term well-being, as it will reveal information about incomes at other dates in the past and future. This is because incomes (including those of the poor) often vary over time in fairly predictable ways, and in such circumstances there are typically consumption-smoothing opportunities available to the poor -- such as through saving and community-based risk sharing.  1.05 In this section, we examine the long-run trends in consumption poverty incidence in Sri Lanka in the period since 1953 to 1985 using a less rigorous approach, which nevertheless appears to be robust, and which clearly points out to a significant decline in consumption poverty over this period. The approach simply combines data on the evolution over time of private consumption per capita in real terms with published household income distribution data, which are available from the series of Central Bank surveys. We first consider the household income distribution data. Trends in the Distribution of Household Income 1953-1986/87 1.06 Table 1 . I below shows the evolution over time of the distribution by decile of the total one-month income of spending units (households) as estimated from the various Central Bank surveys. Of special interest for the present study of poverty is the evolution over time of the share of the bottom half of the household income distribution. The data in Table 1.1 show that, with the exception of 1973, the share in question, as estimated from the various Central Bank surveys, has moved within a fairly narrow range of about 20-23%. Household income distribution in Sri Lanka apparently became less skewed between 1963 and 1973, but later returned to the pattern observed in the 1950s and 1960s. For the entire period 1953-87, the shares of the lowest three, four and five deciles at the end of the period appear to have been about the same as at the beginning of the period. Table 1.1: Percentage of Total One Month Pre-tax Income Received By Each Ten Percent of Ranked Spending Units, 1953 to 1986/87 Decile 1953 1963 1973 1978/79 1981/82 1986/87 Lowest 1.90 1.50 2.79 2.61 2.19 1.93 2nd 3.30 2.95 4.38 3.65 3.53 3.13 3rd 4.10 4.00 5.60 4.68 4.31 4.09 4th 5.20 5.21 6.52 5.70 5.22 4.99 5th 6.40 6.27 7.45 6.62 6.32 6.01 6th 6.90 7.54 8.75 7,17 7.03 7.37 7th 8.30 9.00 9.91 9.12 8.66 8.69 8th 10.10 11.22 11.65 11.37 10.74 11.40 9th 13.20 15.54 14.92 14.31 14.61 14.90 Highest 40.60 36.77 28.03 34.63 37.35 37.40 Three Lowest Deciles Combined 9.30 8.45 12.77 10.94 10.03 9.15 Four Lowest Deciles Combined 14.5 13.66 19.29 16.64 15.25 14.14 Five Lowest Deciles Combined 20.9 19.93 26.74 23.26 21.57 20.15 Mean Monthly Income per Spending Unit (Rs.) Current Prices 169 193 311 917 1,652 2,728 Constant 1952 Prices a/ 166 177 188 382 417 433 Consumer Price Index 101.6 108.8 165.4 240.0 395.7 629.4 Notes: [ Deflated by the Colombo Consumer Price Index, 1952 = 100. Sources: Report on Consumer Finances and Socio Economic Survey 1981/82, Central Bank of Sri Lanka, for 1953, 1963, 1973, 1978/79 and 1981/82; Report on Consumer Finances and Socio Economic Survey 1986/87, Central Bank. Information for 1986/87 provided directly by the Central Bank and excludes parts of the Northem and Eastern provinces. Information for previous years pertains to the entire country.  - 3 - International Comparisons of the Distribution of Consumption. How does Sri Lanka compare with other countries in terms of the skewness of its consumption distribution? In order to address this question, the present study computed the Gini coefficient of the distribution of consumption expenditure using data from the 1985/86 Labor Force and Socioeconomic Survey and the 1990/91 Income and Expenditure Survey. To do this, households were ranked according to per capita expenditure. The Gini coefficient can vary between zero and one, and the higher the value of the coefficient the more unequal the distribution. The results obtained for Sri Lanka were Gini coefficients of about 0.32 in 1985/86 and 0.30 in 1990/91. These results can be compared with similar data from a number of other developing countries compiled in a recent Bank publication (Is Poverty Increasing in the Developing World?, by Shaohua Chen, Gaurav Datt and Martin Ravallion, WPS 1146, World Bank, June 1993). The comparison shows that Sri Lanka's Gini coefficient of consumption of between 0.30-0.32 (estimated using per capita consumption as the ranking variable) is very similar to those of other major South Asian countries in recent years (estimated on the same basis): India, 0.32 in 1989/90; Pakistan, 0.31 in 1991; and Bangladesh, 0.29 in 1988/89. A broader comparison using the same comparator data (and same ranking variable) indicates that these South Asian Gini coefficients are relatively good (i.e., reflect a relatively egalitarian distribution) by developing country standards. For example, they compare favorably with consumption Gini coefficients in the following countries: Indonesia (1990) 0.33 Philippines (1988) 0.40 Thailand (1988) 0.44 Bolivia (1990) 0.42 Peru (1985/86) 0.46 Algeria (1988) 0.39 Morocco (1990/91) 0.39 Tunisia (1990) 0.40 Code d'lvoire (1988) 0.34 Ghana (1988/89) 0.37 Tanzania (1991) 0.59 Uganda (1989/90) 0.33 Zimbabwe (1990) 0.57 Box I Trends in Consumption Per Capita 1953-1985 1.07 Next we look at the evolution of consumption per capita in the 1953-85 period, using national accounts figures on total consumption and official mid-year population estimates. National accounts figures are available starting in 1950. Table 1.2 below shows figures for selected years during this period (1993 is also shown). It can be seen from the table that over the entire period there was a large increase in private consumption per capita in constant prices. When consumption per capita is deflated using the CPI, it appears to have more than tripled from 1953 to 1985, from an index of 30.6 in 1953 to 100 in 1985. Even if consumption per capita is deflated by the GDP deflator, which increased at a faster rate over the period, we still get a very significant increase in consumption per capita in constant prices over the entire period --it more than doubles from an index of 43.1 in 1953 to 100 in 1985. Deflating consumption by the CPI rather than by the GDP deflator is  Table 1.2: TRENDS IN PRIVATE CONSUMPTION PER CAPITA 1953-1993 1953 1963 1965 1970 1973 1977 1978 1979 1985 1993 GDP in Current Prices, Rs. Million 4717 7382 8084 13664 18404 36407 42665 52387 162375 498700 Private Consumption in Current Prices, Rs. Million 3670 5337 5877 9879 14079 26686 32084 40390 126490 369999 Ratio of Private Consumption/GDP 0.778 0.723 0.727 0.723 0.765 0.733 0.752 0.771 0.779 0.742 Population at Mid-Year, 000s 8290 10651 11133 12516 13091 13492 14184 14471 15837 17560 Private Consumption Per Capita, Current Prices, Per Month, Rs. 37 42 44 66 90 160 188 233.00 666 1756 CPI (1985 - 100) 18.1 19.4 20 24.6 29.5 36.2 40.6 45 100 250.9 GDP Deflator (1985 = 100) 12.9 14.6 14.7 17 20.5 34.4 37.7 43.5 100 221.6 Private Consumption Per Capita, p. 1985 Prices, Deflated by CPI, Per Month, Rs 204 216 220 268 305 442 463 518 666 700 Index 1985 = 100 30.6 32.4 33 40.2 45.8 66.4 69.5 77.8 100 105.1 Private Consumption Per Capita, 1985 Prices, Deflated by GDP Deflator, Per Month, Rs. 287 288 299 388 439 464 499 536 666 792 Index 1985 = 100 43.1 43.2 44.9 58.2 65.9 69.8 74.9 80.5 100 118.9 GDP Per Capita in Constant 1985 Prices, Deflated by GDP Deflator, Per Year, Rs. 4409 4756 4949 6420 6847 7586 7978 8319 10253 12816 Index 1985 = 100 43 46.4 48.3 62.6 66.8 74 77.8 81.1 100 125 Sources: National Accounts; Official Population Figures; Department of Census and Statistics for CPI figures  - 5 - the usual procedure; divergences between the two indices are supposed to capture movements in the prices of goods and services entering domestic consumption relative to all goods and services in the GDP. However, in the Sri Lankan case, the existence of price controls during this period have tended to distort the CPI figures, since controlled prices were used in computing the CPI. This was especially so in the 1970s, which is also the period when the widest divergence between the evolution of the CPI and that of the GDP deflator can be observed. 1.08 As shown in Table 1.3 below, the pace of growth in per capita consumption was not uniform during the period. From 1953 to 1965 consumption per capita in constant prices deflated by the CPI grew at an annual rate of only 0.6 percent, or 0.4 percent if deflated by the GDP deflator. From 1965 to 1985 consumption per capita grew much faster, at an annual average rate of 5.7 percent if deflated by the CPI, or 4.1 percent if deflated by the GDP deflator; in either case, a remarkable rate of growth. From 1985 to 1990, on the other hand, there was no growth in consumption per capita when the CPI is used as deflator, and very modest growth (1.4 percent per year) if the GDP deflator is used instead. Growth in GDP and consumption per capita picked up considerably in the 1990-93 period (this is further discussed in Chapter II). Table 1.3: Annual Growth of Private Consumption and GDP Per Capita, 1953-93 (Percentaee) Consumption Consumption GDP Per Capita Per Capita Per Capita (CPI Deflator) (GDP Deflator) (GDP Deflator) 1953-1965 0.6 0.4 1.0 1965-1970 4.0 5.3 5.3 1970-1977 7.4 2.6 2.4 1977-1985 5.3 4.6 3.8 1985-1990 0.0 1.4 1.9 1990-1993 1.7 2.3 4.2 Source: Calculated from Table 1.2. Trends in Consumption Poverty 1953-85 1.09 We can now combine the data in the two previous sections to assess the trends in consumption poverty incidence in the 1953-85 period. Table 1. I shows that the income shares of the five lowest deciles of the household income distribution were roughly the same in 1986/87 and 1953. If the household consumption distribution had approximately followed the evolution of the household income distribution, which is a reasonable assumption, the consumption shares of the five lowest deciles of the household consumption distribution in 1986/87 would have been close to the corresponding 1953 shares. Since, overall, private consumption per capita in real terms increased by a factor of between 2.3 (using the GDP deflator) and 3.2 (using the CPI as deflator) between 1953 and 1985, per capita consumption of households in these lower deciles of the household consumption distribution in 1985 must have been two to three times higher than the per capita consumption of households which were similarly positioned in the 1953 household consumption distribution.2 This in Note that if the combined share of total consumption accruing to the five lowest deciles of the household consumption distribution had been unchanged between the two years, and the proportion of the total population contained by these five lowest deciles had also been unchanged, then the proportional increase in per capita consumption for the five lowest deciles  - 6 - turn implies that there must have been a very significant decline in consumption poverty incidence in Sri Lanka in the 1953-85 period. The magnitude of the decline in incidence (and other poverty measures) would vary depending on the precise definition of the poverty line (which would, of course, have to be held constant in purchasing power over the period), but it would undoubtedly be quite significant. B. Consumption Poverty in the 1985-90 Period 1.10 We now turn to the 1985-1990 period, for which we have conducted a more rigorous analysis of the incidence and depth of poverty based on data from two recent household surveys, viz., the Labor Force and Socio-economic Survey (LFSS) of 1985-86 and the Household Income and Expenditure Survey (HIES) of 1990-91 conducted by the DCS. Our main aim will be to quantify and compare the magnitude of absolute poverty over the two survey dates. In doing this, we will use per capita consumption expenditure (excluding expenditure on durables) as our indicator of individual standard of living. The details of our methodology are explained in Annex 2. We will use three poverty measures in our analysis: (i) The headcount index, which is simply the proportion of the population for whom consumption is less than the poverty line; this can also be referred to as the poverty incidence. (ii) The poverty gap index, which is equal to the sum of the distances between the poverty line and the consumption levels of the poor, as a proportion of the poverty line, divided by the total population (or, in other words, the mean proportionate poverty gap in the population, where the non-poor have zero poverty gap). This measure is an indicator of the "depth" of poverty, as opposed to simple incidence as indicated by the headcount index. (iii) The squared poverty gap index, which, unlike the previous two measures, captures the severity of poverty among the poor, by giving greater weight to individuals among the poor the farther their consumption is from the poverty line.' The Results: Poverty in 1990-91 1.11 Our estimates of poverty in Sri Lanka are shown in Table 1.4 below. By our reference poverty line, about 22.4 per cent of the population, or about 3.8 million persons, are deemed to be poor in 1990-91.4 Using a more generous (20 percent higher) poverty line, the fraction deemed poor rises to 35.3 per cent, and the number of poor to about 6 million. The poverty gap indices for the two poverty lines are 4.8 and 8.8 per cent, implying an average poverty deficit of the between the two years would have been exactly the same as the proportional increase in overall per capita consumption. A good discussion of poverty measures is contained in PovertY Comparisons, by Martin Ravallion, unpublished monograph, World Bank 1994. This is using an estimated total Sri Lankan population of about 17 million for 1990-91. The estimated population of the regions covered by the two surveys was about 14 million in 1990- 91, and thus for these regions the number in poverty (by the reference poverty line) is estimated at about 3.1 million in 1990-91.  - 7 - poor (the proportionate shortfall of their average consumption from the poverty line) of 22 and 25 per cent respectively for the reference and higher poverty lines. 1.12 Rural/Urban/Estate Differentials. The incidence of poverty in 1990-91, as measured by the headcount index for the reference poverty line, is the greatest in the rural sector (24.4 per cent), followed by the urban sector (18.3 per cent), and is the least in the estate sector (12.6 per cent). Other poverty measures show a similar pattern. These differences are statistically significant. The ranking of the three sectors remains unchanged for the higher poverty line. The sectoral poverty rankings depend both on differences in mean per capita consumption and relative inequalities within sectors. The latter are depicted in Figure 7, Annex 3, which shows that inequality is lowest in the estate sector and the highest in the urban sector, with the rural sector occupying an internediate position. Table 1.4: Poverty in Sri Lanka by sector: 1985-86 and 1990-91 1985-86 1990-91 Headcount Poverty gap Squared Headcount Poverty gap Squared index index poverty gap index index poverty gap index index Reference poverty line: Rs. 471.20/person/month * Rural 31.67 7.67 2.75 24.41 5.27 1.78 (0.86) (0.27) (0.13) (0.82) (0.23) (0.1 1) Urban 16.43 3.48 1.11 18.31 4.14 1.37 (0.93) (0.25) (0. 1) (0.98) (0.28) (0.12) Estate 14.31 3.81 1.37 12.62 2.11 0.60 (1.95) (0.62) (0.27) (1.89) (0.43) (0.17) Sri Lanka 27.33 6.54 2.31 22.36 4.82 1.62 (0.64) (0.20) (0.09) (0.61) (0.17) (0.08) Higher poverty line: Rs. 565.44/person/month * Rural 45.48 12.81 5.04 38.05 9.55 3.49 (0.92) (0.34) (0.18) (0.92) (0.31) (0.15) Urban 26.78 6.52 2.29 28.43 7.34 2.70 (1.11) (0.34) (0.16) (1.14) (0.37) (0.18) Estate 30.85 6.79 2.54 27.51 5.17 1.50 (2.57) (0.80) (0.40) (2.55) (0.63) (0.27) Sri Lanka 40.60 11.09 4.30 35.34 8.80 3.20 (0.71) (0.25) (0.13) (0.70) (0.23) (0.11) Note: * at 1990-91 Sri Lanka prices. All poverty measures are in percentages. The numbers in parentheses are the standard errors of poverty measures, computed using results in Kakwani (1990). Data exclude the North and East, and refer to the first three rounds of the 1990/91 survey and same months of the 1985/86 survey. 1.13 The estimates show that poverty in Sri Lanka is predominantly a rural phenomenon; the rural sector accounts for about four-fifths of aggregate poverty. The shares of the rural, urban and estate sectors in the total number of the poor (by the reference poverty line) are 79, 17 and 4 per cent respectively. The relative contributions of these sectors to national poverty are largely invariant over different poverty measures and poverty lines.  Change in Poverty Between 1985-86 and 1990-91 1.14 Table 1.4 also shows that, country-wide, the data indicate a statistically significant5 decline in absolute poverty over the two surveys. For the headcount index using the reference poverty line, a decline of about 18 percent is indicated, from 27.3 percent of the population in 1985- 86 to 22.4 percent in 1990-91. The decline in the absolute number of the poor is of course more modest, by about 12 percent, from 4.3 to 3.8 million. The decline in the poverty gap and the squared poverty gap measures is even greater (by about 26 and 30 percent respectively), indicating a significant decline in both the depth and severity of poverty. The percentage decline in the poverty measures is found to be lower for the higher poverty line; for this line, the headcount, poverty gap and squared poverty gap indices declined by about 13, 21 and 26 percent respectively. This suggests that ultra poverty declined somewhat more sharply than poverty in general during this period. I.15 It should be stressed, however, that because of lack of data our calculations exclude the North and East provinces. Because of the civil war, which especially affected this part of the country, it is likely that poverty increased in the period between the two surveys in these provinces. Ilence, if we could include the two provinces in the calculations, the country-wide decline in poverty between the two survey periods would probably be less than indicated in the previous paragraph. The share of the North and East provinces in total population was about 17 percent in 1990, according to DCS estimates. 1.16 Table 1.5 below shows the evolution of consumption per capita in constant 1990-91 prices and of the Gini coefficient (of consumption) between the two survey periods, 1985-86 and 1990-91, in the aggregate and also broken down by place of residence (rural/urban/estate). It can be seen from the table that this was a period of very slow growth of private consumption per capita at the national level; for the entire period the total increase was only about 1.9 percent (or at an annual rate of 0.4 percent). At the same time, there was a significant improvement in the nationwide Gini coefficient of consumption, from 32.0 in 1985-86 to 29.7 in 1990-91. The modest increase in consumption per capita and the improvement in distribution combined to produce a significant decline in poverty over this particular five-year period. Table 1.5: Population shares, mean consumption and Ginis: 1985-86 and 1990-91 1985-86 1990-91 Population share Mean Gini Population Mean Gini consumption coefficient share consumption coefficient (per (Rs./person/mo (per cent) (Rs./person/mo cent) nth)* nth) * Rural 72.46 708.29 29.87 72.50 743.64 27.57 (2919) (2769) Urban 20.81 1038.47 35.68 20.86 990.10 35.39 (1604) (1573) Fstate 6.73 763.69 24.49 6.64 749.93 20.17 (324) (308) Sri Lanka 100 780.73 32.04 100 795.48 29.66 (4847) (4650) Note: * at 1990-91 Sri Lanka prices. The number of sample households is given in parentheses. The data exclude the North and East provinces, and refer to the first three rounds of the 1990/91 survey and same months of the 1985/86 survey. The standard errors of poverty measures are based on formulae in Kakwani (1990).  1.17 The change in poverty between the two survey periods is uneven for rural, estate and urban areas. This can be seen by looking at the estimates for the reference poverty line; the pattern is similar for the higher line (Table 1.4). The incidence of poverty declined substantially in the rural sector by 23 percent, more moderately in the estate sector by 12 percent, but in the urban sector it increased by 11 percent. The pattern changes somewhat for other poverty measures. For example, the estate sector shows the greatest decline in the severity of poverty (by over 50 percent). I.18 The uneven decline in poverty by place of residence between the two survey periods can be "explained" by the different evolution of per capita consumption and Gini coefficients for the three residence categories. For rural residents, there was an increase in per capita consumption in constant prices between the two survey periods, of about 5 percent, and also an improvement of more than two points in the Gini coefficient of consumption (Table 1.5). For urban residents, by contrast, there was a decline of near 5 percent in per capita consumption, and hardly any change in the Gini coefficient; a combination that led to an increase in urban poverty between the two survey periods. For estate residents, there was a 2 percent decline in per capita consumption, but this was more than compensated (in relation to its impact on poverty) by a very large improvement in the Gini coefficient, of over four points. 1.19 This intuitive interpretation is confirmed by formal decomposition of the changes in poverty in the 1985-86 to 1990-91 period into a growth component and a redistribution component, following the procedure developed by Datt and Ravallion (1992). The results of this decomposition (given in Tables 5-7, Annex 3) indicate that for the rural sector the observed decline in poverty is roughly evenly split between the growth and the redistribution components, with a somewhat higher contribution of the latter. In the urban sector, the observed increase in urban poverty is fully accounted for by the decline in mean consumption. Further details are given in Annex 2.6 Changes in Poverty between 1985 and 1990 and the National Accounts 1.20 It was noted above that Table 1.5 shows that, in the aggregate, consumption per capita in constant prices, as calculated from the two household surveys, was barely 2 percent higher in 1990- 91 than in 1985-86 (using the national-level "General Price Indices" in Table 1, Annex 3, to express both figures in constant 1990-91 prices), i.e., Rs. 795.5 in 1990-91 versus Rs. 780.7 in 1985-86. These figures are not strictly comparable to national account consumption figures, because the latter refer to the entire calendar year while the survey figures are monthly averages for the three-month period of June-July-August (of 1985 and 1990 respectively); also, the survey data excludes 8 districts. However, a calculation from national accounts figures shows consistency with the consumption trends obtained from survey data. Using the CPI to put national accounts private consumption figures for 1985 and 1990 in constant prices, and official population figures, we obtain the result that private per capita consumption in 1990 for the country as a whole was practically identical to the corresponding figure for 1985.7 (It may be noted that between 1985 and 1990 the CPI grew faster than the implicit GDP deflator, i.e., 12.5 percent versus 10.6 percent per annum). The robustness of the ordinal results obtained for the comparison between poverty measures for urban/rural/estate "sectors", as well as for the change in poverty over the 1985-86 to 1990- 91 period, is generally confirmed by dominance analysis. For details, see Annex 2. 7' The figures for the calculation are as follows: Private Consumption in Current Prices in Rs. million: 126,503 in 1985 and 244,288 in 1990; CPI, base 1952=100: 561.2 in 1985 and 1,008.6 in 1990; population in mid-year, in thousands: 15,837 in 1985 and 16,993 in 1990.  - 10 - 1.21 As pointed out above, the survey data in Table 1.5 shows that while monthly private consumption per person in rural areas increased by 5 percent between the two survey periods, in urban areas it declined by about 4.7 percent. This divergent evolution between the two survey periods of urban and rural per capita consumption estimated from survey data cannot be directly checked against national accounts data because the latter are not broken down by rural/urban residence. The result does seem rather counter-intuitive, since there is a general perception in Sri Lanka that economic activity in the 1980s accelerated in urban areas to a greater extent than in rural areas. Value added in agriculture in 1990 was about 6 percent higher than in 1985 in constant prices, or slightly lower than the growth in rural population in the period.8 Total GDP, on the other hand, was about 18 percent higher in 1990 than in 1985; but it should be kept in mind that agriculture is not the sole activity in rural areas, and furthernore that some rural residents could be commuting to jobs in urban areas. 1.22 The data suggest that non-agriculture activities in rural areas grew faster than agriculture in the 1985 to 1990 period. The survey data in Table 10, Annex 3, indicate that the share of the total (rural and non-rural) working population employed in agriculture declined from 45.1 percent in 1985-86 to 40.5 percent in 1990-91; at the same time, the share of rural (including estate) population over total population remained constant (Table 1.5). The combination of these data suggest that over the period the working rural population diversified out of agricultural employment to a significant extent. Another implication is that value added per person employed in agriculture must have increased (since, as already noted, the rate of growth of agricultural value added in 1985-90 was about the same as the rate of growth of the rural population, and since employment in agriculture grew at a slower rate than the rural population). This seems consistent with the fact that from 1984 to 1990 real wages in agriculture appear to have increased in real terms (Table 1.8). It is also consistent with the finding, reported below, that poverty for individuals working in the agriculture sector (and their dependents) declined between the two survey periods. To summarize, then, what appears to have happened in the 1985-90 period is that, while agricultural value added grew at a slower rate than the economy as a whole, and at about the same rate as total rural population, workers were moving out of agriculture during the period, and those who remained in agriculture increased their per capita contribution to value added and their wages. 1.23 It is also possible that there have been changes in the pattem of transfers (private as well as public) between the two survey periods which relatively favored rural residents, as well as possible differential changes in the propensity to save (and consume) income for rural and urban residents. These possibilities were not explored. It is also not clear why the distribution of consumption would have improved between the two survey periods in rural areas but not in urban areas. C. Consumption Poverty in the 1990-93 Period 1.24 In the 1990-93 period, total private consumption in constant prices (deflated by the CPI9) increased at an annual average rate of 2.8 percent, which translates into a per capita annual rate National population grew by about 7% between 1985 and 1990, and the figures in Table 1.5 show almost identical rural population shares in the two survey periods, 1985-86 and 1990-91. 2' The increase in the CPI over this three-year period was somewhat higher than the increase in the GDP deflator. The annual percentage changes in the CPI were: 12.2% in 1991; 11.4% in 1992; and 11.7% in 1993. The corresponding figures for the GDP deflator were: 11.8%, 9.9% and 11.5%.  - 11 - of about 1.7 percent'0. The annual growth of total real private consumption was only about half the growth of real GDP in the 1990-93 period: 2.8 percent versus 5.3 percent. The difference is accounted for by the faster growth in the CPI as compared to the GDP deflator, and also by a decrease in the ratio of private consumption to GDP, from 75.9 percent in 1990 to 74.2 percent in 1993. 1.25 The annual rate of growth of per capita consumption in 1990-93, of 1.7 percent, is of course a considerable improvement over the approximately zero rate of the 1985-90 period, although it is modest by comparison to the consumption growth achieved in the 1965-85 period. More importantly, it represents an important achievement if one takes into account the substantial macroeconomic adjustment that was implemented in this period. 1.26 What pace of consumption poverty reduction would have been observed in the 1990- 93 period with an annual rate of growth of per capita consumption of about 1.7 percent? The answer, of course, depends on what happened to the distribution of consumption. If there had been no changes in that distribution, an annual rate of growth of per capita consumption of 2 percent would lead to a decline in the incidence of poverty nationwide from 22.4 percent in 1990 to 16.2 percent in 1995. This is shown in Table 1.6 below. The assumption that the distribution of consumption may not have deteriorated in the 1990-93 period appears to be reasonable in view of the broad-based nature of growth in this period. In the first place, the manufacturing sector played the most dynamic role; much of this resulted from a surge in labor-intensive garment exports. Secondly, the services sector, which is also relatively labor-intensive in Sri Lanka, also played a substantial role in raising the growth rate. Finally, although agriculture generally lagged behind, there was a recovery in this sector in 1993. There is also evidence that real wages of unskilled workers went up between 1990 and 1993 (see Table 1.8). The policy changes in the 1990-93 period and their implications for poverty are discussed in greater detail in Chapter II. 1.27 The importance of achieving a higher rate of economic growth for consumption poverty reduction can be further illustrated by looking at the simulations in Table 1.6. These are simulations of the decline in poverty incidence that would occur in the 1990-2010 period under different rates of growth of per capita consumption in real terms, taking the poverty situation in 1990 as the base. It must be stressed, however, that in reality changes in the distribution of consumption could cause either a faster or a slower reduction in poverty incidence. This indicates the need to ensure that growth is broad-based, with emphasis on relatively labor-intensive activities. L0/ The 1993 World Development Report projects that Sri Lanka's population would grow at an average annual rate of 1.1% in the 1991-2000 period.  - 12 - Table 1.6: Simulations of Economic Growth and Poverty Incidence Incidence of Poverty (Headcount Index) (%) Altemative Assumed Annual Real Rates of Growth of 1990 1995 2000 2010 per Capita Private Consumption (base year) 2% 22.4a' 16.2 11.5 5.5 3% 22.4a/ 13.8 7.5 2.7 4% 22.4a' 11.5 5.6 0.9 5% 22.4" 9.3 3.8 0.2 '' From Table 1.4 Source: Mission Estimates. The calculations assume that the distribution of private consumption remains identical to that of the base year throughout the projection period. The poverty line used in the simulations is the reference poverty line defined in this chapter. D. Sources of Income Growth 1.28 The previous sections examined trends in consumption poverty using data on growth of value added and consumption from the national accounts, and on household consumption expenditure and income distribution from household surveys. This section examines the complementary question of what have been the sources of income growth for poor households. 1.29 Poor income earners can be classified into two broad groups: (i) those who derive their incomes from paid employment, i.e., wage earners; and (ii) those who are self-employed, i.e., micro-entrepreneurs. In Sri Lanka, the poverty literature suggests that in the post-Independence period the main groups in (i) have been landless laborers in (non-estate) rural areas, estate workers, workers in small enterprises in the informal sectors of manufacturing and services, and household servants. The main groups in (ii), on the other hand, have included small farmers (whether they own the land they till or are tenants), small-scale traders, and self-employed providers of certain services such as carpenters and barbers." For the entire employed population (poor and non-poor), the composition by employment status in 1985/86 according to the Labor Force and Socio-Economic Survey of that year was as follows: employees, 58.2 percent; employers, 2.4 percent; unpaid family workers, 13.4 percent; and own-account workers, 26 percent. When considering poor households, however, the picture becomes more complex, especially if more than one income earner is present. In that case, the head of the household could for example be a small farmer but the second income earner could be, say, a worker in the informal sector. In addition, many poor households would receive transfers. 1.30 Data on the sources of household income growth are sketchy. However, there is evidence that the proportion of income earners in the total population rose sharply since the early 1960s; that real wages of unskilled workers have increased significantly in the post-independence Li See, for example, Henry Bruton et al., Sri Lanka and Malaysia: The Political Economy of Poverty. Eguity and Growth World Bank Comparative Study, Oxford University Press, 1992, p. 165.  - 13 - period; and that private transfers from abroad became a very significant source of household income after the mid-1970s. 1.31 The Proportion of Income Eamers. Table 1.7 below summarizes available information on the size of the labor force, employment and unemployment for various years, from 1953 to 1990. The table shows a significant decline in labor force participation and in the employment ratio (the ratio of the employed to the total population) from 1953 to 1963. This is consistent with the data in Table 1.3, which indicate little economic growth in this period. From 1963 to 1990, on the other hand, the employment ratio rose sharply, from 27.3 percent to 35.1 percent. As shown in the table, this was the result of: (i) a higher labor force participation rate (which was mostly due to increased labor force participation of women); (ii) an increase in the ratio of population aged 10 and over relative to total population (a consequence of declining fertility); and (iii) an approximately constant unemployment rate. 12 Table 1.7: Sri Lanka: Labor Force, Employment and Unemployment, 1953-90 1953 1963 1969/70 1980/81 1985/86 1990 1. Labor Force (000s) 3,254 a/ 3,317 b/ 4,168 c/ 5,595 d/ 5,972 e/ 6,968 f/ 2. Employed (000s) 2,714 2,859 3,609 4,738 5,132 5,964 3. Unemployed (000s) 540 a/ 458 b/ 559 c/ 857 d/ 840 e/ 1,005 f/ 4. Unemployment Rate (%/6) 16.6 13.8 13.4 15.3 14.1 14.4 (3) / (1) 5. Population Age 10 and Over (000s) 5,803 7,626 8,867 12,032 12,573 13,073 6. Labor Force Participation Rate (%) 56.1 43.5 47.0 46.5 47.5 53.3 (1) / (5) 7. Total Population (000s) 8,098 10,463 12,384 14,867 15,977 16,993 8. Employed / Total Population (%) 33.5 27.3 29.1 31.9 32.1 35.1 (2) / (7) Notes: a/ Consumer Finance Survey, Central Bank, 1953. b/ Consumer Finance Survey, Central Bank, 1963. c/ Socio-Economic Survey 1969/70, Department of Census and Statistics. d/ Socio-Economic Survey 1980/81, Department of Census and Statistics. e/ Labor Force and Socio-Economic Survey 1985/86, Department of Census and Statistics. f/ Labor Force Survey, Department of Census and Statistics, First Quarter 1990. Includes the Northem and Eastern Provinces. This is the last quarterly labor force survey for which information pertaining to the entire country was collected. Subsequent surveys exclude the Northern and Eastern Provinces 12/ It should be noted that the data in Table 1.7 refer to the total population, not to poor households (however defined). Also, the table does not include data on workers who emigrated, and there are definitional differences among the various labor force surveys and other problems which render comparability difficult.  - 14 - 1.32 Real Wages. Table 1.8 below presents several time series of nominal and real wages of unskilled workers for the 1952-93 period. Items I and 2 pertain to what is usually referred to as the private organized sector, whose wages are subject to the authority of the Wages Boards (employees of public corporations are also covered under the Wages Boards legislation). There are 38 Wages Boards covering nearly all the important trades and about one million workers.'3 Item I pertains to plantation agriculture while Item 2 is an average of all sectors covered by the Wages Boards, i.e., plantation agriculture, industry and services. The Wages Boards data refer to minimum legal wages, rather than actual wages. To the extent that minimum legal wages are respected in practice, it is likely that they would apply to the least skilled from among each trade, and hence data on the evolution of minimum legal wages may be appropriate for our present purposes. Item 3 pertains to wages of unskilled government employees. Finally, Items 4, 5 and 6 pertain to market wages of unskilled workers in the unorganized private sector, which is that segment of the private sector not covered by the Wages Boards (e.g., the paddy sector, small-scale tree crops sector, small- scale enterprises in industry, commerce, construction, and so on). Data on more categories of occupations are available than shown in the table but the three shown suffice to represent the observed trends for unskilled workers in the unorganized private sector. These data are only available from 1979 onwards. 1.33 Items I and 2 in Table 1.8 indicate that there was a significant total gain in minimum real wages for the organized private sector (and public corporations) in the 1952-93 period: when deflated by the CPI, about 165 percent for plantation agriculture workers and 128 percent on average for all workers in the organized private sector (and public corporations). There have also been lower real wage gains for unskilled government employees in the same period (68 percent). The gains in real wages are more modest if nominal wages are deflated by the GDP deflator. As discussed earlier, there are reasons to believe that the CPI has been underestimated, especially in the 1970-77 period. The newer series on market wages in the unorganized private sector shows total real wage gains of between 16-31 percent in the 1979-93 period for the three occupational categories shown in the table, corresponding to average annual rates of growth of between 1 .1-2.0 percent. The gains are higher if nominal wages are deflated by the GDP deflator. 1.34 Private Transfers. Net Private transfers from abroad (mostly remittances from Sri Lankans who went to work in the Middle East) were negligible before the mid-1970s. By 1980, however, they had reached the equivalent of 3.4 percent of GDP, and continued to climb throughout the 1980s. There was a temporary decline during the Gulf war, but transfers have subsequently recovered and reached the equivalent of 4.8 percent of GDP in 1992 and 5.3 percent in 1993. The extent to which private transfers from abroad went to households in the lowest deciles of the household income distribution is not known. However, it is known that many migrants come from poor families, and hence it is likely that the poverty-reducing impact of these transfers has been quite significant. 1 31 If the percentage of employees over total persons employed in 1990 was approximately the same as in 1985/86, i.e., 58.2%, in 1990 there would have been about 3.5 million employees in the country. Hence, the Wages Boards probably cover about thirty percent of all employees.  Table 1.S: Sri LI -ka: Selected Wage Indicators of Unkilled Workers, 1952-93 Rate of 1952 1957 1967 1974 1979 1984 1989 1990 1991 1993 Growth 1. Minimum Average Daily Wage in Plantation Agriculture (Rs.) a/ 1.96 2.13 2.36 4.11 11.21 22.79 39.70 47.07 52.57 73.14 Real Wage Index 1952=100 (CPI) b/ 100.0 105.6 105.1 112.8 226.5 210.2 243.9 238.3 237.0 265.0 2.4% f/ Real Wage Index 1952=100 (GDP Def.) c/ 100.0 95.7 100.3 101.0 164.1 153.9 191.8 189.6 191.3 219.2 1.9% f/ 2. Minimum Average Daily Wage in All Wage Boards Sectors (Rs.) a/ 2.04 2.20 2.49 4.33 11.30 21.87 37.10 43.30 49.53 65.47 Real Wage Index 1952=100 (CPI) b/ 100.0 104.9 106.4 114.2 219.6 193.6 219.1 210.3 214.6 227.9 2.0% f/ Real Wage Index 1952=100 (GDP Def.) c/ 100.0 94.9 101.7 102.2 158.9 141.9 172.2 167.6 173.2 188.5 1.6% f/ 3. Minimum Nominal Wage Rate Index of Minor Government Employees, 1952=100 d/ 100.0 116.0 136.2 220.5 372.4 845.6 1444.4 1664.5 1887.4 2360.4 Real Wage Index 1952=100 (CPI) b/ 100.0 112.8 118.6 118.7 147.6 152.9 174.0 165.0 166.8 167.6 1.3% f/ Real Wage Index 1952=100 (GDP Def.) c/ 100.0 102.1 113.5 106.2 106.9 112.0 136.8 131.4 134.7 138.7 0.8% f/ 4. Daily Male Average Wage Rate in Paddy Cultivation, Harvesting (Rs.) e/ NA NA NA NA 16.06 37.81 55.88 71.11 84.82 104.3 Real Wage Index 1979=100 (CPI) b/ NA NA NA NA 100.0 107.4 105.7 110.7 117.8 116.3 1.1% g/ Real Wage Index 1979=100 (GDP Def.) c/ NA NA NA NA 100.0 108.6 114.8 121.8 131.3 133.0 2.1% g/ I-,, 5. Daily Male Average Wage Rate in Tea Cultivation, Land Preparation (Rs.) e/ NA NA NA NA 13.09 26.47 39.96 54.88 68.08 89.18 Real Wage Index 1979=100 (CPI) b/ NA NA NA NA 100.0 92.3 92.8 104.9 116.0 122.0 1.4% g/ Real Wage Index 1979=100 (GDP Def.) c/ NA NA NA NA 100.0 93.3 100.7 115.4 129.3 139.5 2.4% g/ 6. Daily Average Wage Rate of Unskilled Construction Workers, Carpentry (Rs.) e/ NA NA NA NA 14.27 33.31 53.92 64.86 77.42 104.55 Real Wage Index 1979=100 (CPI) b/ NA NA NA NA 100.0 106.5 114.8 113.7 121.0 131.3 2.0% g/ Real Wage Index 1979=100 (GDP Def.) c/ NA NA NA NA 100.0 107.7 124.7 125.1 134.9 150.0 2.9% g/ 7. CPI, 1952=100 100 102.8 114.8 185.8 252.3 553.1 830.2 1008.6 1131.5 1408.3 8. CPI, 1979=100 100.0 219.2 329.1 399.8 448.5 558.2 9. ODP Deflator, 1952=100 100 113.6 120.0 207.6 348.5 755.3 1056.2 1266.4 1401.7 1702.1 10. GDP Deflator, 1979=100 100.0 216.7 303.1 363.4 402.2 488.4 Notes: a/ Average of minimum wages set by Wage Boards. Central Bank Anmual Reporta and Department of Labor Gazettes. b/ Deflated by the Colombo Consumer Price Index (CPT). c/ Deflated by the GDP Deflator. d/ Central Bank, Annual Reports. e/ Information collected by the Central Bank. Data pertain to actual wages in the 'unorganized' sector and are available from 1979 onwards f/ Cumulative annual rate of growth for the 1952-93 period (41 years). g/ Cumulative annual rate of growth for the 1979-93 period (14 years)  - 16 - E. Trends in Physical Indicators of Human Welfare 1.35 The discussion until now focussed narrowly on one dimension of poverty, namely, private consumption expenditure of households. To provide a more complete picture, it is necessary to examine other key indicators of human welfare and deprivation. In this section, Sri Lanka's indicators in the areas of education, health, fertility and nutrition are briefly discussed. 1.36 While the Sri Lankan economy is still classified as low-income in the World Bank's standard classification of countries by per capita income (i.e., as presented in the World Development Report), its population is much better off in terms of the basic indicators of education, health and fertility than what could be expected given its level of per capita income. This is shown in Table 1.9. This table presents recent figures from Sri Lanka for several indicators, as well as the averages for three categories of countries identified in the Bank's World Development Report: low-income economies; lower-middle-income economies; and middle-income economies."4 The table consistently shows a superior performance of Sri Lanka relative to these comparators, in terms of primary and secondary education enrollment, adult literacy, life expectancy at birth, infant mortality, total fertility rate and the declining rate of population growth (past and projected). This remarkable achievement undoubtedly reflects the country's long-standing commitment to universal provision of basic health and education services. 1.37 The evolution of selected social indicators from 1950 to 1990 is shown in Table 1.10 below. Sri Lanka's progress in terms of health outcome indicators since 1950 has been remarkable, as can be seen from the table: an increase of 22 percent in life expectancy, and sharp declines in the infant, neonatal and maternal mortality rates, together with a halving of the crude birth and death rates. The table also shows the large gains achieved in terms of adult literacy and school enrollments since 1950. Education, Health and Fertility 1.38 In spite of obvious progress in health, education and fertility indicators as shown in Table 1.10, there has been some controversy as to how good Sri Lanka's performance in the past several decades has been, by international standards, in terms of changes in those indicators. A recent study by P. Isenman, P. Glewwe and H. Aturupane throws new light on this issue.'5 The authors note that linear improvements in social indicators are evidently harder to achieve as a country approaches maximum attainable levels. Hence, in order to measure relative performance in any given period, a suitable procedure would be as follows. First, using cross-sectional data from a sample of countries, one would regress the end-year value of an indicator (such as the primary enrollment rate) on the initial-year value (either by itself, or adding other plausible explanatory variables, such as the change in GDP per capita). One could then measure relative performance of any given country by calculating the divergence between the actual value of the indicator for that country in the end-year with the expected value of the indicator in that year as predicted from the estimated equation. Countries with actual values above expected values could then be said to have done better than average in terms of change in the indicator; and the greater the difference between the two figures, the better the performance (this assumes that the nature of the indicator is such that a higher figure is preferable to a lower one, such as it would be the case with primary enrollment; for other indicators 141 The latter is the average (weighted by population) of lower-middle-income economies and upper-middle-income economies. 15/ Aturupane, Harsha, Paul Glewwe and Paul Isenman, "Poverty, Human Development and Growth: An Emerging Consensus?". HROWP 36, World Bank, August 1994.  - 17 - such as infant mortality, the reverse would be true). Conversely, countries with actual values below expected values could be said to have done worse than average. Table 1.9: Sri Lanka and Other Lower and Middle-Income Countries A Comparison of Some Basic Indicators Low-Income Lower Middle Middle-Income Indicators Sri Lanka ' Economies -' Income Economies-' Economies -' GNP Per Capita 1990 (US$) 470 350 1530 2220 Average Annual Rate of 2.9 2.9 1.5 2.2 Growth of GNP Per Capita, 1965-90 (Percent) Average Annual Rate of Growth of GDP (Percent) a. 1965-80 4.0 4.9 5.5 6.3 b. 1980-90 4.0 6.1 2.6 2.5 Life Expectancy at Birth, 71 62 65 66 1990 (Years) Adult Illiteracy, 1990 (Percent) 17 52 32 27 a. Female 12 40 25 22 b. Total Average Annual Rate of Growth of Population (Percent) a. 1965-80 1.8 2.3 2.4 2.3 b. 1980-90 1.4 2.0 2.2 2.0 c. 1989-2000 1.1 1.8 2.0 1.9 Total Fertility Rate, 1990 2.4 3.8 4.0 3.7 Infant Mortality Rate, 1990 19 69 51 48 (Per 1,000 live births) Primary Education Net 100 68 " 86 89 Enrollment Rate, 1989 (Percent) Secondary Education Gross 74 38 54 55 Enrollment Rate, 1989 (Percent) " Sri Lanka is classified as a low-income economy in World Bank statistics. bl Weighted averages " Excluding China and India Source: World Development Report 1992, The World Bank  - 18 - Table 1.10: Sri Lanka-Major Long-Term Trends in Social Indicators Indicators 1950 1960 1970 1980 1990-91' Life Expectancy at Birth (years) 71.0 Female 57.5 61.4 67.1 71.7 Male 58.8 61.9 64.2 67.8 Infant Mortality Rate" 82.0 57.0 47.5 34.4 19.3 Neonatal Mortality Rate" 49.2 34.2 29.7 22.7 15.7 Maternal Mortality Rate" 5.6 3.0 1.5 0.6 0.4 Crude Birth Rated' 40.4 36.6 29.4 28.4 20.6 Crude Death Rated' 12.6 8.6 7.5 6.2 5.5 Adult Literacy (%) 69.0 77.0 82.4 86.5 88.0 School Enrollment, Ages 5-14 (%) 58.0 65.0 86.0 89.0 Total Fertility Rate' 5.0 4.2 3.7 2.4 Notes: "a Latest estimates of matemal and neonatal mortality rates are for 1987. b/ Data are for the years 1953, 1962, 1971, 1981 and 1990, respectively. " Rate per 1,000 live births. dl Rate per 1,000 population. Average number of live births per married woman over entire reproductive age. Sources: Ministry of Health, Annual Health Bulletin, 1991, p. 14; Ministry of Policy Planning and Implementation, Sri Lanka in Figures 1992, p.25; Priority Issues and Policy Measures for the Alleviation of Rural Poverty MARGA Institute, 1993. 1.39 Using a sample of 71 countries, including Sri Lanka, they conducted such an analysis for the Infant Mortality Rate (IMR) in the 1960-90 period. Their results indicate that not only did Sri Lanka perform better than average in terms of change (decline) in the IMR, but it ranked very highly in the sample, i.e., 19th out of 71. A disaggregation of this result for the two sub-periods 1960-78 and 1978-90 further showed that performance was better in the second sub-period, i.e., Sri Lanka's rank on change in infant mortality was 42 out of 71 for 1960-78 and 7 out of 71 for 1978-90. A similar analysis conducted for Net Primary Enrollment, using a sample of 43 countries, also yielded excellent results for Sri Lanka, which ranked 8th out of 43 for the period 1960-90. Thus, as the authors point out, it is clear that Sri Lanka's performance in recent decades in terms of changes in infant mortality and primary enrollment is outstanding by intemational standards. 1.40 It is also noteworthy that Sri Lanka has made a special effort to target women in its health and education programs. While illiteracy is more prevalent among women than men, a reflection of the differential school enrollment rates in the past, currently boys and girls are enrolled in primary and secondary education at about the same rates (and more girls than boys succeed in the A- level exam at the end of secondary education).'6 This successful effort to educate a great majority of women, together with the success in reducing infant mortality and the provision of family planning services, appear to have paid off handsomely in terms of an acceleration of the demographic transition: '6/ In Sri Lanka in 1990, there were 93 females per every 100 males in primary school, and 105 females per every 100 males in secondary school. In contrast, in Pakistan the corresponding figures were 52/100 and 41/100. See World Bank, World Development Report 1993, Table 32.  - 19 - there has been a sharp decline in the rate of population growth since 1965, and the current total fertility rate is close to the level that would lead to a stationary population. 1.41 As pointed out in a recent article by Amartya Sen'7, the experience of Sri Lanka (and the neighboring Indian state of Kerala) challenges the often-aired opinion that a developing country cannot afford expenditures for basic health care and education until it is richer. Sri Lanka promoted literacy and schooling programs early in this century; it massively expanded medical services in the 1940s. Poor countries need not wait to get rich before they can reduce the incidence of premature deaths and raise life expectancy, and achieve fertility levels close to those of developed countries. Education and health are labor intensive, and these services cost much less to produce in a cheap labor economy than they do in a wealthier country. Hence, although a poor country has less to spend on these services, it also needs to spend less on them (especially if it concentrates on basic services). Nutrition 1.42 One dimension of human welfare where Sri Lanka has not fared as well is nutrition status of young children. Evidence from several large surveys indicates substantial nutritional deficiencies when Sri Lankan children are compared to the standard international reference population of well-nourished children in common use for anthropometric nutrition studies (the standards developed by the U.S. National Center for Health Statistics around 1970, and later endorsed by WHO, usually referred to as the NCHS standards). In Sri Lanka, the incidence in young children in the critical age bracket of 6-36 months of both wasting (low weight-for-height, reflecting acute undernutrition) and stunting (low height-for-age, reflecting chronic undernutrition) is fairly high. 1.43 Table 1 .11 below presents estimates of wasting and stunting in Sri Lanka derived from four national-level surveys: the 1979/82 Nutritional and Socio-Economic Survey (conducted by the Food and Nutrition Policy and Planning Division, Ministry of Planning and Implementation, where data was collected over a period of three years); the 1987 Demographic and Health Survey (conducted by the DCS); the 1988/89 Nutritional and Socio-Economic Survey (conducted by the Food and Nutrition Policy and Planning Division); and the 1993 Demographic and Health Survey (conducted by the DCS). Since the surveys were not conducted on exactly the same basis, inferences about trends should be made with caution. 1.44 The general pattern shown by the figures in Table 1.11 is that the incidence of undernutrition in young children seems to have decreased somewhat between 1979/82 and 1987, but then rose again between 1987 and 1988/89. The year 1988/89 may have been off-trend, however, because at that time there were difficulties in the movement and marketing of food and other problems associated with worsening social disturbances.'8 This could have worsened the prevalence of wasting (but less so stunting, which is a longer term indicator). After 1988/89, there seems to have been a significant decline in the incidence of undernutrition, both wasting and stunting. The decline in the incidence of stunting is especially noteworthy, as this index was some five percentage points lower in 1993 than in 1987. The figures for wasting in all three surveys show the familiar pattern for developing countries, with incidence peaking in the 12-23 months age bracket. 17/ The Economics of Life and Death, by Amartya Sen, in Scientific American, May 1993. 18/ As noted in Nutrition and Health Status of Children 1993, Nutrition and Poverty Policy Division, Ministry of Policy Planning and Implementation, p. 36.  - 20 - Table 1.11: Prevalence of Wasting and Stunting Among Children 6-36 Months 1979/82, 1987, 1988/89 and 1993 1979/82 1987 1988/89 1993 Wasting at 6-11 Months 14.2 3.9 13.3 6.8 12-23 Months 21.6 19.3 20.7 17.7 24-36 Months 10.9 13.3 17.4 15.1 3-36 Months 12.9 6-36 Months 15.9 18.2 14.3 Stunting bh 6-11 Months 18.6 15.2 21.9 11.9 12-23 Months 34.1 31.1 38.2 25.9 24-36 Months 33.7 34.1 32.8 23.9 3-36 Months 27.5 6-36 Months 30.5 33.2 22.0 Notes: ' Percentage of children in respective age bracket with weight-for-height lower than -2 standard deviations from the median of the reference (NCHS) population. bl Percentage of children in respective age bracket with height-for-age lower than -2 standard deviations from the median of the reference (NCHS) population. Sources: The Health and Nutrition of the Poor in Sri Lanka by M. Khan, A. Zerfas, and N. Beller, USAID, August 1989, page 45; Food and Nutrition Statistics 1950-1990, Nutrition and Janasaviya Division, Ministry of Policy Planning and Implementation, pages 122, 127, 129, 134. Demographic and Hlealth Survey 1993, Preliminary Report, DCS. 1.45 Further evidence that undernutrition among young children may have declined in recent years is provided by another survey conducted in 1993, namely, the First Nutrition and Health Survey carried out by the Nutrition and Poverty Policy Division of the Ministry of Policy Planning and Implementation. According to this survey, the percentage of underweight children age 3-36 months (i.e., below two standard deviations from the weight-for-age standards) was 32 percent in 1993, as compared to 38.1 percent estimated from the 1987 Demographic and Health Survey. An estimate of percentage underweight from the 1993 Demographic and Health Survey is also available. This estimate is 32.6 percent, which is quite consistent with the corresponding figure estimated from the 1993 First Nutrition and Health Survey. 1.46 What is the significance of the figures in Table 1 . 11 ? There has been some controversy about genetic differences between populations and the appropriateness of references from Westem countries for children in developing countries. From a variety of studies, however, which have measured the actual growth of upper-class (thus, unconstrained by income) children of various ethnic groups, it seems that well-nourished young children have approximately the same growth potential on the average regardless of ethnic group, although some exceptions have been found (e.g., Japanese children appear to have a lower growth potential).'9 The reference (NCHS) population of well-nourished children is normally distributed with only 2.3 percent of children falling below 2 standard deviations below the median (whether what is being measured is weight-for-height or height- 19t Martorell, R. and J.P. Habitch (1986), "Growth in Early Childhood in Developing Countries", in Human Growth: A Comprehensive Treatise, Second Edition, Volume 3, New York and London, Plenum Press.  - 21 - for-age). Hence, if in a given population significantly more than 2.3 percent of children fall below this level (in either weight-for-height or height-for-age), there is probably an under-nutrition problem.20 The under-nutrition incidence figures for Sri Lanka, as shown in Table 1.1 1, exceed the 2.3 percent level by a large margin, with the exception of the figure for wasting among 6-11 months old in 1987. 1.47 There is also a significant incidence of low birth-weight in Sri Lanka, i.e., below 2,500 grams (5.5 pounds). A random study of 36,578 pregnant women done by the Ministry of Health/CARE in 1984 gave an incidence of 18.9 percent. More recent data collected by the Family Health Bureau give low birth-weight incidence as averaging 22.8 percent for the 1986-89 period. Low birth-weight is usually associated with under-nutrition in mothers and inadequate weight gain during 2 1 pregnancy. 1.48 Health Consequences of Undernutrition. While, as noted above, anthropometric studies suggest the presence of significant undernutrition (wasting and stunting) among Sri Lankan children by comparison with the reference NCHS population, these results have to be put into perspective by noting once again that infant and child mortality rates in Sri Lanka are very low by international standards. In most developing countries, it is the combination of under-nutrition with high prevalence of untreated infectious diseases that accounts for most of the deaths among infants and children under five. In Sri Lanka, under-nourished children apparently do receive medical treatment when afflicted by infectious diseases (and immunization for common childhood diseases is also widespread). What seems to make the difference relative to most other developing countries is the combination in Sri Lanka of generally good levels of education of the mothers (who can thus recognize life-threatening episodes of illness in their children) and good coverage of government- provided, free basic health services (see Box 2). 1.49 Thus, while anthropometric studies indicate significant undernutrition among young children in Sri Lanka by comparison with the reference NCHS population, it is less clear what the actual welfare consequences of such measured undernutrition are. As already noted, such consequences are not apparent in terms of deaths of infants and children. It is possible that there may be other negative consequences, in terms for example of impaired development of intellectual abilities among undernourished children which is not made up later. This issue requires more research. 1.50 Causes of Undernutrition. More research is also needed in order to assess the causes of undernutrition in Sri Lanka. In general, experience from around the developing world indicates that there are three main causes of undernutrition in young children: (i) private consumption poverty of households, sometimes aggravated by an unbalanced distribution of food consumption within the household, which results in low caloric intake of children; (ii) the prevalence of infectious and parasitic diseases in children, which reduce the efficiency of the body in absorbing and retaining nutrients; and (iii) inadequate feeding practices due to inadequate knowledge on the part of mothers. In most countries all three causes are present, but the relative importance of the causes varies from country to country. 20/ Global Indicators of Nutritional Risk (II). by Rae Galloway, World Bank, Human Resources Development and Operations Policy Working Paper, June 1993, page 4. 21/ See A Profile of Child Development in Sri Lanka, Human Resources Development Division, National Planning Department, Colombo, 1991, page 30. This report also indicates that, in addition to protein-calorie malnutrition, there is evidence of significant incidence of iron and iodine deficiencies.  - 22 - Survival. Poverty and Under-Nutrition A longitudinal survey of pregnant women and their newborns in a region which adequately represented Sri Lanka's poor rural areas was conducted by the Department of the Census and UNICEF in 1988, and provides some illuminating insights into the prevailing situation in those areas. The physical environment--housing, water supply and sanitation-- was poor. The large majority of families lived in houses with mud walls, mud floors, and palm thatched roofs. Slightly more than half had no toilets. More than 40 percent used unprotected sources of water, and 25 percent lived in abject poverty with periodic food shortages and lacking even basic consumable items such as rice, sugar and soap. Yet the rate of infant mortality, as revealed by the survey, was 27.5 per thousand births for the sample, a figure not significantly different from the rate reported for the entire region in 1986. In these deprived conditions, the rate of female literacy was 83 percent, and 24 percent of the women in the sample had some years of secondary education. Eighty percent of the births had taken place in medical institutions, 82 percent of the mothers were immunized for tetanus, and 88 percent had visited a clinic at least once, although the level of care and services received varied widely. Immunization of infants was quite high for all groups, nearly 90 percent. Slightly more than two-thirds of the women in the sample had basic knowledge of oral rehydration salts. Therefore, in this poverty-stricken environment, the availability of a basic health infrastructure, together with a literate generation of mothers with several years of formal education, was producing an unusual capacity for survival. ' Reproduced from Godfrey Gunatille, Government Policies and Nutrition in Sri Lanka: changes During the Last Ten Years and Lessons Learned. Pew/Cornell Lecture Series on Food and Nutrition Policy, Cornell University, Ithaca, New York, 1989. Box 2 1.51 In the case of Sri Lanka, the relative importance of the above three causes of undernutrition in young children is not known. The analysis of consumption poverty in this chapter indicated that there is still a large proportion of households which are poor, a fact that must explain in part the observed undernutrition. With regard to infectious and parasitic diseases, as pointed out in Chapter III, the incidence of these diseases in the general population and among children is substantial and appears to be on the rise. 1.52 That consumption poverty falls short of fully explaining undernutrition among young children in Sri Lanka is suggested by the figures in Table 1.12 below. The figures were estimated from a survey conducted in 1980-82 by the government's Food and Nutrition Policy and Planning Division. The second column shows the percentage of pre-school children that were stunted, by per capita expenditure quintile. The third column shows the percentage of pre-school children that were wasted. The table shows that, while the incidence of undernutrition generally declines from lower to higher expenditure quintiles, substantial undernutrition remains in the higher quintiles. 1.53 It can be preliminarily concluded, then, that undernutrition among young children in Sri Lanka cannot be attributed solely to consumption poverty. Other factors must be at work, such as poor feeding practices. The latter can play a major role especially in the process of weaning, which in turn highlights the critical importance of improving maternal education concerning nutrition. This is an area where more research is clearly needed.  - 23 - Table 1.12: Proportion of Pre-School Children Nutritionally At Risk by Expenditure Class, 1980-82 Per Capita Expenditure Percent Percent Quintile Stunted Wasted I (lowest) 49.2 14.9 2 36.7 11.0 3 31.7 13.9 4 30.3 9.5 5 (highest) 22.0 8.6 Source: Intemational Conference on Nutrition, Sri Lanka Country Paper, Nutrition and Janasaviya Division, Ministry of Policy Planning and Implementation, Colombo, January 1992. F. Characteristics of the Poor From Household Survey Data 1.54 This section presents a profile of the characteristics of the poor which emerges from our analysis of the Labor Force and Socio-Economic Survey of 1985/86 and the Household Income and Expenditure Survey of 1990/91. The focus is on what can be learned from the analysis about certain key characteristics of poor households, and their differences relative to non-poor households. Characteristics analyzed include regional differences in poverty, poverty by sector of employment, female headship of households, poverty by household size, poverty by level of educational achievement, poverty in relation to labor force participation and employment, and the household expenditure pattern. Regional Poverty Profile 1.55 For the purposes of constructing a regional poverty profile we have merged the estate sector with the rural sector. This is done for two reasons: (i) the estate sector is insignificant or non- existent in some regions, and (ii) the sample size for the estate sector is relatively small in the two surveys. The regional poverty estimates for 1990-91 and 1985-86 (for the reference poverty line) are presented in Tables 8 and 9, Annex 3. It should be stressed, however, that the available data exclude the Northern and Eastern provinces.22 The following observations can be made on the main results. (i) The results for 1990-91 indicate only limited regional variation in poverty. Only in the Westem region are the poverty levels found to be significantly lower than the other regions. Poverty estimates for the other four regions are quite similar. Limited variation in regional poverty levels carries the implication that the contribution of different regions to aggregate (national) poverty is roughly proportional to their population shares.23 22/ The corresponding excluded districts are Jaffna, Kilinochchi, Mannar, Vavuniya, Mullaitivu, Batticaloa, Amparai, and Trincomalee. 23/ It is possible that more marked geographical differences in the incidence of poverty exist at the district level. Estimates at this level would be feasible using all twelve rounds of the DCS' 1985/86 and 1990/91 surveys.  - 24 - (ii) The regional variations in poverty were distinctly more pronounced in 1985-86, and over time there has been a move toward convergence. Thus, in 1985-86 the Southern and South Central regions had significantly greater poverty than the Central and the North Western & North Central regions, which in turn were significantly poorer than the Western region. Between 1985-86 and 1990-91, while poverty in the Western region increased a little (mainly in the urban sector), it declined significantly in the Southern and the South Central regions (mainly in the rural sector).24 (iii) Within a region, rural poverty is generally higher than urban poverty. There has also been a considerable narrowing of rural-urban poverty differentials within regions over the two survey periods. Poverty by Sector of Employment 1.56 Table 10, Annex 3, presents the poverty profile by the sector of employment using the reference poverty line. In constructing such profiles, it is commonplace to classify households according to the principal occupation of the head of the household. This neglects occupational diversity within the household. We address this issue by first assigning working individuals to different sectors of employment; poverty measures are then computed assuming each individual's consumption is given by the per capita consumption of the household to which he belongs (which is consistent with the standard of living indicator we have been using), and each individual's weight given by the ratio of household size to the number of working individuals in the household. The main features of the estimates in Tables 10 and 11, Annex 3, may be described as below.25 (i) In 1990-91, the sectors reporting levels of poverty above the national average are: agriculture, forestry, fishing; mining and quarrying26; construction; and the unclassified group comprising of those whose industry of employment is not adequately defined. The highest levels of poverty are reported for the unclassified group which is likely to have a high proportion of casual laborers in irregular employment. Poverty in the manufacturing, electricity, gas and water sector is about the same as the national average. The remaining sectors have lower poverty, with the lowest levels observed for transport, storage and communications sector. The pattern is similar for 1985-86. (ii) In terms of changes over time, the largest decline in poverty (by all poverty measures) occurred in the agriculture, forestry and fishing sector (Table 11, Annex 3). Substantial decline in poverty was also witnessed by the manufacturing, electricity, gas and water sector, the mining and quarrying sector and the unclassified group. There was a significant increase in poverty in the trade, hotels, finance, insurance, real estate sector, and the commercial and social services sector. 24/ There was also a moderate decline in poverty in the other two regions. 25/ The estimates are based on 93 % of the sample households for 1990-91 and 92 % of sample households for 1985-86; the remaining households were excluded due to missing information on the sector of employment. 26/ The estimates for this sector have large standard errors reflecting its small sample size; these should be suitably interpreted.  - 25 - Female Headship and Poverty 1.57 About 20 per cent of the households in Sri Lanka are female- headed, and they account for about 17 per cent of the total population. Are female-headed households poorer? Estimates of poverty (for the reference poverty line) by the gender of the head of the household for 1985-86 and 1990-91 are given in Tables 12 and 13, Annex 3. For the country as a whole, in 1985- 86, there was no statistically significant difference between the incidence (head count) of poverty in female- and male- headed households. On the other hand, the poverty gap and the squared poverty gap indices for female-headed households were significantly higher than for male-headed households. In 1990-91, however, the difference in poverty between the two groups of households was not statistically significant for all three indices (head count, poverty gap and squared poverty gap). 1.58 Gender and Poverty. The above implies that individuals living in female-headed households are about as likely to be poor as those living in male-headed households. In some other developing countries, female-headed households have a higher incidence of poverty (however, recent Bank work in Indonesia and Viet Nam yielded conclusions similar to those in this report). The fact that this apparently is not the case in Sri Lanka is probably a reflection of the attention paid to educating women. It is still possible, however, that for all households the incidence of poverty is higher among women than among men; but investigating this point would require data on the intra- household distribution of consumption, which are not available. If women were systematically discriminated against in the intra-household distribution of consumption, then the incidence of poverty would be higher among women than among men. It is noted below that women have a lower labor force participation rate and a higher unemployment rate than men, but these labor market gender differentials need not translate into a higher incidence of poverty for women; the critical factor is the rules for intra-household distribution of income and consumption. This is an important subject on which further research would be desirable. Ethnicity and Poverty 1.59 Estimates of poverty measures by ethnic group are presented in Table 14, Annex 3. In 1990-91, the incidence of poverty for Sri Lankan Tamils in the sample was about the same as for the majority Sinhalese. However, it must be stressed that our sample excluded the North and East, where most Sri Lankan Tamils live; moreover, poverty may have been higher in these areas because of the civil war. The table also indicates that poverty incidence was much lower for Indian Tamils and higher for Sri Lankan Moors and other smaller groups than for the majority Sinhalese. The finding of lower poverty for Indian Tamils is consistent with the estimates in Table 3, Annex 3, which suggest that poverty is lower in the estates than in the rest of the country (a finding also reported in several other poverty studies). This finding, however, does not take into account the fact that calorie needs, and hence minimum required food expenditure, may be greater for Indian Tamils on account of their concentration in the plantations which involves hard physical work. Household Size and Poverty 1.60 The main results on household size and composition by level of poverty for 1990-91 are presented in Table 15, Annex 3. The overall average household size for Sri Lanka was just under 5 in 1990-91. Not surprisingly, household size declines as we move from the poorest to the richest strata in the population. Thus, the average size of a household in the poorest group (below 50 per cent of the poverty line) is about 7. It is around 4.5 for the non-poor, and 3.6 for the richest group (above 4 times the reference poverty line). This positive relation between household size and poverty is a very common finding among developing countries. The positive relationship is of course accentuated by the use of per capita consumption as the standard of living indicator, which does not  - 26 - allow for economies of scale in consumption. However, such economies would have to be fairly substantial to reverse the conclusion that poorer households tend to be relatively larger." Education and Poverty 1.61 Sri Lanka is well-known for its progress in literacy and basic education. The overall literacy rate (defined as the percentage of literates in the population with age above 10 years) in 1990- 91 was about 87 percent, 90 percent for males and 84 percent for females. The literacy rates of course vary inversely with the level of poverty, but the variation is not large (Table 17, Annex 3). Over the entire range of groups ranked by per capita expenditure as percentage of the reference poverty line, the male literacy rate increases from about 81 percent for the poorest group (below 50 percent of the poverty line) to near universal literacy (about 97 percent) for the richest (with per capita expenditure more than 4 times the poverty line); the increase is quite rapid over the bottom end of the distribution, reaching a level of 91 percent by 120-150 percent of the poverty line. The range for female literacy rate is similar from about 80 to 94 percent from the poorest to the richest groups, though for females it increases relatively slowly (to about 84 percent) up to a per capita expenditure of twice the poverty line. The pattern is similar for the average years of schooling (last two columns of Table 17, Annex 3). The overall average is about 7 years of schooling per person 10 years or older, the average for males being slightly higher (7.2 years) than for females (6.9 years). For both males and females, this average ranges from about 5 years of schooling for the poorest group to about 10 years for the richest. Both the literacy rates and the average years of schooling are lower in the rural (including estate) sector than in the urban sector (Table 18, Annex 3). Labor Force Participation, Employment, Unemployment and Poverty 1.62 Table 22, Annex 3, gives the 1990-91 distribution of the population aged above 10 years by employment status (employed, unemployed or outside the labor force) by percentage of the poverty line; Table 23, Annex 3, gives the same information disaggregated by age, and Table 24, Annex 3, by gender. The following observations can be made on the results in these tables. (i) Overall, in 1990-91 nearly half of those above 10 years of age participated in the labor force; the labor force participation rate (LFPR) is defined as the percentage available for work (the employed plus the unemployed) in total population above 10 years of age.28 (ii) The poorer groups have somewhat lower LFPRs. Thus, for example, for those at 80- 100% of the reference poverty line, the LFPR in 1990-91 was 45.3%, while for those at 120-150% of the reference poverty line it was 49.2%. Some of the observed differences are traceable to variations in the age-composition of poor and non-poor households (i.e., the poor households have proportionately more members in their early teens). 27/ Obtaining credible estimates of economies of scale in consumption is a difficult problem. There are not many credible attempts to go by. The usual approach is through a demand system; the main problem is to separate out the economies of scale from other effects of household size (see Lanjouw and Ravallion, 1993). 2S/ As has been the practice in Sri Lanka, we use 10 years (instead of 15) as the age threshold for defining labor force participation in view of the fact many of those in the 10-15 age group do participate actively in the labor force, particularly in the rural sector.  - 27 - (iii) As may be expected, labor force participation rates differ greatly by gender (see Table 24, Annex 3). In 1990-91, while more than two-thirds of males above 10 years of age participated in the labor force, the proportion for females in the same age group was less than one-third.29 For both men and women, labor force participation increases with per capita expenditure, and at comparable rates. (iv) The average unemployment rate (defined as the ratio of the unemployed to those participating in the labor force) for 1990-91 was around 16 per cent (Table 22, Annex 3).3 The results also indicate a strong positive relationship between unemployment and poverty. The unemployment rate declines from about 25 per cent for the poorest group (below 80 per cent of the poverty line) to 5.5 per cent for the richest (above 400 per cent of the poverty line). (v) The unemployment rates are significantly higher for women than men, 24 per cent as against 13 per cent on average (Table 24, Annex 3). Unemployment issues are further discussed in Section II.D of this report. Household Expenditure Pattern and Poverty 1.63 The expenditure pattem by the level of poverty for 1990-91 is shown in Tables 27, Annex 3, for the food group and Table 28, Annex 3, for the non-food group. Consistent with Engel's law, food share of total consumption expenditure (excluding consumer durables) declines steadily from about 80 per cent for the poorest group to about 40 per cent for the richest. The poor (i.e., all those below the reference poverty line) devote more than three-quarters of their total budget to food, with rice being the single most important consumption item accounting for about one-fourth of their total expenditure. Though food looms large in the poor's budget, items such as pulses, milk and dairy products do not figure significantly in their consumption. Within the non-food group, fuel and light, housing, and clothing are the relatively important items for the poor. Housing is clearly identified as a luxury, its budget share rising steadily for higher per capita expenditure groups. Health and personal care accounts for under 3 per cent of total expenditure of the poor, and education accounts for little over I per cent. 1.64 To summarize the results presented in this section, it can be said that: o Poor households are more likely to be found in rural than in urban areas, and less likely to be found in the Westem region than in other regions of the country; o Households are more likely to be poor if their working members are employed in agriculture and other primary production activities, construction, or are casual laborers; o Poverty among female-headed households is about the same as among male-headed households; 29/ As noted in Section I.D., however, the labor force participation rate of women has been increasing. 30/ The definition of unemployment is in terms of being available for work but not employed for most of the reference period. The reference period is the last 12 months. The unemployed do not have to be seeking employment, but are defined as those "ready to work when an opportunity is given" (DCS, 1987, p. 12).  -28 - o Poor households are larger; o Their members are less educated than those of non-poor households, but the differences are not very pronounced3"; o Members of poor households have somewhat lower labor force participation rates, which is partly explained by the higher proportion of children (age 10+) in them; o Members of poor households who are in the labor force are more likely to be unemployed than those of non-poor households; and o The poor, as elsewhere in the world, devote a very large proportion of their consumption expenditure to food, with rice alone accounting for fully 25 percent of their total consumption expenditure. 31/ However, one suspects that there could be significant differences in the quality of education acquired by members of poor and non-poor households (e.g., for the non-poor, household education inputs are likely to be larger, private tutoring may be available, and the quality of schools attended is also likely to be better); this effect cannot be captured by the survey measures we are reporting here.  - 29 - CHAPTER II - ECONOMIC GROWTH AND POVERTY ALLEVIATION 2.01 The evidence and analysis in the first chapter clearly indicate that post-Independence Sri Lanka has made impressive progress in reducing the incidence of poverty, reflected in terms of improvements in human development indicators as well as in consumption terms. This progress was made possible by a combination of two factors: (i) a strong push on public provision of health and education services together with income transfer programs that enabled higher consumption by the poor; and (ii) a relatively good long-term growth performance, with real per capita GDP growth of about 2.5 percent per year on average for the period 1950-93. 2.02 Despite this progress, as also shown in Chapter I, poverty remains substantial. While Sri Lanka has achieved human development indicators far superior to most other developing countries, it is still a low income economy. Though its long-term growth rate in per capita terms compares favorably with most of the developing world, in the last three decades it has fallen well short of the growth rate achieved by the high-performing East Asian economies. This can be seen in the following figures. In 1960, Sri Lanka's per capita GNP, estimated at US$141 in current (1960) dollars, was substantially higher than that in Thailand (US$96) and Indonesia (US$51), about the same as in Korea (US$156), and only 50 percent lower than in Malaysia (US$273). By 1992, Sri Lanka's per capita GNP in current (1992) dollars had risen to US$540, as compared with US$6,790 for Korea, US$2,790 for Malaysia, US$1,840 for Thailand and US$670 for Indonesia. It is arguable that a stronger growth performance along the lines of the high-performing East Asian economies would have allowed a much deeper reduction in Sri Lanka's poverty indicators. 2.03 The main reason why Sri Lanka's economic growth lagged behind the high-performing East Asian economies is that policies were less supportive of economic growth. Problems with macroeconomic management have constrained domestic saving and investment and fueled inflation; until recently, exports growth was weak due to inward looking trade policies, frequent difficulties with exchange rate management and interventions in the labor market; and the agriculture sector's performance has been constrained by pervasive controls on land ownership and use, marketing and pricing. These considerations suggest that the chief focus of Sri Lanka's future poverty reduction strategy would need to be the acceleration of broad-based economic growth, with emphasis on agriculture and labor-intensive manufacturing. At the same time, public spending for human development should be protected (and the efficiency and equity of this spending should be increased). 2.04 In this chapter, we first briefly assess Sri Lanka's growth performance in the post- Independence period, and the changing economic policies in the period. We then discuss in further detail the reforms in the post-1989 economic reform program and possible effects on poverty, followed by a brief discussion of policy directions to ensure sustained, broad-based growth. The rest of the chapter deals with two important topics that are closely related to the general growth prospects of the Sri Lankan economy, and that are also directly related to prospects for further poverty reduction because of their relationships to the characteristics of the poor identified in Chapter I. The first of these two topics is that of unemployment and labor market policies; as explained in Chapter I, we have found that there is a strong positive relationship between unemployment and poverty. The second topic is that of constraints to the growth of agriculture, including issues related to the liberalization of agricultural trade. This topic is of great direct relevance to poverty reduction because: (i) almost 80 percent of poor households in 1990-91 were in rural areas, where agricultural production is of course the engine of growth, either directly or indirectly; (ii) households are more likely to be poor if their working members are employed in agriculture than in other major sectors of activity (para. 1.64); and (iii) the poor devote a large proportion of their consumption expenditure to food (para. 1.71), so that any measures that lowered the price of food in real terms would be of direct benefit to the poor as consumers.  - 30 - A. Sri Lanka's Growth Performance 2.05 As shown in Table 1.9 in Chapter 1, the average annual rate of growth in real terms of Sri Lanka's GDP in the 1965-90 period was 4.0 percent, which was below the average for low-income countries overall (and, in the 1965-80 period, relative to those countries now classified as middle- and lower-middle-income). However, the difference narrows when growth of per capita GDP is considered, since Sri Lanka's average annual rate of population growth in the 1965-90 period was about 0.5 percent lower than that of all low-income countries (the difference was even higher with respect to lower-middle-income countries). Sri Lanka's growth performance looks even better when it is measured in terms of per capita gross national product (GNP): in terms of this indicator, Sri Lanka, with an average annual rate of growth of 2.9 percent in the 1965-90 period, did as well as the average of all low-income countries, and substantially better than either the middle-income or lower- middle-income countries (Table 1.9). In general, per capita GNP is a better indicator of a country's welfare than is per capita GDP, since GNP measures the total domestic and foreign value added claimed by its residents. 2.06 The discussion in Chapter I clarified the mechanics of the relationships between GDP growth, population growth, growth of consumption per capita, changes in distribution and declines in the incidence of absolute consumption poverty. The data clearly suggest that there has been a major decline in consumption poverty in Sri Lanka starting in the mid-1960s, driven by moderate growth of GDP and a rapid deceleration of population growth. At the same time, an examination of the economic policies pursued in this period of poverty decline suggests that Sri Lanka could have done much better in terms of growth and poverty reduction, had it followed more sustainable and efficiency-oriented policies --i.e., more in line with the policies pursued during that period by the high-performing East Asian economies. Unsustainable Policies and Missed Opportunities for Growth 2.07 The 1970-77 Period. Following a period of rapid growth in the second half of the 1960s, growth of the Sri Lankan economy as measured by GDP per capita slowed down in the 1970- 77 period (Table 1.3). The slowdown was due in part to exogenous supply-side shocks, particularly the 1973 oil price increase and a series of droughts in the years from 1971 to 1973 and again from 1975 to 1976 which severely affected agricultural production, especially paddy (rice).' This in turn led to large rice imports at a time when the world price of rice was increasing sharply.2 Performance of the tea and rubber subsectors was also weak; their outputs in 1977 were below the 1970 levels. Yields of these tree crops fell, apparently on account of reduced availability of fertilizer. The latter was related to persistent balance of payments difficulties during the period, and to problems with the largely public distribution system. Production of tree crops may also have been affected by nationalization (more than 60 percent of the area under tea cultivation was transferred from private to public ownership between 1972 and 1975). 2.08 The difficulties in agriculture were not compensated by a dynamic growth of manufacturing. While manufactured exports grew rapidly in the 1970-77 period from a very small base, manufacturing value added as a whole remained at about 15 percent of GDP throughout the period. Thus there was little structural transformation of the economy. The industrial policies of the -' There was also a serious armed insurrection in the Spring of 1971. In 1975, outlays on rice imports were five times their value in 1972 and amounted to 20 percent of total imports.  - 31 - 1970-77 period were similar to those of the 1960s and emphasized import substitution, with its attendant anti-export bias. 2.09 The government's macroeconomic policy response was not able to compensate for the adverse developments in agricultural production and external terms of trade. In response to balance of payments difficulties and low availability of reserves, the government resorted to tightening of import restrictions and increased reliance on quantitative controls rather than devaluing the domestic currency in real terms. Private consumption as a percentage of GDP was allowed to increase sharply, from 0.72 in 1970 to 0.76 in 1973, and remained at that level through 1976. At the same time, the ratio of investment to GDP, which had reached a peak of 19 percent in 1969, declined through 1973 and then remained at around 15 percent until 1978, when it rose at a rapid rate.3 One important lesson of the 1970-77 period is that although it is temporarily possible to allow per capita consumption to grow faster than per capita income, this is not sustainable over the long term. 2.10 The government in this period had the improvement of equity in society as an explicit goal. This was pursued through a variety of policies, such as generous public spending on social welfare programs and price controls to protect the consumers. The land policies of the period also sought to achieve greater equity and did benefit many of the rural poor directly. The Land Reform Law of 1972 capped the size of private paddy holdings at 25 acres and mixed holdings (paddy and other crops) at 50 acres. As a result of its implementation in 1973-74, about 560,000 acres were acquired by the Land Reform Commission. In the second stage launched in 1975, all estates held by large companies were nationalized, and a further 418,000 acres passed to government ownership. About 12 percent of the lands acquired by the government during these years (about 115,000 acres) were distributed to some 350,000 families of landless poor in small plots, mainly to be used as homesteads. The rest of the acquired land, comprising mostly plantation estates, remained in government ownership under two large public corporations. While this land was not distributed to poor families, the living standards of estate workers improved significantly following nationalization. The early 1970s also saw the launching of the Accelerated Mahaweli Development Program, where about 60,000 landless families have so far been settled in irrigated plots. 2.11 The 1977-89 Period. By 1977, it had become apparent that a change in economic strategy was needed to revitalize growth and speed up the economy's structural transformation. In that year, the Government of Sri Lanka initiated a new economic policy that sought to increase the role of markets and the private sector by reducing restrictions on pricing, investment and external trade and payments. Annual growth of GDP accelerated to 6-7 percent in the 1978-80 period, and remained at an average of close to 5 percent through 1985. However, difficulties developed as the reform effort slowed down in the early 1980s; in particular, the reform program did not address the restructuring of the overextended public sector, including the public enterprises. Serious ethnic and political conflict that began in 1983 and reached its peak in 1987-89 also hurt economic performance; the conflict led to a sharp increase in military and police expenditure as a percentage of GDP, from 0.7 percent in 1973-77, to 1.2 percent in 1978-82, 2.7 percent in 1983-87, and 3.7 percent in 1988-93. Unsustainable macroeconomic imbalances emerged in the second half of the 1980s, with the fiscal deficit reaching 15.7 percent of GDP in 1988; the external current account deficit was about 9 percent of GDP during 1985-89. Growth of GDP slowed down after 1985 and was only 2.9 percent per annum during 1987-89. Thus, by 1989, there was again a realization that major changes in economic policies were needed. 2.12 The 1989-93 Period. Sri Lanka's economy has experienced a rebound in growth since 1989, spurred by renewed adjustment efforts and enhanced private sector confidence. The new 3/ For further details of developments in the 1970-1977 period, see Henry Bruton et al, 1992.  - 32 - economic reform program launched in 1989 focussed on reducing macroeconomic imbalances and improving the incentives for the private sector through privatization and liberalization of trade and external payments. The rebound was broad-based with strong growth in industry and services and a rapid expansion of domestic and foreign private investment. Real GDP rose by an average of 5 percent from 1989 to 1992, compared to 2.1 percent during the prior three years, with significant expansion of exports. In 1993, preliminary estimates indicate that real GDP grew by about 7 percent, mainly as a result of strong growth in manufacturing, and a recovery in agriculture from the 1992 drought. Exports and remittances continued to grow rapidly. The current account deficit declined to about 5 percent of GDP, while official reserves increased to 5 months of imports. The budget deficit was still high at 8.1 percent of GDP. 2.13 While economic recovery got well underway, the restoration of macroeconomic balances has proved difficult. There are also some concerns about the quality of the fiscal adjustment that took place. Although the budget deficit was reduced sharply, from almost 16 percent of GDP in 1988 to 8.1 percent in 1993, the adjustment was largely through contraction in public investment. Mounting pressures on current expenditures, in order to accommodate the escalation in spending on defense, civil service wages and debt service obligations, prevented a stronger, more sustainable fiscal adjustment. 2.14 On the structural side, substantial progress was made during 1989-93 in the liberalization of the exchange and trade regimes (except for agriculture), and easing of regulations on foreign direct investment. The government also made commendable progress on the privatization of small- and medium- size firms, but was largely unable to tackle major loss-making public enterprises. Also, an initially successful but costly effort to reduce the size of the civil service could not be sustained. Tax holidays proliferated eroding the tax base. Poorly targeted income transfer programs also increased in scope (this is further discussed in Chapter IV). Last but not least, no progress has been achieved in reaching a solution to the costly civil conflict that continues to rage unabated. 2.15 Developments in 1994. In 1994, the adjustment effort has suffered a setback with the announcement in April-May of a wide-ranging, fiscally expensive package of new economic measures. If implemented fully, the fiscal cost of the package could be as large as 4-5 percent of GDP on an annual basis. The measures would, in the short-run, raise the incomes of many low- and middle- income households, including some that may have been bypassed by the benefits of growth in recent years. At the same time, the package could worsen macroeconomic imbalances significantly, increase inflationary pressures, and severely damage private confidence both domestically and abroad. Thus, the new measures could have a large cost in terms of lost GDP growth and reduced pace of poverty reduction. B. The Post-1989 Economic Reform Program and Effects on Poverty 2.16 In this section, the main reforms instituted since 1989 are briefly summarized, and their likely short-term effects on poverty are considered. We first consider the aggregate effects of the program, and then the effects of the specific reforms that took place in the areas of fiscal adjustment, privatization, and liberalization of the external sector. Aggregate Effects 2.17 In Chapter I.C. we examined the evolution of per capita private consumption in the 1990-93 period. The evolution in the 1989-93 period is similar; per capita consumption grew at an average annual rate of 1.8 percent over this four-year period (a total increase of about 7 percent). This was significantly below the annual average rate of increase of GDP per capita in the same period of about 4.4 percent. As in 1990-93, the reasons for the divergence between the two rates were: (i) a  - 33 - significant decline in the ratio of Private Consumption to GDP, from 77.3 percent in 1989 to 74.2 percent in 1993, reflecting the implementation of the adjustment program; and (ii) a faster rise in the CPI as compared with the GDP deflator. Thus, on the aggregate, the post-1989 period has seen a modest but significant increase in private consumption per capita in the first four years of the program. This implies that, unless there have been offsetting adverse changes in the distribution of consumption, poverty must have continued to decline. The broad-based nature of the economic recovery and the protection of social spending (see below), suggest that the distribution of consumption is unlikely to have been adversely affected by the post-1989 adjustment program. Effects of Specific Reforms 2.18 Fiscal Adiustment. A large fiscal adjustment took place since the peak of the government deficit in 1988 through 1993. The budget deficit declined from 15.7 percent of GDP in 1988 to 7.4 percent in 1992, and then increased to 8.1 percent in 1993. The bulk of the adjustment was borne by capital expenditure, which declined from 13.7 percent of GDP in 1988 to 6.3 percent in 1992 and then increased to 7.4 percent in 1993. Current expenditure as a percentage of GDP actually increased from 20.8 percent in 1988 to an average of about 22 percent in the subsequent three years, and then declined to 20.7 percent in 1992 and further to 19.9 percent in 1993. 2.19 That the brunt of the fiscal adjustment so far has fallen on capital expenditure is a cause of concern because of possible adverse effects on long-term growth, but it seems unlikely that these cuts have had adverse short-term effects on the poor. The cuts in capital expenditures especially affected capital transfers to public corporations and statutory bodies and net lending (much of the latter had also been going to public corporations, especially the Ceylon Electricity Board). It is difficult withouth having much more detailed data to gauge the impact that these cuts may have had on the poor; however, it would seem that in the short term they probably would not have much of an impact, since they would mostly affect the rate of capitalization of the recipient entities. To the extent that the cuts signal a new financing policy for these entities, whereby future capitalization will be financed out of internal cash generation rather than general government revenues, there may be some negative effects on the poor in the longer termn with regard to those corporations and utilities that produce services consumed by the poor (e.g., Railways, Ceylon Electricity Board, National Water Supply and Drainage Board). This point relates to the more general issue of what should be the pricing policies for government corporations and utilities and the weight that should be given to poverty considerations in formulating such policies; this is an important issue, but one that has not been analyzed in detail for this report. 2.20 Another way to look at possible adverse short-terrn effects on the poor of the fiscal adjustment is to examine the evolution of spending on Social Services since 1988. This is shown in Table 2.1 below, as shares of GDP. Table 2.1: Total Government Spending on Social Services, 1986-93 (% of GDP) Avge. 1986-88 1988 1989 1990 1991 1992 1993 Education 2.8 2.9 3.2 3.0 2.5 2.3 2.8 Health 1.6 1.8 1.8 1.5 1.4 1.6 1.5 Welfare 3.1 3.4 3.9 3.7 5.2 4.2 3.9 Housing 0.3 0.3 0.3 0.2 0.1 0.2 0.2 Community Services 0.2 0.1 0.1 0.1 0.1 0.1 0.2 Total Social Services 8.0 8.5 9.4 8.5 9.3 8.4 8.6 Source: Central Bank of Sri Lanka.  - 34 - 2.21 It can be seen that on the aggregate spending on social services did not decline after 1988 as a share of GDP. There were some changes in composition, chiefly an increase in the share of "Welfare"4 from 3.4 percent in 1988 to 3.9 percent in 1993 and a decline in the share of Health from 1.8 percent in 1988 to 1.5 percent in 1993. While the share of Health declined, in absolute terms spending on health in constant prices was about 6 percent higher in 1993 than in 1988, which implies that per capita spending remained approximately constant. Thus it seems clear that the fiscal adjustment per se did not have an adverse effect on government spending on social services. There are, however, serious efficiency and equity shortcomings affecting that spending (not necessarily originating in the period after 1988), which are discussed in detail in Chapters III and IV. 2.22 In addition to changes in government spending, tax reform has also been part of the agenda since 1990. Concerning direct taxes, the government is seeking to lower rates while at the same time widening the tax base. The goal is to reduce the maximum personal and corporate tax rate to 35 percent inclusive of all surcharges, which the government expects to achieve with its 1995 budget. However, in order to preserve (and, if possible, increase) revenues from income taxes as a proportion of GDP, a major effort will be needed to expand the income tax base. This will require curtailing the pervasive use of corporate tax exemptions and holidays. Concerning indirect taxes, the main priority is the introduction of a Value Added Tax (VAT). The government, with Bank and IMF assistance, is working on the design of the VAT. Its successful introduction would permit the elimination of most other indirect taxes and possibly compensate for any loss of revenue resulting from the rationalization and lowering of import duties. 2.23 What is the likely impact of these tax reforms on the poor? Concerning direct taxes, provided the reforms result in an increase in the share of income taxes in total tax revenues --as they should if efforts to widen the tax base are successful--, they should favor the poor by making the overall tax system more progressive. Concerning indirect taxes, the main concern would be that the poor be protected from the regressive impact of the VAT by exemptions for consumption items important to the poor, such as rice, wheat flour and coconut oil.5 With this proviso, introduction of the VAT may help the poor indirectly by increasing efficiency in the economy while avoiding a negative direct impact on poor households' consumption levels. 2.24 Privatization of State Owned Enterprises (SOEs). Privatization has been another major thrust of the post-1989 economic reform program. The government launched its privatization program in 1990; as of October 1993, 37 SOEs had been privatized. Divestiture has concentrated on small to medium size enterprises in industry and services. In addition, in July 1992, the government shifted management of 449 government-owned estates in the plantation sector to 22 private management companies on a profit-sharing basis. 2.25 Three types of beneficial effects are expected to result from the privatization program, namely: (i) a more efficient use of the assets and primary factors of production tied up in the SOEs; (ii) for loss-making enterprises, elimination of the burden imposed on government finances and on the financial system; and (iii) efficiency gains from enhanced competition. The third type of beneficial effect arises because in many instances the government imposed burdensome restrictions on private sector firms in order to protect the interests of SOEs. 2.26 The privatization program would tend to raise the rate of growth of the economy and hence indirectly speed up poverty reduction. If retrenched employees are not compensated, however, a This includes the transfer programs discussed in Chapter IV and government pensions. As recommended in the World Bank's Poverty Reduction Handbook, 1992, p. 2-4.  - 35 - direct effect of privatization may be to enlarge the ranks of the poor in the short term. This issue is important because at the end of 1992 there were approximately 120,000 people employed in SOEs, not including the govemment-owned estates. Moreover, the levels of labor redundancy in these SOEs appear to be high.6 2.27 A rough estimate of retrenched employees from SOEs divested up to April 1993 is about 24,000, of which some 16,000 belonged to the Sri Lanka Transport Boards (which became the "peoplized" bus companies). The bulk of these 24,000 employees were retrenched before the divestment took place, thus receiving their compensation directly from the government. The minority that were retrenched after divestment received their compensation from the new owners. In general, government policy has been to require from buyers of privatized SOEs a commitment to maintaining existing levels of employment. In most cases, retrenchment has been voluntary, i.e., the retrenched employees self-selected after being offered a compensation package; involuntary retrenchment has only taken place in a few cases where SOEs were divested by asset liquidation rather than by sale as going concerns. In all of the divested SOEs retrenched employees were compensated for the loss of their jobs; different compensation packages, however, provided uneven benefits. In future it would be desirable for the govemment to adopt a more uniform compensation method for retrenched employees. 2.28 Concerning the tree crop estates whose management was privatized, there has been no retrenchment of workers except for small numbers of white collar employees. This no-retrenchment policy was agreed as part of the privatization deal. The idea was that, although initially there may be excess workers, as private investment takes place they would eventually be productively employed. Thus, there has been no direct impact of the privatization of estates on their workers via loss of employment. However, there has been some concern that health services provided to estate workers and their families could deteriorate with the transition to private management. This issue is discussed under the Health section of this report in Chapter III. 2.29 Trade Reform and Other Measures to Increase Openness. The third major thrust of the economic reform program in recent years has consisted of measures to increase openness of the economy. The reform of the trade regime started in 1977, when most quantitative import restrictions were replaced by tariffs, but after the initial round of reforms almost a decade elapsed before renewed efforts to liberalize trade were initiated. The govemment's objectives are to replace specific duties with ad valorem tariffs; limit discretionary authority in setting tariffs; eliminate quantitative trade restrictions; and reduce the level and dispersion of tariffs. The benefits of these reforms would take the form of improved resource allocation efficiency in the economy and greater transparency of trade taxation. 2.30 The pace of reform has been especially fast in the last four years, as the maximum ad valorem tariff has fallen from over 100 percent in 1989 to 45 percent with the 1994 budget. At the same time, with few exceptions, the minimum tariff has been brought up to 10 percent. By the end of 1993 almost all the specific duties had been replaced by ad valorem tariffs. Progress has also been made in reducing discretionality with the elimination in 1994 of all tariff certifications (whereby line ministries determined the tariff rate to be applied on particular imports). There has also been considerable progress in eliminating non-tariff barriers. With few exceptions, import licenses have been eliminated. Apart from licenses on wheat and rice (discussed in more detail later), and a few other items, the bulk of the remaining restrictions are justified on health, security, moral or For SOEs in industry and services (including utilities), they could be as high as 40-50%, according to preliminary estimates contained in a recent Bank report on Labor Retrenchment and Redundancy Compensation in State Owned Enterprises: The Case of Sri Lanka, by Ariel Fiszbein, December 1992.  - 36 - environmental grounds. Particularly noteworthy is the fact that, with the 1994 budget, quotas on the importation of several important foodstuffs (potatoes, onions, chilies, black gram, lentils and maize) were removed.' On the export side, all taxes have been eliminated, and licensing requirements are retained for a few items for environmental protection reasons. 2.31 Sri Lanka has also made notable progress in liberalizing the external payments system. In 1992, the government liberalized the repatriation of profits and dividends. The facility to open foreign currency accounts was extended to all residents, and limits on capital transfers to emigrants were raised. In early 1993, payments for business travel, tourism, education and medical treatment abroad were fully liberalized. Moreover, the export surrender requirements were eliminated so that exporters can keep the proceeds from exports in foreign currencies. 2.32 The government has also been reversing its policy of heavy regulation of foreign direct investment. This reversal started in 1990, when the case-by-case approval of foreign direct investment was replaced by a new system that provides for automatic approval in most cases; the restriction that foreign direct investment should be located in the Export Processing Zones was eliminated; some restrictions on joint venture activities and ownership structures were removed; and responsibility for handling investment applications by all foreign firms were consolidated under the Board of Investment. These changes in policies and regulations appear to have already had a positive impact on foreign direct investment flows, judged by the data on government approvals of such investments in 1991 and 1992, which were sharply up from previous years.8 2.33 What are the implications of these external sector reforms for poverty? In the long- tern, the reforms are likely to raise the sustainable rate of growth of the Sri Lankan economy, thus accelerating poverty reduction. The pattern of growth is also likely to be more pro-poor, since a more neutral trade regime is likely to lead to a more labor intensive production system. This is because, as a rule, the greater the degree of protection of a given sub-sector, the greater the capital intensiveness of production in that sub-sector will be. In India in 1986, for example, manufacturing firms enjoying a high degree of effective protection had 53 percent of total fixed capital in manufacturing but only 19 percent of total employment; firms enjoying a medium degree of effective protection had 4 percent of fixed capital and 3 percent of employment; and firns with low effective protection had 43 percent of fixed capital and 78 percent of employment.9 2.34 The external sector reforms would also have short term effects on the poor, both beneficial and harmful. Data requirements and analytical complexity prevent a detailed analysis of such direct short term effects. A priori, however, there does not seem to be any reason to believe that liberalizing the external payments system or deregulating foreign investment would hurt the poor. Lowering protection to import-substituting manufacturing could, however, temporarily displace workers from affected firms, thus leading to short-term hardship. The severity of such hardship would partly depend on how quickly displaced workers are able to switch to those sectors were incentives would have improved as a result of the trade reforms, i.e., export industries. This is partly a function of the flexibility of the labor market, a topic which is discussed in section II.D. below. In agriculture, lowering of protection to domestic production could also cause hardship for a segment of the farning population; on the other hand, there should be a decline in the real price of food (at least relative to what it would be if protection was not lowered), which would benefit all consumers --and especially 71 However, temporary restrictions on the importation of onions were reintroduced in July 1994. 8/ Sri Lanka: Industry Sector Report, Preliminary Draft, World Bank, 1993, p. 12-14. 2' World Development Report 1990, World Bank, p. 61-62.  - 37 - the poor, for whom food expenditure is such a large share of total consumption. These issues are discussed further under section II.E. below. C. Ensuring Sustained Growth 2.35 According to the analysis in Chapter 1, the incidence of poverty is still high (24 percent in rural areas and 18 percent in urban areas in 1990-91). Since poverty affects a large proportion of the population, broad-based growth at a high and sustained pace will be necessary in order to achieve a rapid decline in poverty. Growth has been high in the 1989-93 period, but as already noted there is a serious danger that it will be derailed by the recently announced package of government spending initiatives. Even before this package was approved, there were concerns about inflationary pressures. To address the problem of inflation, fiscal deficits need to be reduced further. So, to ensure that unsustainable macroeconomic imbalances do not develop and inflationary pressures do not accelerate, the govermnent should reconsider these initiatives and return to the fiscal adjustment course it was embarked upon before the new initiatives were announced. 2.36 The government should also continue to pursue its agenda of structural reforms vigorously. This includes completing the rationalization of the trade regime by eliminating remaining quantitative restrictions and moving towards a tariff range with a low mean and dispersion. Under the present structure, tariffs on intermediate inputs for manufacturing are low and exemptions frequently made, while tariffs on finished manufacturing goods are still high. As a result, effective protection to finished manufacturing goods is still high. This distortion neeo. to be addressed, in order to increase efficiency and help promote domestic production of intermediate inputs. 2.37 Other structural reforms to be pursued include tax reforms; rationalizing the structure of public expenditure (e.g., as discussed later in this report for health, education, water supply and transfer programs); deepening the privatization of public enterprises; deepening the deregulation of foreign direct investment; pursuing reforms leading to a more flexible labor market; and tackling land market rigidities and other high-priority areas of reform in agriculture. The latter two sets of reforms (in labor and agriculture) seem especially critical to ensure that growth is broad-based. Because of this, they are discussed in greater detail in the concluding two sections of this chapter. 2.38 Sound macroeconomic policies and structural reforms along the lines specified above would create the conditions for more rapid sustainable growth of the private sector of the economy, which is the only realistic avenue to poverty reduction for the great majority of the poor. Rapid growth of the private sector, together with the low growth of population expected in the coming years, should lead to progressive tightening of the labor market and lowering of unemployment rates. This process would be aided by structural reforms aimed at making the labor market more efficient and flexible (as discussed in section II.D. below). Given the positive correlation between unemployment and poverty found in this report, a tighter labor market should be especially beneficial to the poor. D. Unemployment and the Labor Market 2.39 This section discusses the issue of high unemployment rates in Sri Lanka and possible remedies. While the debate on the nature and causes of this high unemployment continues, it would appear that factors contributing to high unemployment include: excessively restrictive labor legislation that raises the cost of labor to firms; poor quality of education acquired by many workers coupled with high labor market aspirations; and the government sector's tendency to hire large numbers of unemployed graduates, thus increasing incentives for unemployment (in addition, the official definition of the unemployed in use in Sri Lanka is very broad). These factors result in widespread queuing for rationed jobs in the regulated sector of the economy, both public and private. Such queuing, in turn, is made possible by the generalized availability of private transfers to support the unemployed while  - 38 - queuing. This interpretation suggests that in order to reduce the unemployment rate significantly it would be necessary to create good jobs (i.e., such as those in the regulated sector) at a faster pace than that of labor force growth, improve the flexibility of the labor market by easing regulatory restrictions, lower aspirations of entrants to the labor market and improve the quality of education and training. Subsidies to the acquisition of capital goods by firms should also be avoided in order to facilitate job creation. Relying on the public sector to create jobs for the unemployed beyond the needs of an efficient service, however, would be fiscally unsustainable. 2.40 The unemployment rate in Sri Lanka has exceeded 14 percent on average since 1980, which is high by intemational standards. Moreover, unemployment spells are long; in 1990, 85 percent of the unemployed had spent more than one year searching for a job. Also, as discussed in Chapter 1, our analysis of consumption poverty using the 1990-91 Household Income and Expenditure Survey indicates a strong positive relationship between poverty and the incidence of unemployment. This is shown in Table 2.2 below. The poorest group, that of individuals with consumption below 80 percent of the poverty line, shows the highest unemployment rate at 25 percent. The unemployment rate declines with increasing consumption, but remains high for all groups except the top one (over 400 percent of the poverty line). There is also a major gender differential: the overall unemployment rate for women in 1990-91 was 24 percent versus 13 percent for men (Table 24, Annex 3). The higher unemployment rate for women may be related to the fact that labor force participation of women increased by almost 50 percent between 1980 and 1990.'0 Finally, there are also important age differentials: in 1990, the unemployment rate for workers aged 10-14 was 11 percent; 15-19, 30 percent; and 20-29, 27 percent; but it was 9 percent for those aged 30-39, 3 percent for those aged 40- 49, and 2 percent for those aged 50-59. A Suggested Explanation for High Unemployment Rates 2.41 There has been much debate about the causes of persistent high unemployment rates in Sri Lanka. The main theories have been surveyed in a 1992 paper by Patricia Alailima," and more recently by Martin Rama as part of the work carried out for the Sri Lanka Private Sector Assessment.'2 The most plausible explanation for the high rates of unemployment is that most of the unemployed are in search of rationed jobs in the regulated segment of the labor market, and are not interested in available jobs in the non-regulated segment. 2.42 Labor markets in Sri Lanka could broadly be grouped into two segments: a regulated market subject to non-market wage and employment interventions, and a non-regulated market where outcomes are determined by the free interplay of demand and supply. The non-regulated segment of the labor market employs about two-thirds of all workers (four million out of a total of six million). 2.43 The regulated segment of the labor market is composed of govemment-owned plantation estates with about 7 percent of total employment (0.4 million workers), central and local IO/ However, because of migration to the Gulf, total labor supply increased by only 2.2% per year over this period. Li See Education-Employment Linkages: The Macro Profile, by Patricia J. Alailima, Sri Lanka J.S.S. 1992 15 (1&2). L2/ Sri Lanka Private Sector Assessment, Report No. 12514 CE, World Bank, 1994. The discussion in this section is partly based on Martin Rama's analysis, which has been separately published in his article entitled Flexibility in Sri Lanka's Labor Market, Policy Research Working Paper 1262, World Bank, March 1994.  - 39 - government employees with about 11.7 percent of total employment (0.7 million workers), and private firms and state corporations with about 15 percent of total employment (0.9 million workers). The government-owned estate sector is segregated from the rest of the labor market (there is little labor mobility in or out of the estates) and its wages and employment policies have been regulated, resulting in over-staffing and relatively low levels of unemployment (e.g., 7.8 percent in 1985-86). Wages and benefits in central and local government and in the state corporations have traditionally been higher than in the private sector for given levels of skill (or at least of education), and there is a large excess demand for public sector jobs, since there are many applicants to each available vacancy. 2.44 Wages of private firms (as well as state corporations) are subject to a floor or minimum wage set by the Wages Boards for different skill classes, but it is generally believed that actual wages paid by these firms exceed the minimum wages in most cases. To the extent that minimum wages are binding, they would lead to excess demand for jobs in private firms too. Since there are no data on vacancies and applications in this sector of the labor market, it is difficult to assess the extent of possible excess demand for jobs. Employment policies of private firms are also constrained by government intervention, especially the Termination of Employment of Workmen Act (TWA). This is discussed later in this section. 2.45 To summarize, then, and ignoring the almost perfectly segregated estate sector, it appears that there is excess demand for jobs in the regulated segment of the labor market, i.e., jobs are rationed there and wages exceed market-clearing levels. However, this is a relatively small segment of the total labor market, i.e., less than 30 percent if the estate sector is excluded. The rest of the labor market is unregulated, and presumably demand equals supply at the prevailing wages in this segment of the labor market. Table 2.2: Labor force participation, employment and unemployment by poverty groups 1990-91 Percentage of poverty Share in population above 10 years (%) Unemployment rate line (%) Employed Unemployed Not in labor force [2/(1+2)] [1] [2] [31 0-80 34.7 11.8 53.5 25.4 80-100 36.8 8.5 54.7 18.8 100-120 39.0 8.5 52.5 17.9 120-150 41.1 8.1 50.8 16.5 150-200 42.4 8.1 49.5 16.0 200-250 41.8 6.9 51.3 14.2 250-300 47.3 5.7 47.0 10.8 300-400 45.4 5.6 49.0 11.0 > 400 48.0 2.8 49.2 5.5 All 41.0 7.9 51.1 16.2 Source: World Bank 2.46 The above characterization of the labor market suggests that a plausible explanation for the high rates of unemployment is that most of the unemployed are in search of rationed jobs in the regulated sector of the economy (public and private), but are not interested in available jobs in the non-regulated sector. To test this hypothesis, Martin Ratna analyzed the responsiveness of real wage  - 40 - changes in the regulated and non-regulated sectors to changes in the unemployment rate. If the "search hypothesis" is correct, one would expect to find an inverse relationship between the unemployment rate and wage increases in the regulated sector, since according to the hypothesis the unemployed are competing for jobs in this sector, rather than in the non-regulated sector. On the other hand, since a higher unemployment rate would be associated, other things equal, with lower manpower availability for the non-regulated sector, wage increases in this sector should be larger the higher the rate of unemployment. His econometric results, using data from the 1980-92 period, appear to support the "search hypothesis".'3 2.47 The above explanation is also consistent with the evidence that unemployment rates increase with educational achievement, up to A-level graduates. In 1990, the unemployment rate was only about 4 percent for those with no schooling or with schooling up to fourth grade; it was about 14 percent (or about equal to the overall average) for those who had completed grades 5-10; about 27 percent for 0-level graduates; and about 32 percent for A-level graduates. It then fell to about 8 percent for college graduates. Unemployed 0-level and A-level graduates comprised almost one-half of all the unemployed.'4 A plausible interpretation is that these graduates are people who have developed a fair amount of labor market aspirations by virtue of their educational achievements, and are not contented with taking up low-paying, low-prestige jobs in the unregulated sector. 2.48 People searching for jobs still have to consume; so a relevant question is how do they finance their consumption while searching. Here the answer seems clear: they mostly receive transfers from relatives. According to the 1985-86 Labor Force and Socio-Economic Survey, 87 percent of unemployed males and 96 percent of unemployed females were being supported by their families.'5 For the one-quarter or so of those below 80 percent of the poverty line who were unemployed in 1990 (para. 2.39) such family support is evidently at a very low level. Thus, Sri Lankan society would appear to be somewhat unusual in this respect, that even very poor people are willing to wait for long periods of time at very low consumption levels for a good job in the regulated sector. 2.49 High (by international standards) rates of unemployment in Sri Lanka may also be due in part to the wide definition of unemployment used by the DCS. A person is considered unemployed if he has been available for work but not employed for most of the reference period of 12 months, which implies that someone who did work for, say, five months in the reference period would be classified as unemployed. Moreover, the unemployed do not have to be seeking employment, but are defined as those "ready to work when an opportunity is given". 2.50 The wide definition of unemployment may also help "explain" the seeming paradox of large numbers of poor people idly queuing for good jobs. It is likely that many of these people may be actually working in the non-regulated sector, albeit for less than six months in the year. This is consistent with the fact that in 1990-91 almost 80 percent of the poor in Sri Lanka lived in rural areas,16 where many of the jobs are seasonal or casual. 3/ For details, see Martin Rama, Op. Cit., p. 15-17. 4! P. Alailima, Op. Cit., p. 16. Ls/ P. Alailima, Op. Cit., p. 34. 6/ As implied by Tables 2 and 3, Annex 3.  - 41 - The Relative Price of Labor to Capital 2.51 The labor market issues discussed above can also be viewed in the context of the relative price of labor to capital goods. As stressed in the poverty literature, countries should avoid policies that artificially raise the relative price of labor to capital goods faced by enterprises. This is because such policies would induce a less labor intensive pattern of development than would otherwise prevail. The policies in question can be classified into two groups: (i) policies that raise the price of labor, and (ii) policies that lower the price of capital goods. 2.52 Policies That Raise the Price of Labor. Although labor market regulations presently only apply to a small proportion of the employed, they do affect many of the more dynamic firms in the manufacturing and services sector, and may be a deterrent to foreign direct investment.'7 Particularly burdensome for businesses is the Termination of Employment of Workmen Act (TWA), which prevents any retrenchment on non-disciplinary grounds without the written consent of the displaced workers in firms with 15 or more employees. This legislation artificially raises the effective price of labor faced by enterprises. This effect takes place through two mechanisms, i.e., by encouraging absenteeism (which is known to be high in both the public and private sectors), and by preventing downsizing when the latter would be normally required by business fluctuations. Minimum wages set by the Wages Boards, on the other hand, may be not binding in most cases. There are also non-wage labor costs in the form of social security contributions. Contributions by employers to social security legally amount to 15 percent of payrolls (12 percent to the Employee Provident Fund, and 3 percent to the Employee Trust Fund). However, Sri Lanka's system of social security is a "funded" system, where old-age benefits are supposed to be equal to the capitalized value of contributions (by both employers and employees). Thus, as noted by Martin Rama, contributions can be perceived by employers and employees as a partial substitute for current wage earnings, which implies that in the absence of the social security system the price of labor might not be much lower than with the present system.'8 Moreover, it is generally believed that there is widespread avoidance of payment of social security contributions. 2.53 Policies That Lower the Price of Capital Goods. These policies include credit subsidies and accelerated depreciation allowances (in the past, overvaluation of the exchange rate has been at times an additional factor that artificially lowered the price of capital goods paid by enterprises). Credit subsidies to agriculture have been channeled through several directed credit schemes over the past several decades. As in other countries, these subsidies have taken the dual form of below-market interest rates and high rates of default. In the Sri Lankan case, however, it is likely that the effect of the subsidies on the labor intensiveness of agriculture has been small. This is because most directed agricultural credit has been for the financing of working capital rather than agricultural machinery. Moreover, the percentage of farmers with access to subsidized credit at any given time has been very small (e.g., about 2-3 percent in 1990).'9 2.54 In industry, the bulk of formal long-term lending (for financing of plant and machinery) has been provided by the two Development Finance Institutions (DFIs), namely, the DFCC L7/ The manufacturing and service firms that were surveyed for the Private Sector Assessment ranked labor regulations very high as a constraint to expanding or diversifying their operations (third out of 33 possible constraints). 181 Martin Rama, Op. Cit., page 7. 9' A discussion of rural credit can be found in Strategic Issues in the Development of Sri Lanka's Agricultural Sector, draft report, World Bank, 1991, Annex II.  - 42 - and the NDB. The resources of these institutions, in turn, have come from the government and foreign donors. Until very recently the DFIs' lending rates had been held at artificially low levels due to government pressure (NDB was government-owned until 1993, and DFCC is partly government- owned), and had at times been negative in real terms. They were able to follow this policy because of their access to subsidies in the form of government-provided equity and tax holidays, and their access to donors' funds. However, the policy also meant that they have not been able to access and intermediate private domestic savings, which throws doubt on the sustainability of their operations.20 Since 1993, the DFIs have been moving towards market determined lending rates. This is a positive development that should help long-term sustainability of the DFIs and prevent encouraging a pro- capital bias in firms' choice of technology. 2.55 Capital goods have also been subsidized through accelerated depreciation allowances. Particularly since 1990, Sri Lanka has been offering generous tax incentives to promote exports and encourage foreign direct investment. In addition to tax holidays, these incentives include in many cases generous depreciation allowances (in general, three year writeoffs for machinery and 14 years for buildings). Such allowances create an incentive for firms to choose technologies that are more capital intensive. Policy Implications 2.56 In order to provide the conditions for rapid creation of jobs in the private sector, the governmenit would need to continue pursuing vigorously its program of economic reform. Maintenance of macroeconomic stability through prudent fiscal and monetary policies is especially critical. A Private Sector Assessment -- a detailed analysis of constraints to the growth of the private sector in Sri Lanka-- has been recently carried out by the Bank and IFC.2' 2.57 The reduction of unemployment would also be aided by policies conducive to labor market flexibility and which avoid artificially raising the price of labor relative to capital. In this regard, and as recommended in the Private Sector Assessment, the government should consider amending the TWA by setting an upper bound to severance pay while providing for a reasonable minimum period of advance notification of severance to the employee. While amending the TWA is not likely to reduce the rate of unemployment in the short run, it would tend to reduce the average duration of unemployment, thus leading to a fairer distribution of the unemployment burden.22 It would also help to prevent the "overkill" of protected industries as trade liberalization deepens, and therefore reduce the short-term job losses on this account. Moreover, in a long-run perspective, it should lead to greater job creation in the private regulated sector, since the current high costs of retrenchment must be a deterrent for investment in labor-intensive projects.23 2.58 It would also be helpful if policies are designed in such a way that capital goods are not subsidized. This point applies especially to policies in the financial sector, where it would be helpful from the point of view of efficiency and job creation if directed credit schemes did not incorporate subsidies. 20/ A detailed discussion of the operations of the DFIs can be found in Sri Lanka: Financial Institutions Study, Volume II, World Bank Report No. 9339-CE, February 1991. 21/ Sri Lanka: Private Sector Assessment, Report No. 12514 CE, World Bank/IFC, 1994. 22! For a detailed analysis of this issue, see Martin Rama, Op. Cit. 23/ Martin Rama, Op. Cit.  - 43 - 2.59 On the supply side of the labor market, it is critical that Sri Lanka's good performance in the area of education be continued and indeed improved upon, so that good jobs becoming available in the private sector can be matched by suitably qualified applicants. Issues in general education and in vocational education and training are discussed in Chapter III. 2.60 Finally, it needs to be stressed that a policy of trying to reduce unemployment by creating large number of jobs in the public sector would be self-defeating. Such a policy would tend to reduce the rate of growth of the economy, and would encourage queuing for public sector jobs by large numbers of new entrants to the labor force. E. Constraints to Agriculture Sector Growth 2.61 In agrarian societies such as Sri Lanka, the expansion of agriculture is the driving force behind rural development, which in turn lays the foundation for broadly based, poverty-reducing growth. The poor benefit directly if they are farmers; if they are not farmers, they benefit indirectly from growth in the demand for farm labor and for the products of the rural non-farm activities. An expanding farm sector generates demand for transportation, trading and storage services, which can be provided by small rural enterprises, which are usually labor intensive. Moreover, rising farm incomes lead to greater demand for consumer goods and services, some of which can also be provided by rural enterprises. Finally, as consumers, the poor (and others) would benefit if rising productivity in domestic food production leads to a decline in the real price of foodstuffs (i.e., for non-traded foodstuffs, and for traded foodstuffs if rising productivity caused the country to switch from a net importer to a net exporter). 2.62 The agricultural sector is still the most important sector, accounting for about 26 percent of GDP in 1992, and over 40 percent of total employment. The breakdown of agricultural value added in 1992 was as follows: paddy (rice), 18.9 percent; tea, 5.2 percent; rubber, 1.7 percent; coconut, 9.6 percent; other crops and livestock, 46 percent; forestry products, 8.3 percent; and fishing, 10.2 percent. In the 1982-92 period agricultural value added in constant prices grew at a modest annual average rate of 1.9 percent, as compared with an average growth rate of total GDP of 4.1 percent. 2.63 An important question for poverty reduction is thus what actions can the government take in order to accelerate agricultural growth. This is a difficult question, and a detailed analysis exceeds the scope of this report. However, there are three sets of issues that stand out in this context and will be discussed in this section: (i) issues related to the absence of a well-functioning land market; (ii) issues related to the privatization of the tree crop estates; and (iii) issues related to the reform of the agricultural trade regime. Beyond these issues, a further agenda of government action exists, on which we will only briefly comment. Absence of a Well-Functioning Land Market 2.64 The total area of Sri Lanka is about 6.5 million ha, of which about 2.2 million ha are under cultivation (cropland; 1991 figure); the rest consists of forests, wildlife reserves, grasslands and shrubs, waterbodies, wasteland and urbanized areas. An approximate distribution of cropland by subsector in 1989-91 is as follows: main tree crops (tea, rubber and coconut), 53 percent; paddy, 25 percent; crops other than paddy for which the country is a net importer (sugarcane, onions, potatoes, chilies, green gram and maize), 6 percent; annual export crops and minor tree crop exports, 12 percent; and other crops for domestic consumption, 4 percent. 2.65 Sri Lankan agriculture is predominantly of the small-holding type, with the major exceptions of the production of tea and rubber and to a lesser extent coconut. By legislation, there is  - 44 - a ceiling of 50 acres (20 ha) to the amount of land that may be owned by a family (defined as husband, wife and children under 18; children over 18 are entitled to a maximum of 50 acres separately).24 Thus, legally, no private holding devoted to the production of annual crops may exceed 50 acres, and most are much smaller; for paddy holdings in 1982 (the year of the latest Agricultural Census), 94 percent of holdings were under five acres, and they occupied 73 percent of total paddy land. On the other hand, holdings of more than 50 acres, all of which are owned by the government, accounted in 1982 for 78 percent of tea land, 59 percent of rubber land, and 16 percent of coconut land (equivalent to about 40 percent of all land under tea, rubber and coconut, or about 21 percent of all cropland). It must be noted, however, that an amendment to the Land Reform Law introduced in 1978 does allow the government to lease out tracts of land larger than 50 acres to prospective investors, for any type of commercial agriculture, for periods ranging up to 99 years. 2.66 As pointed out, government holdings of tree crop estates over 50 acres represent about one-fifth of all cropland. In addition, the government holds title to many smaller farm holdings which are privately operated (see next paragraph). The government also owns most of the land not under crops (two-thirds of the country's land area). Hence the Sri Lankan government is by far the dominant landlord in the island.25 2.67 Land Distribution Schemes. The main schemes have been those conducted under the 1935 Land Development Ordinance and subsequent related legislation. Under these schemes, several hundred thousand acres have been distributed in small holdings to landless peasants to be operated as family farms. The distributed lands were mainly located in newly established dry zone irrigated colonization schemes. As of 1992, 144,000 families had been settled in 105 schemes; the largest one, the Accelerated Mahaweli Development Program, has not yet been completed. Each family was given an irrigable plot for paddy cultivation and a "highland" plot for subsidiary crops under rainfed conditions and other uses.26 Initially, the holdings were leased out to the beneficiaries through "permits", with the government retaining ownership. In 1981, however, the government introduced a scheme to convert the leases to land grants, but even where the conversion has taken place, restrictions remain on transferability of the land through sale or other means. 2.68 In the case of the Mahaweli program, settlers were given land permits, but the project authority retains a lien on the land, severely limiting tenure rights until the development costs of the irrigated plot are amortized. In the meantime, there appears to be no legal mechanism for transactions involving encumbered land rights. The result is that settlers are virtually locked into their holdings. While a thriving underground market for illegal land-related transactions has reportedly developed, 24/ This ceiling was introduced by the Land Reform Law of 1972. All private lands over and above the ceiling were acquired by the State. In the second phase of the land reform, in 1975, all tree crop estates held by public companies, including foreign companies, were nationalized. 25/ A generally accepted estimate in Sri Lanka is that the government owns about 82% of all the land in the country (cropland and other). If this estimate were correct, and if all of the non- crop land were included in this figure, then the government would own about one million ha of cropland (about 45% of total cropland); however, since some of the non-crop land is privately owned, the true figure for government owned cropland would be higher. 26/ See "Rural Poverty in Sri Lanka: Priority Issues and Policy Measures", by Godfrey Gunatilleke et al., Asian Development Review, Vol. 10, No. 1, 1992, Asian Development Bank.  - 45 - such markets are deficient in terms of transparency and security of contracts and do not operate as efficiently as fully legalized markets.2" 2.69 Because of the long-standing policy of pursuing self-sufficiency in rice, farmers in the colonization schemes have been forced to devote their irrigated land to paddy, which has been an obstacle to agricultural diversification. In fact, government pressures on farmers to grow paddy have been more general than the conditions attached to colonization schemes, as the Agrarian Services Act of 1979 made it illegal to grow anything but paddy on "paddy land", defined as land which at any time had produced paddy. This provision was nullified by the Agrarian Services Amendment Act of 1991. 2.70 Weak Legal and Institutional Infrastructure for Land Market. A well-functioning land market for cropland is not only handicapped by the large proportion of cropland under government ownership and restrictions imposed on land transfers in colonization schemes and on land use allocation. Private land transactions are hampered by the absence of a unified and transparent system of title registration. This problem is compounded by the existence of widespread conflicting ownership claims on individual land holdings (especially in the wet zone) which the courts are unable to resolve in an expedient manner.28 Because of all these problems, agricultural land cannot be allocated to various crops in an optimal manner, which is a constraint on the sector's diversification and growth. Moreover, the uncertainty arising from conflicting ownership claims must be an important disincentive to fixed investment, including investments to maintain and improve the land itself. 2.71 The government is in the process of preparing legislation to address some of these problems. It would enable the start of a program consisting of: (a) preparation and maintenance of Cadastre Map by the Surveyor General's Department; (b) investigation and decisions regarding titles by the Commissioner of Title Settlement; and (c) maintenance of the Title Registry by the Registrar General's Department. This is a welcome development. In addition, it would be desirable for the government to conduct a review of the relevant legislation and court processes with an aim to speeding up the resolution of conflicting ownership claims on land. 2.72 Steps Towards a Functioning Land Market. The previous discussion suggests the steps towards a well-functioning agricultural land market: (i) Move ahead as quickly as possible with revamping of the title registration system and related legislative and court system reforms. (ii) For new settlers in colonization schemes, give outright land grants without any subsequent restrictions on use and transferability of the land; for existing settlers, convert all permits to land grants as soon as possible and eliminate restrictions on land use and transferability. 27/ Strategic Issues in the Development of Sri Lanka's Agricultural Sector, World Bank, draft report, February 1991. 2S/ A 1986 UNDP/FAO report estimated that as much as one-half of the privately owned farmland in the country was subject to conflicting ownership claims, with the average number of claimants to a plot estimated at nine; in extreme cases, as many as 200 people claimed shares in a single plot. Some 90% of the litigation in the courts appeared to be related to land disputes, which frequently lasted 15 years or more.  - 46 - (iii) Make it clearly known to farmers that current government policy concerning land use is that it is up to the farmer to decide on land use allocation in his farm. (iv) Prepare a program to divest to the private sector all government-owned land presently devoted to agriculture, except for tracts where agriculture use may not be sustainable (which should be converted over time to forest reserves or some other non-agricultural use). Privatization of Tree Crop Estates 2.73 As already noted, large tree crop estates are owned by the government as a result of nationalization, and in 1982 they accounted for 78 percent of tea land, 59 percent of rubber land, and 16 percent of coconut land. The public tree crop sector has been severely handicapped by a variety of policies and management practices. A detailed discussion and proposed solutions are contained in a recent Bank report on which the present section is based.29 2.74 The performance of the public tree crop sector in the 1980s was very disappointing. In tea, while world production increased by 33 percent between 1980 and 1989, output from Sri Lanka's state plantations declined. Production of the private smallholder tea sector increased by almost 300 percent in the same period, although this was partly due to the fact that many of the private holdings were approaching maturity during this period. In rubber, production of both the state and private smaliholder sector declined, as the world market became more concentrated among the large producers (Malaysia, Indonesia, Thailand) which could deliver more quality-consistent products than Sri Lanka. In addition, the two state corporations managing the government-owned tea, rubber and coconut estates suffered huge losses from 1985 onwards. This contrasted with a healthy financial situation for the private tree crop sector, in spite of many restrictions and other adverse policies. 2.75 The causes of the difficulties of the public estate sector are not difficult to discern. The main ones are: (a) the loss of experienced private management upon nationalization; (b) adverse pricing policies in the form of export taxes, overvalued exchange rates (in the 1980s) and industrial protection; (c) setting of wages, benefits and employee rights at unrealistic levels, which drove production costs upward; (d) inadequate attention to efficient agronomic practices and other technological improvements; (e) severe constraints on the marketing options of the estates' managers; and (f) poor performance of the rubber and tea research institutions in improving production and processing technology. 2.76 Recent Reforms. In June 1992, the government removed most tree crop estates from the two state corporations that had been holding them (except for those that were deemed non-viable), and placed them under 22 newly created and fully state-owned companies. These companies in turn contracted out the management of their estates to 22 private firms under a profit sharing plan. These management contracts were given initially for five years and are renewable depending on performance. In the same year the government eliminated export taxes. 2.77 While these reforms are encouraging, they do not go far enough. The expectation of the government was that by privatizing management they would be able not only to run the plantations more efficiently, but also attract private investment. However, the short time horizon of the arrangement, added to the refusal of the government to divest any part of its ownership, combined to kill private incentives to invest. Moreover, the government has continued to interfere with management-labor relations in the estates, mandating a 30 percent wage raise in 1993 and refusing to 29/ Sri Lanka: Tree Crops Strategy Paper. World Bank, 1993.  - 47 - permit any changes on burdensome labor regulations. In general the financial situation of the government-owned estates has not improved much and the sector is still short of capital. 2.78 It is clear that a deeper reform is needed. First, the government should give the private sector a bigger stake. To do so, management contracts need to be of a much longer duration (e.g., 50 years), and contractors should be allowed to purchase a capital share. Secondly, it should stop interfering with management-labor relations, and agree to a more realistic package of employee benefits and rights. Third, it should eliminate all constraints on marketing of tree crop products. And, fourthly, the government should restructure the research institutes, including giving a greater representation to the private sector in the decision-making process. With these reforms, the tree crop sector should be able to improve productivity and accelerate the pace of product innovation, thus making an important contribution to growth and poverty alleviation. Reform of the Agricultural Trade Regime 2.79 The agricultural trade regime impacts on the poor as either producers or consumers. In the Sri Lankan case, the most important issue in this context is the importation regime for rice. Because Sri Lanka has pursued a policy of self-sufficiency in rice, it has traditionally restricted its importation. These restrictions raise the domestic price of rice at the wholesale and retail level, and of paddy at the farmgate level, above the corresponding free-trade import parity prices. As such, the policy benefits paddy producers, but at the same time it hurts rice consumers. Conversely, eliminating import restrictions would hurt rice producers (at least in the short term) and benefit rice consumers. 2.80 Rice Importation Regime. A complex rice importation regime was introduced in 1990. Three private international companies were licensed to import rice, subject to several restrictions. First, the government through the Food Commissioner would set an annual import quota, and parcel it out among the three importers. Secondly, the importers were required to maintain certain buffer stocks. Thirdly, there was a minimum wholesale price (floor price) which the importers had to comply with. In addition, there was an import duty which is changed from time to time and at the end of 1993 was set at 20 percent. Recently, other private importers have been allowed to import rice, and it appears that import quotas have been relaxed, but the minimum wholesale price remains in place. The marketing system is competitive at the wholesale level and below, with many firms buying and selling both imported and domestic rice, and many millers competing to buy paddy. Rice imports in recent years have represented about 10 percent of total consumption. 2.81 Effects of Rice Import Restrictions. Restrictions on rice imports tend to raise domestic prices of rice. A recent study by C. Edwards estimated that in 1990 the domestic wholesale price of rice exceeded the level that would have prevailed in the absence of government intervention (no import restrictions or duties and no minimum wholesale price) by 30 percent; the comparable figure for 1991 was 15 percent.30 More recent calculations of nominal protection for rice are not available. However, as pointed out above, at the end of 1993 the import duty on rice was 20 percent, implying that nominal protection must have been at least 20 percent at that time. 2.82 These price effects of protection are quite damaging to those among the poor who are not paddy producers, since rice accounts for about one-fourth of their total consumption (para. 1.67). We have also estimated that, in 1990/91, 79 percent of all poor households were not paddy producers. Moreover, of the 21 percent of poor households which produced paddy, only about one-fourth had Protectionism and Trade Policy in Manufacturing and Agriculture in Sri Lanka, by Chris Edwards, Institute of Policy Studies, October 1993.  - 48 - sales of paddy exceeding in value their purchases of rice during the survey period. Thus, it is clear that for every poor household benefiting from rice protection, many more poor households are being "taxed" by the protection. Only about 5 percent of poor households would be negatively affected by a reduction of rice protection (at least on first round effects).3' 2.83 The net short-term effect on poverty incidence of reducing or eliminating rice protection is however difficult to calculate, because: (i) a certain number of farm households which are not now poor would become poor if rice protection was eliminated; and (ii) conversely, a number of non-farm households (and farm households that are not net sellers of paddy/rice) which are now poor would rise above the poverty line on account of the gain in real income resulting from eliminating rice protection. In the longer term, farm households hurt by the reduction or elimination of rice protection would gradually adjust by changing their product mix, thus recovering the lost income at least in part. 2.84 Reforming the Rice Importation Regime. Because rice is a product whose price affects the welfare of so many people in Sri Lanka, it would be good public policy to make rice protection more transparent. This would enable a more informned public debate concerning the degree of protection which society really desires --a debate where the interests of both rice producers and consumers can be explicitly taken into account. To achieve transparency, all quantitative import restrictions and the minimum wholesale price should be eliminated. That is, importation of rice by the private sector should be allowed without any restrictions or quotas, and importers should be free to set their selling prices. Protection should be given only through an ad valorem tariff, which could be set at a moderate level of 10-20 percent. These reforms should be accompanied by government assistance to rice producers to help them increase their productivity in rice production and to assist them in diversifying into other crops and livestock products. A forthcoming Bank study (Agricultural Sector/Crop Diversification Policy Study, expected to be completed in 1995) will examine the constraints to such diversification. 2.85 Other Food Import Restrictions. As already noted, import quotas have been recently eliminated for several important foodstuffs. Nevertheless, quantitative import restrictions still remain for wheat/wheat flour, sugar and milk (in addition to rice). 2.86 The case of wheat/wheat flour is special because Sri Lanka does not produce any wheat. Hence, in this case, no domestic producers are being directly protected, although under certain circumstances increasing restrictions to imports of wheat and wheat flour could raise domestic rice prices (for a given level of direct rice protection).32 The government, through its Cooperative Wholesale Establishment (CWE), has a monopoly on the importation of wheat and flour. Milling of wheat into flour is the contractual monopoly of a private firm, PRIMA, but the CWE retains ownership of the product. CWE also uses the services of the Food Commissioner's Department (FCD) for flour storage and distribution purposes. Below FCD, the flour distribution system is private and competitive, and there are no price controls. There are no import duties on wheat and flour (since January 1992), and the government-set price of wheat flour for sales by CWE (through FCD) to wholesalers is based on the cost of the imported product plus subsequent expenses, so that in theory Li / The estimates in this paragraph are based on the reference poverty line defined in Chapter 1. L2/ However, with the reforms suggested in para. 2.78, and as long as Sri Lanka remains a net importer of rice, increasing restrictions on imports of wheat/flour would not affect paddy producer prices, but only reduce imports of wheat/flour while increasing imports (and consumption) of rice.  - 49 - the government breaks even. Thus, if these arrangements functioned smoothly, the retail price of flour should currently approximate the price that would prevail under free trade of wheat and flour.33 2.87 There is a danger, however, that rents could be created along the marketing chain below FCD. This would occur if the quantity of flour supplied falls short of the quantity demanded by final consumers at retail prices involving no rents, i.e., retail prices that incorporate "normal" marketing and processing margins. This situation could arise because of error on the part of CWE in assessing the demand for flour, or because of pressure from rice producers on CWE to reduce imports of wheat and flour. In such a situation, the margins and retail prices would adjust upwards in order for the flour market to clear. The rents thus created would accrue to wholesalers, retailers and other agents in the flour distribution and processing chain below FCD, who would gain this extra income at the expense of consumers. Many would view such a situation as highly undesirable from the point of view of the general welfare. Another flaw of the present system is that consumers are forced to consume a single type of multi-purpose flour, while under free trade various types of flour would be imported and consumed. 2.88 To forestall the emergence of rents and to enable consumers a wide choice of products, the government should consider liberalizing the importation of wheat and flour (with PRIMA compensated for any loss arising from the changes). As recommended for rice, free importation and pricing of these products by the private sector should be allowed; a tariff of 10-20 percent could be levied for fiscal reasons. Further Agenda for Government Action in Agriculture 2.89 Govemment policies and programs in agriculture in recent decades have been dominated by two themes, namely: (i) the substitution of public for private management in the tree crop sub-sector, coupled with heavy taxation of the sub-sector; and (ii) the drive for rice self- sufficiency and the associated public investment in expanding land under irrigation (and related colonization schemes) for this purpose. This strategy is no longer adequate and needs to be replaced with a new strategy. This has already begun for tree crop3, as discussed above, although much remains to be done. As far as rice is concerned,-the country has almost achieved self-sufficiency, although the proportion of imports over total consumption (and total consumption itself) would be higher if there were no protection. The important point to note, however, is that there is relatively little scope left for further growth in rice production; unless, of course, there are major improvements in productivity that enable Sri Lanka to become a competitive rice exporter. 2.90 For the agriculture sector to achieve high sustained rates of growth, then, the sources of such growth will have to become more diversified. Undoubtedly, if the government presses ahead with the reform of the tree crop sub-sector, this sub-sector could become again an important contributor to overall agricultural growth. But a strategy for the growth of agriculture should be much broader-based; opportunities for growth in crops other than rice and the traditional tree crops need to be explored, together with opportunities in the non-crop sub-sectors (livestock, fisheries, forestry). 2.91 While exploiting such growth opportunities must be the responsibility of the private sector, certain supporting public goods and services will have to be provided by the government. It is beyond the scope of this report to go into details in this subject. Generally speaking, there are three broad areas: technology generation and transfer (research, extension); public infrastructure (roads and other transport infrastructure, power, telecommunications, irrigation); and human resource development 33/ For the period 1985-91, Edwards (Op. Cit.) has estimated that the domestic price of flour was on average about 10% above the border price equivalent.  - 50 - (general education, agricultural training, basic health services). Some govemment support in the area of rural financial services development may also be warranted. 2.92 One particlarly important instance where institutional reforms are needed concerns the institutional arrangements for the Mahaweli project area, where about 60,000 families have been settled (out of a final target of 70,000), and where about 40 percent of the total irrigated land of the country is located. The Mahaweli Authority of Sri Lanka (MASL), is in urgent need of restructuring. Its functions are too broad, including, in addition to settlement, the operation and maintenance of the irrigation system, the provision of support services and several commercial ventures. Its performance has been deficient. Irrigation structures have been deteriorating for lack of sufficient maintenance and farmers organizations remain underdeveloped. The extension staff includes large numbers of poorly qualified field agents, while in the rest of the country the extension service is being upgraded. The Bank has suggested that the MASL be transformed into a river basin authority responsible for upstream management and environmental protection, thereby clearing the way for the private sector (including farmers organizations) to become the main agent of growth, supported by national line agencies. This recommendation is reiterated in this report.  - 51 - CHAPTER III - HUMAN RESOURCE DEVELOPMENT 3.01 In addition to pursuing policies conducive to rapid sustained growth (discussed in Chapter II), further poverty reduction in Sri Lanka requires that the country keep a strong focus on human resource development. Indeed, rapid sustained growth itself is unlikely to be feasible in the absence of such focus. The two elements of the strategy to reduce poverty are mutually reinforcing. Good policies that promote the productive use of the poor's labor need to be matched by investments in health and education that enable the poor to take full advantage of the emerging opportunities. In the present chapter, we review the major issues facing the provision of basic social services in Sri Lanka and make suggestions for improving the efficiency and equity of those services. A. Education and Training' 3.02 The importance of the education and training sector for the poor has two aspects. First, an efficient and equitable education and training sector at all levels (primary, secondary, higher, and vocational and technical education) is crucial for Sri Lanka's growth prospects, and hence, indirectly, for poverty reduction. Secondly, poverty reduction would be facilitated by the easy availability and good quality of those education and training services that the poor themselves are more likely to demand. From this second aspect, the highest priority is to ensure the availability of universal, tuition-free and high quality primary and lower secondary education (grades 1-8) throughout the country. This could be complemented with vocational and technical courses for school leavers, to the extent that such courses are closely tailored to the needs of the labor market. 3.03 As noted in Chapter I, Sri Lanka has done very well in terms of educational achievement. Moreover, the free public education system has been an important source of upward mobility in the society. This section highlights the main issues facing the education and training sector, which Sri Lanka needs to face .quarely in order to continue being a leader in this area. Government Expenditure on Education 3.04 Total government expenditure on education as a percentage of GDP, excluding transfers under the Mid-Day Meal program, averaged about 2.6 percent in 1980-88, reached a high of 3.2 percent in 1989, and averaged 2.7 percent in 1990-93. The latter figure is low compared to a number of other Asian countries such as China, India, Indonesia, Korea, Thailand and Malaysia.2 It is also below the levels attained in the 1970s (e.g., 3.0 percent in 1976, 3.9 percent in 1970). There are also significant inefficiencies and imbalances in the distribution of expenditure. These need to be addressed as a matter of priority. Once sufficient progress in this respect has been achieved, it would probably make sense for the country to target an increase in government expenditure in education as a percentage of GDP. This section is based on work conducted in connection with the Sri Lanka Education and Training Sector Strategy Review, World Bank, 1993. Measured in constant prices using the GDP deflator, government education expenditure per student in 1992 was about 3% higher than in 1980 (while total enrollment at all levels increased by about 27% in the period). This calculation includes general education (excluding transfers under the Mid-Day Meal Program) and universities but excludes vocational education and training institutions.  - 52 - 3.05 An examination of the education and training sector indicates that there are important opportunities to provide for greater efficiency and equity. This could be achieved by a combination of reforms within the public education and training system and redefining the respective roles of the public and private sectors (currently, the government has a near-monopoly in the provision of general and higher education). In the rest of this section, we discuss the major issues in general, higher and vocational and training education, and point to desirable reforms in government policies. Issues in General Education (Primary and Secondary) 3.06 In the past 50 years, the general education system grew from a small, restricted system to a large one with relatively open access. Education is tuition-free from primary through university, the state-financed school system provides school places at primary level (grades 1-5) to practically the whole population, and widely available places at secondary (grades 6-1 1) and upper secondary (grades 12-13) levels, the language of instruction is the pupils' home language, there is a common curriculum in the basic cycle and textbooks are free up to grade 8. 3.07 As noted in Chapter I, enrollment rates for primary and secondary education compare very favorably internationally. Universal enrollment for primary education has been nearly achieved, and for secondary education the gross enrollment rate in 1989 was a high 74 percent. It is likely that those children not in school tend to be from the poorest families. Despite the relatively good coverage, there is reason to believe that there is ample room for improving the quality of general education services, even within the present overall budget. To judge by the mainly poor examination results or by employers' criticisms of the education levels of new recruits, it would seem that general education is providing poor value. 3.08 Improving the Quality of General Education. Recent Bank sector work on general education points out to a number of measures that could be taken to improve its quality: (i) The government should engage in a major push to improve teacher training, in terms of the content and methods of training, and of the number of teachers trained per year. The latter needs to be increased in view of recent massive levels of recruitment of young inexperienced teachers. A teacher training action plan should be developed as soon as possible. This plan should emphasize a re-thinking of how teaching is actually conducted in the classrooms. (ii) The government should review the general education curriculum with an aim to make it more relevant to the needs of the economy and society. This would imply inter alia a stronger emphasis on mathematics, science and English language teaching. (iii) The government also needs to curb the growth of salaries in the sector by reducing the number of teachers. This should be complemented by measures to increase teachers' productivity, such as reducing absenteeism, increasing the teaching load, and redistributing teachers from surplus to deficit schools. Better control over overall numbers recruited is necessary, since the average student-teacher ratio is already low by international standards, and the number of students entering the system has started to decline (because of the decline in fertility). The Ministry of Education should work with Provincial administrations to ensure maximum utilization of the existing teacher contingent. (iv) Provided that GDP continues to grow at historical rates or better, substantial additional resources for general education should be available over time if public expenditure on general education is kept constant as a percentage of GDP or increased. These  - 53 - additional resources should be applied to providing schools with more and better complementary inputs (i.e., other than teachers, including adequate funds for repairs and maintenance of school buildings, instructional materials, etc.), which are critical for a continuous improvement in the quality of education. (v) The govemment should conduct a careful analysis of the management and administration of the general education system, in the light of the profound decentralization that has taken place since 1990 (i.e., devolution of most schools to provincial administrations). Appropriate remedial measures should be designed and adopted following the study. 3.09 Improving the Eguitv of General Education. At present there is a wide discrepancy in resource allocation per student between smaller and larger schools, and this is only partly explained by the tendency of larger schools to have a higher proportion of students at the secondary level. There are also wide discrepancies in the allocation of general education resources among provinces. In general, poorer provinces (as measured by per capita income) have proportionally more unqualified teachers and tend to spend less on instructional and teaching materials. There are also teacher shortages in some areas. A reduction of these inequalities is likely to be strongly pro-poor: it would tend to increase the participation of poor children at the secondary level, and raise the quality (and hence market value) of general education for them. Central and provincial administrations should work together in the preparation of an action plan to reduce inequalities in the allocation of resources among schools, both within and across provinces. University Education 3.10 University education presently absorbs about 10 percent of total current govemment expenditure on education (primary education absorbs 38 percent, and secondary education 52 percent). While the number of university students as a percentage of the corresponding age group, at about 2 percent, is low for a country aspiring to middle-income status3, govemment spending on university education in real terms increased by a high annual rate of growth of 12 percent in the 1987-92 period, and would continue to grow at about 9 percent per year up to the end of the decade under present plans. Rapid capacity expansion is taking place, including expansion in the existing nine universities and the establishment of a number of two-year Affiliated University Colleges. Pressure for such expansion is great, as there is presently a large unsatisfied demand for university places. The university sub-sector is about 95 percent state funded: it obtains some foreign grants but has negligible income from endowments and minimal cost recovery. At present all tuition in the universities at undergraduate level is free for Sri Lankan students, regardless of parental income. 3.11 While an increase in the percentage attending universities seems warranted, there is a risk that uncontrolled growth of the public university sector could have negative effects for society at large. Given the free tuition policy, it is quite possible that overinvestment in university education could take place, with the economy unable to absorb graduates productively. This would not only constitute a waste of resources in its own right, but could also enhance conflicts in society and lead to enormous pressures for increases in public sector employment not warranted by a rational definition of that sector's functions. Some of the Newly Industrialized Countries (NICs) have as many as 8% of their university- age population attending post-secondary institutions, and for advanced industrialized countries the corresponding figure is in the 25-30% range.  - 54 - 3.12 To forestall these risks, it is recommended that the government reconsider its policies concerning financing of university education and the related issue of the division of labor between the public and private sectors in higher education. A recent Bank study of the experiences of the fast- growing East Asian countries makes the point that these countries followed a strategy of concentrating public education spending on the primary and secondary levels.4 At the post-secondary level, they focussed public spending on science and technology education (including engineering), while allowing the demand for other types of higher education to be met by a self-financed private university system. This strategy would be more efficient than the one being followed by Sri Lanka until now.5 It may also be more equitable, especially if it includes a well-designed program of loans for poor students wishing to attend private universities. Technical and Vocational Education and Training (TVET) 3.13 At least six ministries provide technical and vocational education and training (TVET), through a variety of institutions. In general, the TVET sector is afflicted by low internal efficiency, uneven relevance to labor market needs, duplication and lack of coordination. Some of the main issues in the sector are: (i) Some technical colleges suffer from high dropout rates. (ii) Public sector TVET facilities are under-utilized, and staff vacancies and inexperienced teaching staff constitute a serious weakness in many institutions. (iii) Data on employment of trainees after graduation is sketchy, but it is known that many fail to find employment in their field of training. (iv) Resources in the public TVET sector appear to be spread too thinly. Unit costs in technical colleges average about US$125 per trainee per year, as compared with US$2,500 in a high quality private course. (v) While there are more than 2,000 private training institutions, only a few are well- equipped and staffed. Most enterprises lack training capabilities. 3.14 The government should proceed to rationalize the public TVET sector and to place much greater emphasis on training linked to employment opportunities. Coordination among the various government agencies providing TVET needs to be enhanced, duplication eliminated through consolidation of programs, and quality of training upgraded. To move in this direction, the government recently established the Tertiary and Vocational Education Commission, with wide responsibilities in policy advice, resource allocation, training standards, and monitoring of policy implementation. The Commission is a promising initiative, but its membership needs to be made more representative of employers. 3.15 The government is also considering the establishment of a Skills Development Fund to finance priority training identified by employers. The Fund would initially be financed by the The East Asian Miracle: Economic Growth and Public Policy, World Bank, 1993. Recent policy changes in Sri Lanka have allowed a very limited degree of development of private university institutions. A few small colleges now exist, but the only major development attempted, a private medical school, met intense political opposition and was taken over by the State.  - 55 - government, with part of the cost later transferred to industry (if the Fund proves successful). Such funds in other countries have had mixed results. Where governments have set them up without clear plans and full participation of employers, they have usually been failures. However, if well planned and implemented, the potential for success is high. B. Health 3.16 III health has a pronounced negative effect on the welfare of the poor. For those in the labor force, illness is likely to have a negative effect on their productivity, which in general would translate into a reduction in their incomes. Moreover, illness also leads to loss of work time, which again translates into reductions in income.6 These reductions in income are harder to bear in the case of poor households. In addition to these income effects, ill hlealtlh very directly reduces the welfare of the poor (and others), as do premature deaths --even though these welfare losses are not captured in national income accounts. 3.17 As indicated in Chapter 1, Sri Lanka's health outcome indicators of mortality/life expectancy are exceptionally good for a country at its level of per capita income. Good performance in the health area is due to a large extent to a long-standing government commitment to providing free medical care to the population at large, including both preventive and curative services. In addition to programs in the health sector, programs in the education sector and the consequent achievements in terms of basic education coverage and literacy must have also contributed to good health indicators. For example, data from many countries indicate that there is a strong inverse relationship between education of mothers and infant mortality, which remains even after statistically controlling for other variables. This relationship is partly explained by the fact that educated women have fewer (and better spaced) children, an effect that is also evident in Sri Lanka, with its exceptionally low fertility indicators. 3.18 A key element of a strategy for poverty reduction is to ensure low cost access of the poor to basic health services of good quality. At present, government health facilities already provide good coverage throughout the country. Total government expenditure on health as a percentage of GDP averaged 1.5 percent in 1990-93, a figure which was about average as compared with most South Asian and East Asian Countries.7 In absolute terms, total government expenditure on health facilities and services almost doubled in real terms from 1975 to 1992 (i.e., measured in constant 1982 prices, from Rs. 1,043 million in 1975 to Rs. 2,039 million in 1992), which implied an increase of about 52 percent in real per capita terms. 3.19 As a share of GDP, however, there has been a decline in government expenditure on health from the levels prevailing in the 1970s (1.9 percent in 1970, 1.7 percent in 1976). At the same time, because of changing epidemiological patterns (see below), the average cost of effectively treating These losses can be quite substantial. For example, household survey data from Ghana, Indonesia and Peru indicate that workers on the average loose about one day a month of work (compared to 1/4 in the United States), and that this translates into a loss in potential monthly income of between 3-7%. See Social Development is Economic Development, by Nancy Birdsall, Human Resources Development and Operations Policy Working Paper, World Bank, March 1993. See Charles Griffin, "Health Care in Asia: A Comparative Study of Cost and Financing", World Bank Regional and Sectoral Studies, Washington, D.C., Table 4.2. In the 1980s, government expenditure on health in Sri Lanka as a percentage of GDP averaged 1.5% in 1980-87 and reached a peak of 1.8% in 1988 and 1989.  - 56 - episodes of illness may be going up. This suggests that it may be reasonable, as a medium-term goal, to target an increase in government expenditure on health as a percentage of GDP, to ensure that Sri Lanka continues to be a leader in this area among developing countries. But it must be stressed that allocating more public resources to the health sector will not be productive unless Sri Lanka can effectively address several important issues concerning the composition of government spending in the health sector. Confronting these issues is necessary to ensure that the poor, and others using the public health services, have access to effective health services. 3.20 As with education, the decline in fertility provides an opportunity for using future real increases in government health expenditure to improve the quality of services, as opposed to expanding capacity to keep up with population. However, the decline in fertility together with rising life expectancy also imply changes in the age structure of the population, which is changing the nature of the challenges facing the public health services. Thus, while the number of children under five will decline in the 1990s, the number of people over 60 is expected to increase from about 1.4 million in 1991 to about 2.3 million by 2002. This latter development will increase the system's load of diseases with a high treatment cost, such as cancer and cardiovascular diseases. 3.21 Private health services are an important complement to public services. In 1986/87, private expenditures for health- related goods and services represented an estimated 38.4 percent of overall national health expenditures.8 The importance of private health services highlights the need to formulate an adequate regulatory framework for these services.9 It may also be desirable to redefine the division of labor between the public and private sectors in both the provision of services and the financing of access to services. For example, the public sector could concentrate more on preventive care, health education and communicable disease control, leaving most curative services to the private sector. However, such a model would require solving the problem of financing access to curative services in general for the poor, and for the non-poor with regard to chronic and catastrophic illness. 3.22 The changes in epidemiological patterns and the growth in private services call for a review of the role of government in the sector. As a result of this review, a plan for the long-term development of the public health services should be formulated, in order to guide future public investment in the sector. This process has already started, with the preparation of a National Health Policy Paper by a Presidential Commission in 1992. The policy review should also define the characteristics of a suitable regulatory framework for the private health sector.'" Detailed discussion of these topics exceeds the scope of this report. In what follows, attention will be focussed on several important issues concerning the composition of public expenditures in the health sector. The section then closes with a brief discussion of estate health services in the transition to privatization. National (public plus private) expenditure in health-related goods and services as a percentage of GDP was an estimated 3.4% in 1986-87, which was about average by comparison to most other South and East Asian countries. 2/ This should include an adequate regulation of the activities of doctors in government employment who also treat private patients, which they are legally entitled to do in recent years. There are allegations that many of these doctors are using public resources in treating their private patients, without making any payments for these resources. LO/ The regulatory framework would include not only the rules (legal framework) for the functioning of the private sector, but also the institutions to supervise and enforce the rules (institutional framework).  - 57 - Patterns of Government Expenditure on Health Services 3.23 A recent World Bank review of government expenditure on health services in Sri Lanka identified several important issues concerning the composition of that expenditure, which are briefly discussed below. 3.24 Insufficient Attention to Preventive Services. While mortality rates from infectious and parasitic diseases (i.e., malaria, dysentery, typhoid, amoebiasis, infectious hepatitis, etc.) continue to decline, the incidence of morbidity from these preventable diseases appears to be constant or, in the case of malaria, rising. Data on morbidity are limited to those seeking care at public health facilities." After a decline in the 1970s, the number of cases per year of malaria reported by public health facilities rose from about 51 thousand in 1980 to 129 thousand in 1992, with a corresponding increase in incidence from 345 per 100,000 population to 740 per 100,000 population. If the reported morbidity is representative of overall trends, the data suggest that the preventive aspects of the public health system need to receive greater emphasis. 3.25 Additional public expenditure directed to reducing the incidence of these diseases would have a high economic payoff, through reduction of the burden they impose on curative health services and of days of productive work lost by affected adults. It must be noted, however, that only part of this expenditure ought to be in health services as such. Most of these diseases are transmitted through contaminated food and water, which in turn are often the result of inadequate water and sanitation facilities. Provision of water supply and sanitation facilities will be discussed in the next section. 3.26 Unbalanced Functional Distribution of Current Expenditures. A rapid expansion of higher-level public health facilities has been taking place and it has been driving the priorities in the current budget. In the 1981-91 period, when total current government expenditure in health services increased in real tenns at an average annual rate of about 5 percent, the sub-set of that expenditure going to higher-level health facilities (Colombo area hospitals, including transfers to the quasi-public Sri Jayerwardenepura hospital; 16 specialized hospitals; 15 provincial hospitals; and 109 base and district hospitals) increased at an average annual rate of about 9.5 percent. In contrast, current expenditures for facilities at the primary level (100 peripheral units, 109 rural hospitals, and 620 central dispensaries and maternity homes) and preventive programs actually declined in real terms during the 1981-91 period. In 1981, current expenditure in higher-level facilities accounted for about 53 percent of total current health expenditure, while expenditure in primary level facilities and preventive programs accounted for almost 30 percent of the total; by 1991 current expenditure in higher-level facilities had risen to 80 percent of the total, while current expenditure in primary level facilities and preventive programs accounted for only 13 percent of the total. 3.27 A rough measure of the quality of services that an inpatient may expect from different types of facilities is given by the ratio of annual current expenditure per bed. While it is natural that higher-level facilities will have higher values of this ratio than lower-level facilities, the differences in Sri Lanka are enormous. For 1991, the average yearly current expenditure per bed for five teaching hospitals in the Colombo area was Rs. 92,612, and Rs. 77,082 for seven general district hospitals. In contrast, the average current expenditure per bed for 100 peripheral units and 109 rural hospitals was Li For 1986/87, it was estimated that about one-half of all sick people with symptoms of acute illness sought attention at public health facilities; see 1986/87 Demographic and Social Aspects Survey, Department of Census and Statistics, Colombo, 1992.  - 58 - Rs. 731 1.2 The authorities should study these ratios carefully to determine if efficiency gains could be obtained by reducing disparities across types of facilities; it is possible that redirecting resources towards improving quality of services at primary level facilities may reduce the present overcrowding of higher-level facilities and reduce overall health care costs. Such a redirection, moreover, is likely to be pro-poor, since poor people could be expected to have easier access to primary level facilities than to higher-level facilities."3 3.28 Declining Current Expenditure Relative to Capital Stock. The current expenditure implications of investments appear not to have been adequately budgeted for in recent years. Even taking into account that some proportion of annual investment represents replacement of existing capital stock, it is clear that the ratio of annual current expenditures to the health capital stock (the value of government health facilities and equipment at a point in time) has declined sharply since the mid-1970s, from an average of 0.9 in 1975-77 to 0.4 in 1990-92.'4 While it is difficult to determine what an adequate level of this ratio would be at any point in time, the decline raises the concern that an imbalance may have developed between spending on expanding facilities and spending on operating and maintaining facilities (more on maintenance below). This issue requires careful examination by the authorities. 3.29 Insufficient Expenditure on Repairs and Maintenance. Expenditure on repairs and maintenance of health facilities is clearly insufficient. An internationally accepted norm for such expenditures is 1.5 percent of the value of the capital stock per annum. In Sri Lanka, the value of the capital stock in public health facilities in 1992 was approximately Rs. 10 billion in current (1992) prices. Thus, an adequate allowance for repairs and maintenance in 1992 would have been about Rs. 150 million. In recent years, however, actual current expenditure for repairs and maintenance has amounted to only about one-fourth of the above norm, or about one-third if the rehabilitation and improvements component of the public investment budget is also taken into account. This low level of expenditure on repairs and maintenance of health facilities, if allowed to continue, will inevitably result in deterioration of these facilities and a consequent decline in the quality of services rendered. Estate Health Services 3.30 A separate discussion of health services for the estate sector is important for two reasons. First, the estates have had a separate system of health care. Secondly, their health indicators have compared unfavorably with those of the population at large.5 Table 3.1 below presents some selected estate health indicators for 1980, 1985 and 1990. In general, it can be seen that substantial progress was accomplished in the 1980-90 decade: there were large reductions in infant mortality; the percentage of births in institutions increased, and that of low birth weight babies declined. The L2/ It should be noted, however, that the discrepancies in expenditure would be less if expressed in terms of expenditure per inpatient-day rather than per bed, because higher-level facilities have higher occupancy ratios. L3J In Sri Lanka, access to the better hospitals is not based on patients' ability to meet price, since the services are free of charge. Instead, rationing occurs at the zero price, and such rationing is likely to be based on personal influence, which is directly correlated with income. L4' Mission estimates based on official data. L5/ Total estate population in 1981 (the year of the last census) was about 933 thousand, or 6.3% of the country's total.  - 59 - exception is maternal mortality, which appears to have increased during the period. This is puzzling, in view of the increase in the percentage of institutional deliveries (which should reduce risk to the mothers). It is possible that the increase merely reflects better reporting of maternal deaths that has accompanied the increase in institutional deliveries. What seems clear, however, is that health indicators continue to be worse in the estates than in the population at large, at least with regard to infant and maternal mortality. This can be seen by comparison with the corresponding figures in the last column of Table 1.10. 16 Table 3.1: Selected Health Indicators of the Estate Population in the 1980s 1980 1985 1990 Sri Lanka State Plantations Corporation (SLSPC) Estates Infant Mortality Rate a/ 72.7 44.9 31.0 Matemal Mortality Rate c/ 0.90 1.67 2.20 Institutional Births (%) 49.0 bx 68.0 90.00 Low Birth Weight (%) 36.0 b1 42.0 15.0 Janatha Estate Development Board (JEDB) Estates Infant Mortality Rate a/ 76.7 53.7 44.6 Maternal Mortality Rate c/ 1.00 0.71 1.60 Institutional Births (%) n.a. 53.5 70.8 Low Birth Weight (%) n.a. 48.7 d/ 20.4 Notes: / Number of deaths in first year of life per thousand live births. b1 Figure refers to 1982. C/ Number of maternal deaths per thousand live births. d/ Figure refers to 1983. Sources: Annual Health Returns 1978-86; Monthly Family Hlealth Returns 1987-90. 3.31 The improvement in the health indicators in the estates can be partly traced to the effects of nationalization in the mid-1970s, which was followed by the introduction of better living conditions and a comprehensive primary health care package.'7 Both government corporations in the estates, the JEDB and the SLSPC, had a Social Development Division which could intervene if the need arose, to ensure that appropriate health services were delivered. In 1992, 22 private companies were allocated most of the estates previously under the JEBD and SLSPC, under management contracts. In order to address the provision of health and other social services under the new organization, the companies and several other interested parties established the Plantation Housing and Social Welfare Trust.'8 The recurrent expenditure of the Trust is met through a levy charged on the 22 private companies. L6/ Again, however, it could be that part of the difference in the indicators of estates and the population at large is due to more accurate reporting in the estates. This requires further investigation. These efforts were assisted by the Social Welfare Program I, supported by the Norwegian and Dutch governments. A second phase of this program will be implemented in the 1993-97 period. 18/ Education services, however, were taken over by the government in the estates now under the 22 private companies.  - 60 - 3.32 However, under its articles of association, the Trust is only a coordinating body, and it does not have any authority to formulate, implement or enforce health policies in the estates. Such authority has been placed under the management of the companies. This situation has given rise to a concern among some in the medical community and government that health services in the estates might deteriorate. In the pre-nationalization era, it had been observed that private managers tended to overlook preventive services such as maternal and child health services, and concentrated in providing curative care instead (after nationalization, priorities were changed to emphasize preventive services). 3.33 The Ministry of Health has prepared interim guidelines for the provision of health services in the estates. Further, it plans to appoint a number of Medical Officers of Health to work full-time on the estates. These measures are a step in the right direction. However, the real solution may lie in the government taking over the health services in the estates, as it has already done with education. This may be the only way to ensure that the estate population has access to the same quality of services as the rest of the population. Nutrition 3.34 As discussed in Chapter 1, the incidence of undernutrition among young children appears to be high by reference to international standards. It was also noted that consumption poverty appears to be only part of the explanation for undernutrition among young children. 3.35 Remedies to Undernutrition. Generally speaking, remedies to undernutrition in young children fall into two categories, namely, those that aim at the long-run reduction of the problem, and short-term palliative programs. The long-run remedies include: (i) economic reforms that raise real incomes of households and lower the retail price of food in real terms (e.g., liberalizing food imports); (ii) interventions that reduce the incidence of infectious and parasitic diseases, such as universal immunization, universal access to safe water and adequate sanitation facilities, and hygiene education; and (iii) improving feeding practices through education of the mothers. Short-term palliatives are programs of food supplementation directed to at-risk children and pregnant and lactating women. In Sri Lanka, the main directed program is the Triposha program, which was introduced in 1973. 3.36 The Triposha Program'9 This is a food supplementation program based on the free distribution by government of a manufactured foodstuff (itself called "triposha") whose main ingredients are corn, soya and non-fat dried milk. Target groups are the nutritionally-at-risk from among the following: pregnant mothers for a period of six months from the third month of pregnancy; lactating mothers up to six months from delivery; infants from six months of age to twelve months; and pre-school children from one through four years of age. Beneficiaries are selected by health workers at government clinics based on certain guidelines which reflect risk of malnutrition (e.g., for children, being below a certain percentage of the normal weight-for-age, or not gaining weight adequately). Selected beneficiaries are registered and provided with a monthly ration of "triposha" which they can take with them and consume at home. Every daily intake contains 200 calories, 10 grams of proteins and essential micronutrients. 3.37 The number of beneficiaries is large. In 1988, for example, the program reached nearly 600,000 people, broken down as follows: 84,000 pregnant women (49% of the total number of pregnant women in that year), 60,000 lactating women (35% of total), 96,000 children aged 7-12 months (67% of all children in that age bracket), and 354,000 children aged 13-59 months (26% of all 191 This section is mostly based on A Framework For a Comprehensive Social Welfare Policy, Marga Institute, Colombo, 1991.  - 61 - children in that age bracket). Annual average cost per beneficiary in that year was about Rs. 200 (about US$6). 3.38 There have been several large-scale evaluations of the program, the last in 1983. Evaluations have been positive on the whole, finding evidence of improvement of nutritional status after participation in the program. However, there have been no longitudinal studies to evaluate the long-term impact on the nutritional status of participants. The evaluations also concluded that most of the ostensible beneficiaries do belong to the target groups specified in the program. However, there is clear evidence that "triposha" is freely shared among members of the family, which reduces the program's impact on the intended beneficiaries. Moreover, as discussed in Chapter 1, the incidence of low birth weight in Sri Lanka continues to be high, and so is the incidence of undernutrition in young children as measured by anthropometric indicators (although the two recent 1993 surveys indicate a significant decline). While it is possible that these indicators of undernutrition would have been worse if the Triposha program had not been introduced, it seems clear that the program in its present form is not up to the task of eliminating the problem.20 3.39 A micro-level survey conducted by the Marga Institute in 1990 identified some important deficiencies in the Triposha program. In addition to the tendency, mentioned above, of beneficiaries (especially mothers) to share the food supplement with other members of the household, the survey found that only a small proportion participate in the program with the regularity needed to obtain the program's full benefits. Also, monitoring of weight gain and growth by the health services was not satisfactory. 3.40 In conclusion, while the Triposha program addresses an important social need, it would seem that it is not making a significant dent on the problem of undernutrition of pregnant and lactating women and pre-school children, and that the program itself suffers from some deficiencies which are in part due to behavior of the beneficiaries and their families and in part to behavior of the health workers administering the program. 3.41 A Strategy to Address Undernutrition. In addition to revisiting the Triposha program with an aim to improve its effectiveness, a strategy for addressing undernutrition should focus on the long-run remedies identified in para. 3.35. Access to safe water and sanitation is discussed in the next section. Other interventions to be emphasized are nutrition education for mothers, hygiene education for mothers and children, and universal immunization against infectious diseases. The leadership in strengthening these interventions must be taken by the Ministry of Health, but other government institutions (e.g., schools) should also participate actively. C. Water Supply and Sanitation 3.42 Availability to households of safe sources of potable water and adequate sanitation (excreta disposal) facilities are a critical complement to the provision of health services. As with those services, the government should play a role in the provision of safe water and sanitation facilities, with an aim to ensuring universal coverage of the population. This objective has been adopted in Sri Lanka. A committee established by the government in 1980 set the objective of achieving complete coverage in both urban and rural areas by 1995. Funding and absorptive capacity constraints have caused the target date to be subsequently moved to the year 2000. 20/ One important side benefit of the Triposha program, however, is that it helps to bring children to the clinics, thus facilitating the provision of other health services such as immunization.  - 62 - Present Coverage 3.43 It has been estimated that at the end of 1990 around 67 percent of the population had access to some type of formal water supply facility.2' A formal source is defined as a piped system for the urban sector and either a piped system, protected open well, or handpump for the rural sector. Access was better for urban areas than for rural areas (76 percent versus 64 percent), and there were considerable differences among provinces, as indicated in Table 3.2 below. Using the same definition, in 1981 only about 54 percent of the population had formal water facilities, which shows considerable progress achieved during the 1980s. About 59 percent of the population (68 percent in urban areas, 56 percent in rural areas) was estimated to have access to satisfactory sanitation facilities in 1989 (Table 3.2), up from 48 percent in 1981.22 3.44 The above figures on water coverage should be qualified to note that such coverage not always provides fully satisfactory service. If "satisfactory quality" is defined as a 24 hours a day supply for piped systems; a functioning handpump which produces water of acceptable taste and quality; and an open well which is provided with an adequate perimeter wall, of the 11.6 million persons served with some kind of formal water supply facility in 1990, only about 42 percent had access to fully satisfactory service.23 3.45 In sum, this brief discussion of water supply and sanitation coverage indicates that: (i) while substantial progress has been made in recent years, there is still a long way to go to achieve universal coverage with fully satisfactory services; and (ii) there are important disparities in coverage among provinces and between urban and rural areas. Government Strategy 3.46 As indicated above, the government has a target of providing adequate water supply and sanitation facilities to the entire population by the year 2000. This would include not only serving households not now covered, but also taking care of a large proportion of the existing facilities which are in need of rehabilitation, particularly piped water supply schemes. 3.47 Estimates by the National Water Supply and Drainage Board (NWSDB) conducted in 1991 indicate that to achieve universal water supply coverage by the year 2000, the numbers of people that would have to be provided with either new or rehabilitated facilities in the 1992-2000 period are approximately as follows: piped supply, 3.1 million (of which 54 percent new); handpump, 1.8 million (of which 58 percent new); and protected open well, 8.5 million (of which 55 percent new). About 86 percent of the people that would need to be provided with either new or rehabilitated water supply facilities in the period live in rural areas.24 To achieve the target of universal coverage 21/ National Water Supply and Drainage Board (NWSDB), Corporate Plan, Colombo, 1991. 221 National Planning Department, A Profile of Child Development in Sri Lanka, Colombo, 1991, page 49. 23/ Ibidem, page 47. Based on estimates of the National Water Supply and Drainage Board, 1991. 24/ This assumes that the percentage of rural population will continue to be about 80% throughout the 1992-2000 period. The NWSDB's estimates of target population to be provided with new or rehabilitated water supply facilities appear to be too high, however, because they assume an implicit annual rate of population growth of about 2.1% in the 1990-2000 decade, while according to Bank projections this rate would be only about 1.1%.  - 63 - with adequate sanitation facilities, NWSDB estimates are that about 1.8 million new individual facilities would be needed in the 1992-2000 period, mainly pour-flush latrines. This does not take into account rehabilitation of existing sanitation facilities.25 3.48 The total capital cost of achieving these targets in constant 1991 prices was estimated by NWSDB at about Rs. 12 billion for water supply, of which Rs. 5.7 billion for piped water supply schemes, Rs. 0.8 billion for handpump wells, and Rs. 5.5 billion for protected open wells. For sanitation, the capital cost of achieving universal coverage was estimated to be about Rs. 8.8 billion in 1991 prices. Thus the total capital cost of achieving universal coverage by the year 2000 for both water and sanitation according to the NWSDB would be about Rs. 21 billion in 1991 prices (about US$500 million). Table 3.2: Coverage of Water Supply and Sanitation Facilities, 1990 Water Supply Percent Percent Total Percentage Population Population Water of House- Served in Served in (%) Holds With Province Urban Areas Rural Areas Latrines Westem 82 63 72 76.2 Southern 83 58 62 64.9 Central 100 66 70 71.0 Sabaragamuwa 77 52 54 67.4 North-Westem 81 58 59 44.0 North-Central 78 86 85 45.0 Uva 94 59 61 65.0 Northern 49 74 68 41.5 Eastern 44 68 63 30.3 Total 76 64 67 59.0 Sources: National Water Supply and Drainage Board, Corporate Plan, 1991 (for water); National Health Development Plan, 1989, Ministry of Health (for sanitation). Public Investment in Water Supply and Sanitation 3.49 The section on Water Supply and Sanitation in the 1992-96 Public Investment Program shows a total of Rs. 26.2 billion in constant 1992 prices allocated to capital expenditures in the sector, of which Rs. 9.6 billion for ongoing schemes and Rs. 16.6 billion for new schemes. Even taking into account construction cost increases from 1991 to 1992, this overall amount should suffice to cover the capital cost of achieving universal coverage as estimated by the NWSDB (see previous paragraph). In reality, however, the expenditures programmed under the Water Supply and Sanitation section of the Public Investment Program reflect the investment program of the NWSDB, which concentrates on piped water supply schemes (urban schemes and major rural schemes) and to a much lesser extent on handpumps and sanitation facilities. It is not involved in assisting with construction of protected open wells. The strong urban bias of the NWSDB's investment program can be appreciated by noting that two new urban water supply projects (the Towns South of Colombo Water Supply 25/ See National Planning Department, A Profile of Child Development in Sri Lanka, page 50.  - 64 - Project and the Towns North of Colombo Water Supply Project) account for three-fourths of total investment for new projects in the NWSDB's 1992-96 investment program. 3.50 The data in the Public Investment Program is not organized by province. However, an attempt was made by the mission to obtain the provincial distribution of NWSDB's investment program. This is shown in Table 3.3 below. The figures show that the provincial distribution is quite inequitable. The Western Province and, to a lesser extent, the Central Province, are programmed to receive per capita investments much higher than the other provinces; moreover, both of these provinces already have water supply coverage better than the national average of 67 percent. On the other hand, the province with the lowest water supply coverage, Sabaragamuwa, is programmed to receive the second lowest per capita investment. Table 3.3: NWSDB Public Investment Program 1992-96 by Province Total PIP Population Per Capita % Coverage 1992-96 1991 PIP Water Province Rs. Million (OOOs) Rs. Supply Western 14,961 4,442 3,368 72 Central 2,795 2,220 1,259 70 North Central 666 1,035 643 85 North Westem 921 2,026 455 59 Southern 956 2,239 427 62 Uva 338 1,069 316 61 Sabaragamuwa 500 1,692 296 54 Northern & Eastern 667 2,524 264 66 Allocation Unknown 4,377 Total 26,181 17,247 1,518 67 Source: Mission estimates, based on 1992-96 Public Investment Program. 3.51 The above NWSDB investment program is not the total public investment on water supply and sanitation envisaged for the 1992-96 period. Also active in this area are local governments (provincial councils, municipal councils, urban councils); the Ministry of Health, which encourages good sanitation practices by households through its health education programs; and the Mahaweli Development Board, which is responsible for provision of water supply and sanitation facilities in the Mahaweli area. There are, moreover, water and sanitation components under the various Integrated Rural Development Projects implemented under the auspices of the Ministry of Policy Planning and Implementation. 3.52 It should also be noted that the Ministry of Housing, Construction and Public Utilities, which houses the NWSDB, has started to place a greater emphasis on serving rural areas and the urban poor. Through its Community Water Supply and Sanitation Program, it has begun to assist rural communities in three districts to establish adequate water and sanitation facilities, including the development of systems and procedures for operation and maintenance of the facilities by the communities. This program would be expanded nationwide starting in 1996. A new Clean Settlement Program is also being started by the Ministry, focussing on poor urban areas. It would help communities in urban slums to upgrade their water and sanitation facilities and effect other environmental improvements.  - 65 - 3.53 In addition to public investment, there are private household expenditures on water and sanitation facilities, and a number of indigenous and international Non-Governmental Organizations are active in the sector. A complete data base of all these investment expenditures, or just of the public ones, is not currently available. Hence, it is difficult to assess the speed with which the country is moving towards achieving the target of universal water supply and sanitation coverage. The information base, and coordination within the sector, might improve as a result of the recent creation of a National Water and Sanitation Sector Coordination Committee within the Ministry of Housing and Construction. 3.54 The new Coordination Committee should work closely with the local governments, which are likely to be the key to achieving universal water and sanitation coverage in rural areas. Local governments will need considerable technical assistance in the water and sanitation area if they are going to play an effective role in expanding coverage and ensuring adequate maintenance of facilities. The government should also consider the possible scope for diverting some of the funds in the NWSDB investment program to expanding coverage in rural areas, and it should take steps to improve the equity of the provincial distribution of the NWSDB investment program.  - 66 - CHAPTER IV - TRANSFERS, SAFETY NETS AND NGOs 4.01 In Sri Lanka, as in many other developing countries, a private safety net in the form of private transfers exists and is an important mechanism for alleviating the effects of economic misfortune; we noted in Chapter 1I that most of the unemployed receive private transfers from relatives. Another important component of the private safety net is comprised of the activities of a large number of Non-Governmental Organizations (NGOs). The government has attempted to complement this private safety net with a variety of transfer programs. In this chapter, the main transfer programs are critically reviewed. The chapter also briefly discusses the activities of NGOs, and what can be done by the public sector to enhance their effectiveness. A. Transfer Programs 4.02 The Sri Lankan government has an array of programs to assist certain groups of the population with cash or food coupons, with an aim to achieve certain equity or social goals. The main programs, which together amounted to government spending equivalent to about 2.5 percent of GDP in 1992, are: the Poor Relief Food Stamp Program (FSP); the Janasaviya Program (JSP); the Mid-Day Meal Program (MDMP); Public Assistance; and programs of assistance to population displaced by the civil conflict. Below, the programs are briefly described. An assessment of the programs and some suggestions for their restructuring is then presented. The Poor Relief Food Stamp Program (FSP; 0.7% of GDP in 1992) 4.03 In the mid-1970s, food subsidies had attained a fiscal cost equivalent to about 5 percent of GDP. The main food subsidy program was the rice ration scheme, but there were also generalized subsidies on bread and wheat. The rice ration scheme provided every citizen with a ration of the staple food, rice, at subsidized prices. By the late 1970s the government concluded that the cost of the subsidies was unsustainable. In 1978, it restricted the coverage of the rice ration scheme to the poorest half of the population. In 1979, it replaced the rice ration scheme with the FSP, which covered approximately the same group of the population but entailed a somewhat lower total transfer. By 1980, the government had removed all other subsidies on food. As a result of these changes, the fiscal cost of food subsidies declined to just 1.3 percent of GDP by 1984. 4.04 The FSP is intended to improve the nutritional status of poor households by increasing their disposable income with monthly coupons that can be exchanged for food in special cooperative stores. Initially, the program was targeted to households with self-reported incomes of less than Rs. 300 a month, and in the beginning the rolls included some 1.55 million households, or about 50 percent of all households. The eligibility ceiling (or "poverty line") of Rs. 300 per household per month was estimated to correspond to a caloric consumption equivalent to 60 percent of adequacy, under the assumption that these households spent 75 percent of their total income on food. The stamps had a fixed cash value which was differentiated according to the age of the beneficiary, with children under 8 years receiving the stamp with the highest value.' 4.05 It is clear that, from the start, the income ceiling of the program was not respected, as the following calculation suggests. From 1979 to 1982, the CPI rose by a total of about 65 percent. If this increase is applied to the ceiling for the FSP program, it follows that by 1982 the ceiling should have been about Rs. 500 per household per month. When this new theoretical ceiling is compared with the income distribution of households estimated from the Central Bank's 1981/82 Consumer One important side benefit of the Triposha program, however, is that it helps to bring children to the clinics, thus facilitating the provision of other health services such as immunization.  - 67 - Finances survey, it turns out that the percentage of households in the sample with monthly incomes below Rs. 500 in 1981/82 was only about 10 percent.2 Moreover, the number of beneficiaries steadily increased during the first year of the program (1979), and a freeze on new entries was put into effect in March 1980. By 1986, the number of recipient households had declined to 1.48 million, and some major problems had become apparent: there were sizeable leakages of program benefits to the non-poor, while the newly poor could not enter the program; moveover, inflation had cut the real value of food stamps by more than half. 4.06 In 1986, the income limit for eligibility was raised to Rs. 700 per household per month, and food stamps started to be issued on the following basis: households with monthly incomes of Rs. 300 and below, for all members; households with incomes between Rs. 301-399, for up to four members; households with incomes between Rs. 400-599, for up to three members; and households with incomes between Rs. 600-700, for up to two members. These criteria for eligibility have remained unchanged since then. The value of the stamps was not changed, however, until January 1989, when it was doubled (and has remained at that level since then). 4.07 In 1991 a re-targeting exercise was attempted. The screening was essentially based on the JSP screening process (see below). As a result of this exercise3, it was determined that about 634 thousand households were ineligible to receive food stamps. On the other hand, it was determined that there were about 132 thousand households that should have been receiving food stamps but were not. However, as a result of political pressures and a drought that occurred around that time, the ineligible households were not removed. By the end of 1991 the number of beneficiary households in the FSP was still very high, at about 1.3 million households, excluding the Northern and Eastern provinces. This represented about 45 percent of all households excluding those provinces. 4.08 It is clear that the magnitude of "leakages" in the FSP continues to be very high. The CPI rose from 60.6 in 1979 to 302.9 in 1992 (with 1982=100), or almost exactly a five-fold increase. Thus, if the initial household income ceiling of Rs. 300 per month had been updated using this index, it would have amounted to Rs. 1,500 per month in 1992. This would have been lower than the top of the second lowest decile of the household income distribution (about Rs. 1,620).4 Excluding the Northern and Eastern provinces, each decile contains about 293 thousand households (in 1992); hence, the lowest two deciles contain about 586 thousand households. According to this calculation, then, if the original income ceiling for program eligibility had been held constant in real terms, the number of beneficiary households in the FSP in 1992 should have been less than 586 thousand, excluding the Northern and Eastern provinces. Thus, the actual number of beneficiary households in 1992 was more than twice as high as would have been justified by an updating of the original household income ceiling. 4.09 An analysis of the distribution of the benefits of the FSP between "poor" and "non- poor" recipients was also conducted for this report, using figures from the 1990-91 Household Income See Central Bank of Sri Lanka, Report on Consumer Finances and Socioeconomic Survey 1981/82, Colombo, October 1984, page 182. The Northern and Eastern provinces were excluded from the exercise because of the civil conflict. These figure is obtained by taking the corresponding estimate for 1990 from the Preliminary Report of the 1990/91 Household Income and Expenditure Survey of the DCS, and inflating it by 24% (the increase in the CPI from July 1990 to July 1992). The Preliminary Report is based on the first three rounds of the Survey, which were taken in June-July-August of 1990.  - 68 - and Expenditure Survey of the DCS. According to this analysis, in 1990 about 43 percent of the population (excluding the North and East) belonged to households receiving FSP benefits. Forty percent of the recipients were "poor" by their pre-transfer per capita consumption, using the reference poverty line (Table 1.4); 60 percent were "non-poor" by this criterion. Of all the poor, about 70 percent were FSP recipients, while some 30 percent were not being reached by the FSP. In sum, this analysis confirms that there are substantial leakages to the non-poor in the FSP, but it also shows that the program does reach a large majority of the poor. 4.10 At the same time, the real value of benefits per household has declined sharply since 1979. As indicated above, the value of food stamp entitlements was increased only once in the period since 1979, i.e., it was doubled in 1989; while the CPI increased five-fold from 1979 to 1992. The average household benefit in 1992 was about Rs. 170 per month. If entitlements had retained their value in real terms, that figure should have been around Rs. 425 instead. 4.11 Currently, the FSP is undergoing substantial change as large numbers of households are being moved from that program into the JSP. Once JSP has begun in a given Assistant Government Agent (AGA) division, the only households remaining on FSP rolls are (at least in theory) those households receiving Public Assistance (see section on Programs for the Destitute Poor, below); these households are not eligible for participation in the JSP because it is assumed they would have no chance to lift themselves out of poverty if they did. In this sense, FSP is increasingly becoming a narrowly-targeted income transfer program that reinforces the Public Assistance program. Nevertheless, targeting problems persist in those areas where the JSP has not yet arrived. The Janasaviya Program (JSP; 0.7% of GDP in 1992) 4.12 This program was introduced in October 1989. Its basic concept was to provide poor households with fixed monthly cash grants for a period of two years, in the expectation that this would provide a "breathing space" that would enable some of their members to become microentrepreneurs, or alternatively acquire skills that would enable them to get better paying jobs, thus enabling the households to exit poverty. The JSP is being phased in throughout the country in successive, geographically-defined "rounds", with the least affluent areas (as measured by proportion of households in FSP) being covered first. A total of eleven rounds are planned; five have already been launched, and the remaining six rounds would be launched within the next several years.5 Households selected for inclusion in the JSP are taken off the FSP rolls.' 4.13 Monthly grants under JSP are fixed at Rs. 1,458 per household,' comprising a cash "consumption" transfer of Rs. 1,000 released every month and a "savings" component of Rs. 458 which is only made available to the households at the end of the two-year period on a lump sum, to be used for investment projects. Moreover, JSP beneficiary households were also supposed to get a lump sum grant of Rs. 25,000 at the end of the two-year period. Because of fiscal constraints this has not happened, but a monthly "interest" payment of Rs. 250 per household is being paid instead for the indefinite future to households that have completed the two-year period (unless a member of the The fifth round was launched in May 1994. -' However, this procedure is reportedly not being followed in the recently launched fifth round of the JSP. This is approximately equal to the average monthly household income of the third lowest decile of the household income distribution in 1990 (Rs. 1,403), according to the preliminary results of the 1990/91 Household Income and Expenditure Survey.  - 69 - household gets a government job, or a job in the government sponsored 200-garment factory program, in which case the payment is discontinued). 4.14 In return for the cash payments, beneficiary households must provide 20 days of labor per montlh during the two-year period, which can either be spent on training or private investment projects of the beneficiaries themselves, or working on community infrastructure projects. This latter aspect gives the JSP some resemblance to a public works employment program; however, in a well- designed public works program the wages paid would be very low to encourage self-targeting by the poorest segments of the labor force. 4.15 Withi regard to the eligibility criteria, officially the JSP is limited to households with monthly incomes below Rs. 700, i.e. the same ceiling as (in theory) in the FSP. If strictly enforced, this would have been quite restrictive, since in 1990 the top monthly household income of the lowest decile of the household income distribution was Rs. 874.8 However, the ceiling was not applied in practice. The Women's Bureau of Sri Lanka (an agency within the Ministry of Health and Women's Affairs) conducted a random survey of 299 Round I beneficiary households shortly after they had completed the 24 monthls of the JSP. The survey found that nearly 30 percent of these households had incomes above Rs. 1,500 per month prior to entering the JSP.9 4.16 Subsequent rounds, however, have made increasing use of community participation methods to identify eligible households. In practice, the official eligibility ceiling of Rs. 700 per household per month has been replaced by a set of detailed, region-specific criteria issued by the Janasaviya Commission and framed in terms of such "observable" variables as ownership of land and consumer durables and sources of income. Village-level selection committees prepare lists of eligible houselholds that are publicized in the village, with households ranked in descending order of poverty according to these criteria. Anonymous complaints are then gathered and evaluated by the committee. This method appears to yield substantial improvements in targeting relative to past experiences with the FSP (which was based on self-reporting of income). 4.17 The latter statement can be illustrated with the following calculation. Round I of JSP (started in October 1989) included 169,000 households, of which 42,000 in the North and East were later removed. Round 2 (started in December 1990) included 104,000 households, and Round 3 (started in February 1992) included 101,000 households. Round 4 (started in March 1993) includes about 100,000 households. Round 5 (launched in May 1994) is expected to cover 136,000 households, but it would cover some 26 AGS districts as compared to 19 districts in Round 4. Assuming that the remaining six rounds have an average of 90,000 households each (less than the early rounds because they would cover more affluent areas), the grand total of households covered by the program would be roughly around 1,100,000, or somewhat less than one-third of all households existing at present.'