S P D I S C U S S I O N PA P E R NO. 0821 44927 Population Aging and the Labor Market: The Case of Sri Lanka Milan Vodopivec Nisha Arunatilake August 2008 POPULATION AGING AND THE LABOR MARKET: THE CASE OF SRI LANKA Milan Vodopivec (World Bank) Nisha Arunatilake (Institute for Policy Studies, Colombo) August 2008 Abstract Sri Lanka's population is predicted to age vary fast during the next 50 years, bringing a slowdown of labor force growth and after 2030 its contraction. Based on a 2006 representative survey of old people in Sri Lanka, the paper examines labor market consequences of this process, focusing on retirement pathways and the determinants of labor market withdrawal. The paper finds that a vast majority of Sri Lankan old workers are engaged in the informal sector, work long hours, and are paid less than younger workers. Moreover, the paper shows that labor market duality carries over to old age: (i) previous employment is the most important predictor of the retirement pathway; (ii) older workers fall into two categories: civil servants and formal private sector workers, who generally stop working before they reach 60 because they are forced to do so by mandatory retirement regulations, and casual workers and the self-employed, who work until very old age (or death) due to poverty and insufficient income and who stop working primarily because of poor health; and (iii) the option of part-time work is used primarily by workers who held regular jobs in their prime age employment, but not by casual workers and self-employed. JEL Classification: J11, J14, J26 Keywords: population aging, labor supply of old workers, labor demand for old workers TABLE OF CONTENTS 1. INTRODUCTION ........................................................................................................ 1 2. LABOR SUPPLY AND GROWTH IMPLICATIONS OF POPULATION AGING.. 5 3. WORKING IN OLD AGE............................................................................................ 9 4. PATHWAYS TO RETIREMENT................................................................................ 7 5. DETERMINANTS OF WORK ACTIVITY OF THE ELDERLY............................ 12 6. OBSTACLES FOR EMPLOYMENT OF OLD WORKERS .................................... 18 7. CONCLUDING REMARKS...................................................................................... 23 TABLES Table 1: Labor force participation by age group from 1992-2004 (%)................................................10 Table 2: Unemployment rate by age group from 1992-2004 (%) ........................................................12 Table 3: International comparisons of labor force participation rates for elderly ................................14 Table 4: Reasons for complete withdrawal from work of old workers..................................................5 Table 5: Descriptive statistics of variables used in multinomoial logit estimation ..............................14 Table 6: Multinomial logit estimates of transition from full-time employment1 .................................16 FIGURES Figure 1: Demographic aging, Sri Lanka and world regional averages (percentage of population over 60 years old, 2000-2050)................................................................................................................1 Figure 2: Trends in dependency ratios, Sri Lanka 2001-2101................................................................5 Figure 3: Labor force projections, 2001-2070........................................................................................6 Figure 4: Age composition of standard labor force projections, by gender (2001-2070) ......................8 Figure 8: Unemployment rates by age and gender (1992 and 2004.....................................................11 Figure 9: Retirement status by age and gender (percent) .......................................................................1 Figure 10: Average number of hours worked, by age and gender,.........................................................2 Figure 11: Age wage profile, by sector and gender...............................................................................2 Figure 12: Old workers by type of employment status and type of employer .......................................4 Figure 13: Old workers by sector of work and occupation ....................................................................4 Figure 14: Retirement status, by prime-age employment type and age .................................................9 Figure 15: Percentage of retirements by age groups ............................................................................10 Figure 16: Sector of work, by the prime-age employment type and age..............................................11 Figure 17: Reasons for retirement from full time work........................................................................12 1 THE LABOR MARKET AND POPULATION AGING: THE CASE OF SRI LANKA* 1. INTRODUCTION In slightly more than two decades, Sri Lanka's population will be as old as Europe's or Japan's today, but its per capita income will be much lower. Indeed, while population aging is a universal Figure 1: Demographic aging, Sri Lanka and world regional phenomenon, it looms averages (percentage of population over 60 years old, 2000- particularly large for 2050) Sri Lanka: not only is 35 Sri Lanka 30 its population among As ia 25 the oldest in the non- t OECD 20 developed world, but cenreP15 Latin America and Caribbean Eas tern Euro pe Sri Lanka is also one of 10 and fo rmer SU 5 Middle Eas t and the fastest aging No rth Africa 0 Sub-Saharan countries in the world 2000 2010 2020 2030 2050 Africa (Figure 1). Spectacular growth would be Source: De Silva (2007), for Sri Lanka; World Bank (1994), for required for Sri Lanka other countries to catch up the developed countries' current aging profile at a level of per capita income these countries currently enjoy. * The authors are Milvan Vodopivec, The World Bank Group, mvodopivec@worldbank.org and Nisha Arunatilake, Institute for Policy Studies, Colombo. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. The paper is part of a broader World Bank study investigating the consequences of population aging in Sri Lanka. Partial funding under the TFESSD trust fund grant is gratefully acknowledged. The authors are thankful for useful comments by Gordon Betcherman, Ralitza Dimova, Ravi Peiris, Ravi Rannan-Eliya, Shyamali Ranaraja, Mansoora Rashid, and Nimnath Withanachchi, and for excellent research assistance to Priyanka Jayawardena In developed countries, challenges of population aging are associated primarily with the negative impact of aging on economic growth, and the need to plan for additional public and private outlays for old age income support and healthcare. In Sri Lanka, a developing country, these challenges are compounded by the fact that many old people, now and in the future, lack social security coverage and may therefore face poverty; that the effectiveness of traditional, family support of old people may be reduced; that many old people are deprived of choice, as they are forced to work late in their life and only retire because of health reasons; and that the health system is insufficiently focused on the healthcare needs of elderly and is constrained by the lack of resources and their inequitable distribution. An aging population will therefore create serious challenges for Sri Lanka on several fronts. First, rapid aging and modernization are likely to strain the traditional family support to old people. The modernization of Asian societies is associated by several mechanisms that affect family support of old persons (Hermalin 2002): (i) lower fertility translates into fewer children available to provide familial support; (ii) higher education levels increase the information gap between old parents and children, leading to differences in attitudes and perceptions of obligations to provide familial support; (iii) increasing female labor force participation decreases the number of caregivers available to provide support to older family members; and (iv) rural-urban migration is drawing younger persons out of rural areas and into urban areas to find employment. While modernization is affecting Asian societies in different ways, there is evidence of declining levels of familial support in some Asian countries (including China, India, as well as more developed countries such as Japan and South Korea). For Sri Lanka, World Bank (2007a) finds that lower fertility is significantly related to lower likelihood of co-residence and receipt of intergenerational transfers; that old persons in Sri Lanka aged 60-69 are more likely to value independence compared to elderly aged 80 and above; and that adult children with higher education are increasingly likely to set up separate residence from their parents after marriage. Second, an aging population will increase costs associated with aging (old age income support, health care and long-term care). Future increases of public old age income support costs are particularly worrisome, as the fiscal sustainability of existing social security schemes is questionable. At 2 percent of GDP, Sri Lanka has the highest pension spending on 2 civil service pensions in the region absorbing more than 10 percent of government revenues (World Bank 2007a). With the exception of the Employee's Provident Fund (the main mandatory retirement scheme for the formal private sector) which is self-financing by definition), existing schemes for informal sector workers will also prove to be financially unsustainable if benefits, currently fixed in nominal terms, are adjusted in the future ­ or they will become irrelevant, if the benefits are not adjusted. After pensions, the largest fiscal impacts of aging in terms of increased social outlays arise from healthcare and long-term care costs. World Bank (2007a) estimates that, given current trends and expenditure patterns, population aging itself may add up to 1 percent of GDP by 2050 to the current healthcare expenditure level of 4.2 percents of GDP. A separate health policy question is how to foster healthy aging, in particular, how to deal with the increases in non-communicable diseases and mortality. And third, population aging will not only increase the burden of economically active population, but it may also act as a hindrance to GDP and overall welfare growth. Population aging is predicted to slow down its growth and, within two to three decades, start to reduce labor supply ­ one of the key production inputs. Population aging will also dramatically change the structure of labor force, the change that may affect productivity of the labor force on its own right. This paper focuses on the last challenge ­ the impact of population aging on the labor market. The analysis focuses on the current labor market outcomes of old workers and attempts to identify both the work incentives of old people and the constraints faced by them (such as poor health and lack of employment opportunities).1 The paper addresses the following key questions: how to promote the employment of workers who want to work in old age? On the other hand, do social policies and practices force workers to work in their old age, and how to help those workers? In face of increased longevity, will the increase of the need for long-term care of elderly translate to increased pressures on 40-60 year-old women 1The primary source for the analysis is the World Bank 2006 Sri Lanka Aging Survey conducted in early 2006. The total number of interviewed persons was 2413, selected randomly among the population of 60 year olds and above. The sample was drawn based on pre-listing of old persons in randomly selected 260 primary sampling units (North-East region was excluded). The questionnaire consisted of several modules, with the present paper using the following ones: background and basic data, activities in which old people have been engaged, income and assets, and health and health care use. 3 to abstain from labor market participation, particularly as falling fertility have reduced the number of children per family? What changes in the labor market legislation and other labor market policies and practices are needed to accommodate the slowdown of the labor force growth ­ and, after 2030, its shrinking? Driving this analysis is a concern to promote choices available to old workers, choices that would both improve job prospects of those who would like to work as well as reduce pressures on those who are forced to work. The main findings of the paper can be summarized as follows. In Sri Lanka, a vast majority of old workers is engaged in the informal sector, where they work as self-employed or casual workers. Most of them have full-time rather than part-time jobs, work long hours, and are paid less than younger workers. Moreover and most notably, the labor market duality, a hallmark of labor market in developing countries, carries over to old age and importantly determines the fate of old people: (i) the type of previous employment is the most important predictor of the retirement pathway; (ii) older workers generally fall into two categories: civil servants and formal private sector workers, who stop working earlier than they would like (before they reach 60) because labor market rigidities force them to do so, and casual workers and the self-employed, who work until very old age (or death) due to poverty and insufficient income and who stop working primarily because of poor health; and (iii) the option of part-time work is also used primarily by workers who held regular jobs in their prime age employment, but not by casual workers and self-employed. The paper is organized as follows. To set the stage, Section 2 presents long term labor force projections and discusses economic implications of the slowdown of labor force growth and its shrinking. The next sections analyze labor market outcomes of old workers: Section 3 describes key employment and pay outcomes for old workers and the patterns of their withdrawal from working life; Section 4 examines pathways to retirement, that is, typical ways in which workers transit from full-time employment to complete retirement; and Section 5 explores the determinants of work activity of old workers with multinomial logit models. Section 6 discusses possible obstacles for employment in old age, including labor market legislation and wage-setting practices. Section 7 concludes with policy implications. 4 2. LABOR SUPPLY AND GROWTH IMPLICATIONS OF POPULATION AGING Sri Lanka's population is set to age rapidly during the next 50 years, with the proportion of old people increasing substantially. The main drivers of this process are low and declining fertility and increasing life expectancy. The total fertility rate (the number of children that the average woman bears during her lifetime) fell below the replacement level of 2.1 by 1994 and it has continued to fall, reaching 1.7-1.9 in the 2000s. Increasing life expectancy is another driver of Sri Lanka's population aging. Currently life expectancy for men is 69 years and for women is 77 years, higher than in most developing countries, and present trends indicate that life expectancy will reach the current average OECD level of 77.8 years by 2050 (OECD, 2005). Central to Sri Lanka's demographic aging will therefore be an increase in the share of the old people, Figure 2: Trends in dependency ratios, Sri Lanka with the share of peopled aged 2001-2101 60 years and above projected to 100% 93% 90% Child dependency increase from current 11 80% Old age dependency Total dependency 76% percent to 16 percent in 2020 iont 70% 60% 63% and to 29 percent in 2050 popula 54% 50% of 50% (Figure 1). According to De ge tane 40% 37% Silva (2007), the share of very rc 30% 29% Pe 26% 20% old people (80 years and more) 16% 10% will also strongly increase. 0% 2001 2011 2021 2031 2041 2051 2061 2071 2081 2091 2101 Because in the next 2-3 Source: De Silva (2007) decades the increasing old age dependency ratio (the proportion of population aged 60 years or more versus the proportion aged 15-59 years) will be roughly matched by the fall in the child dependency ratio (proportion of population aged less than 15 years versus the proportion aged 15-59 years), the overall dependency ratio is projected to start increasing rapidly after 2035 (Figure 2). Population aging will translate into a slowdown of labor force growth and later into its contraction, and its composition will also be affected. Labor force projections show that Sri Lankan labor force will continue to grow for another two decades, and it will shrink 5 thereafter in a foreseeable future.2 Applying constant current labor force participation rates to population projections, Figure 3 shows that labor force will stop growing around 2030 (in 2027, 2032, and 2037 for the low, standard, and high population scenario, respectively), and will thereafter start to shrink, dropping to the current size of the labor force in about 30 years under the standard population scenario.3 Population aging will also significantly change the age composition of the labor force, with the share of workers younger than 30 years significantly shrinking and the share of those older than 50 years strongly increasing (Figure 4). Some caveats need Figure 3: Labor force projections, 2001-2070 to be made regarding the 12000 above labor force 10000 projections. First, will LFP 8000 standard rates remain constant as 6000 low assumed by projections? 4000 high Over the 1992 to 2004 2000 0 period, the age and gender specific LFP trends in Sri 2001 2007 2013 2019 2025 2031 2037 2043 2049 2055 2061 2067 Lanka have indeed been Source: Calculated using population projections by De Silva (2007) and largely constant (see labor force survey data by the Census and Statistics. below). However, comparisons with developed countries in Asia (Japan, Malaysia and Thailand) show that Sri Lanka's LFP rates are lower for all age groups for females of as well as for 50 to 65 year old males (and similar for other groups). If these groups gradually catch up with developed countries and increase their participation rates, the shrinking of labor force would be delayed by a few years (see below). Second, will changes in temporary migrations have an effect o the labor force? According to Sri Lanka Bureau of Foreign Employment, there have been about 200,000 registered yearly departures for temporary work outside Sri 2For most developed countries, projections show that their labor forces will stop growing in next decade unless participation rates rise (OECD 2006, World Bank 2007). 3These calculations assume constant future labor force participation rates and apply age and gender-specific labor force participation rates (averages for 2000-04 period) to population projections. 6 Lanka in recent years.4 As labor force participation rates are calculated only for individuals working within the country, these individuals are not reflected in the labor force statistics of the Department of Census and Statistics. If temporary migration trends continue to increase at the current rate, the labor force would start to shrink a few years earlier. However, given that at present the number of temporary migrants is about 3 percent of the labor force each year, changes to temporary migration trends would not affect labor force projections drastically. And third, will permanent migration trends in the country affect labor force calculations? Although there is little information about the number of permanent migrants from Sri Lanka, these numbers in all likelihood are much smaller than those on temporary migrations from Sri Lanka, and as such they are likely to affect labor force projections only marginally. With labor force growth being one of the sources of economic growth, the slowdown of labor force growth and its shrinking after 2030 may contribute to a slowdown in GDP growth. While simulations that would determine the aging-induced "tax" on GDP growth is beyond the scope of this paper, it is instructive to present some estimates of this tax for other countries. Because of population aging, the forecast of GDP per capita growth in the OECD area is reduced to around 1.7 percent per year over the next three decades, bringing about 30 percent reduction in comparison to the 1970-2000 growth rate (OECD 2006). Martins et al (2005) estimate that in the next half century, the decline of their workforces will contribute to a reduction of the per annum growth rates by 0.2 to 0.5 percentage points in France and Germany, and by 0.8 percentage points in Japan. Moreover, because of the raising share of inactive population, the economic growth in Japan, France and Germany will slow down by additional 0.2 to 0.3 percentage points, compared to growth under a stable age structure of the population. For comparison, about one quarter of the growth of OECD countries in recent decades has been attributed to labor force growth (Martins et al 2005). 4Since not all migrant workers register with the Bureau the actual number of temporary out migrants is somewhat larger. 7 Figure 4: Age composition of standard labor force projections, by gender (2001-2070) Males Females 7000 3500 6000 3000 5000 2500 4000 2000 3000 1500 2000 1000 1000 500 0 0 Year 2001 2006 2011 2016 2021 2026 2031 2036 2041 2046 2051 2056 2061 2066 Year 15-29 30-39 40-49 50-59 60-69 70+ 15-29 30-39 40-49 50-59 60-69 70+ Source: Calculated using population projections by de Silva (2007) and labor force survey data by the Census and Statistics. Population aging, however, may not necessarily reduce economic growth, if countervailing factors prevail. A drag on economic growth created by population aging could be countered by higher overall Figure 5: Education structure of the labor force, (multifactor) productivity, 1990 and 2004 increases of labor quality, or increased labor force 50 45 participation. While it is hard to 40 35 30 predict the changes of 1990 25 20 2004 multifactor productivity, recent 15 10 5 trends show that education level 0 Below grade 5 Grade 5-9 O level A level and of workforce has improved above significantly in the recent 15 Source: Department of Census and Statistics (2004b). years (as shown in Figure 5, the share of workers with finished A level or above has increased by 5 percentage points, and with commensurate reduction of workers with less than 5 years of schooling), and the quality of workforce is likely to improve further in the future. Another plausible countervailing factor is the increase of labor force participation rates ­ for example, simulations show that 8 the increase of women's participation rate by 20 percent could delay the reduction of labor force by 15 years.5 3. WORKING IN OLD AGE Below we describe key labor market outcomes for old workers, focusing on the timing of retirement, work load, manner of retirement (how quickly people transit from full- time employment to complete retirement and the use of part-time work before reaching complete retirement), changes in the sector of work and occupation, and reasons for retirement. (a) Labor force participation and unemployment of old people During 1992 - 2004, labor force participation of old workers remained stable, men's participation rate by far exceeded women's, and the withdrawal from a labor market occurred rather late in the life cycle (Figure 6). The only exception in regard with stable trends was a rising LFP rate of females aged 50-59, increasing from 26 percent in 1992 to 33 in 2004; in contrast, LFP for males in that age show no pronounced trend and remained in the 51-56 percent range (Figure 7). Workers were withdrawing from the labor market rather late in the working career ­ starting with the age group 60-69, for males, and with the 50-59, for females (Table 1). For males, LFP rates for all age groups in the prime age (20-59) exceeded 80 percent, dropping to the 50s in the age group 60-69 and to around 20 percent in the age group over 70. Withdrawal for the women starts somewhat earlier ­ in 2004, LFP rates in the age group 40-49 were about 44 percent, dropping to 33 percent in the age group 50-59, to 14 percent in the age group 60-69, and to 3 percent in the age group over 70. 5Simulations for developed countries show that the most powerful countervailing effects can be expected from the increased labor force participation of old workers. Martins et al (2005) show over the long run, the impact of increased retirement age in line with longevity gains prevails over impact of improved labor quality, and that this effect is particularly large for countries currently having low participation rates of older workers. For example, an increased labor force participation scenario in France would turn a decline in labor supply into an increase in labor supply over the next four decades. In Japan, however, the intensity of aging is such that even the combined effect of quality adjustments and increases in participation rates are sufficient to prevent an absolute decline in labor supply by 2050. Similarly, projections for Eastern and Central European countries show that an increased labor force participation of workers aged 40 to 59 years as well those over 60 years old significantly slows down the shrinking of labor force (Muenz 2007). 9 Table 1: Labor force participation by age group from 1992-2004 (%) (a) Male Year 15-19 20-29 30-39 40-49 50-59 60-69 over 70 Total 1992 32.1 89.3 96.7 95.0 80.1 54.9 21.4 75.2 1993 33.8 88.7 96.9 94.6 78.5 56.0 25.6 75.6 1994 34.0 87.9 96.0 95.8 82.7 53.6 23.1 76.4 1995 31.2 87.7 96.3 94.5 82.0 51.4 22.5 74.8 1996 31.0 88.6 95.8 94.7 82.5 52.5 23.8 75.4 1997 31.0 88.6 95.8 94.7 82.5 52.5 23.8 75.4 1998 35.4 89.6 97.0 95.5 84.2 56.3 23.1 77.0 1999 36.7 89.1 96.7 95.0 84.9 53.5 25.7 76.7 2000 32.9 87.7 96.6 95.6 84.2 54.3 22.1 76.1 2001 28.8 88.1 96.0 95.6 84.8 52.6 21.2 75.2 2002 30.1 88.5 96.9 95.3 85.4 54.8 23.6 76.7 2003 29.3 87.9 96.4 94.8 83.6 53.4 22.4 76.3 2004 29.5 88.1 96.3 94.7 83.9 51.9 18.2 76.0 (b) Females Year 15-19 20-29 30-39 40-49 50-59 60-69 over 70 Total 1992 21.1 52.5 45.6 40.8 25.8 11.5 3.6 36.2 1993 22.8 54.2 48.7 45.9 27.1 11.4 3.4 38.2 1994 21.2 53.2 47.4 41.2 27.5 11.6 3.4 36.7 1995 19.8 51.7 46.0 44.3 28.2 14.0 3.8 36.3 1996 20.1 49.9 45.4 43.6 29.6 14.5 3.2 36.1 1997 20.1 49.9 45.4 43.6 29.6 14.5 3.2 36.1 1998 23.0 54.8 51.3 50.9 34.8 17.8 5.1 40.9 1999 21.3 50.6 48.7 48.0 34.5 15.4 3.4 38.2 2000 18.2 50.1 50.4 47.3 34.8 15.6 4.2 37.9 2001 17.2 49.0 44.6 45.1 34.0 15.2 3.3 35.7 2002 19.0 50.0 45.4 47.2 35.6 13.5 5.0 37.4 2003 18.6 44.5 44.0 44.1 34.3 13.7 5.0 35.1 2004 18.1 45.4 43.7 44.4 33.1 14.0 3.4 35.0 Figure 6: Labour force participation, by Figure 7: Labor force participation of old age and gender, 1992 and 2004 (in percent) workers, 1992-2004 (percent) 120 60.0 100 80 40.0 60 20.0 40 20 0.0 0 1992 1994 1996 1998 2000 2002 2004 15-19 20-29 30-39 40-49 50-59 60-69 70 + Men 1992 Men 2004 Males 60-69 Males 70+ Women 1992 Women 2004 Females 60-69 Females 70+ Source: Own calculations based on various Labor Force Surveys of Sri Lanka. 10 Unemployment among old Sri Lankans is a rather uncommon phenomenon. During 1992-2004, unemployment rates for males aged 60-69 were typically well below 1 percent and virtually 0 for those over 70 years; unemployment rates for old women were even lower (Table 2). These figures reflect both the fact that when losing their jobs, old workers tend to withdraw altogether from the labor market (the phenomenon called "hidden unemployment" by OECD 2006). Unemployment rates are rapidly falling with age for both men and women (Figure 8). In 2004, 15-19 year olds had the highest unemployment rates, amounting to 26 and 37 percent, respectively, for men and women, with unemployment rates for older age groups monotonically decreasing and reaching single digit numbers already in the age group of 30-39 year olds. More rigorous analysis confirms strong association of labor force participation and unemployment with age. The above associations were obtained from cross-sectional data and not from observing the same set of individuals through time, and changes in the composition of persons observed in different age categories may influence the results. Accounting for such differences, probit regressions, run separately for males and females, confirm that age has a strong influence on the labor force participation and unemployment of individuals.6 Relative to labor force participation Figure 8: Unemployment rates by age and of individuals aged 20-24, the gender (1992 and 2004) marginal effect of age is very strong 50 in the late 20s, remaining flat till 45 40 about the late 40s, and becoming 35 30 increasingly negative for both males 25 20 and females (not shown). Similar 15 10 probit regressions also confirm the 5 0 association of age with 15-19 20-29Men 1992 40-49 30-39 50-59 Men2004 60-69 over 70 Women 1992 Women 2004 unemployment rates, with age accounting for a much larger Source: Own calculations based on various Labor Force Surveys of Sri Lanka. reduction of unemployment rate between early 20s and late 30s for women than for men. 6The explanatory variables used in the regression include age, marital status, education, province, and number of dependent children and old people in the household. 11 Table 2: Unemployment rate by age group from 1992-2004 (%) (a) Male Year 15-19 20-29 30-39 40-49 50-59 60-69 over 70 Total 1992 38.5 20.4 4.9 2.2 1.8 1.1 0.8 10.7 1993 34.5 18.2 4.9 2.3 2.3 0.2 0.2 9.6 1994 35.5 18.4 5.0 2.6 1.9 0.6 0.2 9.6 1995 31.7 18.4 4.7 2.8 0.6 0.1 0.0 9.0 1996 30.1 16.9 3.5 1.0 0.7 0.7 0.0 7.6 1997 30.1 16.9 3.5 1.0 0.7 0.7 0.0 7.6 1998 25.5 13.7 2.5 1.3 0.9 0.9 0.0 6.5 1999 25.5 14.3 2.5 1.5 1.6 0.5 0.4 6.8 2000 22.1 13.5 2.2 0.8 0.8 0.2 0.0 5.8 2001 28.9 14.3 1.8 1.3 0.7 0.1 0.0 6.2 2002 26.9 15.7 2.2 1.4 1.0 0.3 0.0 6.6 2003 27.1 14.3 2.0 1.2 0.8 0.4 0.0 6.0 2004 26.3 14.3 2.3 1.4 0.8 1.1 0.6 6.2 (b) Females Year 15-19 20-29 30-39 40-49 50-59 60-69 over 70 Total 1992 46.4 39.4 12.9 4.2 0.8 1.4 0.0 22.4 1993 44.9 36.6 15.1 5.3 1.9 0.0 2.1 21.7 1994 49.6 34.6 12.5 2.6 0.2 1.3 1.0 20.1 1995 48.4 33.6 12.7 2.9 0.3 0.0 0.0 18.9 1996 39.6 32.0 8.3 2.8 0.7 0.4 0.0 16.2 1997 39.6 32.0 8.3 2.8 0.7 0.4 0.0 16.2 1998 30.6 28.2 9.1 3.3 0.6 0.0 0.0 14.2 1999 34.1 27.0 8.0 1.7 0.9 0.2 2.2 13.2 2000 26.3 24.3 6.3 2.3 1.6 0.0 0.0 11.1 2001 31.4 25.9 6.5 1.6 0.4 0.0 1.5 11.5 2002 35.4 27.2 7.2 1.7 1.1 1.2 1.9 13.0 2003 35.6 29.1 7.5 2.4 0.6 0.4 0.0 13.2 2004 37.4 29.4 6.3 2.4 0.7 0.8 0.0 13.1 International comparisons show that Sri Lanka's labor force participation rates for old people lag behind the rates of its regional comparators while mostly exceeding those in developed countries. Labor force participation of older men is lower than in other countries in the region (India, Thailand, Bangladesh and Pakistan, Table 3), and so is participation of women, except that women's participation rate in Sri Lanka exceeds the one in predominantly Muslim countries (Pakistan and Bangladesh). Interestingly, the LFPs of elderly in Sri Lanka also lag behind the rates of Japan, but they exceed other developed countries. 12 Unemployment rates of elderly Sri Lankans are in line with those of regional comparators and lag substantially behind developed countries. Unemployment rates for elderly ­ both males and females ­ in Sri Lanka are similar to those in India and Thailand, where open unemployment among old workers is also virtually unknown. Reflecting a different institutional set-up ­ above all, the existence of specialized programs providing cash benefits to unemployed that stimulate labor force attachment ­ unemployment rates of old workers in developed countries are significantly higher than in developing countries (although they are still much lower than youth unemployment rates in these countries). In 2004, unemployment rates for workers aged 50-64 ranged from 2 percent in Norway to 11 percent in Germany (OECD 2006). (b) Employment status, work load and pay of old workers Most of the old workers have full-time rather than part-time jobs. As shown above, as they age, fewer and fewer workers remain employed. But if they do, they tend to work in full-time jobs rather than part time. In our survey of old people, 30 percent of males were employed, of whom two thirds were in full time employment and only a third in part-time 13 Table 3: International comparisons of labor force participation rates for elderly Age group Sri India Bangla- Paki- Thai Philip- Japan Hong US France Ger- Nor- Switzer-la Lanka desh stan -land pines Kong many way Year of data 2004 2001 2003 2004 2004 2004 2004 2004 2004 2004 2004 2004 2004 Males 50¬54 89.2 92 99.2 94 91.6 95.7 89.8 85.4 90.8 90.2 88.2 93.7 55¬59 76.4 97.3 88.2 88.6 93.2 75.7 77.6 67.7 80.3 81.7 89.4 60¬64 57.7 69.7 87.8 73.1 52 70.7 43.6 57 19 37.7 64.3 66 65¬69 43.0 66.1 47.5 54 45.6 9.6 32.6 4.1 7.2 26.5 22.8 70-74 25.0 49.3 29.3 19.4 1.4 3.7 5.6 12.9 Over 75 16.8 14.6 9 0.4 1.5 6.1 Over 80 7.4 34.6 Females 50¬54 37.9 40.9 19.9 18.6 70.3 68.4 52.1 74.5 77 76.4 81 78.7 55¬59 25.7 17.1 14.5 57.3 59.6 34.8 65.0 56.2 61.7 70.9 70.1 60¬64 16.9 26.3 13.4 13.0 28.1 39.7 12.6 45.4 16.2 19.7 52.7 38.9 65¬69 10.1 8.7 10.2 30.2 24.0 1.9 23.3 2.5 4.1 19.1 10.1 70-74 5.6 13.3 14.9 12.0 0.7 1.5 3.6 5.4 75-80 2.1 5.8 4.3 0.2 0.5 2 Over 80 1.3 8.6 Note: Merged cells contain data for the combined age groups. For comparison purposes, labour force participation rates for 65 plus and 75 plus age group Sri Lanka are, males - 27.53 and 12.02 and females 5.90 and 1.72, respectively Source: ILO, LABORSTA Database 14 employment; only 5 percent of women were employed, with 60 percent of them in full-time employment. The proportion of part-time employment among old workers was the lowest in the 60-64 age group and then increased for older groups of workers, with the share of part time workers exceeding the share of full-time workers in the 75-79 age group (see Figure 9).7 Figure 9: Retirement status by age and gender (percent) (a) Men (b) Women 100 100 80 80 retired retired 60 60 work part-time % work part- % 40 time 40 work full-time work full- 20 20 time not relevant not relevant - 0 60-64 65-69 70-74 75-80 80+ 60-64 65- 69 70-74 75-80 80+ Age group Age group Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey The most prevalent way of withdrawing from the labor market is "overnight," that is, without engaging in part-time employment.As implied by falling LFPs and the fact that part- time work is relatively infrequent, most workers exit from full-time employment directly to complete retirement. In our sample of individuals who held stable jobs in their early fifties (see selection details below), a vast majority (89 percent) of workers who did not work full time any longer had retired "overnight," that is, they did not work part time before retirement and they completed transition to inactivity in a very short time. The number of hours worked is also reduced by age. Reflecting retirement patterns, old workers as a group are providing less and less labor to the economy. The average weekly number of hours of work provided by old male workers decreased from 23 for 60-63 year olds to below 10 for the 72-75 year olds; and by old female workers, from 5 and 2 for the same are groups (Figure 10a). At the same time, workers staying active reduce the number of 7The question on retirement status was not relevant for 66 per cent of elderly females. This arises from the fact that many females do not participate in market based economic activities in younger ages. According to DCS, Labor Force Survey data 60 percent of females 60 to 64 years old are unavailable work, because they are engaged in household activities. Although, this percent decreases for elderly age-cohorts even at 80 plus years of age, 6% of elderly females state that they are not available for work due to household work. 1 hours worked as they age, men from the average of 47 in their early sixties to 36 in the 72-75 year group, and women from 35 to 29 for the same age groups (Figure 10b). Despite these reductions, the average numbers of work for those workers who remain employed remain large even in high age. Old workers in part time jobs also work long hours ­ in our sample (SLAS 2006), they on average work only 12 hours less than old workers in full-time jobs (36 hours compared to 48). Figure 10: Average number of hours worked, by age and gender, (a) including fully retired, (b) employed workers only (full- and part-time workers) 25 50 keew 20 ekew 40 15 Men 30 pers Women M en 10 20 Women uroH 5 persruoh 10 0 of# 0 60-63 64-67 68-71 72-75 60-63 64-67 68-71 72-75 Age Age IFO aunt am jumpeda state sue JE am in new song you saw Lucia new share and an claimed a new show used a new star jean inchan clear Schiro ushijima jiao you so you show you should it as a theory as F. new show on our show Ju so JAMA where she use is a is in new star jean F. K. are she didn't do so you share show you starship.new song you so you show used to do so you should do so new channel you saw jumped in the new show where she wrapped JAMA Pay of old workers In comparison to younger Figure 11: Age wage profile, by sector and workers, old worker are paid less, gender particularly in the public sector. Wages Index (wages for 25-29 year olds = 100) of workers of 65 years and above are 170 150 only a fraction of wages of the best 130 index 110 paid group of workers ­ workers in e 90 their late 50s and early 60s, in the Wag 70 50 public sector, and in their 30s, in the 30 25- 30- 35- 40- 45- 50- 55- 60- 65 + private sector (Figure 11). Particularly 29 34 39 44 49 54 59 64 Age group strong reduction of wages of older Public sector male Public sector female Private sector male Private sector female workers occurs in the public sector, Source: Own calculations based on various Labor Force Surveys of Sri Lanka. 2 with workers aged 65 and over earning barely over 1/3 of what the workers in the 60-65 group are earning, and only about 20 percent of what workers in the 55-59 year group are earning (the latter applies for women only). In the private sector, the reduction of wages for workers above 65 years of age is more modest. Interestingly, except for women in the public sector, wages of workers in the 60-64 year group are higher than wages of the 55-59 year group.8 One salient feature of the age-wage profiles is the fact that the both men's and women's profiles of the private sector are much flatter than the ones in the public sector. Many OECD countries have hump-shaped age wage profiles, but there are many exceptions, reflecting different institutional wage-setting and other arrangements (see OECD 2006). (c) Where and as what do elderly work? A vast majority of old workers are self-employed or casual workers engaged in the informal sector. More than half of the working elderly males and females are self- employed (Figure 12a). Majority of the rest of workers are casual workers, with less than 3 percent of elderly working as regular workers, mostly workers in their early sixties. Vast majority of old workers are engaged in informal private sector (the main exception being men in their early sixties, with small share of them working for the government or in the formal private sector) ­ see Figure 12b. 8Note that these profiles are derived from cross-section data and may not necessarily reflect life-cycle profiles (that is, age-wage profiles of the same workers traced through time), because these profiles may be heavily influenced by the selection of workers: other things equal, high-paid workers are more likely to withdraw from the labor market later in their career than low-paid workers. The increase of relative wages for workers in their early 60s compared with workers in their late 50s in the private sector thus reflects two opposing effects: the selection effect (that is, a higher likelihood that better paid workers stay employed) as well as the negative effect of age on wages of life-cycle wage profile, with the former prevailing (that is, better-paid workers self- select and continue working while lower-paid workers transfer to inactivity). 3 Figure 12: Old workers by type of employment status and type of employer (a) By type of employment status (b) By type of employer 70 120 60 100 50 80 40 % 60 30 % 20 40 10 20 0 0 60-64 70-74 60-64 70-74 60-64 70-74 60-64 70-74 Male Female Male Female Regular Casual Self employed Government Formal private Informal private Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. Old workers, both males and females, are working mostly as skilled workers in agriculture, manufacturing, and wholesale and retail trade (Figure 13). A fair proportion of them, particularly of women, are working also at the bottom of occupational ladder (in elementary occupations), and a small proportion as professional workers. Not surprisingly, the composition of workers by occupation changes little with age, as does sector of work, except that the share of workers in agriculture is reduced. While sector of work and status for old workers working part-time does not differ much from full-time workers, a larger share of them are professional and skilled workers. Figure 13: Old workers by sector of work and occupation (a) Old workers by sector of work (b) Old workers by occupation 70 45 40 60 35 50 30 40 25 % % 20 30 15 20 10 10 5 0 0 60-64 70-74 60-64 70-74 60-64 70-74 60-64 70-74 Male Feamle Male Female Agriculture Manufacturing Services Professionals Skilled workers Elementary occupations Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. 4 (d) Reasons for retirement and part-time work The main reasons for retirement are on the push side: ill-health and reaching mandatory retirement age, for both males and females (see Box 1 for a distinction between pull and push factors influencing work and retirement decisions). Nearly half of retired old workers stated that health reasons were very important for their decision to retire, and about 40 percent of them quoted reaching the mandatory retirement as being the second most important reason (Table 4). Among the push factors, work and travel stress, as well as the closure of businesses were also important factors affecting the decision to retire. Among the pull factors, "doing other things (hobbies)" was an important reason for retirement, especially for females (Table 4). Only females found family obligations a very important reason for complete retirement; however, it was a relevant factor only for a small proportion of females. This indicates that obligations of child-care and care for old persons do not have a large effect on retirement decisions. Table 4: Reasons for complete withdrawal from work of old workers Very Moderately Not important Don't important important at all know Total Male 48 17 34 0 100 Health reasons Female 47 23 30 0 100 To do other things Male 8 27 63 2 100 (hobbies) Female 27 31 40 2 100 Male 0 11 84 4 100 Family obligations Female 3 12 78 7 100 Male 5 8 86 2 100 Disliked work Female 9 12 74 4 100 Reached mandatory Male 37 18 43 2 100 retirement Female 40 16 39 5 100 Traveling to work Male 7 17 73 3 100 was stressful Female 12 23 60 4 100 Male 7 5 81 7 100 Business closed Female 11 3 74 11 100 Work was too Male 8 13 77 2 100 stressful Female 11 20 65 4 100 Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. Earning additional income was the main reason for part-time employment for most workers. Three in four elderly in part-time employment are continuing to work to earn 5 additional income. More females than males consider the additional income from part-time employment as the main reason for continued employment. A small percent also worked part time to keep them selves occupied (6 percent) and to be in touch with their profession (3 percent). A large share of elderly working part-time does so involuntarily, as 80 percent of those who retired after doing some part-time work indicated that they would have liked to stop work completely after retirement from full-time work. (On the other hand, over a quarter of the surveyed old people indicated that they would have liked to continue doing some paid work when they retired from their full time jobs.) Box 1: Push and pull factors of labor market withdrawal Several "pull" or "push" factors influence work and retirement decisions of old workers. Pull factors are associated primarily with financial incentives facing the individual, and push factors with circumstances that restrict job opportunities available to an individual. Among the most common factors that pull workers into retirement are financial incentives provided by pension schemes and other formal or informal early retirement schemes. In this regard, the age at which pension funds are accessible, their generosity and how additional years of work change the present value of net pension wealth play key roles in influencing retirement decision of workers. In many developed countries other welfare schemes ­ such as unemployment benefits and disability benefits ­ are also often used by workers as a means of early exit from work. In addition, private pension arrangements and joint retirement decisions of couples may also pull workers into retirement in some countries, particularly for those who value leisure. In developing countries, family obligations ­ such as helping with childcare of grand children ­ also pull workers into retirement. "Push" factors include both firm and individual circumstances that restrict suitable job opportunities. At the firm level these include negative perceptions about the capacities of old workers, labor costs exceeding contribution to output by old workers, and difficulties firms face in adjusting employment as a result of employment protection rules. At the individual level these include skills mismatch in the face of technological and structural changes in labor demand, The reasons for retirement differ substantially from those in OECD countries. Similar to workers in OECD countries, a share of workers retires upon reaching mandatory retirement age. However, unlike in OECD countries, the single most important reason for withdrawal from the labor market is poor health. A smaller share of workers is also pulled into retirement by family obligations, work stress, travel stress and work dislike. Moreover, in OECD countries there are many institutional arrangements that entice old workers to retirement or early retirement, ranging from pre-retirement pension schemes to disability and 6 unemployment insurance benefits (OECD 2006) ­ the arrangements largely missing in Sri Lanka. 4. PATHWAYS TO RETIREMENT Above we analyzed the patterns of withdrawal from working life as people age, treating the population of old people as a whole. Remembering a key motivation for this investigation ­ identifying ways to promote choice for old workers in both improving job prospects as well as reducing pressures on those who are forced to work ­ it is helpful to find out whether the working careers themselves determine the fate of workers in their old age. Recognizing the deep segmentation of the Sri Lankan labor market, we do so by tracing retirement paths ­ typical sequences of labor market arrangements between full-time work late in the working career and a complete retirement/withdrawal from active working life ­ for different groups along these segmentation lines. We therefore focus on the following four groups of workers:9 Regular public sector workers (i.e., regular workers of public sector, public corporations or boards and the co-operative sector), Regular private sector workers (i.e., regular workers of both the formal and the informal private sector), Casual workers (i.e., casual and contractual workers), and Self-employed (i.e., employers and the self-employed). In our analysis of retirement pathways and exploiting the data from the SLAS 2006, we focus on the following aspects (the most relevant ones among the aspects studied in the previous subsection): timing of retirement, manner of retirement ("overnight" vs. via part- time employment), changes of sectors of work and occupation, and reasons for retirement. Note that for this analysis we exploited a panel nature of the SLAS 2006 ­ the fact that in addition to being asked asking about their current employment status, respondents were retrospectively asked also about the age of retirement from their full- and part-time job, as well as their employment status when whey were in their early fifties, that is, at the end of their prime-time work career. Based on this information, we selected 1060 workers with 9Unpaid family workers were not included in the analysis, due to small (20 observations) sample size. 7 stable jobs in their early fifties (they held full-time jobs and were continuously working when they were 51-54 years old), and traced their work history continuously into their old age.10 Our results confirm that pathways to retirement show strong dual patterns. The majority of regular workers in public and private sector retire early, most of them before they reach 60, and for work-related reasons (including mandatory retirement); in contrast, a large share of self-employed and casual workers continues to work full-time into very old ages, and most of them withdraw from employment for health reasons. The option of part-time work is also used primarily by workers who held regular jobs in their prime age employment, but not by casual workers and self-employed. The results show that labor market duality, the hallmark of labor market in many developing countries, carries over to old age and importantly determines the fate of old people. We present detailed results below. (a) Timing of retirement: regular public and private sector workers withdraw from the labor market much earlier than casual and self-employed workers. More than two thirds of workers holding regular jobs in the public sector in their prime age, and 57 percent of workers holding jobs in the private sector, completely retired when they were 60 years old. In contrast, only 23 and 18 percent of workers who were casual workers or self-employed in their prime age retired when they reached the age of 60, with the gap in retirement status widening with the age of workers (Figure 14)! When they reach 69, virtually none of the workers who have spent their careers in the formal sector is still working full time; in contrast, nearly half (47 percent) of their counterparts who were casual workers and self- employed are still working full time.11 Note that in terms of occupations, the first two categories comprise mostly of white-collar workers (officials, managers, clerks and professional), in contrast to predominantly blue-collar workers comprising the casual and self-employed (who were mostly skilled agriculture and fishery workers, craft and related workers, and workers in elementary occupations). The observed pattern of retirement is thus contrary to the one found for OECD countries, where blue-collar workers and less-skilled workers are more likely to retire earlier (see OECD, 2006). 10According to the pre-retirement employment status, the total sample of 1060 individuals is divided into 276 regular public sector workers, 138 regular private sector workers, 223 casual workers, and 424 self-employed (the data are right censored). 11A log rank test showed that the survival curves for the four sectors of employment ­ that is, regular public sector workers, regular private sector workers, casual workers and self employed ­ were statistically different. 8 Figure 14: Retirement status, by prime-age employment type and age (a) Regular public sector w orkers (b) Regular private sector workers 100% 100% 80% 80% 60% 60% 40% 40% 20% 20% 0% 0% 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 Retired Work part_time Work full_time Retired Work part_time Work full_time (c) Casual workers (d) Self-employed 100% 100% 80% 80% 60% 60% 40% 40% 20% 20% 0% 0% 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 Retired Work part_time Work full_time Retired Work part_time Work full_time Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. Note: The employment status refers to employment status at age 54. All workers were working full time between ages 51 to 54. Work related reasons: work stress, travel stress, mandatory retirement, completion of contract, business closed; Health reasons: health reasons or illness, not feeling physically well. Personal reasons: retirement incentives and family obligations. 9 Another way of looking at differences at retirement patterns is to observe retirement ages. Figure 15 shows retirement from full-time employment for the selected sample of individuals who held stable jobs in their early fifties described above. The distribution of retirement age is similar for regular Figure 15: Percentage of retirements by age public and private sector workers, with groups the 95 percent of workers from the 70 60 Publicsector first group and 84 percent from the 50 40 Private sector % 30 second group retiring between ages 55 20 Casual workers 10 and 64 (for public sector workers, Self_employed 0 55-59 60-64 65-69 70-75 76-80 80+ there are clear peaks at ages 55 and 60, Age group due to mandatory retirement ­ not Source: Own calculations based on World Bank shown on the graph). In contrast, the 2006 Sri Lanka Aging Survey. distribution of retirement age for Note: Retirements from full-employment to either complete retirement or part-time work. informal sector workers spreads in much higher age, with 16 and 18 percent of casual and self-employed workers, respectively, retiring when aged 76-80. (b) Manner of retirement: the majority of workers across all groups retire overnight, with labor market withdrawal via part-time employment being more common for regular workers. In our sample of individuals who held stable jobs in their early fifties, 11 to 13 percent of regular public or private sector workers held part-time jobs by the time of the survey, and only 6 percent of casual and self-employed workers did. (Note that at the time of the survey, around 40 percent of casual and self-employed still held full time jobs and very few among the regular workers.) (c) Industry and occupational composition: changes occur primarily for regular but not for casual and self-employed workers. Proportions of elderly in different industrial sectors remain constant for casual and self-employed workers (see Figure 16). In contrast, among workers who held regular jobs at the end of their prime-age working career, a proportion in agriculture increases with age (after age of 67, workers with jobs in public sector tend to withdraw from the agriculture and their share in industrial sector increases). Similarly, occupational composition for casual and self-employed workers shows little 10 change with time, while among workers who held regular jobs at the end of their prime-age working career, the share of skilled workers increases and the share of professional workers decreases with age (not shown). Figure 16: Sector of work, by the prime-age employment type and age (a) Regular public sector workers (b) Regural private sector workers 100% 100% 80% 80% 60% 60% 40% 40% ` 20% 20% 0% 0% 55 57 59 61 63 65 67 69 71 73 75 55 57 59 61 63 65 67 69 71 73 75 age age Agriculture Industry Services Agriculture Industry Services (c) Casual workers (d) Self-employed 100% 100% 80% 80% 60% 60% 40% 40% 20% 20% 0% 0% 55 57 59 61 63 65 67 69 71 73 75 55 57 59 61 63 65 67 69 71 73 75 age age Agriculture Industry Services Agriculture Industry Services Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. (d) Reasons for retirement: most workers are pushed away from jobs, but while for regular public and private sector workers the main reason is mandatory retirement age, for casual and self-employed workers is ill-health. For regular public sector workers, the main reason for retirement at any age is work related, primarily reaching mandatory retirement age (see Figure 17). The same holds true for regular private sector workers, except for those aged 61-70. Overall, a slightly over a quarter of regular workers in the private sector cited reaching mandatory retirement age as a reason for retirement from full time work. On the other hand, by far the most compelling reason for retirement at all ages for casual workers and self employed was ill health (a small exception are casual workers aged 55-59, where 11 work-related reasons slightly overweigh health reasons). Interestingly, personal reasons figured more prominently for the groups of casual workers and self employed than for regular workers. Figure 17: Reasons for retirement from full time work (a) Regular public sector workers (b) Regular private sector workers 120 120 100 100 80 health 80 health % 60 w rk_related % 60 w rk_related 40 personal 40 personal 20 20 0 0 55 56-60 61-65 66-70 71-75 55 56-60 61-65 66-70 71-75 Age of retirement age of Retirement (c ) Casual workers (d) Self-employed 100 100 90 90 80 80 70 70 60 health 60 health % 50 w rk_related % 50 w rk_related 40 personal 40 personal 30 30 20 20 10 10 0 0 55 56-60 61-65 66-70 71-75 55 56-60 61-65 66-70 71-75 age of Retirement age of Retirement Source: Own calculations based on World Bank 2006 Sri Lanka Aging Survey. 5. DETERMINANTS OF WORK ACTIVITY OF THE ELDERLY The above analysis is mainly descriptive, examining the retirement from active working life through a "single lens," that is, one aspect at the time (section 3), and contrasting retirement patterns across workers with different employment status during their working life (section 4). In this section, we extend the approach to multivariate analysis, that is, to the identification of the separate influence of individual variables on retirement decisions, the influence that may be clouded in a simple, univariate analysis. Speaking formally, below we report the results of multinomial logit models identifying factors that contribute to the withdrawal from the labor market. The dependent variable is the change of the labor market status of individuals holding stable, full-time jobs at age 54 years old as compared to their status when they were 61-65 year old (1060 persons 12 in the SLAS). According to work status during ages 61 to 65, we allow three possibilities for this change (that is, our dependent variable can take three values): Completely retiring upon reaching 61 years and not working when aged 61-65 (in our sample, 382 individuals); this transition was taken as a baseline. Working at least some time during ages 61-65, either part time, or full-time if not throughout the period (in our sample, 208 observations). Working continuously full-time during ages 61-65 (in our sample, 470 observations). Among explanatory variables, labor market factors, as well as family, demographic and personal characteristics are included (see Table 5 for descriptive statistics of independent variables). Of the sources of income considered in the study, inclusion of financial or in-kind family assistance could be endogenous, if one believes that the amount of help given by family vary according to other sources of income, and health status of elderly. Hence, instead of using family assistance, we use a proxy variable, which indicates whether individuals have access to help if they need it. Statistically significant results of this analysis are highlighted below. Employment status at age 55 is the most important factor determining work activity of old workers. The results show that keeping other things constant, compared to regular public sector workers, all other types of workers are more likely to be working full-time during ages 61­65 (Table 6). These effects are large for workers of all three groups ­ regular private sector workers, and particularly for casual and self-employed workers: workers from the first group are 40 percent more likely to work in a full-time job in their early 60s than workers who left regular public sector jobs, and casual and self-employed workers 66 and 69 percent more likely, respectively. Interestingly, while casual and self-employed workers are more likely to work full time, but they are less likely to work part time as compared to workers who were regular public sector workers at the end of their working career. In other words, while the option of working part time is open to previous regular public sector workers, this option seems to be less available to casual and self-employed workers. 13 Table 5: Descriptive statistics of variables used in multinomoial logit estimation Completely Retired Working part Full time All time working 382 208 470 1060 Mean Std. Mean Std. Mean Std. Mean Std. Dev Dev Dev Dev Male 0.70 0.03 0.72 0.04 0.75 0.02 0.72 0.02 Individual Sinhala 0.83 0.03 0.86 0.03 0.86 0.02 0.85 0.01 characteristics Head of Household 0.72 0.03 0.71 0.04 0.76 0.02 0.73 0.02 Married 0.75 0.03 0.68 0.04 0.74 0.02 0.73 0.02 Previously married 0.23 0.03 0.28 0.04 0.24 0.02 0.24 0.02 Marital status Never married 0.02 0.01 0.04 0.02 0.02 0.01 0.02 0.01 No schooling 0.12 0.02 0.14 0.03 0.11 0.02 0.12 0.01 Less than primary 0.23 0.03 0.36 0.04 0.39 0.03 0.33 0.02 Lower secondary 0.26 0.03 0.22 0.03 0.37 0.03 0.30 0.02 Passed O/Ls 0.21 0.02 0.17 0.03 0.11 0.02 0.16 0.01 Education Passed A/Ls 0.17 0.02 0.10 0.02 0.02 0.01 0.09 0.01 Very good health 0.08 0.02 0.12 0.03 0.12 0.02 0.11 0.01 Good health 0.62 0.03 0.55 0.04 0.58 0.03 0.59 0.02 Health - 1 Poor health 0.30 0.03 0.33 0.04 0.30 0.03 0.31 0.02 Ever had chronic 0.68 0.03 0.56 0.03 0.62 0.02 Health - 2 illness 0.65 0.04 Receiving 0.38 0.03 0.04 0.01 0.19 0.01 pension 0.19 0.03 Family Transfers 0.78 0.03 0.79 0.02 0.79 0.02 Sources of support (proxy) 0.80 0.03 Adequate wealth 0.35 0.03 0.32 0.04 0.24 0.02 0.29 0.02 Wealth Own a house 0.56 0.03 0.55 0.04 0.65 0.03 0.60 0.02 Well 0.05 0.02 0.02 0.01 0.04 0.01 0.04 0.01 Preparation for Prepared 0.33 0.03 0.26 0.04 0.23 0.02 0.27 0.02 retirement Not 0.62 0.03 0.72 0.04 0.73 0.02 0.69 0.02 14 Table 5: Descriptive statistics of variables used in multinomoial logit estimation (Cont'd) Completely Retired Working part Full time All time working 382 208 470 1060 Mean Std. Mean Std. Mean Std. Mean Std. Dev Dev Dev Dev Public regular 0.51 0.03 0.33 0.04 0.03 0.01 0.26 0.02 Private regular 0.19 0.03 0.15 0.03 0.07 0.02 0.13 0.01 Casual/ contractual 0.10 0.02 0.22 0.03 0.30 0.03 0.21 0.01 Pre-retirement Employer/ 0.20 0.02 0.60 0.03 0.40 0.02 employment status Self employed 0.31 0.04 Location Western province 0.40 0.03 0.35 0.04 0.33 0.03 0.36 0.02 Urban 0.15 0.02 0.13 0.02 0.11 0.02 0.13 0.01 Rural 0.75 0.03 0.79 0.03 0.87 0.02 0.81 0.01 Sector Estate 0.10 0.02 0.08 0.02 0.02 0.01 0.06 0.01 Source: Own calculations using World Bank 2006 Sri Lanka Aging Survey. 15 Table 6: Multinomial logit estimates of transition from full-time employment1 Variable Description1 Marginal Effects Working part time Working full time Coef2 Robust Coef Robust Comp- Working Working std err std err letely part- full-time Retired time Male 0.48 * 0.29 0.43 0.30 -0.11 0.06 0.05 Individual Sinhala 0.47 0.45 -0.42 0.38 0.02 0.12 -0.14 characteristics Head of Household -0.09 0.27 -0.04 0.26 0.01 -0.01 0.00 Marital status Previously married 0.60 ** 0.29 0.41 0.30 -0.11 0.08 0.03 (currently married)3 Never married 0.97 0.70 0.90 0.74 -0.18 0.09 0.09 No schooling 0.15 0.52 0.02 0.64 -0.02 0.03 -0.01 Education Less than primary 0.26 0.44 0.27 0.61 -0.06 0.02 0.04 (A/levels or Lower secondary -0.30 0.45 0.16 0.59 0.01 -0.07 0.07 above) Passed O/Ls 0.01 0.42 0.13 0.61 -0.02 -0.01 0.03 Good health -0.64 * 0.35 -0.96 ** 0.35 0.18 -0.03 -0.15 Health - 1 (very Poor health ** good) -0.84 ** 0.41 -1.45 * 0.40 0.27 -0.04 -0.23 Health - 2 (no Ever had chronic chronic illness) illness -0.13 0.25 -0.53 ** 0.24 0.08 0.03 -0.11 Pension income ** ** -0.91 * 0.34 -1.37 * 0.38 0.28 -0.07 -0.21 Sources of Some one to help support financially 0.04 0.28 -0.02 0.28 0.00 0.01 -0.01 Adequate wealth 0.23 0.27 0.11 0.27 -0.04 0.04 0.00 Wealth Own a house -0.23 0.24 0.08 0.24 0.01 -0.05 0.04 16 Table 6: Multinomial logit estimates of transition from full-time employment1 (Cont'd) Variable Description1 Marginal Effects Working part time Working full time Coef2 Robust Coef Robust Comp-letely Working Working std err std err Retired part-time full-time Preparation for Some 0.59 0.69 -0.03 0.85 -0.06 0.13 -0.07 retirement Not (well) 0.67 0.68 -0.24 0.83 -0.03 0.15 -0.12 Private regular ** -0.04 0.37 1.69 * 0.49 -0.22 -0.18 0.40 Pre-retirement Casual/ ** employment contractual 0.83 ** 0.37 3.66 * 0.47 -0.43 -0.23 0.66 status (public Employer/Self ** regular) employed 0.55 * 0.33 3.72 * 0.45 -0.45 -0.24 0.69 Location Western province -0.09 0.26 0.11 0.25 -0.01 -0.03 0.03 Rural -0.10 0.29 0.26 0.37 -0.02 -0.05 0.07 Sector (urban) Estate -0.21 0.61 -1.32 ** 0.57 0.17 0.06 -0.23 No obs. 1060 Pseudo R2 0.2144 Notes: 1. Dependent variable takes three values : completely retiring upon reaching 61 years and not working when aged 61-65 (taken as a baseline); working part time (at least some time during ages 61-65, either part time, or full-time if not throughout the period, and working continuously full-time during ages 61-65. 2. Significance at 1, 5 and 10 per cent levels are indicated by `***','**' and * respectively. 3. Omitted category is given in parenthesis. Source: Own calculations using World Bank 2006 Sri Lanka Aging Survey 17 The probability of working in old age decreases progressively with (self-perceived) health problems and the presence of chronic illnesses. The results show that relative to individuals who reported very good health, the probability of full-time or part-time work of individuals who reported good or poor health is significantly smaller ­ for those in poor health, the chances for full-time work are 23 percent smaller, and for part time work 4 percent smaller, in comparison to those in very good health. Similarly, the probability of full- time (but not part-time) work of individuals who reported a chronic illness is also significantly smaller (by 11 percent) in comparison to individuals who did not report a chronic illness. Pension income is a major factor reducing labor supply of elderly. The results of the multinomial model show that pension income has a large and negative effect on work activity in old age. Receiving a pension reduces the probability of full-time work by 21 percent and that of part-time work by 7 percent. Location also affects elderly work activity, but the effects are small. Estate sector workers are less likely to work full-time during ages 61-65, relative to urban sector workers. This could possibly be due to the high level of strenuous physical activity associated with estate sector work. Interestingly, individual characteristics ­ gender, ethnicity, marital status, being a head of a household ­ have no significant effect of the studied labor market transitions. The only exceptions are that men are more likely then women to make a transition to part-time jobs, as do previously married workers as compared to currently married. 6. OBSTACLES FOR EMPLOYMENT OF OLD WORKERS The above results allow us to probe into the possible obstacles for employment of workers ­ labor market legislation, relative wages of old workers, and work stress and weak employability of old workers. (a) Inflexible labor market legislation A wide range of outcomes in the Sri Lankan labor market is suggestive of a presence of rigidities in working hours and type of employment faced by old workers. Above we showed 18 that old workers long relatively long working hours ­ according to the 2006 SLAS, only 12 percent of old workers in part-time employment and 8 percent in full time employment were able to adjust their working hours. Moreover, only 8 percent of retired workers had the option of reducing the number of working hours in their last full-time job, and about a quarter of the old workers who retired overnight would have preferred to do some paid work before completely retiring. Other findings of the analysis above complement this picture: we showed that the main reason for retirement of many regular workers was mandatory retirement, that only a small part of workers take part-time employment jobs, and that LFP rates of old workers in Sri Lanka are lower than in regional comparators. Can we conclude from that that inflexibility of employment legislation and barriers to reduce regular work hours or to part-time work? There may be other reasons for apparent rigidities in labor market outcomes of old workers, and it is not clear how much labor market legislation contributes to it. While lack of adjustment of working hours, low LFP rates, and the fact that people are retiring despite wishing to continue working in paid jobs are suggestive of inflexibilities, there may be other reasons for that. Employers may be reluctant to hire old workers on part-time basis because doing so provokes difficulties of coordinating work in teams; it may generate demands of other groups of workers to offer reduced working hours to them as well; and, possibly, runs against outdated attitudes of employers about part-time employment. Note that the obligation to pay gratuity does not work against old workers, because their relative pay is low (see the discussion of age wage profiles above). Sri Lanka's strict employment protection legislation, embodied in the 1971 TEWA system, however, may be a source of disincentives for employment of old workers (see Box 2 on the description of the TEWA). In principle, it is not possible to predict the impact of strict employment protection legislation (EPL) on labor market outcomes for old workers, as it may lead to both greater retention of old workers as well as to the reduction in their hiring. While the empirical evidence is weak, some recent studies report negative effects of strict EPL on employment of old workers. For example, a cross-section analysis of OECD countries shows a statistically significant association of strict EPL with reduction of employment and hiring rate of workers aged 50-64 (OECD 2006). There is also a number of 19 country studies which identified strict EPL as a barrier to the hiring or retention of older workers. Box 2: The generosity of Sri Lanka's severance pay system Sri Lanka's severance pay system--embodied in the Termination of Employment of Workman Act (TEWA) of 1971--is one of the costliest and most restrictive severance pay systems in the world. The TEWA requires employers with more than 14 workers to seek the authorization of the Commissioner of Labor for intended layoffs. It not only requires that employers pay high compensation to laid off workers, but its discretionary nature and lengthy procedures further restrict the ability of employers to lay off workers. By ex-ante determining the amount of severance pay, the recently introduced formulas somewhat reduced arbitrariness, but payments remain well above international norms. The March 2005 formula starts with a factor of 2.5 of monthly wages per year of service for workers with 1-4 years of service and total benefits increase with the years of service, but at a decreasing rate, with an overall maximum of 48 monthly wages for service of 34 years and above. Judged by international standards, the formula is very generous: a Sri Lankan worker with 20 years of experience is awarded by a severance pay of 39 monthly wages, in contrast to average severance pay of 16.3 monthly wages in other Asian countries, 11.9 in Latin American, 7.1 in African, 6.4 in OECD, and 4.4 in transition countries ­ a staggering difference! Source: World Bank (2007b). Mandatory retirement stipulations are provoked by strict employment protection legislation. While in Sri Lanka there is no legally binding retirement age (except for civil servants), employment contracts typically do specify the retirement date. The reason is simple: similar to other countries with large separation costs, setting the retirement data avoids costly layoffs.12 Because Sri Lankan employers can extend employment contracts of old workers beyond pre-set retirement dates, or hire old workers who retired from other firms, without being liable to severance payments (except gratuity), the necessary flexibility in employing old workers is achieved. Increasing the mandatory retirement age for civil servants, set currently at 55, would probably help keeping these workers employed for a longer time, but there is a host of other consideration that should be taken into account (above all, setting consistent hiring practices and considering revisions of wage scales) for this measure to be useful rather than just to increase the costs of bureaucracy. 12For example, OECD (2006) reports that firms in Japan and Korea are massively resorting to mandatory retirement age in order to evade strict employment protection legislation, because this legislation does not apply to mandatory retirements. 20 Does the absence of anti age-discrimination legislation reduce the chance of employment of old workers? While low LFP of old workers may reflect negative attitudes and age discrimination among employers, it is not clear that age discrimination laws would help. Such laws may not always be effective, and they may have significant unintended effects.13 OECD (2006) thus argues for the combination of antidiscrimination laws and information campaigns and guidelines. The UK Age Positive campaign is one example ­ the campaign is promoting age diversity at work through a practical guide for business in the areas of recruitment, selection of candidates, promotion, training and development, layoffs, and retirement, with detailed explanation of the guidelines and indicators for assessing conformity with them (UK 2007). (b) Are wages of old workers set too high? Setting the level of wages too high could work against the employment of old workers. More precisely, if wage increases associated with age exceed the increases in worker productivity, employers would prefer hiring young workers and would be reluctant to retain old workers.14 Indeed, in many countries wages by age show a hump-shaped profile, the profile that is most consistent with high level of employability of old workers (see OECD 2006 for a description of such profiles in OECD countries). The hump-shape profile could be explained by rising productivity of workers as they gain experience and falling productivity after a certain age (see Box 3 on the productivity of old workers). In Sri Lanka, relative costs of old vs. young workers appear not to create employment disincentives. As shown above, the age wage profile for the private sector reaches its peak for the 30-34 age group, shows a slow decrease thereafter and a sharp drop after the age of 60 for all workers. While this drop may reflect the selection issues that are not accounted for in age-wage profiles based on cross-section data, the magnitude of the drop of wages of old workers nonetheless suggests that the level of wages of old workers ­ and thus the wage- setting mechanism ­ is not a disincentive to hire or retain old workers in Sri Lanka. 13For example, in US a law prohibits age-based discrimination of hiring, firing, layoffs, and compensation, but as one researcher noted, in practice, "the law may have only limited benefits and significant costs. Although the laws provide a boon for older men who remain in their jobs and are more difficult to fire, they harm those seeking new employment." 14 Based on cross-country analysis, OECD (2006) provides some evidence of a negative impact of high relative wages on employment opportunities of old workers (drop in retention and hiring rates of male older workers). 21 Box 3: Are old workers less productive than young ones? Empirical evidence suggests so, although the relationship is often very non-linear and there are large differences across individuals and tasks. Studies find strong evidence that after the age of 50, several aspects of physical and mental abilities (such as reasoning and numerical capabilities) start to decline, although this decline is often progressive and can substantially vary from one person to another. In contrast, some verbal abilities remain virtually unchanged also late in the life cycle. It is also very important to note that apart from ability to perform a certain task, it is the ability to adjust to changes in the environment and to be able to work in teams that are important -- the ability that old workers may lack in comparison to young workers. A confirmation for such claims comes from productivity studies based on matched employer-employee data sets. For example, studies of the productivity of French and American firms found that workers aged 55 and over are less productive than younger workers. Source: OECD (2006). (c) Other factors Work stress and stress of traveling to work may also be obstacles to elderly employment. As shown above, a fair share of elderly have stopped work due to work stress and the stress of traveling to work. Work stress may be related to many factors, including inappropriate working conditions, discrimination of old workers, or the perception of the old workers about losing their ability to cater the needs of the market. OECD reports that large shares of old workers in OECD countries is exposed to physically unpleasant conditions (43 percent of men and 26 percent of women aged 50-64), and the conditions in Sri Lanka are probably no better. Another likely obstacle for employment of old workers may be their outdated skills and negative perceptions of employers about the adaptability and productivity of older workers. Even in OECD countries, improving training opportunities and the willingness to take advantage of them have been identified as a key area for strengthening employability of old workers. More research is needed to determine the needs ­ and the likely impact ­ of such training in Sri Lanka. And to change negative employer attitudes, a voluntary approach fostering public information campaigns and guidelines about employment of elderly may help. 22 7. CONCLUDING REMARKS The paper showed that in the coming decades, population aging in Sri Lanka will bring a slowdown of labor force growth and after 2030 its contraction. By studying labor market outcomes of old workers, the paper also demonstrated that labor market duality carries over to old age: while civil servants and formal private sector workers stop working early (generally before they reach 60 years) because they are forced to do so by mandatory retirement provisions, casual workers and the self-employed, in contrast, work until very old age (or death) because they are poor and have to earn their living, and stop working when they are forced to do so for health reasons. How should labor market and other policies respond to such circumstances? To counter the prospects of labor force reduction, policies could be aimed at increasing participation rates, particularly of women and older workers. To increase labor force participation of women, the introduction of family-friendly policies (including increasing part-time working opportunities) would be helpful, as would complementary policies aimed at reducing their burden as primary care givers to old persons (see World Bank 2007). To increase labor force participation of old workers, retirement age should be made more flexible. More research is needed to examine other possible obstacles for employment of old workers (such as inflexible labor laws) and to determine whether employers' negative perceptions about the adaptability and productivity of older workers create work disincentives for old workers ­ and if so, what should be done about it. Another area is improving productivity of the labor force. One important task is to improve skills of older workers through investments in life-long learning. While it is true that older workers undergo training less frequently, increasing the expected duration of their working lives will help them to be more motivated to take training ­ and it will also increase the payoffs of training to employers. Moreover, OECD (1998) argues that if workers undergo continued training throughout the career, it is unlikely that they would experience a decline in their trainability in the old age. In the context of developing country, the second important route to make labor more productive is formalization of the economy. On this front, one key task is making the labor market legislation less restrictive, so that workers could shift towards better, more productive jobs ­ and, at the same time, jobs that offer improved social 23 security. And there is also a third channel of raising productivity: improving health outcomes of old workers, the channel that contributes both to higher earning capacity and to longer working careers of workers. The last point underscores the importance of improving the choices available to old workers. As documented above, many workers ­ including a vast majority of informal sector workers ­ stop working only when they are prevented from doing so by ill health, so improving the health status of old people would significantly relieve financial pressures on them. Old peoples' choices would also significantly improve by providing them with an independent source of income, which underscores the need to extend the coverage of old age income support systems.15 In a country where extending working lives is key to relieve workers of financial pressures, our finding that ill health ­ more precisely, according to World Bank 2007b, chronic illness ­ is a major cause of withdrawal from the labor market, deserves special attention. While it highlights the need to formulate a coherent policy for addressing the health challenges of aging, it also points to the special needs of low socioeconomic groups, because these groups, as shown by this study, are most likely to be forced to work long into their old age. Indeed, World Bank (2007a) shows that in Sri Lanka, the burden of ill-health from untreated non-communicable diseases (NCDs) affects the poor more than the non-poor: the poor have worse access to medical services and so they are less likely than the non-poor to have their NCDs detected and treated. But the need to re-focus health policies to cater better to the needs of old people is clear also from the fact that physical disability rates among the old people in Sri Lanka in the past two decades have increased, in contrast to many other health outcomes. 15World Bank (2007a) points to the weaknesses of the current schemes for informal workers in Sri Lanka and lays out an alternative strategy to expanding social security to informal sector workers. 24 References Hermalin, A. I., 2002. The Well-Being of the Old Persons In Asia: A Four-Country Comparative Study, Ann Arbor: University of Michigan Press. Institute for Health Policy. 2007. Socio-Economic and Health Profile of Elderly in Sri Lanka. Background paper prepared for the project on population aging in Sri Lanka, processed, Institute for Health Policy, Colombo. Martins, J.O., F. Gonand, P. Antolin, C. de la Maisonneuve, and K.Y Yoo. 2005. The Impact of Aging on Demand, Factor Markets and Growth, OECD, Economics Working Papers No. 420. OECD. 2005. Health at Glance: OECD Indicators. Paris, France: OECD. OECD. 2006. Live Longer, Work Longer. Paris, France: OECD. de Silva, W. Indralal. 2007. A Population Projection of Sri Lanka for the New Millenium, 2001- 2101: Trends and Implications. Colombo, Sri Lanka: Institute for Health Policy. World Bank. 1994. Averting the Old Age Crisis. Policies to Protect the Old and Promote Growth. Washington, DC: The World Bank. World Bank. 2007a. Preventing the Old Age Crisis in Sri Lanka. SASHD (forthcoming). World Bank. 2007b. Sri Lanka: Strengthening Social Protection, World Bank: Human Development Unit, Report No. 38197-LK. 25 Social Protection Discussion Paper Series Titles No. Title 0821 Population Aging and the Labor Market: The Case of Sri Lanka by Milan Vodopivec and Nisha Arunatilake, August 2008 (online only) 0820 China: Improving Unemployment Insurance by Milan Vodopivec and Minna Hahn Tong, July 2008 (online only) 0819 Management Information Systems in Social Safety Net Programs: A Look at Accountability and Control Mechanisms by Cesar Baldeon and Maria D. Arribas-Baños, July 2008 (online only) 0818 Guidance for Responses from the Human Development Sectors to Rising Food Prices by Margaret Grosh, Carlo del Ninno and Emil Daniel Tesliuc, June 2008 (online only) 0817 Levels and Patterns of Safety Net Spending in Developing and Transition Countries by Christine Weigand and Margaret Grosh, June 2008 (online only) 0816 Labor Regulation and Employment in India's Retail Stores by Mohammad Amin, June 2008 (online only) 0815 Beyond DALYs: Developing Indicators to Assess the Impact of Public Health Interventions on the Lives of People with Disabilities by Daniel Mont and Mitchell Loeb, May 2008 0814 Enforcement of Labor Regulation and Firm Size by Rita Almeida and Pedro Carneiro, May 2008 (online only) 0813 Labor Markets Lending and Analytical Work at the World Bank: FY2002- 2007 by Milan Vodopivec, Jean Fares and Michael Justesen, May 2008 0812 Risk and Vulnerability Analysis in the World Bank Analytic Work: FY2000- 2007 by Valerie Kozel, Pierre Fallavier and Reena Badiani, May 2008 0811 Pension Lending and Analytical Work at the World Bank: FY2002-2007 by Richard Hinz, Melike Egelmelzer and Sergei Biletsky, May 2008 (online only) 0810 Social Safety Nets Lending and Analytical Work at the World Bank: FY2002-2007 by Margaret Grosh and Annamaria Milazzo, May 2008 0809 Social Funds as an Instrument of Social Protection: An Analysis of Lending Trends - FY2000-2007 by Samantha De Silva and June Wei Sum, July 2008 0808 Disability Activities at the World Bank: FY2002-2007 by Jeanine Braithwaite, Richard Carroll, and Karen Peffley, May 2008 0807 Migration, Labor Markets, and Integration of Migrants: An Overview for Europe by Rainer Münz, April 2008 (online only) 0806 Is the Window of Opportunity Closing for Brazilian Youth? Labor Market Trends and Business Cycle Effects by Michael Justesen, April 2008 0805 Disability and Poverty: A Survey of World Bank Poverty Assessments and Implications by Jeanine Braithwaite and Daniel Mont, February 2008 0804 Poverty Traps and Social Protection by Christopher B. Barrett, Michael R. Carter and Munenobu Ikegami, February 2008 0803 Live Longer, Work Longer: Making It Happen in the Labor Market by Milan Vodopivec and Primoz Dolenc, February 2008 (online only) 0802 Disability in Kazakhstan: An Evaluation of Official Data by Ai-Gul S. Seitenova and Charles M. Becker, February 2008 (online only) 0801 Disability Insurance with Pre-funding and Private Participation: The Chilean Model by Estelle James, Augusto Iglesias and Alejandra Cox Edwards, January 2008 0719 The Life-Course Perspective and Social Policies: An Issues Note by A.L. Bovenberg, November 2007 0718 Social Safety Nets and Targeted Social Assistance: Lessons from the European Experience by Chris de Neubourg, Julie Castonguay and Keetie Roelen, November 2007 (online only) 0717 Informality and Social Protection: Preliminary Results from Pilot Surveys in Bulgaria and Colombia by Franco Peracchi, Valeria Perotti and Stefano Scarpetta, October 2007 (online only) 0716 How Labor Market Policies can Combine Workers' Protection with Job Creation: A Partial Review of Some Key Issues and Policy Options by Gaëlle Pierre and Stefano Scarpetta, October 2007 (online only) 0715 A Review of Interventions to Support Young Workers: Findings of the Youth Employment Inventory by Gordon Betcherman, Martin Godfrey, Susana Puerto, Friederike Rother, and Antoneta Stavreska, October 2007 0714 Performance of Social Safety Net Programs in Uttar Pradesh by Mohamed Ihsan Ajwad, October 2007 0713 Are All Labor Regulations Equal? Assessing the Effects of Job Security, Labor Dispute and Contract Labor Laws in India by Ahmad Ahsan and Carmen Pagés, June 2007 0712 Convention on the Rights of Persons with Disabilities: Its Implementation and Relevance for the World Bank by Katherine Guernsey, Marco Nicoli and Alberto Ninio, June 2007 0711 Reaching the Poor and Vulnerable: Targeting Strategies for Social Funds and other Community-Driven Programs by Julie Van Domelen, May 2007 0710 The Macedonia Community Development Project: Empowerment through Targeting and Institution Building by Caroline Mascarell, May 2007 0709 The Nuts and Bolts of Brazil's Bolsa Família Program: Implementing Conditional Cash Transfers in a Decentralized Context by Kathy Lindert, Anja Linder, Jason Hobbs and Bénédicte de la Brière, May 2007 (online only) 0708 Globalization and Employment Conditions Study by Drusilla K. Brown, April 2007 0707 The Kosovo Pension Reform: Achievements and Lessons by John Gubbels, David Snelbecker and Lena Zezulin, April 2007 (online only) 0706 Measuring Disability Prevalence by Daniel Mont, March 2007 0705 Social Safety Nets in World Bank Lending and Analytic Work: FY2002- 2006 by Annamaria Milazzo and Margaret Grosh, March 2007 (online only) 0704 Child Labor and Youth Employment: Ethiopia Country Study by Lorenzo Guarcello and Furio Rosati, March 2007 0703 Aging and Demographic Change in European Societies: Main Trends and Alternative Policy Options by Rainer Muenz, March 2007 (online only) 0702 Seasonal Migration and Early Childhood Development by Karen Macours and Renos Vakis, March 2007 0701 The Social Assimilation of Immigrants by Domenico de Palo, Riccardo Faini and Alessandra Venturini, February 2007 (online only) 0616 Pension Systems in Latin America: Concepts and Measurements of Coverage by Rafael Rofman and Leonardo Lucchetti, November 2006 (online only). Also available in Spanish. 0615 Labor Market Outcomes of Natives and Immigrants: Evidence from the ECHP by Franco Peracchi and Domenico Depalo, November 2006 (online only) 0614 The Relative Merits of Skilled and Unskilled Migration, Temporary and Permanent Labor Migration, and Portability of Social Security Benefits by Johannes Koettl under guidance of and with input from Robert Holzmann and Stefano Scarpetta, November 2006 (online only) 0613 The Limited Job Prospects of Displaced Workers: Evidence from Two Cities in China by Gordon Betcherman and Niels-Hugo Blunch, October 2006 0612 Unemployment Insurance in Chile: A New Model of Income Support for Unemployed Workers by Germán Acevedo, Patricio Eskenazi and Carmen Pagés, October 2006 0611 Evaluating Social Fund Impact: A Toolkit for Task Teams and Social Fund Managers by Sarah Adam, October 2006 0610 Risk and Vulnerability Considerations in Poverty Analysis: Recent Advances and Future Directions by Carlo Cafiero and Renos Vakis, October 2006 0609 Comparing Individual Retirement Accounts in Asia: Singapore, Thailand, Hong Kong and PRC by Yasue Pai, September 2006 (online only) 0608 Pension System Reforms by Anita M. Schwarz, September 2006 (online only) 0607 Youth Labor Market in Burkina Faso: Recent Trends by Daniel Parent, July 2006 0606 Youth in the Labor Market and the Transition from School to Work in Tanzania by Florence Kondylis and Marco Manacorda, July 2006 0605 Redistributing Income to the Poor and the Rich: Public Transfers in Latin America and the Caribbean by Kathy Lindert, Emmanuel Skoufias and Joseph Shapiro, August 2006 (online only) 0604 Uninsured Risk and Asset Protection: Can Conditional Cash Transfer Programs Serve as Safety Nets? by Alain de Janvry, Elisabeth Sadoulet, Pantelis Solomon and Renos Vakis, June 2006 0603 Examining Conditional Cash Transfer Programs: A Role for Increased Social Inclusion? by Bénédicte de la Brière and Laura B. Rawlings, June 2006 (online only) 0602 Civil-service Pension Schemes Around the World by Robert Palacios and Edward Whitehouse, May 2006 (online only) 0601 Social Pensions Part I: Their Role in the Overall Pension System by Robert Palacios and Oleksiy Sluchynsky, May 2006 (online only) To view Social Protection Discussion papers published prior to 2006, please visit www.worldbank.org/sp. Summary Findings Sri Lanka's population is predicted to age vary fast during the next 50 years, bringing a slowdown of labor force growth and after 2030 its contraction. Based on a 2006 representative survey of old people in Sri Lanka, the paper examines labor market consequences of this process, focusing on retirement pathways and the determinants of labor market withdrawal. The paper finds that a vast majority of Sri Lankan old workers are engaged in the informal sector, work long hours, and are paid less than younger workers. Moreover, the paper shows that labor market duality carries over to old age: (i) previous employment is the most important predictor of the retirement pathway; (ii) older workers fall into two categories: civil servants and formal private sector workers, who generally stop working before they reach 60 because they are forced to do so by mandatory retirement regulations, and casual workers and the self-employed, who work until very old age (or death) due to poverty and insufficient income and who stop working primarily because of poor health; and (iii) the option of part-time work is used primarily by workers who held regular jobs in their prime age employment, but not by casual workers and self- employed. HUMAN DEVELOPMENT NETWORK About this series... Social Protection Discussion Papers are published to communicate the results of The World Bank's work to the development community with the least possible delay. The typescript manuscript of this paper therefore has not been prepared in accordance with the procedures appropriate to formally edited texts. The findings, interpretations, and conclusions expressed herein are those of the author(s), and do not necessarily reflect the views of the International Bank for Reconstruction and Development /The World Bank and its affiliated organizations, or those of the Executive Directors ofThe World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. For free copies of this paper, please contact the Social Protection Advisory Service,TheWorld Bank, 1818 H Street, N.W., Room G7-703,Washington, D.C. 20433-0001.Telephone: (202) 458-5267, Fax: (202) 614-0471, E-mail: socialprotection@worldbank.org or visit the Social Protection website at www.worldbank.org/sp.