Kazakhstan Living Standards During the Transition

Transcription

1 Report No KZ Kazakhstan Living Standards During the Transition March 23, 1998 Hunian Developiment Sector Unit Europe and Central Asia Region Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Document of the World Bank

2 CURRENCY EQUIVALENTS (As of December 31, 1997) Currency Unit = Tenge (T) TI = US $ US $ I T AVERAGE EXCHANGE RATE Tenge per US $ ABBREVIATIONS AND ACRONYMS CA Child Allowances ES Employment Services FBS Family Budget Survey FSU Former Soviet Union GDP Gross Domestic Product ILO International Labor Organization KLSS Kazakhstan Living Standards Survey LFPR Labor Force Participation Rate LSMS Living Standards Measurement Survey MLSP Ministry of Labor and Social Protection MOF Ministry of Finance NGO Non-governmental Organization PSRMAL Public Sector Resource Management Adjustment Loan Q 1-Q4 Quarter 1 - Quarter 4 SIF Social Insurance Fund SM Subsistence Minimum (Prozhitochnyi Minimum) UB Unemployment Benefit UJNDP United Nations Development Program KAZAKHSTAN FISCAL YEAR January 1-December 31 Vice President: Johannes Linn Director: Christopher Lovelace Sector Leader: Michal Rutkowski Staff Member: Mamta Murthi

3 ACKNOWLEDGEMENTS This report is based on the findings of a World Bank mission that visited Kazakstan from April 7-19, A major source of the analysis is the Kazakstan Living Standard Survey (KLSS) conducted in July 1996 under a Bank-financed Technical Assistance Loan. Preliminary results were discussed at a joint Ministry of Labor and Social Protection-World Bank workshop held in Almaty on April 11-12, 1996 which was attended by officials of the Government Apparat, the Ministry of Labor and Social Protection, the Ministry of Finance, the former Ministry of Economy and Trade, the Ministry of Education and Culture, and the Committee on Statistics and Analysis (CSA). The workshop was facilitated by EDI/SPRITE. Special thanks are owed to the Ministry of Labor and Social Protection and the Committee on Statistics and Analysis (CSA) for their role in ensuring the success of the workshop, and in particular to the CSA for making available the raw data from the KLSS. Mamta Murthi is the task manager and principal author of the report. It is based on background papers by Kinnon Scott (poverty profile) and Menno Pradhan (targeting of cash benefits). Alan Thompson (pensions and social assistance), Alan Abrahart (labor markets), and Elca Rosenberg (structure, logistics) made valuable contributions to the design and conduct of the workshop. John Innes was senior advisor to the project. Zhanar Abdildina provided support from the Resident Mission. The report has benefited from the comments of Emily Andrews, Jeanine Braithwaite, Betty Hanan, Emmanuel Jimenez, Christine Jones, Peter Hansen, Peter Lanjouw, Phillip O'Keefe, and Helena Ribe. Annie Milanzi and Ian Conachy provided invaluable production assistance.

4 CONTENTS EXECUTIVE SUMMARY.... ivi 1. DEVELOPMENTS IN THE ECONOMY AND LABOR MARKET...1 INTRODUCTION : FROM CRISIS TO RECOVERY : FISCAL CRISES : LABOR MARKET DEVELOPMENTS : FACTORS EXPLAINING LABOR MARKET STATUS... 8 CONCLUSIONS THE PROFILE OF THE POOR IN KAZAKHSTAN... l 1 INTRODUCTION MEASURING POVERTY IN KAZAKHSTAN I POVERTY IN KAZAKHSTAN IN A PROFILE OF THE POOR IN KAZAKHSTAN EXPENDITURE PATTERNS STRUCTURE OF CASH INCOME AND INFORMAL SECTOR ACTIVITY CONCLUSIONS THE ROLE OF THE SOCIAL SAFETY NET IN KAZAKHSTAN INTRODUCTION : EVOLUTION OF THE SOCIAL SAFETY NET : TRENDS IN SOCIAL SAFETY NET EXPENDITURE THE IMPORTANCE OF CASH TRANSFERS IN KAZAKHSTAN THE TARGETING OF CASH BENEFITS CONCLUSIONS REDUCING POVERTY IN KAZAKHSTAN INTRODUCTION THE IMPORTANCE OF GROWTH FOR POVERTY REDUCTION IN KAZAKHSTAN IS GROWTH SUFFICIENT? FISCAL CAPACITY AND POVERTY REDUCTION STRENGTHENING SOCIAL PROTECTION REGIONAL POLICY AND POVERTY REDUCTION CONCLUSIONS APPENDICES APPENDIX 1: DATA AND METHODOLOGY APPENDIX 2: SENSITIVITY ANALYSIS APPENDIX 3: SOCIAL PROTECTION IN KAZAKHSTAN APPENDIX 4: POVERTY AND ECONOMIES OF SCALE IN CONSUMPTION... 73

5 Kazakhstan: Living Standards During the Transition Executive Summary 1. This report examines the impact of the transition on living standards in Kazakhstan, assesses the scope for burgeoning growth to reduce poverty, and makes recommendations on how to strengthen the social safety net. There are three principal findings. First, over a third of the population lived below a 'subsistence minimum' living standard in This is significantly lower than estimates of poverty based on Family Budget Survey (FBS) income distribution data, and is due to problems with the measurement of income in the FBS. Second, given the proportion of the population that can be considered poor, it should be clear that the problem of poverty cannot be eliminated by public transfer programs. Indeed, the social safety net has shrunk substantially since independence under the combined impact of falling revenues and the need to maintain fiscal discipline. Strong and sustained growth will therefore be key to poverty reduction in Kazakhstan. Third, given that it is unrealistic to expect a significant expansion of social protection expenditures in the near future, assisting the poor will hinge on making existing social programs more effective. This should include maintaining an adequate income floor for pensioners, increasing the coverage of unemployment benefit, and reducing the leakage of child allowances and other social assistance to the non-poor. Developments in the macroeconomy and the labor markets 2. Although poverty was not unknown in Soviet times, Kazakhstan was one of the least poor central Asian republics. However, the sharp contraction in output since independence has contributed significantly to an increase in poverty. In cumulative terms, the decline in output in Kazakhstan since 1991 has been of the order of 40 percent. Although employment has fallen, the burden of adjustment has fallen on the price of labor. Wages have fallen by over 50 percent in real terms, with serious implications for household income and consumption. Open unemployment remains low relative to the scale of output decline. Unemployment (ILO definitions) stood at around 4.5 percent-at the end of 1994, rising to around 6 percent in Workers who are the most severely affected by unemployment appear to be new labor force entrants, those with low education or vocational and technical skills and, to a lesser extent, women. 3. At independence, Kazakhstan inherited an economy deeply dependent on Soviet supply and trade networks. The demise of central planning, the loss of established patterns of supply and trade, and the loss of transfers from Moscow following independence all contributed to large falls in output. The need to bring inflation under control squeezed the economy further. Stabilization efforts were however successful in reducing inflation which fell from over 1800 percent in 1994 to 29 percent in 1996 and 11 percent in The aftermath of separation from the Soviet Union appeared to have been finally thrown off in 1996 when, after declining continuously for five years, the economy began to grow again although at a modest pace. Economic growth during 1997 is estimated at 2 percent. 4. Another feature of macroeconomic decline has been a fall in the government's ability to mobilize tax revenues. As a result, tax revenue as a share of GDP fell from 39 percent in 1992 to i

6 24 percent in Poor revenue performance has contributed to disproportionate declines in the social safety net, in public investment, and the funding of public services. In an effort to keep the budget deficit under control, the government has accumulated large wage and cash benefit arrears. At the end of 1996, arrears in wages stood at 6 percent and arrears in pensions at 2 percent of GDP. Pension arrears were fully cleared in 1997 at the cost of considerable belttightening in other areas. The profile of the poor in Kazakhstan 5. This report makes use of the Kazakhstan Living Standards Survey (KLSS), a representative household survey conducted in July 1996, to examine the extent and main correlates of poverty in Kazakhstan. Households whose per capita consumption is less that the Government's 'subsistence minimum' (prozhitochnyi minimum) are considered to be poor.l Nearly 35 percent of the population was found to be living in poverty in July While by no means low, the estimate is significantly lower than poverty estimates based on income distributions generated by the Family Budget Survey (64 percent in 1995 and over 80 percent in 1996), largely due to problems with the measurement of income. Income data are often subject to a under-reporting, particularly in the case of income from own business and informal activity. While there are differences in rural and urban poverty rates, the more striking difference in Kazakhstan is by region. The poverty rate in the south of the country (69 percent) is nearly twice the national average, while that in the north (9 percent) is many times lower. 2 In terms of number of poor, nearly two out of three poor people live in the south or east of the country. This has the strong implication that reducing poverty in Kazakhstan is largely a question of improving the opportunities for people in these two regions. 6. The poverty profile highlights correlates of poverty that are similar to that observed in other parts of the former Soviet Union (FSU). Large households with many children or many dependents (whether young or old) tend to be poorer, as do households with unemployed members. There is also a distinct correlation between low levels of education, and technical and vocational training, and poverty. Pensioners are also found to be at risk of poverty, although they are relatively well-protected from extreme (bottom quintile) poverty. 7. The analysis of household expenditures points to the emergence of own-production as a very important source of food consumption in Kazakhstan. Over a third of all food consumed on average in July 1996 was derived from home-production. In addition to growing their own food, individuals engage in other informal activity, including working in second jobs and running their own businesses. There are large, private inter-household transfers in Kazakhstan -- mostly from family and other relatives -- which are another important source of income. The growth of I The subsistence minimum is based on a consumption basket, 70 percent of which is food and 30 percent non-food goods. The food basket is derived from official nutritional norms. In dollar terms, it is equivalent to around US $70 per capita per month at purchasing power parity (and around US $40 at the market rate of exchange). Although more generous than many poverty lines, the subsistence minimum does have the advantage of being recognizable. 2 The regions are defined as: North: Kostanai, Kokshetau, Pavlodar, and North Kazakhstan oblasts; Center: Zhezkazgan, Karaganda, Akmola and Torgai oblasts; West: Manghystau, Atyrau, Aktyubinsk, West Kazakhstan; East: Semi Palatinsk, East Kazakhstan, Taldykorgan and Almaty oblasts and Almaty city; South: Kzyl-Orda, South Kazakhstan and Zhambyl oblasts. In April 1997, the number of oblasts was reduced from nineteen to fourteen. ii

7 informal incomes does not come as a surprise given the sharp fall in main wages and formal social support. But equally, informal incomes significantly compromise the ability of the governmento effectively means-test available social assistance. The role of the social safety net in Kazakhstan 8. Kazakhstan's social safety net has undergone major transformation since The need to maintain fiscal discipline in the context of a sharp contraction in revenues meant that spending on the social safety net as a share of GDP fell by over a third from 11.2 percent in 1992 to 6.6 percent in Although new forms of social protection have been introduced -- such as unemployment assistance -- there has been a general decline in program beneficiaries assisted by a tightening of eligibility criteria. Real benefit levels have declined substantially, with the exception of means-tested child allowances. Although they have increased in real terms in the last two years, pensions remain at less than 50 percent of 1993 levels. 9. Pensions dominate social protection expenditures (over 75 percent), while child allowances come a distant second. The housing allowance scheme, which is currently quite small, is expected to expand in the coming years. Despite the general contraction of the system, there remain a multiplicity of small benefits which are complex to administer. 10. The 'thinning' of the social safety net is evident from the fact that over 60 percent of poor households in Kazakhstan receive no public transfers at all. This is partly related to pure 'gaps' in the social safety net, and partly due to non-payment of benefits to eligible households due to resource limitations. In the lowest quintile, pensions and other transfers contribute to about 20 percent of consumption on average. 1I. Among existing social assistance programs, targeting of benefits towards the lowest quintiles is generally poor. In the case of the means-tested child allowance scheme, over 50 percent of the program expenditure leaks to the non-poor (defined as the top 60 percent). Better targeting outcomes have been achieved in other countries such as Uzbekistan, Albania and Chile. Social assistance which is targeted using categorical criteria generally does farworse in terms of leakage than the means-tested child allowance program as the categories in use are inherited from Soviet times and do not appear to correspond to the present realities. Reducing poverty in Kazakhstan: growth and the social safety net 12. Given the scale of output decline, and the rise in poverty, it should be clear that broadbased growth that both raises wages and creates employment will be key to reducing poverty in Kazakhstan. Growth is critical not just for raising incomes but also for providing the resources for a fiscally sustainable safety net to help those who might be temporarily or chronically disadvantaged. 13. In general, the prospects for a sustained increase in output over the medium term in Kazakhstan are good, both because it is one of the stronger reformers among the FSU countries, and because it is well-endowed with natural resources. 3 The World Bank estimates that the 3 Natural resources can be a mixed blessing. The competitiveness of non-oil sectors of the economy has been found to be a problem in almost every country that has experienced a rapid expansion in foreign iii

8 economy will grow modestly in the short-run (2-3 percent), picking up speed (4-6 percent) early in the next century. However, there are several critical areas that need attention if growth in Kazakhstan is to be sustained: (i) further development of the legal and regulatory framework for private sector development, (ii) stronger regulation and oversight of the financial sector, (iii) deepening of capital markets in conjunction with the careful management of pension reform, (iv) sound management of the process of enterprise restructuring to encourage output and productivity growth, (v) improved management of public resources, including an expansion in the level and the quality of the public investment program, and judicious handling of likely future natural resource revenues, (vi) further restructuring and improvement in the quality of health and education services, and (vii) improved governance through the development of a professional, well-remunerated civil service. 14. While growth is a necessary condition for the reduction of poverty in Kazakhstan, it is not sufficient. Whether poverty is reduced by growth is a function of how income growth is distributed in society, and the opportunities offered by growth to the poor to improve their income earning opportunities. The restructuring that has yet to take place (notably in agriculture, but also mining, transportation, utilities, and public administration) is likely at least in the short run to swell the ranks of the unemployed. The evidence from the KLSS on the returns to education suggests that those with low education or mismatched skills may firld it difficult to find work. The social safety net will be critical for these and other groups who are not well-positioned to benefit from growth. 15. While important for maintaining the living standards of the poor, severe constraints have been imposed on the social safety net by poor revenue performance and the need to keep the budget deficit under control. Moreover, the transition from a public pay-as-you-go pension system to a private fully funded system starting in 1998 will add a significant burden to the budget for many years to come due to the reduction in payroll taxes contributed to the public system. 4 A national poverty reduction strategy should therefore aim to make social programs more effective within existing fiscal constraints. This report makes five main recommendations in this respect: 16. Ensure the adequacy of pensions under the pension reform: From a poverty perspective, the strength of the existing pension system in Kazakhstan is that pensioners do not face increased risks of extreme (bottom quintile) poverty. The 1998 reform provides for a minimum pension of 2,400 Tenge which is about 70 percent of the 'subsistence minimum' and in keeping with World Bank recommendations. However, a key concern is whether the minimum pension will be maintained in real terms. A second issue is whether state pensions will be paid on time. It will be important for the Government to maintain the minimum pension in real terms to protect the living standards of poor pensioners, and to keep to its commitment to prevent the emergence of new pension arrears. A medium-term issue under the reformed system is pension adequacy for low-income workers, those with short work histories, and women, all of whom typically receive lower pensions from systems which rely on personal saving. There are a exchange earnings from petroleum exports. In a recent study covering growth in ninety-seven developing countries during the period , Sachs and Warner (NBER Working Paper, 1995) found that countries with high export earnings based on natural resource wealth grew more slowly than those without such natural resources. 4 In the first year of the reform, the payroll tax funding the public system would fall from 25 percent to 15 percent. iv

9 number of options here for improving outcomes, including: higher contribution rates, contributions from other social insurance funds to cover periods of temporary inability to work (owing to sickness, maternity or unemployment), joint treatment of contributions, equalization of retirement age for men and women, and the use of standard life tables when issuing annuities. The Government should develop the capacity to analyze likely pensions for workers with nonstandard work-history and other characteristics and develop options to ensure adequate replacement rates. 17. Improve the targeting of social assistance: Targeting of the main programs of social assistance in Kazakhstan (e.g. child allowances) has largely relied on formal income-testing. But given the structure of incomes in Kazakhstan, alternative approaches clearly need to be developed. One way to improve targeting at low cost would be to use additional household characteristics that are well-correlated with poverty, such as the presence of young children or unemployed members. More sophisticated proxy means-testing could also be considered, with attention both to administrative capacity and costs. Attention should also be paid to features of the housing allowance scheme that tend to exclude rural households (who are equally needy) and to replacing directed subsidies in utilities with targeted cash transfers to those who are in genuine need. 18. Rationalize social assistance programs: In addition to main programs which are meanstested, there a number of small programs (e.g. other child allowances, local social assistance programs) which are categorically targeted at pre-identified "vulnerable" groups. Data from the KLSS show that these programs largely benefit the middle income groups. These resources should be redirected towards those who are truly in need. A simple way to do this would be to continue the existing trend of merging existing programs with means-tested ones. Where this is not feasible, it would be worth considering making benefits subject to an income test, which although it has its limitations, would be an improvement over present criteria. 19. Strengthen the system of unemployment compensation: Although not directed at the poor, unemployment benefit in Kazakhstan is relatively well-targeted. Attention should be paid to increasing coverage, including through better provision of information to the unemployed and, possibly, through raising compensation levels. At the same time, rules governing receipt -- especially with respect to job refusals -- should be tightened so as not to escalate costs. 20. Consider assisting the unemployed through regionally targeted temporary employment schemes: Given the administrative costs and complexity of improving the targeting of individual programs, the scope for exploiting the self-targeting potential of public works programs deserves closer attention. Monitoring progress with poverty reduction 21. Whatever improvements are introduced, it will be essential to monitor changes in poverty and targeting effectiveness. This report recommends developing the Family Budget Survey into an appropriate instrument for such a monitoring exercise. Improvements to the FBS being initiated under the under the World Bank's Technical Assistance Loan should be sustained. The Government should maintain a commitment to tracking the impact of growth and policy changes on living standards and developing the capacity to undertake analyses of alternative reform options. v

10 Future agenda 22. Finally, this report does not address three types of policies that will need to complement the strengthening of the social safety net to improve living standards in Kazakhstan: (i) Labor market policies designed to facilitate labor market restructuring, encourage mobility, and equip workers for changed economic conditions; (ii) Policies aimed at reducing regional imbalances; (iii) Health and education sector policies to maximize the poverty reduction impact of reduced expenditures. The first of these is to be the subject of a subsequent study to be initiated during A start has been made on the second, through the World Bank's study of the Kzyl-Orda oblast which is affected by the Aral Sea crisis. Further work, particularly pertaining to the role of the republican government in reducing imbalances, and role of regional development policies, is clearly warranted. Restructuring the health and education systems so as to maximize quality and efficiency from reduced resources will be an ongoing challenge over the medium term. A Bank pilot project in the health sector should be followed closely for lessons for the reform of the medical care system throughout the country. vi

11 1. DEVELOPMENTS IN THE ECONOMY AND LABOR MARKET Introduction 1.1 Poverty, in the sense of low or inadequate income, is not a new phenomenon in Kazakhstan, but has been aggravated by economic decline. As in other transition economies, economic activity has contracted significantly since independence. Although macroeconomic performance has been improving -- output growth resumed in 1996 for the first time since independence -- GDP remains at a little over sixty per cent of 1990 levels. The fall in output has resulted in a substantial decline in real wages and incomes, while inflation has virtually destroyed household savings. The weakened fiscal position of the state has limited its ability to mitigate the impact of these changes on household welfare, and has reduced the quality and quantity of public services. 1.2 Although poverty was not unknown in Soviet times, Kazakhstan was the least poor of the central Asian republics. Estimates from the 1989 Family Budget Survey (FBS) suggest that around 15 percent of the population had incomes below the 'socially acceptable minimum' compared to percent in other central Asian SSRs. 1 The size of the country however meant that in central Asia, it was second only to Uzbekistan in terms of its contribution to the total number of poor in the FSU. There were significant differences in the incidence of low incomes among population sub-groups. Collective farm households, for example, were over twice as likely to be poor as enterprise employees. 2 Overall inequality in the distribution of income was comparable to other central Asian republics, as there was typically little variation in inequality within individual republics. 1.3 Although households may have suffered low incomes during the Soviet period, the severity of poverty was limited by state action. Low administered prices for a range of key goods and services (food, rent, utilities) kept the cost of living strictly under control. Guaranteed employment, and a relatively generous system of pensions, family allowances, sick pay and maternity benefits insulated households from major income shocks. Subsidies to child-care helped maintain a fairly high level of participation among women. Access to education and healthcare was universal. This system of social provision underlies the strong achievements in the field of human development that is characteristic of the FSU countries. At independence, most human development indicators were higher in Kazakhstan than in countries with comparable levels of income (see Table 1.1), an achievement which largely persists despite recent difficulties. I Atkinson and Micklewright, 1992, Economic Transformation in Eastern Europe and the Distribution of Income. At the time, the all-union poverty line stood at 75 rubles per capita per month. The figures should be treated as indicative as the FBS sample is not nationally representative. 2 Valuation of agricultural produce at state-controlled prices and underreporting of income from private plots probably overstates poverty on collective farms. For a further development of this line of argument, see Atkinson and Micklewright, ibid. 1

12 Table 1.1: Comparison of selected human development indicators in 1991 Kazakhstan OECD Middle-income countries GNP per capita (US$) 2,470 21,530 2,480 Adult literacy (%), Tertiary graduates' Infant mortality' Maternal mortality- 53 n.a. 107 Life expectancy (years): Male Female as a percent of potential graduates per 1,000 live births. The figure for Kazakhstan has been adjusted for international comparability. per 100,000 live births Source: World Bank (1993), Kazakhstan Transition to a Market Economy; UNDP (1994), Human Development Report 1.4 This chapter reviews the main developments since independence focusing on output and inflation, fiscal adjustments, and developments in the labor market. In doing so, it attempts to provide a background against which to evaluate changes in living standards during the transition : From crisis to recovery 1.5 The most important factor affecting living standards in Kazakhstan since independence has been the unprecedented fall in output. After declining continuously for five years, GDP expanded modestly in 1996, but remained below 60 percent of pre-independence levels. The main developments in the macroeconomy since 1991 are summarized in Table The quality of official statistics has suffered during the transition, partly as a result of disruptions to traditional data recording mechanisms and partly on account of the growing unrepresentativeness of existing data-gathering techniques which fail to capture the full extent of private and informal sector activity. Official series may therefore overstate output decline. However, the figures in Tables 1.2 incorporate significant improvements in coverage and technique which have resulted in an upward revision in the level of GDP. 3 Even allowing for some margin of error in these revised estimates, there can be little doubt that there has been a sharp and significant decline in output and incomes in Kazakhstan. 3 The revised estimates of GDP are based on the production approach and are available up to They were developed as part of a joint World Bank-National Statistical Agency project to strengthen national income accounting in Kazakhstan. See National Statistical Agency, Report on National Accounts, February The post-1995 estimates are preliminary. 2

13 1.7 The contraction in output in Kazakhstan since 1991 is related to the twin challenges of adjusting to the dissolution of the FSU while at the same time managing the transformation of a planned to a market economy. Kazakhstan was part of a Union wide division of labor and inherited an economy dominated by agriculture, mining (coal and metals), and energy. The Russian hinterland was (and still remains) the main source of inputs and the main market for output. Transport and other infrastructure was designed with the view to meeting these needs, not those of the local economy. 4 Social spending was high, maintained in part by transfers from Moscow. Central budgetary support was estimated to be of the order of 12 percent of GDP in the early 1990s. 5 These arrangements were seriously disrupted by the dissolution of the FSU. The demise of central planning, the disruption of trade and other economic arrangements, and the end of transfers all contributed to large falls in output. Table 1.2: Selected macroeconomic indicators in Kazakhstan since independence Real growth in GDP(%) Index of real GDP (1990=100) Index of real consumption (1993=100) n.a. Inflation (end-of-year change in CPI) 90 1,381 1,662 1, Fiscal deficit (% GDP) Export growth n.a. Nominal rate of exchange (ann. avg. T/$) Index of real exchange rate (1993=100) Direct foreign investment ,107 1,289 (current million US $) Sources: GDP, Exports, Direct foreign investment: Committee on Statistics and Analysis; Inflation, Fiscal deficit: IMF; Consumption, Exchange rates: World Bank staff estimates. Figures for 1997 are preliminary 1.8 Massive inflation, associated with uncontrolled deficit financing and loose directed credit, was an additional problem in the early years (see Table 1.2). In an attempt at stabilization, the authorities introduced a new currency (the tenge) late in It devalued almost immediately following an unsuccessful attempt to clear inter-enterprise arrears and loss of monetary control. After this somewhat rocky start, stabilization efforts proved more successful. The budget deficit was gradually reined in and money supply growth brought under control. In response, inflation dropped to 11.5 percent in 1997 after peaking at 1880 percent in The inevitable short-run impact of monetary and fiscal restraint was to squeeze output further. 1.9 Starting in 1992, the authorities undertook a number of steps to liberalize and transform the structure of the economy. Price liberalization was initiated soon after independence, and with the 4 For example, the lack of pipeline capacity is a key constraint to the export of oil and gas outside the FSU. 5 See World Bank, 1994, Kazakhstan Economic Report. Transfers partly compensated for low administered prices for energy. 3

14 exception of a few goods, was virtually complete by Currently, the only administered prices in Kazakhstan are for utilities and transport which are being adjusted gradually to cost recovery levels although with a substantial regime of privileges in place. In contrast, trade liberalization occurred more slowly. However, by end-1996 most restrictions on exports -- whether in the form of quotas, licenses, or state marketing arrangements -- had largely been removed Progress has also been made with creating competitive market conditions, most notably with the dismantling of most trading monopolies and holding companies in 1995 and To date, about two thirds of small- and medium-scale enterprises have been privatized. Case-by-case privatization of large enterprises -- which has attracted considerable foreign interest -- has however experienced significant delays. A bankruptcy law is in place 7 and progress has been made with banking and financial sector reform. The international business community has responded favorably to the economic conditions in the country, and in particular to opportunities in the oil and gas sector, with large and growing inflows of foreign direct investment (see Table 1.2). As a result, Kazakhstan is the largest recipient of foreign direct investment in per capita terms among the CIS countries. Foreign investment is mostly concentrated in the areas of energy and metals with a small proportion in food-processing and tobacco As indicated earlier, after contracting for five years the economy appeared to turn the corner in Inflation continued to fall, while output growth resumed. The economy grew by over 1 percent in 1996, and over 2 percent in The sectors with the most significant growth are agriculture, services (trade and housing), and energy. Export performance has improved with exports expanding by 5 percent in 1995 and 9 percent in 1996 despite a real appreciation in the tenge (Table 1.2). The position of the current account has improved significantly On a less positive note, the reduction in the deficit has come at the cost of timely payment of budgetary obligations. Wage and pension arrears mounted steadily from around 4 percent of GDP at the end of 1995 to 6 percent at the end of 1996 despite progress with clearing the bulk of central government wage obligations during the year. Delays in the payment of wages and pensions have become a significant source of public dissatisfaction with the economic transition (see Box 1.1). In response, the government launched a serious effort to clear all pension arrears during 1997 and, to its credit, completed this successfully. To some extent, however, this may have resulted in arrears in other (non-pension related) areas. The backlog of other government arrears, including wage arrears, is expected to be cleared over the next two years More generally, there has been a massive build-up of payments arrears in the economy with arrears and barter arrangements substituting to a large extent for money and credit. Economy wide interenterprise arrears have grown steadily and were estimated at 38 percent of GDP in Privatization, the growth of the informal sector, arrears, and barter have put many activities beyond the reach of the tax authorities with negative implications for the government's ability to collect revenues. This is discussed further below. Despite the resumption of growth, public opinion remains firmly pessimistic, perhaps because the recovery has been relatively modest. 8 6 However, the liberalization of input prices before output prices in agriculture resulted in a massive financial squeeze in the sector and a sharp contraction in output in The agricultural sector is however outside the purview of the law. 8 In December 1996, 82 percent of adults polled said they were dissatisfied, and 38 percent very dissatisfied with the overall economic situation. A year earlier, the figures were 77 percent and 32 percent respectively. See International Foundation for Election Systems (1996), Public Opinion in Kazakhstan. 4

15 Box 1.1: Public dissatisfaction with the economic transition in December 1996 Sample size: 1,500 Question: What are the reasons why you are dissatisfied with the situation in Kazakhstan today? Most frequently volunteered answers ( percent of respondents) Delay in wage/pension 18 Fall in income 17 Unemployment 11 Macroeconomic decline 6 Source: International Foundation for Election Systems (1996), Public Opinion in Kazakhstan : Fiscal crises 1.14 The transition in Kazakhstan has been accompanied by a growing incapacity of the government to mobilize resources. As a result, government revenues have fallen as a share of GDP from about 39 percent in 1992 to 24 percent in 1996 (see Table 1.3). Combined with the fall in GDP, this has resulted in a situation where public expenditures are running at about a third to a half of pre-independence levels in real terms. The cuts have fallen disproportionately on public investment (which has virtually ceased), social protection, and within public services, on operation and maintenance expenditures. 9 Table 1.3: Fiscal adjustment in Kazakhstan since independence Govemment revenues and grants (% GDP) n.a. Social protection expenditures (% of GDP) Index of 'base enumerate' (1992=100) Index of real per capita exp. on healthcare (1992=100) Index of real per capita exp. on education (1992=100) Notes: The 'calculation base' is a nominal floor to which many transfers are pegged. Sources: World Bank (1996), Republic of Kazakhstan Transition of the State, and staff estimates. Figures for 1997 are preliminary The social safety net has worn thin under the impact of the fall in GDP and poor tax collection. Public expenditure on social protection -- pensions, unemployment benefits, family allowances, sick pay, maternity benefits, and local social assistance -- fell from about 11.2 percent of GDP in 1992 to 6.6 percent in In 1997, expenditures rose to 7.9 percent as the government cleared pension arrears of the order of 1.5 percent of GDP during the year. Most spending is financed out of extra-budgetary funds 9 See, World Bank (1996), Republic of Kazakhstan Transition of the State. 5

16 -- the Pension Fund being the most significant -- which receive fixed proportions of a payroll tax, and local budgets. The transition has affected social protection payments in two ways: first, it has eroded their real value. This is perhaps best illustrated in terms of the 'calculation base', a nominal floor to which many public transfers are pegged."' In 1997, it stood at 16 percent of its 1991 value in real terms. Second, the system is unable to meet the all the claims placed upon it. This has resulted in the build-up of arrears. At the end of March 1997, for example, pension arrears stood at 26 billion tenge (equivalent to three months' payments) while arrears in the payment of child allowances were 4 billion tenge (equivalent to six months' payments). It took a concerted effort, as well compression of other expenditures, to pay off pension arrears during The change in fiscal circumstances has resulted in a sharp fall in the resourcing of public goods such as education and healthcare. Public outlays in education and healthcare were running at aboutone third to one half of 1991 levels in real per capita terms (see Table 1.3). Although there has been some restructuring with the closure and/or merger of facilities, it has been small compared to the scale of resource decline. The burden of adjustment has fallen on recurrent costs, leading to sharp falls in the quality of service.'" The situation in education has been compounded by enterprise divestiture of kindergartens. Private out-of-pocket expenditure on education and healthcare (both formal and informal) has expanded, raising questions of access and equity : Labor market developments 1.17 The fall in demand for labor in Kazakhstan has largely been accommodated by a contraction in real wages. According to official estimates, real wages have fallen by more than 50 percent since 1991 (see Table 1.4). There are numerous limitations with official wage and labor market data. The data relate mostly to employees of large and medium enterprises (excluding joint ventures), and cover the amounts that employees were entitled to, not the actual amounts received. 12 Arrears are common, thus actual income will in most cases depart from reported wages. The data do not reflect the impact of putting the workforce on unpaid 'administrative' leave or short hours as has been the common practice. They are thus indicative of broad trends rather than being entirely accurate From a poverty perspective, trends in the distribution of wages are equally important. Official data on earnings differentials across sectors suggest little change over time. This is related to the system of wage setting for public sector employees in Kazakhstan which fixes the level of the minimum wage, as well as the structure of wages (in relation to the minimum wage) by skill and industry. These industry and skill coefficients have changed little relative to each other over time. In 1996, average earnings in the highest paid sector (financial ) were 3.79 times those in the lowest (agriculture). These data provide little guidance on dispersion in individual earnings both because of their aggregative nature and because of the coverage problems alluded to earlier. According to FBS data, earnings inequality was growing in Kazakhstan during the 1980s. The decile earnings ratio (the ratio of earnings in the top most to the bottom 10 These include minimum unemployment benefit, family allowances and social pensions (paid to those who have no work history). Old-age pensions are based on past average earnings. 11 For trends in expenditure, and quality issues, see World Bank(1996), Republic of Kazakhstan Transition of the State. 12 An added problem is a break in the series in 1995 when attempts were made to include cooperatives and private businesses. Coverage amongst these groups is known to be weak. 6

17 most decile) increased from 3.14 in 1981 to 3.28 in Evidence from the KLSS suggests a continuation of this trend. According to the KLSS, the decile earnings ratio in July 1996 was Table 1.4: Labor market developments in Kazakhstan since independence Index of real average wage (1992=100) Official unemployment (%) Unemployment (ILO definitions) n.a. n.a. 4.6 n.a. 6.0 n.a. Labor force participation rate (ILO definitions): Male 80.3 n.a n.a n.a. Female Sources: Index of real wages, Official unemployment: Committee on Statistics and Analysis; Unemployment (ILO) 1994, LFPR 1992 and 1994: Falkingham et. al. (1997); Unemployment (ILO) 1996, LFPR 1996: KLSS. Figures for 1997 are preliminary 1.19 Measured job losses in Kazakhstan (around 15 percent between 1991 and 1996) have been smaller than the decline in GDP. The decline in recorded employment is heavily concentrated in the industrial sector where the number of wage employees fell by 36 percent, from 2.3 million in 1991 to 1.3 million in 1996, over three fourths of the decline in wage employment in the economy as a whole Despite the fall in output and employment, open unemployment in Kazakhstan is not particularly high by international standards (see Table 1.4). Starting from a negligible level, open unemployment (ILO definitions) rose to 4.6 percent towards the end of According to the KLSS, it was 6 percent a year and a half later. 16 This trend is similar to that suggested by official statistics. After having stayed below I percent till 1994, official unemployment rose to 2.1 percent in 1995, jumping thereafter to 4.1 percent in Official statistics tend to underestimate the true extent of unemployment because they are entirely based on Employment Services registrants, and exclude many people who would be regarded as unemployed by standard international definitions What factors explain the slow emergence of open unemployment in Kazakhstan? (i) First, the data do not reflect hidden or disguised unemployment, namely, those who are ostensibly 'at work' but are either on involuntary leave, or work short hours due to shortages of various kinds. Official estimates suggest that hidden unemployment in 1996 (4.5 percent) was as significant as recorded unemployment (4.1 percent). These data almost certainly overestimate hidden unemployment in sampled enterprises because they do not take account of whether those who are on leave or working short 13 Atkinson and Micklewright (1992), ibid. 14 Includes all cash income. 15 See Klugman and Scott in Falkingham et. al. (1997), Household Welfare in Central Asia. This estimate is based on the Kazakhstan Labor Force Survey (KLFS) conducted under World Bank auspices in November Due to various problems in the implementation of the KLFS, the sample may underrepresent younger and urban individuals, see Klugman and Scott (1997). To the extent that these individuals are more likely to be unemployed, the KLFS may understate total unemployment and comparing the two surveys may overstate the increase in unemployment between 1994 and

18 hours are available for additional or full time work i.e. they are not 'fully' employed elsewhere. Nevertheless they provide some guidance on the amount of 'down' time among those who are apparently employed. (ii) Second, for many workers who lose their jobs, inactivity is simply not an option. These people may be engaged in informal sector activity (kitchen-gardening, petty commerce) and may simply not be 'actively seeking work'. Although employment in the informal sector is hard to quantify, there is strong and visual evidence of its growth, especially in the oblast capitals. (iii) A third factor underlying the slow rise in open unemployment is movement out of the labor force. Labor force participation rates were maintained at high levels during Soviet times, encouraged, in the case of women, by the provision of subsidized childcare. Since 1992, many men and women have withdrawn from active participation in the labor force, and contrary to popular perception, the fall appears to be greater in the case of men than women (see Table 1.4). The fall in participation is partly related to the marked increase in early retirees, and to a lesser extent individuals claiming disability pensions, since 1992 (see Chapter 3). Non-participation may also be related to discouragement. 1996: Factors explaining labor market status 1.22 What factors explain unemployment and labor force participation in Kazakhstan? According to the KLSS, the most important personal characteristics appear to be education, gender, and age Educaticn: While there is not much difference in terms of overall years of schooling, the employed in Kazakhstan are more likely to have university education than vocational or technical education. Mean years of schooling are relatively high in Kazakhstan, over 11 years on average (see Table 1.5). Differences between the employed, the unemployed, and the inactive are not that significant in terms of number of years of schooling. However, the employed are much more likely than the unemployed or the inactive to have additional schooling (beyond grade 9). Over 80 percent of the employed had additional schooling compared to 63 percent of the unemployed and 57 percent of the inactive. The employed are twice as likely to have graduated from university and much less likely to have vocational and technical training than the unemployed. Table 1.5: Education and labor market status in Kazakhstan Employed Unemployed Inactive All Mean years of schooling % with more than basic schooling Degree if more than basic (%) Vocational Technical (no secondary) Technical w/ secondary Technical college University Post graduate Note: Basic schooling is grades 1-9. Source: KLSS

19 1.24 Gender: Women in Kazakhstan are more likely to withdraw from the labor force than men, and those who remain face a marginally greater probability of unemployment (see Table 1.4). The fall in labor force participation among women can be attributed to three factors. First, participation was unusually high during the Soviet period. Many women may have withdrawn in the face of a reduced imperative to work. Second, reductions in the provision of childcare may have had a negative impact on the ability of women to work, although the evidence on this is somewhat mixed." 7 Third, the departure of many ethnic Russians after independence would have lowered participation rates because of the traditionally higher participation among Russian women Unemployment rates among women (6.4 percent) are marginally higher than among men (5.7 percent). This may be partly related to the fact that women are less likely than men to have additional schooling (beyond grade 9) -- a factor that is associated with a higher probability of unemployment. This is reflected in the greater concentration of women in semi-skilled jobs especially in social services -- canteens, health posts and clinics, kindergartens, etc.. Many of these facilities have been subject to closures and the predominantly female work force laid off. Despite being less likely to have additional schooling, women in Kazakhstan have moreyears of schooling on average than men. According to the KLSS, women have 11.6 years of schooling compared to 11.4 years for men. This is due to the fact that women who continue their education in Kazakhstan are more likely than men to obtain university degrees Age: Age plays a significant role in explaining unemployment in Kazakhstan. The young are particularly vulnerable to unemployment (see Table 1.6). This is a likely reflection of the sharp contraction in economic opportunities facing new entrants into the labor force. Table 1.6 Age and labor market status in Kazakhstan (ILO definitions) Age Age Age Age Age Age Unemployment (%) Labor force participation rate (%) Note: Labor force participation is the sum of all working and unemployed as a share of working age population (16-59). The unemployed are all those not working but seeking work. Source: KLSS 1996 Conclusions 1.27 This chapter reviewed the changes in the economy and the labor markets since independence with the view to analyzing the main implications for living standards in Kazakhstan. Several conclusions emerge: * There has been a sharp fall in output and real wages since independence. High rates of inflation in the early years added to the general decline in living standards by destroying the cushion of savings. In addition to general inflation, movements in relative prices have taken their toll. 17 See, for example, Klugman et. al. in Falkingham et. al. (1997), ibid.. 9

20 * Delays in the payment of wages and pensions also constrain household consumption. * Public expenditures have fallen sharply in real terms, affecting both the degree of social security provided by the state and the quality of public services. * Although it remains low, unemployment is rising. Higher risks of unemployment are faced by the young, those with low levels of education or mismatched skills (vocational and technical education), and, to limited extent, women. All of these factors would be expected to lead to an increase in poverty. This is explored in greater detail in the next chapter. 10

21 Introduction 2. THE PROFILE OF THE POOR IN KAZAKHSTAN 2.1 This chapter draws a profile of the poor in Kazakhstan. The huge changes since independence have affected not just the general standard of living, but have created new groups of people who are vulnerable to poverty. This chapter investigates which groups have suffered the most from the prevailing economic circumstances and examines the main correlates of poverty. 2.2 While drawing on other source material, it largely relies on the results of the KLSS conducted in July Summer and early autumn are generally the best times of the year as own production of food - - the most important item of consumption -- is generally high. The quantitative estimates presented in this chapter may therefore understate overall poverty as well as understate rural-urban differentials in poverty as rural households are more likely to grow their own food. The estimates also probably overstate the importance of own-produced food in total food consumption. Box 2.1: Advantages of the KLSS The KLSS belongs to the family of nationally representative, multi-topic surveys conducted with World Bank assistance in over fifty countries since There are three distinct features of the KLSS: (i) It is nationally representative. Thus it permits conclusions about the population as a whole e.g. what proportion are poor, what proportion have access to public services etc.. The KLSS uses a probability sample of households in contrast to the annual Family Budget Survey (FBS) which samples on a quota basis. (ii) The survey instrument covers sources of consumption in detail. It can thus be used to construct a more complete picture of consumption than would otherwise be the case, including consumption derived from self-production and gifts. (iii) It is an integrated, multi-purpose survey. In addition to the measurement of variables such as consumption, unemployment, or labor force participation, it can be used to study inter-linkages among a range topics. This report, for example, examines the relationship between education and employment (Table 1.5), and gender and poverty (Table 2.9). The data may be used to study the impact of actual or simulated government policies as in Chapter 3 (Tables ) where the targeting effectiveness of public transfers is estimated. Measuring poverty in Kazakhstan 2.3 In analyzing poverty, the chapter focuses on monetary indicators of household welfare -- in particular, the money value of household consumption. Poverty is a multi-dimensional phenomenon which encompasses non-monetary dimensions such as access to public services. The declining quality of public services and the emergence of differential access are issues of tremendous public concern in Kazakhstan.' By focusing on the money value of consumption, this report does not mean to suggest that I See International Foundation for Election Systems (1996), Public Opinion in Kazakhstan. Over 86 percent of adults polled in November-December 1996 stated they were dissatisfied with the quality of healthcare. Sixty eight percent expressed dissatisfaction with the system of education. 11

22 non-monetary aspects are unimportant. However, there is good reason to believe that the non-monetary aspects of poverty reinforce monetary inequities. While deteriorating public services affect both rich and poor, the rich are in a better position to insulate themselves against the worst effects. The monetary value of consumption can therefore be taken as a proxy for other aspects of well-being. 2.4 To measure household welfare, the results of the KLSS were used to construct an estimate of household consumption including food and non-food goods and services. Food consumption includes food produced and consumed by the household as well as food received as a gift or as payment in kind for services rendered. Non-food goods include utilities, transport, clothing, cleaning, leisure, healthcare, and education. An imputed value for the services of durables was also included, but not imputed rent for owner-occupied dwellings. The consumption estimate was adjusted to take account of regional differences in prices. Full details on the data and methodology employed are provided in Appendix In order to construct poverty rates, this study makes use of the Govemment's "subsistence minimum" (Prozhitochnyi Minimum) as a poverty line -- see Box 2.2. The analysis of poverty inevitably entails a number of methodological choices of which the poverty line is particularly controversial. There is currently no official poverty line in Kazakhstan, and more than one informal poverty line in use, including various social benefit eligibility thresholds. In choosing among several candidates this study utilizes the average subsistence minimum (SM) which stood at tenge per capita in July The advantages of this measure are that it is akin to an absolute poverty line, is based on consumption not income, and is updated monthly. The SM is approximately equal to US $70 per capita per month at purchasing power parity (and US $40 at the market rate of exchange). While this would make it somewhat more generous than poverty lines used elsewhere, it does have the advantage of local ownership As the estimates may be sensitive to the chosen poverty line, this study examines how the proportion of the population in poverty varies with the poverty line. There is often a tendency to focus on the number of poor yielded by a poverty line, whereas for policy purposes what is equally relevant is the identifying characteristics or determinants of poverty. These are typically unaffected for a range of possible lines. To reinforce this message when investigating the characteristics of the poor, this study will look more broadly at the group of less well-off households, not just at those below the chosen poverty line. 2.7 The poverty line as defined in this report is intended as a policy tool not as social benefit eligibility threshold. There is a tendency, especially within the FSU, to think of the poverty line as a threshold below which households become eligible for social assistance and other benefits. In fact, the Govemment has resisted finalizing a poverty line for precisely this reason. However, a poverty line and a social benefit eligibility threshold are two different concepts which should be kept separate. In less welloff countries, a poverty line that is high enough to measure poverty in a meaningful way may be too high for everyone who is below the poverty line to receive social assistance. By the same token, a social benefit eligibility threshold that is low enough to be affordable may understate and thereby minimize the seriousness of poverty. Another danger in conflating the two concepts is the tendency to think of poverty reduction in terms of social transfers alone rather than in terms of a wider set of policies including growth, access to income-generating assets, and human capital investments. 2 A common yardstick employed in international comparisons of poverty is US $1 per capita per day. For comparisons among middle-income countries with richer populations, this is often scaled up to US $2 per day. The SM is more generous than the latter. 12

23 Box 2.2: The Subsistence Minimum (SM) in Kazakhstan The Subsistence Minimum (Prozhitochnyi Minimum) has a long history. It was developed in Soviet times to provide a measure of what would be minimally acceptable level of consumption under a socialist system. Although it has undergone a number of revisions, in methodological terms it remains essentially the same. The starting point for the determination of the SM is a food basket which incorporates nutritional guidelines as developed by the National Institute of Nutrition. These have become less generous over time (see below). The basket varies by region (five) and demographic group (nine). There is an average basket for the nation as a whole. The National Statistical Agency is charged with valuing these baskets on a monthly basis at local (i.e. regional) prices. Price data collected from both rural and urban areas are used for this exercise. Changes in the composition of the average subsistence minimum food basket Per capita kilos per year Meat Fish Milk Eggs* Sugar Fats Potatoes Veg Fruit Bread & Diary 1988/ *: Numbers not kilos To reflect the cost of non-food goods and services an allowance is added to the value of the food basket. Currently the weight of non-food goods and services in the average basket is 30 percent. The SM is similar in concept to an absolute poverty line. Unlike the old Soviet measure which was infrequently adjusted for inflation, the SM is updated monthly. The weights for food and non-food goods is in keeping with the shares of these goods in household consumption as revealed by the KLSS. However, as the prices of utilities are raised further it is likely that the weight of non-food goods will need to be raised. 2.8 There is also the tendency to relate the poverty line to the minimum or average wage. This is problematic for two reasons. First, it ignores household size. A new labor force entrant who earns a wage that is less than poverty line may not be poor if he or she lives in a household where there are other wage earners. By the same token, a family can be poor even if the main wage earner earns more than the poverty line income if there are numerous household members dependent on the one wage. Second, it confuses consumption with income. Formal sector wages in transition economies tends to be a fraction of household consumption. Households with low wage incomes may nevertheless consume at levels that place them well above the poverty line when other sources of consumption, such as earnings from a second job or household production, are taken into account. In Kazakhstan, according to the KLSS, wages are 56 percent of cash incomes and 26 percent of consumption (see Table below). 13

24 Poverty in Kazakhstan in What is the magnitude of poverty in Table 2.1: Poverty in Kazakhstan in Kazakhstan? Using the SM as the poverty line, 34.6 percent of the population lives in Headcount Poverty P2 poverty (see Table 2.1.) While comparisons ratio, H gap, PG are invariably clouded by the comparability of poverty lines, for indicative purposes it is Kazakhstan noted that this figure is over twice the preindependence level of poverty. It is lower than Source: KLSS 1996 the 1995 figure for Russia (41.1 percent) but higher than the estimate for the Ukraine (27.8 percent) Is poverty deep or shallow? Poverty is considered deep if average consumption of the poor falls well below the poverty line. In Kazakhstan, average consumption of the poor is nearly a third below the poverty line. 4 For comparison, the average shortfall of consumption from the poverty line in Russia is 42.6 percent, Ukraine 32.5 percent, Hungary (1993) 11.7 percent, and Poland (1993) 13.2 percent. While poverty is not as deep as in Russia, neither is it as shallow as observed in certain Eastern European countries Is poverty higher in rural areas? Table 2.2: Urban versus rural poverty in According to the KLSS, poverty is higher in Kazakhstan rural than in urban areas, although the differences are smaller than might be expected Head count Poverty P 2 on the basis of pre-1991 estimates. Urban ratio, H gap, PG areas include all large cities, rural areas all rural districts and rural settlements (towns). Urban Thirty percent of the urban population is poor Rural while the figure for rural areas is 39 percent source KLSS 1996 (see Table 2.2). Rural poverty is somewhat deeper than urban poverty: poor urban residents consume 30 percent less than the poverty line on average, while the equivalent figure for rural residents is 35 percent. These findings are consistent with poverty assessments elsewhere in the FSU which have typically found higher poverty and lower standards of living in rural areas. 5 Information on demographic and health indicators reinforces this picture of greater rural deprivation. Rural areas in Kazakhstan show higher infant- and child- mortality relative to urban areas. The incidence of stunting (low height for age) and wasting (low weight for height), and anemia in children, is also higher. 6 Reflecting higher poverty rates, rural areas make a higher 3 Three measures of poverty are used. The head count ratio, H, is the proportion of individuals below the poverty line. The poverty gap, PG, is a measure of the depth of poverty. It measures average consumption shortfall in the population (the non-poor have zero shortfall) as a proportion of the poverty line. P 2 is a measure of the severity of poverty. It is calculated in same way as the PG but gives more weighto poorer households. For further details, see Appendix 1. 4 Average consumption shortfall among the poor is given by (PG/H)xIOO=(I )xlOO= This picture of relative rural deprivation is somewhat weakened when we allow for economies of scale in consumption -- see appendix 4 for details. 6 See National Institute of Nutrition, Kazakhstan (1996), Kazakhstan Demographic and Health Survey

25 contribution to overall poverty. Over 57 percent of the poor live in rural areas compared to 42 percent for urban areas These aggregates hide important differences in the quality of poverty between rural and urban areas. The poor in rural areas have bigger families, produce more of their own food and have more meatrich diets. They have poorer housing and public services. The average poor rural household has about 5 members whereas the average poor urban household has 3.7. Thus, poverty is more a matter of large household size in rural than in urban areas. Cash incomes are lower in rural areas, however, the poor are have greater access to cultivable land and are more able to fall back on self-production to meet food needs. There are significant differences in the quality of the diet. The rural poor consume a larger proportion of meat, fish, dairy products and eggs while the urban poor are far more reliant on bread and cereal products. Housing quality is significantly lower in rural areas. 7 As a result, the urban poor are generally better housed than the rural poor. The quality of public services is also generally poorer in rural areas. Medical facilities are at greater distances. According to the KLSS, rural residents spend nearly five times as much as urban residents on health-care related transport While poverty in urban areas may be lower than in rural areas, one group of urban dwellers -- residents of one-company towns -- are generally believed to be poorer than most. One company towns were dependent on one or a few enterprises during Soviet times which have since closed or are closed for all practical purposes. There are 57 such towns in Kazakhstan. Local unemployment rates are extremely high, and public services (formerly financed by the enterprises) vastly under-funded. Public utilities face particular problems due to the lack of effective demand. Tekali, a mining town of around 27,000 people in Taldykorgan, provides a good example. Official unemployment is 20 percent. Another 40 percent of the labor force is subject to short hours or administrative leave. The sub-sample of households from onecompany towns in the KLSS is too small to be representative. This report was therefore not able to examine the poverty rates in these towns separately Are there regional differences in poverty rates? There are substantial differences in poverty rates across regions in Kazakhstan Poverty incidence is the highest in the South of the country. Over two out of three people live in poverty in the South compared to a national average of around one out of three. Poverty is also significantly deeper in the South than elsewhere (see Table 2.3). These findings reflect the lower resource endowment of the region, limited employment opportunities, and the disproportionate impact of fiscal contraction. 9 Fifty-five percent of the population in the South lives in rural areas compared to a forty-five percent in the nation as a whole. In contrast to the South, one out of eleven people live in poverty in the North. Not only is poverty more limited in the North, it is less deep and far less severe than in other parts of the country. Lower poverty may be related to greater opportunities for cross-border trade with Russia in this region In terms of contribution to national poverty, over 40 percent of the poor reside in the South of the country (see figure below). The next largest contribution is from the East. Although the East has a lower 7 Per capita living space is lower in rural than in urban areas. Nearly 59 percent of urban households have in-house toilets compared 5.6 percent of rural households. Access to piped drinking water is 78 percent and 17 percent, and to central heating, 70.4 percent and 10.2 percent, in urban and rural areas respectively. See National Statistical Agency (1996), Living Standards Measurement Survey Kazakhstan 1996 Final Report. 8 In July 1996, average expenditures in rural and urban areas for health-care related transport were 344 T and 88 T respectively. 9 For further discussion, see for example World Bank (1997), Kzyl-Orda: Regional Development Priorities. 15

26 poverty rate than the West, a larger proportion of total population resides in the East. 10 Together the South and the East make up nearly two-thirds of all the poor in the country although less than half the population live there. The main implication is that reducing poverty in the South and East of the country will be key to reducing poverty nation-wide. Table 2.3: Regional dimensions of poverty in Kazakhstan Regional Contribution to National Poverty in Kazakhstan Region Head count Poverty ratio, H gap, PG P 2 North 5% centre 14% North Center South 42% East West West % South North: Kostanai, Kokshetau, Paviodar, and North Kazakhstan East oblasts; Center: Zhezkazgan, Karaganda, Akmola and Torgai 24% oblasts; West: Manghystau, Atyrau, Aktyubinsk, West Kazakhstan; East: Semi Palatinsk, East Kazakhstan, Taldykorgan and Almaty oblasts and Almaty city; South: Source: KLSS 1996 Kzyl-Orda, South Kazakhstan and Zharnbyl oblasts. In April 1997, the merger of oblasts reduced the total number from nineteen to fourteen How sensitive are the poverty rates to Table 2.4: The head count ratio in Kazakhstan the poverty line? Since poverty is relatively under alternative poverty lines shallow, we can expect small increases or decreases in the poverty line to have a Head count ratio, H magnified impact on the proportion of population in poverty. The effect on the Subsistence Minimum 34.6 poverty rate of changes in the poverty line are 5 percent higher 37.5 detailed in Table 2.4. A five percent increase in the poverty line raises the head count ratio 5 percent lower 32.1 to 37.5 percent (an 8 percent increase), while a 10 percent higher 40.7 ten percent increase raises it 40.7 percent (an 10 percent lower percent increase). Lowering the poverty line by 5 percent reduces the proportion of the Source: KLSS 1996 population in poverty to 32.1 (an 8 percent decrease), while lowering it by 1O percent reduces it to 29.6 percent (a 17 percent decrease) Variations in the poverty line do not affect the broad picture of urban-rural, and regional, differences in poverty discussed above. Although there are differences in detail, the broad thrust of the 10 The shares of the regions in total population are as follows: North 20 percent, Center 19 percent, West 13.8 percent, East 26.8 percent and South 20.4 percent. 16

27 findings is maintained. 11 The South and the West have the highest poverty rates irrespective of where the poverty line is drawn, while the North has the lowest. The ranking of the Center versus the East is less robust, depending on the precise location of the poverty line. Rural poverty is generally higher than urban poverty except at high poverty lines (roughly double the SM). This reversal of rankings is a symptom of the greater inequality in the distribution of consumption in rural areas compared to urban areas (see below). Box 2.3: Poverty estimates based on the FBS The poverty rate estimated using the KLSS is significantly lower than estimates based on the FBS. The 1996 Human Development Report, for example, applies the subsistence minimum to the size distribution of income derived from the FBS. The report estimates that 64 percent of the population was living in poverty in August 1995, including 84 percent of the urban population and 44 percent of the rural population. If the same methodology were applied to size distribution of income in 1996 as estimated by the FBS, the poverty estimate -- around 80 percent of the population -- is even higher. What explains the difference in estimates? First, as indicated above, the surveys are not comparable. Second, and more significantly, the measurement of income through surveys is generally quite difficult owing to a number of problems. Income is often under-reported, especially where respondents have strong incentives to understate for fear of tax and other implications. Perhaps equally important is that fact that for households involved in agriculture or in running their own business, personal and business incomes and expenditures are hopelessly intertwined. Such respondents often have no need for an income concept, and their responses may be inaccurate. In the United States, for example, non-farm self-employment income is over 20 percent lower when estimated through surveys (see Deaton (1997), The Analysis of Household Surveys). In the United Kingdom, income from self employment is 25 percent lower and investment income (where there may be a tendency for concealment) 50 percent lower than that obtained from national accounts (see Atkinson and Micklewright (1992)). More generally, surveys that attempt to collect both kinds of information generally obtain estimates of income which are lower than estimates of consumption. To the extent that the FBS, like other surveys, understates household income, it is likely to result in an overstatement of absolute poverty. Difficulties in measuring income is one reason why, where possible, analysts prefer to use consumption as a measure of household standard of living. There are conceptual grounds for preferring consumption to income. Income tends to vary over time due to seasonal and other factors. Households tend to smooth their consumption through saving and borrowing. Thus current consumption typically provides a better indicator than current income of the standard of living. Inequality 2.18 Inequality in Kazakhstan appears to have increased since the break-up of the FSU. Available estimates from the late 1980s suggest that the Gini for income was around Although there are significant measurement problems, the Gini for consumption is likely to have beenlower, as consumption tends to be more equally distributed than income. According to the KLSS, the Gini coefficient for consumption is Although inequality may have increased in the post-soviet period, consumption in Kazakhstan appears to be more equally distributed than in neighboring countries. Recent estimates from Russia and the Kyrgyz Republic place the Gini for consumption in both countries at around Within Kazakhstan, consumption is more equally distributed in urban (0.33) than in rural areas (0.38). Among the regions, inequality is highest in the West (0.37) and lowest in the North (0.28). 11 See Appendix 2 for details. 12 See Atkinson and Micklewright (1992), Economic Transformation in Eastern Europe and the Distribution of Income. 17

28 A profile of the poor in Kazakhstan 2.19 This section examines the individual and household characteristics that tend to be correlated with poverty. This is useful both for understanding the causes of poverty and for establishing characteristics for targeting the poor. In place of focusing on those below the poverty line, we will examine the issue more broadly, comparing the characteristics of individuals in different consumption quintiles.' 3 If a particular quality, such as being elderly, or living alone, or living in a household headed by a female, is strongly associated with being poor, we would expect to see a greater proportion of such individuals in the lower than the upper quintiles. An equal distribution across quintiles suggests no correlation between a particular characteristic and poverty The main highlights are: poverty in Kazakhstan is associated with large households and the presence of many children or many dependents (whether young or old). Pensioners are also at risk of poverty. The poor are more likely to be unemployed, or to not participate in the labor force, than the nonpoor. Education beyond basic schooling decreases the likelihood of being poor. Gender (in particular, female-headship) does not appear to be significantly associated with poverty. Poverty and household composition 2.21 Looking at household size, we see that poor households in Kazakhstan are larger. Households in the bottom quintile have an average of 4.5 members while those in the top most quintile have slightly less than 3 members (see Table 2.5). Most of the difference is driven by higher average number of children in poor households. In the bottom quintile, households have on average one additional child compared to households in the top quintile. Table 2.5: Household composition and poverty in Kazakhstan Consumption quintile Bottom Second Third Fourth Top Total Average household size Average no. of children Average no. of elderly Dependency ratio Note: Dependency ratio is the sum of all persons below 16 years or above 59 years divided by those of prime age (16-59 years) Source: KLSS Standard dependency ratios are lower for the well-off in Kazakhstan. The dependency ratio indicates the number of household members who depend on each prime age adult (16-59). Among households in the top quintile, on average, each prime age adult supports 0.8 dependent persons; in the lowest quintile each such adult supports slightly more than one person (see Table 2.5) Focusing on particular family groups who may be vulnerable to poverty due to life-cycle factors, we see that single elderly people or elderly couples face lower than average risks of being in the bottom quintile (see Table 2.6). Three percent of households in the bottom quintile consist of only single elderly 13 Since 34.6 percent of the population is poor using the subsistence minimum as the poverty line, the bottom two quintiles (the bottom 40 percent) are roughly equivalent to the poor as defined in the earlier section of this chapter. 18

29 people, compared to 4.9 percent in the population as a whole. Likewise, 6 percent of bottom-quintile households consist of only elderly residents, compared to 9 percent in the population as a whole. However, households with only elderly residents are over-represented in the second quintile. This suggests that while elderly households may face lower risks of extreme poverty, they are more vulnerable to living at or close to subsistence levels. Table 2.6: Poverty incidence among selected demographic groups (percent of households in quintile and in total) % of households which are: Consumption quintile Bottom Second Third Fourth Top Total One person, elderly* Only elderly* One adult w/ dep. children W/ children < 6 years *: An elderly person is anyone 60 years and over Source: KLSS 2.24 Single parent households are less susceptible to low levels of consumption than is commonly believed. Only 3.8 percent of households in the bottom quintile belong to this group compared to 4.9 percent of households in the sample as a whole. This may be due to a variety of reasons including lower dependency ratios (fewer children than average), greater access to social assistance, and a greater tendency among the more educated, higher wage workers to seek separation or divorce. Further investigation is needed to determine the importance of different factors Families with young children are however disproportionately represented in the lower quintiles. Over 40 percent of bottom quintile households have small children whereas they form less than a quarter of all households The analysis in this section does not take account of differences in needs or economies of scale in consumption -- the use of the per capita consumption as the measure of household well-being assumes in effect that all individuals in the household have equal needs and there are no economies of scale. Allowing for greater economies of scale in consumption weakens some of the observed correlation between poverty and household size and overturns conclusions about elderly households and singleparent households which tend to be small and appear more poor as we allow for a greater scale economies (see Appendix 4 for details). However, a finding that is robust to the introduction of scale economies is that households with many children, and high dependency ratios (whether the dependents are young or old) face a greater risk ofpoverty Although households with young children are over-represented among the poor, the extent of malnourishment in Kazakhstan appears to be limited. According to the Kazakhstan Demographic and Health Survey conducted in 1995, 3 percent of all children under three years were wasted (i.e. had weight for height below international norms). Moderate or severe stunting (low height for age) was found in 16 percent of children in the same age group. Although high, the figure for stunting is lower than in Russia. 14 But, more to the point, the prevalence of both wasting and stunting in children in Kazakhstan 14 See World Bank (1995), Poverty in Russia, An Assessment. 19

30 was found to be lower than what might be expected in a normally nourished population. Micro studies focused on poorer regions of the country suggest that even the less well off populations have escaped from the emergence of serious malnutrition. 15 Box 2.4: Do old-age pensioners face greater risks of poverty? Pensioners in Kazakhstan are not an easy group to classify in terms of categories used in this chapter. Although there is a significant overlap between the elderly and old-age pensioners, conclusions about the elderly cannot automatically be applied to old-age pensioners. Over 32 percent of old-age pensioners in the KLSS are below 60 and would fall outside the elderly category used in this chapter. The legal retirement age in Kazakhstan was 60.5 for men and 55.5 for women in July This partly explains why a large proportion of pensioners are below 60. The pension system also grants fairly generous benefits to early retirees which encourage people to take up their benefits before the legal retirement age. Many pensioners both work and draw pensions. As a group, old-age pensioners include people younger than 60 years, many of whom are working. We would, therefore, expect pensioners to have a better profile than the elderly. However, they appear quite similar to the elderly (see table below). While they face a lower risk of extreme (bottom quintile) poverty, they are more likely to be found in the second quintile than in the population as a whole. Thus, pensioners face less risk of extreme deprivation but are more vulnerable to subsistence level existence. Employment greatly reduces the risk of poverty among pensioners. Less than I percent and 2 percent of households in the bottom two quintiles respectively have working pensioners compared to nearly 5 percent in the population as a whole. Old-age pensioners and poverty in Kazakhstan ( percent of households in quintile and total) Percent of households with: Bottom Second Third Fourth Top Total Non-working pensioner Working pensioner Looking at poverty rates among pensioners we see that 32.1 percent of them live in poverty compared to 34.6 for the population as a whole. This would suggest that pensioners face lower risks of poverty than the population as a whole. However, this finding is not very robust. If we raise the poverty line by a little under 10 percent, we find that poverty rates among pensioners are about the same as in the population as a whole. Thus, when the poverty rate is 40 percent for the sample as a whole, it is 40 percent for pensioners. And if we allow for small economies of scale in consumption, poverty rates amnong pensioners become higher than average (see Appendix 4). Thus, while pensioners are less likely to be found in the bottom quintile, they appear to face the same, or greater, risks of poverty than the population as a whole. From a poverty perspective, resources are better expended by targeting poor pensioners than pensioners as a whole as not all pensioners are poor. Poverty and labor force status 2.28 In Kazakhstan, as elsewhere in the world, unemployment is strongly associated with poverty. As seen in Table 2.7 individuals in the higher quintiles are more likely to be economically active (working or seeking work) than those in lower quintiles. But what truly separates the poorer individuals from the others is the level of unemployment. The poorest individuals suffer an unemployment rate of 9.4 percent while their counterparts in the top quintile of the population have an unemployment rate of only 2.6 percent. 15 A representative survey of three rayons in Kzyl-Orda in 1994 and 1995 found no evidence of wasting (0.5 percent), and limited evidence of stunting (15.0 percent) in the region. For further details see Ismail and Hill in Falkingham et. al. (1997), Household Welfare in Central Asia. 20

31 2.29 The importance of employment for living standards is best illustrated by looking at the number of household members being supported by each employed person. In contrast with the standard dependency ratio, dependence on employment (the number of household members being supported by each employed person) varies substantially by quintile. In the top quintile, each employed person supports 1.4 other household members, including children and unemployed or inactive adults. In the bottom quintile, each employed person supports 2.5 persons (see Table 2.7). Even if the average worker in the poorest households earned the same amount as his or her counterpart in the top quintile, consumption per capita would be significantly lower due to the greater dependency on that income. Table 2.7: Labor force status and poverty in Kazakhstan (ILO definitions) Consumption quintile Bottom Second Third Fourth Top Total Labor force participation rate Unemployment rate Dependency on employment Note: Dependency on employment is the number of household members being supported by each employed person. Source: KLSS While unemployment is strongly associated with poverty, the limited rise in open unemployment means that as a proportion of all working age individuals the number of unemployed is relatively small. In the bottom decile, 46.3 percent of working age individuals are employed, 4.8 percent are unemployed and 48.8 percent are inactive (neither employed nor actively seeking work). In the bottom two deciles, the figures for employed, unemployed and inactive are 48.3 percent, 4.7 percent and 47.0 percent respectively. Thus many of the poor are 'working poor' i.e. are at work but are poor either because they have too many dependents to support adequately, work in low-wage occupations, are subject to involuntary leave without pay or short hours, are paid irregularly, or any combination of these factors. Another large proportion of the poor are inactive, including women who stay at home and those who are too discouraged to look for work. Poverty and education 2.31 In accordance with international experience, education and poverty are inversely correlated in Kazakhstan. Individuals in the lowest quintile have less schooling on average than those in the top most quintile, though the difference (1.2 years) is relatively small (see Table 2.8). A stronger correlation appears to exist between poverty and level of schooling (basic or additional). Individuals in the top quintile are much more likely to have additional schooling (beyond grade 9): 79 percent of the individuals in the top quintile had additional schooling compared to 55 in the lowest quintile. University degrees are positively correlated with consumption -- a greater proportion of individuals in the top most quintile have university education. Vocational and technical degrees are negatively correlated with consumption i.e. those with such training are more likely to be found in the lower quintiles.' 6 This is perhaps a reflection of the collapse in the industrial employment in the post-1991 period, and the greater versatility of university education in the changed circumstances. 16 Technical college degrees appear to be an exception bearing no correlation with living standards. 21

32 Table 2.8: Education and poverty in Kazakhstan Consumption quintile Bottom Second Top Years of schooling % with more than basic schooling Degree if more than basic schooling (%): Vocational Technical, no secondary Technical w/ secondary Technical College University Post-graduate Note: Basic schooling is grades 1-9. Source: KLSS 1996 Poverty and gender 2.32 The gender of the household head is not correlated with poverty (see Table 2.9). Female headed households, who form a large proportion of all households in Kazakhstan (40 percent), are underrepresented among the poor. Thirty percent of households in the bottom quintile are female headed compared to 40 percent in the population as a whole. The lower vulnerability of female-headed households to poverty appears to be related to four factors. First, female-headed households tend to be smaller than male-headed households. Second, dependency on employment is lower especially in the lower quintiles i.e. each employed adult in a female headed household support fewer other members. Third, the household head tends to be older. Finally, female-headship may be acting as a proxy for region and location. In the southern part of the country (which is largely Kazakh) and in many rural areas, the head would typically be a man Although female-headed households face no greater risk of poverty, this is not to suggest that there are no gender inequalities in Kazakhstan. Women's work was traditionally associated with low pay in the FSU, as in other parts of the world. Much of this difference has persisted -- in Kazakhstan in 1996, official wages in the female-dominated sectors (education, health-care) were about half the official average wage. Women appear to face a marginally greater risk of unemployment than men, and possibly higher rates of long-term unemployment. However, inequality in access to education and healthcare and intra-household inequality in consumption, are not believed to be significant. 22

33 Table 2.9: Gender and poverty in Kazakhstan Consumption quintile Bottom Second Top % of female headed households Poverty, health and disability Female headed households: Household size Dependency on employment Age of household head Years of schooling of head Male headed households: Household size Dependency on employment Age of household head Years of schooling of head Note: Dependency on employment is the number of household members supported by each employed person. Source: KLSS Disability (as measured by eligibility for a disability pension) is positively correlated with poverty. Disability however affects a small proportion of the population: according to the KLSS, fewer than 5 percent of households have disabled members More generally speaking, poverty and health appear to be positively correlated in Kazakhstan. Individuals in the upper quintiles are less healthy in the sense that they report more sickness and injury than the poor. Nearly 24 percent in the top quintile reported sickness or injury in the month prior to the survey, compared to less than 13 percent of individuals in the bottom quintile." 7 This finding is typical of self-reported health data. Poor health and injury appear to be luxuries that the well-off can more easily afford. Of those who reported poor health status, the poor were marginally more likely to seek medical attention than the rich -- nearly 57 percent in the bottom quintile sought medical attention compared to 51 percent in the top quintile -- perhaps an indication of the relative acuteness of their health ailments. Poverty and other characteristics 2.36 Migrants and refugees are a very small group in numerical terms, but may well be very vulnerable to poverty. Little can be said about them on the basis of the KLSS or any other nationally representative survey. The transition has been accompanied by a sharp increase in migration in Kazakhstan. In addition to substantial emigration, Kazakhstan has received immigrants, including ethnic Kazakhs from other CIS countries and Mongolia, and refugees escaping conflict in other parts of the CIS, 17 The figures include both acute and chronic ailments as well as accidents. The sample average is 19 percent. 23

34 and Afghanistan. Over 150,000 Kazakhs have arrived in Kazakhstan since The number of refugees is more difficult to estimate. Ethnic Kazakhs have tended to settle in the southern half of the country while the south eastern oblasts have received refugees, from the Afghan and Tajik conflicts. In thinking about the living standards of the two groups, it is important to distinguish between ethnic Kazakh immigrants, who are entitled to state assistance, and refugees, who are not typically covered by the social welfare system. Within each group there are likely to be those who are more well-placed to cope with changed circumstances. Expenditure patterns 2.37 Turning to consumption patterns, we see that food dominates consumption in Kazakhstan (see Table 2. 10) The differences between quintiles are surprisingly insignificant. Table 2.10: Structure of consumption and poverty in Kazakhstan Consumption quintile Source of expenditure Bottom Second Third Fourth Top Total Food Food purchased Food self produced Foodeaten out Food as gift Utilities Durables Transport Cleaning Clothing Leisure Health (chronic) Education Total Total Avg. per capita consumption (tenge) 1,524 2,610 3,684 5,245 9,562 4,545 Note: Owing to data limitations, total health and education expenditures are underestimated. See Appendix 1. Source: KLSS Self-production of food is a significant source of food consumption in Kazakhstan, and is more important to the rich than the poor. Over a third of food consumed in Kazakhstan on average comes from own-production. The rich have greater access to own-produced food -- self-production is over twice as high in the top quintile than in the bottom. The poor in Kazakhstan are more reliant on food purchases with over two-thirds of food consumed in the bottom quintile coming through purchases." 18 Recall, however, likely rural-urban differences to this pattern. 24

35 The importance of food purchases to the poor has been noted in other countries and contexts. The greater share of self-production in the upper quintiles is however distinct from experience in other parts of the world though not in the FSU More is spent on the education of children (6-16) in the upper quintiles than the lower quintiles (see Table 2.10). However, the share of education in per capita consumption expenditure is low and the nominal sums involved are small. Primary and secondary education in Kazakhstan is wholly public and largely funded out of general budgetary resources. Textbook charges were introduced in 1996, with exemptions for children from poor households. 19 In addition, there are charges for special classes and some extra-curricular activities. The bulk of household expenditure on the education of children is found to be related to text-books. The proportions spent on fees and transport is relatively small. Textbook charges were incurred on over half the children in the sample during the month of the survey, including 41 percent of children in the bottom quintile and 68 percent of children in the top-most quintile The correlation of expenditures on healthcare with poverty status is less clear, although this may be due to data problems (see Table 2.10). Health expenditures includes expenditure on chronic illnesses and medical devices but not preventive or acute care. 20 However, as in the case of education, inequities have appeared in the system. Of all individuals who were prescribed medicines by their doctor, a greater proportion of individuals in the lowest quintile (71 percent) said they could not afford to buy them than in the top quintile (54 percent). 21 Individual expenditures associated with episodes of hospitalization -- notionally free under the system of health care in Kazakhstan but paid for 'informally' by the patient -- appear extremely high (over 4000 tenge per month) and may well be beyond the means of many individuals. Paying to see a doctor, also notionally free, appears to be fairly common and expensive (12 percent of individuals, who paid 644 tenge for the seeing the doctor on average in July 1996). Structure of cash income and informal sector activity 'Formal' incomes -- from the main occupation or from pensions and other transfer payments -- vary significantly by quintile. In general, wages from the primary or main occupation are the most important source of income accounting for around a half of all cash income in Kazakhstan. Households in the top quintile are slightly more dependent on this source percent of cash income in the top quintile is derived from the main wage compared to 45 percent in the bottom quintile (see Table 2.11). Pensions and other transfer payments are the second most important source of cash income. Unlike wage income however, pensions are generally more important in the lower quintiles. The same is true of income from other public transfers. The share of pensions in cash income falls from 16 percent in the bottom quintile to 9 percent in the top quintile, while the share of other transfers falls from 6 percent to 3 percent. 19 Exemptions are based on eligibility for the main social assistance program -- the means-tested child allowance. Households who receive this benefit are eligible for a 50 percent discount on the cost of new text-books. 20 See Appendix I for details. 21 All outpatient drugs are paid for by the patient in Kazakhstan. 22 Self-reported income data is generally less reliable than consumption data. However, the sum of cash income and imputed income from self-production does not appear that different from household consumption -- see Tables

36 Table 2.11 Structure of cash incomes in Kazakhstan Consumption quintile Bottom Second Third Fourth Top Total Primary wage Pensions Other public transfers Secondary/other occupation Own business Sale of food Private cash transfers Other TOTAL Avg. per capita cash income (tenge) 957 1,535 2,167 3,067 4,116 2,370 Avg. per capita income (including 1,371 2,412 3,229 5,156 9,473 4,743 imputed income from own-production) Note: Other public transfers includes all transfer payments excluding pensions. Private transfers are gifts and donations from private parties. Source: KLSS Reflecting the growth of informal sector, income from a second job and from own business activity contribute significantly to cash income in Kazakhstan. Trade is an important informal sector activity -- see Box 2.5. Income from informal activity is more important to the poor than the rich -- nearly 14 percent of cash income in the lowest quintile is derived from a 'second' job or own business activity compared to 12 percent in the top quintile. While the sale of food is a common income generating activity, its contribution to overall householdcash income is relatively small on average. As discussed earlier, own production is however a very important source of income in kind Private transfers including assistance from immediate and extended family, other households, charities, cultural and religious organizations, and NGOs are a significant source of cash income in Kazakhstan. Reflecting extended family networks, parents are the most important source of private cash transfers, then other relatives. In terms of overall magnitude, private transfers are smaller than pensions but are significantly larger than the sum of all other public transfers. In fact, private assistance is over twice as large as the sum of all government transfers excluding pensions. This reflects the erosion in the value of benefits and the impact of arrears. 23 Not all own businesses and trade are informal but a good proportion are. 26

37 Box 2.5: Trade in Almaty Market and street trade are important economic activities in Kazakhstan. In the city of Almaty, there are an estimated 20,000 people involved in trading activity, the majority of whom are women. The Asian Development Bank (ADB) conducted a survey of a hundred women engaged in trade in Almaty city in The results suggest that trade is not a part-time or marginal activity for these women but a full time occupation. Many of those surveyed used to work in 'regular' jobs but had either been dismissed or had decided to start their own business because of low or irregular -payment of salaries. A small proportion of those interviewed -- about 7 percent -- retained their forner jobs so as not to lose access to work-related benefits. Many of the interviewees had formerly worked as professionals. Nearly 30 percent had higher education, 25 percent had specialized technical education and 25 percent vocational training. Around a fifth had formerly been school teachers. The ADB survey revealed that trading is typically not linked to own-production. Over a half the interviewees were engaged in wholesale trade. Nearly a third were selling goods on commission. About 15 percent were selling produce which had been imported directly from abroad. Less than one-fifth were selling own-produced food. A large proportion of those surveyed relied on help from others to run their businesses. This was typically for transporting goods, book-keeping and purchasing supplies and goods. Both relatives and friends were involved in this support. Despite difficulties with obtaining credit, most of the interviewees considered their businesses to be successful. Monthly tumover was reported to be between 10,000 tenge and 100,000 tenge, with a median 15,500 tenge for street trade, 24,600 tenge for trade in food markets and up to 40,000 tenge in non-food markets. Interviewees declared an average net income of percent of turnover. This implies a monthly income of 4,800 tenge (US $70) from street trade, 7,400 tenge (US $109) in food markets, and 8,400 (US $ 123) in non-food markets. Income from trading thus compares favorably with average monthly wages of about 7,000 tenge (US $103). Coping mechanisms 2.44 The structure of consumption and income suggest different ways in which households in Kazakhstan have responded to the changed economic circumstances. The ability to cultivate one's own plot is clearly a vital coping strategy. Access to land is generally higher in rural areas. According to the KLSS, nearly 80 percent of rural households cultivate a private plot compared to around 50 percent of urban households. The poor in Kazakhstan are those who are not able to meet a large proportion of their food consumption needs from self-production and need to rely instead on purchases Other informal activities, such as holding a second job, or engaging in informal business activity, also offer alternative ways of making ends meet in Kazakhstan. Gifts and transfers between households function as an informal social safety net. In Kazakhstan, gifts of food appear, somewhat perversely, to be a larger share of consumption in the upper quintiles than in the lower quintiles (see Table 2.10). Gifts of money and other private cash transfers are also important, contributing around 15 percent of cash incomes on average. Conclusions 2.46 This chapter has several conclusions: * First, using the 'subsistence minimum' as poverty line, over nearly 35 percent of the population of Kazakhstan lived in poverty in July Given what we know about the pre-independence situation, this represents a significant increase in poverty. * Second, rural-urban differences in poverty in Kazakhstan are not as significant as regional differences. Over two in three people live in poverty in the southern oblasts of Kzyl-Orda, South 27

38 Kazakhstan and Zhambyl. Together the south and the east (Semi Palatinsk, Taldykorgan, East Kazakhstan and Almaty oblasts) account for over two out of three poor people in Kazakhstan. * Third, there is a strong association between family size and poverty. The transition has been very hard on households who have to support a large number of children or a large number of dependents (whether young or old). * Fourth, pensioners face higher risks of poverty than the population as a whole. However, they appear relatively well-protected from extreme (bottom quintile) poverty. * Fifth, poverty is strongly correlated with unemployment. Poor households are characterized by high dependency on employed members. * Sixth, workers with university degrees have generally fared better during the transition than those with vocational or technical degrees or low levels of schooling. * Seventh, own-production has emerged as a significant source of consumption during the transition accounting for over a third of all food consumption in Kazakhstan in July Being able to grow one's own food is an important coping strategy. * Eighth, informal labor market activity is an important source of cash income for poor households. * Ninth, there are significant private, inter-household transfers in Kazakhstan which serve as an informal social safety net. 28

39 3. THE ROLE OF THE SOCIAL SAFETY NET IN KAZAKHSTAN Introduction 3.1 This chapter discusses the role of the social safety net in maintaining living standards in Kazakhstan. It will attempt to answer three questions. How has the social safety net in Kazakhstan evolved since independence? What role does it play in protecting the income and consumption levels of poor households? To what degree are cash transfers targeted at poor households? 3.2 Kazakhstan's social safety net incorporates elements of both social assistance and social insurance.' Social assistance includes: (i) child allowances, of which the major program is the meanstested child allowance, (ii) means-tested housing allowances, and (iii) benefits in cash and kind aimed at groups believed to be poor or vulnerable. Social insurance includes: (i) pensions, (ii) unemployment compensation, and (iii) allowances related to events such as births and deaths and temporary incapacity to work including sickness and maternity. 2 The weak link between contributions and benefits implies that in practice social insurance has a large assistance element. Box 3.1 provides an overview of the system of social protection. Pensions are the most important program. In 1996, spending on all social protection programs in Kazakhstan was 6.6 percent of GDP of which the bulk (75 percent) was spent on pensions. During 1997, pensions rose to nearly 90 percent of all social protection expenditures as arrears were cleared. The second largest program is the means-tested child allowance. The housing allowance scheme is small but is expected to expand substantially in the near future : Evolution of the social safety net 3.3 At independence, Kazakhstan inherited a comprehensive system of social security which combined guaranteed employment with a generous system of pensions, sickness and maternity benefits and subsidies for food, fuel, transport and a range of consumer items. Normal pension ages were low by international standards and significant groups of the population were entitled to early retirement on full pensions. Social assistance relied on universal child allowances and a range of benefits targeted at specially identified vulnerable groups including single pensioners, veterans of war, victims of environmental disaster, nuclear testing, and political and social repression. 3.4 The inherited system was in need of reform to make it more consistent with the needs of a market economy. To its credit, the Government took a number of steps to make the social safety net both less generous and, in the case of general subsidies and cash social assistance, more targeted. Almost the first to go was the guarantee of employment. As unemployment emerged, the government established a system of employment services including cash benefits, training and job placement. The generalized subsidies were gradually removed starting with the explicit subsidy on food. In 1996, the price of utilities I In discussing the social safety net, we exclude general production and consumption subsidies. 2 A fourth component is health insurance which this report will not consider in detail. For a more detailed description of individual programs see Appendix 3. 29

40 began to be raised, although the subsidies to special groups -- the so-called privileges -- were retained. To cushion the impact on poor households, a housing allowance scheme was introduced with the aim of compensating households for utilities and maintenance expenditure above a certain proportion of household income for a pre-determined maximum floor space. In the same year the decision was taken to raise pension ages for men and women. Further pension reform is envisaged from 1998 onwards when the existing pay-as-you-go system will be gradually phased out in favor of a privately managed fully funded pension system with a minimum state guarantee. The main child allowance scheme was made more targeted by the introduction of a means-test in A. Social Assistance Box 3.1 Kazakhstan 1997: the system of social protection at a glance 1. Child allowances a. Means-tested child allowance (children < 18 years of age) b. Child allowances for servicemen on active duty c. Allowances for disabled children d. Allowances for children or adults with HIV e. Temporary allowances for children whose fathers evade alimony f. Allowances for unemployed mothers with 4 or more children < 7 years old. 2. Means-tested housing allowances 3. System of privileges in housing and municipal service fees, fuel, transportation, medical equipment etc. for socially vulnerable groups such as invalids, veterans of war and labor, low-income pensioners and others 4. Allowances and social services for the care of the elderly and disabled and other low income groups 5. Social pensions 6. Allowances for returning Kazakhs 7. Old age, invalidity and mental health institutions B. Pensions (Pension Fund) 1. Labor pensions (old age and length of service) 2. Disability pensions (Groups 1, 11 and III in decreasing order of physical disability) 3. Survivor pensions 4. Others (special services to the Republic, military) C. Unemployment (Employment Fund) 1. Unemployment benefit 2. Job search and other employment services D Other benefits (Social Insurance Fund) 1. Sick pay 2. Maternity benefits 3. Birth allowances 4. Funeral allowances 5. Sanitarium vouchers This definition does not include health benefits. 3.5 Perhaps more controversially, there was a disproportionate contraction in social safety net expenditures during Expenditure on all programs fell from around 11.2 percent to 4.6 percent of GDP during (by 60 percent) at a time when government revenues fell from 39 percent to 24 percent (40 percent). Pensions took a sharp hit falling from around 8.2 percent in 1992 to less than 4 30

41 percent in The real value of unemployment compensation was allowed to erode so much that many of the unemployed were simply not registering. A subsequent rise in the value of the minimum benefit towards the end of 1995, from one to three times the calculation base, was accompanied by a sharp increase in registered unemployment, particularly in rural areas. 3 Although social safety net expenditures as a share of GDP have since risen, the system continues to be bedeviled by arrears. This results in 'gaps' in coverage. At the end of March 1997, pension arrears stood at 26 billion tenge (equivalent to three months' payments) while arrears in the payment of child allowances were 4 billion tenge (equivalent to six months' payments). Ironically, payment arrears are more significant in the poorer oblasts where the reliance on the system of social security is per force higher. 4 As a prelude to its program of systemic pension reform, the government cleared the backlog of all overdue pensions during Taxes on labor which fund the main social insurance programs continue to be high. Up from 26 percent at the beginning of the decade, they rose to 38 percent in 1993, before falling to 32 percent. 5 As a part of the pension reform program, the 25.5 percent payroll tax that funds the pension system was reduced by 0.5 percent to 25 percent, and was divided into two parts: 15 percent will go to fund the staterun pension system, while 10 percent will be paid into privately managed individual retirement accounts. The 15 percent payroll tax is expected to be reduced in a phased manner and finally eliminated, although this is likely to happen over the medium to long run due to fiscal reasons. 3.7 Despite commendable efforts to trim the system, there continue to be a multiplicity of benefits. Child allowances are a case in point. In addition to the means-tested allowance, there are allowances for numerous other categories, including children of: (i) single mothers, (ii) fathers who fall behind in alimony payments, (iii) unemployed mothers with 4 or more children less than 7 years, and (iv) those in active military duty, and children with (v) disabilities, and (vi) HIV. The 1998 budget merged the allowance for single mothers with the main means-tested benefit. In addition, there are numerous privileges and related allowances awarded to special groups. 6 While the benefits may not be large, they are complicated to keep track off and potentially complex to administer : Trends in social safety net expenditure 3.8 Table 3.1 records trends in social safety net expenditure (excluding consumer subsidies) as a share of GDP since The aggregate picture reveals the sharp contraction between 1992 and 1994 referred to earlier, and the subsequent expansion. A similar picture emerges from an examination of the 3 The calculation base is a notional floor determining the level of a number of public transfers. Its value (in tenge): 1996: Ql: 320, Q2: 380, Q3: 460, Q4: 530; 1997: Ql: 550, Q2: 565, Q3: 585, Q4: 620; 1998: Ql: 630, Q2: 640, Q3: 650, Q4: For example, in the case of the means-test child allowances, the highest arrears were observed in the oblasts identified in Chapter 2 as being the most poor: South Kazakhstan (I billion tenge), Kzyl-Orda (600 million tenge) and Zhambyl (400 million tenge). 5 The contributions to the different programs or Funds is as follows: Pensions 25.5 percent, Health 3 percent, Social Insurance (which pays short-term benefits) 1.5 percent, and Employment 2 percent. 6 Subsidies and allowances are awarded to: invalids of war and their 'equivalent', war veterans and their 'equivalent', war widows, invalids (Groups I, II, III), children with disabilities, special pensioners, 'hero' mothers, low income pensioners, and civilian war workers. Goods and services covered are: housing and communal services, electricity, communication, sanatoria treatment, cars, petrol, wheelchairs, transportation, medicines, prosthetic and hearing devices, funeral expenses, food, recreation and personal services. Entitlements, both in terms of categories of goods and services and degree of subsidy (which can be as much as 50 percent), varies by group. 31

42 main programs with one exception -- labor market programs -- which have expanded continuously. Labor market programs however remain small in overall size. Table 3.1: Social safety net expenditures in Kazakhstan as a percent of GDP, Total social safety net expenditure of which: 1. Pensions Labor market programs (including unemployment benefits) 3. Social Insurance Fund benefits of which: Sickness Maternity Social Assistance of which: Means tested child allowances GDP (million tenge) 1,213,616* 29, ,484 1,019,722 1,425,000 1,742,000 *: million ruble Source: MLSP, World Bank staff estimates. Figures for 1997 are preliminary. 3.9 As seen in Table 3.1, the composition of social safety net expenditure between pensions, other social insurance, and social assistance has remained by and large unchanged. The exception is 1997 when, as a result of the payment of pension arrears, pensions rose from around three-quarters to nearly 90 percent of total social safety net expenditure With the exception of pensions, there have been changes in composition within the broad grouping of pensions, other social insurance, and social assistance (see Table 3.2). (i) Within labor market programs, expenditure on unemployment benefit has grown, from 5 percent in 1992 to 23 percent in 1995 and 57 percent in 1996, to become the largest single item of expenditure. Significantly, the share of administration (spending on the Employment Services) halved during from 36 percent to 18 percent of all expenditures. In 1993, over 43 percent of the Employment Fund was being spent on assistance to returning Kazakh nationals. This has since fallen to less than 3 percent. (ii) Of the benefits charged to the Social Insurance Fund, sickness and maternity have grown in importance, while sanatoria have fallen sharply. (iii) Within social assistance, the means-tested child allowance program has fallen in significance. This is largely explained by the growth of the housing allowance scheme which, after its introduction in May 1996, grew to 12 percent of all social assistance expenditure. Other child allowances -- awarded on categorical criteria -- which grew from 3 percent of all social assistance spending in 1992 to 10 percent in 1996, fell to 7 percent in Other forms of social assistance -- traditionally targeted at the elderly and the disabled living alone but also other 'low income' groups identified by the local authorities -- grew in importance till 1994 before dropping off considerably. 32

43 Table 3.2: Composition of social safety net expenditures in Kazakhstan Selected years, Pensions Old age n.a Disability 8 8 n.a. Loss of breadwinner 7 8 n.a. Social 2 2 n.a. 2. Labor market programs Unemployment benefit n.a. Administrative costs n.a. Assistance to migrants n.a. Training n.a. Public works n.a. 3. Social Insurance Fund benefits Sickness Maternity Sanatoria Social Assistance Means tested child allowances Other child allowances Local social assistance 0 0.4* 6 0 Housing allowances i The figure for 1994 is 17 Source: MLSP, World Bank staff estimates. Figures for 1997 are preliminary 3.11 What do these trends reflect in terms of number of beneficiaries and benefit levels? With a few exceptions -- notably, labor market programs where numbers have been rising -- the number of beneficiaries has fallen since independence, in some cases quite significantly. Real benefits, with the exception of the means-tested child allowance, have contracted sharply (see Table 3.3). (i) In the case of pensions, the number of pensioners declined from 2.83 million in 1993 to 2.67 million in In real terms, pensions declined by over two-thirds between 1993 and 1995 after which they have seen a modest increase. 7 Among pensioners, the number of old-age pensioners fell from 2.11 million to 2.00 million during the same period. However, the number of persons receiving length-of-service pensions - special pensions that are available after completing required years of service -- increased from 19,496 to 38,931. The number receiving disability pensions also rose from 261,000 to 263,

44 (ii) Like pensions, minimum unemployment benefit contracted sharply in real terms between 1993 and 1995 before being revised upwards. Minimum benefits are currently 1.27 times earlier benefit levels in real terms. (iii) The introduction of the means-test for the basic child allowance in 1994 reduced the number of program beneficiaries by over 30 percent. By 1997, the number of beneficiaries had been roughly halved. This enabled a rise in per beneficiary expenditure in real terms -- in 1997, real expenditure per beneficiary stood at nearly 2.5 times its 1993 level. Table 3.3: Program beneficiaries and average benefit levels in Kazakhstan, Selected programs 1. Pensions Number of beneficiaries (OOOs) 2,829 2,814 2,819 2,771 2,670 Index of real average benefits (1993=100) Labor market programs -- UB Registered unemployed (OOs) Index of real minimum benefits (1993=100) Social Insurance Fund benefits - Maternity Number of beneficiaries (OOOs) n.a. n.a. n.a. Index of average real benefit n.a. n.a. n.a. 4. Social Assistance - Means tested CAs Number of beneficiaries (OOOs) 2,417 1,674 1,795 1,469 1,232 Index of average real benefit Note: Beneficiaries refer to total program beneficiaries. Source: MLSP and World Bank Staff estimates. Figures for 1997 are preliminary The importance of cash transfers in Kazakhstan 3.12 The next few sections focus on the incidence and impact of social protection expenditure in Kazakhstan using the KLSS. 8 This section first considers who does and who does not receive transfers and the importance of transfers to maintaining consumption levels in Kazakhstan. The next section focuses on how well targeted transfers are to poor individuals and regions Thirty-seven percent of households in Kazakhstan receive at least one transfer (see Table 3.4). This proportion does not depend on whether household consumption falls below the subsistence minimum or not. Over sixty percent of poor households receive no cash benefits at all. Of the different transfers, pensions are the most common, received by over 28 percent of households. The other transfers s This study does not include housing allowances as this information was not collected by the survey. However, the uptake of housing allowances during the survey period would have been small both because the scheme had only just got underway at the time and because the survey was conducteduring a summer month. 34

45 are relatively uncommon. Child allowances are received by 4 percent of households, including 6 percent of the poor and 3 percent of the non-poor, while unemployment benefits are received by around 1.5 percent of households (2 percent of the poor and I per cent of the non-poor). Average benefits are modest compared to the subsistence minimum, especially for unemployment benefit. Table 3.4: Coverage and average value of cash transfers in Kazakhstan, 1996 Poor Non-poor Average benefit Proportion of Receipt Proportion of Receipt per household households in rate I households in rate I (tenge per month) receipt of benefit receipt of benefit Pensions l Old age ,619 Disability ,586 Loss of breadwinner ,683 Unemployment benefit ,382 SIF - Sick pay ,353 SIF - Maternity benefit ,063 Child allowances Means-tested ,603 Other I n.a. 2 n.a. 776 Other social assistance 4 n.a. 4 n.a. 2,345 ALL BENEFITS *: Receipt rate is proportion of eligible households who received benefits during the month of the survey. Eligibility is selfreported with the exception of mean-tested child allowances. Notes: Subsistence minimum in the survey month was 2861 tenge per capita. Source: KLSS A large proportion of households appear not to receive the transfers they are entitled to. Table 3.4 presents receipt rates -- the ratio of receiving to eligible households. Eligibility for the means-tested child allowance was determined by applying the means test to the income per capita of households with children. 9 For other transfers, eligibility is self-reported. Receipt rates are the highest for pensions and the lowest for maternity benefits. In the case of the means-tested child allowances, the major social assistance program, only one in three eligible households actually received payments during the month of the survey. Receipt rates were slightly higher for unemployment benefit which was received by around two in five eligible households. The proportion of beneficiary households, and the impact of the transfers, 9 The income concept used is the one described in the Resolution of the Government of the Republic of Kazakhstan, No. 604, May which lists the types of incomes per family member used to target the child allowance and other forms of social assistance. It includes labor compensation, income from self-employment, property income, pensions, scholarships, allowances (except allowances for families with children), payments to single pensioners, birth and funeral allowances, disability and sickness benefits, and prizes in cash or kind. 35

46 would obviously be significantly higher if a greater proportion of eligible households were to in fact receive benefits For the majority of transfers, the lack of resources was the most common reason given for nonreceipt. In the case of pensions for example, over 90 percent of those who did not receive their pensions during the month of the survey cited 'shortage of cash' with the local authorities as the main reason for non-receipt.1 There is a distinct locational pattern to non-receipt of transfers -- it is higher in rural areas and in the South and the East than in other parts of the country. The incidence of non-receipt in Almaty city is however small Despite the fall in benefit levels and non-receipt, cash transfers do play an important role in protecting living standards in Kazakhstan. Pensions and other transfers account for nearly 20 percent of consumption of the bottom quintile and over 20 percent of consumption in the second quintile (see Table 3.5).11 In terrms of household cash income, these shares are higher (see Chapter 2). Public transfers fall to a little over 7 percent in the top quintile. For the population as a whole, transfer payments account for around 15 percent of total consumption. In terms of relative importance, pensions are the most important transfer while child allowances come second. In the lowest quintile, pensions contribute over six times as much to household resources than child-allowances While important to the income and consumption of the poor, the social programs in Kazakhstan - - especially social assistance -- are clearly not as generous as in some parts of the FSU. Transfers contribute less to household income in Kazakhstan that in Russia or Estonia both of whom spend a larger proportion of GDP on social protection.' 2 This is largely related to the sharper contraction in government revenues in Kazakhstan and the need to maintain fiscal discipline. In response to the 'thinning' of the social safety net, self-production has emerged as an important coping strategy in Kazakhstan and family networks have also become an important source of income for the poor (Chapter 2). It is however difficult to reach any firm conclusions on whether these alternative responses are more significant in Kazakhstan than in, say, Russia. Table 3.5: Share of public transfers in consumption Kazakhstan Consumption quintile Bottom Second Third Fourth Top Total Pensions Mean-tested child allowances Unemployment benefit All others Note: All others includes other child allowances, other social assistance, sickness and maternity benefits Source: KLSS Another 3 percent did not collect pensions because they thought it too small to claim, and 5 percent did not have the necessary documentation. 11 The bottom two quintiles are roughly equivalent to the poor as defined earlier section and in Chapter However, transfers are more important than in Azerbaijan or the Kyrgyz Republic. 36

47 3.18 The limitations imposed by the overall public resource constraint underline the importance of consolidating programs and targeting assistance on those most in need. This is taken up further below. The targeting of cash benefits 3.19 Table 3.6 presents information on the targeting of existing public transfers. It shows the proportion of transfers received by different consumption quintiles. If a higher proportion of a transfer is received by the lower quintiles, then the transfer is well-targeted at the poor (or, is progressive). Table 3.6 Proportion of program expenditures reaching different quintiles in Kazakhstan Consumption quintile Bottom Second Third Fourth Top All pensions Old age pensions I Disability Loss of bread-winner* Unemployment benefit* SIF -Sick pay* SIF -- Maternity benefit* All child allowances Means-tested child allowances Other child allowances Other social assistance *: Owing to the small number of sampled households who receive these transfers, the results should be treated as indicative. Source: KLSS In general, we are more interested in the targeting of social assistance transfers specifically aimed at the poor. However, the targeting of social insurance payments is also of interest from a poverty perspective as the insurance nature of many of these payments has been eroded giving them a strong social assistance (redistributive) element. The picture is muddied by the non-receipt of benefits. If nonreceipt is higher among poor households (as we might expect given the information on regional and rural/urban non-receipt patterns), then it is likely to make a program appear less progressive than it would be if it were meeting its expenditure commitments fully Pensions: Pensions in Kazakhstan are not progressive. The benefits flow largely to the middle income groups with the bottom quintile receiving 1 1 percent of total program expenditure. This is in keeping with the fact that pensioners are not over-represented in the bottom quintile but are concentrated in the second, third and fourth quintiles (Chapter 2). Of course, the pension system in Kazakhstan does not purport to target the poor Child allowances: Child allowances in Kazakhstan are marginally progressive. Slightly more than 20 percent of the benefits go to the bottom quintile, and over 45 percent to the bottom two quintiles. The means-tested scheme favors the poor more than the categorical allowances which largelyflow to the middle quintiles. 37

48 3.23 How does the performance of the means-tested child allowance scheme in Kazakhstan compare with targeted schemes from other parts of the world? In general, the scheme performs quite poorly in international terms in directing benefits towards the poor. Table 3.7 lists the performance of other targeted schemes in terms of proportion of benefits retained by the bottom quintile compared to the benefits that leak to the top two quintiles. In terms of the bottom quintile, the Kazakhstan scheme performs worse than the mahalla scheme in Uzbekistan, 13 the locally administered social assistance scheme in Albania (Ndihme Ekonomika) and the housing allowance scheme in Chile, where 31 percent, 52 percent and 28 percent of program expenditures respectively reach the bottom quintile. The Kazakhstan scheme is also less successful than the Albanian scheme and the Chilean scheme in preventing leakage to the top 40 percent, but better than the Uzbek scheme on this count. On balance there would appear to be further scope for reducing the leakage to the non-poor and increasing the share of the poor. Table 3.7 Comparative performance of targeted social assistance schemes (Proportion of benefits by quintile) Consumption quintile Bottom 20% Top 40% Kazakhstan means-tested child allowance Uzbekistan mahalla Chile housing allowances Albanian social assistance scheme (Ndihme Ekonomika) 52 2 Source: KLSS 1996, Coudouel et. a]. (1997), Grosh and Baker (1995), Alderman (1998) 3.24 Why does the means-test not direct more program expenditure towards the poor? The means-test in Kazakhstan is weak because measurable income is a poor proxy for household consumption. This is only expected given the importance of self-production and its positive correlation with household consumption in Kazakhstan, and the significance of private inter-household transfers which are inherently difficult to measure. Using the income concept used in the means-test, the KLSS was used to construct an estimate of per capita household income. This is plotted against per capita household consumption in Figure 3.1. The two measures, per capita income, and per capita consumption show little correlation. For any given level of per capita income, there is a substantial variation in per capita consumption. 14 This suggests that there might be a case for supplementing income with additional information on living standards in order to better target the poor. The scope for strengthening the means-test is discussed in further detail in the following chapter. 13 The mahalla scheme is a program of local assistance channeled through local community organizations. 14 Less than 2 percent of the variation in per capita consumption can be explained by variation in per capita income. 38

49 Figure 3.1: Per capita income versus per capita consumption in Kazakhstan g 0 0 o 0 co, 0 o 0 00 co co RD oo ~ oc co C 0000% C ~ 00 poor 5000e: T t o0 14 pe 0 0 Q) C) 0 (Do 0 0 c 0 O's) (p0~~~~~~~~0 O D Q )0- u e0o0 of c 0orrelates ,1V ec&-,6~00 0. ~~~~~~~~o 5co0 (9 )- co per capia monthl icome 30C 3.26~~~( Unmlomn benfit Unmlomn beeft wihotbengitetonal C agee larel Noowte: the onc Oenratino points on the leefit becaue14vern is of thouehboldsm qinthle. srerpothzer isnome.rriin childn allowancesd whcharelcategoricallyn taretpoyedtadpvrycatr2. Thesml proportion ofthsaloncsrcigte botteomqntles in the caspe ofa categrall rchild alemlowymest isenegligile Thean chattegoriesuintse arshousdb porecreated s ofdi lowiper. aiacnupin 3.26 Uicnemsloymnt baerneit: UenemploymTent benefits withoutt beinageintentionallby tarete largelyr The small numbers caveat applies with equal force to these benefists Other social assistance: 47 percent of all benefits go to the top quintile. Other social assistance consists of a range of local programs of which the elderly and the disabled living alone have been the traditional focus but there are other groups as well. Note that the elderly living alone were identified as a heterogeneous group (not uniformly poor) in Chapter 2, so we would expect this benefit to have high leakage. Many oblasts also target the 'low income' population though the basis for this targeting is 39

50 unclear." 5 Russia. These transfers show a marked bias towards the non-poor. A similar finding was observed in 3.29 The geographical targeting of transfers: Given the strong regional dimension of poverty, do a higher proportion of public transfers flow to poor areas? In general, no. This is because eligibility for public transfers in Kazakhstan is tied not to poverty status, but to the demographic and beneficiary characteristics of regions. Pensions, the dominant program, is a universal program. Thus, the regional disbursement of pensions is related to the distribution of beneficiaries across regions not to regional poverty rates. Arrears may however make the system appear regressive in regional terms because of the higher incidence of non-receipt in rural areas and poor oblasts. Child allowances, on the other hand, do largely flow to the poor areas. Over 60 percent of means-tested allowances disbursed in July 1996 went to the South and the East, and over a quarter to the West. If we look at the bottom two quintiles only, over 60 percent of disbursements were again in the South and East. This pattern is related to the fact that those eligible for child allowances tend to be poor. Or, to put it differently, child allowances favor poor regions because they are targeted at one of the characteristics of poor households, namely, the presence of many children. Conclusions 3.30 This chapter has several conclusions: * Kazakhstan's social safety net has undergone major transformation since The need to maintain fiscal discipline in the context of a sharp contraction in revenues meant that spending on the social safety net as a share of GDP fell by over a third from 11.2 percent of GDP in 1992 to 6.6 percent in Although new forms of social protection have been introduced -- such as unemployment assistance -- there has been a general decline in program beneficiaries assisted by a tightening of eligibility criteria. Real benefit levels have declined substantially, with the exception of means-tested child allowances. * The composition of expenditures among social insurance and social assistance has remained broadly unchanged. Pensions dominate social protection expenditures (accounting for over 75 percent of all social protection expenditures in 1996, and nearly 90 percent in 1997). Social assistance schemes, of which the means-tested child allowance is the most important, are fairly small. * Despite the general contraction of the system, there remain a multiplicity of benefits and privileges, generally small in value, which are complex to administer. * Over 60 percent of poor households receive no public transfers at all. This is partly related to pure 'gaps' in the social safety net, and partly due to non-payment of benefits to eligible households due to resource limitations. * Despite the falls in benefit levels and non-receipt, cash transfers play an important role in protecting living standards in Kazakhstan. Pensions and other transfers account for nearly 20 percent of consumption in the bottom quintile, and over 22 percent in the second quintile. * Among existing social assistance programs, targeting of benefits is generally poor. In the case of the means-tested child allowance scheme over 50 percent of the program expenditure leaks to the nonpoor (top 60 percent). Better targeting outcomes have been achieved in other countries and considerable scope remains for improvements in Kazakhstan. 15 Benefits from these programs consists of both cash and in-kind assistance. In-kind benefits include 'in-home' assistance, especially for the old and/or disabled who live alone, food, assistance with the purchase of vehicles (for the invalid) and medical devices. 40

51 * Social assistance which is targeted using categorical criteria at pre-identified "vulnerable" groups generally does far worse in terms of leakage than the means-tested child allowance program as the categories in use do not appear to be well correlated with low consumption. 41

52 4. REDUCING POVERTY IN KAZAKHSTAN Introduction 4.1 The first three chapters of the report focused on the recent past in Kazakhstan. Chapter I summarized the main developments in the macroeconomy and labor markets. Chapter 2 examined the impact of these developments on living standards in Kazakhstan with particular reference to the poor. The third chapter looked at the social safety net, its evolution during the transition, and its targeting effectiveness. This chapter will address the policy implications that flow from the foregoing analysis. The importance of growth for poverty reduction in Kazakhstan 4.2 Economic activity in Kazakhstan has contracted by over 40 percent since independence leading to a sharp fall in wages and a rise in unemployment. Over a third of the population was living in poverty in Given the scale of output decline, and the rise in poverty, it should be clear that broad-based growth that both raises wages and creates employment will be key to reducing poverty in Kazakhstan. Growth is critical not just for raising incomes but also for providing the resources for a fiscally sustainable safety net to help those who might be temporarily or chronically disadvantaged. 4.3 Fortunately, the early years of crisis following the dissolution of the FSU have passed. Inflation is under control and output is no longer contracting. The first, and key, element of a poverty reduction strategy for Kazakhstan thus consists of maintaining and nurturing the current policy environment. 4.4 In general, the prospects for a sustained increase in output over the medium term in Kazakhstan are good, both because it is one of the stronger reformers among the FSU countries, and because it is well-endowed with natural resources.' The World Bank estimates that the economy will grow modestly in the short-run (2-3 percent), picking up speed (4-6 percent) early in the next century. The sectors which are likely to see the most rapid growth are agriculture, natural resources, and non-government services. As in other transition economies, the private sector is likely to provide the main impetus for growth. In addition to macroeconomic stability and low inflation, there are several areas where further development of the policy environment will be critical to fostering growth. They are: (i) further development of the legal and regulatory framework for private sector development, (ii) stronger regulation and oversight of the financial sector, (iii) deepening of capital markets in conjunction with the careful management of pension reform, (iv) sound management of the process of enterprise restructuring to encourage output and productivity growth, (v) improved management of public resources, including an expansion in the level and the quality of the public investment program, and judicious handling of likely future natural resource I Natural resources can be a mixed blessing. The competitiveness of non-oil sectors of the economy has been found to be a problem in almost every country that has experienced a rapid expansion in foreign exchange earnings from petroleum exports. In a recent study covering growth in ninety-seven developing countries during the period , Sachs and Warner (NBER Working Paper, 1995) found that countries with high export earnings based on natural resource wealth grew more slowly than those without such natural resources. 42

53 revenues, (vi) further restructuring and improvement in the quality of health and education services, and (vii) improved governance, through the development of a professional, well-remunerated civil service As with other oil-exporting countries, Kazakhstan will need to adjust to large inflows of foreign exchange and increased competition from imports. Over the medium run, it will be essential for the government to give attention to policies that improve the competitiveness of the non-oil sectors, particularly as these sectors are vital from the viewpoint of employment creation. Is growth sufficient? 4.6 While growth is a necessary condition for the reduction of poverty in Kazakhstan, it is not sufficient. The experience from other economies in transition (e.g. Poland) suggests that consecutive growth over even three or four years may do little to reduce the proportion of the population in poverty There is no reason to expect Kazakhstan to be different. 4.7 First, growth will not necessarily result in net increases in employment, at least in the short run. Further restructuring needs to occur in the non-service industries, notably in agriculture, but also mining, transportation and utilities. This is likely to involve significant net reductions in employment, given inherited levels of over-staffing and the limited adjustment of employment to falls in output since independence. 3 Although many services will expand, non-market services (government) are, if anything, likely to contract. Over 6 percent of the population in Kazakhstan is employed in government compared to less than 3 percent even among the less affluent members of the European Union. The government is committed to a trimming down of the civil service, with the intention of bringing public services in line with fiscal capacity and raising remuneration levels for remaining staff. 4 Information from the Kazakhstan Living Standards Survey (KLSS) on employment probabilities suggests that differences between those with university education and those with technical and vocational training have already emerged. Many of the semi-skilled, and those with mismatched skills, are likely to become vulnerable to poverty through unemployment. 4.8 Second, growth is unlikely to be distributionally neutral. Wage and earnings differentials were suppressed under the former system with skilled manual labor compensated relatively well compared to other professions. 5 With the growth of the private sector and the market economy, eamings differentials are likely to diverge. In other words, wage growth cannot be expected to be distributed evenly across the population. 4.9 Third, many of the poor may not benefit from growth because they do not benefit directly from the labor market e.g. the elderly, discouraged workers, etc. (Chapter 2). 2 For an elaboration of these arguments, see the following World Bank documents: Kazakhstan Country Economic Memorandum (1994), Kazakhstan Agricultural Sector Review (1994), Kazakhstan Transition of the State (1996), and Kazakhstan Country Assistance Strategy (1997). 3 Employment norms that are several times levels found in other countries are still common in Kazakhstan. In Kzyl- Orda, for example, sheep farms typically employ about one person per 200 animals compared to the international average of one person per 600 animals. In water utilities there are 50 staff members per 1,000 connections, compared to employees per 1,000 connections in relatively inefficient utilities, and 2-3 employees per 1,000 connections in efficient ones. See World Bank (1997), Kzyl-Orda Regional Development Priorities. 4 The reform of the civil service is being supported under the Bank's Public Sector Resource Management Adjustment Loan (PSRMAL). 5 See Nick Baff (ed.) (1994), Labor Markets and Social Policy in Central and Eastern Europe. 43

54 4.10 There is thus a clear need for forward thinking regarding two types of policy interventions: policies designed to facilitate labor mobility and equip workers to adapt to changed circumstances (labor market policy), and policies designed to strengthen the safety net and provide transfers to reduce income losses (social protection policy). The two policies will perforce interact. For groups who are not well placed to benefit from growth, both sets of policies will have a vital role to play in maintaining living standards. The recommendations of this report are essentially focused on the second set of policies. A proposed companion study on labor market policy is aimed at examining the first. Fiscal capacity and poverty reduction 4.11 While recognizing that the social safety net will be important for maintaining the living standards of the poor, it is important to recognize that severe constraints have been imposed by the overall fall in economic activity since independence and the need to keep the budget deficit under control. As discussed earlier, under favorable policy conditions, the economy is expected to grow by 2-3 percent in the shortrun, and 4-6 percent after the year The immediate growth in revenues under this scenario is limited. 6 Moreover, the move from a public pay-as-you-go pension system to a private fully funded system in 1998 is likely to add a significant burden to the budget for many years to come. 7 This is likely to limit the scope for substantial real increases in social protection expenditures A second element of a poverty reduction strategy would therefore be making social programs more effective within existing fiscal constraints. These are spelt out in greater detail in the sections that follow. The proposals include: (i) ensuring the adequacy of pensions under the new pension reform legislation; (ii) improving the targeting of social assistance; (iii) rationalizing social assistance programs so that they are both less fragmented and more effective, (iv) strengthening the system of unemployment compensation, and, (v) assisting the poor through regionally targeted temporary employment schemes. These proposals are presented as options for the authorities to consider and refine While the gains to revenue from growth may be modest in the short run, there remains much unexploited potential in system of tax collection. The share of public revenues in GDP has fallen sharply since independence (Chapter 1). Improvements in collections rates therefore deserve serious attention. A recent IMF technical assistance mission has identified a number of areas in which tax administration could be strengthened. These should be speedily implemented. The collection of the payroll taxes that fund social security, traditionally subordinated to the other tax liabilities, should also be improved. In the short run, this would involve addressing the difficulties currently being experienced by the relevant authorities (the State Center for Pension Payments, SCPP) in collecting pension contributions under the reformed system. Over the medium to long term, it would be worth examining the options for extending the mandate of the SCPP to other social security contributions. The economies of scale in this approach 6Moreover, social protection expenditures are not the only pro-poor use that additional resources can be put to. Public investment has a vital role to play in generating growth and employment which in turn would reduce the call on social protection programs. Public investment in Kazakhstan has fallen substantially since The proposed reform would add to the budget deficit by 2 percent of GDP in the first year of reform. The implicit debt of the PAYG system is calculated at 110 percent of 1997 GDP, assuming a discount rate of 3 percent per annum. 44

55 would be considerable. Where necessary, the base for social insurance taxes should be broadened by unifying tax rates across sectors. This is particularly relevant to the contributions to the Employment Fund from agriculture and the budgetary sector which should be raised from 1 to 2 percent to equal other sectors. Improvements in collections would allow a shift to lower tax rates over time, which would in turn strengthen compliance and further raise collections. Table 4.1 Social protection reform agenda Reform Reform Time-frame Impact on government finances Area Pensions No new accumulation of arrears Short term Negative Maintain value of minimum pension in Short term Negative, to keep costs under real terms control government should flatten benefit structure Explore options for raising replacement Medium term Neutral rates under the funded scheme, especially for women and those with short work histories Restore disability and survivor's Medium term Neutral insurance Social Develop mechanisms for improved Short/medium Neutral assistance targeting of means-tested child term allowances Improve targeting effectiveness of Short/medium Unclear housing allowances, reduce urban bias in term scheme Reduce preferential tariffs for selected Medium termn Positive population groups Rationalize social assistance programs Short/medium Neutral term Introduce poverty benefit Medium term Negative; neutral if it replaces current local programs of social assistance Labor Increase coverage of UB Short term Variable by method employed market Develop options to assist the long-term Short/medium Negative programs unemployed termn Public works programs Short term Negative Inter-governmental fiscal relations 4.14 An area that warrants particular attention over the medium term is the role played by intergovernmental fiscal relations in reducing resource mobilization. The present revenue sharing arrangements between the republican and oblast authorities -- with poor incentives for local tax inspectors to collect taxes that do not benefit the oblast authorities -- do not promote effective revenue collection. The pattern of arrears in wages and cash benefits across oblasts (and to a lesser degree between urban and rural areas) is intimately tied to this system of revenue sharing among different levels of government. An effective and stable system of revenue sharing with incentives for local tax collection would need to be established. 45

56 Strengthening social protection A. Ensuring the adequacy of pensions under the new pension reform legislation 4.15 From a poverty perspective, the strength of the existing pension system in Kazakhstan is that pensioners are relatively well-protected from extreme (bottom quintile) poverty. A key distributional issue in any proposed reform would therefore be its effect on the living standards of pensioners. The 1997 pension reform legislation essentially preserves the pension entitlements of existing pensioners and those with sufficient service history in the former system (25 years for men, 20 for women), while phasing in a greater reliance on mandatory private saving as a source of old-age income for future generations. Pension levels are protected from the vagaries of investment returns through a stateguaranteed minimum pension, and for those who do not qualify for the minimum pension, the legislation allows for social assistance in old age. The new system comes into effect on January 1, Paying pensions on time and maintaining the level of the minimum pension in real terms: In terms of pension adequacy, the two key issues under the proposed reform are: (i) whether the Government holds to its resolution to allow no new accumulation of arrears, and (ii) the level of the minimum pension. Over the short-run, when most pensioners will draw pensions from the state system, paying pensions on time will be key to maintaining living standards among the old. The level of the minimum is equally relevant. For 1998, the minimum pension has been set at 2,400 tenge (about US $32 at market exchange rates), which is roughly 70 percent of the Subsistence Minimum. At percent of the economy-wide average wage, the level is in keeping with international practice and is consistent with the World Bank's recommendations on the subject. The Government intends to set the level of the minimum pension every year. Over time, there is likely to be pressure to reduce the cost of the transition to the new pension system by letting inflation erode the real value of state pensions. However, from a poverty perspective, a containment of costs is best achieved through compression of the pension distribution above a fully indexed minimum rather than an erosion of the minimum benefit. It would therefore be important that the minimum be fully indexed to prices as a rule and any desired flexibility be achieved through less than full indexation of higher pensions. The system would thus gradually develop into one with a relatively flat structure of benefits before being eventually phased out Examining other options for the indexation of the minimum: The Government may wish to examine other options for indexing the minimum, including some level of indexation to wages. This would allow the minimum to reflect not just a minimum absolute standard, but to rise gradually in real terms in response to rising living standards in society as a whole. The implications of the indexation choice for costs and fiscal sustainability should be thoroughly explored prior to any decisions Assuring adequate replacement rates: A major characteristic of pension systems that rely on personal saving are that they tend to offer low replacement rates to low-income workers and those with short or interrupted work histories, many of whom are women. Low-income workers may find it difficult to accumulate sufficient pension capital for an adequate income in old age. Short contribution histories (reduced by spells of unemployment for example) would also lead to significant reductions in the replacement rates offered by the new system. In Kazakhstan, shorter work histories for women would be compounded by the fact they retire five years earlier than men, and live nearly twice as long after retirement. The combined effect of these factors is that women can expect replacement rates which are significantly lower than that of men under the new system There are a number of options for improving the outcomes for low-income workers, those with short earnings histories, and women. Higher contribution rates are one alternative. Contributions from 46

57 other social insurance funds to cover periods of temporary inability to work (owing to sickness, maternity or unemployment) could also be considered. International experience suggests that it will be important for the population to be fully informed about likely pensions under different earnings and employment scenarios so as not to raise false expectations, and to encourage voluntary saving. The equalization of retirement age would allow women a longer period during which to accumulate retirement savings, while the use of standard life tables when issuing annuities would reduce the 'disadvantage' of higher longevity. Pooling the contributions of husbands and wives is also likely to help women. Many of these considerations are further down the road as over the short term, most retirees will receive grandfathered state pensions. However, it will be important for the Government to develop the capacity to analyze likely outcomes for both average and non-standard working histories, the likely call on the minimum pension and social assistance, and to devise and examine policy options to ensure adequate replacement Restoring disability and survivor's insurance: The 1997 pension reform legislation transfers disabled workers and survivors from the pension system to the social assistance system, under which they would receive a flat-rate benefit funded out of general tax revenues. This idea is not in concert with international practice. Disability and death are insurable events with insurance (either social or private) allowing for an element of consumption smoothing. The proposed benefit levels also set up perverse incentives for low wage workers to take up the more generous disability pensions. In the absence of a private insurance industry in Kazakhstan there would appear to be a case for the government to continue to provide this form of insurance (at least for those with the most severe disability, and in the event of the loss of the breadwinner) till a private industry develops that could implement such insurance in complement with the private pension funds. B. Improving the targeting of social assistance in Kazakhstan 4.21 Regarding social assistance, the authorities have already taken efforts to improve targeting and increase coverage, notably through the introduction of the means-test for the basic child allowance, and the introduction of a means-tested housing allowance scheme. However, there is scope for the implementation of further reforms to increase the allocation of resources to the poor Strengthening means-testing of child allowances through the use of additional information: Means-tested child allowances are targeted at a group that is at a greater risk of poverty i.e. large families. Despite this, and the presence of a means test, a large proportion of the benefit (over 50 percent) leaks to the non poor. Reducing leakage to, say, half of present levels, would make an additional 25 percent of program expenditure (around 2 billion tenge or US $30 million at 1996 prices) available to the poor An important issue raised by the KLSS is the suitability of targeting child allowances on the basis of income alone. One way to improve targeting at low cost would be to use additional demographic characteristics that are well-correlated with poverty. One criterion that suggests itself on the basis of the poverty profile discussed in Chapter 2 is the presence of children less than six years of age. The KLSS was used to simulate the effect of targeting child allowances at families using this alternative criterion (see Table 4.2). Restricting child allowances to children less than 6 years raises the proportion of program expenditure going to the bottom quintile to 42 percent, and the bottom two quintiles to 58 percent. Leakage to the non-poor (or top three quintiles) falls by 18 percent from 51 percent to 42 percent. 47

58 Table 4.2: Distribution of program expenditure under different targeting assumptions Means-tested child allowance program Consumption quintile Bottom Second Third Fourth Top A. All households with children < 6 yr B. Means test + children < 6 years Existing means-tested program Other child allowances Other social assistance Note: Under option A, the existing program budget is distributed among all households with children less than 6 years. Under option B, it is distributed among households with children less than 6 years who satisfy the means-test. Source: KLSS An extension of this idea would be to use a group of indicators including income to determine whether a household should be entitled to assistance or not. This is the basis forproxy-means testing used to target a number of social assistance programs in various Latin American countries e.g. the Ficha CAS program in Chile. In the FSU, a proxy means test is used to target humanitarian assistance in Armenia, and various targeting improvements, including proxy means-testing, are being piloted in Russia under the World Bank's Social Protection Adjustment Loan. Proxy-means tests vary in complexity but typically involve the use of a small number of household characteristics (such as income, location, number of children, education, and ownership of durables) which are strongly correlated with poverty. The characteristics need to be verifiable -- this is usually undertaken by a social worker -- and not subject to easy change. They are weighted to arrive at a 'score' which determines eligibility for a program. Weights typically represent degree of correlation with poverty status. The strength of correlations, and weights, may be investigated using a multi-purpose survey like the KLSS. In principle, a proxy-means test has the potential for improving the targeting of social assistance considerably. While potentially attractive, the usefulness of introducing proxy-means testing must be weighed against the cost of introducing a more complex system. At the very least, leakage to the non-poor should fall sufficiently to justify the costs of introducing a new system. Lessons from the Russian pilots will be particularly relevant in this context Strengthening housing allowances through the use of additional targeting information, improved design and the elimination of privileges: Given the likely significance of the housing allowance program, and the limitations of income-testing, it will be important to examine the housing allowance scheme in Kazakhstan from a targeting perspective. 8 The assessment should cover how well existing criteria perform, as well as the scope for making improvements. As in the case of child allowances, the use of additional information that is well correlated with household poverty may prove necessary Two additional features of the housing allowance scheme are worthy of attention. (i) First, the scheme has a distinct urban bias. It is designed to compensate households for increased prices charged by utility companies or housing maintenance organizations (zheks). It is 8 It was not possible to undertake such an investigation using the KLSS. It will be essential to include questions that would allow for such analysis in any subsequent surveys. 48

59 therefore likely to exclude many poor rural households who typically rely on non-standard forms of heating (LPG, wood) which are also subject to price increases. This feature has important regional dimensions, as the poorer regions in Kazakhstan tend to be more rural. By the end of 1996, for example, Kzyl-Orda and South Kazakhstan oblasts identified by the poverty profile as being the poorest had the second and third lowest number of beneficiaries from the housing allowance scheme. 9 This suggests a need for closer consideration of alternative billing systems that would extend program coverage. (ii) Second, although a housing allowance scheme has been introduced to protect low-income households from utility price increases, the government still maintains preferential tariffs for selected groups in the population, who typically pay 50 percent or less than the administered price. These privileges are directed at groups considered vulnerable or deserving under the Soviet system. External targeting mechanisms are generally considered to be more effective at reaching the poor than internal mechanisms that rely on differential tariffs, and the Soviet targeting categories have been shown to have their limitations. There would appear to be little reason to retain these privileges once a well-functioning housing allowance scheme is in place. The Government agrees with the need for a closer examination of the issue and is committed to preparing proposals for revisions to the existing system under the World Bank's PSRMAL. C. Rationalizing social assistance programs 4.27 Rationalizing other child allowances: In addition to the main means-tested program, Kazakhstan has several other child allowance programs ostensibly targeted at vulnerable households to whom benefits are awarded on categorical grounds. The rationale for these programs is not apparent given the existence of the means-tested benefit. The categories in use do not appear to be well correlated with poverty (see Chapter 3). Moreover, the number of beneficiaries under certain programs are too small to justify their existence. 10 In 1996, there was one child benefiting from the program for children with HIV. There would thus appear to be a strong case for continuing the existing trend to streamline and/or merge these programs with the means-tested benefit Rationalizing other social assistance programs: Non child-related programs of social assistance in Kazakhstan appears to be particularly poorly targeted. A closer investigation of the underlying factors is warranted. One reason would appear to be that the categories of people who receive this assistance were identified as vulnerable during Soviet times, but in the changed circumstances there are others who are more deserving. While a complete investigation of the programs encompassed under this head is beyond the scope of this report, there are two principles which should guide the redirection of this assistance. First, objective guidelines should be introduced to determine eligibility. This could be done without an income test, but if formal income testing is considered desirable, the income test used for the child allowance would appear to provide a stronger basis for targeting than the categories currently in use. Second, as far as possible, this assistance should be provided in cash rather than in kind so as to allow poor household to assess their needs themselves. There is also a case for supporting the efforts of NGOs and charities in the area of social assistance to both augment local budgets and administrative capacity.'] 9 Of a total of 150,000 beneficiary households, 1601 and 1539 were in Kzyl-Orda and South Kazakhstan respectively. 10 See Appendix 3 for full details on number of beneficiaries and program expenditures. 11 The IFES survey conducted in December 1996, indicated a wide interest among the population in joining NGOs which provide assistance to the needy. More generally, as shown in Chapters 2 and 3, informal social safety nets are an important source of income for the poor in Kazakhstan. 49

60 4.29 Giving consideration to the introduction of a poverty benefit: One option worthy of consideration is the introduction of targeted income support for the poor. There is currently no targeted benefit for those who are poor but do not qualify for public means-tested assistance through the child allowance scheme or the housing allowance scheme. The lack of benefits for people who are generally poor and do not fall into one of the pre-determined categories partly explains why coverage among social assistance programs is so low (see Table 3.4). This would suggest that the various non means-tested and local programs are perhaps best replaced with a targeted poverty benefit. In any such scheme the level of the benefit would need to be carefully considered to preclude adverse incentives. At the very minimum, the benefit should be lower than the level of other social insurance payments. It should also provide an encouragement for people to help themselves. This would suggest that the level of the benefit should be austere, and given the considerable reliance on self-production and informal safety nets in Kazakhstan, about a third to a half of the level deemed necessary for minimum subsistence would not be inappropriate. Significantly, the poverty benefit would not be aimed at bringing the individual or household up to the subsistence level but to provide a helping hand to the process. For administrative efficiency, the program could be administered by the local departments of social protection. D. Strengthening the system of unemployment compensation 4.30 Increasing coverage of unemployment benefits: Unemployment benefit is a well targeted element of Kazakhstan's social safety net. After eroding sharply in value, the level of benefit was restored to its 1993 level in real terms, accompanied by a large increase in the number of registered unemployed. Yet coverage remains low. A large proportion of the unemployed (over 40 percent) simply do not register. This warrants closer investigation. According to the KLSS, among those who were unemployed but did not register, over 8 percent did not know how to register, another 10 percent said they did not rely on the Employment Services (ES), while nearly a third said they were not interested. The unwillingness to register may be related to the fact that the number of vacancies offered by the ES is typically smaller than the number of job seekers, and the vacancies are often for very low paying jobs. It is also a likely reflection of the low level of unemployment benefit relative to wages. Although UB was restored to over 1.25 times its 1993 level in real terms, it is roughly equivalent to 20 percent of the economy-wide average wage, and 50 percent of the subsistence minimum (Chapter 3). Increasing information, and raising incentives for registration, including considering raising unemployment compensation in real terms, would thus be important to extending coverage. At the same time, it will be important to ensure that only those who are genuinely unemployed remain eligible. Enforcing rules -- especially regarding job refusals -- will be important Considering policy options regarding the long-term unemployed: In extending coverage, one question that arises is whether the long-term unemployed should be entitled to benefits once eligibility for UB has expired. This is already happening to some extent in the case of individuals who move to training after having been on UB for six months. In general, there would appear to be a case for providing benefits to the long-term unemployed provided rules governing eligibility are strictly enforced. The case for extending benefit duration is stronger in regions which have suffered mass layoffs or have very depressed demand conditions. In Kazakhstan, there is the additional problem that there is no social assistance program that such people could be automatically shifted to, if this was considered desirable. However, the impact of such a policy of extending unemployment benefit on the financial viability of the Employment Fund would have to be carefully considered. More generally, it will be important to consider policy options to combat long-term unemployment, including social assistance, training, recruitment incentives, direct employment, and counseling and guidance. 50

61 E. Strengthening public works programs 4.32 Public works programs: The observed correlation between poverty and unemployment, and difficulty of targeting transfers in Kazakhstan, suggests a role for public works programs in targeting assistance through self-selection among the poor (see Box 4. 1). Public works programs operate on a very limited scale in Kazakhstan at the moment, accounting for less than 1 percent of Employment Fund expenditure. These schemes could however be expanded significantly and play an important role in targeting assistance to the poor. Given the possible association of such schemes with Soviet labor camps, it may however to important to emphasize community services at the expense of heavy manual labor. This would also favor women who face a slightly higher risk of unemployment in Kazakhstan. Community services worthy of consideration for inclusion in public works programs include cleaning and sanitation activities, home-care for the aged, and informal child-care. Box 4.1: Targeting the poor: the role of public works programs Public works programs can help the unemployed poor, provided they are willing to work for low wages. The low wages act as a self-targeting mechanism because only the needy will accept them. These schemes are particularly appropriate during nation-wide recessions when other job opportunities are unavailable. They are also well suited for rural areas during the slack season and can have a secondary benefit of building or maintaining important infrastructure assets. In the United States, in what was perhaps the most significant use of public works programs among industrial countries to date, the Works Progress Administration provided employment for up to a fifth of all workers during the depression of the 1930s. Many developing countries have also made extensive use of public works programs. The Food for Work Program in Zimbabwe and the Emergency Social Fund in Bolivia provide a basic level of employment and consumption for some of the poor. The Employment Guarantee Scheme in the Indian state of Maharashtra provides employment to the poor in rural areas. Chile introduced large government financed urban public works programs during the recessions of the 1970s and 1980s. At the bottom of the recession of the early 1980s, they absorbed more than 10 percent of the labor force. By the end of 1988, a year after Chile's vigorous economic recovery, that share had fallen to less than 0.1 percent. In transition economies, public works programs have played a role in improving the targeting of benefits and as a bridge between jobs. In Estonia, the targeting of unemployment benefit was improved substantially when it was made conditional on participation in part-time public employment schemes. In Albania, public works programs are widely used to cushion the transition between jobs. Public works programs have a greater chance of success where labor is more mobile and where there is a tradition of community work. When workers are able to move, the location of the program can be determined mainly by the quality of the projects. But low mobility makes the programs less effective -- some studies indicate that obstacles to mobility in China have made it more difficult for the country's public works programs to reach the poor. Community participation greatly enhances the chances of success. The level of wages in such programs is important in determining their success at targeting the needy. High wages may attract better off workers and, given limited budgets, lead to fewer jobs being created for the truly needy. The importance of wage determination for the success of the public works programs is demonstrated by the Maharashtra program in India. In its initial fifteen years, the program maintained wages on a par with prevailing wages for unskilled casual agricultural labor. But in 1988, wages were raised in line with a doubling of the statutory minimum wage. The higher wage, combined with budgetary pressures, effectively eliminated the program in the year after the wage increase. Source: World Development Report 1995, Workers in an Integrating World, World Bank: Washington D.C It is important to stress that these schemes are intended to be temporary in nature, not permanent employment generating programs. Their primary purpose would be social assistance. To be selftargeting, the wage should be set at a rate to promote participation among the poor. Given the strong regional nature of poverty in Kazakhstan, it would probably be more effective for the schemes to be focused on a few poor oblasts than to be undertaken on a nationwide basis. Implementation of some 51

62 schemes by private businesses and NGOs should also be considered to encourage private sector development. Monitoring progress with poverty reduction 4.34 Whatever improvements are introduced to the system of social protection, it will be essential to retain an instrument for monitoring changes in living standards and targeting effectiveness. The Family Budget Survey (FBS) is the logical choice for such an instrument, although there is a need for significant improvements, most notably in sampling methodology and questionnaire design (see Box 4.2). The sampling frame and questionnaire developed for the KLSS would be a good starting point for these improvements. What is needed at this stage is to graft the improvements in sample and questionnaire design achieved under the KLSS on to the FBS. This is being taken up under the World Bank's Technical Assistance Loan. Box 4.2: Strengthening the FBS for poverty monitoring In order to serve as an instrument for monitoring progress with poverty reduction, the FBS would require strengthening in two critical areas. First, a probability sample would need to be developed so as to make the survey representative. Second, the FBS questionnaire would need to be modified in several ways: Linking: To measure household consumption and income, it must be possible to link data from every household for all twelve months of the year. This should be done by giving each household a unique identification code for the year. Methodology for administering questionnaires: Several types of data (income, health status) are not adequately provided by the head of the household. Such information should be obtained directly from the individual members themselves. Expanding the questionnaire: The survey should be expanded to collect information on: (i) access to and use of publicly provided services such as education, healthcare and utilities; (ii) who receives pensions, social assistance, other government transfers and non-governmental assistance; (iii) the age of durables and other inputs necessary to compute the use value of housing and other durables; (iv) person by person expenditures on hard-to-measure items; (v) local, especially rural, price data. Timing: Some types of questions may need to be asked more often than is presently the case. Unemployment data is one example where annual data is not adequate From a poverty monitoring perspective, there are some areas that deserve attention in subsequent surveys. It will be important to allow for comparability to the KLSS with respect to timing. At the same time, future survey work should build in an element of seasonality to improve understanding of seasonal differences is living standards. Future surveys should allow better estimation of private expenditures on education and healthcare which are likely to be significant sources of household expenditure and are key to devising policies to encourage human capital investments among the poor. In addition, necessary training to develop capacity within the Ministry of Labor and Social Protection and the National Statistical Agency to undertake policy analysis using the data is essential. These data and analyses would provide an ongoing and objective basis for strengthening the system of social protection. 52

63 Regional policy and poverty reduction 4.36 The strong regional dimension of poverty indicates the need for further thought on the impact of growth on regional differences in living standards in Kazakhstan, and the policy options for dealing with regional imbalances. This report has not touched upon regional development issues. The World Bank has however recently completed a study of the development priorities in the oblast of Kzyl-Orda which is affected by the Aral Sea crisis -- see Box 4.3. There is clear need for further work, particularly to: (i) take a global view of existing and emerging regional imbalances and enunciate an appropriate role for the republican government in reducing inter-regional disparities, and (ii) develop strategies for other poor regions which have not attracted the same degree of public attention. Box 4.3: Kzyl-Orda: Regional Development Priorities The Aral Sea crisis has attracted public attention -- of the World Bdnk as well as other international institutions (e.g., UNDP) -- to the fate of Kzyl-Orda oblast. In 1997, the Bank attempted to assess the potential for improving living standards in the region. This study was motivated by a desire to better understand the challenges facing the region and the options for its future development. The oblast currently posts living standards achievements which, beyond the immediate shore of the Aral Sea (Aralsk rayon), do not compare particularly unfavorably with those observed in the rest of the country. Such achievements are all the more remarkable given the fact that, as typical in the southern rim of Kazakhstan, income per capita in the oblast is only about 60 percent of the national average. They are a reflection of the powerful equalization mechanisms at play in the economy, particularly in the form of inter-regional transfers of public resources. But these transfers have also made Kzyl-Orda overreliant on the public sector, exposing the region more than others to adverse fiscal developments. The future economic and social development of the oblast will have to rest on different, non-public sector sources. Given the dire fiscal straits the country is in, there is little prospect that such transfers of resources could remain a driving force of development over the foreseeable future. Rather, the oblast will need to rely mainly on improving the productivity of its own -- albeit limited -- resources. The positive impact of ongoing reforms shows that this is indeed possible. Agriculture, traditionally the mainstay of Kzyl-Orda's economy, is already beginning to reap the benefits of trade and price liberalization. Foreign investment is helping to develop the oblast's oil resources. And the ongoing reform of utility tariffs is opening the door to better service provision. The challenge is to nurture and amplify these incipient trends. The main thrust of the authorities' efforts in this respect should be to complete the shift in the role of the public sector from direct interventions in production and trade to the provision of key public goods (e.g., research, heat, water), with priority being given to (i) maintaining and selectively rehabilitating existing public infrastructure, supported by adjustments in public tariffs, and (ii) invigorating labor productivity and mobility by maximizing the impact of reduced health and education expenditure. Source: World Bank (1997): Kzyl-Orda Regional Development Priorities Investments in human capital 4.37 A key element in any national poverty reduction strategy is the quality and efficiency of health and education services. The performance of the health and education systems in Kazakhstan, their organization, financing and delivery are topics outside the scope the present report. The discussion here is limited to a brief overview of the direction of education and healthcare reform from a poverty perspective. ] 12 Many of the relevant issues are raised in World Bank (1996), Kazakhstan Transition of the State, and Kazakhstan Health Sector Note. 53

64 4.38 The health and education sectors are facing a severe financing crisis which threatens both the coverage and the quality of services provided. Reductions in capacity have not kept pace with reductions in real resources, with the consequence that while access to services is nominally unaffected, quality and effective coverage are greatly reduced. In education, where resource allocation patterns among levels of education appears appropriate, priority should be given to: (i) Further consolidation of classes and institutions in the school sector and closure of specializations for which enrollment is declining in higher education; (ii) Increasing teaching loads in schools, including through introducing multi-topic and multigrade teaching methods in low density areas; (iii) In vocational and higher education institutions, development of curricula in high growth skill areas; and (iv) Improvements in the quality and supply of textbooks. In health, the resource allocation patterns show a bias towards inpatient and tertiary care. Recommendations to rationalize resource allocation patterns to cost-effective uses include the following: (i) Introducing medical protocols emphasizing ambulatory care where appropriate; (ii) Strengthening diagnostic and treatment capacity at the level of ambulatory and polytechnic institutions and devoting to them a higher proportion of spending; (iii) Establish a basic package of services, for which access should be free, with clear financing arrangements for different services Conclusions 4.39 This chapter discussed elements of a poverty reduction strategy for Kazakhstan, focusing on the role of growth and strengthening of the social safety net. The summary recommendations are: * Maintain and strengthen the current policy environment so as to provide an impetus for economic growth. This would include: (i) further development of the legal and regulatory framework for private sector development, (ii) stronger regulation and oversight of the financial sector, (iii) deepening of capital markets in conjunction with the careful management of pension reform, (iv) sound management of the process of enterprise restructuring to encourage output and productivity growth, (v) improved management of public resources, including an expansion in the level and the quality of the public investment program, and judicious handling of likely future natural resource revenues, (vi) further restructuring and improvement in the quality of health and education services, and (vii) improved governance through the development of a professional, well-remunerated civil service. * Give attention to the competitiveness of the non-oil sectors so as to avert serious Dutch disease, especially over the medium-term; * Improve tax collection -- particularly social security taxes -- through strengthened tax administration, extension of the tax base and improved incentives for compliance; 54

65 * Maintain the minimum pension in real terms, and prevent any new accumulation of arrears; * Develop the capacity to undertake policy analysis of replacement rates under the new pension system, develop and implement options to raise replacement rates for low-income workers, those with short work-histories and women; * Improve the targeting of means-tested allowances by using additional information on household characteristics strongly correlated with poverty. This would include characteristics such as the presence of children less than 6 years of age, and the presence of at least one unemployed member in the household; * Make available the resources for the full implementation of the housing allowance scheme. Improve its design to increase coverage among households in rural areas. Redirect savings from a reassessment of privileges to social assistance; * Rationalize other programs of social assistance which are targeted at pre-identified categories of people and which currently largely benefit middle income groups; * Improve poverty monitoring and targeting through institutionalizing improvements to the Family Budget Survey and developing the capacity within the Ministry of Labor and Social Protection for undertaking policy analysis; * Increase the coverage of unemployment benefit; * Consider the expansion of self-targeted public works programs; * Undertake further study of regional impact of growth and the policy options to reduce regional imbalances; * Maximize the impact of reduced health and education expenditures through more cost-effective provision of services. 55

66

67 APPENDICES Appendix 1: Data and methodology 1. This appendix summarizes the key features of the data set and methodology employed to construct the poverty estimates discussed in Chapter 2. It is intended to serve as a companion to Chapter 2, designed to aid in the interpretation of the estimates. It explicitly addresses data and methodological limitations. Data 2. The household survey data used in this report are from the Kazakhstan Living Standards Survey (KLSS). The KLSS is a multi-purpose household survey, akin to a Living Standards Measurement Survey, and was implemented by the National Statistical Agency in July 1996 under a Bank-financed Technical Assistance Project. The information collected from sampled households includes income, expenditures, the nature and quality of housing, and other household characteristics such as number and age of household members, labor force status, educational attainment, health status, use of medical facilities etc.. A community questionnaire was used to collect information on local conditions, including prevailing local prices for food. 3. The sample consists of 1,996 households and 7,223 individuals and is designed to be nationally representative. Although the sample is not designed to be representative at the sub-national level, it is sufficiently large to obtain indicative results for the five main regions of the country: north, west, central, south and east. 4. The KLSS was implemented with technical assistance from the Sigma Institute, and represents a significant departure from the Family Budget Survey in terms of sample and questionnaire design. Significantly, it employed random as opposed to quota sampling techniques. As such, it provides the first opportunity to evaluate living standards and government interventions in the area of poverty reduction and social services during the transition. Choice of welfare indicator 5. This report evaluates household welfare by comparing the money value of household consumption per capita with the poverty line. Information on income was not used for a number of reasons. First, income data are often subject to under-reporting, particularly income from private and informal sector activity. Underreporting is related to a number of factors including the reluctance to disclose incomes due to tax implications, the growth of self-employment, and the importance of second jobs and other informal activities which have developed as a part of coping strategies. While consumption too may be underreported, the problem tends to be less severe than for income. 6. Second, consumption varies less than income, and therefore tends to be a more reliable estimate of living standards at a single point in time. Incomes, particularly in agricultural areas, tend to vary substantially over the course of the year. Households can, and do, smooth their consumption by saving and borrowing. Thus a household which receives a large post-harvest income may save some or a large part of it to meet consumption needs during lean months. Measuring household welfare using income runs the risk of classifying an otherwise poor household as not-poor because it received a large income 57

68 that month. This would reduce the overall estimate of poverty, and create problems when evaluating, for example, how well social assistance is targeted. Equivalence scales 7. As households differ in size and composition, simple comparisons of total household income or consumption would be misleading about the standard of living of individuals in those households. Allowances for differing sizes and compositions is typically made by expressing household income or consumption inper capita terms. However, this is an adjustment that gives all household members equal weight. A household which is twice as large is assumed to require twice the income or consumption in order to be as well off. This does not allow for differences in needs (a child needs fewer calories, hence food, than an adult) and economies of scale in consumption (the fact that two can often live as cheaply as one). 8. Equivalence scales are a sophisticated way of adjusting for differences in needs in households of different types. The choice of equivalence scale reflects judgments about differences in needs. For example, the per capita adjustment incorporates the extreme judgment that all household members have equal needs irrespective of age. Other judgments may be used, and a variety of scales are used in practice depending on country and context. There is no accepted equivalence scale in Kazakhstan or the FSU. Because of the difficulty in justifying the use of a particular equivalence scale, the best practice of examining the sensitivity of poverty estimates to a range of plausible scales has evolved. 9. Owing to time and resource limitations, this report was not able to investigate the impact of alternative equivalence scales on the estimates of poverty. In general, allowing for lower needs for children vis-a-vis adults would in all likelihood reduce the percentage of the poor in the country (by making large families appear less poor), and make the south (where people have larger families) appear a little less poor relative to the north. It would also weaken the correlation between poverty and household size. Appendix 4, discusses the related but distinct matter of economies of scale. Defining consumption 10. Consumption as defined in this report is the money value of all food and non-food goods and services consumed by the households. It includes all purchases, as well as the value of home-produced goods (such as food) and goods and services received in kind. Non-food goods cover clothing, cleaning, transport, leisure, utilities, health, education, and durables. Consumption is expressed in tenge per capita per month. 11. In defining total household consumption, four issues had to be decided: the consumption of home produced food, the valuation of healthcare and education, the treatment of consumer durables, and the treatment of owner-occupied housing. 12. Home-produced food is a very important source of consumption in Kazakhstan, as in other parts of the FSU. Excluding home-produced food would therefore bias consumption estimates, as well distort comparisons between households with and without access to plots of land. As this produce is often not traded or sold, the question arises as to how to value it. In this report, home production is valued at the prevailing local price for the food item. Local prices were obtained from the community questionnaires and were cross-checked for consistency with median unit values for items which were both purchased and produced at home. 58

69 13. Education and healthcare are important publicly provided goods. However, the fiscal crises in Kazakhstan has necessitated the introduction of charges (e.g. for outpatient drugs and pharmaceuticals), and anecdotally one learns of extensive private payments, especially in healthcare. The KLSS included a range of questions on out-of-pocket payments for these services which provided a basis for evaluating expenditures in these areas. Expenditure on healthcare was defined as all expenditures for chronic health problems (covering items such as medicines and visits to the doctor) and health-related devices (such as spectacles and hearing aids). It does not include expenditures on preventive care (e.g. vaccinations, check-ups etc.) or acute care and accidents (e.g. hospitalization) as the former were not covered and the latter were measured only over a short reference of thirty days. Including the latter would run the danger of making households with, for example, a medical emergency in the previous month appear uncharacteristically well-off. Education includes expenditure on schooling for children aged 6-16 years, and specific adult classes. It excludes university students and higher level secondary and technical school students as data were not collected on these items. 14. Durables are lumpy items, purchased infrequently. Including them in the estimate of consumption would exaggerate the long-run resources of the purchasing household and lead to misclassification of its poverty status. Ideally, however, we would like to include an estimate of the flow of services from the ownership of durables. There are a variety of ways of doing this. For the purposes of this report services from durables was estimated as the depreciation of the goods during the year of the survey. This was estimated from the net present value of the goods, which was calculated on the basis of the expected life of the goods, the age of the goods, and their actual value. The opportunity cost of owning the good (which should be set equal to the interest rate) was assumed to be zero. 15. Housing is typically a big ticket item in Western economies. The bulk of housing in Kazakhstan is privately owned. 1 Therefore evaluating the services from housing would largely involve imputing rent for owner-occupied housing. The rental market for housing in Kazakhstan is very thin, and where it does exist (e.g. Almaty) provides a questionable basis for evaluating the services from owner-occupied housing. It was therefore decided to exclude this item from the consumption estimate. Owing to data limitations, rent could not be included. Housing subsidy 16. The KLSS excludes the housing benefit which since its inception in mid-1996 has emerged as a substantial benefit received by over 150,000 households. The benefit covers a portion of monthly costs on utilities and maintenance, and goes not to the household but to the utility and/or the housing maintenance organization (zhek). Excluding the benefit from our estimates undoubtedly leads some downward bias in the estimate of household expenditure. However as the benefit is likely to have been small in July 1996 both because the scheme was in its infancy, and because of seasonal factors, we do not believe that exclusion of the housing benefit leads to undue bias. Choice of poverty line 17. In order to measure poverty it is necessary to specify a level of consumption below which individuals are considered to be poor. This can either be an absolute minimum, leading to an absolute poverty line, or a relative minimum, defined in relation to average national income or earnings, which would evolve over time. The advantage of an absolute poverty line is that it can be used to monitor trends in poverty with respect to an unchanging yardstick. In transition economies such as Kazakhstan I Over 90 percent of urban residents and 92 percent of rural residents live in their own homes (National Statistical Agency (1996), Living Standards Measurement Survey Kazakhstan, Final Report). 59

70 which have experienced sharp falls in the standard of living, the concern is not so much with the protection of relative standards but with the maintenance of certain minimum absolute standards. This would suggest the use of an absolute poverty line. 18. Defining an absolute poverty line is not as straightforward as it seems due to the numerous policy decisions that need to be taken as to what constitutes the minimum necessary for survival. No poverty line can be wholly objective. Two themes emerge from countries with a history of well-established absolute poverty lines:first, in order to command acceptance, a poverty line must have some grounding in empirical reality while at the same time being relatively simple to compute and explain (see Box 2.1). Transparency is important because the choice of a poverty line will invariably involve judgments which are open to question. Second, time spent agonizing over where exactly to draw a poverty line is often wasted as there is almost always a range of possible lines. Moreover, most of the critical policy decisions related to poverty reduction would be unaffected by the choice of any line within a plausible range. Box A1.1: The Tale of Two Poverty Lines The poverty line in the United States provides an example of a well established absolute poverty line. Derived in the early 1 960s by Orshansky (Social Security Administration), the first step in calculating the US poverty line was to take a low cost food plan' developed by the Department of Agriculture, the composition of which was observed to be consistent with the 1955 household survey. Because food represented only one third of the typical household budget in the 1955 survey, this food poverty line was multiplied by three to derive the final poverty line of around US$ 3,000 for a non-farm family in Deaton (1994) describes how the Orshansky poverty line became widely-accepted as a meaningful and empirically well-grounded poverty line. Yet he points out that this line is far from perfectly objective and non-arbitrary: the food plan itself was only one of several devised from the observed consumption patterns; the scaling of the food poverty line by a factor of three was based on an arbitrary choice of working with the food share at the mean of the expenditure distribution rather than at some other (equally reasonable) point. In fact it is not unlikely that the Orshansky poverty line received official sanction mainly because it happened to be quite close to a poverty line which was already in unofficial use within the US Federal Government at the time. The poverty line in India also dates back to the 1960s. It was originally based on an average calorie norm of 2,250 calories per capita per day for both rural and urban areas. Using National Sample Survey data on consumer expenditure, the Working Group of Eminent Economists (1962) estimated the average per capita expenditure at which households consumed 2,250 calories per person per day together with any such non-food items as they chose. An amended poverty line was endorsed in 1993 which essentially used the same methodology with the modification that a separate calorie cut-off was used for rural and urban areas (2,400 and 2,100 calories respectively). The poverty line in India seems to satisfy some notion of objectivity even if there exist many elements which are essentially arbitrary. For example, calories are taken as sole inputs into nutritional status, and there exist alternative approaches to incorporating non-food needs than the implicit approach adopted. The ultimate challenge in deriving poverty lines is to arrive at thresholds which receive broad endorsement. A line which has some empirical grounding clearly stands a better chance of acquiring that acceptance than a line which is simply plucked out of thin air. What also seems clear from the US and Indian examples, however, is that although established poverty lines are linked to certain objective criteria (such as specified nutritional standards and observed consumption patterns), there are many respects in which they remain essentially arbitrary. Yet the remaining elements of arbitrariness do not appear to undermine their credibility. Moreover, the relatively straightforward way in which they have been calculated is likely to have contributed to their intuitive interpretation. The question to be asked when thinking about defining a poverty line is whether it strikes the right balance between transparency and simplicity on one hand, and an adequate degree of objectivity and empirical grounding on the other. Source: Peter Lanjouw, The Purpose and Importance of a Poverty Line in Central Asia, in Falkingham et. al. (1997) Household Welfare in Central Asia. 60

71 Regional price adjustments 19. Kazakhstan is a large country with substantial spatial variation in prices both across regions of the country and between urban and rural areas. Failure to account for differences in costs of living could make households in urban areas, where prices tend to be higher, appear to be more well-off than they actually are. Likewise households in regions which experience higher prices may be misclassified relative to other households. 20. Price data on all major food items were available from the KLSS, collected on site at each of selected primary sampling units. These were used to calculate ten price indices, one urban and one rural for each region. The price indices are given by the regional cost of the minimum living standards consumption basket relative to its national value. These indices were used to adjust the consumption estimates. As no information was available through the KLSS on non-food goods, the assumption made was that non-food price relativities are similar to food price relativities. While this may not generally be the case (e.g. one would typically expect rural and/or agricultural regions to have low prices for food and related goods while having higher prices for industrial goods), food is an overwhelmingly large proportion of household consumption. We therefore expect any resulting distortions from the use of food price indices to adjust total consumption to be small. This is confirmed by examining poverty rates both with and without adjustments. 21. In Table AI.1 the Table A1.1: Poverty Rates Using Various Assumptions first column provides estimates of the poverty Head count ratio (H) rate using food price (1) (2) (3) indices to adjust total Using food price Using food price No consumption. In column indices to adjust indices to adjust adjustment (2), the indices are applied total consumption only food only to the food portion of Kazakhstan total consumption, while Center columnn (3) makes no West allowances at all for price East differences. The most South significant change comes from adjusting North: Kostanai, Kokshetau, Pavlodar, and North Kazakhstan oblasts. Center: consumption for food Zhezkazgan, Karaganda, Akmola and Torgai oblasts; West: Manghystau, Atyrau, i.e. wfhe Aktyubinsk, West Kazakhstan; East: Semi-Palatinsk, East Kazakhstan, Taldykorgan and price differences, I.e. when Almaty oblasts and Almaty city; South: Kzyl-Orda, South Kazakhstan and Zhambyl moving from column (3) oblasts In April 1997, the nineteen oblasts were merged into fifteen. to (2). The North appears significantly less poor and the South more poor as a result. (Households pay less on average in the North than the South for the 'subsistence minimum' food basket). When moving from column (2) to column (l), although there is a rise in the national poverty rate by two percentage points, there are no significant re-rankings of regions. As one might expect, the emerging regional pattern are reinforced. 22. Although adjustments were made for regional price differences, the same poverty line ( tenge per capita) was used throughout. 61

72 Choice of poverty measures 23. There is a large literature on measures on poverty. Three measures are used in this report which are the members of the a class of decomposable poverty measures proposed by Foster, Greer, and Thorbecke (1984). The are: (i) the head count ratio, H, which measures the extent of poverty, (ii) the poverty gap index, PG, which measures the depth of poverty, and (iii) P2, which measures the severity of poverty. 24. The head count ratio which is the simplest measure is given by the percentage of people for whom consumption per capita is less than the poverty line. If q people have consumption per capita below the poverty line and there are n persons in total, then n Although H conveys information on how many people are poor, it does not say how poor they are. It assigns the same weight to a poor person whether s/he has a level of consumption which is I percent of the poverty line or 99 percent of the poverty line. If poor people get richer over time but their consumption remains below a stationary poverty line, then H would give the misleading impression of no change in the condition of the poor. 25. In order to account for the depth of poverty we use the poverty gap index. It distinguishes among the poor according to how far below the poverty line their consumption falls. Ifyi is consumption of the ith poor individual and z is the poverty line, then PG = i (z _ y,) n j=] z PG measures the average poverty 'shortfall' in the population (the non-poor have zero shortfall) as a proportion of the poverty line Thus it is a measure of the depth of poverty. PG is sometimes used to compute the fiscal costs of eliminating poverty, assuming perfectly targeted transfers. The fiscal cost is calculated as the sum of poverty shortfalls, usually expressed as a percent of GDP. In practice, leakages and incentive effects would raise actual fiscal costs of poverty elimination through transfers. 26. The disadvantage of PG is that it is not sensitive to distribution below the poverty line. The P2 measure gives more weight to individuals who are further away from the poverty line. It is the weighted sum of the poverty gaps, where the weights are the poverty gaps themselves. 2 The unit of analysis 27. The unit of analysis in this report is the individual. Our measures of poverty, for example, pertain to individuals. Thus, the head count ratio indicates the percentage of people, not households, who live in 2 See Ravallion (1994), Poverty Comparisons. 62

73 poverty. Our data, however, provides us with estimates of household consumption. 3 In order to compute estimates of poverty, household expenditure per capita was attributed to each individual. This would appear to be relatively unproblematic, since what is being done is counting all individuals equally regardless of the size of the household they belong to. 28. However, attributing consumption per capita equally to all household members does incorporate the judgment that consumption is shared equally within the household, and that household members are equally well-off. This may not always be the case, and women, children or the elderly may enjoy a lower standard of living than other household members. In other words, there may be poor individuals in nonpoor households. Assuming that household consumption is shared equally within the household may thus lead to an underestimate of the overall degree of poverty. 4 However, in the absence of information on the intra-household allocation of resources (a problem in most countries not just Kazakhstan) it is not clear that anything can be done to investigate the matter further. The limitations of a snapshot 29. While the KLSS have major advantages over the FBS, including a representative sample and a survey instrument that permits the investigation of a wide range of issues to international standards, it provides a 'snapshot' at one point in time of living standards in Kazakhstan. Thus it does not permit us to observe trends in poverty, or to track the persistence of poverty among particular groups in the population. Experience from Russian and other transition economies suggest that not only does the percentage of the poor in poverty vary over time, there is substantial movement of households into and out of poverty. 30. It will be important to track the impact of policy reforms on living standards, as well as develop a better understanding of both those who a subject to transitional poverty as well aspersistent poverty in Kazakhstan. This report, and the methodological and analytical exercise that underlies it, should be regarded as a first step in developing the tools and techniques for such a policy endeavor. Timely, reliable data are a vital ingredient to this effort. 3 A household is defined as a group of individuals living together and sharing incomes and consumption 4There may also be non-poor individuals in poor households though this is likely to be less of a problem. 63

74 Appendix 2: Sensitivity Analysis 1. The figure below presents the cumulative density function (cdf) for per capita consumption. As can be seen from its slope, the cdf is such that small changes in the level (or value) of the poverty line lead to disproportionately large changes in poverty. In fact, a 10 percent increase in the value of the poverty line increases the proportion of the population in poverty to 40.7 percent, almost an 18 percent increase. Fig 4: Cdf of Per Capita Annual Expenditure, _ ~ - _ CD - ox 6/ C >~~~~~~ 74 m. - E.2 / Per Capita Annual Expenditure 2. Changing the value of the poverty line would affect poverty most in urban areas and the least in the rural ones. As can be seen in Table A2.1, a ten percent increase in the value of. the poverty line would increase the head count ratio in urban areas by 26 percent, from 30 to 37 percent. The same change in the poverty line would only increase poverty rates in rural areas by 11 percent. Essentially, there is more clustering around the SM in urban areas and towns than in rural areas. The table also includes rural settlements under a separate category. These are lumped together with rural areas in the main text. Table A2.1 Sensitivity Analysis of Poverty Measures to Change in Poverty Line % of Population In Poverty Urban Town Rural National Poverty at SM: If change in the value of SM: Increase by 5% Decrease by 5% Increase by 10 % Decrease by 10% Source: KLSS 64

75 3. Changing the value of the poverty line could change the poverty ranking of locations. Using the SM as the poverty line shows rural poverty to be higher than in either towns or urban areas. But as can be seen in the figure below, the rural cumulative distribution function crosses the town and urban lines at higher levels of expenditures. If the real value of the poverty line were just slightly above the 10 percent increase seen here, poverty would appear to be higher in towns than in rural areas. In fact, it is not clear whether rural areas and towns should be considered to have different levels of poverty. They are treated as one in the main text. The rural cdf crosses the urban cdf at a level of expenditure that is around twice the SM. At poverty lines below this level, poverty is higher in rural than in urban areas. Fig 5: Cdf of Per Capita Annual Expenditure, >-}:'-"u'a ) ' -of. o ).00 Per Capita Annual Expenditures Note: The vertical line on the left represents a poverty line set 10 percent lower than the SM The vertical line on the right represents a poverty line set at IO percent more than the SM. The middle vertical line is the SM (annualized). The distribution has been truncated to allow more detail around the poverty line. 4. Changes in the poverty line do not affect the ranking of regions to the same degree. As can be seen in the figure below, the slope of the curves varies among the regions and will lead to different percentage changes in poverty rates given any change in the value of the poverty line. But, unlike the urban, town and rural differences, there is a clearer pattern on povert-y in regions. The South has the highest poverty rate regardless of where the poverty line is set and the North the lowest. Additionally, except at the very ends of the distribution, the West has the second highest poverty rate. The crossing of the lines of the other two regions, however, makes it more difficult to rank them robustly. 65

76 Fig 6: Cdf of Per Capita Annual Expenditure, 1996 U) = Per Capita Annual Expenditure Note: The vertical line on the left represents a poverty line set 10 percent lower than the SM The vertical line on the right represents a poverty line set at 10 percent more than the SM. The middle vertical line is the SM (annualized ). The distribution has been truncated to allow more detail around the poverty line. 66

77 Appendix 3: Social Protection in Kazakhstan 1. This appendix provides a brief summary of the main programs of social protection in Kazakhstan, considering first social assistance followed by social insurance. Table A3.1 highlights the main programs, sources of finance and basis for benefit calculation. Table A3.1: Main social protection programs in Kazakhstan Sources of finance and eligibility rules Type of benefit Source Base Means-tested child Local budget For families with incomes < 2 base enumerate: allowance 1.25 x calculation base for each child < 3 years 1.30 x calculation base for each child between 3-18 years Means-tested Local budget, but Formula based on gap between actual expenditure on utilities housing allowance earmarked for and maintenance and 30 percent of household income for republican transfer maximal floor space. Labor pension Pension Fund Formula based on past earnings and work experience with adjustments for other activities and special circumstances. Minimum pension is floor, maximum pension is ceiling. In April 1996, minimum = 1452 tenge (3.8 x base enumerate), maximum = 3960 tenge, average = 2321 tenge. Unemployment Employment Fund Formula based on past earnings and duration of unemployment. benefit 3 x calculation base is floor. Average benefit in 1996 was 1736 tenge. Sick pay Social Insurance Fund Formula based on past earnings and work experience. Replacement rate is 100 percent for service of 8 years or more. Average annual usage in 1995 was 6.5 days. Planned usage in 1996 was 3.25 days. Maternity Social Insurance Fund Full wage continuation for a maximum of 126 days (140 days allowance for complicated or multiple births). Notes: Calculation base 1996: Ql: 320 T, Q2: 380 T. Q3: 460 T, Q4: 530 T; 1997: Ql: 550 T, Q2: 565 T, Q3: 585 T, Q4: 620; 1998: Ql: 630, Q2: 640, Q3: 650, Q4: 660. Source: MOF, MLSP Social Assistance 2. Child allowances: Child allowances tend to have a broader objective than poverty relief, however, the program in Kazakhstan is explicitly targeted at low income households and others who on categorical grounds may be poor (with the exception of allowances for children of those in active military duty). The total budget for these allowances as a share of GDP was 0.5 percent in 1996 and 0.25 percent in The means-tested benefit is the largest item of expenditure accounting for over 85 percent of all child allowances in 1996 and over 90 percent in Trends in expenditure on the main allowances is summarized in Table A3.2. The allowance for single mothers was merged with the means-tested benefit in

78 Table A3.2: Trends in expenditure on child allowances, (million tenge) All child allowances 170 2,073 6,196 8,000 4,519 Means-tested child allowances 157 1,871 5,481 7,100 4,227 Child allowances for single mothers Child allowances for active servicemen Allowances for disabled children Allowances for children with HIV.... Allowances in lieu of alimony Allowances for unemployed mothers with children < 7 years negligible Source: MLSP 3. The means-tested child allowance has increased steadily in real value over time facilitated by a restriction in the number of beneficiaries following the introduction of the income-test in The average value of the child allowance rose by nearly 150 percent in real terns between 1993 and Like all other payments from the budget, child allowances are subject to arrears. In some oblasts, child benefits are in arrears by several months. During the month of the KYLSS, around a third of eligible households received the means-tested child allowance. 1 Outstanding arrears on means-tested child allowances at the end of March 1997 amounted to over 4 billion tenge, including arrears of more than I billion tenge in South Kazakhstan oblast, 600 million tenge in Kzyl-Orda, and 400 million tenge in Zhambyl. Another problem has been the tendency to draw upon resources in the local Pension Funds to meet child allowances. This results in backlogs in pension payments. 5. Means tested housing allowance: Since the transition began, there has been a significant reduction in consumer price subsidies. All subsidies on food were eliminated late in In 1996, the authorities began to implement a policy of full cost recovery on all housing-related fees, including maintenance and repair, water and sanitation, and energy (heating and gas). Before 1996, the amount collected from residents for housing-related charges were estimated to cover between a quarter to a third of the cost of these services. This large, untargeted subsidy was a significant burden on sub-national budgets. 6. In order to cushion the impact of the increase in housing-related fees on the poorest segment of the population a targeted housing allowance scheme came into effect from May 1, The scheme had been piloted for over a year and a half in Semi Palatinsk. The philosophy underlying the scheme is that no household should have to pay more than 30 percent of its income in housing-related costs. The scheme consists of a discount from the amount that the household pays to the utility company or housing maintenance organizations (zhek), rather than a cash payment to the residents. Utilities and zheks are supposed to be reimbursed in full for these discounts by the oblast authorities. I Eligibility was determined by applying the incorne test to all households with children. See Chapter 3. 68

79 7. The introduction of an explicitly targeted housing subsidy in place of a general subsidy is a move in the right direction. However, there appear to be many respects in which the means-testing and administration of the housing subsidy need to be strengthened. By end of 1996, over 152,000 households were benefiting from the subsidy which was on average equal to a little over 1900 tenge per beneficiary household. The minimum benefit was 31 tenge while the maximum was over 14,000 tenge. The scheme has got off to an uneven start in regional terms. This may be to the detriment of households in poor oblasts where the reliance on the scheme is likely to be greater. By the end of 1996, for example, Kzyl- Orda and South Kazakhstan oblasts identified by the poverty profile as having very high poverty rates had the second and third lowest number of beneficiaries (1601 and 1539 beneficiary households respectively). It is also possible that specific design features may exclude many rural households as heating may rely on LPG, wood and other means for which the billing system does not apply. While it is early days yet, it seems clear that further research would need to be undertaken to establish whether the housing benefit is indeed helping the poorest households and what measures could improve targeting. A start could be made by closer examination of whether the characteristics of the beneficiaries correspond to the characteristics identified in the profile. Unfortunately, the timing of the KLSS means that there is no information in the survey on those receiving the subsidy. Other social assistance programs 8. In addition to the programs described above, there are other forms of social assistance provided to vulnerable groups by the local authorities on a case-by-case basis. The elderly and the disabled living alone have traditionally been the target of these programs but there are other groups as well. Note that the elderly living alone were not identified as a group especially vulnerable to extreme poverty in Chapter 2, so these benefits might be poorly targeted. Benefits consist of both cash and in-kind transfers. In-kind benefits include 'in-home' assistance, especially for the old and/or disabled who live alone, food, assistance with the purchase of vehicles (for the invalid) and medical devices. Many oblasts also target the 'low income' population though the basis for this targeting and the assistance provided is unclear. 9. Local authorities are also expected to bear the cost of social pensions which are paid to those who do not have sufficient creditable work experience to qualify for a labor pensions. The age of qualification for these pensions, 63 for men and 58 for women, is higher than the current retirement age. In 1996, there were 35,000 old-age pensioners drawing on such pensions which were equal to 1,072 tenge on average. In addition to old age pensions, there are social invalidity pensions, paid to those born with chronic conditions or whose work history does not make them eligible for standard labor invalidity pensions. In 1996, such pensions were worth 1,090 tenge on average, and were paid to 78,000 individuals. There is a small category of social survivor pensions which include dependents who do not qualify for full-fledged survivor pensions. These numbered 3,000 individuals in 1996 and received 1,027 tenge on average. 10. In terms of total expenditure, other social assistance amounts to a small fraction of expenditure on all social assistance and has fallen drastically in recent years. In 1995, these expenditures were 9.8 million tenge (15 percent of all child allowances), falling to 5.2 million tenge in 1996 (6.5 percent of all child allowances). 69

80 Social Insurance 11. The insurance function of social insurance in Kazakhstan has weakened because generous eligibility in the face of eroding contribution compliance has largely destroyed the benefit-contribution link. 12. Pensions: 2 Till 1998, Kazakhstan had an unfunded PAYG pension system financed by a 25.5 percent payroll tax. There are around 2 million pensioners. In addition to old-age pensions, the system provides disability and survivor's benefits. Like other transfer payments, pensions have fallen in real terms as a result of sporadic indexation. Pensions have also fallen relative to the average wage. The ratio of average pensions to average wages halved between January 1994 and December 1995 (falling from 0.56 to 0.27), after which it climbed slightly to 0.33 in June While the pension structure is compressed, there are some pensions which are relatively generous. Special services pensioners receive the highest pensions while Group III invalidity pensioners the lowest (excluding social pensions). According to official figures, about 9 percent of all pensioners (and 12 percent of early retirees) also earn a wage income. The KLSS estimates working pensioners to be 18 percent of all pensioners, Although pensions are the most regular of all transfers, there is significant regional and urban/rural differences in pension receipts. According to the KLSS, rural pensioners were 50 percent more likely to have pensions paid on time, while those in the South, East and Center were at a disadvantage relative to the other parts of the country. 13. The state pension system in Kazakhstan is in financial difficulties. This is not because replacement rates are high but because contribution compliance is low. Combined with fairly generous eligibility criteria, this has created a situation where the system is dependent on a narrow range of contributors. By 1996, there was one pensioner for every two contributors, compared to one pensioner for everyfive people of working age. In response to a growing inability to meet pension commitments, the authorities have allowed pensions to erode and benefits to be compressed. Taxation of the payroll is already at a high level. This forecloses the option of raising the payroll tax, even if it can be assumed that this would not lead to further erosion of compliance. The 1996 Pension Law raised the statutory retirement age. However, as it fails to restrict early retirement at favorable terms, it is unlikely that it will succeed in reducing the number of beneficiaries relative to contributors. The law also allowed for some increase in benefits. This appears to have precipitated the payments crisis early in By March, total overdue pensions amounted to T 26 billion, equivalent to 4 months of pension payments and 2 percent of GDP. 14. The government intends to stem the hemorrhage of contributors by closely aligning benefits to contributions through a system of mandatory private saving. The Pension Law of June 1997 mandates a shift from the current PAYG system to a privately managed, funded system akin to the one introduced in Chile in the 1980s. The new system will take effect from January 1, The accumulated rights of current pensioners and those who have paid into the current system would be respected. The current 25.5 percent payroll contribution would be split into two: 15 percent (declining to zero) going towards financing 'old' pensions, together with general budget resources, and 10 percent being accumulated with private pension funds to finance 'new' ones. The existing system will not disappear immediately but would be phased out over time. The role of the state under the new regime would be to (i) enforce mandatory participation in private pension plans; (ii) regulate private pension funds, asset managers, and custodian banks; (iii) guarantee a minimum level of pension ('old' and 'new' pensions combined); and 2 For a more detailed description see, Ministry of Labor and Social Protection-DAI (1997): The Pension System of the Republic of Kazakhstan. 70

81 (iv) provide certain types of insurance (disability and survivor), a poverty benefit for those with limited contributions. At the time of writing, many details of the new system remain to be worked out. Figure A3.2: Pension expenditures by category, June 1996 Survivor Social Other 9% 4% 1% Invalid III Invalid II11l Invalid 1 6%fE 1% Old Age 59% Old age- early 17% 15. Unemployment assistance: Unemployment benefits in Kazakhstan are paid by the Employment Fund. The main of expenditure in 1996 were unemployment benefit (56 percent) and training (14 percent). There is a small public works program. The program of assistance for the unemployed in Kazakhstan is small and unemployment benefits low. The majority of the unemployed receive the basic benefit. Those on training programs tend to receive lower than the average benefit. This weakens the incentives for people to enroll in these programs. In 1996, the ratio of average benefits to average wage was 20 percent. Many of the unemployed do not receive benefits: the KLSS suggests that two in five of those entitled to unemployment benefits actually receives it.3 The regional nature of financial arrangements in Kazakhstan means that receipt of the benefit is partly a function of where one lives. Many of the unemployed do not register, partly on account.of the low level of benefits. A rise in the minimum benefit in 1995 to 3 times the base enumerate was accompanied by a sharp rise in registered unemployment. 16. Although registered unemployment remains low compared to open unemployment, the Employment Fund is finding it increasingly difficult to meet its commitments. The Employment Fund is funded by a 2 percent tax on the payroll. In addition to a shrinking tax base, the financial position of the Fund has suffered from the failure to extend the payroll tax to the agricultural and budgetary sectors. In 1997, both sectors became tax-liable but at reduced rate of 1 percent. An additional problem appears to be that benefits continue well beyond the basic eligibility period (6 months). For people who will reach the age of retirement within two years, eligibility is extended by 3 months. Extended benefits may also be granted to persons who have been out of employment for more than a year. The benefits of workers who were employed for more than 20 years are extended by a month of every five years of employment. The system of unemployment benefit thus appears to function more like a system of social assistance rather than a labor market measure. Part of the, problem is that there is no system of income support or social assistance that people can fall back upon leaving the system of unemployment insurance to perform more than one function. 3 Eligibility is defined as those registered at the Employment Center and looking for work. See Chapter 3. 71

82 17. Sickness and maternity benefits: These are paid for out of the Social Insurance Fund. The SIF is funded by 1.5 percent tax on the payroll. In 1996, nearly two-thirds of SIF expenditure went towards sick pay and nearly a quarter was spent maternity benefits. Coverage rates are very low. According to the KLSS, less than 15 percent of eligible households receive sick pay, while less than 5 percent of eligible households benefit from maternity leave.' Among those who receive sick pay, the average payment, over 10,000 tenge per month (median: 4,000 tenge), appears quite high, compared to the level of other transfers. Sanatoria vouchers, which till 1995 were the second largest item, have shrunk considerably in recent times. Figure A3.3: Structure of SIF payments in Kazakhstan, 1996 ; ~~~~Birth all._il. 2% Sanatoria Funeral all. 10% Maternitya 19% X ( l F~~~~Sick pay 68% 4 Eligibility is self-reported. 72

83 Appendix 4: Economies of Scale in Consumption and the Demographic Profile of Poverty in Kazakhstan I. Household surveys, such as the Kazakhstan Living Standards Survey, typically yield most information at the level of the household rather than the individual. While many surveys include individuallevel data on the age and gender of family members (and sometimes on educational achievement, health and nutritional status, and economic activities of family members) most information on consumption and income typically pertains to the household as a whole. The ultimate interest in poverty studies, however, is in comparing welfare levels of individuals. This means that when we ask about average consumption levels per person, we have to adopt a rule for allocating a household's expenditures to the individuals which make up that household. By far the most widespread "rule" applied in practice is that all family members receive the samne fraction of household expenditure. Household expenditure is converted into household per capita expenditure which is then attributed to all members of the household. Thus individual well-being, y, is proxied by total household consumption, Y, divided by household size, N: y=y/n 2. The decision to proceed in this way is prompted not so much by a belief that all family members do in fact require the same share of the household's total resources in order to reach the same welfare level, or that household resources are in fact allocated equally across all household members, but rather because there is no widely accepted alternative to the simple equal-sharing convention. 3. But there is an important assumption implicit in the decision to work with per capita consumption: when household consumption is divided by household size, the implicit assumption is made that no economies of scale exist in household consumption. 1 Each additional member of the household reduces the per capita consumption level of all household members proportionately. 4. Is this assumption reasonable? Where might economies of scale come from? There are many reasons to think that the per capita cost of reaching a consumption level associated with a specific level of utility might decline as a household gets larger. For example, among the items typically consumed within a household, some fraction of these could be regarded as public goods. For example, the cost of a radio or water tap within a household will not vary whether one person or a number of persons use it. In a single person household, the cost to that person of consuming the services provided by the radio is equal to the price of the radio itself. In a two person household, the cost per person of consuming the same services, is only half that. As more family members consume radio services the per-capita cost of these declines accordingly. Presumably, at some point, there will be rising congestion costs, but these may not be significant before the family is very large. The degree to which household expenditures are devoted to I The use of a per capita measure also takes no account of differences in needs between individuals. These are usually corrected for on the basis of equivalence scales. We will not address such scales in this note, although a similar treatment to the one applied here could be used to investigate the impact of alternative equivalence scales on the poverty profile (see Buhmann et. al. (1988), Equivalence Scales, Well-being, Inequality and Poverty", Review of Income and Wealth, and Coulter et. al. (1992 ), Equivalence Scale Relativities and the Extent of Inequality and Poverty, Economic Journal.) 73

84 public goods as opposed to private goods is thus going to influence the extent to which there are economies of scale in consumption. In addition, the larger the household associated with a particular level of consumption on public goods, the lower the per capita cost of consuming the public good's services. 5. In transition economies, food is typically a large share of measured household consumption. And, generally, economies of scale are not associated with food consumption. This is one of the main reasons why the assumption of no economies of scale has traditionally been non-controversial. However, our interest is in total consumption (not just that which can be easily incorporated into an empirical measure of consumption) and a sizable proportion of total consumption is devoted to non-food items. Although clearly not all non-food items can be regarded as "public goods" or some elements thereof, there will be some association between the share of non-food expenditures and the presence of public goods. Moreover, there might even be some economies of scale aspect of food consumption. Often the cost per food item declines with the quantity purchased at a time. The presence of bulk-purchase discounts is not of itself sufficient to indicate the existence of economies of scale, but if certain important food items are perishable, and the cost of storing them is high, then large households may be better able to take advantage of the discounts than small households. In a similar vein, even if the consumption of food itself does not entail economies of scale, there may be some economies of scale in converting purchased or grown food items into prepared foods (e.g. through cooking time, or the use of cooking fuel) We test below the sensitivity of the results from the analysis of poverty in Kazakhstan (Chapter 2) with respect to economies of scale assumptions. It seems most useful, for this purpose, to focus on rankings within the consumption distribution rather than try to re-estimate poverty measures. 7. A simple manner in which to test for sensitivity is to postulate a relationship between individual welfare, y, and household consumption and household size, Y and N, respectively, as described above. We now add a parameter,o, which enters into the relationship in a non-linear fashion: y=y/no The term, 0, can be interpreted as a index of the degree of economies of scale which we assume to exist. 3 When 0=1, the equation reduces to our original formulation; individual welfare is best proxied by household per capita consumption. When 0=0, on the other hand, economies of scale are so pervasive that the best proxy for individual welfare is provided by total household consumption. Values of 0 between 1.0 and 0 can then be interpreted as different degrees of economies of scale. 8. In the table below we abstract away from issues associated with the specification of a poverty line. 2 Note that attempts to measure poverty based on subjective assessments in developed as well as developing countries have found evidence of considerable economies of scale (See Pradhan and Ravallion (1997), Measuring Poverty Using Qualitative Perceptions of Welfare, mimeo, Development Research Group, the World Bank for an example in developing countries). These studies indicate, for example, that a household with four members will typically report needing far less than four times the consumption level of a single-person household for its members to judge themselves at a comparable standard of living. 3 Lanjouw and Ravallion (1995), discuss the interpretation of 0 in greater detail, as well as the prospects of empirically estimating the true value of 0 for a given population. 74

85 We impose the assumption that 40 percent of the population is poor, irrespective of whether or not there exist economies of scale in consumption. 4 This assumption seems acceptable because it allows us to focus our attention on how our comparisons of poverty among different groups in the population might vary with different assumptions about 0. There are grounds for arguing that these comparisons are of greatest importance to policy makers concerned to identify sub-groups in the population who can be targeted for poverty alleviation purposes, or who are clearly under-represented among the poor. Because we can expect to see the greatest sensitivity to assumptions of economies of scale in demographic aspects of the poverty profile, we focus our attention on this set of household characteristics. 9. Table 1 provides a very simple "poverty profile" for Kazakhstan based on a selection of mainly demographic characteristics. There are some striking findings. First, when we assume that there are no economies of scale in consumption (0=1.0), pensioner households, defined as households with no males below the age of 59 years and no females below the age of 55, are far less likely to be poor than an average household (the incidence of poverty in the population belonging to pensioner households is 29 percent relative to 40 percent if an individual were randomly selected from the population). If we allow for only slight economies of scale (0=0.9), the incidence of poverty among pensioners rises to By the time we allow for a 0 of 0.7, pensioners are considerably more likely to be poor than an average person. 10. With no economies of scale, the incidence of poverty among households consisting of couples with young children (below the age of 16) increases sharply with the number of children. Thus, the incidence of poverty in the population belonging to households comprising couples with three children is 54 percent, with four children is 77 percent and with five or more children is 100 percent. As economies of scale are allowed to enter, the degree to which the number of children affects poverty rates declines. For example, with 0=0.7, poverty among households comprising couples with three children is 49 percent, with four children is 67 percent and with five or more children is 74 percent. While the number of children adds to the probability of poverty, the strength of this relationship declines as more economies of scale are assumed. With perfect economies of scale, it would make virtually no difference how many children belonged to the household. 11. The key factor in determining whether poverty associated with different household characteristics changes as a result of different assumptions regarding economies of scale, is whether those household characteristics are correlated with household size. We can see from Table I that the average household size of poor households declines monotonically as greater economies of scale are assumed present. Thus we can expect that the relationship between poverty and household characteristics which are strongly correlated with the size of households (such as whether or not we are looking at pensioners) will also change. 12. An interesting finding is that while the average child ratio among poor households declines with assumptions of increasing economies of scale, the average dependency ratio among poor households doesn't vary. The point here is that pensioner households are typically small households (so they become more poor as we assume economies of scale) while households with more children are typically large (they become less poor). The net effect, when we label both old people and children as dependents is that the average 4 To assess the impact of different assumptions of economies of scale on measured poverty rates one would, in fact, have to recalculate poverty lines for each assumed degree of economies of scale. Moreover, different poverty lines would have to be specified for households of different size categories. 75