Report No. AAA-38-KG SOCIAL SAFEY NET IN THE KYRGYZ REPUBLIC CAPITALIZING ON ACHIEVEMENTS AND ADDRESSING NEW CHALLENGES May 20, 2009 Human Development Sector Unit Europe and Central Asia Region Document of the World Bank CURRENCY EQUIVALENTS (Exchange Rate Effective June 1, 2008) Currency Unit = Kyrgyzstan Som (KGS) 40.97 KGS = US$1 FISCAL YEAR January 1 - December 31 ABBREVIATIONS AND ACRONYMS CASE Kyrgyzstan Center for Social and Economic Research CCA Commissions on Child’s Affairs CCT Conditional Cash Transfer CDS Country Development Strategy DPT Diphtheria, Pertussis, Tetanus DSP Department for Social Protection ECA Europe and Central Asia FCSD Family and Children Support Departments GDP Gross Domestic Product GMCL Guaranteed Minimum Consumption Level GoKG Government of Kyrgyzstan HCI Headcount Index HMT Hybrid Means Test JCSS Joint Country Support Strategy KIHBS Kyrgyz Integrated Household Budget Survey MTBF Medium-term Budget Framework MCB Minimum Consumption Budget MDG Millennium Development Goals MICS Multiple Indicator Cluster Survey MLSD Ministry of Labor and Social Development MOF Ministry of Finance MOH Ministry of Health MSB Monthly Social Benefit MT Means Test NDC Notional Defined Contribution NSC National Statistics Committee OECD Organization for Economic Co-operation and Development PMT Proxy Means Test TLU Tropical Livestock Unit UMB Unified Monthly Benefit UNICEF United Nations Children’s Fund Vice President: Shigeo Katsu Country Director: Annette Dixon Sector Director: Tamar Manuelyan Atinc Sector Manager: Kathy Lindert Task Team Leader: Boryana Gotcheva TABLE OF CONTENTS ACKNOWLEDGEMENTS .......................................................................................................................... 6 EXECUTIVE SUMMARY .......................................................................................................................... 7 CHAPTER 1. INTRODUCTION AND OVERVIEW ............................................................................ 16 Objectives and main issues ..................................................................................................................... 16 Methodology ........................................................................................................................................... 16 Structure of the report ............................................................................................................................. 17 CHAPTER 2. ECONOMIC AND SOCIAL CONTEXT AND POVERTY PROFILE .......................... 19 Economic and social indicators .............................................................................................................. 19 Poverty trends ......................................................................................................................................... 22 Where are the poor? ................................................................................................................................ 23 Who are the poor? ................................................................................................................................... 25 Conclusions ............................................................................................................................................. 28 CHAPTER 3. THE CURRENT SYSTEM OF NON-CONTRIBUTORY SOCIAL BENEFITS ........... 30 Current system of social benefits ............................................................................................................ 30 The Unified Monthly Benefit.................................................................................................................. 35 The Monthly Social Benefit .................................................................................................................... 39 Categorical in-kind subsidies and benefits.............................................................................................. 43 Conclusions ............................................................................................................................................. 45 CHAPTER 4. TARGETING PERFORMANCE AND POVERTY IMPACT OF CURRENT BENEFITS ........................................................................................................................ 46 Distribution of beneficiaries and benefits ............................................................................................... 46 Poverty impact ........................................................................................................................................ 51 Cost efficiency ........................................................................................................................................ 54 Differences in outcomes based on administrative and household survey data ....................................... 56 Conclusions ............................................................................................................................................. 59 CHAPTER 5. FURTHER STRENGTHENING THE UMB – OPTIONS FOR CONSIDERATION .... 62 Modifying the existing means-testing formula ....................................................................................... 62 Introducing indicators: categorical targeting and proxy-means testing .................................................. 63 Combining elements of a means test and a proxy-means test approach: a hybrid means test ................ 69 Improving the administration .................................................................................................................. 75 Conclusions ............................................................................................................................................. 83 CHAPTER 6. REFORMING THE OVERALL SAFETY NET – OLD CHALLENGES AND NEW OPPORTUNITIES............................................................................................................ 86 Monetization and phasing out of the categorical benefits....................................................................... 86 Using the safety net to respond to crises ................................................................................................. 89 Linking benefits to services .................................................................................................................... 95 Conclusions ............................................................................................................................................. 99 LIST OF LITERATURE .......................................................................................................................... 101 ANNEXES ………………………………………………………………………………………….103 Annex 1: Measuring poverty in the Kyrgyz Republic ......................................................................... 104 Table A 1:1. Extreme poverty indices for individuals by region and topography ........................... 105 Table A 1.2: Extreme poverty indices for individuals by oblast....................................................... 105 Table A 1 3: Extreme poverty indices for individuals by demographic characteristics.................... 105 Table A 1.4: Extreme poverty indices for individuals by composition of household ....................... 106 Table A 1 5: Poverty indices for individuals by composition of household ..................................... 107 Table A 1.6: Poverty indices for individuals by employment status of household head .................. 107 Table A 1.7: Extreme poverty indices for individuals by employment status of household head .... 107 Table A 1.8: Poverty indices for individuals by housing characteristics .......................................... 107 Table A 1.9: Extreme poverty indices for individuals by housing characteristics ............................ 108 Table A 1.10: Poverty indices for individuals by housing amenities and consumer durables .......... 109 Annex 2: Categorical benefits and subsidies: types, beneficiaries, financing ...................................... 110 Annex 3: Assessing the peformance of the social protection system. Data and methodology ............ 112 Data ................................................................................................................................................... 112 Methodology ..................................................................................................................................... 113 Table A 3.1: Marginal propensity to consume out of social protection transfers. Dependent variable: annual per capita consumption.......................................................................................................... 116 Table A 3.2. Benefit incidence of social protection benefits and private transfers, percentages, 2005: 100% substitution of consumption in the absence of transfers ......................................................... 117 Table A 3.3. Benefit incidence of social protection benefits and private transfers, percentages, 2005: 0% percent substitution of consumption in the absence of transfers ................................................ 117 Table A 3.4: Distribution of beneficiaries from social protection and private transfers, percentages, 2005: 100% substitution of consumption in the absence of transfers ............................................... 117 Table A 3.5: Distribution of beneficiaries from social protection and private transfers, percentages, 2005: 0% substitution of consumption in the absence of transfers ................................................... 118 Table A 3.6: Distribution of social protection benefits and private transfers, 2005: 100% substitution of consumption in the absence of transfers ....................................................................................... 118 Table A 3.7: Distribution of social protection benefits and private transfers, 2005: 0% substitution of consumption in the absence of transfers ........................................................................................... 119 Table A 3.8: Impact of social protection benefits and private transfers on poverty, 2005: 0% substitution of consumption in the absence of transfers ................................................................... 119 Table A 3.9: Relative change in poverty after social transfers in EU countries, 2005 ..................... 121 Annex 4: Estimating presumptive income from livestock (Tesliuc & Leite, 2008) ............................. 122 Table 4.1. Regression model for presumed livestock income........................................................... 122 Once the information on the increase livestock herds is collected from a household, the predicted (presumptive) income is estimated as the sum of the increase/decrease in the number of animals times respective coefficient: ........................................................................................................................... 123 Table 4.2. Example of predicted livestock income ........................................................................... 123 Annex 5: Methodology for a proxy-means test for the Kyrgyz Republic............................................. 124 Table A 5.1 Proxy-means indicators for household consumption per capita*.................................. 125 Table A 5.1 Proxy-means indicators for household consumption per capita*.................................. 125 Table A 5.2: Comparing actual and predicted poverty status ........................................................... 126 Annex 6: Tables simulations scenarios ................................................................................................. 127 Table A 6.1. Performance indicators simulated scenarios (coverage, inclusion, exclusion, leakage), percentages ........................................................................................................................................ 127 Table A 6.2. Poverty reduction impact different scenarios, percentages .......................................... 127 Table A 6.3. Average benefits and total program costs for the simulated scenarios ........................ 128 Table A 6.4: Summary of main privileges, comparing actual and legal allocations, 2008............... 128 Annex 7: Comparing HMT targeting approaches for the UMB .......................................................... 129 4 Tables Table 2.1: Kyrgyz Republic: progress towards achievement of the MDGs................................................ 21 Table 4.1: Benefit coverage of social protection benefits and private transfers, percentages, 2005 .......... 47 Table 4.2: Distribution of beneficiaries from social protection and private transfers................................. 48 Table 4.3: Distribution of social protection benefits and private transfers, 2005 ....................................... 48 Table 4.4: “Benefit Adequacy:� Share of benefits in total household consumption, 2005......................... 49 Figures Figure 2.1: Real GDP growth, IMF estimates (2007) ................................................................................. 19 Figure 2.2: Regional comparison of key economic and social indicators................................................... 20 Boxes Box 2.1 Concepts of poverty measurement ................................................................................................ 22 Box 3.1: Safety net reform vision and priorities ......................................................................................... 33 Box 3.2: Calculation of the UMB per household........................................................................................ 36 Box 5.1 Qualitative assessment of the trends in UMB administrative costs 2005 - 2008 .......................... 76 Box 5.2 Targeting Inspections to High-Risk Profile Groups ...................................................................... 82 5 ACKNOWLEDGEMENTS This report is the second output of two-year programmatic analytical work and technical assistance provided to the Ministry of Labor and Social Development and the Social Fund of Kyrgyszstan. It was prepared by a team consisting of Boryana Gotcheva (Team Leader), Franziska Gassmann, Katarina Szecsi Asbrink, Phillippe Leite and Stephan Siebert, under the direction of Kathy Lindert (Sector Manager, Social Protection). Peer reviewers are Aline Coudouel, William Wiseman and Emil Tesliuc. Elena Glinskaya and Hermann Von Gersdorff provided helpful comments on the report at different stages of its preparation. The study benefited substantially from feedback and contributions from representatives of the Ministry of Labor and Social Development, the Ministry of Finance and the Office of the President of the Kyrgyz Republic at different stages of the preparation of the report. The Bank team would like to extend its gratitude and appreciation to the Deputy Prime Minister and previously Minister of Labor and Social Policy Uktomkan Abdullayeva and her team for their excellent collaboration and partnership on all stages of the preparation of this work. The Bank team would also like to extend its gratitude and appreciation to the Minister of Labor and Social Policy Nazgul Tashpaeva for the excellent collaboration at the stage of the finalization of the report and its public discussion. The study incorporates findings from a qualitative study undertaken in the capital and Northern parts of Kyrgyzstan by the NGO CASE Kyrgyzstan under the leadership of Roman Mogilevsky. The report also incorporates ideas and agreements reached at discussions on the safety net reform directions held with representatives of UNICEF and the European Commission in Kyrgyzstan and Central Asia. The team would also like to recognize the useful inputs from joint analytical and advisory work, and two workshops held in June 2008. The first was jointly organized with the Government of the Kyrgyz Republic, the IMF, World Bank ECA PREM and HD sectors and was dedicated to the food crisis and possible responses with safety net instruments. The second involved World Bank teams from ECA and the Human Development Network, and consultants from RAND Europe and the UK Department of Work and Pensions and was dedicated on detection and prevention of error and fraud in the social assistance system. The team would like to recognize also the strategic recommendations and views on the safety net reform expressed by the representatives of the central, oblast and ayil okmotu level admoinistrations, the National Statistical Institute and other agencies during the closing dissemination events held in May 2009. 6 EXECUTIVE SUMMARY The present report was prepared upon the request of the Government of the Kyrgyz Republic to inform policy decisions on reforming the existing safety net. The report aims at providing analytical underpinning of the country’s ongoing safety net reform, along with venue for policy discussions with the Government and stakeholders on the immediate as well as on the longer-term challenges related to designing and implementing of a modern national safety net. The report’s focus is on the analysis of the poverty reduction impact, targeting accuracy and coverage of the existing non-contributory social benefits, the poverty-targeted Unified Monthly Benefit (UMB) in particular, with the objectives of recognizing the strengths on which the reform can capitalize, identifying opportunities for improvement and fostering consensus building among Kyrgyz policymakers regarding the options for reforming the UMB and the broader vision of public policy in social assistance and the feasible instruments for its implementation. The focus of the report on the UMB is also determined by the need of identifying safety net instruments for mitigating the adverse effects of the economic shocks which are already affecting ECA countries, including the Kyrgyz Republic, on the poor and vulnerable. The safety net in the Kyrgyz Republic faces multiple objectives: (a) to redistribute income to the poorest and most vulnerable, with an immediate impact on poverty and inequality; (b) to help households manage risks by maintaining consumption and assets in the outburst of shocks, such as the food, fuel, financial and economic crises; and (c) to protect the livelihoods of the negatively affected by structural reforms aimed at economic growth and efficiency gains. Poverty Context for the Safety Net in Kyrgyzstan The poverty reduction objective of the safety net is primary because large groups of the population continue to be poor and extremely poor despite gains in overall poverty reduction over the past decade. In 2005, over 43 percent of the population was living in poverty and 11 percent was not able to meet even their food needs. Moreover new risks of rising poverty and vulnerability are emerging following the increase in energy and food prices globally in 2007-08 leading to increase in domestic prices, energy tariffs and utility fees. The global financial and economic crisis is also starting to affect the country, with transmission channels playing out through financial markets (savings, assets, credit), employment, and product markets (slower growth of agricultural production and merchandize exports, contracting construction, changes in relative prices, inflation). Most notably, the crisis is affecting employment and incomes both at home and abroad, with particularly adverse impacts expected for migrant workers and remittances, given that income from remittances accounts for 17 percent of GDP in the Kyrgyz Republic. Kyrgyzstan has also been affected by adverse effects of climate change (such as severe winters and flooding) and other natural disasters, all of which are expected to exacerbate risks of increased poverty. The safety net also faces the challenge of responding to certain distinct patterns of poverty: • Child Poverty. Child poverty is very pronounced with 52 percent of the under-18 year old and 57.7 percent of the under-6 year old living in poverty and with 14.8 percent of the under-18 year old and 18.9 percent of the under-6 year old living in extreme poverty. The highest poverty rates are estimated for households with three and more children - almost 70 percent of individuals living in such households are found to be poor. • Large Families. Poverty is concentrated in large and multi-generational households – a half of the total poor live in households with four and more adults, while a half of the extreme poor live in big households of seven and more members. 7 • Spatial Dimensions. Poverty in Kyrgyzstan has distinct spatial and non-income dimensions. It is a predominantly rural phenomenon with over 50 percent of the rural population being poor, including close to 14 percent being extremely poor. Poverty incidence varies by regions with Batken having the highest rate of moderate poverty and Naryn having the highest rate of extreme poverty while in the Jalalabat region both moderate and extreme poverty are among the highest. Poverty is also higher for the households living in mountainous areas where the poverty rate is 60 percent. • Education and Other Characteristics. Poverty and extreme poverty are strongly positively correlated with the level of education of the household head, quality of dwelling, housing amenities and possession of consumer durables. The Overall Safety Net in Kyrgyzstan: Mix of the “Old� and the “New� Kyrgyzstan spends annually between 1 and 1.5 percent of GDP on social assistance benefits and services in the 2000s.This is lower than the average spending in ECA (25) which amounts to 1.7 percent of GDP but comparable with relative spending-to-GDP shares in countries of similar level of economic development. After 2005, spending is declining in relative terms, and in 2009 for the first time is expected to fall below 1 percent of GDP (0.84 percent as per the MTBF 2009-2011). The share of the spending targeted to the poor and vulnerable (the Unified Monthly Benefit, UMB, and Monthly Social Benefit, MSB) is falling- from 0.64 percent of GDP in 2008 to 0.49 percent of GDP in 2009. In relative terms, spending on UMB and MSB has been cut in half compared to the first half of the 2000s. At the same time, the spending on poorly targeted subsidies for certain categorically eligible groups – though also declining in relative terms - remains highly weighted against the number of beneficiaries and spending with similar objectives in other ECA countries. Table 1: Public expenditure on non-contributory social protection programs, 2007-2009 2007 2008 2009* GDP (billion KGS) 141.9 185.0 253.6 Social assistance / non-contributory social 1835.4 2137.2 2140.6 benefits as percentage of GDP (total) 1.29 1.16 0.84 - state benefits (UMB and MSB) 1060 1188 1234.7 as percentage of GDP 0.74 0.64 0.49 - categorical benefits and subsidies** 592.4 688.7 639.9 as percentage of GDP 0.42 0.37 0.25 - pregnancy and child delivery benefits, ritual 102.7 138.0 194.7 services benefits, health recovery activities, social care services, including in institutions as percentage of GDP 0.07 0.09 0.08 *Planned spending as per the MTBF 2009-2011; **Only republican budget. Source: MLSD and MOF The social safety net is weaved together with a patchwork of “old-style� categorical benefits (“privileges�) and two new targeted benefits. Non-contributory benefits are in the midst of a reform aimed at improving their targeting to the poor and vulnerable. The country inherited from the Soviet times a safety net consisting of a large number of categorical poorly targeted and costly benefits (“privileges�). In the middle of the 1990s, the government initiated reforms aimed at simplifying the benefit programs, reducing the overall costs and reaching the neediest of an increasingly impoverished population. Hence, the present non-contributory part of the social protection system in Kyrgyzstan is a mix of “new� and “old� programs. 8 On one hand, the safety net includes a plethora of categorical benefits (“privileges�) 1 that represent a legacy-of-the-past. These include cash benefits and in-kind subsidies for categories of citizens who are not necessarily poor. Spending on categorical benefits and subsidies although declining is still significant, at 0.42 percent of GDP in 2007 and 0.37 percent of GDP in 2008. On the other hand, two new targeted cash transfer programs were introduced after independence in 1991. These include: the Unified Monthly Benefit (UMB) and the Monthly Social Benefit (MSB), which together account for 0.74 percent of GDP in 2007 and 0.64 percent of GDP in 2008. Spending on UMB and MSB is also going down to less that 0.5 percent of GDP in 2009. • The UMB was introduced via a Presidential Decree in January 1995 as part of an important step in reforming the old-style safety net. After several amendments and improvements on how to assign the new benefit, the UMB was finally regulated in the Law on State Benefits, approved by the Kyrgyz Parliament, the Jorgorku Kenesh, and signed by the President of the Kyrgyz Republic on 5 March, 1998. The UMB is a last-resort poverty-targeted variable cash benefit program that is mean-tested and categorically-targeted to children from 1.5 to 16 years of age (or up to 21 years of age when still studying) from low-income families. The UMB also includes payment of a fixed birth grant and allowance for children under 1.5 years old. In 2007, 97 percent of the UMB was allocated to children under 16 years of age. According to household survey data, UMB is received by 14.6 percent of the population. According to MLSD administrative data, in 2007 UMB was received by 145.1 thousand low income families (18 percent of all families with children) with 421 thousand children. In 2008, UMB was received by 387 thousand eligible beneficiaries. • The MSB, was also established by the Law on State Benefits, and is a cash income-replacement program that is categorically targeted (but not means-tested) to disadvantaged groups, including children with disability up to 18 years of age and other categories of people with disabilities, also orphaned children, mothers of large families and elderly who do not qualify for pensions from the social security system. According to household survey data, MSB is received by 6.5 percent of the population. According to administrative data, in 2008, close to 60 thousand people were receiving MSB. Performance of the UMB: Strong with Room for Improvement The current UMB performs well compared to poverty-targeted programs in other countries in Eastern Europe, and taking into account the country situation (large agricultural sector, high level of subsistence agriculture, informal economy, remittances). It is a progressive transfer (larger share received by the poor), although performance in 2005 declined compared to 2001. The UMB is also relatively cost-efficient transfer compared to other social transfers. The UMB is fairly accurate in channeling benefits to poor, who account for 43 percent of the population: • In terms of targeting accuracy (benefit incidence with the population ranked by per capita consumption), 38 percent of the total benefit amount goes to those in the poorest quintile, and another 43 percent goes to those in the next quintile, such that the poorest 40 percent of the population receives 81 percent of total benefits paid out. Leakages to the non-poor are low: only 19 percent of benefits go to those in the top three quintiles; • In terms of beneficiary incidence, close to 75 percent of the UMB beneficiaries belong to the poorest 40 percent of the population; • In terms of program compliance (ranking the population according to per capita income calculated according to the methodology used to administer the program), the UMB performs 1 These programs are referred to as ‘privileges’ (l’goti) in some of the quoted literature. 9 even better: 60.5 percent of the beneficiaries belong the poorest 20 percent based on administrative income per capita receiving 77 percent of total UMB benefits; • In terms of program coverage (based on per capita consumption) UMB is received by 28.2 percent of the poorest quintile and 25.6 percent of the second poorest quintile. This coverage is not particularly strong and implies leakage of part of the benefit amount to non-poor. 2. • In terms of benefit value, the UMB is not a generous transfer. The value of the Guaranteed Minimum Consumption Level (GMCL) which determines UMB eligibility is lower than the extreme poverty line. The UMB accounts only for 7 percent of total household consumption in poorest households receiving the UMB. The UMB reduces the extreme poverty rate from 11.3 to 10.9 percent, equaling to a relative reduction of just 3 percent. While the impact with respect to the absolute poverty line is limited, the UMB is more successful in reducing the extreme poverty gap - by 12 percent. • In terms of cost efficiency, poverty reduction is achieved with relatively lower UMB transfers. The UMB costs 1.6 and 4.7 KGS to reduce the absolute and extreme poverty gap by 1 KGS respectively (administrative costs are not included). Table 2 below presents a Summary Scorecard of the main indicators of program performance across the non-contributory benefits in the Kyrgyz safety net. Pensions are included for comparative purposes and illustration that higher coverage and more adequate benefit level could increase substantially the poverty impact even when the main purpose of the program is to avert old-age risk by providing replacement of work-related income. The comparison across safety net programs indicates the strongest performance of the UMB compared to the MSB and especially to other benefits. The MSB performs relatively well. It is a smaller program in terms of beneficiary incidence and budget outlays but has a higher poverty impact in terms of relative reduction of the extreme poverty rate and comparable accuracy of targeting the poorest quintile. Similarly to the UMB, the MSB accounts only for 7.6 percent of total household consumption in poorest households receiving the MSB, however its contribution to relative reduction of the extreme poverty rate is higher - 6.4 percent. In terms of targeting accuracy, the MSB without being explicitly targeted to the poor, delivers one third of total MSB amount to those in the poorest quintile. At the same time the leakage to the non-poor is high with close to 30 percent of MSB going to those in the top two quintiles. 2 When assessing the UMB coverage gap, one should bear in mind that the UMB is a last resort program that complements a number of other, some of which more generous, social protection programs, notably pensions, and to a lesser extent other social insurance benefits, and scholarships. These benefits are received by 66.2 percent in the poorest and by 56.9 percent in the second poorest quintile. Any kind of contributory or non-contributory social transfer is received by 74.1 percent of the poorest quintile and 61.8 percent of the second poorest quintile receive any social transfer, and. Consequently, those who are not covered by any social protection program represent 25.9 percent of the poorest 20 percent of the population. Part of this consists of households where the household members are not entitled to pensions, disability benefits or categorical benefits and subsidies. Other part consists of childless households. In both cases these households are not eligible for social benefits by program design. There are also households which meet the eligibility criteria but do not receive UMB. 10 Table 2: Social Safety Net: Summary Performance “Scorecard� Program Spending / Coverage Targeting Benefit Poverty Poverty Cost share of share of accuracy / generosity / impact / impact / efficiency/ GDP poorest Q share of share of relative relative cost per 1 (%, 2007) (%, 2005) benefits benefit in reduction reduction KGS captured consumpti of extreme of extreme reduction of by Q1 (%, on (%, poverty poverty poverty and 2005) 2005) gap (%, rate (%, extreme 2005) 2005) poverty gap UMB 0.53 28.2 38.0 7.2 11.6 3.1. 1.6 / 4.7 MSB 0.21 12.7 33.3 7.6 6.3 6.4 3.0 / 5.5 Categorical benefits and 0.6 16.3* 12.6* 1.1* 0.8* 0.5* 6.3 / 27.3 subsidies Pensions 5.1 55.5 28.5 25.0 47.8 32.6 3.3 / 10.2 *Only for utilities and housing subsidies. Source: Staff calculations based on KIHBS 2005 and administrative data from MOF, MTBF. The relative performance of the UMB summarized in Table 2 also suggests that means testing can work in a low income country context. Nonetheless, it also signals the program’s main weak points, including: (i) low coverage and, respectively, significant errors of exclusion and (ii) low benefit generosity leading to a limited poverty impact. With the existing foundation of respectable performance, there is still room to further strengthen the UMB performance. This can be done by expanding coverage / reducing the exclusion errors, confining leakage to the non-poor and by increasing the level of the benefit (benefit adequacy or generosity) which would eventually strengthen its poverty reduction impact. • Coverage could be expanded through improved outreach methods and active involvement of social workers in information dissemination and rapid assessment of the poverty and vulnerability in order to reach more eligible households. Social Passports can be used for mapping all poor and vulnerable, for monitoring vulnerability and for policy planning. Coverage could be also improved with enhanced fiscal efforts (preferably with funding coming from downsizing and consolidation of other non-targeted to the poor social programs). Furthermore, coverage could be increased by extending eligibility for a UMB also to poor households without children. • Finally, by allocating more resources, the benefit value of the UMB could be raised thereby increasing the impact on poverty reduction. Improving the UMB: targeting and implementation matter There are a number of measures that could further strengthen the UMB in terms of its targeting, effectiveness, efficiency and impacts in addition to the actions aimed at improving coverage and benefit adequacy. Targeting alternatives involve (i) improving the design of the existing targeting mechanism (the means test), or modifying it more specifically by including in the overall household income the presumptive income from livestock – a type of hybrid means test (HMT), and (ii) considering replacing the current combination of means test and categorical targeting with alternative targeting methods: (a) a purely categorical approach to targeting (universal child benefit), (b) using a set of indicators / determinants of household poverty status - proxy-means testing, and (c) a combination thereof. The choices of specific alternative approaches for consideration are driven by the country’s medium-term social policy objectives conveyed by the Country Development Strategy and Social Development Concept. These strategy papers call for increasing the generosity of the delivered non- 11 contributory benefits while at the same time improve the methods for targeting the poor so that the errors of inclusion are reduced and more space becomes available for including eligible for UMB beneficiaries. This study ‘translates’ the vision and objectives into possible targeting approaches which have been used by other countries in similar circumstances and models their targeting accuracy, benefit coverage and poverty impact. The choice of considered options also draws on policy proposals launched by development partners and representatives of the academic community that look at options for reducing the errors of inclusion and exclusion, and have been the subject of public discussion in recent years. Last but not least, the choice was influenced by the increasing risks of worsening poverty at the outset of the economic crisis and the mounting need of preparing the safety net for providing quick and adequate response to this risk. Improving / modifying the existing means testing instrument. Modifying the design of the existing UMB means test by adding presumptive income from livestock may marginally improve the targeting efficiency of the UMB program. Using the existing instrument and definition of income under the UMB program, one would correctly identify 56 percent of the poorest 10 percent households as compared with the yardstick of full household income estimated in the household survey. This compares to 60 percent of the poorest decile under the revised UMB income specification that incorporates imputations for income from livestock. A comparison across three HMT modifications of the targeting options for the UMB indicates that the option which combines the administrative income assessment with presumptive income from livestock would be relatively stronger in terms of targeting accuracy and coverage. However, the overall performance would be dependent on how the HMT is implemented. The cost of establishing, implementing and regularly updating the methodology for accounting for livestock income has to be considered against the potential benefits. Moreover, the more precise accounting for household income can reduce the error of inclusion without necessarily contributing to reducing the error of exclusion which is a much bigger concern for the UMB targeting. Alternative targeting instruments. Given the Government’s emphasis on improving targeting as a policy objective, the performance of three alternative targeting instruments – categorically universal child benefits, proxy means testing (PMT), and a combined PMT-child benefit – were simulated and compared with “perfect implementation� of the existing UMB means testing instrument.3 Importantly, the analysis suggests that it would be difficult to outperform “perfect implementation� of the existing instrument. Compared with the means test for the UMB when operating under optimal circumstances, the three alternative targeting options (universal child benefit, PMT and their combination) display higher inclusion error and leakage of benefits to non-poor. Only the universal child benefit is comparable in terms of coverage of the poorest 20 percent of the population. These findings make a case that means testing can work better than targeting with indicators in the context of a low income country and provide strong argument in favor of focusing policies on improving the current targeting instrument, as well as the implementation of the UMB program. Going forward, one way of building on the strengths of the UMB targeting methodology would be to combine it with additional screening instruments / indicators – a HMT with the objective of improving targeting and reducing leakage to non-poor. The proxy indicators would be selected taking into account two criteria: (i) positive correlation between the per capita consumption and the predictor, which will define the accuracy of the prediction and/or verification, and (ii) verifiability of the predictor which will determine the accuracy of the information used to estimate the per capita consumption. In Kyrgyzstan, the HMT approach could be applied in several ways: (i) as an integral part of the UMB program eligibility criteria to determine / predict which families will be accepted in or reject from the program; (ii) as a verifier of the eligibility of UMB program beneficiaries already receiving the UMB for inspection and audit purposes and eventual detector of error and fraud in the system; and (iii) as an identifier of poor households which are subject to exclusion when only formal income is considered. 3 Since the results of the simulated scenarios cannot be compared with the actual performance of the UMB as they assume perfect implementation and 100 percent take-up, the report uses an optimal UMB as the counterfactual or theoretical benchmark for the alternative scenarios. 12 In parallel with improving the UMB targeting methodology and design, improvements in the UMB program implementation and administration are equally important for program performance, and especially for better outreach to the poor. The main strong features of UMB implementation are on- demand applications, using community-level knowledge in combination with administrative targeting procedures, systematic verification of eligibility, regularly scheduled recertification procedures. The UMB targeting is implemented in a transparent and uniform throughout the country manner. Nevertheless, there are ways that the implementation of the UMB could be strengthened for improved program outcomes and efficiency. Areas for improvement include but are not limited to adequate funding of program administration, adequate staffing of social protection departments, provision of regular training and methodological support to field social workers, better case management and beneficiary tracking with a management information system and electronic databases; limiting the scope for discretionary decision-making and collusion when deciding on UMB eligibility and payment amounts; and reducing the susceptibility of the UMB design and administration to errors and fraud. Another area of UMB improvement is the computerization of the Social Passports of low income families which – if filled and updated in a timely manner - can be used for mapping all poor and vulnerable, for monitoring vulnerability and for planning of social interventions at local and central government level. Improving the safety net: old challenges and new opportunities There are old challenges and new opportunities for reforming the safety net in the Kyrgyz Republic within the existing benefit system but also going beyond the design and implementation of the current non-contributory benefits. Addressing them with adequate policies could bring savings which are very much needed in the context of the pressure for budget constraints coming from the emerging economic crisis. Dealing with them could contribute to improving the effectiveness of the safety net programs as instruments for immediate response to the crisis to protect the consumption of the poor and vulnerable, and could also protect and promote the human capital. Within the existing safety net, the UMB has the potential to serve as an enhanced tool for social policy, including for: (i) using the UMB (and also the MSB) to further consolidate the safety net (merging other programs, mostly certain monetized categorical benefits and subsidies, into it); (ii) linking the UMB to human capital incentives (conditional cash transfers); (iii) linking UMB and MSB beneficiaries to social services; (iv) introducing work-related aspects, such as work requirements for able-bodied members of families which receive the UMB and/or links to job support services (job search, training, etc.). Last but not least, the UMB can be considered for delivering additional support in response to the food, fuel and financial crises in the short run, and for more effective poverty risk mitigation in the context of the developing economic crisis. The two ‘new’ safety net programs – the UMB and MSB - are well suited to facilitate the reform of the poorly targeted categorical benefits and subsidies. This reform is on-going, and most of the benefits are already monetized. The next steps would involve abolishment of certain already irrelevant benefits and – more importantly – targeting or the remaining ones to those who are poor and vulnerable. Targeting can be achieved by using the UMB means test approach to the current beneficiaries and including those who remain under the income threshold in the UMB program. For the beneficiaries with disability, categorical test might apply to place them in the MSB program. The resources ‘freed up’ from the poorly targeted benefits would strengthen financially the UMB and MSB programs. In the short run, the UMB and MSB programs are well suited also for providing safety net response to increased poverty risks triggered by the food, fuel and financial crises. Due to their fairly well- working targeting mechanisms, the UMB and also the MSB are well-positioned to redistribute income to the poor and vulnerable and channel additional resources. • The UMB is the single means-tested program with national coverage which is delivered in cash on a monthly basis and under defined conditions for flow of funds. The MSB is the only cash 13 support for a distinct category of people with specific vulnerabilities – reduced capacity to work certified with medical assessment, and also poor or nearly poor. At time of crises both programs can be easily scaled up to protect incomes, help avoid irreversible losses of physical assets and human capital. • Due to design features, the UMB can automatically expand during crisis to cover more beneficiaries and serve as counter-cyclical stabilizers. The drawback would be that the expansion will not cover poor who live in families without children. On the other hand, since the UMB targets only children, the risk of creating disincentives to work is and will continue to be negligible. • In fact, the UMB has already been tested as a response mechanism to the food crisis. The monthly benefits paid out under this program are being temporarily “topped-up� in value to help compensate for food price increases (first with World Bank support – the Global Food Response Program, afterwards to be sustained with possible support by other members of the development community or the state budget). The benefit scale up was complemented by a program for micronutrients delivery to especially vulnerable groups of the population. Similar type of scaling up the MSB can be applied to protect the consumption of the MSB beneficiaries. • These policy actions require close monitoring of program outcomes and impacts. During crisis, targeting is more challenging – family circumstances change more often, the informal economy is expanding, and the profile of the poor changes. They also require increased ability for rapid identification of those who are most in need of support and for effective outreach. The UMB and MSB are equally well suited to mitigate the increasing risks and vulnerabilities following from the economic crisis. As already mentioned, they both deliver one third or higher share of the benefit to the poorest quintile. At the same time they both suffer from low coverage and low benefit adequacy: the amounts they transfer are not sufficient to enhance consumption substantively. The crisis response capacity of the UMB and MSB can increase first and foremost if benefit coverage and generosity expand counter-cyclically. Program outcomes could improve through greater outreach to those who are eligible but not receiving UMB and MSB; improved program design to reduce the errors of inclusion to free up space for eligible beneficiaries; and strong program implementation. The social assistance system can start designing and subsequently piloting design and implementation innovations which could improve the performance of the UMB. The human capital outcomes of social assistance would increase if the increased cash transfers or the top-ups only (UMB in particular) are made conditional to positive behavioral changes (regular school attendance, health and nutrition interventions, health checks and immunizations). Last, but not least in the crisis context closer monitoring and analysis of UMB and MSB program indicators by policy makers would provide early warning signals about changing household welfare, profile of poverty and vulnerability, and inform policy making. Again within the existing safety net there is room for coping with an ‘old’ challenge – the reform of the in-kind benefits and subsidies which are a legacy from the Soviet past, are not targeted to the poor and do not contribute to an effective and efficient safety net. The process of monetizing these benefits has already started and the reform can go deeper with more radical approaches. The study proposes four approaches which do not exclude but could rather complement each other: (i) grandfathering / phasing out over time following natural attrition of the remaining beneficiaries combined with gate keeping to the granting of new entitlements and reconsidering the relevance of certain benefits vis-à-vis the needs of recipients; (ii) consolidation through provision of one lump sum / single cash benefit instead of several fragmented entitlements going to the same beneficiary category without undermining the overall amount of the cash received before the reform; (iii) integration / mainstreaming through including the in-kind benefit into the regular pension or MSB where applicable as a one-time increase, and stop existing as a separate benefit; (iv) scaling down through (a) abolishment of certain already irrelevant categorical benefits and (b) targeting of the remaining benefits through a means test of the existing beneficiaries and 14 shifting those under the threshold to the UMB and a categorical test for disability and inclusion in the MSB program. In the medium-term, additional impact on the poor could be achieved through linking cash benefits to social care and / or employment and activation services which are still limited in scope. Social care services exist now mainly in the form of institutional care and home-based support for elderly and people with disabilities (paying of bills, purchase of food and medicines, housekeeping). These services are provided by the state and local governments, and on a very limited scale – by NGO providers. Recently the GoKG established the legal framework for licensing of non-government providers for the provision of social services. As a part of the safety net reform, licensing and / or accreditation of providers needs to be put in place along with standards for quality of services and mechanisms for financing the non- government providers in a sustainable manner. 15 CHAPTER 1. INTRODUCTION AND OVERVIEW OBJECTIVES AND MAIN ISSUES The present report was prepared upon the request of the Government of the Kyrgyz Republic to strengthen the analytical underpinning of the country’s ongoing safety net reform. The report analyzes the poverty reduction impact, targeting accuracy and coverage of the existing non-contributory social benefits, with a particular focus on the poverty-targeted Unified Monthly Benefit (UMB) with the objectives of recognizing the strengths on which the reform can capitalize, identifying opportunities for improvement and fostering consensus building among Kyrgyz policymakers regarding the vision of public policy in social assistance and the feasible instruments for its implementation. The report assesses the performance of the non-contributory social benefits - the UMB and to a lesser extent the Monthly Social Benefit (MSB) - against empirical indicators of performance as coverage, adequacy and benefit incidence; and the existing delivery and targeting mechanisms. More specifically, the report answers the following policy questions: • Who and where are the most poor and vulnerable groups in the Kyrgyz Republic who are in need of social assistance? • To what extent, do the current social benefits - the UMB in particular - reach the poorest and most vulnerable groups and reduce poverty? • How much does the current UMB program cost in terms of administration costs and costs for applicants? • Given budget and administrative constraints – and low coverage and benefit levels -- what potential policy options could be considered for improving and modifying the targeting effectiveness of the UMB and for improving its poverty reduction impact? • In what areas and how the UMB administration could be improved to reduce administrative costs, enhance the outreach to the poorest and hardest to reach and reduce the risks and likelihood for errors and fraud while assigning and delivering the UMB? • To what extent and under what conditions the UMB can be used for mitigating the poverty risks arising from the increase in the world food and energy prices? • What would be the medium-term perspective for reforming the safety net to cope with existing problems and make use of new opportunities? What are the opportunities for broadening the safety net and enhancing its role in human capital development by linking the UMB provision to social care and activation services. METHODOLOGY The report takes a comprehensive view of the social safety net in the Kyrgyz Republic that links multiple areas rather than addressing them in fragmented fashion. This approach implies (a) reviewing a broad range of social protection programs (but with an emphasis on social assistance benefits); and (b) integrating the report findings and hypotheses in broader discussions and considerations about the growth, fiscal and poverty impact of the global food and energy prices increase on Kyrgyzstan. A range of data sources and methodologies were used – including qualitative and quantitative methods such as: • Quantitative analysis. The main data source for the report is the most recent (in terms of data readiness for analytical purposes) Kyrgyz Integrated Household Budget Survey (KIHBS) fielded 16 in 2005. The quantitative analysis provides information on the annual household consumption per capita after transfers, the counterfactual consumption in the absence of transfers, the annual benefit values and national poverty lines. The scope of the analysis is extended to coverage (benefit incidence), distribution and adequacy, and poverty impact (poverty headcount and poverty gap). More detailed description of the data, methodology and sequence of the quantitative analysis of the performance of the non-contributory social benefits is presented in chapter 4; • Analysis of statistical and administrative data provided by the Ministry of Labor and Social Development and the Ministry of Finance on program scope and trends (types of programs, numbers of beneficiaries, size of individual benefits, etc.), analysis of the MBTF, the individual program budgets, including financing by sources and programs, of the adequacy of funding and funding flows; • High level meetings with policy makers, including the Minister of Labor and Social Development and the Minister of Finance, the Office of the President, to discuss issues and preliminary report findings, and to conceptualize solutions and actions towards improving the design, management and effectiveness of the system; • Field observations of the work of social assistance offices and social workers (work load, time spent on different stages of program implementation, field work and outreach, problems encountered in every day work, program design and implementation issues from the perspective of social workers, reasons for non-rightful inclusion and reasons for exclusion, perceptions of the scope of exclusion and inclusion errors in their constituencies), as well as ad hoc analysis of primary household level data collected through social passports; • Qualitative survey (focus groups and in-depth interviews) with UMB program beneficiaries, applicants to whom the UMB was not assigned and / or whose benefit provision has been terminated, recipients of social services, policy makers, benefit administrators and social workers to assess their perceptions of the access to, and the quality of the safety net, and to test attitudes to prospective policy changes. The findings from the qualitative survey are summarized in a background report to this report; • Review of legal documents – laws, regulations and instructions – governing the design and organization of the safety net, the identification of beneficiaries, determination of benefit eligibility and benefit levels, implementation regulations, monitoring and control; • Summary of evidence of program impacts (where available); • Desk review of the status of the social service provision, including study of documents related to individual projects implemented by NGOs under the framework of ‘state order’, and MLSD reports. Review and use of available research on the Kyrgyz social protection system, including World Bank poverty assessment reports and comparative studies on the coverage, effectiveness and administrative costs of social assistance in ECA; reports prepared by CASE Kyrgyzstan, UNICEF, the European Commission and other sources. The report was prepared in a consultative manner with the Ministry of Labor and Social Development (MLSD), preliminary findings were discussed with the Minister of Labor and Social Development and the Minister of Finance and their teams. The preliminary findings were discussed and the final report results were arrived at after two detailed presentations to ministerial staff and social workers from Bishkek and Chui, and to the oblast administration and deputy mayors responsible for social affairs in the Osh oblast. STRUCTURE OF THE REPORT 17 This report is organized in the following way. Chapter 2 analyzes the macroeconomic and social context of the safety net, and answers key empirical questions for social policy: how many, and what kinds of, people are poor and extreme poor in the Kyrgyz Republic. Chapter 3 provides a synopsis of the current system, including an overview of existing social benefits, legislative background, objective, design and administration, coverage and generosity based on administrative data, as well as the budget envelope and financing mechanisms. Chapter 4 analyzes the targeting performance and poverty impact of the current benefits, in particular the distributional impact in terms of beneficiaries and benefits, the impacts of various transfers on poverty and the cost efficiency of the impact of these transfers, as well as the differences in outcomes based on administrative and household survey data. Chapter 5 simulates a number of targeting options for the UMB, including modifying the existing means testing formula and alternatives based on introducing of indicators (categorical targeting and proxy means testing) with the objective of improving its effectiveness and efficiency. The chapter also discusses issues related to of UMB administration. Chapter 6 outlines challenges and new opportunities for reforming the safety net as completion of the monetization and phasing out of privileges, using the safety net for protecting the consumption of the poor in the food, fuel and financial crisis and the onset of the economic crisis in the Kyrgyz Republic. It also discusses the ongoing and envisaged reforms aimed at improving the access to, as well as the scope and quality of social services and linking the provision of benefits to services and investment in human capital. Each chapter concludes with summary of findings and policy recommendations. 18 CHAPTER 2. ECONOMIC AND SOCIAL CONTEXT AND POVERTY PROFILE This chapter analyzes the macroeconomic and social context of the safety net focusing on the economic growth trends, the progress towards the achievement of the millennium development goals (MDGs), and the trends in poverty with the objective of answering key empirical questions for social policy as how many, and what kinds of, people are poor and extreme poor in the Kyrgyz Republic; how poverty and inequality change with economic growth; where poverty remains most severe and what are its main determinants. ECONOMIC AND SOCIAL INDICATORS The Kyrgyz economy is growing despite country constraints. GDP growth was close to 4 percent per annum at average in 2000-2007. GDP growth became steadier and accelerated after 2005 to reach 8.3 percent in 2007 - considerably higher than the projected growth of around 7 percent (IMF estimate, Figure 2.1), and 6.6 percent in the nine months of 2008 (NSC data).. In the 2000s economic growth has also became more broad-based and less susceptible to short-term policy reversals (JCSS, vol.1) however it is challenged by certain country characteristics: the country is landlocked, mountainous and difficult to access, with small domestic market which increases the cost of services and undermines competitiveness. It is resource-poor, with only 7 percent of the land arable, dependent on imports of staple foods and remittances of out-migrating workers which represent 17 percent of GDP. Figure 2.1: Real GDP growth, IMF estimates (2007) Real GDP Growth Source: IMF 2007 15 10 annual percent change 5 Georgia Kazakhstan Kyrgyz Republic Russia Tajikistan 0 Uzbekistan 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 -5 -10 Source: Compiled by staff based on IMF data. The Kyrgyz Republic has been hit by multiple shocks in recent years, including food price increases (Kyrgyz had the highest food inflation in the region), fuel / energy price increases, and a recent earthquake which are suppressing economic growth and exacerbating poverty risks. The financial crisis hit the country when it was already vulnerable. The first negative effects of the global economic crisis are the end of the construction boom, slower non-gold real GDP growth, and modest growth of agricultural outputs The likely transmission mechanisms for the global crisis include: financial (credit, liquidity, savings), heavy dependence of the domestic market on imported staple foods (vulnerability to food-price fluctuations), heavy dependence of merchandize exports on the already contracting external demand, 19 slowing of remittances from migrant workers abroad (especially in Kazakhstan and Russia) – which represent 17 percent of GDP – and weaker economic growth and labor demand (employment). Regardless of the economic growth in 2000-08, in terms of GDP per capita the Kyrgyz Republic remains the second poorest country in the region of Europe and Central Asia, but its human capital stock is high by regional comparison. Income per capita which can be a strong predictor of social indicators, is low as already mentioned, and well below the average for the ECA region and even for Central Asia, however some of its key social indicators are much better than would be expected judging from the per capita GDP (Figure 2.2). For example, life expectancy and tertiary enrollment rates are close to the regional average while labor force participation is even a bit above it (World Bank, 2007). Figure 2.2: Regional comparison of key economic and social indicators Income per capita, 2005 Life Expectancy, 2005 12,000 76 10,000 74 72 GDP per capita in US$ 8,000 70 in years 6,000 68 66 4,000 64 2,000 62 0 60 Kyrgyz Central Caucuses Balkans EU-8 Other Kyrgyz Central Caucuses Balkans EU-8 Other Asia FSU Asia FSU GDP per capita ECA Average Life Expectancy ECA Average Tertiary Enrollment Rates, 2004 Labor Force Participation Rate, 2005 70 70 69 60 in percent of 15-64 year olds 68 50 67 in percent 40 66 30 65 64 20 63 10 62 0 61 Kyrgyz Central Caucuses Balkans EU-8 Other FSU Kyrgyz Central Caucuses Balkans EU-8 Other Asia Asia FSU Tertiary Enrollment Rates ECA Average Labor Force Participation Rates ECA Average a. Central Asia includes Kyrgyz Republic, Kazakhstan, Tajikistan, Turkmenistan, and Uzbekistan. b. Caucuses includes Armenia, Azerbaijan, and Georgia. c. Balkans includes Albania, Bosnia and Herzogovenia, Bulgaria, Croatia, FYR Macedonia, Romania, and Serbia and Montenegro. d. EU-8 includes Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic, and Slovenia. e. Other FSU includes Belarus, Moldova, Russia, and Ukraine. Source: World Bank (2007) Compared to other low income countries, the MDGs indicators in Kyrgyzstan are significantly better, but the progress towards their achievement shows mixed results. As seen in Table 2.1, selected indicators of hunger, schooling, gender empowerment, and life expectancy are closer to those of the Europe and Central Asia average rather than other low income countries despite the difference in GDP per capita. A few important MDGs have shown deterioration over the last decade and half, such as women’s employment in non-agriculture activities and primary school enrollment rates. This is also true of gross secondary enrollment rates which fell from 100 percent in 1990 to 86 percent in 2005. On the 20 other hand, mortality indicators for infants and children have improved reaching respectively 38 and 44 per thousand live births in 2006 (MISC, 2006)as have adult mortality rates after 1990, despite the retrenchment of the state in the provision of key child care and health care services. Other positive developments during this period include the Government’s ability to maintain high immunization rates for measles and DPT and the decline in adolescent fertility rates. Last, certain outcomes have stagnated or did not display a distinct trend: e.g., the incidence of tuberculosis increased between 1990 and 2005 though in more recent years they have shown some decline; the life expectancy has remained at 68 years since 1990 despite the improvements in mortality rates during this period; maternal mortality rate is estimated at 104 deaths per 100,000 live births, and showing no sign of improvement over the last 10-15 years (MICS, 2006). The prevalence of poorer human development outcomes is correlated with poverty in general, and specifically with poverty in rural areas and in certain geographical regions. Table 2.1: Kyrgyz Republic: progress towards achievement of the MDGs Kyrgyz Republic LIC /1 ECA /2 Earliest Latest 1990-1995 2001-2004 2001-2004 Hunger Prevalence of undernourishment (% of population) 21 4 24 6 Schooling School enrollment, primary (% net) 92 90 80 90 Gender Share of women employed in the nonagricultural sector 48 43 23 47 (% of total nonagricultural employment) Health Mortality rate, infant (per 1,000 live births) 68 58 80 28 Mortality rate, under-5 (per 1,000) 80 68 122 34 Incidence of tuberculosis (per 100,000 people) 55 122 224 83 Environment - CO2 emissions (metric tons per capita) 3 1 1 7 Improved sanitation facilities (% of population with access) 60 59 38 85 Improved water source (% of population with access) 78 77 75 92 Connectedness Fixed line and mobile phone subscribers (per 1,000 people) 71 106 71 536 Internet users (per 1,000 people) - 52 24 138 Other - GNI per capita, Atlas method (current US$) 350 400 507 3,307 Life expectancy at birth, total (years) 68 68 59 69 Gross capital formation (% of GDP) 24 14 27 23 Trade (% of GDP) 79 95 50 83 Source: World Development Indicators database, September 2006 Figures in italics refer to periods other than those specified. 1. LIC refers to low income countries. 2. ECA refers to Europe and Central Asia (excluding high income countries). 21 POVERTY TRENDS Poverty measured through poverty headcount, gap and severity continues to be a serious issue for the Kyrgyz Republic. In 2005, an estimated 43.1 percent of the Kyrgyz population was living below the poverty line, and 11.1 percent of the population was even unable to fulfill basic food needs. The estimated poverty gap, which describes the mean aggregate shortfall in consumption of the population relative to the poverty line, amounts to 10.5 percent. Taking into account inequality among the poor, we obtain a poverty severity estimate of 3.6 percent. The gap and severity of extreme poverty in the Kyrgyz Republic amount to 1.5 and 0.4 percent, respectively (Figure 2.3). Figure 2.3: Poverty indicators for the Kyrgyz Republic, 2005 60.0 50.0 43.1 40.0 percent 30.0 20.0 10.5 11.1 10.0 3.6 1.5 0.0 0.4 Poverty Poverty gap Poverty Extreme Extreme Extreme headcount severity poverty poverty poverty headcount gap severity Source: Staff calculations based on KIHBS 2005. Box 2.1 Concepts of poverty measurement Poverty headcount index (HCI): The poverty headcount is the percentage of the people whose consumption is low poverty line. Correspondingly, the extreme poverty headcount is the percentage of the people of a given population consumption is lower than the extreme poverty line (or food poverty line). The HCI says nothing about how poor th Poverty gap: The poverty gap (or poverty depth) is the average distance between actual consumption and the pove line. The score is obtained by taking the aggregate consumption shortfall relative to the poverty line of all the poor subsequently dividing it by the size of the whole population. The practical implication of this measure of poverty is express how much resources would be needed to lift all the poor above the poverty line (if perfect targeting is assum If the poverty gap is for example 10 percent, the cash transfer needed to eliminate poverty would be 10 percent of t poverty line on average per person. In the hypothetical situation of the mean national income being twice the pover line, the transfer would amount to 5 percent of the country’s mean income. Poverty severity: The poverty severity is similar to the poverty gap, with the difference that it takes into account inequality among the poor. This is achieved by squaring all individual consumption shortfalls of the poor and subsequently dividing their sum by the total population, as was the case with the poverty gap. By doing so, greater shortfalls are given a higher weight. Thus, an increase in the poverty severity index indicates that income distributi worsened among the poor (assuming that the poverty headcount has remained unchanged). However, the poverty severity is more difficult to interpret than the poverty gap in the sense that it provides no information on the exact amount of resources needed to eradicate poverty. A description of the measurement of the poverty line is given in Annex 1.1. Source: World Bank staff, Foster, Green and Thorbecke (1984). The rates of total poverty and extreme poverty are declining in the 2000s. Over the time period from 2000 to 2005, both total and extreme poverty rate have shrunk by about 20 percentage points. More recent consumption based poverty rates calculated by the NSC indicate further decline in poverty headcount from 43.1 percent in 2005 to 40 percent in 2006 and 35 percent in 2007. Extreme poverty also declined – from 11.1 percent in 2005 to 6.6 percent in 2007. 22 Figure 2.4: Poverty rates from 2000 to 2005 Source: World Bank, 2007. WHERE ARE THE POOR? There are pronounced spatial disparities in poverty rates in the Kyrgyz Republic where poverty is predominantly a rural phenomenon. Two important lines of demarcation are identified when analyzing spatial distribution of poverty, namely whether one lives in an urban or in a rural settlement and whether one lives in a mountainous or a plain area. While 50.8 percent of all people living in rural areas find themselves below the poverty line, this only holds for 29.8 percent of the urban population (Figure 2.5). This finding is reflected in the fact that the rural poor constitute almost three quarters of the total poor, while, at the same time, their share in the total population is less than two thirds. Figure 2.5: Poverty incidence and distribution of the poor by rural/urban sector Poverty headcount index (% of population) Distribution of the poor (% of all poor) 60 Total Poverty: 50.8 50 Urban Rural extremely moderately poor poor 40 Total poverty: 29.8 Moderate 30 Poverty: 37.0 Urban extremely poor Moderate 20 Poverty: 23.3 10 Extreme Extreme Poverty: 13.8 Poverty: 6.5 Rural 0 moderately poor Urban Rural Source: Staff estimates based on KIHBS 2005. Poverty is also concentrated in mountainous areas, with more than half of the people living in highly or moderately mountainous areas being poor. As a result, their share in the total poor clearly exceeds their share in the total population. Concerning extreme poverty, an even more pronounced picture of the poor living in rural and mountainous areas emerges: almost four fifths of the extremely poor live in rural areas. Furthermore, more than 20 percent of the people living in highly mountainous areas live below the food poverty line, which is almost twice the national average (Table 2.2). 23 Table 2.2: Poverty indices for individuals by region and topography Poverty Poverty Poverty Share in total Share in total headcount gap severity poor population by region: Urban 29.8 7.1 2.4 25.4 36.7 Rural 50.8 12.4 4.3 74.6 63.3 by topography: Highly mountainous 63.8 14.3 5.0 15.3 10.3 Moderately mountainous 51.0 13.0 4.3 14.3 12.1 Plain 39.1 9.6 3.3 70.4 77.6 Extreme Extreme Extreme Share in total poverty poverty poverty extremely Share in total headcount gap severity poor population by region: Urban 6.5 1.1 0.4 21.4 36.7 Rural 13.8 1.8 0.4 78.6 63.3 by topography: Highly mountainous 20.5 2.2 0.6 19.0 10.3 Moderately mountainous 8.5 1.3 0.2 9.2 12.1 Plain 10.3 1.5 0.3 71.8 77.6 Source: Staff calculations based on KIHBS 2005. The oblasts of the Kyrgyz Republic significantly differ in terms of their topographies and degrees of urbanization, which directly translates into considerable differences in oblast-level poverty rates. People in Batken, Osh, and Jalalabat are especially susceptible to being poor. Poverty rates in these oblasts fall into the range from 55 to 59 percent. Considering that Naryn and Issyk-kul have only marginally lower poverty rates, we find that precisely those oblasts that have the highest poverty incidence also have the most mountainous surface. At the other extreme, poverty rates are particularly low in the city of Bishkek and the fairly plain oblast of Chui. With the slight exception of Osh, the above findings also hold for poverty gap, poverty severity and extreme poverty (Table 2.3). In terms of extreme poverty, the oblasts of Naryn, Batken and Jalalabat report food poverty rates of about 20 percent. This results in the fact that their shares in the total extremely poor are almost twice their population shares. Analogously to the extreme poverty headcount, extreme poverty is also deepest and most severe (alongside the oblast of Talas) in these three oblasts. As mentioned above, the case of Osh is a special one in this respect. While a lot of people are poor in this oblast (and 10 percent even extremely poor), poverty and extreme poverty seem to be less deep and severe there. 24 Table 2.3: Poverty indices for individuals by oblast Poverty Poverty Share in total Share in total headcount Poverty gap severity poor population by oblast: Issyk-kul 51.5 14.3 5.1 10.0 8.3 Jalalabat 55.9 16.3 6.1 24.3 18.7 Naryn 51.2 14.4 5.4 6.2 5.2 Batken 59.1 17.3 6.1 11.2 8.1 Osh 55.9 10.4 3.2 32.8 25.3 Talas 44.4 11.7 4.2 4.3 4.2 Chui 22.0 4.5 1.3 7.5 14.6 Bishkek 10.8 1.9 0.4 3.9 15.5 Extreme Extreme Share in total poverty Extreme poverty extremely Share in total headcount poverty gap severity poor population by oblast: Issyk-kul 14.2 2.1 0.4 10.6 8.3 Jalalabat 18.0 3.0 0.7 30.2 18.7 Naryn 20.0 2.7 0.7 9.4 5.2 Batken 19.0 2.5 0.6 13.9 8.1 Osh 10.2 1.3 0.3 23.1 25.3 Talas 14.1 2.1 0.6 5.3 4.2 Chui 5.3 0.3 0.1 6.9 14.6 Bishkek 0.4 0.0 0.0 0.6 15.5 Source: Staff calculations based on KIHBS 2005. WHO ARE THE POOR? Predominantly children and households with children are at risk of being poor or extreme poor. Table 2.4 it indicates that 52 percent of the under-18 year old and even 57.7 percent of the under-six year old live in poverty in the Kyrgyz Republic. Similarly, 14.8 percent of the under-18 year old and 18.9 percent of the under-six year old live in extreme poverty, which is in both cases significantly higher than the national average of 11.1 percent. The highest poverty rates were estimated for households with three and more children: almost 70 percent of individuals living in such households are found to be poor (see Annex 1.2). These people alone account for more than half of the total poor in the Kyrgyz Republic. Logically, households with at least six members figure disproportionately in the distribution of the poor, given their share in the total population. A good illustration of the above is the example of households with three adults and three or more children. It was estimated that 63.2 percent of the individuals living in such households are poor. Further, it was found that 44.9 percent of the total poor live in households with at least four adults. This figure is more than 10 percent higher than the share of such households in the total population. Extreme poverty is even more related to household size: half of all individuals that fall below the extreme poverty line live in households with seven or more members. This is a very strong pattern given the fact that less than 20 percent of the population live in such households. Similarly, 65 percent of the extremely poor live in households with three or more children, which is more than twice the population share of such households. Not surprisingly, extreme poverty is also by far deepest and most severe among individuals living in such very large households with many children. Males and females are not significantly different in terms of poverty status. 25 Table 2.4: Poverty indices for individuals by demographic characteristics Poverty Poverty Share in total Share in total headcount Poverty gap severity poor population by sex:* Male 42.5 10.3 3.6 46.8 48.2 Female 43.7 10.6 3.6 53.2 51.8 by age (three groups): Under 18 years 52.0 13.0 4.6 51.5 38.9 18-59 years 38.0 9.0 3.0 43.0 53.4 60 years or older 33.8 8.0 2.7 5.5 7.8 by age (seven groups): Under 6 years 57.7 15.1 5.3 18.9 11.1 6-15 years 51.5 12.7 4.4 28.1 22.9 16-20 years 38.8 9.4 3.2 8.2 10.2 21-40 years 43.4 10.5 3.6 30.1 28.6 41-60 years 30.1 6.6 2.1 9.4 19.6 61-70 years 32.0 7.1 2.3 2.2 4.1 71 years or older 36.9 9.3 3.3 3.2 3.5 Extreme Extreme Share in total poverty Extreme poverty extremely Share in total headcount poverty gap severity poor population by sex:* Male 10.8 1.6 0.3 47.5 48.2 Female 11.4 1.5 0.3 52.5 51.8 by age (three groups): Under 18 years 14.8 2.0 0.4 46.9 38.9 18-59 years 9.0 1.3 0.3 47.0 53.4 60 years or older 7.8 1.2 0.3 6.1 7.8 by age (seven groups): Under 6 years 18.9 2.3 0.5 14.9 11.1 6-15 years 13.7 1.9 0.4 27.4 22.9 16-20 years 8.9 1.3 0.3 9.2 10.2 21-40 years 11.7 1.6 0.4 28.8 28.6 41-60 years 5.3 0.8 0.2 13.6 19.6 61-70 years 6.0 0.9 0.3 3.1 4.1 71 years or older 10.2 1.7 0.3 3.0 3.5 * Not significant at the 10 percent level. Source: Staff calculations based on KIHBS 2005. Poverty and extreme poverty in particular, is prominently positively correlated with the level of education of the household head. Of the people living in households headed by a person holding a higher education degree, only 18.4 percent are poor, and their share in the total poor is 7.3 percent, which is considerably lower than their population share of 17.1 percent. Individuals living in households headed by a person holding less than a higher degree experience a higher poverty risk: their shares in the total poor are significantly higher than their population shares. The findings are even more convincing in the case of extreme poverty, with individuals living in households headed by a person possessing a higher degree accounting for only 2.2 percent of the total extremely poor. Somewhat surprisingly, having a household head with no education at all does not translate into a higher risk of being extremely poor. Yet, extreme poverty in this segment is (if present) most severe. Other characteristics of the household head that matter are his or her marital status and nationality. In terms of the former, widow(er)s experience the highest poverty incidence, while living apart without being divorced is related to deeper and more severe poverty. Concerning nationality, Russians experience a significantly lower poverty risk than Kyrgyz nationals. The gender of the household head has no effect on poverty; the same applies to the employment status of the household head (see Annex 1). 26 Table 2.5: Poverty indices for individuals by characteristics of household head Poverty Poverty Poverty Share in total Share in total headcount gap severity poor population by education: Higher degree 18.4 4.0 1.1 7.3 17.1 Secondary education 47.5 11.3 3.9 75.4 68.4 Primary education 48.4 14.5 5.4 12.1 10.8 No education, illiterate 60.2 13.0 4.4 5.1 3.7 by marital status: Legally married 44.5 11.0 3.8 73.8 71.5 Married under common law 33.3 5.6 2.0 2.5 3.2 Divorced 26.1 5.6 1.6 2.7 4.4 Living apart, not divorced 39.5 13.7 5.0 1.2 1.3 Widow(er) 46.8 11.1 3.8 19.5 17.9 Has never been married 11.6 2.8 0.8 0.5 1.7 by nationality: Kyrgyz 46.9 11.3 3.9 67.9 62.4 Russian 10.6 1.7 0.4 2.9 11.9 Other 48.7 12.6 4.4 29.1 25.8 by sex:* Male 44.4 10.5 3.6 74.4 72.3 Female 39.7 10.4 3.6 25.6 27.7 Extreme Extreme Extreme Share in total poverty poverty poverty extremely Share in total headcount gap severity poor population by education: Higher degree 1.5 0.2 0.0 2.2 17.1 Secondary education 12.3 1.7 0.4 75.3 68.4 Primary education 21.3 2.6 0.4 20.6 10.8 No education, illiterate 5.4 1.8 0.8 1.8 3.7 by marital status: Legally married 11.7 1.7 0.4 75.0 71.5 Married under common law 5.9 1.3 0.4 1.7 3.2 Divorced 4.0 0.5 0.1 1.6 4.4 Living apart, not divorced 33.9 0.6 0.1 3.9 1.3 Widow(er) 10.9 1.5 0.4 17.5 17.9 Has never been married 2.3 0.1 0.0 0.4 1.7 by nationality: Kyrgyz 12.5 1.6 0.4 70.0 62.4 Russian 0.6 0.1 0.0 0.7 11.9 Other 12.7 2.0 0.5 29.3 25.8 by sex:* Male 11.0 1.5 0.4 71.5 72.3 Female 11.4 1.6 0.4 28.5 27.7 * Not significant at the 10 percent level. Source: Staff calculations based on KIHBS 2005. Poverty is very much linked to housing characteristics, such as the amenities of ones house, and the possession of consumer durables. Concerning the former, it is especially the material of house’s roof and its walls, as well as the main water source of the household that matter. Individuals living in houses built of bricks or concrete are less often found to be poor than those living in constructions made of airbricks, clay, earth or other materials. Similarly, living in a house with a roof made of concrete is related to a lower poverty incidence. This finding is confirmed in the case of extreme poverty. Yet, a remarkable observation in this respect is that 12 percent of the extremely poor have roofs that are neither made by concrete nor by roofing slates (see Annex 1). This hints at the fact that alternative roofing techniques rely on inferior material and are correlated to the risk of being extremely poor. Not surprisingly, individuals living in households disposing over running water or (to a lesser extent) a 27 private pump are less prone to be poor than their counterparts that rely on public water sources such as public pumps or rivers and lakes. However, those using rivers and lakes, in turn, appear to be less susceptible to suffer from extreme poverty and their poverty is not as deep as the one of those obtaining their water from public pumps. This indicates that living in areas with little natural water sources is a further significant burden on those who are poor anyways. Furthermore, water consumed in poorer areas (Southern oblasts) is often not entirely clean, which can (among others) give rise to diseases such as trachoma, cholera, and typhoid (MICS, 2006). As far as other housing amenities and services are concerned, we estimated that predominantly those living in houses without central heating and hot water supply are poor (the latter being related to the above discussion on running water). The case of electricity is a special one, as by now almost all houses in the Kyrgyz Republic are equipped with electricity. Against this background, the figure of 99.8 percent poverty incidence among those living in houses without electricity should not be overestimated. In terms of consumer durables, we observe that particularly the possession of a telephone, television, a washing machine, a car, and a fridge are linked to lower poverty incidences. The exact estimations with regards to housing characteristics and consumer durables can be found in Annex 1. Disparities are often more pronounced in the reliability of public services provision than in the on-the-paper access to them (World Bank, 2007). The prevalence of living in privately owned houses and land ownership are very high but the access to housing and land somewhat differs for urban and rural poor. Almost all of the rural population lives in separate houses, whether they are poor or not. In urban areas, the poor are more likely to live in separate houses than the non poor, because the poor resort to informal housing in peri-urban areas to a much larger extent. Both poor and non-poor own land but in rural areas the poor have smaller land holdings than the non-poor. On average, poor rural households that have access to own land have plots of about 1,800 square meters, or less than 0.2 hectares, while non-poor households have plots of about 2,800 square meters. Because urban poor are less likely to live in the most densely populated central areas of cities, less likely to live in apartment buildings, and more likely to need to rely on subsistence farming, they have more access to land than the non-poor. The average area available is very small, however, measuring about 20 by 20 meters (World Bank, 2007). CONCLUSIONS With economic growth becoming steadier and more broad-based in 2000-2007, total and extreme poverty has declined, however: • Large parts of the Kyrgyz population continue to be poor. Not only the share of the poor remains considerable, but also new risks of rising poverty and vulnerability emerge following the increase in energy and food prices globally leading to increase in domestic prices, energy tariffs and utility fees. Adverse climate change effects as severe winters, floods, and other natural disasters are exacerbating the poverty risks. Finally, the global financial and economic crisis threatens to hurt the Kyrgyz economy, particularly given its openness and dependence on economic slowdown in the region (Russia and Kazakhstan in particular) and subsequent contraction of employment and remittances from migrant workers; • Child poverty is very pronounced, with 52 percent of the under-18 year old and 57.7 percent of the under-6 year old living in poverty and with 14.8 percent of the under-18 year old and 18.9 percent of the under-6 year old living in extreme poverty. The highest poverty rates are estimated for households with three and more children - almost 70 percent of individuals living in such households are found to be poor. • Poverty is concentrated in large and multi-generational households – a half of the total poor live in households with four and more adults, while a half of the extreme poor live in big households of seven and more members. • Along with being child-related, poverty in Kyrgyzstan has distinct spatial and non-income dimensions. It is a predominantly rural phenomenon with over 50 percent of the rural population being poor, including close to 14 percent being extremely poor. Poverty incidence varies by regions with Batken having the highest rate of moderate poverty and Naryn having the highest rate of extreme poverty while in the Jalalabat region both moderate and extreme poverty are 28 among the highest. Poverty is also higher for the households living in mountainous areas where the poverty rate is the highest in the country - 60 percent. • Poverty and extreme poverty are strongly positively correlated with the level of education of the household head, quality of dwelling, housing amenities and possession of consumer durables. 29 . CHAPTER 3. THE CURRENT SYSTEM OF NON- CONTRIBUTORY SOCIAL BENEFITS This chapter describes the current social benefits system in the Kyrgyz Republic emphasizing on the non- contributory benefit with their eligibility criteria, benefit formulae, financing mechanisms and allocations, drawing mainly on legislative documents and administrative data provided by MLSD and MOF. It sets the context for the follow-up analysis of the non-contributory benefits from the perspective of their main social policy objectives: to assist households and individuals in smoothing consumption over lifetime by providing supplementary income in the event of chronic or transient poverty. CURRENT SYSTEM OF SOCIAL BENEFITS Social protection in the Kyrgyz Republic includes both contributory (social insurance) and non- contributory (social assistance) benefits. Collectively, these schemes aim at reducing the risks of poverty and vulnerability associated with temporary or permanent decline in earning capacity by providing contributory (pensions, unemployment) and non-contributory benefits (state benefits, categorical in-kind subsidies and benefits). Social Insurance (Contributory Benefits) Social insurance benefits account for just below six percent of GDP in Kyrgyzstan. This is slightly less than the average for the ECA region (8.3 percent) but fairly high for low-income countries. The old- age related risks are addressed by the current pension system which is being shaped as a result of an ongoing since 1997 reform 4, and at this point is characterized by the following parameters: • A basic pension component which combines element of a contributory and a social pension - on one hand, it is based on residency and contribution history, and a shorter length of service results in proportional reductions to the basic pension; on the other hand, the basic pension is not earnings-related meaning that everyone with the minimum contribution history of 25/20 years (men/women) is entitled to a flat-rate of basic pension set at 12 percent of the average wage. • A transition component which is a legacy of the pre-reform pay-as-you-go defined benefit social insurance scheme that honors pension rights accrued under that regime. Benefit assessment is based on the best five consecutive years of contribution prior to 1996 and benefit levels follow the wage index. Since no new rights accrue under the transitory system, it is expected that from approximately 2030 all new benefits will comprise of only two components – the base pension and annuities accrued in the notionally defined contribution (NDC) pension pillar. • A notionally defined contributions (NDC) component with individualized accounts which notionally accrue contributions and are credited with a notional interest equal to 75 percent of wage growth. The scheme is pay-as-you-go financed and contributions are used to pay concurrent benefits, therefore the NDC scheme only accrues notional capital to be translated into annuities at retirement. • A mandatory defined contribution scheme initially with 2 percent of the contributions to be channeled to individual accounts has been considered for introduction. At this point with a Presidential Decree 5 the concept and the action plan for the introduction of this scheme are put in place, along with a medium- and long-term pension reform agenda aimed at achieving financial stability of the Pension Fund and broadening the financing base of the pension system.. 4 Detailed analysis of the pension system is provided in the first part of this programmatic analytical work: Pension Policy Note – Policy Considerations and Practical Proposals. 5 Presidential Decree “On Measures for Introduction of Fully Funded Scheme to the Pension System�, September 24, 2008, No. 339. 30 • Pension contribution rate at 27 percent of the payroll. Contributions are collected by the Social Fund which pays out all three types of benefits out of them, with 24 percent being credited to the NDC accounts. • The retirement age for men and women increases gradually from 60/55 before the transition to 63/58 in 2008. The process was reverted to 60/55 in 2007 leading to an expected inflow of additional 75,000 and 50,000 pensioners in 2007 and 2008 respectively, but the gradual retirement age increase was resumed in 2008. According to administrative data, in 2007, 8 percent of the population was of pension age (427.2 thousand people), and 352.2 thousand receive old-age pensions. The pension system provides certain coverage against old age risks. At the same time the basic pension is with low coverage, level and replacement rate. The replacement rates in the NDC pillar are also low, while the benefit distribution is compressed and with little relationship to contribution performance. The Social Fund’s revenues are being strengthened as a result of the reform which is driving average benefits down, the economic growth and the resulting wage growth, as well as due to efforts to collect arrears. However, the number of contributors and coverage are quite low - only about 15 percent of the working age population is accruing any pension rights, due to low labor force participation rate (44-45 percent) and low compliance: out of 2.4 million people in the labor force in 2008, only 0.8 million are registered as contributors. In 2007, the average old age pension amounted to one third of the average wage, with projected decline of projected replacement rates by at least 10 percentage points by 2040. The pension benefit levels are particularly low for farmers who make up approximately 55 percent of the population and one third of the registered contributors. Their contributions make up less than 1 percent of total contribution revenues, at the same time retired farmers represent approximately 65 percent of the beneficiary population and absorb close to 60 percent of the pension expenditures. With time the farmers’ pension replacement rate is expected to decline further and eventually asymptote to the basic pension’s replacement rate. Social Safety Net (Non-Contributory Social Assistance Benefits) The social safety net is weaved together with a patchwork of “old-style� categorical benefits and subsidies (“privileges�) and two new targeted benefits. Non-contributory benefits are in the midst of a reform aimed at increasing their targeting to the poor and vulnerable. The country inherited from the Soviet times a safety net consisting of a large number of categorical poorly targeted and costly benefits (“privileges�). In the middle of the 1990s, the government initiated reforms aimed at simplifying the benefit programs, reducing the overall costs and reaching the neediest of an increasingly impoverished population. Hence, the present non-contributory part of the social protection system in Kyrgyzstan is a mix of “new� and “old� programs. On one hand, the safety net includes a plethora of categorical benefits (“privileges�) 6 that represent a legacy-of-the-past. These include cash benefits and in-kind subsidies for categories of citizens who are not necessarily poor. Spending on categorical benefits and subsidies is still significant, at 0.42 percent of GDP in 2007 (Table 3.1) but expected to decline. These are discussed in more detail below. On the other hand, two new targeted cash transfer programs were introduced after independence in 1991. These include: the Unified Monthly Benefit (UMB) and the Monthly Social Benefit (MSB), which together account for about 0.7 - 1.0 percent of GDP annually in the period from 2001 to 2008 (0.74 percent in 2007). • The UMB was introduced via a Presidential Decree in January 1995 as part of an important step in reforming the old-style safety net. After several amendments and improvements on how to assign the new benefit, the UMB was finally regulated in the Law on State Benefits, approved by the Kyrgyz Parliament, the Jorgorku Kenesh, and signed by the President of the Kyrgyz Republic on 5 March, 1998. The UMB is a last-resort poverty-targeted cash benefit program that is means-tested and categorically-targeted to children from low-income families. 6 These programs are referred to as ‘privileges’ (l’goti) in some of the quoted literature. 31 • The MSB, was also established by the Law on State Benefits, and is a cash income-replacement program that is categorically targeted (but not means-tested) to disadvantaged groups, including orphans, people with disabilities and elderly who do not qualify for old-age pensions. Social assistance outlays also include spending on social services and institutional care, which accounts for around 0.1 percent of GDP. These services are limited in scope and are provided mostly in state institutions, and – on small-scale pilot basis – in community and family environment. A substantive part of the spending is allocated for maintenance of 14 social care institutions and provision of care and board of their residents who are elderly, people with disabilities and children deprived of parental care, including children with disabilities. In recent years, the budget for social services increases due to increased financing standards for food and maintenance, refurbishment and for new pilot initiatives7. Finally, after 2005, when the Social Insurance Fund was abolished, the safety net took over the payment of maternity benefits for working mothers, birth grants, payments for ritual services and health recovery activities. The new cash transfer programs target only specific categories of beneficiaries meaning that the safety net could mitigate only part of the poverty and vulnerability risks throughout the individual’s life cycle. As mentioned, the UMB apples two filters to identify prospective beneficiaries: first, only the households with per capita income below the Guaranteed Minimum Level of Consumption (GMCL) and second, children in these low-income families. The MSB targets certain population groups with increased vulnerability - people with disabilities, elderly who are not eligible for work-related pensions and those that have lost the breadwinner in the family. At the same time social assistance programs are not designed in a way to protect able-bodied individuals without children that might fall into extreme poverty in case of unemployment, loss of assets, price shocks or other idiosyncratic risks. The support for the unemployed with benefits and active employment measures (re-training, public works, micro credits) is quite limited and became available only after 2005. The UMB and MSB not only reach specific groups of beneficiaries only, but also due to their low value, they cannot have significant immediate impact on poverty reduction. Their longer-term impact on poverty reduction is also limited because the benefit provision is not linked to services which could motivate employment and self- employment, improve employability and promote investments in children, e.g. through positive parenting behavioral changes that promote human capital development through healthy nutrition patterns, school attendance, compliance with immunizations schedules and regular health checks. Short- and medium-term safety net reform objectives The Government of the Kyrgyz Republic demonstrates commitment to complete the reform of social assistance and to improve the design and implementation of the safety net programs with the overall objective to alleviate the most severe manifestations of poverty and increase the capacity of the safety net to protect the poor against risks emerging from crises and change. The stabilized and broader-based economic growth after 2005 was conducive for increasing some of the benefits in real terms, to monetize (cash out) some of the poorly targeted categorical benefits and subsidies, and to eliminate the arrears in benefit payments. The main directions of the forthcoming reforms are determined by the Country Development Strategy of the Government of the Kyrgyz Republic 2007-2010 (CDS), and its update for 2009-2011. The CDS was adopted in 2006, and underscores Kyrgyzstan’s ambition to sustain and accelerate growth, reduce further poverty and become a middle-income country. The updated strategy’s overarching objective is to improve the quality of life through improving the quality of economic growth, public services and environment. The establishment of an effective system of social protection is among the instruments for quality of life enhancement. The CDS update which is currently under preparation emphasizes sustainability of social spending in the outset of the food, fuel and financial crises, and calls for improving the targeting of the social transfers to the poorest and for making the transferred amounts more meaningful. Also in 2006, the Kyrgyz Government adopted Social Protection Strategy for 2006-2008, with more specific short to medium-term reform objectives emphasizing on improving the effectiveness of social protection, especially for the vulnerable population, by reinforcing the targeting of benefits, stopping proliferation of privileges and monetization of utility benefits; reinforcing activation – active labor market policies and promotion of entrepreneurship and independent 7 Social services are discussed in greater detail in Chapter 6. 32 job creation; and expansion of the scope of social services for the elderly, disabled and children and families at risk. In parallel, a longer-term vision for social reforms is being currently elaborated in the Social Development Concept 2008-2014, with the support of the development partners of Kyrgyzstan. Box 3.1: Safety net reform vision and priorities The Country Development Strategy underscores Kyrgyzstan’s ambition to become a middle-income country and takes a comprehensive approach to national development which should be focused on rapid and sustainable economic growth and on sustained poverty reduction, improvement of living standards and human capital enhancement. To achieve significant reductions in poverty, the strategy focuses reform efforts on sectors and industries that can provide the maximum cumulative contributions to GDP growth, also on improving labor productivity, diversification of the economic base, and accelerated growth. The CDS puts the safety net in the context of broader social protection and basic social services reforms, and makes a strong argument that they are dependent on and interrelated with: • the quality of the country’s labor force; • the access to affordable education and health services; • the population’s living conditions, including access to water, sanitation and basic services; • the government’s effectiveness and performance in delivering basic social services. The CDS Update for 2009-2011 is under preparation. It emphasized the need of significant poverty reduction, especially in rural areas, including through improved targeting of the social benefits and enhanced social protection. It also accentuates the need of increasing the average social benefit level. The main result of the implementation of the updated CDS are expected to be: (i) UMB average level reaching the extreme poverty line; (ii) MSB and pensions reaching the minimal consumer budget; (iii) development of social services system at the local level, including, in particular, new services for children from groups at risk; (iv) increase in the financial resources in the contribution-based pension schemes. The Social Protection Strategy and the Social Development Concept outline a more specific vision for the safety net reform in the short- and medium term respectively, more specifically they call for: • Increasing the generosity of the non-contributory social benefits through increasing the value of the GMCL and sustaining or increasing their value relative to the GMCL; • Enhancing the targeting of the UMB to the poor by more accurate accounting of all available incomes; • Linking the MSB with services relevant to the specific needs of the beneficiaries, and extending its coverage to mothers engaged in home care for disabled children; • Reforming further the categorical benefits and subsidies and introduce targeting • Developing partnerships with the civil society sector for the provision of social care services outside the state institutions and to increase the quality of institution-based care. Source: Government of Kyrgyz Republic, 2006. Country Development Strategy 2007-2010; Country Development Strategy update 2009-2011 (draft), Social Protection Strategy 2006-2008. Fiscal effort Kyrgyzstan spends annually between 1 and 1.5 percent of GDP on social assistance benefits and services in the 2000s.This is lower than the average spending in ECA and OECD but comparable with relative spending-to-GDP shares in countries of the same level of economic development. After 2005, spending is declining in relative terms, and in 2009 for the first time is expected to fall below 1 percent of GDP (0.84 percent as per the MTBF 2009-2011). Table 3.1: Public expenditure on non-contributory social protection programs, 2007-2009 2007 2008 2009* GDP (billion KGS) 141.9 185 253.6 Social assistance / non-contributory social 1835.4 2137.2 2140.6 benefits as percentage of GDP (total) 1.29 1.16 0.84 - ..state benefits (UMB and MSB) 1060 1188 1234.7 as percentage of GDP 0.74 0.64 0.49 - ..categorical benefits and subsidies** 592.4 688.7 639.9 as percentage of GDP 0.42 0.37 0.25 - ..pregnancy and child delivery benefits, ritual 102.7 138.0 194.7 33 services benefits, health recovery activities, social care services, including in institutions as percentage of GDP 0.07 0.09 0.08 *Planned spending as per the MTBF 2009-2011; **Only republican budget. Source: MLSD and MOF The structure of social spending reflects societal needs and forthcoming challenges only partially. In the 2000s, the spending on all safety net programs – UMB, MSB, categorical benefits and subsidies, pregnancy and child delivery benefits, funeral grants, one-time emergency support and social care services (home- and institution-based) – ranges between 1 and 1.5 percent of GDP. This is lower than the average spending in ECA (25) which amounts to 1.7 percent of GDP but comparable with relative spending-to-GDP shares in countries of similar level of economic development. After 2005, spending is declining in relative terms, and in 2009 for the first time is expected to fall below 1 percent of GDP (0.84 percent as per the MTBF 2009-2011). However, as indicated in Table 3.1, within this category, the share of the spending targeted to the poor and vulnerable (UMB and MSB) is lower - at 0.64 percent of GDP in 2008 and declining to 0.49 percent of GDP in 2009. In relative terms the spending on UMB and MSB has been cut in half compared to the first half of the 2000s. At the same time, the spending on poorly targeted subsidies for certain categorically eligible groups – though also declining in relative terms - remains high weighted against the number of beneficiaries and spending with similar objectives in other ECA countries. 34 Figure 3.1: Expenditures on social assistance in Kyrgyzstan (2008) and ECA countries* (% of GDP) Public Spending on Social Assistance, % of GDP Bosnia- Slovak Ukraine Belarus Kosovo Czech Republic Kazakhstan Turkey Armenia Uzbekistan Croatia Russia Moldova Azerbaijan Georgia Serbia Macedonia Latvia Bulgaria Albania Romania Kyrgyzstan Poland Tajikistan - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 *Data is the most recent available for the respective country and is subject to further update. Source: Lindert (2008) The commitment of the MoF to financing social assistance as indicated in the MTBF is to further increase in the state defined financing standards which predetermine the future non-contributory benefits levels and the overall spending on social benefits and services till 2011, and stop the fall of the UMB and MSB relative to financing standards. The Guaranteed Minimum Consumption Level (GMCL) which amounted to KGS 200 in 2008 and KGS 240 in 2009 will increase to KGS 280 in 2010. In parallel the share of the GMCL in the Minimum Consumption Budget (MCB) is planned to increase compared to 2008 (5.6 percent of MCB) and stabilize at 6.3 – 6.4 percent in 2009-2011. This will halt the trend from 2005-2007, when the share of the UMB and MSB in the MCB was declining. Table 3.2: Dynamics of social financing standard and average benefits, 2005-2011 Years Measur 2005 2006 2007 2008 2009 2010 2011 Name e/ unit actual actual actual actual forecast forecast forecast MCB KGS 1837,0 2377,2* 2795,9* 3570,9* 3800,0 4360,0 5020,0 GMCL size KGS 140,0 175,0 175,0 200,0 240,0 280,0 320 GMCL of MCB percent 7,6 7,4 6,3 5,6 6,3 6,4 6,3 Average size of UMB KGS 88,4 124,0 121,7 127,5 165,2 205,2 245,2 UMB as percentage of MCB percent 4,8 5,2 4,4 3,6 4,3 4,7 4,8 Average size of MSB KGS 364,7 457,0 504,4 717,2 858,84 1100,98 1145,12 MSB as percentage of MCB percent 19,9 19,2 18,0 20,1 22,6 25,2 22,8 Source: MLSD, MOF, MTBF 2009-2011, Kyrgyz National Statistics Committee *calculated according to the methodology approved in Government Resolution № 333 of 15.08.2007 THE UNIFIED MONTHLY BENEFIT Definition. The UMB is a means-tested as well as categorically targeted cash benefit aimed at poor households with children. The UMB is a ‘poverty gap’ program which amount per eligible household member is calculated as the gap between his/her average monthly per capita income and an income 35 threshold called the Guaranteed Minimum Consumption Level (GMCL). Thus the UMB is a variable benefit ensuring all households with eligible household members an income no less than a defined minimum social standard which applies to the eligible household members only. Benefit formula. The formula used to determine the variable UMB, is described in Box 3.2. Box 3.2: Calculation of the UMB per household The amount of UMB due to each household with eligible beneficiaries is arrived at as follows: UMB= (GMCL-(HItot/ Nh))*Nel, where: GMCL= Guaranteed Minimum Consumption Level HItot= Total household income per month Nh= Number of persons in the household Nel= Number of eligible members of the household Categorical eligibility criteria According to the Law on State Benefits, the following household members are eligible for the variable UMB: • Children under 16 years (and pupils still in general educational institutions until graduation, however, maximum until the age of 18) • Pupils of primary vocational schools and students of secondary and higher vocational educational institutions (professional colleges) up to the age of 21 8 • People with disabilities 9 Total household income Total household income includes net income by all household members from all sources, cash as well as in-kind. Hence it includes income from among others; employment, bonuses, patented private commercial activities, leases, income from assets and deposits, crops (estimated by productivity coefficients 10), pensions, private transfers, scholarships and inheritance. Income from livestock however is not included and neither are unemployment benefits, the MSB (see below) or single transfers such as funeral allowances or childbirth benefits. Guaranteed Minimum Consumption Level The GMCL is a cash social standard established by the GoK in 1998 and adjusted annually based on the following formula: 11 GMCL= BF+12*PUMB*APMI (PUMB+1.39*PSMB)*12 where: BF = earmarked budget funds for social protection PUMB = predicted number of UMB recipients APMI = predicted average per capita monthly income of UMB recipients PSMB = predicted number of SMB recipients UMB and SMB (Social Monthly Benefits) are to be paid out during 12 month, hence the number and 1.39 is an adjustment coefficient. Source: Law of the Kyrgyz Republic on State Benefits. According to Article 4 in the Law on State Benefits, the annual revisions of the GMCL aim at gradual convergence between the GMCL and the Minimum Consumption Budget (MCB). 12 The amount assigned for the GMCL in 1998 was 100 KGS, which at that time was almost half of the extreme poverty line, i.e. the cost of a basket of food comprising 2100 Kcal per day, and more then 12 percent of the MCB. In 2008, the GMCL was defined at 200 KGS, which represents less than a third of the extreme poverty line and less than 6 percent of the MCB. 13 It is obvious that the objective to close the gap between the GMCL and the MCB is far from being realized. Instead this gap is steadily increasing (see Figure 3.2). The fact 8 Excluded are students studying per correspondence, during evenings or on a contractual basis 9 According to Article 3 this includes children under sixteen, persons with disabilities confirmed by a social expert commission and people of pension age (without pension) 10 The coefficients differs by regions and also reflects the type of land (arable or irrigated) 11 Provision on guaranteed minimum consumption level, approved by regulation #231 of the Government of the Kyrgyz Republic on April 29, 1998 12 The average minimum consumption budget consists of food, nonfood, services and taxes, dues and payments 13 MCB figure for 2008 is a projection, 36 that the GMCL supports only about a third of the minimal nutritional requirement in a food basket also indicates that this threshold is far away from a social consumption minimum. Figure 3.2: Development of GMCL, the extreme poverty line and the minimum consumption basket, 1998-2008 4000 3500 3000 GMCL 2500 KGS Extreme Poverty Line 2000 1500 Minimum Consumption 1000 Basket 500 0 98 00 02 04 06 08 19 20 20 20 20 20 Source: Kyrgyz National Statistics Office, MLSD administrative data The GMCL formula further reveals that the GMCL and subsequently the amount of the UMB (and MSB) rather are determined by fiscal considerations than by the needs of the poorest. Barrientos and Davies (2008) also point to the problem of the pro-cyclical nature of the GMCL formula. An efficient social safety net should rather be counter-cyclical in order to shield the poorest from negative economic shocks. 14 In addition to the variable UMB calculated according to the above formula, there are also the following fixed UMB benefits granted to poor families whose per capita income does not exceed the GMCL: Table 3.3: Categories of UMB beneficiaries and UMB amounts in 2007 Category of UMB beneficiary UMB amount Benefit at child birth 300% of GMCL (one-time) Benefit to children under 1.5 years 100% of GMCL (monthly) Benefit to twins aged 0-3 years 100% of GMCL per child (monthly) Benefit to triplets and higher multiple births aged 0-16 years 150% of GMCL per child (monthly) Source: MLSD. Eligibility. Determining eligibility for the fixed UMB is categorical. Proof is the birth certificate. Determining eligibility for the variable UMB involves two filters: a means-test and a categorical criterion. First, only households where per capita income is less than GMCL are eligible for the program. Second, only certain family members are eligible for the UMB: children under 16 or enrolled students up to 21 years old. For the purpose of eligibility determination, the household income is calculated as the sum of cash and imputed income from land plot ownership. To calculate per capita income, family/household income is divided by the number of family members. • Incomes that count include: (i) income from main job or study – wage plus all benefits, one-time top-ups, awards and bonuses; pensions and scholarships; (ii) incomes from jobs different from the main one – honoraria, patent remunerations, royalties, profits from commercial activities, interests earned on bank deposits and treasury bills; cash or in-kind inheritance and donations; incomes from sales of immobile and mobile property; dividends; alimony; (iii) incomes from entrepreneurial activities, including renting of land and breeding of animals for hunting); (iv) estimates / imputed income from ownership of agricultural land which is considered arable 14 S. Ibragimova (2008). Assessment of State Benefits to Children and Families in the Kyrgyz Republic, Observations and a Way Forward (mimeo). 37 (irrigated or non-irrigated), estimated income from allotments and gardens. Coefficients / norms which vary by oblast, type of land and its quality are used for the imputation. • Incomes which do not count include: state benefits (MSB), funeral grants, other one-time benefits, in-kind and cash categorical benefits and subsidies (“privileges�); maternity leave compensations; per diems, insurance premiums in case of work accidents, and similar. Seasonal incomes from land are not considered. • Seasonal variations in incomes, if available, are not captured well with the current estimate of total income. Changes in incomes for different reasons should be reported by the beneficiary family as ‘change in the material circumstances’ compared to the period when eligibility for the UMB has been determined. Financing: The UMB is entirely financed from the republican budget. 15 The UMB expenditures have been rather volatile neither following the trends in beneficiary numbers nor in inflation (Figure 3.3). Further, the spending on UMB has not been counter-cyclical in order to mitigate the adverse impacts of stagnation and / or economic decline on the poorest. The small average amount of the UMB also undermines its ability to mitigate poverty risks. In 2005, the average UMB was 88.4 KGS which covered 4.8 percent of the MCB and in 2008, the average UMB was 127.5 KGS (USD 3.5) covering 3.6 percent of the MCB. The average UMB in 2009 is expected to be 151.2 KGS (USD 3.7). The financing for the UMB is expected to rise to allow increasing the average UMB amount to 6.3 percent of the MCB in 2011 (MTBF, 2009). Figure 3.3: Actual and real expenditures on UMB, 1998-2008 900 800 700 600 Mln. KGS 500 Actual expenditures UMB 400 Real expenditures UMB 300 200 100 0 98 99 00 01 02 03 04 05 06 07 08 19 19 20 20 20 20 20 20 20 20 20 Source: MLSD administrative data. Coverage. As of 2008, almost 8 percent of the Kyrgyz population receives UMB and thereby the UMB program is the biggest in terms of coverage. Since the start of the program the number of beneficiaries has not followed any particular trend and has both increased and decreased. Starting from 2007, however, the number of UMB beneficiaries has decreased substantially from 451,300 in 2007 to 387 thousand in 2008 (Figure 3.4). 16 The main reasons for the reduction of the number of UMB beneficiaries since 2007 are the overall income increase and the update of the coefficients for calculation of income from land plots. Figure 3.4: Number of UMB beneficiaries, 1998-2008 15 The Kyrgyz state budget consists of the republican and the local budget. 16 The MLSD forecast for 2009 is 380.000 UMB beneficiaries 38 Numbe r of UMB be ne ficiarie s, 1998-2008 (thousands) 700 600 500 400 Number of UMB beneficiaries (thousands) 300 200 100 0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Source: MLSD administrative data The absolute majority of the UMB recipients are children under 18 years (87 percent in 2007). In terms of geographical distribution, most of the UMB recipients reside in the southern oblasts of Jalalabat, Osh and Batken. THE MONTHLY SOCIAL BENEFIT Definition. The MSB is a categorically targeted benefit to disadvantaged groups that are considered having few opportunities to earn their own living. It is an income maintenance program providing a ‘social pension’ for people with disabilities, orphans, elderly without pension rights, mothers of three or more children. The MSB is extended irrespective of the income of the beneficiaries. Benefit formula. The MSB is calculated as a percentage of the GMCL and varies between 75 percent and 300 percent of the GMCL. Since 2003, a top up of 20 percent is provided in addition to the original percentages and since 2007 additional absolute top-ups are provided as well. Following the additional financing provided by the World Bank to top up the UMB, the government also decides to further top-up the MSB by 35 KGS per beneficiary. Table 3.4 lists the different categories of beneficiaries and the size of the MSB as of 2008, including the different top-ups since 2003. 39 Table 3.4: Categories of MSB beneficiaries and MSB amounts in 2008 Category of MSB beneficiary MSB amount/ percentage of the GMCL and top ups People with disabilities Children with disability (infantile cerebral palsy (ICP) (until the age of 18) 300% of GMCL +20%+335 KGS Children with disabilities (until the age of 18) 225% of GMCL +20%+235 KGS Children infected HIV/AIDS (until the age of 18) 225% of GMCL +20%+335 KGS Children up to 18 months with HIV infected mothers 225% of GMCL +20%+335 KGS People with disability since childhood – 1st group 300% of GMCL +20%+335 KGS People with disability since childhood – 2nd group 225% of GMCL +20%+235 KGS People with disability since childhood – 3rd group 150% of GMCL +20%+135 KGS People with disability as a result of general disease 1st group (without pension 225% of GMCL rights) +20%+235 KGS People with disability as a result of general disease 2nd t group (without pension 150% of GMCL rights) +20%+135 KGS People with disability as a result of general disease 3rd group (without pension 75% of GMCL rights) +20%+135 KGS Children Survivor benefit to children without pension rights 150% of GMCL +20%+135 KGS Full orphans (lost both parents) without pension rights 225% of GMCL +20%+235 KGS Elderly Aged citizens without pension rights 105% of GMCL +20%+135 KGS Aged citizens of mountainous regions without pension rights 150% of GMCL +20%+135 KGS Mother-heroines (more than 3 children) without pension rights 225% of GMCL +20%+235 KGS Source: Ibragimova (2008) and MLSD administrative data. Financing. Similarly to the UMB, the MSB is entirely financed from the republican budget. Following the increase in GMCL and the several top-ups starting from 2003, the average MSB has been increasing since its introduction and reached 365 KGS per month in 2005 and 717,2 KGS ($20) in 2008. The average size of the MSB increases almost in parallel with the administratively defined food poverty line but remains below it (Figure 3.5). 40 Figure 3.5: Food poverty line and average MSB, 1998-2008 Source: MLSD administrative data. Contrary to the UMB, which represents a ‘poverty gap’ program, the MSB is an income maintenance program. However, since the average MSB since its introduction has remained lower than the extreme poverty line, it is not able to guarantee minimum consumption (minimum food basket) to its beneficiaries. Coverage. The number of MSB beneficiaries is much lower than the number of UMB recipients however it is constantly increasing since 2004, to reach almost 60 thousand in 2008 (Figure 3.6). Figure 3.6: Dynamics of MSB recipients, 1998-2008 Number of MSB beneficiaries (thousands) 70 60 50 Thousands 40 Number of MSB beneficiaries 30 (thousands) 20 10 0 00 01 02 03 04 05 06 07 08 20 20 20 20 20 20 20 20 20 Source: MLSD administrative data The main reason for the rising numbers of MSB beneficiaries is the growth of the number of people with disabilities who qualify for the MSB, especially the number of children with disabilities from birth whose share among MSB beneficiaries is the highest. The increased disability incidence among children (Figure 3.7) is explained with the deterioration of the health care of pregnant women allowing early diagnostics and prevention of disability from birth, deterioration of nutrition and lack of vital nutritive components in the diet of pregnant and lactating women and young children, deteriorated living conditions of poor households, limited access to health care. According to qualitative research, part of the explanation is also the increased awareness of certification of the families with children with disabilities related to receiving the benefit and other supplementary support and benefits related to disability (CASE, 2008). Figure 3.7: Number of children with disabilities from birth, 2000 - 2007 41 Disabled children from birth 45000 40000 35000 30000 25000 Number of disabled 20000 children from birth 15000 10000 5000 0 2000 2001 2002 2003 2004 2005 2006 2007 Source: MLSD administrative data. Actual expenditures for UMB and MSB: The total public expenditures for UMB and MSB as a share of GDP have stayed under 1 percent of GDP since the adoption of the Law on State Benefits in 1998. The actual expenditures for state benefits (UMB and MSB) in 2007 was 1.06 billion KGS, (0.74 percent of GDP), a decrease in nominal as well as in real terms compared to 2006 (Figure 3.8). In 2008, however, the actual expenditures for state benefits increased to 1.188 billion KGS, which despite continued high inflation constitutes an increase in both nominal and real terms. 17 As a percentage of GDP, however, the expenditures on UMB and MSB decreased to 0.64. Figure 3.8: Actual and real expenditure on state benefits, 1998-2008 Public Expenditures, 1998-2008, State Benefits (UMB and SMB) 1400 1200 1000 Actual public expenditures, state Mln. KGS 800 benefits (UMB and SMB) Real public expenditures, state 600 benefits (UMB and SMB) 400 200 0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Source: MLSD administrative data. Over time the distribution of the allocated budget between the UMB and the MSB has changed substantially. In 1998, almost 88 percent of the budget was allocated to the UMB. In 2007, however, the MSB received over 34 percent of the budget while covering less than a tenth of the number of beneficiaries receiving UMB. In 2008, the allocations for UMB and MSB converged further (Table 3.9). 17 The 2008 expenditures include the UMB top-ups financed by the World Bank Additional Financing. 42 Figure 3.9: Dynamics of actual expenditures on UMB and MSB, 1998 - 2008 Public Expe nditure s, 1998-2008, UMB and MSB 900 800 700 600 Mln. KGS 500 UMB 400 MSB 300 200 100 0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Source: MLSD and MOF administrative data. CATEGORICAL IN-KIND SUBSIDIES AND BENEFITS Definition. As a legacy from the Soviet era, the Kyrgyz Republic also operates a number of in-kind subsidies and benefits (“privileges�) for specific categories of the population. The categories are either considered vulnerable or having contributed to society in a way that they qualify for special support. There are no less than 38 different categories of beneficiaries, ranging from war veterans and handicapped citizens to pensioners with merits and families living in mountainous areas. The list of the categories and the number of beneficiaries in each category in 2007 as well as in 2009 is presented in Annex 2.1. The benefits take the form of subsidies for energy, utilities, transport, sanatorium treatment, medicines, medical aids etc. Benefit formula: The benefits are differentiated by beneficiary category. Also, some categories receive the full benefit (100 percent) while others only are eligible for a certain percentage of the benefit. The over 30 categorical benefits and subsidies can be divided into five main types: municipal services, transport, sanatorium and travel, medical benefits and other benefits. Included in the municipal services are subsidies for coal, central heating, electricity, water and sewage. The transport subsidy covers within- city public transportation as well as transportation within the country, and trips once a year within the CIS. The sanatorium voucher covers trips to sanatoriums for medical treatment and rehabilitation and the medical benefits cover medicines, hearing aids, optical glasses and dental work. Other benefits include special cash bonuses extended on the occasion of 9 May 18. Depending on the category, the beneficiaries receive different ‘packages’ of benefits. Categories such as war veterans and former prisoners of concentration camps are entitled to all types of benefits. The biggest category in terms of numbers of beneficiaries - families living in mountainous areas- however, are only eligible for one type of electricity subsidy. Thus, the types and number of categorical benefits and subsidies vary substantially between the different categories of beneficiaries. In addition, also the percentage of the particular type of benefit or subsidy received varies. While, war veterans, for example receive 100 percent of all different types of benefits and subsidies, the beneficiaries from the category merit pensioners only receives 50 percent of the benefits.19 For energy subsidies, these percentages are 100, 50, 40, 30 and 25 percent depending on the category. Eligibility. Eligibility is regulated in ten laws and two resolutions introduced in the 1990s. All ‘privileges’ are categorical (eligibility based on certain “categorical� characteristics), and four are also means-tested, along with being categorical. Eligibility is proved with documents as stipulated in the legal 18 WW2 Victory Day 19 They entitled to municipal services, sanatorium trips, medicines, dental work and public transport 43 acts regulating the respective privilege. The eligibility is verified by the social workers who keep track of the beneficiaries in the respective rayon or village. The beneficiary lists are public. Financing. Historically all the categorical subsidies and benefits were granted in-kind, however, starting from 2003, a gradual monetization has taken place. As of 2008, the majority of the benefits are delivered as cash transfers, except for glasses, hearing aids, dental work and the trips within CIS. 20 Monetization is discussed in more detail in Chapter 6 of the report. Public spending on these “privileges� accounted for about 0.6 percent of GDP till 2005 and declined to 0.42 percent of GDP in 2007 and 0.37 percent of GDP in 2008 (Table 3.1). Further decline is planned for 2009 – up to 0.25 percent of GDP. The vast majority of the 38 categorical benefits and subsidies are financed through the republican (central government) budget. The categories receiving benefits from the local budgets are mothers with more than 3 children and the survivors of the Chernobyl NPP accident. Despite the shrinking number of beneficiaries, the public expenditures from the republican budget for the benefits increased substantially in nominal terms between 2001 and 2005, followed by a decline in 2006, and resumed growth thereafter, to reach 688.7 million KGS in 2008. The allocations for these benefits are expected to increase further following the monetization, where financing is linked to eligibility and not to the actual consumption of the respective privilege (eligibility and actual take-up will converge). Also, the increasing energy tariffs contribute substantially to the increasing expenditures. Detailed republican budget allocations across different types of categorical in-kind benefits and subsidies are given in Annex 2. They indicate that despite the decreasing number of beneficiaries by around 30 percent in 2006 compared to 2001, the spending has almost doubled, most notably after the start of their monetization. Figure 3.10: Actual and real expenditures on categorical in-kind benefits and subsidies, 2001 – 2008 800 700 600 Actual expenditures, 500 subsidies and benefits Mln. KGS (republican budget) 400 Real expenditures, subsidies 300 and benefits (republican budget) 200 100 0 2001 2002 2003 2004 2005 2006 2007 2008 Source: MLSD administrative data, Kyrgyz National Statistics Office Coverage. The number of beneficiaries has been steadily decreasing in the 2000s. The total number of eligible beneficiaries in 2008 was 284500 persons. In 2007, the persons who received at least one type of categorical benefit or subsidy were 280,000 down from over 400,000 in the beginning of the 2000s. The biggest group among the recipients is the families living in mountainous areas - 154,658 persons in 2008. Figure 3.11: Recipients of categorical in-kind benefits and subsidies, 2001 – 2008 20 The trips to sanatoriums are also still most often received in the form of a voucher form but can be received in cash as well 44 Number of beneficiaries of privileges 450000.0 400000.0 350000.0 300000.0 250000.0 Number of beneficiaries of 200000.0 privileges 150000.0 100000.0 50000.0 0.0 2001 2002 2003 2004 2005 2006 2007 2008 Source: MLSD administrative data. CONCLUSIONS In summary, the current non-contributory safety net in the Kyrgyz Republic contains one instrument (the means-tested UMB program) that is specifically designed to assist in smoothing the consumption of poor households with children in case of chronic and transient poverty. The system also contains instruments to assist in smoothing the consumption of certain categories of vulnerable groups – an income substitution program (the categorical MSB) for adults and children with disabilities, orphaned children, elderly without access to work-related pension and mothers of many children who did not work outside the home, along with subsidies and benefits (privileges) for categories of citizens with special contribution to the society who are also considered vulnerable. However, the safety net lacks certain elements allowing supporting households and individuals in smoothing consumption over lifetime and reducing poverty and providing a safety net of last resort for chronically or transitory poor households without children. The administrative data suggests that the UMB is not adequately funded and provides for only marginal contribution to the income and consumption of the poor families with children. Moreover, the public spending is volatile and not counter-cyclical in order to protect the poorest from risks and shocks. Given the scope and persistency of child poverty, the role of the safety net, and the UMB in particular, for mitigating the risks of child poverty and vulnerability is instrumental and of increasing importance. The UMB can potentially assume also the functions of a risk-reducing program. Currently the safety net is more important for coping with the poverty and vulnerability in rural areas where 14 percent of the population lives in extreme poverty and even to a larger extent in high mountainous areas where 21 percent of the population (1.8 times the national average) lives in extreme poverty. With the increase in food, energy and utility prices, however the risk of urban poverty – among those who are net consumers of food stuffs, users of district heating services and with higher energy consumption – could also increase. This requires a critical assessment of the UMB targeting design and implementation mechanisms to increase its relevance to the new challenges to the safety net, as discussed in subsequent chapters. 45 CHAPTER 4. TARGETING PERFORMANCE AND POVERTY IMPACT OF CURRENT BENEFITS This chapter assesses the performance of the current social protection system in the Kyrgyz Republic. To what extent do the current social protection transfers reach the poor? Are the transfers cost-efficient and adequate to ensure a minimum living standard for the most needy and reduce poverty? The chapter discusses the distributional effectiveness of social transfers by analyzing benefit coverage and the distribution and adequacy of benefits. Then the poverty reduction impact is estimated followed by an analysis of the cost efficiency of the various transfers. The targeting efficiency of the main social assistance programs is compared with similar programs in other countries to conclude that their performance is relatively good, but that there is room for improvement by reducing the exclusion errors, confining leakage to the non-poor and by increasing the level of the benefit which would eventually strengthen their poverty reduction impact. The analysis is based on data from the 2005 Kyrgyz Integrated Household Budget Survey (KIHBS) collected by the National Statistics Committee (NSC) of the Kyrgyz Republic. It provides full survey information for a nationally-representative sample of 4,771 households and 19,136 individuals. Average per capita household consumption as calculated by the NSC is used as the main household welfare indicator. The welfare quintiles to assess benefit performance are based on counterfactual household consumption per capita in the absence of transfers (25 percent substitution rate). 21 The first section reviews the distributional impacts in terms of beneficiaries (coverage, distribution) and benefits (targeting accuracy and benefit adequacy) using household survey data reported for a range of benefits (pensions, scholarships, MSB, UMB, other social insurance, and utility and housing subsidies/privileges). The distribution of transfers from private sources is included for comparative purposes. The second and third sections examine the impacts of the various transfers on poverty, as well as the cost efficiency of poverty impacts of these transfers. The final section of this chapter discusses the differences in outcomes based on administrative and household survey data. It shows that the choice of the indicators against which the performance of the targeted benefits are measured matters. In terms of implementation, social protection department staff manages to assess benefit eligibility based on income rather well. However, in terms of program outcomes, which are assessed based on household consumption, the results are slightly weaker. DISTRIBUTION OF BENEFICIARIES AND BENEFITS In order to assess the distribution and inequality-reducing effect of social transfers, we compare benefit coverage, distribution and adequacy along the welfare distribution of the Kyrgyz population. It is important to note that poverty reduction and redistribution are not the main objective for all social transfers. Pensions and other social insurance benefits aim at smoothing income over a lifetime and mitigate the risks associated with becoming old, disabled or sick. 22 Nonetheless, these benefits do redistribute income and help avert poverty among these population groups. As such, it is worth analyzing their redistributive (this section) and poverty impacts (next section).23 21 See Annex 3 for more details on the data and methodology. 22 Transfers from pensions and other social insurances have both contributory and non-contributory aspects as part of the transfers are from public government resources and social funds. 23 A few words of caution are warranted at the outset, in particular with respect to the results for coverage and poverty impacts of transfers, which tend to be under-estimated when analyzed using household survey data. There are several reasons for this, including: (a) survey samples that are intended to be representative of the national population (and certain socio-economic strata) but not specifically of beneficiaries of specific social programs; and (b) incomplete respondent recall both on receipt of transfers (participation, coverage) and on the value of income transfers received. In particular, the information on categorical subsidies and benefits (utility and housing subsidies) needs to be treated with caution as there is evidence of incomplete capture of such benefits in household survey data. 46 Coverage of Social Programs. Table 4.1 presents coverage rates for the various transfers across per capita consumption quintiles assuming a substitution rate of 25 percent. Coverage rates indicate the proportion of the population living in households receiving a certain transfer. In general, almost half of the Kyrgyz population lives in a household receiving some form of social transfer. Coverage is significantly higher among the poorer households with 74 percent of the first quintile receiving transfers compared to 23 percent of the richest quintile. Pensions have the highest coverage (34 percent of the population), while scholarships reach only three percent of the population. Coverage by pensions is three times as high for the poorest quintiles compared to the top quintile. Other social insurance benefits are less important. They cover five percent of the population in total. Coverage by social transfers is higher in rural areas and among female-headed households and increases with age of the household head. The latter is clearly driven by pension entitlements. Table 4.1: Benefit coverage of social protection benefits and private transfers, percentages, 2005 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Any social transfer 74.1 61.8 47.8 33.4 22.7 48.0 Pensions 55.5 42.9 27.6 26.1 17.0 33.9 Scholarships* 2.9 4.4 3.9 2.1 3.1 3.3 Monthly social benefit (MSB) 12.7 4.9 6.8 4.6 2.7 6.4 Unified monthly benefit (UMB) 28.2 25.6 14.2 4.1 0.9 14.6 Other social insurance benefits 7.8 9.6 4.2 3.0 1.3 5.2 Utility and housing subsidies (privileges) 16.3 10.3 9.6 10.5 6.2 10.6 Money from relatives 48.7 31.3 21.0 34.9 34.3 34.1 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 25 percent. * not significant at the 10 percent level Source: Staff calculations based on KIHBS 2005. Of the two main social assistance programs, the UMB has higher overall coverage and coverage of those in the poorest quintile than the MSB or any other non-contributory public transfer. In total, 14.6 percent of the population lives in households receiving benefits from the UMB, and 28.2 percent of the population belonging to the poorest quintile is covered by the UMB, with lower coverage for richer households. Less than one percent of the top quintile receives a UMB. Overall coverage of the MSB is lower (6.5 percent), though coverage is similarly higher among poorer households (with 12.7 percent of those in the poorest quintile receiving MSB benefits). Coverage of social assistance benefits is higher in rural areas. With respect to the UMB, relatively high beneficiary participation levels are found in Naryn, Batken and Osh oblasts, and very low coverage in Chui and Bishkek. The presence of children is a strong indicator for receipt of UMB benefits.24 Coverage by the UMB is especially high among households with three and more children. This is fully in line with the poverty profile which suggests that large households with children are more likely to be poor. Households with a head aged 20-40 or 70 and older and male-headed households have also higher UMB incidence rates. Coverage of private transfers (the ’informal safety net’) is higher than the coverage of any non- contributory social benefit. For comparative purposes, Table 4.1 also shows the coverage of private transfers / money from relatives, i.e. remittances from family members working abroad or domestically, and other money transfers. Overall, private transfers are almost equally widespread as pensions, about one third of all households reporting receipt of private transfers. An even higher share of households in the poorest quintile receives private transfers (close to half). Private transfers cover a relatively larger share of the population in Issyk-kul, Naryn and Talas oblast. Households with either very young or very old heads are also more likely to receive private support, possibly due to their added demographic vulnerability. Female-headed households are relatively more often recipients of private transfers. 24 This is by definition the case as UMB eligibility is tied to the presence of children in families. 47 Distribution of Beneficiaries. Considering the distribution of beneficiaries along the welfare distribution, Table 4.2 indicates that, except for scholarships, relatively more individuals of the poorest quintile receive social transfers. Of all social transfer recipients, 31 percent belong to the bottom quintile. The shares are highest for the MSB and the UMB. 40 percent of MSB and UMB beneficiaries belong to the bottom quintile. Table 4.2: Distribution of beneficiaries from social protection and private transfers (percentage, 2005 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Any social transfer 31.0 25.7 20.0 13.9 9.4 100 Pensions 33.4 24.9 16.3 15.4 10.0 100 Scholarships* 17.7 26.8 23.8 12.6 19.1 100 Monthly social benefits (MSB) 40.4 15.4 21.5 14.4 8.4 100 Unified monthly benefit (UMB) 39.2 34.6 19.4 5.6 1.2 100 Other social insurance benefits 30.6 36.7 16.1 11.5 5.2 100 Utility and housing subsidies (privileges) 31.1 19.2 18.2 19.9 11.7 100 Money from relatives 28.6 18.4 12.4 20.5 20.1 100 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 25 percent. * not significant at the 10 percent level Source: Staff calculations based on KIHBS 2005. Targeting accuracy (distribution of benefits). Whereas coverage rates and the distribution of the beneficiaries (Tables 4.1 and 4.2) assess program participation in terms of people (beneficiaries), in order to assess the targeting performance of the various social protection programs, we need to consider the allocation of the benefit amounts across the welfare distribution. Table 4.3 presents the distribution of the benefits of social transfers across the quintiles. Based on Table 4.2 (above), one would be inclined to conclude that the poorer groups are benefiting relatively more from social protection programs. After all, the share of beneficiaries is highest in the poorest quintiles. Considering Table 4.3 (below) , this conclusion still holds for the distribution of benefits, though to a lesser extent. Almost 30 percent of the value of all social transfers is received by the poorest 20 percent, while the richest receive only 16 percent of total social transfers. The distributional incidence of social transfers varies significantly by program, however. Categorical in-kind benefits and subsidies (privileges) and scholarships are regressive with richer households benefit proportionally more from privileges. The top quintile captures 28 percent of the value of utility and housing subsidies and 30 percent of scholarships. The distribution of pension benefits varies considerably across the spectrum, with the bottom quintile receiving the largest share (close to 29 percent of benefits) and other quintiles receiving between 15-21 percent. The distribution of private transfers resembles a U-curve. The poorest and richest households capture the largest share of private transfers (a quarter each), while the middle-income group receives only ten percent. Table 4.3: Distribution of social protection benefits and private transfers, 2005 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Distribution of social protection benefits and private transfers across groups Total benefits from social transfers 28.9 18.7 16.8 19.3 16.4 100 Pensions* 28.5 17.5 15.2 21.4 17.5 100 Scholarships 12.9 30.0 13.2 14.4 29.5 100 Monthly social benefits (MSB) 33.3 10.0 27.4 13.8 15.5 100 Unified monthly benefit (UMB) 38.0 42.5 14.9 3.4 1.2 100 Other social benefits* 16.6 32.5 12.9 22.9 15.2 100 Utility and housing subsidies* 12.6 8.7 24.2 26.5 28.1 100 Money from relatives 25.8 18.5 10.1 17.9 27.7 100 Total consumption 10.2 13.5 16.4 22.0 37.8 100 48 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 25 percent. * not significant at the 10 percent level. Standard errors in parentheses. Source: Staff calculations based on KIHBS 2005. The two non-contributory social assistance programs – UMB and MSB - are better targeted towards the poor households. The lion’s share of benefits paid out under the UMB (80 percent) is received by those in the poorest 40 percent of the population. Table 4.4: “Benefit Adequacy:� Share of benefits in total household consumption, 2005 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Benefit adequacy (ratio of benefit/consumption) for all households (including non-beneficiaries) Total social transfers 20.9 (1.6) 10.0 (0.8) 6.9 (0.8) 5.7 (0.6) 3.2 (0.4) 9.4 (0.5) Pensions 17.4 (1.7) 8.1 (0.7) 5.3 (0.6) 5.6 (0.7) 3.0 (0.4) 7.9 (0.5) Scholarships 0.1 (0.1) 0.2 (0.1) 0.1 (0.0) 0.1 (0.0) 0.1 (0.0) 0.1 (0.0) Monthly social benefits 1.0 (0.3) 0.2 (0.0) 0.4 (0.1) 0.2 (0.0) 0.1 (0.0) 0.4 (0.1) Unified monthly benefit 2.0 (0.5) 1.5 (0.4) 0.4 (0.1) 0.1 (0.0) 0.0 (0.0) 0.8 (0.1) Other social benefits 0.2 (0.1) 0.3 (0.1) 0.1 (0.0) 0.1 (0.1) 0.1 (0.0) 0.2 (0.0) Utility and housing subsidies 0.2 (0.1) 0.1 (0.0) 0.2 (0.1) 0.2 (0.1) 0.1 (0.0) 0.2 (0.0) Money from relatives 11.2 (1.4) 5.5 (1.3) 2.4 (0.4) 3.8 (0.4) 3.5 (0.6) 5.3 (0.4) Total consumption 100 100 100 100 100 100 Benefit adequacy (ratio of benefit/consumption) for beneficiary households (excluding non-beneficiaries) Total social transfers 21.1 (1.8) 16.8 (1.5) 16.5 (1.4) 22.4 (1.7) 22.9 (1.7) Pensions 25.0 (2.2) 21.4 (1.7) 19.3 (2.0) 25.4 (1.7) 26.7 (2.0) Scholarships 3.9 (1.2) 1.7 (0.5) 3.6 (2.0) 2.2 (0.4) 2.3 (0.6) Monthly social benefits 7.6 (1.4) 4.5 (0.9) 4.0 (0.9) 6.3 (1.8) 5.7 (1.8) Unified monthly benefit 7.2 (1.1) 5.4 (1.0) 4.3 (1.1) 2.1 (0.4) 1.8 (0.5) Other social benefits 0.2 (0.1) 0.1 (0.0) 0.3 (0.1) 0.2 (0.1) 0.1 (0.0) Utility and housing subsidies 1.1 (0.2) 1.3 (0.4) 2.2 (0.9) 2.0 (0.5) 2.3 (0.4) Money from relatives 15.0 (1.8) 13.6 (1.8) 16.0 (2.3) 17.5 (3.1) 15.1 (1.5) Note: Quintiles based on annual per capita consumption after transfers. Standard errors in parentheses. Source: Staff calculations based on KIHBS 2005. Benefit Adequacy (Relative Incidence). Table 4.4 assesses the “adequacy� or relative importance of social assistance transfers as a share of average consumption in each quintile. It is divided into two panels, with the upper panel presenting results for all households (including non-recipients) and the lower panel presenting results for beneficiary households only (excluding those recording “zero� transfers). Relatively speaking, overall social transfers remain more important for the poor (relative progressivity). They constitute 21 percent of total counterfactual household consumption for those in the poorest quintile, as compared to three percent for the richest households.25 The same applies to pensions. Although the poorest 20 percent receive a smaller share of the pensions, they account for 17 percent of the household budget. With regards to the targeted social assistance programs, both the UMB and the MSB are progressive in relative terms, with benefits contributing a higher share of total consumption for those in the lower quintiles than those in the richer quintiles. However, the relative magnitude of this contribution is quite low for both transfers due to low overall coverage and inadequate benefit values. The MSB constitutes less than one percent of total household consumption for all groups. The benefits of the UMB are even smaller on average, but they account for two percent of total household consumption of the bottom quintile. These transfers are, however, important to those who receive them. Considering only recipient households (after all transfers) the UMB and MSB each account for around 7 percent of total consumption among recipient households in the poorest quintile (Table 4.4). Clearly, private transfers are significantly more important for poor households. They constitute 11 percent of total household consumption in the bottom quintile. This dependence on private transfers is noteworthy since 25 This refers to the upper panel in Table 4.4 and includes all households (including non-recipients). 49 it is likely that remittances from workers abroad could fall with the onset of the global crisis thus weakening the informal safety net. Redistributive Impact. In order to assess the redistributive impact of social transfers, Tesliuc (2004) proposes three related concepts. A transfer is regressive if the poorest group receives a smaller share of total benefits than the share of the group in total consumption. A mildly progressive transfer allocates a larger share to the poorest group compared to their share in total consumption, but less than the share of the group in the total population. A transfer can be classified as being highly progressive if the poor receive a larger share of program benefits than the share of the group in the total population. Following this classification, social transfers in general are progressive. The poorest 20 percent receive almost 30 percent of total allocated social transfers. This exceeds their share in total consumption (10.2 percent), and is more than their share in the total population (20 percent). Scholarships, other social insurance benefits and privileges are all mildly progressive. Both the MSB and UMB are highly progressive social protection programs. The poorest 20 percent capture 33 and 38 percent respectively of total allocated transfers. How does the UMB targeting performance compare with other countries? The targeting performance of the Kyrgyz UMB compares favorably in terms of coverage and distribution of benefits to the bottom quintile, but remains behind in terms of benefit adequacy. Only Romania, Lithuania and Hungary manage to allocate a higher share of means-tested social assistance benefits to the poorest 20 percent of the population (Figure 4.1). 50 Figure 4.1: Targeting accuracy of the UMB compared to targeted social assistance programs, selected ECA countries* Targeting Accuracy of Social Assistance: % of Total Benefits Received by Poorest Quintile (All Non-Contributory Transfers) Source: World Bank Analysis of Household Surveys, Various Years 70% 60% 50% 40% 30% 20% 10% 0% Bu ine gia R a Ro nia Es an Hu vo P o ia er ry Ge ia jik va za an Be d tia A r ia ac ia K o ia ia Cr s rg ia b e ia Lit ijan an M ina Uk n i u lan tv n rb n an a on Az nga M an Uz uss Ky lgar so t do lar or Ka yzst oa ra st ist to e k is ba v La Se ba m m hu go ed kh ol Al rze Ta He a-i sn Bo *Data is the most recent available for the respective country and is subject to further update. Source: Lindert (2008) POVERTY IMPACT Methodology. The previous section analyzed the coverage and distributive effectiveness of the social protection benefits. In this section, we consider the impact of the transfers on absolute and extreme poverty. The poverty lines used to analyze the poverty reduction impact of social and private transfers are those developed by the NSC. We use the absolute and extreme poverty line. The value of the extreme poverty line is equivalent to the costs needed to cover a minimum of 2,100 kcal per capita per day. The absolute poverty line includes an allowance for non-food goods and services deemed necessary to cover basic needs. Both poverty lines were calculated by the NSC in 2003 and have since been updated annually using the consumer price index. We compare poverty rates before and after the respective transfers. Poverty rates (poverty headcount rate and poverty gap) are calculated for after-transfer per capita consumption and the counterfactual pre-transfer consumption. 26 The poverty headcount rate measures the percentage of the population living below the respective poverty line. The poverty gap measures the average consumption shortfall of the poor as a percentage of the poverty line. It measures the depth of poverty. The change in the poverty headcount accounts for every individual that crossed the respective poverty line due to income from a social protection transfer. By targeting individuals close to the poverty line, a social transfer can be very effective in reducing the poverty headcount without changing the average poverty gap significantly. Since the poverty gap says something about the average distance of the poor to the respective poverty line, a social transfer is more effective in reducing the poverty gap if it manages to reach the poorest and close their poverty gap fully or to some extent. Strictly speaking, a social transfer can be very effective in closing the poverty gap without changing the poverty headcount rate. Table 4.5 presents both the absolute reduction in percentage points and the relative reduction as a percentage of the rates before transfers. In the absence of any social transfers, poverty rates would be much higher. Both in absolute and relative terms pensions are most effective in reducing poverty. Due to their nature of replacing income in old age, pensions are significantly higher than non-contributory transfers, especially if the value of the pension is related to previous income. Pensions are the largest social transfer program in terms of resources allocated from general government budget and the social fund. However, poverty reduction is 26 We use the standard Foster-Greer-Thorbecke family of poverty measures (Foster et al., 1984). 51 not the only objective of pension programs. They are primarily meant to redistribute income over the life cycle. Pensions foresee the elderly with an income after retirement. They mitigate the risk of living in poverty in old age after the working life. The second most important transfer in terms of poverty reduction impact are private transfers. Money received from relatives reduces the poverty headcount by 7 percent and achieves a relative reduction of the poverty gap of 15 percent. Private transfers are even more effective in reducing the extreme poverty gap. Thanks to these transfers, the extreme poverty gap is reduced by 40 percent. This finding is of particular interest in the current context, since remittances (one form of private transfers) are likely to fall with the onset of the global crisis and its impact on migrant workers. In that situation, targeted safety net transfers (such as under an expanded UMB) may be needed to compensate for these possible losses in incomes under the crisis. In their current form, targeted social assistance benefits are less effective in reducing poverty due to inadequate benefit values and coverage, although they out-perform all other social protection benefits besides pensions. While the impact with respect to the absolute poverty line is limited both for the UMB and the MSB (both reduce the headcount rate by less than one percent), the UMB succeeds in reducing the extreme poverty gap by 12 percent. Note that the value of the GMCL determining UMB eligibility is lower than the extreme poverty line. 52 Table 4.5: Impact of social protection benefits and private transfers on poverty, 2005 Poverty rate Poverty rate Absolute Relative Type of benefit before benefit (%) after benefit (%) reduction reduction (%) Total social transfers 49.1 43.4 5.7 11.6 Pensions 48.0 43.4 4.7 9.7 Monthly social benefits 43.5 43.4 0.1 0.2 Unified monthly benefit 43.8 43.4 0.5 1.1 Other social benefits 43.5 43.4 0.2 0.4 Utility and housing subsidies 43.4 43.4 0.1 0.1 Money from relatives 46.7 43.4 3.3 7.1 Poverty gap before Poverty gap after Absolute Relative benefit (%) benefit (%) reduction reduction (%) Total social transfers 14.0 10.4 3.6 25.7 Pensions 13.3 10.4 2.9 21.9 Monthly social benefits 10.6 10.4 0.1 1.2 Unified monthly benefit 10.8 10.4 0.4 3.6 Other social benefits 10.5 10.4 0.1 0.7 Utility and housing subsidies 10.5 10.4 0.0 0.3 Money from relatives 12.2 10.4 1.8 14.5 Extreme poverty Extreme poverty rate before benefit rate after benefit Absolute Relative (%) (%) reduction reduction (%) Total social transfers 17.2 10.9 6.3 36.6 Pensions 16.2 10.9 5.3 32.6 Monthly social benefits 11.6 10.9 0.7 6.4 Unified monthly benefit 11.3 10.9 0.4 3.1 Other social benefits 10.9 10.9 0.0 0.1 Utility and housing subsidies 11.0 10.9 0.1 0.5 Money from relatives 13.0 10.9 2.1 16.1 Extreme poverty Extreme poverty gap before benefit gap after benefit Absolute Relative (%) (%) reduction reduction (%) Total social transfers 3.5 1.6 1.9 54.3 Pensions 3.1 1.6 1.5 47.8 Monthly social benefits 1.7 1.6 0.1 6.3 Unified monthly benefit 1.8 1.6 0.2 11.6 Other social benefits 1.6 1.6 0.0 1.2 Utility and housing subsidies 1.6 1.6 0.0 0.8 Money from relatives 2.7 1.6 1.1 40.1 Note: Poverty rates before transfers, assuming a marginal propensity to increase consumption of 25 percent. Source: Staff calculations based on KIHBS 2005. Taking the reduction of the extreme poverty headcount, the Kyrgyz Republic compares favorably with other Eastern European countries, but remains below the EU average. The Kyrgyz system performs below most of the presented countries if we take the reduction of the absolute poverty rate as the comparative standard. Figure 4.2 summarizes the relative impact of social and private transfers with respect to the poverty rate and gap both for the absolute and extreme poverty line. Figure 4.3 gives an indication of the poverty reduction effectiveness of social transfers in countries of the European Union. The results are not directly comparable with the results presented for the Kyrgyz Republic in Figure 4.2 and Table 4.6, but give an indication of the extent of the poverty reduction impact of other social policy systems. The methodology to measure household consumption differs. Furthermore, poverty lines to assess pre- and post transfer poverty are based on a relative standard, and pre-transfer consumption assumes zero substitution of foregone income. Figure 4.2: Relative reduction of poverty headcount and poverty gap for different transfers, 2005 53 60.0 50.0 40.0 30.0 20.0 10.0 0.0 Headcount Poverty gap Headcount Poverty gap Absolute poverty Extreme poverty percent Pensions Monthly social benefits Unified monthly benefit Other social benefits Utility and housing subsidies Money from relatives Source: Staff calculation based on KIHBS 2005 Figure 4.3: Relative reduction of poverty headcount*, selected EU countries, 2005 Poverty reduction, EU countries, 2005 HUN CZR SLK POL LAT EST LIT SWE NOR DMK FIN FRA AUT NL ICE BEL GER LUX EU15 EU25 IRE EU… MAL CYP POR IT ESP GRE 0 10 20 30 40 50 60 70 80 percentage (relative change) Source: Staff calculations based on data from Eurostat, *Compares poverty rates before and after social transfers. Before social transfer income is including pension, but excluding all other social transfers COST EFFICIENCY Methodology. Cost efficiency measures for each program how much it costs to reduce the poverty gap by 1 KGS. It is a simplified cost-benefit analysis, where the benefits (that is the reduction of the poverty gap in KGS) are weighted against the program costs. The program costs consist of the actual transfers made, the costs incurred to administer the programs and other (hidden) costs incurred by the 54 beneficiary. 27 The KIHBS data only provide information on the transfers made. We lack information both on the administrative costs and the costs incurred by the beneficiaries. Therefore, the cost-benefit ratios presented in Figure 4.4 are calculated by dividing the actual program costs (total value of transfers per benefit) by the reduction of the poverty gap (difference between the counterfactual and observed poverty gap in KGS). The resulting ratio gives the amount of KGS spent per 1 KGS reduction in the poverty gap. The lower the cost-benefit ratio, the more efficient the program is. We provide estimates both for the absolute and extreme poverty line. The social assistance programs are most efficient in reducing the poverty gap, and the poverty reduction is achieved with relatively lower transfers. Figure 4.4 presents data that confirms that the social assistance programs are most efficient in reducing the poverty gap, although the total amount of transfers is far lower than, for example, for pensions. The UMB costs 1.6 and 4.7 KGS to reduce the absolute and extreme poverty gap by 1 KGS respectively. Note that these ratios present a lower bound since we did not include administrative and other costs. Categorical in-kind benefits and subsidies and other social insurance benefits are much less efficient in reducing the poverty gap. The results also indicate that it is more costly to reduce the extreme poverty gap. This is related to the fact that the extremely poor are hard to target and that the current means-test is less successful in identifying the poorest households. Figure 4.4: Cost-benefit ratios of social protection benefits Absolute Poverty Other social benefits 7.6 Privileges 6.3 Pensions 3.3 MSB 3.0 UMB 1.6 -10.0 10.0 30.0 50.0 Som spent per 1 Som reduction in poverty gap Extreme Poverty Other social benefits 44.2 Privileges 27.3 Pensions 10.2 MSB 5.5 UMB 4.7 0.0 20.0 40.0 60.0 Som spent per 1 Som reduction in poverty gap Source: Staff calculations. Comparing the results with earlier studies (Tesliuc, 2004), we find that the cost efficiency has decreased. In 2001, the cost-benefit ratio for the UMB was 1.2 (1.7) KGS to close the absolute (extreme) poverty gap. The ratio for pensions was 1.9 (3.7) KGS respectively. The results are not directly comparable, 27 For example, transport costs to visit the local office, costs to obtain certificates, opportunity costs of time spent to apply for the benefits. 55 though. Pensions, for example, have been increased several times over the four years and to a larger extent than the value of the poverty lines. DIFFERENCES IN OUTCOMES BASED ON ADMINISTRATIVE AND HOUSEHOLD SURVEY DATA The analysis based on household survey data is not without limitations. The KIHBS is a very complex survey. It consists of seven different forms, some of which are applied annually and others quarterly. Households are visited four times a year and they have to keep diaries during several weeks. Some data files seem to indicate that there is ‘survey fatigue’ towards the end of the year.28 Aggregate values drop significantly in the fourth quarter for certain variables. Ideally, the results obtained from the survey in terms of number of beneficiaries and total amount spent per transfer match the data from administrative sources. This is not the case as Table 4.6. shows. Several reasons cause the observed discrepancies. First, it is not always possible to identify the individuals in a given household eligible for a certain benefit. For example, the MSB is for disabled and elderly without other pension entitlements, but the survey does not allow identifying disabled household members. The same applies to individuals eligible for privileges. Other benefits, such as the UMB use the family as unit to assess benefit eligibility. The KIHBS provides information at the household level. One household can consist of more than one family in terms of the UMB. The data do not allow to clearly distinguish between different families within one household. Secondly, income from transfers is underreported, which is not unusual for household surveys. Respondents forget transfers, report not the full amount, or are simply not aware of all transfers received by different household members. Especially problematic is the information on privileges and subsidized goods and services. The KIHBS collects data on housing and utility benefits, but the benefit amounts are difficult to separate from the total bills paid. No information is available on privileges related to medical care, transport, education, etc. Therefore, we suspect that the estimates presented above significantly underestimate the value of privileges and subsidies. 28 This observation was confirmed by the NSC (oral communication, 17 October 2008). 56 Table 4.6: Comparison of program expenditures according to administrative and survey data, 2005 Administrative data Survey data Type of benefit (million KGS) (million KGS) Absolute Difference Pensions 4748.7 4735.5 13.2 Monthly social benefits 219.0 188.7 30.3 Unified monthly benefit 412.1 313.6 98.5 Categorical in-kind subsidies and benefits 694.8 110.4* 584.4 * In the survey data only utility and housing subsidies are included. Source: MOF and Social Fund for administrative data; survey data based on KIHBS 2005. The choice of the performance indicator matters: program compliance versus program outcomes 29. Earlier sections in this chapter assessed the performance of the UMB in terms of program outcomes using per capita consumption as the main welfare indicator. Program compliance measures how well a program is implemented given its own rules for calculating income and valuing assets, and determining eligibility. UMB eligibility is assessed based on the administrative definition of household income. 30 In principle, income from all sources is taken into account for the eligibility assessment. However, certain income is disregarded (MSB, subsidies and income from livestock), and income from land is estimated based on the land owned by the household using land coefficients. Using annual values, one would expect a close match between the values of per capita consumption and income. In the Kyrgyz Republic, the correlation between annual per capita consumption and administrative income is moderate. The Pearson correlation coefficient is 0.65 for administrative income and consumption.31 The fact that the UMB is evaluated against consumption partly explains the inclusion error. Table 4.7 compares the targeting performance using both per capita consumption and administrative income per welfare decile. 32 Program coverage of the poorest ten percent of households is considerably higher using administrative income as performance indicator. Close to half (53 percent) of the households from the poorest (per capita income) decile receive an UMB compared to 31 percent when using per capita consumption. Some 60 percent of all the beneficiaries belong to the poorest twenty percent of households according to the administrative measure. They receive 77 percent of the total UMB allocated. These values are significantly higher than the values for per capita consumption. 29 This section draws on input from E. Tesliuc and P. Leite (Background note on Livestock income, July 2008). 30 Administrative income differs from a purely economic definition which would include all income sources, including a direct estimate of agricultural income from crops and livestock. It represents a second-best option to assess eligibility. 31 A perfect match would have a correlation coefficient equal to 1. 32 Administrative income used here represents a proxy as it is based on the survey data. 57 Table 4.7: Program compliance versus program outcomes Deciles of Welfare Poverty Status Total D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 EP AP NP Coverage Per capita consumption 14.6 31.2 23.8 19.2 21.7 27.2 11.7 7.1 2.1 0.8 1.0 n.a 31.2 12.7 Per capita UMB income 14.6 53.0 34.9 15.4 12.9 12.9 6.4 3.4 3.3 2.7 0.9 56.8 29.3 4.9 Distribution of beneficiaries Per capita consumption 100.0 21.7 16.3 13.0 14.8 18.6 8.1 4.8 1.4 0.5 0.7 n.a 22.1 77.9 Per capita UMB income 100.0 36.3 24.2 10.5 8.8 8.7 4.4 2.3 2.2 1.8 0.6 14.3 79.9 20.1 Distribution of benefits Per capita consumption 100.0 25.6 11.0 16.0 14.2 18.6 9.2 3.0 1.2 0.6 0.6 n.a 27.0 73.0 Per capita UMB income 100.0 57.0 19.8 5.7 5.1 6.0 2.9 1.1 1.6 0.5 0.3 23.4 87.6 12.4 Note: EP = extreme poor; AP = absolute poor; NP = not poor. UMB income is administrative income estimated using the KIHBS 2005. Source: Tesliuc & Leite (2008), based on KIHBS 2005. In terms of program compliance (based on administrative income per capita) the UMB performs even better than in terms of program outcome (based on per capita consumption). The discrepancy with program outcome is due to the correlation between household consumption and income. Reported income is lower than consumption, especially for the poorer households. To a certain extent this is due to the omission of incomes from livestock and the use of a coefficient for land ownership instead of actual income from crop production. Furthermore, income tends to be underreported. For example, total income from social transfers reported in the survey is significantly lower than the resources allocated based on administrative data. From an administrative perspective, income is difficult to be verified by social workers and may not fully reflect the true living standard of the households. Figure 4.5 distinguishes between easy- and hard-to-verify incomes along the income distribution. Easy- to-very income includes those income sources which can be easily confirmed with documentation (wages, pensions, stipends, government transfers). Hard-to-verify income is the one which is either received in-kind (income from land, live stock, unpaid household work, in-kind private transfers) or is in cash but not reported (income from sale of agricultural production, earnings in the informal sector, private transfers outside the banking system). Especially in rural areas, a significant share of total household income is difficult to verify, except for the very poor households (poorest 5 percent). In urban areas, the situation is different as most households rely on income from employment which is easier to verify. The analysis illustrates the difficulties a social worker is confronted with when assessing benefit eligibility. Although the poorest households are less likely to depend on hard-to-verify income, this does not apply to other households, especially in rural areas. The share of easy-to-verify income in households belonging to the second to the fifth decile is around 50-60 percent, which makes it difficult to reduce the inclusion error with a means test that is mainly based on easy-to-verify incomes. 58 Figure 4.5: Share of hard-to-verify income per income ventile (NB: the numbering of the figures below will be corrected in the final version) Figure 1a: Share of Easy-to and Hard-to-verify incomes: Kyrgyzstan Figure 1b: Share of Easy-to and Hard-to-verify incomes: Kyrgyzstan - Urban 100 100 80 80 Share (%) Share (%) 60 60 40 40 20 20 0 0 Easy-to-Verify Hard-to-verify Easy-to-Verify Hard-to-verify Figure 1c: Share of Easy-to and Hard-to-verify incomes: Kyrgyzstan - Rural 100 80 Share (%) 60 40 20 0 Easy-to-Verify Hard-to-verify Source: Staff calculations. CONCLUSIONS Program participation (coverage and distribution of beneficiaries) in social protection is extensive in the Kyrgyz Republic. Almost half of the population receives some form of social transfer either from social insurance or social assistance. Coverage is significantly higher among the poorer households with three quarters of the households in the poorest quintile receiving transfers compared to less than one quarter in the richest quintile. Pensions have the highest coverage. Coverage by social transfers is higher in rural areas and among female-headed households and increases with age of the household head indicating the importance of pension entitlements. Compared to the formal safety net, the private transfers (the ’informal safety net’) have higher coverage than the coverage of any non-contributory social benefit. Among the safety net programs, the UMB has the highest overall benefit coverage and coverage of those in the poorest quintile. In total, 14.6 percent of the population lives in households receiving UMB. 28.2 percent of the population belonging to the poorest quintile is covered by the UMB compared to less than one percent of UMB recipients in the richest quintile. Benefit coverage by the UMB is especially high among households with three and more children which is fully in line with the poverty profile. Benefit coverage is also higher in rural areas, also in line with the predominantly rural profile of the poor. Households with a head aged 20-40 or 70 and older and male-headed households have also higher UMB incidence rates. The current UMB performs well compared to poverty-targeted programs in other countries in Eastern Europe, and taking into account the country situation (large agricultural sector, high level of subsistence agriculture, informal economy, remittances). It is a progressive transfer (larger share received by the poor), although performance in 2005 declined compared to 2001. The UMB is also relatively cost- efficient transfer compared to other social transfers. The UMB is fairly accurate in channeling benefits to poor, who account for 43 percent of the population: 59 • In terms of targeting accuracy (benefit incidence with the population ranked by per capita consumption), 38 percent of the total benefit amount goes to those in the poorest quintile, and another 43 percent goes to those in the next quintile, such that the poorest 40 percent of the population receives 81 percent of total benefits paid out. Leakages to the non-poor are low: only 19 percent of benefits go to those in the top three quintiles; • In terms of beneficiary incidence, close to 75 percent of the UMB beneficiaries belong to the poorest 40 percent of the population; • In terms of program compliance (ranking the population according to per capita income calculated according to the methodology used to administer the program), the UMB performs even better: 60.5 percent of the beneficiaries belong the poorest 20 percent based on administrative income per capita receiving 77 percent of total UMB benefits; • In terms of program coverage In terms of program coverage (based on per capita consumption) UMB is received by 28.2 percent of the poorest quintile and 25.6 percent of the second poorest quintile. This coverage is not particularly strong and implies leakage of part of the benefit amount to non-poor. 33. • In terms of benefit value, the UMB is not a generous transfer. The value of the Guaranteed Minimum Consumption Level (GMCL) which determines UMB eligibility is lower than the extreme poverty line. The UMB accounts only for 7 percent of total household consumption in poorest households receiving the UMB. The UMB reduces the extreme poverty rate from 11.3 to 10.9 percent, equaling to a relative reduction of just 3 percent. While the impact with respect to the absolute poverty line is limited, the UMB is more successful in reducing the extreme poverty gap - by 12 percent. • In terms of cost efficiency, poverty reduction is achieved with relatively lower UMB transfers. The UMB costs 1.6 and 4.7 KGS to reduce the absolute and extreme poverty gap by 1 KGS respectively (administrative costs are not included). Table 4.8: Social Safety Net: Summary Performance “Scorecard� Program Spending / Coverage Targeting Benefit Poverty Poverty Cost share of share of accuracy / generosity / impact / impact / efficiency/co GDP poorest Q share of share of relative relative st per 1 (%, 2007) (%, 2005) benefits benefit in reduction reduction KGS captured consumpti of extreme of extreme reduction of by Q1 (%, on (%, poverty poverty poverty and 2005) 2005) gap (%, rate (%, extreme 2005) 2005) poverty gap UMB 0.53 28.2 38.0 7.2 11.6 3.1. 1.6 / 4.7 MSB 0.21 12.7 33.3 7.6 6.3 6.4 3.0 / 5.5 Categorical benefits and 0.6 16.3* 12.6* 1.1* 0.8* 0.5* 6.3 / 27.3 subsidies Pensions 5.1 55.5 28.5 25.0 47.8 32.6 3.3 / 10.2 *Only for utilities and housing subsidies. Source: Staff calculations based on KIHBS 2005 and administrative data from MOF, MTBF. The comparison across safety net programs indicates the UMB has the strongest performance compared to the MSB and especially the rest of the non-contributory benefits. The MSB also performs relatively well. It is a smaller program in terms of beneficiary incidence and budget outlays, has a higher poverty impact in terms of relative reduction of the extreme poverty rate and comparable accuracy of targeting 33 When assessing the UMB coverage gap, one should bear in mind that the UMB is a last resort program that complements a number of other, some of which more generous, social protection programs, notably pensions, and to a lesser extent other social insurance benefits, and scholarships. These benefits are received by 66.2 percent in the poorest and by 56.9 percent in the second poorest quintile. Any kind of contributory or non-contributory social transfer is received by 74.1 percent of the poorest quintile and 61.8 percent of the second poorest quintile receive any social transfer, and. Consequently, those who are not covered by any social protection program represent 25.9 percent of the poorest 20 percent of the population. Part of this consists of households where the household members are not entitled to pensions, disability benefits or categorical benefits and subsidies. Other part consists of childless households. In both cases these households are not eligible for social benefits by program design. There are also households which meet the eligibility criteria but do not receive UMB. 60 the poorest quintile. Similarly to the UMB, the MSB accounts only for 7.6 percent of total household consumption in poorest households receiving the MSB, however its contribution to relative reduction of the extreme poverty rate is higher - 6.4 percent. In terms of targeting accuracy, the MSB without being explicitly targeted to the poor, delivers one third of total MSB amount to those in the poorest quintile. At the same time the leakage to the non-poor is high with close to 30 percent of MSB going to those in the top two quintiles. Of the two “new� non-contributory social assistance programs, the UMB has higher overall coverage and coverage of those in the poorest quintile than the MSB or any other non- contributory public transfer. Overall coverage of the MSB is lower (6.5 percent), though coverage is similarly higher among poorer households with 12.7 percent of those in the poorest quintile receiving MSB benefits. At the same time the UMB performance score card highlights the programs main weak points: low coverage and, respectively, significant errors of exclusion and low benefit generosity leading to a limited poverty impact. With the existing foundation of respectable performance, there is still room to further strengthen the UMB by reducing the exclusion errors, confining leakage to the non-poor and by increasing the level of the benefit (benefit adequacy or generosity) which would eventually strengthen its poverty reduction impact. Coverage of poor households could be expanded through improved outreach methods and active involvement of social workers in information dissemination and rapid assessment of poverty and vulnerability in order to reach more eligible households. Social Passports can be used for mapping all poor and vulnerable, for monitoring vulnerability and for policy planning. Coverage could be also improved with enhanced fiscal efforts (preferably with funding coming from a downsizing and consolidation of other non-targeted to the poor social programs). Furthermore, coverage could be increased by extending eligibility for a UMB to poor households without children. Finally, by allocating more resources, the benefit value of the UMB could be raised thereby increasing the impact on poverty reduction. 61 CHAPTER 5. FURTHER STRENGTHENING THE UMB – OPTIONS FOR CONSIDERATION The previous chapter concluded that the UMB has the strongest targeting performance among the non- contributory benefits and that with the existing foundation of respectable performance, there is still room for its further strengthening by reducing the exclusion errors, confining leakage to the non-poor and by increasing the level of the benefit which would eventually strengthen the poverty reduction impact. The coverage of poor households could be expanded through improved outreach, improved capacity of social workers and more active use of the Social Passport. Coverage could be also improved with enhanced fiscal efforts (preferably with funding coming from downsizing and consolidation of other non-targeted to the poor social programs). Furthermore, coverage could be increased by extending eligibility for a UMB to poor households without children. Finally, by allocating more resources, the benefit value of the UMB could be raised thereby increasing the impact on poverty reduction. Improved targeting would further reduce leakages to the non-poor, thereby freeing up resources for expanded coverage of the poor and/or increased unit benefits. This chapter simulates a number of alternative targeting options for the UMB with the objective of improving its effectiveness and efficiency, and discusses the rationale for them. What policy options would improve the targeting effectiveness and the poverty reduction impact of social protection transfers considering the given budget constraints? Policy options focus on alternative targeting methods, which would reduce the exclusion error; better target the poor (reduce leakage to the non-poor); and increase the poverty reduction impact. Several options to improve targeting are explored. First, we discuss the potential of modifying the current means test by including income sources, which are not currently considered in the UMB means testing formula. Second, we simulate alternative scenarios / targeting options: (i) categorical targeting (universal child benefit); (ii) proxy-means testing (PMT); and (iii) a combination of categorical targeting and PMT. Since the results of the simulated scenarios cannot be compared with the actual performance of the UMB as they assume perfect implementation and take-up, we also simulate and include as a counterfactual an optimal UMB serving as a theoretical benchmark for the alternative scenarios. Last, we argue for considering a hybrid means test as a way of improving the targeting of the UMB and reducing the leakage to non-poor. In terms of assumptions, the simulations assume 100 percent take-up among the eligible population and perfect program compliance and implementation. These assumptions allow us to analyze the impacts of specific policy reforms in isolation, but are clearly unrealistic in any situation since implementation is never perfect and, in the case of the Kyrgyz Republic, given the low level of staffing and the low level of administrative funds, it cannot be expected that these assumptions will be matched after implementation. For a reality check, any reform of the targeting mechanism would have to be tested in pilots in order to gauge the administrative challenges and the actual reform impact in program outcomes. As such, the presented analysis also serves as an example how survey data can be used to simulate alternative policy options in a simplified cost- benefit framework. Finally, the chapter discusses options to improve the administration of the UMB, thereby further increasing the performance of targeted benefits for the poor. MODIFYING THE EXISTING MEANS-TESTING FORMULA The rationale for modifying and improving the existing UMB means-testing formula is reducing the exclusion of eligible recipients from the program and the leakage of benefits to the non-poor. One of the Government’s policy objectives is to improve the existing system of targeting the poorest34 inter alia by better registering the disposable household income. As a first step, in September 2007 MLSD updated the coefficients used to estimate (impute) the value of income from owned land in the aggregate household income. This resulted in a considerable decrease of the number of eligible UMB beneficiaries - from 10 percent of the population before the change to 7 percent in 2008. The objective of the update was mostly to reduce the errors in assessing the presumptive income from land (plots and 34 GoKG, Social Development Concept 2007-2012 62 yards/gardens) by applying more sophisticated and diversified indicators 35. The improved update of imputed income is now based on categorization of the land by region and by quality / productivity (unusable, non-arable, arable, also irrigated and non-irrigated. The categorization is done by specialized bodies of the regional agricultural departments and approved by GoKG. The village administrations (aiyl okmotu) bear the responsibility for validity of the data in the issued certificate with respect to land size and quality. The next step towards more precise assessment of overall household / family income is to include estimates of income / revenues from owned livestock. Imputing livestock income is under consideration by the MLSD and stakeholders 36 for quite some time but is not yet introduced. According to the Law on State Benefits, the means test applied currently to define eligibility for receiving UMB takes into consideration income from wages (main job and supplementary jobs), pensions, scholarships, entrepreneurial activity, income from membership in agricultural associations, income from agricultural land and subsistence farming (yard or garden plot). As stated in chapter 3, other incomes are excluded, e.g. incomes from all non-contributory social transfers, and the income from owned livestock. The proponents of including presumptive income from livestock emphasize on the fact that livestock supports food consumption. Households owning livestock could be poor but it is less probable for them to be extremely poor. The opponents emphasize on the limited possibilities to ‘commercialize’ / trade livestock and its products and generate monetary incomes. The owners of livestock remain poor in terms of cash (CASE, 2008). They also raise the issue of reliability of the data on livestock ownership. The village registers of owned livestock are often out-of-date. The KIHBS data on livestock ownership often shows wide and hard-to-explain variations across different years and across the quarters of one calendar year. Including revenues from livestock in the current methodology of assessing household income is potentially relevant because a large share of the households in the Kyrgyz Republic (41 percent) own livestock. At the same time, this income is hard to verify and susceptible to underreporting. Income from livestock is typical for rural and mountainous areas, where – as shown in the previous chapter - a significant share of total household income is difficult to verify / assess which makes it difficult to reduce the inclusion error with a means test that is mainly based on easy-to-verify incomes 37. If the income from livestock is taken into account, the UMB targeting efficiency would improve as UMB beneficiaries earning income from livestock would be excluded. Undoubtedly, accounting for livestock would exclude some currently eligible households from receiving the UMB, but it would not necessarily decrease the exclusion error. This could happen indirectly – by freeing up resources which can then be allocated to eligible but not receiving UMB households. INTRODUCING INDICATORS: CATEGORICAL TARGETING AND PROXY-MEANS TESTING In this section we simulate three targeting options / scenarios for the ‘variable’ UMB with the objective of improving its effectiveness especially in terms of coverage of the poorest. More explicitly, we are looking for alternative targeting methods, which (i) reduce exclusion errors, (ii) better target the poor (reduce leakage to the non-poor to make space for inclusion of more poor households), and, (iii) increase the poverty reduction impact.38 We recognize that from an administrative point of view, the proposed targeting methodology should be simple to administer, easy to communicate, difficult to manipulate, cost efficient, with low costs for applicants and, last but not least, be politically feasible. We do not simulate options for reforming the MSB (which is increasing over the years in real terms and 35 Regulations “On the order of assessment of the aggregate family (household) income of citizens for entitlement to UMB by disadvantaged families and citizens� and Methodological instructions for the applications of the Regulations, approved by the Jogorku Kenesh of the Kyrgyz Republic as of June 25, 2007. 36 The imputation of the income from livestock has been subject to discussion in working groups in MLSD and with representatives of the President’s Office. Interviews with field social workers (CASE, 2008) indicate that they are convinced in the need of accounting for the revenues from livestock. In some cases they admittedly make own adjustments of household welfare situation depending on the availability of livestock despite of the missing legal grounds for that. Last but not least, the European Commission provides budget support of € 24 million to the GoKG in the context of the Social Protection Support Strategy 2007-2009. Condition 3 of the agreement requires that targeting of social assistance is improved. The reform proposal foresees in (i) updating of the norms for income calculation to exclude non-eligible applicants, (ii) improving access and application procedures to include the poorest, (iii) increasing the level of the allowances, and (iv), improving the eligibility criteria in order to limit overlap with other social assistance schemes. One of the options considered to improve the income calculation has focused on updating the coefficients used to estimate revenues from crops, and including a coefficient for revenues from livestock. 37 More details on livestock ownership and its link to poverty is discussed later in this chapter. 38 Note that additional objectives of a well-designed targeting methodology are the avoidance of negative work incentives and benefit dependency. The available data do not allow assessing the impact of alternative methods in that respect. 63 relative to the UMB) and the lump-sum UMB because the KIHBS does not provide data on who are the beneficiaries of these programs and also because the small number of observations would prevent arriving at statistically robust findings. The main alternatives considered for targeting social assistance benefits to the poor are: categorical targeting (scenario 1), proxy-means testing (scenario 2) and a combination thereof (scenario 3). Essentially, the categorical and proxy-means testing approaches are similar as they are based on the use of indicators that are easy to observe and difficult to manipulate for the applicants. Categorical targeting selects beneficiaries based on a single indicator, while proxy-means testing uses a combination of different indicators that are closely related to the risk of living in poverty. The outcomes of the above targeting options are compared with the hypothetical outcome of the simulation of performance of the existing means test under optimal condition. This comparison allows understanding whether, and which of the simulated options would bring improvement relative to the means test. Categorical targeting (Scenario 1) One of the strongest indicators for poverty in the Kyrgyz Republic is the presence of children in a household. Children have a higher than average risk of living in absolute and extreme poverty. The more children in a household, the higher the poverty risk for all individuals living in this household. Research has shown that family benefits (categorical benefits for children) can be effective in reducing poverty. In some countries, they even outperform targeted social assistance benefits (Grosh et al 2004; Gassmann & Notten 2008). Categorical benefits to children are easy to administer and reduce the cost burden for the applicants as the application procedures are simplified and requiring less documentation. We simulate the targeting performance of a categorical child benefit.39 The value of the monthly benefit per child is based on the available annual budget divided by the number of children of the respective age group. Monthly values are the annual benefit divided by 12. The results are described below. Proxy means targeting (Scenario 2) Allocating cash transfers based on a single indicator – such as children (categorical targeting) -- is simple and limits the administrative costs involved. However, both poor and non-poor households are eligible for the transfer thereby resulting in a sizeable inclusion error. If the budget is given, more recipients mean lower benefits on average. An argument against categorical cash transfers frequently refers to the latter. An alternative approach to target poor households is to combine several indicators and develop a proxy means test. Chile was the first country to implement a proxy means test in 1980 (Grosh et.al. 2004). Since then, the tool has been refined and is currently used in a number of countries across the world, mainly in Latin America, but also, for example, in Armenia. Proxy-means tests (PMT) use a limited number of indicators that are easily observable, difficult to manipulate by households and correlated with poverty. Based on statistical analysis household characteristics (demographic composition, labor market status, possession of assets, housing characteristics, location) are identified that strongly correlate with poverty. These indicators are used to calculate a score indicating a household’s economic welfare. The score can be used to determine eligibility for social transfers and services and the size of the individual benefit. Annex 4 provides a detailed description of the methodology for estimating proxy-means tests and comes up with a PMT formula. As with the standard means-test, a cut-off point has to be selected below which households become eligible for social transfers. For the purpose of this simulation, we identify the poorest 10 percent of households as eligible for a cash transfer. This is the same percentage as currently covered by the UMB based on the survey data. The simulated cash transfer is a flat rate benefit for each household member. 39 An alternative to targeting by age is to target benefits geographically. Based on a detailed poverty map, the poorest aiyl okmotu could be selected for social assistance benefits. Within the aiyl okomotu eligible beneficiaries are then selected based on another targeting method. This can be based on a means-test, other categorical indicators, proxy means tests or by community targeting, where the community (for example the council of the elder) decides which households are most in need. Geographical targeting cannot be simulated based on the survey data only as we lack the necessary information on household welfare at the local level. 64 While the current simulation use the proxy-means test as a stand-alone targeting method, its features can also be used to improve the current means test. Proxy means testing could be used as a means of verification of self-reported family income, or a proxy-means test could be used to introduce additional filters for a pre-selection of potentially eligible households40. Targeted child benefit based on PMT (Combination of 1 and 2, Scenario 3) For the final scenario we combine a categorical benefit for children with the proxy-means test. This means we limit the eligibility for child benefits to the poorest households, which are identified by proxy means. Combining proxy-means targeting with categorical targeting methods is not uncommon. In Mexico, the PROGRESA program uses geographical targeting combined with a proxy-means test to identify eligible families within the targeted regions. In the Chilean unified family subsidy program, proxy-means testing is combined with categorical targeting of children (Grosh, et.al. 2004). For the present simulations, we assign a flat rate benefit to each child younger than 16 (or 21 if attending school) living in an eligible household (bottom 10 percent). Again, the flat rate monthly benefit is set at a budget neutral level. Total annual resources are divided by the number of children living in eligible households as identified by the proxy-means test. Evidently, the larger the target group, the smaller the individual benefit. A comparison of the performance of alternative scenarios Methodology. For the simulations of alternative targeting methodologies, we keep the available budget constant (Table 5.1). According to the MLSD Strategy for 2006-2008, the UMB budget for 2007 is 1,008 million KGS. 41 The pre-transfer household income per capita is based on total household consumption minus 75 percent of the UMB, assuming a substitution rate of 25 percent in the absence of current transfers. Simulated household consumption after transfer is computed by adding 75 percent of the new transfer to pre-transfer household consumption, respecting the same 25 percent substitution rate. As the data from the KIHBS refer to 2005, we adjust household consumption values to 2007 with the consumer price index. Although this does not take into account economic growth, the estimates provide a more realistic and timely impression of the actual costs involved. Table 5.1: Simulated scenarios: average monthly benefits Eligibility Monthly benefit per Average monthly Total annual child (KGS) benefit per household benefits (million member (KGS) KGS) Scenario 1 All children below 16 years 48 21 1,008 Scenario 2 All household members living in 93 93 1,008 10% poorest households based on PMT Scenario 3 Children below 16 years living in 176 94 1,008 poorest 10% of households based on PMT Counterfactual Perfect UMB (GMCL: 705 KGS), 139 72 1,010 targeted at the poorest 15% of households Source: Staff calculations. For the assessment of the alternative targeting methodologies we use a number of performance indicators. For all scenarios we assume 100 percent take-up rate. This neglects any issues related to the actual implementation of such alternatives. Consequently, the presented results are not comparable with the actual performance of the current UMB. This is why, for comparative purposes (counterfactual) we include a simulated UMB under optimal circumstances: assuming perfect implementation and 100 40 These possibilities are discussed in greater details further in the chapter in the context of the argument for hybrid means testing. 41 Note that actual expenditures for UMB in 2007 amount only to 695 million KGS. 65 percent take-up rate. 42 The level of the underlying GMCL had to be adjusted to 705 KGS per month to approximate the same budget allocation (1,008 million KGS). As a result, it would cover the poorest 23 percent of the population with children. It serves as a benchmark for the alternative targeting methodologies. Figure 5.1 compares coverage and targeting efficiency of the different scenarios. Since we are interested in the poorest part of the population we compare the performance for the poorest 20 percent of the population (quintile I) measured before the allocation of the UMB and assuming a 25 percent substitution rate. Figure 5.1: Coverage and targeting efficiency of categorical and proxy-means targeting 100 90 80 70 60 Percent 50 40 30 20 10 0 Perfect UMB Scenario 1 (Cat 0-15) Scenario 2 (PMT all) Scenario 3 (PMT 0-15) Population of Q1 covered Beneficiaries belonging to Q1 Share of benefits going to Q1 Source: Staff calculations based on KIHBS 2005. Coverage of the bottom 20 percent of the population is especially high for the categorical child benefit. Assuming a 100 percent take-up rate, a categorical child benefit for all children up to the age of 16 would cover almost 100 percent of the population belonging to the poorest quintile. There are very few households without children that are extremely poor. As a result, coverage rates for the two alternatives based on proxy-means testing hardly differ. Up to half of the population of the poorest quintile would be covered. With respect to targeting efficiency, i.e. the share of benefits going to the poorest quintile, categorical child benefits perform less favorably compared to the current system. Due to the universal character of the benefit, leakage to non-poor is considerable under the assumption that the take-up rate is 100 percent. Considering the size of the monthly benefit, it can be expected that wealthier households will not bother applying for the child benefit. As a result, the effective targeting efficiency rates may be higher than presented here. On the other hand, problems with birth registration may exclude some of the poorest households from applying for, and receiving a universal child benefit. The results for the proxy-means test simulation indicate improved targeting efficiency compared to the universal child benefit but none of the alternative targeting methodologies can match the results of a “perfectly implemented UMB means test,� with 100 percent take-up and an actual budget of 1,010 million KGS spent, except for the categorical child benefit which matches the level of inclusion of the poorest households. 42 The assumption of a ‘perfectly implemented UMB means test’ as the counterfactual is necessary for comparing the simulations. However, we acknowledge that in reality the current means test has a more complex design and would be subject to greater errors as compared to the much simpler proxy means testing and categorical targeting systems. 66 Figure 5.2: Exclusion errors, inclusion errors and benefit leakage for categorical and proxy-means targeting 80 70 60 50 Percent 40 30 20 10 0 Perfect UMB Scenario 1 (Cat 0-15) Scenario 2 (PMT all) Scenario 3 (PMT 0-15) Inclusion error (individuals Q1) Exclusion error (individuals Q1) Leakage to Q2-Q5 Source: Staff calculations based on KIHBS 2005. Figure 5.2 presents the results from the perspective of inclusion errors (the percent of beneficiaries wrongly receiving a benefit), exclusion errors and leakage of benefits to the population not belonging to the bottom quintile. The exclusion error mirrors the coverage rates presented in the previous figure. In terms of leakage to the non-poor, i.e. the share of total benefits going to non-poor households, purely categorical benefits to children have the highest leakage rate, while with a proxy-means test, either combined with an age limit or not, leakage to less poor households would be around 40 percent of the total benefit sum. Inclusion error and leakage in all three simulated scenarios are much higher compared to the means-tested UMB if operating under optimal circumstances. The main objective of the UMB is to reduce poverty and enable poor households to cover a minimum level of consumption. Figure 5.3 compares the poverty reduction impact of the simulated budget- neutral policy options. We compare the relative reduction of the poverty incidence rates and the poverty gaps both for extreme poverty. 43 The simulated policy options can be effective in reducing extreme poverty. Targeting the 20 poorest households using a proxy-means test and assigning a budget neutral flat rate benefit either to all household members or children only would achieve a reduction of the extreme poverty gap of more than 70 percent. However, this is the optimal scenario under the assumption of full compliance and take-up. Figure 5.3: Relative reduction of extreme poverty 43 The relative reduction is the difference between the pre- and post transfer poverty rate (gap) as a percentage of the pre-transfer poverty rate (gap). 67 100 90 80 Relative reduction (%) 70 60 50 40 30 20 10 0 Perfect UMB Scenario 1 (Cat 0-15) Scenario 2 (PMT all) Scenario 3 (PMT 0-15) Reduction extreme poverty incidence Reduction extreme poverty gap Source: Staff calculations based on KIHBS 2005. Figure 5.4 puts the poverty reduction impact in relation to the total amount spent on the benefits (cost efficiency for poverty reduction). The total budget is the same for all tested options. All options assume perfect implementation with 100 percent take-up rate. No administration costs have been included.44 Under these assumptions, the differences between the targeting scenarios (categorical, proxy-means testing and the combination of them) are minimal in the case of the cost of reduction of the absolute poverty gap. There is virtually no difference in their effectiveness with respect to the reduction of the absolute poverty gap. Compared with them, the counterfactual ‘perfect’ means-tested UMB is more effective in reducing the extreme poverty gap, but considerably less effective in reducing the gap in absolute poverty. However, the cost efficiency of the proxy means targeting might be underestimated in the above simulation as it assumes a flat rate of the delivered PMT benefit. A variable PMT benefit (which is difficult to simulate while keeping the simulation budget neutral) would correct for this underestimation / increase the cost efficiency performance of the proxy means testing targeting system. Figure 5.4: Cost efficiency of simulated targeting scenarios 44 Assuming same administrative costs will not affect the comparative results. Differences will emerge in case of assuming different administrative costs – higher for the means-testing and lower for the proxy-means testing. However, we lack a good foundation for a valid assumption of what rates of administrative costs to choose. 68 Perfect UMB 2.09 Extreme CAT 0-15 (a) 2.46 COMB 1 2.50 PMT 1 2.50 Perfect UMB 3.18 Absolute CAT 0-15 (a) 0.68 COMB 1 0.51 PMT 1 0.47 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 Som spent per 1 Som reduction of poverty gap Source: Staff calculations Who would lose from the alternative targeting methodologies? Table 5.2 compares the population eligible under the simulated targeting methodologies with households currently receiving UMB. Especially relevant are the poorest 20 percent of households. Under a universal child benefit for all children up to the age of 16, no household would lose the rights for the UMB. With respect to the proxy- means test, 13 percent of households in the poorest quintile would no longer be entitled to a UMB. The change of the targeting methodology would hardly have any effect on the poorest households in the urban north. Slightly higher drop-out rates are found if the bottom 10 percent households are targeted with a proxy-means test in the south, both in urban and rural areas. Table 5.2: Individual and households losing UMB entitlements under new scenarios, percentage of current beneficiaries, 2005 Losing benefits – total Losers in quintile I Eligibility Individuals Households Individuals Households Children 0-15 0.1 0.2 0.0 0.0 Bottom 10% of households (PMT)* 8.2 6.7 11.0 13.2 Note: * the same rates apply to the combined scenarios where only children are eligible for benefits in households identified through a proxy- means test. Source: Staff calculations based on KIHBS 2005. COMBINING ELEMENTS OF A MEANS TEST AND A PROXY-MEANS TEST APPROACH: A HYBRID MEANS TEST The main conclusion from simulating and comparing alternative targeting methodologies is that they can hardly match the results of a “perfectly implemented UMB means test.� Inclusion error and leakage in all three simulated scenarios are much higher compared to the means-tested UMB if operating under optimal circumstances. In this context replacing the current UMB targeting methodology with a completely new approach is not advisable from the perspective of potential accuracy. The UMB does perform well in terms of targeting the poor, but lies behind in covering a large share of poor and vulnerable households and having a significant impact on poverty. This indicates the need for shifting the focus of any policy reform from the reduction of the inclusion error (which is very much the prevailing agenda reflected in policy documents and research papers), to reducing the exclusion error in order to achieve a broader coverage of the poorest with the UMB. Strengthening outreach activities with communication campaigns and pro-active social workers, a better use of the Social Passport, better management of information and increasing the allocated budget may solve some of the shortcomings of the current system. Moreover, in the onset of the economic crisis, enhanced outreach to the poor becomes even more important. 69 Defining hybrid means testing A possible way of further strengthening the targeting of the UMB could be the hybrid means test (HMT). The HMT formula is a combination of elements from an income test system with those of a proxy means tested system. HMT methods are those where multidimensional indicators are used in combination with a means test either to predict or to validate targeting outcomes. A comparison of data requirements and eligibility criteria for the MT, PMT and HMT, along with their main advantages and disadvantages, is presented in Table 5.3. Table 5.3: A Spectrum of Targeting Instruments Based on Individual Assessment Data Eligibility Criteria Advantages/Disadvantages Means Testing • Self-reported income & • Income < Threshold • ADV: Can be very accurate (MT) assets collected through Income Cut-off Level (especially with verification); interviews • Sometimes establish a also: more responsive to • Verified with higher cut-off level for transient changes (e.g., in certification, public program “exit� crisis) information, cross-checks • DISADV: Administratively demanding; challenging with informality; potential for work disincentives Proxy Means • Alternative indicators of • Score = ά + ßX + ßX + • ADV: Useful in situations Testing (PMT) living standards ßX with high degrees of • Develop models usually • Predicted values can informality; less potential for with Household Surveys establish weights and work disincentives; allows to to identify indicators that eligibility cut-offs capture multi-dimensional are correlated with (thresholds) aspects of poverty (not just poverty + scoring formula income poverty) • Collect data on indicators • DISADV: Administratively through interviews and demanding; eligibility criteria (usually) home visits may need to change regularly as people learn to “game� the system; doesn’t capture changes quickly (less responsive in crisis) Hybrid Means • Combination of the • Predict incomes using: • ADV: Can be very accurate; Testing methods above o Easily measured optimizes use of information; (HMT) income possible with informality; less o Imputed incomes work disincentives (?); (using proxies or objective/verifiable; other imputation responsive to changes (e.g., in methods) times of crisis) o And/or use proxies • DISADV: Administratively to validate or demanding cross-check data on reported incomes • Estimated/predicted income < Threshold Cutoff Level Source: Lindert (2008) A HMT formula is computed for each applicant household based on two factors: (i) declared income that can be verified by cross-checking the information reported by the UMB claimants with other public databases, and (ii) presumptive income, or welfare predictors derived from a PMT formula. The proxy indicators / predictors can be selected taking into account two criteria: (i) positive correlation between the per capita consumption and the predictor, which will define the accuracy of the prediction and/or verification, and (ii) verifiability of the predictor which will determine the accuracy of the information used to estimate the per capita consumption. The types of predictors and verifiers of the level of welfare of the household can be such as location variables, housing quality variables, household demographic, 70 educational and other characteristics, ownership of durable goods and productive assets, capacity of the household to sustain big expenditures like utility bills, phone and cell phone bills, electricity bills, students’ tuition fees. The HMT is being developed in some middle income countries in ECA and elsewhere. It is particularly relevant for situations characterized with high informality of the economy and inability of the formal income test to capture large proportion of cash and in-kind incomes outside the formal sector. Its main advantage is the ability to combine the clarity and exactness of the verifiable income with indicators (household characteristics) which are good predictors and /or verifiers when informal income is estimated. A HMT approach for the UMB The HMT approach could be used as a targeting instrument for the UMB in Kyrgyzstan in several ways. First, it can be an integral part of the UMB program eligibility criteria to determine which families will be accepted in or reject from the program. The eligibility could be assessed with a scoring formula combining income and welfare predictors, or – alternatively and more easy to apply, the eligibility will be assessed with the means test and certain number of indicators which are welfare predictors can be used as additional filters to eligibility. Second, it can be a verifier of the eligibility of UMB program beneficiaries (already receiving the UMB) for inspection and audit purposes and eventual detector of error and fraud in the system. Beneficiary households that qualify for UMB based on income but at the same time avail of characteristics that do not predict poverty, would be considered as ‘risky’ recipients where more frequent and thorough inspections would be appropriate. Third, it can be an identifier of poor households which are subject to exclusion based on formal income. Given that the GMCL is much lower than the extreme poverty line, it is possible for some household with per capita income above the GMCL to be very poor and in need of social assistance, especially if all their income is the reported income. It is also possible that poor households do not qualify for the UMB because they cannot provide certain documents due to administrative, legal and bureaucratic reasons. In such cases, the hybrid means test can be applied to verify eligibility and allow perhaps temporarily receiving the UMB without submission of full set of required documents. For Kyrgyzstan, we have simulated a proxy means test model which predicts the per capita consumption of each household from the KIHBS 2005 (Annex 4). For a HMT, the following variables could serve as predictors: (i) for urban areas, whether the household has hot water, durable goods (telephone, color TV, refrigerator, car or motorbike); (ii) for rural areas, whether the household owns agricultural equipment. Household size and number of children are also very strong predictors of actual household welfare, but they would not be useful as filters as they are taken into account when calculating income. Relation between livestock possession and poverty Kyrgyzstan already implements some form of HMT when identifying UMB beneficiaries. Eligibility for UMB is determined based on verifiable (and verified) income, complemented with presumptive income from land. The intended inclusion of presumptive income from owned livestock, the way top estimate it and its effects on the targeting accuracy of the UMB, is also a part of the HMT approach. The issue is whether the possession of livestock is indeed a relevant variable for the assessment of a household’s living standard. Do families owning livestock have a lower poverty risk? What is the relationship between livestock ownership and poverty. Livestock is a generic term for very different animals that are kept for production and consumption purposes. The possession of a cow cannot straightforwardly be compared with the ownership of poultry, for example. The KIHBS provides data on livestock ownership per type of animal and per quarter. In principal, one can calculate the animal stocks and flows. The quality of the data is however not consistent over the four quarters. Especially, data for the fourth quarter are seriously flawed. We therefore use only the information on livestock for the first quarter and develop an index that accounts for the differences between different types of animals. 71 The methodology to develop an ‘animal index’ is based on the concept of Tropical Livestock Units developed by the FAO. 45 It provides a convenient method for quantifying a wide range of different livestock types and sizes in a standardized manner. It provides a single figure expressing the amount of livestock owned by a household irrespective of the composition of the livestock. 46 Different species are compared in relation to a common unit, Livestock Units (LU). There exist different formulae for estimating LUs depending on the region. Since no comparative material exists for Central Asia, we base the conversion units on Maltsoglou and Taniguchi (2004) that adapted the methodology to Nepal. We make no difference between animals held for consumption or for production.47 Table 5.4: Livestock conversion units Species Survey categories KIHBS LU per head Cattle Cows, heifers, calves, bulls, oxen, yaks 0.7* Sheep and goats Sheep, lambs, goats, goat kids 0.1 Pigs Pigs, swine, piglets 0.2 Horses Horses, young horses 0.7** Donkeys (young) donkeys, (young) mules, (young) hinnies 0.5 Poultry Old and young poultry 0.01 Source: Based on Maltsoglou & Taniguchi (2004), except: *yaks are considered equivalent to cattle;** own judgment. Overall, 41 percent of Kyrgyz households own livestock, and almost half of the poor households own livestock. The average herd is equivalent to 2.4 LUs. Among livestock owners, cattle and poultry are most prevalent. There are considerable differences in the proportion of households owning livestock across oblasts and type of livestock they own. The highest livestock incidence is observed in Naryn and Batken oblast, where three out of four households own livestock. On average households have larger herds in Naryn. Cattle is widespread in both oblasts. Four out of five livestock owners have cattle. Naryn oblast has the highest incidence of horse ownership in the whole country. 40 percent of livestock owners have horses, compared to only 2 percent in Batken. Livestock ownership is lowest in Talas, except for Bishkek that due to its urban environment is less adaptive for animal husbandry. Interesting distinctions can be found comparing the rural areas in the north and south of the country. Although livestock incidence and heard size are similar, cattle is more widespread in the rural south, while poultry, and to a lesser extent, horses are more prevalent in the rural north. Livestock ownership is especially prevalent in highly mountainous areas. Herd sizes are relatively large on average. Livestock owners keep mainly cattle, sheep and goats. About a third also owns horses. The incidence of livestock ownership is higher among poor households. This transpires from the analysis of livestock ownership by poverty status (Tables 5.5 and 5.6). However, non-poor households own slightly larger herds. The smallest herds are owned by the extremely poor. The richer the households are, the lower the share owning livestock, but the larger the herd size. Table 5.5: Livestock ownership, herd size and type of cattle owned, 2005* Livestock Herd size Ownership of different types of livestock ownership (LU)** (only livestock owners) % of Cattle Sheep Horses Poultry households Mean (st.dev.) (%) (%) (%) (%) Total 41.1 2.38 (0.14) 70.2 40.2 14.7 75.1 by oblast:: Issyk-kul 58.7 2.28 (0.47) 69.9 37.9 27.2 75.1 Jalalabat 40.8 2.18 (0.18) 85.8 43.3 9.0 71.7 Naryn 77.0 4.95 (0.67) 86.1 73.5 39.1 83.8 Batken 74.6 2.41 (0.18) 80.3 50.9 1.7 56.4 Osh 46.0 2.90 (0.43) 75.2 43.7 19.8 67.8 Talas 32.5 1.89 (0.22) 67.5 30.6 18.7 68.9 45 See, for example, FAO (2002), Maltsoglou & Taniguchi (2004), or http://virtualcentre.org/en/dec/toolbox/mixed1/tlu.htm. 46 Using a composite indicator has drawbacks as detailed information is lost. 47 For example, horses, donkeys, mules etc. might be used to support agricultural work. 72 Chui 53.0 1.42 (0.13) 47.8 21.8 5.6 91.9 Bishkek 4.1 0.69 (0.35) 18.9 9.4 0.0 90.6 by region: North-urban 11.2 0.84 (0.12) 22.6 20.5 2.9 88.9 North-rural 63.1 2.53 (0.22) 67.7 37.7 20.4 83.8 South-urban 16.3 0.99 (0.11) 48.9 38.9 0.3 48.1 South-rural 63.2 2.75 (0.24) 83.3 46.0 13.7 68.6 by topography: Highly mountainous 62.5 5.06 (0.63) 84.5 75.8 32.5 58.6 Mod. Mountainous 49.0 2.30 (0.16) 74.6 47.5 21.8 71.9 Plain 37.7 1.97 (0.15) 66.9 32.7 10.3 78.6 by poverty status: Poor 48.8 2.27 (0.19) 77.2 44.6 16.2 62.6 Not poor 37.4 2.50 (0.21) 66.0 37.6 13.9 82.8 Extremely poor 40.5 1.95 (0.27) 72.4 52.0 19.3 67.1 Not extremely poor 47.9 2.46 (0.16) 70.0 39.1 14.3 75.9 *Household level. ** Only for livestock owners. Source: Staff calculations based on KIHBS 2005 Table 5.6: Livestock ownership and herd size by quintile, 2005 Livestock ownership Herd size (LU)* % of households Mean (st.dev.) Quintile I 50.8 2.00 (0.17) Quintile II 51.5 2.28 (0.19) Quintile III 46.7 2.37 (0.31) Quintile IV 37.7 2.51 (0.36) Quintile V 30.7 2.78 (0.42) Note: quintiles based on per capita consumption. * of livestock owners only. Source: Staff calculations based on KIHBS 2005 Including livestock in income assessment 48 Would the inclusion of income from livestock further improve UMB program outcomes? Consultants of the EC Technical Assistance Unit in the Ministry of Finance drafted a proposal for the calculation of potential income from livestock in 2006. Net potential revenue is defined as the number of livestock slaughtered times the price of the meat, plus the produce from the animals time the price, minus the expenditures. Average potential revenues are calculated per type of animal and oblast. This section estimates the precision lost by calculating the administrative definition of income used in determining eligibility for the UMB instead of the full income (the inclusion error associated with the use of a second- best income indicator). Similarly, it is important to estimate by how much this imprecision will fall if the presumptive income from livestock will be taken into account. To estimate the presumptive livestock income we develop a simple regression model relating the 2005 livestock income (as reported in KIHBS) to the increase in livestock herd over that period and a set of regional dummies to account for regional differences in the prices of output and inputs.49 The coefficients estimated by the model could be used in a similar fashion as the land coefficients, to determine the livestock income using a presumptive method. When added to the UMB income, it will improve the administrative definition of income. The Pearson correlation coefficient of actual annual livestock income and predicted livestock income is 0.799. 48 This section draws on Tesliuc & Leite (Background note, July 2008). 49 See Annex 3 for the model. 73 Table 5.7 compares the household rankings based on full income versus administrative income (upper panel) and administrative income including presumptive livestock income (lower panel) to assess the losses or gains in precision. Table 5.7: Precision gains and losses in ranking households: administrative income with and without predicted income from livestock Household per capita income Correlation: 0.9457* 1 2 3 4 5 6 7 8 9 10 Total 1 5.6 0.9 1.0 0.7 0.7 0.5 0.6 0.0 0.0 0.0 10 2 3.5 3.2 1.6 0.2 0.6 0.3 0.1 0.6 0.0 0.0 10 3 0.7 4.5 1.6 1.1 0.3 0.4 0.7 0.4 0.0 0.1 10 4 0.1 1.2 4.2 2.1 1.3 0.7 0.3 0.1 0.1 0.0 10 Per capita household 5 0.2 0.2 1.3 4.0 2.1 0.5 1.1 0.5 0.2 0.0 10 UMB/adm income 6 0.0 0.1 0.3 1.5 4.4 1.5 0.5 0.8 0.3 0.6 10 7 0.0 0.0 0.0 0.3 0.7 5.4 1.8 0.7 0.9 0.2 10 8 0.0 0.0 0.0 0.1 0.1 0.2 4.6 3.7 1.1 0.3 10 9 0.0 0.0 0.0 0.0 0.0 0.0 0.2 3.4 5.8 0.7 10 10 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 1.6 8.1 10 Total 10 10 10 10 10 10 10 10 10 10 100 Correlation: 0.9443* 1 6.0 0.7 0.8 0.7 0.6 0.5 0.6 0.0 0.0 0.0 10 2 3.2 3.4 1.6 0.3 0.5 0.4 0.0 0.6 0.0 0.0 10 3 0.3 4.9 2.3 0.9 0.7 0.4 0.3 0.1 0.1 0.0 10 4 0.2 0.4 4.2 2.9 0.9 0.5 0.5 0.4 0.1 0.0 10 Per capita household 5 0.2 0.5 0.8 3.6 2.1 0.6 1.2 0.8 0.1 0.0 10 UMB/adm income + estimated livestock 6 0.0 0.1 0.1 1.2 5.0 2.0 0.7 0.4 0.1 0.6 10 7 0.1 0.1 0.1 0.2 0.3 4.8 2.6 0.6 1.0 0.1 10 8 0.0 0.0 0.1 0.1 0.1 0.3 3.9 4.5 0.7 0.4 10 9 0.0 0.0 0.0 0.0 0.0 0.1 0.1 2.7 6.3 0.7 10 10 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.1 1.5 8.2 10 Total 10 10 10 10 10 10 10 10 10 10 100 Source: Tesliuc & Leite (2008) based on KIHBS 2005. Overall, the addition of livestock on household UMB income may slightly improve the targeting accuracy of the program given that the new UMB specification is a better proxy for actual household income. Table 5.4 shows that adding predicted livestock income to actual UMB income marginally affects the correlation of incomes. Under the actual UMB income definition, we would correctly identify 5.6 percent of the poorest 10 percent households (in other words, 56 percent out of the poorest 10 households, bottom decile equals 100 percent) according to full household income compared to 6 percent out of 10 percent (meaning 60 percent of all) under the new UMB income specification. We also compare three different HMT approaches to defining household income (Annex 7). HMT Option 1 (HMT 1) takes into account the current UMB administrative income and presumptive income from land ownership. HMT Option 2 (HMT 2) includes the current UMB administrative income and presumptive income from livestock. HMT Option 3 (HMT 3) includes estimated income – verifiable income and predicted non-verifiable income, which includes income from land ownership and livestock. The comparison of targeting performance across the three options is slightly in favor of HMT 2. • In terms of benefit incidence / targeting accuracy, the HMT 2 (administrative income and presumptive livestock income) delivers the highest share (45.2 percent) of all UMB to the poorest quintile. The benefit incidence of HMT 1 (administrative income and presumptive land income) is quite close to the scenario with highest targeting performance HMT 2. Targeting is the poorest in the case of the case of HMT 3 (verifiable income plus presumptive income from land and livestock) delivers the lowest - 41.5 percent, and the leakage is respectively the highest in this case. 74 • In terms of beneficiary incidence, among the three HMT scenarios, HMT 2 (administrative income and presumptive livestock income) displays the highest beneficiary incidence in the poorest quintile (41.8 percent) while HMT 3 (verifiable income plus presumptive income from land and livestock) has the lowest (39.6 percent). In terms of program coverage, a UMB targeted with the HMT 2 approach would be received by the highest share of the population of the poorest quintile (41.8 percent) followed by HNT 1 scenario (41 percent)> Coverage would be the lowest with HMT 3 targeting option (37 percent). In terms of generosity, the benefit amount delivered is almost the same in the case of HMT 1 and HMT 2 and slightly higher for HMT 3. The ultimate effect of inclusion of the predicted income from livestock in the overall household income would depend on how the imputation is implemented. The Government of the Kyrgyz Republic has proposed to use a coefficient at oblast level similar to the coefficients used for imputation of the income from land ownership. As with the land coefficient, potential revenue will be measured without taking into account the actual household situation and may lead to targeting errors.50 The costs of establishing, implementing and regularly updating a methodology to account for income from livestock have to be considered against the potential benefits. The data needs are extensive as they require detailed information on the number of livestock per household, prices for products derived from the animals, estimates for expenditures for livestock keeping per type of animal, the quantity of output produced by animal, and many more variables. At this point the social workers have only the information of livestock ownership because it is recorded in the Social Passport. IMPROVING THE ADMINISTRATION Improving eligibility criteria is not the only way to strengthen program performance. Indeed, international evidence (see Grosh, del Ninno, Tesliuc & Ouerghi (2008); Grosh &Tesliuc (2007); Castaneda & Lindert), and others) suggests that implementation also matters for good program performance. In the case of the UMB, its success is the result of both good design and good implementation. This section reviews both administrative costs and processes and concludes that the cost of administration is comparable with the cost of administration of programs with similar objectives and design in other parts of the world. The section also reviews the different stages of UMB implementation and concludes that there is room to improve the implementation of the UMB, especially at the stages of determining and verification of eligibility, also with respect to the payment process, the monitoring and tracking of beneficiaries with adequate registries and management information systems, as well as with respect to reducing the susceptibility of the system to error and fraud in the social assistance system. Administrative costs of the UMB So far, this study has not yet considered administrative costs. In general it is very difficult to assign administrative costs to separate programs as few public financial management systems allow the breakdown per program and type of cost. Administrative costs for means-tested benefits can be substantial, but empirical evidence is scarce. Grosh & Tesliuc (2007) present estimates for administrative costs for a small number of countries. The share of administrative costs in total program costs range from 6.5 percent in Lithuania to 11 percent in the United States. According to Tabor (2002), administrative costs of means-tested social assistance programs account for 7 to 11 percent of total program costs in OECD countries. According to the Social Protection Strategy 2006-2008 of the GoKG, total administrative costs of the MLSD account for 3.2 percent of the allocated budget. The MOF (MTBF 2007-2010) estimates administrative costs incurred by the MLSD at 2.2 percent of the respective budget. For the Pension Fund, 50 Tesliuc & Leite (2008) estimate that the land coefficients considerably underestimate actual income from land ownership. Actual estimated crop income accounts for 10 percent of total income compared to 6 percent based on predicted income. 75 administrative expenses account for 2.6 percent. These are the budgeted resources for administrative outlays for the program. However, actual costs of administering the program are estimated to be significantly higher than this. The Kyrgyzstan Center for Social and Economic Research (CASE) estimated the administrative costs for the UMB at 11.5 percent of total program costs in 2005 (CASE, 2006), as discussed in more detail in Box 5.1 below. This means that insufficient resources are being provided for quality implementation. As discussed below, implementation quality matters, and there are many ways that it can be improved in the case of the UMB. Box 5.1 Qualitative assessment of the trends in UMB administrative costs 2005 - 2008 The first and only quantitative assessment of the UMB administration costs was conducted in 2005-2006 according to methodology proposed by the World Bank and adjusted to the Kyrgyz realities by experts of the CASE-Kyrgyzstan Consulting Agency (Grosh & Tesliuc 2007). The applied methodology envisaged analysis of: (i) the number of UMB beneficiaries by region, and by rural and urban areas; (ii) the budgets of the MLSD’s central office, regional, district and municipal social protection departments, payment centers and postage; (iii) the time allocated by social protection staff at different levels for work with the UMB. According to this survey, the cost of UMB administration equaled 11.5 percent of the total UMB program cost in 2005. In 2008, CASE made an effort through qualitative survey work to identify the factors which could have contributed to possible changes in the cost of administration of the UMB program. The following county-wide and regional factors were identified. Nation-wide, the increase of the GMCL from 175 to 200 KGS resulting in reducing the number of eligible beneficiaries and the abolishment of the regional (oblast) level social protection departments, work towards reduction of the administrative costs. At the same time, the increase of the salaries of social workers as of January 1, 2008 has the opposite effect. Regionally, in the rural North, the cost of administration of the UMB should be further decreasing following the decrease of the number of new applicants for UMB after the introduction of new land income quotas; and also following the attrition of part of the former recipients of UMB for the same reason - mainly in Naryn, Chui and parts of Issyk-kul regions where the rural communities have large land allotments. In the North urban areas, despite the considerable flow of migrants to Bishkek and Tokmok from Naryn and the Southern parts of the country, the cost of administration of the UMB should not be increasing because most of these migrants have land allotments which prevent them from being eligible for UMB in two ways: (i) because of the land ownership itself which limits the access to UMB; (ii) because of the administrative requirement to report on the owned land when applying for UMB in the new place of residence. In this case the applicants are required to submit certificates of land ownership from the aiyl okmotu and the social protection authorities which is often a requirement they fail to comply with. In the South, the situation is unclear since the households there a larger, and if the land allotments are not big, they will not necessarily affect the overall number of beneficiaries served by the social assistance system. A new quantitative assessment of the administrative costs of the UMB will shed light on the actual changes. The appropriate timing for it would be the end of 2008, when the aggregate impact of the GMCL increase, the introduction of new land quotas (which impute return on land to the aggregate household income), and the increase in social workers’ salaries and budgets of social protection departments and payment centers. Source: CASE (2008) Figure 5.5: Share of administrative costs in total program costs 76 16 13.4 14 12 11.0 10.5 10.0 9.9 Percent 10 7.8 8 6.5 7.1 6.0 6 4 2.2 2 0 Albania Bulgaria Mexico Lithuania United Kingdom United States Colombia Romania Armenia Brazil Source: Grosh & Tesliuc (2007) Improving implementation Implementation matters for good targeting performance for many reasons: (a) to improve program outcomes, such as coverage (reduced errors of exclusion) and targeting accuracy (reduced errors of inclusion of the non-poor); (b) to improve administrative and program cost efficiency; and (c) to improve transparency and reduce fraud and errors. 51 With respect to the latter, it is important to note that all transfer programs invoke some degree of fraud and errors, the goal is to improve processes to minimize, monitor, detect and remedy these errors. It is also important to note that not all errors are fraud; indeed, some errors are unintentional (arising out of ignorance, lack of clarity on processes, or measurement error by claimants or officials). The UMB program has been implemented for a decade already in a transparent way and similarly throughout the country. The main administrative processes are defined and articulated in MLSD regulative documents. The UMB is administered by the main department of social protection of MLSD with devolved oblast and rayon departments. The Kyrgyz Post Office executes the benefit payment. The process of applying and receiving the UMB broadly consists of six stages / processes: (i) application; (ii) verification and determining of the UMB benefit level; (iii) verification and payment; (iv) processing the payment; (v) recertification; (vi) registration of poor, applicants and recipients and data management. The processes take place at several levels of government – local government / aiyl okmoty, rayon and central / national levels. The main unit of support is the family defined as people who live together and share incomes (common budget) and household assets. Application is on an on-demand basis - any member of a low-income family can initiate it and at any point of time. The initial point of contact is the social worker. The social worker is tasked with monitoring poverty and vulnerability in his/her constituency by filling Social Passports for all local residents living below different administrative poverty lines: on the first place for those living below the GMCL / extreme poverty line of KGS 200 in 2008, then for those living below the low poverty line of KGS 558, and finally also for those living below the high poverty line of KGS 860. The Social Passport is a planning and monitoring instrument and not an obligatory precondition for applying for UMB. It can be filled at the start of the UMB application procedure by the social worker in the presence of the applicant. If the potential beneficiary lives in a village, she/he applies in the ayil okmotu, if living in town she/he turns to the Department of Social Protection (DSP) of the rayon. A permanent residence permit is 51 This section draws from a recently completed by Christian Van Stolk study51 for the World Bank where the UMB administration in Kyrgyzstan is analyzed from the perspective of the program administration experience of the Fraud Investigation Service of the UK Department of Work and Pensions and the comparative OECD-wide experience of RAND Europe. The findings this study have been confirmed by the qualitative research (CASE, 2008a), and in discussions with social workers in the field. 77 required to apply. If the citizen is registered in one rayon (city), but actually lives in another place of this city or rayon, UMB is applied for at the place of registration, and a certificate from rayon DSP at the residence is required to prevent duplicative applications. For citizens who live without registration, UMB could be granted at the place of actual residence on the basis of community / quarter committee certificate confirmed by neighbors. The application process requires three visits to the rayon or aiyl okmotu. The first visit is for information and initial screening – the prospective applicant checks on potential eligibility in a conversation / interview with the social worker, on the procedures and required documents in support of his/her case. The social worker conducts initial screening, and if the applicant appears eligible, advises on the list of supporting documents to be collected. The applicant is responsible for collecting the documents which include copies of personal identification cards, residence permit / registration, certificate of family composition and number of children, birth, marriage and / or divorce certificates, work record book issued by the Employment Office, certificate of school enrollment issued by the respective school, document of land ownership issued by the local government based on data from the land registry, employment verification with salary statement issued by the employer, documents on taxable incomes and taxes paid for the self-employed, etc. Document for livestock ownership issued by the local government based on data from the livestock registry is also collected but only for the purpose of filling and update of the Social Passport. During the second visit the applicant presents the collected documents to the social worker, and; for completion of the application form (opening of the applicants’ case / folder with documents proving eligibility). A third visit is usually required for provision of missing documents. The process can take place in one month conditional to how quickly the documents are collected and verified. This is fairly impressive by international standards. Verification of eligibility is a multi-stage process which involves robust data verification and community-level eligibility pre-screening. If the application is initiated in a village, the applicant’s documents are verified first by the social worker in the respective aiyl okmotu. This is done by scrutinizing signatures and stamps, phone calls and personal visits of the issuing institutions and registries (land and livestock registers, courts, employers / firms, employment offices, schools, etc.). The community pre-screening is done by a Poverty Inspection Committee consisting of permanent members - the head and the secretary of the aiyl okmoty, the head of the village social committee and the local social worker, and ad hoc members - usually community participants / neighbors of the applicant. The committee usually works pro bono and with a one-year mandate. The committee will pay a home visit to examine observable characteristics of the household (quality of the dwelling, number of rooms, size of the house, availability of furniture, number of children and adults, agricultural land, garden, livestock, well-to-do relatives, etc.) and recommend continuation or suspension of the application. When the applicants pass verification and community screening in aiyl okmotu, their documents will be sent to the rayon DSP for additional verification and subsequent sending to the payment center. If the application is initiated in a town or city, it is submitted directly to the rayon DSP. The responsible social worker there will verify the documents supplied and will cross-check against information gathered from other sources. In both cases after verifying eligibility, the rayon will determine the benefit level and transfer the beneficiary files to the one of the six oblast level payment centers. The payment center registers the protocols and creates an electronic database of UMB recipients which appear in the protocols. Meanwhile the applicants will be informed about the rayon decision and can appeal them in the rayon DSP. The access to UMB is usually granted for one year, but the period could be shortened if closer monitoring of eligibility is required If family circumstances change meanwhile, the recipient is obliged to report on them in the ayil okmotu or DSP. Not less than once a year, officials of the rayon DSP will check UMB recipients for changes in the family social and economic circumstances. The list of UMB recipients is public and announced in the ayil okmoty of DSP for community control purposes. Payment of the UMB is executed by the Kyrgyz Post Office upon the order of the oblast-level payment centers after they undertake a further number of checks on the materials provided by the rayon (whether the rayon has calculated the benefit correctly, or will assess whether the application is complete). Once the payment center officials approve the documents of the UMB recipient, they will open an account, issue a payment order to the recipient, and forward it to the Post Office of the Kyrgyz Republic. The latter will execute the payment using state budget funds transferred to its accounts. The benefit will be 78 collected by the UMB recipient in a local post office. Each month the Post Office will submit a report to the payment center on the amounts paid and amounts outstanding. Payment through the banking system (debit cards or bank accounts) is being explored as an alternative quicker and more transparent method, especially in urban areas where banking services are readily available. Recertification of beneficiaries is conducted after a period of 12 or 6 months (the standard period of UMB extension), and in exceptional cases for 3 months. Scheduled recertification will take place upon expiration of the benefit period. Ad hoc recertification will take place in case of change of address, personal identification documents, economic or demographic characteristics of the family. In rural areas re-certification is implemented by the social workers of aiyl okmotu and in urban area – by specialists of rayon DSP. The UMB recipients are expected to initiate their re-certification and re-submit updated documents. Registry of UMB claimants and databases. The files consisting of application forms and supporting documents are paper-based. They are kept in the rayon DSP. Electronic databases are created in the payment centers based on the paper files presented by the social workers from the rayons. Then the data is transmitted to the post offices to guide the payments. The rayon databases are fragmented and not aggregated into a nation-wide centralized database in the MLSD with capacity to control and standardize the procedures for the registration for social assistance, eligibility determination, assignment of benefits, reasons for rejection, payment of the benefits and the statistical and management reporting. The registration of claimants is not based on a unique personal number / code as in the case of the registry of the Social Fund. There is no unified document management system to manage the flow, processing and storing important documents including history of application, received social assistance, reasons for refusals (if any), etc. Finally, there is no inter-ministerial electronic interface for exchange of data with other relevant institutions, such as the MOH, the Social Fund, the Office on Employment and Migration, MOF, the Post Office, etc. Moreover, files are kept only for the actual UMB recipients. Those whose applications have been turned down before the actual opening of the case / file are not recorded. The GoKG is making efforts to collect information for those who live below the poverty line but with incomes that are higher than the eligibility thresholds for UMB. This is done with the Social Passport – a household level questionnaire which is used to identify those who are eligible for UMB but for different reasons have not applied and for a broader synopsis of poverty and vulnerability in the respective ayil okmoty or rayon. In summary, there are many strong features of the implementation arrangements for the UMB, including: on-demand applications, verification of eligibility, and regularly scheduled recertification procedures. Combining administrative targeting procedures with community targeting allows better accounting of local knowledge and assessment of specific household circumstances. The public announcement of beneficiaries is conducive for community control. The benefit level is determined in the rayon and verified in the payment center thus preventing collusion. Nonetheless, there are ways that the implementation of the UMB could be strengthened, for improved program outcomes, efficiency and reduced fraud and error. Improvements could address all or some of the procedures related to determining UMB eligibility and the capacity of the social workers in the field, and more specifically: • Reduce the number of required documentation and minimize the number of documents which have to be submitted for re-certification; • Ensure that all the documents attesting eligibility are provided free-of-charge, and that as much as possible the flow of documents is through administrative channels and do not require beneficiaries to travel extensively to collect and submit them; • Increase the capacity and the incentives for social workers to reach and identify the poor and vulnerable, including through expanding the use of the Social Passport for initial identification and self-targeting of the poor; • Strengthen the control over the decision making on eligibility for UMB benefit by enforcing a requirement for the social workers and committees to write down the reasons for rejection; 79 • Disseminate more information about the access to UMB program; • Improve the screening of welfare characteristics during the home visits by more structured observations; • Expand the community involvement in identifying poor. With respect to fraud and error, there is some sense that the UMB is more susceptible to fraud and errors than other safety net programs in the Kyrgyz Republic because of its more complex eligibility criteria (categorical and means-tested criteria) and decentralized implementation. Error and fraud 52 are not uncommon in social programs with similar means-tested design throughout the OECD countries 53. Policy makers tend to find a correlation / causal relationship between eligibility complexity and the occurrence of fraud and error 54. Table 5.8 gives an overview of the number and geographical distribution of home visits and inspections of applications according to data for the fourth quarter of 2007. It indicates that the major part of the errors happens at the initial stage of application, and that the rate of rejection differs substantial across oblasts. Nationwide, 70 percent of UMB claimants receive the UMB, but this ratio is much lower at 40 percent in Talas, and extremely low at 7.5 percent in Chui. 55 The data from Table 5.8 suggests on one hand that there is complexity of eligibility requirements that makes them difficult to understand for staff and claimants, and allow discretion in judgment on eligibility and, on the other hand, that the processes put in place to verify eligibility allow staff and claimants to make mistakes or commit fraud. If the eligibility criteria are difficult to understand, this can have two consequences: increased error within the social protection system as claimants and social protection staff inadvertently introduce error into benefit claims; and increased transaction costs as social workers have to deal with a large volume of ineligible claims, which are ultimately rejected. If there is discretion of judgment on eligibility, and potential for arbitrary decisions, this can cause exclusion of potentially eligible UMB recipients, or will burden the system with processing of non-eligible applications. Table 5.8: Quarterly information on home visits and inspections of employers, fourth quarter 2007 Issyk- Djalal- Region Bishkek Chui Talas Naryn Osh Batken Total Kul Abad Number of applicants 1,322 1,818 15,654 1,560 7,125 3,276 28,772 6,185 65,712 Total received benefit applications 1 159 287 8,793 1,476 5,703 2,692 24,137 5,742 49,989 Total entitlements 1 129 138 6,268 1,467 5,619 2,401 23,756 5,708 46,486 Home visits 963 261 1 358 836 5,528 1 512 4,068 4,809 19,335 in% to received 83.1% 90.9% 15.4% 56.6% 96.9% 56.2% 16.9% 83.8% 38.7% Total number applications of visits Inspections of 708 1 675 240 763 1,579 5,118 2,986 12,070 employers in% to received 61.1% 0.3% 7.7% 16.3% 13.4% 58.7% 21.2% 52.0% 24.1% applications 52 In the safety net systems of OECD countries, the unintentional mistakes which UMB claimants or staff might make at different stage of the application for and delivery of the benefit due to limited knowledge or incorrect understanding of certain complex program requirements or processing difficulties are considered as errors. Fraud is defined as intentional behavior on behalf of the applicants for UMB aimed at securing access to the benefit by deliberately providing inaccurate, incomplete or falsified information in the application form or supporting documents, or by hiding information (for example, hiding information on income sources and income levels, in-kind transfers, remittances, availability of assets, making payments for study, etc.). Corruption relates to the staff behavior which attempts to intentionally exploit the benefit system. 53 An international benchmarking study puts the range of fraud and error in the social security systems in OECD at 2-5 percent of overall state social security expenditures. Quoted from Christian Van Stolk. Tackling Benefit Fraud and Error in the Kyrgyz Republic. Prepared for the World Bank by RAND Europe, June 2008 (mimeo). 54 Ibid. 55 It is important to note that the subsequent number of applications tends to be lower after the initial enquiry – 76 percent of initial enquiries with the social worker turn into applications of which 92 percent turn into benefits to recipients. 80 Issyk- Djalal- Region Bishkek Chui Talas Naryn Osh Batken Total Kul Abad Number of rejections in the course 163 1,531 6,861 84 1,422 584 4,635 443 15,723 of interview Rejections after home visits 8 35 411 3 28 0 5 17 507 Rejections after inspections of 0 0 11 3 0 0 0 9 23 employers Rejection for exceeding allowed 14 57 929 3 34 93 76 6 1212 income (GMCL) Total rejected applications 193 1,680 9,386 93 1,506 875 5,016 477 19,226 Total amount of savings after on- 9,698 39,200 595,289 7,200 32,845 0 19,825 31,200 735,257 site visits, KGS Source: MLSD data derived from CASE-Kyrgyzstan report, quoted from: Christian Van Stolk. Tackling Benefit Fraud and Error in the Kyrgyz Republic. Prepared for the World Bank by RAND Europe, June 2008 (mimeo). Errors and fraud are also possible at the stage of document collection and verification. Some of the family characteristics are difficult to verify with documents, e.g. parents without or with invalid IDs cannot register their children, single parents cannot verify separation, employment and earning verification could take long, be costly and cumbersome in case of labor mobility. As a result the application form cannot be completed. Verification of documents can induce error because some information sources could contain outdated or fraudulent information. There is anecdotal evidence that in certain places the land and agricultural registers are not regularly updated. There is no legal liability for the claimants in case of submission of documents with erroneous of falsified data. The only sanction would be return of the received cash if the UMB is awarded and fraud and error are revealed ex post (CASE, 2006). The institutions and individuals issuing falsified documents do not bear legal responsibility either. In addition, much of the verification work is done by hand and repeated at various levels. Given the large volume of applications and the large number of staff involved from ayil okmotu level to the payments centers and post office, errors might be introduced along the way. In many cases it is objectively difficult to assess true family income in the Kyrgyz Republic. The large agricultural sector, informal income, remittances and subsistence agriculture put high costs on the verification of the information provided by the client. Applicants could underreport income, and social workers do not have the possibility to verify the provided information. Therefore, the effectiveness of the social safety net in general, and the UMB in particular may benefit from reforms outside the sector, such as the computerization of land, agricultural and real estate registries, and the use of this information to cross-check the information provided by the applicants. At the stage of payment, the handling of the payment of the UMB by the post office is seen as an important cause of fraud, error, and even corruption. The MLSD statistics on financial infringements in 2006 bears this out. At central post office level, the main problem is that the date on which the post office reports to the MLSD on payments is the same as the payment date (when payments are made to recipients by the post office). This leads to a large ‘carry forward’ of funds, which may encourage these funds to be used inappropriately by the post office. At the interface with beneficiaries, the post offices are reported to delay cash payments, underpay, link the payments to imposed purchase (e.g. of old newspapers, CASE, 2008), appropriate non-claimed benefits. The system is partially equipped to detect, deter and prevent fraud and error. Internal controls exist, exercised by the Internal Control and Audit department of MLSD. It operates as a division within the central office of MLSD since 2006 with functions with respect to monitoring and control of the financial and economic activities of subordinated to MLSD units to ensure effective use of public resources. With respect to UMB, the internal control and audit is tasked with detection of fraud, conducting inspections and documentary audits of how the benefit is awarded and paid. It also does joint inspections with the Chamber of Accounts, controls over the recovery of overpayments owed to the state, monitors performance indicators and develops methodology for detection, deterrence and prevention of error and fraud in the social benefit system. However, the staff capacity and the budget for home visits, random spot checks and inspections are limited. Also, the audit methodology has not been able up to now to effectively detect errors and fraud. Despite that 20-40 percent of benefit files are inspected by the Internal 81 Control and Audit Department, and that in addition, about 40 percent of the households are subject to a home visit to verify material circumstances, the infringements detected in the UMB are only a small fraction of a percent of total payments in 2007 and a third of a percent in 2006. These numbers are very small compared to most OECD countries, with ranges between 2 to 5 percent of overall expenditure, suggesting underestimation of errors (Van Stolk, 2008). Implementation and administration could help improve outcomes and reduce the likelihood of fraud and errors through a number of corrective measures, including: • Formally and legally defining errors (and types of errors), fraud and corruption in social assistance,; • Formally empowering the Control and Audit Department to sanction infringements with administrative measures and even to recover unlawful benefit payments; • Developing a methodology to profile (or “target�) high risk beneficiary groups for inspection in addition to conducting random inspections (which are useful for monitoring overall fraud and error rates and overall program performance). Box 5.2 Targeting Inspections to High-Risk Profile Groups There are three ways in which inspections can be strengthened to “target� high risk profile groups. Firstly, they can be done on the basis of a risk assessment, which is based on building up profiles of clients who are at risk of committing benefit fraud. Secondly, they can be done on the basis of hybrid means-test which can help with cross-checking and imputing incomes to identify those who do not meet eligibility requirements of means-testing or whose personal characteristics do not match up with the means-test. Finally, inspections could occur on the basis of systematic intelligence gathered through data-matching (intelligence-led approaches). Risk-based inspections. The MLSD could start building up capacity within its central structures to analyze and review existing files and on this basis arrive at client and staff profiles of those most at risk of committing benefit fraud. These could be claimant and staff groups on the basis of age, profession, and family circumstances among others. This analysis would have to be specific to benefit types and can also be made region-specific. Risks then have to be prioritized and can be fed to the inspectors. The inspectors would match these risk criteria against the characteristics of claimants and inspect on that basis. HMT. The MLSD could strengthen the HMT of the UMB to improve the targeting of benefit claimants. The system is in use in a number of other middle-income countries and has improved the targeting of benefits to the needy. The MLSD could instruct the social workers and inspectors to use the HMT or the results of the tests to respectively assess eligibility and target inspections. In other words, the HMT could inform which households fall outside of the eligibility requirements and where an inspection might be appropriate. Data-matching and cross checks. The MLSD could expand the number of data sources that it matches against and start putting in place infrastructure to exchange data electronically with other services. This would entail the difficult task of building up a greater understanding of the issue across Government and promoting joint working with other agencies and ministries. Other sources of information could include utility bills, the passport agency, savings information held by the tax administration, mobile phone providers, bureaux de change, land and agricultural registries, credit agencies and others. Studies have shown that expanding data-matching to include additional sources of data and also private sector data increases the effectiveness of data-matching substantially. Source: Christian Van Stolk (2008). A number of improvements of UMB administration can help prevent error and fraud, and at the same time improve the access to UMB and efficiency of targeting. The main improvements could be in the area of (i) information dissemination about program eligibility and application procedures; (ii) empowering and guiding social workers to apply the criteria and procedures according to good practices and in an uniform manner across the country; and (iii) improving quality of data and data management. More specifically, the implementation of the UMB will benefit from: • Collecting and / or verifying and cross-checking the majority of the administrative data in the applications by the social workers through direct (on-line) access to other administrative 82 databases. This is how work and pension records can be checked, also employment status, disability certification, whether family members are receiving full state support in social care residential institutions, etc. , without burdening the applicant with providing written documents to certify each of these circumstances and with taking the cost of that; • More broadly providing uniform information on UMB eligibility criteria and application procedures across regions, to reduce the administrative burden of dealing with speculative enquiries, ‘just to try’ applications and those that are filed due to limited information on the purpose of the UMB and limited transparency around eligibility requirements; • Intensifying investments in the human factor – social workers. Inadequate public resources for administration may impede social workers from performing their functions effectively, and even become conducive to exploiting opportunities to commit fraud; • Introducing training, guidance, check lists and operation manuals for verification of eligibility through home visits and following these uniformly across regions to reduce arbitrary and discretionary judgments; • Introducing incentives and material conditions (vehicles, travel allowances) so that the social workers can fill Social Passports for all extreme poor and map poverty and vulnerability even in the most remote rural and mountainous communities; • Emphasizing reductions in errors of exclusion by providing incentives and performance criteria for social workers to actively seek eligible families and help them overcome obstacles that prevent them from proving with documents that they meet the UMB eligibility criteria; • Formalizing how benefit applicants are notified about a rejected application to improve transparency. At the moment, decisions are communicated in a conversation with the social worker. Using letters and more formalized approaches seems advisable; • Improving accountability of social workers and community committees for turning down potential applicants. This can be done by including standardized reasons for rejection in the informing letters; • Strengthening UMB payment methods and instruments, e.g. through bank debit cards where possible. • Introduction of a centralized and uniform across the country social assistance management information system with unified registry of applicants and beneficiaries, and possibilities for inter-agency data exchange. The effectiveness of the social safety net in general, and the UMB in particular may also benefit from reforms outside the sector, such as the computerization of land, agricultural and real estate registries, and the use of this information to cross-check the information provided by the applicants. CONCLUSIONS The analysis reveals that “perfect implementation� of the existing means-testing instrument would outperform the simulated alternatives and suggests focusing on improving the current MT rather than looking for alternative targeting methodologies. The comparison of the simulated alternative approaches to UMB targeting (categorical targeting / universal child benefit, PMT and a combination thereof), reveals that: • the universal child benefit stands out with a very high coverage of the poor (bottom 20 percent of the population) – close to 100 percent and twice as high as the simulated coverage of other two options. This responds to the fact that the presence of children in a household is one of the strongest indicators of poverty and that the poverty risk increases as the number of children in the household goes up; • with respect to targeting efficiency (share of benefits going to the poorest quintile) the categorical child benefits perform less favorably. Due to the universal character of the benefit, 83 leakage rate to non-poor would be considerable and more than two times higher compared to the rest of the options under the assumption that the take-up rate is 100 percent; • the PMT, either combined with a limit to children or not, displays very similar outcomes in terms of coverage and targeting because most of the poor households are with children. Leakage to less poor households would be around 40 percent of the total benefit sum; • the simulated policy options can be effective in reducing extreme poverty. Targeting the 20 poorest households using a PMT and assigning a budget neutral flat rate benefit either to all household members or children only would achieve a reduction of the extreme poverty gap of more than 70 percent However, this is the optimal scenario under the assumption of full compliance and take-up; • consequently, when comparing the outcomes of the simulated targeting options with the current UMB means test functioning under ‘perfect’ conditions as counterfactual, it becomes obvious that inclusion error and leakage in all three of them are much higher compared to the means- tested UMB if operating under optimal circumstances. An important implication of the above simulations is that improvements in the performance of the UMB need to come from extension in coverage. The existing means test already performs fairly well in terms of targeting accuracy (and improvements in implementation – rather than design – could further strengthen the application of that potential, as discussed below). Assuming perfect application of the means test in practice, the existing instrument outperforms simulated alternative design options. This indicates the need to shift the emphasis on policy reforms away from a focus on reducing errors of inclusion to one that emphasizes improving coverage among the poor (reducing errors of exclusion). Improvements in coverage depend on stronger outreach to encourage potentially eligible poor families to apply for benefits and expanded fiscal efforts (possibly coming from a consolidation of other programs). The analysis also suggests that modifications to the eligibility identification instrument would marginally improve the targeting performance of the UMB. Specifically, a modified MT that incorporates presumptive income from livestock in the overall income (a type of HMT) would only marginally improve the targeting efficiency of the program, increasing the share of correctly identified households in the poorest quintile from 56 percent under the current instrument to 60 percent under the new UMB income specification. Moreover, the ultimate effect of the inclusion of the predicted income from livestock would depend on how the imputation is carried out in practice. For example, using oblast- level coefficient (similar to the ones used for imputation of the income from land ownership) may lead to targeting error because the estimated revenue will not take into account the actual household situation. In addition, the cost of establishing, implementing and regularly updating the methodology for accounting for livestock income has to be considered against the potential benefits. These considerations confirm the importance of implementation and do not doubt or rule out the use of the HMT approach. The HMT approach can have different applications in the safety net of Kyrgyzstan. Along with being an integral part of the UMB program eligibility criteria to determine which families will be accepted in or reject from the program, the HMT can be used as a verifier of the eligibility of UMB program beneficiaries already receiving the UMB for inspection and audit purposes and eventual detector of error and fraud in the system. The HMT can become an identifier of poor households which are subject to exclusion based on formal income. Finally, implementation matters considerably for good UMB program performance. Since design options for reforming the program would have limited impacts on improved performance, efforts need to focus on strengthening implementation. There are many strong features of the implementation arrangements for the UMB, including: on-demand applications, verification of eligibility, and regularly scheduled recertification procedures. Nonetheless, there are ways that the implementation of the UMB could be strengthened, for improved program outcomes, efficiency and reduced fraud and error. Improvements could involve: • Reducing the number of required documentation and minimize the number of documents which have to be submitted for re-certification; 84 • Ensuring that all the documents attesting eligibility are provided free-of-charge, and that as much as possible the flow of documents is through administrative channels and do not require beneficiaries to travel extensively to collect and submit them; • Increasing the capacity and the incentives for social workers to reach and identify the poor and vulnerable, including through expanding the use of the Social Passport for initial identification and self-targeting of the poor; • Strengthening the control over the decision making on eligibility for UMB benefit by enforcing a requirement for the social workers and committees to write down the reasons for rejection; • Disseminating more information about the access to UMB program; • Improving the screening of welfare characteristics during the home visits by more structured observations; • Expand the community involvement in identifying poor. A number of improvements of UMB administration can help prevent error and fraud, and at the same time improve the access to UMB and efficiency of targeting. The main improvements could be in the area of (i) information dissemination about program eligibility and application procedures; (ii) empowering and guiding social workers to apply the criteria and procedures according to good practices and in an uniform manner across the country; and (iii) improving quality of data and data management. 85 CHAPTER 6. REFORMING THE OVERALL SAFETY NET – OLD CHALLENGES AND NEW OPPORTUNITIES This chapter discusses the main challenges and options for reforming the safety net for a more consolidated, rationalized, and consistent package of benefits. More specifically, the chapter advocates for completing the monetization of categorical in-kind subsidies and benefits. Their gradual scaling down and mainstreaming into the poverty-targeted UMB and categorically targeted MSB would eventually allow increasing the generosity of the UMB and MSB and scaling up their poverty reduction impact. This reform is gaining even increased importance in the context of the increasing food and overall inflation which undermines the purchasing power of cash benefits. The chapter thus also addresses the potential role and reforms of the safety net within the context of the food, financial and fuel crises (of 2007-08) and the emerging effects of the global economic crisis. Finally, the chapter reviews the legislative pre- conditions and implementation status of social services and makes the case that improving the design and delivery of non-contributory social benefits is not sufficient for effective mitigating the risks of falling into poverty or becoming vulnerable due to old age, disability, incapacity to work, losing parent or breadwinner in the family, or other reason. Social services and professional social work need to be developed and linked to the provision of social benefits to address the needs of the poor and vulnerable in a sustainable manner, and to promote investing in human capital through conditional cash transfers and activation policies. The Government’s vision of safety net reforms is being shaped by the strategic priorities of the Country Development Strategy and the Social Development Concept which call for: • Increasing the generosity of the non-contributory social benefits through increasing the value of the GMCL and sustaining or increasing their value relative to the GMCL; • Enhancing the targeting of the UMB to the poor by more accurate accounting of all available incomes; • Linking the MSB with services relevant to the specific needs of the beneficiaries, and extending its coverage to mothers engaged in home care for disabled children; • Reforming further the categorical benefits and subsidies and introduce targeting • Developing partnerships with the civil society sector for the provision of social care services outside the state institutions and to increase the quality of institution-based care. With respect to the non-contributory social benefits, the envisaged reforms could be qualified on one hand as ‘parametric’ and ‘second generation’ reforms because they build on the already established in the past decade foundation of the safety net. The reforms would deal with improvements in the design and implementation of the current system and focus on fixing the ‘nuts and bolts’ of the existing fairly well targeted anti-poverty programs (UMB and MSB). At the same time, the safety net reforms go beyond the ‘nuts and bolts’ by advocating for a holistic approach to social strategy development and for linking social benefit provision with social care and employment services. Their pace and scopes are challenged by the unfolding economic crisis which slows down economic growth and state budget revenues while at the same time putting pressure on the safety net to respond with counter cyclical increase in funding and with better outreach to the poorest. MONETIZATION AND PHASING OUT OF THE CATEGORICAL BENEFITS Categorical in-kind subsidies and benefits are legacy entitlements for people in defined categories to the free delivery of goods and services or to a broad and variable array of price discounts (such as for within- city transport, travel within the country, communications, energy, medicines, health services, housing, vacations, utilities). By design they are not targeted to the poor - the access is categorical, no income or means test is required, except for four out of 38 types of them. Indeed, the analysis confirms that these benefits are regressive. Although their coverage is rather high among the poorest households, the largest share of benefits is captured by wealthier households. Poor families are the most numerous group among the recipients. They account for 77 percent of all recipients, but receive quite low amounts - on average 86 78 KGS per month. World War II veterans and other war veterans and their families form the second largest group accounting for 13 percent of all recipients. Their entitlement is most generous - according to the law, they are entitled to more than 4,000 KGS per month if all types of eligible subsidies and benefits are taken at the maximum value. Due to partial take up of non-monetized benefits and subsidies, the actual monthly amounts received are lower. Based on data from the MLSD and the Ministry of Finance, we estimate that about 20 percent of non-monetized categorical subsidies are actually taken up by the beneficiaries. Historically, categorical in-kind subsidies and benefits have been granted in kind. The first attempt for monetization was made in 1996, by converting the benefit for rehabilitation and sanatorium treatment into a cash basis payment. Systematic monetization with the objective of monetizing all categorical in- kind subsidies and benefits is in place since 2003 when the benefits for coal and medicines were monetized throughout the whole country. As of July 2006, cash transfers for utilities and energy became optional. From January 1 2008, total monetization applies to transport and municipal services, except for the beneficiaries in Bishkek who still can opt to get them in kind. The option is being kept because beneficiaries fear that with time the monetary value of the cash transfers will decline thus reducing the scope of and undermining the access to the respective services if the cash transfers are not regularly indexed with inflation. Currently, the process of monetization is close to completion. The majority of categorical in-kind benefits are monetized except for glasses, hearing aids, dental work and travel within the CIS countries, 56 as well as the utility and housing subsidies for the citizens of Bishkek. The monetization is taking place gradually, building on lessons learned from Russia and Kazakhstan. The budget for cash compensations is provided in a timely manner, and the benefit amounts are updated following the increase in the prices of coal (differentiated by regions), energy tariffs, transportation and utility costs. The monetization was undertaken after consultations with think tanks and civil society organizations with respect to social impact and methodology for valuation of different kinds of services based on their average cost, frequency of use and other criteria. Monetization itself is not going to solve the complexity of operating numerous benefit types and beneficiaries, and will not lead to reducing the budget allocations for these programs. On the contrary, administrative data suggests that it – especially initially – acts towards increasing budget allocations for the categorical in-kind subsidies and benefits due to increase in take up. The most pronounced budget increase was observed exactly at the start of monetization, when in 2004 it grew by 71 percent compared to 2003, without major change in the number of beneficiaries and without major adjustment of benefit value for inflation. The budget increase was mostly driven by the higher uptake of sanatorium vouchers, and to a much lesser extent to the uptake of compensations for transportation, communication, electricity, thermal power and utility benefits. At the same time some benefit payments which were made before to the utility companies in full (at 100 percent irrespective of de facto usage rate), have been reduced in accordance with the actual take up of the respective benefit. This mostly applied to the electricity and energy benefits for population living in high mountainous areas, where the actual take up of certain utility benefits is less than 100 percent, and sometimes as low as 25 percent.57 For the forthcoming years, the MBTF assumes annual update of the benefit levels with inflation and reduction of the benefits categories (currently 38) by around 10 percent anticipated due to natural attrition of beneficiaries. Under these assumptions, the budget for categorical benefits and subsidies will not be increasing in 2008-2010. In 2008, 688.7 million KGS were allocated, representing an increase of 16 percent compared to 2007. In 2009, the planned allocation of 639.9 million KGS is by 7 percent lower than in 2008. In 2010, the republican budget will allocate 674.2 million KGS which in nominal terms is close to the 2008 republican allocation for the same purpose. Reforming categorical in-kind subsidies and benefits is a politically sensitive issue. There is a deep- rooted belief that there are groups in the society that deserve to be privileged, like war veterans, participants in the Chernobyl clean-up, senior citizens with special contribution to science and art. Despite this, MOF and MLSD are committed to start a reform acknowledging their regressive nature and costly administration. Given the sensitivity, a reform needs to be designed thoroughly. It could start with synopsis of the current situation, possibly based on a bigger sample of beneficiaries and improved 56 The trips to sanatoriums are also still most often received in the form of a voucher but can be received in cash as well 57 MOF data. 87 questionnaire module on privileges in the KIHBS. A booster can be introduced in the survey sample to capture more observations for a more representative analysis. Following on the analysis, scenarios can be developed taking into consideration the impact of the reform on the recipients’ welfare status. The process need to be consultative – to take into account the views of the affected beneficiaries, and also to take stock of other countries’ successful experiences. After the ‘losers’ are identified, the reform should envisage mitigation measures in terms of services, special cash compensations or introduction of poverty targeting. The reform of categorical in-kind subsidies and benefits cannot be standalone. It needs to be put into the context of wage and pension reforms, energy and price and utility fee liberalization, disability reform, reform of social assistance and social services, and many more. Finally, it needs to be based on a new legislative bill to be subject to public discussion and ex ante impact evaluation. Possible options for a stepwise reform of the system of categorical in-kind subsidies and benefits could look at several options which are not mutually exclusive but rather complementary to one another: (i) gradual phasing out (grandfathering); (ii) consolidation of similar privileges or privileges for same beneficiary category into a lump-sum payment / single benefit; (iii) integration or mainstreaming / merging into other benefit programs (basic pension, UMB or MSB) and, (iv) scaling down through poverty targeting with an income and asset test or with other methods. Moreover, that four out of the existing 38 types of categorical benefits and subsidies are means-tested already. It should be made clear that the objective is not to deprive recipients from getting categorical benefits but to adjust them more flexibly to the changing composition of recipients, their needs and ability to use the respective benefit. Finally, a less complicated system of categorical benefits and subsidies can reduce the cost of administration while maintaining and improving quality. The main concepts/approaches on how to reform them are presented in Table 6.1 below. Table 6.1: Possible reform options for the categorical in-kind subsidies and benefits Option Definitions Examples / reform options Phasing out / Phasing out over time following • All kinds of privileges for war veterans, solders Grandfathering natural attrition of the remaining internationalists, participants in the Chernobyl beneficiaries combined with (i) gate clean-up could qualify for this option keeping to the granting of new • Sanatorium benefits need to be analyzed to entitlements and (ii) reconsidering understand why the vouchers are cashed more the relevance of certain benefits vis- often than used, and to be phased out if found à-vis the needs of recipients irrelevant or difficult to access • Benefits for purchase of a horse and cart also need to be reconsidered because they aim at increasing earning capacity while many of the recipients are of retirement age Consolidation Provision of one lump sum / single • Certain yearly allowances granted now to war cash benefit instead of several veterans such as for medicines, horse and cart, fragmented entitlements going to the heating sanatorium treatment, transportation and same beneficiary category without housing can be combined / consolidated into a undermining the overall amount of single cash benefit (veteran/military allowance) the cash received before the reform. • The price subsidy for the purchase of coal and the Separate groups of beneficiaries can one for other sources of heating can be be subject to consolidation as well. consolidated / merged together given their The approach envisages that the common objective consolidated benefit is added as top- • In the case of people with disabilities, several up to other income received by the small disability benefits can be lumped together, beneficiary (basic pension, disability or the people with disabilities can be regarded in pension, wage where applicable) broader categories, instead of having special benefits for deaf, blind, etc. people (disability compensation benefit, social integration benefit) Integration / Including the in-kind benefit into the • The benefits currently received by pensioners mainstreaming regular pension as a one-time based on merit can be made part of the pension as 88 Option Definitions Examples / reform options increase. The approach envisages one-time lump sum increase that the benefits become part of the • The pensioners’ road transportation benefit can be pension or MSB where applicable, do added to the basic pension not exist separately any more • Benefits for deaf and blind could be added to their monthly disability pensions • The annual compensation for medication, also the value of the sanatorium voucher can be added to the MSB • The annual compensation for fuel, and the monthly energy support can be added to the UMB Scaling down Could be achieved in two ways • The redundancy of the allowance for horse and which are complementary to one cart to which the World War II veterans are another and could be applied entitled might be considered because the current separately or in combination: (i) beneficiaries are already out of the work force and abolishment of certain already not expected to work. As a result the benefit might irrelevant categorical benefits and (ii) be captured by a non-eligible individual within or targeting of the remaining benefits outside the veteran’s household. A through (a) means test of the existing veteran/military pension could be the substitute beneficiaries and shifting those under • The practice of extending merit based pensions the threshold to the UMB and (b) can be phased out by linking more tightly categorical test for disability and pensions to individual payroll and contribution inclusion in the MSB program record • The existing beneficiaries if different subsidies could be required to pass a means-test. Those who are poor should be included in the UMB program. Means testing is already applied in the case of four types of categorical benefit sand subsidies. Possible ‘candidates’ for means testing could be the benefits for people living in high mountainous areas where 60 percent of the population is indeed poor, but the 40 percent who are not poor also receive the same package of categorical benefits • Those who qualify for categorical assistance because of disability should be included in the MSB program. The budget for the MSB and UMB will be scaled up to (e.g. with the current allocation of 200 million KGS) to compensate for the lost income from the categorical benefit. In summary, monetizing and reforming of the categorical in-kind subsidies and benefits are expected to have positive impact on benefit administration (simplification and cost reduction) and to increase the benefit utility for the recipient. Monetization increases the efficiency of benefit delivery - cash is easier, cheaper and simpler to deliver than in-kind entitlements. Delivery of cash for a multi-purpose benefit increases the flexibility of its use, allows choice and prioritization and eventually increases the overall utility of the benefit. This is particularly valid for cases of unnecessarily “atomized� benefits. At the same time the cost of the programs potentially increases when / if the intake is based on eligibility, and not on actual usage / take-up like in the case of in-kind benefits. With consolidation, integration and scaling down of the benefits, their number is going to be reduced and eligibility allegedly simplified. The same budget could be allocated in a more targeted manner for categorical benefits, or partially reallocated for clearly progressive benefit programs like the UMB and the MSB. A comprehensive review of all kinds of categorical benefits, beneficiary categories, eligibility criteria and their verification, and spending patterns will shed light on their relevance and impact on consumption, poverty reduction and quality of life of the beneficiaries and compare the impact with the associated administrative burden. USING THE SAFETY NET TO RESPOND TO CRISES 89 Worldwide the fuel, food and financial crisis, and the developing global economic crisis disproportionately affect the poor and near poor. These adverse impacts can be temporary, but also with long-term consequences. The crises establish a new role for safety nets: to protect the incomes and consumption of the poor; to help them avoid irreversible losses of physical assets and human capital; to maintain social and political stability and to serve as counter-cyclical stabilizers (to automatically expand with crisis) when necessary. The safety net of the Kyrgyz Republic is already faced with the need to respond to crises. Expected impacts of the crises The welfare of the population has been already undermined by multiple impacts of the crises on the product, financial and labor markets. The product markets face slowdown in the growth of domestic agricultural production and sales, contraction of the construction sector (by 4 percent in 2008 compared to 20-percent growth in 2007), increase in imports and subsequently higher trade deficit, higher inflation and changing relative prices. The financial markets crisis is already cutting back the value of savings and assets and curbing the access to credit which started reducing the fixed investments in the economy. The labor market offers less opportunity for productive employment which affects adversely incomes, especially in the urban parts of the country. The depressed demand for labor in the countries which used to be and still are the main recipients of Kyrgyz migrant workers (Russia and Kazakhstan) is expected to slow down remittances which are now quite high as share of GDP (17 percent) and are more important for improving household welfare than social transfers. The Kyrgyz Republic was first hit by the food price increase. In 2007 and the first half of 2008, the country had the highest food inflation in Central Asia. Food prices drove a 20-percent pick up of the consumer price index in 2007 and another 30-percent pick up in the first half of 2008.. In 2007, the food prices in Kyrgyzstan used to increase more sharply compared to Russia, Ukraine and Kazakhstan, and their impact on boosting up inflation was the strongest compared to the above mentioned middle-income CIS countries. Higher food prices undermine the food security affecting both availability and affordability of foodstuffs. The Kyrgyz Republic is a net importer, and the dependence of imported and more expensive foods is increasing. The wheat grain balance is increasingly achieved through imports from Kazakhstan. 87 percent of the food import comprise of staple foods: cereals (40 percent), sugar (15 percent), coffee and tea (13 percent), meat (8 percent), and fruits and vegetables (8 percent). The costs of these food imports rose by 50 percent between 2006 and 2007, from about US$35 per person to about US$53 per person. Thus food imports contributed to increasing the current account deficit and fiscal vulnerabilities Food inflation was the principal driver of overall inflation till the middle of 2008 (Figure 6.1). Compared to 2006, in 2007, the prices of bread and cereals increased three times faster than all prices and twice compared to food prices. In 2008, the prices of key staple foods like wheat and milk continued to increase and in the end of 2008 were substantially higher than in the end of 2007. In the second half of 2008 food prices somewhat stabilized reflecting the downward trend in international commodity prices. However, the prices of some major imported food items continued to rise: sugar by 31 percent, grains – 2 times 58. In parallel, at that point of time the country started experiencing the adverse impacts of the fuel and financial crises. The imports of non-food consumer goods and energy tripled in a year, driven mainly by increased prices. As a result the domestic prices of services grew by 33.5 percent driven primarily by the hike in the tariffs for utilities, transport and education. Consumer inflation remained high – 19 percent annual growth (November 2008 compared to November 2007). Figure 6.1: Contribution of food inflation to overall inflation (Jan. 2006 – May 2008) 58 NSC data. Kyrgyz Republic: Recent Economic Developments, December 16, 2008. 90 Figure 9. Kyrgyz Republic: CPI Components (12-month change, percent) 29 Food and non-alcoholic beverages Electricity,gas and other fuels Others 23 17 11 5 -1 Jan-06 May-06 Sep-06 Jan-07 May-07 Sep-07 Jan-08 May-08 Source: GoKG, World Bank and IMF Conference, Bishkek, June 25, 2008. The high food prices raise the risk of falling into poverty or retaining the poverty status for a long time (chronic poverty). • Higher food prices reduce purchasing power as less food can be purchased with the same budget, and erode consumption. Impact is significant because the food share in total consumption is 50 percent or more for all income levels except the most affluent quintile if the society, for whom it is 41 percent (Figure 6.2). In the poor households (two bottom quintiles) respectively 60 percent and 55 percent of the household budget is spent on food. • The consumption of the poor is even more at risk given that almost 30 percent of the food expenses are only on cereals. Flour and bread, along with cooking oil, are prevalent in the consumption basket of all poor, especially the rural poor, and the drastic increase of their prices presents an immediate and high food security risk (CASE, 2008). Particularly hard hit are those who are both poor and net food consumers. Such is 20 percent of the population. • In the context of Kyrgyzstan, where poverty is mostly rural phenomenon, the rising food prices would increase risk of poverty in urban areas and Bishkek in particular, where the net food consumers prevail over the net food producers, and thus change the poverty profile. • The higher food prices would also undermine the quality of food consumption due to possible substitution / shift to products which are cheaper but with inferior nutritional value. • Finally, an increased household budget for food could crowd out other expenditures for basic needs such as health care, education or housing.) Figure 6.2: Food share in consumption by population quintiles 91 Food share in total consumption, in % 70 60 60 55 52 50 50 41 40 30 20 10 0 1 2 3 4 5 Population Quintiles Source: GoKG, World Bank and IMF Conference, Bishkek, June 25, 2008 Ability to leverage safety net programs in response to crises The safety net response is a part of the overall policy response to crises and complementary to other sector policies. It can support food and energy security, limit the negative impact on poverty and inequality, ensure access to health and education. The main policy questions would be first, whether the existing non-contributory benefits could be used to deliver additional resources to the current beneficiaries and identify new poor, and, second, what adjustments to the design and implementation of the existing programs need to be made to increase their relevance for safety net response to the food and energy price increase. Ideally an existing benefit program should be selected for the purpose because instituting a new one would require time and resources for design and instituting. The time frame for launching a new safety net program is estimated at 9 to 12 months. The main characteristics which could make an existing program suitable as a crisis response instrument should be its good targeting performance and implementation (established administrative and payment structures). In the safety net in the Kyrgyz Republic, two safety net programs – the UMB and MSB - have been already identified as relatively well targeted. UMB is clearly progressive (the poorest 20 percent capture 38 percent of total benefits going to those in the poorest quintile, and another 43 percent going to those in the next quintile) and fairly well targeted by international standards. This makes the UMB the logical choice of a safety net program which can be used to mitigate the poverty risk arising from the food, fuel and financial crises. Such assessment is similarly applicable to the MSB which is categorically-targeted (but not means-tested) to disadvantaged groups (disabled, orphans, the elderly without pension rights, mothers of three or more children, etc.) whose vulnerability also increases during a crisis. Without the means-testing, the MSB is not as well targeted to the poor as the UMB, but is also a progressive and relatively well targeted transfer with 33 percent of benefits going to those in the poorest quintile. At the same time the suitability of both programs could be improved substantially to make the policy response to the crises more effective. Having well functioning safety net programs is a window of opportunity. Along with good targeting, programs used for crisis response with the objective of protecting the consumption of the poor, need to reach as much poor as possible, to be adequately funded and to benefit from good capacity for management and administration. At present both programs have limited coverage: 14.6 percent and 6.4 percent of the total population for UMB and MSB respectively, and 28.2 percent and 12.7 percent of those in the poorest quintile respectively. Even more important challenge is that the levels of benefits are low and represent only a very small share of total consumption of the poor, e.g. as already stated the UMB delivers low benefit amount that equals to 7 percent of the consumption of the poor (bottom 92 quintile) in 2005 and even lower - 4 percent in 2006 59. The UMB program is lagging behind other ECA and CIS countries in terms of public spending as share of GDP. Hence, increasing the funding for the UMB program is of immediate importance to compensate for already lost purchasing power. In a situation of tightening budget expenditures the necessary increased fiscal outlays could come from consolidating other less-well targeted benefit programs. At the very minimum, budgets should be protected for the well-targeted programs (as well as for spending on basic health and education) (Lindert, 2008). Greater adequacy of both programs could be achieved in two main ways: first, through temporary scaling-up / topping up of the value of the UMB and MSB combined with expanding of coverage through increased outreach efforts, and, second, through indexation of the GMCL, and the UMB and MSB respectively, with the consumer price index. The scaling up of the UMB and MSB is being introduced already. Currently the UMB is being topped up with one USD per month for 10 months starting in October 2008, thanks to a World Bank grant from the GFRP for UMB additional financing, with possible extension with financing from other development partners or state budget resources. The top up is kept separate from the main benefit amount and equal for all beneficiaries to allow scaling down when the crisis response is not needed any more. To increase the impact of the top up it is being complemented by delivery of micronutrient supplements for specific very vulnerable groups for whom the risk of contraction of consumption and reducing of its quality is very high – pregnant and lactating women, young children. The top up is also complemented by interventions aimed at increasing the outreach to the poor, for rapid assessment to identify new poor who for different reasons are not yet included in the UMB program, and for improved monitoring and assessment of program impacts. The MSB is topped up with state budget resources. The topping up of the UMB and MSB has distinct advantages: • It is administratively easy to introduce and deliver, especially when one applies a flat rate (calculated as a share of the average benefit amount and delivered as uniform top-up to all recipients); • .If the top up is indicated as a temporary measure, it could be phased out smoothly; • The top up can be introduced as a flat or variable rate. The variable rate would incur additional administrative costs for calculation of marginal variations in the top up by the social staff. The uniform increase is also more suitable from the perspective of keeping the top-up separate from the basic benefit which emphasizes on its temporary/ transitory character. • In addition, it can be introduced quickly once the resources are made available. The approach has also its disadvantages: • Only beneficiaries who are eligible under the current program criteria will benefit from the topping up. Poor households without children who are not eligible for the UMB will remain without access to it even if poor. • Despite of the ex ante ‘separation’ of the top up from the ‘main body’ of the benefit, it may be politically sensitive to withdraw the top-up once it has been installed. However, despite the stated disadvantages this is the best possible safety net instrument for protecting the consumption of the poor in the short run. It is already implemented by the Government of the Kyrgyz Republic. The main questions are what amount of the top up would be meaningful for the recipients and affordable for the government of Kyrgyzstan, and for how long should the additional benefit be granted. Figure 6.3 shows the development of total costs for topping up the existing UMB and MSB with one or two USD (36 or 72 KGS) per month. The additional costs for an additional dollar per UMB and MSB beneficiary per month is 16.2 million KGS and 2.1 million KGS respectively. 59 Based on KIHBS 2006, households receive the UMB on average during 7 months. The average benefit size per eligible recipient per months is 56 KGS. According to administrative data, the average UMB was 124 KGS per month on 2006. 93 Figure 6.3: Development of total benefit costs over time: topping up the UMB and MSB (based on administrative data) 1200 1000 800 add 1 USD to UMB Mio KGS add 1 USD to MSB 600 add 2 USD to UMB add 2 USD to MSB 400 200 0 0 1 2 3 4 5 6 7 8 9 10 11 12 Duration (months) Source: Staff calculations The second approach - to protection the purchasing power of the UMB and MSB by raising the GMCL – is also been applied. Both the UMB and the MSB are linked to the GMCL which serves as the basis for determining and increasing their benefit levels. The GMCL has been raised already a few times over the past years, but these adjustments have had an ad-hoc nature and reflected rather the budgetary resources availability than the needs of the poor. As a result, the value of the GMCL has remained relatively low. In 2006, the GMCL was 175 KGS per month, representing 31 percent of the extreme poverty line. Further adjustment was introduced as of January 1, 2008 when the GMCL was raised to 200 KGS but this did not fully compensate for the inflation (as seen from Table 6.2).. Table 6.2: Annual adjustment of the GMCL with the CPI 2007 2008 2009 2010 Inflation (end of year) % 20.1 15.0 7.0 6.5 GMCL - ad-hoc KGS 175 200 GMCL – CPI KGS 175 210 242 259 Average UMB KGS 128 154 151 189 Average MBS KGS 514 617 718 759 Source: Staff calculations based on administrative data from the MLSD; inflation: IMF (2008) Indexing the GMCL to the CPI on a regular basis and taking into account the full scope of price increase has the definite advantage that it can ensure the regular and sustainable adjustment of the benefit levels to price changes provoked by crises. As a result of regular indexation, the average values of the UMB and MSB would increase as well and their adequacy will improve. However, taking the commitment for regular (annual) indexation of the GMCL with the CPI will have fiscal consequences which will be difficult to forecast. It will be difficult to assess the impact of the regular increases on the number of eligible beneficiaries. Increasing the GMCL may increase the number of eligible households for the UMB if the incomes at the lower end of the income distribution to not develop at the same pace as the CPI. This introduces some uncertainty for the administrator in terms of budgetary needs (may not be fiscally sustainable). Furthermore, local social staff has to recalculate the benefit values annually. Also, since the increase is integrated in the overall benefit amount, and does not 94 appear as top up, and follows a pre-defined principle, it will be politically difficult to withdraw from the commitment. Finally, as in the case of the UMB top up, poor households without children will remain without access to the benefit. At the time of the crisis targeting can be challenging. Family circumstances change more often and require higher administrative effort for managing same number of cases. The informal economy expands as well as the share of hard-to-verify income. Generally, in such circumstances increases the number of those trying to enter last resort income support programs or disability benefit programs. At the same time the outreach to the extreme poor remains difficult. For the topping up and/or indexation of the UMB, it is particularly important to quickly remedy as much as possible the error of exclusion by increasing the effort to identify the new poor and to include the current poor who are left behind because cannot comply with certain (mainly administrative) requirements for benefit eligibility (details of implementation issues were already highlighted in chapter 4). Improving the coverage to include all households below the extreme poverty line (perfect targeting) with an average UMB per household per month of KGS 260 (as in 2006) would bring the coverage of the UMB to 60 percent of the poorest quintile and at the same time will require increasing the cost of the UMB program with 205 million KGS annually. The impact of the short-term solutions related to expanding the coverage and improving the generosity of the available well performing safety net programs will be higher if combined with medium term measures to strengthen targeting (introduction of “proxy indicators� to use as “validators� or predictors of income to enhance means-testing) and introduce a management information system. LINKING BENEFITS TO SERVICES With the developing global crisis, the risk of reducing the levels of consumption and decreasing the ability to smooth fluctuations in consumption are expected to be combined with increased risks of irregular school attendance and early dropping out (as poor families pull their children out of school to work), change in the stricture of food consumption which becomes less diverse and biased towards cheaper and lower quality foodstuffs, lower household investments in child health and nutrition (with often irreversible long-term consequences for cognitive and physical development), and a general worsening of household welfare, social indicators and poverty. In this context, additional medium-term policy challenges emerge related to how to use the safety for investing in human capital. One way of enhancing the human capital promotion impact of the social benefits is to link them to social services. More specifically, one might consider to: (i) link the UMB (or the top-up to it / bonus) to human capital incentives (conditional cash transfers); (ii) link the provision of cash to the UMB and MSB beneficiaries with provision of targeted social care services; and (iii) introduce work-related aspects, such as work requirements and/or links to job support services (job search, training, etc.) for adult able bodied persons living in families which receive the UMB. Current status of social care services Social care services are a newer area of social policy development and implementation, where the efforts of MLSD will be focused in the next years to transform the pilots into sustainable government policies. In Kyrgyzstan, they exist as a system of social support to people at risk. Generally, services are provided by the state with the support of the local authorities. Only in recent years NGO provision is being piloted in different parts of the country with financing from the state or the development community. The needed basic legal and regulatory framework is in place. The types of social services provided by the state, the financing standards / norms and the eligible beneficiaries for free-of-charge services or for services with co-payments are regulated by the Law on Foundation of Social Care in the Kyrgyz Republic and the Law on Minimal State Social Standards. Social services are defined broadly as services related to improving housing and living conditions, provision of food, home care and organization of free time, socio-medical and recreation services, provision of medical and prosthetic-orthopedic devices for people with disabilities, social and on-the-job rehabilitation of people with disabilities, access to education for people with disabilities, legal services, and funeral services. The state has defined norms for service utilization and has set a network of state service providers. 95 Social care services are provided in different settings: residential institutions, at home or in day-care centers. MLSD is responsible for 15 institutions for elderly, and people with disabilities and children with disabilities, hosting altogether close to 2,500 residents. More boarding houses and orphanages exist under the MES. The financing of institutions (including for capital repair and purchase of special equipment) has increased from 84 million KGS in 2006 to 115 million KGS in 2007, or close to 1.4 times. Home-based care is the responsibility of the social departments of the MLSD at oblast and town level, and the aiyl okmotu. Home-based care (social patronage) takes care of home repair works, delivery of food, medicines, other purchases, payment of utility bills on behalf of the beneficiaries. Around 750 dedicated social workers are employed in the social patronage, implying average caseload of 13 per social worker. The average caseload is not high however there are significant variations across oblasts and communities meaning that the access to services and the availability of social workers is not even across the county. The main target group of the social care services are the elderly and the people with disabilities, followed by children at risk. At present, 9,748 elderly and people with disabilities are benefiting from home and community-based social care; 7,800 single elderly are identifies as in need of long-term home care. The demand for services for the elderly is higher than the present capacity for service provision. The current system needs reforms in the areas of financing, access and quality. The existing facilities need improvement but the financing standards are low and do not allow capital investments. For example, the residential care for elderly needs facility refurbishment, new medical, transportation and other equipment which are not affordable with the current financing. They also need updated norms for maintenance costs and the cost of labor to be able to attract qualified social workers, medical workers and psychologists. The financing for food is also low – at the level of 30 KGS per person per day it does not allow compliance with nutrition norms 60. The access of people with disabilities to regular medical and social assessment, rehabilitation and labor market integration services is an issue. Medicines and medical checks are expensive and not affordable by the institutions even in the cases where co-payment is allowed. The state guaranteed package of medical services and medicines is far from being able to secure regular checks, timely treatment and rehabilitation. As a result, the number of neglected cases of disability is increasing and the number of certified people with grave and uncared for cases is increasing. The social services for children target different groups at risk: children with special needs, children suffering from chronic diseases, working children, children in institutions, drop-outs from the school system. The provision of such services is gaining momentum, especially after the introduction of the Child Code emphasizing on protecting the rights of children and promoting a new institutional framework for child protection.According to different sources, there are currently more than 10,000 neglected children, more than 23,000 working children and 20,000 children not attending school. There are currently 19,400 children with disabilities below 18 years (20 percent out of total number of people with disabilities), 62 percent of whom live in rural areas where the social services network is much less developed, poverty is higher, and the quality of medical and diagnostic services is lower. Nowadays 83 percent of children of early age suffer from anemia, 40 percent of school-age children have chronic diseases, around 70 percent of children have iodine deficiency and 33 percent are with vitamin A deficiency. Child labor is one of the most difficult problems after independence. Since the country is agriculturally oriented, the biggest share of working children is concentrated in agriculture. In the rural area almost all children are working seasonally at plots belonging to the family. Child labor also expands in transportation and trade, mostly road and city green markets. Currently, there are over 5,000 institutionalized children in the country. To prevent further institutionalization, MLSD is developing alternative services for children and families at risk as family-type homes, centers for family and child support, child patronage / adaptation centers. They address new risks emerging from poverty and migration which often entail family dysfunction and child abandonment. The state launched the “New Generation� program to pilot until 2010 new social services for children and establish Family and Children Support Departments in two rayons. The future of social care services is conditional on policy decisions aimed at diversification of providers, reforming the financing, introducing standards of quality and improving the capacity of those who manage and implement them. The diversification of providers was launched with the “New Generation� 60 UNICEF data. 96 government program, where the state started purchasing services for vulnerable groups from NGO providers under the conditions of ‘social order’ 61. In 2006-2008, MLSD was allocating 10 million KGS per year for contracting out child welfare services to NGOs. However, the NGO capacity for service provision (the supply side) seems to be still limited as indicated by the fact that in 2007, the actual absorption of budget outlays for purchase of social services from NGOs was around 4.6 million KGS, or less than half of the allocated. The outsourcing of social services to non-government providers requires standards for services quality, as well as mechanisms for monitoring and controlling compliance with the standards. It also requires the establishment of management information services to assess local need and plan services; analyse and evaluate the impact of service delivery to clients; adjust and develop services on the basis of evidence of impact and effectiveness; assess costs and report back to government. Given the limited capacity of the NGO sector and the need of prudent and transparent management of public resources, MLSD will need to enact and enforce social contracting legislation (qualifying criteria for providers, licensing procedures, procurement procedures, etc.) and to assign responsibilities for financing, provision and monitoring of services to stakeholders in the state, local government and non- government sectors. The targeting of social services / identification of the beneficiaries will imply a decision on whether MLSD will continue to keep the existing eligibility criteria and norms which put strong emphasis on the categorical approach, or will move towards new targeting methods allowing identifying the needy that are excluded when the present targeting is applied. The availability of the social passports seems encouraging to this end. However, now the information obtained from each family in need of social benefits and services is not entered on a computer system, and only aggregate information from social passports is available, which is of limited use for targeting purposes. Computerization of the social passports, along with developing of expertise at local level to collect accurate beneficiary, costs and outcomes related data and to enter this data in electronic format would contribute to improving the targeting and eventually the access to social services. At MLSD level, expertise needs to be developed to analyze information and recommend where redistribution of resources could lead to better service outcomes. The social workers will need additional skills and qualification for identification of the target groups with different assessment tools, case management, ability to understand the continuum of care needed for each case, and ability to link social services with other social policy instruments, including social benefits. International evidence has demonstrated that linking benefit provision to educational, medical, nutritional and social (e.g. good parental practices, early childhood development interventions) investments in children is the more viable and cost effective alternative compared to crisis interventions. Conditioning the UMB to behavioural changes (Conditional Cash Transfers) Conditional cash transfers (CCT) provide money to the poor, including social benefits, contingent to them making investment in human capital, such as keeping children in school (measured by enrolment and attendance rates), providing nutritional supplements and / or taking them to health care centers on a regular basis. The CCT change the concept of social protection – it is no longer understood only as a way to mitigate risks of poverty and vulnerability, but is rather regarded as a way to increase productivity and earning capacity of the beneficiaries in the long run. There is a strong productivity case for investing in young children: returns are higher than for any other age group, impacts are multiple – on educational attainment, health status, cognitive ability, academic performance, tenure within the school system and higher incomes. Because of their dual objective to reduce the current level of poverty and at the same time to promote investments in the human capital of the poor, thus overcoming the intergenerational poverty trap, CCTs are becoming increasingly popular and widespread all over the world. The ‘classic’ CCT programs emerged and gained popularity first in Latin America, but are now popular in South and South East Asia, and are being piloted in the USA, Africa countries and elsewhere. In Europe, a CCT program in being implemented in Turkey, while Macedonia is designing one. The Czech Republic and Bulgaria provide child allowances upon evidence of school attendance; Estonia and Serbia link them to enrollment. Some ‘old’ EU countries have a ‘history’ of linking provision of social assistance to certain behaviors. For example, in Austria, conditioning of cash transfers to attendance of health facilities by young children and their mothers is in place for over 30 years. In Central Asia, Kazakhstan is piloting a CCT program with health and nutrition conditionalities for young children, pregnant and breast feeding 61 Law of the Kyrgyz Republic “On the State Social Order�, July 2008, No, 162. 97 mothers, and with schooling and social care conditionalities for children with disabilities and their families is under preparation. The UMB or its increments (cash top ups) can be linked to health and / or educational conditions. These are options which might be considered for modifying the UMB delivery with the objective of increasing its overall impact. Linking to regular school attendance would reduce the risk of child labor in rural areas and early dropping out in urban areas. This is of growing importance given that crises as a rule exacerbate such risks. Linking to immunizations and regular visits of health clinics for pregnant and breast feeding women and young children would work towards prevention and early detection of health care problems which affect the individual’s development, health status and productivity in the long run. Linking UMB to the intake of nutritional supplements could work towards reducing the risk of anemia, iodine deficiency, stunting. If combined with supply of micronutrients, the UMB conditional cash transfer could help reduce nutritional vulnerability of the poor. The “conditions for conditioning� of the UMB are: • The program to be well targeted to the poor, which is the case as shown by the previous analysis of the targeting of the UMB to the poor and extreme poor, to be adequately funded and to cover the predominant part of the poor children. Generosity and coverage, as already mentioned, are issues for the UMB, which could be overcome by increasing the budget and improving the outreach to the poor; • The constraints to human development outcomes to be identified, along with the behavioral changes which need to be promoted (types of conditionalities). For example, administrative data indicates that school enrollment is quite high in the Kyrgyz Republic, especially at the lower levels of educations and is not an issue. At the same time research (MISC) and anecdotal data suggests that child labor is an issue, especially during the crop collection season in Southern Kyrgyzstan. A CCT type of intervention using the UMB which is predominantly targeted to children in rural areas might become instrumental for reducing child labor. Similarly, the provision of the MSB to children with disabilities could be linked to requirements for school attendance, home schooling, health checkups and regular take up of rehabilitation services; • Compared to unconditional cash transfer, the CCT requires higher administrative capacity for monitoring the compliance with the conditionalities, which usually involves several actors in central and local government and is time sensitive, along with capacity for information dissemination on the impact of the desired behavioral changes; • Last but not least, the introduction of CCT will only make sense if the supply side is in order, e.g. if health and educational services exist and are accessible for the beneficiaries who in order to comply, will be raising the service use. However, children’s health and learning outcomes will only improve if the quality of services they receive is adequate, but the CCT does not inherently address it. Employment and activation services In Kyrgyzstan, employment services were virtually non-existent till 2004. Since 2005, the republican budget started making allocations for passive and active labor market policies. The overall allocations grow from 52.5 thousand KGS in 2005, to 72 thousand KGS in 2006 and 77.7 thousand KGS in 2007 but remain negligible as a share of the overall budget for social assistance (between 2 and 2.5 percent). Close to half of the budget is for compensations, one third – for public works and very little for training and micro-credits. In 2008, the State Committee on Labor and Migration’s target was to increase the number of formally employed by 2.4 percent compared to 2007. For that purpose, around 5,300 unemployed received job orientation, motivation and search services; 20,100 were offered participation in public work programs organized jointly with the local authorities; 12 million KGS are budgeted for micro credits (400-600 recipients of micro credits with average value of 20,000 – 30,000 KGS). The mismatch between labor supply and demand is being addressed through creation of an employers’ database of job vacancies, public-private partnerships for adult vocational training and re-qualification, and amendments to the labor legislation (expected by end-2008) to regulate labor migration and temporary employment of foreign workers in the border areas. 98 Employment activation generally assumes the incentives for work as the very heart of social policy. This implies (i) greater emphasis on work as a way to ensure that individuals in their prime age are not excluded from mainstream society; (ii) greater effort on behalf of the social and employment services in helping benefit recipients overcome the obstacles to entering into paid work, and (iii) greater effort on behalf of the benefit recipients to enter or re-enter the job market. Activation services are not offered now in the Kyrgyz Republic because (i) they are resource-intensive; (ii) require an established network of labor market services, social services, health services, housing sector, communities, etc., and (iii) job placement could be successful only if labor markets are reasonably ‘tight’ in terms of demand for labor. The policy which is closest to activation relates to the attempted UMB extension as an annual lump sum. It was launched in 2002 on a pilot basis to mobilize the human capital of poor families, promote entrepreneurship and self-help aimed at getting the participating families out of poverty. The initiative was also conducive for promotion of social capital building because it encouraged pooling the resources of several beneficiary families around a joint business project. In a five year period the number of volunteer families increased quite rapidly - from 25 in 2002 to close to over 9,000 in 2006, but not without administrative push. In 2007, the number of participants in the pilot declined abruptly. Anecdotal evidence explains the loss of motivation and interest in the pilot with difficulties related to the implementation of a joint self help initiative by several poor families, and – more importantly - by the fact that the poorest need the UMB cash support to meet immediate and intermitting consumption needs rather to invest in future income streams. The activation piloted with the UMB delivery on an annual basis has not been evaluated. This is why now it is not possible to understand why the beneficiaries preferred to invest the lump sum in subsistence farming, what UMB amount could have been adequate to acquire assets that would be sufficient to explore non-farm opportunities, what business climate improvements and technical assistance would have helped the pilot to succeed. In the future, the UMB could build on recent practice in the USA and Europe of a more active approach to encouraging “graduation� from social assistance. Apart from being extended only when there are children in the family, the UMB provision could be made conditional to active job search and / or workfare requirement for the unemployed adults in the family. The design of the workfare requirement of the UMB could follow well performing models in other countries. 62. Workfare reduces dependency and cost of benefit provision by tightening the access to benefits. It also provides a mechanism for attrition from the cash benefit program of those who are employed informally. These are important considerations in the crisis context when the pressure for budget constraints mounts. CONCLUSIONS There are old challenges and new opportunities for reforming the safety net in the Kyrgyz Republic within the existing benefit system but also going beyond the design and implementation of the current non-contributory benefits. Addressing them with adequate policies could bring savings which are very much needed in the context of the pressure for budget constraints coming from the emerging economic crisis. Dealing with them could contribute to improving the effectiveness of the safety net programs as instruments for immediate response to the crisis to protect the consumption of the poor and vulnerable, and could also protect and promote the human capital. In the safety net there is room for coping with an ‘old’ challenge – the reform of the in-kind benefits and subsidies which are a legacy from the Soviet past, are not targeted to the poor and do not contribute to an effective and efficient safety net. The process of monetizing these benefits has already started and the reform can go deeper with more radical approaches. The study proposes four approaches which do not exclude but could rather complement each other: (i) grandfathering / phasing out over time following natural attrition of the remaining beneficiaries combined with gate keeping to the granting of new entitlements and reconsidering the relevance of certain benefits vis-à-vis the needs of recipients; (ii) consolidation through provision of one lump sum / single cash benefit instead of several fragmented entitlements going to the same beneficiary category without undermining the overall amount of the cash received before the reform; (iii) integration / mainstreaming through including the in-kind benefit into 62 In 1996 US Welfare Reform introduced incentives and sanctions to encourage work; most of ECA countries apply workfare requirements in social assistance when the beneficiaries are able bodied adults; UK uses an “activation� model balancing customized contacts with individual work programs and active labor market programs; Chile’s Solidario program and customized contacts targeting the entire family. 99 the regular pension or MSB where applicable as a one-time increase, and stop existing as a separate benefit; (iv) scaling down through (a) abolishment of certain already irrelevant categorical benefits and (b) targeting of the remaining benefits through a means test of the existing beneficiaries and shifting those under the threshold to the UMB and a categorical test for disability and inclusion in the MSB program. In the safety net there exist instruments / programs which can be used for protecting the consumption of the poor and vulnerable in the context of the food, fuel and financial crisis, and the ascending economic crisis. With its two fairly well-targeted UMB program, the Kyrgyz Republic is well-positioned to channel resources and help protect the poor from the adverse impacts of the crises. In fact, the UMB has already offered an opportunity to respond to the food crisis, and the monthly benefits paid out under this program are being temporarily “topped-up� in value to help compensate for food price increases. These measures are temporary, and increased funding needs to be integrated into the budgetary process for longer-term sustainability of impact. Building on the UMB and MSB, there is scope for further consolidating the safety net, particularly through completion of the monetizing of privileges. Additional impact on the poor could be achieved through increased outreach (including through social passports and pro-active dissemination of information about the UMB program), measures to strengthen targeting (introduction of “proxy indicators� to use as “validators� or predictors of income to enhance means-testing), strengthening control over eligibility decisions by requiring social workers to document reasons for rejection, and establishment of a management information system (MIS, registry) to expand the database for decision-making and oversight and controls mechanisms. In the medium-term, additional impact on the poor could be achieved through linking cash benefits to social services. The payment of the UMB can be made conditional to positive behavior (like regular school attendance and intake of nutrition supplements) and thus mitigate the respective risks for the quality of human capital. Enhancing the promotional role of the safety net in terms of positive behaviors and investments in children and labor market related skills is of increasing importance during the crisis when families will be faced with increased risks and vulnerability. 100 LIST OF LITERATURE Barrientos, A. and M. Davies (2008). Assessment of State Benefits to Children and Families in the Kyrgyz Republic: Observations and a Way Forward, on behalf of UNICEF Biskek (mimeo). Carraro, L. (2000), Effectiveness of Social Assistance in the Kyrgyz Republic, National Statistics Committee of the Kyrgyz Republic (mimeo). CASE (2005), Public Expenditure Review on Social Sector in the Kyrygz Republic, CASE Kyrgyzstan, mimeo. CASE (2008), Kyrgyz Republic Social Safety Net Policy Note, Background report, CASE Kyrgyzstan, mimeo. European Commission (2007), Sector Policy Support Programme Kyrgyzstan 2007-2009. European Commission (2007), Technical Notes and Proposed Timeframe for the Implementation of the SPSP 2007-2009 – Policy Reform Matrix. FAO (2002) Cattle and Small Ruminant Systems in Sub-Saharan Africa, www.fao.org/docrep/005/Y4176E/y4176e04.htm, accessed February 2008. Foster, J., Greer, J. and E. Thorbecke (1984), A Class of Decomposable Poverty Measures, Econometrica, 52(3), pp. 761-765. Gassmann, F. (2004), Republic of Tajikistan: Poverty and Social Protection – Update, Background paper for the Tajikistan Poverty Assessment Update (World Bank, 2004). Gassmann, F., and G. Notten (2008), ‘Size Matters: Poverty reduction effects of means-tested and universal child benefits in Russia’, in European Journal of Social Policy, 18(3). Grosh, M., Coady, D. and J. Hoddinott (2004), Targeting of Transfers in Developing Countries; Review of Lessons and Experience. IBRD and The World Bank. Grosh, M., and E. Tesliuc (2005), Program Implementation Matters for Targeting Performance: Evidence and Lessons from the ECA Region. Presentation at Bucharest Seminar, June 6-7, 2005. Grosh, M. and E. Tesliuc (2007), The Design and Implementation of Targeting Systems, Presentation for the UNICEF Learning Programme of the Maastricht Graduate School of Governance. Grosh, M., C. del Ninno, E. Tesliuc and A. Ouerghi (2008), For Protection and Promotion. The Design and Implementation of Effective Safety Nets. The World Bank. Harwin, Judith (2007). Development of Child Protection in Kyrgyz Republic, Consultancy Report, UNICEF Ibragimova, Shamsiya (2007). Cash transfers to families and children in the Kyrgyz Republic - assessment of effectiveness (mimeo). Ibragimova, Shamsiya (2008). Evaluation of Efficiency of Benefit Payments to Families and Chikdren in the Kyrgyz Republic (mimeo). Kyrgyz Republic. Joint Country Support Strategy 2007 – 2010. Asian Development Bank, Swiss Cooperation, UK Department for International Development, the United Nations Group, the World Bank Group. January 2007. Vol. 1. Kyrgyz Republic: Recent Economic Developments. December 16, 2008. World Bank Country Office in the Kyrgyz Republic. LAW of the Kyrgyz Republic on Employment Promotion of June 27, 1998, last amended June 30, 2005, No. 119. LAW of the Kyrgyz Republic on State Social Benefits……. LAW of the Kyrgyz Republic on State Social Minimum Standards of August 7, 2006, No.149. 101 Lindert, K. (2008). Economic Shocks in ECA: The Social Side of the Global Crisis. November 2008, World Bank. Ministry of Labor and Social Development, Strategy of the Ministry of Labor and Social Protection (now Social development) of the Kyrgyz Republic for the period of 2006 – 2008. Maltsoglou I. and K. Taniguchi (2004), Poverty, Livestock and Household Typologies in Nepal, PPLPI Working Paper #13, Pro-Poor Livestock Policy Initiative. National Statistical Committee of the Kyrgyz Republic and UNICEF. Monitoring the situation of children and women. Multiple Indicator Cluster Survey. Kyrgyz Republic, 2006. Pokhrel, Subhash (2007). Technical Assistance to Ministry of Labor and Social Development in costing of social services for vulnerable children and their families. UNICEF, Kyrgyz Republic (mimeo). Staines, Verdon (2003). Social Assistance Programs and Poverty in the Kyrgyz republic. The World Bank, January 2003 (mimeo). Tabor, S. (2002). Assisting the Poor with Cash: Design and Implementation of Social Transfer Programs, Social Protection Discussion Paper Series, #0223. World Bank. Tesliuc, E. D. (2004), Mitigating Social Risks in Kyrgyz Republic, Social Protection Discussion Paper Series, # 0408. World Bank. Tesliuc E. D. and P. Leite. (2008). Assessment of Program Compliance. World Bank (mimeo). Tesliuc E. D. and P. Leite. (2008). Estimating Presumptive Income from Livestock. Background note on Livestock income, July 2008. World Bank (mimeo). Virtual Centre, Tropical Livestock Units (TLU), virtualcentre.org/en/dec/toolbox/mixed1/tlu.htm, accessed February 2008. World Bank (2007), Pension Policy Note – Policy Considerations and Practical Proposals (mimeo). World Bank (2007), Kyrgyz Republic Poverty Assessment. Volume 1: Growth, Employment and Poverty, Report # 40864-KG. The World Bank. World Bank (2007), Social Assistance in Central Europe and the Baltic States, draft, The World Bank. World Bank (2006), Dimensions of Urban Poverty in the Europe and Central Asia Region. Washington DC: World Bank World Bank (2002). Kyrgyz Republic: Enhancing Pro-poor Growth, August 2002. 102 ANNEXES Annex 1: Measuring poverty in the Kyrgyz Republic Annex 2: Categorical benefits and subsidies: types, beneficiaries, financing Annex 3: Social protection system assessments: data and methodologies Annex 4: Estimating presumptive income from livestock Annex 5: Methodology for a proxy-means test Annex 6: Tables simulated scenarios Annex 7: Comparing HMT targeting approaches for the UMB 103 ANNEX 1: MEASURING POVERTY IN THE KYRGYZ REPUBLIC Annex 1.1. Methodology This report follows the standard World Bank method of estimating a “basic needs� poverty line for consumption. All those who live in households where per capita consumption is below the poverty line are considered poor. The steps involved are as follows: • Compute a consumption aggregate based on food expenditures (including home produced food) and non-food expenditures. The consumption aggregate includes the computed user value of durables, but excludes housing rent. • Identify a reference group whose food consumption pattern can be used as a basis. The third, fourth and fifth consumption deciles were used. • Based on the consumption shares of this reference population, 2,100 calories per day is allocated across the most important food items. The resulting expenditure level is equivalent to the food poverty line that also delineates extreme poverty. • To determine the allowance for non-food consumption, the share of non-food consumption in total consumption of those individuals whose food consumption is just above the value of the food poverty line is used. The resulting per capita consumption poverty line is used to compute mainly three consumption poverty indicators: (i) the poverty headcount index (ii) the poverty gap, and (iii) the poverty severity, all of which belong to the class of Foster, Green and Thorbecke poverty measures: • The extreme/total poverty headcount index (HCI) describes the percentage of population that consumes below the food/total poverty line. Those whose consumption falls below the food poverty line are considered extremely poor. Those whose consumption falls between the food and total consumption poverty line are considered moderately poor. The HCI says nothing about how poor the poor are. • The poverty gap or poverty depth addresses this problem: the poverty gap is the average distance between the actual consumption of the poor and the poverty line, expressed as a proportion of the poverty line. It thus measure the average poverty short fall (the poor have zero poverty short fall) in the population. The poverty gap is not sensitive to the distribution among the poor, however. • The poverty severity index has a less straightforward interpretation than the other two poverty measures, but is constructed to give a higher weight to the poorer households. Thus, an increase in the poverty severity index indicates that income distribution has worsened among the poor. Source: World Bank; Foster, Green and Thorbecke (1984). 104 Annex 1. 2: Poverty tables Table A 1:1. Extreme poverty indices for individuals by region and topography Extreme Extreme Extreme Share in total poverty poverty poverty extremely Share in total headcount gap severity poor population by region: Urban 6.5 1.1 0.4 21.4 36.7 Rural 13.8 1.8 0.4 78.6 63.3 by topography: Highly mountainous 20.5 2.2 0.6 19.0 10.3 Moderately mountainous 8.5 1.3 0.2 9.2 12.1 Plain 10.3 1.5 0.3 71.8 77.6 Source: Staff calculations based on KIHBS 2005. Table A 1.2: Extreme poverty indices for individuals by oblast Extreme Extreme Share in total poverty Extreme poverty extremely Share in total headcount poverty gap severity poor population by oblast: Issyk-kul 14.2 2.1 0.4 10.6 8.3 Jalalabat 18.0 3.0 0.7 30.2 18.7 Naryn 20.0 2.7 0.7 9.4 5.2 Batken 19.0 2.5 0.6 13.9 8.1 Osh 10.2 1.3 0.3 23.1 25.3 Talas 14.1 2.1 0.6 5.3 4.2 Chui 5.3 0.3 0.1 6.9 14.6 Bishkek 0.4 0.0 0.0 0.6 15.5 Source: Staff calculations based on KIHBS 2005. Table A 1 3: Extreme poverty indices for individuals by demographic characteristics Extreme Extreme poverty Extreme poverty Share in total Share in total headcount poverty gap severity extremely poor population by sex:* Male 10.8 1.6 0.3 47.5 48.2 Female 11.4 1.5 0.3 52.5 51.8 by age (three groups): Under 18 years 14.8 2.0 0.4 46.9 38.9 18-59 years 9.0 1.3 0.3 47.0 53.4 60 years or older 7.8 1.2 0.3 6.1 7.8 by age (seven groups): Under 6 years 18.9 2.3 0.5 14.9 11.1 6-15 years 13.7 1.9 0.4 27.4 22.9 16-20 years 8.9 1.3 0.3 9.2 10.2 21-40 years 11.7 1.6 0.4 28.8 28.6 41-60 years 5.3 0.8 0.2 13.6 19.6 61-70 years 6.0 0.9 0.3 3.1 4.1 71 years or older 10.2 1.7 0.3 3.0 3.5 Total 11.1 1.5 0.4 100.0 100.0 * Not significant at the 10% level. Source: Staff calculations based on KIHBS 2005. 105 Table A 1.4: Extreme poverty indices for individuals by composition of household Extreme poverty Extreme Extreme poverty Share in total Share in total headcount poverty gap severity extremely poor population by household composition: 1 adult, no children 0.2 0.0 0.0 0.0 2.6 1 adult, 1-2 children 3.1 0.4 0.1 0.6 2.3 1 adult, 3 or more children 11.5 1.6 0.4 1.5 1.5 2 adults, no children 1.5 0.1 0.0 0.7 5.5 2 adults, 1-2 children 4.9 0.5 0.1 7.4 16.9 2 adults, 3 or more children 25.1 2.9 0.6 33.9 15.1 3 adults, no children 1.0 0.3 0.1 0.4 4.6 3 adults, 1-2 children 2.3 0.3 0.1 2.2 10.7 3 adults, 3 or more children 13.6 2.2 0.6 7.7 6.3 Other 14.6 2.3 0.5 45.4 34.7 by number of children (<16): None 1.5 0.2 0.1 2.8 20.2 One child 6.1 1.1 0.3 13.5 24.8 Two children 8.9 0.9 0.2 18.7 23.6 Three or more children 23.0 3.2 0.7 65.0 31.5 by household size: One member 0.2 0.0 0.0 0.0 2.6 Two members 1.5 0.2 0.0 0.9 6.8 Three members 1.5 0.3 0.1 2.0 14.1 Four members 5.1 0.6 0.1 9.3 20.3 Five members 7.5 0.7 0.2 13.5 20.0 Six members 15.6 2.2 0.5 24.3 17.3 Seven or more members 29.3 4.5 1.0 50.1 19.0 Source: Staff calculations based on IHBS 2005. 106 Table A 1 5: Poverty indices for individuals by composition of household Poverty Poverty Share in total Share in total headcount Poverty gap severity poor population by household composition: 1 adult, no children 2.5 0.4 0.1 0.2 2.6 1 adult, 1-2 children 21.5 4.6 1.3 1.1 2.3 1 adult, 3 or more children 63.5 17.8 5.9 2.2 1.5 2 adults, no children 6.5 1.1 0.4 0.8 5.5 2 adults, 1-2 children 26.6 5.8 1.7 10.5 16.9 2 adults, 3 or more children 65.3 17.6 6.5 22.9 15.1 3 adults, no children 16.1 2.3 0.6 1.7 4.6 3 adults, 1-2 children 26.9 5.4 1.5 6.6 10.7 3 adults, 3 or more children 63.2 14.8 5.0 9.2 6.3 Other 55.7 13.8 4.8 44.9 34.7 by number of children (<16): None 14.2 2.2 0.6 6.6 20.2 One child 28.5 7.1 2.4 16.4 24.8 Two children 47.8 10.7 3.3 26.1 23.6 Three or more children 69.7 18.2 6.7 50.8 31.5 by household size: One member 2.5 0.4 0.1 0.2 2.6 Two members 6.4 1.2 0.4 1.0 6.8 Three members 17.5 3.2 0.9 5.7 14.1 Four members 29.3 6.4 1.9 13.8 20.3 Five members 41.9 9.0 2.7 19.4 20.0 Six members 63.0 15.8 5.4 25.3 17.3 Seven or more members 78.3 21.6 8.1 34.6 19.0 Source: Staff calculations based on KIHBS 2005. Table A 1.6: Poverty indices for individuals by employment status of household head Poverty Poverty Poverty Share in total Share in total headcount gap severity poor population by economic status:* Employed 41.3 9.6 3.2 59.3 61.9 Unemployed 40.8 10.1 3.4 11.5 12.1 Pensioner (age) 49.6 12.3 4.3 23.4 20.3 Pensioner (disabled) 45.6 14.0 4.8 3.1 2.9 Other 43.1 14.7 6.1 2.8 2.8 * Not significant at the 10% level. Source: Staff calculations based on KIHBS 2005. Table A 1.7: Extreme poverty indices for individuals by employment status of household head Extreme Extreme Extreme Share in total poverty poverty poverty extremely Share in total headcount gap severity poor population by economic status:* Employed 9.8 1.4 0.3 54.5 61.9 Unemployed 9.6 1.2 0.3 10.4 12.1 Pensioner (age) 13.9 1.9 0.5 25.3 20.3 Pensioner (disabled) 14.8 1.5 0.3 3.9 2.9 Other 24.3 3.8 1.0 6.0 2.8 * Not significant at the 10% level. Source: Staff calculations based on KIHBS 2005. Table A 1.8: Poverty indices for individuals by housing characteristics 107 Poverty Poverty Poverty Share in total Share in total headcount gap severity poor population by material of house's walls: Bricks 31.4 8.6 3.2 17.3 23.8 Concrete 16.9 3.4 1.1 4.1 10.3 Airbricks 45.5 11.0 3.8 43.9 41.5 Earth, clay 62.5 14.5 4.7 29.3 20.2 Other 57.0 14.0 4.4 5.4 4.1 by material of house's roof: Roofing slates 46.9 11.5 3.9 88.6 81.4 Concrete 14.9 3.0 1.0 3.9 11.2 Other 43.5 10.5 3.7 7.5 7.4 by main water source: Running water 20.5 4.5 1.4 13.6 28.5 Private pump 42.3 10.0 3.5 23.7 24.2 Public pump 55.1 15.5 5.6 36.3 28.4 River, lake 61.6 11.0 3.2 20.5 14.3 Other 54.7 17.0 6.4 5.9 4.6 Source: Staff calculations based on KIHBS 2005. Table A 1.9: Extreme poverty indices for individuals by housing characteristics Extreme Extreme poverty Extreme poverty Share in total Share in total headcount poverty gap severity extremely poor population by material of house's walls: Bricks 11.6 1.6 0.4 24.7 23.8 Concrete 2.3 3.5 0.1 2.1 10.3 Airbricks 11.0 1.8 0.4 40.9 41.5 Earth, clay 15.4 1.8 0.3 28.0 20.2 Other 11.7 0.9 0.2 4.3 4.1 by material of house's roof: Roofing slates 11.7 1.7 0.4 85.2 81.4 Concrete 2.8 0.5 0.1 2.9 11.2 Other 17.9 1.6 0.4 12.0 7.4 by main water source: Running water 4.0 0.5 0.2 10.2 28.5 Private pump 9.6 1.7 0.4 20.7 24.2 Public pump 17.9 2.6 0.6 45.6 28.4 River, lake 11.3 0.9 0.2 14.5 14.3 Other 21.5 2.7 0.6 9.0 4.6 Source: Staff calculations based on KIHBS 2005. 108 Table A 1.10: Poverty indices for individuals by housing amenities and consumer durables Poverty headcount Share in total poor Share in total population by heating system: Central heating 12.8 3.9 13.1 Individual heating 47.9 82.1 82.1 Neither 44.1 4.8 4.8 by hot water supply: Yes 6.9 1.7 10.5 No 47.4 98.3 89.5 by telephone: Yes 22.8 15.6 29.5 No 51.6 84.4 70.5 by electricity: Yes 43.0 99.7 99.9 No 99.8 0.4 0.2 by television: Yes 24.0 18.2 32.7 No 52.4 81.8 67.3 by washing machine: Yes 29.5 31.5 45.9 No 54.6 68.5 54.1 by car: Yes 26.8 8.1 13.0 No 45.5 91.9 87.0 by fridge: Yes 27.8 34.5 53.6 No 60.8 65.5 46.4 Source: Staff calculations based on KIHBS 2005. 109 ANNEX 2: CATEGORICAL BENEFITS AND SUBSIDIES: TYPES, BENEFICIARIES, FINANCING Annex 2.1: Categories of in-kind benefits and subsidies and number of recipients as of January 1, 2007 and January 1, 2009 Number of beneficiaries Categories of beneficiaries 2007 2009 1 Disabled veterans of the Great Patriotic War (WW2) 1809 1329 2 Veterans of the Great Patriotic War (WW2) 4282 3183 3 Disabled Soviet Army veterans including 1086 1155 disabled Afghan War veterans including 193 176 4 Workers from reserved occupations 14422 11766 disabled workers from reserved occupations 2120 1860 5 Heroes of the Kyrgyz Republic of the Soviet Union 3 3 6 Leningrad blockade survivors 56 49 7 Former minor prisoners of concentration camps 78 66 8 Participants in the Hungarian events 28 23 9 Veterans of international combat operations 6290 6225 10 Rehabilitated participants of the “national labor service army� (civilian duty) 36 30 11 Surviving members of families of the deceased military personnel, including 864 731 Soldiers from WW2 393 248 12 Surviving spouses of WW2 veterans with disability 4604 3886 13 Surviving spouses of WW2 veterans with later disability 2583 2264 14 Surviving spouses of Leningrad survivors 3 4 15 Chernobyl NPP accident survivors 86-87 352 320 16 Chernobyl NPP accident survivors 88-89 83 90 17 Disabled survivors of the Chernobyl NPP accident 1041 892 18 Families with lost breadwinner in Chernobyl NPP accident 195 189 19 Beneficiaries of merit pensions 2272 2301 20 Persons decorated with State awards of the Kyrgyz Republic 22 19 21 Retired military personnel 1423 1485 22 Retired law enforcement personnel 5206 4973 23 Families of law enforcement personnel that perished in duty 159 178 24 Personnel of the criminal justice system 864 846 25 Retired personnel of the criminal justice system 198 502 26 Households living in mountainous areas (not receiving privileges under other categories) 1539349 154658 27 Single pensioner with pension of KGS 825 or less (owner-occupied house) 1223 1010 28 Single pensioner with pension of KGS 1,525 or less (owner-occupied house) 3337 3292 29 Families with unemployed pensioner and disabled member with income less than KGS 2,350 440 448 30 Unemployed disabled pensioners 26303 29604 31 Families with disabled children < 18 years 12736 13882 32 Unemployed recipients of survivor benefits 10634 9241 33 Unemployed pensioner with pension below the base pension of 450 KGS 372 253 34 Rehabilitated citizens 3101 3044 35 Hearing impaired persons 3142 3225 36 Visually impaired persons 9190 10258 37 Heroine mothers (<3 children) 15748 15179 38 Honorary donors 413 628 TOTAL amount of beneficiaries of subsidies and benefits 290446 284571 Financed from the republican budget 274202 269302 Financed from the local budget 16244 15269 Source: MLSD administrative data. 110 Annex 2.2: Republican budget allocations on certain categorical in-kind benefits and subsidies No Expenditure categories 2001 2002 2003 2004 2005 - 2006 2007 Actual Actual Actual Actual Actual Actual Actual Republican budget "Categorical in- kind subsidies and benefits" including: 305029.0 352918.7 385476.3 543511.5 595417.3 503092.1 592425.8 One-time allowance to Victory Day 1 Anniversary 38139.7 130000.0 21729.0 Annual one-time allowance to 2 Victory Day 12000.0 39062.5 9400.0 6843.2 6618.5 6100.0 4350.0 3 Fuel subsidies 368.5 438.6 193.9 110.5 148.0 98.0 100.5 4 Subscription privilege 783.1 3164.8 843.1 883.4 676.7 577.2 5 Sanatorium vouchers 24828.4 31615.8 32981.7 31360.3 29454.0 24587.4 26069.4 Compensation for unused 6 sanatorium vouchers 4331.0 17940.3 25277.3 79302.7 86640.2 109065.6 119647.7 Communication services and 7 telephone line installation 4375.6 2860.4 1660.7 3586.5 5733.3 5512.6 3878.8 Travel by all transport means, 8 including 48337.1 42929.5 35960.8 47593.8 51617.6 40891.9 70408.4 Monetization of travel by auto transport For travel Natural gas, liquid gas and meter 9 installation 36931.8 28914.2 25854.2 25670.8 23591.0 24613.3 38867.5 10 Electricity 70110.1 80013.5 77401.5 84734.1 84170.7 137248.8 157916.0 11 Utilities and thermal power 23701.1 30213.1 29740.9 18320.2 18202.6 22435.5 22949.9 Cash compensation for utilities (monetization) 23196.6 One-time compensation for 12 rehabilitated people 33.8 103.7 61.2 51.0 168.6 166.2 Non-refundable portion of loan 13 issued to Chernobyl disaster fighters 137.8 237.5 10.1 88.0 84.2 133.5 One-time compensation for due to 14 Chernobyl move out 15 Fuel (for coal purchase) 46811.5 21666.7 16894.9 2949.5 76.6 16 Cash compensation for coal 37666.6 38737.6 40029.7 38509.9 26662.9 17 Horse and light carrier purchase 6368.5 4977.3 2367.8 3058.6 3858.9 11319.7 9818.6 Health services and medicines 18 including: 8461.0 6909.3 15004.1 13532.6 14502.0 21466.5 23038.0 compensation for medicines 10086.4 11795.8 11511.3 16547.1 19347.5 deaf aids, optical glasses, dental work 4919.4 3690.5 19 Wheel-chair 1000.0 1900.0 4352.0 1508.7 2374.2 1618.4 3000.0 20 Dental work 12248.0 15696.3 15739.8 15457.8 21999.9 26124.9 30049.1 Reference indicators Number of population (thousand) 4921.0 4994.0 5000.0 5085.0 5140.0 5190.0 5240.0 Number of beneficiaries (thousand) 427.1 427.0 384.0 346.0 311.0 295.0 280.0 As % of the total population 8.7 8.6 7.7 6.8 6.1 5.7 5.3 Average amount received per person per year (KGS) 714.2 826.5 1003.8 1310.7 1914.5 1705.4 2115.8 Average amount received per person per month (KGS) 59.5 68.9 83.7 109.2 159.5 142.1 176.3 Source: MLSD administrative data. 111 ANNEX 3: ASSESSING THE PEFORMANCE OF THE SOCIAL PROTECTION SYSTEM. DATA AND METHODOLOGY Data The analyses in the present and the following chapter are based on data from the 2005 Kyrgyz Integrated Household Budget Survey (KIHBS) collected by the National Statistics Committee (NSC) of the Kyrgyz Republic. The KIHBS is implemented quarterly covering a sample of 5,000 households. The survey consists of several questionnaires and diaries collecting information on the household composition, education and health, migration, employment, housing, land and livestock possession and household income and expenditures. Households are visited four times per year. The collected data are representative at the national and oblast level. The 2005 KIHBS provides full survey information for 4,771 households and 19,136 individuals. The household-level descriptive statistics of the recipients of social transfers is as follows: Program Mean benefit per Program share No. of participation recipient HH Coef. of in total SP Type of benefit cases (%) (KGS/year) variation spending (%) Total social transfers 2397 45.5 9175 0.73 100.0 Pensions 1872 34.0 10799 0.61 87.9 Scholarships 125 2.9 1488 1.05 1.0 Monthly social benefits 315 5.8 2534 0.32 3.5 Unified monthly benefit 486 10.3 2355 0.43 5.8 Other social insurance benefits 193 4.5 1631 0.28 1.7 Source: Staff calculations based on KIHBS 2005. 112 Methodology The analysis covers all social and private transfers for which information is available in the KIHBS. First, we consider the distribution of benefits and beneficiaries across welfare quintiles and other socio- economic household characteristics, and secondly, assess the impact of the various transfers on poverty reduction. The first step is to define a welfare indicator that allows ranking the population from poorest to richest. We use per capita household consumption as calculated by the NSC. The advantage of using this measure is comparability with other studies, such as the latest poverty assessment (World Bank, 2007). This might seem inconsistent taking into account that the assessment of eligibility for means-tested social assistance benefits, the Unified Monthly Benefit (UMB), is based on family income. It raises the question of the appropriate indicator against which to assess program performance. If the aim is to evaluate the program outcomes, i.e. whether the benefits achieve the objectives of supporting the poorest households and reduce poverty, a consumption-based indicator is more appropriate as it better reflects the living standards of households over time. If the objective is to assess whether the program is properly implemented, household income is the preferred indicator. This assumes, however, that the income data collected with the KIHBS represents the same income sources taken into account by social protection officers when assessing benefit eligibility. A few issues deserve attention with respect to the latter: benefit eligibility depends on average family income. The KIHBS uses the household as the main unit. It is not possible to distinguish between different families living together at one household. Administrative income differs from income reported in the survey. Income from livestock is not taken into account and income from land is imputed using predetermined coefficients. Furthermore, income data stemming from household surveys are often not very reliable, especially in countries with a high prevalence of subsistence farming, a large agricultural sector and a sizeable informal economy. Respondents have difficulties assessing their net income and may be reluctant to reveal their true income to officials from the NSC. In order to establish the counterfactual consumption in the absence of transfers, assumptions have to be made. The strongest assumption and widely used in benefit incidence studies is that households do not compensate the lack of transfers with income from other sources. That means that households do not replace lost income through savings, a second job, money from relatives or other behavioral changes in the absence of transfers. This assumption is not very plausible. The opposite is to assume that households would fully compensate foregone transfers. Pre-transfer consumption is equal to after-transfer consumption. Again, this assumption is not plausible, especially for poorer households that may not have sufficient alternative income generating capacities. The issue here is to find a plausible replacement rate in the absence of social transfers. Using the data from the survey, it is possible to estimate the share of income that would be replaced.63 The marginal propensity to replace consumption is estimated at 25 percent based on the data for 2005. This means, for every 100 KGS not received in the form of social transfers, households would replace 25 KGS with income from other sources. This is half the value as estimated by Tesliuc (2004) based on the 2001 data. The choice of the substitution parameter has a considerable impact on the distribution of the benefits.64 Assuming 100 percent substitution, the richest quintile receives the largest share of social transfers (31 percent). This also holds for pensions, scholarships, privileges and private transfers. Assuming a marginal propensity to substitute foregone consumption of 25 percent, the distribution of transfers across the welfare distribution changes significantly. The poorest quintile captures almost 30 percent of all social transfers and pensions. The results are more robust to the choice of the parameter for substituting foregone transfers with other income when assessing the progressiveness of the transfers. The results change slightly for privileges and private transfers, which are regressive under the assumption of 100 percent substitution. Total social transfers and pensions become less progressive. 63 See Annex 3.1 for the description of the model. 64 The results for 0% and 100% marginal propensity to consume are presented in Annex 3.2. 113 Coverage, distribution of beneficiaries and benefits, and benefit adequacy are the concepts used to assess the distributional effects of social and private transfers. The second part of the analysis assesses the impact of social and private transfers on poverty reduction. Poverty indicators are estimated before and after transfers. The poverty lines used to analyze the poverty reduction impact of social and private transfers are those developed by the NSC. We use the absolute and extreme poverty line. The value of the extreme poverty line is equivalent to the costs needed to cover a minimum of 2,100 kcal per capita per day. The absolute poverty line includes an allowance for non-food goods and services deemed necessary to cover basic needs. Both poverty lines were calculated by the NSC in 2003 based on the new KIHBS which was first introduced then, and have been updated annually using the consumer price index. The constructed variables relevant for poverty analysis, such as aggregate per capita consumption and poverty lines are provided by the NSC as daily values. However, information on income from different sources is for the year as a whole. For the subsequent analysis we use annual values. Daily per capita consumption and the poverty lines have been adjusted to annual values by multiplying with 365. As a result, poverty rates may slightly differ from the ones provided in chapter 2. And most importantly, analyzing survey data provides only best estimates. Poverty rates (poverty headcount rate and poverty gap) are calculated for after-transfer per capita consumption and the counterfactual pre-transfer consumption. 65 The poverty headcount rate measures the percentage of the population living below the respective poverty line. Changes in the headcount rate indicate that the transfers managed to pull people over the poverty line. The poverty gap measures the average consumption shortfall of the poor as a percentage of the poverty line. It measures the depth of poverty. Changes in the poverty gap before and after a transfer allow conclusions on the effectiveness of the transfer to close the poverty gap. All results are calculated at the individual level (measuring ‘people’ not households). Sampling weights are used in order to render the results nationally representative, unless otherwise noted. Modeling the behavioral response of households to public transfers Following the methodology used by Tesliuc (2004), we estimate the marginal propensity to substitute consumption in the absence of social transfers. The model aims at estimating the proportion of social transfers households would replace with income from other sources in the absence of these transfers. It is based on the assumptions that households would adjust their behavior if they did not receive any social transfers. Household members may, for example, take a second job, engage in subsistence farming, migrate or ask relatives and friends living outside the household for assistance. Not all behavioral changes may be positive. Households may take children out of school, send them to work or engage in hazardous activities. The estimated model is similar to a typical consumption regression. Per capita consumption is the dependent variable to be explained by the model. In addition to variables for household characteristics, income from social transfers is included as an explanatory variable on the right hand side of the model: yi = α + βSTi + γXi + εi where yi is household consumption per capita (in KGS), STi income from social transfers (in KGS), Xi a vector of household characteristics, γ a vector of coefficients to be estimated, α a constant and εi the error term representing the variance not explained by the model. β is the coefficient we are looking for as it expresses the marginal propensity to consume out of social transfers. We estimate several models testing the sensitivity of the estimator. Model 1 and 3 (see Table A3.1) include a number of household characteristics as control variables, while model 2 and 4 regress per capita consumption on social transfers only ( yi = α + βSTi + εi ). Model 1 and 2 are estimated using standard 65 We use the standard Foster-Greer-Thorbecke family of poverty measures (Foster et al., 1984). 114 OLS, while model 3 and 4 account for fixed rayon effects. The estimated marginal propensities vary between 0.07 and 0.48, depending on the estimated model. The simple arithmetic average is 0.25, which we will take for the analysis. This is most probably still overestimated taking into account that in two of the four models the coefficient for Social Transfers is not significantly different from zero. It means that for every 100 KGS foregone in transfers, a household would substitute 25 KGS with income from other sources. Stated differently, a 100 KGS reduction in social transfers will reduce per capita consumption with 75 Som. Ideally, marginal propensities would be calculated for each type of transfer separately. Transfers like the UMB are not exogenously determined, as pensions or privileges. They depend on pre-transfer income. Propensities may also differ across socio-economic groups. The estimate used for the analysis is by no means precise. Using a different counterfactual consumption measure has an impact on the distribution of the benefits, the share captured by the poor and the level of progressiveness as tables A2 to A7 show below. 115 Results from the modeling the behavioral response of households to public transfers Table A 3.1: Marginal propensity to consume out of social protection transfers. Dependent variable: annual per capita consumption OLS OLS with rayon fixed effects Regression model (1) (2) (3) (4) Coef. Std. Err. P>t Coef. Std. Err. P>t Coef. Std. Err. P>t Coef. Std. Err. P>t Per capita SP transfers 0.11 0.09 0.21 0.48 0.07 0.00 0.07 0.08 0.39 0.36 0.06 0.00 Household size -2079.06 132.28 0.00 -1899.22 118.27 0.00 Age household head 67.48 21.61 0.00 68.19 20.56 0.00 Female household head -363.64 446.63 0.42 -349.49 373.90 0.35 Education of the head Secondary degree -3256.90 536.15 0.00 -2974.94 445.41 0.00 Primary degree -3398.06 714.33 0.00 -3591.60 681.55 0.00 Below primary -3709.95 994.59 0.00 -3537.78 843.03 0.00 Economic status of head Unemployed -148.00 674.12 0.83 -1208.19 595.26 0.04 Old age pensioner -2748.25 698.97 0.00 -2498.38 664.11 0.00 Disability pensioner -2012.68 685.24 0.00 -1375.21 764.71 0.07 Other 425.22 903.84 0.64 -94.96 950.92 0.92 Rural area -1851.86 446.97 0.00 -426.58 550.59 0.44 Using land -1282.77 561.34 0.02 467.68 553.59 0.40 Tropical livestock units 576.22 188.79 0.00 731.61 200.96 0.00 Constant 24462.87 1231.01 0.00 12167.56 259.17 0.00 20957.29 1114.10 0.00 12293.50 214.75 0.00 # of observations 4771 4771 4771 4771 R-squared 0.37 0.02 0.48 0.27 Note: Reference case omitted variables: male head, employed and with higher degree, urban area, no land use. Source: Staff calculation based on KIHBS 2005. 116 Table A 3.2. Benefit incidence of social protection benefits and private transfers, percentages, 2005: 100% substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Total social transfers 63.1 58.8 49.9 37.3 30.5 48.0 Pensions 40.3 39.1 34.9 30.5 24.7 33.9 Scholarships* 2.9 3.7 4.4 2.1 3.2 3.3 Monthly social benefits 12.4 5.0 6.2 4.7 3.6 6.4 Unified monthly benefit 27.5 20.5 19.4 4.6 0.9 14.6 Other social insurance benefits 7.3 4.9 8.8 3.3 1.5 5.2 Utility and housing subsidies 15.7 10.4 9.6 10.7 6.5 10.6 Money from relatives 38.2 27.7 25.5 38.1 41.0 34.1 Note: Quintiles based on annual per capita consumption after transfers, assuming a marginal propensity to increase consumption of 100%. * not significant at the 10% level. Source: Staff calculations based on KIHBS 2005. Table A 3.3. Benefit incidence of social protection benefits and private transfers, percentages, 2005: 0% percent substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Total social transfers 77.9 62.0 49.3 30.7 19.7 48.0 Pensions 60.6 43.3 28.4 22.7 13.9 33.9 Scholarships* 2.9 4.4 4.0 2.0 3.1 3.3 Monthly social benefits 12.8 5.1 6.6 4.6 2.6 6.4 Unified monthly benefit 31.2 22.9 14.1 4.1 0.9 14.6 Other social insurance benefits 8.0 9.4 4.2 3.0 1.3 5.2 Utility and housing subsidies 16.3 10.3 9.6 10.6 6.1 10.6 Money from relatives 54.3 28.9 20.8 34.7 31.5 34.1 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 0%.* not significant at the 10% level. Source: Staff calculations based on KIHBS 2005. Table A 3.4: Distribution of beneficiaries from social protection and private transfers, percentages, 2005: 100% substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Total social transfers 26.5 24.4 20.9 15.5 12.7 100 Pensions 23.9 22.9 20.7 17.9 14.5 100 Scholarships* 17.7 22.6 27.2 12.8 19.7 100 Monthly social benefits 39.1 15.5 19.6 14.6 11.2 100 Unified monthly benefit 37.9 27.9 26.7 6.3 1.2 100 Other social benefits 28.2 18.9 34.4 12.8 5.7 100 Utility and housing subsidies 29.9 19.5 18.3 20.1 12.3 100 Money from relatives 22.6 16.2 15.0 22.2 24.0 100 Note: Quintiles based on annual per capita consumption after transfers, assuming a marginal propensity to increase consumption of 100%.. * not significant at the 10% level Source: Staff calculations based on KIHBS 2005. 117 Table A 3.5: Distribution of beneficiaries from social protection and private transfers, percentages, 2005: 0% substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Total social transfers 32.6 25.9 20.6 12.7 8.2 100 Pensions 36.6 25.2 16.6 13.4 8.2 100 Scholarships* 17.7 26.8 24.4 12.1 19.0 100 Monthly social benefits 40.8 15.9 20.6 14.4 8.3 100 Unified monthly benefit 42.7 31.3 19.3 5.6 1.2 100 Other social benefits 31.4 36.0 16.0 11.4 5.2 100 Utility and housing subsidies 31.2 19.3 18.0 20.0 11.6 100 Money from relatives 32.0 16.9 12.2 20.6 18.3 100 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 0%. * not significant at the 10% level Source: Staff calculations based on KIHBS 2005. Table A 3.6: Distribution of social protection benefits and private transfers, 2005: 100% substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Distribution of social protection benefits and private transfers across groups Total social transfers 14.8 15.5 16.5 22.1 31.1 100 Pensions 12.7 14.8 15.5 23.4 33.6 100 Scholarships 12.9 10.1 31.9 11.9 33.3 100 Monthly social benefits 30.3 10.2 14.0 21.3 24.1 100 Unified monthly benefit 36.6 30.2 27.8 4.2 1.2 100 Other social benefits 14.1 12.8 26.8 27.0 19.2 100 Utility and housing subsidies 9.3 10.0 21.1 27.3 32.3 100 Money from relatives 9.2 8.7 11.9 26.6 43.7 100 Total consumption 9.4 13.1 16.6 22.4 38.6 100 Benefit adequacy (ratio of benefit/consumption) Total social transfers 13.3 (1.4) 9.9 (1.0) 8.2 (0.9) 8.4 (0.9) 7.0 (0.7) 9.4 (0.5) Pensions 10.1 (1.5) 8.3 (1.1) 6.7 (0.8) 7.8 (0.9) 6.6 (0.7) 7.9 (0.5) Scholarships 0.1 (0.1) 0.1 (0.0) 0.2 (0.1) 0.1 (0.0) 0.1 (0.0) 0.1 (0.0) Monthly social benefits 0.9 (0.3) 0.2 (0.1) 0.3 (0.1) 0.3 (0.1) 0.2 (0.1) 0.4 (0.1) Unified monthly benefit 2.0 (0.5) 1.1 (0.3) 0.8 (0.3) 0.1 (0.0) 0.0 (0.0) 0.8 (0.1) Other social benefits 0.2 (0.1) 0.2 (0.0 ) 0.3 (0.1) 0.2 (0.1) 0.1 (0.0) 0.2 (0.0) Utility and housing subsidies 0.2 (0.0) 0.1 (0.1) 0.2 (0.1) 0.2 (0.1) 0.2 (0.0) 0.2 (0.0) Money from relatives 5.7 (0.9) 3.8 (0.6) 4.1 (0.7) 6.7 (1.4) 6.2 (0.8) 5.3 (0.4) Total consumption 100 100 100 100 100 100 Note: Quintiles based on annual per capita consumption after transfers, assuming a marginal propensity to increase consumption of 100%. * not significant at the 10% level. Standard errors in parentheses. Source: Staff calculations based on KIHBS 2005. 118 Table A 3.7: Distribution of social protection benefits and private transfers, 2005: 0% substitution of consumption in the absence of transfers Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Distribution of social protection benefits and private transfers across groups Total social transfers 36.1 20.6 17.0 14.4 12.0 100 Pensions* 36.0 18.8 17.4 15.5 12.4 100 Scholarships 12.9 30.0 13.9 13.9 29.3 100 Monthly social benefits 34.3 10.6 26.2 13.6 15.4 100 Unified monthly benefit 46.4 34.4 14.6 3.4 1.2 100 Other social benefits* 18.0 31.2 12.8 22.8 15.2 100 Utility and housing subsidies* 12.6 8.7 24.1 26.8 27.8 100 Money from relatives 39.9 11.5 9.6 18.1 20.9 100 Total consumption 11.2 13.6 16.5 22.1 36.7 100 Benefit adequacy (ratio of benefit/consumption) Total social transfers 23.8 (1.5) 10.1 (0.8) 6.4 (0.7) 4.2 (0.5) 2.2 (0.3) 9.4 (0.5) Pensions 19.9 (1.7) 8.0 (0.8) 5.5 (0.7) 3.9 (0.5) 2.0 (0.3) 7.9 (0.5) Scholarships 0.1 (0.1) 0.2 (0.1) 0.1 (0.0) 0.0 (0.0) 0.1 (0.0) 0.1 (0.0) Monthly social benefits 1.0 (0.3) 0.2 (0.0) 0.4 (0.1) 0.2 (0.0) 0.1 (0.0) 0.4 (0.1) Unified monthly benefit 2.4 (0.5) 1.2 (0.3) 0.4 (0.1) 0.1 (0.0) 0.0 (0.0) 0.8 (0.1) Other social benefits 0.3 (0.1) 0.3 (0.1) 0.1 (0.0) 0.1 (0.1) 0.1 (0.0) 0.2 (0.0) Utility and housing subsidies 0.2 (0.1) 0.1 (0.0) 0.2 (0.1) 0.2 (0.1) 0.1 (0.0) 0.2 (0.0) Money from relatives 14.4 (1.8) 3.5 (0.5) 2.2 (0.4) 3.7 (0.6) 2.6 (0.3) 5.3 (0.4) Total consumption 100 100 100 100 100 100 Note: Quintiles based on annual per capita consumption before transfers, assuming a marginal propensity to increase consumption of 0%.* not significant at the 10% level. Standard errors in parentheses. Source: Staff calculations based on KIHBS 2005. Table A 3.8: Impact of social protection benefits and private transfers on poverty, 2005: 0% substitution of consumption in the absence of transfers Poverty rate before benefit Poverty rate Absolute Relative reduction Type of benefit (%) after benefit (%) reduction (%) Total social transfers 50.9 43.4 7.5 14.8 Pensions 49.6 43.4 6.2 12.5 Monthly social benefits 43.7 43.4 0.3 0.7 Unified monthly benefit 44.0 43.4 0.7 1.5 Other social benefits 43.5 43.4 0.2 0.4 Utility and housing subsidies 43.4 43.4 0.1 0.2 Money from relatives 47.5 43.4 4.2 8.8 Poverty gap before benefit Poverty gap Absolute Relative reduction (%) after benefit (%) reduction (%) Total social transfers 15.7 10.4 5.2 33.5 Pensions 14.7 10.4 4.3 29.2 Monthly social benefits 10.6 10.4 0.2 1.6 Unified monthly benefit 10.9 10.4 0.5 4.8 Other social benefits 10.5 10.4 0.1 0.9 Utility and housing subsidies 10.5 10.4 0.0 0.5 Money from relatives 13.1 10.4 2.7 20.7 119 Extreme poverty rate Extreme before benefit poverty rate Absolute Relative reduction (%) after benefit (%) reduction (%) Total social transfers 18.8 10.9 7.9 42.0 Pensions 17.8 10.9 6.9 38.8 Monthly social benefits 11.7 10.9 0.8 6.6 Unified monthly benefit 11.3 10.9 0.4 3.8 Other social benefits 11.2 10.9 0.3 2.4 Utility and housing subsidies 11.0 10.9 0.1 0.7 Money from relatives 14.7 10.9 3.8 26.0 Extreme poverty gap Extreme before benefit poverty gap Absolute Relative reduction (%) after benefit (%) reduction (%) Total social transfers 4.8 1.6 3.2 66.6 Pensions 4.2 1.6 2.6 61.5 Monthly social benefits 1.8 1.6 0.2 8.6 Unified monthly benefit 1.9 1.6 0.3 15.1 Other social benefits 1.6 1.6 0.0 1.7 Utility and housing subsidies 1.6 1.6 0.0 1.1 Money from relatives 3.4 1.6 1.8 53.0 Note: Poverty rates before transfers, assuming a marginal propensity to increase consumption of 0%. Source: Staff calculations based on KIHBS 2005. 120 Table A 3.9: Relative change in poverty after social transfers in EU countries, 2005 Poverty incidence rate before social transfers after social transfers Absolute change Relative change GRE 23 20 3 13.04 ESP 24 20 4 16.67 IT 24 19 5 20.83 POR 26 20 6 23.08 CYP 22 16 6 27.27 MAL 21 15 6 28.57 EURO 25 16 9 36.00 IRE 32 20 12 37.50 EU25 26 16 10 38.46 EU15 26 16 10 38.46 LUX 23 13 10 43.48 GER 24 13 11 45.83 BEL 28 15 13 46.43 ICE 20 10 10 50.00 NL 22 11 11 50.00 AUT 24 12 12 50.00 FRA 26 13 13 50.00 FIN 28 12 16 57.14 DMK 31 12 19 61.29 NOR 29 11 18 62.07 SWE 29 9 20 68.97 LIT 26 21 5 19.23 EST 24 18 6 25.00 LAT 26 19 7 26.92 POL 30 21 9 30.00 SLK 22 13 9 40.91 CZR 21 10 11 52.38 HUN 29 13 16 55.17 Source: before and after transfer poverty rates: EUROSTAT; change: own calculation. Note: income before transfer is including pensions. Income after transfers includes all other social transfers. 121 ANNEX 4: ESTIMATING PRESUMPTIVE INCOME FROM LIVESTOCK (TESLIUC & LEITE, 2008) To estimate the presumptive livestock income we develop a simple regression model relating the 2005 livestock income (as reported in KIHBS 2005) to the increase in livestock herd over that period, and a set of regional dummies to account for regional differences in the prices of output and inputs. We proceed as follows: • First, we estimate the revenues from the livestock business for each household. These correspond to the increase in the herd – for each type of livestock – over the year, times the price of an animal (here, we use regional median prices derived from the survey). The net increase in herd is calculated as entries (heads slaughtered plus heads sold over the year) minus exit (heads fallen and purchased over the year). • Second, we estimate the costs associated with the livestock production, by summing up the production expenses related to fodder and other expenses with livestock. • Third, we estimate the livestock income by netting out production costs from revenues. Theoretically and practically, this income is the sum of the profit and wages earned by the household from the livestock business. • Last, we regress livestock income, Li, on the increase in livestock herd by type of animal, Hij, and regional dummies (oblasts). The model is the following: J K Li = α + ∑ β j H ij + ∑ γ k ⋅ 1(oblast i = k ) + ε i j =1 k =1 where j=1,..,J are type of animals (cows, pigs,..) and k=1,..,K are oblasts of residence Table 4.1. Regression model for presumed livestock income Annual Livestock income coef. p-value Head Increase Cows 2,102.91 (-1.96)** Heifer and Calves (more than 1 year old) 2,907.09 (5.96)*** Bulls and oxen (more than 1 year old) 1,178.61 -0.99 Calves (less than 1 year old) 1,264.26 (2.02)** Pigs (including swines and piglets) 916.146 -1.05 Sheep (more than 1 year old) 1,456.17 (6.68)*** Lambs (less than 1 year old) 884.558 (3.11)*** Goats (more than 1 year old) 851.577 (4.91)*** Goat kids (less than 1 year old) 617.772 (4.79)*** Horses 5,445.60 (2.86)*** Old poultry 42.096 (2.27)** Young poulty 0.966 -0.09 Oblast Issykkul 1,463.06 (2.00)** Jalalabat 1,715.59 (2.58)** Naryn 1,363.13 (-1.96)** Batken 1,412.90 (2.11)** Osh 1,674.53 (2.49)** Talas 1,283.19 (-1.89)* Chui 529.281 -0.78 Bishkek 0 (.) Constant -1751.968 (2.72)** Observations 1117 122 R-squared 0.49 Note: Robust t statistics in parentheses; *significant at significant at 10%; ** significant at 5%; *** significant at 1%. Source: Staff calculation; KIHBS 2005. Once the information on the increase livestock herds is collected from a household, the predicted (presumptive) income is estimated as the sum of the increase/decrease in the number of animals times respective coefficient: J K Li = α + ∑ β j H ij + ∑ γˆ k ⋅ 1(oblast i = k ) . ˆ ˆ ˆ j =1 k =1 Table 4.2 presents information of 3 households from KIHBS 2005 as an example of the predicted annual livestock income of the households. Table 4.2. Example of predicted livestock income Dependent variable: Annual Livestock income Household-id from KIHBS 2005 Coefficient 40395 40711 40595 Head Increase Cows 2,102.91 0 0 0 Heifer and Calves (more than 1 year old) 2,907.09 0 0 1 Bulls and oxen (more than 1 year old) 1,178.61 0 0 0 Calves (less than 1 year old) 1,264.26 0 0 0 Pigs (including swine and piglets) 916.146 0 0 0 Sheep (more than 1 year old) 1,456.17 0 0 2 Lambs (less than 1 year old) 884.558 0 0 0 Goats (more than 1 year old) 851.577 0 2 -1 Goat kids (less than 1 year old) 617.772 0 0 0 Horses 5,445.60 0 0 0 Old poultry 42.096 -6 0 0 Young poultry 0.966 0 0 0 Oblast Issykkul 1,463.06 0 0 0 Jalalabat 1,715.59 0 0 0 Naryn 1,363.13 1 1 1 Batken 1,412.90 0 0 0 Osh 1,674.53 0 0 0 Talas 1,283.19 0 0 0 Chui 529.281 0 0 0 Bishkek 0 0 0 0 Constant -1751.968 PREDICTED ANNUAL LIVESTOCK INCOME -641.42 1,314.31 4,579.00 Actual Annual livestock income -360.00 1,858.89 7,622.22 Source: Staff calculations, KIHBS 2005. These coefficients may be used in a similar fashion as the land coefficients, to determine the livestock income using the presumptive method. When added to the UMB income, it will improve the administrative definition of income. The Pearson correlation of actual annual livestock income and predicted livestock income is 0.799. 123 ANNEX 5: METHODOLOGY FOR A PROXY-MEANS TEST FOR THE KYRGYZ REPUBLIC We use a stepwise OLS regression model in order to find the most powerful indicators that are highly correlated with poverty among a long list of indicators. The dependent variable is the natural logarithm of per capita consumption before UMB and MSB transfers (25 percent substitution assumed). We have excluded the MSB as well as it does not qualify for the assessment of the family income for the UMB according to the GoKG resolution on the methodology to calculate family income. Normalizing the absolute consumption values by taking the logarithm reduces the estimation bias caused by the skewed distribution of per capita consumption. The estimated model can be written as follows: Ln(yi) = α + βXi + εi where yi is pre-transfer household consumption per capita, α a constant, β a vector of coefficients, Xi a vector of household indicators, and εi the error term. We estimate the models separately for urban and rural areas accounting for the different living conditions. The selection criterion for inclusion of the indicators in the stepwise regression was set at p < 0.15. A stricter selection (p < 0.05) would further reduce the number of indicators, though only marginally. Table A 5.1 presents the estimation results. As a result of the stepwise procedure, indicators not significant at a level of 15 percent were excluded from the final model. The remaining model explains more than 50 percent of the variance among households in terms of household consumption per capita. Although the model has been estimated for urban and rural areas separately, a number of indicators are relevant in both locations. Household size and the number of children are negatively correlated with household consumption. The same applies for the age of the household head, although the size of the coefficient is relatively small. On the other hand, the presence of elderly people in the household has a positive effect on household consumption, all else being equal. The sex of the household head only matters in urban areas. Having a female head has a negative effect on the level of household consumption. If we are looking for potential filters that could be used to fine-tune the current means-test, indicators related to the dwelling and assets are indicative for the welfare position of a household. In urban areas, the availability of hot water, the presence of a phone, color TV, washing machine, refrigerator and the possession of a car indicate a higher welfare standard for a given household. In rural areas, with the exception of the washing machine, the same indicators are relevant. In addition, the possession of agricultural equipment may serve as an indicator for well-being. 124 Table A 5.1 Proxy-means indicators for household consumption per capita* Urban areas Rural areas Coef. Std. Err. P>t Coef. Std. Err. P>t Household demographics Household size -0.119 0.006 0.000 -0.110 0.006 0.000 Number of children <6 -0.100 0.012 0.000 -0.103 0.012 0.000 Number of children 6-15 -0.074 0.008 0.000 -0.023 0.008 0.006 Number of elderly >60 0.017 0.010 0.106 0.063 0.008 0.000 Dependency ratio 0.017 0.009 0.052 Characteristics of the head Age of the head -0.005 0.001 0.000 -0.002 0.001 0.000 Sex of the head -0.065 0.015 0.000 Head has low education** -0.083 0.028 0.003 Head is unemployed -0.052 0.021 0.014 -0.034 0.023 0.143 Dwelling Roof of dwelling not with slates 0.047 0.015 0.002 0.105 0.030 0.000 Has hot water 0.090 0.017 0.000 0.688 0.080 0.000 Has central heating -0.216 0.068 0.001 Assets & Durables Has phone 0.129 0.015 0.000 0.070 0.022 0.002 Has b&w TV -0.111 0.017 0.000 Has color TV 0.161 0.014 0.000 0.071 0.021 0.001 Has washing machine 0.062 0.015 0.000 Has refrigerator 0.156 0.017 0.000 0.080 0.018 0.000 Has car 0.240 0.021 0.000 0.193 0.023 0.000 Has motorbike 0.179 0.086 0.038 0.216 0.059 0.000 Has bicycle 0.053 0.026 0.039 Land in use (ha) 0.010 0.004 0.011 Livestock (LU) 0.023 0.003 0.000 Has agricultural equipment 0.113 0.022 0.000 Location Mountainous area -0.141 0.024 0.000 Constant 9.987 0.033 0.000 9.839 0.039 0.000 Observations 2915 1856 Adj. R-square 0.568 0.534 * Dependent variable: natural logarithm of household consumption per capita before UMB and MSB (25% substitution). ** Incomplete secondary education or less. Source: Staff calculations based on KIHBS 2005. The estimated model coefficients are the scores for the proxy-means test. For each household, a score is calculated multiplying the indicators with the coefficients and returning the logarithmic values back to absolute values. For example, in urban areas the calculation would be as follows: Scorei = e(a + b*Xi) Score for household i = exp [ -0.119* household size - 0.1 * children younger than 6 - 0.074* children between 6 and 15 + 0.017* elderly living in the households + 0.017* dependency ratio - 0.005* age of the head - 0.065 for a female head - 0.083 if head has low education - 0.052 if head is unemployed + 0.047 if roof is not with slates + 0.090 if dwelling has hot water + 0.129 if household has a phone + 0.161 if household has color TV + 0.062 if household has washing machine + 125 0.156 if household has refrigerator + 0.240 if household has a car + 0.179 if household has a motorbike + 0.053 if household has a bicycle] Any household with a score below a defined cut-off level will be eligible for a certain transfer or service. For the purpose of the policy simulations, we selected two cut-off points: the first selects the 10 percent households with the lowest score, and the second cut-off point selects the 20 percent lowest scoring households. In order to compare the predicted consumption levels with the original poverty status, we select the bottom 44 percent and bottom 12 percent of individuals based on the proxy (predicted) consumption level. Table A 5.2: Comparing actual and predicted poverty status Absolute poor by proxy* Absolute poverty Not poor Poor Not poor 80.34 19.66 100 Poor 24.79 75.21 100 Extreme poor by proxy* Extreme poverty Not Poor Poor Not poor 92.83 7.17 100 Poor 52.42 47.58 100 * Poverty status based on predicted consumption (see Annex for the model) Source: Staff calculations based on KIHBS 2005. As Table 5.2 indicates, 75 percent of the absolute poor based on actual per capita consumption are also identified as poor based on predicted consumption. The match is considerably lower when considering extreme poverty. Less than half of the extremely poor are identified likewise when using the predicted values. This may indicate that it is especially difficult to identify the extremely poor among the poor households. The model to estimate household proxy scores can be further improved, for example, by limiting the sample to the bottom 50 percent of households. 126 ANNEX 6: TABLES SIMULATIONS SCENARIOS Table A 6.1. Performance indicators simulated scenarios (coverage, inclusion, exclusion, leakage), percentages PMT 1COMB 1CAT 0-15 Recipient households 10 10 53.3 Population living in recipient households 17.5 17.5 65.9 Population of Q1 covered 49.7 49.7 83.2 Beneficiaries belonging to Q1 57.6 57.6 25.7 Share of benefits going to Q1 57.6 56.4 25 Inclusion error, individuals (20% poorest, b25umb) 42.4 42.4 74.3 Exclusion error, individuals (20% poorest, b25umb) 50.3 50.3 16.8 Leakage 42.4 43.6 75 Quintiles based on household consumption per capita before UMB, assuming 25 percent marginal propensity to consume; individuals. Source: Staff calculations based on KIHBS 2005 Table A 6.2. Poverty reduction impact different scenarios, percentages Scenario 1 (Cat 0-15) Scenario 2 (PMT all) Scenario 3 (PMT 0-15) Absolute poverty incidence 35.3 28.9 30.4 Absolute poverty gap 6.9 5.8 6.1 Extreme poverty incidence 3.9 4.0 4.0 Extreme poverty gap 0.5 0.5 0.5 Absolute reduction (percentage points) Absolute poverty incidence 8.5 14.9 13.4 Absolute poverty gap 3.9 5.1 4.8 Extreme poverty incidence 7.4 7.3 7.3 Extreme poverty gap 1.3 1.3 1.3 Relative reduction (as % of before rate) Absolute poverty incidence 19.4 34.0 30.7 Absolute poverty gap 36.0 46.8 44.0 Extreme poverty incidence 65.8 65.0 65.0 Extreme poverty gap 74.4 73.3 73.3 *Poverty rates before UMB, assuming 25 percent marginal propensity to consume; Source: Staff calculations based on KIHBS 2005 127 Table A 6.3. Average benefits and total program costs for the simulated scenarios CAT 0-15 (a) PMT 1 COMB 1 Average monthly transfer per eligible person KGS 48 93.3 176.4 Average monthly transfer per capita (receiving households) KGS 20.5 93.3 94.4 Total transfers, annual million KGS 1008 1008 1008 Source: Staff calculations based on KIHBS 2005; GDP 2007 from IMF (World Economic Outlook, 2007). Table A 6.4: Summary of main privileges, comparing actual and legal allocations, 2008 % of % of annual tota monetiz % of monetized Clients total costs l ed total (average) non-monetized* "000 per per per per Category "000 KGS KGS year month year month Privileges for war veterans 37,300 12.9 682,611 66.4 207,011 30.3 5,550 462 2550 213 Chernobyl participants 1,524 0.5 17,620 1.7 5,744 32.6 3,769 314 1558 130 Merit pensioners 13,397 4.7 73,527 7.2 56,740 77.2 4,235 353 251 21 Blind and deaf 12,855 4.5 46,278 4.5 46,278 100 3,600 300 0 0 Poor families 222,958 77.4 207,667 20.2 207,667 100 931 78 0 0 Total based on law (MLSD) 288,034 1,027,703 Actual based on MOF 274,200 628,100 Monetized (based on law) 523,979 Non monetized based on law 504,262 Estimated take up of non- monetized privileges 20.6 *assuming take-up of 20 percent Source: MLSD and MOF administrative data and own estimates 128 ANNEX 7: COMPARING HMT TARGETING APPROACHES FOR THE UMB The simulation includes three versions of HMT as follows: Table A 7.1: Proposed HMT scenarios Targeted population: all children under 16 living in poor households (poorest 20 percent HMT according to per capita administrative UMB household income). The UMB definition of Scenario 1 income includes current administrative UMB income and presumptive income from land Benefit: budget neutral benefits - Flat rate benefit of KGS per child per month Targeted population: all children under 16 living in poor households (poorest 20 percent HMT according to per capita administrative UMB household income plus presumptive livestock Scenario 2 income) Benefit: budget neutral benefits - Flat rate benefit of KGS per child per month Targeted population: all children under 16 living in poor households (poorest 20 percent according to per capita presumptive income. The presumptive income includes verifiable HMT income plus predicted non-verifiable income, presumptive income from land and livestock Scenario 3 included Benefit: budget neutral benefits - Flat rate benefit of KGS per child per month Source: Staff calculations based on KIHBS 2005 The main outcomes of the simulated scenarios are as follows: Table A 7.2: Basic statistics of proposed HMT scenarios - cost, number of beneficiaries and per child transfer Annual cost Monthly Number of Monthly Scenarios (million Number of average beneficiaries average KGS) beneficiaries benefit per (direct + benefit per (direct) child indirect) households HMT 1 1,008 489,600 172 978,700 86 HMT 2 1,008 481,918 174 975,660 86 HMT 3 1,008 472,697 178 952,259 88 Source: Staff calculations based on KIHBS 2005 The main performance characteristics of the simulated scenarios are as follows: Table A 7.3: Comparative performance of simulated HMT approaches to targeting the UMB, 20 % poorest households Absolute Absolute Under- Scenarios Coverage incidence of incidence of Generosity Leakage coverage beneficiaries benefits HMT 1 41.0 42.7 44.4 18.33 59.9 57.3 HMT 2 41.8 43.7 45.2 18.41 59.1 56.3 HMT 3 37.0 39.6 41.5 19.26 63.9 60.4 Source: Staff calculations based on KIHBS 2005 129