TO WHAT EXTENT DOES THE EXISTING SAFETY NET PROTECT THE POOR? Franziska Gassmann June 28, 2011 1 Table of Content INTRODUCTION .................................................................................................................................. 3  CURRENT SYSTEM OF NON-CONTRIBUTORY SOCIAL BENEFITS .......................................... 3  Unified Monthly Benefit ................................................................................................................. 4  Reforms per January 2010 and beyond ............................................................................................... 6  Monthly Social Benefit ................................................................................................................... 7  Reforms per January 2010 .................................................................................................................. 7  Categorical State Benefits ............................................................................................................... 8  Reforms per January 2010 ................................................................................................................ 10  TARGETING PERFORMANCE AND POVERTY REDUCTION IMPACT .................................... 11  Methodology ................................................................................................................................. 11  Coverage, distribution and benefit adequacy ................................................................................ 12  Poverty Reduction Impact............................................................................................................. 16  CHALLENGES AND THE ONGOING REFORM ............................................................................. 18  References ............................................................................................................................................. 22  List of Tables Table 1: Government spending on non-contributory social transfers, 2005-2010 .................................. 4  Table 2: UMB categories eligible for fixed amounts .............................................................................. 5  Table 3: MSB: before and after reform (changes introduced per January 2010) .................................... 8  Table 4: 38 Categories of categorical state benefit recipients................................................................. 9  Table 5: Examples of state categorical benefit package: most and least generous, KGS per month, 10/2009 ................................................................................................................................................. 10  Table 6: Categorical state benefits as per January 2010 according to President’s Decree ................... 11  Table 7: Benefit coverage of social protection benefits and private transfers, percentages, 2008 ....... 13  Table 8: Distribution of beneficiaries from social protection and private transfers across groups, %, 2008 ...................................................................................................................................................... 14  Table 9: Distribution of social protection benefits and private transfers across groups, %, 2008 ........ 14  Table 10: Benefit adequacy: Share of benefits in total household consumption, 2008 ........................ 16  Table 11: Poverty reduction impact of safety net reforms, percentage of individuals, 2008 ................ 19  Table 12: Benefit coverage and distribution after the reform, percentages, 2008 ................................ 19  Table 13: Summary table: all criteria included at the different levels of the GMI, number of observations .......................................................................................................................................... 21  List of Figures Figure 1: Development of GMI (GMCL) and extreme poverty line, 1998-2010 ................................... 5  Figure 2: Dynamics of safety net beneficiaries, 2000-2010 ................................................................... 6  Figure 3: Dynamics of safety net spending, 2000-2010 ........................................................................ 7  Figure 4: Categorical state benefits (l’goti): costs and beneficiaries .................................................... 10  Figure 5: Benefit coverage, 2005 and 2008 .......................................................................................... 13  Figure 6: Distribution of selected social transfers, 2005 and 2008 ....................................................... 15  Figure 7: Targeting accuracy of the UMB compared to targeted social assistance programs, selected ECA countries* ..................................................................................................................................... 15  Figure 8: Extreme Poverty Reduction (left: incidence; right: gap), 2008 ............................................. 17  Figure 9: Relative reduction of extreme poverty incidence rate, 2005 and 2008 ................................. 18  Figure 10: Relative reduction of extreme poverty gap, 2005 and 2008 ................................................ 18  2 INTRODUCTION Within the set of poverty reduction strategies, social protection plays an important role. An effective social safety net protects households and prevents them from destitution. It cushions the potentially impoverishing effect of covariate or idiosyncratic shocks. It prevents households from adopting strategies, such as selling their last assets, which may result in irreversible losses. The objective of social protection policies is to help families, individuals and communities with the prevention and mitigation of social and economic risks. Social protection policies are a range of public interventions to support the poor and vulnerable, and assist families, individuals and communities to better manage social and economic risks. Within the whole range of social protection policies, the social safety net focuses on the provision of support to poor and vulnerable households. It is distinct from social insurance policies targeted to workers in the formal labor market, as eligibility does not depend on the work history, but is related to the immediate needs of the poor and vulnerable population not covered by formal social protection measures. The objective of this paper is to assess the effectiveness of the Kyrgyz social safety net in protecting the poor using data from the 2008 KIHS. The analysis focuses on non-contributory social benefits targeted to poor and vulnerable households. The next section provides an overview of the current system of social benefits and recent reforms. Section three presents the results of the empirical analysis, focusing on coverage, distribution, adequacy and impact of the current non-contributory social transfers. Section four discusses the potential impact of the reform and further challenges. CURRENT SYSTEM OF NON-CONTRIBUTORY SOCIAL BENEFITS1 The Kyrgyz social benefit system includes contributory (social insurance) and non-contributory (social assistance) benefits. The social insurance component protects citizens against the risks of old- age, disability, loss of breadwinner and unemployment. It provides benefits upon the manifestation of the risks to those with a formal employment record, or contribution history. Just below six percent of GDP is annually allocated to the financing of the social insurance system (World Bank, 2009, p. 30). Non-contributory benefits are allocated independently of former contributions made and cover categorical state benefits, also called privileges (l’goti), monthly social benefits and a unified monthly benefit targeted to poor households with children. Total spending on non-contributory benefits nominally increased from KGS 1.3 billion in 2005 to KGS 2.2 billion in 2009, which represents about one percent of GDP. Over the same period, the number of beneficiaries decreased from 975,000 to 708,000. 1 This section draws heavily on chapter 3 of the Social Safety Net Note (World Bank, 2009). 3 Table 1: Government spending on non-contributory social transfers, 2005-2010 2005 2006 2007 2008 2009 2010 est GDP* bio KGS 141.9 188.0 196.4 229.1 Total spending on non-contributory benefits mio KGS 1,323 1,605 1,670 1,905 2,204 3,790 as percentage of GDP % 1.18 1.01 1.12 1.65 - UMB mio KGS 508 773 695 673 755 1,147 as percentage of GDP % 0.49 0.36 0.38 0.50 - MSB mio KGS 220 329 365 542 619 1,033 as percentage of GDP % 0.26 0.29 0.32 0.45 - Categorical state benefits (l’goti) mio KGS 595 503 609 689 829 1,609 as percentage of GDP % 0.43 0.37 0.42 0.70 Population** mio 5.14 5.20 5.25 5.31 5.37 5.44 Total number of beneficiaries '000 975 972 837 789 708 487 as percentage of the population % 19.2 18.9 16.1 15.0 13.3 9.1 - UMB '000 482 481 475 434 362 396 as percentage of the population % 9.4 9.3 9.0 8.2 6.7 7.3 - MSB '000 54 57 59 59 61 65 as percentage of the population % 1.1 1.1 1.1 1.1 1.1 1.2 - Categorical state benefits (l’goti) '000 439 434 303 296 285 26 as percentage of the population % 8.5 8.3 5.8 5.6 5.3 0.5 average amount of UMB per month KGS 89 124 125 120 135 205 average amount of MSB per month KGS 367 456 461 656 715 1,295 Source: MLSP 2006, 2007, ASP 2009, World Bank (GDP), IMF (Population) * GDP 2009: preliminary; ** Population: estimates for 2007-2010. Unified Monthly Benefit The Unified Monthly Benefit (UMB) is a means-tested and categorical non-contributory benefit targeted to poor households with children. The UMB is a variable benefit and covers the gap between the Guaranteed Minimum Income (formerly called Guaranteed Minimum Consumption Level) and the average per capita family income for eligible beneficiaries. Until 2009, the following household members were eligible for the UMB2:  Children under the age of 16 (and pupils under the age of 18 still attending general educational institutions);  Pupils of primary vocational schools and students of secondary and higher vocational institutions up to the age of 21;  Persons with disabilities (confirmed by medical commission); and  People of pension age without other pension entitlements. Total household income includes net income of all household members from all sources, cash as well as in kind.3 Not included is income from livestock, unemployment benefits, MSB, or single transfers such as funeral allowances or child birth grants. 2 Law on State Benefits. 3 Income included, among others: from employment, bonuses, commercial activities, leases, income from assets and deposits, imputed income from land (based on productivity coefficients), pensions, private transfers, scholarships and inheritance. 4 The Guaranteed Minimum Income (GMI) is a social standard established by the GoKG in 1998 and adjusted regularly. The level of the GMI depends on available budgetary resources and the predicted number of beneficiaries.4 Upon its establishment in 1998, the value of the GMI (GMCL) was set at KGS 100, which at that time was about half the value of the extreme poverty line, i.e. the costs of a food basket comprising 2,100 kcal per capita per day. Over the years, the gap between the GMI and the extreme poverty line widened further, contrary to the objective of the GoKG to gradually converge towards the Minimum Consumption Budget (MCB)5. In 2008, it was KGS 200, representing only 20 percent of the extreme poverty line (figure 1). Figure 1: Development of GMI (GMCL) and extreme poverty line, 1998-2010 1200 1000 800 KGS per month 600 400 200 0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Guaranteed minimum consumption level (GMCL) som a month Extreme poverty line. som a month Source: MLSP/ASP Eligibility for the UMB is determined in two steps. First, only households whose average per capita income is below the set standard are eligible for the program. Secondly, only household members falling into one of the eligible categories are entitled to the benefit. In addition to the variable UMB, fixed benefits are granted to poor families upon submission of the appropriate birth certificate: Table 2: UMB categories eligible for fixed amounts Category of UMB beneficiary Fixed UMB (- 2009) Fixed UMB (2010 -) Benefit at child birth 300% of GMCL 300% GMI Benefit to children under 1.5 years 100% of GMCL Benefit to children under 3 years 200% GMI Benefit to twins aged 0-3 years 100% of GMCL per child 100% GMI Benefit to triplets and higher multiple births 150% of GMCL per child 150% GMI aged 0-16 years 4 Until 2009, both expected UMB and MSB beneficiaries determined the level of the GMI. Starting from 2010, the GMI only applies to the UMB. 5 The MCB basked contains food, non-food, services, taxes and other payments. In 2008, its value was KGS 3,570 per month. 5 The UMB is entirely financed from the republican budget. Spending has been rather volatile in the past, neither related to the number of beneficiaries nor to economic growth or inflation (figure 2 and 3). The available budget, and subsequently the level of the GMI (and average UMB), is based on fiscal considerations and lacks a regular adjustment mechanism. As a result, the average value of the UMB is very small, undermining its potential to mitigate hardship and protect the poorest families (table 1). In 2009, the average UMB was KGS 135 per month (USD 2.956). Since 2007, the number of UMB beneficiaries has been decreasing substantially from 475,000 to 362,000 beneficiaries in 2009. While increasing incomes (prior to the onset of the financial and economic crisis in 2008) may partly explain the decline, the main reason for the reduction of the number of beneficiaries is the revision of the land coefficients used to impute income from land. The majority of UMB beneficiaries are children (87 percent in 2007). Most of the beneficiaries reside in the southern oblasts of Jalalabad, Osh and Batken. Reforms per January 2010 and beyond The GoKG adopted a new Law on State Benefits in the Kyrgyz Republic in December 2009. The revised UMB will be targeted to children only. The few other categories eligible under the previous law have been removed. The formula for the calculation of the GMI has been slightly modified, but essentially follows the same principle of budgetary dependence. In response to the energy tariff increase in January 2010, the GoKG raised the GMI from KGS 240 to KGS 282 per month.7 In order to improve the targeting efficiency of the UMB, the GoKG is planning further reforms of the UMB. It is currently testing the introduction of additional filters such as possession of livestock and of durable assets. At the end of 2009, it was foreseen to exclude households, which possess two of more cows or 25 or more sheep (goats), from UMB eligibility. This filter accounts for income from livestock which currently is not included in the calculation of family income. Secondly, by excluding households possessing more than three durable assets, the GoKG aims at reducing the inclusion error even further8. Additional plans envisage a larger role for ail okmotus in eligibility assessments, especially with respect to the valuation and use of land. Figure 2: Dynamics of safety net beneficiaries, 2000-2010 6 00 5 00 499 493 482 474 47 5 481 475 458 460 443 439 434 4 34 43 2 424 4 00 3 96 s e i 362 r a i c i f 303 3e00 2 96 n 285 e b 0 0 0 ' 2 00 1 00 57 59 59 61 65 45 48 51 54 37 40 26 0 20 00 2001 200 2 2003 2004 2005 2006 2 007 2008 20 09 2 010 est UM B recipients (thds) MSB rec ipient s (t hds ) Source: MLSP/ASP. 6 Exchange rate: 1 USD = 45.7 KGS (May 2010) 7 Note that the energy price reform was reversed after the events in 2010. 8 Accountable durable assets: car, truck, agricultural machinery, color TV, fridge, washing machine, mobile phone. 6 Figure 3: Dynamics of safety net spending, 2000-2010 1800 1600 1609 1400 1200 1147 1000 S 1033 G K o i 829 m800 7 73 755 668 695 689 673 600 59 5 609 619 568 542 507 50 8 5 03 454 400 386 365 353 3 29 282 305 265 200 206 220 22 0 106 144 53 0 2000 2001 20 02 2 003 2004 2005 2006 2007 2008 2009 2010 est Total f unds allocat ed for UMB per year. mil lion som tot al funds all ocated f or M SBper year, mio som Total f unds allocat ed for pri vi leges per year, mio som Source: MLSP/ASP. Monthly Social Benefit The Monthly Social Benefit (MSB) is a categorical benefit targeted to vulnerable groups with limited income generating opportunities. It is an income maintenance program providing a ‘social pension’ to people with disabilities, orphans, and elderly without pension rights. The MSB is assigned irrespective of the income of the beneficiary. Until 2009, the value of the MSB was based on the GMI. It varied between 75 and 300 percent of the GMI depending on the category. In 2003, a top-up of 20 percent was added to the original values. More top-ups have been added since. Similar to the UMB, the MSB is entirely financed from the republican budget. Due to its link to the GMI and the various top-ups, the average value of the MSB has been increasing from KGS 367 per month in 2005 to KGS 715 in 2009 (USD 15.65) (table 1), but it still remains below the value of the extreme poverty line. The number of MSB recipients has been gradually increasing from 37,000 in 2000 to 64,000 beneficiaries in 2009. According to the Agency of Social Protection, the main reason for this increase is the growing number of children with disabilities from birth. Deteriorating health care of pregnant women, deterioration of nutrition and lack of vital nutritive components in the diet of pregnant and lactating women and young children, poor living conditions and limited access to health care are potential reasons for the growing incidence of disability among children (CASE, 2008). 54 percent of all MSB beneficiaries are children. While the number of MSB recipients increased 1.64 times, spending increased in the same period with a factor 11.8 (table 1 and figure 3). Reforms per January 2010 The new Law on State Benefits introduced some major changes to the MSB program. First and foremost, the MSB is no longer tied to the GMI. The ‘Decree on determining the amounts of state benefits’ defined flat rate benefits for 15 categories of former MSB recipients. The new flat rates vary between KGS 1,000 and KGS 2,000 per month depending on the category. The average monthly MSB increased to KGS 1,295 (USD 28.3) per month. 7 Table 3: MSB: before and after reform (changes introduced per January 2010) Average monthly benefit (KGS) persons % of total old new Disabled children with CPIP 3,864 6.0 1164 2000 Disabled children 18,013 27.8 848 1500 Children with HIV or AIDS 120 0.2 948 2000 Children born from mothers with HIV/AIDS 108 0.2 948 2000 Disabled from childhood - I category 4,012 6.2 1164 2000 Disabled from childhood - II category 15,449 23.8 848 1000 Disabled from childhood - III category 5,189 8.0 532 1000 Disabled - I category 444 0.7 848 2000 Disabled II category 2,144 3.3 532 1500 Disabled III category 672 1.0 316 1000 Elderly citizens 1,602 2.5 402 1000 Elderly citizens of high-mountainous areas. 442 0.7 532 1000 Hero-mothers 167 0.3 848 2000 Children, in the event of breadwinner loss 12,237 18.9 532 1000 Orphans without both parents 332 0.5 848 2000 Total beneficiaries 64,795 100 Source: ASP. Categorical State Benefits Categorical state benefits (also called privileges – l’goti) are a legacy form the former Soviet era. Specific categories of privileged or vulnerable citizens are eligible for state subsidies and benefits. Until 2009, the system knew 38 different categories of beneficiaries and 14 different types of benefits and subsidies (tables 4 and 5). The value of the benefits is differentiated for different categories. Some groups of beneficiaries, e.g. disabled veterans of World War II, receive 100 percent of the entitlement, while others receive only a certain percentage. The largest group of beneficiaries, families living in mountainous regions are only entitled to a specific electricity subsidy. Eligibility is mainly categorical and independent of household income. Historically, all categorical benefits were granted in kind. Except for the utility subsidies, which were directly transferred to utility providers, take-up of the various entitlements was limited. Based on expenditure data, we estimate that actual take-up of benefits other than utility subsidies was in the range of 20 percent. Starting from 2003, a gradual monetization has taken place. In 2008, the majority of the benefits were monetized with the exception of glasses, hearing aids, trips within the CIS and sanatorium vouchers. The large majority of categorical benefits is financed from the republican budget (95 percent). Despite the shrinking number of beneficiaries, expenditures from the republican budget have been increasing substantially over the years, not the least as a result of the monetization of benefits (Figure 4). 8 Table 4: 38 Categories of categorical state benefit recipients # of beneficiaries as % of total Categories 2007 2008 2009 2009 1 Disabled from WW2 1809 1645 1329 0.47% 2 Participants of WW2 4282 3919 3183 1.12% Disabled Soviet Army, of which: 1086 1127 1155 0.41% 3 Disabled DRA (Afghanistan) 193 196 176 0.06% Home front workers of which: 14422 13700 11766 4.13% 4 Disabled 2120 2029 1860 0.65% 5 Military heroes of KR in the USSR 3 3 3 0.00% 6 Survivors of the Leningrad siege 56 55 49 0.02% 7 Under aged survivors of Concentration camps 78 76 66 0.02% 8 Participants of the Hungarian “events” 28 27 23 0.01% 9 Internationalists 6290 6336 6225 2.19% 10 Military workers later rehabilitated. 36 35 30 0.01% 11 Families of fallen soldiers of which 864 825 731 0.26% Soldiers in WW2 393 338 248 0.09% 12 Widows of disabled soldiers in WW2 4604 4443 3886 1.37% Widows of soldiers in WW2 with later 13 disabilities 2583 2543 2264 0.80% 14 Widows of Leningrad siege survivors 3 3 4 0.00% 15 Participants of clean up of Chernobyl 86-87 352 342 320 0.11% 16 Participants of clean up of Chernobyl 88-89 83 76 90 0.03% 17 Disabled due to Chernobyl 1041 988 892 0.31% 18 Families that lost breadwinner in Chernobyl 195 194 189 0.07% 19 Personal pensioner due to medals etc 2272 2285 2301 0.81% Recipient of orders/medals by the Kyrgyz 20 Republic ... 22 20 19 0.01% 21 Pensioners from the military 1423 1501 1485 0.52% 22 Pensioners from MOI 5206 5231 4973 1.75% 23 Families of MOI staff that died in duty 159 134 178 0.06% 24 Staff of criminal justice system 864 918 846 0.30% 25 Pensioners of criminal justice system 198 224 502 0.18% Families in mountainous regions not receiving 26 other privileges 159349 161288 154658 54.35% Alone living pensioners with less than 660 27 soms in pension 1223 1002 1010 0.35% Alone living pensioners with less than 1220 28 soms in pension 3337 3511 3292 1.16% Alone living pensioners with less than 29 1880soms in pension (gas) 440 463 448 0.16% Non working pensioners; of which 26303 27795 29604 10.40% 30 Alone living (gas) 3134 3071 Families with disabled children up to the age 31 of 18 years 12736 13916 13882 4.88% 32 Non working pensioner that lost breadwinner 10634 11558 9241 3.25% Non-working pensions with pensions less than 33 the base pension of 363 soms 372 354 253 0.09% 34 Rehabilitated and victims 3101 3084 3044 1.07% 35 Deaf people 3142 3241 3225 1.13% 36 Blind people 9190 9614 10258 3.60% 37 Heroine mothers (=>3 children) 15748 15652 15179 5.33% 38 Blood donors 413 561 628 0.22% TOTAL 290446 295879 284571 100% From the republican budget 274202 280151 269302 From the local budget (16+37+38) 16244 15728 15269 Source: ASP. 9 Table 5: Examples of state categorical benefit package: most and least generous, KGS per month, 10/2009 Lump sum Addition Free Free Free travel Compen cash Free Compensa Subscription al cash glasse Hearin Free travel Sanatoriu within CIS sation benef Denta travel tion for to benefit s g aid public withi Utilities m/resort once a year for it for l within transportat newspapers for 9th once once in transpor n the vouchers (round trip medicine 9th of work subur ion of May in 3 3 years t count ticket) s May b services occasion years ry occas ion Example 1: Disabled veterans of World War II 1040 867 3195 66.6 50 250 11.1 116.6 140 80 210 40 67 34.6 Example 2: Families living in mountainous regions 60 Source: ASP. Figure 4: Categorical state benefits (l’goti): costs and beneficiaries 900 900 800 800 thds of beneficiaries 700 700 600 600 mio KGS 500 500 400 400 300 300 200 200 100 100 0 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 Total number of beneficiaries ('000) Total costs (mio som) Source: ASP. Reforms per January 2010 The GoKG introduced major reforms to the system of categorical state benefits. The presidential decree ‘On providing monetary compensations to selected categories of citizens in connection with energy prices’ replaced the previous system of in-kind subsidies and benefits by flat cash benefits and reduced the eligible groups of beneficiaries. New benefit values vary between KGS 7,000 (USD 153) per month for war veterans and KGS 1,000 (USD 22) per month (table 6) and are all paid in cash. As a result of this reform, the number of beneficiaries is expected to drop with 240,000, leaving about 26,000 entitled beneficiaries. However, government expenditures are expected to double to KGS 1.6 billion per year as a result of the full monetization and the benefit increase due to the energy tariff reform. 10 Table 6: Categorical state benefits as per January 2010 according to President’s Decree Monthly compensation Category (KGS) 1 WW II disabled people 7000 2 WW II Veterans 7000 3 Disabled of Soviet Army and Afghanistan war 7000 4 Heroes of KR and Soviet Union 7000 5 Workers of rear services with group of disability 2000 6 Underage prisoners of concentration camps 7000 7 Leningrad siege survivors 7000 8 Participants of international conflicts 6000 9 Labor service people 2000 10 Families of fallen soldiers (FFS) 1000 Widows of WW II disabled and participants, siege participants 11 with disabilities 1000 12 Workers of rear services 1000 13 Persons suffered in Chernobyl NPP in 1986-1987 2500 14 Disabled of Chernobyl NPP 3000 15 Persons suffered in Chernobyl NPP in 1988-1989 1000 16 Families of dead disabled in Chernobyl NPP 1000 17 Children under 18 years of victims of Chernobyl NPP 1000 18 Rehabilitated citizens 1500 19 Hearing impaired 1000 20 Visually impaired 1000 21 Honorable blood donors 1000 Source: ASP. TARGETING PERFORMANCE AND POVERTY REDUCTION IMPACT This section assesses the performance of social protection benefits in terms of coverage, distribution and adequacy of the transfers. The analysis is based on data from the 2008 Kyrgyz Integrated Household Survey and covers all social and private transfers for which information is available in the KIHS. 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. While the GoKG has implemented a number of reforms in January 2010, this assessment considers the system as it was applicable in 2008. The implications of the recent reforms will be discussed in the next section. Methodology Average household consumption per capita as calculated by the National Statistics Committee (NSC) is the main welfare indicator. The population is ranked and grouped into quintiles (five groups of equal size) based on counterfactual household consumption per capita in the absence of social transfers (25 percent substitution rate).9 We then analyze the coverage, distribution and benefit adequacy for the different transfers. The second part of the analysis assesses the impact of social and 9 See Annex 1 for more details. 11 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. Poverty rates (poverty headcount rate and poverty gap) are calculated for after-transfer per capita consumption and the counterfactual pre-transfer consumption.10 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. 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. A few words of caution are warranted: the results of coverage and poverty impact in particular tend to be under-estimated when using household survey data. The main reason is incomplete respondent recall both on the receipt of a given transfer (coverage) and on the amount received (poverty impact, adequacy). In particular the results on the MSB and on housing and utility privileges have to be treated with caution as they are severely underreported in the current survey. Coverage, distribution and benefit adequacy In 2008, almost half of the population was receiving a social transfer, either contributory or non- contributory (table 7). This is slightly less than in 2005 (figure 5). The share of social benefit recipients is largest among the poorest 20 percent of the population, where almost two out of three individuals benefit from a transfer. Compared with 2005, coverage of the poorest with social transfers declined from 74 to 65 percent, while it increased from 23 to 29 in the richest quintile. Pensions account for the largest share in terms of coverage. More than one third of the population lives in a household receiving pensions (one in two of the poorest 20 percent). Other social insurance benefits are much less important with a very low coverage rate. The UMB as the main social safety net program explicitly targeted to the poor has an overall coverage rate of 10 percent (15 percent in 2005). Compared to 2005, coverage of the poorest decreased from 26 to 18 percent. Coverage of the MSB according to the KIHS is exceptionally low with two percent.11 Utility and housing subsidies reach almost one fifth of the population. Coverage rates are similar across the welfare distribution.12 For comparative purposes, the analysis in this section also includes informal transfers, money received from relatives.13 30 percent of the population benefit from informal transfers, a reduction with four percentage points compared to 2005. 10 We use the standard Foster-Greer-Thorbecke family of poverty measures (Foster et al., 1984). 11 This is not in line with administrative data reporting an increase of MSB beneficiaries. As mentioned earlier, MSB receipt seems to be severely underreported. Results have to be treated with caution. 12 Note that compared to 2005, coverage with privileges almost doubled. This result should be treated with caution due to measurement problems in the survey. 13 This also includes remittances received from abroad. 12 Table 7: Benefit coverage of social protection benefits and private transfers, percentages, 2008 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Any social transfer 65.2 56.7 39.1 41.3 28.8 46.3 Pensions 50.7 45.4 30.3 36.8 25.2 37.7 Scholarships* 1.9 1.7 0.9 1.1 2.8 1.7 Monthly Social Benefit 3.5 2.3 2.3 1.8 0.2 2.0 Unified Monthly Benefit 17.9 14.5 10.4 3.7 1.4 9.6 Other social insurance benefits* 2.4 3.5 1.4 1.3 0.6 1.8 Utility and housing subsidies* 21.7 16.8 20.2 18.5 15.1 18.5 Money from relatives* 28.3 34.5 30.9 26.1 29.6 29.9 Note: Quintiles are based on annual per capita consumption before transfers, assuming a marginal propensity of 25 percent. * Differences between groups are not significant at the 10 percent level (Chi-square test). Source: own calculation based on KIHS 2008. Figure 5: Benefit coverage, 2005 and 2008 60 50 48.0 46.3 40 37.7 percent of population 33.9 34.1 29.9 30 20 18.5 14.6 10.6 9.6 10 6.4 5.2 3.3 1.7 2.0 1.8 0 Total social Pensions Scholarships Monthly social Unified monthly Other social Utility and Money from transfers benefits benefit benefits housing relatives subsidies 2005 2008 Source: Own calculation based on KIHS 2005 and 2008. The distribution of beneficiaries along the welfare distribution (table 8) slightly worsened compared to 2005 to the benefit of richer households. In 2008, 29 percent of social transfer beneficiaries belonged to the poorest 20 percent of the population (a reduction of two percentage points compared to 2005), while 12.5 percent belong to the richest quintile (three percentage points increase compared to 2005). Considering different types of transfer separately, the largest change occurred among pension recipients. In 2008, 27 percent belonged to the poorest quintile compared to 33 percent in 2005. The shares are highest among the poorest 20 percent of the population for the MSB and the UMB. 38 percent of UMB and 35 percent of MSB recipients belong the poorest quintile. The higher on the welfare distribution, the lower the share of beneficiaries, especially with respect to the UMB, the only means-tested safety net program targeted to the poor. Compared to 2005, a clear shift in the distribution of beneficiaries took place with respect to informal money transfers. In 2008, the distribution across welfare quintiles is rather uniform, while in 2005, the largest share of beneficiaries (29 percent) belonged to the poorest quintile. 13 Table 8: Distribution of beneficiaries from social protection and private transfers across groups, %, 2008 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Any social transfer 28.7 24.0 16.9 17.9 12.5 100 Pensions 27.4 23.6 16.1 19.5 13.4 100 Scholarships* 22.8 20.2 10.3 12.9 33.7 100 Monthly Social Benefit 35.1 22.0 22.6 18.0 2.3 100 Unified Monthly Benefit 38.0 29.6 21.7 7.7 3.0 100 Other social insurance benefits* 26.8 36.9 15.3 14.2 6.8 100 Utility and housing subsidies* 23.9 17.9 21.8 20.1 16.4 100 Money from relatives* 19.3 22.7 20.7 17.5 19.8 100 Note: Quintiles are based on annual per capita consumption before transfers, assuming a marginal propensity of 25 percent. * Differences between groups are not significant at the 10 percent level (Chi-square test). Source: own calculation based on KIHS 2008. The distributional incidence of social transfers varies significantly by program (table 9). Scholarships, utility and housing subsidies and informal transfers are regressive, with richer households benefiting proportionally more than poorer households. 34 percent of utility and housing subsidies are allocated to the richest 20 percent households, while the poorest 40 percent receive only 24 percent of the total benefit value. Compared with the distribution of beneficiaries in table 8, one can conclude that richer households benefit from significantly higher privileges than the poor households. This is most probably due to the fact that (i) privileged citizens are not predominantly poor, and (ii) richer households consume more gas and electricity and benefit proportionally more from central heating. A similar pattern applies to informal transfers. While the share of the population receiving informal transfers is rather uniform across the welfare distribution, richer households receive larger informal transfers. Almost half of all informal transfers are going to the richest 20 percent. Table 9: Distribution of social protection benefits and private transfers across groups, %, 2008 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Total Any social transfer 26.9 21.9 14.8 17.3 19.1 100 Pensions 25.7 22.0 14.7 17.8 19.8 100 Scholarships 16.7 21.1 8.5 16.0 37.8 100 Monthly Social Benefit 52.6 5.0 17.1 19.7 5.7 100 Unified Monthly Benefit 51.9 22.7 19.7 4.0 1.7 100 Other social insurance benefits 39.0 24.9 7.8 8.2 20.2 100 Utility and housing subsidies 13.2 11.1 22.5 19.6 33.6 100 Money from relatives 7.2 14.0 13.4 19.3 46.1 100 Total consumption 9.8 14.0 17.6 22.7 35.9 100 Note: Quintiles are based on annual per capita consumption before transfers, assuming a marginal propensity of 25 percent. Source: own calculation based on KIHS 2008. In terms of targeting accuracy (distribution of benefits), the UMB and MSB manage to transfer the majority of the funds to the poorest households. More than half of total transfers are accrued by the poorest quintile. This constitutes a significant improvement compared with 2005, when only 38 percent (33 percent) of the UMB (MSB) was allocated to the poorest 20 percent (Figure 6). Compared to other countries in the region, the UMB performs very well in terms of targeting accuracy, similar to social assistance programs in Macedonia and Lithuania (Figure 7). 14 The redistributive impact of social transfers can be assessed by comparing the share of benefits captured by the poorest quintile with their share in total consumption and their size in the population (Tesliuc, 2004). Taken together, social transfers are highly progressive. The poorest quintile receives 27 percent of all transfers. This is more than their share in total consumption (ten percent) and as a group of the total population (20 percent). Scholarships and privileges are mildly progressive, while pensions, UMB, MSB and other social insurance benefits can be classified as highly progressive. Only informal transfers exhibit a regressive pattern. The poorest quintile captures only seven percent of total informal transfers, which is even less than their share in total consumption. Figure 6: Distribution of selected social transfers, 2005 and 2008 Distr ibution of transfers, selected benefits 100 % 80% 60% Quint ile V Quint ile IV 40% Quint ile I II 20% 0% 2005 20 08 2005 20 08 2005 20 08 2005 20 08 2005 20 08 2005 20 08 Total social Pensio ns Mo nth ly socia l Un ifie d mon thl y Utili ty and hou sing Mo ney from tran sfers bene fits be nefit subsidi es rela tive s Source: Own calculation based on KIHS08. Figure 7: Targeting accuracy of the UMB compared to targeted social assistance programs, selected ECA countries* *Data is the most recent available for the respective country and is subject to further update. Source: Lindert (2008). 15 Table 10 presents the accuracy of benefits, or formulated differently, the relative importance of social transfers as a share of average consumption in each quintile. The upper panel presents results for all households (including non-recipient households), and the lower panel shows the results for recipient households only. Relatively speaking, total social transfers are slightly more important for poorer households (relative progressivity). They constitute 10 percent of after transfer household consumption for those in the poorest quintile, compared to six percent for the richest households. However, the importance of social transfers as a share of total household consumption of the poorest group considerably declined compared to 2005, when social transfers accounted for 21 percent of the total. Pensions remain the most important social transfer. They account for 9 percent of total household consumption in the poorest households (17 percent in 2005). Considering only recipient households, pensions cover one fifth of total household consumption. This is four percentage points less than in 2005. The importance of pensions has mainly declined in wealthier households. While pensions accounted for one quarter of total household consumption in the top 40 percent in 2005, their share declined to 17 percent in 2008 among recipient households. Income from informal transfers remains the second most important transfer for all households. They contribute between 14 and 21 percent to total household consumption in receiving households. The UMB and MSB, both targeted to poor and vulnerable households, are progressive in relative terms, with benefits representing a higher share of total household consumption for those in the lower quintiles. Still, the magnitude of the transfers remains low due to their low coverage and inadequate benefit values. Especially the UMB further lost consumption power compared to 2005. Table 10: Benefit adequacy: Share of benefits in total household consumption, 2008 Type of benefit Quintile I Quintile II Quintile III Quintile IV Quintile V Benefit adequacy (ratio of benefits/consumption) for all households (including non- beneficiaries) Total social transfer 10.4% 7.9% 6.2% 6.4% 5.6% Pensions 8.6% 7.5% 5.8% 6.2% 5.5% Monthly Social Benefit 0.3% 0.0% 0.0% 0.0% 0.0% Unified Monthly Benefit 1.3% 0.4% 0.2% 0.0% 0.0% Other social insurance benefits 0.2% 0.1% 0.0% 0.0% 0.0% Utility and housing subsidies 0.2% 0.1% 0.2% 0.2% 0.2% Money from relatives 3.9% 4.2% 4.1% 4.4% 6.5% Benefit adequacy (ration of benefits/consumption) for beneficiary households (excluding non- beneficiaries) Total social transfer 18.2% 15.2% 14.9% 15.1% 14.6% Pensions 21.1% 17.9% 17.5% 16.7% 15.5% Monthly Social Benefit 8.0% 2.0% 2.9% 2.8% 2.5% Unified Monthly Benefit 6.7% 2.6% 2.6% 1.1% 0.6% Other social insurance benefits 0.2% 0.1% 0.0% 0.0% 0.0% Utility and housing subsidies 0.9% 0.8% 0.9% 1.2% 1.1% Money from relatives 13.8% 12.4% 13.2% 16.7% 21.2% Note: Quintiles are based on annual per capita consumption after transfers. Source: own calculation based on KIHS 2008. Poverty Reduction Impact In this section we compare extreme poverty rates (poverty incidence and poverty gap) before and after transfers. The poverty rates are calculated for after-transfer per capita consumption and the counterfactual pre-transfer consumption. The extreme poverty line is adjusted to represent the annual value. The change in the poverty incidence rate accounts for every individual that crossed the poverty line due to income from a social transfer. By targeting individuals close to the poverty line, a social 16 transfer can be very effective in reducing the poverty rate without changing the poverty gap significantly. In terms of the poverty gap, a social transfer is more effective if it manages to reach those households with the largest distance to the poverty line and close their poverty gap fully or to a certain extent. Strictly speaking, a social transfer can be very effective in closing the poverty gap without changing the poverty incidence rate. In the absence of any social transfer, extreme poverty rates would be considerably higher (figure 8, left panel). The extreme poverty rate is reduced with four percentage points when accounting for all social transfers. Social transfers reduce the extreme poverty gap from 2.3 percent to 1.2 percent (figure 8, right panel). The extreme poverty gap is almost halved. Pensions are the most important transfers in terms of poverty reduction, although poverty reduction is not their main objective. They are primarily meant to redistribute income of the life cycle. However, pensions provide significantly higher transfers than other social transfers, as they are often related to previous income. Pensions are the largest social transfer program in terms of allocated resources (government budget and social fund). Informal transfers are the second most important transfer in terms of poverty reduction. They reduce the extreme poverty rate by two percentage points. Non-contributory benefits are far less effective in reducing extreme poverty. Their impact on the poverty incidence and gap are limited. Nonetheless, the UMB remains the most effective targeted transfer. Its impact on poverty reduction could be further increased by increasing coverage and benefit levels. Note, that the GMI determining UMB eligibility is lower than the extreme poverty line. Figure 8: Extreme Poverty Reduction (left: incidence; right: gap), 2008 Any social transfer 12.0 Any soc ial transfer 2.5 10.0 2.0 Money from relatives 8.0 Pensions Money from relatives 1.5 Pensions 6.0 4.0 1.0 2.0 0.5 0.0 0.0 Utility and housing lity and housing Monthly Soc ial Bene Monthly Social Benef subsidies subsidies Other soc ial insurance Unified Monthly Benefit Other social insurance benefits Unified Monthly Benefit benefits B efo re transfer A fter transfer Before transfer After transfer Source: Own calculation based on KIHS08. Comparing the poverty reduction performance of 2008 with 2005, figures 9 and 10 present the relative poverty reduction of different transfers in terms of poverty incidence (figure 9) and poverty gap (figure 10). Overall, the effectiveness of reducing the extreme poverty rate increased slightly compared to 2005. This improvement is driven by the UMB, which achieved a reduction of extreme poverty of 13 percent in 2008. The MSB lost out compared to 2005, but this result has to be interpreted with caution considering the very low number of observations in the 2008 KIHS. In terms of poverty gap reduction, all social transfers are less effective compared to 2005. Taken together, one may conclude that social transfers became more effective in targeting those households close to the poverty line, but less effective in reaching the extremely poor. 17 Figure 9: Relative reduction of extreme poverty incidence rate, 2005 and 2008 60 50 40 37.5 36.6 % relative reduction 32.6 30.8 2005 30 2008 24.1 20 16.1 13.4 10 6.4 3.1 1.3 0.5 0.5 0.1 0.0 0 Any social transfer Pensions Monthly Social Unified Monthly Other social Utility and housing Money from Benefit Benefit insurance benefits subsidies relatives Source: Own calculation based on KIHS08. Figure 10: Relative reduction of extreme poverty gap, 2005 and 2008 60 54.3 50 47.8 44.9 40.1 39.2 40 % relative reduction 28.8 2005 30 2008 20 11.6 9.7 10 6.3 2.9 1.2 1.9 0.8 0.7 0 Any social transfer Pensions Monthly Social Unified Monthly Other social Utility and housing Money from Benefit Benefit insurance benefits subsidies relatives Source: Own calculation based on KIHS08. CHALLENGES AND THE ONGOING REFORM Following the analysis in the previous section, one can conclude that the poverty reduction impact of social safety net transfers is rather limited. Coverage with transfers targeted to the poor is extremely limited. Overall, only ten percent of the Kyrgyz population benefits from the UMB. The exclusion error among the poorest 20 percent is more than 80 percent. Nevertheless, the allocated monies reach 18 the poorest. Coverage with MSB is even more limited with two percent of the population living in beneficiary households. More widespread are categorical state benefits (privileges). Almost one in five individuals benefits from a utility or housing subsidy. But these subsidies are not very well targeted to the poor and benefit rich households as well. The social safety net reforms introduced by the GoKG in January 2010 primarily aimed at mitigating the impact of the energy tariff increase for the poor and vulnerable households. While the energy tariff reform was reversed mid-2010, the changes to safety net remained. The reforms described above (increase of UMB, flat MSB benefits and changes to the categorical state benefits) have an impact on the coverage and distribution of benefits and the poverty reduction impact. Using data from the KIHS 2008, we simulate the potential impact of the reforms as if they were introduced in 2008. Overall, the change in beneficiaries and benefit levels due to the reform has no significant impact on poverty (table 11). Table 11: Poverty reduction impact of safety net reforms, percentage of individuals, 2008 Poverty incidence Poverty gap Before reform After reform Before reform After reform Absolute poverty 31.5 (1.7) 28.8 (1.7) 7.6 (0.7) 7.0 (0.7) Extreme poverty 6.4 (1.1) 5.8 (1.1) 1.2 (0.4) 1.1 (0.4) Standard errors between parentheses. Source: own calculations based on KIHS08. The reforms of categorical state benefits have the largest impact on benefit coverage and distribution. Coverage decreases from 18 to 6 percent (table 12). 83 percent of beneficiaries will lose their entitlements for categorical state benefits. Coverage and distribution remain rather uniform across the welfare distribution. The reforms of the MSB have no impact on coverage and distribution of beneficiaries, as the groups remain unchanged.14 The reform of the UMB (increase of GMI and limitation of eligibility to children) has a minor impact on the performance indicators. Limiting eligibility to children only has no significant impact on coverage rates. Less than two percent of individuals would loose UMB entitlements. This is due to the fact that 95 percent of the poorest quintile lives in a household with children eligible for the UMB. Table 12: Benefit coverage and distribution after the reform, percentages, 2008 Quintile I Quintile II Quintile III Quintile IV Quintile V Total Coverage after reform UMB 17.3 14.4 10.4 3.5 1.4 9.4 Categorical state benefits* 8.4 4.8 5.6 6.0 5.0 6.0 Distribution of beneficiaries UMB 37.3 30.0 22.1 7.5 3.1 100 Categorical state benefits* 28.7 15.9 18.7 20.2 16.6 100 Distribution of benefits UMB 47.7 24.9 20.3 4.6 2.6 100 Categorical state benefits* 16.7 20.8 25.5 19.3 17.7 100 * not significant at 10% level. Source: Own calculation based on KIHS08. 14 The impact of the reform from variable to flat rate benefits cannot be assessed using KIHS data. It is impossible to distinguish the different groups. Therefore, an average increase of MSB benefits has been assumed, leaving the initial distributions unchanged. 19 It remains questionable to what extent the introduced changes to the safety net are financially sustainable in the medium term. Since public finances are already stretched, there is little room for the GoKG to increase government expenditures even further. Economic growth has slowed down considerably since 2008, and revenues are expected to be lower than projected. Thanks to a substantial grant from the Russian Government, the revenue shortfall was softened. However, total revenues are expected to decline in 2010, further limiting Government expenditures (table 13). Table 13. General Government Budget, as a percentage of GDP 2008 2009 2010 2011 2012 2013 2014 Act. Prel. Proj. General government finances (in percent of GDP) 1/ Total revenue and grants 29.8 33.0 30.1 28.6 28.1 28.5 28.8 of which: Tax revenue 23.0 22.7 22.0 22.1 22.2 22.9 22.7 Total expenditure (including net lending) 29.2 37.1 38.2 36.3 34.9 34.0 32.3 of which: Current expenditure 24.8 29.1 29.0 28.0 29.2 29.9 27.8 Capital expenditure 4.0 5.2 6.6 6.3 5.8 4.2 4.6 Overall fiscal balance 0.0 -3.2 -8.1 -7.7 -6.8 -5.5 -3.5 Primary balance 0.8 -2.3 -7.2 -6.9 -5.7 -4.2 -2.0 Primary balance excluding grants -1.0 -7.6 -9.9 -8.5 -6.8 -5.0 -3.3 Total public debt 48.4 52.2 60.5 67.5 71.2 72.7 68.3 Source: World Bank 1/ General government comprises state government, Social Fund and development fund (from 2009) finances While the increase of the GMI is a first steps towards a more generous UMB, the measure is not appropriate to reduce exclusion and inclusion errors. The European Union (EU) is supporting the Agency of Social Protection (ASP) with the reform of the UMB. During 2009, alternative targeting criteria have been tested based on administrative and survey data from seven pilot districts. The team simulated the impact of the following criteria, separately and combined (Delarue & Nikaj, 2009):  Reduction of inclusion error: o Exclusion of households with livestock above a specified minimum; o Exclusion of households possessing three out of five durable assets;  Reduction of exclusion error: o Exclusion of pension from income assessment for families with three or more children; o Higher GMI thresholds. The simulations show that the inclusion of livestock as additional criteria has only a minor effect on the inclusion error. In the pilot districts, the number of potential beneficiaries would decrease with less than three percent. This confirms the conclusions of the Social Safety Net Note (World Bank, 2009). Using the possession of durable goods as an additional filter would have a larger impact reducing the number of potential beneficiaries with 16 percent compared to the current situation. Excluding pensions from the calculation of family income would significantly increase the number of eligible households by 46 percent. Increasing the GMI threshold extends coverage as well. Table 13 summarizes the expected impact of the new criteria. The financial impact depends on the level of GMI. Only applying the new criteria without changing the GMI level (base case: KGS 200), would increase the costs with 10 percent. The new criteria combined with higher GMI levels increase costs significantly more. Under the most generous scenario with a GMI at KGS 485 per month, financial needs would increase more than five times compared to the base scenario (Delarue & Nikaj, 2009, p. 15). 20 Table 13: Summary table: all criteria included at the different levels of the GMI, number of observations GMI 200 GMI 280 GMI 485 Base +new criteria Update to +new criteria Update to +new criteria Scenario applied 280 applied 485 applied Number of eligible 447 582 555 688 772 881 households Number of potential 1239 1773 1532 2103 2090 2672 beneficiaries Source: Delarue & Nikaj (2009, p. 14). The reform effort of the GoKG with respect to the social safety net was highly welcome. The reform of categorical state benefits increases the transparency of public expenditures and allows beneficiaries to spend the cash benefits according to their own preferences. The increase of the GMI is a first step towards strengthening the poverty reduction impact of the UMB. However, major challenges are lying ahead. The complete monetization and increase of categorical state benefits requires double the amount of funding in 2010 compared to 2009, raising doubts about the sustainability of the reform. Categorical state benefits are the most expensive program, though with the least number of beneficiaries (table 1). Even more, these benefits are not targeted to the poorest households. The program is highly regressive. The UMB is essentially the only benefit purposefully targeted to the poorest households. While it performs very well in terms of targeting efficiency, the program is too small to have a meaningful impact on poverty. Coverage is limited and benefits are extremely small, especially when compared with the generous categorical state benefits. The GoKG needs to focus on the exclusion error and the low generosity of the program if the program is to have a meaningful impact as mitigation measure in times of crises and beyond. More funds should be allocated to the UMB, allowing the GMI getting closer to the extreme poverty line over time. More concretely, in order to improve the targeting efficiency of the social protection measures and the effective support of the poor and vulnerable households, we recommend completely abolishing the remaining categorical benefits and reallocating the budget to the UMB. This would substantially increase the funds available for the UMB. Increasing the GMI would also allow extending the coverage of the poorest with the UMB and protect the GMI from inflationary erosion in the future. 21 REFERENCES Delarue, J. and B. Nikaj (2009), Report on the Assessment of the Implementation of the Unified Monthly Benefit (UMB) Scheme of the Republic of Kyrgyzstan, mimeo. Tesliuc, E.D. (2004), Mitigating Social Risks in Kyrgyz Republic, Social Protection Discussion Paper Series, # 0408, World Bank, Washington DC. World Bank (2009), Social Safety Net in the Kyrgyz Republic. Capitalizing on Achievements and Addressing New Challenges, Report No. AAA-38-KG, The World Bank, Washington DC. 22 Annex 1: Modeling the behavioral response of households to public transfers 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. It assumes that households do not replace lost income through savings, a second job, money from relatives or other behavioral changes in the absence of transfers. The opposite is to assume that households would fully compensate foregone transfers. Pre-transfer consumption is equal to after-transfer consumption. Both assumptions are not plausible, especially for poorer households that may not have sufficient alternative income generating capacities to fully replace foregone benefits. Following the methodology used by Tesliuc (2004), we estimate the marginal propensity to substitute consumption in the absence of social transfers. Using the data from the survey, we estimate the share of income that would be replaced. 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 Som), STi income from social transfers (in Som), 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. 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 half the value as estimated by Tesliuc (2004) based on the 2001 data, but matches the findings from 2005 data (World Bank, 2009).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 Som foregone in transfers, a household would substitute 75 Som with income from other sources. Stated differently, a 100 Som reduction in social transfers will reduce per capita consumption with 25 Som. The results are sensitive to changes in the rate of substitution. In case of zero substitution, targeting accuracy of the UMB is further increasing. The opposite is the case if we assume 100% substitution of foregone transfers (see also World Bank, 2009). 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. 23