The effect of means-tested social transfers on labor supply: heads versus spouses

The effect of means-tested social transfers on labor supply: heads versus spouses An empirical analysis of welfare dependence in the Kyrgyz Republic Franziska Gassmann and Lorena Z. Trindade 1 DRAFT (v29/5/15) Abstract This paper empirically estimates potential labor disincentives of means-tested social transfers in the Kyrgyz Republic for adults with different household positions. Using data from the Kyrgyz Integrated Household Survey 2012, the analysis compares labor market outcomes for household heads, spouses and other adult household members. Quasi-experimental methods are applied to assess the effect of the Monthly Benefit for Poor Families with Children (MBPF) on labor supply. The analysis indicates that overall beneficiaries have on average higher labor market participation rates when compared to non-beneficiaries, but they are more exposed to seasonal effects. Results differ when analyzing different household members separately. Household heads in beneficiary households are less likely to be economically active than similar non-beneficiaries. Yet, spouses are more likely to be economically active. Key words: social transfers, labor disincentives, Kyrgyz Republic JEL Codes: 1 Both authors: UNU-MERIT/MGSOG, Maastricht University; franziska.gassmann@maastrichtuniversity.nl 1

1. Introduction According to economic theory, under competitive market conditions, income transfers reduce labor supply at the margin by raising the wage needed to attract these workers into the labor market. The empirical literature on the relationship between social assistance programs and work disincentives suggests that negative labor supply effects remain limited. Evidence from developing countries indicates that targeted social assistance programs lead to a re-allocation of labor within households resulting in an increase of adult labor participation (ILO, 2010). In most countries in Eastern Europe and Central Asia, which introduced or intensified social assistance programs in the early 1990s, as poverty increased during the economic transition, popular perceptions prevail that social assistance leads to welfare dependence. Still, the analysis of labor supply effects in the region is scarce and most of the studies do not find evidence of welfare dependence (see, e.g., Levin and Ersado, 2011). The Kyrgyz Republic was the first Central Asian country to adopt a national poverty benefit in 1995 (World Bank, 2000). The Monthly Benefit for Poor Families with Children (MBPF) is the only social assistance transfer in the Kyrgyz Republic specifically targeted at extremely poor households with children. It is a means-tested transfer whereby eligibility depends on average family income being below the Guaranteed Minimum Income (GMI). Due to the strict application of income criteria, poverty-targeted cash transfers may negatively affect the labor market participation of beneficiaries. The latter may occur as result of reduced job search intensity of transfer recipients or the willingness to accept work if the expected wage is only marginally higher than the social transfer (Guzi, 2013). Benefit receipt may negatively affect the labor market participation of other household members, or influence the choice between formal and informal sector jobs. This is particularly an issue where benefit eligibility is assessed based on formal income (Packard et al., 2012). If income tests focus on formal income, beneficiaries may be pushed into the informal sector in order to maintain their benefit eligibility (Tesliuc et al., 2014). Particularly those with low skills or working in low paid jobs have fewer incentives to graduate into fulltime (formal) work (Gotcheva & Sundaram, 2013). In situations with a high tax wedge on earned (formal) income, beneficiaries may opt for not working at all or staying in the informal sector to avoid withdrawal from social assistance (Koettl, 2013; Koettl & Weber, 2012). Labor reallocation may also occur among working-age adults in response to social transfers. Depending on the household position, adults may respond differently to the receipt of benefits. Therefore, intra-household decisions on labor supply may be reconsidered once a household is eligible for transfers. Studies on intra-household decisions on labor supply became popular after Gary Becker (1965) seminal work A Theory of the Allocation of Time, in which it is assumed that the household is a rational economic unit. According to Becker s approach, individual choices within the household and the relationships between these decisions and the production of wealth can be explained by the theoretical assumptions of neoclassical economics, such as maximizing behavior, market equilibrium and stable preferences. The general formulation of Becker, however, did not differentiate leisure activities from domestic activities (Heckamn, 2015). Gronau (1977) proposed an extension to the Becker model incorporating time as an input in the production of goods and services consumed in the household. This model became known as the Becker-Gronau model and was a fundamental contribution to detach domestic labor from leisure. In 2

this study, Gronau confirms several stylized facts about the distribution of labor supply within the household. It shows, for example, that changes in the socioeconomic context, such as income, education and number of children have different effects on the allocation of time among household members. The author shows that an increase in the wage rate of spouses may cause an increase in their labor supply and consequent reduction of time devoted to domestic activities and leisure. A change in the income of the spouses does not affect the labor supply of household heads, but if the change refers to the household head's wage rate, there is an increase in labor supply and the same reduction in the spouses labor supply. The author also concludes that the presence of children induces the mother to reallocate their time in the labor market to domestic activities. Regarding fathers, the main effect of the presence of children is the increased labor supply and time devoted to household chores, especially for parents with higher education levels and from non-paternalistic societies. The collective model for household consumption behavior, developed by Chiappori (1988, 1992) has lately become popular for analyzing intra-household decision making processes. Within this framework, household members interact with each other to make decisions through an exogenous unobservable process that produces a Pareto efficient allocation. Household preferences can be described by a weighted sum of individual preferences, in which the individual weights represent the bargaining power of household members. These preferences can be modified according to changes in prices, wages or household income from other sources. However, other factors can affect the intra-household allocation process, such as personal income from other sources. Chiappori (1992) developed the sharing rule function, which described how the income from other sources is distributed within the household and can be used to identify the individual preferences and intra-household allocation process. Several extensions of the collective model have been developed based on Chiappori s (1988, 1992) seminal work. These extensions contributed to the analysis of relevant aspects related to decisions on household labor supply. Chiappori et al. (2002) derived conditions to determine the sharing rule for household income from other sources between spouses. Chiappori (1997) and Apps and Rees (1997) formulated the collective model of labor supply and domestic production, in which home production was taken into account, implying that all unused time in the labor market is not interpreted as leisure. Empirical studies on intra-household allocation decisions are still limited. However, given that in paternalist societies intra-household allocation decisions often reflect gendered outcomes, empirical findings from studies on the impact of social transfers on male and female labor supply are relevant for the present analysis. Guzi (2013) finds evidence of a welfare trap created by the tax and social security system in the Czech Republic: individuals, and especially women, who receive relatively higher social benefits, have a higher probability to remain unemployed. In Tajikistan, social assistance transfers had a positive effect on adult employment rates in female-headed households, indicating the importance of the safety net for the transition from inactivity or informality into employment (Arias & Sanchez-Paramo, 2014). In the Kyrgyz Republic, paternalistic structures are common within the families and in wider society. Current market institutions and prevailing socio-cultural attitudes do not support the active involvement of women in economic activity. In addition, women s economic behavior is restrained by the lack of opportunities and the requirement to carry out their traditional family roles (World Bank, 2011). In this context, assuming that intra-household allocation decisions might be highly 3

correlated to gender, it is essential to investigate whether means-tested social transfer have distinct effect on labor supply of adults with different household positions. The Kyrgyz labor market is characterized by insufficient labor demand, a changing composition of employment by economic sector, high informality and a sizeable gender gap. Over the last decade, the labor force frequently grew more rapidly than the annual growth of employment. Over the same period, the structure of employment changed considerably. The percentage of the labor force employed in agriculture decreased from 49 percent in 2000 to 28 percent in 2011, while employment in other sectors, notably services, grew with ten percentage points over the same period reaching 45 percent in 2011. Approximately 70 percent of employment is informal, and almost two thirds of those working in the informal sector are self-employed. Under-employment is particularly an issue in rural areas. Although unemployment rates are lower in rural areas, the work intensity is less and seasonal employment is more frequent. The employment-to-population rate is lower for women than for men. While men s employment rate increased from 66.4 percent in 2003 to 70.7 percent in 2010, female employment levels remained unchanged. The same pattern holds for the younger generations (15-24 years old) that are entering the labor market: young men increased their share while young women significantly lost position. Furthermore, women account for higher shares of employment in activities and occupations involving lower wages (Schwegler-Rohmeis, 2013). The purpose of this paper is to estimate potential labor disincentives of the Monthly Benefit for Poor Families with Children (MBPF) for adults with different household positions in Kyrgyz Republic. Using data from the Kyrgyz Integrated Household Survey 2012, the analysis compares labor market outcomes for household heads, spouses and other adult household members. Quasiexperimental methods are applied to assess the effect of the Monthly Benefit for Poor Families (MBPF) on labor supply. Contrary to other studies, the analysis applies matching procedures separately to household heads and spouses in order to estimate average treatment effects on the two groups separately. The analysis indicates that overall beneficiaries have on average higher labor market participation rates when compared to non-beneficiaries, but they are more exposed to seasonal effects. Results differ when analyzing different household members separately. Household heads in beneficiary households are less likely to be economically active than similar nonbeneficiaries. Yet, spouses are more likely to be economically active. The remainder of this paper is structured as follows: the next section describes the MBPF in more detail and critically assesses potential work disincentives arising from the design of the MBPF. Section 3 describes the data and methodology used for the empirical analysis. Section 4 contains the results of the empirical analysis and Section 5 concludes. 2. MBPF and potential work disincentives The social protection system of the Kyrgyz Republic comprises both contributory (i.e. social insurance) and non-contributory (i.e social assistance) benefits which play an important role in helping individuals and families cope with income shocks. The MBPF is a means-tested, noncontributory benefit targeted to poor households with children. It is a variable benefit and covers the gap between the Guaranteed Minimum Income (GMI) and the average per capita family income for each child up to the age of 18 in eligible households. The income assessment includes both formal and informal income. In addition to the recipient household s formal and informal monetary 4

income, in-kind incomes (e.g., from agriculture), and the possession of family assets (e.g., durable goods, draft animals) are also taken into account to determine benefit eligibility (Hasanov & Izmailov, 2011). In 2012, the GMI was equal to KGS 370 per month 2, which reflected less than 30 percent of the extreme poverty line. 3 Despite its potential to alleviate extreme poverty and contribute to equalizing opportunities in the years of childhood, coverage of the extremely poor remains limited. The number of MBPF beneficiaries has been declining since 2005, when the system counted over 480 thousand beneficiaries, representing nine percent of the population. In 2011, the system registered 377,000 beneficiaries (World Bank, 2014). Based on estimates derived from the Kyrgyz Integrated Household Survey (KIHS), 7.4 percent of the population lived in a beneficiary household in 2012 (Table 1). The coverage is highest among the poorest households with 13.3 percent of the population of the first (poorest 10 percent) and 19.4 percent of the population of the second decile covered. On the other hand, there are a large number of equally poor households that do not receive the MBPF transfers. Table 1. MBPF coverage and distribution indicators across decile, 2012 in percent Indicators Deciles Total 1 2 3 4 5 6 7 8 9 10 Coverage 13.3 19.4 8.4 6.6 13.7 4.1 4.5 3.8 0.2 0.1 7.4 Distribution of beneficiaries 18.0 26.2 11.3 8.9 18.6 5.5 6.1 5.1 0.3 0.1 100.0 Distribution of benefits 22.7 33.8 10.0 5.9 15.2 4.3 4.6 3.0 0.4 0.1 100.0 MBPF as % of total consumption 1.4 1.8 0.4 0.3 0.6 0.2 0.2 0.1 0.0 0.0 0.4 MBPF as % of total consumption (only recipients) 10.9 10.4 5.2 5.0 3.5 3.4 2.9 3.3 5.4 1.9 6.2 Note : Deciles are based on annual per capita consumption before transfers with substitutions effects. Source : Estimates based on KIHS 2012. Although the allocation of the MBPF is progressive, (more than 50 percent of total transfers are reaching the poorest 20 percent), more than 80 percent of extremely poor children below the age of 18 are not covered. Furthermore, the benefit provided by the program is not sufficient to meet the most basic needs. In 2012, MBPF transfers accounted for about 10 percent of total household income in the poorest households. Therefore, the impact on poverty is limited due to its low coverage and benefit levels. Income from wage remains the most important income source for Kyrgyz households. This applies both to the total population and to the group of MBPF beneficiaries. Income from wage accounts for71.8 percent of total household income for MBPF beneficiary households, followed by pensions (14.1 percent). On average, the share of wage income is slightly higher for the poor and extremely poor, compared to the MBFP recipients (Table 2). This suggests that MBPF beneficiaries might behave differently on the labor market compared to other households that are equally poor. Table 2 also hints at the importance of informal transfers. In MBPF households, money from 2 The equivalent of 19.36 USD PPP in 2012 (currency equivalent in international from IMF WEO, April 2015). 3 The extreme (food) poverty line was KGS 1,286 per capita per month in 2012. The absolute poverty line was KGS 2,182 (NSC). The GMI is set by government decree depending on the available financial resources provided by the Ministry of Finance and expected number of beneficiaries. 5

relatives and work abroad accounts for 18.2 percent of total household income, which is almost twice the share compared to total population (11.7 percent). Considering the income composition of all individuals, income from remittances alone account for 9.5 percent of total income. However, if only the income composition of recipient households is accounted, the share of remittances represents 42.5 percent of total income, and among the group of MBPF beneficiaries receiving remittances, they account for 52.3 percent of total household income. Table 2. Household income composition, by poverty status and MBPF beneficiaries: Kyrgyz Republic, 2012 in percent Components of total income Total Poor Extremelypoor beneficiaries MBPF population Total Wage 67.7 66.4 68.5 57.1 Agriculture activities 12.8 12.4 8.2 20.6 Pension 14.8 16.2 18.6 11.1 Money from relatives and work abroad 1 11.7 12.9 22.8 18.2 MBPF 0.4 1.0 0.9 6.2 Other social benefits 0.2 0.2 0.3 0.4 Other income 0.9 1.0 1.4 0.7 Note: 1) Money from relatives and work abroad includes reliefs received by the household from relatives or friends and income from employment outside Kyrgyzstan. Source: Estimates based on KIHS 2012. To understand the potential for work disincentives, the GMI and average MBPF transfer has to be compared with local salaries and minimum wage regulations. The average monthly wage in the formal economy across sectors was KGS 10,884 in 2012 and the lowest average wage was reported for the agriculture, hunting and forestry sector (KGS 5,454) (NSC). The situation is different in the informal sector. Informal labor in agriculture pays considerably less. The average monthly wage for persons involved in planting or animal production is less than KGS 500, which is below the official minimum wage (KGS 760 per month in 2012) (NSC). Given the low level of the GMI, the amount of MBPF benefits a household is maximally entitled to is considerably less than the wage this household could earn in the formal and informal labor market. It is therefore unlikely that the MBPF a priori creates work disincentives, a conclusion which is also supported by Gotcheva and Sundaram for the Western Balkans (2013). Table 4 compares different hypothetical households with respect to their earning potential and the GMI, which reflects the amount a household can receive per child if it has no income at all. In households with one working adult, the MBPF only exceeds the minimum wage if this household has three or more children. In case of two working adults and assuming that both earn the minimum wage, income from work exceeds transfer income. Employment in the formal and informal sector always pays off, with the exception of informal work in the agricultural sector, where the earning potential is clearly below a full MBPF transfer. Hence, based on the comparison of these standard situations, there is a potential for work disincentives for adults working informally in agriculture. The design of the MBPF, whereby the transfer amount is equal to the gap between total family income per capita and the GMI, in principal imposes a 100 percent marginal tax rate on each additional Som of reported income exceeding the GMI. However, the actual marginal tax rate may be different. It entails that the means test is strictly applied with automatic recertification in the 6

event that the family income situation has changed. While this might be feasible for income from formal employment or social transfers, it is administratively much more challenging to assess changes in informal income if households do not inform the local authorities of such changes. The fact that each additional Som earned would reduce the MBPF transfer by one Som may induce beneficiaries not to report additional income and may deter them from accepting work if the expected wage only marginally exceeds the GMI. This potential poverty trap is further enhanced by several in-kind benefits MBPF and other poor households are entitled to. MBPF recipients are exempted from paying state charges and fees, are eligible for free legal support, and are exempted from paying for the registration of a birth or issuance of a passport. Financially disadvantaged citizens are exempted from co-payments for hospital care and children from poor families with four or more children are also eligible for free outpatient care. Local governments have the right to provide textbooks free of charge and exempt children from paying school fees. Finally, in Bishkek pre-school children from poor families are entitled for free school meals. 4 Table 3. Comparing potential incomes for model families in KGS Hypothetical HH1: 1 Working Hypothetical HH2: 2 Working Sources of income adult + adults + 1 Child 2 Children3 Children 1 Child 2 Children3 Children Full or maximum MBPF 370 740 1,110 370 740 1,110 Minimum wage 760 760 760 1,520 1,520 1,520 Average wage - formal sector 10,884 10,884 10,884 21,768 21,768 21,768 in agriculture, hunting and forestry activities 5,454 5,454 5,454 10,908 10,908 10,908 Average wage - informal sector 3,879 3,879 3,879 7,758 7,758 7,758 in planting production 230 230 230 460 460 460 Source : Own calculations based on KIHS 2012 and NSC. 3. Data and Methodology The 2012 Kyrgyz Integrated Household Survey (KIHS) is the study s main source of data. The KIHS is an annual household survey implemented by the National Statistics Committee of the Kyrgyz Republic, collecting data since 2003. It has a unique structure as every year up to one quarter of the sampled households are replaced;. However, households are not tracked over time if they are dissolved or move within the country. In addition, while household identification (ID) codes are unique and consistent over the years, the individual ID codes are not unique to individuals. There are concerns about the representativeness of the cross-sectional samples as well as the panel sub-samples (Esenaliev, Kroeger & Steiner, 2011). 4 Law of the Kyrgyz Republic on State Charges and Fees; Law of the Kyrgyz Republic on Legal Assistance Guaranteed by the State; Presidential Decrees on Issuance of Kyrgyz International and Internal Passport as per the 2004 Design and Provision of Exemption from State Charges to Certain Categories of Citizens; Law of the Kyrgyz Republic on Education; Provisions on Allowances for Catering in Pre-school Education Facilities approved by the Bishkek City Council Resolution #84, 30 June 2009; State Guarantee Program for Provision of Healthcare Aid to Citizens of the Kyrgyz Republic, Government Resolution #388, 28 June 2013. 7

The 2012 KIHS contains information on 5,000 households and is representative at the national and regional level. Households participating in the KIHS are visited four times per year. The survey provides detailed information on the demographic composition of the household, incomes, expenditures, housing, assets and labor. The labor module of the KIHS collects labor force data for present and migrant individuals 15 years and older at quarterly level. Due to seasonal events, it is likely that an individual s participation in the labor market varies throughout the year. Therefore, the information used for the labor market analysis is based on quarterly data, but only considers present individuals in the household. Although quarterly data is used for the analysis, the weights used for the estimations are annual weights. This does not compromise the analysis and might only lead to minor differences in estimations. Labor force participation refers to the proportion of the working age population that is economically active: all individuals who supply labor for the production of goods and services, or are ready to work. In assessing the impact of cash transfers on work disincentives, the analysis considers different labor market outcomes. Using the labor force module of the KIHS, the NSC identifies whether an individual is employed, unemployed or inactive. Economically active individuals (active labor market participation) includes both employed and unemployed individuals Although the NSC definition for employed follows the ILO convention, it does not consider individuals engaged in unpaid informal activities. The analysis therefore uses a second definition for employment, referred to as extended employment, which also includes individuals engaged in informal unpaid activities, such as working on the plot, in the forest, or the own production of goods. The analysis of the labor market outcomes of beneficiaries and non-beneficiaries will focus on the able-bodied working age population (aged 18-62 years old), excluding full time students 5, living in households with children under 18 years old, unless indicated otherwise. This group represents 39.8 percent of the total population in the Kyrgyz Republic. Household heads that are part of the reference group represent 19.9 percent of the country s total population, while spouses represent 14.1 percent. Within the reference group, household head is a position more common among men (66.8 percent), while spouse is more common among women (96 percent) (Table 4). The treatment group in the subsequent analysis comprises able-bodied adults aged 18-62 years old, not studying, living in a household with children under 18 years old and receiving the MBPF. Note that the term MBPF beneficiary used throughout the text refers to all individuals living in a MBPF recipient household, unless otherwise indicated. The analysis in this study empirically assesses the existence of adverse incentives for labor force participation for adults MBPF beneficiaries with different household positions. In order to understand whether certain characteristics are more likely to induce the presence or absence from the labor market, different econometric models will be applied. 5 Adults indicating to be day students when reporting their social status were excluded from the analysis given the importance of education for human capital development. 8

Table 4. Share of the reference group in total population and distribution of reference group by gender: Kyrgyz Republic, 2012 in percent Population Share of total population Adult aged 18 to 62 years old 55.8 Able-bodied adult aged 18 to 62 years old 54.7 Able-bodied adult aged 18 to 62 years old, not full time student 51.1 Able-bodie adult aged 18 to 62 years old, not full time student, living in household with children under 18 years old Household head, able-bodie adult aged 18 to 62 years old, not full time student, living in household with children under 18 years old Spouse, able-bodie adult aged 18 to 62 years old, not full time student, living in household with children under 18 years old As a first step, we apply simple binary models to test whether able-bodied adults living in MBPF beneficiary households are more or less likely to participate in the labor market. The analysis focuses on three different outcomes: (i) active labor market participation, (ii) extended employment (including unpaid work), and (iii) participation in informal work (including unpaid work). The results will indicate which individual or household characteristics are positively or negatively correlated with labor market outcomes and to what extent seasonal influences matter. For the latter purpose, the models are estimated for quarter 1 (winter) and quarter 3 (summer). In the binary (probit) model, the probability of an individual to be active or to be employed is assumed to be a function of: MBPF beneficiary status; gender; age; level of educational attainment; status for residing in the Southern part of the country, rural and mountainous areas. The models further include household size; number of children under six years old in the household; number of children between 6 and 18 years old in the household; and whether a household receives remittances from relatives or work abroad. The models are estimated separately for household heads and spouses. The model for informal work is estimated only for employed household heads and spouses in the sample, including those engaged in unpaid work (extended definition of employment). At last, the three models are estimated again, but this time only including the bottom 40 percent of the welfare distribution, assuming that the determinants of labor market participation may differ for low-income households. 39.8 19.9 14.1 Share of reference group Household head, able-bodie adult aged 18 to 62 years old, not full time student, living in household with children under 18 years old Male 66.8 Female 33.2 Spouse, able-bodie adult aged 18 to 62 years old, not full time student, living in household with children under 18 years old Male 4.0 Female 96.0 Source : Estimates based on KIHS 2012. 9

Given the limitations of the binary models to draw conclusions about the impact of MBPF receipt on labor force participation, a quasi-experimental design is used to establish an adequate counterfactual to isolate effects from program participation. According to Khandker, Koolwal & Samad (2009, p.8), several quasi-experimental approaches can be used to evaluate the effect of programs and estimate whether changes in well-being or behavior are results of the program intervention and not of other factors. Given the cross-sectional nature of the data, this study applies Propensity Score Matching (PSM), whereby individuals from the participant group are matched with individuals from the nonparticipant group who share similar characteristics. By comparing the labor market behavior of these two groups, we will be able to analyze whether or not the MBPF creates work disincentives. The Propensity Score Matching model allows the estimation of the Average Treatment Effect (ATT) of the treated (i.e. MBPF receipt) on labor market outcomes 6. The quasi-experimental analysis estimating the impact of MBPF receipt on labor market outcomes will be estimated separately for each quarter of 2012, as seasonal effects are likely to have an effect on labor market outcomes. In addition to the quarterly analysis, the model will also consider the hypothesis that labor market outcomes of MBPF beneficiaries differ for heads and spouses, given their position in the household. To test the hypothesis that the MBPF program might have different impacts according to the location of a household, the sample will be further divided into households living in the North and households living in the South of the country. The PSM model considers individual and household characteristics to estimate the impact of receiving MBPF on labor market outcomes. In order to establish control groups for heads and spouses, the model considers age, gender, relation to the household head and educational attainment 7. For household characteristics, the models includes household size, number of children under 6 years old in the household, number of children between 6 and 18 years old in the household, household living conditions 8, size of land 9, durable assets 10 ; and household location 11. Table 5 provides summary statistics for the variables included in the analysis. 6 The Average Treatment Effect (ATT) of the treated is estimated by Stata s program teffects psmatch. 7 Complete or incomplete higher degree and complete general secondary degree compared to primary or incomplete secondary degree and no education. 8 House walls not made from brick or concrete; house roof not made with roofing slates or concrete; house with no running water. 9 Land bigger than 500 sq meters. 10 Household has car and/or refrigerator 11 Rural and mountainous area. 10

Table 5. Summary statistics for variables of interest Variables * Household head Spouse Obs Mean Std. Dev. Min Max Obs Mean Std. Dev. Min Max MBPF beneficiary 2688 0.068 0.253 0 1 2076 0.071 0.257 0 1 Male 2688 0.733 0.443 0 1 2076 0.027 0.162 0 1 Age 2688 45.560 8.518 18 62 2076 42.340 9.051 19 62 Education - higher or general secondary degree 2688 0.964 0.187 0 1 2076 0.970 0.170 0 1 Household size (ln) 2688 1.450 0.340 1 2 2076 1.538 0.272 1 2 HH number of children under 6 2688 0.610 0.781 0 4 2076 0.645 0.776 0 4 HH number of children above 6 and below 18 2688 1.566 1.038 0 6 2076 1.621 1.062 0 5 South oblasts 2688 0.399 0.490 0 1 2076 0.387 0.487 0 1 Rural 2688 0.434 0.496 0 1 2076 0.451 0.498 0 1 Mountainous area 2688 0.344 0.475 0 1 2076 0.363 0.481 0 1 House walls not made from brick or concrete 2661 0.624 0.485 0 1 2059 0.650 0.477 0 1 House roof not made with roofing slates or concrete 2688 0.949 0.221 0 1 2076 0.949 0.220 0 1 House with no running water 2688 0.498 0.500 0 1 2076 0.474 0.499 0 1 Household has car 2661 0.238 0.426 0 1 2059 0.277 0.448 0 1 Household has car and/or refrigerator 2661 0.639 0.480 0 1 2059 0.644 0.479 0 1 Land bigger than 500 sq meters 2688 0.584 0.493 0 1 2076 0.618 0.486 0 1 Active Q1 2688 0.876 0.330 0 1 2076 0.653 0.476 0 1 Q2 2658 0.883 0.321 0 1 2067 0.663 0.473 0 1 Q3 2621 0.878 0.327 0 1 2045 0.677 0.468 0 1 Q4 2585 0.872 0.334 0 1 2024 0.651 0.477 0 1 Employed (extended definition) Q1 2688 0.838 0.369 0 1 2076 0.674 0.469 0 1 Q2 2658 0.864 0.343 0 1 2067 0.706 0.456 0 1 Q3 2621 0.872 0.334 0 1 2045 0.706 0.456 0 1 Q4 2585 0.862 0.345 0 1 2024 0.680 0.467 0 1 Note : (*) The variables who are not described by quarter refer to quarter 1. Source : Estimates based on KIHS 2012. 4. Results Labor force participation rates for all able-bodied adults, aged 18-62 years old, excluding fulltime students, and living in households with children range between 77.7 and 79.5 percent over the year. In the first quarter, 74.3 percent of able-bodied beneficiaries aged 18 to 62 years old are economically active. During the second and the third quarters, in which agricultural activities are more likely to happen, this share increases to 78.5 percent and further to 79.2 percent, after which it decreases to 75.2 percent in the fourth quarter. The higher level of activity in the second and in the 11

third quarters and given that the majority of beneficiaries lives in rural areas suggests that beneficiaries are more likely to be involved in agricultural activities than non-beneficiaries. MBPF beneficiaries have higher participation rates compared to poor non-beneficiaries. When compared to non-beneficiaries belonging to the richest 60 percent, beneficiaries have lower participation rates throughout the year. However, for the second and the third quarters, the difference between these groups decreases by half, from 6 to 3 percentage points (Figure 1, left panel). Unemployment rates in 2012 started at a relatively high 9.4 percent in the first quarter and declined to 5.3 percent at the end of the year. This reflects the overall economic situation in the Kyrgyz Republic. The economy only picked up speed in the second half of 2012 and 2013. MBPF beneficiaries were particularly hard hit with unemployment in the first quarter of 2012 compared to the rest of the population (16.4 percent compared to 9.9 and 8.1 percent for poor and non-poor non-beneficiaries). However, in the second half of 2012, unemployment rates for beneficiaries were slightly lower than for the others. MBPF beneficiaries living in the North have lower labor force participation rates compared to non-beneficiaries and to beneficiaries in the South. This holds for all four quarters. The labor force participation of MBPF beneficiaries living in the South is slightly higher when compared to nonbeneficiaries. Beneficiaries living in the North are less likely to be active. Seasonal employment appears to have a higher influence on the participation rate of able-adult beneficiaries living in the South: in the Northern oblasts, the share of employed beneficiaries varies from 53.2 percent in the fourth quarter to 62 percent in the second quarter, in the South it varies from 58.6 percent in the first quarter to 78.2 percent in the third quarter. However, during the third quarter, when agriculture activities are expected to be higher, a higher share of MBPF beneficiaries from the North have been engaged in farm activities, 54.9 percent against 43.3 percent for beneficiaries living in the south oblasts. When compared to MBPF beneficiaries living in south oblasts, those living in north oblasts are less likely to be employed at firms, institutions or collective farms, but more likely to be selfemployed and wage workers. MBPF beneficiaries living in the north oblasts are also more likely to be engaged in informal economic activities than those living in south oblasts. Housekeeping and childcare are the main occupations for inactive MBPF beneficiaries. According to the social status of the respondents, the main occupation for inactive MBPF beneficiaries is housekeeping and taking care of children, followed by retirement due to old age, and retirement due to disability. Inactivity due to housekeeping and taking care of children is also the most important reason for non-beneficiaries, though less pronounced. Retirement as reason for inactivity score higher among non-beneficiaries. This pattern is also observed among MBPF beneficiaries living in south or north oblasts. Household heads are more active in the labor market than other adult household members and they are less affected by seasonal variations. While household heads rates remain above 80 percent throughout the year, the rate for spouses varies from 64.1 percent (Q1) to 75.8 percent (Q3) and the rate for other relatives varies from 70.1 (Q4) to 76.3 (Q2). Compared to poor nonbeneficiaries, household heads from MBPF recipient households tend to have slightly lower participation rates, but their spouses have higher activity rates. Overall, participation rates of spouses in poor households are clearly lower than in households belonging to the top 60 percent of the welfare distribution (Figure 1, right panel). 12

Figure 1. Labor force participation components: able-bodied adults aged 18-62 years old, not students, living in households with children, Kyrgyz Republic, 2012 Q1-Q4 Determinants of labor outcomes Source: Estimates based on KIHS 2012. Overall, able-bodied adult household heads and spouses living in households with children and receiving MBPF transfers are neither less nor more likely to be active in the labor market (Table 6). 12 The only exception are household heads in MBPF recipient households, which in the third quarter are less likely to be active when the analysis only includes the bottom 40 percent of the welfare distribution. With respect to employment (extended definition), household heads that are MBPF beneficiaries are less likely to be employed in the first quarter of the year and in the third quarter when the analysis only includes the bottom 40 percent (Table 7). This result is most probably driven by the significantly higher unemployment rate for MBPF beneficiaries in the first quarter of 2012. For spouses, MBPF receipt has no effect on their likelihood to be employed or engaged in unpaid work. Spouses in MBPF households are between 14 and 30 percentage points more likely to be active in the informal sector, while the effects for household heads are not significant (Table 8). With 96 percent of spouses being female, this result may reflect gender-based preferences. One of the concerns about targeted social assistance programs relates to the potential risk of such transfers to push recipients into the informal sector (Tesliuc et al., 2014). While such effects have been found, for example in countries of the Western Balkans (Gotcheva & Sundaram, 2013; Koettl & Weber, 2012), it is not clear whether similar mechanisms explain the high likelihood of female MBPF beneficiaries to work informally in the Kyrgyz Republic. Informal employment accounts for 70 percent, which is significantly higher than in most countries in the Western Balkan. The argument that particularly individuals with low levels of education and low skills are pushed into informality (Koettl, 2013) does not hold either. Compared to individuals with only completed primary education, incomplete general secondary or less, adults that completed general secondary education or higher are neither less nor more likely to work in the informal sector. Furthermore, the means test 12 Full models are provided in the Annex. 13

in the Kyrgyz Republic includes income from all sources, whether formal or informal, and also accounts for potential income from land ownership. Table 6. Determinants of active labor force participation estimated by probit models: ablebodied adults, aged 18-62 years old, not students and living in household with children, Kyrgyz Republic, 2012 Marginal effects Bottom 40 % of annual All Variables Household head Spouse per capita comsumption Household Spouse head Active - Quarter 1 MBPF beneficiary -0.039-0.003-0.131-0.014 rural 0.009 0.0499** 0.0724* 0.023 south oblasts -0.0757*** -0.029-0.050-0.021 mountainous area 0.017-0.021 0.032-0.004 receives money from relatives or work abroad -0.006-0.0638** 0.065-0.0842** Other control variables omitted Active - Quarter 3 MBPF beneficiary -0.073 0.039-0.148** 0.080 rural -0.006 0.0702*** 0.0558* 0.100* south oblasts -0.0413* -0.019 0.002-0.056 mountainous area 0.0690** 0.002 0.031-0.051 receives money from relatives or work abroad -0.0602** -0.0725*** 0.039-0.123** Other control variables omitted Note : Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01. Source : Estimations based on KIHS 2012. Table 7. Determinants of employment estimated by probit model: able-bodied adults, aged 18-62 years old, not students and living in household with children, Kyrgyz Republic, 2012 Marginal effects Bottom 40 % of annual All per capita comsumption Variables Household Household Spouse Spouse head head Employed (extended definition) - Quarter 1 MBPF beneficiary -0.134** -0.063-0.220*** -0.076 rural 0.016 0.0974*** 0.068 0.060 south oblasts -0.0962*** -0.015-0.047-0.031 mountainous area -0.020 0.0545* 0.024 0.066 receives money from relatives or work abroad -0.036-0.0664** 0.046-0.069 Other control variables omitted Employed (extended definition) - Quarter 3 MBPF beneficiary -0.073 0.052-0.123* 0.087 rural 0.038 0.123*** 0.0943* 0.115*** south oblasts -0.0798** -0.011-0.019-0.008 mountainous area 0.046 0.028 0.053 0.024 receives money from relatives or work abroad -0.043-0.0727*** 0.031-0.073 Other control variables omitted Note : Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01. Source: Estimations based on KIHS 2012. 14

Table 8. Determinants of informal work estimated by probit model: employed (extended definition), able-bodied adults, aged 18-62 years old, not students and living in household with children, Kyrgyz Republic, 2012 Marginal effects Bottom 40 % of annual All per capita comsumption Variables distribution Household head Spouse Household head Spouse Informal work - Quarter 1 MBPF beneficiary 0.055 0.146** -0.076 0.135* rural 0.018 0.171*** 0.030 0.118* south oblasts -0.0456* 0.012-0.053-0.048 mountainous area -0.106*** -0.037-0.102* -0.024 receives money from relatives or work abroa -0.044 0.023 0.013 0.033 Other control variables omitted Informal work - Quarter 3 MBPF beneficiary 0.082 0.260*** -0.052 0.312*** rural 0.019 0.192*** 0.028 0.122** south oblasts -0.035 0.002-0.034 0.019 mountainous area -0.141*** -0.053-0.155*** 0.029 receives money from relatives or work abroa -0.0648** 0.029-0.025 0.105* Other control variables omitted Note : Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01. Source: Estimations based on KIHS 2012. The place of residence is significant to explain labor force participation and employment, however the significance level of the marginal effects are not consistent throughout the samples. Household heads living in Southern oblasts are less likely to be economically active and employed. But they are less likely to be engaged in informal work only in quarter 1. Living in a rural area increases spouses likelihood of being active, employed and being engaged in informal work. Income from remittances is negatively correlated with the probability of being economically active and employed, particularly for spouses. Working spouses in poor households that receive remittances are between 22 (Q1) and 37 (Q3) percentage points more likely to be work informally. The number of children in the household plays a significant role as determinant of labor force participation, but with different effects for household heads and spouses. Given that the size of the MBPF depends on the number of children in eligible households, the presence of more children may result in negative work incentives. The analysis indicates that the number of children indeed matters, but that the age of the children plays a role as well. The number of children below the age of six reduces the likelihood of spouses to be active in the labor market, irrespective of the specific outcome analyzed. On the other hand, an increasing number of children aged 6 to 18 has a positive effect on the labor market participation of household heads. Given that spouses are mainly female, inactivity may be a deliberate choice in order to care for the small children. Assuming that the costs of having school-aged children are higher, household heads need to work in order to support the family. The effect of children on informal work only plays a role for household heads but does not matter for spouses. With respect to other individual and/or household characteristics, men are more likely to be economically active and employed, as are individuals with higher education. 15

Furthermore, the likelihood to be economically active and employed increases with age up to a certain age, after which it decreases, but it plays no role in determining the probability of working informally. Table 9. Average treatment effect (ATT) of MBPF participation on labor market outcomes: ablebodied adults, aged 18-62 years old, not students and living in household with children, Kyrgyz Republic, 2012 Sample Household head Outcome variable for the ATT Active Spouse Employed (exteded definition) Household head Spouse All Q1-0.127*** -0.0580*** -0.243*** -0.159*** (0.0064) (0.0044) (0.0082) (0.0045) Q2-0.128*** 0.0196*** -0.168*** 0.0244*** (0.0037) (0.0061) (0.0039) (0.0057) Q3-0.110*** 0.0420*** -0.138*** 0.0730*** (0.0040) (0.0048) (0.0042) (0.0047) Q4-0.0739*** 0.0463*** -0.121*** 0.0560*** (0.0068) (0.0069) (0.0072) (0.0063) North Q1 0.0327*** -0.289*** -0.0834*** -0.375*** (0.0059) (0.0123) (0.0074) (0.0123) Q2-0.0202*** -0.265*** -0.0520*** -0.375*** (0.0046) (0.0119) (0.0074) (0.0126) Q3-0.0437*** -0.0993*** -0.0195*** -0.271*** (0.0035) (0.0093) (0.0045) (0.0103) Q4 0.0603*** -0.270*** -0.0741*** -0.383*** (0.0080) (0.0133) (0.0121) (0.0133) South Q1-0.0573*** 0.248*** -0.194*** -0.00189 (0.0045) (0.0051) (0.0052) (0.0052) Q2 0.0379*** 0.335*** -0.0325*** 0.309*** (0.0032) (0.0053) (0.0052) (0.0048) Q3-0.0452*** 0.254*** -0.0998*** 0.237*** (0.0045) (0.0044) (0.0045) (0.0044) Q4-0.0280*** 0.0498*** -0.0724*** 0.117*** (0.0059) (0.0046) (0.0060) (0.0048) Note: Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01. Source: Estimations based on KIHS 2012. Average treatment effect Household heads who are MBPF beneficiaries are less likely to be economically active than similar non-beneficiaries throughout the year, on the other hand, spouses are more likely to be economically active for most part of the year. The effect on household heads of participating in the MBPF varies between -7.4 p.p. in the fourth quarter, to -12.7 p.p. in the first quarter (Table 9). Participation in the MBPF program positively affects the likelihood of a household head being 16

economically active when considering beneficiaries from rural areas, except during the third quarter. MBPF beneficiaries living in the northern oblasts are also more likely to be economically active during the first and the fourth quarters when compared to similar non-beneficiaries. For those living in the southern oblasts, MBPF benefits contribute to higher probability of being economically active only during the second quarter. For spouses, the effect of participating in the MBPF varies from -5.8 p.p. in the first quarter, to a higher likelihood of participation of 4.6 p.p. in the fourth quarter (Table 9). Participation in the MBPF program negatively affects the likelihood of spouses being economically active in the northern oblasts. The impact of MBPF participation on the probability of being employed is negative. For household heads, the impact of MBPF benefits on the probability of being employed is negative, except for beneficiaries living in rural areas or south oblasts during the second quarter. On the other hand, spouses that are MBPF participants have a higher likelihood of being employed or engaged in unpaid work for most of the quarters. The effect is particularly strong for spouses living in the southern oblasts (Table 9). 5. Conclusion The Monthly Benefit for Poor Families with Children (MBPF) is the only social assistance transfer in the Kyrgyz Republic specifically targeted at extremely poor households with children. It is a means-tested transfer whereby eligibility depends on average formal and informal family income being below the Guaranteed Minimum Income (GMI). Although the design of the MBPF implies a 100 percent marginal tax rate for every additional Som earned above the GMI, the likelihood of potential work disincentives is limited considering the low benefit levels. In 2012, MBPF benefits represent less than 30 percent of the extreme poverty line; they are considerably below the official minimum wage and even more to what adults can earn in the formal and informal sector. While social assistance benefits can create labor market disincentives, this is mainly applicable in higher income countries and when benefits are close to wages for low-paid jobs. In most countries in Eastern Europe and Central Asia, such concerns are not warranted given the low generosity (and generally also low coverage) of social assistance benefits (Arias & Sanchez- Paramo, 2014). According to the descriptive analysis, household heads in MBPF households have higher activity rates when compared to spouses and other members of the household. This pattern is confirmed in the subsequent analysis for household heads, but the effect varies across quarter and per location. Overall, the analysis in this report indicates that the MBPF does creates disincentives for labor force participation for household heads living in Southern or Northern oblasts However, for spouses, the effect is positive throughout the year, except for those living in the North. Seasonal effects have a significant impact on labor force participation. This is closely related to the fact that MBPF beneficiaries predominantly live in rural areas and are engaged in farming activities, which are rather seasonal. Inactive MBPF recipients are mainly engaged with housekeeping and childcare. When analyzing the whole sample, the effect of MBPF participation changes according to the quarter in focus. Therefore, the effect of participation in the MBPF differs according to ones household positions. Regional and seasonal effects on labor market outcomes should be further investigated. The analysis in this paper indicates that the labor market in the Kyrgyz Republic follows a seasonal 17