Report No. 99775-KG Kyrgyz Republic Poverty and Economic Mobility in the Kyrgyz Republic Some insights from the ‘Life in Kyrgyzstan Survey’ April 27, 2015 Poverty Global Practice Europe and Central Asia Region Document of the World Bank CURRENCY AND EQUIVALENT UNITS (Exchange Rate Effective as of March 6, 2015 Currency Unit = Kyrgyz som (KGS) US$ = 61.41 WEIGHTS AND MEASURES Metric system ABBREVIATIONS CASE Center for Social and Economic Research DIW Deutsches Institut fuer Wirtschaftsforschung ECA Eastern and Central Asia KIHS Kyrgyz Integrated Household Survey LiK Life in Kyrgyzstan Survey LFS Kyrgyz Labor Force Survey NSC National Statistics Committee of the Kyrgyz Republic Vice President : Laura Tuck Country Director : Saroj Kumar Jha Practice Director : Ana Revenga Sector Director : Jean-Michel Happi Sector Manager : Carolina Sanchez-Paramo Task Leader : Sarosh Sattar ii ACKNOWLEDGEMENTS This report is part of the Kyrgyz Republic poverty analysis program led by Sarosh Sattar (Task Team Leader). The analysis was carried out by Anita Guelfi (consultant; GPVDR) and Sarosh Sattar (senior economist; GPVDR). The team has benefitted from advice and contributions from Maria E. Davalos (GPVDR), Aibek Uulu, Saida Ismailakhunova (GPVDR), and Moritz Meyer (GPVDR). The team is grateful to the Deutsches Institut fuer Wirthschaftsforschung (DIW) in Berlin for making their dataset available and providing assistance and clarifications when needed. The team is grateful to the UK’s Department for International Development (DFID) for their financing of the CAPAS program in Kyrgyz Republic that made this report possible. Any errors and omissions remain the responsibility of the authors. CONTENTS EXECUTIVE SUMMARY …………………………………………………………… i CHAPTER 1. BACKGROUND ……………………………………………………… 1 CHAPTER 2. THE DATA …………………………………………………………… 5 CHAPTER 3. ECONOMIC MOBILITY: THE EMPIRICAL STRATEGY ………… 7 A. BUILDING THE PANEL DATA-SET ………………………………………....... 7 B. IDENTIFICATION OF THE POVERTY POOL …………………….. …………….7 CHAPTER 4. ECONOMIC MOBILITY: MAIN FINDINGS ………………………. 9 A. SOME PRELIMINARY DESCRIPTIVE RESULTS ………………………….......... 9 B. PROFILE OF HOUSEHOLDS BELOW/ABOVE THE BOTTOM 40% THRESHOLD ...11 C. PROFILES OF ECONOMIC MOBILITY BY TRANSITION GROUP …………….… 15 CHAPTER 5. DRIVERS OF ECONOMIC MOBILITY: FIRST INSIGHTS FROM THE DIW PANEL DATA ………… …………………………………………..… 18 A. CORRELATES OF ECONOMIC MOBILITY: THE ROLE OF HH INVARIANT AND HH HEAD’S CHARACTERISTICS …………………………………………........ 18 B. CORRELATES OF ECONOMIC MOBILITY: THE ROLE OF INDIVIDUAL CHARACTERISTICS OF ALL MEMBERS OF EACH HOUSEHOLD ..……………. 24 CHAPTER 6. FOCUS ON PERSISTENCY INSIDE THE BOTTOM 40%: CHRONIC POVERTY? ……………………………………………………………………….... 28 CHAPTER 7. PRELIMINARY CONCLUSIONS AND STEPS AHEAD ……...… 32 APPENDIX ……………………………………………………………………….... 34 REFERENCES ……………………………………………………………………... 42 iv EXECUTIVE SUMMARY 1. In the last decade, sustained economic growth significantly contributed to reduce poverty across all segments of the population in the Kyrgyz Republic. In about ten years the share of people below the national poverty line fell indeed from almost 70% to about 37%, with most of the improvements especially benefiting the population living in rural areas. Rural poverty rates indeed declined dramatically between the early 2000s (73%) and 2008 (about 40%), also thanks to the increasing number of migrant workers1 and the corresponding rise in the inflow of remittances. 2. Starting from 2009, however, the pace of poverty reduction in the Kyrgyz Republic began to stagnate. Poverty rates stopped decreasing after 2008 and started to reverse in 2010, in a context of still sustained but rather unstable economic growth. In particular, most of the deterioration took place in the urban areas, where the share of poor people rose from about 22 to about 35% between 2009 and 2012. 3. Given this trend, one of the present challenges at the national level is to determine how to revive the poverty reduction process in the current less favorable economic environment. To this end it is then crucial to make a new effort to better understand the evolution and main characteristics of poverty dynamics in the Kyrgyz Republic in order to determine: (i) who is persistently poor as opposed to those moving out of or falling into poverty over time; (ii) what are the main determinants of the movement (or the absence of movement) observed in and out of the poverty status. 4. In this context, this report aims at shedding some light on the reasons lying behind the stagnation observed in poverty reduction since 2009. It does so by moving beyond the traditional static poverty analysis to adopt a “truly” dynamic approach to investigate individual wellbeing and behavior. It indeed focuses on the degree and main characteristics of economic mobility in the Kyrgyz Republic. Mobility analysis can indeed help to unveil what happened below the surface of poverty stagnation, thereby enabling to understand whether the reduced pace in poverty eradication was associated with high or low mobility in and out of the poverty status and to obtain a profile of the corresponding groups. In this study we are particularly interested in analyzing the role of labor market outcomes in moving people (or preventing them from moving) out of poverty. 5. This report could benefit from the availability of “true’ longitudinal data at the individual level. Mobility analysis was indeed carried out resorting to a relatively new data set of micro-data, made available by the Deutsches Institut fuer Wirtschaftsforschung (DIW) in Berlin. Unlike other available data sources, the DIW data-set consists of actual panel data, collected through a survey (the “Life in Kyrgyzstan Survey”2), which was able to track the same individuals over time. In particular, results presented here are based on the first two waves of data, referring to the years 2010 and 2011. The third wave, relative to the year 2012, was not available at the time this report was drafted and will be included in the analysis in the forthcoming months. 1 See Yang et al., 2015. 2 See chapter 2 for further details on the “Life in Kyrgyzstan” project a nd data characteristics. 6. In particular, this report focuses on the economic mobility of individuals (and corresponding households) belonging to the bottom 40% of the Kyrgyz population. This is indeed the target population chosen by the World Bank Group for the achievement of its second (and twin) goal of shared prosperity (the first one remaining poverty reduction). Moreover, in the specific case of the Kyrgyz Republic total poverty rates in the period under analysis ranged between 34 and 37%, thus making the poverty and bottom 40% pools almost identical. That is why, throughout the whole report we will often speak interchangeably of poverty and bottom 40% groups of the population. 7. Overall, the evidence suggests the existence of relatively high economic mobility between the bottom 40% and top 60% of the Kyrgyz population. Churning (i.e. movements of people in and out the poverty pool) between these two welfare groups was indeed relevant between 2010 and 2011: about 40% of those belonging to the bottom 40% in 2010 managed indeed to enter the upper deciles in the following years; conversely, about 19% of those who were in the top 60% in 2010 fell down into the bottom 40% in 2011. 8. For each selected transition group, the report depicts a detailed profile of corresponding main socio-economic characteristics. Four transition groups were in particular identified: (i) those who were found to be stuck in the bottom 40% in both 2010 and 2011; (ii) those who managed to move up from the bottom 40% in 2011; (iii) those who entered the bottom 40% in 2011, and (iv) those who managed to stay in the top 60% in both years under analysis. The evidence shows a few significant differences among these four groups in terms of both households’ invariant and individual characteristics. 9. Households’ size and dependency ratios appear to be significantly different between the selected transition groups. The rural/urban divide also appears to be a strong factor of differentiation. Households persisting in the bottom 40% are characterized on average by a relatively higher number of both members and dependents compared with those managing to move up to the top 60% in 2011. Symmetrically, both size and dependency ratios are significantly lower in households managing to stay in the top 60% in both years compared with those falling into the bottom 40% in 2011. As far as the geographical location is concerned, rural households appear to be more likely to persist in the bottom 40% compared with urban ones. 10. Focusing on individual characteristics, summary statistics suggest a relevant role of both education and labor market outcomes. When the selected transition groups are broken down by the individual characteristics of the household head, the evidence shows that heads with a relatively higher educational attainment are more often found in households either moving out and up from the bottom 40% in 2011 or staying in the top 60% in both years. Similarly, households escaping poverty appear to be characterized by a relatively higher employment rate of their household head; this looks particularly true if the head is employed in the service sector. 11. To test the robustness of these first summary results, a multivariate regression analysis was carried out to investigate the correlates of economic mobility for each of the selected transition groups. A linear probability model was used, regressing the probability of belonging to each of the four selected transition groups against both households’ invariant and individual characteristics. Results confirm many of the features highlighted by the simple summary statistics, but also look different from them in some important respects. 12. Household size and dependency ratio confirmed as significant drivers of economic mobility. Regression analysis, carried out with different specification models, confirm that larger households with a relatively higher number of dependents are more likely to be stuck in the bottom ii 40% over time. Conversely, multivariate analysis did not find a clear-cut and significant correlation between the probability of persisting in poverty/moving out of it and households’ location in rural vs. urban areas (though some strong correlations seem to characterize some specific regions/oblasts of the country). 13. Households’ reliance on remittances seems to play a positive role for economic mobility. Receiving remittances is indeed found to be associated on average with a relatively higher probability of leaving the bottom 40%, but it also appear to enhance the chance for top 60% households to remain in the upper deciles in the following period. 14. No clear-cut correlation emerges instead for the availability of social benefits. The evidence shows indeed that receiving social benefits does not appear to influence either persistency into the bottom 40% or the movement out of it, whereas a significant correlation is found between social benefits and the probability of remaining in the top 60% in both years. 15. Educational attainment emerges as key driver of economic mobility. Data analysis highlights a strong link between the educational attainment of each member of a household and the corresponding outcome in terms of economic mobility. Households belonging to the bottom 40% in 2010 with a relative higher share of members with a tertiary educational attainment are indeed found to face a significantly higher chance of leaving poverty in the following year. 16. Sector and quality of employment appear to be relevant for economic mobility. As far as labor market outcomes are concerned, being employed (as opposed to being either unemployed or inactive) does not seem to matter for economic mobility. What appears to be more significant in terms of correlates of mobility is instead the specific sector and quality of the jobs people are employed into. Working in agriculture and mining enhances indeed the chance of persisting in the bottom 40% in both years, whereas being employed in the service sector seems to significantly increase the probability of staying in the upper deciles and avoid poverty. A similar finding emerges for jobs of different quality, with households with a relatively higher share of members employed in low-skilled types of occupations facing a significantly higher probability of being stuck in poverty over time. 17. Finally, a section of the report was devoted to address the characteristics of a specific sub-group of households belonging to the bottom 40%, i.e. those who turned out to be stuck in the very first deciles of the welfare distribution. We focused in particular on those households who were found to persist in either the very first or the second deciles of the distribution as well as those who moved down from the second to the first decile between 2010 and 2011; in this report, these households were defined as the “chronic poors”. 18. Chronic poor households more often found in urban areas and depending on (low?) wage incomes. They are often unemployed or inactive. Compared with the rest of households belonging to the bottom 40%, chronic poor ones are more likely to reside in urban areas and resort on a relatively higher share of wage income compared with other income sources (working poors?). They also claim to rely on a significantly higher proportion of income deriving from alimony, inheritance, and scholarships. As far as their labor market status is concerned, chronic poor appear to have a significantly higher probability of being non-employed (i.e. either unemployed or inactive). iii 1. BACKGROUND 1.1 In the last decade, the Kyrgyz Republic achieved significant improvements in poverty reduction. The Kyrgyz Republic is one of the poorest countries in the ECA region: about two-thirds of its population lives in rural areas. Like most of other ECA countries, it benefited from a decade of sustained growth, which significantly contributed to reduce poverty across all segments of the population. Between 2003 and 2008 the share of people below the national poverty line fell indeed from 49.9% to 31.7%, with significant reductions both in the urban and rural areas of the country (Chart1). 1.2 However, the pace of poverty eradication began to loose strength in the more recent years. Starting from 2009 poverty reduction began indeed to stagnate, and started to reverse in 2010: according to the last available data, the national poverty rate amounted to about 38% in 2012, in a context of still sustained but rather unstable growth in the total economy (Chart 2). In particular, urban areas appear to have been relatively more affected by this deterioration compared with the rural ones (though poverty rates are still significantly higher in the rural regions; Chart 1). 1.3 This paper tries to throw some light on what could have happened and why. It does so by providing a more thorough investigation of poverty dynamics and its determinants in the Kyrgyz Republic. It indeed tries to move beyond the traditional static poverty analysis to adopt a “truly” dynamic approach to investigate individual wellbeing and behavior. 1.4 This was made possible by the availability of a relatively new data-set of micro-data at both the household and individual levels, produced by the Deutsches Institut fuer Wirtschaftsforschung (DIW) in Berlin, as part of a wider research project3. Unlike the Kyrgyz Integrated Household Survey (KIHS), the DIW survey (called “Life in Kyrgyzstan- LiK” survey) is a true panel survey, which tracks individuals over time, thereby allowing to move from static to truly dynamic analyses of individual well-being and behavior. The availability of longitudinal data is indeed expected to enable us to shed light on the transitions in and out of the poverty status, as well as on the persistence into it, also providing some insight into its main determinants4. 3 The project – “Economic Transformation, Household Behavior and Well-Being in Central Asia: The Case of Kyrgyzstan”, was funded by the Volkswagen Foundation and carried out in collaboration with Humboldt University of Berlin, the Center for Social and Economic Research (CASE-Kyrgyzstan), and the American University of Central Asia. Overall, it aims at (i) collecting nationally representative panel survey data in Kyrgyzstan; (ii) investigating well-being and behavior of individuals and households in this country; and (iii) improving research capacity within Central Asia (for more details, see: http://www.diw.de\kyrgyzstan). 4 See chapter 2 for a more detailed description of the LiK dataset. Figure 1.1: Recent trends in absolute poverty rates, 2003-2012 (In % of total population) 70.0 60.0 50.0 49.9 40.0 45.9 43.1 38.0 39.9 30.0 35.0 36.8 31.7 31.7 33.7 20.0 10.0 0.0 2003* 2004 2005 2006 2007 2008* 2009 2010 2011* 2012 Total Urban Rural *Years when poverty lines were re-estimated (otherwise indexed). Source: Kyrgyz National Statistics Committee. Figure 1.2: Sustained (but unstable) growth and stagnation in poverty reduction after 2009 60.0 15.0 14.0 13.0 50.0 12.0 11.0 40.0 10.0 9.0 Real GDP 8.0 30.0 7.0 Poverty rate 6.0 5.0 20.0 4.0 3.0 10.0 2.0 1.0 0.0 0.0 -1.0 20 20 20 20 20 20 20 20 20 20 20 03 04 05 06 07 08 09 10 11 12 13 Poverty rate 49.9 45.9 43.1 39.9 35.0 31.7 31.7 33.7 36.8 38.0 37.0 Extreme poverty rate 16.2 13.4 11.1 9.1 6.6 6.1 3.1 5.3 4.5 4.4 2.8 Real GDP (right axis) 7.0 7.0 -0.2 3.1 8.5 8.4 2.9 -0.5 6.0 -0.1 10.5 Source: Kyrgyz National Statistics Committee 2 1.5 The focus of this report will then be on the economic mobility of households and its members. Mobility analysis5 can indeed help to unveil what happened below the surface of poverty stagnation. We will indeed investigate how many and what types of households tend to: (i) persist in poverty, (ii) move out of poverty, (iii) move into poverty, and (iv) stay out of poverty. This will enable us to understand whether observed stagnation in poverty reduction has been associated with high or low mobility in and out of the poverty status and to obtain a profile of the corresponding groups. At the same time the wide set of households’ and individuals’ characteristics collected by the DIW survey will help us to identify some of the key determinants of the observed mobility patterns. In this study we are particularly interested in analyzing the role of labor market outcomes in moving people (or preventing them from moving) out of poverty. 1.6 The rest of the paper is organized as follows. After a short description of the DIW dataset and its main features (chapter 2), chapters 3 is devoted to explain in detail our empirical strategy, whereas chapter 4 contains our main findings. Finally, the last chapter offers some conclusions. 5 See Cancho et al., 2015, as well as Davalos and Meyer, 2015 for other recent research work on economic mobility in the ECA region. 3 2. THE DATA 2.1 Data are derived from the Life in Kyrgyzstan (LiK) Survey, carried out by the DIW between 2010 and 2012, as part of the research project “Economic Transformation, Household Behavior and Well-Being in Central Asia: The Case of Kyrgyzstan”6. The LiK survey is a multi-purpose, socio-economic survey covering many different topics of interest in the analysis of well-being and behavior of individuals and households. Data cover information on demographics, assets, expenditure, migration, employment, agricultural markets, shocks, social networks, and subjective wellbeing7. 2.2 It is an individual panel survey. All adult members (i.e. aged 18 and older) of the sampled household are interviewed and followed over time. This means that: “if a member of an original household leaves the household (e.g. to form an own family), she is still part of the sample. If relevant, other members (e.g. spouse and children) of the new household are then included in the sample as well”8. 2.3 The LiK survey is representative at the national, rural/urban, as well as south/north levels of the country. It collects data in all seven Kyrgyz oblasts and the cities of Bishkek and Osh. 2.4 Three waves of this survey have been carried out so far, covering respectively the years 2010, 2011 and 2012. At present only the first two waves of data have been made available. In the original 2010 sample, households were drawn through stratified two-stage random sampling9. In 2010 the survey was carried out on 3,000 households and 8,160 individuals. In the following year interviewed households amounted to 2,863 (for a total of 8,066 individuals), 95.2% of which had already been interviewed in 2010. Table A1 in the appendix provides an example of the main sample characteristics with respect to the 2010 wave of data. 2.5 The survey consists of three main questionnaires: (i) a household questionnaire, to be filled in by the most informed household member; (ii) an individual questionnaire, to be filled in by all adults aged 18 and over; (iii) a community questionnaire, to be filled in by a representative of the local administration. For our research purposes, the following variables of interest were selected: (a) From the household questionnaire we extracted most of the available information on each HH’s composition and related demographic characteristics (e.g. number of components, gender, age, ethnicity, as well as geographical location). 6 See footnote 3, page 1. 7 Each adult individual was administered a household, an individual and a community questionnaire. Information about children was included in the household questionnaire. 8 . For further methodological details see DIW (2010). 9 A detailed description of the LiK survey can be found in Brück et al. (2012), “Household Survey Data for Research on Well-being and behavior in Central Asia”, IZA DP No. 7055, November. (b) From the household questionnaire we also took the information about each household’s consumption and expenditure, as well as data on main income sources. (c) We then turned to the individual questionnaire to gather information about each adult’s educational attainment and all available data referring to her/his labor market status. In particular, it was possible to distinguish between employed and non-employed and among the employed it was possible to know each adult’s sectorial affiliation, her/his wage status (e.g. employer, own-account worker, employee, etc.), her/his occupational status (e.g. managers, professionals, clerks, production workers, etc.). It was more difficult to distinguish the pool of non-employed between unemployed and inactive but some rough estimation is possible. (d) We have not used any information contained in the community questionnaire so far10. 10 We are however planning to use the community questionnaire in the forthcoming second phase of this research project, especially focusing on information on access to services and infrastructure. 6 3. ECONOMIC MOBILITY: THE EMPIRICAL STRATEGY A. BUILDING THE PANEL DATA-SET 3.1 The first step of our empirical strategy consisted in linking the available waves of data by exploiting the panel structure of the survey. This was done so far only on the first two waves, i.e. those covering the years 2010 and 2011, since the 2012 data have not yet been made available11. 3.2 By merging the 2010 and 2011 waves, we ended up with 8,658 individuals, whose characteristics were tracked in both years. However for some of them (22 in 2010 and 23 in 2011) no information was provided on food and nonfood consumption expenditure by the household (HH) head. Thus we decided to eliminate these individuals/households from the sample. We also observed that some individuals younger than 18 were still part of the sample, despite the claim that only people aged 18 and older had been tracked overtime: we thus chose to eliminate everybody younger than 16 years old. As a result, our mobility analysis will be based on a panel sample made of 8,541 individuals, corresponding to 2,841 households. B. IDENTIFICATION OF THE POVERTY POOL 3.3 In this study the focus will be on the economic mobility of the bottom 40% of the Kyrgyz population. This is indeed the target population chosen by the World Bank in order to achieve its second (and twin) goal, which is to boost shared prosperity, i.e. to act so that the income of this segment of the population grows faster than the rest of it. Moreover, in the specific case of the Kyrgyz Republic total poverty rates in the period under analysis ranged between 34 and 37%, thus making the poverty and bottom 40% pools almost identical. That is why, throughout the whole report we will often speak interchangeably of poverty and bottom 40% groups of the population12. 3.4 Given this choice, the second step of our empirical strategy then consisted in identifying who belongs to the bottom 40% of the population. From the previous chapter we already know that the DIW survey only tracks individuals aged 18 and over; thus, if we computed a transition matrix among these individuals between 2010 and 2011, we would be focusing on the bottom 40% of these individuals (which we can observe) and not on the entire population. 3.5 To cope with this issue (intrinsic to the nature of the data we have) we chose to adopt the following two-step empirical strategy: 11 At the moment of drafting this report, their release is about to take place, so that new data will be included in the analysis in the following months. 12 In addition, this choice helps to overcome eventual differences arising in the magnitude of the poverty pool, due to underlining differences in the way poverty lines have been computed according to different data sources (a) Infer the actual distribution of the entire surveyed (panel) population (observed and non-observed individuals, i.e. 2,841 households including 13,175 individuals in 2010 and 13,595 in 2011), by imputing households’ information on per-capita consumption to each household’s member (we indeed know each household’s size). This allowed us to identify the 40% cut-off point and determine who lay in the bottom 40%. As a result: 5,267 individuals turned out to belong to the bottom 40% in 2010 and 5,435 in 2011. (b) Once we knew the cut-off point at the individual level, we could then project these people back to the households they belonged to, and identify how many households lie below the 40% line. In our case, results indicated that about one third of all households (31.7% in 2010; 31.9% in 2011) belonged to the bottom 40% in each of the selected years. We could then compute a transition matrix at the household level between 2010 and 201113. 3.6 With this strategy in mind, we can then turn to mobility analysis and its main results. 13 Although mobility is captured by the DIW survey at the individual level (on people aged 18 and over), the welfare measure, i.e. per-capita consumption, is indeed measured at the household level. 8 4. ECONOMIC MOBILITY: MAIN FINDINGS 4.1 This section presents the report’s main findings. The first part displays some basic descriptive statistics providing: (i) a profile of households belonging to the bottom 40% vis-à-vis those beyond this threshold; (ii) a profile of households in each of the transition groups highlighted after computing our mobility matrix (see below). The second part will then focus on the analysis of correlates of economic mobility. 4.2 Throughout both sections we chose to start focusing first on households’ invariant characteristics (e.g. HH size, HH dependency ratio, HH location, as well as the composition of its income sources), together with some individual ones but only related to the HH head (e.g. age, gender, ethnicity, education, labor market status), and only then move to investigate the role of each member’s individual characteristics. We indeed want to look separately at how each household’s poverty status and the corresponding mobility trend does change as a function of each household’s common characteristics as compared with how it changes as a function of each member’s individual characteristics. A. SOME PRELIMINARY DESCRIPTIVE RESULTS 4.3 Households’ mobility across different deciles of the welfare distribution between 2010 and 2011 is displayed in Table 4.1. As already explained in the previous section, household mobility is here deduced from the sample of adult members who have been tracked in both years, but the bottom 40% threshold was derived from the entire surveyed population using information on each household size. 4.4 Each cell percentage represents the share of households transitioning between different deciles of the consumption distribution between 2010 and 2011, due to the movements observed in their individual adult members. For example, 31% of households belonging to the first decile of the distribution in 2010 were still there in the following year, while 20.2% of them moved up to the second decile; on the other hand, households’ persistency in the richest decile of the population (in terms of consumption) amounted to 46.4%, while 21.9% of them moved down to the 9th decile. 4.5 If we now focus on the economic mobility of those belonging to the bottom 40% of the population (our target group in term of poverty reduction) vis-à-vis those beyond this threshold, our sample can be broken down into four different transition groups of households (Table 4.1): (a) Those belonging to the bottom 40% in 2010 and still there in 2011 (poverty persistence; red shade); (b) Those belonging to the bottom 40% in 2010 but moving beyond this threshold in 2011 (exit from poverty; green shade); (c) Those being beyond the 40% threshold in 2010 but moving down the threshold in 2011 (entry into poverty status; brown shade); and (d) Those staying beyond the 40% in both years (out of poverty persistence; yellow shade). Table 4.1: Decile mobility matrix of households (%) DECILES 2011 DECILES 2010 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Total decile decile decile decile decile decile decile decile decile decile 1. decile 31.0 20.2 15.8 9.9 7.9 6.4 4.9 1.5 1.0 1.5 7.1 2. decile 17.0 13.5 18.8 14.3 11.2 10.3 5.4 6.3 1.3 1.8 7.8 3. decile 10.5 14.4 14.8 12.2 15.3 11.4 7.0 7.4 4.8 2.2 8.1 4. decile 8.9 10.5 12.1 17.0 15.4 8.9 10.5 6.9 7.7 2.0 8.7 5. decile 8.7 9.8 9.8 13.6 11.0 14.8 11.0 8.3 7.2 5.7 9.3 6. decile 3.7 7.0 7.4 10.0 9.2 15.1 10.7 15.9 12.9 8.1 9.5 7. decile 3.9 6.2 5.2 8.2 10.8 5.9 17.0 17.0 10.5 15.1 10.7 8. decile 2.0 2.3 4.1 5.5 11.0 9.9 16.2 15.9 16.2 16.8 12.1 9. decile 1.1 3.6 4.9 2.5 6.8 8.5 10.7 13.7 25.4 23.0 12.9 10. decile 0.3 2.1 0.8 1.8 2.3 5.7 6.2 12.6 21.9 46.4 13.7 Total 7.2 7.8 8.3 8.6 9.6 9.5 10.3 11.3 12.5 14.9 100.0 Source: Own computations on DIW. Note: Households remaining in the bottom 40% (poverty persistence). Households exiting the bottom 40% threshold in 2011 (exit poverty) Households entering the bottom 40% from above (from non-poverty to poverty) Households remaining in the top 60% 2011 (persistence out of poverty) 4.6 Overall, the evidence suggests the existence of a relatively high economic mobility between the bottom 40% and top 60% of the Kyrgyz population. Numbers displayed in Table 1 show in particular that about 60% of households (corresponding to 537 households) who were in the bottom 40% in 2010 persisted in the same condition in the following year, whereas the remaining 40% (365 households) managed to move upwards and climb over the 40% threshold. On the other hand, about 19% (370) of households who were in the top 60% in 2010 moved down to the bottom 40% in 2011, while the remaining 81% remained in the upper deciles. This is shown in a more summarized version in Figure 4.1 below. Figure 4.1: Household mobility between bottom 40% and top 60%, (%) 100% 90% 19.1 80% 40.5 70% 60% 50% 40% 80.9 30% 59.5 20% 10% 0% In bottom 40% in 2010 In top 60% in 2010 Stayed Moved out Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 10 B. A PROFILE OF HOUSEHOLDS BELOW/ABOVE THE BOTTOM 40% THRESHOLD 4.7 But who are these households? We tried to better characterize households below and above the 40% threshold. In particular, by exploiting the wide set of characteristics collected by the DIW survey we tried to draw a profile of those belonging to the bottom 40% vs. those who being beyond this threshold. Furthermore we went on by investigating in more detail the characteristics of households in each of the transition group highlighted above, i.e. (i) stayed in bottom 40%; (ii) moving upwards from bottom 40%; (iii) entering bottom 40% from above; (iv) staying above bottom 40%. 4.8 As already explained before, we chose to start by focusing only on households’ invariant characteristics, plus some individual characteristics related to the household’s head. We indeed want to start investigating how each household’s poverty status and the corresponding mobility trend does change as a function of each household’s common characteristics. 4.9 A first profile of households belonging to the bottom 40% (vis-à-vis the top 60%) is displayed in Figure 4.2 (here data refer to 2010; the following table 4.2 displays instead both 2010 and 2011, which however tell about the same story). As expected, relatively poorer households tend to be of larger size, display on average higher dependency ratios, and appear to be significantly more concentrated in rural areas. Significant differences also emerge, when data are broken down by geographical region (oblast). If we then focus on the different sources of households’ income, the most significant difference emerges when comparing income from household enterprises and income from wage employment, with the bottom 40% households receiving a relatively higher share of their income from the former and the top 60% ones relatively more resorting to wage employment. Figure 4.2: Profile of HHs below/beyond bottom 40% by main HH characteristics 2010 80.0 70.0 60.0 50.0 Bottom 40% 40.0 Top 60% 30.0 20.0 10.0 0.0 HH enterprises Batken Naryn Osh city Bishkek Village Wage employment HH average size City Chui Pensions Remittances* HH dependency ratio (<14 & >65) Property income Aid from people in Kyrgyzstan Other Issyk-Kul Osh Djalal-Abad Talas Other social transfers Demogr. Rural/urban Geographical region/Oblast Source of income Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 11 Table 4.2: Profile of HHs below/above the bottom 40% threshold by main HH characteristics 2010 2011 Bottom Top Bottom Top 40% 60% 40% 60% Demographics HH size (# of members) 5.8 4.1 6.0 4.2 HH dependency ratio (<14 & >65) 36.9 27.7 44.0 34.5 Urban vs. rural location City 23.7 48.7 24.4 48.4 Village 76.3 51.3 75.6 51.6 Geographical region/Oblast Issyk-Kul 9.6 9.2 10.9 8.6 Djalal-Abad 20.6 14.6 20.4 14.7 Naryn 4.1 4.5 2.9 5.0 Batken 10.4 6.5 7.2 8.1 Osh 25.6 12.2 26.4 11.8 Talas 6.7 3.1 7.3 2.8 Chui 12.5 18.8 13.2 18.5 Bishkek 6.9 26.4 7.1 26.3 Osh city 3.5 4.7 4.7 4.2 Source of income (as % of total income) HH enterprises 38.4 33.8 34.7 28.8 Property income 1.6 1.3 0.5 0.8 Pensions 17.1 18.9 16.7 18.0 Other social transfers 2.8 1.2 2.3 1.0 Remittances* 1.3 1.9 5.0 5.6 Aid from people in Kyrgyzstan 3.2 3.7 2.7 3.2 Wage employment 33.7 37.8 37.4 41.2 Other 1.8 1.4 0.7 1.4 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 4.10 Figure 4.3 describes instead the main differences between bottom 40% and top 60% households in terms of some individual characteristics referred to their HH head. In particular, Figure 4.3 focuses on differences in education, labor market status, specific sectors of employment, as well as the wage status (further characteristics are shown separately in Table 4.3). 4.11 Educational attainment strongly differentiates between the two groups of households. Households in the bottom 40% display a higher share of HH heads with a relatively lower educational attainment: about 75% of them completed in 2010 at most a secondary general level of education (compared with about 50% of top 60% households); the share of poor HH heads with a tertiary degree was 9% in 2010, compared with 22.9% of non-poor HH ones. 4.12 Employment rates also look significantly different. HH heads in the bottom 40% appear to have a relatively lower employment rate compared with the top 60% ones (56.5% vs. 12 59.2%). Detailed data in Table 4.3 show in particular that, while the proportion of non- employed people who have retired looks the same in the two groups of households, poorer households have a higher share of non-employed people who are either unemployed or inactive (different from retired). Figure 4.3: Profile of HHs below/beyond bottom 40% by HH head’s characteristics 2010 70.0 60.0 50.0 Bottom 40% 40.0 Top 60% 30.0 20.0 10.0 0.0 Public services Tertiary education Agriculture Low-skilled Industry without construction High-skilled Employed Private services Other Up to lower secondary Medium-skilled Non-employed Own-account worker Secondary general Primary and secondary technical Construction Employer Employee Member of producers' Contributing family workers cooperative Education Labor mkt Sector Wage status Skill level status Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 4.13 Sector of employment as well as the specific wage status also different between the two welfare groups. Employed HH heads in the bottom 40% appear to be relatively more concentrated in agriculture and construction; they are prevalently own-account workers and employees. 13 Table 4.3: Profile of HHs below/above the bottom 40% threshold by HH head’s characteristics 2010 2011 Bottom Top Bottom Top 40% 60% 40% 60% Age Up to 24 0.6 2.1 0.4 1.6 from 25 to 34 12.0 10.4 11.5 9.5 from 35 to 44 25.7 21.2 24.8 20.1 from 45 to 54 27.6 30.5 28.7 30.7 from 55 to 64 16.4 19.3 17.0 22.0 65+ 17.7 16.6 17.6 16.3 Education Up to lower secondary 19.2 12.8 17.3 12.2 Secondary general 55.3 37.5 55.9 38.5 Primary and secondary technical 16.4 26.7 17.7 26.1 Tertiary education 9.0 22.9 9.1 23.2 Labor market status EMPLOYED 56.5 59.2 61.9 63.5 of which: Employer 0.2 1.2 0.2 1.8 Own-account worker 31.8 24.5 34.8 26.5 Employee 23.4 32.0 24.6 33.0 Member of producers' cooperative 0.1 0.3 0.0 0.2 Contributing family workers 1.0 0.8 1.8 1.8 Other 0.0 0.4 0.4 0.4 NON-EMPLOYED 43.5 40.8 38.1 36.5 of which: Retired 16.7 16.7 10.5 10.1 Other non-employed (this includes the unemployed) 26.7 24.1 27.7 26.4 Sector Agriculture 46.1 23.9 49.0 26.4 Industry without construction 7.5 11.2 6.4 9.8 Construction 13.1 7.6 12.7 8.1 Private services 19.4 33.3 18.9 32.7 Public services 13.9 24.0 13.0 23.0 Type of occupation Legislator, senior official, manager 1.0 3.7 2.0 5.8 Professional 5.7 8.4 5.5 11.0 Technician, associated professional 5.3 6.3 3.9 7.4 Clerk 2.0 9.4 1.4 5.4 Service worker, shop or market sales worker 8.8 14.7 10.0 17.2 Skilled agricultural or fishery worker 12.9 8.5 10.7 5.0 Craft and related trades 12.9 16.3 8.7 12.1 Plant or machine operator or assembler 0.8 1.6 1.4 1.6 Unskilled worker 50.2 30.9 56.1 34.1 Armed forces 0.4 0.2 0.2 0.4 Skill level High-skilled 12.0 18.4 11.4 24.2 Medium-skilled 23.7 32.7 22.1 27.6 Low-skilled 63.9 48.8 66.3 47.8 Armed forces 0.4 0.2 0.2 0.4 14 C. PROFILES OF ECONOMIC MOBILITY BY TRANSITION GROUP 4.14 Figures 4.4 and 4.5 focus instead on the transitions observed between 2010 and 2011. They indeed provide a profile of households in each of the following transition statuses: (i) stayed in bottom 40%, (ii) moved down to bottom 40%, (iii) moved out from bottom 40%, and (iv) stayed in top 60%. Figure 4.4: Profiles of economic mobility by main transition group: share of HHs in each group by HH invariant characteristics, 2010 90.0 80.0 70.0 Stayed below 40% Down to 40% Up from 40% Stayed up 60% 60.0 50.0 40.0 30.0 20.0 10.0 0.0 Remittances HH average size Pensions Rural Wage employment Other HH enterprises Property income Aid from people in HH dependency ratio (<14 Other social transfers Kyrgyzstan & >65) Demogr. Rural/Urban Source of income Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 4.15 HH size and dependency ratios significantly different. As before, the first chart analyzes the role played by households’ invariant characteristics and shows that, as expected, larger households with higher dependency ratios are found to be associated with a relatively higher chance of being stacked in the bottom 40% in both years under analysis. The share of households persisting in the poverty status tends also to be relatively higher in rural areas compared with those living in the urban ones. 4.16 Interesting results also emerge with respect to the different sources of income accruing to each household. Results indicate indeed that households either persisting or moving down into the poverty status tend to have a relatively higher share of income deriving from HH enterprises; on the contrary, households either moving out or staying out of poverty in 2011 tend to resort relatively more to income from wage employment. No clear-cut results do instead emerge with respect to the other sources of income. 4.17 Figure 4.5 looks instead at the association between our four transition groups and some selected individual characteristics referred to each household’s head. Both education and labor market outcomes emerge as potentially determinant factors in helping households skipping poverty. As far as education is concerned, the evidence shows indeed that when households’ heads have completed tertiary education they have a relatively higher probability of moving out or persisting out of poverty. Households 15 persisting in poverty are instead associated to a very large extent with heads having at most a secondary general diploma. 4.18 Potentially interesting results emerge as far as labor market outcomes are concerned. First of all, both those moving out of the bottom 40% and those staying out of it in 2011 display a relatively higher employment rate; at the same time the percentage of non- employed people (which includes both the unemployed and the inactive) turns out to be relatively higher in those households, which either persist or fall into poverty. Thus, at first sight, we could start thinking that having a job can help avoiding poverty. Let’s however try to be more specific and start by looking at what happens if we break our data down by different economic sectors. In this case our findings suggest that only working in some sectors does actually help escaping poverty. In particular, more than 60% of those staying out of the bottom 40% in 2011 claim to work in the service sector14, while the same share associated with households persisting in the bottom 40% is less than 30%. The majority of people staying in poverty in both years work indeed in agriculture; they are also relatively more present in mining and construction. Non clear-cut results do seem to emerge for manufacturing with its employment rate relatively higher for both those staying out of poverty and falling into it. Figure 4.5: Profiles of economic mobility by main transition group: share of HHs in each group by HH head’s characteristics 70.0 Stayed below 40% Down to 40% Up from 40% Stayed up 60% 60.0 50.0 40.0 30.0 20.0 10.0 Member of producers'… 0.0 Manufacturing* Retired Mining Tertiary Services High-skilled Secondary general Other nonemployed Construction Employer Employee Employed Medium-skilled Low-skilled Agriculture Less than secondary general Own-account worker Contributing family workers Education LM status Sector (employment Employment type Skill level rate) Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 14 Macroeconomic data show that structural shifts have occurred in the Kyrgyz Republic over time. Historical poverty reduction has indeed been associated with relevant employment declines in agriculture and corresponding increases in the service sector. This shift was associated with a strong movement of workers from rural to urban areas, but also with a rise of informal jobs. 16 4.19 Other interesting results emerge if we look at households’ heads with different types of employment. The type of employment seems indeed to be associated with different transitions in terms of poverty. The vast majority of households persisting into poverty or falling into it in 2011 is indeed associated with a head working as an own-account worker, while most households exiting or staying out of poverty have a head working as an employee. 4.20 If the type of employment appears to be significantly linked with the poverty status, another relevant differentiating factor seems to be represented by the type of specific occupation. Occupations are collected by the DIW survey according to the international classification of occupations (ISCO), which distinguished jobs by the intrinsic required skill level of the involved worker. Taking this characteristic into account, we re-grouped employment data by occupation into three main categories: (i) high-skilled occupations; (ii) medium-skilled occupations; and (iii) low-skilled occupations. If being an employee is positively associated with leaving or persisting out of poverty, data by skill level show additionally that only jobs with either a medium or high skill level do enhance the chance of avoiding poverty. More than 60% of households persisting in poverty in 2011 have indeed a household’s head working in an unskilled job position. However the proportion of unskilled workers among household’s head is about 60% also for those moving out of poverty in 2011: this could mean either that being skilled is not the driving force to escape poverty or that it is not enough to look at the characteristics of the HH head, since the labor market characteristics of the other members of the household could exert a relevant impact on the overall result. That is why we need to go deeper into our investigation and turn to multivariate regression analysis. 17 5. DRIVERS OF ECONOMIC MOBILITY: FIRST INSIGHTS FROM THE DIW PANEL DATA 5.1 In this section we try to go deeper into the analysis in the attempt to identify some of the drivers of the transitions observed above. To this end a multivariate regression analysis has been carried out in order to estimate the correlates of economic mobility in the 2010-11 period. In particular, we used a linear probability model, where the dependent variable was alternatively one of the following four transition groups: (i) stayed in bottom 40% in 2011; (ii) moved out of bottom 40% in 2011; (iii) moved into bottom 40% in 2011; and (iv) stayed out of bottom 40%. Following the line of reasoning adopted in the previous section, here again we started by regressing our dependent variables only against HH invariant as well as HH head’s individual characteristics. These results are shown in the section A. We then turn to analyze the role eventually played by each HH’s member’s individual characteristics. A. CORRELATES OF ECONOMIC MOBILITY: THE ROLE OF HH INVARIANT AND HH HEAD’S CHARACTERISTICS 5.2 The first set of regressions investigated the correlates of economic mobility for each of the selected transition groups only focusing on HH invariant and HH head’s characteristics. The explanatory variables included in this part of the analysis are: HH size, HH dependency ratio, HH location (urban vs. rural), as well as HH income sources; gender, age, education, ethnicity, labor market status, as well as type and sector of employment, all referred to each HH’s head. 5.3 The analysis was carried out for the biennium 2010-11, based on HHs’ initial characteristics. Detailed findings from different model specifications are displayed in Tables A2- A5 in the Appendix, while the following Figures 5.1-5.3 summarize some of the main outcomes; as above, for every transition group we will distinguish between the role of HH invariant characteristics (panel (a) of each graph) and the one played by individual characteristics of the HH head (panel (b) of each graph). In each graph, blue histograms represent the coefficients associated with each displayed characteristic, while the black bars indicate the corresponding confidence intervals. 5.4 HH demographics highly significant for economic mobility. Unsurprisingly (and consistently with what we already observed in the previous unconditional evidence), HH size and HH dependency ratios turn out to be significantly correlated with economic mobility: (a) For households in the bottom 40% in 2010, the larger the size, the higher the probability of persisting in the poverty status in the following year (Figure 5.1, panel (a)). The same turns out to be true for HHs’ dependency ratios (but with an even stronger coefficient). Symmetrically, smaller households with a lower share of dependent members are associated on average with a higher chance of getting out of the poverty 40% (Figure 5.2, panel (a)). (b) Similar (and opposite) results apply for households beyond the bottom 40% in 2010: larger households with higher dependency ratios do face on average a higher chance of moving down into the poverty pool. Figure 5.1: Correlates of economic mobility: persistence in the poverty pool Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Talas Remittances HH size Rural Dependency ratio Chui HH enterpr income -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH head's characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Occupation Age Tertiary educ Employed Mining Private services Dungan Tenure -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 5.5 HH location (whether in rural or urban regions) only significant for mobility of HHs who were not initially poor. Transitions of HHs who in 2010 were in the bottom 40% do not show any correlation with the different location of people between rural and urban areas (see panel (a) of 19 both Figures 5.1 and 5.2). On the contrary, they emerge as significant drivers of mobility for households who in 2010 were beyond the poverty threshold. More specifically, our evidence shows that non-poor households in 2010 who live in rural regions do indeed face a 10% higher risk of becoming poor in 2011. Figure 5.2: Correlates of economic mobility: moving out of poverty Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Rural Remittances HH size Chui Talas Dependency ratio HH enterpr income -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH head's characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Mining Occupation Tertiary educ Age Employed Tenure Dungan -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 5.6 Different results emerge instead if HH location is not broken down by rural/urban areas but rather by different regions/oblasts. In this case, indeed, HH transitions appear to be significantly correlated with some specific regions/oblasts: 20 (a) Poor HHs in 2010 living in the Talas oblast do indeed appear to have a 26% higher chance of remaining poor in the following year. The opposite is true for poor HHs living the Chui oblast, who display a 21% higher chance of moving out of poverty in 2011. (b) For non-poor HHs, instead, the probability of skipping poverty in the following year turns out to be significantly higher if they live in the Naryn, Batken, Chui and Bishkek oblasts. 5.7 Households’ reliance on remittances seems to play a positive role. Households’ availability of remittances coming from members working abroad appear to significantly contribute to avoid poverty: (a) For households in the bottom 40% in 2010, receiving remittances enhances their chance of leaving poverty in the following year by about 10 percentage points (Figure 5.2). (b) Receiving remittances do also appear to contribute positively and significantly to avoid poverty for households who were already beyond the poverty threshold in 2010, though to a lesser degree (about 5 percentage points) compared with households who were poor in 2010 (Figure 5.3). 5.8 No clear-cut correlation emerges instead with the availability of social benefits (see Box 5.1 for a brief overview on the system of social benefits in the Kyrgyz Republic ). Receiving social benefits as part of their income does not have an unambiguous correlation with HH mobility: (a) No correlation emerges between mobility/persistency of households who were poor in 2010 with respect to the availability of social transfers (neither pensions nor other types of social benefits). (b) On the contrary, the availability of social benefits (excluding pensions) seems to have enhanced the chance of non-poor households of staying out of poverty by slightly less than 10 percentage points. 5.9 HH head’s educational attainment as significant driver of economic mobility for all transition groups. The educational attainment of the HH head turns out to be significantly correlated with mobility for both those who were and those who were not initially poor: (a) For a household in the bottom 40% in 2010, having a head with a tertiary education reduces on average the probability of staying poor in the following year by about 15% points (compared with a HH whose head has less than a secondary level of education). (b) For a household beyond the poverty threshold in 2010, having a head with a tertiary education reduces the probability of becoming poor in 2011 by about 8 percentage points. 5.10 Gender (and age) of the HH head does not seem to show a significant correlation with household mobility, while some correlation appears with her/his age. Being employed, 21 instead, appears to be significantly linked to mobility outcomes, but only for households who were non-poor in 2010. The evidence displayed in Figures 5.1-5.3 shows indeed that: (a) For households who belonged to the bottom 40% in 2010, being employed (as opposed of being either unemployed or inactive) does not appear to influence the probability to exit poverty nor that of persisting into it. (b) For HHs beyond the 40% threshold, instead, having an employed head appears to reduce the probability to enter into poverty by about 7 percentage points. Box 5.1: Social Benefits in the Kyrgyz Republic The Kyrgyz Republic is characterized by a rather complex system of social benefits. The benefit system consists indeed of several pillars, the most relevant being the following ones: (i) the monthly social benefits (MSB); (ii) the monthly benefits for poor families with children (MBPF); (iii) the categorical benefits; and (iv) the energy benefits. MBPF is the only poverty focused, means-tested benefit, covering about one third of households from the first quintile, indicating less than perfect targeting efficiency. About half of the country population lives in a household receiving any type of social protection transfer. Old age and disability pensions are by far the most important programs, covering 43 percent of total population. All other social transfers, with the exception of energy compensations, are relatively small and reach between one and seven percent of the population. Private transfer, including remittances from migrants abroad, also play a major role given that one fifth of the population regularly receives money from relatives. In 2012, the Kyrgyz Republic spent about 5.7 percent of GDP on social protection programs; however, concerns remain with respect to the effectiveness of the programs. While Government social expenditures are growing, the performance of the system in terms of coverage of the poor, targeting accuracy, as well as poverty impact did not significantly improve. The increase in spending has favored programs largely benefitting individuals and households who are financially better off (Gassmann, 2013). This might be also reflected in panel data, where there is low correlation between social benefits and mobility. 22 Figure 5.3: Correlates of economic mobility: staying out of poverty Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Remittances HH size Rural Naryn Batken Chui Bishkek Dependency ratio Social benefits -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH head's characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Employed Uigur Occupation Public services Tertiary educ Private services Dungan -0.1 Tenure -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 5.11 Other labor market characteristics. Panel (b) of Graphs 8-10 also focuses on some selected labor market characteristics of the HH head in the attempt to investigate whether labor market outcomes do indeed represent potential determinants of the mobility observed between 2010 and 2011. In particular, results show that working in specific sectors of the economy as well as having a specific type of occupation can play a significant role. Indeed: (a) Poor households with a head working in the mining sector face a significantly higher probability of persisting in the poverty status in the following year. 23 (b) On the contrary, working in the service sector (both private and public) seems to enhance significantly the chance of skipping poverty for those who were already non- poor in 2010. (c) Some correlation finally emerges also between mobility and the type of occupation held by the HH head, though coefficients are generally rather small in magnitude. Data show in particular that the lower the skill level15 associated with the type of occupation the higher the probability for a household either to persist in poverty (Graph 8) or to fall into the poverty status in 2011 (Figure 5.3). B. CORRELATES OF ECONOMIC MOBILITY: THE ROLE OF INDIVIDUAL CHARACTERISTICS OF ALL MEMBERS INSIDE EACH HOUSEHOLD 5.12 We finally wanted to look at the correlates of economic mobility taking into account the relative contribution – in terms of their respective individual characteristics – of all adult HH members, in order to better understand the role played by education and labor market outcomes. To this end a linear probability model has been estimated by collapsing all relevant individual characteristics at the household level, so that the contribution of each member emerges as a relative share. Thus, for every household we will have, for example, the share of members with a tertiary educational attainment, or the proportion of members who are employed, etc. Detailed results of the different model specifications are shown in Table A6-A7 in the Appendix, while main findings are shown in Figures 5.4 and 5.5 below and can be summarized as follows. 5.13 Education is a key ingredient of economic mobility. Once the relative contribution of each HH members’ characteristics is explicitly taken into account, the evidence confirms and further reinforces the link between the educational attainment of each HH members and the outcome in term of economic mobility: (a) HHs in the poverty pool with a relative higher share of members with a tertiary educational attainment face a significantly higher chance (between 19 and 24 percentage points depending on the model specification; see Table A6) of leaving poverty in the following year. (b) Non-poor households with a tertiary educational attainment face a significantly lower probability (about 10-11 percentage points lower) of entering poverty in 2011. 5.14 Older people less likely to persist in poverty or entering into it. If HH mobility is broken down by the relative importance of the age groups of its members, it generally emerges a relatively more positive position of the oldest cohorts of the population. Indeed: 15 Employment data by different types of occupation are usually collected referring to the ISCO international classification of occupations, which classify jobs into 10 main groups of occupations, in decreasing order of the amount of skills deemed necessary to carry out the corresponding task. Thus the occupation variable ranges from 1 to 10, with 1 representing occupations with the highest skilled workers and 9 being the unskilled ones (the 10th group actually includes armed forces and was treated differently). 24 (a) Poor households with a relatively higher proportion of people older than 64 face a 40% higher chance to leave poverty in the following year. (b) Non-poor households have a 10% higher chance to remain non-poor the highest the incidence of member older than 64. At the same time, the higher the share of young members (25-34 years of age), the higher the probability for non-poor households to become poor in 2011. Figure 5.4: Economic mobility and HH members’ characteristics: persisting into poverty (vis-à-vis moving out of poverty) Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Remittances HH size Chui Dependency ratio Talas Pensions -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH members' characteristics 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 Tertiary educ Women Agriculture Mining Employed Employer Members 65+ Low-skilled job -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -0.8 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 25 Figure 5.5: Economic mobility and HH members’ characteristics: moving down into poverty (vis-à-vis staying out of poverty) Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Wages Remittances HH size Naryn Chui Bishkek Batken Dependency ratio Social benefits -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH members' characteristics 0.6 0.6 0.4 0.4 0.2 0.2 0 0 Private services Own-account worker Member 25-34 Tertiary educ Employer Retired Members 65+ -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 5.15 HH members’ labor market characteristics relevant but no clear-cut message seems to emerge so far. Once the labor market outcomes of each member of the HH are taken into consideration, our evidence generally confirms what we already found in the previous section, though some differences seem to emerge: 26 (a) Individual labor market characteristics do indeed confirm the existence of a significant correlation between HH economic mobility and the specific sector of employment of its members. Poor households with a higher share of members working in the agricultural and mining sectors are found to face a relatively higher chance of persisting in the poverty status, whereas households who are already non-poor and whose members are prevalently employed in the private service sector have a relatively higher probability of skipping poverty in the following year. (b) Being employed as opposed to be either unemployed or inactive does not seem to influence HH mobility in any of the four transition groups. The only partial exception is represented by non-poor households with a relatively higher presence of retired members, who face on average a relatively lower risk of falling into poverty in the following year. (c) Likewise, the specific wage status of the HH members, i.e. whether self-employed or employees, does not show a particularly significant correlation with HH mobility. (d) Having a low-skilled type of occupation turns instead to be significantly correlated with a relatively higher chance for poor households to persist into poverty. The same cannot be said symmetrically for non-poor households: for them no correlation seems to emerge between mobility and type of occupation carried out by HH members. 27 6. FOCUS ON PERSISTENCY INSIDE THE BOTTOM 40%: CHRONIC POVERTY? 6.1 The idea of this section is to focus on the mobility/persistency observed inside the poverty pool. In Table 4.1 we could indeed observe a relatively high mobility between the first and fourth deciles of the distribution, but also a relatively significant share of poor families stuck in the very first deciles of it. That is why we decided to devote a section of this study to the analysis of this specific segment of the population, with the aim to draw a first picture16 of who tends to be stuck in the lowest end of the welfare spectrum. We will focus in particular on those households who are found to persist in either the very first or the second deciles of the distribution as well as those who moved down from the second to the first decile between 2010 and 2011; we will call them here the “chronic poors”. 6.2 According to the DIW sample, this group of extremely poor households amounts to 131 families, which represents about one fourth of the entire poverty pool (537 households belonging to the bottom 40%). Detailed data show in particular that (Figure 6.1), slightly less than a half of them turn out to be stuck in the very first decile of the welfare spectrum, about 23% stayed in both years in second decile, while the remaining 29 per cent fell down from the second to the first decile of the distribution. Figure 6.1: Structure of households in the pool of chronic poverty, 2010-11 (%) 29.00763359 Stuck in 1. decile 48.1 Stuck in 2. decile Moved down from 2. to 1. decile 22.90076336 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 16This section is only intended as an introductory overview to the analysis of this specific segment of the poverty pool in order to evaluate whether it could be relevant to investigate it further. 6.3 Figure 6.2 provides some descriptives of the profile of the chronic poors, compared with the other households belonging to the poverty pool. Following the line of reasoning adopted in previous sections, data in panel (a) only refer to the role eventually played by HH invariant characteristics, while panel (b) focuses on the individual characteristics of the HH head. Figure 6.2: Profile of households in the bottom 40%: chronic vs. non-chronic poors Panel (a): HH invariant characteristics 80.0 70.0 Chronic poors 60.0 Non-chronic poors 50.0 40.0 30.0 20.0 10.0 0.0 Batken Naryn Osh city Bishkek HH dependency ratio (<14 & Village Wage employment HH average size City Chui Pensions Remittances* HH enterprises Property income Aid from people in Kyrgyzstan Other Issyk-Kul Osh Djalal-Abad Talas Other social transfers >65) Demogr. Rural/urban Geographical region/Oblast Source of income Panel (b): HH head's characteristics 70.0 60.0 50.0 Chronic poors 40.0 Non-chronic poors 30.0 20.0 10.0 0.0 Private services Secondary general Agriculture Employed Low-skilled Tertiary education Other High-skilled Mining Industry without construction Up to lower secondary Medium-skilled Construction Own-account worker Non-employed Primary and secondary technical Employee Public services Employer Member of producers' Contributing family workers cooperative Education Labor mkt Sector Wage status Skill level Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 6.4 HH size and dependency ratio. Our evidence further reinforces our previous findings that relatively larger families with a relatively higher incidence of children and older people do generally face a much higher probability of staying or becoming poor over time; this result also applied for the poorest segment of the population, who turn out to be characterized by larger 29 families and a higher number of dependents compared with the rest of households in the poverty pool (Figure 6.2). 6.5 Geographical location and main sources of income. Chronic poors are found to be located more often in urban areas (with respect to the other poors) and turn out to resort on average on a relatively higher share of wage income compared with other income sources (working poors?). It is also interesting to note that, compared with rest of households belonging to the bottom 40%, chronic poor households also claim to rely on a significantly higher proportion of income deriving from alimony, inheritance, and scholarships. If we indeed have a look at the individual characteristics of the HH head (panel (b)), we notice that compared with the rest of the poverty pool, chronic poors are more often non-employed (i.e. unemployed or inactive), whereas the employed ones are more often found to work in the mining and construction sectors (as well as in the public services). 6.6 Wage status and type of occupation. If we look at their specific wage status, we also find that they tend to work relatively more often as own-account workers and employees but the difference does not seem very significant. What could be interesting to better investigate is the fact that, according to our evidence chronic poors appear on average to be relatively better educated than the other poors; this is also confirmed by the distribution by type of occupation, with the chronic poors working relatively more in high-skilled positions compared with the rest of the poverty pool. 6.7 Multivariate analysis. In order to better understand and qualify this type of findings, we also carried out a preliminary regression analysis focusing on the correlation between the probability of belonging to the pool of chronic poors and the usual set of HH invariant and HH head’s characteristics we already introduced in the previous sections. 6.8 Results are shown in Table A8 in the Appendix, for three different model specifications, while main results are displayed in Figure 6.3. Though this only represents a preliminary and tentative analysis of the correlates of persistency into the pool of chronic poors, some interesting results seem indeed to emerge, which could deserve some future investigation. First of all, the probability of being stuck in the lower end of the welfare distribution looks strongly linked to the demographic structure of the poor households, with the most disadvantaged ones being especially characterized by a significantly higher number of dependent members (whereas the household size turns out to be only marginally larger). No significant difference seems to emerge with respect to the urban/rural divide, though chronic poors are relatively more likely to be found in the region of Talas. 6.9 Further results. Econometric evidence also confirms that chronic poors tend to rely relatively more on income sources such as alimonies, scholarships and inheritances. Their incidence among the non-employed is higher compared with the other poor, whereas the employed ones are more frequently found to work in mining and construction. As already observed in the descriptive section, no correlation seems to emerge between being chronic poor and the specific educational attainment of the HH head. 30 Figure 6.3: Correlates of probability of belonging to the pool of chronic poors Panel (a): HH invariant characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 HH size Rural Talas oblast Other incomes Dependency ratio -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Panel (b): HH head's characteristics 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Construction Tertiary educ Mining Industry Age Nonmployed -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 31 7. PRELIMINARY CONCLUSIONS AND STEPS AHEAD 7.1 This report aims at characterizing economic mobility in the Kyrgyz Republic, in the attempt of better understanding recent poverty dynamics and its determinants. It did so by resorting to a relatively new data-set of micro-data, made available by the Deutsches Institut fuer Wirtschaftsforschung (DIW) and consisting of “true” longitudinal data at the individual level, thereby allowing to carry out dynamic analyses of individual well-being and behavior. 7.2 Overall, data analysis was constrained by the availability so far of only two subsequent waves of micro-data, covering the years 2010 and 2011. As a consequence, only preliminary and partial conclusions are possible so far, and will be improved as soon as new data are made available. Nonetheless available data allow characterizing in some detail the different profile of households with different mobility patterns. 7.3 Throughout the paper, our focus has been on the economic mobility of households belonging to the bottom 40% of the population, which we assumed here to be our target group in terms of poverty reduction. As a consequence, four main transition groups were explicitly addressed in our analysis: (i) households persisting in the bottom 40% (and hence in poverty) in both 2010 and 2011; (ii) households moving out of bottom 40% in 2011 (and hence leaving poverty); (iii) households who entered the bottom 40% (and hence the poverty status) in 2011; and (iv) households staying above the bottom 40% (and hence skipping poverty) in both years. 7.4 With these transitions groups in mind, we drew a profile of each of them in terms of both households’ invariant characteristics (e.g. household size, dependency ratio, geographical location, etc.) and individual characteristics referred to each household’s head. We then turned to multivariate regression analysis to investigate in some more detail the correlates of economic mobility for each transition group in the attempt to highlight some of its key determinants. In particular, we were highly interested in studying the role eventually played by individual’s labor market outcomes. 7.5 Overall our evidence suggests that household invariant characteristics such as household size and dependency ratio are strongly associated with mobility in and out poverty, with larger households characterized by a relatively higher number of dependents facing a significantly higher chance of persisting in poverty or entering into it. The rural or urban location of a household, instead, does not seem to influence economic mobility in any direction (though some strong correlation seems to characterize some specific regions/oblasts of the country). The different origin of each household’s income shows some interesting correlation with economic mobility, though some clear-cut conclusions seem to emerge only for some selected income sources: receiving remittances, for example, is found to be associated on average with a relatively higher probability of leaving the poverty pool, but it also enhances the chance for non-poor households to stay non-poor in the following period; on the other hand, the availability of social benefits does not appear to influence either persistency into poverty or the movement out of it, whereas a significant correlation is found between social benefits and the probability of avoiding poverty in both years. 7.6 Turning to individual characteristics, education emerges as the key driver of economic mobility in the (short) period under analysis. Having a tertiary educational attainment (both referred to the household head and to the sum of adult members in each household) seems indeed to significantly enhance the probability of moving out of poverty, as well as that of persisting out of poverty in both years. 7.7 Some interesting results finally emerge when looking at individual’s labor market characteristics, though they do seem to generally play a key role in determining economic mobility between 2010 and 2011. In particular, our evidence does not suggest that being employed by itself does significantly enhance the chance of leaving the poverty pool. However, working in specific sectors as well as having a specific type of occupation appears to exert a significant role. Working for example in the mining and agricultural sector, enhances the probability of persisting in poverty, whereas being employed in the private service sector reduces the risk of becoming poor in the following year. 33 APPENDIX: FIGURES AND TABLES Table A1 - Life in Kyrgyzstan Survey 2010: Basic Sample Characteristics Source: Staff computations on DIW, Life in Kyrgyzstan Survey, 2010-11. 34 Table A2 – Probability of persisting in poverty in 2011, by HH invariant and HH head’s characteristics: different model specifications Model 1 Model 2 Model 3 Model 4 Stayed in bottom40% coeff. t coeff. t coeff. t coeff. t Female 0.0393 0.91 0.0359 0.84 0.0308 0.73 0.0439 0.66 HH head's socio- characteristics demographic Age -0.0167 -1.87 -0.0148 -1.66 -0.0144 -1.62 -0.0310 -1.75 Age2 0.0001 1.60 0.0001 1.28 0.0001 1.22 0.0003 1.54 Secondary general education 0.0333 0.67 0.0181 0.37 0.0249 0.51 -0.0573 -0.71 Technical education -0.0113 -0.19 -0.0181 -0.31 -0.0174 -0.29 -0.0538 -0.58 Tertiary education -0.1130 -1.53 -0.1438 -1.96 -0.1496 -2.11 -0.2110 -1.95 Rural location -0.0428 -0.99 HH geographical location Region 2: Djalal-Abad 0.0101 0.16 -0.0008 -0.01 0.0064 0.08 Region 3: Naryn -0.0822 -0.88 -0.0963 -1.03 -0.0563 -0.45 Region 4: Batken -0.0392 -0.54 -0.0184 -0.26 -0.0345 -0.41 Region 5: Osh 0.0456 0.70 0.0387 0.59 0.2256 2.83 Region 6: Talas 0.2635 3.16 0.2678 3.24 0.4016 4.19 Region 7: Chui -0.2131 -2.80 -0.2221 -2.93 -0.1529 -1.75 Region 8: Bishkek 0.0306 0.37 0.0053 0.06 0.0998 0.98 Region 9: Osh city 0.1080 0.97 0.0855 0.78 0.2773 2.21 demo ncomp10 0.0452 5.24 0.0459 5.29 0.0464 5.35 0.0529 4.01 Labor market characteristics of HH gr. depend_ratio10 0.1823 2.12 0.1924 2.25 0.1981 2.32 0.2903 2.30 Agriculture_10 0.0451 0.94 0.0249 0.52 Mining 0.2246 1.53 0.2740 1.86 Industry -0.0138 -0.14 0.0282 0.28 the HH head Construction 0.0031 0.04 0.0190 0.27 Private services -0.0887 -1.48 -0.0821 -1.40 Public services -0.0015 -0.02 -0.0009 -0.01 Employed 0.0031 0.08 Type of occupation 0.0209 1.43 Job tenure 0.0027 0.80 High-skilled 0.1051 0.99 Kyrgyz -0.0106 -0.09 -0.1284 -1.03 -0.1370 -1.10 HH head's ethnicity Uzbek 0.0542 0.42 -0.0936 -0.70 -0.1106 -0.83 Russian 0.0698 0.50 0.0067 0.05 0.0028 0.02 Dungan 0.1810 1.22 0.2611 1.76 0.2548 1.72 Uigur 0.2111 1.37 0.0820 0.53 0.0713 0.46 Tajik 0.0180 0.07 0.0185 0.07 0.0148 0.06 Kazakh -0.3918 -1.60 -0.3723 -1.54 -0.4064 -1.68 HH enterprise 0.0948 2.38 0.0526 1.32 0.0500 1.28 0.0314 0.58 HH sources of Wages 0.0618 1.68 0.0426 1.16 0.0501 1.39 0.0104 0.21 income Pension 0.0546 1.21 0.0553 1.24 0.0656 1.47 0.0239 0.40 Other social benefits 0.0150 0.32 -0.0154 -0.33 -0.0128 -0.27 -0.0892 -1.49 Remittances -0.0728 -1.51 -0.0983 -2.02 -0.1011 -2.09 -0.1474 -2.10 _cons 0.6388 2.37 0.7395 2.74 0.7366 2.73 0.7816 1.78 Source: Own computations on DIW, Life in Kyrgyzstan Survey 2010-2011. 35 Table A3 – Probability of moving out of poverty in 2011, by HH invariant and HH head’s characteristics: different model specifications Model 1 Model 2 Model 3 Model 4 Moved out of poverty coeff. t coeff. t coeff. t coeff. t Female -0.0393 -0.91 -0.0359 -0.84 -0.0308 -0.73 -0.0439 -0.66 HH head's socio- characteristics demographic Age 0.0167 1.87 0.0148 1.66 0.0144 1.62 0.0310 1.75 Age 2 -0.0001 -1.60 -0.0001 -1.28 -0.0001 -1.22 -0.0003 -1.54 Secondary general education -0.0333 -0.67 -0.0181 -0.37 -0.0249 -0.51 0.0573 0.71 Technical education 0.0113 0.19 0.0181 0.31 0.0174 0.29 0.0538 0.58 Tertiary education 0.1130 1.53 0.1438 1.96 0.1496 2.11 0.2110 1.95 Rural location 0.0428 0.99 HH geographical location Region 2: Djalal-Abad -0.0101 -0.16 0.0008 0.01 -0.0064 -0.08 Region 3: Naryn 0.0822 0.88 0.0963 1.03 0.0563 0.45 Region 4: Batken 0.0392 0.54 0.0184 0.26 0.0345 0.41 Region 5: Osh -0.0456 -0.70 -0.0387 -0.59 -0.2256 -2.83 Region 6: Talas -0.2635 -3.16 -0.2678 -3.24 -0.4016 -4.19 Region 7: Chui 0.2131 2.80 0.2221 2.93 0.1529 1.75 Region 8: Bishkek -0.0306 -0.37 -0.0053 -0.06 -0.0998 -0.98 Region 9: Osh city -0.1080 -0.97 -0.0855 -0.78 -0.2773 -2.21 demo ncomp10 -0.0452 -5.24 -0.0459 -5.29 -0.0464 -5.35 -0.0529 -4.01 Labor market characteristics of HH gr. depend_ratio10 -0.1823 -2.12 -0.1924 -2.25 -0.1981 -2.32 -0.2903 -2.30 Agriculture_10 -0.0451 -0.94 -0.0249 -0.52 Mining -0.2246 -1.53 -0.2740 -1.86 Industry 0.0138 0.14 -0.0282 -0.28 the HH head Construction -0.0031 -0.04 -0.0190 -0.27 Private services 0.0887 1.48 0.0821 1.40 Public services 0.0015 0.02 0.0009 0.01 Employed -0.0031 -0.08 Type of occupation -0.0209 -1.43 Job tenure -0.0027 -0.80 High-skilled -0.1051 -0.99 Kyrgyz 0.0106 0.09 0.1284 1.03 0.1370 1.10 HH head's ethnicity Uzbek -0.0542 -0.42 0.0936 0.70 0.1106 0.83 Russian -0.0698 -0.50 -0.0067 -0.05 -0.0028 -0.02 Dungan -0.1810 -1.22 -0.2611 -1.76 -0.2548 -1.72 Uigur -0.2111 -1.37 -0.0820 -0.53 -0.0713 -0.46 Tajik -0.0180 -0.07 -0.0185 -0.07 -0.0148 -0.06 Kazakh 0.3918 1.60 0.3723 1.54 0.4064 1.68 HH enterprise -0.0948 -2.38 -0.0526 -1.32 -0.0500 -1.28 -0.0314 -0.58 HH sources of Wages -0.0618 -1.68 -0.0426 -1.16 -0.0501 -1.39 -0.0104 -0.21 income Pension -0.0546 -1.21 -0.0553 -1.24 -0.0656 -1.47 -0.0239 -0.40 Other social benefits -0.0150 -0.32 0.0154 0.33 0.0128 0.27 0.0892 1.49 Remittances 0.0728 1.51 0.0983 2.02 0.1011 2.09 0.1474 2.10 _cons 0.3612 1.34 0.2605 0.96 0.2634 0.98 0.2184 0.50 Source: Own computations on DIW, Life in Kyrgyzstan Survey 2010-2011. 36 Table A4 – Probability of moving down into poverty in 2011, by HH invariant and HH head’s characteristics: different model specifications Model 1 Model 2 Model 3 Model 4 Moved down into poverty coeff. t coeff. t coeff. t coeff. t Female 0.0225 1.10 0.0251 1.25 0.0214 1.07 0.0445 1.66 HH head's socio- characteristics demographic Age -0.0010 -0.26 -0.0026 -0.67 -0.0025 -0.64 -0.0045 -0.78 Age2 0.0000 -0.18 0.0000 0.21 0.0000 0.19 0.0000 0.51 Secondary general education 0.0105 0.35 0.0059 0.19 0.0048 0.16 -0.0057 -0.11 Technical education -0.0472 -1.47 -0.0483 -1.52 -0.0529 -1.67 -0.0301 -0.58 Tertiary education -0.0774 -2.28 -0.0813 -2.39 -0.0869 -2.58 -0.0683 -1.23 Rural location 0.0879 4.22 HH geographical location Region 2: Djalal-Abad -0.0244 -0.65 -0.0283 -0.76 -0.0924 -1.97 Region 3: Naryn -0.2369 -4.87 -0.2309 -4.76 -0.2136 -3.36 Region 4: Batken -0.1851 -4.12 -0.1855 -4.13 -0.1875 -3.45 Region 5: Osh 0.0203 0.52 0.0183 0.47 0.0541 1.10 Region 6: Talas -0.0424 -0.75 -0.0399 -0.70 -0.0630 -0.96 Region 7: Chui -0.1399 -3.89 -0.1425 -3.98 -0.1592 -3.74 Region 8: Bishkek -0.1879 -5.51 -0.1983 -5.91 -0.2381 -5.77 Region 9: Osh city -0.0476 -0.96 -0.0609 -1.24 -0.0620 -1.06 demo ncomp10 0.0425 8.11 0.0413 7.89 0.0420 8.02 0.0284 3.94 Labor market characteristics of HH gr. depend_ratio10 0.0916 2.43 0.0713 1.91 0.0716 1.92 0.1434 2.48 Agriculture_10 -0.0459 -1.50 -0.0149 -0.50 Mining -0.0299 -0.32 -0.0785 -0.86 Industry -0.0566 -1.36 -0.0660 -1.61 the HH head Construction -0.0330 -0.73 -0.0158 -0.35 Private services -0.0892 -3.18 -0.0817 -2.95 Public services -0.0598 -1.90 -0.0575 -1.85 Employed -0.0505 -2.26 Type of occupation 0.0115 1.65 Job tenure -0.0003 -0.27 High-skilled 0.0249 0.59 Kyrgyz 0.0719 1.72 0.0325 0.75 0.0301 0.70 HH head's ethnicity Uzbek 0.0317 0.64 -0.0508 -0.98 -0.0568 -1.10 Russian 0.0652 1.42 0.0458 1.00 0.0457 1.01 Dungan 0.2097 3.07 0.2539 3.76 0.2536 3.76 Uigur 0.2985 3.58 0.2697 3.26 0.2724 3.30 Tajik -0.0373 -0.44 -0.0596 -0.68 -0.0614 -0.70 Kazakh -0.0050 -0.04 -0.0214 -0.19 -0.0285 -0.25 HH enterprise -0.0040 -0.18 -0.0095 -0.43 -0.0066 -0.30 -0.0451 -1.57 HH sources of Wages -0.0290 -1.32 -0.0293 -1.35 -0.0310 -1.46 -0.0477 -1.67 income Pension -0.0367 -1.52 -0.0294 -1.23 -0.0281 -1.18 -0.0087 -0.30 Other social benefits -0.0909 -2.47 -0.0987 -2.69 -0.0983 -2.68 -0.1559 -3.05 Remittances -0.0346 -1.32 -0.0581 -2.21 -0.0564 -2.15 -0.0239 -0.68 _cons 0.0477 0.43 0.2834 2.46 0.2832 2.48 0.3050 1.92 37 Table A5– Probability of staying/persisting out of poverty in 2011, by HH invariant and HH head’s characteristics: different model specifications Model 1 Model 2 Model 3 Model 4 Staying out of poverty coeff. t coeff. t coeff. t coeff. t Female -0.023 -1.10 -0.0251 -1.25 -0.0214 -1.07 -0.0445 -1.66 HH head's socio- characteristics demographic Age 0.001 0.26 0.0026 0.67 0.0025 0.64 0.0045 0.78 Age2 0.000 0.18 0.0000 -0.21 0.0000 -0.19 0.0000 -0.51 Secondary general education -0.011 -0.35 -0.0059 -0.19 -0.0048 -0.16 0.0057 0.11 Technical education 0.047 1.47 0.0483 1.52 0.0529 1.67 0.0301 0.58 Tertiary education 0.077 2.28 0.0813 2.39 0.0869 2.58 0.0683 1.23 Rural location -0.088 -4.22 HH geographical location Region 2: Djalal-Abad 0.0244 0.65 0.0283 0.76 0.0924 1.97 Region 3: Naryn 0.2369 4.87 0.2309 4.76 0.2136 3.36 Region 4: Batken 0.1851 4.12 0.1855 4.13 0.1875 3.45 Region 5: Osh -0.0203 -0.52 -0.0183 -0.47 -0.0541 -1.10 Region 6: Talas 0.0424 0.75 0.0399 0.70 0.0630 0.96 Region 7: Chui 0.1399 3.89 0.1425 3.98 0.1592 3.74 Region 8: Bishkek 0.1879 5.51 0.1983 5.91 0.2381 5.77 Region 9: Osh city 0.0476 0.96 0.0609 1.24 0.0620 1.06 demo ncomp10 -0.042 -8.11 -0.0413 -7.89 -0.0420 -8.02 -0.0284 -3.94 Labor market characteristics of HH gr. depend_ratio10 -0.092 -2.43 -0.0713 -1.91 -0.0716 -1.92 -0.1434 -2.48 Agriculture_10 0.046 1.50 0.0149 0.50 Mining 0.030 0.32 0.0785 0.86 Industry 0.057 1.36 0.0660 1.61 the HH head Construction 0.033 0.73 0.0158 0.35 Private services 0.089 3.18 0.0817 2.95 Public services 0.060 1.90 0.0575 1.85 Employed 0.0505 2.26 Type of occupation -0.0115 -1.65 Job tenure 0.0003 0.27 High-skilled -0.0249 -0.59 Kyrgyz -0.072 -1.72 -0.0325 -0.75 -0.0301 -0.70 HH head's ethnicity Uzbek -0.032 -0.64 0.0508 0.98 0.0568 1.10 Russian -0.065 -1.42 -0.0458 -1.00 -0.0457 -1.01 Dungan -0.210 -3.07 -0.2539 -3.76 -0.2536 -3.76 Uigur -0.298 -3.58 -0.2697 -3.26 -0.2724 -3.30 Tajik 0.037 0.44 0.0596 0.68 0.0614 0.70 Kazakh 0.005 0.04 0.0214 0.19 0.0285 0.25 HH enterprise 0.004 0.18 0.0095 0.43 0.0066 0.30 0.0451 1.57 HH sources of Wages 0.029 1.32 0.0293 1.35 0.0310 1.46 0.0477 1.67 income Pension 0.037 1.52 0.0294 1.23 0.0281 1.18 0.0087 0.30 Other social benefits 0.091 2.47 0.0987 2.69 0.0983 2.68 0.1559 3.05 Remittances 0.035 1.32 0.0581 2.21 0.0564 2.15 0.0239 0.68 _cons 0.952 8.62 0.7166 6.22 0.7168 6.27 0.6950 4.36 Source: Own computations on DIW, Life in Kyrgyzstan Survey 2010-2011. 38 Table A6– Probability of persisting into poverty, by HH invariant and individual characteristics: model specifications Model 1 Model 2 Model 3 Model 4 Stayed in bottom40% coeff. t coeff. t coeff. t coeff. t Female -0.1132 -1.19 -0.1375 -1.49 -0.0642 -0.70 -0.0577 -0.63 HH head's socio-demographic Age 25-34 -0.0583 -0.65 -0.0524 -0.59 -0.0292 -0.32 -0.0334 -0.37 Age 35-44 -0.1151 -1.23 -0.1098 -1.18 -0.1027 -1.09 -0.0948 -1.02 characteristics Age 45-54 -0.1366 -1.22 -0.1253 -1.12 -0.1252 -1.11 -0.1332 -1.19 Age 55-64 -0.1957 -1.56 -0.1988 -1.59 -0.1241 -0.97 -0.1484 -1.18 Age 65 and more -0.4311 -2.75 -0.4396 -2.81 -0.2898 -1.53 -0.4177 -2.67 Secondary general education 0.0445 0.65 0.0518 0.76 0.0178 0.26 0.0040 0.06 Technical education -0.0476 -0.53 -0.0545 -0.60 -0.0731 -0.83 -0.0694 -0.79 Tertiary education -0.1891 -1.97 -0.2000 -2.15 -0.2421 -2.60 -0.2140 -2.33 Region 2: Djalal-Abad -0.0018 -0.03 -0.0070 -0.11 -0.0056 -0.09 -0.0016 -0.03 HH geographical location Region 3: Naryn -0.1192 -1.28 -0.1341 -1.45 -0.1432 -1.53 -0.1318 -1.41 Region 4: Batken -0.0618 -0.85 -0.0379 -0.53 -0.0444 -0.61 -0.0429 -0.60 Region 5: Osh 0.0411 0.64 0.0361 0.56 0.0536 0.85 0.0617 0.98 Region 6: Talas 0.2587 3.21 0.2608 3.27 0.2583 3.16 0.2676 3.31 Region 7: Chui -0.2116 -2.84 -0.2203 -2.98 -0.1006 -1.43 -0.1010 -1.44 Region 8: Bishkek 0.0396 0.46 0.0237 0.29 0.0902 1.10 0.0949 1.17 Region 9: Osh city 0.1065 0.97 0.0912 0.84 0.1058 1.07 0.1076 1.09 demogr. ncomp10 0.0362 4.16 0.0365 4.20 0.0361 4.13 0.0376 4.36 HH depend_ratio10 0.4416 4.02 0.4513 4.13 0.4127 3.76 0.4060 3.69 Agriculture_10 0.1228 1.88 Mining 0.5258 1.84 Labor market characteristics of the HH head Industry 0.0683 0.43 Construction 0.1069 0.90 Private services 0.0517 0.56 Public services 0.0491 0.49 Employed 0.0940 1.68 Employer -0.7841 -1.85 Ownaccount worker 0.0580 0.73 Employee 0.1215 1.56 Member of coop 0.8606 1.42 Family worker 0.0733 0.72 Retired -0.1579 -1.12 Low-skilled 0.1153 2.08 Kyrgyz -0.1931 -2.29 -0.1991 -2.38 HH head's ethnicity Uzbek -0.1631 -1.62 -0.1708 -1.71 Russian 0.0000 0.0000 Kazakh -0.1553 -0.41 -0.1728 -0.46 HH enterprise 0.0206 0.51 0.0174 0.44 0.0360 0.85 0.0286 0.73 HH sources of Wages 0.0227 0.60 0.0232 0.64 0.0110 0.28 0.0346 0.97 income Pension 0.0863 2.00 0.0937 2.19 0.0970 2.23 0.0892 2.08 Other social benefits -0.0154 -0.33 -0.0144 -0.31 -0.0103 -0.22 -0.0246 -0.53 Remittances -0.0873 -1.86 -0.0927 -1.98 -0.0924 -1.96 -0.0948 -2.02 39 Table A7– Probability of entering poverty in 2011, by HH invariant and members’ individual characteristics: different specifications Model 1 Model 2 Model 3 Model 4 Entering bottom40% coeff. t coeff. t coeff. t coeff. t demographic characteristics Female -0.0025 -0.07 -0.0167 -0.44 -0.0145 -0.38 -0.0162 -0.43 Age 25-34 0.0816 1.76 0.0804 1.74 0.0851 1.84 0.0777 1.69 HH head's socio- Age 35-44 0.0278 0.60 0.0256 0.55 0.0313 0.67 0.0154 0.33 Age 45-54 0.0607 1.23 0.0568 1.15 0.0643 1.31 0.0546 1.11 Age 55-64 -0.0392 -0.73 -0.0372 -0.69 -0.0034 -0.06 -0.0265 -0.50 Age 65 and more -0.1536 -1.95 -0.1534 -1.95 -0.0948 -1.13 -0.1441 -1.84 Secondary general education 0.0250 0.67 0.0258 0.69 0.0413 1.11 0.0373 1.00 Technical education -0.0591 -1.47 -0.0714 -1.78 -0.0433 -1.09 -0.0539 -1.36 Tertiary education -0.0982 -2.35 -0.1127 -2.74 -0.0878 -2.13 -0.1011 -2.48 Region 2: Djalal-Abad -0.0123 -0.33 -0.0163 -0.44 -0.0466 -1.27 -0.0423 -1.16 Region 3: Naryn -0.2323 -4.75 -0.2288 -4.68 -0.2311 -4.73 -0.2347 -4.80 HH geographical Region 4: Batken -0.2077 -4.70 -0.2097 -4.74 -0.2195 -4.98 -0.2148 -4.87 location Region 5: Osh 0.0387 0.99 0.0356 0.91 0.0154 0.40 0.0165 0.42 Region 6: Talas -0.0225 -0.40 -0.0208 -0.37 -0.0258 -0.46 -0.0239 -0.42 Region 7: Chui -0.1210 -3.37 -0.1250 -3.50 -0.1046 -3.01 -0.1087 -3.14 Region 8: Bishkek -0.1675 -4.87 -0.1818 -5.39 -0.1718 -5.13 -0.1766 -5.29 Region 9: Osh city -0.0265 -0.53 -0.0421 -0.85 -0.0643 -1.32 -0.0706 -1.45 demo ncomp10 0.0322 5.57 0.0333 5.77 0.0307 5.28 0.0328 5.78 Labor market characteristics of the HH head HH gr. depend_ratio10 0.1662 3.00 0.1686 3.04 0.1880 3.39 0.1773 3.20 Agriculture_10 0.0384 0.92 Mining -0.1161 -0.67 Industry -0.0277 -0.47 Construction 0.0890 1.11 Private services -0.0997 -2.46 Public services -0.0565 -1.26 Employed -0.0285 -0.92 Employer -0.2617 -1.62 Ownaccount worker -0.0793 -1.85 Employee -0.0262 -0.66 Member of coop -0.3171 -0.98 Family worker 0.0600 0.87 Retired -0.1007 -1.89 Low-skilled -0.0023 -0.08 Kyrgyz 0.0909 2.87 0.0981 3.12 HH sources of HH head's ethnicity Uzbek 0.0000 0.0000 Russian 0.1214 2.94 0.1321 3.21 Kazakh 0.0833 0.69 0.0884 0.73 HH enterprise -0.0139 -0.61 -0.0103 -0.46 -0.0078 -0.33 -0.0175 -0.81 Wages -0.0339 -1.46 -0.0389 -1.76 -0.0478 -1.95 -0.0445 -2.16 income Pension -0.0006 -0.03 0.0023 0.10 0.0101 0.42 0.0070 0.30 Other social benefits -0.1013 -2.77 -0.0974 -2.66 -0.0951 -2.60 -0.0918 -2.51 Remittances -0.0647 -2.53 -0.0638 -2.49 -0.0620 -2.42 -0.0607 -2.37 40 Table A8– Probability of belonging to the chronic poverty pool in 2011, by HH invariant and HH members’ individual characteristics: different model specifications Model 1 Model 2 Model 3 Chronic poverty coeff. t coeff. t coeff. t Female -0.0432 -0.86 -0.0388 -0.77 -0.0388 -0.77 HH head's socio- characteristics demographic Age 0.0176 1.74 0.0173 1.71 0.0173 1.71 Age2 -0.0002 -1.81 -0.0002 -1.85 -0.0002 -1.85 Secondary general education 0.0543 0.96 0.0479 0.85 0.0479 0.85 Technical education -0.0074 -0.11 -0.0082 -0.12 -0.0082 -0.12 Tertiary education 0.0526 0.55 0.0402 0.42 0.0402 0.42 Rural location -0.0683 -1.32 HH geographical location Region 2: Djalal-Abad -0.0790 -1.05 -0.0790 -1.05 Region 3: Naryn -0.1440 -1.24 -0.1440 -1.24 Region 4: Batken 0.0623 0.72 0.0623 0.72 Region 5: Osh -0.0117 -0.15 -0.0117 -0.15 Region 6: Talas 0.2348 2.62 0.2348 2.62 Region 7: Chui -0.0187 -0.17 -0.0187 -0.17 Region 8: Bishkek 0.0163 0.16 0.0163 0.16 Region 9: Osh city 0.0303 0.25 0.0303 0.25 demog ncomp10 0.0128 1.29 0.0169 1.68 0.0169 1.68 Labor market characteristics of HH r. depend_ratio10 0.2282 2.08 0.2553 2.32 0.2553 2.32 Agriculture_10 -0.0357 -0.67 -0.0873 -1.63 0.1364 1.16 Mining 0.3068 2.10 0.3031 2.05 0.5268 2.94 Industry -0.2570 -2.18 -0.2237 -1.91 0.0000 the HH head Construction -0.0078 -0.10 -0.0042 -0.05 0.2195 1.71 Private services -0.0908 -1.23 -0.0869 -1.19 0.1368 1.09 Public services 0.0063 0.07 -0.0095 -0.11 0.2142 1.62 Nonemployed 0.2237 1.91 Type of occupation Job tenure High-skilled Kyrgyz 0.0448 0.30 0.0077 0.05 0.0077 0.05 HH head's ethnicity Uzbek 0.0203 0.13 0.0187 0.11 0.0187 0.11 Russian 0.0467 0.27 0.0130 0.07 0.0130 0.07 Dungan 0.2547 1.48 0.2197 1.26 0.2197 1.26 Uigur 0.2880 1.64 0.2632 1.42 0.2632 1.42 Tajik -0.0804 -0.24 -0.1025 -0.31 -0.1025 -0.31 Kazakh -0.2473 -0.55 -0.2285 -0.51 -0.2285 -0.51 HH enterprise 0.0178 0.37 -0.0137 -0.28 -0.0137 -0.28 HH sources of Wages 0.0175 0.41 -0.0088 -0.20 -0.0088 -0.20 income Pension 0.0024 0.05 0.0043 0.08 0.0043 0.08 Other social benefits -0.0495 -0.96 -0.0810 -1.53 -0.0810 -1.53 Remittances -0.0232 -0.40 -0.0249 -0.43 -0.0249 -0.43 Other (alimony, scholarships) 0.3214 3.67 0.2915 3.29 0.2915 3.29 41 REFERENCES Cancho C. et al., 2015. Economic Mobility in Europe and Central Asia. Exploring Patterns and Uncovering Puzzles. Policy Research Working Papers, # 7173, January. Davalos Maria E. and Moritz Meyer, 2015. Moldova. A Story of Upward Economic Mobility. Policy Research Working Papers, # 7167, January. The World Bank Group. DIW (2010), “Codebook for the ‘Life in Kyrgyzstan’ Survey”. Gassman Franziska, 2013. Kyrgyz Republic: Minimum Living Standards and Alternative Targeting methods for Social Transfers: A Policy Note. World Bank, Washington DC. Yang, Judy, Aibek Uulu and Sarosh Sattar, 2015. Labor Migration and Household Welfare in Kyrgyzstan. The World Bank.