LABOR MIGRATION AND WELFARE IN THE KYRGYZ REPUBLIC ( )

Transcription

1 Report No KG LABOR MIGRATION AND WELFARE IN THE KYRGYZ REPUBLIC ( ) May 8, 2015 Poverty Global Practice Europe and Central Asia Region Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Document of the World Bank

2 CURRENCY AND EQUIVALENT UNITS Exchange Rate Effective as of April 13, 2015 Currency Unit = Som (KGS) US$1 = 63.9 Som FISCAL YEAR January 1 December 31 ACRONYMS AND ABBREVIATIONS ADB Asian Development Bank KGZ the Kyrgyz Republic CALISS Central Asia Longitudinal Inclusive KIHS Kyrgyz Integrated Household Survey Society Survey (World Bank) CIS Commonwealth of Independent States LAC Latin America and the Caribbean ECA Europe and Central Asia LFS Kyrgyz Labor Force Survey ECAPOV Europe and Central Asia Poverty Data NSC National Statistics Committee of the Kyrgyz Republic EDB Eurasian Development Bank PPP Purchasing Power Parity EEU Eurasian Economic Union RUS Russia EU European Union GDP Gross Domestic Product SA Social Assistance HH Household TJK Tajikistan KAZ Kazakhstan USD U.S. Dollars KGS Kyrgyz Som Vice President : Laura Tuck Country Director : Saroj Kumar Jha Country Manager : Jean-Michel Happi Practice Director : Ana Revenga Practice Manager : Carolina Sanchez-Paramo Task Leaders : Sarosh Sattar (GPVDR) ii

3 TABLE OF CONTENTS EXECUTIVE SUMMARY... vi CHAPTER 1 INTRODUCTION... 1 CHAPTER 2 PROFILING LABOR MIGRANTS... 3 A. THE NUMBER OF LABOR MIGRANTS... 3 B. CHARACTERISTICS OF LABOR MIGRANTS... 5 CHAPTER 3 HOUSEHOLDS AND HOUSEHOLD MEMBERS A. REGIONAL AND ECONOMIC DISTRIBUTION OF HOUSEHOLDS B. CHARACTERISTICS OF THE HOUSEHOLD AND HOUSEHOLD MEMBERS CHAPTER 4 THE WELFARE IMPACT OF LABOR MIGRATION A. INCOME B. EXPENDITURES C. POVERTY CHAPTER 5 CONCLUSION ANNEX 1.2: KIHS Data ANNEX 2: Panel Households ( ) ANNEX 3: Impact of Rotation on Survey Results REFERENCES List of Tables Table 3.1: Differences in Household Characteristics (non-migrants migrant) Table 3.2: Characteristics of domestically employed workers Table 3.3: Employment of Domestically Located Household Members Table 4.1: Trends in Remittance Income, Migrant Households Table 4.2: Predictors of Per Capital Daily Consumption List of Figures Figure 2.1: Number of Labor Migrants... 3 Figure 2.2: Employed Workers (15 years of age +), by location of work... 4 Figure 2.3: Age Distribution, by location of work (2013)... 6 Figure 2.4: Educational Distribution, by location of work (2008 & 2013)... 6 Figure 2.5: Tertiary Enrollment Rates... 6 Figure 2.6: Profession Distribution of Labor Migrants Abroad, Figure 2.7: Distribution of Profession by location of work, Figure 2.8: Average hourly wages in the Previous Month (2008 & 2013)... 9 Figure 2.9: Increasing domestic wages, selected sector... 9 iii

4 Figure 2.10: Declining wage premium from working abroad Figure 2.11: Percent of Firms Identifying an Adequately Educated Workforce as a Major Constraint Figure 2.12: Average Monthly Wage Income of Labor Migrants Abroad Figure 3.1: Migrant and Non-Migrant Households Figure 3.2: Households with Labor Migrants Abroad, Rural-Urban Figure 3.3: Percent of Households with Labor Migrants Abroad, by Region Figure 3.4: Employment Rate of Non-Migrant Head of Households (15-64 years of age) Figure 3.5: Employment and Enrollment Rates of year olds Figure 3.6: Number of Migrant Households, by income quintile Figure 4.1: Average Annual household income per capita (ECAPOV, 2005 KGS) Figure 4.2: Income Ratio (Non-Migrant Household / Migrant Household) Figure 4.3: Average Monthly Pension Income HH Heads Figure 4.4: Total remittance income (2005,PPP) Figure 4.5: Annual remittance income per capita, by pre-migration income quintile Figure 4.6: Income Decomposition, Migrant Households Only (2008 to 2013) Figure 4.7: Average Household Expenditures and Savings per capita (NSC, 2005 KGS) Figure 4.8: Average Migrant Education Expenditures per capita (2005 KGS) Figure 4.9: Average Migrant Household Expenditures and Savings per capita Figure 4.10: Shared Prosperity Figure 4.11: Regional Poverty and Migration Rates (2013) Figure 4.12: National Absolute Poverty Rate ( ) Figure 4.13: National Absolute Poverty Rate, by Rural/Urban ( ) iv

5 ACKNOWLEDGMENTS This report is part of the Kyrgyz Republic poverty analysis program led by Sarosh Sattar (Task Team Leader). This report is principally authored by Judy S. Yang (ETC, GPVDR), and supervised by Sarosh Sattar (Senior Economist, GPVDR). Valuable comments were received from Aibek Baibagysh Uulu (STC, GPVDR), Saida Ismailakhunova (Economist, GPVDR), Jennifer Keller (Senior Economist, GPVDR), and Erwin Tiongson (Professor, Georgetown University). This note also benefitted from comments from an internal BBL and meetings with local migration related agencies in Bishkek. The team is grateful to the National Statistics Committee for the access to the Kyrgyz Republic s Integrated Household Survey, and the Kyrgyz Republic s Labor Force survey that were drawn upon for the analysis within. The team is also grateful for consultations with the government officials of the Kyrgyz Republic, primarily with the representatives of the Ministry of the Economy, Ministry of Labor, and the National Statistics Committee of The Kyrgyz Republic. The team is grateful to the UK s Department for International Development (DFID) for their financing of the CAPPAS program in Kyrgyz Republic that made this report possible. Any errors and omissions remain the responsibility of the authors. v

6 EXECUTIVE SUMMARY This paper examines the impact of labor migration 1 from a welfare and social development perspective. Rather than focusing on regulatory and legal aspects determining migration, this note centers on the impacts of migration on the domestic welfare of households 2 in the Kyrgyz Republic. The profiling of labor migration and identification of knowledge gaps are used to inform the development of strategies for more effective and sustainable welfare impacts from labor migration and remittances. Trends The report begins by taking stock of the number and characteristics of labor migrants. What types of individuals are leaving the country? Since labor migration is fundamentally a household decision, the characteristics of the household are examined next. Are migrant households benefitting from remittances and labor migration? How are households faring in terms of housing and the education of children? How are household members faring in terms of their labor choices, wages, and what are the educational choices of the youth? Lastly, what are the monetary benefits of labor migration? Are households with migrants better off than households without migrants 3? Labor migration has been an increasingly large phenomenon beginning in the early 2000s. During the period where survey data is available 4, roughly 200,000 labor migrants are counted in the survey, a number much lower than other estimates 5, but still about 10 percent of the working population. Labor migrants are not a random segment of the population, and are motivated by economic considerations. Minimum wages in the Kyrgyz Republic is the lowest among all CIS countries, even lower than its poorer neighbor Tajikistan. Similar to other countries, migrants are mostly young males with low education and from the poorer southern rural regions. They, work in low-skilled occupations and over time skilled manufacturing workers have reduced while unskilled workers are more common. Since 2008, hourly wages abroad (by sector and profession) have slowly declined, while real wages domestically has increased. These two factors lead to a declining wage differential from working abroad. This trend is not fully explained by changes in demographic characteristics. A declining wage differential can strongly impact incentives to migrate as well as remittance income. In fact in the data, we do see a decline in the number of labor migrants in recent years, however, it is not clear if either of these trends will be long-term. Migrant households have low levels of domestic sources of income and social assistance. In recent years, households with members who are labor migrants are becoming more dependent on remittance income, and therefore more vulnerable to economic shocks. For example in 2008, 34.4 percent of total income in migrant households were from remittances compared to 46.1 percent in It appears at least from the data that 1 Labor migration in this report refers to migration abroad, and in the case of the Kyrgyz Republic, migration mainly to Russia. The welfare of internal migration is not extensively discussed. Since there is no official migration survey, this report uses publicly available Labor Force and Household Budget Surveys. This obviously yields technical measurement and representativeness challenges that are discussed in detail in the Annex. 2 Households are differentiated between being migrant and non-migrant, which has some advantages over classifying households as either remittance or non-remittance receiving. Differentiating households by those who have migrants, takes into account that a labor migrant reduces the domestic labor supply of a household and that some households with migrants may not receive remittances at all. Therefore, this captures a more holistic measurement of the welfare impacts from labor migration. 3 For this question, it is important to take into account the selection process of becoming migrant households. 4 The primary data source for this report is the Kyrgyz Integrated Household Survey At the time of this report the 2014 data is not yet available but would reveal interesting trends in light of recent regional economic challenges. 5 Other estimates range from 500,000 to even a million labor migrants. vi

7 there is an increasing divide between households that receive enough income through domestic channels and do not need any additional sources of income, and households who are becoming more dependent on remittances as an income source and safety net. Poorer regions have higher rates of migration. In terms of welfare, migrant households are getting poorer. In fact, in 2013, the poverty rate of migrant households exceeded non-migrant households (39 vs 36.7 percent respectively 6 ). This is in sharp contrast to just 5 years earlier where the poverty rate of migrant households was much lower than non-migrant households (23.9 vs 33.9 percent respectively). Many of the temporal determinants and drivers of these trends are difficult to decipher using the KIHS for reasons that are explained more fully in the Annex. To put shortly, the analysis in this report is completed using data that is not ideal for studying migration in both its structure and variable content. A comprehensive migration survey would be incredibly insightful and useful. Current Context Recent events have made labor migration in the Kyrgyz Republic a central topic of interest. The slowdown of the Russian economy, devaluation of the ruble, and pending accession into the Eurasian Economic Union (EEU) all have implications on the impact of labor migration towards domestic welfare. National accounts data have shown large declines in remittances in November and December of 2014 (NBKR). It will be important to corroborate this with household data from 2014 as national accounting confounds transfers that are remittances and transfers that are business transactions (Mogilevsky and Atamanov, 2008). The impacts from accession into the EEU is one of the main points of interest for policy makers. Umurzakov, Poletaev, and Hasaova (2014) provide some insights to potential impacts based on 16 expert interviews. In terms of welfare, there are many implications as Kyrgyz labor migrants will not be subject to tightening Russian migration controls and regulations. With accession to the EEU, Kyrgyz migrants will have the same rights as other laborers in Russia. At the time of this report, the Kyrgyz Republic has yet to join the EEU, and has delayed entry a number of instances. The timing of its accession is a point of uncertainty. If the Kyrgyz Republic does ascend, labor migration will have few bottlenecks as member countries will enjoy free movement of labor. However, as long as the Kyrgyz Republic is not a member, benefits from labor migration will diminish as the minimum cost to migrate climbs to 317 USD including recurring costs such as monthly fees for patents (Titov, Prokhorova, and Chavez, 2015). In the short-run, the event of a large flow of return migration to the Kyrgyz Republic seems unlikely. The latest migration figures show that return migration to the Kyrgyz Republic is much lower than its neighbor Tajikistan, which will not join the EEU in the short run. Tajik labor migrants will surely face these new Russian migration laws. On the other hand, labor migrants from the Kyrgyz Republic may be waiting for ascension after which these new costs will no longer apply to them. Over the period March , the stock of Kyrgyz migrants in Russia declined by only 0.97 percent while Tajik migrants reduced by 6.8 percent. However, labor migration is still policy relevant. As recent migrant households are more vulnerable than before since they have higher dependency on remittances; policies to increase the development impacts of remittances are also more relevant than ever. Moreover, in 2013, migrant households became poorer than non-migrant households for the first time over the period with available household data. How do we increase the development impacts of remittances? This question can be framed from the perspective of how to increase the amount of potential labor earnings through pre-migration policies, how to make it easier to send remittances home, and how to educate migrant households to make more productive use out of these 6 Using the national consumption aggregate and national poverty line. vii

8 remittances. McKenzie and Yang (2014) review the evidence of policies that increase the development impact of international migration. Policy Recommendations From the regulatory side, the amount of government involvement in labor migration has been small, especially considering how labor migration is a very large phenomenon. This is an area where the government may choose to participate more. Many non-regulatory policy recommendations are also suggested. 1. Government facilitation of labor migration: Government intervention and oversight can improve welfare and migration outcomes. In the Kyrgyz Republic, migrants abroad are less educated and rely on networks rather than government programs when finding work abroad (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013). This can lead to underpayment and informal work environments. Moreover, the government may pursue bilateral labor agreements with ideal destination countries. Agreements can ensure the placement of labor migrants under well-known conditions. The facilitation of labor migrants can also be built in a before, during, and after framework. Before-migration policies include training, skills development, and job placement. During migration facilitation can be monitoring and protecting migrants. After migration policies can involve placement in the domestic labor force and utilizing new skills to spur domestic entrepreneurship. 2. Learning from International Experiences: Examining the Philippines 7 case and their relationships with other countries can offer valuable expertise in managing labor migration policies. The Overseas Workers Welfare Administration in the Philippines was established in 1977 to provide social and welfare services. It is highly regarded as a very successful program. Regional lessons can also be drawn from existing World Bank projects in other regions. 3. Diversifying destinations: To reduce dependence and regional risk, migrants should diversify their choice of destination countries. The Kyrgyz Republic s reliance on Russia and Kazakhstan may largely be a language issue. The institution of language schools can be a solution to increase the choice of destination countries. 4. Guarding against shocks: Migrant households ex-ante have lower levels of domestic income (social assistance, pension, and domestic wage incomes) and use remittances as an income shock absorber. Income and consumption patterns of migrant households are more volatile than patterns in non-migrant households. Moreover, the poverty rates of migrant households are increasing over time. A stronger study on the vulnerabilities and social status of labor migrants would be useful. The government may also wish to help migrant households think of innovative ways to protect against shocks by developing appropriate financial instruments. 5. Financial Literacy: Migrant households in the Kyrgyz Republic underspend remittances in investment (Akmoldoev and Budaichieva, 2012). This paper also finds that migrant households do not spend any more on education than non-migrant households. The development of financial instruments for migrants such as micro-insurance or savings products to invest at home can help enhance the impact of remittances. Knowledge on financial literacy is powerful for reducing the cost of sending remittances home, increasing the returns from remittances through investment and savings, and increasing the sustainability of remittances as a source of income. Financial training such as budgeting and money management can also be a useful tool especially for pre-migrants. 7 Philippines Overseas Workers Welfare Administration viii

9 For example, in the Philippines, the government provides financial training before labor migrants depart as well as deposit options to minimize the cost of sending remittances. 6. Skills: Pre-migration training and government support can create pathways for labor migrants into better jobs in higher skilled sectors. The bulk of Kyrgyz labor migrants work in low skilled occupations and the majority have completed only general secondary education. Training, upgrading of skills, and recognition of skills in countries of destination would help migrants secure better employment in host countries, earn more income, and obtain better jobs when they return home. 7. Data monitoring of return migrants: Labor migration unarguably plays a large role in the economic welfare of Kyrgyz citizens. While this report profiles labor migrants and their households to the best extent possible 8, there are still important knowledge gaps that need to be filled that cannot be answered using existing surveys. For example, what are return migrants doing and where do they work? Do young migrants return to school? These are all important questions to understand the longer run impacts of labor migration. Are the number of labor migrants really declining? International organizations and the government the Kyrgyz Republic are familiar with many of the recommendations listed above. There is great progress on many fronts as the Kyrgyz Republic works to better serve its labor migrants. For example, the IOM has helped launch employment centers in Bishkek and Osh. The World Bank has facilitated technical assistance and expert consultations on improving the legal and institutional framework for migration management. These efforts should continue as well as filling knowledge gaps for the purposes of enhancing the development impacts of labor migration. 8 There are many challenges to studying migration without a proper survey honed to examine this topic specifically. For example, the lack of sample rotation in the KIHS does not ensure year to year representativeness. ix

10 CHAPTER 1 INTRODUCTION 1.1 Kyrgyzstan is a sending country in terms of international labor migration, and the impacts of labor migration are of interest for a number of reasons. One is the sheer size of remittances as a proportion of GDP; over a third in There are a large number of individuals who are labor migrants; estimates ranging from half a million to one million. Another point of interest is the evidence from other countries that migration and remittances are welfare enhancing (World Bank, 2010ab; Adams Jr. and Page, 2005; Acosta et al, 2008). Yet, contrarily at the macro-level, remittances have not been found to promote economic growth (Clemens and McKenzie, 2014). 1.2 Understanding Kyrgyz migration and its relationship to employment and domestic welfare is extremely important in the current period. The recent Russian crisis and devaluation of the Russian ruble has put a strain on labor migration and remittances, which are channels for safety nets and income generation. Since the majority of Kyrgyz migrants work in Russia, the Russian economic downturn is expected to impact the Kyrgyz Republic a great deal. How will the welfare of households be impacted as income from remittances start to decline? 1.3 Income, economic reasons, and unemployment are leading indicators for migration. The vast majority of Kyrgyz labor migrants aboard are in Moscow (92 percent), and most of the remaining are located in Kazakhstan (8 percent) according to an Eurasian Development Bank report (2013). The difference in incomes that a labor migrant earns in Moscow compared to in the Kyrgyz Republic is substantial, as much as 10 times more a month. The majority of labor migrants are employed in low skilled occupations. 1.4 A large proportion of the domestic population benefits from some amount of remittances. In the national surveys, about 20 percent of households have a member who is working abroad. As the crisis in Russia begins to unfold, and return migration increases, it is important to evaluate who labor migrants are, what their families are like, and what are their characteristics. These insights will help the country prepare for the amount of return migration. 1.5 This is not the first time that the Russian economy faced an economic crisis affecting remittance flows to the Kyrgyz Republic. In 2009, remittances to the Kyrgyz Republic dropped as the Russian GDP contracted by 7.8 percent. This did not affect remittance flows drastically, which is consistent with existing literature that finds remittances flows to be resilient in times of crisis, and also be consumption smoothing. (Sirkeci, Cohen, and Ratha, 2012; Kose, Islamaj, Ratha, De, and Yousefi, 2015). 1.6 However, common reasons why remittances may be resilient during times of crisis may not apply to the current Kyrgyz context and the current Russian economic downturn. Remittance flows in November and December of 2014 have already shown visible dips. Foremost, in the current economic conditions, there are added challenges such as currency devaluation, tougher migration policy, and regulations that are making labor migration to Russia less lucrative, easy, or attractive. The cost of passports to go abroad, permits, new requirements for proficiency tests in Russian language and history are all new costs to migration that did not exist during the 2009 financial crisis period. There is an absence of diversification of destinations for Kyrgyz migrants; in 2014, the vast majority of migrants are in Russia. A component of Kyrgyz migrants are seasonal and more vulnerable to shocks. The majority of Kyrgyz migrants are also low skilled. 1

11 1.7 Existing literature on Kyrgyz labor migration often focuses on the amount of remittances received rather than welfare of the domestic households. ADB 2008 study (Mogilevsky and Atamanov, 2008) provides a comprehensive study of remittances in Central Asia with a focus on financial sector and fiscal considerations. However, there is less a focus on welfare aspects. An EDB 2013 study focuses on push-pull factors of migration but also does not focus on its impact on domestic poverty (Beishenaly, Levant, Ormonbekova, Shamshiev, 2013). A recent World Bank migration survey on skills has a migration component but few migration questions (CALISS, 2013). 1.8 Labor migrant is a strategy for individuals and households to improve their welfare when it cannot be achieved domestically. Since not all labor migrants from the Kyrgyz Republic choose to remain in their host country and remittances are sent to family, the intention is fundamentally to improve household welfare. This note examines trends in labor migration and ultimately at the household level. Many studies on the benefits and resilience or remittances has been carried out using macroeconomic data. The collection of national aggregates of remittances is fundamentally different than its collection at the household level. Since macro trends in remittance recordings often confound private and enterprise related transfers, they are less ideal for studying household and individual welfare. Collection at the household level is more revealing of household-level welfare impacts. 1.9 The following sections are arranged as follows. The note begins with a brief description of the data in Section 1. Section 2 examines the number and characteristics of labor migrants. Section 3 examines and compares the distribution and characteristics of households with and without labor migrants. The welfare impacts of labor migration are explored in Section 4, namely the impact of remittances on poverty reduction and socioeconomic outcomes. Section 5 concludes. Annexes also provide important information on the data; its structure and limitations. 2

12 # Migrants Abroad CHAPTER 2 PROFILING LABOR MIGRANTS 2.1 As a proportion of the population, the number of labor migrants abroad in the Kyrgyz Republic is very high. In 2013, the country had a population of approximately 5.7 million people. In some estimates, the number of labor migrants is as high as one million. Since such a large proportion of the working age population is abroad, it is important to track this figure, understand who they are, what they are doing, and how this is impacting domestic development and growth. 2.2 The Labor Force Survey (LFS) in the Kyrgyz Republic Integrated Household Survey (KIHS) 9 allows for the identification of labor migrants as household members who are absent and working abroad. An employed individual is identified to be either a domestic worker, an internal migrant, or a migrant abroad based on their location of work in the quarterly LFS. When these workers are absent from the household during the time of the interview, other household members can report on their employment and wage details. A. THE NUMBER OF LABOR MIGRANTS 2.3 Over the period 2008 to 2013, the number of labor migrants captured by the Kyrgyz Republic Integrated Household Survey averaged around 200,000. The number of labor migrants increased from 2008 to 2011, but then fell in 2012 and , both in the number of internal and international labor migrants. The largest increase in the number of labor migrants abroad was the period from right before the financial crisis. The Russian economy rebounded in 2010 after the financial crisis, but since then, GDP growth has been declining. In 2012, the country experienced a year of zero growth. Continued declines in migration abroad may be influenced by Russia s poor economic growth and currency devaluation. In 2013, the growth rate of GDP in Russia was 1.3 percent. 300,000 Figure 2.1: Number of Labor Migrants 250, , , , , , , , , ,000 50,000 0 Q1 Q2 Q3 Q4 Annual Average Notes: Quarterly labor force population weights are applied (vesakv) for quarterly data, and annual weights are applied for annual statistics (vesa1). Source: KIHS , LFS (questionnaire #4) 9 The 2008 to 2012 data is based on sampling that originated in 2003 using the 2001 Census. In 2013, a new sample was drawn. 10 As a caution, it should be noted that poor household rotation in the sample may exaggerate some trends. For example, if the survey is not sufficiently rotated, average age would increase. 3

13 % Change 2.4 Seasonality in migration is present since migrants tend to return home during the winter (Error! Reference source not found.). This is corroborated in the LFS data since there are generally the ewest number of workers in quarters 1 and 4 of each year when the weather is colder. Many labor migrants work in construction, and they would be least likely to work these times of the year. 2.5 The number of labor migrants in the KIHS is lower than estimated by other sources. The National Statistical Committee of the Kyrgyz Republic estimates the number of labor migrants abroad to be 457,000 in 2011 and the Eurasian Development Bank (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013) estimates over a million. Other expert estimates of 25 percent of the labor force is abroad (Umurzakov, Poletaev, and Hasanova, 2014). It is possible that labor migrants who have been away from the household for a prolonged period of time are not included in the KIHS roster. Adult relatives that have moved away to work may not be considered by respondents to be in the household, even though these households receive remittances from them. Another possible reason for the low count of labor migrants is that entire families who have moved abroad would be missing from the KIHS. Russian administrative data on the number of migrants are not clear since there is a break in 2011/2012 on how migrants are counted. 2.6 In 2012, the Kyrgyz economy experienced a period of zero growth, which affected all types of household absenteeism. Low growth should be a push-factor that increases the number of labor migrants abroad. However, the number of labor migrants actually decreased in One possibility is that as it became relatively more expensive to travel abroad and existing migrants abroad simply started sending more money home. Absence from the household also decreased in the Kyrgyz Republic for other reasons. The overall decrease in absent household members in 2012 is substantial as the number of members living with relatives dropped more than half, and the number of students abroad also dropped. 2.7 Workers abroad comprise about 10 percent of the total employed work force (Figure 2.2). While domestic employment levels has been slowly increasing since 2009, employment of internal and international labor migrants have experienced larger fluctuations and sharp declines beginning in Figure 2.2: Employed Workers (15 years of age +), by location of work Domestic Internal Migrant Migrant Abroad ,899,185 68, , ,897,441 80, , ,912,084 89, , ,941,254 91, , ,968,233 87, , ,891 61, , Domestic Internal Migrant Migrant Abroad Notes: International labor migrants are workers who list their location of work to be outside of the Kyrgyz Republic. Population weights are applied. Panel households are those that are present in all five years. Source: KIHS , LFS (questionnaire #4) 11 While there is a sharp decline in internal migration in 2013, there are no real indications that internal labor migration has experienced a decline, and this trend may be a result of resampling in Since internal migrants is a relative small proportion of the total sample of individuals, these estimates may not be robust 4

14 B. CHARACTERISTICS OF LABOR MIGRANTS 2.8 Labor migration is an individual or collective household choice often made under economic considerations. Individuals who have difficulty finding work in their current location will choose to relocate where they think they can find a job or receive higher returns. Households with low domestic income in terms of wages or social assistance will seek income sources from abroad. For example, younger and more able-bodied members may go work if older members are not earning enough or not receiving sufficient social assistance. In the Kyrgyz Republic, the wage premium from working in Russia and Kazakhstan is a pull factor for the poor and rich alike. 2.9 Labor migrants working abroad are more likely to be male, younger, and less likely to be married than workers who remain in the Kyrgyz Republic. Error! Reference source not found. illustrates basic haracteristics of labor migrants as well as domestic and internal migrants for comparison. Labor migrants are not likely to be the head of household, but rather the sons and daughters in the household. Migrants who are heads of their own households tend to participate in internal migration, likely so they are not away from their families for too long. While international migrants live with their parents, about half are also married, suggesting that labor migrants are part of intergenerational households Males are much more likely to leave the country for work. In 2010, this gender gap closed a bit, when the proportion of female labor migrants abroad was at its largest share over this period, 29.4 percent. In 2010, the number of labor migrants increased in total as well. The addition of females could be similar to the added-worker effect that is seen by other countries during times of crisis as well. In Turkey, during the financial crisis, the proportion of married females participating in the labor force increased, to supplement household wages Social networks are key in facilitating migration. The presence of family and friends abroad is a strong determinant for migrants deciding to emigrate, and the expat community also provides assistance and support in looking for a job and finding accommodations. The presence of immigrant occupational networks has been repeatedly observed across different countries and migrants of different nationalities. Networks facilitates in job search. Not only that, migrants who succeed in certain occupations or have tried and tested the labor market have many advantages. Patel and Vella (2013) find that immigrants in occupations with a large existing network of fellow countrymen in the same occupation earn higher wages than immigrants in occupations without a network. Age and Education 2.12 Labor migration is a common activity among individuals who are young. In 2013, on average migrants abroad were 28.2 years of age compared to 36.7 years for domestic workers. In 2013, about 17 percent of workers between 15 and 24 years of age were working abroad. Not only is this a large proportion, it also means that these individuals are not enrolling in higher levels of education. It is unclear if these migrants are only delaying their education or never return to school to pursue higher education. The proportion of young workers who were abroad peaked in In 2011, 21 percent of working youths aged were abroad at some point in the year. 5

15 School enrollment, tertiary (% gross) Number of workers Figure 2.3: Age Distribution, by location of work (2013) Age Notes: Annual LFS weights are used. Source: KIHS , LFS (questionnaire #4) Domestic Internal Migrant Migrant Abroad 100% Figure 2.4: Educational Distribution, by location of work (2008 & 2013) 80% 60% 40% 20% 0% 7.7% 0.2% 7.1% 0.1% 0.0% 10.1% 7.4% 0.1% 6.70% 0.10% 4.7% 0.0% 44.4% 27.5% 52.7% 52.4% 20.1% 21.3% 26.2% 20.2% 19.9% 16.1% 18.2% 48.2% 63.40% 79.4% 21.00% 7.6% 8.80% 8.3% Domestic Internal Migrant Migrant Abroad None General Secondary Higher Primary Specialized Secondary 2.13 Labor migrants abroad are less educated. About two-thirds of labor migrants have completed a general secondary degree, which is expected from a young pool of migrants who work in low skilled jobs. Domestic workers are much more educated than both internal and international migrant workers; over a third have completed a specialized secondary or tertiary degree. Perhaps enticed by relatively high earnings for low-skilled work, the composition of labor migrants abroad has become less educated over time (Figure 2.4). The average of labor migrants has not changed significantly over this period Out-migration of youths is similar to some extent to other country experiences. Migrants have lower safety nets and are more likely to be early school leavers. In the EU countries, this is based on migrant vulnerabilities compared to the native population (Nouwen, Clycq, Ulicna, 2015). While this is a different scenario since all Kyrgyz are natives, the EU case is comparable in the sense of who are the socially vulnerable Low returns to education abroad can discourage youths from enrolling in higher education. In fact, enrollment in tertiary education in the Kyrgyz Republic has been flat since The dominance of migration among the population may lead to under development of human capital in the country. Chiquiar and Hansen (2005) find similar trends of migrants from Mexico to the United States. McKenzie and Rapaport (2005) also illustrate that migration and low returns to education has a disincentive effect on education attainment Figure 2.5: Tertiary Enrollment Rates ECA (all income levels) ECA (developing only) KAZ KGZ RUS TJK Source: World Bank, data.worldbank.org, indicator SE.TER.ENRR 6

16 2.16 Forward looking, one important policy and development consideration is to increase the understanding of the welfare of young workers and their families after migration. When the young are spending their youth as labor migrants, this puts pressure on the quality of the labor force in later years. If the prospects of migration are affecting the educational decisions of the youth, this could be a severe social problem. The majority of labor migrants have intentions to return to the Kyrgyz Republic; only 6 and 9 percent want to stay in Russia and Kazakhstan respectively (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013). Therefore, their educational and career plans upon return are important to understand as well. Using the current survey, since individuals cannot be tracked, it is not possible to understand the dynamics of return migrants. Sector and Labor Market Characteristics 2.17 Over time, the composition of professions of labor migrants abroad has stayed primarily in manufacturing, though the proportion of manufacturing workers has shifted from skilled to unskilled (Figure 2.6). In 2008, an estimated 112,700 labor migrants worked in skilled manufacturing and 31,468 worked in unskilled manufacturing. By 2013, only 72,245 worked in skilled manufacturing and 88,917 worked in unskilled manufacturing. The proportion of labor migrants working in unskilled manufacturing increased from 15.8 percent in 2008 to 44.6 percent in The professions of migrants differ from those of domestic workers (Figure 2.7). In contrast to domestic workers, the majority of migrant workers abroad are mainly employed in manufacturing and lower skilled work. The top three professional classifications, which includes 96.4 percent of migrants are skilled manufacturing, unskilled manufacturing, and service workers. The majority of internal labor migrants, though to a smaller degree, also work in these three professional areas. The primary differences between domestic workers and those who migrate are the shares of workers in high medium professions and agriculture. Figure 2.6: Profession Distribution of Labor Migrants Abroad, , , ,441 49,537 68,361 79, ,919 68, ,269 97, ,031 46,230 46,466 44,150 46,479 35,820 3,681 3,135 3,653 3,506 2,528 1,860 High Medium Services Agriculture Skilled manuf Unskilled manuf Notes: Employed 15+. Number of workers labeled in graph. Source: KIHS , LFS (questionnaire #4) 88,917 72,245 Figure 2.7: Distribution of Profession by location of work, % 7% 27% 28% 46% 0% 20% 31% 21% 16% 18% 1% Domestic Internal Migrant Migrant Abroad Notes: Employed 15+ Source: KIHS , LFS (questionnaire #4) 45% 36% 0% High Medium Services Agriculture Skilled manuf Unskilled manuf 2.19 Comparing the sectors of employment, migrants abroad primarily work in construction and retail sectors. Construction is the top sector of employment of migrant workers abroad, though the share of workers in this sector has declined from 2008 to 2013 (49.3 to 35.3 percent respectively). In 2013, the 7

17 percent of labor migrants in retail is 35 percent, if the trend continues, construction will no longer be the top sector of employment One good trend is that the proportion of labor migrants who are working in informal contracts or verbal agreements is declining. In 2008, 58.4 percent of migrants abroad worked under verbal agreements compared to 37.6 percent in 2013 (Error! Reference source not found.). Moreover, he decline in informal contracts is present across all professions Some literature finds that return migrants are more productive, spur innovation, or hold more patents than workers who never migrated. These are all potential benefits of migration. A negative aspect is occupational downgrading of migrants when abroad. These are all important questions which are not easily answered using the KIHS and LFS. Principally because individuals cannot be tracked across years or quarters. Return migrants cannot be identified, and pre-migration labor history cannot be tracked. These are just some reasons why a migration specific survey would be indispensable. Wages 2.22 Economic reasons are the primary motivation to migrate. An EDB report on labor migration in the Kyrgyz Republic finds this to be almost universally the reason in poorer regions, and to a small degree lesser in wealthier regions or Bishkek (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013). In the EDB survey, very few cited other reasons for migration such as political instability of skill acquisition. The official minimum wage in the Kyrgyz Republic is lower than even in Tajikistan; $17 per month and $52 per month respectively 12. At these levels, a household with two adults earning minimum wage would be living below poverty Average monthly wages across almost all sectors are higher abroad, but barely so by Figure 2.8 illustrates average wages by location of work. Average hourly wages are shown for the top three sectors and professions in 2008 and Wages are predominantly from work in Russia. Wages earned by workers abroad are two or three times higher than domestic workers in manufacturing. In the sectors that are most common to labor migrants, workers earn about double that of domestic workers. Construction is the most popular sector for labor migrants abroad, however, average earnings of labor migrants is lower than earnings of natives in host countries. In 2012, construction workers of Russian nationality earned an average of 25, Russian rubles in the previous month, or about 33,676 Soms Wages are higher abroad, however the premium from working abroad has diminished over time (Figure 2.9). From 2008 to 2013, the difference between domestic and international hourly wages have diminished. Wages are becoming more similar because both domestic wages are increasing and because foreign wages are declining or flat. This could also be related to changes in the composition of workers as the Kyrgyz labor migrant becoming younger and less skilled, though composition is not fully explanatory of the diminishing wage gap Using the RLMS data: ruadult.dta: sum iuwagelm if iupriind==6&iunation==1 8

18 Change in Wages (%) 2005 KGS 2005 KGS 2005 KGS 2005 KGS Figure 2.8: Average hourly wages in the Previous Month (2008 & 2013) Sectors Construction Retail Services Professions Skilled manuf Unskilled manuf Services Construction Retail Services Skilled manuf Unskilled manuf Services Domestic Internal Migrant Migrant Abroad Notes: The numbered columns rank the share of workers by sector or profession. Non-zero wages only; zero income workers are excluded. Sample of year olds who are employed. The reference period for hours worked is the last week while the period for wages is the last month. Source: KIHS , LFS (questionnaire #4) 40% 30% 20% 10% 0% -10% -20% -30% -40% Figure 2.9: Increasing domestic wages, selected sector Manufacturing Services % 20% 10% 0% -10% -20% -30% Domestic Internal Migrant Migrant Abroad Notes: Sample of year olds who are employed. The wage gap = ln(wage_foreign wage_domestic). Monetary values are in 2010 KGS, adjusted by the CPI. Zero income is removed. Source: KIHS , LFS (questionnaire #4) 2.25 Figure 2.10 illustrates the wage premium from working abroad for the top 3 professions across all years. The decline in the gap between foreign and domestic hourly wages has been shrinking in manufacturing from 2008 to 2013 almost year to year. Domestic wages in the last period increased faster than foreign wages. In the most recent period of survey data, domestic wages have increased in the two most popular migrant sectors; construction and services. This may coincide with the presence of domestic skill shortages The percent of domestic services firms citing an inadequately educated workforce as a major constraint has also increased from 2009 to 2013 (Figure 2.11). In the latest 2014 administrative data, over 9

19 2010 KGS Foreign Wage - Domestic Wage (2005 KGS) Percent of firms half of migrants in Russia are in the services sector. The lack of adequacy can be related to the shortage of skills as more and more labor migrants are working in the services sector abroad. The proportion of manufacturing firms that site the constraint to be a major one has increased at a much smaller degree. Figure 2.10: Declining Wage Premium from Working Abroad Services Skilled manuf Unskilled manuf Notes: Sample of year olds who are employed. The wage gap = ln(wage_foreign wage_domestic). Hourly wages are in 2005 KGS, adjusted by the CPI. Zero income is removed. The reference period for hours worked is the last week while the period for wages is the last month. Source: KIHS , LFS (questionnaire #4) Figure 2.11: Percent of Firms Identifying an Adequately Educated Workforce as a Major Constraint Source: Enterprise Surveys Manufacturing Retail and Services 2.27 Macroeconomic conditions influence the number of labor migrants as well as their earnings. While the number of labor migrants was highest during the crisis, average income was also the lowest (Figure 2.12). By 2013, average monthly earnings of labor migrants was much lower than in 2008 with the number of labor migrants also declining. The number of workers with zero income was higher in 2009 and This suggests that the crisis did not deter the volume or quantity of workers, but there were price effects that resulted in lower returns to labor migrants. The lower returns actually increased the number of migrants as more migrants were needed to earn the same amount of income. Figure 2.12: Average Monthly Wage Income of Labor Migrants Abroad (excluding zero wages) 9000 Q1 Q2 Q3 Q4 Annual Average $6,919 $6,142 $5,299 $5,163 $5,493 $5,502 Notes: Sample = workers aged 16-64, and working abroad. Wages are income from primary work (zarplata). Quarterly labor force population weights are applied (vesakv) for quarterly data, and annual weights are applied for annual statistics (vesa1). Monetary values are in 2010 KGS, adjusted by the CPI. Zero income is removed for coding reasons. In the KIHS, zero reported income captures different scenarios. Zero can be because there is zero income, the household does not want to respond, or the household does not know. Source: KIHS , LFS (questionnaire #4) 10

20 Number of households % of households with labor migrants CHAPTER 3 HOUSEHOLDS AND HOUSEHOLD MEMBERS 3.1 Since many labor migrants are abroad to provide income for their families and not solely for individual gain, it is important to observe migration trends at the household level as well. Understanding the type of households that benefit from migrants is also a useful profiling exercise. It is very possible that certain types of households have a higher propensity to migrate than others. For example, households with low domestic income through wages or social assistance are more likely to have members of the household who are migrants. 3.2 Migrant households are defined as those who reported at least one household member whose location of work is abroad in any quarter from the Labor Force Survey 14. This is a better indication than defining migrant households as only those who receive positive remittances. To illustrate the full impact of welfare from labor migration, it should be accounted the instances where households have migrants but did not receive remittances as well as the difference in the domestic household labor supply. A. REGIONAL AND ECONOMIC DISTRIBUTION OF HOUSEHOLDS 3.3 Over time, the proportion of households with labor migrants has declined (Figure 3.1). In 2008, 19 percent of households had at least one labor migrant abroad, compared to 12.6 percent of households in Note that if an entire household has migrated, then this household would not be a candidate in the KIHS. This is one possibility for the underreporting of migration in the KIHS compared to national statistics. Figure 3.1: Migrant and Non-Migrant Households 1,600,000 1,400,000 1,200,000 1,000, , , , , % 20.0% 15.0% 10.0% 5.0% 0.0% # non-migrant households # migrant households % households with labor migrants Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad at any time over the year. Household weights are applied. Source: KIHS Rural regions have higher proportions of migrant households than urban areas. Over time, the number of migrant households has dropped dramatically in rural areas, and only marginally in urban areas. The percent of migrant households in rural regions ranged from 26 percent in 2008 to 19 percent in 14 Another source of information to identify workers is the household roster. However, internal and international migrants cannot be differentiated using the household roster. Only in the LFS, can labor migrants be differentiated between internal and international migrants. There are small instances where absent household members reported in the Q1 household roster are not present in the LFS, but this is a very small proportion. Since the LFS is available for four quarters, it is a better instrument for using to identify households with migrant workers. 11

21 Percent of Households with Labor Migrants Percent of Households with Labor Migrants Abroad Number of Households 2013 (Figure 3.2). This large decline translates into approximately 150,000 to 200,000 households annually. About 9-11 percent of urban households had migrant workers abroad; or about 50,000 households, a number that remained relatively stable over the period of analysis. Figure 3.2: Households with Labor Migrants Abroad, Rural-Urban 30% 26% % 20% 15% 10% 23% 23% 23% 11% 10% 10% 10% 19% 19% 9% 9% % % 0 Rural Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad at any time over the year. Bishkek does not have any rural areas. Household weights are applied. Source: KIHS Urban 3.5 The incidence of labor migration varies across regions. It is interesting why there is virtually no migration in Naryn while its neighboring region Osh has one of the highest rates of labor migration. Network effects are undocumented but can influence the concentration of labor migrants. For example, if certain labor migrants in Issykul became very successful abroad, they may attract even more migrants from the same region. Migrants abroad will often live together with other members from the same village. 3.6 Southern regions with the largest number of poor, also tend to have the highest rate of migrant households. In both urban and rural areas, Osh, Jalal-Aba, and Batken have the highest rate of migrant households; and at rates much higher than other regions (Figure 3.3). Southern regions are some of the poorest regions. This is not surprising since economic reasons such as unemployment and low domestic wages are the primary drivers of migration. These regions don t necessarily have the highest poverty rate, but do have some of the largest number of poor (Error! Reference source not found.). Figure 3.3: Percent of Households with Labor Migrants Abroad, by Region Rural Urban 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% Batken Bishkek Chui Issykul Jalal-Aba Naryn Osh Talas Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad at any time over the year. Bishkek does not have any rural areas. Household weights are applied. Source: KIHS

22 3.7 There are regions that are poor with also very little international migration. For example, the Dzhety-Oghuz district in the Issy kul oblast has a very high number of poor as well but the Issy kul region has a low rate of international migration. There regions also do not have high levels of internal migration. 3.8 The drop in the number of labor migrants in appears to be driven by a termination of migration activity rather than migrant households moving to out of the country. This can be seen by examining a sub-sample of panel households that are in the sample continuously from 2008 to Error! Reference source not found. in the Annex illustrates the proportion of households exiting and ntering the survey. The highest proportion of migrant households exiting the survey was between 2009 and 2010, where 13.3 percent of households were migrant households. Moreover, there was also a large amount of migration activity in the population since 8.7 percent of the newly sampled households were also migrant households. However, overall migration activity began to slow down. Between 2011 and 2012, only 9.9 of exiting households were migrant households and 5.1 percent of replacements were migrant households. 3.9 As stated earlier, possible reasons for a slowdown in migration are improving wages domestically and lower wage premiums from working in Russia. However, estimating the flow of migrants is difficult and often imprecise, with levels varying from source to source. Moreover, the structure of the KIHS survey, with poor household rotation, results in a data set that is more longitudinal rather than cross-sectional. Therefore, there may be some bias in the year to year trends. However, it is some comfort that the number of migrants continues to decline in 2013 when the sample of households was refreshed. This seems to indicate a real decline in the proportion of households that are migrating. Though, the true nature of this is best revealed in a survey that is specialized with migration questions where these dynamics can be asked more directly. B. CHARACTERISTICS OF THE HOUSEHOLD AND HOUSEHOLD MEMBERS 3.10 The composition of households and migrants themselves is an important point of study. Many studies on remittances and labor migrants are conducted using small scale qualitative focus groups (Tynaliev and McLean, 2011; Isabaeva, 2011) or is based on earlier data (Agadjanian et al, 2008; Mogilevsky and Atamanov, 2008). After such recent events as the 2010 revolution and financial crisis, a current study of labor migrant profiles is useful. Remittances are highly poverty reducing among migrant households. Due to the current economic down turn, both remittances and labor migration is expected to decline Labor migration, both internal and foreign impacts social aspects of life. Labor migrants in the recent period are extremely different than those twenty years ago. The types of households that have migrant household members are also changing. The migrant members are less likely to be the heads of households, but rather adult children. Current migrants are young, and often skip the prospects of a college education and immediately immigrate to Russia after high school Error! Reference source not found. illustrates the many differences in the characteristics of igrant and non-migrant households. Consistently across all years, households without migrants are smaller, however their dependency ratio is higher indicating that households without migrants have larger shares of children or elderly. The dependency ratio for migrant households is lower. It appears that the choice of migration is not necessarily investment into the youth. Instead there is more evidence that remittance income is being spent on assets. Households without migrants also are better educated. A larger proportion of heads of household in non-migrant households have a higher education, and less likely to have lower levels of education. 13

23 3.13 Error! Reference source not found. shows regression results on the determinants of being a migrant household controlling for a variety of characteristics simultaneously. Migrant households have lower social income and lower dependency ratios, lower levels of education, and are located in rural areas. The migrating youths from these households are low skilled and their households are poor. Migrant households have lower pension and social assistance income, or social pre-migration income. The choice to migrate could is driven by vulnerability and to earn additional income to supplement low levels of income that is acquired domestically. Contrary to research in other countries, the sheer size of households itself is not so much a determining factor of being a migrant household versus the dependency ratio of the household. Table 3.1: Differences in Household Characteristics (non-migrants migrant) Age of HH head 1.883*** 2.380*** 2.212*** 1.849*** 1.490** (3.43) (4.51) (4.02) (3.42) (2.64) (0.79) Male HH head ** * * ** (2.69) (2.16) (2.47) (0.86) (1.04) (2.61) # of household members *** *** *** *** *** *** (including absent members) (-21.28) (-18.93) (-18.17) (-17.78) (-17.91) (-16.34) # of Children (0-14) *** *** *** ** *** *** (-6.68) (-4.33) (-3.43) (-3.29) (-3.92) (-5.01) # of adults (15-64) *** *** *** *** *** *** (-22.84) (-22.70) (-22.49) (-22.47) (-21.73) (-20.58) Dependency Ratio ** *** *** *** ** (based on present members) (2.61) (4.13) (5.21) (3.97) (3.27) (1.64) Dependency Ratio *** 0.108*** 0.109*** 0.102*** *** *** (including absent members) (9.05) (11.28) (11.37) (10.77) (9.28) (6.88) Highest Education of Household Head Higher 0.101*** 0.104*** 0.100*** 0.114*** 0.124*** *** (7.74) (7.98) (7.43) (8.95) (9.73) (6.72) Incomplete Higher *** ** ** (4.26) (2.68) (0.34) (2.76) (0.54) (0.48) Secondary Professional (-0.38) (-0.69) (-0.79) (1.53) (0.86) (1.91) Professional Technical (-0.84) (0.06) (-1.09) (-1.96) (-1.20) (1.72) General Secondary (completed) *** *** *** *** *** *** (-4.94) (-6.08) (-3.69) (-5.56) (-5.11) (-7.16) General Secondary (incomplete) (-0.22) (0.30) (0.92) (0.97) (0.92) (1.36) Elementary * (0.87) (2.56) (0.08) (-0.23) (-0.85) (0.13) Observations Notes: Difference is non-migrant household migrant household. t statistics in parentheses* p<0.05, ** p<0.01, *** p< Table 3.2 compares the characteristics of employed family members working domestically 15 in migrant and non-migrant households. In terms of profession, domestic workers in migrant households are 15 Excluding internal migrants. 14

24 much more likely to be working in agriculture and have unspecified labor contracts. In non-migrant households, a larger proportion of employed members work in high medium occupations. Over time, domestic workers in non-migrant households are more likely to be females. The average education level of domestic workers in migrant households are also higher, which highlights the selection of younger less educated workers traveling abroad to earn higher wages, whereas individuals with higher education may earn sufficient wages domestically. Table 3.2: Characteristics of Domestically Employed Workers Non-Migrant Household Migrant Household Population 1,357,997 1,645, , ,713 Male 44% 40% 48% 51% Female 56% 60% 52% 49% Head 43% 41% 36% 35% spouse 22% 21% 21% 20% son/daughter 26% 29% 32% 31% son/daughter in law 5% 6% 8% 11% father/mother 0% 0% 0% 0% spouse's parents 0% 0% 0% 0% brother/sister 1% 1% 0% 0% grandson/daughter 2% 1% 2% 2% grandfather/mother 0% 0% 0% 0% Other 1% 0% 1% 0% Age Not Married 31% 29% 29% 31% Married 69% 71% 71% 69% None 22% 21% 12% 9% Primary 28% 21% 23% 14% General Secondary 42% 50% 54% 71% Specialized Secondary 7% 7% 10% 6% Higher 0% 0% 0% 0% Not specified 7% 4% 13% 7% at enterprise, organization, collective 37% 34% 27% 20% at a peasant farm 17% 22% 30% 50% individual commercial activities 1% 0% 0% 0% on an individual basis 16% 18% 17% 14% wage work for private individuals 22% 21% 13% 8% Not specified 41% 46% 60% 72% open-end written labor contract 36% 29% 25% 18% fixed-term written labor contract 4% 3% 2% 1% labor or other civil law contract 2% 3% 1% 2% verbal arrangement without execution of contract 17% 18% 12% 8% Agriculture & Mining 28% 29% 46% 59% Manufacturing 10% 8% 7% 3% Utilities 2% 2% 2% 1% Construction 6% 9% 5% 6% Retail 16% 15% 12% 8% Transport 7% 7% 5% 3% Services 7% 9% 4% 5% Other 23% 20% 19% 15% High Medium 26% 23% 17% 12% Services 21% 20% 18% 16% Agriculture 26% 25% 42% 54% Skilled manufacturing 23% 29% 19% 13% Unskilled manufacturing 5% 4% 4% 5% Source: KIHS

25 3.15 Heads of household in migrant households are less likely to be employed than their counterparts in non-migrant households. In migrant households, members are working less, either because they are unable to or there is no need to because of large remittance income. The gap in the employment rate in urban areas has widened over time, the gap has narrowed in rural areas (Figure 3.3). Figure 3.4: Employment Rate of Non-Migrant Head of Households (15-64 years of age) Rural Urban 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 74% 75% 75% 78% 77% 68% 64% 66% 69% 62% 73% 66% Non-Migrant Household 81% 81% 81% 82% 82% Notes: Sample includes only workers who work in the home oblast aged or non-employed living at home. Migrant households are defined based on the employment file, if at least one person in the household was working abroad at any time over the year. Bishkek does not have any rural areas. Population weights are applied. Source: KIHS Table 3.3 illustrates the determinants of employment for household members that are located domestically, focusing on if these individuals are more likely to work if they are in a migrant or non-migrant household. From , domestic working age household members in migrant households were less likely to be working than similar household members in non-migrant households. Starting in 2011, they were more likely to be working. The reason for the increased employment of domestic workers may be related to the lower power of remittances and reduction in wages from abroad. Note that the wages from abroad should be an upper bound of remittances sent home since some foreign wages are spent abroad. Domestic workers in migrant households earn less than those in non-migrant households. This is true even after controlling for a number of individual and regional characteristics (Error! Reference source not ound.). Table 3.3: Employment of Domestically Located Household Members Migrant Household *** *** *** 0.244*** 0.168*** 0.441*** ( ) ( ) ( ) ( ) (0.0101) ( ) Male 0.145*** 0.264*** 0.305*** 0.603*** 0.204*** 0.270*** ( ) ( ) ( ) ( ) ( ) ( ) Married *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) General Secondary *** *** 0.170*** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) Specialized Secondary *** *** 0.233*** 0.143*** *** 0.268*** ( ) ( ) ( ) ( ) ( ) ( ) Tertiary Education *** *** 0.115*** 0.236*** *** *** ( ) ( ) ( ) ( ) (0.0111) ( ) Urban 0.274*** 0.759*** 0.130*** 0.639*** 1.227*** *** ( ) ( ) ( ) ( ) (0.0124) ( ) Constant 2.526*** 1.447*** 2.132*** 1.847*** 2.986*** 2.884*** (0.0209) (0.0148) (0.0167) (0.0169) (0.0312) (0.0226) Number of Observations 7,562,363 7,512,148 7,521,448 7,636,099 7,720,560 7,673,940 Mc-Fadden's R-square % 80% 70% 60% 50% 40% 30% 20% 10% 0% 66% 67% 63% 62% 58% 76% 51% Migrant Household 16

26 Notes: A logit regression. Dep. Variable = employment dummy. Other controls include gender, age group, relation to head of household, education, oblast, marital status, occupation, and profession. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p< In migrant households, employment rates for year olds are much higher than in nonmigrant households (Figure 3.5). While young people are completing basic primary and secondary, they are not pursuing higher education. Kroeger and Anderson (2014) find that during the volatile period of , boys aged from remittance receiving households were less likely to be in enrolled in school than others. Though the effects they find are small, remittances did not improve education outcomes of male children in remittance-receiving households. 30% 25% 20% 15% 10% 5% 0% Figure 3.5: Employment and Enrollment Rates of year olds Employment Rate Enrollment Rate 18% 17% 17% 15% 15% 13% 14% 14% 13% 12% 14% 25% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 84% 83% 85% 85% 81% 77% 76% 77% 78% 69% Non-Migrant Household Migrant Household Notes: Based on year olds who are present in the household. Migrant households are defined based on the employment file, if at least one person in the household was working abroad at any time over the year. Individual weights are applied. Source: KIHS There are many positive economic aspects of migration, but also many social costs. For example, divorce rates have increased, especially in southern regions where migration rates are high. Disunion of the family negatively impacts children. Children are often left in the care of relatives or grandparents. This trend has led to the term social orphanship, often present in households where both adults are absent and children do not have access to education. Migrant households do not spend more on education than non-migrant households. Analyzing the set of households who remained in the sample across all five years, while being a migrant household does correlate with higher consumption in a given year, it does not correlate with higher education expenditures (in total or in per capita terms) Overtime, wealthier households (net of remittances) are migrating less. To determine how wealth is related to migration, it is useful to rank households based on domestic income excluding remittances. This is illustrative of household conditions that are a determining factor on whether additional income from remittances would be beneficial. Two measures of domestic wealth are used: total income net of remittances, and only domestic wage income. A decline of migration is apparent in households at all income quintiles except for the bottom 20 th, when wealth is measured as total income net of any remittance income or only domestic wages (Figure 3.6) Over time, households that have labor migrants are becoming more dependent on remittances, as it starts to comprise larger shares of total household income. The percentage of migrant households in the bottom 20 is increasing over time, which can be interpreted that a large number of households are becoming increasingly dependent on remittance income. The trend in migration and wealth is even more evident when income quintiles are based only on domestic wages alone. In households where income is already high, there may be no need to earn additional income or to send household members abroad. Or it may be the case that households with the lowest amount of remittance income have simply 17

27 stopped migrating as domestic wages have started to increase. Whatever the reason, the current pool of migrant households are more vulnerable to shocks given such a large proportion of their income is from remittances rather than domestic sources. Figure 3.6: Number of Migrant Households, by income quintile 1) Quintiles = Total income net of remittances 2) Quintiles = Domestic wages only Bottom 20th percentile 40th 60th 80th Top Bottom 20th percentile 40th 60th 80th Top 20 Notes: Income quintiles are based on 1) total household income per capita income excluding remittances, and 2) only domestic wage income. Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. Incomes are from the KIHS household level income file and harmonized according to ECAPOV guidelines. Panel households are those that are in the sample all five years. Source: World Bank calculations, KIHS

28 CHAPTER 4 THE WELFARE IMPACT OF LABOR MIGRATION 4.1 There are many potential welfare benefits from labor migration for the household. In some countries, the education enrollment of children is higher and child work declines. These are all potential benefits from remittances. On the other hand, labor migration may be present because the family is poor. 4.2 It is not obvious that migrant households are absolutely better off than non-migrant households. Migrant households are poorer ex-ante, and their expenditures can be spent on luxury goods rather than important investment pieces. Since reporting of large purchases in housing tend to be underreported, it is not clear if there is investment by migrant households. 4.3 The ADB (Mogilevsky and Atamanov, 2008) finds that most migrant households receive remittances only for a year and only 10 percent of households receive remittances for more than five years. This suggests that as a household decision, migrants have a temporary role for fast wealth accumulation. As individuals, migrants may choose to remain abroad. Some labor migrants return to the Kyrgyz Republic, primarily for family concerns or to start their own business. An EDB (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013) study reports the most common reason of emigrants from Russian returning to the Kyrgyz Republic is for family concerns. 4.4 This section compares the welfare of migrant and non-migrant households. Income and consumption patterns are analyzed and compared. In terms of income, receipts of remittances are of most interest. Income and consumption are often not reported in a consistent manner, and some effort is made to at least explain the gaps between reporting. For example, the poverty rate of migrant households is higher despite the large reporting of remittance income. A. INCOME 4.5 In this section on the study of household income, household-level income data is used. Household income (questionnaire 6) is listed for individuals 7 years and older for the last 1, 2, or 3 months as designated by the respondent. Compared to individual income from the Labor Force Survey, household income data includes additional categories of income from property and household production. Additionally, if remittances are perceived as gifts, this would also be captured in the household income module but not the Labor Force Survey. The LFS is used to assess labor income of migrants, however in terms of household welfare, the household income file is used. 4.6 Reported income is higher than consumption reporting in the KIHS. This is unusual compared to other countries since usually reported monetary income is less than consumption expenditures in household budget surveys. Underreporting in income is an expected reality in household surveys due to high nonresponse rates. One example of this is the high rate of zero reporting in the wage variable. In the Labor Force Survey, earnings of absent household members or the labor migrants are reported by household members who may not know their income and this is left blank. Moreover, households would be reluctant to report any income that was acquired illegally or informally. In general, consumption values tend to exceed income reporting. Households may be more open about daily food and non-food expenditures than about how much they are earning. 19

29 2005 KGS 2005 KGS 4.7 The contribution of different types of income to total household income per capita is illustrated in Figure 4.1. Income values are taken from the household income file. Remittances are equal to the sum of household income from employment outside of the Kyrgyz Republic and assistance from family or friends who live outside of the Kyrgyz Republic. Figure 4.1: Average Annual household income per capita (ECAPOV, 2005 KGS) Non-Migrant Households Migrant Households 25,000 25,000 20,000 20,000 15,000 15,000 10,000 10,000 5,000 5, Other Domestic Wages Social Assistance Unemployment Property Pension Remittances Notes: Household income components are in annual per-capita terms. Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. Incomes are from the KIHS household level income file and harmonized according to ECAPOV guidelines. Source: World Bank calculations, KIHS Migrant households report much higher income than non-migrant households, primarily due to remittances 16. Household remittance income per labor migrant has been cyclical from 2008 to Migrant households not only have substantial remittance income, it constitutes over half of total income in many cases. All other components of income is lower in migrant households (other, domestic wages, social assistance, unemployment, property, and pension incomes). One important distinction is that wages are domestic wages, excluding labor income from abroad. Labor income from abroad is classified under remittances. 4.9 Over time, migrant household income has been more volatile than non-migrant household income 17. Among non-migrant households, per capita household income steadily increased from 2008 to However among migrant households, income declined, bottomed out in 2010 during the crisis and then began to increase again Aside from the remittance component of income, all other components of income are lower in migrant households. Other income components include domestic wages, social assistance, unemployment, property, pension, and other. The largest gap is in property income. In 2012, non-migrant households had property income that was three times higher than in migrant households. However in 2009 and 2011, this difference was even higher. 16 Household remittance income in includes gifts from friends and family, which is why migrant labor income is an insufficient characterization of remittances. 17 Some of these factors could be associated with the changing composition of migrant households as this data is not a panel and nor it is stratified by migrant status. 20

30 2005 KGS Income ratio 4.11 The domestic wage gap between non-migrant and migrant households has increased over time. In 2008, migrant and non-migrant households had similar wage income from domestic employment (a ratio close to 1). By 2013, the wage income of non-migrant households was 2.1 times higher than migrant households. This is consistent with previous statistics that poorer households are becoming migrant households. It appears that as domestic wages increase, wealthier households are staying in the country. Figure 4.2: Income Ratio (Non-Migrant Household / Migrant Household) Domestic Wages Social Assistance Pension Notes: Income is in per capita terms. Migrant households are those that reported an absent member due to work in the Q1 household roster. Incomes are from the KIHS household level income file and harmonized according to ECAPOV guidelines. Source: KIHS Since migrant households in the Kyrgyz Republic have lower levels of domestic income, remittances do contribute to development. For example in the case of Tajikistan, remittances have been referred to as a social shock absorber, which is a substitute for social protection when the domestic labor market or social assistance is poor (Kireyev, 2006). This paper also finds that pension income is much lower among the elderly in migrant households than non-migrant households. Migration is strongly influenced by a household s economic state and lack of resources and assistance domestically. Figure 4.3: Average Monthly Pension Income HH Heads ,000 12,000 10,000 8,000 6,000 5,022 4,000 1,586 2, ,124 Non-Migrant Household 3,165 8,679 8,584 Migrant Household 10,814 3,909 3,905 4,239 12,398 5,822 Notes: The average pension income is individual (not household per capita) pension income for individuals aged 50+. Individual income is from the Labor Force Survey. Migrant households are defined based on the employment file, if at least one person in the household was working abroad over ] year. Source: KIHS Pension income is much higher in non-migrant households. In 2013, persons over 50 years of age in non-migrant households on average received 12,398 KGS per person in pensions. Compared to those in migrant who received 5,822 KGS per person. The average age of the head of household in a migrant household is less than heads in non-migrant households, but only by a couple of years. This suggests that migrant households may be more vulnerable and remittances are a necessary source of income when pension income is lacking. 21

31 4.14 Fiscally, there is an underlying issue of pension disbursement for labor migrants who return to the Kyrgyz Republic after age 40. Under the current regime, labor migrants in Russia pay 10-13% income tax, 17% non-resident tax, so 30% of their income remains in Russia. The non-resident tax is paid for the first 6 months, and afterwards labor migrants pay taxes like a Russian citizen. However, because there is so much seasonality, the 6 month period restarts a lot. For seasonal migrants, they have no rights to social pensions in Russia, and also no pensions in the Kyrgyz Republic. Governments are currently working on this to get taxes transferred to the Kyrgyz Republic. There is already an agreement like this between Russian and Belarussia. There might be changes with the customs union as well Over time, migrant households exhibit an increasing dependency on remittance income and declining shares of income in domestic wages. From 2008 to 2013, the proportion of remittances as a share of household income has been increasing in migrant households. This holds both in terms of total and per capita income levels. On the other hand, the ratio of domestic wages to foreign remittances has nearly halved over this period. Table 4.1: Trends in Remittance Income, Migrant Households Per-Capita level Remittances as a Share of Total Income (per capita) Domestic Wage to Remittance Ratio Remittances as a Share of Total Income Household level Domestic Wage to Remittance Ratio % 114.4% 34.4% 63.4% % 68.6% 44.8% 48.3% % 83.8% 37.7% 45.7% % 85.5% 39.7% 55.8% % 58.1% 47.4% 42.9% % 59.4% 46.1% 36.8% Source: World Bank calculations, KIHS In the Kyrgyz Republic, personal remittances as a percentage of GDP are one of the highest in the world. In 2012, the Kyrgyz Republic ranked 2 nd in the world in this regard, remittances comprised percent of GDP, or valued at just under 2 billion USD. Though national accounts may be overestimating remittances. Only Tajikistan had a higher rate of remittances as a proportion of GDP, at 47.5 percent (World Bank). However, in terms of value, many other countries receive much larger sums of remittances. India, China, and Mexico receive over 20 billion USD in remittances (World Bank Remittance Trends 2007) In 2013, total household remittances computed using the KIHS totaled just under one billion USD (2005). This is lower than official estimates, by about half, but underreporting of remittance data in a national household survey is expected. The accuracy of national accounting itself is also somewhat questionable. National accounting makes the assumption that any money transfers made by individuals or amounts less than $3000 are remittances. National accounting may be overestimating remittances when some transfers are actually forms of trade. When comparing the distribution of denominations of money transfers through financial institutions and households, household report receiving much smaller denominations. Therefore, larger denominations counted in national accounts are likely not private remittances (Mogilevsky and Atamanov, 2008) At the household level, remittances can have a large impact on welfare. In other countries, remittances and international migration has also been found to reduce poverty (Adams Jr. and Page, 2005). In Latin America and the Caribbean, international remittances have also been found to reduce poverty and inequality (Acosta, Calderon, Fajnzylber, and Lopez, 2008). 22

32 2005 KGS Total Remittances (2005 USD) in billions Share of GDP mln. USD per year Figure 4.4: Total remittance income (2005,PPP) KIHS Survey National Accounts % 35% 30% 25% Dynamics of remittances in KGZ As share of GDP mln. USD % % 10% 5% Notes: Incomes are from the KIHS household level income file and harmonized according to WBG ECAPOV guidelines. Source: KIHS % Assuming av. size of transfer of 370 USD and 6 transactions per year the number (or flow ) of migrants is 1.1 mln. in almost three times of similarly estimated 370 thous. in 2008) 4.19 The poorest 18 migrant households receive the largest amount of remittances per capita. Poverty ranking is based on non-remittance income. In 2013, the poorest 20 percent of migrant households received 15,246 KGS in remittances per person annually, substantially higher than the wealthier 80 percent (Figure 4.5). However, the difference between remittances per capita received by poor and rich households has narrowed over time. It should not be that surprising that household remittance income has not dropped drastically in the data despite a slowing Russian economy since remittance declines begin only in November Figure 4.5: Annual remittance income per capita, by pre-migration income quintile $20,000 $18,000 $16,000 $14,000 $12,000 $10,000 $8,000 $6,000 $4,000 $2,000 $- Bottom 20 40th percentile 60th percentile 80th percentile Top 20 Average Notes: Income quintiles are based on total household income per capita income excluding remittances. Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. Incomes are from the KIHS household level income file and harmonized according to WBG ECAPOV guidelines. Source: KIHS In the Kyrgyz Republic, remittance income plays a very large poverty reduction role for migrant households. From 2008 to 2013, income poverty for migrant households declined a net of 5.6 percentage points. Among migrant households, remittances played the largest role in poverty reduction. Domestic wages on the other hand contributed to explaining an increase in poverty. Incomes of migrant households experienced a net decrease of 311 KGS from 2008 to 2013, with remittances contributing to an 18 Income poverty based on pre-migration income. 23

33 Change Change (2005 KGS) increase of 2808 KGS. The large positive contribution of remittance income helped offset the large negative effects from wage of domestic household members and the household employment share. Figure 4.6: Income Decomposition, Migrant Households Only (2008 to 2013) National Absolute Poverty Rate Poverty Gap Index Squared Poverty Gap Index Mean Bottom (40) Adult Share Share of Employed Wage (per employed) SA (per adult) Pension (per adult) Other (per adult) Remittances (per adult) Unemployment (per adult) Property (per adult) Notes: Annual household income components are in household per-capita terms, and 2005 KGS. Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. Incomes are from the KIHS household level income file and harmonized according to ECAPOV guidelines. The poverty line is the national absolute poverty line. Source: KIHS On the other hand for non-migrant households, domestic income sources rather than remittances had the largest poverty reduction impact. Among these households, pension and domestic wage income contributed the most to their poverty reduction and income growth (Error! Reference source ot found.). However, employment is not contributing to any poverty reduction. The domestic labor market still shows rigidities In some other countries, households receiving remittances have been shown to have higher expenditure on education than non-remittance receiving households (Egypt and Morocco, WB2010a), reduces the probability of child work (Egypt, WB2010a). Yet in the Kyrgyz Republic, children aged in migrant households are less likely to be enrolled in school and more likely to be employed than children in non-migrant households Globally, there is little evidence of an impact from remittances onto economic growth at the national level (Clemens and McKenzie, 2014, World Bank, 2006). As Clemens and McKenzie summarize, Migration and remittances clearly have first-order effects on poverty at the origin.but detecting their effects on growth of the origin economy is likely to remain elusive. Some reasons for this is the use of remittances on immediate consumption needs rather than investment In the case of the Kyrgyz Republic, Akmoldoev and Budaichieva (2012) find that remittances are mostly spent on consumption rather than investment, and therefore reducing its impact on growth. There are some anecdotal evidence that migrant families have improper use of remittances. Assistance on family budget regulation could be very useful to promote productive spending on education, children rather than luxury items. 24

34 2005, KGS B. EXPENDITURES 4.25 In terms of the amount of household expenditure/consumption per capita, migrant and nonmigrant households consume at similar levels; although over time consumption levels for both groups have declined. Figure 4.7 illustrates the trend in average household expenditure per capita between migrant and non-migrant households. The trend in expenditures are similar between non-migrant migrant households; over time average expenditures is declining. Figure 4.7: Average Household Expenditures and Savings per capita (NSC, 2005 KGS) Non-Migrant Households Migrant Households Real Estate Savings Durables Services Non-Food Food Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The welfare aggregate is the national consumption aggregate by the NSC. Source: KIHS One immediate observation is that while income is much higher among migrant households than non-migrant households, average consumption between these two groups are almost identical. Migrant and non-migrant households have similar consumption patterns even though income in migrant households was much higher. So why are consumption patterns so much more similar? There are a couple explanations for this. First migrant households are more likely to make large purchases that are infrequent and not well captured in the survey. Only about 1 percent of the sample reported any purchases of real estate including land, houses, apartments, or cottages. Among those who did report real estate purchase, expenditures among migrant households were higher. However, overall consumption of migrant households is still lower. Another possibility of lower expenditures is higher savings Another reason for the difference between income and consumption reporting among migrant households is higher savings rates. Migrant households are more likely to save than non-migrant households. According to FINDEX (2011), 38 percent of individuals 15 years of age and older in the Kyrgyz Republic saved in the last 12 months. In the KIHS, household reporting of savings is much lower, but migrant households are two to three times more likely to save. An ADB report (Mogilevsky and Atamanov, 2008) also finds savings rates to be higher in migrant households; 59.9 percent of remittance receiving households save compared to 37.3 percent of non-remittance households Table 4.2 shows that migrant households have higher consumption per capita than nonmigrant households controlling for a variety household characteristics. This is important in establishing that both consumption and income are higher among migrant households. However, over time the average per capita consumption of migrant households have reduced and become more similar to non- 25

35 migrant household levels. In recent years, migrant households are also more likely to be from the poorest segments of society. Table 4.2: Predictors of Per Capital Daily Consumption Migrant Household= ** *** *** * ** ** (0.0473) (0.0213) (0.0199) (0.0267) (0.0168) (0.0107) Household Size *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) Age of Head of Household ( ) ( ) ( ) ( ) ( ) ( ) Gender of Head of Household ** ** (0.0201) (0.0285) (0.0116) (0.0164) (0.0191) ( ) Urban ** (0.0343) (0.0265) (0.0160) (0.0151) (0.0308) (0.0505) Share of Children *** *** *** *** *** *** (0.0741) (0.0901) (0.0294) (0.0366) (0.0373) (0.0300) Share of Elderly ** ** (0.136) (0.0614) (0.0251) (0.0359) (0.0209) (0.0658) Primary or No Education 0.256*** 0.256*** 0.270*** 0.189*** 0.148*** 0.134*** (0.0561) (0.0470) (0.0538) (0.0378) (0.0421) (0.0348) General Secondary 0.120** 0.123*** 0.115*** * (0.0509) (0.0411) (0.0376) (0.0266) (0.0255) (0.0485) Specialized Secondary (0.0413) (0.0264) (0.0337) (0.0369) (0.0364) (0.0265) Constant 4.666*** 4.767*** 4.967*** 5.145*** 5.269*** 5.264*** (0.0829) (0.0499) (0.0919) (0.0898) (0.0879) (0.1000) Observations 4,995 4,984 4,951 5,010 5,005 5,013 R-squared Notes: Dep. Var = log(per Capita Daily Consumption) Regions also controlled. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p< One motivation of migration is to accumulate money for large purchases such as housing, property, or to open a business (Taylor et al, 1996). Tynaliev and McLean (2011) conduct a small qualitative study and find that migrant workers are likely to start their own business when returning to the Kyrgyz Republic and also invest in housing and real estate. This can generate employment for neighbors and increase welfare at home Home improvement can be a large area of expenditures but may not be well captured in the KIHS. Higher income may lend itself to purchases or consumption of larger goods and even housing. In urban areas, migrant households are more likely to make repairs to their home. There is some evidence of spending on housing in migrant households. Migrant households tend to have larger living areas, live in newer dwellings. However they are more poorly connected to public grids such as electricity and energy, and also have rougher material for walls and roofs (See Error! Reference source not found. for housing tatistics) Differences in housing materials between migrant and non-migrant households are more apparent in urban areas. In urban areas, materials for roofs and walls in migrant households are much simpler than in non-migrant households. In 2012, 55.4 percent of migrant households had walls made of crude airbricks (adobe) compared to 33.1 percent of non-migrant households. The same is true of wall materials; in urban areas 7.7 percent of migrant households have roofing made from roofing slates compared to 26 percent of non-migrant households (Error! Reference source not found.). In rural areas, he materials for housing between migrant and non-migrant households are more similar Migrant households are not spending more on education. When similar regressions are estimated using education expenditure per capita or total education expenditure as a dependent variable, 26

36 2005 KGS education expenditures do not significantly vary between migrant and non-migrant households. Kroeger and Anderson (2014) also find that remittances did not improve education outcomes of male children in remittance-receiving households between 2005 and On average, migrant households in rural areas have higher educational expenditures than non-migrant households, whereas the converse is true in urban areas. Over time, only in urban areas are education expenditures increasing. This could be for a number of reasons, that families with children move to urban areas with better schools or urban areas have intrinsically higher educational costs. Figure 4.8: Average Migrant Education Expenditures per capita (2005 KGS) Total Annual Education Expenditures Annual Education Expenditures per capita (2005, KGS) (2005, KGS) $3,000 $2,500 $2,000 $1,500 $1,000 $500 $- $2,449 $1,135 $1,237 $920 $734 $769 Migrant Rural Migrant Urban Non-Migrant Rural $1,722 $1,472 Non-Migrant Urban $500 $400 $300 $200 $100 $- $428 $254 $184 $273 $210 $154 Migrant Rural Migrant Urban Non-Migrant Rural $378 $444 Non-Migrant Urban Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. Since there was very little sampling rotation, observing the trend of educational expenditures may be inaccurate as education expenditures change as schooling is completed. Therefore we compare only the earliest and latest years available. Source: KIHS Over time, the average consumption of both rural and urban migrant households have declined, though much for migrant households in urban areas. The spending patterns of migrant households is an important activity that requires sharper analysis. In 2012 and 2013, urban migrant households have lower average consumption per capita than rural migrant households. The deterioration of the welfare of urban migrants is important to understand. Some information that would be very informative to help differentiate different spending behavior is asking households explicitly their reason for having labor migrants. This would be revealing about why certain types of expenditures are highest. Figure 4.9: Average Migrant Household Expenditures and Savings per capita (2005 KGS), Rural vs. Urban $25,000 Real Estate Savings Durables Services Non-Food Food $20,000 $15,000 $10,000 $5,000 $- Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The welfare aggregate is the national consumption aggregate by the NSC (pccddc). 27

37 Growth (%) Source: KIHS C. POVERTY 4.34 The welfare aggregate is household consumption per capita following national definitions. There are some advantages to measuring welfare using consumption as opposed to income. Daily consumption is a better representation of welfare than income that can be lumpy over the year. While income may be smoothed, consumption is a more accurate measure of welfare of living standards throughout the year and the regularity of consumption Shared prosperity is the measure of the mean consumption growth of the bottom 40 percent of the population. Figure 4.10 illustrates the growth of mean consumption for 4 groups of the population; non-migrant and migrant households in the Top 60 and Bottom 40. The relative growth of migrant household consumption very much depends on the period. For example, the growth of migrant households in the bottom 40 was higher than non-migrants from and , but was lower in other years. One period where migrants suffered was between 2011 and In this period, consumption of migrant households declined, while non-migrant households still experienced some amount of positive consumption growth. 25% 20% 15% 10% 5% 0% -5% -10% Figure 4.10: Shared Prosperity 23% 21% 14% 13% 11% 9% 9% 9% 10% 7% 8% 8% 5% 5% 2% 3% 1% 2% -4% -5% Top 60 - Non-Migrant Top 60 - Migrant Bottom 40 - Non-Migrant Bottom 40 - Migrant Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The Bottom 40 and Top 60 percent of the population are defined based on annual household consumption per capita. Source: KIHS Poorer regions have higher rates of migration (Figure 4.11). In 2013, Batken had the highest poverty rate as well as the highest proportion of the working population that were abroad. This does not mean that migration itself is not welfare enhancing since the poverty rates of non-migrant households in Batken might be very high while the poverty rate of migrant households are low. Though it still poses some questions as to how welfare enhancing labor migration can be or if there are sufficient spillovers from labor migration since some researchers believe that wealthy migrant families spend remittance income on neighborhood projects and employing other locals The poverty rate of migrant households has increased over time. Inequality is decreasing among migrant and non-migrant households as well, unfortunately due to increasing poverty. Recall in Figure 3.6, there was descriptive evidence that migrant households were increasingly likely to be from the lower pre-remittance income distribution. Migrant household incomes consist of large shares of remittances that can be sensitive to macroeconomic conditions; without remittance income, they are much poorer than non-migrant households. The poverty rates of non-migrant households have remained relatively stable. However in migrant households, poverty rates have been steadily increasing from 2008 to 2013 (Figure 28

38 Poverty Rate National Poverty Rate 4.12). In 2010, the increase in poverty rate may be related to the revolution or the financial crisis. In 2012, the country experience a period of zero growth and poorer households may have started to send labor migrants abroad. Figure 4.11: Regional Poverty and Migration Rates (2013) 60% Batken 50% Naryn 40% Jalal-Abad Osh 30% 20% Issykul Bishkek Chui Talas 10% 0% 0% 10% 20% 30% 40% 50% 60% Migrants Abroad, % of Population Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The poverty line is pline_dc, national absolute poverty, reflects threshold of meeting food and nonfood needs. The welfare aggregate is the national consumption aggregate by the NSC (pccddc). Source: KIHS The convergence in the poverty rates between migrant and non-migrant households is mainly due to increasing poverty in migrant households. First, it is important to note that poverty here is based on consumption. While migrant households showed much higher incomes than non-migrant households, discrepancies between income and consumption may lead to a different welfare story. Remittances are poverty alleviating for migrant households, yet households that are choosing to migrate are initially (premigration) poorer over time. As noted earlier the number of migrant households are declining as well, so one possibility is that the wealthiest migrant households over time has stopped migrating or has completely moved out of the country. Figure 4.12: National Absolute Poverty Rate ( ) National Absolute Poverty Rate Migrant Households Non-Migrant Households Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The poverty line is pline_dc, national absolute poverty, reflects threshold of meeting food and nonfood needs. The welfare aggregate is the national consumption aggregate by the NSC (pccddc). Source: KIHS

39 Poverty Rate Poverty Rate 4.39 In rural areas, non-migrant households are poorer than migrant households, while in urban areas the opposite is the case. The difference in poverty between migrant and non-migrant households is more apparent in rural areas (Figure 4.13). Migrant households are much better off than non-migrant households in rural areas. However, in urban areas, poverty between migrant and non-migrant household has been much more similar until There is a large divergence in poverty between migrant and nonmigrant households in In rural areas, migrant households become much less poor, and in urban areas these households become much poorer relative to non-migrant households Figure 4.13: National Absolute Poverty Rate, by Rural/Urban ( ) Rural Urban National Absolute Poverty Rate Migrant Households Non-Migrant Households Notes: Migrant households are defined based on the employment file, if at least one person in the household was working abroad over the year. The poverty line is pline_dc, national absolute poverty, reflects threshold of meeting food and nonfood needs. The welfare aggregate is the national consumption aggregate by the NSC (pccddc). Source: KIHS Urban poverty peaked in 2012, and to a larger degree for migrant households was a year of zero growth in the Kyrgyz Republic. While the spike in 2012 urban migrant poverty is substantial, it is consistent with some other indicators from the data. For example, when the value of dwellings for urban migrant households has declined significantly in 2012; valuations in 2012 were a bit more than half of valuations in The wage gap (of non-migrants) is also increasing (Figure 4.2). The rise of urban migrant poverty can be related to many other corroborating statistics: increase in the wage gap with domestic workers, lower employment rate of members in migrant households, lower dwelling values, lower education enrollment rates. It is not clear how much of the increase in urban poverty is due to internal migration. When examining just the panel set of households who are in the survey between 2008 and 2012, this subgroup also experienced sharp increases in poverty in The rise in urban migrant household poverty is driven by Jalal-Abad, Batken, and Osh regions. Contrary, it is not driven by Bishkek since in , there are no migrant households surveyed in Bishkek city. While Bishkek does have a large internal migrant population, it seems that the migrant population in general may be under sampled in this city Trends in poverty are much more volatile among migrant households, which may be related to churning of migrant households. Concerning the length of receiving remittances, over half of households received remittances for one year of less, while only 10 percent of households received remittances for at least five years (Mogilevsky and Atamanov, 2008). Therefore, a household that was a migrant household in one year could easily become a non-migrant household in the following year. There is a factor of selection that contributes to the volatility of welfare indicators. Naturally households that are poorer or are in need of income will become migrant households. However, the magnitude of the spike in urban poverty for migrant households is still puzzling. 30

40 4.43 In the sample of households of the KIHS, the number of migrant households interviewed has been steadily declining, but by an incredibly large proportion in 2012, from 591 households in 2011 to 507 households in Since the sample is not stratified, and sample rotation is small, it is difficult to assess how representative the decline of migrant households is of the actual situation. Further investigation of why the sample of migrant households declined so dramatically and if this is linked with the jumps in poverty trends is an important follow-up activity. The recent volatility of labor migrant welfare should also be a priority for future study. 31

41 CHAPTER 5 CONCLUSION 5.1 Kyrgyzstan is home to a large number of labor migrants and the amount of remittances as a percentage of GDP is one of the highest in the world. Labor migration increased substantially in the early 2000 s. To ensure that this activity is sustainable in the long-term, it is important to work to enhance the impact of remittances and to properly facilitate labor migration in all phases, before, during, and after migration. Policies at all stages are relevant as the impact of labor migration and its trends in remittances are vulnerable to regional events and the economic health of Russia. 5.2 In recent years, migrant households are becoming more vulnerable to economic shocks since they have higher dependency on remittances. Moreover, in 2013, migrant households are poorer than nonmigrant households. Therefore, policies to increase the development impacts of remittances are more relevant than ever. How do we increase the development impacts of remittances? This question can be framed from the perspective of how to increase the amount of potential labor earnings through premigration policies, how to make it easier to send remittances home, and how to educate migrant households to make more productive use out of these remittances. McKenzie and Yang (2014) review the evidence of policies that increase the development impact of international migration. 5.3 Based on what could be analyzed about labor migrant households and comparing them relative to non-migrant households, a number of policy recommendations are suggested both along regulatory and non-regulatory dimensions. 5.4 From the regulatory side, a stronger government presence is important. Government intervention and oversight can improve welfare and migration outcomes. In the Kyrgyz Republic, migrants abroad are less educated and rely on networks rather than government programs when finding work abroad (Beishenaly, Levant, Ormonbekova, and Shamsiev, 2013). This can lead to underpayment and informal work environments. To reduce dependence and regional risk, migrants should diversify their choice of destination countries. The Kyrgyz Republic s reliance on Russia and Kazakhstan may largely be a language issue. The institution of language schools can be a solution to increase the choice of destination countries. Moreover, the government may pursue bilateral labor agreements with select destination countries. Agreements can ensure the placement of labor migrants under well-known conditions. 5.5 There are many learning opportunities that can be offered through international experiences and through conducting migration surveys and data monitoring of return migrants. The Philippines can offer valuable experience and expertise in managing labor migration policies, where the government is very interactive with the management and facilitation of labor migrants, before, during, and after. Regional lessons can also be drawn from existing World Bank projects in other regions. Labor migration unarguably plays a large role in the economic welfare of Kyrgyz citizens. While this report profiles labor migrants and their households to the best extent possible 19, there are still important knowledge gaps that need to be filled that cannot be answered using existing surveys. For example, what are return migrants doing and where do they work? Do young migrants return to school? Are the number of labor migrants really declining? These are all important questions to understand the longer run impacts of labor migration. A migration survey would be valuable to accurately understand the welfare of migrant households, and to overcome the challenges in using a household budget and labor force survey that is not well suited to answering migration questions. A targeted survey will also make it possible to follow return migrants and understand their 19 There are many challenges to studying migration without a proper survey honed to examine this topic specifically. For example, the lack of sample rotation in the KIHS does not ensure year to year representativeness. 32

42 integration into the domestic labor market or lack thereof. This is especially important if the government needs to anticipate a potentially large return migration of young and low educated workers. These return migrants may not prefer to enroll in school but continue to be bread winners for their families. 5.6 Policies along the lines of financial literacy, skills training, and guarding against shocks also areas that can improve the well-being of migrant households. Migrant households in the Kyrgyz Republic underspend remittances in investment (Akmoldoev and Budaichieva, 2012). The development of financial instruments for migrants such as micro-insurance or savings products to invest at home can help enhance the impact of remittances. Knowledge on financial literacy is powerful for reducing the cost of sending remittances home, increasing the returns from remittances through investment and savings, and increasing the sustainability of remittances as a source of income. Financial training such as budgeting and money management can also be a useful tool especially for pre-migrants. For example, in the Philippines, the government provides financial training before labor migrants depart as well as deposit options to minimize the cost of sending remittances. Pre-migration training and government support can create pathways for labor migrants into better jobs in higher skilled sectors. The bulk of labor migrants work in low skilled occupations and the majority have completed only general secondary education. Training, upgrading of skills, and recognition of skills in countries of destination would also help migrants when they return home to get better jobs. Income and consumption patterns of migrant households are more volatile than patterns in non-migrant households. Moreover, the poverty rates of migrant households are increasing over time. A stronger study on the vulnerabilities and social status of labor migrants would be useful. The government may also wish to help migrant households think of innovative ways to protect against shocks by developing appropriate financial instruments International organizations and the government the Kyrgyz Republic are familiar with many of the recommendations listed above. There is great progress on many fronts as the Kyrgyz Republic works to better serve its labor migrants. For example, the IOM has helped launch employment centers in Bishkek and Osh. The World Bank has facilitated technical assistance and expert consultations on improving the legal and institutional framework for migration management. These efforts should continue as well as filling knowledge gaps for the purposes of enhancing the development impacts of labor migration. 33

43 ANNEX A1: Tables and Figures Table A1: Characteristics of Workers, by Location of Work Domestic Internal Migrant Labor Migrant Abroad Age Female 45.0% 44.0% 43.3% 43.2% 43.0% 41.3% 28.1% 30.6% 28.9% 33.3% 33.8% 39.0% 24.3% 26.6% 29.4% 28.2% 25.4% 25.8% Male 55.0% 56.0% 56.7% 56.8% 57.0% 58.7% 71.9% 69.4% 71.1% 66.7% 66.2% 61.0% 75.7% 73.4% 70.6% 71.8% 74.6% 74.2% Head 39.7% 39.4% 39.1% 38.7% 38.6% 39.7% 24.0% 26.5% 23.6% 24.1% 26.8% 22.5% 7.3% 4.6% 5.4% 7.6% 6.6% 6.0% Spouse 21.9% 21.2% 21.1% 20.9% 21.1% 20.7% 12.3% 13.9% 12.4% 11.4% 11.8% 9.7% 13.8% 13.2% 12.0% 10.4% 12.0% 10.2% Children 28.8% 30.4% 30.3% 30.6% 30.6% 29.6% 54.4% 50.1% 55.9% 56.7% 55.7% 57.8% 65.9% 65.3% 67.8% 67.6% 68.1% 70.9% Other 9.6% 9.0% 9.5% 9.8% 9.7% 10.1% 9.3% 9.6% 8.1% 7.8% 5.7% 9.9% 13.0% 16.9% 14.8% 14.4% 13.3% 12.9% Not Married 31.1% 31.6% 31.0% 30.8% 29.1% 29.0% 51.9% 50.1% 52.3% 54.2% 54.4% 54.0% 50.8% 53.4% 49.5% 49.5% 49.6% 53.2% Married 68.9% 68.4% 69.0% 69.2% 70.9% 71.0% 48.1% 49.9% 47.7% 45.8% 45.6% 46.0% 49.2% 46.6% 50.5% 50.5% 50.4% 46.8% Higher 18.4% 18.5% 18.6% 19.1% 19.6% 19.2% 16.7% 14.7% 11.6% 16.1% 19.2% 18.3% 8.8% 10.3% 10.5% 10.0% 10.3% 8.3% Specialized Secondary 26.1% 26.0% 26.7% 25.3% 25.1% 19.6% 21.4% 26.6% 25.9% 25.9% 25.0% 26.7% 21.0% 17.7% 15.5% 19.6% 16.3% 7.6% General Secondary 47.3% 47.1% 46.4% 47.5% 47.6% 53.8% 52.0% 48.9% 54.1% 48.2% 46.1% 47.7% 63.4% 63.6% 65.4% 61.6% 66.9% 79.4% Primary 8.0% 8.2% 8.0% 7.9% 7.4% 7.3% 10.0% 9.7% 8.4% 9.8% 9.7% 7.3% 6.7% 8.4% 8.6% 8.7% 6.5% 4.7% None 0.2% 0.2% 0.3% 0.3% 0.3% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.1% 0.1% 0.0% 0.0% 0.2% 0.0% 0.0% Agriculture & Mining 38.9% 37.7% 36.6% 36.2% 34.8% 36.1% 2.0% 0.8% 1.4% 1.2% 2.0% 1.7% 1.9% 1.4% 1.3% 0.6% 1.0% 0.1% Manufacturing 8.2% 7.5% 7.5% 7.3% 7.4% 7.0% 12.1% 17.9% 17.8% 17.0% 16.7% 18.1% 6.6% 7.2% 6.9% 6.2% 6.1% 3.5% Utilities 1.9% 1.9% 2.0% 1.8% 2.2% 1.9% 2.0% 2.4% 2.7% 2.3% 3.0% 1.1% 0.2% 0.1% 0.2% 0.3% 2.1% 3.5% Construction 5.6% 6.3% 6.7% 7.3% 7.9% 8.3% 25.3% 22.2% 22.5% 18.8% 16.1% 16.6% 49.3% 48.2% 40.9% 39.4% 41.3% 35.3% Retail 13.4% 13.1% 13.6% 13.0% 13.0% 13.9% 22.2% 18.0% 19.4% 24.3% 25.4% 22.9% 24.9% 23.4% 26.6% 30.6% 31.3% 35.0% Transport 6.0% 6.8% 7.0% 6.9% 6.2% 6.5% 9.9% 8.3% 8.5% 8.1% 6.8% 8.2% 6.2% 4.1% 2.9% 2.4% 4.3% 3.8% Services 5.6% 6.3% 6.1% 6.1% 8.7% 7.8% 7.5% 10.9% 10.9% 10.1% 11.3% 14.6% 7.9% 11.2% 14.2% 14.2% 11.8% 14.2% Other 20.4% 20.4% 20.6% 21.3% 19.7% 18.6% 18.9% 19.5% 16.9% 18.2% 18.7% 16.9% 3.0% 4.5% 6.9% 6.1% 2.0% 4.6% High Medium 21.9% 21.9% 21.8% 21.8% 21.5% 20.5% 16.2% 15.3% 13.3% 15.1% 15.6% 16.3% 1.9% 1.4% 1.6% 1.5% 1.2% 0.9% Services 17.9% 17.9% 18.0% 18.0% 18.7% 18.9% 27.8% 25.4% 24.6% 28.9% 31.7% 30.5% 23.6% 20.8% 20.4% 19.3% 21.6% 18.0% Agriculture 36.2% 34.6% 33.5% 33.1% 32.2% 31.6% 4.6% 3.3% 2.2% 0.4% 1.1% 0.4% 1.7% 1.0% 1.1% 0.2% 0.1% 0.3% Skilled Manufacturing 20.0% 21.3% 21.7% 21.6% 22.2% 25.4% 41.6% 45.6% 45.6% 39.7% 39.7% 45.7% 57.0% 54.5% 46.9% 44.5% 45.4% 36.2% Unskilled Manufacturing 4.0% 4.3% 5.0% 5.5% 5.5% 3.5% 9.7% 10.4% 14.3% 15.8% 12.0% 7.1% 15.8% 22.2% 30.0% 34.4% 31.7% 44.6% Open-end written labour contract 62.0% 60.2% 59.5% 58.1% 55.1% 55.1% 36.0% 35.7% 29.1% 34.3% 37.7% 39.9% 12.0% 10.5% 15.8% 16.7% 7.1% 3.7% Fixed-term written labour contract 6.7% 6.5% 8.0% 9.7% 10.7% 6.1% 10.0% 6.8% 10.6% 8.4% 6.4% 2.4% 10.8% 9.0% 7.7% 7.2% 8.4% 6.4% Labour or other civil law contract in writing 3.4% 3.4% 3.0% 2.8% 4.4% 5.5% 3.2% 4.7% 6.3% 5.9% 9.1% 7.7% 18.8% 21.0% 26.2% 21.8% 39.1% 52.3% Verbal arrangement 27.9% 30.0% 29.6% 29.4% 29.8% 33.3% 50.8% 52.7% 54.0% 51.4% 46.7% 50.0% 58.4% 59.5% 50.3% 54.3% 45.5% 37.6% Notes: Employed workers (during the week of the survey) years of age Source: KIHS

44 Table A2: Logit, Dep. Variable = Migrant Household ln(pension) -8.00e-05** -4.12e e e-05** -1.30e-05* -1.89e-05 (3.19e-05) (3.42e-05) (1.07e-05) (7.57e-06) (6.77e-06) (1.34e-05) ln(sa) ** *** -2.80e e e-05 ( ) ( ) ( ) (3.12e-05) (8.42e-05) (4.12e-05) Household Size ** ** (0.0845) (0.0711) (0.0651) (0.0607) (0.0607) (0.0341) Age of Head of HH ** ** *** ( ) ( ) ( ) ( ) ( ) ( ) Gender of Head of HH *** *** *** ** *** (0.404) (0.209) (0.0725) (0.0876) (0.163) (0.0948) Urban *** *** ** ** ** *** (0.158) (0.176) (0.277) (0.296) (0.221) (0.151) Share of Children *** *** *** *** *** (0.329) (0.321) (0.0917) (0.263) (0.393) (0.387) Share of Elderly *** ** *** *** *** * (0.481) (1.299) (0.480) (0.308) (0.378) (0.458) Primary or No Education * * (0.337) (0.229) (0.643) (0.602) (0.613) (0.265) General Secondary (0.267) (0.284) (0.509) (0.493) (0.634) (0.347) Specialized Secondary *** (0.388) (0.486) (0.445) (0.474) (0.526) (0.142) Constant *** *** *** *** *** *** (0.728) (0.530) (0.645) (0.544) (0.521) (0.340) Observations 1,363,535 1,405,929 1,363,934 1,416,609 1,453,357 1,426,116 Mc-Fadden's R-square Notes: Education of the head of household, and regions also controlled. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<.1. Weights applied Table A3: Individual wages of domestic workers Migrant Household *** *** *** * ** * (0.0132) (0.0123) (0.0134) (0.0133) (0.0143) (0.0173) Male 0.219*** 0.161*** 0.140*** 0.147*** 0.152*** 0.137*** (0.0118) (0.0113) (0.0115) (0.0117) (0.0114) (0.0119) Married *** *** *** *** *** *** (0.0126) (0.0120) (0.0118) (0.0119) (0.0115) (0.0121) General Secondary 0.156*** 0.129*** 0.114*** 0.132*** 0.130*** 0.116*** (0.0183) (0.0169) (0.0170) (0.0172) (0.0168) (0.0190) Specialized Secondary 0.212*** 0.181*** 0.152*** 0.186*** 0.182*** 0.171*** (0.0192) (0.0178) (0.0177) (0.0180) (0.0177) (0.0204) Tertiary Education 0.435*** 0.398*** 0.375*** 0.346*** 0.334*** 0.345*** (0.0212) (0.0196) (0.0195) (0.0198) (0.0193) (0.0214) Constant 6.896*** 7.183*** 7.366*** 7.265*** 7.301*** 7.571*** (0.0307) (0.0288) (0.0298) (0.0304) (0.0310) (0.0332) Observations 18,969 19,339 19,188 19,607 19,452 18,153 R-squared Notes: Dep. Var = annual wages. Other controls include gender, age group, relation to head of household, education, oblast, marital status, occupation, and profession. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 35

45 Figure A1: Poverty Map 2009 Poverty Rate Number of Poor Notes: WB and NSC Poverty Mapping Exercise Source: Mahadevan, Yoshida, Praslova (2013) 36