Remittances and labor supply of the left-behind youth: Evidence from Kyrgyzstan

Abstract

This paper examines whether remittances from international migration impact on the occupational choice of left-behind youth in Kyrgyzstan. Labor supply is analyzed both at the extensive and intensive margins using cross-sectional data for 2011. To overcome endogeneity concerns, an instrumental variable approach was implemented. Findings demonstrate that migration, rather than remittances, pushes the left-behind youth to become unpaid family workers. This is explained by the substitution effect as the youth left behind are called upon to replace the migrant labor. Moreover, this effect is heterogeneous – female youth are more inclined to becoming unpaid family workers both at the extensive and intensive margins.

Keywords

remittances labor supply youth Kyrgyzstan

Introduction

The welfare impact of migrants' remittances has been widely acknowledged by the economic literature. Empirical studies on the impact of remittances on poverty show that they contribute considerably to sustaining the livelihood of households (Acosta, 2007; Adams, 2006; Gupta et al., 2009). This is particularly important in Kyrgyzstan, a country exhibiting high migration rates and among the poorest in Eastern Europe and Central Asia. Because of high migration trends over the last seven years, mostly to Russia and Kazakhstan, remittance inflows into the Kyrgyz Republic have significantly increased, placing the country not only among the top remittance-receiving countries but also as a country where remittances account for 30 percent of its gross domestic product (Ratha et al., 2016). However, this might not have enhanced the labor market conditions of the left-behind population, particularly the youth (i.e., those in the 15–29 age group), which accounts for almost one-third of the total population and almost half of the unemployed (National Statistical Committee of Kyrgyz Republic, 2016). Such a large share of the youth population requires high and sustained job creation rates in order to keep youth unemployment at low rates. However, the country's economic performance and job creation have lagged behind since it attained independence in 1991 (Baumann et al., 2013; Esengul et al., 2012; Schwegler-Rohmeis et al., 2013). These factors make the youth population a vulnerable segment of Kyrgystan's labor force and raise questions regarding the role of remittances on youth labor market outcomes.

According to the literature, migration can exert different effects on the labor supply of the left-behind household members. On the one hand, the migration of a household member may cause reallocation of labor within the household to replace the migrants' labor or income. On the other hand, remittances may dissuade household members from labor market participation through a reservation wage channel (for example, see Amuedo-Dorantes and Pozo, 2006; Binzel and Assaad, 2011; Mendola and Carletto, 2009).

Although several previous studies on remittances in Kyrgyzstan focused on household welfare and agricultural productivity effects, there are no empirical studies addressing the relationship between remittances and labor market outcomes of the left-behind youth. Thus, the main objective of this research is to study the impact of remittances from international migration on the labor supply of left-behind youth at the intensive and extensive margins.¹ This paper also contributes to the empirical literature by exploring such relationship in one of the most remittances-dependent countries, Kyrgyztan, which is also suffering severe youth unemployment rates. The empirical analysis builds on the “Life in Kyrgyzstan” household survey for 2011. The main methodological concern relies on the endogeneity of remittances which is addressed by implementing an instrumental variable approach. One key finding from the study is that remittances exert a labor substitution effect for the left-behind young people, that is, that they are more likely to become an unpaid family worker in remittance-receiving households.

The paper is structured as follows. The next section provides an overview of the literature on remittances, previous studies on the impact of remittances in Kyrgyzstan, and the impact of remittances on the labor supply of those left-behind. This is followed by a section on descriptive statistics on migrants, remittance-receiving households and employment. The next three sections discuss the econometric model and estimation methodology, the estimation results and finally, the conclusions and policy implications.

Literature review: Impact of remittances on households

Empirical studies on the impact of remittances on poverty using household level data show that it reduces poverty and mitigates social tensions (Adams, 2004; Airola, 2007; Brown and Jimenez, 2008; Lokshin et al., 2007; Zhu and Luo, 2008). By increasing household income, remittances ease financial constraints and, hence, may generate investment that goes into education and health or stimulate new entrepreneurial activities (Edwards and Ureta, 2003; Woodruff and Zenteno, 2007). Nevertheless, arguments on the economic consequences of remittances are controversial. A classical and widely accepted view is that remittances are rarely used for productive investments, but mostly are channelled to current consumption (Chami et al., 2003).

Also, according to the neoclassical model of labor–leisure choice, remittances reduce the labor supply (at the intensive margin) of remittance-receiving individuals through a non-labor income effect (Killingsworth, 1983). Along with this effect, remittances may reduce the likelihood of the labor force participation of the left-behind household members (Airola, 2008; Cabegin, 2013; Funkhouser, 1992; Justino and Shemyakina, 2012; Kim, 2007; Lucas, 1987; Rodriguez and Tiongson, 2001). However, other studies found little to no significant effect of migration or remittances on labor supply and decisions on labor force participation (Görlich et al., 2007; Piracha et al., 2013).

Findings also pointed out that labor supply effects of remittances may differ by gender and location characteristics. In particular, for households that receive remittances, left-behind women are likely to shift from wage work to informal and non-paid economic activities. This effect may be more evident in rural areas (Amuedo-Dorantes and Pozo, 2006; Binzel and Assaad, 2011; Lokshin and Glinskaya, 2009).

Several studies have addressed the specific question of remittances and labor supply by age groups, particularly as they effect the youth. For instance, Braga (2009) indicates that in Albania, remittances decrease labor force participation of youth. Analogously, Petreski and Mojsoska-Blazevski (2015) find that Macedonia's left-behind youth have a considerably larger probability to establish their own businesses.

There are only few studies focused on remittances in the Kyrgyz Republic. Some of these studies have examined the impact of remittances on the expenditure of households on durable goods, children's education and health (Kroeger and Anderson, 2014; Ukueva, 2010). Atamanov and Van den Berg (2012) looked into the effect of remittances on crop income in Kyrgyzstan. They found a higher negative impact of lost labor from permanent migration than the positive impact of remittances, while seasonal migration positively affects crop production. However, little is known about the impact of remittances on the youth labor supply in the Kyrgyz Republic. To the best of our knowledge, this is the first study attempting to quantitatively evaluate the impact of remittances on left-behind youth at the intensive and extensive margins.

In summary, the economic literature suggests that migration and remittances may have heterogeneous effects on the labor supply of remittance-receiving households. We can classify these different effects in four groups. First, remittances may create an income effect for the left-behind household members through increase of the reservation wage and discourage them from participating in the labor market. This translates into a decrease in labor supply observed through higher unemployment or fewer working hours. Second, migration may cause a labor force substitution effect in migrant-sending households. Non-migrant members of a household may increase their labor supply in order to supplement the loss of workforce caused by migration. In Kyrgyztan, this may be reflected by an increase of unpaid family members contributing to work due to the predominance of household agricultural activities across the country. Third, remittances from migrant workers may act as a financial resource pushing left-behind members to self-employment activities that require funding. Fourth, remittances may have a neutral impact on the labor supply if none of the abovementioned effects dominate.

Data and descriptive statistics

As mentioned earlier, data for this study came from the 2011 “Life in Kyrgyzstan” (LiK) survey,² which includes a sample of 3,000 households and 8,066 individuals. It is representative at the national level and for urban and rural areas. This survey was conducted by DIW Berlin in collaboration with Humboldt University of Berlin, the Center for Social and Economic Research and the American University of Central Asia. The Law of the Kyrgyz Republic of 31 July 2009 No. 256 on fundamentals of the state youth policy defines the youth population as those in the ages 14–28 years. Respondents in the LiK survey, however, were from 15 years old and older. Thus, this paper defines the youth population as those aged between 15 and 28. The sample includes the active youth labor force; those enrolled in full time education or people with disabilities are excluded. The final sample consists of 1,633 youth.

The survey includes a wide range of data, including information on household characteristics, assets, shocks, social networks, income and expenditure. More importantly, the survey contains a special module on migration and remittances. The cross-sectional data used in this study do not allow us to analyze the impact of migration and remittances over a longer period of time. In the absence of administrative data on the migrant labor force, estimates offered by experts suggest that some 10 to 20 percent of the active population may have migrated, mainly to Russia (National Institute for Strategic Studies of the Kyrgyz Republic and International Organization for Migration, 2016: 58).

Many reports on migration in Kyrgyzstan (see, for example, International Organization for Migration, 2015: 27) mention the economic situation in the country as one of the key factors for migration and for receiving remittances. In our dataset, 67 percent of migrants are male and more than 60 percent are in the 16–28 age range. Moreover, most labor migrants from Kyrgyzstan are unskilled workers with secondary education, whose main destination is Russia. Ninety-three percent of current migrants are working in Russia and 80 percent are unskilled workers in construction, trade, etc.

The share of households receiving remittances accounts for almost 16 percent of the total sample.³ Basic household characteristics presented in Table 1 indicate that most of the remittance-receiving households come from the southern part of the country: Jalal-Abad, Osh and Batken regions. The rural population are more inclined to migrate, hence, rural households are more likely to receive remittances. Remittance-receiving households have more members – specifically, they have, on average, two young adults able to work compared to one among non-receiving households. Despite the important contribution of remittances to recipient households (where they account for 30 percent of total income), remittance-receiving households are the poorest.

Table 1.

Basic characteristics of remittance-receiving and non-receiving households.

		Remittance-receiving households		Non-receiving households
	Total	Number	%	Number	%
Total households	2,831	451	15.9	2,380	84.1
Households by region
South (Osh, Jalal-Abad, Batken)	1,279	367	28.91	912	71.08
Central (Chui, Bishkek city)	1,046	57	5.2	989	94.72
North (Talas, Naryn, Issyk-Kul)	506	27	5.5	479	94.4
Residence
Urban	1,153	148	13.65	1,005	86.35
Rural	1,678	303	18.69	1,375	81.30
Household head ethnicity
Kyrgyz	1,883	324	17.9	1559	82.1
Uzbek	322	73	23.0	249	77.0
Russian	358	23	6.6	335	93.4
Other	268	31	11.6	237	88.4
Household structure
Household size (mean)	4.86	5.92	−	4.64	−
Child ratio (aged between 0–5 years, mean)	0.104	0.093	−	0.106	−
Male adult number (15–65 ages, mean)	1.52	2.14	−	1.40	−
Youth labor force (mean)	1.19	1.98		1.03
Total household expenditure per capita (mean, in Kyrgyzstan Som)	38,512	28,394	−	40,546	−

Source: Authors' calculations.

Almost 30 percent of the sample are young individuals 15–28 years old; 64 percent of the labor force are concentrated in the rural areas. The distribution of the labor force by employment status is given in Table 2. About 37 percent of both old and young in the sample are salaried employees. Own-account workers comprise 29 percent of the old population, while it is considerably lower for the youth. The largest proportion of the youth population is employed as unpaid family workers, while unemployment is high for adults and the youth population. Moreover, the share of employers and private business owners among both old and young groups, is very low.

Table 2.

Employment status of individuals by age group (in %).

	Young (15–28)			Old (29–65)
	Total	Male	Female	Total	Male	Female
Salaried employee	35.45	35.33	35.63	37.26	35.46	36.95
Own-account worker	15.27	21.32	6.00	29.16	43.48	14.07
Unpaid family worker	23.50	23.17	24.01	9.2	4.23	15.84
Employer	0.29	0.28	0.29	1.03	1.45	0.42
Other	0.25	0.11	0.46	0.48	0.58	0.40
Member of producers' cooperatives	0.25	0.21	0.32	0.26	0.13	0.41
Unemployed	25.00	19.58	33.29	22.60	14.67	31.9
Total	100	100	100	100	100	100

Source: Authors' calculations.

The survey data show variations in employment status by gender: the unemployment rate is considerably higher among females in both the young and old age groups (15–28 and 29–65). Male employment rates as own-account workers are considerably higher in the old age group. Among the old age group, women work more as unpaid family workers than men. Among young individuals, men compared to women are employed as own-account workers at higher rates and are less unemployed. They have similar participation as salaried and unpaid family workers. In both young and old age groups, there are no gender differences in the rates for salaried employment. In general, it may be noted that salaried employees and own-account workers constitute the bulk of the labor force in the old age group and only a small proportion are employed as unpaid family workers, while young individuals are mostly employed as salaried employees and unpaid family workers, and less as own-account workers, particularly the women.

Overall, the descriptive statistics indicate that migration in Kyrgyzstan is mainly explained by economic factors. Migration mostly occurs in those communities where economic conditions are poor and people have difficulty in finding jobs. There is a high incidence of unemployment and unpaid family worker employment among the youth, particularly in remittance-receiving households. Also, there are gender gaps in youth employment.

Under these conditions, the inflow of remittances from migrants abroad may have several effects on the labor supply of the left-behind youth. Firstly, it may create an income effect and discourage their labor participation, thus increasing youth unemployment. Or it may cause a labor force substitution effect to replace migrant workers in agricultural households. This may translate into an increase of unpaid family workers. Finally, a large inflow of remittances may contribute to the development of entrepreneurship initiatives at sending communities, that is, an increase in self-employment activities.

Econometric specification

Remittances are expected to impact on the intensive and extensive margins of the left-behind youth population through a non-labor income effect. A higher non-labor income at the household level may increase household members' reservation wage, discouraging them from employment or supplying more hours of work (Airola, 2008; Amuedo-Dorantes and Pozo, 2012; Binzel and Assaad, 2011; Justino and Shemyakina, 2012). This is grounded on the static neoclassical model of labor supply and leisure where individuals maximize their utility subject to budget constraints, that is, the allocation decision space is bounded by total (labor and non-labor) income resources. If leisure is a normal good, an increase in non-labor income will decrease labor supply (at the intensive margin) while increasing leisure. Similarly, the participation decision is explained by the reservation wage, which indicates the minimum wage at which individuals are willing to work. Thus, non-labor income raises the reservation wage and, therefore, dampens incentives to participate in the labor market (Killingsworth, 1983).

Therefore, we analyze labor supply response first with respect to the youth labor supply decision among several occupational choices (extensive margins), followed by an assessment of the remittances impact on working hours (intensive margins).

Individual labor supply at the extensive margin (occupational choice) is a categorical nominal variable determined by observable and unobservable factors (such as unobserved abilities or choice specific characteristics). Thus, it is modeled by a multinomial probit model (MNP). By assuming an underlying multivariate normal distribution, the MNP allows for common unobservable factors to affect the many choice probabilities. Imposing a non-correlated error structure,⁴ when the data generating process is characterized by a full structure, would imply biased and non-consistent estimations (Winship, 2003). The ‘independent’ variable of interest is a binary indicator ( $y_{2 i}$ ) equal to one for remittance-receiving households and zero otherwise. Household and community level explanatory variables suggested by the literature are included as controls ( $x_{1 i}$ ). Formally, this is represented by the following equations (baseline-model-1):

v_{i j'} = η_{i j'} - η_{ik} = Z_{i} γ_{j'} + ɛ_{i j'}

(1)

where

η_{i j'}

stands for i-th individual unobserved random of utility that results from choosing status

j'

. Thus

v_{i j'}

is a differential utility with respect to a reference alternative (k). Occupational choice alternative ranges from 1 to 4 for a sample of n individuals, that is,

j = 1, 2, 3, 4

;

i = 1, \dots, n

. The unobserved utility factors are modeled by normally distributed random components

ɛ_{i j'}

∼N(0,1). These are independent and identically distributed within a given category but may be correlated across alternatives which captures the fact that some occupational choices may be related to factors beyond the observable ones (

Z_{i}

Thus, the probability that the i-th individual chooses outcome k is defined by the probability of getting the highest random utility from the k-th alternative ( $η_{ik} > η_{i j'}$ for $j' \neq k$ ):

Pr (y_{1 i} = k) = Pr (v_{i 1} \leq 0, \dots, v_{i, J - 1} \leq 0)

(2)

y_{2 i} = x_{i} θ + ξ_{i}

(3)

where

y_{1 i}

is the multinomial dependent variable indicating i-th (young) individual labor market status or occupational choice (own-account worker, employee, unpaid family worker and unemployed);

y_{2 i}

is the remittances dummy equal to one if the i-th young individual belongs to a remittance-receiving household and zero otherwise and

Z_{i} = (y_{2 i}, x_{1 i}), x_{i} = (x_{1 i}, x_{2 i})

, and

γ

is a vector of structural parameters.

Addressing endogeneity in remittances

However, unobserved factors – such as unobserved skills or attitudes to risk – that determine migration and remittance decisions may also drive the labor market decisions of those left-behind. Lower skilled households may be prone to migration while having higher preferences for a given occupational choice. Similarly, risk-taking households may be inclined to migration while exhibiting a preference toward riskier occupational choices such as self-employment. This can be formalized within a household decision-making framework where both migration decision and the occupational choices of the left-behind youth are performed simultaneously. Furthermore, household labor market outcomes (low employment quality) may clearly motivate members' migration decisions leading to reverse causality of our relation of interest. Hence, the effect of remittances on labor supply would face an endogeneity issue as acknowledged by the literature (Adams and Cuecuecha, 2010; Amuedo-Dorantes and Pozo, 2006; Bettin et al., 2012).

Thus, the multinomial probit model must assess the endogeneity of remittances by including equation (3) into the structural specification. For presentation purposes, equation (3) is written as a linear probability model, but estimated from a probit specification. Then, as defined above, $x_{i}$ contains $x_{1 i}$ — the vector of exogenous variables at individual, household and community levels; and $x_{2 i}$ — the vector of instrumental variables that allow for the identification of remittances effect on labor status.

The instrumental variable used to correct for endogeneity should be correlated with remittance-receiving status and meet the exclusion restriction, that is, it should not affect the labor supply choice of young household members, except through its effect on remittances. We define our instrument as the share of households in the community that have at least one member who lived abroad in the past five years for labor purposes. This share is a measure of migration networks at the community level in the last five years. A higher share implies more information for potential migrants and lesser migration costs, which in turn facilitates the migration process and increases the possibility of being a remittance-receiving household. Similar instruments have been widely used in the literature (see, for example, Acosta, 2007; Atamanov and Van den Berg, 2012; Binzel and Assaad, 2011; Lokshin and Glinskaya, 2009). One may argue that this instrument (share over the community) is mechanically linked to the migration dummy. Nevertheless, such a link tends to disappear by avoiding small samples within communities. This instrument may be thought of as a natural experiment where community-level variation in migration networks is exploited to identify remittances effects on our labor market outcomes of interest.

Occupational choice

A major concern is that migration networks, our instrumental variable, may also have an impact on the occupational choices of the left-behind youth. This may correlate with unobserved factors at the community level that may be relevant to explain labor market outcomes, such as, occupational choice. We attempt to address this concern by introducing relevant community level variables. It is noteworthy that the non-linearity between the remittances dummy $(y_{2 i})$ and its determinants $(x_{1 i}, x_{2 i})$ through the normal cumulative distribution function (probit) may ease the identification of the remittances effect on occupational choices $(y_{1 i})$ . An analog identification approach is available for Heckman models without exclusion restrictions (Bushway et al., 2007; Leung and Yu, 2000; Puhani, 2000).

Since our sequential model consists of a multinomial dependent variable and a binary endogenous regressor under given parametric assumptions, the model is estimated by Limited Information Maximum Likelihood following Roodman (2011). The estimation procedure is available in Stata through the CMP (Conditional Mixed Process) package.

Remittances may exhibit heterogeneous effects on occupational choice decisions depending on recipients and their household characteristics. Thus, additional models, other than baseline-model-1, are specified by including remittances interactions with gender, age and household poverty status. Equation (4) shows i-th individual's probability of being at labor status k, which is a non-linear function (g) of the explanatory variables

Pr (y_{1 i} = k) = g (b_{0} + b_{1} x_{1 i} + α_{i} y_{2 i})

(4)

where g is the multivariate cumulative distribution function and the impact parameter

α_{i}

depends on individual, household and community characteristics

(x_{ij})

as follows:

α_{i} = \sum x_{ij} θ_{j} = x_{i 1} θ_{1} + x_{i 2} θ_{2} + \dots + x_{ij} θ_{j}

(5)

where

θ_{j}

are parameters. A particular case where

α_{i}

is determined by two variables of interest (

x_{1 i}

and

x_{2 i}

) is given below,

Pr (y_{1 i} = k) = g (b_{0} + b_{1} x_{1 i} + (x_{i 1} θ_{1} + x_{i 2} θ_{2}) y_{2 i})

(6)

Pr (y_{1 i} = k) = g (b_{0} + b_{1} x_{1 i} + θ_{1} y_{2 i} x_{i 1} + θ_{2} y_{2 i} x_{i 2})

(7)

Working hours

Labor supply may be characterized not only by occupational choices, but also by working hours within a particular occupation. Therefore, the second approach for analyzing the impact of remittances on youth labor supply is to investigate its impact at the intensive margin. Because of the zero-inflated nature of worked hours, an ordinary least squares estimate would yield biased and inconsistent results (Amuedo-Dorantes and Pozo, 2006). A zero-inflated behavior results from the fact that many individuals in the sample are not employed, which implies an excess of zero hours of work with respect to a parametric framework (normally distributed working hours). To address this issue and the concern that remittances are correlated with the error term, we specify a Tobit model with endogenous regressors or IV-Tobit. This model enables us to fit estimation with censored dependent variables and endogenous regressors. Formally, our baseline model 2 is (Wooldridge, 2010):

{w_{1 i}}^{*} = w_{2 i} β + x_{1 i} α + u_{i}

(8)

w_{2 i} = x_{1 i} Π_{1} + x_{2 i} Π_{2} + v_{i}

(9)

The dependent variable ${w_{1 i}}^{*}$ is the total hours of work supplied by youth during the last seven days; $w_{2 i}$ is the endogenous regressor, that is, the amount of remittances received by a household during the last 12 months in the national currency (Kyrgyzstani Som or KGS); $x_{1 i}$ is a vector of exogenous control variables at the individual and household levels; $x_{2 i}$ is the instrumental variable, migration network.

The model assumes that $(u_{i}, v_{i})$ follow a bivariate normal distribution with zero mean. Finally, $β, α$ are vectors of structural parameters while $Π_{1}$ and $Π_{2}$ is a vector of reduced-form parameters.

Non-linear effects on working hours

We also explore potential non-linear effects of remittances on working hours since standard labor supply theory suggests the presence of non-linear relationships between working hours and non-labor income (remittances). We use a semi-log labor supply equation following Airola (2008) and Posso (2012). Furthermore, we examine the labor supply response for males and females by augmenting baseline model 2 (equations 8 and 9).

The literature indicates different explanatory variables within

x_{1 i}

(Justino and Shemyakina, 2012; Mendola and Carletto, 2009; Rodriguez and Tiangson, 2001). In this study, we use individual, household and community level characteristics as explanatory variables of youth labor supply (see Table 3). At the individual level, it includes the individual's age, gender, marital status and educational attainment.

Table 3.

Definition of variables.

Dependent variables:
Labor participation	1 = individual is unemployed 2 = individual is own-account worker 3 = individual is employee 4 = individual is unpaid family worker
Working hours	The number of hours worked during the last 7 days
Explanatory variables
Remittances	1 = household receives remittances, 0 = otherwise
Remittances per capita	The amount remittances received by household per capita
Individual characteristics
Gender	1 = male; 0 = female
Age	Age (years)
Marital status	1 = ever married; 0 = single
Educational category
Basic, secondary education	1 = individual has basic and secondary education; 0 = otherwise
Technical education	1 = individual has technical education; 0 = otherwise
Tertiary education	1 = individual has tertiary education; 0 = otherwise
Household characteristics
Household's head occupation
Employee	1 = household's head is employee, 0 = otherwise
Own-account worker	1 = household's head is own-account worker, 0 = otherwise
Family contributing worker	1 = household's head is unpaid family worker, 0 = otherwise
Retired or handicapped	1 = household's head is retired or handicapped, 0 = otherwise
Unemployed	1 = household's head is unemployed, 0 = otherwise
Household size	The total number of household members
Child ratio	The ratio of children in household, aged between 0 and 5 years
Quality of home construction	1 = the wall of main dwelling is made of mud; 0 = otherwise
Non-labor income	The total of non-wage income the household receives (in logarithm)
Regional dummies
Issyk-Kul	1 = the household resides in Issyk-Kul oblast, 0 = otherwise
Jalal-Abad	1 = the household resides in Jalal-Abad oblast, 0 = otherwise
Naryn	1 = the household resides in Naryn oblast, 0 = otherwise
Batken	1 = the household resides in Batken oblast, 0 = otherwise
Osh	1 = the household resides in Osh oblast, 0 = otherwise
Talas	1 = the household resides in Talas oblast, 0 = otherwise
Chui	1 = the household resides in Chui oblast, 0 = otherwise
Bishkek and Osh cities	1 = the household resides in Bishkek or Osh cities, 0 = otherwise
Community characteristics
Share of male adults with regular work	The proportion of the male population of working age with regular job
Presence of bank	1 = presence of bank or place for savings deposit, 0 = otherwise
Presence of major factory	1 = presence of major factory/other source of employment, 0 = otherwise
Migration network	Share of households which had labor migrants abroad in the past 5 years of total households in the community

Likewise, the vector $x_{1 i}$ incorporates a variety of household and regional characteristics that are used in shaping the labor supply decisions of its members, such as household composition, the level of other sources of non-labor income, assets of the household and dummy variables reflecting the locations of households. As Mendola and Carletto (2009) note, these household characteristics do not directly affect the potential market wages, but influence the decision to stay home by increasing or decreasing the individual reservation wage. The rationale behind the inclusion of the control variables is described below.

While household size controls for the economies of scale effect on labor supply, the child ratio controls for the children effect on adult's reservation wages and labor supply (Amuedo-Dorantes and Pozo, 2012). Non-labor income is expected to reduce the likelihood of being at work according to neoclassical theory (Airola, 2008; Binzel and Assaad, 2011; Justino and Shemyakina, 2012). To control for household's wealth effect on labor supply, the quality of home construction is used following previous studies (Acosta, 2006; Amuedo-Dorantes and Pozo, 2012).

Regional labor market heterogeneity is partly accounted for by including regional dummies. Following the other studies, employment opportunities at the community level are also proxied by the inclusion of the share of male adults with regular jobs in the community (see, for example, Binzel and Assaad, 2011). The existence of a major factory (employer) is also accounted for by a dummy regressor. Finally, the community's access to financial resources, which may foster entrepreneurial activities and employment, is accounted for by introducing a dummy variable on the existence of commercial banks in the community.

Estimation results

Table 4 reports baseline-model-1 estimates as marginal effects where participation in the labor market, the dependent variable, is defined as: own-account worker, employee, unpaid family worker and unemployed (extensive margin). An MNP and its Instrumental Variable version (IV-MNP) are estimated through the CMP that does not impose the independence of irrelevant alternatives assumption and allows for an endogenous regressor. The results discussed below refer to the consistent IV-MNP model estimates unless said otherwise.

Table 4.

Estimation results for baseline model 1 (marginal effect estimates).

	Own-account worker		Employee		Unpaid family worker
	MNP	IV MNP	MNP	IV MNP	MNP	IV MNP
Remittances (1 = household receives remittances)	0.0204	0.1134	0.0852	0.2673	0.1172	0.4204***
Individual characteristics:
Gender (1 = male)	0.4854***	0.4694***	0.2905**	0.3072***	0.2570***	0.2522***
Age	0.0582***	0.0580***	0.0961***	0.0853***	0.0201**	0.0212**
Marital status (1 = ever married)	0.1541**	0.1441**	−0.0739	−0.0607	0.0527	0.0386
Educational category (reference: tertiary education)
Basic, secondary education	0.2025***	0.1960***	−0.6139**	−0.3990*	0.1195	0.1288
Technical education	0.3691***	0.3542***	−0.0804	−0.0281	0.1876	0.1720
Household characteristics:
Household's head occupation (reference: unemployed)
Employee	−0.0235	−0.0082	0.4009**	0.3134**	0.1763	0.2139*
Own-account worker	0.1506	0.1449	−0.0848	−0.0311	0.4035***	0.3780***
Unpaid family worker	0.3781**	0.3342*	−0.1299	−0.0341	0.6344***	0.5969***
Retired or handicapped	0.0803	0.0819	0.3808*	0.2881*	0.1805	0.1772
Household size	−0.0032**	−0.0331**	−0.0403*	−0.0418**	−0.0161	−0.0232*
Child ratio	0.6588***	0.6664***	0.4688	0.5453*	0.3590*	0.4601**
Quality of home construction (1 = mud)	−0.0021	−0.0002	−0.0250	−0.0045	0.1706***	0.1774***
Non-labor income (log)	0.0609***	0.0605***	−0.0986**	−0.0597	0.1026***	0.1067***
Regional dummies (reference: Bishkek and Osh cities)
Issyk-Kul	0.2215	0.2539**	−0.3102	−0.3784***	0.5076***	0.5273***
Jalal-Abad	0.1104	0.1013	−0.3382**	−0.3642***	0.5742***	0.5245***
Naryn	0.2065	0.2294	−0.3038	−0.4243***	0.2848	0.2969*
Batken	0.3166***	0.2954**	−0.2919*	−0.3181**	0.5792***	0.5292***
Osh	0.2663**	0.2375**	−0.1734	−0.1142	0.6636***	0.5936***
Talas	−0.0179	−0.0214	−0.2299	−0.2128	0.6750***	0.6478***
Chui	−0.1505*	−0.1283	0.0321	0.0155	0.2102	0.2070
Community characteristics:
Share of male adults with regular work	0.0159***	0.0154***	0.0150***	0.0153***	0.0188***	0.0187***
Presence of bank	0.1219*	0.1086*	−0.0631	−0.0603	−0.0906	−0.1011*
Presence of major factory	−0.1361**	−0.1231*	0.1929	0.1037	−0.2314***	−0.2306***
Number of observations	1,633	1,633
Log likelihood	−1627.72	−2225.34
Wald χ2 (probability)	449.76 (0.000)	724.64 (0.000)

Notes: *, ** and *** show statistical significance at the 10, 5 and 1% level, respectively. Reference group for dependent variable: Unemployed youth.

Source: Authors' calculations.

Extensive margin: Effects of remittances

While the MNP shows that no significant relation exists between employment status and remittances received, the IV-MNP model, that controls for endogeneity, shows that, remittances have a positive impact on the ‘unpaid family worker’ choice. This ‘migration’ rather than ‘remittances’ effect is consistent with previous studies (Görlich et al., 2007; Piracha et al., 2013) that find that left-behind household members tend to work in family businesses in order to replace the migrant worker. The first stage results of the IV-MNP shows the high statistical significance of migration networks, our instrument, on the probability of receiving remittances (its t-statistic being 8.8). The conditional correlation between the disturbance terms of the second (labor participation) and first stage (receiving remittance) is also controlled for by the estimated conditional correlation parameter.⁵

Heterogeneous effects of remittances

The effects of remittances may differ according to individual characteristics. Thus, Table 5 presents the results of alternative estimations, which assume heterogeneous effects conditioned on income levels, age and gender. According to these results gender and age interaction variables appear to be significant. Young people, under the age of 22, in remittance-receiving households show higher probability to become unpaid family workers, whereas, female young people in remittance-receiving households are prone to self-employment and unpaid family work.⁶ These results are not different from the general ones but are informative in terms of the higher intensity of remittances effects on the younger and female population.

Table 5.

Estimation results for baseline model 1 with gender, age, poverty and sectoral interactions (coefficient estimates).

	Own-account worker		Employee		Unpaid family worker
	MNP	IV-MNP	MNP	IV-MNP	MNP	IV-MNP
Age interaction model¹
Remittances × age young	0.34 (0.27)	0.58 (0.42)	0.58 (0.60)	1.04 (2.13)	0.46* (0.25)	1.61*** (0.41)
Remittances × age old	−0.20 (0.26)	−0.17 (0.45)	−0.05 (0.51)	0.62 (2.57)	0.16 (0.25)	−0.03 (0.43)
Gender interaction model¹
Remittances × female	0.49 (0.30)	0.89** (0.42)	0.31 (0.65)	−0.03 (0.78)	0.58** (0.27)	1.66*** (0.36)
Remittances × male	−0.25 (0.25)	−0.50 (0.40)	0.17 (0.66)	0.01 (0.72)	0.06 (0.25)	0.71 (0.44)
Poverty interaction model¹
Remittances × non-poor	0.3008 (0.2934)	0.3151 (0.6281)	0.1034 (0.5554)	−0.0129 (1.1922)	0.2926 (0.319)	0.7892 (0.6544)
Remittances × poor	−0.1131 (0.2457)	0.2707 (0.4303)	0.3653 (0.5282)	0.7332 (1.0511)	0.3169 (0.218)	0.5899 (0.4549)
Sectoral interaction model²
Remittances × non-agricultural sector	0.4616** (0.2165)	−0.3617 (0.6054)	−	−	1.5519*** (0.2619)	−0.8253 (0.6290)

Notes: *, ** and *** show statistical significance at the 10, 5 and 1% level, respectively. Standard errors presented in parentheses.

Reference group for dependent variable: unemployed youth.

Reference group for dependent variable: wage employed youth

Each model includes control variables presented in the baseline model, such as individual, household and community characteristics, but coefficients for these variables are not reported. Full estimation outputs and first stages of the models are available from the authors upon request.

As mentioned above, the remittances effect on labor supply may be heterogeneous across sectors. For instance, households involved in small-scale agricultural activities may be more concerned about labor substitution effects. Thus, we extended our analysis by including an interaction between remittances and a non-agricultural sector dummy and found no significant results. This verifies that there are no labor substitution effects beyond agricultural households.

Effects of control variables

The estimated coefficients of other explanatory variables have the expected signs and significance levels. The results show that gender differences are statistically significant. Thus, young men's participation as an own-account worker is significantly higher than for women. Similar gender gaps in occupational choices were identified by Mendola and Carletto (2009), who note that being female decreases the probability of working in the remunerative labor market and increases the likelihood of being an unpaid worker.

Age is a significant factor in explaining every employment status (i.e., own account worker, employee and unpaid worker) though the effect is smaller for the last status. This is expected given that young people at later ages are better able to find a job due to having acquired work experience and other skills.

These results also imply that working as an own-account worker among young people may not require tertiary education because engaging in small-scale trade does not require high skills. As such, young persons facing difficulties in finding paid work may likely get into self-employed activities. As Piracha and Vadean (2010) noted in the Albanian case, own-account workers have characteristics closer to those who are not participating in the labor market, for example, having lower education levels, while entrepreneurship is related to higher education levels. As mentioned above, in Kyrgyzstan, most of the individuals who own land are considered as own-account workers. This means that even if individuals are not employed in any other sectors (but just work as family unpaid agriculture workers with low productivity), they are considered to be own-account workers. Having basic and secondary education reduces the probability of being an employee, as wage employment, especially in government organizations, requires higher education; those with these levels of education are more likely to become unpaid family workers. In general, it can be concluded that individuals with more education are likely to have higher labor participation in remunerative work.

The occupational characteristics of household heads have a symmetrical impact on youth's occupational choice. This evidence underlines the fact that in the labor market, where job opportunities are restricted, there is a high probability that the youth labor supply is path-dependent on the parental occupation.

The composition of the household appears to significantly influence youth labor participation. Household size (correlated with low welfare) increases unemployment likelihood. Child density (child ratio) increases the probability of being an unpaid family worker which can be explained by the need for childcare assistance that pushes some youth members to stay at home and contribute to care work. Family wealth may have significant influence on the youth labor supply. Higher wealth may support a young person's choice to be a salaried or own-account worker and reduce the household need for unpaid family workers. From our data, wealth is proxied by a binary indicator of whether the main material of the walls of the family house is constructed with mud (low wealth). As expected, our results show that households whose main dwelling is made of mud walls are more likely to have young people as unpaid family workers. The non-labor income of a household increases the probability of youth labor force participation as unpaid family workers and own-account workers.

Regional dummies show that young people in almost all regions out of Bishkek and Osh cities have a higher probability of being unpaid family workers or own-account workers. These results underline the uneven distribution of economic opportunities across regions. The cities of Bishkek and Osh have higher concentrations of enterprises, while the low economic activity in other regions implies limited employment opportunities for young people. Other variables indicating the importance of the existence of employment opportunities for occupation choice are the share of male adults in the community who have a regular job and a dummy variable on the existence of a major factory offering employment in the community. The estimates indicate that communities characterized by greater employment opportunities are associated with less youth unemployment and more participation in paid jobs.

Financial facilities (lack of financial constraints) are proxied by a binary indicator that reflects the existence of a commercial bank at the community. This has positive effect on the own-account worker choice, while reducing the unpaid family work choice of youth. This result implies that financial facilities contribute to the labor participation of young people into entrepreneurship.

Labor supply at the intensive margin: Effects of remittances

Remittances may have an impact not only on occupational choice but may also change the intensity of labor supply (intensive margin) measured by working hours. Because of the endogeneity of remittances, a Tobit and IV-Tobit models of working hours are specified (baseline model 2). Since remittances may exhibit a non-linear effect on working hours, a log-linear relationship is specified. The results shown in Table 6 below refer to the IV-Tobit unless stated otherwise.

Table 6.

Estimation results for baseline model 2.

	Employee		Own-account worker		Unpaid family worker
	Tobit	IV-Tobit	Tobit	IV-Tobit	Tobit	IV-Tobit
Remittances per capita	−0.0001 (0.0002)	0.0006 (0.0010)	0.00004 (0.0004)	0.0001 (0.0016)	0.00038* (0.0002)	0.0035*** (0.0010)
Individual characteristics:
Gender (1 = male)	6.5327** (2.6779)	6.7109** (2.6991)	37.5530*** (5.5886)	37.617*** (5.6176)	2.2232 (3.1200)	3.4464 (3.3308)
Age	2.2574*** (0.5258)	2.3011*** (0.5311)	3.4072*** (0.9399)	3.4165*** (0.9438)	−1.0066* (0.5864)	−0.9254 (0.6246)
Marital status (1 = ever married)	−3.8250 (2.9674)	−4.2782 (3.0361)	11.2351** (5.4765)	11.1315** (5.5538)	1.7433 (3.4846)	0.1638 (3.7302)
Educational category (reference: tertiary education)
Basic, secondary education	−15.8297*** (3.6139)	−15.925*** (3.6352)	25.9797** (7.9397)	25.9438** (7.9467)	14.8226** (5.7485)	14.2241** (5.9676)
Technical education	−2.1023 (4.7794)	−2.2179 (4.8081)	28.9611** (10.0209)	28.9170** (10.0302)	8.0693 (7.5093)	7.3999 (7.8135)
Household characteristics:
Household's head occupation (reference: unemployed)
Employee	9.3640** (4.3390)	10.1354** (4.4686)	−17.6414** (8.9003)	−17.4793* (9.0161)	10.8648 (7.1003)	14.2465* (7.4297)
Own-account worker	−7.7288* (4.6670)	−7.9211* (4.6936)	2.7127 (8.3455)	2.6597 (8.3590)	25.8205*** (6.5165)	24.9101*** (6.7835)
Unpaid family worker	−18.5686* (10.6632)	−21.0318* (11.1573)	0.7392 (15.6779)	0.1747 (16.4658)	29.5129** (9.0705)	18.5874* (10.3132)
Retired or handicapped	7.8809* (4.7385)	8.0914* (4.7675)	−3.3108 (8.7897)	−3.2806 (8.7943)	7.2587 (6.8528)	7.6833 (7.1303)
Household size	−0.6187 (0.6185)	−0.7443 (0.6414)	−1.4014 (1.0933)	−1.4265 (1.1163)	0.7667 (0.7151)	0.2346 (0.7773)
Child ratio	8.8967 (9.8396)	12.5039 (10.8906)	38.4781** (17.8166)	39.2903** (19.2373)	−4.3795 (12.5295)	12.5143 (14.2028)
Quality of home construction (1 = mud)	−2.7727 (2.9370)	−2.4975 (2.9711)	−6.4847 (5.3762)	−6.4424 (5.3899)	11.8355** (3.7318)	12.6531** (3.9477)
Non-labor income (log)	−3.7641*** (0.8880)	−3.5626*** (0.9275)	4.7247** (1.7080)	4.7785** (1.7749)	6.9779*** (1.3138)	8.3434*** (1.4519)
Regional dummies (reference: Bishkek and Osh cities)
Issyk-Kul	−31.4450*** (6.5543)	−30.358*** (6.7194)	20.9782** (10.1444)	21.2268** (10.3860)	37.309*** (9.3341)	41.562*** (9.7505)
Jalal-Abad	−23.5015*** (5.4577)	−24.197*** (5.5543)	6.6579 (9.8630)	6.5303 (9.9277)	45.010*** (8.9386)	42.419*** (9.1183)
Naryn	−28.506*** (7.6789)	27.291*** (7.8576)	35.0432** (11.0517)	35.3381** (11.3637)	24.9965** (10.6695)	29.9253** (11.1787)
Batken	−22.8284*** (6.1419)	−23.869*** (6.3103)	21.5481** (10.6235)	21.2761* (10.8972)	44.519*** (9.3102)	38.352*** (9.6801)
Osh	−13.6545** (4.5871)	15.1516** (4.9845)	8.9964 (8.6928)	8.6241 (9.3036)	48.345*** (8.3617)	40.672*** (8.7490)
Talas	−20.1152** (9.3046)	20.1214** (9.3401)	−28.2272 (19.5694)	−28.2445 (19.5718)	61.388*** (10.7055)	60.144*** (11.1878)
Chui	1.9984 (4.3383)	2.5471 (4.4151)	−9.2780 (8.8556)	−9.1497 (8.9293)	26.3107** (8.6028)	28.0716** (8.7592)
Community characteristics:
Share of male adults with regular work	0.1636* (0.0914)	0.1798* (0.0942)	0.5221** (0.1684)	0.5264** (0.1729)	0.9238*** (0.1169)	0.9886*** (0.1248)
Presence of bank	−0.5282 (3.4056)	−0.4364 (3.4210)	18.5639** (5.7168)	18.5840** (5.7203)	−12.077*** (3.6429)	−12.7723** (3.8901)
Presence major factory	7.1717** (3.5254)	6.6708* (3.5956)	−6.6538 (6.4273)	−6.8027 (6.5639)	−17.419*** (4.7881)	−19.982*** (5.0828)
Constant	−13.4781 (17.1446)	−18.833 (18.5217)	281.669*** (35.1404)	283.01*** (37.1358)	−220.09*** (24.9979)	−244.90*** (27.5516)
Number of observations	1,630	1,630	1,631	1,631	1,631	1,631
Likelihood ratio χ2 (probability)	416.70 (0.000)	−	217.91 (0.000)	−	543.32 (0.000)	−
Log likelihood	−3579.502	−20177.71	−1774.1182	−18381.76	−2184.816	−18780.86
Pseudo R²	0.0550	−	0.0579	−	0.1106	−
Wald χ2 (prob.)	−	350.0 (0.000)	−	140.63 (0.000)	−	256.71 (0.000)
Wald test of exogeneity	−	0.62	−	0.01	−	9.73***

Notes: *, ** and *** show statistical significance at the 10, 5 and 1% level, respectively. Standard errors presented in parentheses.

Source: Authors' calculations.

Per capita remittances (in log scale) increase the working hours of young unpaid family workers only (see Table 7). Therefore, remittances are shown to have analogous effects both at the intensive and extensive margins in terms of the substitution effect.

Table 7.

Estimation results for baseline model 2 with log and gender interactions.

	Employee		Own-account worker		Unpaid family worker
	Tobit	IV-Tobit	Tobit	IV-Tobit	Tobit	IV-Tobit
Semi-log baseline model 2
Log of remittances per capita	0.0940 (0.9744)	2.3535 (3.7762)	−0.8010 (1.7419)	0.5390 (6.4308)	1.6695 (1.0296)	14.1894*** (4.0831)
Semi-log baseline model 2 with gender interactions
Log of remittances per capita *female	−1.1625 (1.4673)	−2.5528 (4.7154)	2.2888 (2.8785)	10.057 (8.488)	1.8356 (1.5145)	17.4657*** (4.8181)
Log of remittances per capita *male	0.9768 (1.2318)	5.4228 (4.2452)	−2.2106 (2.0681)	−3.7418 (7.067)	1.5504 (1.3028)	11.4665** (4.7036)

Notes: *, ** and *** show statistical significance at the 10, 5 and 1% level, respectively. Standard errors presented in parentheses.

Source: Authors' calculations.

Heterogeneous effects of remittances

To account for the gender gap remittance effect at the intensive margin, a model with gender interaction dummies with log remittances per capita is estimated (see Table 7). The results confirm the estimation of baseline model 2. Gender dummies' interaction with log per capita remittances are statistically significant for unpaid family workers only. These results show that both male and female youth in remittance-receiving households work more hours as an unpaid family worker than their counterparts in non-receiving households. More specifically, women tend to increase their working hours more than men, which is in line with the high extensive margin effect found for women.

Effects of control variables

Individual characteristics have significant effects on the working hours of salaried and own-account workers. For instance, males are more inclined to work longer hours as own-account workers and, to a lesser extent, as salaried workers. The significant gender gap of own-account workers may be related to the specifics of this occupational choice, which requires physical strength. Young individuals with basic and secondary education have reduced working hours as employees, while young people with the same educational characteristics in other occupation choices exhibit higher working hours.

The variable used as proxy for household wealth – construction of the family house from mud – shows that less wealth results in more working hours for young unpaid family workers. On the other hand, another proxy for household wealth, non-labor income, shows a significant positive impact on the working hours of unpaid family workers.

Regional dummies show that employed young people from almost all regions compared to Bishkek and Osh cities, have less working hours, while unpaid family workers demonstrate more working hours. This result may be explained by the fact that in regions where employment opportunities are restricted, wage employment is mostly employment in government organizations. The private sector is not sufficiently developed to generate more working hours from employees in these regions. Moreover, because of limited job opportunities, young people who are unpaid family workers have more working hours. This is analogous to the labor participation effect of community characteristics such as the existence of a commercial bank and factory. The presence of such employment opportunities reduces the intensity of unpaid family work and provides possibilities for salaried employment or as own-account workers. These findings are also in accordance with findings on labor participation effects.

Conclusions

This paper sought to investigate the impact of remittances on the labor supply of young people who are left behind, which was analyzed in terms occupational choices and working hours (extensive and intensive margins). This study contributes to the literature by its focus on the left-behind youth since previous studies mostly analyzed all members of the left-behind household (Braga, 2009; Petreski and Mojsoska-Blazevski, 2015).

Our main finding suggests that migration, rather than remittances, pushes the left-behind youth to becoming unpaid family workers, mainly in agriculture. A substitution effect occurs as the left-behind youth is called upon to replace migrant workers. Moreover, this effect is heterogeneous since women and younger household members showed a higher propensity toward becoming unpaid family workers. The substitution effect is also verified at the intensive margin where women were found to increase their working hours as unpaid family workers at a higher rate than men.

Unlike previous studies (Airola, 2008; Justino and Shemyakina, 2012; Kim, 2007), we did not find any strong evidence of remittance-dependency behavior by the left-behind household members. That is, we did not find significant evidence of remittance effects on participation decisions through higher reservation wages or self-employment or entrepreneurship activities. This result is in line with Görlich et al. (2007).

From a policy perspective, the migration substitution effect toward unpaid family work employment is not expected to enhance the productivity of the youth labor force. On the contrary, it may threaten skills development and labor market participation in sectors other than agriculture. Thus, policy should focus on the promotion of the productive use of remittances to increase the labor market opportunities for young people, particularly young women.

Aside from remittance effects, our estimates provide insightful evidence about Kyrgyzstan's youth labor market. Wealthier labor force participants are less likely to be unemployed or to stay as own-account or unpaid family workers, whereas agricultural households are rather characterized by their propensity toward unpaid family work. This can be explained by the lack of entrepreneurial skills among the youth. Moreover, the findings suggest that the lack of financial facilities in the community could explain the propensity to remain as unpaid family workers. Higher economic activity is strictly related to lower unemployment among young people in the regions. Our results also verify that gender gaps exist in terms of occupational choices: overall, young men are more likely to participate in the labor market as own-account workers or salaried workers than young women.

Footnotes

Acknowledgments

The authors are grateful to Luca Tiberti, Paola Ballon, Francis Teal, Guy Lacroix and Jean-Yves Duclos for technical support and guidance, as well as for their valuable comments and suggestions.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this paper.

Funding

This research work was carried out with financial and scientific support from the Partnership for Economic Policy (), with funding from the Department for International Development of the United Kingdom (or UK Aid), the Government of Canada through the International Development Research Center.

Labor supply at the extensive margin refers to the occupation choice of individuals while intensive margin refers to the working hours in a given occupation.

For more details about the survey, see http://lifeinkyrgyzstan.org/

The household survey follows a stratified random sampling where households are sampled according to the weight of each region in the overall population.

Also known as the independence of irrelevant alternatives (IIA).

First stage estimation results are available from the authors upon request.

In order to examine further factors that explain why females are prone to becoming unpaid family workers, we estimated several specifications that interacted gender with ethnicity, marital status and presence of children at home. However, these specifications did not yield significant results. These estimations are available from the authors upon request.

References

Acosta P (2006) Labor Supply, School Attendance, and Remittances from International Migration: The Case of El Salvador. Policy Research Working Paper 3903. Washington, DC: World Bank. Available at: https://www.cbd.int/financial/charity/elsalvador-remittance.pdf.

Acosta

(2007) Entrepreneurship, labor markets, and international remittances: Evidence from El Salvador. In: Özden

Schiff

(eds) International Migration, Economic Development and Policy. Trade and Development, Washington, DC: World Bank and Palgrave Macmillan, pp. 141–159.

Adams R Jr (2004) Remittances and Poverty in Guatemala. Policy Research Working Paper 3418. Washington, DC: World Bank. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.615.1237&rep=rep1&type=pdf.

Adams

Jr (2006) Remittances, poverty and investment in Guatemala. In: Özden

Schiff

(eds) International Migration, Remittances and the Brain Drain, Washington, DC: World Bank and Palgrave Macmillan, pp. 53–80.

Adams

Jr Cuecuecha

(2010) Remittances, household expenditure and investment in Guatemala. World Development 38(11): 1626–1641.

Airola

(2007) The use of remittance income in Mexico. The International Migration Review 41(4): 850–859.

Airola

(2008) Labor supply in response to remittance income: The case of Mexico. The Journal of Developing Areas 41(2): 69–78.

Amuedo-Dorantes

Pozo

(2006) Migration, remittances, and male and female employment patterns. The American Economic Review 96(2): 222–226.

Amuedo-Dorantes

Pozo

(2012) Remittance income volatility and labor supply in Mexico. Southern Economic Journal 79(2): 257–276.

10.

Atamanov

Van den Berg

(2012) Heterogeneous effects of international migration and remittances on crop income: Evidence from the Kyrgyz Republic. World Development 40(3): 620–630.

11.

Baumann A, Jansova E and Saar E (2013) Transition from School to Work in Kyrgyzstan: Results of the 2011/12 Transition Survey. Turin, Italy: European Training Foundation. Available at: http://www.etf.europa.eu/webatt.nsf/0/F468B191CAD95B8BC1257BBE005530DF/$file/School-to-work%20transition_Kyrgyzstan.pdf.

12.

Bettin

Lucchettia

Zazzaro

(2012) Endogeneity and sample selection in a model for remittances. Journal of Development Economics 99(2): 370–384.

13.

Binzel

Assaad

(2011) Egyptian men working abroad: Labor supply responses by the women left behind. Labor Economics 18(Supplement 1): 98–114.

14.

Braga M (2009) When the manna comes from abroad. Remittances and youth labor market behavior in Albania. Paolo Baffi Centre Research Paper 2009-41. Milan, Italy: Bocconi University.

15.

Brown

RPC

Jimenez

(2008) Estimating the net effects of migration and remittances on poverty and inequality: Comparison of Fiji and Tonga. Journal of International Development 20(4): 547–571.

16.

Bushway

Johnson

Slocum

(2007) Is the magic still there? The use of the Heckman two-step correction for selection bias in criminology. Journal of Quantitative Criminology 23(2): 151–178.

17.

Cabegin

(2013) Gendered labor supply response to Filipino spouses' overseas migration and remittances. Asian and Pacific Migration Journal 22(2): 147–175.

18.

Chami R, Fullenkamp C and Jahjah S (2003) Are Immigrant Remittances Flows a Source of Capital for Development? Working Paper WP/03/189. Washington, DC: International Monetary Fund. Available at: https://www.imf.org/external/pubs/ft/wp/2003/wp03189.pdf.

19.

Edwards

Ureta

(2003) International migration, remittances, and schooling: Evidence from El Salvador. Journal of Development Economics 72(2): 429–461.

20.

Esengul

Mamaev

Yefimova-Trilling

(2012) Youth and Public Policy in Kyrgyzstan, Berlin, Germany: Youth Policy Press.

21.

Funkhouser

(1992) Migration from Nicaragua: Some recent evidence. World Development 20(8): 1209–1218.

22.

Görlich DM, Toman O and Trebesch Ch (2007) Explaining Labor Market Inactivity in Migrant-Sending Families: Housework, Hammock, or Higher Education. Working Paper 1391. Kiel, Germany: Kiel Institute for World Development. Available at: https://www.cgdev.org/sites/default/files/archive/doc/events/9.12.08/Hammock_Kiel_Working_Paper_final.pdf.

23.

Gupta

Pattillo

Wagh

(2009) Effect of remittances on poverty and financial development in Sub-Saharan Africa. World Development 37(1): 104–115.

24.

International Organization for Migration (IOM) (2015) Mapping on Irregular Migration in Central Asia 2014. Astana, Kazakhstan: IOM. Available at: http://iom.kg/wp-content/uploads/2015/01/IDFPublicationeng.pdf.

25.

Justino

Shemyakina

(2012) Remittances and labor supply in post-conflict Tajikistan. IZA Journal of Labor & Development 1(8): doi:10.1186/2193-9020-1-8.

26.

Killingsworth

(1983) Labor Supply, Cambridge, UK: Cambridge University Press.

27.

Kim N (2007) The Impact of Remittances on Labor Supply: The Case of Jamaica. Policy Research Working Paper 4120. Washington, DC: World Bank. Available at: http://documents.worldbank.org/curated/en/926051468044132562/pdf/wps4120.pdf.

28.

Kroeger

Anderson

(2014) Remittances and the human capital of children: New evidence from Kyrgyzstan during revolution and financial crisis, 2005–2009. Journal of Comparative Economics 42(3): 770–785.

29.

Leung

(2000) Collinearity and two-step estimation of sample selection models: Problems, origins, and remedies. Computational Economics 15(3): 173–199.

30.

Lokshin M, Bontch-Osmolovski M and Glinskaya E (2007) Work-Related Migration and Poverty Reduction in Nepal. Policy Research Working Paper 4231. Washington, DC: World Bank.

31.

Lokshin

Glinskaya

(2009) The effect of male migration on employment patterns of women in Nepal. The World Bank Economic Review 23(3): 481–507.

32.

Lucas

(1987) Emigration to South Africa's mines. American Economic Review 77(3): 313–330.

33.

Mendola M and Carletto G (2009) International Migration and Gender Differentials in the Home Labor Market: Evidence from Albania. Policy Research Working Paper 4900. Washington, DC: World Bank.

34.

National Institute for Strategic Studies of the Kyrgyz Republic (NISSKR) and International Organization for Migration (IOM) (2016) Kyrgyzstan Extended Migration Profile 2010–2015. Bishkek, Kyrgyzstan: IOM. Available at: https://auca.kg/uploads/Migration_Database/Publications/Migr_Profile_KG_2012_eng.pdf.

35.

National Statistical Committee of Kyrgyz Republic (NSCKR) (2016) Employment and Unemployment, Bishkek, Kyrgyzstan: National Statistical Committee of Kyrgyz Republic.

36.

Petreski

Mojsoska-Blazevski

(2015) Youth self-employment in households receiving remittances in Macedonia. Czech Journal of Economics and Finance 65(6): 499–523.

37.

Piracha

Vadean

(2010) Return migration and occupational choice. World Development 38(8): 1141–1155.

38.

Piracha M and Randazzo T and Vadean F (2013) Remittances and Occupational Outcomes of the Household Members Left-Behind. IZA Discussion Paper 7582, August. Bonn: IZA. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2322087.

39.

Posso

(2012) Remittances and aggregate labor supply: Evidence from sixty-six developing nations. The Developing Economies 50(1): 25–39.

40.

Puhani

(2000) The Heckman correction for sample selection and its critique. Journal of Economic Surveys 14(1): 53–68.

41.

Ratha

Plaza

et al. (2016) Migration and Remittances – Recent Developments and Outlook. Migration and Development Brief 26, Washington, DC: World Bank.Available at: http://pubdocs.worldbank.org/en/661301460400427908/MigrationandDevelopmentBrief26.pdf.

42.

Rodriguez

Tiongson

(2001) Temporary migration overseas and household labor supply: Evidence from urban Philippines. International Migration Review 35(3): 709–725.

43.

Roodman

(2011) Fitting fully observed recursive mixed-process models with CMP. The Stata Journal 11(2): 159–206.

44.

Schwegler-Rohmeis

Mummert

Jarck

(2013) Labor Market and Employment Policy in the Kyrgyz Republic: Identifying Constraints and Options for Employment Development, Bishkek, Kyrgyzstan: Gesellschaft für Internationale Zusammenarbeit (GIZ)Available at: http://zentralasien.ahk.de/uploads/media/20130301_KIRG_Study_on_Labour_Market_policy_EN.pdf.

45.

Ukueva N (2010) Migration, remittances, and growth. PhD Thesis, Duke University, Durham, NC. Available at: https://pdfs.semanticscholar.org/96df/3c65cf8cf483e60a8e9e4794544170080aaa.pdf.

46.

Winship

(2003) Logit and probit: Ordered and multinomial models. Journal of the American Statistical Association 98(463): 775–776.

47.

Woodruff

Zenteno

(2007) Migration networks and microenterprises in Mexico. Journal of Development Economics 82(2): 509–528.

48.

Wooldridge

(2010) Econometric Analysis of Cross Section and Panel Data, 2nd edn. Cambridge, MA: MIT Press.

49.

Zhu N and Luo X (2008) The Impact of Remittances on Rural Poverty and Inequality in China. Policy Research Working Paper 4637. Washington, DC: World Bank. Available at: https://openknowledge.worldbank.org/bitstream/handle/10986/6597/wps4637.pdf?sequence=1&isAllowed=y.