The impact on remittances on labor supply decisions in Nepal

(1)

The Impact of Remittances

on

Labor Supply Decisions in Nepal

Master Thesis Economics

July, 2017

Author: Dikshya Parajuli 11375736 Supervisor: 

(2)

Statement of Originality

This document is written by student Dikshya Parajuli who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Table of Contents Abstract ………3 1. Introduction ……….4 2. Literature Review ………7 3. Data ………..8 3.1 Description of Data ……….….8 3.2 Descriptive Statistics ……….….10 4. Methodology ………..….13 5. Findings ……….…….……16

5.1.Adult Labor Supply ……….……16

5.2. Child Labor ……….…….18

5.3. Elderly Labor Supply ……….…………. 21

6. Discussions ……….…..23

7. Conclusion ………24

References ….………26

(4)

Abstract

Due to increasing rates of emigration, international remittance flows to Nepal have been rising steadily in recent years. This paper examines the impact of receiving remittances on labor supply decisions of working age adults (16- 60) and elderly population (61-75) as well as its effects on child labor in Nepal using cross sectional data from household survey in 2011. Instrumental variable approach is used to address concerns of endogeneity in the model. The results of this analysis find no significant effects of remittances on adult and elderly labor supply. However, remittances are associated with significant reduction in instances of child labor but not the intensity of child labor, as measured by hours worked by children.

(5)

1. Introduction

Remittances are transfer of funds sent by a foreign worker back to his or her home country. With the rapid increase in migration from poor countries, remittance flows have become an important source of income for the sending economy. It is estimated that migrants sent home a total of 441 billion USD to developing countries and this is larger than official development assistance and more stable than private capital flows (World Bank 2016). Besides the financial flows, social remittances acquired in the form of new skills, knowledge and ideas can promote development when integrated in the countries of origin.

In most of the developing countries, remittance income accounts for a significant portion of the GDP. In 2015, remittance accounted for 31.8 percent of the total GDP of Nepal, which was the highest remittance to GDP ratio in the world (World Bank 2016). Nepal is one of the poorest countries in South Asia with a per capita GDP of 743 USD (World Bank 2015). Lack of employment opportunities, absence of quality higher education and persistent political instability in the country has led to an accelerated rate of external migration for high skill as well as low skilled labor. Between 1995 and 2004, the total remittances increased from 1.2 to 11.3 percent of the overall GDP (World Bank 2016). It has been estimated that one-fifth of the poverty reduction in Nepal that occurred in this period can be attributed to increased levels of work-related migration and remittances sent home (Loshkin et al 2010). Remittance has played a key role in improving the standard of living of many rural and urban Nepalese population. Acharya and Gonalez (2013) find that consumption is higher in remittance receiving households, and Basnak and Chezum (2009) conclude that remittance receipt is associated with higher likelihood of children attending schools in Nepal. The additional income from remittance also makes healthcare more affordable. A recent world bank study shows that the poorest migrant households prioritize health expenditures after consumption, education and loan payments (Ferrari et al 2007).

The impact of remittance also varies across different sectors of the economy. Tuladhar et al (2014) find that agricultural output is negatively affected in remit receiving households in Nepal.

(6)

Shortage of agricultural labor induced by rising migration rates can have an adverse effect on the agricultural sector which comprises of approximately 33% of the total GDP of the country (World Bank 2015). To investigate the economy wide macro effects of remittances, Brussolo and Medvedev (2008) create a computable general equilibrium model encompassing 22 different sectors of Jamaican economy. Their findings suggest that a 10 percent remittance shock raises household income by 2 percent which leads to total decline in labor supply by 8000 workers. Although, this number is less than one percent of the total employment level, if concentrated within a single sector it can eliminate a small industry entirely. From this perspective, there is a huge economic impact of remittances. While remittance and diaspora investments contribute to the economic development, it can also create dependency in recipient countries.

Although there hasn’t been a study that directly examines the effect of remittances on labor force participation in Nepal, Glinskaya and Lokshin (2009) examine the effects of male migration on female employment patterns using NLSS survey from 2003. The results show that female labor market participation is negatively affected by male migration in the migrant sending household. They cite that increase in household income from remittances to be one of the causes of this reduction. Drawing upon these findings, I examine the link between labor supply decisions of not only females, but the entire working age population, elderly population and also child labor. To the extent of my knowledge, past research on migration and remittances in Nepal uses data from 2003. Between 2003 to 2011, remittance transfers to Nepal has increased significantly from 822 million to 4.22 billion US dollars (World Bank 2016) Given this increment, studying the uses and effects of remittances using the latest figures from 2011 can enhance our understanding on how remittance flows affect the Nepalese economy. In this paper, I use Nepal Living Standard Survey data from 2011 to examine how remittances affect labor supply decisions of the individuals that stay back home.

In my analysis, I divide the sample into three groups and study the effect of remittance receipt on labor participation within these three subgroups separately. The first is the main sample which comprises of individuals of working age population (16 - 60 years). Remittance income is likely to influence the recipients decision to allocate their labor. Typically, earnings abroad are higher

(7)

than earnings at home. If remittance is considered as a substitute for labor income, then recipient families can be anticipated to increase their consumption of leisure. As families desire to consume more leisure and the working age group withdraw themselves from the labor market, then it is likely that they will continue to depend on these transfers to fulfill their needs. Remittance transfers also impose a moral hazard problem as the remitter is not able to monitor effectively the household decisions of recipients(Adolfo et al 2009). Despite the intended use of remittance transfers, recipients can be inclined to increase consumption of leisure and reduce effort into labor market activities. Understanding household responses under these circumstances can be crucial in the policy making space as generating productive use of remittance is key to economic development.

The second sample consists of only individuals that are between 61 and 75 years of age. With mostly youth migrating out and elder members of the household staying back, remittances can act as a old age insurance or pension income for the older group. Lower labor force participation amongst the individuals in households receiving remittance compared to non-recipients can signal for use of remittance transfers as retirement income. The third sample is restricted to only children under 15 years of age in order to investigate the impact of remittances on child labor. Remittance income can help ease financial constraints on households, which allow them to pull children from the workforce, thus reducing child labor supply.

Using past district level remittance rates as an instrument for current remittance receipts, I use IV estimation to study effects of remittances on labor supply. The main findings of this analysis present no effect of remittances on labor force participation of the working population of 16-60 years and elderly population of 61-75 years. The child labor results show that the remittances reduce instances of child labor while do not affect the total intensity of the children’s labor supply. The rest of this paper is organized as follows. Section 2 is an overview of the existing literature on the effects of remittances on labor supply. Section 3 provides the description of the data used. Section 4 explains the methodology used in this research paper. Section 5 presents the results and findings. Finally, section 6 includes discussions and section 7 concludes the paper.

(8)

2. Literature Review

The literature on remittances and labor market participation has found a lot of mixed evidences. Using cross-sectional data from household survey in El Salvador, Acosta(2006) finds that remittance decreases instances of child labor and female labor supply, but has no effect on adult male labor supply. In order to correct for the selection bias in the sample, he employs a propensity score matching to measure outcomes for recipients and non-recipients after matching them based on the probability of receiving remittances. In addition, remittances are instrumented with village migrant networks and the number of international migrants who returned two or more years ago. Similar results have been established for the Phillipines (Rodriguez and Erwin 2001). In a study based in Mexico, Amuedo-Dorantes and Pozo (2006) also address the issue of endogeneity by employing instrumental variable technique to study how employment status and the hours worked by men and women vary with remittance income. They instrument remittance with total number of western union offices in the state during the previous year. Their conclusion suggest that a 100 peso increase in remittance income induces 15 percent reduction in formal sector work by men in both urban and rural areas. However, they find that the same amount of remittance increases working hours by 14 percent and 13 percent in the urban and rural areas respectively.

The decision of a household member to emigrate and remit are not randomly assigned, and is likely to be an outcome of a joint household decision based on a number of unobservables. Therefore, it is likely that confounding factors lead to bias when measuring effect of remittances on outcomes. To reduce selection bias and confounding, Cox et al (2009) restrict their sample to include only households that receive remittances. Using similar matching technique as Acosta (2009), their methodology identifies individuals in households that receive persistent remittances and individuals that do not receive persistent remittances, but are equally likely to have received it. Comparing the labor force participation of these groups, they find that remittances increases labor force participation among urban women only, with no significant effect on urban men or rural men and women. Using panel data, Funkhouser(2006) measures the difference in household labor market outcomes over period between 1998 and 2001 for households that have emigrants

(9)

to the United States. He establishes a negative relationship between teenager labor supply and remittances, while finding no effect on the labor force participation of adults in Nicaragua.

The effect of remittances on labor force participation on individuals is also likely to be depend on the type of work they are involved in. Due to imperfect credit markets in developing countries, households might have to rely on remittance to gain access to credit and ease their liquidity constraints. In such instances, remittance flow may be correlated with increased self employment and increased labor supply for those engaged in such activities.. There is a lot of literature that aims to disentangle this link. Funkhouser (1992) finds that while remittance reduces labor force participation, it increases participation in self-employment activity. Woodruff and Zenteno (2001) survey micro-enterprises in Urban Mexico to examine their sources of investment and find that remittances are responsible for 20% of the total capital. However, an alternative view suggests that remittances are not used for capital accumulation but to increase consumption. Amuedo-Dorantes and Pozo(2006) evaluate the link between remittance receipt and business ownership in Dominican Republic. Acknowledging that remittance receipt can influence start of a business but at the same time having a household business may induce emigrants to send money back home, they construct a simultaneous probit model that examine the likelihood of both outcomes; business ownership and remittance receipts. Their results show that remittances actually reduce the likelihood of business ownership suggesting that by increasing reservation wages, remittances are likely to induce individuals to consume more leisure and work less.

3. Data

3.1. Description of Data

The data for this paper comes from the third Nepal Living Standard Survey (NLSS) carried out by the Central Bureau of Statistics of Nepal in 2011. NLSS is a nation wide household survey carried out in three different periods; 1995, 2003 and 2011. The NLSS practices the Living Standards Measurement Survey (LSMS) methodology developed by the World Bank. The survey provides information on key measure of living standard such as consumption, income,

(10)

expenditures, education, housing, health, access to facilities, employment, remittances, migration, credit and anthropometry.

This paper uses the data from the third round of NLSS in 2011. The survey uses two stage stratified sampling procedure. First, the country is divided into 14 strata, comprising of all 75 districts. The strata formed were as follows: mountains, urban areas of the Kathmandu valley, other urban areas in the hills, rural eastern hills, rural central hills, rural western hills, rural mid-western hills, rural far-mid-western hills, urban Terai, rural eastern Terai, rural central Terai, rural western Terai, rural mid-western Terai, and rural far-western Terai. Each strata then breaks down into Primary Sampling Units(PSU). PSUs are the smallest administrative unit called wards. A number of households are then randomly selected in each PSU. NLSS III consists of two independent samples: the first is a cross sectional sample and the second is a panel. The panel sample consisted of PSUs and households previously enumerated in one or both of the past two rounds of the survey. For the purpose of this paper, the two samples are pooled together to form a single cross section sample of 2011. NLSS III consists of total 6923 households.

It should be reiterated that households that send migrants abroad are not a random sample. For instance, higher migration costs can discourage poor population from emigrating or migrants can be restrained by their educational background or skills. Potential sample selection bias arises when comparing migrant households to non migrant households, as the two are explicitly different. Assuming that the selection mechanism occurs as a result of the decision to migrate, I drop all households which do not have a migrant from the sample. Restricting my sample size to only households with migrants, I consider households that receive remittance as the treatment group and the households that do not receive any remittance as the control group. Therefore, I am interested in assessing the impact of the treatment on household labor supply decisions. The outcome of interest is the labor supply indicator, which is calculated by taking daily average of the total hours worked by the person in the past year.

(11)

3.2. Descriptive Statistics:

Out of the total households surveyed in NLSS III, 3499 households have a migrant abroad. After dropping 3424 households that have no migrant, the remaining sample consists of 3499 households. For the remainder of the paper, I refer to the households with migrants as the total sample. Table 1 presents descriptive statistics for the main variables of interest in this analysis. Panel A and B display means at individual level, while panel C is calculated at household level.

Note: Total hours worked by an individual in a day is measured as labor supply in panel A and B. In panel C, child labor supply is measured at HH level by taking average of total hours worked by children aged 5-15 in a day and N is the total number of households.

Table 1: Descriptive Statistics of main variables

Mean Standard Deviation

A. 16-60 years Share of individuals that

receive remittances

0.3339 (0.4716)

Labor Supply 3.292 (2.895)

N= 8680 B. 61-75 years Share of individuals that

receive remittances

0.3485 (0.4767)

Labor Supply 2.172 (2.459)

N= 1147 C. 5-15 years Share of HH that receive

remittances

0.3376 (0.4730)

Share of HH with child labor 0.3817 (0.4859)

Labor Supply 0.2954 (0.6411)

(12)

Table 2 shows the means of variables in the remittance receiving as well as non receiving sample of working age population aged 16-60 years. The results allow to compare recipient and non-recipient individuals across a number of household and individual characteristics. These covariates are likely to influence remittances as well as labor force participation of the individuals. To capture an individual’s level of education, three dummies are created for the highest level of education completed; namely primary, secondary and university education. Higher number of individuals in the remittance receiving group have completed secondary level of education. Secondly, a variable measuring the maximum education in the household is also included in the model. The highest level of education in the household is likely to capture unobservables that influence household decision making. Highest level of household education is 1.5 years more in remittance receiving group. The number of women in the recipient group is also much higher. The share of work force in the household represents total number of individuals aged 16-60 years of age over the total household size. Having a larger work force is likely to reduce the hours worked by an individual member as it allows for sharing responsibility and reducing work load on a single person. The estimates on the difference in means for this variable are however not significant. Nepal is divided into three geographical regions Mountains, Hills and Terai. Besides distinct geography and climate, these regions differ significantly in cultural, ethnic and socio-economic aspects. For the purpose of this paper, I have constructed three regional dummies that capture these differences. Table 2 shows that more individuals in the Terai region receive remittance, whereas share of individuals not receiving remittance is higher in the Hills. Expenditure and consumption is likely to be highly correlated with the remittances received by households in the past year, which makes this proxy of household income endogenous. In order to capture household’s wealth as well as non-labor income sources, variable capturing land ownership is measured. This is measured in terms of Ropani, which is the most widely used Nepalese unit of land measurement. 8 Ropani is equivalent to the acre. Higher landownership signals for higher household wealth.

(13)

Note : * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level. Land ownership measured in terms of Ropani, 8 Ropani = 1 Acre.

Table 2: Comparison of means across the remittance recipients and non-recipients (16-60 years)

No Remit Remit Difference

(p-value) Primary Education 0.251 0.267 0.5332 Secondary Education 0.126 0.161 0.0247** University Education 0.051 0.067 0.3789 Highest Household Education 8.06 9.54 0.0275** Age 37.06 36.43 0.2430 Male 0.375 0.332 0.0129** Married 0.728 0.722 0.7635 Self-Employed 0.301 0.278 0.5438 Region Hill 0.519 0.314 0.000*** Terai 0.390 0.614 0.000*** Mountain 0.089 0.070 0.4384 Land Ownership No land 0.141 0.162 0.34 < 1 Ropani 0.202 0.225 0.5026 1≤ and<5 Ropani 0.172 0.254 0.0193** 5 ≤ and<10 Ropani 0.135 0.073 0.0013** >10 Ropani 0.348 0.282 0.047** Share of Working Population in HH 0.651 0.639 0.5082 Rural 0.863 0.863 0.9883 N 5931 2733

(14)

4. Methodology:

Although, the sample is restricted to only migrant households, it should be considered that remittances can be correlated with unobserved factors which influence the outcome. While it is possible to control for factors such as person’s education and age, it is not possible to capture a person’s preferences for certain working conditions. If the decision to remit is based on unobservables such as this, then it is likely that remittance receipt is actually influenced by the household’s decision to allocate their time and labor. To account for the endogeneity and reverse causality in remittances, I use instrumental variables in my estimation. Following the works of Hanson and Woodruff (2003) and Pablo A. Acosta et al (2007), I construct instrumental variables that serve as a proxy for the regional migrant network. Using the NLSSII survey, I calculate the past remittance rate in 2003 ( number of households receiving remittances divided by the total number of households in that district). This remittance rate is used as an instrument for current remittance receipts in this analysis. An instrumental variable must be related to the endogenous regressor, which in the context is the receipt of remittances and it should not directly influence the outcome of interest, which is labor supply of individuals.

Past remittance rate is a proxy for availability of services that facilitate remittance transfer in the district, which is correlated with the current remittance rates. Past remittances also signal the extent of migrant network in the region. Higher past remittances mean higher past migration, which in return influences the current outflow of migrants from the district and thus, levels of current remittance. As more information becomes available about opportunities in destination countries through migration from a region, more and more people are likely to migrate from the region. This widely accepted theory of chain migration is called “the cumulative causation of migration” (Massey 1990). The validity of the instrument rests on this idea.

Additionally, past remittance rates in the district is not likely to influence labor supply decisions of individuals at present. One key assumption is that rate of remittances received in the past in a given district, is not correlated with other characteristics of the region which are likely to influence labor force participation of individuals that reside in the region.

(15)

The empirical equation used for the working age population model is as follows:

Laborsupplyijt = α + β1Remitijt + β2Xi + β3HHj + β4 Regioni + ɛijt (Eq 1)

where i = 1…N, j = 1…H such that N is the total number of individuals and H is the total number of households. Remit is a binary variable that takes on value 1 if the individual belongs to a household that receives remittances, 0 otherwise. X is a set of controls at the individual level which includes highest education level completed, age, sex, martial status and a dummy for whether the person is engaged in self-employment activities.

HH is set of control at the household level that includes highest level of education completed in the household, land ownership of the household as a proxy for asset ownership/wealth and the share of working age population in the household. The regional controls include dummies for the three geographical regions of the country and a dummy for rural location.

The first stage and second stage 2SLS equation is below:

Remitijt = α1 + β5RemittanceRatet-1 + β6Xi + β7HHj + β8 Regioni + ɛ1ijt (Eq 2)

where the instrument is the district level remittance rate calculated from the NLSS II survey in 2003. The labor supply of elderly population is also modeled using the equations 1 through 3. The coefficient of interest is β33 which measures how labor supply changes with remittance

receipts.

While the labor supply of adults and elderly is calculated at the individual level, child labor is calculated at the household level. The decision to work in employment is not made by the children themselves but it is determined by the characteristics of the household they belong to. Households where the highest level of schooling is less is likely to have children engaged in labor market activities. It is also likely to capture unobservables such as a family’s preference or

(16)

constraints for attaining higher education. Additionally, family’s share of working population possibly impacts child labor as higher share means that children do not need to sacrifice school for work. It is a well known fact that poverty forces children into work. Besides total household wealth as measured by land ownership, regionals controls and rural area dummy aim to capture these economic differences between households. The below equations are used to model effects of remittances on child labor.

Remitjt = α3 + β12Remitt-1 + β13HHj + β14Regionj + ɛ3jt (Eq 4)

where j=1…H such that H is the number of households in the sample. Past remittance rates in the district is used to instrument current remittance receipts. HH is the set of household characteristics which includes the highest level of education completed in the household, land ownership, share of working age population, share of boys aged 5-15 in the household. The regional controls which include dummies for the three geographical regions and dummy for rural location.

Two different models are developed that measure effect of remittances on the instances as well as intensity of child labor. Average daily hours worked by a child within a household is calculated by summing up the total hours worked by children divided by the number of children in the household. For the first model, a household is reported as practicing child labor if the average daily hours worked by a child in employment is non-zero. Therefore, the outcome is a binary variable that takes the value 0 or 1, where 1 means that child labor exists in the household. For the second model, child labor is measured as average hours worked by a child (5-15) in the household in a day. In most of the households, the total hours worked by children is reported as 0, therefore, this is estimated using a tobit model.

(17)

5. Findings 5.1 Adult labor

Table 3 presents the 2SLS results from Equation 3 above. The coefficient on the remittance receipt measures the effect of remittances on labor supply for adults. Column 1 shows that the total hours worked by an individual in a given day is positively correlated with the status of remittance receipts. Although the coefficient is positive, it is very small and not statistically significant which doesn’t allow to draw any substantial conclusions. The value of F-statistic shows that the instrument is relevant (Appendix A1). Moreover, the coefficient on the remittance rate is positive and statistically significant at the 5% level in the first stage showing that the instrument is highly correlated with the endogenous regressor.

In column 1, the coefficient on sex is highly significant which shows that while holding remittance status constant with other control variables, men work significantly longer than women in the sample. As males and females have different rates of labor force participation, their labor supply responses to receiving remittances is also likely to be different. Therefore, I split this sample of working population into two based on gender and analyze the effects of remittances separately. Understanding the gender breakdown of the effect of remittances can be crucial from a policy perspective in a country like Nepal, where majority of the international migrants are male (Glinskaya and Lokshin 2009) and women are typically discouraged from entering the labor markets, mostly so in rural areas.

For the females only sample, the instrument remains relevant with a F-statistic higher than 10 (Appendix A1). Although the 2SLS results show that the effect of remittances are not significant on female labor supply decisions, the coefficient on remittance receipt is much higher and positive. Female labor supply increases by an almost an hour as a result of remittances with a 95 percent confidence interval [ -.757648 , 2.587856].

(18)

Note : * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level. Land ownership measured in terms of Ropani, 8 Ropani = 1 Acre.

Table 3: Effects of Receiving Remittances on Labor Supply of Working Age Population (16-60 years)

Hours worked in a day All Adults(IV) Females Only(IV) Males Only(IV)

Coefficient Standard

Errors Coefficient Standard Errors Coefficient Standard Errors

Remittance Receipt 0.546 (1.053) 0.915 (0.851) -0.077 (2.607) Primary Education 0.127 (0.153) 0.0006 (0.191) 0.098 (0.292) Secondary Education 0.138 (0.231) -0.282 (0.289) 0.439 (0.323) University Education 0.664** (0.329) 0.525 (0.483) 0.680 (0.434) Highest Household Education -0.020 (0.024) -0.004 (0.020) -0.033 (0.057) Age 0.032*** (0.005) 0.021** (0.006) 0.026** (0.009) Male 1.723*** (0.147) Married 0.937*** (0.162) 0.394** (0.192) 2.000*** (0.244) Self-Employed 1.189*** (0.174) 1.327*** (0.175) 0.868 ** (0.250)

Region (reference: Hill)

Terai -0.277 (0.277) -0.601** (0.256) 0.223 (0.537)

Mountain 0.671** (0.314) 0.699** (0.274) 0.413 (0.440)

Land Ownership (reference:landless HH)

< 1 Ropani -0.185 (0.254) 0.306 (0.274) -0.966** (0.407) 1≤ and < 5 Ropani -0.460 (0.281) 0.262 (0.352) -1.774*** (0.3800) 5 ≤ and <10 Ropani -0.383 (0.288) 0.291 (0.328) -1.619** (0.511) >10 Ropani -0.386 (0.264) 0.254 (0.317) -1.411** (0.427) Share of Working Population in HH -0.213 (0.322) -0.222 (0.303) -0.162 (0.518) Rural 0.301 (0.199) 0.270 (0.223) 0.328 (0.356) N 8664 5472 3192

(19)

On the contrary, many previous research have suggested that remittances actually have negative effect on female labor force participation (Acosta(2006), Glinskaya and Loshkin (2009). However, it is necessary to consider that Glinsakaya and Loshkin isolate the effects of male migration in their research in Nepal. It should be noted that in this analysis, there is no information available about attributes of the remitter, which is likely to have resulted in inconclusive findings if remittances affect household labor supply allocation through channels such as the remitter’s gender. In the next column (3) , the results are presented on the males only sample. The 2SLS estimates in Table 3 show that coefficient on remittance receipts is negative suggesting that remittance receipt is negatively related to labor force participation for men. However, the estimates are very small and not significant concluding that remittances do not affect adult male labor supply. This conclusion is in line with past findings from Acosta (2006), (Rodriguez and Erwin 2001) and Cox et al (2009). Additionally the first stage results for this sample shows that the instrument is weak with a F-statistic much lower than 10 (Appendix A1) The coefficient on the remittance rates are positive and statistically significant at only 10% level. Past remittance rate is a poor predictor of current remittance receipts for the sample of males, therefore it is not likely to strongly predict the labor supply outcomes in the second stage regression.

5.2 Child Labor

Nepal has the highest rate of child labor among South Asian countries. It is estimated that 34 percent of children aged between 5 - 15 years work in employment. (ILO 2010). A vast majority of this takes place in rural areas. General consensus on child labor is that it declines as families pull themselves out of poverty and improve their standard of living. As discussed earlier, remittance has played a key role in improving the welfare of Nepalese households. As remittances have shown to increase educational enrollment of children (Basnak and Chezum 2009), it can be inferred that there should be reduction in child labor as there is a trade-off between the two. Using the Equation 4 and 5, I examine how is child labor affected by remittances.

(20)

The findings are reported in Table 4. Column 1 presents the 2SLS findings on impact of remittances on the prevalence of child labor in the receiving household. The outcome variable is 1 if the household has a child engaged in child labor. The coefficient on remittance receipt is negative and highly significant at 5% , indicating that remittances reduce the prevalence of child labor in households by 76.68%. These findings are also corroborated by results from Acosta(2006) and Funkhouser(2006). The significance of other control variables in the table allow for a better understanding of where child labor occurs more. Child labor is reportedly higher in rural areas and lower amongst households that have a higher share of working adults. Households where the highest level of education completed is only primary and secondary schooling also have higher instances of child labor.

Past remittance rates are a valid instrument with the F-statistic higher than 10 (Appendix A2). Current remittance receipts are highly correlated to the past rates as the coefficient on the instrument from the first stage is positive and highly significant at 5% level. To examine closely how much does remittance reduce the intensity of child working hours, the outcome variable measured is the average hours worked by a child aged 5- 15 in a household in a given day. In contrast to column 1, the IV tobit estimates do not support that remittances significantly reduce the intensity of child labor. The estimates are negative, but not highly significant. Receiving remittances reduces child working hours by 1.54 hours per day with a 95% confidence interval of [-3.42789, 7.9764913].

(21)

Note: Robust standard errors in parentheses.* Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level. Land ownership measured in terms of Ropani, 8 Ropani = 1 Acre.

Table 4: Effects of Remittances on Child Labor Supply

Child Labor Supply Instances of Child labor (IV) Intensity of Child labor (ivtobit)

Coefficient Standard errors Coefficient Standard errors

Remittance Receipt -0.766** (0.383) -1.54 (0.961)

Highest Household Education

Primary Schooling 0.266*** (0.068) 0.540** (0.222)

Secondary Schooling 0.202** (0.075) 0.366 (0.243)

University Schooling -0.042 (0.141) -0.429 (0.464

Land Ownership (reference: landless HH)

< 1 Ropani 0.227*** (0.056) 0.578** (0.200)

1≤ and < 5 Ropani 0.134** (0.063) 0.408 * (0.227)

5 ≤ and <10 Ropani 0.179** (0.083) 0.565* (0.290)

> 10 Ropani 0.269*** (0.075) 0.785** (0.250)

Rural 0.159** (0.064) 0.547** (0.214)

Region (reference: Hill)

Terai 0.046 (0.093) 0.182 (0.254)

Mountains 0.012 (0.108) 0.154 (0.247)

Share of Working Population -0.366 ** (0.158) -0.628 (0.435)

Share of boys 0.070 (0.053) -0.011 (0.140)

(22)

5.3 Elderly Labor Supply

Table 5 presents the findings from the elderly labor supply model which includes individuals from 61-75 years of age. The outcome measured is the total hours worked by an individual in a day. The coefficient on remittance receipt is negative but not statistically significant. Therefore, it should be inferred with caution that remittances reduce the labor force participation of the elderly population. People who belong in households that receive remittances in this sample typically work 1.9 hours less per day. The chosen instrument remains valid with F-stat higher than 10 (Appendix A3). The coefficient on remittance rates are highly significant at 1% level indicating that the probability of an individual receiving remittances is correlated with the past remittance rates which proxy for the migrant network and remittance transfer facilities in the district.

Workers in the informal sectors of employment are not entitled to services such as employe based retirement pension or public pension income and are more likely to continue working beyond the age 60, which puts a physical strain on them. Column 1 shows that the labor force participation is positively affected if the person has only completed primary level of education. Lower levels of education are also associated with informal sectors of work which can help explain this correlation. It is important to note that there is no information available on the physical and health conditions of people in this age group. Due to old age, this population group is vulnerable to chronic health concerns and the status of their health is correlated with how much they are able to work. Moreover, remittances are associated with better health care (Ferrari et al 2007), thus enabling them to be more physically fit for work. In such a case, the effects of remittances on the labor force participation of old are likely to be underestimated. Remittances are likely to ease financial constraints on households but the results do not suggest that they significantly lessen the burden on the older family members to participate in work related activities.

(23)

Table 5: Effects of Remittances on Labor Supply of Old Age Population (61-75 years)

Hours worked in a day 61-75 Years (IV)

Coefficient Standard Errors

Remittance Receipt -1.951 (1.345)

Primary Education 1.364 ** (0.546)

Secondary Education 0.706 (0.715)

University Education -0.787 (0.858)

Highest Household Education -0.043 (0.028)

Age -0.135*** (0.028)

Male 1.062*** (0.241)

Married 0.538** (0.249)

Self-Employed 0.841** (0.325)

Region ( reference: Hill)

Terai -0.059 (0.307)

Mountain 0.374 (0.497)

< 1 Ropani 1.279** (0.442)

1≤ and < 5 Ropani 0.800* (0.468)

5 ≤ and <10 Ropani 0.377 (0.642)

> 10 Ropani 0.571 (0.364)

Share of Working Population in HH -0.367 (0.536)

Rural 0.308 (0.420)

(24)

6. Discussions:

The results show that while remittances have no effect on adult labor supply or elderly labor supply, they are associated with reduction in prevalence of child labor. This indicates that remittance income ease financial burden on households but do not induce adults to consume more leisure time by increasing reservation wage. Children are often forced into work due to stark poverty. This paper finds that remittance receipts reduce the instances of child labor while it has no impact on the intensity of the hours worked by children in employment. The contrast in the two findings suggest that there needs to be further research into this subject and challenges the generalizability of these results. If the findings of this paper are corroborated by additional research, a key take away is for child labor reduction efforts to also focus on families that receive remittances and aim to improve their access to mediums that facilitate these remittance transfers.

Just as poverty induces child labor, the lack of savings and asset-accumulation over their lifetime forces older people in developing countries to join the labor force. Improving the living standards of the elderly requires that they are allowed income transfers to ease the financial burden on themselves and their families. In absence of a solid public pension system or provision of old age benefits and insurance, elderly people are likely to suffer. This analysis did not find any significance of remittance in reducing labor force participation amongst the elderly group. However, the findings are limited to only 1147 individuals. As discussed earlier, the model also doesn't account for the health of these individuals which is a strong determinant of labor supply. This analysis points in the direction of need for further research to disentangle this link. If indeed remittances lead to reduction in work by the old age population, increasing migration and remittance flows can particularly raise the living standards of the elderly.

In order to fully discern these results, it is also necessary to understand the motivation for an individual to migrate and send transfers back home, which are not captured in this model. While summing up the discussion on the findings, it should be noted that there is a limitation in this method of analysis due to absence of data on the characteristics of the migrant who sends remittances. It cannot be ruled out that it can influence how the remittance impacts households. A

(25)

highly educated and skilled migrant will have higher income abroad, send large sums of money back home and have a higher bargaining power in the household to direct how remittance income should be used. As most households in Nepal have extended families living in the same house, the relationship of the remitter to the individual is also likely to influence decisions of time and labor allocation. In order to obtain reliable estimates and clear results, it is necessary to account for this omitted variable in similar models of labor supply.

7. Conclusion

The objective of this paper was to analyze how remittances affect labor force participation of households in the context of Nepal. Using cross sectional data from Nepal Living Standard Survey in 2011, three models of labor supply were constructed for the individuals belonging to age group of 5-15 years, 16-60 years and 61-75 years. In order to account for possible endogeneity and reverse causality in the estimations, an instrumental variable approach was applied. Past remittance rates in the district, which serve as a regional migrant network and proxy for availability of channels that make remittance transfers possible, is used to instrument current remittance receipts in the district. It was hypothesized that remittance eases financial constraints on households which result in reduction in child labor practices and minimizes the labor force participation of the elderly population. On the other hand, remittances were postulated to negatively affect labor supply of the working age population by increasing their reservation wage and enabling them to afford leisure.

The findings for the working age adults show that remittance do not have an impact on their labor supply. Examining these effects by gender show that remittances negatively affect male labor force participation, while they positively affect labor force participation of females. However, these results are not significant so they should be interpreted with caution. With regards to the model for children, it is observed that remittance reduces prevalence of child labor in households and these estimates are significant at 5% level.These results can also be explained using past findings. Prior research has shown that remittances are related to poverty reduction and better living standards. It is known that incidence of child labor is higher in poverty stricken

(26)

areas. As families life themselves out of poverty, the need for these children to engage in employment is also lowered. Interestingly, while measuring the total hours worked by children in a household, remittances appear to reduce the hours but the estimates are not significant to support the hypothesis of child labor and remittances. Lastly, it is found that remittances do not have any effect on labor supply of the old age population.

Migration not only increases the inflow of remittance which plays a significant role in the Nepalese economy but it also reduces the potential workforce within the country. While this paper did not find any significant impact of remittances on adult and elderly labor supply, future research is required to investigate this relationship and possibly explore further labor market implications including labor force and unemployment.

(27)

References:

1. Acharya, Chakra P., and Roberto Leon-Gonzalez. "The impact of remittance on poverty and inequality: A micro-simulation study for Nepal." Asian Journal of Empirical Research 3.9 (2013): 1061-1080

2. Acosta, Pablo. “Labor supply, school attendance, and remittances from international migration: the case of El Salvador”. Vol. 3903. World Bank, Development Research Group, Trade Team, 2006.

3. Amuedo-Dorantes, Catalina, and Susan Pozo. "Remittance receipt and business ownership in the Dominican Republic." The World Economy 29.7 (2006): 939-956.

4. Amuedo-Dorantes, Catalina, and Susan Pozo. "Migration, remittances, and male and female employment patterns." The American Economic Review 96.2 (2006): 222-226.

5. Bansak, Cynthia, and Brian Chezum. "How do remittances affect human capital formation of school-age boys and girls?" The American Economic Review 99.2 (2009): 145-148

6. Barajas, Adolfo, et al. "Do workers' remittances promote economic growth?” Working Paper Series (2009).

7. Bussolo, Maurizio, and Denis Medvedev. "Do remittances have a flip side? A general equilibrium analysis of remittances, labor supply responses and policy options for Jamaica." Journal of Economic Integration (2008): 734-764

8. Central Bureau of Statistics (2011): Nepal Living Standards Survey 2010/2011: Statistical Report Vol. 1 and 2. Kathmandu: Government of Nepal.

9. Cox-Edwards, Alejandra, and Eduardo Rodríguez-Oreggia. "Remittances and labor force participation in Mexico: an analysis using propensity score matching." World Development 37.5 (2009): 1004-1014

10. Ferrari, Aurora, Guillemette Jaffrin, and Sabin Shrestha. Access to financial services in Nepal. Washington, DC: World Bank, 2007.

11. Funkhouser, Edward. "The effect of emigration on the labor market outcomes of the sender household: a longitudinal approach using data from Nicaragua." Well-Being and Social Policy 2.2 (2006): 5-25

12. Funkhouser, Edward. "Migration from Nicaragua: some recent evidence." World development 20.8 (1992): 1209-1218.

13. International Labour Organization “Eliminating Child Labour in Nepal, Facts, figures, commitments and action”. 2010

(28)

14. Lokshin, Michael, and Elena Glinskaya. "The effect of male migration on employment patterns of women in Nepal." The World Bank Economic Review (2009):

15. Lokshin, Michael, Mikhail Bontch-Osmolovski, and Elena Glinskaya. "Work-Related migration and poverty reduction in Nepal." Review of Development Economics 14.2 (2010): 323-332.

16. Massey, Douglas S. "Social structure, household strategies, and the cumulative causation of migration." Population index (1990): 3-26.

17. Rodriguez, Edgard R., and Erwin R. Tiongson. "Temporary migration overseas and household labor supply: evidence from urban Philippines." International Migration Review 35.3 (2001): 709-725

18. Woodruff, Christopher M., and Rene Zenteno. "Remittances and microenterprises in Mexico." (2001)

19. World Bank Group. 2016. Migration and Remittances Factbook 2016, Third Edition. Washington, DC: World Bank

20. World Bank 2015. World Development Indicators Retrieved from http://data.worldbank.org/ indicator/NV.AGR.TOTL.ZS?locations=NP

21. Tuladhar, Raju, Chandan Sapkota, and Naveen Adhikari. "Effects of migration and remittance income on Nepal’s agriculture yield." ADB South Asia Working Paper Series 2014

(29)

Appendix:

Appendix A1: First Stage Results for Working Age Population (16-60 years)

Remittance Receipt All adults sample Males only sample Females only sample

Coefficient Standard

Errors Coefficient Standard Errors Coefficient Standard Errors Past Remittance Rate

(2003) 0.0066** (0.0019) 0.0047* (0.0024) 0.0077*** (0.0018) Primary Education 0.0434 (0.0303) -0.026 (0.035) 0.0835** (0.041) Secondary Education 0.0812** (0.036) 0.0200 (0.0528) 0.1023** (0.0425) University Education 0.0805 (0.0694) -0.0015 (0.0931) 0.1150 (0.0724) Highest Household Education 0.0085** (0.0040) 0.0165*** (0.0049) 0.0056 (0.0041) Age 0.0014* (0.0007) -0.0019 (0.0012) 0.0031** (0.001) Male -0.069*** (0.0194) Married -0.0211 (0.0253) 0.0417 (0.0389) -0.0261 (0.0295) Self-Employed -0.0243 (0.0369) -0.0363 (0.0445) -0.0167 (0.0343)

Region (reference Hill)

Terai 0.2081*** (0.0433) 0.1677*** (0.0515) 0.2322*** (0.0434) Mountain 0.0716 (0.0654) -0.0338 (0.0519) 0.1163 (0.0725) Land Ownership (reference landless HH) < 1 Ropani -0.0741 (0.0590) -0.0386 (0.0604) -0.0876 (0.0603) 1≤ and < 5 Ropani 0.0428 (0.0565) 0.0773 (0.0634) 0.0276 (0.0571) 5 ≤ and <10 Ropani -0.0695 (0.0619) -0.0434 (0.0632) -0.0777 (0.0654) > 10 Ropani -0.0609 (0.0541) 0.0010 (0.0547) -0.0922 (0.0580) Share of Working Population in HH -0.0309 (0.0573) -0.0110 (0.0901) -0.0445 (0.05781) Rural 0.0283 (0.0504) 0.0061 (0.0580) 0.0467 (0.0495) F- statistic 11.52 3.83 17.71 N 8664 3192 5472

(30)

Appendix A2: First Stage Results for Children’s Labor Supply Model

First Stage Results Children (5-15 years)

Coefficient Standard error

Past Remittance Rate (2003) 0.0064** (0.0020)

Highest Household Education

Primary Schooling 0.1156** (0.0395)

Secondary Schooling 0.1285** (0.0470)

University Schooling 0.0008 (0.1236)

< 1 Ropani 0.0436 (0.0581)

1≤ and < 5 Ropani 0.0603 (0.0583)

5 ≤ and <10 Ropani -0.0008 (0.0603)

>10 Ropani 0.02315 (0.0567)

Rural 0.03236 (0.0507)

Terai 0.1912*** (0.0410)

Mountains 0.04150 (0.0538)

Share of Working Population -0.0651 (0.1006)

Share of boys 0.0363 (0.0390)

F-statistic 10.23

(31)

Appendix A3 : First Stage Results for old age population (61-75 years)

Remittance Receipt 61-75 Years

Coefficient Standard errors

Past Remittance Rate

(2003) 0.0106*** (0.0029)

Primary Education 0.191** (0.076)

Secondary Education -0.0052 (0.1504)

University Education -0.0115 (0.0912)

Highest Household Education 0.0018 (0.0063)

Age 0.0041 (0.0065)

Male -0.0607 (0.0475)

Married 0.0457 (0.0476)

Self-Employed -0.0702 (0.0681)

Terai -0.0012 (0.0680)

Mountain 0.08963 (0.1142)

< 1 Ropani 0.0740 (0.1059)

1≤ and < 5 Ropani -0.0017 (0.1065)

5 ≤ and <10 Ropani -0.2265** (0.1085)

>10 Ropani -0.0769 (0.0981)

Share of Working Population in HH -0.0702 (0.0681)

Rural 0.0915 (0.0661)

F-statistic 12.72