15
REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS Alberto POSSO School of Economics, Finance and Marketing, RMIT University, Melbourne, Australia First version received November 2010; final version accepted September 2011 Do remittances reduce labor supply in recipient economies? This paper addresses this question with aggregate level data for a panel of sixty-six developing countries from the Middle East and Africa, Asia and the Pacific, and Latin America and the Caribbean over the period 1985 to 2005. The results exhibit a positive and significant relationship between remittances and aggregate labor supply. The effect is clearly driven by men in each of the three regions. Three potential explanations are put forward to explain these empirical findings: (1) non-migrating household members are likely to increase their labor supply in order to defray migration-related expenses; (2) neighboring households increase their labor supply to help family members migrate after they become more aware of the benefits of remittances; and (3) remittances overcome credit constraints, thus generating employment. Keywords: Remittances; Labor supply; Panel data; Developing countries JEL classification: J21, F24, O1 I. INTRODUCTION A s remittances have grown to become the main source of external financing for a large number of developing countries, they have attracted substantial attention from economists (Acosta et al. 2006). Authors have focused on a number of key issues, such as remittances and health, education, economic growth, inequality, and poverty to name a few. 1 This paper addresses a question that has received considerably less attention, “Do remittances lower labor supply in recipient economies?” A number of authors The author would like to thank Simon Feeny, Andrew Leigh, Ashton Da Silva, Tim Fry, Lisa Farrell, George Tawadros, Julia Farrell, and participants in the RMIT Economics seminar series for their valuable comments on earlier drafts. This paper also benefited from the useful comments of I-Ling Shen, Peter Petri, Peter Drysdale, Hugh Patrick, Nobuaki Yamashita, Shiro Armstrong, and other participants at the PAFTAD 33 Conference, Taipei, Taiwan October 5–8, 2009. The usual caveat applies. 1 World Bank (2006), Giuliano and Ruiz-Arranz (2009), and Acosta et al. (2007). The Developing Economies 50, no. 1 (March 2012): 25–39 © 2012 The Author The Developing Economies © 2012 Institute of Developing Economies doi: 10.1111/j.1746-1049.2011.00153.x

REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

Embed Size (px)

Citation preview

Page 1: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

REMITTANCES AND AGGREGATE LABOR SUPPLY:EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

Alberto POSSOSchool of Economics, Finance and Marketing, RMIT University, Melbourne, Australia

First version received November 2010; final version accepted September 2011

Do remittances reduce labor supply in recipient economies? This paper addresses thisquestion with aggregate level data for a panel of sixty-six developing countries from theMiddle East and Africa, Asia and the Pacific, and Latin America and the Caribbean overthe period 1985 to 2005. The results exhibit a positive and significant relationshipbetween remittances and aggregate labor supply. The effect is clearly driven by men ineach of the three regions. Three potential explanations are put forward to explain theseempirical findings: (1) non-migrating household members are likely to increase theirlabor supply in order to defray migration-related expenses; (2) neighboring householdsincrease their labor supply to help family members migrate after they become moreaware of the benefits of remittances; and (3) remittances overcome credit constraints,thus generating employment.

Keywords: Remittances; Labor supply; Panel data; Developing countriesJEL classification: J21, F24, O1

I. INTRODUCTION

A s remittances have grown to become the main source of external financingfor a large number of developing countries, they have attracted substantialattention from economists (Acosta et al. 2006). Authors have focused on

a number of key issues, such as remittances and health, education, economicgrowth, inequality, and poverty to name a few.1

This paper addresses a question that has received considerably less attention,“Do remittances lower labor supply in recipient economies?” A number of authors

The author would like to thank Simon Feeny, Andrew Leigh, Ashton Da Silva, Tim Fry, Lisa Farrell,George Tawadros, Julia Farrell, and participants in the RMIT Economics seminar series for theirvaluable comments on earlier drafts. This paper also benefited from the useful comments of I-LingShen, Peter Petri, Peter Drysdale, Hugh Patrick, Nobuaki Yamashita, Shiro Armstrong, and otherparticipants at the PAFTAD 33 Conference, Taipei, Taiwan October 5–8, 2009. The usual caveatapplies.1 World Bank (2006), Giuliano and Ruiz-Arranz (2009), and Acosta et al. (2007).

The Developing Economies 50, no. 1 (March 2012): 25–39

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

doi: 10.1111/j.1746-1049.2011.00153.x

Page 2: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

have addressed this issue by looking at household level data in countries in Asiaand Latin America. Rather than focusing on household level data, the present studyuses aggregate level data in order to uncover movements in aggregate labor supply.Household level data is better suited to test whether remittances, by increasing agiven household’s income, can raise budget constraints and increase reservationwages, thus reducing both the likelihood of employment and the hours worked byremittance-receiving individuals alone. However, changes in household behaviordo not necessarily translate to significant changes in aggregate labor supply. It ispossible that an increase in remittances reduces the labor supply of the recipienthousehold, while simultaneously increasing the labor supply of other householdsin the same vicinity (which become aware of the benefits of remittances, such asbuying a house or starting a business) that are seeking to help a family membermigrate. And, it is aggregate labor supply that is more pertinent than changes inrecipient-household behavior when thinking about more general macroeconomicoutcomes that have significant effects on entire nations. It is changes in aggregatelabor supply that affect unemployment, GDP, and real wages in an economy.

Therefore, the evidence from previous case studies cannot be readily generalizedfor entire nations. Previous studies have found that remittances decrease femalelabor supply, with mixed evidence for males. Using household survey data from ElSalvador, Acosta (2006) identified that remittances are negatively related to childlabor and adult female labor supply, while on average adult male labor forceparticipation is unaffected. A similar survey conducted in Nicaragua found thatremittances increase male self-employment, while reducing female labor supply(Funkhouser 1992). Additionally, in a similar study on the Philippines, Rodriguezand Tiongson (2001) discovered that men are more likely than women to reducetheir labor supply choice after receiving remittances, although the effect is nega-tively significant across both genders. Similarly, focusing on household level datafrom Haiti, Jadotte (2009) found that remittances are associated with a (statisticallyinsignificant) decline of labor supply. In a similar study, Kim (2007) showed thatJamaican households that receive remittances significantly reduce their labor forceparticipation.

Mexico has attracted disproportionate attention in this literature for two reasons.First, there is ample accessible Mexican household survey data for the study of theremittance-labor market nexus. Second, about 2.5 million Mexicans migrated tothe United States during 1997–2002, with 1.6 million of them sending remittancesto their families. Cox-Edwards and Rodríguez-Oreggia (2009), for instance, findlittle evidence of labor force effects of remittances in Mexico. Similarly, Amuedo-Dorantes and Pozo (2006) found no evidence of reduced labor effort resulting fromgreater remittance incomes among Mexican men. Instead, they note that remit-tances only seem to alter the allocation of male labor supply across various typesof employment. In contrast, women appear to work less in response to greater

26 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 3: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

remittance receipts although this was found exclusively in the informal sector andnon-paid work in rural areas. Similarly, Airola (2008) found that remittancereceipts are associated with fewer work hours for households in general. Finally,Hanson (2007) found that over the 1990s women, but not men, in high-migrationMexican states, become less likely to work outside the home than women inlow-migration states after an inflow of remittances.

Using aggregate level panel data from sixty-six developing nations over theperiod 1985 to 2005, the present study revisits this issue. This paper finds a positiveand significant relationship between remittances and aggregate labor supply. Sup-porting the evidence in Funkhouser (1992), it is found that this effect is driven bymen in each of the three regions. The positive relationship is found to be robust tovarious specifications.

Two potential explanations or hypotheses have been provided by previousstudies to shed light on these empirical results. First, Amuedo-Dorantes and Pozo(2006) argue that non-migrating household members are likely to increase theirlabor supply in order to defray the migration-related expenses of other familymembers. Second, Giuliano and Ruiz-Arranz (2009) suggest that remittances havebeen shown to provide an additional source of credit, which in turn generatesemployment. Additionally, this paper argues that this result may also stem from theuse of aggregated data. These data can capture the effect of remittances on bothrecipient and non-recipient households. Essentially, it is possible that an increasein remittances for recipient households raises the labor supply of non-recipienthouseholds that wish to help a family member migrate once they realize thefinancial benefits of remittances. The fact that these data can perhaps capture thispossible cross-household effect gives additional impetus to the use of aggregatedata.

II. DATA

This paper uses international data sources in order to study the effect of remittanceson labor supply choice among a sample of sixty-six developing nations. Thesample is limited by data availability; however, most major developing economiesin the Middle East and Africa (Africa), Asia and the Pacific (Asia), and LatinAmerica and the Caribbean (LAC) are represented. The countries studied arepresented in Table 1. The quality of these data poses an additional limitation. It iswell understood that remittance flows through informal channels are believed to bevery large although the rise of credit unions and other financial institutions hasreduced this to an extent. Additionally, many countries report remittance inflows as“other transfers,” which makes it difficult to disentangle remittances from inflowssuch as tourism receipts and nonresident deposits. Moreover, the collection ofremittance data during the 1980s was very poor, but has increased in quality

remittances and aggregate labor supply 27

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 4: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

through time. Therefore, the sample uses a series that is potentially measured witherrors and that has become more accurate through time (Maimbo and Ratha 2005).In fact, the large increment in remittances discussed in the literature and in thispaper may be a reflection of better measurement on the part of receiving govern-ments and a larger portion of transfers going through the formal system (seeFigure 1). As discussed below, year fixed effects are included to try to account forsome of these data problems.

Data on remittances come from the World Bank’s World Development Indica-tors (WDI) database. The WDI presents these data as workers’ remittances andcompensation received by employees, measured in current US dollars. Thismeasure is converted to real 2008 US dollars using consumer price index data fromthe US Department of Labor, Bureau of Labor Statistics. The data are then log-transformed in order to down-weight outliers.

Following the previous literature on this subject, this paper analyzes the rela-tionship between remittances and both male and female labor force participationrates. The analysis is separated according to gender, but also includes a combinedsample. These measures are readily available from the United Nations StatisticsDivision’s Gender Info 2007 database. Labor force participation is defined as thenumber of individuals of working age (15–64 years) who participate in the laborforce either as workers or actively looking for work, as a share of their cohort.

Even though remittance and labor force data are generally available from the1970s, coverage only becomes adequate for econometric analyses from the 1980s.As a result, this study focuses on the twenty years leading up to 2005. It is believedthat this choice of time frame presents sufficient variation in the data for anadequate econometric analysis. Figure 1 shows time trends for the averages oftotal, female, and male participation rates as well as remittances for sixty-six

TABLE 1

List of Sixty-Six Countries Assessed in Study, by Region

Asia and the Pacific (Asia)Fiji, India, Indonesia, Republic of Korea, Kyrgyzstan, Malaysia, Maldives, Pakistan, Papua New

Guinea, Philippines, Samoa, Sri Lanka, Thailand, Tonga, Turkey, and Vanuatu.

Latin America and the Caribbean (LAC)Argentina, Barbados, Belize, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic,

Ecuador, El Salvador, Guatemala, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Paraguay,Peru, Suriname, Trinidad and Tobago, and Venezuela.

Middle East and Africa (Africa)Algeria, Benin, Botswana, Burkina Faso, Cameroon, Cape Verde, Comoros, Côte d’Ivoire, Egypt,

Ghana, Guinea, Jordan, Kenya, Lebanon, Lesotho, Madagascar, Mali, Morocco, Mozambique,Niger, Nigeria, Oman, Senegal, South Africa, Sudan, Swaziland, Syria, Togo, and Tunisia.

28 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 5: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

developing countries. Participation rates are presented in percentages andremittances are measured in 2008 US dollars. Total and female participation ratesas well as remittances exhibit an upward trend. Total and female labor forceparticipation increase from 64.2 to 65% and 45.5 to 49.5% respectively during thistime period. However, there is significantly more variation in the trend for totalparticipation rates. Remittances increase from around 40 billion 2008 US dollars in1985 to just over 140 billion in 2005. The male participation rate, on the otherhand, trends downwards. Male participation rates fell by 2 percentage points, from82.5 to 80.5%. To avoid the estimation of spurious relationships and to reduce theeffect of large annual fluctuations, data are averaged over five years in the econo-metric analysis.2

2 It is also the case that identifying whether a variable has a unit root is difficult with fewobservations.

Fig. 1. Average Remittances and Labor Force Participation Rates in Sixty-Six DevelopingCountries, 1985–2005

64.2

64.4

64.6

64.8

65

(%)

1985 1990 1995 2000 2005

Year

Total Participation

45

46

47

48

49

50

(%)

1985 1990 1995 2000 2005

Year

Female Participation

80.5

81

81.5

82

82.5

(%)

1985 1990 1995 2000 2005Year

Male Participation

0.5

1

1.5

2

2.5 2

008

US

$ (B

illi

on)

1985 1990 1995 2000 2005Year

Remittances

Source: Labor force participation rates come from the United Nations Statistics Division’sGender Info (2007). Remittances data comes from the World Development Indicators, WorldBank; it is converted to real 2008 US dollars using CPI data from the United States Departmentof Labor, Bureau of Labor Statistics.

remittances and aggregate labor supply 29

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 6: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

To give some sense of the patterns prevalent in the data, Figure 2 plots male andfemale labor force participation rates versus the log of remittances. The data areaveraged by country over the period from 1985 to 2005. The figure shows mixedevidence of the connection between remittances and participation rates. Countrieswith higher levels of remittances seem to have lower female participation rates,while there is no obvious relationship between these variables for males. A simplecorrelation exercise confirms these relationships—the correlation coefficientbetween remittances and female and male participation rates is -0.36 and -0.007,respectively.

In the econometric analysis, this paper also includes a number of control vari-ables. It is well understood in microeconomic theory that an individual has areservation wage, as such it is expected that an increase in real wages will result inan increase in the supply of labor. Due to data availability, manufacturing wage datais employed as a proxy for the general wage level in each country. Wage data comes

Fig. 2. Male and Female Participation Rates and Log Remittances in a Selection of DevelopingCountries, Averaged by Country over 1985–2005

6070

8090

100

14 16 18 20 22 24

Male

2040

6080

100

14 16 18 20 22 24

Female

Par

tici

pati

on r

ate

(%),

ave

rage

d by

cou

ntry

Log remittances, averaged by country

Source: Remittances data comes from the World Development Indicators, World Bank; it isconverted to 2008 US dollars using CPI data from the United States Department of Labor,Bureau of Labor Statistics. Labor force participation data comes from Gender Info (2007),United Nations Statistics Division.

30 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 7: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

from the United Nations Industrial Development Organization (UNIDO); it isconverted to 2008 US dollars using the same methodology applied toremittances.

A country’s level of economic development can also determine whether a persondecides to participate in the labor market. Real wages and GDP per capita may behighly correlated, in order to avoid multicollinearity, the model employs GDP percapita relative to the United States (Y-YUS) as an explanatory variable. Similarly,the consumption share in GDP per capita can be positively related to a nation’saggregate labor supply. Societies with higher consumption rates are expected toneed to supply higher quantities of labor in the market. The investment share inGDP per capita is used as a demand side factor. Investment can be an importantemployment generator, and ceteris paribus, higher demand for workers could leadto an increase in supply. These data come from Heston, Summers, and Aten (2011)and are calculated using purchasing power parity GDP measures.

Finally, this paper also controls for net migration, which can have conflictingeffects on labor force participation rates. On the one hand, it can immediately causea contraction in labor supply. On the other, as individuals migrate a labor vacuumcan potentially remain in the household, thus having a positive effect on partici-pation rates. Migration was initially considered as an instrument for remittances;however, the aforementioned effects on labor force participation render this aninadequate instrumental variable.3 Net migration is obtained from the WDI and ismeasured in thousands of people. Table 2 presents the summary statistics of thedata described above.

3 This was also confirmed by the use of a C-test.

TABLE 2

Summary Statistics

Variable Obs. Mean Std. Dev. Min. Max.

Labor force participation (%) 1,386 64.64 9.76 44.09 89.22Female participation (%) 1,386 47.49 16.77 14.40 87.80Male participation (%) 1,386 81.63 5.55 62.70 95.30Remittances (2008 bn US$) 1,356 1.10 2.46 0.001 25.47Real manufacturing wages per worker

(2008 bn US$)858 6,240.31 4,776.48 121.77 35,712.73

PPP converted GDP per capita Relativeto the United States (Y-YUS)

1,386 13.15 11.45 1.03 72.24

Net migration (1000s) 1,386 -0.12 0.39 -3.98 2.20Consumption/GDP per capita 1,386 73.23 18.40 14.80 189.48Investment/GDP per capita 1,386 22.55 10.47 0.51 86.34

remittances and aggregate labor supply 31

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 8: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

III. EMPIRICAL SPECIFICATION

The empirical model used in this study involves estimating an equation thatprovides for capturing the impact of remittances on the labor force participationrates of males, females, and the population as a whole using five-year averageddata. This can be analyzed using the following equation:

LF R Xi t i t i t i t i t, , , , ,= + + + +β β α λ ε1 2 (1)

where LFi,t represents the dependent variables, which refers to the labor forceparticipation of the total population, as well as for males and females, for countryi at time t. Ri,t is the log of remittances, measured in 2008 billion US dollars, Xi,t isa vector of time-varying country characteristics, ai is a country fixed effect, lt is ayear fixed effect, and ei,t is a normally distributed mean-zero error term.

The model is estimated using lagged differenced dependent variable as instru-ments to control for an autocorrelation problem of order 1 and 2.4 A Baltagi and Wu(1999) locally best invariant (LBI) test statistic confirmed the presence of positiveserial correlation.5 Aside from controlling for an AR(1) and AR(2) process, usinga lagged dependent variable as an instrument can also account for the path-dependent nature of labor force participation rates, which are dependent onstructural and institutional factors that are very slow to change. Moreover, atwo-step system equation is specified, where the standard covariance matrix isrobust to panel-specific autocorrelation and heteroscedasticity. The tables belowshow Arellano–Bond tests for autocorrelation, AR(1) and AR(2). Using a two-stepsystems estimation technique suggests that the Hansen J statistic must be employedto diagnose goodness of fit. Note also that following a Durbin-Wu-Hausman testfor endogeneity, all the independent variables are treated as exogenous.6 Countryfixed effects are used because it is believed that unobserved country characteristicsthat are constant through time, such as regional and institutional factors, are likelyto be correlated with the explanatory variables. A Hausman specification test wasundertaken to corroborate this intuition. Similarly, year fixed effects are used tocontrol for unobserved effects that vary over time. This is of particular importancein light of coverage in the media suggesting that remittances are susceptible toeconomic downturns. Moreover, year fixed effects take a step toward correcting forthe aforementioned data problems, such as improved measurement of remittanceinflows through time. An F-test for the use of time fixed effects found that these arejointly significant at the 1% level.

4 The rhos for male, female, and total participation rates are 0.911, 0.948, and 0.939, respectively.5 This test accounts for unbalanced panels with unequally spaced data. The LBI test statistic for male,

female, and total participation rates are 0.42, 0.37, and 0.38, respectively.6 This issue is revisited in Section IV. C.

32 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 9: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

IV. EMPIRICAL RESULTS

A. General Results

Table 3 presents the results of equation (1). The columns differ according todependent variable included and by whether Y-YUS is included. This variable issometimes excluded as it is potentially jointly determined with formal labor marketparticipation. The dependent variables are total (columns 1 and 2), female(columns 3 and 4) and male (columns 5 and 6) labor force participation rates. Theresults suggest a positive relationship between remittances and the total participa-tion rate. Columns (1) and (2) suggest that an increase in remittance inflows by 1%results in a statistically significant increase in the total labor force participation rateby 1 percentage point. To put these results into context, in 2005 the total labor forceparticipation for these developing countries was 65.03%. The results suggest thatan increase in remittance inflows by 1% could lead to an increase in the total laborforce participation rate to 66.03%. This positive relationship is consistent across

TABLE 3

Instrumental Variable Regression (Dependent Variables: Total, Female, and MaleParticipation Rates)

Model:(1) (2) (3) (4) (5) (6)

Total Total Female Female Male Male

Log remittances 1.10** 1.24*** 0.066 0.17 2.10*** 2.26***[2.47] [2.74] [0.093] [0.24] [6.52] [6.89]

Log wage 3.44*** 2.75*** 2.78* 2.24* 4.17*** 3.36***[3.60] [3.46] [1.75] [1.79] [6.35] [5.66]

Y-YUS -0.15 -0.12 -0.18**[-1.10] [-0.49] [-2.17]

Net migration(millions)

2.66** 2.72* 3.35* 3.40* 2.00** 2.07**[2.06] [2.00] [1.90] [1.87] [2.06] [2.06]

Consumption/GDP 0.22** 0.24*** 0.31** 0.33** 0.13** 0.16***[2.45] [2.69] [2.34] [2.51] [2.55] [2.84]

Investment/GDP -0.039 -0.060 -0.013 -0.029 -0.090 -0.12[-0.31] [-0.49] [-0.068] [-0.16] [-1.32] [-1.63]

Year fixed effects Yes Yes Yes Yes Yes YesAR(1) p-value 0.97 0.93 0.41 0.43 0.75 0.77AR(2) p-value 0.11 0.17 0.12 0.12 0.09 0.20Hansen p-value 0.95 0.84 0.81 0.90 0.75 0.62Observations 152 152 152 152 152 152

Note: Robust t-statistics in brackets. Bias correction initialized by Arellano-Bond estimatorusing lagged values of the dependent variable as instruments. *, **, and *** represent statisticalsignificance at the 10%, 5%, and 1% level, respectively.Y-YUS is dropped from the specificationsin columns (2), (4), and (6) due to possible simultaneity.

remittances and aggregate labor supply 33

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 10: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

both the female and male participation rates. However, from the latter, it is onlyfound to be statistically significant for males (columns 5 and 6), which is consistentwith the findings in Funkhouser (1992).

Three potential explanations can be provided to shed light on the positiverelationship between remittances and labor force participation. First, Amuedo-Dorantes and Pozo (2006) argue that the relationship between remittances andlabor supply is theoretically ambiguous. They explain that the receipt of remit-tances is usually preceded by the out-migration of working-aged householdmembers, which may induce changes in the labor supply of non-migrating house-hold members in order to compensate for foregone income or to defray migration-related expenses. This study controls for net migration; however, the effect ofmigration-related expenses could be driving more women to work. The secondexplanation was flagged above as resulting from the use of aggregate-level data,which may capture a possible cross-household effect of remittances (withoutdirectly testing for them). Essentially, when non-migrating households becomeaware of the benefits of remittances, they increase their labor supply in order tohelp a family member migrate. Third, remittances have been shown to overcomecredit constraints in developing countries (Giuliano and Ruiz-Arranz 2009). Inturn, additional credit generates further employment.

The remaining explanatory variables are, for the most part, found to be statis-tically significant across the six different specifications. Not surprisingly, anincrease in real wages by 1% is associated with an increase in the participation rateby 3 percentage points. Similarly, higher rates of consumption are associated withgreater labor supply—an increase in the consumption share in GDP by 1 percent-age point leads to an increase in the total labor force participation rate by 0.2percentage points. Investment is found to be insignificant; however, an F-testrevealed this to be jointly significant with consumption. Finally, net migration isassociated with a positive and significant effect on participation rates.

B. Results: Regional Dummies

It is of particular interest to discuss the results above in terms of regionalcompositional differences for at least three reasons. First, there are major differ-ences in the actual size of remittances inflows going into each region. For example,it is well known that Latin America is the largest recipient of remittances in theworld (Fajnzylber and López 2008). Second, the structure of labor markets isgenerally quite different across these regions, with Asia generally exhibiting agreater degree of flexibility than LAC and African nations (Forteza and Rama2006). Third, there is evidence indicating differences in the skill composition ofmigrants across the three regions, which in turn may affect the way remittances areused by recipients (Docquier, Lohest, and Marfouk 2007; Faini 2007). Table 4follows a similar structure to that of Table 3, with the exception that it includes

34 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 11: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

interactive regional dummies for Africa, Asia, and LAC. The dummies interactwith the log of remittances measured in 2008 US dollars. The table exhibitssurprisingly similar results to those presented above. This suggests that aftercontrolling for unobserved country characteristics in the fixed effects model, it ispossible to highlight a cross-country causal relationship between remittances andlabor force participation.

C. Robustness Checks

This section summarizes a number of robustness tests performed. Table 5summarizes the results. For ease of exposition it only exhibits remittance-related

TABLE 4

Instrumental Variable Regression with Regional Interactive Dummies (Dependent Variables: Total,Female and Male Participation Rates)

Model:(1) (2) (3) (4) (5) (6)

Total Total Female Female Male Male

Log remittances*dummies

*Africa 1.00* 1.19** -0.23 -0.0087 2.18*** 2.37***[1.84] [2.22] [-0.28] [-0.011] [5.67] [6.16]

*Asia 1.09** 1.22*** 0.017 0.18 2.13*** 2.26***[2.39] [2.71] [0.024] [0.25] [6.48] [6.75]

*LAC 1.08** 1.19** 0.035 0.16 2.19*** 2.29***[2.27] [2.44] [0.046] [0.21] [6.40] [6.29]

Log wage 3.53*** 2.81*** 3.00* 2.17 4.06*** 3.36***[3.37] [3.39] [1.72] [1.61] [5.84] [5.32]

Y-YUS -0.18 -0.21 -0.18*[-1.21] [-0.79] [-1.76]

Net migration(millions)

2.86** 2.75* 3.96* 3.84* 1.97* 1.85*[2.07] [1.97] [1.95] [1.87] [1.97] [1.87]

Consumption/GDP 0.22** 0.24*** 0.33** 0.35*** 0.13** 0.15**[2.54] [2.79] [2.52] [2.72] [2.42] [2.67]

Investment/GDP -0.029 -0.062 0.020 -0.018 -0.087 -0.12[-0.23] [-0.50] [0.11] [-0.097] [-1.16] [-1.59]

Year fixed effects Yes Yes Yes Yes Yes YesAR(1) p-value 0.86 0.92 0.61 0.57 0.78 0.87AR(2) p-value 0.11 0.16 0.12 0.12 0.09 0.20Hansen p-value 0.99 0.85 0.72 0.85 0.74 0.57Observations 152 152 152 152 152 152

Note: Robust t-statistics in brackets. Bias correction initialized by Arellano-Bond estimatorusing lagged values of the dependent variable as instruments. *, **, and *** represent statisticalsignificance at the 10%, 5%, and 1% level, respectively. Y–YUS is dropped from thespecifications in columns (2), (4), and (6) due to possible simultaneity.

remittances and aggregate labor supply 35

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 12: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

coefficient estimates, the remaining estimates were, for the most part, found toremain unchanged to these exercises.

The first test accounts for the possibility that remittances have a quadratic impacton labor force participation. Including a squared term renders the remittancecoefficient estimate insignificant for total and female participation rates. However,a significant quadratic relationship is found between remittances and the maleparticipation rate. The coefficient estimates suggest that the positive effect ofremittances on male labor force participation will peak at around 19%. In otherwords, if remittances increase by more than 19%, males in these developingnations are likely to substitute away from labor into leisure. That is, by increasing

TABLE 5

Robustness Tests (Dependent Variables: Total, Female, and Male Participation Rates)

Model:(1) (2) (3) (4) (5) (6)

Total Total Female Female Male Male

QuadraticLog remittances 3.66 2.37 -4.29 -4.12 10.8*** 9.87***

[0.90] [0.46] [-0.43] [-0.33] [5.01] [4.26]Log remittances

squared-0.12 -0.082 0.052 0.052 -0.26*** -0.24***

Observations 152 152 152 152 152 152AR(1) p-value 0.12 0.12 0.34 0.83 0.79 0.96AR(2) p-value 0.15 0.15 0.06 0.68 0.46 0.96Hansen p-value 0.49 0.49 0.34 0.80 0.75 0.73

Non-linearLog remittances 1.12*** 1.12*** 1.77*** 1.77*** 0.50** 0.50**

[4.20] [4.20] [3.96] [3.96] [2.29] [2.29]Observations 152 152 152 152 152 152

Endogeneity:Methodology GMM IV GMM IV GMM IVLog remittances 2.22* 2.95** 0.69 3.70 3.44*** 2.27**

[1.97] [2.13] [0.41] [1.51] [3.65] [2.43]Observations 152 151 152 151 152 151Hansen p-value 0.19 0.33 0.14 0.35 0.18 0.16

LaggedLog remittances 0.074 0.068 0.23 0.19 0.20 0.20

[0.21] [0.20] [0.41] [0.34] [0.79] [0.80]Observations 106 106 106 106 106 106R-squared 0.21 0.21 0.14 0.14 0.19 0.19

Note: Robust t-statistics in brackets. *, **, and *** represent statistical significance at the 10%,5%, and 1% level, respectively.

36 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 13: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

an individual’s income, remittances can potentially also increase the reservationwage, thus decreasing the amount of labor they are willing to sell in the market.

A second robustness test was included to verify that the results hold undernonlinear assumptions. A fixed effects Tobit regression that accounts for truncateddependent variables was estimated. The results remained robust to this exercisewith coefficient estimates changing only marginally.7 Interestingly, a positiveeffect is found for women under this nonlinear specification.

The third robustness exercise revisits the important issue of endogeneity. Eventhough the Durbin-Wu-Hausman test did not find evidence of this problem, itremains the case that this test performs better with larger samples. Moreover, wecan intuitively expect some degree of endogeneity. For instance, a family membermay choose to increase remittances if those who remain in the household areneither working nor looking for work (Naiditch and Vranceanu 2009). The robust-ness test engineered to deal with this issue takes a two-tiered approach. On the onehand, equation (1) is reestimated using two step system GMM instrumentingremittances with their lagged levels (Dollar and Kraay 2003). Simple C andBreusch et al. (1999) tests found lagged remittances to be inadequate instruments.Therefore, a second approach is undertaken where remittances are instrumentedusing three variables: inflation, the US dollar to local currency exchange rate, andthe log of GDP per capita (in PPP). These variables were all found to insignifi-cantly affect labor force participation rates, thus successfully meeting the criteriafor good instruments. The regression results are once more consistent with thosefound in Tables 3 and 4.

The last robustness exercise in Table 5 tests for the effect of past remittances oncurrent labor force participation rates. The aim of this exercise is to test whetherremittances have a long-term effect on labor force participation rates. The lastsection of Table 5 reveals that this is not the case.

V. CONCLUSION

An increase in remittance inflows may generate negative labor supply effects byincreasing the reservation wages of recipients. This relationship has been testedpreviously in a number of case studies focusing on developing countries. Thispaper revisits this question with the use of aggregate level data. Household leveldata does not necessarily capture movements in aggregate labor supply, which ismore relevant for thinking about general macroeconomic outcomes. Therefore, astudy using aggregate level data can provide conclusions that may be better suitedto policymakers. Using a panel of sixty-six developing nations over the period

7 The dependent variables are truncated between 0 and 100. These results are available from theauthor upon request.

remittances and aggregate labor supply 37

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 14: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

1985 to 2005, this paper identifies an overall positive and significant relationshipbetween remittances and labor force participation rates. An inverted-U relationshipis also found for males, suggesting that once remittances pass a certain threshold,they can potentially induce recipients to substitute away from labor and intoleisure.

It is proposed that three alternative hypotheses or explanations can be used toshed light on this positive causal relationship. First, it may result from the fact thatnon-migrating household members are likely to increase their labor supply in orderto bear the expenses of migration-related costs. Second, by overcoming creditconstraints in developing countries, remittances may also be generating additionalemployment opportunities. Third, the use of aggregate data in this paper means thatit is impossible to distinguish between those households that receive remittancesand those that do not. It is likely that an increase in remittances raises the laborsupply of non-recipient households who wish to help a family member migrateonce they realize the benefits of remittances from observing their neighbors. Thefact that these data may capture this possible cross-household effect gives addi-tional impetus to the use of aggregate-level data.

Overall, this paper finds that the results from previous case studies should not begeneralized across developing nations. Instead, it is suggested that different house-holds are likely to respond differently in the aftermath of a surge in remittances.Consequently, further analysis of the effect of remittances on labor force partici-pation rates of specific countries is warranted. In particular, these studies shouldalso aim to uncover whether a cross-household effect of remittances exists. In orderto undertake that exploration, a regional approach would be needed, focusing onthe effect of remittances on labor supply adjusting for the concentration of remit-tances in a certain region or community.

REFERENCES

Acosta, Pablo A. 2006. “Labour Supply, School Attendance, and Remittances from Inter-national Migration: The Case of El Salvador.” World Bank Policy Research WorkingPaper no. 3903. Washington, D.C.: World Bank.

Acosta, Pablo; Cesar Calderón; Pablo Fajnzylber; and Humberto López. 2006. “Remit-tances and Development in Latin America.” World Economy 29, no. 7: 957–87.

———. 2007. “What Is the Impact of International Remittances on Poverty and Inequalityin Latin America?” World Development 36, no. 1: 89–114.

Airola, Jim. 2008. “Labor Supply in Response to Remittance Income: The Case ofMexico.” Journal of Developing Areas 41, no. 2: 69–78.

Amuedo-Dorantes, Catalina, and Susan Pozo. 2006. “Migration, Remittances, and Maleand Female Employment Patterns.” American Economic Review 96, no. 2: 222–26.

Baltagi, Badi H., and Ping X. Wu. 1999. “Unequally Spaced Panel Data Regressions withAR(1) Disturbances.” Econometric Theory 15, no. 6: 814–23.

38 the developing economies

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies

Page 15: REMITTANCES AND AGGREGATE LABOR SUPPLY: EVIDENCE FROM SIXTY-SIX DEVELOPING NATIONS

Breusch, Trevor; Hailong Qian; Peter Schmidt; and Donald Wyhowski. 1999. “Redun-dancy of Moment Conditions.” Journal of Econometrics 91, no. 1: 89–111.

Cox-Edwards, Alejandra, and Eduardo Rodríguez-Oreggia. 2009. “Remittances and LaborForce Participation in Mexico: An Analysis Using Propensity Score Matching.” WorldDevelopment 37, no. 5: 1004–14.

Docquier, Frédéric; Olivier Lohest; and Abdeslam Marfouk. 2007. “Brain Drain in Devel-oping Countries.” World Bank Economic Review 21, no. 2: 193–218.

Dollar, David, and Aart Kraay. 2003. “Institutions, Trade, and Growth.” Journal of Mon-etary Economics 50, no. 1: 133–62.

Faini, Riccardo. 2007. “Remittances and the Brain Drain: Do More Skilled Migrants RemitMore?” World Bank Economic Review 21, no. 2: 177–91.

Fajnzylber, Pablo, and J. Humberto López, eds. 2008. Remittances and Development:Lessons from Latin America. Washington, D.C.: World Bank.

Forteza, Alvaro, and Martín Rama. 2006. “Labor Market “Rigidity” and the Success ofEconomic Reform across More Than 100 Countries.” Journal of Policy Reform 9, no.1: 75–105.

Funkhouser, Edward. 1992. “Migration from Nicaragua: Some Recent Evidence.” WorldDevelopment 20, no. 8: 1209–18.

Giuliano, Paola, and Marta Ruiz-Arranz. 2009. “Remittances, Financial Development, andGrowth.” Journal of Development Economics 90, no. 1: 144–52.

Hanson, Gordon H. 2007. “Emigration, Remittances and Labor Force Participation inMexico.” INTAL-ITD Working Paper 28. Buenos Aires: Institute for the Integration ofLatin America and the Caribbean.

Heston, Alan; Robert Summers; and Bettina Aten. 2011. “Penn World Table Version 7.0.”Center for International Comparisons of Production, Income and Prices at the Univer-sity of Pennsylvania, May.

Jadotte, Evans. 2009. “International Migration, Remittances and Labour Supply: The Caseof the Republic of Haiti.” UNU-WIDER Research Paper no. 2009/28. Helsinki: UNUWorld Institute for Development Economics Research.

Kim, Namsuk. 2007. “The Impact of Remittances on Labor Supply: The Case of Jamaica.”World Bank Policy Research Working Paper no. 4120. Washington, D.C.: World Bank.

Maimbo, Samuel M., and Dilip Ratha. 2005. Remittances: Development Impact and FutureProspects. Washington, D.C.: World Bank.

Naiditch, Claire, and Radu Vranceanu. 2009. “Migrant Wages, Remittances and RecipientLabour Supply in a Moral Hazard Model.” Economic Systems 33, no. 1: 60–82.

Rodriguez, Edgar R., and Erwin R. Tiongson. 2001. “Temporary Migration Overseas andHousehold Labor Supply: Evidence from Urban Philippines.” International MigrationReview 35, no. 3: 709–25.

World Bank. 2006. Economic Implications of Remittances and Migration. Washington,D.C.: World Bank.

remittances and aggregate labor supply 39

© 2012 The AuthorThe Developing Economies © 2012 Institute of Developing Economies