Accounting for Remittance and Migration Effects on Children’s Schooling

Catalina Amuedo-Dorantes

Department of Economics

San Diego State University

5550 Campanile Drive

San Diego, CA 92182

(619) 594-1663

camuedod@mail.sdsu.edu

Susan Pozo

Department of Economics

Western Michigan University

1903 West Michigan Ave.

Kalamazoo, MI 49008

(269) 387-5553

susan.pozo@wmich.edu

May 2010

Abstract: We examine the impact of remittances on children’s school attendance in the

Dominican Republic. To isolate the effect of remittances from the effect of sometimes

concurrent household out-migration, we focus on children in households without members

currently residing in the U.S. While girls’ school attendance rises with the receipt of

remittances, secondary school-age children and younger siblings are the ones who most

decidedly gain from remittances. Additionally, we find that migration negatively impacts the

school attendance of children as it eliminates the positive effect of remittances when we expand

the sample to include children in households with members residing abroad.

Keywords: Latin America, Dominican Republic, school attendance, remittances, migration.

1

1. INTRODUCTION

Because of the development potential of financial inflows, a growing literature has

looked for positive impacts of international remittances on poor economies. Remittances have

been studied to ascertain to what degree these flows serve to raise or even-out consumption

levels, provide for housing, promote business investments, and increase the availability of health

services for individuals with monetary constraints. Yet, some studies have suggested that

remittances primarily raise consumption levels and do not necessarily promote investments in

either physical or human capital.1 We suspect that these competing conclusions about the

benefits of remittances may partially be due to the confounded impacts of remittances and family

migration. While remittances can benefit households by lifting liquidity constraints, migration of

a family member may have a deleterious impact on the household’s well-being. The absence of

a family member may deprive the household of the migrant’s market and non-market production,

possibly making the household worse off. Therefore, it is of interest to ask to what extent the

gains from remittances make up for the losses sustained from family migration.

In this paper, we provide illustrative evidence of the importance of distinguishing

between the presumably negative migration effect and the positive effect of remittances when

examining the impact of remittance inflows on children’s schooling using data from the

Dominican Republic. The Dominican Republic is an interesting case study for various reasons.

First, it is a country that has experienced extensive emigration. According to the World Bank

(2009), twelve percent of the Dominican population has emigrated. Second, remittance receipts

in the Dominican Republic account for about 10 percent of the country’s GDP (World Bank,

2009). Third, children’s school attendance rates are relatively low. The primary school net

enrollment rate was 82 percent in 2007, with secondary enrollment rates falling to 61 percent

2

according to UNESCO (2009), leaving significant room for improvements in educational

attainment. And most importantly, the Dominican Republic is a particularly interesting case

study for the purpose of our analysis due to the substantial variation in emigration and

remittance-receiving patterns across households, which enables us to separate the migration

effect from the remittance effect and to confirm their competing impacts.

We proceed by first distinguishing households with a family member currently in the

U.S. (migrant households) from households without migrants in the U.S. (non-migrant

households).2 Focusing our attention on non-migrant households – to which most of the children

in the sample belong and in which more than half of children in remittance-receiving households

reside, we examine the impact that the receipt of remittance income has on children’s school

attendance. To focus on non-migrant households allows us to isolate the impact of the receipt of

remittances from the effect of family migration on the schooling of children. In order to

empirically sign the effect of family migration on the schooling of children and to gauge whether

the presumably negative migration effect offsets the positive impact of remittance receipt, we

repeat the analysis expanding our sample to include children residing in migrant households. We

then compare our estimates of the impact of remittance receipt in the two instances, i.e. when we

exclude and when we include children residing in migrant households, to gain a better

understanding of how family migration and remittance-receipt affect children’s schooling. To

conclude, we explore variations in the impact of remittance receipt on children’s school

attendance depending on their age, gender and order of birth –characteristics suggested as being

crucial in previous studies of human capital investments on children (e.g. Calero et al., 2009;

Emerson and Portela Souza, 2008).

3

We find that remittances promote children’s school attendance in the Dominican

Republic, particularly among secondary school-age children and higher order of birth siblings.

Additionally, we are able to conclude that migration has a negative impact on the school

attendance of children since the positive effect of remittance receipt effectively disappears when

we expand our sample to include children in migrant households. Overall, the analysis provides

further evidence of the positive impact that remittances can have on remittance-receiving

countries, while also noting that migration of family members may temper and even cancel-out

the positive remittance effect. As is the case with the Dominican Republic, the findings should

prove of interest to developing countries with extensive emigration, striving to promote

educational attainment.

2. BACKGROUND ON MIGRATION, REMITTANCES AND SCHOOLING IN THE

DOMINICAN REPUBLIC

During the reign of Rafael Trujillo (1930-1961), foreign travel of Dominicans was

strictly controlled in order to repress the formation and growth of political opposition (Derby,

2009). Emigration to the U.S. from the Dominican Republic involved only 17,000 Dominican

over the dictatorship years (Sagás and Molina, 2004). The assassination of Trujillo in 1961

however, released social and political tensions and contributed toward demand for emigrating.

The subsequent lifting of controls prompted a large exodus of the population – a trend that

persists to this date. By the conclusion of the 20th

century, it was commonly quoted that over

one-tenth of the Dominican population had emigrated, and according to the OECD (2010), 88

percent were residing in the U.S. The next most popular destination, Spain, accounted for only 5

percent of Dominican emigrants.

4

What do we know about the character of that mass emigration? While it is difficult to

obtain direct evidence on the skill level of Dominican emigrants, we do have some indirect

information from the 2002 Dominican census, which asked household whether any of their

members were currently living abroad. In Figure 1, we tabulate this information by indicating

the percentage of households who claim to have a family member abroad according to the

educational attainment of the household head. About nine percent of households with heads

claiming no education report having at least one family member abroad. The comparable

statistic for heads claiming 24 years of education is 12 percent. Hence, there may be a very

slight bias toward more skilled emigration. Overall, however, emigration seems to be fairly

broadly represented across skill levels in the Dominican Republic – a finding that is not

surprising given the mass emigration that took place from the 1960s and onward.3

FIGURE 1 ABOUT HERE

The widespread growth of migration in the later part of the past century eventually

yielded a dramatic increase in remittances to this island economy. Remittances currently

represent between 8 to 10 percent of Dominican GDP over the past decade (OECD, 2010).

According to a door-to-door survey of 3,000 households in the Dominican Republic in 2004,

recipients claim to use 60 percent of those receipts to pay for daily expenses, 17 percent on

education, 5 percent on business investments, 5 percent on saving, 4 percent to buy property and

6 percent to buy luxury products (IBD, 2004).

Reports of the educational attainment of Dominicans in the Dominican Republic are

somewhat inconsistent. While Barro and Lee (2000) claim that, on average, Dominican adults

(aged 25 and over) had 5.17 years of education in 2000, the Dominican government statistics

point towards a more educated population. Average educational attainment for adults aged 25 to

5

54 is in the order of 7 years, whereas for adults aged 55 and older it is about 7.5 years (Secretaría

de Estado de Educación, 2008). The educational attainment of women exceeds that of men (8.3

years versus 7.9 years in 2007), with rising gender disparities (in favor of women) at higher

grade levels. About three quarters of Dominican students attend public schools and the other

quarter attend private schools (Secretaría de Estado de Educación, 2008). Grade repetition in the

Dominican Republic is common. In primary school, the repetition rate is about 8 percent, while

in secondary school it is about 4 percent (UNESCO, 2009). While education attainment is not

out of line with that of many Latin American nations, improvements need to be made to promote

the economic diversification that the Dominican government has pursued in the past half century.

In terms of assessing the likelihood of raising educational levels, we are interested in the

growing number of studies that have linked migration and education by estimating the impact of

remittances on children’s schooling. Focusing on Latin America, the hypothesis that remittances

raise educational attainment or investments in schooling has received support in a growing

number of studies, including those by Ilahi (2001) for Peru; Cox Edwards and Ureta (2003) for

El Salvador; Borraz (2005) for Mexico; Acosta et al. (2007) for a sampling of Latin American

countries; and Calero et al. (2009) for Ecuador. Standard economic theory suggests that, by

lifting liquidity constraints, remittances should raise investments in human capital of household

members, particularly in developing countries. While schools may often be state-supported,

students in developing nations are often expected to pay for their books, uniforms, supplies and

sometimes even teacher salaries. Furthermore, attending school imposes additional costs on the

family through accompanying reductions in monetary income or household production by the

attendee. Given the competition between school and work, remittances, often amounting to a

sizable fraction of the household budget, have the potential to loosen household liquidity

6

constraints and increase investments in human capital. Yet, perhaps contrary to what one would

expect from the income effect due to increases in non-labor income, Acosta et al. (2007)

conclude that remittances do not raise educational attainment in the Dominican Republic4.

In this paper, we revisit the role of remittances on school attendance in the Dominican

Republic. In addition to carefully addressing the endogeneity of remittance receipt, we take into

account the fact that migration of a family member often accompanies the receipt of remittances

by the household. This migration, in and of itself, may induce changes in the schooling of the

remaining children, independently of remittances, by several avenues. First, children may engage

in market activities to replace the household’s lost income, leaving less time to devote to their

studies. Second, children may skip or quit school in order to attend to necessary household

chores or to care for younger children or elderly family members –activities perhaps formerly

undertaken by the now absent household member. Third, children may devote less time to

studies if they are expecting to migrate too. Consequently, in our view, if we do not adequately

control for family emigration, we may not uncover the true impact of remittances on children’s

schooling. A recent study by Bansak and Chezum (2009) concerning the educational attainment

of Nepalese children, acknowledges the two competing impacts of remittances and family

emigration, which they describe as net remittance and absenteeism effects. However, some

doubts remain as to the ability of the strategy employed in their analysis to separate and properly

estimate the remittance and migration effects.5

3. DATA AND DESCRIPTIVE STATISTICS

To ascertain the impact of remittances on children’s schooling, we use Dominican data

from the Latin American Migration Project survey (known as the LAMP-DR7).6 The LAMP-

DR7 consists of 907 households with 1123 children between the ages of 7 and 18 interviewed in

7

seven Dominican communities in 1999 and 2000. Two of the communities are located within

the capital city, Santo Domingo, while the remaining communities are located in a diversity of

areas including farming communities, impoverished rural towns, and middle-sized cities.

Table 1 displays some descriptive statistics for all children, as well as for children

residing in what we refer to as migrant and non-migrant households. Out of the 1123 children,

140 live in migrant households –households that claim having some members currently living in

the U.S.7 The other 983 children live in non-migrant households –households that do not have

any members currently living in the U.S. A first look at the statistics for all children (first two

columns of Table 1) reveals that the average school attendance rate is 76 percent. Nineteen

percent of all children in the seven communities being surveyed by the LAMP reside in

remittance-receiving households.8 Eighty percent of the children residing with household heads

who are employed and approximately 40 percent of household members are children between 7

and 18 years of age. Twenty-two percent of the children reside in households where the female

spouse claims to have 6 or fewer years of education. The remaining 78 percent have more than 6

years of education. In the aggregate, men are less educated than women, with only 70 percent of

them claiming more than 6 years of schooling.

TABLE 1 ABOUT HERE

When we distinguish according to whether the children reside in a non-migrant as

opposed to a migrant household (consult the middle versus the two far right columns in Table 1),

we find some statistically significant differences in the variables describing the two sub-samples.

For instance, the probability of living in a remittance-receiving household is greater for children

8

in migrant households relative to children in non-migrant households. Sixty-six percent of the

children residing in migrant households receive remittances as opposed to 13 percent in non-

migrant households.9 Migrant households also appear to have fewer preschoolers, more adults,

more assets and less educated female spouses relative to non-migrant households.10

The purpose of our analysis is to learn about the impact that the receipt of remittances by

the household may have on children’s schooling. Since we do not have information on the

household’s history of remittance receipts, we focus on the impact of the current receipt of

remittances by the household on the contemporaneous school attendance of children. One might

argue that it would be preferable to measure educational attainment (perhaps the number of years

of schooling completed by the child in relation to his or her age) and analyze the impact that

remittance receipt has on years of schooling completed. However, educational attainment is

likely dependent on the household’s history of remittance receipts, which is not reported in the

survey. Therefore, we focus on school attendance.

TABLE 2 ABOUT HERE

Simple descriptive statistics for school attendance by basic child characteristics are

displayed in Table 2. School attendance rates are significantly higher for secondary school-age

children (79 percent attend school) than for primary school-age children (71 percent attend

school). Yet, the 2002 Dominican census reports attendance rates for primary school-age

children as being slightly higher than for children in secondary school. The difference between

the LAMP survey and the Dominican census is probably explained by the fact that the LAMP

focuses primarily, on migrant communities. Since the schooling rates of primary school-age

children appear particularly lower in migrant households (see Table 1), it is possible that the

difference being observed between this survey and the national census is due to the population

9

sample being surveyed by the LAMP. However, because migrant communities are the most

impacted by remittance inflows, our sample is still of interest for the study at hand. The figures

in Table 2 also show that girls have a slight edge over boys in school attendance, as is

customarily found in the Dominican Republic. However, the gender difference is not

statistically significant. With respect to birth order, there is some evidence that firstborn children

are provided with more opportunities to attend school. While 78 percent of firstborn children

attend school, only 74 percent of younger siblings are in school. This difference is statistically

significant at the 10 percent level.

What information do we have regarding remittances? We know whether the household

currently receives money from people in the U.S.11

Unfortunately, the survey does not contain

information on the dollar amount received by families, nor on the frequency with which they

receive such transfers nor on how they use the money received. However, if households report

receiving money from people in the U.S., they are asked about the relative size of the remittance

inflow with respect to household income.12

Table 3 tabulates this qualitative information. About

half of remittance-receivers indicate receiving a modest amount when compared to household

income. The other half of recipients is about equally split into those who indicate remittances to

be a medium or a large inflow relative to their household incomes. When we distinguish

households by migrant and non-migrant status, we observe some further differences. In one third

of migrant households, remittances represent a large share of household income. In contrast,

only a fifth of remittance-receiving non-migrant households fall in that category. Yet, as we

shall discuss in what follows, the incidence of remittance receipt among non-migrant households

is still large, suggesting that quite a few Dominican households receive remittances from

10

individuals other than their own household members, such as distant family members and

friends.

TABLE 3 ABOUT HERE

4. EMPIRICAL METHODOLOGY

To analyze the effect of remittance receipt on children’s school attendance, we could

estimate the following benchmark model:

(1) )0(, ***

ififififfifif yyXRyAttendanceSchool

where *ify is the unobserved or latent likelihood of attending school by child i in family f. The

function is the indicator function taking on the value one if *ify > 0 and zero

otherwise. fR is a dummy variable denoting whether the household currently receives

international remittances. The vector ifX includes information on a variety of covariates thought

to be important determinants of children’s schooling according to earlier studies, such as those

by Cox Edwards and Ureta (2003) and Hanson and Woodruff (2003). Some of these

determinants include information on the children’s gender and order of birth to allow for

differential returns to educational investment for boys and girls, as well as for firstborns and

higher order of birth siblings. We account for additional child descriptors potentially affecting

children’s educational attainment, such as age. We also include information on the employment

of the household head, household assets, and household composition (as captured by the number

of adults, number of school-age children and number of preschool-age children) as factors

possibly influencing the household’s financial ability to send children to school.13

Additionally,

we account for the educational attainment of the male as well as the female spouse, factors found

to be correlated with children’s educational outcomes in other studies (Haveman and Wolfe,

11

1995; Schultz, 2002; Breierova and Duflo, 2004).14

Finally, we include a set of community

dummy variables to take into account regional differences across the various Dominican

Republic communities in which the children reside. Communities may differ in per capita

income levels or school infrastructure possibly impacting school attendance rates.

(a) Distinguishing Remittance from Household Migration Effects

There are several econometric problems that could arise in the estimation of equation (1).

Perhaps, the most pressing problem is the fact that household migration has taken place for 48

percent of the children residing in remittance receiving households. Remittances, a source of

non-labor income, may lift budget constraints and, through an income effect, improve the

likelihood that children in remittance-receiving households go to school. However, the presence

of family members abroad may induce changes in school attendance of children in non-migrant

households for a variety of reasons. Children may have less time to devote to schooling because

they engage in market activities to earn income to defray migration-related expenses of

household members or to replace the migrant’s former contributions to the household’s income.

Alternatively, children may leave school to attend to necessary household chores that the absent

migrant no longer attends to. Finally, if children expect to follow their family members and

migrate in the future, they may drop out of school if Dominican education is not generally

rewarded in the destination.15

Therefore, attributing choices in schooling to remittances alone

may not be appropriate if there is a concurrent migration effect.

Distinguishing the disruptive effect of household emigration from the income effect of

remittance inflows on children’s schooling is problematic as it requires the identification of two

separate events that are often driven by similar factors. In the Dominican Republic, however,

there is significant diversity in household emigration and remittance-receipt patterns that we can

12

take advantage of to distinguish the two impacts. In effect, we separate the migration from the

remittance effect by focusing our attention on children residing in non-migrant households.

While the existence of a close family member abroad significantly raises the odds of remittance

receipt, more than half of remittance-receiving households report receiving remittance flows

from individuals other than household members. These may be distant relatives or perhaps

friends.16

Among children, more than half of those residing in remittance-receiving households

(124 out of 217 in the LAMP-DR7) reside in non-migrant households (see Table 4). Therefore,

by examining how these children’s schooling responds to the receipt of remittances by the

household, we can ascertain the impact of remittance receipt independently of household

emigration.17

TABLE 4 ABOUT HERE

(b) Endogeneity of Remittance Receipt

The estimation of equation (1) presents one additional challenge. Specifically, the receipt

of remittances and the error term may be correlated, in which case the coefficient estimate for

remittance receipt is biased. There are two potential sources for this noted correlation. The first

source originates in the presence of unobserved heterogeneity and omitted variable bias. The

receipt of remittances may be inversely related, for example, to household income which, in turn,

may be positively correlated to school attendance.18

In that regard, our estimate of the impact of

the receipt of remittances is likely to be downward biased. The second source of correlation

between the receipt of remittances and the error term in equation (1) results from the potential

joint determination of remittance transfers and children’s schooling. In particular, while it seems

reasonable to expect that remittance receipts facilitate investments in schooling, it may also be

the case that children’s schooling induce remittance inflows, e.g. an aunt may be remitting to a

13

favorite nephew to reward him for his school attendance. In that case, the nephew’s schooling is

determining the aunt’s remittances instead of the reverse.

We first test whether remittance receipt is endogenous to the child’s school attendance

using the augmented Durbin-Wu-Hausman (DWH) test for endogeneity suggested by Davidson

and MacKinnon (1993). After concluding that OLS is inconsistent,19

we estimate equation (1) as

a two-stage linear probability model. We instrument remittance receipt with the 1999-2000

unemployment rate and average real earnings in personal care and service occupations in those

U.S. states where households likely developed networks.20

Migration network information is

obtained from the LAMP survey, which inquires about the past and current location of household

and extended family members in the U.S. When there is no history of migration to the U.S. for

neither household nor extended family members, these instrumental variables take on the

unemployment and wage values for Puerto Rico. In those instances, Puerto Rico –accessible by

boat from the Dominican Republic and, as such, a common destinations of Dominicans going to

the U.S.– is more likely to be the origin of the remittance flows. Table 5 reports the most

common destinations for Dominican migrant locating in the U.S according to the survey. Owing

to the importance of networks in facilitating migration, there is a large concentration of

Dominicans in New York. Nevertheless, we also find a significant fraction of Dominicans in

other states, such as New Jersey and Puerto Rico. Other, less frequent destinations, present in

the dataset include Massachusetts, Florida, Rhode Island and Pennsylvania, as well as

Connecticut, California, Illinois, Kansas, North Carolina and South Carolina. These destinations

are in accord with the major destinations of Dominican emigrants reported by the Inter-American

Development Bank (IDB, 2004).

TABLE 5 ABOUT HERE

14

What is the logic behind our choice of instruments? Current labor market conditions in

U.S. destination areas are likely to be correlated with the sending of remittances by relatives and

friends as those conditions will impact the ability and desirability to remit. Our identifying

assumption is that current U.S. labor market conditions do not affect the school attendance of

children in the Dominican Republic other than through remittances. As is often the case with

instruments, ours could be subject to potential shortcomings. For instance, one potential threat is

that the instruments could be related to household characteristics that affect children’s schooling,

such as household income. Higher income households may have historically placed migrants in

economically more attractive states in the U.S. To address this possibility, we control for as

many household characteristics correlated with household income as we possibly can, including

a measure of the educational attainment of the female spouse and male spouse, the employment

of the household head, and household assets.

A second possible threat to the validity of our instruments could come from the fact that

Dominican migrants from different regions may traditionally send migrants to specific U.S.

states. In that case, the instruments could be simply capturing regional differences across

Dominican communities, such as differences in per capita income levels, school infrastructure or

overall economic development. To account for this possibility, the analysis includes community

dummies.

A final threat to the validity of our instruments (although also related to regional

differences and economic development captured by the community dummies) is if migration

networks alter children’s school attendance rates by directly or indirectly impacting household

wealth or by changing the incentives to acquire an education. Introducing community fixed

effects indirectly accounts for differences in migration networks across communities. In any

15

event, as noted by others in this literature (McKenzie and Rapoport, 2006), the effect of a

community network is likely to be second-order (to the effect of other household characteristics)

in the education decision.

In addition to discussing the theoretical basis and overall rationale for our choice of

instruments, we inspect our instrumental variables to ascertain their validity as instruments from

an econometric standpoint. Specifically, we first check their correlation with the receipt of

remittances by the household –the endogenous regressor to be instrumented. The problem of

“weak instruments” arises when either the instruments are weakly correlated with the

endogenous regressor (i.e. remittance receipt), or the number of instruments is too large (Angrist

and Krueger, 2001). Therefore, we check for the strength of our instruments with the F-test at

the bottom of Table 6 and Table 7. In both instances, the tests indicate that our instruments are

strongly correlated to remittance receipt.21

Additionally, the results from the first-stage

estimation in the bottom panels of Table 6 and Table 7 (to be discussed in the next section) are

reasonable and indicate that the instruments help predict household remittance-receipt.

Because remittance income is being instrumented by two variables, we also use over-

identification tests to examine the exogeneity of the instruments. Due to existing concerns

regarding the low power of these tests in case of general misspecifications (e.g. Newey, 1985),

we use Sargan’s (1958) test as well as a recommended variation of the Basmann (1960) test –the

Basmann-LIML form of the test (see Staiger and Stock, 1997). Both tests examine the

exogeneity of each one of our instruments conditional on the other one being valid. That is, in

both tests, the null hypothesis is that the excluded instruments are uncorrelated with the error

term and correctly excluded from the estimated equation. As such, a rejection of the null

hypothesis casts doubt on the validity of the instruments (Baum et al., 2002; Wooldridge, 2002).

16

As shown by both tests at the bottom of Table 6 and Table 7, we are unable to reject the null

hypothesis regardless of the sample of children considered in the analysis. As such, the

instruments appear to be correctly excluded from the main equation modeling children’s school

attendance.

In what follows, we examine the impact of remittance receipt on school attendance by

estimating equation (1) as a two-stage linear probability model. Relative to a probit or logit

model, the linear probability model allows us to handle instrumental variable estimates with

standard two-stage least squares procedures, facilitating the estimation of standard errors and

model convergence when sample sizes are not large. Subsequently, we assess how remittance

receipt may be affecting children’s school attendance by age, gender and order of birth. All the

analyses compute robust standard errors that take into account data clustering at the household

level.

5. RESULTS

The main objective of our analysis is to assess the impact of remittance receipt on

children’s school attendance by separating the income effect of remittance receipt from the

disruptive impact of contemporaneous household emigration. In addition, we also address the

endogeneity of remittance-receipt. In order to purge the coefficient estimate of remittance-

receipt from any potential disruptive impact of contemporaneous family emigration, we first

estimate our model focusing on children residing in non-migrant households. Additionally, we

address the endogeneity of remittance-receipt through the use of an instrumental variable

approach in the estimation of equation (1). The results from such an exercise are displayed in

Table 6.

TABLE 6 ABOUT HERE

17

Before discussing our findings, note that the test results at the bottom of Table 6 confirm

that each instrumental variable (state unemployment rates and real wages in personal care and

service occupations) appears sufficiently correlated to remittance receipt and, conditional on the

other one being valid, uncorrelated to the error term in equation (1). As shown by the regression

output in the bottom panel of Table 6, higher unemployment rates in the destination states of

migrants are positively correlated with a higher likelihood of remittance receipt, which supports

the view that migrants may be remitting money back to their communities for self-insurance

purposes. Immigrants likely bear higher employment risk during times of rising unemployment,

making it prudent for them to insure against these risks by remitting funds back to their

communities. In this way, migrants maintain “good standing” within the community permitting

them to return (with honor) in the event of an unsuccessful migration experience (Lucas and

Stark; 1985; Amuedo-Dorantes and Pozo, 2006). Yet, migrants may have multiple motives

when remitting money home, including altruism (Becker, 1974; Stark, 1991). If migrants also

remit altruistically, remittances should be directly related to their remitting capacity as reflected

by the positive sign on the average real earnings in personal care and service occupations in

those U.S. states where migrants tend to locate. Likewise, if migrants remit money home to

make a specific purchase (e.g., a plot of land or a house) or for investment purposes (e.g., setting

up a small business), remittances should increase with their real earnings as we observe in the

first-stage results.

Finally, it is also interesting to note that the receipt of remittances is positively related to

the number of preschoolers. By contrast, remittances are negatively related to the number of

school-age children in the household as well as to the number of adults. The presence of

preschoolers limits working options for adults, suggesting that remittances in these cases are

18

compensating for market work. On the other hand, the presence of school-age children and of

more adults provides the household with more flexibility with respect to garnering the necessary

resources for the household, making remittances less crucial in these instances. Similarly, as we

would expect, households with employed heads are also less likely to receive remittances.

(a) Remittance Receipt and Children’s School Attendance

Do remittances promote children’s school attendance? According to the figures in the

top panel of Table 6, an increase in the probability of remittance receipt of 10 percentage-points

raises the likelihood of school attendance by almost 3 percentage-points (i.e. 0.1*0.286=0.029)

from an average of 0.76 to approximately 0.79 (see Table 1 for group averages).22

As such,

remittances help close the non-attendance gap. This is the finding we would anticipate given that

remittances in these households have the potential to lessen liquidity constraints.

Other results are the expected ones. For instance, the positive coefficient on the dummy

variable for female secondary education is consistent with the idea that in households in which

the female head (or female spouse) is more highly educated, the educational attainment of

children in the household is higher. It is interesting that the male spouse’s educational

attainment is not statistically different from zero and suggests that males might not play as

important a role in directing the education of children in the household. This is in line with the

weak family structure, characteristic of the Dominican Republic, where consensual unions

predominate and, as a result, male partners end up having much less influence on family matters

than female spouses (Sana and Massey, 2005). Other determinants of children’s school

attendance include the number of school-age children in the household, which is positively

related to school attendance, thus suggesting that the educational attainment of all children in the

household is positively correlated. In contrast, the presence of younger children, such as

19

preschool-age children, lowers the likelihood of school attendance.23

Finally, the likelihood of

school attendance appears to significantly differ by birth order, with firstborns being 9

percentage-points more likely to attend school than their younger siblings.

In sum, these findings help us gauge the effect of remittance receipt on children’s school

attendance. However, can we say anything about the effect of migration and the need to

distinguish between the two effects? To illustrate the importance of separating the remittance

effect from the disruptive effect of contemporaneous family migration when assessing the impact

of remittance-receipt, we re-estimate our model adding children in migrant households. Table 7

displays the results from this exercise.

TABLE 7 ABOUT HERE

Once we add children in households with migrants currently in the U.S., remittances no

longer have a positive impact on children’s school attendance.24

Hence, the result in Table 7

suggests that the coefficient for remittance-receipt confounds the impact of family migration

with the remittance effect. It should be noted that our analysis only focuses on remittance receipt

and, as such, assumes that the remittance amounts received by migrant remittance-receiving

households and by non-migrant remittance-receiving households are the same. This, however,

only helps strengthen our findings. As shown in Table 3, non-migrant remittance-receiving

households in the LAMP report receiving smaller amounts of remittances (as a fraction of

household income) than migrant remittance-receiving households. Therefore, the migration

effect might actually be stronger than at first glance when using the receipt of remittances as our

key independent variable. In other words, migrant households are blessed with greater levels of

remittances and, yet, their children are less likely to be in school. It is also worth noting that the

result from Table 7 parallels the finding of Acosta et al. (2007), who examine the impact of

20

remittances on the educational attainment of children in the Dominican Republic without

distinguishing between children in migrant as opposed to non-migrant households.

Consequently, they conclude that remittances have no significant impact on the educational

attainment of children. In contrast, we find that, once we include children from migrant families

experiencing a contemporaneous family emigration effect, remittances no longer have the

positive impact on children’s schooling. Therefore, failure to properly separate the family

emigration effect from the remittance effect may underestimate the positive effect of remittances

on children’s schooling.

(b) Remittances and School Attendance According to Various Child Characteristics

To gain a better understanding of the impact of remittance inflows on children’s

schooling, we repeat the analysis for children in non-migrant households distinguishing

according to the child’s age, gender and order of birth children -- characteristics suggested as

being crucial in previous studies of human capital investments on children (e.g. Calero et al.

2009; Emerson and Portela Souza 2008).

Panel A in Table 8 shows the estimated impact of remittances on children’s schooling

according to whether the children are of primary school-age as opposed to of secondary school

age. It is not at all surprising to find that remittances exclusively promote the school attendance

of secondary school-age children. After all, school attendance entails higher explicit and implicit

costs for older children. They often have to travel further to get to school, raising the cost of

schooling. They also usually spend more money on books and related materials than their

younger counterparts. Furthermore, the opportunity cost of attending school is significantly

higher for older children than for the very young. As such, remittances, which help lift

households’ liquidity constraints, can have a significant impact on the school attendance of

21

secondary school- age children. Specifically, a 10 percentage-point increase in the likelihood of

receiving remittances raises the likelihood of school attendance by 3 percentage points, from an

average of 0.79 to approximately 0.82, among secondary school-age children in the household.

The positive finding for secondary school-age children coupled with no impact on primary

school-age children suggests that, when children of primary school-age are not enrolled in

school, it may be due to non-monetary factors. A Chow test of the equality of the estimated

effect of remittance receipt for primary and secondary school-age children reveals that those

coefficients are statistically different.

TABLE 8 ABOUT HERE

We also examine the differential impact of remittances on children’s school attendance

according to gender. According to the figures in Panel B, Table 8, the receipt of remittance

inflows appears to increase school attendance among girls, although the effect is only marginally

significant at the 10 percent level. It is, therefore, not surprising that the Chow test of the

equality of the estimated effect of remittance receipt for girls and for boys finds no evidence of a

statistically significant difference in the effect of remittance receipt on the school attendance by

gender.

Finally, we look at the differential impact that remittances may have on children’s school

attendance according to the order of birth. It is common for firstborns to be placed in a more

privileged position than higher order of birth children in many cultures. In that case, remittances

may be particularly helpful in improving the school attendance of the less privileged children.

Indeed, this is what we find in Panel C, Table 8. The receipt of remittances benefits higher order

of birth children, but not firstborns. Specifically, a 10 percentage-point increase in the likelihood

of receiving remittances raises the likelihood of school attendance by 5 percentage points, from

22

an average of 0.74 to approximately 0.79, among higher order of birth children in the household.

However, it does not appear to have a statistically different from zero impact on the school

attendance of firstborns. As such, the Chow test of the equality of the estimated effect of

remittance receipt for firstborns and for younger siblings indicates that the effect on remittance

receipt on the school attendance of both groups of children is statistically different.

In sum, girls appear to slightly benefit from the receipt of remittances. But it is primarily

secondary school-age children and higher order of birth siblings that gain the most from the

receipt of remittances.

6. Concluding remarks

The present study examines the impact of remittance receipt on the school attendance of

children in the Dominican Republic. We focus our analysis on children residing in households

with heads claiming no close family members in the U.S. –the destination of the vast majority of

Dominican emigrants. This focus is intended to help isolate the impact of remittance receipt

from that of contemporaneous household migration. While non-migrant households constitute a

selected group, children from non-migrant households account for 88 percent of our sample of

Dominican children and drive the average rate of remittance receipt in the sample. After all, 52

percent of children in remittance receiving households in the Dominican LAMP are living in

non-migrant households –defined as households with no members currently in the U.S.

We find that remittances do positively impact children’s school attendance. A 10

percentage point increase in the likelihood of receiving remittances raises the likelihood of

school attendance by approximately 3 percentage-points from an average of 76 to about 79

percent. Secondly, while statistically speaking, girls’ school atttendance appears to rise from the

receipt of remittances, secondary school-age children and younger siblings are the ones who gain

23

the most from the receipt of remittances by the household. A 10 percentage point increase in the

likelihood of remittance-receipt by the household raises their probability of school attendance by

3 and 5 percentage-points, respectively. Finally, we find empirical evidence of a confounding

negative impact of family migration on children’s schooling. The positive impact of remittance

receipt on children’s school attendance effectively disappears when we expand our sample to

include children in migrant households. This result confirms the existence of a disruptive effect

of family migration on children’s schooling. After all, the emigration of a family member often

imposes hardships on the family members left behind, including children, who may need to skip

school and work to make up for the monetary and non-monetary contributions that the migrant

family member made to the household before migrating. Alternatively, if Dominican education

is poorly rewarded in the destination countries of emigrants, expectations of future emigration

may reduce school attendance among children residing in migrant households. Overall, family

migration may temper the positive impact of remittance-receipt, which helps us understand

previous findings. In those cases, measurement of a zero impact of remittance income on

children’s educational attainment may result from confounding the remittance with the migration

effect.

In sum, our findings emphasize the need to account for migration effects when evaluating

the impact of remittances on children’s schooling and on other household activities. This is

especially important when the incidence of remittance-receipt and migration differ across the

population, as we observe in the Dominican Republic. While remittances tend to follow the out-

migration of a household member, it is not always so. The contemporaneous foreign residency

of a household member is not a precondition for the receipt of remittances, as half of the children

in the sample who receive remittances live in a non-migrant household. Therefore, remittances

24

are often forthcoming from individuals who are not considered household members, such as

distant relatives and friends. By the same token, migrant households do not necessarily receive

remittances. Only two-thirds of migrant households in the Dominican LAMP database enjoy

remittances.

From a policy perspective, our results underscore the importance of distinguishing

between the impacts of remittances and migration in policy-making. Specifically, if the

objective is to raise investments in children’s human capital, policies that favor migration when

remittances are more likely to follow and policies aimed at increasing remittance flows (e.g. by

lowering remitting costs or by offering matching funds) can prove particularly helpful for

developing countries impacted with extensive out-migration.

We would like to conclude with some suggestions for future research. This study has

revealed that, while remittances raise school attendance among children in the Dominican

Republic, emigration of family members appears to do the opposite. Yet, the fact that migration

does not help children’s school attendance does not mean migration is necessarily undesirable.

Perhaps migration serves to smooth out household income, as predicted by the New Economics

of Labor Migration. Alternatively, migration may help households maximize income or provide

the financial capital to begin or expand a family business. Working abroad may also facilitate

the accumulation of entrepreneurial capital to succeed in business. If those are the main motives

behind migration, migration may still be a desirable outcome for many households. Therefore, it

would be interesting for future work to examine the impact that migration, separately from

remittance-receipt, has on other household outcomes aside from children’s schooling. Likewise,

it would be useful to get a fuller understanding of the possibly divergent motives for sending

remittances according to whether they are sent to migrant versus non-migrant households. This

25

may help us better track the different impacts of remittances on the various household types.

Finally, better information on the frequency, levels and history of remittance receipt could also

help sharpen our estimates of migration and remittances on household outcomes and on the

overall well-being of households in developing economies.

26

References

Acosta, P., Fajnzylber, P., & Humberto Lopez, J. (2007). The impact of remittances on poverty

and human capital: Evidence from Latin American household surveys. World Bank Policy

Research Working Paper 4247, June 2007.

Amuedo-Dorantes, C., & Pozo, S. (2006). Remittances and Insurance: Evidence from Mexican

migrants. Journal of Population Economics, 19(2), 227-254.

Angrist, J. D., & Krueger, A. (2001). Instrumental variables and the search for identification:

From supply and demand to natural experiments. Journal of Economic Perspectives, 15(4),

69-85.

Barro, R., & Lee, J. W. (2000). Education attainment: Updates and implications. Center for

International Development, Working Paper # 42, April 2000, Available:

<http://www.cid.harvard.edu/ciddata/ciddata.html>.

Bansak, C., & Chezum, B. (2009). How do remittances affect human capital formation of school

age boys and girls? American Economic Association Papers & Proceedings, 99(2), 145-

48.

Basmann, R. L. (1960). On finite sample distributions of generalized classical linear

identifiability test statistics. Journal of the American Statistical Association, 55(292), 650-

659.

Baum, C. F., Schaffer, M. E., & Stillman, S. (2002). Instrumental variables and GMM:

Estimation and testing. Department of Economics, Boston College, Working paper no. 545,

Available: <http://fmwww.bc.edu/ec-p/WP545.pdf>.

Becker, G. (1974). A theory of social interactions. Journal of Political Economy, 82(6), 1063-

93.

27

Borraz, F. (2005). Assessing the impact of remittances on schooling: The Mexican experience.

Global Economy Journal [on-line serial], 5(1) available: bepress.com/gej/.

Breierova, L., & Duflo, E. (2004) The impact of education on fertility and child mortality: Do

fathers really matter less than mothers? NBER Working Paper No. 10513, National

Bureau of Economic Research.

Caceras, L. R. (2008). El Destino de las remesas en El Salvador. Comercio Exterior, 58(1), 27-

40.

Calero, C., Bedi, A. J., & Sparrow, R. (2009). Remittances, liquidity constraints and human

capital investments in Ecuador. World Development, 37(6), 1143-54.

Cox Edwards, A., & Ureta, M. (2003). International migration, remittances, and schooling:

evidence from El Salvador. Journal of Development Economics, 72(2), 429-61.

Davidson, R., & MacKinnon, J. G. (1993). Estimation and inference in econometrics. New

York: Oxford University Press.

Derby, L. (2009). The dictator’s seduction: Politics and the popular imagination in the era of

Trujillo. Durham: Duke University Press.

Durand, J., & Massey, D. S. (1992). Mexican migration to the United States: A critical review.

Latin American Research Review, 27(2), 3-42.

Emerson, P. M., & Portela Souza, A. (2008). Birth order, child labor and school attendance in

Brazil. World Development, 36(9), 1647-64.

Hanson, G. H., & Woodruff, C. (2003). Emigration and educational attainment in Mexico.

Working paper, University of California, San Diego.

Haveman, R. & Wolfe, B. (1995). The determinants of children’s attainments: A review of

methods and findings. Journal of Economic Literature, 33(4), 1829-1879.

28

Illahi, N. (2001). Children’s work and schooling: Does gender matter? Evidence from the Peru

LSMS panel data. The World Bank, Policy Research Working Paper Series, No. 2744.

IDB. (2004). Sending money home: Remittance recipient in the Dominican Republic and

remittance senders in the US. November 23, 2004, Mutilateral Investment Fund/Inter-

American Bank. Available:

<http://idbdocs.iadb.org/wsdocs/getdocument.aspx?docnum=546769>.

Kandel, W. & Kao, G. (2001). The impact of temporary labor migration on Mexican children’s

educational aspirations and performance. International Migration Review, 35(4), 1205-

1231.

Lucas, R. E. B., & Stark, O. (1985). Motivations to remit: Evidence from Botswana. Journal of

Political Economy, 93(5), 901-18.

McKenzie, D & Rapoport, H. (2006). Can migration reduce educational attainments?

Depressing evidence from Mexico. CReAM Discussion Paper Series, no. 01.

Newey, W. (1985). Generalized method of moments specification testing. Journal of

Econometrics, 29(3), 229-256.

OECD (2010). Latin American economic outlook 2010. Available:

<http://www.oecd.org/dataoecd/20/1/44535785.pdf>.

Sagás, E. & Molina, S. E. (2004). Dominican transnational migration. In E. Sagás & S. E.

Molina (Eds.), Dominican migration: Transnational perspectives, Gainsville: University

Press of Florida.

Sana, M. & Massey, D, (2005). Household composition, family migration and community

context: Migrant remittances in four countries. Social Science Quarterly, 86(2), 509-528.

29

Sargan, J.D. (1958). The estimation of economic relationships using instrumental variables.

Econometrica, 26(3), 393-415.

Schultz, T. P. (2002). Why governments should invest more to educate girls World

Development, 30(2), 207-225.

Secretaría de Estado de Educación. (2008). Indicadores Estadísticos 2007-2008. Santo

Domingo, Dominican Republic, Available at:

<http://www.see.gov.do/portalsee/planificacion/documentos/Boletindeindicadores2007-

08.pdf>.

Staiger, D. & Stock, J. H. (1997). Instrumental variables regressions when instruments are weak.

Econometrica, 65(3), 557-586.

Stark, O. (1991). The migration of labour. Oxford: Basil Blackwell.

UNESCO. (2009). Global education digest 2009: Comparing education statistics across the

world. Montreal: UNESCO Institute for Statistics.

Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. (2nd

ed.)

Cambridge, MA: The MIT Press.

World Bank, Migration and remittances factbook. compiled by Dilip Ratha and Zhimei Xu,

Available: <www.worldbank.org/prospects/migrationandremittances>.

30

Source: Computed by the authors from VII Censo nacional de población y vivienda, 2002 (2002

Dominican Census).

Figure 1: Incidence of Household Migration by Education Attainment of the Head

0

.05

.1

.15

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

31

Table 1. Descriptive Statistics for Children in Various Groups of Households

Variable Description All Households Non- Migrant Households Migrant Households

Statistic Mean S.D. Mean S.D. Mean S.D.

School Attendance 0.76 0.43 0.75 0.43 0.79 0.41

Primary school age Children (7-11 years) 0.71 0.45 0.72 0.45 0.65 0.48

Secondary school age Children (12-18 years) 0.79 0.41 0.77 0.42 0.89 0.31

Girls 0.76 0.42 0.75 0.43 0.82 0.38

Boys 0.75 0.44 0.74 0.44 0.75 0.43

Firstborn Children 0.78 0.42 0.77 0.42 0.83 0.38

Higher Birth Order Children 0.74 0.44 0.73 0.44 0.76 0.43

Remittance-receiving Household 0.19† 0.40 0.13 0.33 0.66*** 0.47

Employed Household Head 0.80 0.40 0.81 0.40 0.74 0.44

Female Spouse or Head has 6 years of schooling or less 0.22 0.41 0.20 0.40 0.34*** 0.47

Female Spouse or Head has at least some secondary schooling 0.78 0.41 0.80 0.40 0.66*** 0.47

Male Spouse or Head has 6 years of schooling or less 0.30 0.46 0.30 0.46 0.30 0.46

Male Spouse or Head has at least some secondary schooling 0.70 0.46 0.70 0.46 0.70 0.46

Household Assets (# Parcels of land + # Businesses + # Houses) 1.10 1.04 1.08 1.01 1.26* 1.20

Number of Adults in the Household 3.74† 2.64 3.48 2.33 5.55*** 3.76

Number of School age Children in the Household 2.77 1.47 2.78 1.47 2.69 1.50

Number of Preschool age Children in the Household 0.63 0.93 0.65 0.94 0.49 0.82

Boy 0.49 0.50 0.50 0.50 0.44 0.50

Child’s Age 12.39 3.52 12.35 3.54 12.68 3.33

Firstborn Child 0.47 0.50 0.46 0.50 0.50 0.50

Community no. 1 0.16 0.37 0.14 0.34 0.33 0.47

Community no. 2 0.13 0.33 0.13 0.34 0.06 0.25

Community no. 3 0.08 0.27 0.08 0.28 0.06 0.25

Community no. 4 0.09 0.29 0.10 0.30 0.01 0.12

Community no. 5 0.11 0.31 0.10 0.30 0.16 0.37

Community no. 6 0.17 0.37 0.16 0.37 0.19 0.40

Community no. 7 0.27 0.44 0.28 0.45 0.17 0.38

Unemployment Rate in U.S. Destination State 10.98 0.94 10.98 0.90 11.00 1.14

Earnings in U.S. Destination Statea 8774.40 291.99 8765.02 249.39 8840.26* 493.77

Number of Observations 1123 983 140

Notes: (a) Real Yearly Earnings for Personal Care Service Workers in U.S. Destination State. *, ** and *** signal statistical significant differences in

the means for children in non-migrant and migrant households at the 10%, 5%, and 1% levels, respectively. Means for children in all versus non-

migrant households are not statistically different from each other except for those cases marked with: † in which case they are different at the 5% level.

32

Table 2. School Attendance of All Children in Migrant and in Non-Migrant Households

By Group of Children: N School Attendance Difference t-statistic

All Children 1123 0.755

By Age:

Primary school age 487 0.710 - -

Secondary school age 636 0.789 -0.079*** 3.012

By Gender:

Boys 550 0.745 - -

Girls 573 0.764 -0.019 0.73

By Birth Order:

Firstborns 524 0.777 - -

Higher Birth Order Children 599 0.736 0.040* 1.58

Notes: *** stands for statistical significance at the 1% level and * at the 10% level.

33

Table 3. Size of Remittance Receipt

Size of Remittance Inflow All Recipients Remittance-receiving

Migrant HHs

Remittance-receiving

Non-migrant HHs

Small portion of HH income 0.48 0.43 0.58

Medium portion of HH income 0.24 0.26 0.22

Substantial portion of HH income 0.27 0.31 0.20

Table 4. All School-age Children (Ages 7-18)

Count of:

Living in

Migrant

Households

Living in

Non-Migrant

Households

All Children

All Children 140 983 1123

Children in Remittance-Receiving Households 93 124 217

Table 5. Household Migration Networks

U.S. State Incidence of Networks in U.S. State

New York 0.68

New Jersey 0.12

Puerto Rico 0.08

Massachusetts 0.03

Florida 0.02

Rhode Island, Pennsylvania Between 0.01 and 0.02

Connecticut, California, Illinois, Kansas, North Carolina, South Carolina Less than 0.01

34

Table 6. Two-Stage Linear Probability Model of School Attendance Using Children in Non-Migrant Households

Second-Stage Results (School Attendance) Coefficient S.E.

Household Receives Remittances 0.286** 0.139

Household Head is Employed 0.004 0.069

Potential Educational Attainment of Female Spouse or Head 0.160*** 0.064

Potential Educational Attainment of Male Spouse or Head 0.005 0.060

Household Assets -0.001 0.025

Number of Adults in the Household 0.003 0.011

Number of School Aged Children in the Household 0.037** 0.016

Number of Preschool Aged Children in the Household -0.084** 0.034

Boy -0.036 0.039

Child’s Age -0.002 0.007

Firstborn Child 0.090** 0.037

Community Dummies Yes

First-Stage Results (Household Remittance Receipt) Coefficient S.E.

Household Head is Employed -0.137*** 0.046

Potential Educational Attainment of Female Spouse or Head 0.012 0.042

Potential Educational Attainment of Male Spouse or Head 0.017 0.035

Household Assets 0.001 0.015

Number of Adults in the Household -0.013** 0.006

Number of School Aged Children in the Household -0.035*** 0.009

Number of Preschool Aged Children in the Household 0.035* 0.019

Boy 0.013 0.027

Child’s Age 0.005 0.004

Firstborn Child -0.024 0.031

Unemployment Rate in U.S. Destination State 0.340*** 0.023

Earnings in U.S. Destination State 0.001*** 0.000

Community Dummies Yes

Number of Observations 983

Number of Family Clusters 467

Prob > F 0.000

Correlation of Instruments with Endogenous Variable:

F-statistic 15.03

Prob > F 0.000

Over-identification Test of Instruments:

Sargan Test 0.668

Prob > Chi-square 0.412

Basmann Test 0.655

Prob > F 0.418

Notes: All regressions include a constant term. Standard errors correct for clustering at the household level.

***Significant at the 1% level, **significant at 5% level, and *significant at the 10% level. We use state

occupational wages for personal service providers and state unemployment rates as instruments for remittances.

35

Table 7. Two-Stage Linear Probability Model of School Attendance Using All Children

Second-Stage Results (School Attendance) Coefficient S.E.

Household Receives Remittances -0.097 0.475

Household Head is Employed -0.065 0.081

Potential Educational Attainment of Female Spouse or Head 0.153*** 0.058

Potential Educational Attainment of Male Spouse or Head -0.008 0.052

Household Assets 0.009 0.019

Number of Adults in the Household -0.002 0.010

Number of School Aged Children in the Household 0.017 0.023

Number of Preschool Aged Children in the Household -0.067*** 0.026

Boy -0.060* 0.037

Child’s Age 0.001 0.007

Firstborn Child 0.079** 0.037

Community Dummies Yes

First-Stage Results (Household Remittance Receipt) Coefficient S.E.

Household Head is Employed -0.132*** 0.044

Potential Educational Attainment of Female Spouse or Head 0.019 0.040

Potential Educational Attainment of Male Spouse or Head 0.031 0.036

Household Assets 0.014 0.018

Number of Adults in the Household 0.014** 0.007

Number of School Aged Children in the Household -0.040*** 0.009

Number of Preschool Aged Children in the Household 0.005 0.019

Boy -0.029 0.029

Child’s Age 0.001 0.005

Firstborn Child -0.014 0.033

Unemployment Rate in U.S. Destination State 0.308*** 0.080

Earnings in U.S. Destination State 0.001*** 0.000

Community Dummies Yes

Number of Observations 1123

Number of Family Clusters 542

Prob > F 0.000

Correlation of Instruments with Endogenous Variable:

F-statistic 8.05

Prob > F 0.000

Over-identification Test of Instruments:

Sargan Test 0.186

Prob > Chi-square 0.666

Basmann Test 0.183

Prob > F 0.669

Notes: All regressions include a constant term. Standard errors correct for clustering at the household level.

***Significant at the 1% level, **significant at 5% level, and *significant at the 10% level. We use state

occupational wages for personal service providers and state unemployment rates as instruments for remittances.

36

Table 8. Two-Stage Linear Probability Models of School Attendance by Various Children Characteristics Using Children in Non-Migrant Households

Panel A:

Second-Stage Results (School Attendance)

Primary School age Secondary School age

Coefficient S.E. Coefficient S.E.

Household Receives Remittances -2.256 3.782 0.293*** 0.119

Number of Observations 430 553

Number of Family Clusters 293 323

Prob > F 0.000 0.000

Chow Test of Equality of Coefficients F(1, 466) = 6.87 with Prob > F = 0.009

Panel B:

Second-Stage Results (School Attendance)

Boys Girls

Coefficient S.E. Coefficient S.E.

Household Receives Remittances 0.224 0.175 0.301* 0.211

Number of Observations 489 494

Number of Family Clusters 330 328

Prob > F 0.000 0.000

Chow Test of Equality of Coefficients F(1, 466) = 0.15 with Prob > F = 0.699

Panel C:

Second-Stage Results (School Attendance)

Firstborns Higher Birth Order Kids

Coefficient S.E. Coefficient S.E.

Household Receives Remittances 0.106 0.129 0.532*** 0.139

Number of Observations 454 529

Number of Family Clusters 446 307

Prob > F 0.000 0.000

Chow Test of Equality of Coefficients F(1, 466) = 11.15 with Prob > F = 0.001

Notes: All regressions include a constant term. Standard errors correct for clustering at the household level. *** Significant at the 1% level and *significant

at 10% level. Instruments are: 1) unemployment rates in states from which remittances are likely to originate and 2) yearly real earnings for service

employees in states from which remittances are likely to originate. Originating states were determined by state location of family members who had migrated

in the past.

37

1 See Durand and Massey (1992) for a review of studies suggesting that remittances are used in

“non-productive” ways and Caceras (2008) for a more recent study arguing likewise for El

Salvador.

2 As we shall note in what follows, the U.S. is, by far, the main destination of Dominican

emigrants.

3 This conclusion is consistent with the claim by the OECD (2010) that about 50 percent of the

Dominican-origin population in the U.S. has completed secondary school and with the

observation that educational attainment of Dominicans in the U.S. exceeds, on average, that of

Mexican-origin immigrants in the U.S.

4 They find, however, that remittances do increase educational attainment in the case of the other

Latin American nations in their study.

5 Their study uses past literacy rates and political unrest indicators as instruments for remittances

and absenteeism. While the two instruments are likely to be related to the variables being

instrumented, they are also correlated to children’s educational attainment. Furthermore, to the

extent that both instruments contribute to absenteeism and family remittances, it is not feasible to

separate the two competing impacts.

6 The Latin American Migration Project (LAMP) is a collaborative research project based at

Princeton University and the University of Guadalajara, supported by the National Institute of

Child Health and Human Development (NICHD). The LAMP website is:

http://lamp.opr.princeton.edu/.

38

7 Information on whether any of the household members (listed in Table A of the LAMP-DR7

questionnaire) currently resides in the U.S. is gathered from Table D of the LAMP-DR7

questionnaire and is available for each household member in the PERS file.

8 We only have information on the receipt of remittances by the household, but not on the

amount received.

9 We do not know who in the household receives the remittances. We only know that

remittances are received by the household.

10 We also examine how the descriptive statistics for the entire sample of children differ from

those of children in non-migrant households. For the most part, the two are quite similar, which

is not surprising given that the vast majority of children in the LAMP-DR7 reside in non-migrant

households, that is 983 children out of 1123. However we do find that there is a statistical

significant difference in the proportion of households that receive remittances (0.19 for all

children and 0.13 for children in non-migrant households) and in the number of adult family

members (3.74 for all children and 3.48 for children residing in non-migrant households).

11 Specifically, Table J3 in the LAMP-DR7 questionnaire asks households: “Does this household

currently receive any money from people in the U.S.?”

12 Specifically, respondents are asked the following question in Table J3 in the LAMP-DR7

questionnaire: “Compared to the average household income of your household, would you say

that these remittances constitute a small portion of your household income, a medium portion, or

a substantial portion?”

13 The case can be made that some of these variables are endogenous. Therefore, we have also

estimated more parsimonious specifications that exclude some of these potentially endogenous

variables. The results, available from the authors, prove to be robust to these exclusions.

39

14

Unfortunately, the survey does not allow us to decipher whether the female and male spouse

are, respectively, the mother and father of the child.

15 Kandel and Kao (2001) suggest this to be the case in Mexico. They find that children in

families with high U.S. emigration probabilities are less likely to go to school. Such a response

is logical since Mexican education (particularly at the primary and secondary level) is not likely

to be recognized and rewarded in the U.S. labor market.

16 Credit should be given to the LAMP designers for making it possible to identify both

remittance-receiving and migrant households. Many other surveys that report on both migration

and remittances ask about remittances only in households that claim migrant members. In other

words, they conclude that the migration of a family member is a pre-condition for the receipt of

remittances and design the survey questionnaire in such a way that they never obtain remittance

information from households that do not indicate having a household member currently living

abroad. The LAMP survey results demonstrate that, in many instances, households without

migrant members still receive remittances. It is reasonable to assume that these then are from

more distant family relatives or friends. In short, household migration is not a precondition for

the receipt of remittances.

17 While it may seem logical to also compare children’s school attendance in migrant households

according to their receipt of remittances, the small number of children in migrant households

impedes this exercise. As indicated in Table 1, there are only 140 children in migrant

households and, of these, ninety-three live in households that receive remittances and only 47

residing in households that do not receive remittances. These sample sizes are too small to yield

statistically meaningful results.

18 The survey does not provide us with information on household income.

40

19

The F-statistic is 10.78 and the Prob > F = 0.000. The small p-value indicates that OLS is not

consistent.

20 The assumption is that the current or past U.S. location of family and household members is

likely to be the geographical source of any remittances received by the household today.

Unemployment rates and earnings data corresponding to those locations were obtained from the

Bureau of Labor Statistics at: http://www.bls.gov/.

21 The F-statistic is larger than 10 in both instances.

22 Note that the likelihood of receiving remittances is instrumented using continuous variables,

thus changing the interpretation of its coefficient from the interpretation of a dichotomous

variable to that of a continuous one.

23 School attendance is positively correlated to the number of school-age children in our sample.

The correlation coefficient is equal to 0.2505. In contrast, school attendance is inversely related

to the number of preschool-age children in the household, for which the correlation coefficient

equals -0.0207.

24 A Chow test of the equality of the estimated coefficients for remittance-receipt using all

children and using children in non-migrant households reveals that the two are statistically

different from each other. Specifically, the F-statistic equals 5.92 and is statistically significant

at the 5 percent level.