Educational Selectivity of Migrants and Current School Enrollment of Children Left behind: Analyses in Three African Countries

Abstract

Migration of household members is often undertaken to improve the well-being of individuals remaining in the household. Despite this, research has demonstrated inconsistent associations between migration and children’s well-being across sending areas and types of migration. To understand the degree to which different types of migration and migrants are associated with schooling, we analyze comparable data across three African countries differing in prevalence, type, and selectivity of migration. Results suggest that recent migration is differentially associated with left-behind children’s school enrollment across settings. When analyses are restricted to migrant-sending households, however, migrant selectivity is positively associated with school enrollment.

Keywords

migration children schooling

Introduction

In 2015, the United Nations adopted the Sustainable Development Goals to improve human development around the globe (UNDP 2015b). Goal 4 sets a target of achieving universal primary and secondary education for all children by 2030. To reach such a macro-level goal, it is necessary to understand the individual- and family-level factors that impede and facilitate sending children to school, keeping them in school, and achieving while in school, particularly in contexts where overall education levels are historically low.

Labor migration can be an important family strategy for increasing resources and investing in children’s well-being (Stark 1991). Economic resources generated by a migrating household member (most often a parent) can then be used to enhance children’s access, persistence, and success in formal school settings (Edwards and Ureta 2003; Chen et al. 2009; Amuedo-Dorantes, Georges, and Pozo 2010; Piotrowski and Paat 2012). Yet migration may also create a competing activity that reduces children’s school enrollment, becoming a normative path that reduces incentives for furthering education among young people who intend to become labor migrants themselves (Kandel and Kao 2000; Fox et al. 2012). In this case, migration and education are not complementary but competitive routes to social mobility in migrant-sending communities. The extant literature contains findings consistent with both roles of migration, suggesting positive and negative forces on children’s schooling (Deb and Seck 2009; Yao and Treiman 2011; McKenzie and Rapoport 2011; Meyerhoefer and Chen 2011; Robles and Oropesa 2011).

Assessing migration’s role on children’s well-being is also complicated because selection into migration and returns from migration itself can operate differently across contexts. In well-established settings of labor migration, the incentive for migration may be highest among those with lower stores of human capital (i.e., negative selection on education). In other settings, particularly where the cost of migration is higher and migrant networks are less established, migration occurs more frequently among those with greater human capital (i.e., positive selection on education) (Feliciano 2005; Takenaka and Pren 2010). Thus, there may be great variation in the selection of migrants and migration’s subsequent impact on children from diverse regions.

This paper takes advantage of comparable data across three dynamic settings in Sub-Saharan Africa to address two research questions:

Research Question 1: Is recent migration from the household similarly associated with school enrollment of left-behind children across three different settings characterized by both internal and international labor migration?

Research Question 2: Does the association between recent migration and children’s school enrollment vary depending on whether migrants are positively selected on education?

To answer the latter question, we construct a measure of migrant selectivity based on the migrant’s location in the educational distribution of the population of the same age and gender in the sending context. The social context in which migration occurs may alter the relationship between education and the returns to migration for children left behind by migrants. Thus, we compare patterns of school enrollment among children in sending households in three understudied settings of migration: Burkina Faso, Kenya, and Senegal. We further consider the role of destination (international vs. domestic destinations) and remittances in the relationship between migrant selectivity and children’s school enrollment in these three diverse settings.

Background

Migration is not a random event, and migrants are not randomly selected from across the economic and social realms of sending communities. In the case of labor migration, many migrants come from a middle strata of human capital, where resources are high enough to fund migration trips but low enough to work as an incentive (i.e., a push) to migration (Kaestner and Malamud 2014). The probability of international migration varies depending on the maturity of migrant systems and the degree of educational inequality in sending communities (Takenaka and Pren 2010). Living in a community with a higher prevalence of international out-migration, having family members with migration experience, and coming from households with more human capital (education) are all associated with an increased likelihood of migration (Massey and Aysa 2005). Costs for internal migration may be lower than for international migration, but here too, depending on the setting, human capital, household resources, and community characteristics can help predict who becomes a migrant (De Vreyer, Gubert, and Roubaud 2010). Internal migrants, however, may need lower stores of human capital and resources than international migrants from the same sending communities (Binci and Giannelli forthcoming).

Overall, there is strong evidence of positive selection among many international migrants. Positive selection is evident when individuals who have higher stores of human capital than average for their sending community have greater opportunities through migration than through remaining in their community. In a US context, Feliciano (2005) finds that most immigrants are positively selected on education when compared to their origin countries. Positive selection from developing settings is often equated with “brain drain,” particularly in the case of legal international migration to the United States or other OECD countries (Adams 2003). Among health care workers from Sub-Saharan Africa, for example, demand for skills is high abroad and often poorly compensated in the country of origin (Syred 2011). However, positive selection is not restricted to international migration. There is some evidence of positive educational selectivity for internal migration into capital cities in West Africa (De Vreyer, Gubert, and Roubaud 2010). In particular, female education is predictive of migration from rural to urban areas in Sub-Saharan Africa (Brockerhoff and Eu 1993; Reed, Andrzejewski, and White 2010).

Negative selection, on the other hand, is evident when individuals with lower stores of human capital compared to the average in the sending community are drawn into migration. Negative selection is more likely when local labor markets disfavor poorly educated workers and destination markets, including other rural areas, can offer employment opportunities for those with low levels of education (Choe and Chrite 2014). Negative selection is also more likely where costs and other barriers to migration are low and migration networks are available (Orrenius and Zavodny 2005; McKenzie and Rapoport 2010). In Mexico, for example, international migrants tend to be positively selected from sending communities, but internal migrants, particularly those headed to other agricultural areas within Mexico, are less positively selected on education (Mora and Taylor 2005; Deb and Seck 2009). Although rural-urban internal migrants tend to have higher levels of education than their rural counterparts in countries we compare here, including Kenya and Senegal (de Brauw, Mueller, and Lee 2014), there is less positive selection in Sub-Saharan African countries for internal rural-rural migrants who move to work in agriculture or extraction (i.e., mining). In this case, internal migrants may be more likely to be negatively selected than international migrants who face more structural barriers and distance to travel (Mora and Taylor 2005).

Other characteristics associated with migrant selectivity may also be associated with children’s schooling. Some migrants, for instance, may be better positioned to remit back to origin households. In many sending communities around the world, remittances have a positive effect on children’s entrance to and persistence in school (Edwards and Ureta 2003; Chen et al. 2009; Amuedo-Dorantes, Georges, and Pozo 2010; Piotrowski and Paat 2012). In western China, remittances have a positive effect on the educational performance of children of internal migrants (Hu 2013), and in Ghana, South Africa, and Vietnam, these offset the negative impact of parental absence of children’s school enrollment (Adams, Cuecuecha, and Page 2008; Yao and Treiman 2011).

One approach researchers take to deal with the role of migrant selection in relation to children’s well-being is to adjust for migrant selection statistically (i.e., with instrumental variables, propensity scores, or other techniques) to help articulate the causal relationship between migration and impacts on origin communities. Previous work taking these approaches has yielded very different results for the association between migration and children’s outcomes across a diverse array of sending communities (Halpern-Manners 2011; Antman 2012; Hu 2013). Rather than seeking to explain away the selection effect, we focus here on understanding whether such differential migrant selection can itself help explain some of these varied outcomes for children in migrant-sending households. Our analyses attend to the type of selection (relatively positive or negative), type of destination (internal or international), and receipt of remittances as important mechanisms through which migration may be associated with children’s school enrollment in diverse and understudied settings of migration.

Research Settings

Our study examines whether migrant selectivity is differentially associated with children’s school enrollment in three African countries: Burkina Faso, Kenya, and Senegal (Figure 1). These countries represent various subregions of Sub-Saharan Africa: Burkina Faso and Senegal in West Africa and Kenya in East Africa. They also differ in the prevalence, type, and selectivity of migration as well as education levels, education opportunities, and education systems (République du Sénégal 2003; Plaza, Navarrete, and Ratha 2011; Miller and Elman 2013; UNESCO Institute for Statistics 2013). Though school-age children are expected to attend school in all three countries, many do not because they lack access to schools, have parents who cannot afford school-related expenses, or are needed for paid or unpaid labor. Compulsory education is rarely, if ever, enforced, and children and parents do not suffer sanctions for not attending school.

Figure 1.

Map of Study Countries.

In all three countries, internal migration, primarily rural-urban, is relatively common (de Brauw, Mueller, and Lee 2014). The opportunity to receive higher wages in both the formal and informal sectors attracts rural migrants to cities and towns (de Brauw, Mueller, and Lee 2014). The scarcity of secondary schools in rural areas also increases the likelihood that individuals migrate to urban areas for schooling-related reasons (de Brauw, Mueller, and Lee 2014). International migration, on the other hand, varies considerably across the three countries in prevalence and in the number and distance of countries to which migrants travel (Plaza, Navarrete, and Ratha 2011). These differences are likely a function of past colonialism, geographic location, and available local economic opportunities.

Burkina Faso

Burkina Faso, a landlocked country in West Africa, is one of the poorest and least developed countries in the world. Gross national income per capita (GNI) is $1,591 (2011 PPP $) (UNDP 2015a). According to the Human Development Index (HDI), it ranks 183 out of 188 countries in its level of development. Burkina Faso is also one of the least educated countries in the world: the adult literacy rate is 28.7 percent among the population 15 years and older (UNESCO Institute for Statistics 2013). Although primary education is free and children are expected to attend school for six years, parents are required to pay for books and supplies and contribute to school funds as needed. Parents may also weigh schooling’s long-term benefits against the short-term loss of income if children attend school rather than engage in paid labor. These costs may deter parents from sending their children to school. Approximately 39 percent of five- to 14-year-old children are involved in economic activities in Burkina Faso (UNICEF 2014). School fees required for secondary school can also serve as an obstacle to children’s continuing education. Very few children, however, are eligible to attend secondary school given that most do not finish primary school. In 2006, only 31 percent of children in the relevant age group completed primary school (de Hoop and Rosati 2014).

Approximately 71 percent of Burkina Faso’s population lives in rural areas (United Nations, Department of Economic and Social Affairs, and Population Division 2014) and relies on subsistence agriculture. Frequent droughts, especially in the northern part of the country, make it difficult for people to make a living (Niemeijer and Mazzucato 2002). Labor migration is thus an important strategy that households employ to diversify their risk and protect themselves against crop failure (Wouterse and Taylor 2008). Burkina Faso has a long history of internal and international migration (Cordell, Gregory, and Piché 1996; Konseiga 2007). As in most African countries, internal migration in Burkina Faso is largely rural-urban (de Brauw, Mueller, and Lee 2014) and directed toward large cities, especially Ouagadougou and Bobo Dioulasso (Beauchemin and Schoumaker 2005). Most migration, however, is international and likely influenced by the country’s colonial past. During much of the twentieth century, while under French colonial rule, Burkina Faso experienced a pattern of male workers migrating to Cote d’Ivoire’s cocoa and coffee plantations (Cordell, Gregory, and Piché 1996). Though Burkina Faso gained political independence more than 50 years ago, labor migrants continue to travel to Cote d’Ivoire in search of work. Today, more than 80 percent of international migrants from Burkina Faso migrate to Cote d’Ivoire (Plaza, Navarrete, and Ratha 2011), and most of them have little or no education.

Kenya

Of the three countries in this study, Kenya performs the best on economic, development, and education indicators. It is ranked 145 out of 188 countries in its level of development, and its GNI per capita is $2,762 (2011 PPP $) (UNDP 2015a). Since political independence, Kenya’s government has made education a priority (Somerset 2009). Its total public expenditure on education, 7.2 percent of GNP, is higher than the average for Sub-Saharan Africa (Miller and Elman 2013). The importance placed on education can be observed in Kenya’s relatively high education levels: its adult literacy rate is 72.2 percent (UNESCO Institute for Statistics 2013). Children start school at age six and are expected to attend eight years of primary school and four years of secondary school. Kenya implemented three separate initiatives to expand access to free primary education, the first in 1974, the second in 1979, and the most recent in 2003. The number of children enrolled in school dramatically increased following each initiative (Somerset 2009).¹ The most recent initiative abolished all direct costs, including those for books, school supplies, and school maintenance, and thus removed a major barrier to attending primary school. Although primary school is free and most children attend primary school for at least a few years, not all children pass the primary school leaving exam to continue on to secondary school. Even if children pass this exam, parents may lack funds to pay secondary school fees.

Similar to Burkina Faso, approximately 75 percent of Kenya’s population lives in rural areas (United Nations, Department of Economic and Social Affairs, and Population Division 2014) and engages in subsistence agriculture. To increase livelihood security, many rural households engage in internal labor migration, mainly toward urban centers with more employment opportunities. Circular migration — where a family member migrates to another area to live and work, sends remittances, and eventually returns home — is not uncommon (Bigsten 1996). In such cases, the household head, usually the husband, migrates to an urban area, leaving his wife and children behind (Agesa 2004). There is also a great deal of seasonal male migration from rural areas to other rural areas. Though less common, some Kenyan migrants move to international destinations such as the United States and United Kingdom (Plaza, Navarette, and Ratha 2011). Relatively few, however, migrate to other African destinations. Compared to migrants from Burkina Faso and Senegal, Kenyan migrants have much higher levels of education, suggesting more positive selection particularly in the case of international migration.

Senegal

Senegal, located in West Africa, ranks as one of the poorest and least developed countries in the world, although it performs slightly better than Burkina Faso on a number of economic, development, and education indicators. Senegal’s GNI per capita is $2,188 (2011 PPP $) (UNDP 2015a), and the country ranks 170 out of 188 countries in its level of development (UNDP 2015a). Education levels are relatively low: Only half the population 15 years and older is literate (UNESCO Institute for Statistics 2013). In 2001, Senegal attempted to increase education levels through a country-wide effort to guarantee education to all children (République du Sénégal 2003). With the financial assistance of the United Nations and donors, the Senegalese government increased the supply of schools and teachers, and as a consequence, primary school enrollment grew from 62 percent in 2003 to 73 percent in 2009.² In Senegal, the official starting age is seven years, and school fees are not required for primary or secondary school. Despite free schooling, many children do not attend school, often because parents need them to work. Approximately 17 percent of five- to 14-year-old children are involved in child labor (UNICEF 2014).

Senegal has a long history of migration, initially as a destination for migrants and more recently as a country of emigration (Adepoju 2004). In 2005, nearly half a million Senegalese lived outside Senegal (D. Ratha and Xu 2007). Among these migrants, 40 percent were living in other African countries (Gerdes 2007), representing more varied destinations than migrants from Burkina Faso. The destination countries of Senegalese migrants have changed over time. In the 1960s, the top destinations were Mauritania, Mali, Guinea, and Guinea-Bissau. By the late 1960s, however, Cote d’Ivoire and Gabon began replacing these countries in popularity. Migration to Cote d’Ivoire continued, becoming the top destination for Senegalese migrants moving within Africa, until civil war broke out in Cote d’Ivoire in the early 2000s and disrupted this flow (International Organization for Migration 2009). Now Gambia is the most popular destination.

Senegalese migration to non-African destinations commenced while Senegal was a French colony. During World War I and II, many Senegalese men served in the French army and remained in France after their service, finding employment in Marseille harbor (Robin, Lalou, and Ndiaye 2000; Gerdes 2007). After Senegal gained political independence in 1960, migration toward France intensified with the recruitment of Senegalese workers to France’s growing automobile industry (Pison 1997). In 1985, migration to France became more difficult because Senegalese now needed visas to enter the country. Consequently, Senegalese migration moved toward other European countries, particularly Italy and Spain (Gerdes 2007; Toma and Castagnone 2015). Many international migrants come from some of Senegal’s poorest regions and have little or no formal education (Diatta and Mbow 1999; Auriol and Demonsant 2012); however, a substantial minority, approximately 25 percent, represent the other end of the spectrum with tertiary education (International Organization for Migration 2009).

The Current Study

In this study, we first ask whether recent labor migration is associated with school enrollment of left-behind children when compared to children in nonmigrant households. We examine the extent to which international or internal migration (or both) are positively associated with school enrollment even beyond the educational attainment levels of adults in the sending households when compared to households without migrants. We expect some differences across the three contexts we compare. In Burkina Faso and Senegal, where migrants tend to be less positively selected, we expect a smaller association of migration and children’s school enrollment than in Kenya, when analyses compare migrant households to nonmigrant households.

We next consider whether the educational selection of migrants themselves has an additional role to play in children’s school enrollment. If we observe that children are more likely to be enrolled in school in the presence of a positively selected migrant even when adjusting for the education levels of other adults in the household and household resources, it would suggest that the dichotomous migrant versus nonmigrant distinction misses an important part of migration’s role for those left behind. Rather, such a result would indicate that differential returns to migration for sending households are associated with the relative “quality” of the migrant they are able to send.

We further examine the role of educational selection beyond its impact on the type of migration (i.e., beyond the possibility that more positively selected migrants travel internationally versus internally) or remittances (i.e., beyond the possibility that positively selected migrants are more likely to remit). These analyses are confined to children in migrant-sending households only. Here too, we expect to observe some differences across our three contexts. Economic returns to education in Burkina Faso and Senegal are lower than in Kenya, so having a positively selected migrant (i.e., living in a household sending migrants with relatively higher levels of education than the context in general) will be positively associated with children’s school enrollment in Burkina Faso and Senegal, with a smaller role of migrant selection observed in Kenya.

Data and Methods

There are comparatively few studies of children in sending communities in Africa. We use data from the Migration and Remittances Household Surveys conducted as part of the World Bank’s Africa Migration Project. These cross-sectional surveys, conducted in 2009 and 2010 in six Sub-Saharan African countries (Burkina Faso, Kenya, Nigeria, Senegal, South Africa, and Uganda), collected comparable household-level data on the characteristics of migrants in sending households, remittances sent to their households, and the characteristics of return migration.³ Different methodologies, however, were used to obtain the sampling frames in each country. In Senegal, a nationally representative sampling frame was used while in Burkina Faso and Kenya, sampling frames were representative at the provincial and district levels, respectively. Because the Migration and Remittances Household Surveys were designed to gather information about both internal and international migrant-sending households, each country’s sampling frame needed to include a sufficient number of migrant households. Even in countries with high rates of international migration, it is difficult to find households that have migrants currently living abroad (McKenzie and Mistiaen 2007). To capture a sufficient number of migrant-sending households in the primary sampling units, survey teams first conducted household listings with the purpose of classifying households into one of three categories: nonmigrant, internal migrant, and international migrant. A household was considered to be a migrant-sending household if at least one former household member lived in another village, urban area, or country for at least six months before the time of the survey. Next, survey teams randomly selected equal numbers of households according to their migration status, purposively oversampling internal migrant and international migrant households. In each household, interviewers surveyed the household head on all modules except the return migrant module, for which the return migrant was surveyed. For the purposes of this survey, the household head was required to be a regular household member who was primarily responsible for managing the household’s financial resources, regardless of whether household members usually viewed this member as the household head. Thus, for this survey, household members living outside the household, including migrants, were not eligible to be the household head.⁴

Our analytic sample is composed of school-aged children reported by the household head to be living in the surveyed households at the time of the survey. In Burkina Faso and Kenya, eligible children were aged six to 17 years, and in Senegal, they were seven to 17 years.⁵ We excluded children from the household who were living elsewhere at the time of the survey because information on their schooling status was not collected in all three countries. If children are living outside the household for schooling-related reasons and migrant-sending households are more likely to have children living outside the household for this reason, then our study could underestimate migration’s effect on school enrollment.⁶ We further restricted our sample to children with nonmissing data on all dependent and independent variables: Burkina Faso (N = 6,621), Kenya (N = 2,182), and Senegal (N = 5,220).⁷

Outcome Measure

The data set is primarily focused on migration and the economic status of households with fewer measures available to compare children’s outcomes across countries. Although several measures of education, including educational attainment and number of grades of schooling completed, were collected in the Migration and Remittances Household Survey, we focused on children’s current school enrollment as our primary outcome of interest because it was the only schooling variable consistently measured across the three countries. Furthermore, current school enrollment measures a recent decision made by parents or the household head to send or keep a child in school and thus is more likely to be affected by the recent labor migration of a household member. Educational attainment and grades of schooling completed capture a child’s cumulative schooling history and may be less impacted by recent migration.

We created the variable current school enrollment, using information collected in the household roster. This variable is measured using the household head’s response to the following question: “What is [NAME]’s current work situation?” If the respondent reported “full-time student,” then we coded the child as currently in school.⁸ All other responses were coded as not enrolled in school. We note that this is a conservative measure of school enrollment because children working and going to school may not be coded as being in school.

Migration Variables

Our focus is on recent labor migration out of the focal households. We coded migration using information collected from each household in a roster of former household members reported by the household head. Former household members are coded as migrants if they had lived outside the household for more than six months before the time of the survey.⁹ Household heads answered a series of questions about each migrant’s sociodemographic characteristics and migration experience. The first variable, household migration status, is a binary variable indicating whether a household had a former household member working as a labor migrant. A former household member is defined as a labor migrant if the household head reported work-related reasons as the primary reason that a given household member was living outside the household. Because some labor migrants may have left the household decades ago and had little to no contact with the household, we limited our migrant sample to labor migrants who were reported to be living in his or her current location for the last five years. Ideally, we would have restricted the sample to migrants who left the household in the last five years; however, this information was not collected in Kenya. Instead, we captured duration of migration using the following question: “How long has [NAME] lived in his or her current location?”¹⁰ We coded a household as a migrant household if the household head reported at least one former household member who lived outside the household for work-related reasons and resided in his or her current location for five years or less. All other households were coded as nonmigrant households.

Our study goes beyond the dichotomy of migrant versus nonmigrant household by taking into account the educational selectivity of migrants living outside the households. We consider whether school enrollment is higher for children in migrant-sending households when migrants have higher levels of education than individuals of the same age and gender in their communities. To do so, we followed Ichou’s (2014) methodology to construct a migrant relative education score measuring the migrant’s location in the educational distribution of the population of the same age and gender. Specifically, this variable captures “the percentage of people of the same country of origin, gender, and age who have a lower level of educational attainment, plus half the percentage of the people with the same level of education” (Ichou 2014, 754). Feliciano and Lanuza (2017) rely on a similar measure they refer to as contextual attainment and that we refer to as relative education.

To create this measure, we first coded migrant’s educational attainment into the following categories: none, primary, secondary, or tertiary. We then located migrants’ educational attainment in the distribution of educational attainment of individuals who are of the same age and gender in the origin country using data from the country’s Demographic and Health Survey (DHS)¹¹ that took place closest in time to the Migration and Remittances Household Survey. We constructed a migrant’s relative education score by summing the percentage of the population of the same age and gender who have lower levels of education and adding half the percentage of the population with similar levels of education as the migrant. For example, in Kenya, the distribution of educational attainment among 20- to 24-year-old men is as follows: 3 percent have no education, 49 percent have primary education, 38 percent have secondary education, and 11 percent have tertiary education. Thus, the relative education score for a 23-year-old male migrant with secondary education is equal to 71 percent (3% + 49% + 0.5 × 38%), which indicates that the migrant’s educational attainment is higher than or the same as 71 percent of the population of the same age and gender. If a household has more than one labor migrant, we used the highest migrant relative education score.

We also considered the possibility that school enrollment will vary according to the migrant’s destination, specifically whether the migrant is an internal or international migrant, recognizing that migrant education and destination are also related. Finally, we constructed a variable measuring whether the household reported receiving remittances in the past year. Households that received remittances may have had more financial resources to invest in children’s schooling, which could have increased the likelihood that children were enrolled in school.

Control Variables

We include several control variables associated with children’s school enrollment: age, gender, urban residence, region/province, and number of children in the household. Because family and household resources including education, land, and income are all positively associated with children’s schooling in African countries, just as they are in other contexts (Buchmann 2000), and because these finite resources are shared among household members, we control for household wealth and number of children living in the household. To capture household economic resources, we used data collected on household assets to construct a household wealth index using principal components analysis (Filmer and Pritchett 1999). Household wealth is measured in quintiles and coded in the following manner: poorest, second, middle, fourth, and richest. We also include controls for characteristics of the household head because he or she has influence over whether a child attends school. Age of the household head is a proxy for family life cycle stage that is an important predictor of selection into migration (Durand and Massey 1992). The child’s relationship to the household head (child, grandchild, brother/sister, nephew/niece, other) could also be associated with school enrollment. Children with closer biological ties to the household head are more likely to be enrolled (Case, Paxson, and Ableidinger 2004). Finally, we included a measure capturing the educational attainment of adults in the sending household.¹² This variable measures the highest educational attainment (none, primary, secondary, tertiary) of an adult living in the household. In this way, we can assess the role of the migrant’s education net of the education levels of other adults in the child’s household.

Methods

We first consider whether having a recent migrant from the household is associated with a higher probability of school enrollment, controlling for the educational attainment of adults in the household with separate multivariate logistic regression models for each country. Our first set of models shows how household educational attainment and child- and household-level control variables predict current school enrollment. Next, we assess whether there is an additional role of migration by adding a variable measuring household migration status — no migration, internal migration, and international migration — to the model.

We then follow the example of Ichou (2014) and others to consider the importance of migrants’ own relative education net of household relative education within the subsample of households with recent migration (Ichou 2014; Feliciano and Lanuza 2017). These analyses do not account for unobserved factors that predict both migration and school enrollment beyond the educational attainment of adults in the household. Rather, with the cross-sectional data available, we can assess the importance of one observable source of that selectivity and the extent to which the differential patterns of migrant education selection across these three contexts are consistently associated with school enrollment net of the household’s underlying education levels. In the first set of models, we show how household educational attainment is related to current school enrollment, controlling for child- and household-level control variables. Next, we added migrants’ relative education score and remittances to our models.¹³ Finally, we added the type of migration (internal or international) to assess whether this measure mediates the relationship between migrant selectivity and current school enrollment. Additionally, in all models, we adjust standard errors to take into account the clustering of children within households because some households contribute more than one child to the analytic sample. We also tested for interaction effects between migration variables and children’s age and gender (Creighton and Park 2010; Acosta 2011; Antman 2012). Because none of the interaction terms were statistically significant across all three countries, we do not include them in the models presented in this paper.

In exploratory models, we took into account the survey’s sampling design by incorporating survey weights into our regression models. We obtained similar results to those presented in the current paper. Because Kenya did not include survey weights in their publicly released data set and survey results for Burkina Faso and Senegal do not vary with or without survey weights, we chose to maintain consistency across all three countries by presenting descriptive statistics and regression models without survey weights.¹⁴

Results

Descriptive Statistics

The characteristics of children in the analytic sample by country are presented in Table 1. In all three countries, the mean age of children is approximately 11 years, and girls make up slightly less than half the analytic sample. Approximately 50 percent of sampled children live in urban areas, except in Burkina Faso, where fewer than 5 percent are urban dwellers. In Burkina Faso, children are evenly distributed by household wealth, which is not the case in the other countries. A greater proportion of children live in poorer households in Kenya while the opposite is true in Senegal. The mean number of children living in a household varies greatly by country, ranging from an average of three children per household in Kenya to an average of seven in Burkina Faso and Senegal.

Table 1.

Characteristics of Children, Migration, and Remittances Household Surveys, 2009–2010.

	Burkina Faso	Kenya	Senegal
Mean age (years)	10.9	11.4	11.2
Female (%)	47.9	49.9	48.7
Lives in urban area (%)	3.5	43.2	53.9
Household wealth index (%)
Poorest	23.5	21.8	11.0
Second	17.4	26.5	19.1
Third	23.1	19.1	22.6
Fourth	18.3	18.1	24.3
Wealthiest	17.6	14.6	22.9
Mean number of children living in household	7.2	3.5	7.4
Mean age of household head (years)	51.6	46.9	54.3
Relationship to household head (%)
Child	76.4	75.0	55.9
Grandchild	7.6	15.7	17.1
Brother/sister	3.8	1.5	2.1
Nephew/niece	7.5	3.0	13.1
Other	4.7	4.8	11.9
Household migration status (%)
Nonmigrant	68.0	66.7	61.9
Internal migration	12.8	21.9	15.3
International migration	19.2	11.4	22.8
Household educational attainment (%)
None	53.7	7.1	36.5
Primary	24.2	29.1	23.9
Secondary	21.9	39.8	28.9
Tertiary	0.3	24.1	10.8
Total	6,621	2,182	5,220

We also present characteristics of the household head due to the important role he or she typically plays in the allocation of household resources and decisions regarding schooling in Sub-Saharan Africa. The household head’s mean age ranges from 47 years in Kenya to 54 years in Senegal. In all three countries, the sample is primarily composed of biological children, followed by grandchildren, of the household head.

Across all three countries, approximately one-third of the sample lives in migrant-sending households.¹⁵ In Burkina Faso and Senegal, the majority of these children live in international migrant-sending households. Household educational attainment, which captures the highest educational attainment of an adult household member, varies across the three countries; it is highest in Kenya, where 64 percent have secondary or tertiary education, and lowest in Burkina Faso, where only 22 percent have this level of education.

Our dependent variable, children’s school enrollment, also varies considerably across all three countries (not shown). School enrollment is lowest in Burkina Faso, where 42 percent of school-age children are in school full-time, and highest in Kenya, where 83 percent are in school. In Senegal, 61 percent of children are enrolled in school at the time of the survey. In all three countries, current school enrollment varies significantly by household migration status as well (Figure 2). In Burkina Faso and Senegal, children in international migrant-sending households have lower current school enrollment than those in nonmigrant households. No significant difference is observed in these two countries between children in internal migrant-sending households and nonmigrant households. In Kenya, we observe a different pattern: Current school enrollment is higher in internal and international migrant-sending households than in nonmigrant households.

Figure 2.

Children’s Current School Enrollment by Household Migration Status.

As expected, school enrollment also varies considerably according to the education levels of adults and migrants in these households. Current school enrollment is higher in households with greater levels of household educational attainment (not shown). A similar pattern is observed in migrant-sending households: Current school enrollment is higher in households where migrants have higher relative education scores (Figure 3). In Kenya, however, the proportion currently enrolled plateaus at 85 percent once the migrant’s relative education score reaches 40 or higher.

Figure 3.

Bivariate Association between Migrant’s Relative Education Score and Children’s Current School Enrollment.

School enrollment patterns in Kenya are quite different from those observed in Burkina Faso and Senegal. First, the overall levels of full-time school enrollment are higher. Second, school enrollment levels are highest for children in migrant-sending households. This suggests that selection into migration operates differently across the three countries and the possibility of a differential association between migration and children’s schooling, depending on the type of selection into migration that occurs.

The relationships depicted in Figures 2 and 3 suggest the importance of looking at the type and selectivity of migration when examining the association between migration and children’s schooling. In Table 2, we present characteristics of current labor migrants from migrant-sending households in our sample. In all three countries, migrants are, on average, 30 years of age and overwhelmingly male. Close to half of all migrants in Kenya and Senegal come from urban areas, while only 4 percent of migrants from Burkina Faso do so. The majority of migrants are children of the household head, with another sizable minority reported to be the household head’s sibling. Whereas approximately two-thirds of migrants in Kenya and Senegal sent remittances in the past year, less than half did so in Burkina Faso. Migrants’ destination also varies across the three countries. Most migrants in Kenya are internal migrants, likely engaged in rural-urban migration; in Burkina Faso, however, 60 percent are international migrants, almost exclusively working in another African country. Approximately half of all Senegalese migrants are working internationally and are evenly distributed between other African countries and countries outside Africa (not shown). In Kenya, 31 percent of migrants work internationally.

Table 2.

Characteristics of Current Labor Migrants by Country, Migration, and Remittances Household Surveys, 2009–2010.

	Burkina Faso	Kenya	Senegal
Mean age (years)	27.9	30.7	32.0
Male (%)	95.3	68.4	93.5
From urban area (%)	4.3	38.1	52.1
Relationship to household head (%)
Spouse/partner	3.1	15.9	7.2
Son/daughter	52.0	66.3	58.9
Father/mother	6.6	0.9	0.3
Brother/sister	30.7	10.6	17.7
Grandchild	5.6	2.3	NA
Other	2.1	3.9	15.9
Sent remittances in past year (%)	43.2	63.4	69.6
Destination (%)
Internal	40.2	69.1	50.6
International	59.8	30.9	49.4
Educational attainment (%)
None	75.6	0.7	55.4
Primary	15.4	24.6	22.5
Secondary	8.4	47.1	16.6
Tertiary	0.6	27.6	5.6
Mean relative education score	41.1	68.5	47.6
N	682	433	735

Note: NA = not applicable.

In addition to differences in the destinations of migrants by country, migrants’ educational selectivity varies across the three countries. In Burkina Faso and Senegal, 76 percent and 56 percent of migrants, respectively, had never attended school. Among migrants who ever attended school, most attended only primary school. We further examined migrant selectivity by comparing migrants’ educational attainment to that of individuals of the same age and gender (Figure 4). In Burkina Faso and Senegal, we observed that most migrants are negatively selected: 76 percent and 57 percent of migrants, respectively, had relative education scores that were below 50. The opposite pattern existed in Kenya, where 75 percent of migrants were positively selected.

Figure 4.

Distribution of Migrant’s Relative Education Score by Country.

Regression Analyses

We begin our multivariate analyses by examining whether household labor migration is associated with current school enrollment once we control for the other characteristics of sending households in each country (Table 3). Model 1 includes the household educational attainment of adults along with controls for child and household characteristics. Overall, higher household educational attainment is positively associated with children’s school enrollment in all three countries. Other characteristics matter as well. Child’s age is associated with school enrollment in a nonlinear fashion, as is consistent with the normative increase with entrance to school and attrition from school in adolescence. Girls in Burkina Faso were less likely to be enrolled than boys, but we do not observe this gender difference in Kenya or Senegal. More household wealth and urban residence were associated with school enrollment in Burkina Faso and Senegal. Children in urban areas in Kenya appeared less likely to be “full-time students” than those in nonurban areas. This likely reflects the sample design and the inclusion of children from very poor communities surrounding large urban areas. Finally, we note that in all three countries, not being a close relative of the household head (i.e., “other” relationship) is negatively associated with school enrollment.

Table 3.

Logistic Regression Models Predicting Children’s Current School Enrollment by Household Migration Status, Migration and Remittances Household Surveys, 2009–2010.

Variables	Burkina Faso		Kenya		Senegal
Variables	Model 1	Model 2	Model 1	Model 2	Model 1	Model 2
Household educational attainment (reference = “none”)
Primary	2.10***	2.10***	1.46	1.45	1.95***	1.94***
	(0.17)	(0.17)	(0.50)	(0.50)	(0.26)	(0.25)
Secondary	4.38***	4.41***	1.40	1.43	4.02***	3.97***
	(0.44)	(0.44)	(0.49)	(0.50)	(0.56)	(0.56)
Tertiary	3.07	3.06⁺	1.96⁺	2.03⁺	4.40***	4.40***
	(2.09)	(2.07)	(0.75)	(0.78)	(1.02)	(1.02)
Household migration status (reference = “nonmigrant”)
Internal		1.10		1.51*		1.26⁺
		(0.12)		(0.30)		(0.17)
International		0.88		1.65⁺		1.09
		(0.08)		(0.49)		(0.15)
Child’s age	2.94***	2.95***	2.81***	2.77***	2.58***	2.59***
	(0.20)	(0.20)	(0.39)	(0.38)	(0.21)	(0.21)
Child’s age squared	0.95***	0.95***	0.96***	0.96***	0.96***	0.96***
	(0.00)	(0.00)	(0.01)	(0.01)	(0.00)	(0.00)
Female child	0.83**	0.83**	0.97	0.96	0.98	0.98
	(0.05)	(0.05)	(0.12)	(0.12)	(0.07)	(0.07)
Urban residence	1.71*	1.71*	0.44***	0.44***	1.58***	1.60***
	(0.38)	(0.38)	(0.08)	(0.08)	(0.19)	(0.20)
Wealth index (reference = “poorest”)
Second	1.01	1.02	1.30	1.29	1.61**	1.58*
	(0.10)	(0.11)	(0.30)	(0.29)	(0.29)	(0.29)
Middle	1.19⁺	1.20⁺	1.35	1.30	2.34***	2.33***
	(0.12)	(0.12)	(0.36)	(0.35)	(0.45)	(0.45)
Fourth	1.26*	1.27*	1.68⁺	1.64	2.57***	2.60***
	(0.14)	(0.14)	(0.51)	(0.50)	(0.56)	(0.57)
Richest	1.61***	1.61***	1.27	1.22	3.39***	3.43***
	(0.19)	(0.19)	(0.42)	(0.41)	(0.81)	(0.82)
Number of children in household	0.96***	0.96***	1.07	1.06	0.98	0.98
	(0.01)	(0.01)	(0.05)	(0.05)	(0.01)	(0.01)
Household head’s age	1.00	1.00	1.00	1.00	0.99**	0.99**
	(0.00)	(0.00)	(0.01)	(0.01)	(0.00)	(0.00)
Child’s relationship to household head (reference = “child”)
Grandchild	1.01	1.02	0.97	0.92	1.14	1.13
	(0.13)	(0.13)	(0.25)	(0.24)	(0.16)	(0.16)
Brother/sister	0.65*	0.65*	1.17	1.18	0.76	0.77
	(0.11)	(0.11)	(0.82)	(0.85)	(0.20)	(0.20)
Nephew/niece	0.89	0.89	0.68	0.70	1.00	0.98
	(0.11)	(0.11)	(0.30)	(0.30)	(0.14)	(0.14)
Other	0.38***	0.39***	0.29***	0.29***	0.80	0.79⁺
	(0.07)	(0.08)	(0.08)	(0.09)	(0.12)	(0.11)
Constant	0.00***	0.00***	0.01***	0.01***	0.01***	0.01***
	(0.00)	(0.00)	(0.01)	(0.01)	(0.00)	(0.00)

N (children)	6,621	6,621	2,182	2,182	5,220	5,220
R ²	0.127	0.127	0.102	0.107	0.205	0.206

Note: Robust standard errors, clustered at household level to account for correlation between household members, are shown in parentheses. All models control for region/province.

⁺p < .10. *p < .05. **p < .01. ***p < .001.

Model 2 adds household migration status, which captures whether the household had engaged in recent migration and, if the household did, the type of migration (internal or international). In Burkina Faso and Senegal, little to no association is observed between household migration status and current school enrollment. In Kenya, in contrast, internal and international migration are both positively associated with schooling.

The results in Table 3 suggest that household educational attainment is indeed predictive of children’s school enrollment. Furthermore, in Burkina Faso and Senegal, there is very little difference in school enrollment by household migration status. For children in Kenya, though, being in a household with recent labor migration, both internal and international, is positively associated with school enrollment, even when controlling for household educational attainment. The next step is to consider whether a differential probability of school enrollment exists depending on whether the household sent a positively selected migrant.

To examine this question, the analyses in Table 4 are restricted to migrant-sending households. The first model establishes that the same factors that were predictive of school enrollment for children in all households (i.e., results in Table 3) are also predictive of school enrollment among children in migrant-sending households only. The results are consistent with the models for children in all households. Household educational attainment is positively associated with children’s school enrollment in Burkina Faso and Senegal, where overall levels of education are low. The age pattern of school enrollment is also consistent across all three countries, while being more distantly related to the household head (i.e., “other” relationship) is negatively associated with school enrollment in Burkina Faso and Kenya.

Table 4.

Logistic Regression Models Predicting Children’s Current School Enrollment in Migrant Households, Migration, and Remittances Household Surveys, 2009–2010.

Variables	Burkina Faso			Kenya			Senegal
Variables	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Household educational attainment (reference = “none”)
Primary	2.07***	1.92***	1.92***	2.01	1.89	1.86	2.66***	2.11***	2.11***
	(0.32)	(0.30)	(0.30)	(1.16)	(1.02)	(0.99)	(0.49)	(0.40)	(0.40)
Secondary	3.71***	3.41***	3.45***	1.78	1.42	1.40	3.62***	2.51***	2.51***
	(0.66)	(0.62)	(0.63)	(1.00)	(0.74)	(0.72)	(0.71)	(0.52)	(0.52)
Tertiary	0.88	0.71	0.70	2.90⁺	2.02	1.98	3.47**	2.50*	2.50*
	(1.56)	(1.13)	(1.09)	(1.82)	(1.28)	(1.23)	(1.49)	(1.00)	(1.01)
Migrant’s relative education score		1.01**	1.01**		1.02⁺	1.02⁺		1.02***	1.02***
		(0.00)	(0.00)		(0.01)	(0.01)		(0.00)	(0.00)
International migration (reference = “internal migration”)			0.86			0.90			0.99
			(0.12)			(0.35)			(0.16)
Received remittances in past year (reference = “no remittances in past year”)		1.04	1.05		1.57	1.59		1.04	1.04
		(0.13)	(0.13)		(0.57)	(0.58)		(0.18)	(0.18)
Child’s age	3.23***	3.26***	3.28***	3.96***	3.77***	3.76***	2.99***	3.09***	3.09***
	(0.41)	(0.41)	(0.41)	(1.23)	(1.21)	(1.21)	(0.39)	(0.40)	(0.40)
Child’s age squared	0.95***	0.95***	0.95***	0.94***	0.94***	0.94***	0.95***	0.95***	0.95***
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Female child	0.80*	0.80*	0.80*	1.49	1.51	1.52	1.00	1.03	1.03
	(0.08)	(0.08)	(0.08)	(0.38)	(0.39)	(0.40)	(0.12)	(0.12)	(0.12)
Urban residence	1.14	1.22	1.22	0.57	0.62	0.62	1.32	1.19	1.19
	(0.55)	(0.61)	(0.61)	(0.28)	(0.29)	(0.30)	(0.23)	(0.21)	(0.21)
Wealth index (reference = “poorest”)
Second	0.78	0.82	0.83	1.30	1.32	1.33	1.47	1.34	1.34
	(0.14)	(0.15)	(0.15)	(0.53)	(0.57)	(0.57)	(0.43)	(0.38)	(0.38)
Middle	1.07	1.07	1.08	1.14	1.03	1.05	2.10*	1.91*	1.91*
	(0.20)	(0.20)	(0.20)	(0.51)	(0.48)	(0.49)	(0.67)	(0.60)	(0.60)
Fourth	1.00	1.01	1.02	2.41⁺	1.98	2.02	2.27*	2.02*	2.03*
	(0.20)	(0.20)	(0.20)	(1.29)	(1.09)	(1.12)	(0.74)	(0.66)	(0.66)
Richest	1.74**	1.65*	1.67*	0.86	0.63	0.65	2.15*	1.71	1.71
	(0.36)	(0.34)	(0.35)	(0.56)	(0.44)	(0.45)	(0.78)	(0.62)	(0.63)
Number of children in household	0.96*	0.96*	0.96*	1.16⁺	1.22*	1.22*	0.96*	0.97⁺	0.97⁺
	(0.02)	(0.02)	(0.02)	(0.09)	(0.10)	(0.10)	(0.02)	(0.02)	(0.02)
Household head’s age	1.01	1.01	1.01	1.01	1.01	1.01	0.99	0.99	0.99
	(0.01)	(0.01)	(0.01)	(0.01)	(0.02)	(0.02)	(0.01)	(0.01)	(0.01)
Child’s relationship to household head (reference “child”)
Grandchild	0.86	0.90	0.91	0.66	0.59	0.59	1.27	1.15	1.15
	(0.16)	(0.17)	(0.17)	(0.31)	(0.28)	(0.28)	(0.29)	(0.26)	(0.26)
Brother/sister	1.06	1.09	1.12	0.23	0.16⁺	0.17	0.91	0.84	0.84
	(0.33)	(0.35)	(0.36)	(0.24)	(0.17)	(0.18)	(0.43)	(0.41)	(0.41)
Nephew/niece	1.06	1.05	1.06	NE	NE	NE	0.99	0.96	0.96
	(0.23)	(0.22)	(0.22)				(0.18)	(0.18)	(0.18)
Other	0.40**	0.40**	0.40**	0.17***	0.16***	0.16***	1.15	1.03	1.03
	(0.14)	(0.14)	(0.14)	(0.08)	(0.08)	(0.08)	(0.22)	(0.20)	(0.20)
Migrant’s relationship to household head (reference “spouse/partner”)
Child	0.58	0.60	0.59	1.15	1.46	1.42	1.02	1.00	1.00
	(0.21)	(0.21)	(0.21)	(0.55)	(0.70)	(0.69)	(0.32)	(0.31)	(0.31)
Parent	0.80	0.74	0.74	NE	NE	NE	0.63	0.23	0.23
	(0.36)	(0.33)	(0.34)				(0.57)	(0.21)	(0.21)
Brother/sister	0.61	0.61	0.59	1.86	2.38	2.34	0.91	0.92	0.92
	(0.22)	(0.22)	(0.22)	(1.06)	(1.63)	(1.62)	(0.29)	(0.29)	(0.29)
Grandchild	0.63	0.65	0.61	1.29	2.18	2.19	NA	NA	NA
	(0.30)	(0.30)	(0.29)	(1.09)	(1.96)	(1.95)
Other	0.55	0.57	0.56	0.95	1.16	1.18	1.38	1.40	1.39
	(0.25)	(0.27)	(0.26)	(0.72)	(0.78)	(0.79)	(0.45)	(0.46)	(0.46)
Constant	0.00***	0.00***	0.00***	0.00***	0.00***	0.00***	0.01***	0.00***	0.00***
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.01)	(0.00)	(0.00)

N (children)	2,092	2,092	2,092	680	680	680	1,934	1,934	1,934
R ²	0.132	0.137	0.138	0.147	0.164	0.164	0.204	0.225	0.225

Note: Robust standard errors, clustered at household level to account for correlation between household members, are shown in parentheses. All models control for region/province. NE = not estimated, perfectly collinear; NA = not applicable.

^⁺p < .10. *p < .05. **p < .01. ***p < .001.

Model 2 in Table 4 adds characteristics of migration, including the migrant’s relative education score and whether the household received remittances in the past year, to the baseline model. Here, the analyses suggest that having a migrant with a higher relative education score is indeed positively associated with children’s school enrollment even beyond the role of household education overall. Receipt of remittances is not statistically significant for predicting school enrollment in any of the three countries.

Model 3 tests whether type of migration (internal or international) helps explain the relationship between migrant selectivity and current school enrollment. Results indicate that the type of migration is not significantly associated with school enrollment for children in any of the three countries. Further, including the type of migration in the models does not reduce the size of the coefficients for migrant’s relative education. In sum, migrant’s relative education, net of the educational attainment of adults in the household, is an important predictor of children’s school enrollment that is not explained by the likelihood that positively selected migrants go to different destinations or are more likely to remit than those with lower relative education scores.

Discussion

Migration can be an important household strategy for maximizing economic opportunities and enhancing children’s well-being. There are many circumstances of migration, however, and they may not all be associated with the same outcomes. Previous research across diverse contexts yields very mixed conclusions about migration’s importance for enhancing children’s schooling in the origin household (Deb and Seck 2009; McKenzie and Rapoport 2011; Meyerhoefer and Chen 2011; Robles and Oropesa 2011; Yao and Treiman 2011). The analyses presented here build on this previous work by not only considering whether children’s schooling is associated with recent migration from the household but also examining the potential for migration to have a differential association depending on the educational selectivity of migrants within the household and across diverse contexts. The comparable household-level data collected in three Sub-Saharan African countries (Burkina Faso, Kenya, and Senegal) represent different geographical regions, migration patterns, and migrant selectivity. Burkina Faso offers a context where migrants have lower human capital and fewer resources flowing back to sending households when compared to Kenya and Senegal. Migrants from Senegal, in contrast, tend to be more positively selected on education and send more remittances. Kenyan migrants are the most positively selected, with similar proportions as their Senegalese counterparts sending remittances back to their households.

Overall, there is a weak association between living in a migrant-sending household and current school enrollment across these three different contexts in Sub-Saharan Africa. Rather, children living in households with greater levels of adult educational attainment have higher probabilities of school enrollment in Burkina Faso and Senegal regardless of whether the household sends migrants. Only in Kenya does a strong and positive association exist between household migration and school enrollment after controlling for both child and household characteristics. And here, both internal and international migration are similarly associated with school enrollment, suggesting that labor migration in general is helpful for keeping children in school in this context regardless of destination.

However, looking only at a dichotomous indicator of migration without considering migrant “quality” or relative education misses the more complex relationship between migration and children’s schooling. Our second research question thus addressed the role of migrant selectivity in the relationship between school enrollment and migration net of the educational attainment of adults in the household. In these analyses, restricted to children in migrant-sending households, migrants with higher levels of education relative to the origin country were positively associated with children’s school enrollment in the origin household. The results hold whether households send internal or international migrants.

These results provide some compelling insight for studies that have found very mixed support for migration’s role on children’s well-being. It may be that living in households with positively selected migrants invoked higher aspirations among children who perceive that education and migration are a successful combination (Feliciano and Lanuza 2017). Such expectations of improved lives may explain why the perception that migrants are successful enhances children’s outcomes regardless of the actual amount of economic remittances provided by the migrant (Yabiku, Agadjanian, and Cau 2012). At the other end of the relative education distribution, the success of labor migrants with low levels of education may provide an additional incentive for children to orient themselves toward becoming labor migrants rather than remaining in school (Kandel and Kao 2000; Fox et al. 2012). This seems particularly likely in Burkina Faso and Senegal, where so many migrants have little or no formal education. Children in these poorly resourced settings may also face higher demands for their own labor in the household, thus reducing children’s school enrollment where migrants’ relative education levels are also low (Amuedo-Dorantes, Georges, and Pozo 2010; Antman 2012).

In Kenya, however, where overall education levels are higher and more high-skilled migration is the norm, we find that migration overall is associated with children’s schooling. Children in migrant-sending households are more likely to be enrolled in school than their peers in nonmigrant households. In this case, it may be that observing the returns to education in the form of migration encourages attachment to schooling on the part of children left behind. In other words, our results suggest that migration can be positively associated with school enrollment but that this association is more likely in contexts where more positively selected migration is the norm or where households in poorly resourced settings can take advantage of sending a more educated migrant out. The results also have implications for concerns about “brain drain” where positively selected international migrants reduce the human capital by removing high-skilled migrants from the labor force. Our results suggest that when these migrants leave family members behind in the origin country, growth in human capital — through children’s schooling — continues.

Although our study takes advantage of comparable data in three understudied contexts of migration, the data also have features that limit our ability to draw firmer conclusions. First, the data are cross-sectional, preventing us from making stronger causal arguments about the relationship between migration and children’s schooling. Clearly, longitudinal data from such understudied settings would help further investigate these compelling questions (Binci and Giannelli forthcoming). Second, we could not identify the relationship of the child to the migrant. This is important because the type of relationship between migrant and child may be associated with the extent to which children benefit from migration. For example, a migrant’s biological child may benefit more from migration than his or her niece or nephew.

Third, we limited our migrant sample to labor migrants who were reported to be living in their current location for five years or less rather than those who left the household in the last five years. We did not use the latter definition because the survey did not collect this information in Kenya. We did, however, tabulate the number of labor migrants whose duration since migration was less than five years with labor migrants who had lived in their current location for the same duration. We found that most labor migrants whose duration since migration was less than five years were also coded as living in the current location for the same amount of time. We chose to focus on recent migrants to maintain proximity between the experience of sending labor migrants out and children’s school enrollment.

Fourth, our study focused solely on children reported to be regular household members at the time of the survey. Due to the lack of availability of schools, some children, especially those from rural areas, may have been sent to live with other family members, typically in towns, cities, or larger rural communities, to attend school. Our results cannot reflect the extent to which migrant households were more likely than nonmigrant households to send children to live with other family members to attend school. However, in Kenya, the survey included data on children living outside the household, so we could examine the relationship between migration variables and current school enrollment for children in and outside the household. The results were similar for both sets of children (not shown). We also cannot observe migration of all household members or cases where children migrate. Thus, our results are applicable to the case of children left behind in sending households.

Finally, our study found a positive association between migrant selectivity and children’s current school enrollment in migrant-sending households, but our regression models could not simultaneously control for migrant’s educational attainment. Fortunately, our analyses were able to control for overall household educational attainment, which captures some of the human capital available in children’s homes and is correlated with migrants’ education levels as well. We further attempted to disentangle the relationship by including migrants’ educational attainment and relative education score in the same model; however, high levels of collinearity affected the model’s fit and made the coefficient estimates unstable.¹⁶

Conclusion

Our study reveals the importance of taking into account the context and type of migration and going beyond the use of a dichotomous indicator of household migration status. We examined internal and international migration and children’s current school enrollment across three diverse sending contexts representing various regions of Sub-Saharan Africa and different streams of migration. Similar to previous studies, we obtained an inconsistent relationship between household migration status and children’s schooling. However, once we considered migrants’ educational selectivity, controlling for the educational attainment of adults in the household, we observed that migration was more likely to be positively associated with children’s schooling when migrants were more positively selected on education. These findings point to the need to consider the relative resources of sending households (including migrants’ education) to understand the association between migration and children’s schooling more generally.

Footnotes

Authors’ Note

Sophia Chae’s affiliation is included for informational purposes only. This work was not conducted under the auspices of the Population Council.

Acknowledgments

We appreciate the technical support provided by Robert Highfield and the helpful advice of the editors and anonymous reviewers.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Population Research Institute at The Pennsylvania State University, which is supported by an infrastructure grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD041025) and support from P01HD080659.

Notes

References

Acosta

2011. “School Attendance, Child Labour, and Remittances from International Migration in El Salvador.” Journal of Development Studies 47(6):913–36.

Adams

R. H.

2003. “International Migration, Remittances, and the Brain Drain: A Study of 24 Labor-Exporting Countries.” World Bank Policy Research Working Paper (3069), World Bank, Washington, DC.

Adams

R. H.

Jr Cuecuecha

Page

. 2008. “Remittances, Consumption, and Investment in Ghana.” World Bank Policy Research Working Paper Series, World Bank, Washington, DC.

Adepoju

2004. “Trends in International Migration in and from Africa.” In International Migration: Prospects and Policies in a Global Market, edited by Massey

D. S.

Taylor

J. E.

, 59–76. Oxford: Oxford University Press.

Agesa

Ri. U.

2004. “One Family, Two Households: Rural to Urban Migration in Kenya.” Review of Economics of the Household 2(2):161–78.

Amuedo-Dorantes

Georges

Pozo

. 2010. “Migration, Remittances, and Children’s Schooling in Haiti.” The ANNALS of the American Academy of Political and Social Science 630(1):224–44.

Antman

F. M.

2012. “Gender, Educational Attainment, and the Impact of Parental Migration on Children Left Behind.” Journal of Population Economics 25(4):1187–214.

Auriol

Demonsant

J.-L.

. 2012. “Education and Migration Choices in Hierarchical Societies: The Case of Matam, Senegal.” Regional Science and Urban Economics 42(5):875–89.

Beauchemin

Schoumaker

. 2005. “Migration to Cities in Burkina Faso: Does the Level of Development in Sending Areas Matter?” World Development 33(7):1129–52.

10.

Bigsten

1996. “The Circular Migration of Smallholders in Kenya.” Journal of African Economies 5(1):1–20.

11.

Binci

Giannelli

G. C.

. Forthcoming. “Internal Versus International Migration: Impacts of Remittances on Child Labor and Schooling in Vietnam.” International Migration Review. doi:10.1111/imre.12267

12.

Brockerhoff

. 1993. “Demographic and Socioeconomic Determinants of Female Rural to Urban Migration in Sub-Saharan Africa.” International Migration Review 27(3):557–77.

13.

Buchmann, 2000. “Family Structure, Parental Perceptions, and Child Labor in Kenya: What Factors Determine Who Is Enrolled in School?” Social Forces 78:1349–78.

14.

Case

Paxson

Ableidinger

. 2004. “Orphans in Africa: Parental Death, Poverty, and School Enrollment.” Demography 41(3):483–508.

15.

Chen

Huang

Rozelle

Shi

Zhang

. 2009. “Effect of Migration on Children’s Educational Performance in Rural China.” Comparative Economic Studies 51(3):323–43.

16.

Choe

Chrite

E. L.

. 2014. “Internal Migration of Blacks in South Africa: An Application of the Roy Model.” South African Journal of Economics 82:81–98.

17.

Cordell

D. D.

Gregory

J. W.

Piché

. 1996. Hoe and Wage: History of a Circular Migration System in West Africa. Boulder, CO: Westview Press.

18.

Creighton

Park

. 2010. “Closing the Gender Gap: Six Decades of Reform in Mexican Education.” Comparative Education Review 54(4):513–37.

19.

de Brauw

Mueller

Lee

H. L.

. 2014. “The Role of Rural-Urban Migration in the Structural Transformation of Sub-Saharan Africa.” World Development 63:33–42.

20.

de Hoop

Rosati

F. C.

. 2014. “Does Promoting School Attendance Reduce Child Labor? Evidence from Burkina Faso’s BRIGHT Project.” Economics of Education Review 39:78–96.

21.

De Vreyer

Gubert

Roubaud

. 2010. “Migration, Self-Selection and Returns to Education in the WAEMU.” Journal of African Economies 19(1):52–87.

22.

Deb

Seck

. 2009. Internal Migration, Selection Bias and Human Development: Evidence from Indonesia and Mexico. New York: United Nations Development Programme.

23.

Diatta

M. A.

Mbow

. 1999. “Releasing the Development Potential of Return Migration: The Case of Senegal.” International Migration 37(1):243–66.

24.

Durand

Massey

D. S

. 1992. “Mexican Migration to the United States: A Critical Review.” Latin American Research Review 27(2):3–42.

25.

Edwards

A. C.

Ureta

. 2003. “International Migration, Remittances, and Schooling: Evidence from El Salvador.” Journal of Development Economics 72(2):429–61.

26.

Feliciano

2005. “Educational Selectivity in U.S. Immigration: How Do Immigrants Compare to Those Left behind?” Demography 42(1):131–52.

27.

Feliciano

Lanuza

Y. R.

. 2017. “An Immigrant Paradox? Contextual Attainment and Intergenerational Educational Mobility.” American Sociological Review 82(1):211–41.

28.

Filmer

Pritchett

. 1999. “The Effect of Household Wealth on Educational Attainment: Evidence from 35 Countries.” Population and Development Review 25(1):85–120.

29.

Fox

Santibañez

Nguyen

André

. 2012. Education Reform in Mozambique: Lessons and Challenges. Washington, DC: World Bank Publications.

30.

Gerdes

2007. Focus Migration. Senegal. Hamburg: Hamburg Institute of International Economics.

31.

Halpern-Manners

2011. “The Effect of Family Member Migration on Education and Work among Nonmigrant Youth in Mexico.” Demography 48(1):73–99.

32.

2013. “Does Migration Benefit the Schooling of Children Left behind?: Evidence from Rural Northwest China.” Demographic Research 29:33–70.

33.

Ichou

2014. “Who They Were There: Immigrants’ Educational Selectivity and Their Children’s Educational Attainment.” European Sociological Review 30(6):750–56.

34.

International Organization for Migration. 2009. Migration Au Sénégal: Profil National 2009. Genève: International Organization for Migration.

35.

Kaestner

Malamud

. 2014. “Self-Selection and International Migration: New Evidence from Mexico.” Review of Economics and Statistics 96(1):78–91.

36.

Kandel

Kao

. 2000. “Shifting Orientations: How U.S. Labor Migration Affects Children’s Aspirations in Mexican Migrant Communities.” Social Science Quarterly 81(1):16–32.

37.

Konseiga

2007. “Household Migration Decisions as Survival Strategy: The Case of Burkina Faso.” Journal of African Economies 16(2):198–233.

38.

Massey

D. S.

Aysa

. 2005. “Social Capital and International Migration from Latin America.” Expert group meeting on international migration and development in Latin America and the Caribbean, United Nations Secretariat.

39.

McKenzie

D. J.

Mistiaen

. 2007. “Surveying Migrant Households: A Comparison of Census-Based, Snowball.” Policy Research Working Paper 4419, The World Bank, Washington, DC.

40.

McKenzie

Rapoport

. 2010. “Self-Selection Patterns in Mexico-US Migration: The Role of Migration Networks.” The Review of Economics and Statistics 92:811–21.

41.

McKenzie

Rapoport

. 2011. “Can Migration Reduce Educational Attainment? Evidence from Mexico.” Journal of Population Economics 24(4):1331–58.

42.

Meyerhoefer

C. D.

Chen

C. J.

. 2011. “The Effect of Parental Labor Migration on Children’s Educational Progress in Rural China.” Review of Economics of the Household 9(3):379–96.

43.

Miller

Elman

E.h

. 2013. “Improving the Quality of Education: Kenya’s Next Challenge.” Geography 98:24–32.

44.

Mora

Taylor

. 2005. "Determinants of Migration, Destination and Sector Choice: Disentangling Individual, Household and Community Effects.” In International Migration, Remittances and the Brain Drain, edited by Ozden

Schiff

, 21–52. Washington, DC: The World Bank.

45.

Niemeijer

Mazzucato

. 2002. “Soil Degradation in the West African Sahel: How Serious Is It?” Environment: Science and Policy for Sustainable Development 44(2):20–31.

46.

Nishimura

Yamano

. 2013. “Emerging Private Education in Africa: Determinants of School Choice in Rural Kenya.” World Development 43:266–75.

47.

Orrenius

P. M.

Zavodny

. 2005. "Self-Selection among Undocumented Immigrants from Mexico." Journal of Development Economics 78(1):215–40.

48.

Piotrowski

Paat

Y.-F.

. 2012. “Determinants of Educational Attainment in Rural Thailand: A Life Course Approach.” Population Research and Policy Review 31(6):907–34.

49.

Pison

Hill

Cohen

Foote

. 1997. “Les changements démographiques au Sénégal.” Travaux et documents de l'Institut national d'études démographiques (INED). Cahiers 138. Paris: INED.

50.

Plaza

Navarrete

Ratha

. 2011. “Migration and Remittances Household Surveys: Methodological Issues and New Findings from Sub-Saharan Africa.” Unpublished manuscript, Africa Migration Project, World Bank, Washington, DC.

51.

Ratha

. 2007. “Migration and Remittances in Senegal.” In Migration and Remittances Factbook, edited by World Bank Development Prospects Group. Washington, DC: The World Bank.

52.

Reed

H. E.

Andrzejewski

C. S.

White

M. J.

. 2010. “Men’s and Women’s Migration in Coastal Ghana: An Event History Analysis.” Demographic Research 22.

53.

République du Sénégal. 2003. Programme de développement de l’éducation et de la formation - Education pour tous. Dakar: Ministere de l'Education.

54.

Robles

V. F.

Oropesa

R. S.

. 2011. “International Migration and the Education of Children: Evidence from Lima, Peru.” Population Research and Policy Review 30(4):591–618.

55.

Robin

Lalou

Ndiaye

. 2000. Les Déterminants de l’émigration internationale au Sénégal. Sénégal: Eurostat-IRD-DPS.

56.

Somerset

2009. “Universalising Primary Education in Kenya: The Elusive Goal.” Comparative Education 45(2):233–50.

57.

Stark

1991. The Migration of Labor. Oxford: Blackwell.

58.

Syred

2011. “International Emigration: A Sub-Saharan Perspective.” African Policy Watch 7:1–8.

59.

Takenaka

Pren

K. A.

. 2010. “Determinants of Emigration: Comparing Migrants’ Selectivity from Peru and Mexico.” The ANNALS of the American Academy of Political and Social Science 630(1):178–93.

60.

Toma

Castagnone

. 2015. “What Drives onward Mobility within Europe? The Case of Senegalese Migrations between France, Italy and Spain.” Population 70(1):69–101.

61.

UNDP. 2015a. Human Development Report 2015: Work for Human Development. New York: UNDP.

62.

UNDP. 2015b. Sustainable Development Goals. New York: UNDP.

63.

UNESCO Institute for Statistics. 2013. Adult and Youth Literacy: National, Regional and Global Trends, 1985–2015. Montreal, Canada: UNESCO Institute for Statistics.

64.

UNICEF. 2014. State of the World’s Children 2014: Every Child Counts. New York: UNICEF.

65.

United Nations, Department of Economic and Social Affairs, and Population Division. 2014. World Urbanization Prospects: The 2014 Revision. New York: United Nations.

66.

Wouterse

Taylor

J. E.

. 2008. “Migration and Income Diversification: Evidence from Burkina Faso.” World Development 36(4):625–40.

67.

Yabiku

S. T.

Agadjanian

Cau

. 2012. “Labor Migration and Child Mortality in Mozambique.” Social Science & Medicine 75(12):2530–38.

68.

Yao

Treiman

D. J.

. 2011. “Migration, Remittances and Educational Stratification among Blacks in Apartheid and Post-Apartheid South Africa.” Social Forces 89(4):1119–43.