Migration and Parental Absence: A Comparative Assessment of Transnational Families in Latin America

Abstract

Despite growing research on transnational families and children’s welfare in migrant-sending countries, there is a dearth of information about the prevalence of, what we call, parental absence via migration, especially relative to other sources of parental absence, and a lack of estimates that are comparable across populations and places. This makes it difficult to determine the scale for policy interventions, and to justify future research on transnational families and children’s welfare. Using harmonized surveys covering eight Latin American countries and Puerto Rico, validated by nationally representative census and survey data, we provide the first estimates of the prevalence parental absence via migration that are comparable across populations in contemporary Latin America. We show that between 7 and 21 percent of children live in transnational families where parents are absent because of migration. We compare our estimates to similar estimates of parental absence from other sources, and show that, in several populations, more children are experiencing parental absence due to migration than to union dissolution or parental mortality. Finally, we link our descriptive work to children’s welfare by examining the characteristics of children’s home environments when parents migrate. Children living in families with absent parents due to migration are less likely to coreside with extended family members, and to fare better in terms of household assets, relative to children living in other family forms. We conclude by highlighting the limitations of the data, and underscore the value of attempts to estimate the prevalence of parental absence via migration.

INTRODUCTION

In 2010, the theme of the International Day of Families was “The Impact of Migration on Families around the World” (United Nations, 1993, 2010).¹ Presently, about one billion persons worldwide live outside of their country or, within countries, their region of birth (Démurger, 2015:1), meaning that many families are split up and family members—e.g., children, spouses, and parents (Antman, 2013)—”left behind.” The study of transitional families involving children is particularly important because parental absence has potentially negative implications for children’s welfare in migrant-sending areas (Cortes, 2008; D’Emilo et al., 2007; Mazzucato & Schans, 2011).

Recognizing that there are different sources of parental absence (e.g., migration, union dissolution, and parental mortality), what we refer to as parental absence via migration can generate instability in children’s living arrangements by altering the quantity and quality of material and non-material resources, as well as the roles and responsibilities of remaining family members (Antman, 2013; Assaad, 2010; Dreby, 2010; Falicov, 2007). Given that strategies to offset the potential costs of parental absence via migration (e.g., additional investments in children via migrant remittances) may be ineffective or absent, instability in children’s living arrangements may, in turn, have negative implications for their welfare in such areas as education and health (Démurger, 2015). In light of these important dynamics, the large and growing body of research on the relationship between parental absence via migration and children’s welfare in migrant-sending areas is not surprising (Mazzucato & Schans, 2011).

What is surprising is the dearth of information on the prevalence of parental absence via migration and, with few exceptions (Mazzucato et al., 2015), the virtual absence of cross-nationally comparable estimates. Portes (2001:181; see also Portes et al., 1999) was the first to highlight this problem in studies of transnationalism more generally, and emphasized the need to “clarify the actual scope of the phenomenon.” Cortes (2008:11) later cited problems of data availability, quality, and cross-national comparability that make it “almost impossible” to determine the “magnitude” of parental absence via migration. And, in their review of research on transnational families and children’s welfare, Mazzucato & Schans (2011:704-705) remarked that the “exact prevalence of transnational family arrangements is unknown…because of a scarcity of quantitative evidence caused by the lack of academic and policy attention to this phenomenon.” Prevalence estimates of parental absence via migration are ultimately needed to establish the generalizability of the phenomenon to “justify further research on transnational families and the well-being of children,” as well as to determine the appropriate scope of policy interventions (ibid.; see also Démurger, 2015; Eremenko & González-Ferrer, this issue; Reyes, 2007).

Accordingly, following similar efforts by Mazzucato et al. (2015) in the context of migration from Africa to Europe, we go beyond conventional “back of-the-envelope calculations” (Bryant, 2005:2) to provide the first prevalence estimates of parental absence via migration that are comparable across populations in contemporary Latin America. Using harmonized surveys covering eight Latin American countries and Puerto Rico, validated by nationally representative census and survey data, we begin by situating parental absence via migration at the intersection of migration and family-building processes. We estimate the proportion of migrants who are parents and, thus, physically absent from children’s households. Taking a child-centric perspective, we then estimate the proportion of children with absent migrant parents. Finally, linking our descriptive work on the prevalence of parental absence via migration to children’s welfare, we examine children’s living arrangements and household assets in households with absent migrant parents.

Importantly, acknowledging the dearth of comparative research in this area (Caarls et al. this issue), our work is comparative in two senses. First, it is comparative across populations in Latin America and the Caribbean. Second, following prior research (Creighton et al., 2009; Nobles, 2013), we compare our estimates of parental absence via migration to other sources of parental absence—namely, union dissolution and parental mortality. Doing so serves to benchmark the magnitude of parental absence via migration as a demographic process that influences children’s living arrangements. Taken together, our work provides a solidly comparative and long-overdue point of reference for research on parental absence via migration and the welfare of children left behind in migrant-sending areas, a topic of strong interest to those in both scholarly and policy circles.

BACKGROUND

Parental Absence via Migration and Children’s Welfare in Migrant-Sending Areas

Parental absence via migration has a number of potential implications for resource and responsibility sharing among household members left behind (Antman, 2013; Cortes, 2008; D’Emilo et al., 2007; Démurger, 2015; Mazzucato & Schans, 2011). Many important economic and social aims are achieved or, at least, are strongly influenced by the presence of both parents in the household. One well-studied aspect of household production is the link between coresidence with two parents and intergenerational transfers, including investments made in childrearing (McLanahan & Sandefur, 1994). A large body of research documents the importance of coresidence with two parents for children’s welfare (Fomby & Cherlin, 2007; Booth & Amato, 2001; Seltzer, 2000). To date, these studies unambiguously demonstrate that significant benefits in education, health, and social skills accrue to children who reside with two parents for the duration of childhood (Carlson, 2006).

These insights have informed a large and growing body of transnational research on the relationship between parental absence via migration and children’s welfare in migrant-sending countries and areas therein in Asia (Arguillas & Williams, 2010; Caarls et al., this issue; Gao et al., 2010; Graham & Jordan, 2011; Jordan et al., this issue; Wu and Cebotari, this issue), Africa (Jordan et al., this issue; Lu & Treiman, 2011; Madhaven et al., 2008; Wu and Cebotari, this issue; Yabiku et al., 2012), Europe (Giannelli & Mangiavacchi, 2010), and Latin America (Creighton et al., 2009; Edwards & Ureta, 2003; Halpern-Manners, 2011; Jordan et al., this issue; Moran-Taylor, 2008; Nobles, 2013, 2011; Schmeer, 2009; Smith et al., 2004). Presently, there is strong consensus that parental absence via migration generates instability in children’s living arrangements by altering the presence and composition of authority figures and caregivers, the consistency of care, rules and expectations that govern behavior, and the availability and distribution of material resources (Antman, 2013; Dreby, 2010; Falicov, 2007).

However, there is less agreement on whether and to what extent these changes negatively impact children’s welfare (Cortes, 2008; Démurger, 2015; Mazzucato & Schans, 2011). While some studies suggest that penalties in child schooling and health outcomes can be attributed to parental absence via migration (Creighton et al., 2009; Donato & Duncan, 2011; Heymann et al., 2009; McKenzie, 2005; Nobles, 2011), the link between instability in children’s living arrangements and children’s welfare is ambiguous because it depends on the strategies employed (or not) by families to mitigate the potential costs of parental absence via migration. For example, studies show that migrant remittances (see Arias, 2013; Goldring, 2004; Sana, 2008; Villareal & Shin, 2008) can soften the destabilizing influences of parental absence via migration on children’s welfare, including keeping families out of poverty (Adams, 2008; Arguillas & Williams, 2010; Chant, 2006; Edwards & Ureta, 2003; Goldring, 2004; Kanaiaupuni & Donato, 1999; Lu & Treiman, 2011; Villarreal & Shin, 2008). Other studies have documented the importance of alternative living and caregiving arrangements, including coresidence with extended family (Creighton et al., 2009; De Vos, 1995; Madhaven, 2008; Mier y Terán et al., 2001; Moran-Taylor, 2008). Still other studies have emphasized alternative forms of involvement and caring provided by absent migrant parents themselves, as regular communication can ease the psychological costs of parental departure and maintain absent parents’ roles as authority figures, which, in turn, may reduce behavioral problems and improve schooling outcomes (Dreby, 2010; Jordan et al., this issue; Kanaiaupuni, 2000; Nobles, 2011).

The Prevalence of Parental Absence via Migration and Need for Comparisons

In different ways, the above studies highlight the many nuances of parental absence via migration, including pertinent limiting conditions and intermediary mechanisms that affect whether and to what extent it has negative implications for children’s welfare. It is therefore surprising that we know very little about the prevalence of parental absence via migration, and even less in a comparative perspective (Caarls et al., this issue; Mazzucato et al., 2015). Estimates of the prevalence of parental absence via migration are important for justifying continued research on transnational families (Démurger, 2015; Mazzucato & Schans, 2011). They are also important for determining the scale at which policy interventions should be developed, coordinated, and implemented (Cortes, 2008; D’Emilo et al., 2007; Eremenko and González-Ferrer, this issue; United Nations, 2014), which, as Reyes (2007) noted, can range from highly targeted individual and family counseling to large-scale multi-partner outreach and public awareness campaigns within and across countries.

Presently, definitions and available estimates of parental absence via migration vary widely across countries and data sources (Cortes, 2008). For example, using data from the 1996 Matlab Health and Socioeconomic Survey, Kuhn (2006) estimated that approximately 10% percent of children in Matlab, a rural region in Bangladesh, were living apart from a migrant father. Bryant (2005) used census data to estimate that 10-20 percent of Filipino children and 2-3 percent of children in Indonesia live apart from at least one migrant parent. Data from the Mexican Family Life Survey indicate that about 7 percent of children have absent migrant fathers living in the United States or elsewhere in Mexico (Creighton et al., 2009; Nobles, 2011). Nobles (2013) further showed that about 17 percent of children in Mexico could expect to live in households with absent migrant fathers over the course of their childhoods.

A central difficulty that arises in comparing estimates of parental absence via migration across countries is the fact that they are particularly sensitive to the data source(s) used (Cortes, 2008). There are well-documented problems with the availability, quality, and cross-national comparability of publicly available migration data, even for countries with arguably the best data collection, processing, and reporting infrastructures (for extended discussions of these issues, see Levine et al. 1986; Bilsborrow, 1997; Poulain et al., 2006; Waters & Pineau, 2015; Willekens et al., 2016; Zlotnik, 1987). Consequently, studies of parental absence via migration are routinely limited to single countries because, as Cortes (2008:11) lamented, “international comparisons are almost impossible.”

Despite these difficulties and limitations, comparable estimates of the prevalence of parental absence via migration across countries and populations remain an important goal given their potential for scholars and policymakers to consider whether and how differences in cultures of migration (Ali, 2007; Horváth, 2008; Kandel & Massey, 2002), family-building processes (Falicov, 2007), and their intersection(s) structure parental absence via migration over the life course and across generations in ways that are consequential for children’s welfare in both the short- and long-run (Eremenko and Bennett this issue). Presently, with few exceptions (Mazzucato et al., 2015), these processes have largely been examined within countries (Démurger, 2015:8). The importance of “culturally contextualized understandings” notwithstanding (Graham & Jordan, 2011:784), a clearer understanding of parental absence via migration prevalence, and of transnational families more generally, in comparative perspective is needed (Caarls et al., this issue; Mazzucato & Schans, 2011; Portes, 1999, 2001).

Migration and Family-Building Processes: The Case of Latin America

Complementing ongoing efforts to generate comparative assessment of transnational families in the European and African context (Mazzucato et al., 2015; Caarls et al., this issue) the current study takes as its case parental absence via migration in eight Latin American countries and Puerto Rico. Although Mexico-U.S. migration tracks back well over 100 years, the transformation of Latin America into a region of large-scale out-migration is a fairly recent phenomenon, and is related, in part, to shifts in U.S. immigration policy.² As World War II began, the United States recruited large numbers of unskilled agricultural workers from Mexico via Bracero Programs between 1942 and 1964. Together with the abolition of national origin quotas and other amendments to the Immigration and Nationality Act passed in 1965, legal and unauthorized U.S. immigration surged. As a consequence, U.S. policies governing immigration became increasingly symbolic political displays with many “unintended consequences and unanticipated policy feedbacks” (Massey & Pren, 2012:28; see also Gonzalez-Barrera, 2015). For example, U.S. political and military interventions in Latin America were accompanied by policies that accommodated those seeking refuge from communist regimes in Cuba and Nicaragua in 1959 and 1979, respectively (Alvarado & Massey, 2010). In the 1960s, the U.S. government sought to defuse political pressures in the Dominican Republic by giving out permanent residency cards to as many Dominicans as would accept them (Riosmena, 2010). And, more recently, restrictive U.S. immigration policies that first emerged in the late 1980s have led to an unprecedented rise in border enforcement and policing (Andreas, 2009), a trend that continues to the present day as policymakers scapegoat immigrants for a host of problems including budget shortfalls and terrorism (Massey et al., 2016; see also Faist, 2002).

In countries in Latin America, economic policies and demographic pressures further helped to fuel migration. Combined with fledgling and, in some cases, collapsing economies, total fertility rates of just under six children per woman in the 1960s put substantial pressure on households already beset by inequities stemming from import substitution policies, the privatization of agricultural lands, and generally limited economic opportunities (Aide & Grau, 2004; Durand & Massey, 2010; Pellegrino, 2000). As a result, migration became a necessary and increasingly viable household risk reduction strategy. In the 1960s and 1970s, Latin America lost approximately 5-6 million persons to out-migration worldwide. By the 1980s, this figure increased to about 8 million persons, and climbed to nearly13 million persons by the early 2000s, for a total net loss of 11 million persons during the decade (Abel, 2013, 2015; Abel & Sander, 2014).

According to Durand and Massey (2010), these emigrations took three characteristic forms. The first, and perhaps most accessible to households struggling to make ends meet, was intraregional migration (Portes & Grosfoguel, 1994). Costa Rica, for example, became an important migrant destination for coffee harvesters (Rosero-Bixby et al., 2002). Large-scale indigenous and ethnic migrations were also common (Reid, 2007). Many movements were directed to urban areas, where employment opportunities were more plentiful than in rural areas. That said, patterns of rural-to-urban migration varied substantially across populations. By the early 2000s, three-fourths of the populations in Colombia, Peru, and Mexico were urban-dwellers. In contrast, less than one-half of the populations in Costa Rica, El Salvador, and Guatemala lived in urban areas (Cerrutti & Bertoncello, 2003). In addition to intraregional migration, two other characteristic migration forms included South-North migration and transoceanic migration (Brick et al., 2011; Massey and Espinosa, 1997; Massey et al., 1994; Pellegrino, 2004), with Abel and Sander (2014:1522) recently noting that “[m]igrants from Latin America move almost exclusively to North America and Southern Europe.”

An important feature of these migration patterns was and continues to be their age composition (García & de Oliveira, 2011). Though migration flows typically occur sequentially and cumulate over time, i.e., an initial onset followed by a continuation through family and social networks (Massey & España, 1987; Massey et al., 1998), these movements also unfold with strong age regularities (Rogers & Castro, 1981). Migration typically occurs in young- to mid-adulthood (Durand et al., 2001), and in the context of family-building (Brick et al., 2011; Lindstrom & Saucedo, 2007). In their analysis of moderate- to high-migration communities in Costa Rica, the Dominican Republic, Guatemala, Mexico, and Nicaragua, Lindstrom and Ramírez (2010) showed that more than one-half of pioneer U.S. migrants were married at the time of departure. Importantly, the average age at first departure ranged from 26 to 31 years, well into their childbearing years. Migrants from urban Nicaragua were the lone exception, with an average age of departure of 22 years.

In addition to men’s migration, women have long played an important role in Latin American migration (Donato, 1993, 2010; Donato & Gabaccia, 2015; Donato et al., 2011; Hondagneu-Sotelo, 2003; Pedraza, 1991). Both domestic work and factory labor in urban centers employ large shares of migrant women from rural areas (Durand & Massey, 2010). Women currently make up about 50 percent of international migrants in the region (Donato, 2010; Donato & Gabaccia, 2015; Donato et al., 2011), though this proportion varies considerably across populations. This heterogeneity can be explained, in part, by the fact that some countries, e.g., the Dominican Republic and Puerto Rico, are matrifocal societies with higher levels of divorce, female household headship, and female migration than other countries, e.g., Costa Rica, Mexico, and Guatemala (Massey & Sana, 2003). As a proportion of total migration flows, female out-migration from the Dominican Republic is therefore much larger than from other countries (Donato, 2010). And, compared to men, Dominican women are also more likely to initiate migration outside of unions (Massey et al., 2006).

It is at the intersection of migration and family-building processes that the current paper is situated. Parental absence via migration has a number of important and wide-ranging implications for the region, including, for example, the structure and stability of children’s home environments (Dreby, 2010; Kanaiaupuni, 2000; Suárez-Orozco et al., 2002), economic transfers to and within households in migrant-sending areas (Abrego, 2009; Funkhauser, 1995; Kanaiaupuni & Donato, 1999; Massey & Sana, 2005; Sana, 2008), and transgenerational migration patterns (Durand & Massey, 2010:43), whereby children reach adulthood with the migration-specific capital that strongly influences their own orientation toward migration (Donato & Sisk, 2015; Dreby, 2006; Kandel & Massey, 2002). However, with the exception of the Mexican context (Jordan et al., this issue; Lindstrom & Saucedo, 2007; Nobles, 2013; Suro, 2005; White & Potter, 2013), the intersection of migration and family-building processes is not well-documented. The aim of this paper is therefore to provide the first estimates of the prevalence of parental absence via migration that are comparable across populations in contemporary Latin America, followed by an examination of some of the linkages to children’s welfare.

APPROACH

Data and External Validation

This paper relies on two primary data sources. The Latin American Migration Project and the Mexican Migration Project provide data for eight countries and the territory of Puerto Rico. The Latin American Migration Project began in 1998, and the Mexican Migration Project began in the early 1980s. Both the Latin American Migration Project and Mexican Migration Project are on-going data collection efforts that use an ethnosurvey methodology to collect comparable data from households in countries in Latin America, the Caribbean, and Puerto Rico. The Latin American Migration Project and Mexican Migration Project surveys contain detailed information on the migration experiences of household members, as well as on the social and economic characteristics of households. Because the Mexican Migration Project has a longer history than the Latin American Migration Project, we limit the Mexican Migration Project sample to communities surveyed between 1998 and 2013, the years when communities in the Latin American Migration Project were surveyed. As we show in Table 1, sample sizes range from several hundred households in El Salvador and Puerto Rico to several thousand households in Mexico. Accordingly, descriptive estimates for Mexico are, on average, more precise than those calculated for other countries and Puerto Rico.

Table 1.

Latin American Migration Project and Mexican Migration Project sample descriptions

Country/Territory	Year(s) Collected	Households Surveyed	Children (Age <18)
Colombia	2008-2009	1,400	614
Costa Rica	2000-2002	1,391	1,904
Dominican Republic	1999-2000	907	1,426
El Salvador	2007	382	600
Guatemala	2004	513	896
Mexico	1998-2009	32,000	17,061
Nicaragua	2000-2003	1,598	2,986
Peru	2000-2005	826	884
Puerto Rico	1998	385	426

Notes: Only selected years of data collection from the Mexican Migration Project are shown for Mexico.

Sources: Latin American Migration Project and Mexican Migration Project.

The Latin American Migration Project and Mexican Migration Project surveys present two challenges to our efforts to estimate the prevalence of parental absence via migration. First, although most children—ranging from 66 percent in Nicaragua to 90 percent in Mexico—are children of the household head, the Latin American Migration Project and Mexican Migration Project surveys lack unique identifiers to link the remaining children to parents. We therefore imputed parent-child links using an algorithm that searches for adults whose age-difference with the child is closest to 27 years, which is the mean age of maternity in most countries, including those in Latin America and the Caribbean (United Nations, 2015). This mean falls within the mean age of maternity values for the countries and Puerto Rico in this study, and ranges from 25.8 in the Dominican Republic to 28.0 in Guatemala.

As a further check on this assignment, we used this algorithm to assign parent links to children in the Mexican Family Life Survey, a nationally representative survey with parent link identifiers for all children in the household. The algorithm correctly assigned mothers for 95 percent of children and fathers for 93 percent of children, where a correct match includes assigning no parent when the parent actually lives outside of the household. When limited to children whose mothers or fathers are in the household, the algorithm correctly assigned mothers for 96 percent of children and fathers for 98 percent of children.

Second, the Latin American Migration Project and Mexican Migration Project surveys are not necessarily representative of national populations. In each country in these surveys, communities are stratified by population density, and purposively sampled to achieve sample variation on density. Households are then probabilistically selected into the sample. When weighted—the estimates described below are weighted by the sample-specific weight provided in the Latin American Migration Project and Mexican Migration Project data, which is the inverse of the sampling fraction—the resulting data are representative of the sampled communities (Pren, 2012). Accordingly, we sought to validate the Latin American Migration Project and Mexican Migration Project data with nationally-representative census and survey data from IPUMS-International (Minnesota Population Center, 2015) and the Demographic and Health Surveys. Together, IPUMS-International and Demographic and Health Surveys contain information for each country and Puerto Rico on children of household heads living apart from their parents for reasons other than migration, defined as the second parent’s absence from the household member roster, which we also calculate from the Latin American Migration Project and Mexican Migration Project data. These absences most likely occur following divorce, parental death, or nonmarital childbearing. IPUMS-International data are available for seven countries and Puerto Rico close to the time of the Latin American Migration Project and Mexican Migration Project data collection, and Demographic and Health Surveys data are available for three countries. These estimates are presented in Table 2.

Table 2.

Percent of children of household head living apart from parent(s) for reasons other than migration

	Latin American Migration Project & Mexican Migration Project Surveys		IPUMS-International		Demographic and Health Surveys
Country/Territory	Year(s) Collected	Percent	Year Collected	Percent	Year(s) Collected	Percent
Colombia	2008-2009	31	2005	21
Costa Rica	2000-2002	15	2000	16
Dominican Republic	1999-2000	22	2002	21	2002	13
El Salvador	2007	29	2007	24
Guatemala	2004	8			1998-1999	7
Mexico	1998-2009	8	2000	12
Nicaragua	2000-2003	21	2005	18	2001	14
Peru	2000-2005	12	2007	20
Puerto Rico	1998	37	2000	27

Notes: Migration refers to domestic and international migration.

Sources: Authors’ calculations using data from Latin American Migration Project, Mexican Migration Project, IPUMS-International, and Demographic and Health Surveys.

According to Latin American Migration Project data for Costa Rica, 15 percent of children of household heads lived apart from at least one parent who had been absent for long enough to not be listed in the household roster. The corresponding estimate from IPUMS-International data is 16 percent. Similarly comparable estimates are shown for Guatemala and Nicaragua. Latin American Migration Project data for Colombia and Puerto Rico capture families with higher rates of parental absence relative to the IPUMS-International and Demographic and Health Surveys, respectively; and the Latin American Migration Project and Mexican Migration Project data for Mexico and Peru capture families with lower rates of parental absence. The largest discrepancies are observed for Colombia, Peru, and Puerto Rico. We revisit these discrepancies in the final section of this paper.

Analyses

Our analysis is designed to characterize several important aspects of parental absence via migration. The first is the selectivity of recent migration on parenthood and on child-tied movements. To characterize these movements, we analyze the composition of migrants departing five years prior to the survey date. Narrowing the time period allows us to characterize parenthood among migrants during comparable periods. It also reduces the potential bias introduced when the histories of entire households are lost following completed stage migration. Because the Latin American Migration Project and Mexican Migration Project data contain proxy responses from adults about the departure histories of household members who are migrating at the time of the survey, the estimates are not limited to a characterization of returned migrants. This is an unusual strength of the Mexican Migration Project and Latin American Migration Project data, and one that is essential for generating prevalence estimates. We use migration data for all adults aged 18 and older at the time of the surveys to calculate the proportion of adults who migrated in the five years prior to the survey date. We use the cut-off of 18 years to distinguish adults from children because it signifies legal adulthood in many of the populations studied here. All adults who were migrants at the time of the survey and reported to be residents of surveyed households are included in our estimates.³ Unless otherwise stated, the estimates refer to both international and internal migration, although we distinguish between the two types at several points in our analysis. We then match children to parents in households using information about household members’ relationship to the head and the aforementioned imputed parent-child links to calculate the proportion of recent adult migrants with a child in the home at the time of departure.

We then focus on the scope of parental absence via migration from a child-centric perspective. Using the same parent-child links, we calculate prevalence estimates of parental absence via migration from the perspective of children. Specifically, we calculate the proportion of children who have a migrant parent living away from the household at the time of the survey. Intuitively, these estimates are linked to those assessing selectivity on parenthood among migrants; however, a child-centric perspective will deviate from the above estimates if fertility patterns differ between migrants and non-migrants. After disaggregating these estimates by migration (domestic versus international migration) and parent (mothers’ versus fathers’ migration) type, following prior research (Creighton et al., 2009; Nobles, 2013), we compare our estimates of parental absence via migration to parental absence arising from union dissolution and parental mortality. Doing so serves to benchmark the magnitude of parental absence via migration to other, more commonly studied sources of parental absence. We estimate these sources of parental absence using data on the status of the partners of children’s living parent. If she/he is widowed, then the child is classified as having a deceased parent. If she/he has experienced a union dissolution after the birth of the child, then the child is classified as living apart from a parent for this reason.⁴ Union dissolution is defined by the remaining parent’s stated marital status at the time of the survey. Critically, some parents absent following union dissolution and not defined as household members may also be migrants. To the extent that this happens, our estimates will be conservative estimates of parental absence via migration.

Finally, in an effort to link our descriptive work on the prevalence of parental absence via migration to children’s welfare, like Caarls et al. (this issue), we examine descriptive traits of children’s home environments when parents are migrating. Scholarship on parental absence via migration underscores several mechanisms that may be protective for children in migrant-sending households, including the presence of extended family to care for children and increases in household consumption made possible by remittances (Antman, 2013; Démurger, 2015). Using the Latin American Migration Project and Mexican Migration Project data, we measure extended family coresidence and material welfare in households with migrant parents, and then test for differences in these measures across other types of living arrangements for children. This description does not reflect a causal outcome of parental migration; rather, it identifies whether children whose parent(s) are absent due to migration are living in different conditions than children whose parents are absent due to union dissolution and parental mortality. Our first measure of household characteristics is a dichotomous variable indicating that the child resides with an extended family member, measured as an adult family member that is not the child’s parent. Across all countries and Puerto Rico, between 37.8 percent of children in Puerto Rico and 70.6 percent of children in Nicaragua lived with an extended family member. The second measure is of household assets using a standard asset index (Filmer & Pritchett, 2001; see also McKenzie, 2005, Bredl, 2011, Sanchez-Soto, 2009). We generated a factor score from questions in the Latin American Migration Project and Mexican Migration Project about household ownership using a set of comparable items across countries and Puerto Rico, including residential property, housing materials, housing size, electricity, running water, radios, and televisions (see Supplemental Table A1). Factor scores for each household are mean centered at the household level within each country and Puerto Rico. Sample mean values range from −.37 (S.D.=1.76) in El Salvador to .41 (S.D.=1.71) in Mexico. Deviation of mean values from zero arises because our samples select on households with children under the age of 18.

Table A1.

Asset index factor loadings

	Item
Country/Territory	1	2	3	4	5	6	7	8	9	10	11	12	13	14
Colombia	0.11	0.21	0.27	0.42	0.57	0.49	0.46	0.55	0.54	0.49	0.19	0.46	0.45	0.26
Costa Rica	0.22	0.34	0.24	0.37	0.43	0.38	0.46	0.56	0.41	0.48	0.03	0.37	0.30	0.29
Dominican Republic	0.10	0.36	0.14	0.26	0.18	0.29	0.43	0.66	0.57	0.56	0.36	0.53	0.45	0.37
El Salvador	0.35	0.71	0.35	0.53	0.62	0.52	0.37	0.63	0.27	0.47	0.10	0.55	0.39	0.35
Guatemala	0.23	0.44	0.05	0.53	0.24	0.25	0.18	0.69	0.51	0.63	0.10	0.39	0.63	0.54
Mexico	0.11	0.49	0.08	0.51	0.36	0.32	0.52	0.70	0.62	0.52	0.30	0.53	0.56	0.48
Nicaragua	0.05	0.43	0.05	0.41	0.34	0.34	0.60	0.70	0.35	0.65	0.24	0.48	0.53	0.46
Peru	0.07	0.55	0.07	0.50	0.30	0.54	0.54	0.70	0.57	0.70	0.11	0.27	0.59	0.40
Puerto Rico	0.08	naa	0.17	0.14	0.66	0.37	0.77	0.79	0.35	0.29	0.18	0.36	0.24	0.19

Notes: Items include: (1) Own residential property, (2) Cement, finished, or wood floor, (3) Tile or cement roof, (4) Number of rooms, (5) Electricity, (6) Running water, (7) Stove, (8) Refrigerator, (9) Washing machine, (10) Telephone, (11) Radio, (12) Television, (13) Stereo, and (14) Vehicle.

^aIn the Latin American Migration Project data for Puerto Rico, there is no variation on Item 2, as all households report cement, finished, or wood floor.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

RESULTS

Migration and Parenthood

In Table 3, we display two sets of adult-centric estimates of recent migration and parenthood for domestic and international migration. The first set describes the percent of adults with recent migration experience, defined as migration within five years prior to the survey. The second set describes the percent of recent adult migrants with a child in the home at the time of departure. With respect to domestic migration, between 2.8 and 8.1 percent of adults migrated during the five years prior to the survey. International migration was smaller in magnitude for countries like Colombia and Costa Rica. Consistent with our earlier discussion, international migration was higher than domestic migration in Mexico and Nicaragua. With respect to the former, 7.3 percent of adults had international migration experience, but 4.2 percent had domestic migration experience.

Table 3.

Recent migration and parenthood: An adult-centric view

	Percent of adults with recent (5 years prior to survey) migration experience			Percent of recent adult migrants with child (age <18) in home at time of departure
Country/Territory	Domestic Migration	International Migration	N	Domestic Migration	N	International Migration	N
Colombia	6.8	2.6	2,630	39.4	179	22.7	72
Costa Rica	8.1	3.5	5,352	40.3	417	38.2	234
Dominican Republic	4.4	4.8	4,086	35.1	247	40.8	143
El Salvador	2.8	4.5	1,428	31.9	37	40.5	71
Guatemala	5.4	5.2	1,897	53.9	92	38.5	104
Mexico	4.2	7.3	51,082	35.6	2,324	34.2	3,743
Nicaragua	4.7	6.4	7,861	42.2	381	39.8	511
Peru	3.1	2.7	3,538	44.1	113	29.2	107
Puerto Rico	4.0	3.2	2,359	32.5	84	25.8	61

Notes: For Costa Rica, the Dominican Republic, and Guatemala, international migration in the Latin American Migration Project reflects only U.S. migration. For Puerto Rico, international migration includes U.S. migration.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

In the last two columns in Table 3, a large fraction of migrants were parents at the time of their departure. The findings are consistent with the aforementioned age patterns of migration in these countries and Puerto Rico (Lindstrom & Ramírez, 2010). In general, between about one-third and one-half of domestic migrants left behind a child at the time of their departure. Although smaller shares of international migrants were parents at the time of their departure, these shares are fairly large, ranging from 22.7 percent in Colombia to 40.8 percent in the Dominican Republic. Finally, it is worth noting that, given Mexico’s long history of out-migration, shares of domestic and international migrants who are parents at the time of their departure are the same: about one-third left behind a child in their household at the time of migration.⁵

Prevalence of Parental Absence via Migration

In Figure 1, we present child-centric estimates of parental absence via migration. In each country and Puerto Rico, the prevalence is substantial and ranges from 7.1 percent of children in Peru to approximately 21 percent of children in the Dominican Republic. In El Salvador, Mexico, Nicaragua and Puerto Rico, approximately one in six children under the age of 18 are living apart from a migrating parent. As we noted earlier, small sample sizes in some countries reduce the precision of our estimates. As is evident, this uncertainty is especially apparent for El Salvador and Puerto Rico.

Figure 1.

Percent of children under age 18 living in context of parental absence via migration

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Migration refers to domestic and international migration. Vertical bars correspond to standard errors.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

In Table 4, we further compare our estimates of the prevalence of parental absence via migration to other sources of parental absence.⁶ In several countries, e.g., Guatemala and Mexico, parental absence via migration contributes to about half of parental household absence (e.g., in the case of Mexico, 0.55 = 16.2/(100-70.6)). Importantly, the contribution of migration may be underestimated here: as parents may have migrated prior to a divorce or death, we attribute these absences to divorce or death, and not to migration. Consistent with the increase in divorce in Latin America and Caribbean during the last twenty years (Arriagada, 2007; Cerutti & Binstock, 2009; García & de Oliveira, 2011), our estimates show that upward of one in five children in Colombia are living apart from a parent because of union dissolution. However, excluding Colombia and Puerto Rico, the contribution of parental absence via migration is about equal to or greater than that of union dissolution (Table 4, Col. 5); this positive difference is statistically significant for El Salvador, Guatemala, Mexico, and Peru. Similarly, consistent with the rapid health and mortality transitions in Latin America over the past half century (McEniry, 2014), migration generates considerably more parental absence than parental mortality in the recent period (Table 4, Col. 6), with variation across countries and Puerto Rico largely matching variation in adult mortality (United Nations, 2015).⁷ Finally, between about 40 and 70 percent of children in these countries and Puerto Rico live in a two-parent home. Notably, these percentages indicate a marked change in parental absence since the mid-1970s, when World Fertility Surveys revealed that less than 8 percent of children in Latin American populations lived in a single-parent home, including 5.4 percent in Colombia, 6.5 percent in Costa Rica, 4 percent in Mexico, and 5 percent in Peru (DeVos et al., 1995; Richter, 1988).

Table 4.

Percent of children under age 18 by source of parental absence

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Source of Parental Absence
Country/Territory	Migration	Union Dissolution	Mortality	Two Parents in Household	Difference (1-2)	Difference (1-3)	Difference (1-4)
Colombia	9.5	21.3	2.1	48.3	−11.8	7.4 ***	−38.8 ***
Costa Rica	8.6	10.5	0.8	66.6	−1.9 ***	7.8 ***	−58.0 ***
Dominican Republic	20.8	16.8	1.3	43.1	4.0	19.5 ***	−22.3 ***
El Salvador	15.7	12.6	6.0	43.3	3.1 *	9.7 ***	−27.6 ***
Guatemala	13.9	3.2	5.1	69.9	10.7 ***	8.8 ***	−56.0 ***
Mexico	16.2	4.0	1.7	70.6	12.2 ***	14.5 ***	−54.4 ***
Nicaragua	16.6	17.0	2.7	45.3	−0.4	13.9 ***	−28.7 ***
Peru	7.1	6.1	2.3	70.6	1.0 ***	4.8 ***	−63.5 ***
Puerto Rico	15.6	22.2	5.8	40.5	−6.6 ***	9.8 ***	−24.9 ***

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Migration refers to domestic and international migration. Estimates in columns (1)-(4) for each country and Puerto Rico may not sum to 100 percent given reasons for parental absence other than migration, union dissolution, or mortality (see Appendix Table A2).

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

In light of the substantial prevalence of parental absence via migration, we conclude this subsection by presenting estimates disaggregated by domestic migration, international migration, mothers’ migration, and fathers’ migration in Table 5. Note that these are not mutually exclusive; some children fall into multiple categories. The relative contribution of international migration and domestic migration to parental absence via migration varies across populations. In Colombia, the Dominican Republic, Peru, and Puerto Rico, children are more likely to experience parental absence via migration from maternal migration. In Peru, this is more heavily concentrated among domestic movement. In El Salvador, Guatemala, and Mexico, children are more likely to experience parental absence via migration from paternal migration, which in these populations is more likely to be international.

Table 5.

Percent of children under age 18 by type of parental absence via migration

Country/Territory	Domestic Migration	International Migration	Mothers’ Migration	Fathers’ Migration
Colombia	4.7	5.0	6.4	3.4
Costa Rica	3.7	4.9	4.5	5.0
Dominican Republic	5.1	15.9	14.4	13.3
El Salvador	2.6	13.6	5.1	13.8
Guatemala	4.2	9.6	3.7	11.0
Mexico	3.0	13.5	5.9	13.9
Nicaragua	5.7	11.3	10.3	11.8
Peru	5.4	1.8	4.4	2.8
Puerto Rico	2.8	12.8	11.9	10.5

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Mothers’ and fathers’ migration refers to domestic and international migration. For Costa Rica, the Dominican Republic, and Guatemala, international migration in the Latin American Migration Project reflects only U.S. migration. For Puerto Rico, international migration includes U.S. migration. The four categories in the table are not mutually exclusive.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

Descriptively Linking Parental Absence via Migration to Children’s Welfare

Table 6 summarizes the presence of extended family members in children’s homes by source of parental absence. Column 1 summarizes the percent of children who live apart from a migrant parent and coreside with at least extended family member (grandparents, aunts and uncles, adult siblings, or other relatives). Columns 2 and 3 provide similar estimates for sources of parental absence that include union dissolution and mortality. These estimates can be compared against similar estimates for children residing in households with two parents and at least one extended family member. The remaining columns, 5-7, summarize differences relative to children who live apart from a migrant parent and coreside with at least extended family member.

Table 6.

Percent of children under age 18 coresiding with at least one extended family member by source of parental absence

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Source of Parental Absence
Country/Territory	Migration	Union Dissolution	Mortality	Two Parents in Household	Difference (1-2)	Difference (1-3)	Difference (1-4)
Colombia	65.8	58.9	77.8	45.9	6.9	−12.0	19.9
Costa Rica	35.0	60.6	58.4	39.4	−25.6 ***	−23.4	−4.4
Dominican Republic	50.1	74.9	53.4	41.9	−24.8 *	−3.3	8.2
El Salvador	62.3	44.7	77.2	38.8	17.6	−14.9	23.5 **
Guatemala	46.0	78.2	77.7	43.0	−32.2	−31.7 *	3.0
Mexico	35.6	58.0	68.2	39.9	−22.4	−32.6 **	−4.3
Nicaragua	59.7	80.9	85.0	58.3	−21.2 ***	−25.3 **	1.4
Peru	57.5	77.8	95.9	58.8	−20.3	−38.4 *	−1.3
Puerto Rico	36.3	48.4	14.9	22.8	−12.1	21.4	13.5

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Migration refers to domestic and international migration. Extended family members include any household adult who is not the child’s parent or stepparent.

^*p < 0.05

^**p < 0.01

^***p < 0.001

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

Table 6 shows that extended family coresidence is a relatively common practice. However, the most striking finding is the differential pattern of extended family coresidence by source of parental absence. Children living apart from migrating parents are not more likely to coreside with extended family relatives compared to children in two parent households. In fact, with the exception of Colombia and El Salvador, those living apart from migrant parents are less likely to coreside with extended family members compared to children living apart from divorced or separated parents. Similarly, excluding Puerto Rico, children living apart from migrant parents are less likely to live with extended family members than children with deceased parents. Indeed, in several populations—Costa Rica, Guatemala, Mexico, and Puerto Rico—the modal child living in the context of parental absence via migration is living with a single adult.

Table 7 presents values of asset indices calculated for each country and Puerto Rico. Because these values are country mean-centered, levels are comparable within populations. The results show that, despite living in households with fewer adults, on average, children living in the context of parental absence via migration reside in homes with more assets than children in two-parent homes. The observed higher levels of household assets for children in migrant homes and children in two-parent homes in Mexico is consistent with research on the role of remittances in alleviating poverty among single-mother homes influenced by parental absence via migration (Villareal & Shin, 2008; Arias, 2013). Children living in homes after union dissolution fare the worst: they have access to significantly fewer assets than children in migrant households and, in several cases, children in two-parent households. Children living apart from deceased parents also have access to fewer assets, on average, though the proportion of children affected by parental mortality is small.⁷

Table 7.

Mean household asset index score by source of parental absence for households with children under age 18

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Source of Parental Absence
Country/Territory	Migration	Union Dissolution	Mortality	Two Parents in Household	Difference (1-2)	Difference (1-3)	Difference (1-4)
Colombia	0.4	−1.1	−0.1	−0.2	1.4 ***	0.5	0.6 *
Costa Rica	0.1	−0.3	0.1	0.1	0.4 *	0.0	0.0
Dominican Republic	0.7	−0.1	−0.4	0.1	0.8 ***	1.1 ***	0.6 ***
El Salvador	0.1	−1.0	−0.8	−0.2	1.1 **	0.9	0.3 *
Guatemala	0.0	−0.6	−0.4	0.0	0.6	0.4	0.0
Mexico	0.0	0.0	0.0	0.0
Nicaragua	0.8	−0.2	−0.2	0.4	1.0 ***	0.9	0.4
Peru	0.5	−0.3	0.0	−0.2	0.9 ***	0.5	0.7 ***
Puerto Rico	−0.4	0.1	0.8	−0.1	−0.5	−1.3	−0.4

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Migration refers to domestic and international migration. Asset index constructed by combining information on the presence of 14 household items and using a factor analysis to weight the index items. Higher values indicate a larger number of assets. Index is mean-centered within countries and Puerto Rico. Significance tests of the difference include controlling for the number of years of education acquired by household head.

^*p < 0.05

^**p < 0.01

^***p < 0.001

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

The estimates displayed in Table 7 mask an important source of variation in a few populations. Specifically, in these populations, the link between parental absence via migration and household assets holds primarily for children living apart from fathers (versus mothers) who have migrated. As we show in Figure 2, children living apart from mothers who have migrated live in homes with fewer assets. In Guatemala and Peru, the differences are substantial. Exceptions to this pattern occur in El Salvador and several other countries and Puerto Rico in which household assets are similar when mothers and fathers migrate, a result consistent with Abrego (2009) who reported that children living in El Salvador apart from migrating mothers fare well economically because of mothers’ considerable efforts to provide resources to migrant-sending homes.

Figure 2.

Mean household asset index score by type of parental absence via migration for households with children under age 18

Notes: Asset index constructed by combining information on the presence of 14 household items and using a factor analysis to weight the index items. Higher values indicate a larger number of assets. Index is mean-centered within each country and Puerto Rico.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.

Because differences across families could be attributable to selection into migration, union dissolution, and adult mortality, variation in asset levels may therefore reflect pre-existing socioeconomic differences across family forms, i.e., the kinds of parents that separate/divorce or migrate. Migrant selection on socioeconomic indicators, e.g., education, varies across countries and Puerto Rico. For example, emigrants from the Dominican Republic and Puerto Rico tend to be better educated than those from Costa Rica (Massey & Sana, 2003). Likewise, educational selectivity in Mexico is relatively small (Feliciano, 2005). Accordingly, we further tested whether the comparisons in Table 7 hold after adjusting for variation in adult education levels in children’s households. Measuring the education level of the household head and the education level of the best-educated member of the household, we find that the differences in Table 7 are, in fact, robust to these adjustments.⁸

DISCUSSION

A growing body of research about transnational families focuses on children who are left behind in migrant-sending homes when their parents migrate across international borders or internally within countries. This scholarship takes the perspective that intergenerational effects of adult migration are a central pathway through which parental absence via migration affects children’s welfare in migrant-sending regions. However, despite a growing interest in this phenomenon, comparative assessments of the magnitude and associated implications of parental absence via migration are lacking, due in large part to the challenges of measuring parental migration in most standard data collection efforts (Caarls et al., this issue; Cortes, 2008; D’Emilo et al., 2007; Démurger, 2015; Mazzucato & Schans, 2011). In response, we drew on a unique set of harmonized household surveys collected in eight Latin American countries and Puerto Rico, which we compared to nationally representative census and survey data, to generate adult- and child-centric estimates of the prevalence of parental absence via migration that are plausibly comparable across populations. We subsequently linked our descriptive work to children’s welfare by examining their living arrangements and household assets in households with absent migrant parents.

The primary empirical contributions of the study are threefold. First, we provide unambiguous evidence that children’s living arrangements in contemporary Latin America depend on parental migration more than on parental mortality and at least as much as union dissolution. Second, we show that extended family coresidence is more common in the presence of union dissolution compared to parental migration. In several populations—Costa Rica, Guatemala, Mexico, and Puerto Rico—the majority of children living apart from migrating parents reside with a single adult. Third, children living in the context of parental absence via migration —and fathers’ migration in particular—fare better in terms of household assets relative to children living in other family forms; a finding that becomes more nuanced when one begins to consider which parent, i.e., fathers versus mothers, migrates (Garcia & Rojas, 2002; Lawson, 1998).

Several issues related to the use of cross-sectional, purposively sampled data are crucial to raise. First, the estimates presented here are descriptive and cross-sectional. Though the Latin American Migration Project and Mexican Migration Project data contain detailed migration and marital histories for household heads and spouses, information on extended family coresidence is recorded only at the time of the survey. As such, our work can be understood as providing a static, but comparative, snapshot of parental absence via migration relative to other demographic sources of parental absence. The second limitation concerns the fact that the data were designed to be representative at the community level, and therefore may or may not be nationally representative. Although, at least in the Mexican case, this feature has been shown to be inconsequential for generating certain types of population-level estimates (Massey & Zenteno, 1999), less is known about the representativeness of the Latin American Migration Project data. To mitigate this as a potential issue, we compared what quantities could be generated from the Latin American Migration Project and Mexican Migration Project to similar quantities derived from IPUMS-International and Demographic and Health Surveys. Importantly, we found no evidence that households are substantially different on basic characteristics related to children’s experience of parental absence. Nevertheless, it is possible that the Latin American Migration Project and Mexican Migration Project data miss migration from certain types of communities that have distinct experiences for children, e.g., communities in which divorce is much more common than migration and both are navigated by leveraging child care from extended families. When nationally-representative data are collected that facilitate describing parental absence via migration across Latin American populations, it will be essential to update these estimates.

Our findings lay the groundwork for a number of future lines of research on migration and transnational families. First, the findings emphasize the importance of efforts to study female migration (see also Caarls et al., this issue; Jordan et al., this issue). The estimates here indicate a form of family patterning that is distinct from the emphasis on migrant father/non-migrant mother families. In several countries, children are more likely to be living apart from migrating mothers than they are from migrating fathers. This is consistent with scholarship on women’s migration in Latin America (Abrego, 2009, 2014; Donato, 1993, 2010; Donato & Gabaccia, 2015; Dreby, 2010). Indeed, the present analysis indicates that many of these women, particularly domestic migrants, are mothers at the time of departure.

The findings also underscore the value of understanding strategies for childrearing in the context of parental absence via migration (Jordan et al., this issue; Wu et al., this issue). Despite the tendency to assume that extended family members help non-migrant parents, the evidence here suggests that these networks are less available than is widely assumed. This may arise because the types of families from which a parent migrates are families in which remaining parents have fewer neighboring kin, who may also be migrating. Alternatively, it may be that divorced parents caring for children—most often, mothers—experience both social and economic pressures to move in with natal families, whereas non-migrants who are still married—and often the recipients of remittances— live independently (Paoli, 2008).

Finally, our findings suggest that more research is needed about the benefits of female versus male migrant remittances when children are present (see also Jordan et al., this issue). To date, several studies suggest that men and women send differing amounts of remittances and that these remittances are allocated differently in the household (Antman, 2011; Antman, 2012; Curran & Rivero-Fuentes, 2003; Van Wey, 2004). Moreover, differences in mothers’ versus fathers’ allocation of household resources are well-documented (Thomas, 1990, 1994). Across countries and Puerto Rico, we observe substantial differences in children’s likelihood of living in resource-poor households when mothers (versus fathers) are migrating. The connection between the flow of household resources, shifts in parental decision-making about allocation, and, ultimately, their influences on children’s welfare warrants continued investigation.

Table A2.

Percent of children under age 18 by source of parental absence other than migration, union dissolution, or mortality

	(1)	(2)	(3)	(4)
		Among children with absent parents in (1)
Country/Territory	Parental absence for reasons other than migration, union dissolution, or mortality	Fertility outside of marriage	Lives apart from both parents	Parents alive, partnered, but at least one not coresident
Colombia	18.0	8.1	8.5	0.7
Costa Rica	13.1	5.6	6.7	0.6
Dominican Republic	18.0	5.8	9.5	1.6
El Salvador	21.7	7.0	12.8	0.0
Guatemala	7.9	1.0	2.8	1.9
Mexico	7.4	2.4	3.1	1.4
Nicaragua	18.1	3.3	10.3	3.2
Peru	13.8	4.6	5.4	1.4
Puerto Rico	15.8	3.7	8.8	1.6

Notes: Living arrangements assessed at time of survey with data from household roster and respondent migration histories. Migration refers to domestic and international migration.

Sources: Authors’ calculations using data from Latin American Migration Project and Mexican Migration Project.