Introduction

Between 1970 and 2005—a period characterized as the Great Mexican Emigration—the flow of migrants from Mexico to the United States increased more than 700 % (Hanson and McIntosh 2010). During this period, the larger but slower-growing interstate migration flows within Mexico increased by 20 % (Instituto Nacional de Estadística y Geografía INEGI 2010). Because these changes have had important implications for both sending and receiving communities, various facets of Mexican migration are now well documented (CONAPO 1999; Hill and Wong 2005; Marcelli and Cornelius 2001; Massey et al. 2002; Passel 2006). Nevertheless, a key implication of these changes has been largely ignored: what migration means for children’s family structure in Mexico.

This constitutes a surprising omission; the proportion of children affected by this process is substantial. Mexican migration is not limited to the young, unmarried, or childless. Although parenthood among migrants remains poorly documented, more than 30 % of migrants are older than 35 at departure (Chávez Juárez 2009; Marcelli and Cornelius 2001), well above the mean age at childbearing. In fact, among married couples, births increase the probability of men’s migration (Lindstrom and Giorguli Saucedo 2007). And although some children ultimately move to join migrant parents, families usually migrate in stages (Dreby 2010; Suarez-Orozco et al. 2002). In combination, the patterns point to a rise in a distinct form of family life in Mexico—one in which parents and children can be separated for years, despite the continuity of parents’ marriages.

Documenting the shift in children’s family structure is important because of mounting evidence that parental migration has meaningful implications for child development. Although research on nonmigrant children remains quite limited (particularly in comparison with research on children who move with parents to the United States), several studies suggest that the migration of Mexican parents constitutes a disruptive family transition (Dreby 2010; Falicov 2007),Footnote 1 with significant health and education costs for children who remain in origin homes. The migration of fathers is associated with increased illness (Schmeer 2009), reduced vaccination coverage, reduced breast-feeding yet larger birth weight (Hildebrandt and McKenzie 2005), lower educational aspirations (Kandel and Kao 2001; Kandel and Massey 2002; Nobles 2011), a lower probability of attending high school (Creighton et al. 2009; Halpern-Manners 2011), fewer completed years of schooling (McKenzie 2005), a reduction in study hours (Antman 2011), and a higher probability of behavioral problems (Heymann et al. 2009). Although many migrants send sizable remittances to families, these may not begin for months or years after a migrant’s departure. Moreover, remittances are often important but incomplete substitutes for the emotional, caregiving, and disciplinary roles filled by coresident parents (Dreby 2010; Goldring 2004; Hondagneu-Sotelo 1994).

Given these findings, this study highlights the dearth of population-level estimates on children’s family structure in lower-income countries, such as Mexico, and emphasizes that the form of parental absence most overlooked in developing contexts may now be the form that is most prominent. The research has three goals: (1) to use period data to document how the rise in Mexican migration over the past three decades has changed children’s coresidence with fathers; (2) to use detailed household survey data collected in the past decade to assess the relative contribution of migration, versus other demographic sources of father absence, to children’s residence in single-parent homes over the course of childhood; and (3) to assess whether these processes are disproportionately experienced by children from less-advantaged backgrounds. To this end, I also use migration flow data to document parenthood among adult migrants. Because Mexican men are much more likely than Mexican women to migrate without family members (Cerruti and Massey 2001), I emphasize the departure and return of fathers in this research.

The value of the undertaking arises from a voluminous literature demonstrating that the structure and stability of children’s living arrangements are strong predictors of their later-life outcomes and serve as a central pathway through which changing social phenomena shape child well-being (Amato and Gilbreth 1999; Fomby and Cherlin 2007; Strohschein 2005). Many studies have examined how family structure is shaped by mortality, fertility, and marital processes, identifying at-risk children and quantifying the potential impact of changes in divorce, cohabitation, and nonmarital fertility on population welfare (Bumpass and Lu 2000; Heuveline et al. 2003; Landale and Hauan 1992; Lloyd and Desai 1992; Mier y Terán et al. 2001; Mier y Terán and Rabell 2005). Unfortunately, we still know little about how migration contributes to child-parent coresidence across populations (United Nations Children’s Fund UNICEF 2008) or how migration compares with other demographic processes as a source of family instability.

In the absence of an intersection between migration, marriage, and fertility—if migrants delay fertility until after moving—children’s family structure in sending countries will remain unchanged, and the intergenerational effects of rising migration rates will be muted. By contrast, in contexts with lower divorce and nonmarital fertility rates, migration may be the dominant source of parental absence. In these cases, national estimates provide an important foundation for further research on this understudied group of children. When the absence of migrating parents has demonstrated costs for children, national estimates of shifting family structure are central to understanding the impact of migration on population welfare.

Data and Method

To capture Mexican family structure changes, I use four national data sets: the 1992, 1997, and 2009 cross-sectional Encuesta Nacional de Dinámica y Demográfica (ENADID) as well as the 2002 and 2005 waves of the longitudinal Mexican Family Life Survey (MxFLS). Mid-1970s estimates come from a fifth national data set, the 1976–1977 World Fertility Survey (WFS), and are calculated using tables from Richter (1988) and De Vos (1995). Description of the WFS is available at Ordorica and Potter (1981) and ENADID in Instituto Nacional de Estadística y Geografía (1994, 1999, 2010). The MxFLS collected data from 8,400 households in 150 communities across Mexico. In 2005, the second wave achieved 90 % respondent recontact (Rubalcava and Teruel 2008).

I use mothers’ reports of marital status and spouses’ residential location to generate period estimates of father absence due to migration and other demographic processes. U.S. migrants are fathers living in the United States at the survey date. Domestic migrants are fathers who live away from the household but not in the United States (Creighton et al. 2009).Footnote 2 Both groups exclude fathers who depart the household following divorce and separation, and I attribute these men’s absences to union dissolution.Footnote 3 Because fathers’ assignment as migrants is conditioned on their marriage to mothers, the estimates will, if anything, provide a lower bound on the contribution of migration to family structure. Migration may also contribute indirectly to father absence by influencing divorce, death, or nonmarital fertility (e.g., Frank and Wildsmith 2005).

To examine how migration affects father absence over the course of childhood, I develop a series of period increment-decrement life tables (Palloni 2001; Schoen 1988); these rely on the MxFLS data, the only national data set with sufficient information to estimate age-specific transition rates in children’s family structures. I use household roster information in combination with marital and migration histories to assign paternity to the children and to measure their living arrangements at the date of first interview in 2002 and exactly one year prior in 2001. The sample includes 11,110 children ages 0–14 in 2001 and ages 1–15 in 2002.

If father absence does have costs for children and is differentially borne by those with fewer resources, migration will likely play a role in the exacerbation of regional and social inequality across generations. To investigate differences in the estimates by potentially relevant household characteristics, I stratify the MxFLS sample first on rural residence at birth (67 % of the sample) and then by socioeconomic status—using an indicator that the child’s mother did not obtain any secondary schooling (57 % of the sample)—and generate a new set of life tables. Both characteristics are appropriately fixed at birth.Footnote 4 Finally, I resample the MxFLS data with replacement 1,000 times and use the sampling variability to create 95 % confidence intervals around the life table estimates (Efron and Tibshirani 1994).

Results

Migration, Parenthood, and Changes in Father Absence From 1975 to 2009

Although migration from Mexico to the United States has a long history with smaller increases in the early 1900s, the rise from 1970 to the late 2000s was unprecedented (Hanson and McIntosh 2010). Several studies have documented the changing flow composition during this period (Durand and Massey 2001; Marcelli and Cornelius 2001; Passel 2006); this is a difficult task because of the issues inherent to sampling a highly mobile and marginalized population. Despite frequent attention to age, education, and health selection, selection on parenthood among migrants is rarely documented.Footnote 5

The MxFLS data facilitate an assessment of flow composition because migrants were followed between waves, independent of documentation. An examination of data on adult respondents reveals that 30 % of men who migrated to the United States between 2002 and 2005 and 34 % of men who migrated domestically between 2002 and 2005Footnote 6 had children younger than age 15 living in their origin homes (Table 1). The percentages are substantially larger among older migrants.

Table 1 Adult male migrants between 2002 and 2005
Full size table

Intuitively, then, one would expect that rising migration would have implications for children’s family structure in Mexico. Period estimates confirm this. Between 1976 and 2005, the percentage of children living apart from migrant fathers (either domestic or international) increased by a factor of 5 (Fig. 1). In fact, the proportion with fathers in the United States increased by a factor of 4 between 1992 and 2005 alone. In 1992, 1 in 100 children had a father in the United States; by 2005, 1 in 22 children had a father in the United States. Although data on U.S. migration were not collected in the 1976–1977 WFS, if we assume that the percentage of children with fathers in the United States did not exceed one-fourth of children with either type of migrant father—a condition that holds in the 1990s when data on both forms of father absence are first available—the percentage of children in 1976 living apart from fathers in the United States would not exceed one-third of 1 %. Thus, the increase in the proportion of adult men migrating during the “Great Mexican Emigration” (Hanson and McIntosh 2010) produced an even larger increase in the proportion of Mexican children living apart from emigrant fathers. Notably, the change between 2005 and 2009 in the proportion of children with a father in the United States is consistent with data on post-recession declines in Mexican emigration (Passel et al. 2012).

Fig. 1
figure 1

Percentage of Mexican children living apart from fathers because of migration, divorce, nonmarital fertility, and death, 1976–2009. Percentages from 1976–1977 World Fertility Survey are drawn from Richter (1988). The estimate of children living with married mothers and migrant fathers in 1976 is calculated using De Vos (1995:Table A2) and Richter (1988:Table 5). Other percentages are calculated using data from the 1992, 1997, and 2009 Encuesta Nacional de la Dinámica Demográfica (INEGI 1999, 2010) and the 2002 and 2005 Mexican Family Life Survey (Rubalcava and Teruel 2008). Because of changes to the ENADID 2009 survey, the 2009 estimate of U.S. migration (3.71 %) is calculated by applying the observed ratio of children living apart from fathers in the United States to children living apart from all migrants in 2002 and 2005 to the percentage of children living apart from all migrants observed in 2009. All estimates are weighted

Full size image

Both divorce and nonmarital fertility increased roughly threefold during this period (Cerrutti and Binstock 2009; Instituto Nacional de Estadística y Geografía INEGI 2011),Footnote 7 yet the percentage of children affected by both processes has changed little over the past three decades. Adult mortality—and the percentage of children living with a widowed mother—declined by roughly one-third. As a result of these trends, migration is now the most common form of father absence in Mexico (Fig. 1).

Household Instability and the Temporal Burden of Father Absence

The preceding period estimates measure the percentage of children living apart from their fathers at a given survey date. They capture only a fraction of children whose fathers migrate at some point before the child reaches age 15. Increment-decrement life table estimates of the cumulative probability of fathers’ migration are shown in Table 2. Estimates in column 1 are calculated by taking the product of estimated age-specific transition probabilities (\( {}_1q_x^{ij } \)) of staying in state i.

$$ \prod\limits_0^{14 } {{}_1q_x^{ii }} . $$
Table 2 Cumulative conditional probabilities (p) of making transitions in living arrangements between birth and age 15 for children born into two-parent homes (95 % confidence intervals (CI) in parentheses)
Full size table

Estimates in columns 2 and 3 are calculated using the age-specific probabilities of moving from state i to state j, where state i is the starting state of interest, state j is all other states, and state k is the receiving state (Schoen 1988).

$$ 1-\left[ {\prod\limits_0^{14 }\left( {\left( {\sum\nolimits_j {{}_1q_x^{ij }} } \right)-{}_1q_x^{ik }}\right) } \right]. $$

Eighteen percent of children born in two-parent homes are expected to experience a fathers’ migration by age 15 (Table 2, summing columns 2 and 3). Because some children are born apart from migrating fathers (4 % to domestic migrants and 4.5 % to U.S. migrants), the proportion of all children with migrating fathers is even larger. Nearly 1 in 11 children are expected to experience a father’s U.S. migration, and more than 1 in 5 are expected to experience any type of migration by age 15.Footnote 8

Importantly, the estimates presented in Table 2 represent average probabilities for the population, which likely mask underlying heterogeneity. Clustering of migration among the same children could upwardly bias estimates in Table 2. One means of assessing this issue is to examine experiences of father absence among children aged 15 using the migration history data in the MxFLS. The completed cohort estimate closely matches the period estimate in Table 2. Among children of married parents, 6 % have fathers who migrated to the United States, and 16 % have fathers who migrated to any location between their births and age 15.

As a temporal burden over childhood, migration is also comparable to divorce in Mexico. Figure 2 demonstrates how the pooled person-years lived over childhood are distributed across living arrangements at the population level. Migration accounts for nearly one-third of children’s time spent outside a two-parent home (9 % of 29 %). This proportion appears roughly equally divided among internal and U.S. international migration. Eleven percent of children’s time is spent apart from fathers because of divorce or widowhood, and 7 % is spent apart from fathers because of nonmarital fertility.

Fig. 2
figure 2

Percentage of childhood spent in selected living arrangements between birth and age 15 in Mexico. Estimates are calculated with increment-decrement life tables. Births stratified by region and mothers’ education exclude children living apart from mothers. Source: Mexican Family Life Survey

Full size image

These averages mask significant regional and socioeconomic heterogeneity. Among children born into two-parent homes, the probability of having a father migrate either domestically or to the United States is substantially higher in rural regions and among children with less-educated mothers; in both subgroups, more than 20 % of children who reside with their mothers are expected to experience a father’s absence at some point during childhood because of migration (Table 2, summing columns 2 and 3). The probability of a father’s domestic migration is higher for children of less-educated mothers. By contrast, educational patterning is not observed in fathers’ U.S. migration. This is consistent with previous research on the educational selectivity of internal migrants in Mexico and the relative lack of educational selectivity among international migrants (Arenas et al. 2009; Feliciano 2005).Footnote 9

In rural regions, 14 % of childhood years are lived apart from migrating fathers (Fig. 2); in these areas, migration actually outpaces divorce, death, and nonmarital fertility combined as a contributor to children’s time apart from fathers. Similarly, children of less-educated mothers spend significantly more time with fathers migrating, either to the U.S. or domestically, than do children of mothers with at least some secondary schooling.

The findings are meaningfully large; omitting migration from estimates of children’s living arrangements understates by one-half the extent to which children in rural Mexico spend time in “single” parent homes.

Discussion

Children who remain in origin households are rarely incorporated into debates about the economic ramifications of migration systems (Hochschild 2000; McKenzie 2005; Parreñas 2005). Yet, a large number of children will be affected by the process; the worldwide population of international migrants now exceeds 200 million, and the worldwide population of internal migrants exceeds 700 million (United Nations 2009). The results here indicate that at least 4 % of children under 15 in Mexico—1.3 million children—have fathers living in the United States at present; twice that many are expected to experience father’s departure to the United States at some point before age 15. More than 6 million children will experience a father’s domestic migration by age 15. Given the well-documented effects of family stability for children’s later-life outcomes, one link between migration and socioeconomic change in origin communities will operate through the parenting investments made in the next generation.

Undoubtedly, remittances present an opportunity to offset the costs of instability and father absence. Nevertheless, remittances are not universal (Goldring 2004), and the capacity of financial transfers to offset these costs lacks strong empirical support, although this line of inquiry is still relatively new. The few studies on children’s outcomes test for the net effects of fathers’ departures over the short run; these effects appear to be negative (Creighton et al. 2009; Heymann et al. 2009; Schmeer 2009).

This study thus highlights a second point: the traditional conception of family transitions must be extended to include parental migration as a form of household instability for children in middle- and low-income countries. In Mexico, migration accounts for nearly the same percentage of childhood spent apart from fathers as divorce. In rural areas, migration dwarfs other demographic processes as a driver of father absence.

Documenting the effect of migration on children’s family structure thus serves as an important counterpart to similar undertakings in cohabitation research, which showed that children who appeared to be raised by single parents were actually living with two (Bumpass and Lu 2000). The opposite argument must also be considered: in developing countries, many children who appear to be living in homes with two married parents are actually living with one.

And despite the importance of extended-family caregiving in many societies, Mexican children with emigrant fathers do not disproportionately coreside with extended family relative to children living with both parents (Nobles et al. 2012).

The magnitude of the migration findings is notable because the estimates almost certainly constitute lower bounds on the proportion of children affected by migration. Children whose parents are divorced may also have fathers in the United States. In this study, time apart from these fathers is attributed exclusively to divorce. And although not emphasized here, the number of migrating Mexican mothers appears to be growing (Dreby 2010). The national estimates also mask further regional heterogeneity. Targeted samples from historically important sending states have noted an even larger proportion of children with absent parents.Footnote 10

The study thus provides an important foundation for further research on children raised in the origin regions of major migration flows. That migration can contribute to meaningful gains in microenterprise, capital accumulation, and city infrastructure is undeniable (Kanaiaupuni and Donato 1999; Massey et al. 2002; Woodruff and Zenteno 2007). Parental migration likely generates new opportunities for some children. At the same time, the effects of household instability on children’s subsequent education, mental health, employment, and negotiation of adult relationships are increasingly well established (Chase-Landsdale et al. 1995; Donohue et al. 2010; Halpern-Manners 2011; Sassler et al. 2009; Sun and Li 2008). Whether children of migrants, now a substantial proportion of the Mexican population, reach adulthood ultimately advantaged by remittances and regional development gains remains an important subject for future research.