Abstract
Using data from the birth cohort of the Early Childhood Longitudinal Study (n = 1,200) and the Mexican Family Life Survey (n = 1,013), this study investigated the living arrangements of Mexican-origin preschool children. The analysis examined children’s family circumstances in both sending and receiving countries, used longitudinal data to capture family transitions, and considered the intersection between nuclear and extended family structures. Between ages 0–1 and 4–5, Mexican children of immigrants experienced significantly more family instability than children in Mexico. They were more likely to transition from 2-parent to single-parent families and from extended family households to simple households. There were fewer differences between U.S. children with immigrant versus native parents, but the higher level of single parenthood among children of natives at ages 0–1 and the greater share making transitions from a 2-parent to a single-parent family suggest ongoing erosion of children’s family support across generations in the United States.
Keywords: early childhood, Hispanic Americans, immigration, living arrangements
Families play a pivotal role in immigration and immigrant adaptation. As a key decision-making unit in migration that forms the heart of migrant networks, the family is “the strategic research site for understanding the dynamics of immigration flows (legal and illegal) and of immigrant adaptation processes” (Rumbaut, 1997, p. 4). For children of immigrants, the family is a proximate context that connects origins and destinations. Even though most children of immigrants in the United States are native born, their family environments reflect their parents’ premigration experiences, the process of migration itself, and the challenges faced after arrival (Landale, Thomas, & Van Hook, 2011). Using resources garnered in both origin and destination countries, immigrant families reshape themselves as they adjust to life in a new country.
Studies that integrate data from origin and destination countries are at the forefront of research on the intersection between immigration and family processes (e.g., Landale, 1994; Singley & Landale, 1998; Van Hook & Glick, 2007). Binational studies allow researchers to assess whether family patterns among immigrants closely resemble the family patterns of those who stayed behind or whether they are distinctive due to migration-related factors, such as selective emigration, kin availability, economic uncertainty, or exposure to a new social setting.
Despite the promise of the binational approach, destination-only studies predominate because of a scarcity of data that permit rigorous comparisons between sending and receiving countries. Such studies typically compare the family patterns of three generations within an ethnic group: the foreign born, the native born of foreign parentage, and the native born of native parentage. In studies of immigrants, the foreign born may be disaggregated by duration of residence to examine the effect of cumulative exposure to the destination or by age at arrival to emphasize the timing of onset of exposure. Although useful, destination-only studies can be misleading if researchers assume that immigrants’ behavior is typical of the sending country and a benchmark for examining the extent of assimilation across generations.
At the same time, most data sources that can be used to conduct binational studies are cross-sectional, precluding examination of changes in family circumstances over time. Cross-sectional studies may mask considerable churning in family circumstances, which may vary across origin and destination countries or by generation in the receiving country. Thus, Glick’s (2010) review of research on immigrant families echoes an emerging consensus that it is critical to include both sending and receiving countries in research and to use longitudinal data to capture family processes among immigrant populations.
We took advantage of a unique opportunity to address these research challenges in an analysis of Mexican-origin children’s family structure transitions between birth and age 5. Because of the sharp rise in the number of immigrants from Mexico during the last four decades and the relatively high fertility of those immigrants, a large share (39%) of the children of immigrants under age 5 are Mexican (Passel, Cohn, & Gonzalez-Barrera, 2012). Little is known about the stability of these children’s families, despite the special challenges faced by their immigrant parents. Binational, longitudinal data from the Mexican Family Life Survey (MxFLS; see http://www.ennvih-mxfls.org/en/mxfls.php?seccion=1&subseccion=1&session=) and the birth cohort of the Early Childhood Longitudinal Study (ECLS-B; see http://www.nces.ed.gov/ecls/birth.asp) allowed us overcome this research gap by providing a detailed portrait of two dimensions of family structure (nuclear and extended) among three groups of Mexican-origin preschool children: (a) children in Mexico, (b) U.S.-born children of immigrants, and (c) U.S.-born children of natives.
Examination of changes in family structure among children in immigrant populations is important given recent evidence that family structure transitions—especially multiple transitions—are detrimental to children (Fomby & Cherlin, 2007; Magnuson & Berger, 2009). This conclusion is based on an extensive and increasingly nuanced literature on changes in parental union status (Brown, 2010), but less is known about transitions in extended family living arrangements. Nonetheless, one recent study (Mollborn, Fomby, & Dennis, 2012) showed that a larger share of young children undergo transitions in extended household structure compared to nuclear household structure and that shifts in extended family living arrangements are associated with children’s developmental outcomes.
Our research was motivated by both descriptive and analytic objectives. In keeping with the need for careful description as a prelude to explanation (Abbott, 1998; Bianchi & Casper, 2005), the first objective was to provide a nuanced view of how the prevalence and stability of different family structures during early childhood are shaped by parental immigration status. The analytic objective was to assess the role of immigration status in key family structure transitions using a multivariate framework that considered characteristics of the child, mother, and family on both sides of the border.
Method
Data
The population of interest in this study was Mexican-origin preschool children in the United States and Mexico. We used longitudinal data from surveys conducted in each country in approximately the same years to examine children’s family situations at ages 0–1 and 4–5. The ECLS-B provided data from personal interviews with the parents of a nationally representative sample of approximately 14,000 children born in the United States in 2001. These children were identified from birth certificates and selected with oversampling of some ethnic and birth-outcome groups (e.g., twins, low-birth-weight infants). The parents were surveyed when their children were about 9 months (2001–2002), 2 years (2003–2004), 4 years (2005–2006) and 5–6 years old (2006–2007 or 2007–2008). The weighted overall response rate for the first wave was 74%, and over 90% of the parents responded in each subsequent wave. Sample sizes for the ECLS-B have been rounded to the nearest 50, as required by the confidentiality agreement for use of restricted data.
Children from the ECLS-B were considered to be of Mexican origin if they were classified as Mexican or had a parent who was classified as Mexican. For cases with incomplete survey data on ethnicity, the child’s birth certificate was used as a source of information. About 700 of the Mexican children were U.S.-born children of immigrants, and 600 were U.S.-born children of natives. After selecting one child per household (see below) and eliminating seven higher order births that lacked information on whether they were from the same household, our analysis was based on 650 U.S.-born children of immigrants and 550 U.S.-born children of natives. For ease of presentation, we refer to these two groups of U.S.-born Mexican children as children of immigrants and children of natives, respectively.
Information on children in Mexico comes from the MxFLS, a longitudinal survey administered to a nationally representative sample of Mexican households in the same time frame as the first and third waves of the ECLS-B. Multistage and cluster-based probability sampling methods generated a sample of 8,440 households that supplied 35,000 individual interviews in 2002. A household roster and various survey components provided information on all household members, including children. The MxFLS had response rates of 85% to 95%, depending on the question module, and 90% of households were recontacted in 2005–2006.
To generate an analytic sample comparable to the ECLS-B, we selected the 1,140 children who were ages 0–1 in Wave 1 and present in Wave 2 at ages 4–5. To have a well-defined reference group for comparison with the two groups of Mexican-origin children in the United States, we eliminated 39 children whose parents had prior U.S. migration experience. Such children are potentially of interest because parental migration to the United States is common among children in Mexico (Nobles, 2013), but they were not sufficiently numerous in our sample of infants to be analyzed as a separate group. These selection criteria yielded a sample of 1,101 children. After we eliminated 14 children with no parents present, randomly selected one child per twin pair, and eliminated 17 higher order births, our MxFLS sample included 1,013 children.
The decision to select one child per household as the focal child stemmed from the oversampling of twins in the ECLS-B and the inclusion of multiple children per household in the MxFLS. To avoid well-known problems of statistical inference with clustered data, we randomly selected one child per household from each data set. Another potential option was to use survey analysis procedures to generate robust standard errors, which take clustering and other features of the sample designs into account. This option was not viable because the MxFLS does not release variables with which to identify primary sampling units. Randomly selecting one child per household is a viable alternative and replicates a common sample selection protocol for surveys that focus on children.
Measures
Children’s family structure
Nuclear and extended family structures at ages 0–1 (Time 1 [t1]) and 4–5 (Time 2 [t2]) were defined in terms of persons sharing a household dwelling unit with the child. Nuclear family structure was delimited by the number of coresident parents and their union status. Children were classified as living with two married parents, two cohabiting parents, or a single parent. Extended family structure was defined by the presence of family members other than parents and their offspring in the household. This variable distinguished households that were not extended, vertically extended (grandparents present), horizontally extended (aunts, uncles, or cousins present), and both vertically and horizontally extended. Our descriptive analysis focused on the prevalence of these dimensions of family structure, their intersection, and their stability.
Our multivariate analyses examined six significant transitions in family structure between ages 0–1 and 4–5. Three changes in nuclear family structure were examined: (a) any transition, (b) transition to two married parents among those in single-parent or cohabiting families at t1, and (c) transition to single parent among those in married or cohabiting families at t1. Similarly, three changes in extended family structure were examined: (a) any transition, (b) transition to a nonextended—or simple—family household among those in an extended family at t1, and (c) transition to an extended family household among those not in an extended family at t1. The analysis of each transition was based on those at risk of a particular change.
Migration status
The central predictor of interest was migration status, which was measured in terms of place of residence and parents’ migration experience. Three groups were distinguished: (a) children in Mexico whose parents had never migrated to the United States, (b) U.S.-born Mexican children of immigrants, and (c) U.S.-born Mexican children of natives. Two additional groups of potential interest were (a) children in Mexico whose parents were return migrants and (b) children in the United States who were immigrants themselves. As noted, only 39 children in our MxFLS sample had parents with prior U.S. migration experience. The ECLS-B included only U.S.-born children by design, but only 1% of children under age 4 with parents from Mexico were themselves born in Mexico (The Urban Institute, 2012). Thus, the three migration status groups we examined represent almost all Mexican preschool children in the United States and Mexico.
Control variables
The multivariate analyses included covariates that could be harmonized across the two data sets and were measured at the initial time point. Child characteristics included child age in completed years (1 vs. 0) and gender. Mother characteristics included maternal age in completed years, education (less than high school, high school, beyond high school), and employment status (employed vs. not employed). Because the preliminary analysis failed to reveal evidence of nonlinearity in the relationships between maternal age and the outcome variables, neither a quadratic transformation nor dummy coding was necessary.
Family characteristics included nuclear and extended family structure at ages 0–1 (coded as specified above) and number of children in the family. Another predictor was household income quartile, determined separately for the United States and Mexico because of the dissimilar income distributions and cost of living in the two countries. Household income at t1 was clearly bound up with family composition at t1 and does not represent resources transportable to another family form. Unfortunately, more detailed income data for specific household members were not available in the ECLS-B. Last, we included a measure of rural residence. Families in communities of fewer than 2,500 residents were distinguished from others.
We selected these control variables because prior research shows that family structure is associated with socioeconomic status. Low maternal education, limited work experience, and low income increase the risk of single parenthood and cohabitation. Extended family living arrangements are also associated with economic need. Individuals with low human capital and economic resources may need to share a residence with kin to get by. Expectations regarding family transitions are less well developed and may depend on the specific transition analyzed.
Missing data and weights
We used multiple imputation to deal with missing data (Rubin, 1987). The harmonized ECLS-B and MxFLS data sets were pooled, and five imputed data sets were created to generate plausible values for missing data. We then analyzed the five data sets with standard complete-data methods and results were combined to yield estimates, standard errors, and p values that take uncertainty about the imputed values into account.
The sample designs of the ECLS-B and the MxFLS required that the data be weighted to generate the correct parameter estimates (the MxFLS provided special case weights for analyses of children). Consequently, all analyses were based on weighted data, with weights adjusted to return the total size of each sample to its unweighted N.
Analytic Strategy
Although prior studies have shown variation in family structure by parental nativity among Mexican children in the United States, our binational study design permitted new comparisons with children in nonmigrant families in Mexico. We first considered differences across migration status groups in separate measures of nuclear and extended family structure. We then analyzed the intersection of these dimensions of family structure at ages 0–1 and 4–5 to provide a detailed descriptive portrait of migration group differences in children’s living arrangements and their stability across the preschool years. Next, we conducted multivariate analyses of the major transitions in nuclear and extended family structures with a series of logistic regression models using the pooled longitudinal data from the ECLS-B and MxFLS. These analyses accomplish the descriptive and analytic objectives posed at the outset.
Results
Prevalence and Stability
The prevalence of nuclear and extended family structures among Mexican children shortly after they were born is shown in Table 1. There was little difference between children of immigrants (88%) and children in Mexico (84%) in the likelihood of living with two parents, but there were underlying differences in parental union type. Children of immigrants were less likely to live with married parents (59% vs. 67%) and more likely to live with cohabiting parents (29% vs. 17%) than children in Mexico. Among children in two-parent families in Mexico, there were nearly four times as many children with married parents as with cohabiting parents (67/17 = 3.9). Among their immigrant counterparts, there were only twice as many children with married parents as with cohabiting parents (59/29 = 2.0). In contrast, the children of natives were especially likely to live in a single-parent family. About 28% lived with a single parent, compared to 12% of children of immigrants and 16% of children in Mexico.
Table 1. Prevalence of Nuclear and Extended Family Structures at Age 0–1: Mexican Children in Mexico and the United States.
| Dimensions of family structure | Mexico | United States | |
|---|---|---|---|
| Children with nonmigrant parents (%; N = 1,013) |
U.S.-born children of immigrant parents (%; N = 650) |
U.S.-born children of U.S.-born parents (%; N = 550) |
|
| Nuclear family type | |||
| Two parents | 84.2 | 87.6 | 72.6 |
| Married | 66.9 | 58.6 | 52.6 |
| Cohabiting | 17.3 | 29.0 | 20.0 |
| Single parent | 15.8 | 12.4 | 27.5 |
| Extended family type | |||
| Not extended | 63.5 | 67.3 | 67.3 |
| Extended | 36.6 | 32.7 | 32.7 |
| Vertical | 30.8 | 5.1 | 13.3 |
| Horizontal | 3.2 | 18.6 | 4.6 |
| Both vertical and horizontal | 2.6 | 9.0 | 14.8 |
Note: The number of cases listed for each group is based on unweighted data. Numbers for children in the United States are rounded to the nearest 50 to comply with Early Childhood Longitudinal Study birth cohort requirements. Data for children in Mexico come from the Mexican Family Life Study; data for children in the United States come from the birth cohort of the Early Childhood Longitudinal Study.
Group differences were greater in the type of extended family structure than in the likelihood of family extension per se. Regardless of parental birthplace, roughly one third of U.S. Mexican children lived in some type of extended family at ages 0–1. The comparable figure for children in Mexico was 37%. Consistent with possible limitations in kin availability due to grandparents remaining in the origin country, only 5% of children of immigrants lived in families characterized by vertical extension alone. Another 9% lived in families with both vertical and horizontal extension. This differed sharply from Mexico, where 31% of children lived in vertically extended families and another 3% lived in families with both vertical and horizontal extension. However, children of immigrants (19%) were more likely than those in Mexico (3%) or children of natives (5%) to live in horizontally extended families. The latter group was the most likely to live in a family with both types of extension (15% vs. 3% of children in Mexico and 9% of children in immigrant families).
The intersection of nuclear and extended family structures and the stability of complex family arrangements are shown in Table 2. Three columns are provided for each migration status group. The first column indicates the distribution of children ages 0–1 (t1) across 12 categories defined by the cross-tabulation of nuclear and extended family structure. It also indicates (in boldface type) the percentage of children in each nuclear family type. The second column summarizes children’s family structure when they were ages 4–5 (t2). The third column indicates the percentage of children who remained in the same detailed family type from ages 0–1 to ages 4–5.
Table 2. Prevalence and Stability of Combinations of Nuclear and Extended Family Structures: Mexican Children in Mexico and the United States.
| Family structure | Mexico | United States | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Children with nonmigrant parents (N = 1,013) |
U.S. born children of immigrant parents (N = 650) |
U.S. born children of U.S. born parents (N = 550) |
|||||||
| Age 0–1 (%) |
Age 4–5 (%) |
Stable family structurea (%) |
Age 0–1 (%) |
Age 4–5 (%) |
Stable family structurea (%) |
Age 0–1 (%) |
Age 4–5 (%) |
Stable family structurea (%) |
|
| Two married parents | 66.9 | 68.3 | (90.9) | 58.6 | 62.6 | (95.5) | 52.6 | 55.3 | (88.7) |
| Nuclear | 48.1 | 50.6 | 89.6 | 43.5 | 49.1 | 84.8 | 42.7 | 47.4 | 84.7 |
| Vertically extended | 15.6 | 14.1 | 67.9 | 2.5 | 2.5 | 33.1 | 5.4 | 3.7 | 45.4 |
| Horizontally extended | 1.8 | 2.3 | 69.7 | 9.2 | 8.8 | 36.0 | 2.7 | 2.6 | 33.6 |
| Both types of extension | 1.5 | 1.4 | 76.5 | 3.4 | 2.2 | 13.6 | 1.8 | 1.6 | 9.3 |
| Two cohabiting parents | 17.3 | 16.3 | (61.1) | 29.0 | 23.7 | (68.7) | 20.0 | 14.1 | (49.2) |
| Nuclear | 11.7 | 11.1 | 63.0 | 17.8 | 18.4 | 68.5 | 15.1 | 11.2 | 45.8 |
| Vertically extended | 4.8 | 4.3 | 44.6 | 1.4 | 0.4 | 9.6 | 1.4 | 0.8 | 7.9 |
| Horizontally extended | 0.5 | 0.4 | 62.1 | 7.0 | 3.9 | 35.9 | 0.8 | 1.8 | 83.9 |
| Both types of extension | 0.3 | 0.5 | 19.9 | 2.8 | 0.9 | 6.9 | 2.6 | 0.3 | 8.0 |
| Single parent | 15.8 | 15.4 | (83.2) | 12.4 | 13.8 | (56.7) | 27.5 | 30.7 | (71.2) |
| Nuclear | 3.7 | 3.6 | 55.6 | 6.0 | 6.8 | 41.0 | 9.5 | 13.9 | 40.6 |
| Vertically extended | 10.4 | 8.9 | 76.7 | 1.3 | 2.6 | 63.4 | 6.5 | 9.5 | 55.9 |
| Horizontally extended | 0.9 | 1.5 | 76.8 | 2.4 | 1.6 | 20.1 | 1.1 | 1.6 | 5.5 |
| Both types of extension | 0.9 | 1.4 | 89.3 | 2.7 | 2.9 | 55.9 | 10.4 | 5.8 | 37.3 |
Note: The number of cases listed for each group is based on unweighted data. Numbers for children in the United States rounded to the nearest 50 to comply with requirements of the birth cohort of the Early Childhood Longitudinal Study (ECLS-B). Data for children in Mexico come from the Mexican Family Life Study; data for children in the United States come from the ECLS-B. The percentages of children in each nuclear family type are highlighted in boldface type. The shaded columns show the percentages of individual children who were in the same family arrangement at both ages.
In this column, numbers in parentheses represent stability in the number of parents in the household and parental union type regardless of extended family arrangements. Numbers without parentheses represent stability in household structure taking family extension into account.
Within each migration status group, the aggregate distributions for family structure at ages 0–1 and 4–5 were roughly similar. There was a slight movement toward married-parent nuclear families among both groups of U.S. children and toward single-parent nuclear families among U.S. children of U.S.-born parents, reflecting the dissipation of two-parent cohabiting families over time. Nonetheless, these changes were not substantial, and stability appears to be the rule rather than the exception at the aggregate level.
This apparent stability in the overall distributions is deceptive because it masks considerable churning at the individual level. The shaded columns in Table 2 show the percentages of individual children who were in the same family arrangement at both ages. We first focus on the stability of nuclear family structure (numbers in boldface type in Table 2) without considering extended family type. Children of immigrants who lived with two married parents at ages 0–1 experienced greater family stability (96%) than children in Mexico (91%) or children of natives (89%). A similar pattern was found for those who lived with cohabiting parents. As we will see, however, the pattern for all two-parent families differs because of variation across the groups in the mix of marital and cohabiting unions. Last, children of immigrants were substantially less likely to remain in a single-parent family than others (57% vs. 83% in Mexico and 71% among children of natives).
Much less stability was evident when we jointly considered nuclear and extended family structures, especially among U.S. children. For example, only 33% of children of immigrants and 45% of children of natives in vertically extended married-couple families at ages 0–1 remained in the same type of family at ages 4–5. The comparable figure for children in Mexico was 68%. A similar pattern was evident for horizontally extended married-couple families. Among children in vertically extended single-parent families at ages 0–1, the percentage remaining in the same family structure at ages 4–5 ranged from 56% among children of natives to 77% among children in Mexico. Despite some variation across specific family arrangements, the overall pattern suggests that children in Mexico experienced the greatest family stability. U.S.-born children with immigrant or native parents were more likely to go through changes in the intersection of nuclear and extended family structure across the preschool years.
Transitions
Our multivariate analysis shifted the focus to specific routes to family instability using the harmonized pooled data. We first considered changes in nuclear family structure between t1 and t2: (a) any transition, (b) transition to a married-couple family, and (c) transition to a single-parent family. We then considered shifts in family extension between t1 and t2: (a) any transition, (b) transition to a simple household, and (c) transition to an extended family. Collectively, these describe the major family transitions experienced by Mexican-origin preschool children.
Descriptive statistics for the dependent variables and predictors by migration status group are provided in Table 3. Significance tests use children of immigrants as the reference group. In regard to the dependent variables, children in Mexico differed from children of immigrants on every family transition except one. Children in Mexico were more likely to experience the transition to a married-parent family and less likely to experience the transition to a single-parent family than children of immigrants. These differences offset each other, resulting in little difference between the two groups in “any transition” in nuclear family structure (17% vs. 15%). Regarding family extension, children in Mexico were significantly less likely to experience instability through transitions into or out of extended family living arrangements. About 11% experienced a transition in extended family structure between ages 0–1 and 4–5, compared to 28% of children of immigrants.
Table 3. Descriptive Statistics (Weighted Means/Proportions), Mexican Children in Mexico and the United States.
| Variable | Children in Mexico, nonmigrant parents (N = 1,013) |
U.S.-born children, immigrant parents (N = 650) |
U.S.-born children, U.S.-born parents (N = 550) |
|---|---|---|---|
| Transitions in family structurea | |||
| Any transition in nuclear family structure | 0.15 | 0.17 | 0.24** |
| Transition to two married parents (sample: Unmarried, Time 1) | 0.23* | 0.16 | 0.18 |
| Transition to single parent (sample: Partnered, Time 1) | 0.03*** | 0.08 | 0.15*** |
| Any transition in extended family structure | 0.11*** | 0.28 | 0.27 |
| Transition to simple household (sample: Extended, Time 1) | 0.16*** | 0.47 | 0.42 |
| Transition to extended household (sample: Simple, Time 1) | 0.06*** | 0.12 | 0.13 |
| Child characteristics | |||
| Age | 0.51*** | 0.12 | 0.17* |
| Male | 0.48* | 0.53 | 0.52 |
| Mother characteristics | |||
| Age | 27.79 (6.99) | 27.62 (6.39) | 25.37 (5.68) |
| Education | |||
| Less than high school | 0.80*** | 0.51 | 0.27*** |
| High school | 0.13*** | 0.31 | 0.38 |
| Beyond high school | 0.07*** | 0.18 | 0.34*** |
| Employed | 0.23*** | 0.35 | 0.52*** |
| Family characteristicsb | |||
| Number of children | 2.52*** (1.69) | 2.20 (1.33) | 1.88*** (1.08) |
| Household income quartile | |||
| First quartile (lowest) | 0.26 | 0.28 | 0.20** |
| Second quartile | 0.24*** | 0.34 | 0.21*** |
| Third quartile | 0.25 | 0.25 | 0.28 |
| Fourth quartile (highest) | 0.27*** | 0.12 | 0.30*** |
| Rural residence | 0.26*** | 0.03 | 0.06* |
Note: Data for children in Mexico come from the Mexican Family Life Study; data for children in the United States come from the birth cohort of the Early Childhood Longitudinal Study. Standard deviations for continuous variables are presented in parentheses.
Analysis of each dependent variable is restricted to children at risk of the specific transition. Children in cohabiting-couple families are at risk of transitions to both married and single families.
Parental union status and family extension type at Time 1 were also included as predictors in the multivariate analysis. Descriptive information on these variables is shown in Table 1.
p < .05.
p < .01.
p < .001. (Significantly different from U.S.-born children with immigrant parents.)
The primary difference between Mexican children with immigrant versus native parents was in the transition to single parenthood. About 15% of children of natives made a transition from a two-parent family to a single-parent family before they reached ages 4–5, compared to 8% of children of immigrants. This variation drives the difference between the two groups in “any transition” in nuclear family structure. There were no significant differences in transitions in extended family structure by parental nativity among Mexican children in the United States.
Also noteworthy are the marked differences in maternal education and employment. The share of children with a mother who had less than a high school education ranged from 80% for children in Mexico to 51% for children of immigrants to 27% for children of natives. At the other end of the distribution, the percentage of children whose mothers had an education beyond high school was 7% for children in Mexico, 18% for children of immigrants, and 34% for children of natives. Maternal employment exhibited a similar pattern, with only 23% of children in Mexico having employed mothers, compared to 35% of children of immigrants and 52% of children of natives. By definition, roughly 25% of children in Mexico and in the United States fall into each country-specific income quartile. Nevertheless, a comparison of U.S. children by parental nativity showed that children of immigrants were more likely than children of natives to fall into the lowest quartiles of the income distribution. As might be expected, a relatively high share of children in Mexico lived in rural areas (26% vs. 3%–6% for those in the United States).
In Table 4 we examine changes in nuclear and extended family structure within a multivariate framework. We focus first on results for nuclear family structure shown in columns 1 through 3. Odds ratios are presented from models that included migration status and the full set of child, mother, and family characteristics as predictors. The odds ratios in column 1 indicate that children of natives were significantly more likely than children of immigrants to undergo a transition in nuclear family structure. Their greater family instability cannot be explained by differences in child age and gender, maternal characteristics, or family characteristics at t1. The estimate for children with native-born parents (1.47) indicates that their odds of experiencing any type of transition in nuclear family structure were about 50% higher than the odds for children of immigrants. Children in Mexico did not differ from children of immigrants.
Table 4. Logistic Regression Models Predicting Transitions in Family Structure Between Ages 0–1 and 4–5 (Odds Ratios).
| Transitions in nuclear family structure | Transitions in extended family structure | |||||
|---|---|---|---|---|---|---|
| Independent variables | Any: All cases (N = 2,250)a |
To marriage from unmarried (N = 1,000) |
To single from partnered (N = 1,800) |
Any: All cases (N = 2,250) |
To simple from extended HH (N = 800) |
To extended from not extended HH (N = 1,450) |
| Child characteristics | ||||||
| Migration status | ||||||
| Child in Mexico, nonmigrant parents | 0.88 | 1.24 | 0.30*** | 0.24*** | 0.13*** | 0.35*** |
| U.S.-born child, immigrant parents (ref.) | ||||||
| U.S.-born child, U.S. born parents | 1.47* | 1.24 | 2.16*** | 1.02 | 0.81 | 1.03 |
| Age | 1.35* | 1.35 | 1.21 | 1.27† | 1.49† | 1.36 |
| Male | 0.81† | 0.86 | 0.58** | 0.82† | 0.74† | 0.87 |
| Mother characteristics | ||||||
| Age | 0.95*** | 1.00 | 0.93*** | 0.99 | 0.96* | 1.01 |
| Education | ||||||
| Less than high school (ref.) | ||||||
| High school | 0.99 | 1.01 | 0.85 | 0.75* | 0.91 | 0.54* |
| Beyond high school | 1.32 | 1.85* | 0.94 | 0.81 | 1.23 | 0.52* |
| Work status | ||||||
| Not employed (ref.) | — | |||||
| Employed | 0.99 | 0.86 | 1.17 | 0.87 | 0.59** | 1.05 |
| Family characteristics | ||||||
| Parental union status | ||||||
| Single (ref.) | NA | |||||
| Cohabitation | 1.92*** | 2.18*** | 3.50*** | 0.94 | 1.02 | 0.33*** |
| Married | 0.28*** | NA | 1.09 | 1.88** | 0.31*** | |
| Number of children | 1.21*** | 1.09 | 1.17† | 0.95 | 1.03 | 0.87† |
| Extension type | ||||||
| None (ref.) | NA | NA | ||||
| Vertically extended | 1.37† | 1.28 | 1.66† | 5 43*** | 1.46 | NA |
| Horizontally extended | 0.97 | 1.33 | 0.63 | 8.22*** | 1.23 | NA |
| Both extension | 0.87 | 1.32 | 1.09 | 9.21*** | NA | |
| Rural residence | 0.81 | 1.43 | 0.64 | 1.10 | 0.89 | 1.12 |
| Household income quartile | ||||||
| First quartile (lowest; ref.) | — | |||||
| Second quartile | 0.94 | 1.30 | 0.71 | 0.62* | 0.73 | 1.08 |
| Third quartile | 0.62* | 0.86 | 0.69 | 0.88 | 0.61 | 1.74† |
| Fourth quartile (highest) | 0.92 | 1.72† | 0.82 | 0.58* | 0.70 | 1.03 |
Note: Data for children in Mexico come from the Mexican Family Life Study; data for children in the United States come from the birth cohort of the Early Childhood Longitudinal Study. ref. = reference category; NA = not applicable because the subgroup was not at risk of transition.
Analysis of each dependent variable is restricted to individuals at risk of the specific transition. The number of cases is based on unweighted data and is rounded to the nearest 50. Children in cohabiting-couple families are at risk of transitions to both married and single families.
p < .10.
p < .05.
p < .01.
p < .001.
The results in column 2 of Table 4 show that there were no significant differences between migration status groups in the transition to a married-couple family. The relatively high share of children in Mexico who experienced this transition (see Table 3) was fully accounted for by covariates in the multivariate model. In contrast, there were large and significant differences in the transition to a single-parent family among children in two-parent families at t1 (see column 3 of Table 4). The likelihood of experiencing this transition by age 5 was less than one third as high for children in Mexico as for children of immigrants. On the other hand, the odds ratio for children of natives exceeded 2. It is important to note that the odds of experiencing the transition to a single-parent family were seven times higher for children of natives than for children in Mexico (2.16/.30 = 7.20) after controlling for child, mother, and family characteristics.
There were few consistent patterns for the covariates across the models for transitions in nuclear family structure. The most consistent predictor was nuclear family structure at t1. Not surprisingly, cohabiting families were the least stable, and married-couple families were the most stable. The odds of any type of transition, or the more specific transition to a married-couple family, were twice as high for children in cohabiting families as for children in single-parent families. The odds of shifting to a single-parent family were more than three times higher for children in cohabiting families compared to children in married-couple families. The likelihood of making any transition in nuclear family structure or the transition to a single-parent family was negatively associated with maternal age. Maternal education played little role in nuclear family transitions, except that transitions to marriage were considerably more likely among mothers with education beyond high school (odds ratio [OR] = 1.85).
The remaining columns in Table 4 provide results from logistic regression models for changes in extended family structure. Columns 4 through 6 focus, respectively, on any transition in extended family structure, the transition from an extended family household to a simple household, and the transition from a simple to an extended family household. The highly consistent pattern for migration status across the three models reveals that children in Mexico were significantly and substantially less likely to experience transitions in extended family structure than children of immigrants, net of background factors. The most striking contrast is in the transition from an extended family household to a simple household. The odds ratio for children in Mexico relative to children of immigrants was 0.13. Estimates from models in which the reference group was changed to children in Mexico (results not shown) also indicated that all three transitions in extended family structure were significantly more common among children of natives.
Another key finding for any transition in extended family structure is that instability was high for all types of extended family arrangements compared to nonextended families. Transitions were the most likely for children living in households that were extended both vertically and horizontally (OR = 9.21), followed by horizontally extended households (OR = 8.22) and vertically extended households (OR = 5.43). Again, other covariates had inconsistent patterns by transition type. Transitions from an extended to a simple household were more likely for children in married-couple families than children in single-parent families (OR = 1.88, in column 5), and transitions into extended families were less likely for children in cohabiting or married-couple families than for children in single-parent families (ORs = 0.33 and 0.31, in column 6).
Conclusion
Immigrants carry family histories and family-building strategies with them from their places of origin, but their postmigration family patterns are also influenced by the experience of migration and new realities in the destination setting. Knowledge of how parental immigration shapes children’s living arrangements has been limited by an overreliance on analytic strategies based on cross-sectional data collected in the destination alone. Our study extends prior research on Mexican-origin children by considering their family structure in both Mexico and the United States with longitudinal data that capture the dynamic nature of living arrangements across the preschool years. This period of life is important because early childhood circumstances have lasting effects on numerous outcomes that ultimately reach into adulthood.
Our analysis showed that shortly after birth, Mexican children of immigrants were similar to children in Mexico in terms of coresidence with both parents. Yet their parents were considerably more likely to be in cohabiting unions, a pattern with important implications for later family stability. In contrast, Mexican children with U.S.-born parents were distinct from the other groups in their greater likelihood of living in a single-parent family.
In regard to extended family structure, U.S.-born Mexican children with immigrant versus native parents were similar in one respect: About one third of each group lived with extended family members at ages 0–1. Still, the particular constellations of family members differed. Perhaps because of differences in kin availability, children of immigrants were less likely to live in vertically extended families and more likely to live in horizontally extended families than children of natives. Children in Mexico were the most likely to live in extended families, and the vast majority of such families were vertically extended.
An important finding was that the considerable stability in the aggregate distributions of children across nuclear and extended family structures at ages 0–1 and 4–5 is deceptive. The aggregate patterns mask substantial instability in the living arrangements of individual children and differences in instability across migration status groups. The most striking contrast was in the transition to a single-parent family. Only 3% of at-risk children in Mexico made this transition, compared to 7% of Mexican children of immigrants and 15% of Mexican children of natives. These differences persisted in multivariate models, with children of immigrants differing significantly from both other groups. Even though the group most likely to shift into a single-parent family (children of natives) had the highest levels of maternal education and employment, this early life transition represents a potential loss of resources for children.
Similarly, transitions in extended family structure were relatively common among Mexican-origin children in the United States. In Mexico, children in extended families at ages 0–1 were highly likely to remain in them at ages 4–5. Only 17% of those at risk experienced the transition to a nonextended family. This contrasts sharply with the United States where, regardless of parental nativity, 40% to 50% of Mexican children in extended families exited them by age 5. The contrast between children of immigrants and children in Mexico remains significant and large for all transitions in family extension in multivariate models.
Differences in the type of extension and the resources that are potentially available through it may be partly responsible for these patterns. Economic need, lack of housing, and norms are commonly offered as reasons for young parents to live in vertically extended households in Mexico (García & de Oliveira, 2011). Similar motivations undoubtedly exist for Mexican immigrants, but their living arrangements also reflect the process of international migration itself. Migrant networks span origin and destination communities, and family members of the same generation (e.g., siblings, cousins) often provide housing and other forms of assistance to new arrivals because older family members remain in the country of origin. Although some children may benefit from living in horizontally extended households, such arrangements are often less stable than vertically extended households. This instability may reflect the purpose of this household strategy. Horizontal extension may be accompanied by the expectation that it is a temporary solution to assist relatives who are in need, as well as tensions linked to defining obligations for assistance and indebtedness among siblings (Menjívar, 2000).
Recent studies of family structure and child outcomes have moved beyond a static view of living arrangements to emphasize the role of family transitions in children’s well-being (Fomby & Cherlin, 2007; Magnuson & Berger, 2009). Building on this emergent literature on complex family states and transitions, our longitudinal approach with binational data provides new information about the impact of immigration on the family lives of Mexican-origin children. There was strikingly less stability in the living arrangements of Mexican children of immigrants compared to children in Mexico. On every dimension except the transition to marriage, children of immigrants experienced significantly more family change than their counterparts in Mexico. There were fewer differences between U.S.-born children with immigrant and native parents. Nonetheless, the higher levels of single parenthood among the children of natives and the greater share making transitions from a two-parent family to a single-parent family suggest further erosion of family support for young children.
A limitation of our study is that the full complexity of migration between Mexico and the United States could not be considered. International migration is a selective process, and some differences between immigrant families and families in Mexico may be related to migration selectivity on unobserved characteristics. Moreover, some immigrants to the United States return to Mexico. We were unable to examine family structure among children of return migrants because there were too few such cases in the MxFLS. At the same time, our ECLS-B sample of U.S.-born Mexican children excluded those who had moved to Mexico by ages 4–5. This is unlikely to be a serious concern because relatively few U.S.-born Mexican children migrate to Mexico in the first years of life. Moreover, rates of attrition between ages 0–1 and 4–5 are similar for Mexican children with immigrant and native parents in the ECLS-B (authors’ calculations).
In summary, this study is the first to examine Mexican children’s nuclear and extended family structure using binational longitudinal data. Our research extends the literature in critical ways, but it also reveals opportunities for additional research. Future efforts should focus on disentangling the effects of migration selectivity from the influence of exposure to distinct conditions in origin and destination countries. This is a key methodological issue confronting all studies of immigrants’ outcomes. In addition, future studies should collect more information on family members residing outside the household, including their location and the types of support they provide. Binational research must also turn to the content of relationships among coresident family members and how the various family transitions described above influence the development of Mexican-origin children. These lines of inquiry clearly show considerable promise for advancing scholarship on the intersections among international migration, family structure, and child development.
Acknowledgments
Support for this research was provided by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (5P01HD062498-02) for the Mexican Children of Immigrants Program Project. The research also received support from the Population Research Institute and the Family Demography Training Program at The Pennsylvania State University, both of which are funded by the National Institute of Child Health and Human Development (R24HD041025 and T-32HD007514).
References
- Abbott A. The causal devolution. Sociological Methods & Research. 1998;27:148–181. [Google Scholar]
- Bianchi SM, Casper LM. Explanations of family change: A family demographic perspective. In: Bengston VL, Acock AA, Allen KR, Dilworth-Anderson P, Klein DM, editors. Sourcebook of family theory and research. Sage; Thousand Oaks, CA: 2005. pp. 93–117. [Google Scholar]
- Brown SL. Marriage and child well-being: Research and policy perspectives. Journal of Marriage and Family. 2010;72:1059–1077. doi: 10.1111/j.1741-3737.2010.00750.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fomby P, Cherlin AJ. Family instability and child well-being. American Sociological Review. 2007;72:181–204. doi: 10.1177/000312240707200203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- García B, de Oliveira O. Family changes and public policies in Latin America. Annual Review of Sociology. 2011;37:593–611. [Google Scholar]
- Glick JE. Connecting complex processes: A decade of research on immigrant families. Journal of Marriage and Family. 2010;72:498–515. doi: 10.1111/jomf.12621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Landale NS. Migration and the Latino family: The union formation behavior of Puerto Rican women. Demography. 1994;31:133–157. [PubMed] [Google Scholar]
- Landale NS, Thomas KJA, Van Hook J. The living arrangements of children of immigrants. The Future of Children. 2011;21:43–70. doi: 10.1353/foc.2011.0003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Magnuson K, Berger LM. Family structure states and transitions: Associations with children’s well-being during middle childhood. Journal of Marriage and Family. 2009;71:575–591. doi: 10.1111/j.1741-3737.2009.00620.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Menjívar C. Fragmented ties: Salvadoran immigrant networks in America. University of California Press; Berkeley: 2000. [Google Scholar]
- Mollborn S, Fomby P, Dennis JA. Extended household transitions, race/ethnicity, and early childhood cognitive outcomes. Social Science Research. 2012;41:1152–1165. doi: 10.1016/j.ssresearch.2012.04.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nobles J. Migration and father absence: Shifting family structure in Mexico. Demography. 2013;50:1303–1314. doi: 10.1007/s13524-012-0187-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Passel J, Cohn D, Gonzalez-Barrera A. Net migration from Mexico falls to zero—And perhaps less. 2012 Pew Hispanic Center Report. Retrieved from http://www.pewhispanic.org/2012/04/23/net-migration-from-mexico-falls-to-zero-and-perhaps-less/
- Rubin DB. Multiple imputation for nonresponse in surveys. Wiley; New York: 1987. [Google Scholar]
- Rumbaut RG. The ties that bind: Immigration and immigrant families in the United States. In: Booth A, Crouter AC, Landale NS, editors. Immigration and the family: Research and policy on U.S. immigrants. Erlbaum; Mahwah, NJ: 1997. pp. 3–46. [Google Scholar]
- Singley S, Landale NS. Incorporating origin and process in migration-fertility frameworks: The case of Puerto Rican Women. Social Forces. 1998;76:1437–1464. [Google Scholar]
- The Urban Institute The Urban Institute Data Tool. 2012 Retrieved from http://datatool.urban.org/charts/datatool/pages.cfm.
- Van Hook J, Glick JE. Immigration and living arrangements: Moving beyond economic need versus acculturation. Demography. 2007;44:225–249. doi: 10.1353/dem.2007.0019. [DOI] [PubMed] [Google Scholar]