Family Structure Patterns from Childhood through Adolescence and the Timing of Cohabitation among Diverse Groups of Young Adult Women and Men

Abstract

Family structure changes experienced by children are likely to shape their transitions into young adulthood, including the formation of their own romantic relationships. This study examined links between children’s family structure trajectories from childhood through adolescence and their timing of entry into cohabitation as young adults, a transition with implications for future relationship instability through adulthood. Repeated measures latent class analysis identified configurations of family structures and family structure changes from birth through age 15 among 10,706 young people in the National Longitudinal Survey of Youth 1979 Child and Young Adult Surveys. A Cox proportional hazard model then used the resulting classes to predict timing into cohabitation over the period from age 15 to age 38. Both timing of family structure transitions and the type of transitions (e.g., early transitioning into a stepfamily home) were associated with earlier entry into cohabitation. Notably, links between family structure trajectories and the timing of cohabitation differed by gender and race/ethnicity (Latinx, African-American, White), such as a faster entry into cohabitation by women who experienced early entry into stepfamily structures. Regardless of gender, Latinx and White young adults were faster to enter into cohabitation if they lived in a stepfamily structure during early childhood.

Although most children in the U.S. live with two married parents, those who do not live in a variety of family structures as their parents often re-partner over the course of their childhood. Experiencing their parents enter and exit marriages and cohabiting partnerships can shape children’s social development directly and set the stage for their adult lives (Afifi et al., 2015; Laumann-Billings & Emery, 2000; Weaver & Schofield, 2015). Indeed, children’s exposure to their parents’ relationship instability may be reflected in their own adult relationships in ways that disrupt their social mobility and healthy development (Cherlin, 2009; Johnson, Krahn, & Galambos, 2017; Lehrer & Son, 2017; Amato & Cheadle 2005). For example, children born to unmarried parents who make early transitions to nonresidential parental care, or who experience their parent’s divorce, are more likely to divorce as adults (Amato, 2014; Hoefferth & Goldscheider, 2015; Wolfinger, 2005).

Young adult union formation—or the initial entry of young people into committed relationships—is likely to play a role in the intergenerational transmission of relational instability. Indeed, a child’s exposure to various family structures and family structure changes may translate into young adult relationships that are less likely to endure and less likely to build a foundation for the future. If so, then young adult union formation is a mechanism in the intergenerational transmission of family structure (Crosnoe & Cavanagh, 2010; Cui & Fincham, 2010; Wolfinger, 2000; 2005; Amato, 1996).

In particular, contemporary historical changes in young adult union formation in the U.S. have positioned entry into first cohabitating relationship as a key way to consider the links among family structure trajectories early in life, the transition into adulthood, and inequality. The number of cohabiting couples has increased over the last 20 years, and cohabitation now represents the normative beginning of union formation in the U.S. For some, it is a path to marriage. For others, cohabitation is a substitute for marriage; for many more, it is a form of co-residential dating (Kennedy & Bumpass, 2008; Manning, 2015; Smock, 2000; U.S. Census Bureau, 2018). Notably, as cohabitation has become more common, the median age at which young women enter into cohabitation has fallen considerably into the early twenties while the median age for marrying has increased to the mid to late twenties—although the median age of cohabitation for young women of color and young men has not fallen. These age trends are meaningful because early cohabitation (i.e., cohabiting in the late teens through early twenties) often emerges from social and economic disadvantages growing up and can predict relationship instability through adulthood (Manning & Cohen, 2015; Ryan & Claessens, 2013; Sassler, Michelmore, & Qian, 2018).

This study, therefore, identifies common family structure patterns throughout childhood and adolescence in a contemporary sample of young adults and examines how their childhood family structure trajectories predict the timing of their first cohabiting union in young adulthood, with special attention to gender differences and racial/ethnic variation. These intergenerational linkages are examined by applying repeated measures latent class analysis to data from a representative sample of U.S. youth followed from birth through adulthood in the National Longitudinal Survey of Youth 1979 Child and Young Adult Surveys (NLSY79-CYA). Establishing these linkages can then lead to future research elucidating the underlying mechanisms. Such research can contribute to theoretical development by fine-tuning perspectives on the intergenerational transmission of family structure to provide details about the dimensions of family structure in childhood and adolescence that matter in the long term and honing in on the foundation of young people’s own relational trajectories in adulthood (Amato & Patterson, 2017). It can also identify specific groups of adults at elevated risk for relational instability and its associated risk to wellbeing (Johnson et al., 2017).

The Intergenerational Transmission of Family Structure

Figure 1 presents the conceptual model of this study. We sketch out its basic parameters before elaborating on the theoretical and empirical motivation for each one. Life course theory, based on the foundation that early life experiences lead to developmental trajectories, is especially useful for unpacking developmental processes over time and understanding how early life experiences may shape later adjustment and functioning (Elder, 1998). Following this theory, this model focuses first, on a key developmental ecology that demonstrates continuity and change over a long period (i.e., dynamic family structure experiences that are linked to developmental disruptions during childhood and adolescence; Ryan & Claessens, 2013); second, on the links of these ecological pathways to a key life transition, a concrete change in role and setting that is conditioned by timing and forecasts developmental disruptions during adulthood (i.e., early entry into the first cohabiting relationships in young adulthood; Smock & Greenland, 2010); and third, on the embeddedness of the interplay of trajectory and transition within the gendered and racialized experience of family structure and relationship formation in the U.S. (i.e., differences between boys and girls and/or among White, African-American, and Latinx youth; Cavanagh & Fomby, forthcoming).

Figure 1.

Conceptual model of the intergenerational transmission of family structure through early cohabitation in young adulthood

The clearest empirical support for this conceptualization is evidence that the odds of divorce are higher among adults whose parents divorced (Wolfinger 2005; Amato & Patterson, 2017). Decades of studies have deepened understanding of this intergenerational association. One key insight is that it is driven less by parental divorce than by family structure experiences that often accompany parental divorce, such as being born outside of marriage and/or living with a single parent (Amato & Patterson, 2017; Teachman, 2002). Another key insight is that the timing of transitions into adult relationships mediates the association between parental divorce growing up and one’s own divorce (Amato, 1996; Amato, 2010). The earlier in life that people enter a committed relationship, the less likely they are to remain in it over time (Bumpass, Martin, & Sweet, 1991). This timing effect potentially reflects less informed decision-making about relationships (Oppenheimer, 1988), greater likelihood of being poorly matched with a partner (Lehrer, 2008), and heightened economic insecurity (Kennedy & Bumpass, 2008).

Building on this literature within a life course framework, this study extends the developmental window through early childhood and adolescence (periods in which family dynamics have long-term implications for young people’s romantic lives; Collins, Welsh, & Furman, 2009), conceptualizes a more multidimensional view of these family dynamics that recognizes different ways family structure is developmentally significant (Teachman, 2002), focuses on a young adult relationship transition—earlier entry into cohabitation—that signals particular vulnerability in future relationship trajectories (Kuperberg, 2016), and considers how that same family structure pattern can be experienced differently—with varying long-term consequences—by young people from diverse walks of life (Cavanagh & Fomby, forthcoming).

Conceptualizing Family Structure Trajectories

Beginning with the multidimensional conceptualization of family structure during the early life course, family structure—typically defined in terms of a residential mother’s union status—is a dynamic context of development threading together various types of family experiences over time (Osborne et al., 2007; Osborne & McLanahan, 2007). Commonly discussed family structures include a child living with two biological parents (married or cohabiting), a biological mother and a partner who is not the child’s father (again, married or cohabiting), and living alone with a biological mother, all of which can be compared to children who are in non-parental care altogether or who are in other types of family structures (Crosnoe et al., 2014; Lee & McLanahan, 2015). These family structures can be mixed and matched in various combinations and durations within and across developmental periods. For example, two children may experience a parental divorce followed by time spent in a single parent home but do so at different developmental stages, affecting how they process what is happening and how long they are exposed to inter-parental conflict and economic hardship. As another example, two children may live in a stepfather family structure, but one lives with a stepparent for two years while the other lives with a series of stepparents over 14 years. Those differences then reduce or exacerbate changes in resources and family processes. More fully capturing such complexity over long periods is crucial to elucidating the oft-discussed developmental significance of family structure (Teachman, 2002; Cavanagh, 2008). This study tackles this dynamic complexity through childhood and adolescence along three inter-related dimensions, each of which is relevant to the potentially consequential timing of the first cohabiting relationship in young adulthood.

The first dimension is the amount of family structure change that young people experience growing up. Such changes can shape relational decisions and behaviors in young adulthood by altering youths’ conceptions of romantic partnerships (e.g., something to be desired or avoided; Amato & DeBoer, 2001), affecting expectations for what such partnerships entail (e.g., marriage vs cohabitation), and facilitating the development of behaviors and attributes that shape future union formation (e.g., problematic coping skills; Smock, 2000). In line with these mechanisms, we know from past research that the number of family structure changes experienced early in life is associated with younger entry into first unions, perhaps as young people seek out stability and support but are not well-positioned to secure it (Ryan, et al., 2009).

The second dimension is the timing of family structure transitions. In general, a change in family structure, no matter the type, is more developmentally significant when occurring earlier in life (Ryan et al., 2009). For example, family structure transitions in early childhood, before school entry, predict behavioral problems later in childhood more strongly than transitions that occur later in childhood or adolescence. Especially for boys, family instability before school entry seems to alter how children navigate social relationships during elementary school (Cavanagh & Hutson, 2008; Osborne & McLanahan, 2007). In turn, such social experiences may lead to a trajectory of problematic peer relationships throughout adolescence and young adulthood. These timing effects could reflect links between early family experiences and children’s attachment-related templates and orientations, greater developmental plasticity, and/or age-related differences in maturation, coping skills, and access to extrafamilial support systems, all of which are relevant to how young people approach relationship formation across the transition to young adulthood (Cavanagh & Huston, 2008; Ryan & Claessen, 2013).

The third dimension is the type of family structures in which young people spend time. Two family structures are particularly important to consider. First, entry into and time spent in stepfamilies can be developmentally disruptive, regardless of socioeconomic circumstances (Sweeney, 2010). Scholars suggest that roles and expectations in a newly constructed stepfamily unit are not well defined. These unclear boundaries can undermine family cohesion and consistent parenting (Coleman, Ganong, & Fine, 2000). For example, a stepfather may invest less time scaffolding the academic work of a stepchild or monitoring their behavior. Second, entry into and time spent in single parent families can have similar developmental significance but for different reasons. Key mechanisms include loss of social support, instrumental assistance, and income when a parental partner leaves the home (Osborne et al., 2012). At the same time, single parents often engage in romantic relationships, thereby actively modeling dating roles (Whitbeck, Simons, & Kao, 1994). Still, longer duration in both stepparent and single parent families can provide the opportunity for a new “normal’ to emerge”, allowing adjustment that eases developmental disruptions associated with transitions to stepfamily and single parent households (Amato, 2001; Osborne & McLanahan; 2007).

Notably, these three dimensions—change, timing, type—can converge within dynamic family structure patterns to shape child and adolescent development. The conceptual model contends that this dynamic convergence encourages young people to seek out relational security while not sufficiently preparing them for relationship-formation, speeding up their transitions into cohabitation. The first aim of this study is to test this idea; specifically, that family structure patterns over the early life course characterized by instability lead to transitions into cohabitation earlier in young adulthood, especially when instability occurs early in childhood and involves time in both stepfamily and single parent families. Of course, these associations are not entirely causal. Factors like parental education and poverty status both select young people into different family structure patterns and contribute to the likelihood that they experience instability in their own relationships (Fomby & Cherlin, 2007; Lee & McLanahan, 2015) and so they will be taken into account as we pursue this aim.

Conceptualizing Differential Impact

The conceptual model contextualizes the link between multi-dimensional family structure patterns and earlier young adult entry into cohabitation following the emphasis of life course theory on broad social structures that shape expectations, constraints, and opportunities (Elder, 1998). This macro-level contextualization is defined here by two key elements of social stratification and culture that locate young people within a society: gender and race/ethnicity.

First, family structure trajectories may have different influences on young men and women’s relationship trajectories as they enter into adulthood. Although young women tend to be more interpersonally oriented and more invested in relationships than young men, evidence clearly suggests that both the developmental significance of family structure change (Cavanagh & Fomby, forthcoming; Cavanagh & Huston, 2008; Cooper, McLanahan, & Meadows 2009) and ecological influences on relationship formation are stronger for young men than young women (Cavanagh, Crissey, & Raley, 2008; Crockett & Randall, 2006). Part of this difference is likely tied to the social construction of gender and the ways boys and girls interpret and compensate for changes in parental relationships (Maccoby 1998). Girls, for example, often have intimate friendships with peers whom they can turn to for support in the face of family change. Because boys have fewer intimate friendships, they may have fewer resources to draw upon and may be more likely to compensate for diminished parental support and closeness through romantic relationships. This need for connection may also encourage romantic and sexual relationships at earlier ages (Way, 2013). Thus, the kinds of family structure patterns highlighted in the conceptual model are more likely to hasten young men’s transitions into cohabiting relationships even if young men, on average, are overall less likely to enter (or seek to enter) into such relationships during this period (Manning, Brown & Stykes, 2015).

Second, race/ethnicity is likely to moderate links between family structure growing up and young adult relationship formation. To be sure, race/ethnicity is related to the likelihood of being born into specific family structures and of transitioning between family structures throughout childhood and adolescence, with African-American children more likely to experience particular family structures (e.g., single parent homes) and family structure changes (Fomby & Cherlin, 2007; Osborne, Manning, & Smock, 2003). Although less often studied, Latinx family structure patterns seem to be similar to those of White families, with the exception that their cohabiting relationships tend to be more marriage-like and durable (Landale, Oropesa, & Bradatan, 2006; Osborne et al., 2003). Still, recent evidence suggests that family structure and instability tend to matter most for White youth (Fomby & Cherlin, 2007; Fomby, Mollborn, & Sennott, 2010; Crosnoe & Wildsmith, 2011). For example, family structure change is linked with age at sexual debut and the timing and martial context of childbearing for White but not Black women (Wu & Thomson 2001; Wu & Martinson, 1993); associations among Latinas are less consistent. This heightened susceptibility of White youth relative to others may be tied to their socioeconomic advantage. Relative to Whites, people of color are more likely to experience other types of disadvantage (e.g., economic volatility, housing instability, parental incarceration) (Fomby et al., 2010; Turney & Wildeman 2013). Because White children experience fewer stressors overall, family structure change may be especially disruptive, in part because it is less frequent (Turney, 2015). Similarly, differences in the emotional and instrumental support provided to parents and children from networks of kin and kin-like figures can also play a role, better insulating children of color during periods of change. The organization of families extends beyond the nuclear household and is more likely to involve extended kin households (Sarkisian, 2007). In contrast, White children are less likely to have this buffer of extended family support (Fomby et al., 2010). Thus, if certain family structure patterns during childhood and adolescence lead to early cohabitation, we expect they will likely do so more for White youth even if they are less likely to experience more complex family structure histories.

The second aim of this study, then, is to test the moderation component of the conceptual model involving gender differences and racial/ethnic variability in the links between dynamic and multi-dimensional family structures patterns from birth through adolescence and the timing to first cohabitation during young adulthood. This moderation is expected to take the form of stronger links among young men and White youth.

Method

Dataset and Participants

Sponsored by the U.S. Bureau of Labor Statistics (BLS), the National Longitudinal Survey of Youth 1979 (NLSY79) is a publicly available, nationally representative sample of youth born in the U.S., followed annually until 1994 and then biennially. Beginning in 1986, the Children of the National Longitudinal Survey of Youth (C-NLSY) biennially collected data from the children born to women in the NLSY79. Children were then included in the National Longitudinal Survey of Youth-Young Adult survey (NLSY79-YA) once they were 15 years old. This study, Identifying the Optimal Levels and Timing of Family and School Influences on Child and Youth Development, was determined that it did not meet the criteria for human subjects research as defined by the Common Rule (45 CFR 46) or FDA Regulations (21 CFR 56) by the University of Texas at Austin Institutional Review Board based of the secondary use of deidentified data with no direct or indirect links to identifiers.

Given the focus on the years from birth through young adulthood, this study used data from the C-NLSY and NLSY-YA (i.e., NLSY-CYA) through the most recent wave in 2014. Because respondents had an age range up to 26 years, we restructured the data so that they were organized by respondents’ ages rather than by the year of the survey. The full NLSY-CYA sample (n = 11,512) for the Cox proportional hazards models was restricted to respondents who were between the ages of 15–38 during the observation period. For this analysis, only young adults who were old enough to have graduated high school were retained in the sample. This restriction resulted in a final n of 10,706 for this study. Of note is that this necessary filter reduced the cell size for one class (defined in terms of spending time outside of parental care during childhood or adolescence, see below) too much (n = 173) to support interpretation of hazard ratios or coefficients for this class. We did not drop this class from analyses but also do not offer substantive interpretations for this group of young adults.

Measures

Age at first cohabitation.

Beginning in 1994, NLSY79-CYA respondents reported if they had ever cohabited (“Have you ever lived with someone of the opposite sex to whom you were not married?”). If yes, respondents were asked what month and year they first started cohabiting. From 2000 onward, they reported whether they had cohabited since the previous interview and the month and year the cohabitation began and ended (if applicable). A variable measuring age at first cohabitation was then created using these data from age 15 to age 38. Table 1 presents descriptive statistics for these variables and for all other study variables.

Table 1:

Study demographics means and percentages

	% for Non-Continuous Variables, M (SD) for Continuous Variables
	All	Consistent two biological parent home	Consistent single biological parent home	Early transitioning into single biological parent home	Late transitioning into single parent home	Early transitioning into stepfamily home	Late transitioning into stepfamily home	Returning to two biological parent home	Transitioning from parental into non-parental care
	(n = 10,702)	(n = 4027)	(n = 1608)	(n = 1116)	(n = 1050)	(n = 883)	(n = 734)	(n = 905)	(n = 379)
Young adult characteristics
Gender
Female	0.49	0.48	0.48	0.45	0.49	0.53	0.53	0.51	0.48
Male	0.51	0.52	0.52	0.55	0.51	0.47	0.47	0.49	0.52
Race/ethnicity
Latinx	0.19	0.18	0.13	0.18	0.25	0.20	0.23	0.24	0.22
African-American	0.28	0.13	0.59	0.41	0.19	0.25	0.34	0.28	0.39
White	0.53	0.69	0.27	0.41	0.56	0.55	0.43	0.48	0.39
Education attainment
Less than high school	0.19	0.10	0.27	0.25	0.17	0.22	0.25	0.23	0.38
High school	0.81	0.90	0.73	0.75	0.83	0.78	0.75	0.77	0.62
Age at first cohabitation	21.3 (3.5)	21.9 (3.2)	21.1 (3.4)	21.3 (3.8)	20.8 (3.1)	20.8 (3.4)	20.9 (3.5)	21.4 (4.0)	20.2 (3.3)
Percent who cohabited	0.41	0.38	0.61	0.63	0.70	0.69	0.69	0.76	1.00
Number of maternal partner transitions	2.14 (2.28)	0.45 (0.78)	1.77 (1.58)	3.54 (2.30)	2.23 (1.32)	3.92 (2.17)	4.23 (2.46)	4.36 (2.31)	3.94 (2.52)
Maternal characteristics
Education
No college degree	0.81	0.73	0.89	0.87	0.70	0.87	0.86	0.93	0.95
College +	0.19	0.27	0.11	0.13	0.30	0.13	0.14	0.07	0.05
Age at childbirth	24.3 (5.1)	25.9 (4.1)	23.8 (4.8)	22.9 (5.4)	28.1 (5.3)	21.9 (4.2)	21.8 (4.6)	21.3 (4.6)	21.3 (4.9)

Family structures through childhood and adolescence.

The structures of each child’s family of residence from ages 1 through 15 were tracked using information from demographic surveys given at each wave about the residence of the child and their mother’s cohabiting partner(s). The residence variable was assessed at each wave and provided information regarding with whom the focal child most often resided. If children were adopted before 12 months of age, they were coded as living with biological parents. For each year, we used variables identifying whether the child lived with her or his single parent, two biological parents, or two-parent step or social (mother could have been legally married or cohabiting with her partner) parents. An additional category identified children who were living with other kin or who were in foster or long-term care. For each child age, we created a four-category family structure variable (child lived with both biological parents, single biological parent, biological parent and their partner, someone besides a biological parent). The 1986 and 2006 waves added family residence options. We incorporated these new options into the four levels as appropriate (e.g. temporary living arrangement or reporting homelessness were coded as “other”). Only 1% of participants reported being in any of the additional categories from 1986 to 2014.

Covariates.

Demographic information from the first wave of data collection included child gender (49% female; 51% male), racial/ethnic identification (19% Latinx; 28% African-American; 53% White), maternal education, respondent education, and maternal age at birth of first child (M = 24.3, SD = 5.08). Maternal reports of education were categorized as no college degree (81.25%) or college degree or higher (18.75%). Respondent’s own education was limited to whether they had a high school diploma or GED (80.89%) or not (19.11%) given the youngest were 19 at the last wave, and therefore, not able to complete college in the given time frame.

Plan of Analysis

The first stage of analysis involved the identification of children’s family structure trajectories derived from repeated measures latent class analysis (LCA), a type of LCA using longitudinal panel data. Repeated measures LCA is a person-centered analysis that groups sample members based on the configuration of pre-specified traits. People are not hard-classified into groups but instead given a probability distribution to determine probable class membership. There are two parts to the LCA: the measurement model and structural model (Masyn, 2013).

The measurement model used all variables denoting children’s family structures each year from age 1 to age 15 to inform a latent variable. LCA parameters were estimated using an expectation maximization (EM) algorithm and maximum likelihood (ML) estimation (using MPlus 8.2). Multiple sets of parameter estimates were tested iteratively until a maximum number of iterations had been reached or the model converged. We used log likelihood, BIC, Lo-Mendell-Rubin (LMR), and the bootstrapped likelihood ratio test (BLRT) to determine the model that best fit the data. The LMR was significant through eight classes, and the BLRT remained significant (likely due to sample size). After the four-class model, the smallest class was 3% of the overall sample. The AIC, BIC, and loglikehood continued to trend downward through nine classes. Based on this information and consideration of prior research, we chose the eight-class model (Figure 2). Each class still had a substantial number of people (smallest class n = 391), and, substantively, the eight-class model allowed examination of timing while maintaining excellent model fit. Approximate group sizes are given because one defining feature of LCA is that grouping reflects individuals’ proportional odds of being in one group over another. For example, a participant might have an 85% chance of being in class 1, a 5% change of being in class 3, and a 10% chance of being in class 5. When forced to hard-classify, this person would be defaulted to class 1 but LCA allows that individual to be proportionally represented in all 3 classes.

Figure 2:

Observed family structure trajectories from childhood through adolescence

Next, the structural model explored associations between the family structure variables derived through the LCA and young adults’ timing to first cohabitation in a Cox proportional hazard model (in Stata 15.1). A single multivariate model was estimated, including young adult and parental controls as well as LCA classes. Additional models were estimated post hoc that rotated the reference category to better understand the links between family structure histories and timing of first cohabitation. Beginning at age 15, all respondents became “at risk” of entering a cohabitating union. Young adults who did not enter into a cohabitation by the last interview date (age 38) were censored, including those who went directly into marriage and those who did not cohabit or get married. Censoring occurs when an individual has not experienced the event of interest (cohabitation) but their information is still used in model estimation. A final step interacted children’s family structure classes with gender in one model and race/ethnicity in another model, with contrast analyses testing for significant differences among all possible pair-wise combinations, including within and between classes. We conducted the contrast analyses using the pwcompare and margins commands in Stata 15.1. These analyses calculated the coefficient (and odds ratio) for all possible pair-wise combinations between the family structure history variable (categorical) and the moderator (gender or race/ethnicity).

Consistent with NLSY79 item nonresponse data, there was less than 2% missing data for children’s family structure from year to year. For the LCA, missing data were accounted for with full information maximum likelihood (FIML) in Mplus, which estimated all missing values through multivariate estimation techniques (Muthén & Muthén, 2017). The hazard model included three variables (maternal age, child gender, child race) with less than 1% missing data. Maternal education had the most missing data (35% in last wave). The hazard analysis was weighted to account for differential attrition across waves.

Repeated measures LCA estimates latent groups of people based on response patterns across a single categorical variable (i.e. family structure) over time. The resulting latent classes were, therefore, not observed family structure transitions but instead changes in probabilities of children’s family structures over time. As such, they suggest probable family structure trajectories. Accordingly, we labeled the latent classes by the predominant pattern and probable trajectory of family structure transitions across ages. We believe that this language is consistent with the conceptualization of family structure trajectories and is more easily interpretable.

Results

Family Structure Trajectories from Birth through Adolescence

LCA models identified eight common family structure patterns as defined by whom children lived with from ages 1 through 15: consistently two biological parent home, consistently single biological parent home, early transitioning into a single biological parent home, late transitioning into a single biological parent home, early transitioning into stepfamily home, late transitioning into stepfamily home, returning to two biological parent home, transitioning from parental care to non-parental care. Table 1 presents demographic characteristics for each class.

The first and largest class, consistently two biological parent home, included children with high odds of growing up with both biological mother and father. Any transitions experienced by children in this class occurred very early in childhood and usually involved a transition into a two biological parent family structure (see panel 1, Figure 2). In many ways, these children were socioeconomically advantaged, with relatively high levels of maternal college completion rates (73%) and a high percentage of high school completion rates (90%).

Youth in the second class, consistently single biological parent home, mostly grew up with single mothers. A small percentage experienced a transition from a family structure with two biological parents to a single parent home in early childhood but subsequently were likely to remain stable in a single parent home (see panel 2, Figure 2). Few mothers had a college degree (11%), and almost three quarters of respondents graduated from high school.

The third class, early transitioning into a single biological parent home, included young people who lived in a two biological parent home in early childhood and were likely to have transitioned into a single parent home by middle childhood but were not likely to experience any further transitions (see panel 3, Figure 2). African-American children were over-represented compared to the overall sample (Table 1, column 2), but less so than in the consistently single biological parent class. Three-fourths of respondents had a high school diploma (74%), similar to the consistently single biological parent class.

The demographic profile of the fourth class, late transitioning into a single biological parent home, looked substantially different than the consistently single biological parent or early transitioning into a single biological parent classes (see panel 4, Figure 2). The majority of young people in this class spent their early childhood with both parents, but that proportion began to drop from age 10 to 11 when a larger proportion began residing in a single biological parent home. An overrepresentation of Latinx children (25%) were in this class compared to the overall sample (Table 1, column 2), the consistently single biological parent class, and the early transitioning into a single biological parent class. Their mothers were the oldest at first childbirth (M = 28.1, SD = 5.3), almost 5 years older than mothers in the previous two classes. Mothers and children in the late transitioning into a single biological home class were more similar to those in the consistently two biological parent home class than mothers and children in the consistently single biological parent class or the early transitioning into a single biological parent class.

Children in the fifth class, early transitioning into stepfamily home, were more likely to start out in either biological mother or stepfamily structures during early childhood (see panel 5, Figure 2). In middle childhood, almost 90% of children in this class were in a stepfamily family. By age 15, 70% were still in this family structure while 20% were in single parent family structures. The racial/ethnic composition of this class mirrored the overall sample.

The sixth class, late transitioning into a stepfamily home, included young people who spent a majority of their childhood in single biological parent or two biological parent homes, but a large proportion transitioned into a stepfamily home during late childhood (see panel 6, Figure 2). Compared to the full sample (Table 1, column 2), the racial/ethnic representation of this class was more weighted to Latinx (23%) and African-American (34%) children than Whites (43%). In other words, Latinx and African-American youth made up a larger proportion of this class than of the full sample, even though they were still not the largest groups in the class. Young people’s mothers were in their early 20’s (M = 21.8, SD = 4.6) and 14% had at least a college degree. By age 15, around 80% of children in this class were in stepfamilies and 20% were in single parent families. While the end result at age 15 was similar to those in the early transitioning into stepfamily home class, their early trajectories were different. Children in this class spent more time in other family structures before transitioning into a stepfamily.

Young people in the seventh class, returning to two biological parent home, lived in single biological parent and two biological parent homes during childhood and a majority transitioned back into two biological parent homes during middle childhood (see panel 7, Figure 2). By age 13, over 90% were in two biological parent homes, similar to the children aged 13 in the consistently two biological parent home class, albeit with different histories and demographic biographies. In this class, Latinx children were overrepresented (24%), the representation of African-American children mirrored the overall sample (28%), and White children were underrepresented (48%) relative to their respective proportions of the full sample. Although children in this class did not appear as socioeconomically advantaged as children in other classes, they were more on par in terms of their own education – at least up until high school.

Finally, young people in the transitioning from parental care to non-parental care class spent a majority of their childhood in either a single parent or two biological parent family structure and by middle childhood a large proportion of children were living in family structures without either of their biological parents (see panel 8, Figure 2). Non-parental care consisted mostly of relative care (e.g., grandparent, aunts, uncles) or foster care. Their mothers were in young when they had their first child (M = 21.3, SD = 4.9), and only 5% had a college degree. This class was the smallest, but it was stable. It did not change from the 4-class model to the 8-class model, suggesting that, despite its small size, it was unique from the rest of the sample.

This person-oriented approach identified groups of young people who experienced family structure trajectories that were similar in some ways but differed in others. The primary distinctions involved type of family structure, consistency of a family structure vs. family structure transitions, and the timing of such transitions. Two classes might be similar on one or even two of these dimensions but differ on the third (e.g., two classes transitioned into stepfamily homes but at different times, multiple classes involved transitions out of two biological parent structures but differed in timing and the subsequent family structure).

Linking Family Structure Trajectories to the Timing of Young Adult Cohabitation

Having identified the most common multi-dimensional family structure trajectories through childhood and adolescence, the next step was to examine how such histories were differentially associated with the timing to first cohabiting union early in adulthood. Hazard models were weighted to account for longitudinal attrition. We further conducted an additional attrition analysis (logistic regression) of young adults who were lost from the first wave to the last wave to address generalizability of findings. Overall, young men were more likely to be lost to attrition as well young adults of either gender whose mothers had less education.

Table 2 presents results from Cox proportional hazard models. Hazard ratios (HR) over 1, such as 1.60, can be interpreted as indicating that membership in LCA class X is associated with a 60% increase of entering earlier into a cohabitation. HR results below 1, such as .70, can be interpreted as indicating that membership in LCA class X is associated with a 30% decrease of earlier entry. Young people in the consistently two biological parent class serve as the reference category in the first column; subsequent columns include associations when the reference category was rotated. For example, young people in the late into single parent class did not significantly differ from those in the consistently two biological parent class (HR = 1.08) or early into single parent (HR = .92) class, but they were significantly slower to enter cohabitation than those in the consistently single biological parent (HR = .84) class, net of other factors.

Table 2:

Rotated results from hazard models predicting timing to first cohabitation in young adulthood by family structure trajectories from childhood through adolescence

	HR (SE)
Family structure class reference category:	Consistent two biological parent home	Consistent single biological parent home	Early transitioning into single biological parent home	Late transitioning into single parent home	Early transitioning into stepfamily home	Late transitioning into stepfamily home	Returning to two biological parent home
Family structure trajectories
Consistently two biological parent
Consistently single biological parent	1.28 (.07)***
Early into single biological parent	1.17 (.07)**	0.91 (.06)
Late into single biological parent	1.08 (.07)	0.84 (.07)*	0.92 (.08)
Early into stepfamilies	1.38 (.07)***	1.08 (.07)	1.19 (.07)*	1.28 (.08)**
Late into stepfamilies	1.28 (.08)***	1.00 (.07)	1.10 (.07)	1.19 (.08)*	0.93 (.07)
Returning to two biological parent home	1.01 (.07)	0.78 (.07)***	0.86 (.07)*	0.93 (.08)	0.73 (.08)***	0.78 (.08)**
Non-parental care	1.33 (.15)*	1.04 (.11)	1.14 (.11)	1.24 (.12)	0.96 (.12)	1.04 (.12)	1.33 (.12)*
Young people characteristics.
Male	0.77 (.03)***
Race/ethnicity
Latinx	0.89 (.04)**
African-American	0.76 (.03)***
No high school degree	0.80 (.04)***
Young adult age	1.04 (.01)***
Maternal characteristics
No college degree	0.70 (.03)***
Age at childbirth	0.95 (.01)***

Note:

^*p < .05,

^**p < .01,

^***p < .001.

Consistently two biological parent family structure was the comparison group for the original model. Post hoc analyses were conducted to rotate the reference group. Rotated results are in columns 3–8.

Consistent with the hypothesis that transitions into stepfamilies (in combination with time in single parent families) would result in entry into cohabitation at younger ages, young adults who were likely to experience an early transition into a stepfamily structure based on LCA class membership (early transitioning to stepfamily home) were the fastest (i.e., youngest) to transition into a cohabiting relationship. For example, young adults in the early transitioning to stepfamily home were 38% more likely to enter into a cohabitation earlier, or faster, than those in the consistently two biological parent home. Young people in the early transitioning to stepfamily home class were also significantly more likely to enter into a cohabitation at an earlier age than young adults in the early and late into single parent family classes. Young adults in the late into stepfamily class were quicker to transition into cohabiting relationships compared to young adults in the consistently two biological parent and late transitioning into single parent classes.

Partially supporting the hypothesis that more family structure transitions would lead to early entry into cohabitation, young adults in the consistently two biological parent home (few to zero transitions) and the returning to two biological parent home (multiple transitions) were the slowest (i.e., oldest) to enter into a cohabitation. Young adults in the returning to two biological parent home were 27% less likely to enter into a cohabitation during the risk period than those in the early transitioning to stepfamily home.

Demographic characteristics of young adults and their mothers operated in expected ways, with young adults of color and young adults without a high school degree slower to transition to cohabiting unions relative to others whereas older young adults were more likely to transition early. Older maternal age at first birth and more maternal education were both associated with slower transitions to cohabitation.

Moderators of Links between Family Structure Trajectories and First Cohabitation

The final step was to examine whether the associations between family structure trajectories in the early life course and timing of first cohabiting relationship in young adulthood differed by gender and varied by race/ethnicity (see Table 3). Interactions between the family structure classes and gender suggested such gender differences, although they did not clearly support the hypothesis for more pronounced links among young men (B = −.22, p < .001). Young women seemed to be more affected by the timing of family structure transitions while young men were affected by the type of family structure transitions. Young women following the early transitioning into stepfamily home trajectory were among the youngest to enter cohabitation (Figure 3, panel 5), compared to young women following all other family structure trajectories except for those who followed the consistently single biological parent home (Figure 3, panel 2) and transitioned into non-parental care (Figure 3, panel 8) trajectories. Young women in the early transitioning into single parent family (Figure 3, panel 3) class were older upon cohabitation than young women in the consistently single biological family class (B = −.19, p < .05). For young men, significant differences among family structure trajectories clustered around single parent and stepfamilies compared to families with two biological parents, suggesting that, for them, family structure was more important than timing of the family structure transition. Indeed, those in the consistently single biological parent, early transitioning to single parent home, early transitioning to stepfamily home, and late transitioning to stepfamily home were all significantly younger when they entered into their first cohabiting relationships than young men in the consistently two biological parent home class. Turning to between gender comparisons, women are consistently faster to enter into cohabitating relationships than men. These differences are depicted in the individual graphs in Figure 3. Specifically, women enter into cohabitating relationships significantly faster than men in the consistently two biological parent, consistently single biological parent, late transitioning to single parent, and early to stepfamily family structure history trajectories.

Table 3:

Moderation results from hazard models predicting timing to first cohabitation in young adulthood by family structure trajectories

	Model A (gender)		Model B (race/ethnicity)
	Coefficient	SE	Coefficient	SE
Family structure (ref: Consistently two biological parent home)
Consistently single parent	0.32***	−0.07	0.38**	0.11
Early into single parent	0.13	−0.08	0.35***	0.09
Late into single parent	0.08	−0.09	0.04	0.09
Early into stepparent	0.43***	−0.09	0.37***	0.09
Late into stepparent	0.22*	−0.09	0.35***	0.09
Returning to two biological parent home	−0.03	−0.1	−0.13	0.10
Non-parental care	0.34*	−0.17	0.79**	0.23
Child gender	−0 22***	−0.06
Latinx young adults			−0.08	0.07
African-American young adults			−0.11	0.08
Family structure	ref: men		ref: Latinx
Consistently single parent	0.17*	0.07	0.26*	0.12
Early into single parent	0.18*	0.08	−0.05	0.13
Late into single parent	0.06	0.09	0.12	0.12
Early into stepparent	0.21*	0.08	0.38**	0.13
Late into stepparent	0.29**	0.09	0.24*	0.13
Returning to two biological parent home	0.04	0.09	−0.12	0.13
Non-parental care	0.24	0.15	0.70***	0.19
Family structure			ref: African American
Consistently single parent			0.09	0.09
Early into single parent			0.01	0.10
Late into single parent			0.08	0.13
Early into stepparent			0.11	0.11
Late into stepparent			0.03	0.12
Returning to two biological parent home			0.17	0.11
Non-parental care			−0.21	0.17

Note:

^*p < .05,

^**p < .01,

^***p < .001.

Both analyses controlled for maternal education, maternal age at childbirth, respondent education, respondent age at last interview and either race/ethnicity (Model A) or gender (Model B). Post hoc analyses were conducted to rotate the reference group.

Figure 3:

Timing to cohabitation by family formation history and gender

Note: The y-axis represents the percentage of young adults who have entered in to a cohabitation. The x-axis represents the age at which cohabitation first occurred.

When examining racial/ethnic variability (see Table 3), evidence partially supported the hypothesis of stronger associations for White young adults. Their family structure trajectories tended to more strongly predict cohabitation timing compared to their African-American peers. Latinx young adults, for the most part, trended with White young adults, with notable exceptions as described below. One pattern stood out for Latinx young adults. Latinx young adults who were predominately in early transitioning to stepfamily homes were among the youngest to enter into cohabitation, compared to their Latinx peers who were in the consistently two biological parent (B = 0.38, p < .001), early transitioning into single biological parent (B = 0.43, p < .05), and returning to two biological parent (B = −.50, p < .001; Figure 4) classes. In support of the hypothesis, White young adults appeared most reactive to family structure transitions and time spent outside of two biological parent homes. White young adults in the classes who spent the longest time in two parent family structures (i.e., consistently two biological parent home, returning to two biological parent home, and late transitioning into single parent home) were among the slowest to enter into a cohabitation compared to all other classes (Figure 4).

Figure 4:

Timing to cohabitation by family formation history and race/ethnicity

Note: The y-axis represents the percentage of young adults who have entered in to a cohabitation. The x-axis represents the age at which cohabitation first occurred.

Ancillary Analyses

Of young adults who entered into a relationship during the study time frame, 82% entered into a cohabitation as their first relationship, and 18% entered into a marriage as their first relationship (50% and 11%, respectively, for the full sample). Thus, our focus on cohabitation could have obscured evidence about links between family structure trajectories growing up and transitions into young adult relationships. To that end, we also estimated Cox proportional hazard models for timing to first marriage.

These results, which are available from the authors, revealed that seven classes predicted earlier entry into marriage than the consistently two biological parent class (note: the only non-significant association was for the early transitioning to stepfamily class). The main take-away from these ancillary analyses—that growing up with both parents leads to delays in marriage—is consistent with the prior literature. It is far simpler and less informative than studying timing to first cohabitation, which is now the normative first relationship.

Discussion

Family structure dynamics from childhood into adolescence can influence the transition into young adulthood (Cavanagh, 2008). Research suggests that experiencing a parental divorce, the total number of family transitions, and family structure itself can influence trajectories of relational transitions in young adulthood (Amato & Patterson, 2017; Sassler et al., 2009), offering a window into how relational patterns carry over from generation to generation. This study used life course theory to dig deeper into these intergenerational family dynamics, highlighting specific family structures and family structure transitions over time, the timing of such transitions, and a key young adult relational experience (early timing to first cohabitation) with clear implications for relational advantages and disadvantages across adulthood. Given that young men’s engagement in intimate relationships tend to be more responsive to the family context as well as the relative disadvantage associated with changing family structures for White youth (Cavanagh & Fomby, forthcoming; Crosnoe & Wildsmith, 2011; Fomby & Cherlin, 2007), we also explored gender differences and racial/ethnic variation in the implications of family structure trajectories for the start of one’s own relational trajectories.

The results were consistent with the hypothesis that early transitions into a new family structure as a child would be associated with an earlier entry into cohabitation as a young adult. The timing of family structure transitions also appeared to be important but in ways that differed between young men and young women and among young adults from three major racial/ethnic groups. While men overall were slower to enter into cohabiting unions than women, men who experienced family transitions during childhood were more likely to enter into cohabiting unions at younger ages than their male peers who had spent their childhood in two biological parent family homes. Specifically, young women who transitioned early into a stepfamily were among the youngest women to form a cohabiting union, while young men who transitioned late into a stepfamily formed cohabiting unions at younger ages than young men who spent the majority of their childhood in two biological parent family structures. These nuanced results highlight the importance of measuring family structure patterns to capture timing, type, and number of changes across the early life course. Results also suggest White and Latinx young adults were more vulnerable to dimensions of changing family structures than their African-American peers.

The LCA provided insights into the type and timing of transitions in young adults’ family structures over time. For the full sample, transitioning early into a stepfamily was followed by early entry into a first cohabiting union compared to all other family structure histories, including all trajectories involving time in single parent families. One explanation could be that, in many cases, transitioning into a stepfamily is an additional transition (e.g., partnered to single to partnered, single to partnered), adding to the overall instability and change in the family structure trajectory. Moreover, the addition of at least one person—or multiple people if children came with a stepparent—can create more initial disruption and adjustment in household routines and norms than people leaving the household (Coleman et al., 2000). In light of new family dynamics, additional disruptions and adjustment periods may lead to a desire for young adults to leave home earlier and establish own unions.

An interesting point of discussion concerns the within-family structure differences that we observed, particularly among young men and women who experienced a transition into stepfamilies. Early transitions into a stepfamily significantly predicted earlier cohabitation for young women, compared to early transitioning into a stepfamily for men and late transitioning into a stepfamily for young men and young women. Drawing on the extensive stepfamily literature, this significance of early transitions into a stepfamily for young women could reflect observations of their mothers’ (presumably) dating and modeling behavior (Bandura, 1969; Arocho & Kamp Dush, 2017). Additionally, the quality of the mother-child relationship—particularly for the mother-daughter dyad—in stepfamilies is often lower than in other family structures (Amato, King, & Thorsen, 2016). As new family dynamics replace old dynamics, young women’s relationships with their mothers is likely to change too, possibly resulting in their early home-leaving and union formation.

Consistent with the hypothesis, White young adults appeared to be the most responsive to changes in family structure in terms of their own early entry into cohabitation. This apparent vulnerability—predicated on the assumption that early cohabitation often leads to poor relationship quality and relationship dissolution—was particularly pronounced for young people living in families defined by early transitions into single parent and stepfamily homes. Latinx young adults followed similar patterns as their White peers with a few notable exceptions. Half of White young adults following the early transitioning to a single parent family trajectory entered into a cohabiting union by approximately age 21, but half of African-American and Latinx young adults entered into a cohabiting union by approximately age 23. This pattern could reflect something negative about family structure “deviations” from two biological parent homes for White youth having to do with cultural stigma and norm-breaking, but it could also reflect something positive about the strong kin networks and community supports that African-American and Latinx youth enjoy beyond the nuclear family (Stack, 1975; Sarkisian & Gerstel, 2004; Taylor et al., 2016).

Limitations

The NLSY-CYA was a useful dataset for this investigation of the intergenerational transmission of family structure patterns given the longitudinal nature of its data for both children and their parents. These detailed data allowed us to connect multiple developmental periods on the part of both parents and children, but, as with all data, these benefits came with some limitations. Attrition was more likely to occur for young men than for young women, which could have shaped the results, although we accounted analytically for this potential attrition bias with longitudinal weights. Maternal education was also significantly associated with attrition from the overall sample (Table 4). Respondents from the NLSY-CYA data were born to the women in the NLSY79, they ranged in age up to 26 years by 2014. As cohabitation has become more common over the last several decades, there might be cohort differences among respondents in whether they first cohabited or married. Indeed, young adults in the millennial cohort (born from 1980–1996) entered into a cohabitation at a mean age of 22, and young adults in Gen-X entered into their first cohabitation, on average, around a year older at 23. Furthermore, the nature of data collection for the NLSY-CYA means the data are not fully age representative of the U.S. population (instead, they are representative of younger children born to younger mothers). To help account for this representational issue, all models accounted for age of mother and were ordered by child age, not wave.

Table 4:

Results from logistic model predicting attrition

	Odds Ratio	SE
Child gender	0.81*	0.08
Child race/ethnicity
African-American	0.86	0.12
White	0.87	0.12
Young adult age	0.98	0.01
Birth order	0.96	0.05
Maternal education	1.09***	0.02
Constant	11.67	6.82

^**p < .01;

^*p < .05

Future Research

Next steps in this line of research are expanding children’s early contexts beyond maternal partners, including household and institutional contexts within a child’s ecology. Changing family structures do not occur in isolation; they often coincide with other changes, such as income volatility, residential moves, and school moves (Fomby & Mollborn, 2017). The conceptualization of family structure histories in terms of the nature and composition of household composition is especially important given our argument about the differential impact of family structure by race/ethnicity and how it might reflect supports outside the nuclear family. Another future direction is exploring respondents’ own patterns of family structure changes beyond initial entrance into cohabitation as they continue to age. Moreover, replicating these initial findings with other data would further confirm the importance of early and often family transitions and the importance of continuing to measure family structure from a life course perspective. Additionally, causal inference can be improved by the addition of mechanisms that were left unexplored to establish a baseline link between detailed family structure histories and timing to cohabitation across populations. Potential mechanisms include parental mental health (Williams et al. 2008), financial strain (Sassler, 2004), young adult higher education (Sassler et al., 2018), and parent-child relationship (Sassler, 2004). Future research can contribute by predicting class membership, identifying who is at heightened risk for relationship instability.

Family structure is dynamic and complex, and the field is moving towards a better understanding of how family structure serves as a context of development that reinforces social and economic status across generations. This study contributed to this tradition by conducting a focused analysis of family structure trajectories across childhood and adolescence and examining the potential mechanism of early entry into cohabitation. Its findings can be built on to continue that tradition by examining additional mechanisms of the intergenerational transmission of inequality (e.g. relationship quality, social support, parental mental health) using family structure histories that detail the number, timing, and type of transitions. We argue that doing so is important given the continued emphasis of developmental research on the family as the most proximate setting of development as the continued focus of child-focused social policy on parental marriage and family structure.