Intergenerational Relationships and Union Stability in Fragile Families

Abstract

Using data from the Fragile Families and Child Well being Study (N= 2,656), we examined the association between intergenerational relationships and parents’ union stability five years after a baby’s birth. Results showed that more amiable relationships between parents and each partner’s parents, and more time children spent with paternal grandparents, were associated with increased odds that parents were co-residing by the time their focal child was age five. More time that children spent with maternal grandparents reduced union stability, although this result was not robust to methods that better address selection. These findings underscore the importance of the broader social contexts affecting couple stability. Findings further suggest that even amidst changing demographic conditions, intergenerational family ties are important for couples—and by extension—their children.

Family structure and stability have long been of interest to social scientists, given the fundamental role that families play in rearing and socializing children. Children who spend time in single-parent families, and those who experience multiple family transitions, are at greater risk of having psychological problems, having sex or bearing children at an early age, dropping out of high school, and a host of other negative outcomes—compared to children who grow up with both biological parents (McLanahan & Sandefur, 1994; Amato, 2005; Fomby & Cherlin, 2007). Therefore, both researchers and policymakers have been concerned with the factors that predict union stability, particularly among couples with children.

Research to date has emphasized the role of individual-and couple-level characteristics associated with marital and nonmarital union stability(e.g., Lichter, LeClere, & McLaughlin, 1991; Smock, 2000); however, we do not know much about how extended family relationships influence the stability of couple relationships. A limited body of literature has examined how network ties influence relationships among college students and long-term marriages (Bryant & Conger, 1999; Bryant, Conger, Meehan, & Meehan, 2001; Sprecher & Felmlee, 1992), but these studies focus on unions found early or late in the life course and are limited to small, non-representative samples. The extent to which extended family relationships influence union stability for couples after the birth of a child remains largely unexplored.

In this paper, we use data from multiple waves of the Fragile Families and Child Well being Study (N= 2,656 ) to examine whether intergenerational relationships between mothers and paternal grandparents, fathers and maternal grandparents, and the focal child’s interaction with both grandparents, influence couples’ relationship stability early in a child’s life. We draw on social integration and uncertainty reduction theories to argue that more positive relationships between partners and extended kin likely increase union stability after the birth of a child. Our results show that stronger intergenerational relationships, including greater interaction between the paternal grandparents and the focal child, increased the probability that couples co-resided (either cohabiting or legally married) by the time their child was age five.

THEORETICAL PERSPECTIVES AND PREVIOUS RESEARCH

Theoretical consideration of the importance of social integration dates back to Durkheim’s ([1897] 1951) seminal work on suicide. We learned from Durkheim that the interplay between an individual and a collectivity has the potential to shape individual outcomes, including (his focus) one’s decision to commit suicide. Social isolation, regardless of the cause, separates individuals from the larger collectivity that gives them a sense of belonging. Although Durkheim never actually defined social integration (Moen, Dempster-McClain, & Williams, 1989), it is typically conceived as the connectedness of individuals through a set of shared beliefs and norms constituting a collectivity (e.g., families or religious organizations), legitimizing its members, and providing a sense of purpose. Social approval and emotional support in ongoing relationships contribute to the sense of purpose that social integration affords (Booth, Edwards, & Johnson, 1991).

Even amidst major changes in family demography in recent years, families remain important social institutions governed by norms and shared beliefs, where dyadic relationships affect each other and influence individual outcomes and change (Minuchin, 1988; O’Brien, 2005). Scholars argue that intergenerational relationships are one of the most important aspects of family life, which retain importance even as children become adults—and especially when grandchildren are involved (Hogan, Eggebeen, & Clogg, 1993). The quality of intergenerational relationships is largely dependent on shared norms of mutual obligation across the life course (Lye, 1996), and the initial dependence of children on their parents, the permanence of the parent-child relationship, and the social force toward family identity make intergenerational ties central to social integration (Umberson, 1992). These relationships can have enduring effects on adult children, including their attitudes toward—and behavior in—romantic relationships (Bryant et al., 2001; Veroff, Young, and Coon, 2000).

Spouses are expected to form relationships with their in-laws, even if they are not very similar to them (Bryant et al., 2001), and this expectation can place stress on —or offer important benefits to—couples. Uncertainty reduction theory posits that reducing uncertainty is a driving force in relational development (Berger, 1987). When parents or friends disapprove of, or are at odds with, one’s romantic partner or spouse, it may reduce one’s assurance about the relationship and place strain on it. By contrast, approval from members of one’s social network helps reduce uncertainty, increases confidence about a romantic partner, and gives couples a sense of shared ‘couple’ identity (Sprecher & Felmlee, 1992). When intergenerational relationships are strong, this integration reinforces social norms, serves as social control, and facilitates communication within marital unions (Booth et al., 1991; Stets, 1991).

Family expectations for partners in terms of family commitment and financial contributions may become greater once a couple has children. The transition to parenthood has been shown to both improve and cause stress in adult child-parent relationships (Rossi & Rossi, 1990; Spitze, Logan, Deane, & Zerger, 1994; Kaufman & Uhlenberg, 1998), but some evidence shows that parents provide greater support to their adult children who have young children than those who do not (Eggebeen & Hogan, 1990). This provision of social support may signal grandparents’ intent to contribute to the social mobility of their adult child and family or that the younger family is in need, so the causal impetus is potentially bi-directional.

Until recent decades, the vast majority of individuals married, and most children were born within marriage (Cherlin, 2009); hence, intergenerational ties typically involved a married couple with one or two sets of in-laws. Delayed marriage, high divorce rates, and rising rates of nonmarital childbearing have diversified these family patterns and increased the complexity of intergenerational ties. More children are now born outside of marriage than ever before: fully 40% of births in 2007 were to unmarried couples, with much higher proportions among racial and ethnic minorities (Hamilton, Martin, & Ventura, 2009). Children born to unmarried parents are at greater risk of experiencing multiple family transitions early in life compared to those born to married parents, and this instability is associated with negative outcomes (Fomby & Cherlin, 2007; Osborne & McLanahan, 2007). Between birth and age five, children whose parents were unmarried at the time of birth experienced an average of 2.55 relationship transitions, compared to .67 for married parents (Beck, Cooper, McLanahan, Brooks-Gunn, 2009).

Therefore, nonmarital childbearing complicates the process of reducing relationship uncertainty because these family contexts are less stable. Many unmarried relationships break up within only a few years of the baby’s birth (Center for Research on Child Well being, 2007), and many unmarried parents have had children by prior partners (Carlson & Furstenberg, 2006). The fragility and complexity of these family circumstances may challenge the establishment of strong kinship ties.

Predictors of Union Stability

An extensive body of empirical research has examined the factors that affect union formation and stability, particularly marriage. The primary emphasis has been on the role of individual- and couple-level factors, with little attention to ‘external’ relationships that may affect the couple. Socioeconomic characteristics are important factors that affect the likelihood of marriage (Lichter et al., 1991) or divorce (Cherlin, 2005), and some evidence suggests that high incarceration rates of African American men reduces the likelihood of both union formation and stability (Western, Lopoo, & McLanahan, 2004). Premarital cohabitation and young age at first marriage are also linked to marital instability (DeMaris & Rao, 1992; Smock, 2000; Dush, Cohan, & Amato, 2003), with notable differences between racial groups (Philips & Sweeney, 2005).

Rising cohabitation has increased attention to the stability of nonmarital unions, especially for couples with children. Relationship instability for these couples has been linked to relationship problems such as infidelity (Edin & Kefalas, 2005; Hill, 2007), substance abuse (Reed, 2007), and physical violence and women’s distrust in men (Carlson, McLanahan, England, 2004). Fathers’ multipartnered fertility has also been shown to diminish couple stability overtime (Carlson et al., 2004; Harknett & McLanahan, 2004; Monte, 2007). Difficult financial circumstances are also a major source of stress which may contribute to couple instability (Tach & Edin 2009) and serve as a barrier to marriage (Gibson-Davis et al., 2005).

External Relationships and Union Stability

The literature on union stability has largely ignored the potential influence of external relationships. In recent decades, however, sociologists have begun to examine the influence of larger social contexts, such as extended kin and friendship ties, on several specific types of unions. This research, which we summarize below, has primarily focused on relationships among college students and long-term married couples and has relied on small, non-representative samples.

Existing evidence suggests that young couples in the process of forming new unions may be especially influenced by family and friends. For example, female college students whose family and friends supported their relationship were more likely to stay in it than those whose relationship was not supported (Sprecher & Felmlee, 1992). At that age, however, support from family versus friends may be different, as one study finds that college students stayed together when their families disapproved of their relationships (the so-called ‘Romeo and Juliet effect’) but only if their friends approved of the relationship (Felmlee, 2001). As the author notes, the over-representation of friends versus family in college networks, and the potential that peer group support overrides family support, may explain this outcome. Young couples, particularly college students searching for the right long-term partner, may be less influenced by their parents’ views than couples who are in more serious relationships with long-term expectations. At the same time, it has been shown that young couples themselves may help shape their parents’ views of their relationships, especially as they move toward commitment (Leslie, Huston, & Johnson, 1986), underscoring the bi-directionality of parent-child relationships (Crouter & Booth, 2003).

The influence of extended family ties on couples who are in more serious, adult relationships is likely to have a more lasting effect —both because of the maturity of the relationships and because (grand)children are more likely to be involved (Lye, 1996). Although research on how external relationships affect union formation and stability among adults (i.e., post-college) is limited, two studies have explored how relationships with in-laws influence marital success among long-term married couples with children living in a rural Midwestern state. Using a sample of 451 white married couples in the early 1990s(average marriage duration of 20 years), Bryant and colleagues found that family support for the marital relationship, and discord between spouses and in-laws, significantly influenced marital stability, satisfaction, and commitment over time (Bryant & Conger, 1999; Bryant et al., 2001). This research highlights the important influence of parents on adult children’s relationships across the life course.

It is less clear, however, whether the influence of these kin relationships holds true among more newly married couples with young children or in nonmarital unions with children. Also, we do not know how these associations bear out in socioeconomically-disadvantaged families, where unions tend to be less stable (Osborne & McLanahan, 2007) and multipartnered fertility more common (Carlson & Furstenberg, 2006). Moreover, the couples in Bryant and colleagues’ study were married around 1969, a time when cohabitation and nonmarital childbearing were far less common than they are today (Casper & Bianchi, 2002). This raises the question of whether the changing demographic landscape, i.e. more cohabiting couples and single parents living with children, changes the potential for intergenerational ties to influence relationship stability. In this paper, we extend prior research by examining how relationships between couples and their parents—and between children and their grandparents—influence the stability of couple relationships after a birth in large U.S. cities in the late 1990s.

METHOD

Data

We used data from the Fragile Families and Child Wellbeing Study, a longitudinal birth-cohort study with an over sample of unmarried parents. The study includes 4,897 births—3,710 unmarried and 1,187 married. The weighted sample represents nonmarital births in U.S. cities with populations over 200,000. Baseline interviews with mothers and fathers took place in 75 hospitals in 20 U.S. cities just after the baby’s birth from 1998 to 2000, and follow-up interviews were conducted about one, three, and five years after the birth. Response rates were 88% for unmarried mothers and 75% for unmarried fathers at baseline; 85% of mothers were retained in the study by the five-year interview, and 88% of fathers were interviewed at least once. In this paper, we used data from three waves of interviews with mothers—baseline, one year and five years; we omitted the three-year wave because several of our independent variables of interest were not included in that survey.

We limited our study to 18 cities because three out of six questions used as independent variables were not asked in Oakland and Austin in the one-year survey (reducing the sample by 657 cases). Then, missing data on individual items across the six independent variables reduces the analytic sample by 1,047 cases (819 of these are missing because the maternal and/or paternal grandparents were deceased). Attrition by the fifth survey also resulted in the loss of 534 cases. Finally, four cases were missing information on relationship status at year one. Taken together, these selection criteria resulted in a final sample of N= 2,6 56 couples who had a child together in the late 1990s and who had at least one living parent at the one-year follow-up survey. We address possible selection bias as a result of our sample in the sub-section on Robustness below.

Dependent Variable

At every wave of the Fragile Families Study, mothers reported their current relationship status with the focal child’s father. For our dependent variable, union stability, we coded mother’s report of current relationship with the biological father at the one -and five-year follow-up into two categories of resident (i.e., married or cohabiting)and non-resident (i.e., do not live together). Cohabitors were those who reported that they lived with the baby’s father “all or most of the time” or “some of the time.” We included those who were romantically involved, but living apart, in the non-resident category because there were only 109 of these cases ( 4% of the weighted sample). Note that mothers who were not in a relationship with the biological father may have lived with a new partner, but we focused only on the relationship stability of the focal child’s biological parents.

Independent Variables

At the one-year survey, mothers were asked six questions about intergenerational ties. These measures reflect how well the mother got along with her own mother and father as well as the father’s parents; how well the father got along with the mother’s parents; and how often the focal child saw both the maternal and paternal grandparents. Response choices for the questions asking how well the mother got along with her parents, the father’s parents, and how well the father got along with the mother’s parents ranged from 1 to 3, where 1 =not very well, 2= pretty well, and 3 =very well; therefore, higher scores represented better-quality relationships. Mothers who reported not knowing their fathers (12% of sample) were included in our main analyses at the lowest level of relationship quality. We estimated supplemental models where we assigned these cases to a new category of ‘no relationship’ and the results were consistent with our findings. For the questions about the child’s frequency of seeing their grandparents, response choices ranged from 1 to 5, where 1 =never, 2 = less often, 3 =few times/year, 4 =few times/month, and 5 =one or more times a week.

Factor analyses showed that these items did not load well together on one or more factors. Inter-item correlations ranged from r= .00 to r= .36 (with 9 out of 15 below r= .20). Alpha reliability scores for all six items, and various subsets of items, were always less than 0.5. Therefore, we treated the six items as separate measures of intergenerational relationships.

Covariates

We relied on prior research, particularly studies of union stability following a baby’s birth using the Fragile Families data (Carlson et al., 2004; Harknett, 2008; Harknett & McLanahan, 2004), to select our control variables. All time-invariant covariates were from mothers’ reports at the baseline interview, unless otherwise noted (some were unavailable until the one-year survey). Time-varying covariates were from the one-year and five-year surveys. We used multiple imputation, with the ice command in Stata (Royston, 2004), to estimate missing values on our covariates only; the proportion of cases missing on any covariate was always less than 10%, except for multipartnered fertility (15%), religious attendance (11%), and the constructed income-to-poverty ratio (roughly 25% at baseline, and 10% at follow-up surveys).

For background and socioeconomic variables, mothers’ age was measured in years. A dummy variable reflected whether she lived with both of her biological parents at age 15. We used four dummy variables for mothers’ race/ethnicity: non-Hispanic Black, Hispanic, non-Hispanic White, and non-Hispanic other (mostly Asian and Native American). We also used a dummy variable for whether the mother and father were of different racial/ethnic backgrounds. We used four dummy variables for mothers’ and fathers’ education: less than high school, high school degree, some college, and a college degree or higher. We used a dummy variable for whether the father worked in the week prior to the baseline interview. The income-to-poverty ratio for the mother’s household reflected total household income divided by the Federal poverty line based on household size; higher ratios indicated greater economic resources. Family and fertility characteristics were measured using a continuous variable for mother’s parity (number of other children besides the focal child). Mothers’ and fathers’ multipartnered fertility were measured using dichotomous variables (at the one-year survey), where 1= had a child by another partner.

Our relationship quality measures included a dichotomous variable for whether the mother reported that the father often or sometimes hit or slapped her when he was angry. Also, we included a measure of supportiveness in the couple relationship based on the average of four questions: whether the father was fair and willing to compromise, whether the father showed the mother affection, whether she felt insulted by the father (reverse coded), and whether the father encouraged her (α= .66); response choices ranged from 1= never to 3 =often, where higher scores represented a greater level of supportiveness. We also control led for the number of years that the mother knew the father prior to the birth of their child.

In addition, we included a measure of attitudes toward marriage, based on the average of responses to three questions about: whether it is better to get married than to live together, whether it is better for children if their parents are married, and whether living together is the same as being married (reverse coded) (α= .63); response choices ranged from 1 = strongly disagree to 4 =strongly agree, so higher scores indicated more positive attitudes toward marriage. Other measures included the self-reported health status of the mother (range is 1 = poor to 5 = excellent) and how often the mother attended religious services (range is 1 =never to 6 =one or more times per week).

We also included a dichotomous measure for grandmother co-residence, so that our measure of contact between the grandmother and the child did not simply reflect her living with the mother and child. Finally, to account for the oversampling of nonmarital births within the Fragile Families survey design, we included a dummy variable for marital status at baseline.

Analytic Approach

Our goal was to estimate how external family relationships were associated with couple relationship status over years one through five of the focal child’s life, net of confounding covariates. Because of concerns about selection, we focused on estimates from random and fixed effects models, using repeated observations about intergenerational relationships and union stability pooled across years one and five and taking advantage of the longitudinal design of the data.

Random effects models capture variation both between and within subjects while controlling for unobserved heterogeneity via the composite error term (treated as a random variable, see Allison 2009: 2). Fixed effects models utilize only within-subject variation and reflect how changes in intergenerational ties are associated with changes in union stability. This more conservative technique reduces bias by controlling for fixed unobserved individual characteristics, which may be associated with both intergenerational relationships and union stability. Fixed effects models may also do a better job of controlling for unobserved variables than random effects models because unobserved and observed variables are allowed to correlate, which is not the case in random effects models (Allison, 2009). Even so, fixed effects models do not account for possible unobserved differences between individuals that change over time. Estimates from random and fixed effects models provided different information about our research question; therefore we used both approaches here.

We first ran each model with only the intergenerational relationship variables to provide a baseline comparison before adding covariates because the random and fixed effects models do not include all of the same covariates. The random effects model with controls included both time-variant and time-invariant covariates. The fixed effects model with controls included only time-varying covariates at years one and five; these were mother’s age, health, education, income-to-poverty ratio, and whether the grandmother lived in the household.

RESULTS

Descriptive Statistics

We begin by describing the characteristics of our sample. Table 1 shows the weighted means and percentages for our covariates; we show figures separately by couples’ relationship status at year one because there were notable differences between married mothers and both categories of unmarried mothers. Married mothers were older than cohabiting and non-resident mothers(with mean ages of 29, 24, and 23, respectively), and they were more likely to have lived with both of their biological parents at age 15. Married mothers were more likely to be White (45%) and less likely to be African American (1 4%) than unmarried mothers.

Table 1.

Sample Characteristics by Relationship Status at One Year After Birth of Child (N = 2,656)

	Resident		Non-Residenta
	Married (54%)	Cohabiting (11%)	Other (37%),
Demographic/Background Characteristics
Mother’s Age (mean)	28.76 (5.49)	24.28 (5.97)	23.32 (5.38)
Lived w/Both Biological Parents at Age 15 (yes/no)	69.7	36.8	31.9
Mother’s Self-Reported Health (range = 1–5)	4.11 (.84)	3.89 (.93)	3.82 (1.02)
Years Known Baby’s Father Prior to Birth	7.96 (5.45)	4.39 (4.30)	4.25 (4.46)
Mother’s Race/Ethnicity
Black Non-Hispanic	13.8	47.8	56.2
White Non-Hispanic	44.7	11.1	13.4
Hispanic	31.1	37.0	28.8
Other Non-Hispanic	10.4	3.9	1.5
Father of Different Race/Ethnicity (yes/no)	9.8	13.4	19.9
Socioeconomic Characteristics
Mother’s Education
Less than High School	17.8	32.5	35.3
High School	25.3	36.3	41.1
Some College	20.4	29.6	21.7
College Degree or Higher	36.6	1.7	1.9
Father’s Education
Less than High School	17.7	35.6	70.6
High School	18.1	38.2	33.0
Some College	24.9	16.9	23.7
College Degree or Higher	38.5	5.2	2.0
Mother’s Income-to-Poverty Ratio	4.68 (3.78)	1.81 (1.68)	1.34 (1.35)
Father Worked for Pay Last Week (yes/no)	96.8	80.9	63.2
Household Characteristics
Grandmother Present in Household at 1 Year (yes/no)	10.6	17.4	39.2
Number of Children	1.00 (1.09)	1.08 (1.26)	1.03 (1.37)
Multi-Partnered Fertility
Mother Has Child by Prior Partner (yes/no)	12.8	38.7	34.0
Father Has Child by Prior Partner (yes/no)	13.7	28.7	54.6
Relationship Quality & Attitudes
Father Hits/Slaps When Angry (yes/no)	2.8	1.8	6.1
Mother Feels Supported by Father (range = 1–3)	2.75 (.28)	2.69 (33)	2.45 (.46)
Mother’s Positive Attitudes about Marriage (range = 1–4)	3.14 (.49)	2.58 (.56)	2.66 (.55)
Mothers’ Attendance at Religious Services (Range = 1–6)	3.81 (1.47)	3.12 (1.48)	3.28 (1.64)

Note: All figures are weighted by city sampling weights. Standard deviations shown in parentheses.

^aThis category includes 4% who are romantically involved but living apart and 19% who are not in a romantic relationship.

Educational attainment was much higher among married mothers and fathers, and married mothers had higher income-to-poverty ratios. Married fathers were more likely to be working in the week prior to the baby’s birth. Notably, about 39% of mothers who did not live with the baby’s father co-resided with their mothers, compared to 17% of cohabitors and 11% of married mothers. The average number of other children was similar across groups, but cohabiting and non-resident parents were much more likely to have a child by a prior partner compared to married parents. Physical violence was low across all groups, and relationship supportiveness differed little but was highest among married couples. Married mothers had more positive attitudes toward marriage and attended religious services a little more frequently than unmarried mothers.

Table 2 shows means and percentages for our six intergenerational relationship variables, separated by couple relationship status at year one. Most mothers got along ‘very well’ with their mothers regardless of relationship status, but the figure was 18 percentage points higher for married compared to non-resident mothers. Fifty-six percent of married mothers got along with their fathers ‘very well’ compared to 49% of cohabiting mothers and 39% of non-resident mothers. This pattern was similar for how well mothers and fathers got along with each other’s parents. On the other hand, when we looked at how often children spent time with their maternal grandparents, the pattern changed: 74% of children whose parents were living apart saw their maternal grandparents once or more per week, compared to 46% for those with married parents and 67% with cohabiting parents. The frequency of contact with paternal grandparents was lower across all groups: regardless of relationship status, less than 50% of married and non-resident mothers reported that their child saw their paternal grandparents one or more times per week. Moreover, 31% of children with non-resident parents ‘never’ saw their paternal grandparents compared to a mere 3% in the same category for maternal grandparents. Children of unmarried parents saw both grandparents more frequently than children of married parents.

Table 2.

Means and Percentages for Intergenerational Relationships among Couples at One Year After Birth of Child (N = 2,656)

	Married		Cohabiting		Non-Resident
	M (SD)	%	M (SD)	%	M (SD)	%
Mother gets along with her mother (1–3)	2.74 (.52)		2.72 (.53)		2.45 (.74)
Not well		3.9		3.8		14.7
Pretty well		18.2		20.6		25.8
Very well		77.9		75.6		59.5
Mother gets along with her father (1–3)	2.37 (.79)		2.20 (.86)		2.05 (.85)
Not well		19.7		28.9		33.1
Pretty well		24.0		22.0		28.1
Very well		56.3		49.1		38.8
Father gets along with mother’s parents (1–3)	2.61 (.59)		2.49 (.73)		1.83 (.80)
Not well		5.7		14.4		42.0
Pretty well		28.1		22.4		32.7
Very well		66.2		63.3		25.4
Mother gets along with father’s parents (1–3)	2.52 (.68)		2.53 (.71)		1.99 (.83)
Not well		10.6		12.7		35.2
Pretty well		26.9		21.6		30.7
Very well		62.5		65.8		34.1
How often child sees mother’s parents (1–5)	3.82 (1.37)		4.29 (1.20)		4.53 (.95)
Never		12.8		6.0		3.1
Less Often than Few Times/Year		2.3		6.2		1.3
Few Times/Year		21.3		7.8		9.8
Few Times/Month		17.6		13.5		11.5
Once or More/Week		46.0		66.6		74.3
How often child sees father’s parents (1–5)	3.56 (1.33)		3.97 (1.36)		3.09 (1.59)
Never		13.7		10.6		30.7
Less Often than Few Times/Year		3.4		7.4		3.8
Few Times/Year		30.6		9.1		17.4
Few Times/Month		17.1		21.0		22.2
Once or More/Week		35.2		52.0		25.9

Note: All figures are weighted by city sampling weights. Standard deviations shown in parentheses. Percentages that do not round to 100% are due to rounding.

Our focus is on relationship stability, so we now describe the change in parents’ relationship status from one year after a baby’s birth to the five-year follow-up survey. These figures differed from other studies using the Fragile Families data because we limited our sample to those cases with valid information on intergenerational relationships at year one. Table 3 shows that 85% of those who were married at one year remained married by the five-year follow-up, and 15% had separated or divorced. Among cohabiting couples at year one, 62% were still together at the five-year survey—18% had gotten married, and 44% were still cohabiting, whereas 38% were no longer living together. Among couples living apart at one year, 5% had gotten married, and 7% had moved in together, while the majority (88%) continued to live apart.

Table 3.

Relationship Status One and Five Years After Birth of Child (N = 2,656)

One Year After Birth of Child	Five Years after Birth of Child (percent of row)
	Married (53%)	Cohabiting (11%)	Non-Residenta (37%)	Number of Cases (n ) (100%)
Married (57%)	85%	0%	15%	794
Cohabiting (20%)	18%	44%	38%	812
Non-resident (23%)1	5%	7%	88%	1,050
Number of cases (n )	832	378	1,446	2,656

Note: Percentages are weighted by city sampling weights. Numbers of cases (n’s) are unweighted.

^aNon-resident cases in our weighted sample at one year include 4% in nonresident relationships and 19% not romantically involved; at five years, only 1% were in visiting relationships, and 35% not romantically involved.

Multivariate Analyses

Table 4 reports the results from our random and fixed effects models predicting relationship status as a function of our intergenerational relationship variables (and covariates); the reference category is living apart from the baby’s father. For comparison purposes, we included a Model 1 without covariates, but we focus our interpretation on Model 2. Beginning with our random effects model, the quality of relationship between mothers and their parents decreased the odds (.81) that the couple co-resided five years after their child’s birth. By contrast, the degree to which the father got along with the maternal grandparents increased the odds by 2.25 that the couple co-resided by the fifth year. The same pattern holds for mothers’ relationships with the paternal grandparents; however, the odds were smaller (1.19) and less statistically significant than for fathers who got along with the mother’s parents (Wald Chi-Square test shows that these odds ratios we re significantly different). When we looked at the frequency of contact between children and their grandparents, we found that the child’s more frequent interaction with maternal grandparents (marginally) significantly decreased the odds (.80) that couples co-resided by the fifth year. On the other hand, the odds of union stability increased by 1.18 when children spent more time with their paternal grandparents.

Table 4.

Random and Fixed Effects Results (Odds Ratios): Co-Resident Relationships at Year Five by Intergenerational Relationships One Year After Birth

	Random Effects		Fixed Effectsa
	Model 1	Model 2	Model 1	Model 2
Family Integration
Mother gets along with her parents	1.05	.81 *	.80	.83
Father gets along with mother’s parents	4.28 ***	2.25***	2.23 ***	2.83 ***
Mother gets along with father’s parents	1.40 ***	1.19 *	1.45 **	1.44 *
How often child sees mother’s parents	.53 ***	.80 ***	1.02	.96
How often child sees father’s parents	1.55 ***	1.18 ***	1.65 ***	1.27 *
Demographic/Background Characteristics
Mother’s Age		.95 ***		.77 ***
Lived w/Both Biological Parents at 15 (yes/no)		1.40 **
Mother’s Health (Range = 1–5)		1.07		1.01
Years Known Baby’s Father Prior to Birth		1.04 *
Mother’s Race/Ethnicity (reference = Non-Hispanic White)
Black Non-Hispanic		.66 *
Hispanic		.94
Other Non-Hispanic		1.77
Father of Different Race/Ethnicity (yes/no)		.96
Socioeconomic Characteristics
Mother’s Education (reference = less than HS)
High School		1.05		2.21
Some College		.80		1.47
College Degree or Higher		.99		.90
Father’s Education (reference = less than HS)
High School		.86
Some College		.99
College Degree or Higher		1.22
Mother’s Income/Poverty Ratio (Range = 0–12.5)		1.23 ***		1.27 **
Father Worked for Pay Last Week (yes/no)		1.49 **		1.67
Household Characteristics
Total Number of Children		1.07		1.27
Grandmother Present in Household (yes/no)		.60 **		.36 **
Multi-Partnered Fertility
Mother Has Child by Prior Partner (yes/no)		.98
Father Has Child by Prior Partner (yes/no)		1.00
Relationship Quality & Attitudes
Father Hits/Slaps When Angry (yes/no)		1.19
Mother Feels Supported by Father (Range = 1–3)		.97
Mother-Positive Attitudes about Marriage (Range = 1–4)		1.03
Attendance at Religious Services (Range = 1–6)		.98		.90
N	2,656		2,656

^aOnly cases that change on the independent and dependent variables between years 1 and 5 contribute to the estimate (n = 412).

^*p < .05.

^**p < .01.

^***p < .001.

Turning to our fixed effects models, when we controlled for unobserved heterogeneity by looking at only within-respondent change, the effect of mother’s relationship with her parents on co-residing with the child’s father was no longer statistically significant. The significant decrease in the odds of union stability when children spent more time with their maternal grandparents also diminished. This suggests that the original negative association between contact with maternal grandparents and union instability may be due to some selection factor(s) that are correlated with both the frequency of child-grandparent contact and couple union stability; once individual characteristics were held constant, we no longer observed this unexpected relationship. With respect to children’s contact with paternal grandparents, we found that more frequent contact increased the odds (1.27) that couples co-resided five years after their child’s birth, although the magnitude was smaller once we included covariates.

The results were fairly robust across all models. The fixed effects results show that even when unobserved characteristics that are fixed across individuals were held constant, positive changes in relationships between fathers and maternal grandparents—and increased contact between children and paternal grandparents—increased the odds that couples stayed together. The fact that the point estimates were larger than in the random effects models suggests some negative selection, i.e., unobserved variables were masking part of the positive association between family ties and union stability.

In short, with one important exception (in the random effects model s), we found support for the hypothesis that stronger intergenerational ties positively influenced union stability among couples who have a child together; this was especially true for fathers. In particular, when fathers had increasingly good relationships with the maternal grandparents, and when the focal child spent increasingly more time with the father’s parents, couples were significantly more likely to stay together than to break up by the time their child was five. Positive relationships between mothers and paternal grandparents were also important, but the magnitude was much smaller. We also looked at possible differences by race/ethnicity by running our models separately for Whites, African Americans, and Hispanics(results not shown). The patterns were the same, although the coefficients were sometimes not statistically significant due to smaller cell sizes. Examining differences in how family ties affect couples by race/ethnicity is an important topic for future research.

With respect to our covariates, older age, living with both biological parents at age 15, and knowing the father longer were all significantly associated with greater odds that couples lived together versus apart by the time their child was five years old. We also found that race and socioeconomic factors were significantly associated with the union stability of couples with children. Income and fathers’ employment were both positively associated with co-residing by the fifth year after the baby’s birth. Grandmother co-residence was associated with a reduced likelihood of parents living together, as we expected.

Robustness

To evaluate the robustness of our findings, we conducted additional descriptive analyses to examine two possible sources of sample selection bias: 1) cases excluded where parents reported their parents were deceased; and 2) cases lost due to attrition from the survey. To do this, we compared the baseline characteristics of those who were not in our final sample to those who were in the final sample. The death of parents (or in-laws) of mothers in their twenties points to the possibility of greater disadvantage among the excluded cases, including serious parental health problems and/or deaths due to violence, accidents, or other unnatural causes. In the cases whose parent(s) died, respondents were slightly older and more likely to be African American compared to our sample. Also, parents had less education, fewer fathers worked, the household income-to-poverty ratio was lower, and multipartnered fertility was more prevalent in excluded cases. Fewer respondents who had a deceased parent were married at the time of birth, and they were in slightly worse health; there were other small differences such as baseline relationship quality, but overall the differences were modest. Further, we estimated a bivariate logistic regression model (results not shown) predicting couple stability at five years as a function of whether either focal parent had a deceased parent (mother or father) as of the one-year survey; we found no statistically significant difference in union stability by whether a parent had died.

We found similar patterns comparing those who attrited from the study by year five to those who remained in the study. Overall, cases lost to attrition were more likely to be minorities and have lower socioeconomic resources than those who remained in the study. Although the extent to which this biased our results is not clear, the consistency of the findings in our separate analyses by race suggests that our results we re not strongly biased by our sample selection.

DISCUSSION

This study provides new information about the important role of external family ties for the stability of relationships among couples who had an urban birth in the late 1990s. Most studies of union stability have emphasized individual-or couple-level factors, whereas our research suggests that intergenerational kin ties play a significant role in shaping what happens to families with a new child. This study gives credence to the long-argued notion that social integration is an important predictor of individual-level outcomes—something Durkheim noted decades ago.

A small body of prior research has suggested that friends and family play an important role in relationship outcomes for college students and long-term married couples (Sprecher & Felmlee, 1992; Bryant & Conger, 1999; Bryant et al., 2001). We found evidence that positive family relationships significantly improved the probability that couples co-resided five years after their child’s birth. The results were robust to controls for a number of factors previously shown to influence union stability, such as demographic and socioeconomic characteristics, relationship quality and attitudes, and family and fertility characteristics.

The salience of fathers’ family integration and its influence on couple outcomes is particularly striking in our findings. Our results showed that the relationship between mothers and their own parents did not significantly increase the probability that couples stayed together. On the other hand, fathers’ getting along well with mothers’ parents had the largest association (among the intergenerational relationship predictors) with couples’ stability, followed by the time the focal child spent with the paternal grandparents. This result is consistent with Stack’s (1974) argument that when fathers are well-liked by mothers’ families, the relationship stands a better chance of surviving; it is also consistent with the notion that maternal grandparents may hinder couple relationships when the interaction between the grandmother and the father is not amiable. Here, we reiterate that our reports of family relationships were from mothers’ perspectives, and fathers’ perspectives may be different.

The negative association between how often the focal child spent time with their maternal grandparents and union stability (in the random effects models) was surprising. We expected that time spent with both sets of grandparents reflected greater family integration, promoting couple stability. Yet, children’s greater time spent with mothers’ parents significantly decreased the probability that couples stayed together by their child’s fifth year, even when we controlled for grandmother co-residence (and other factors, including the provision of child care in supplementary analyses not shown). Although this finding appears counter-intuitive, research shows that maternal grandparents often provide child care in low-income families and may substitute for an absent biological parent (Jendrek, 1993; Pruchno, 1999; Uttal, 1999). Therefore, this may reflect reverse causality such that the grandmother takes more responsibility for the child once the couple relationship has broken down. Understanding this association is a useful topic for future research.

A key question is whether stronger intergenerational relationships contribute substantially to couples staying together, or whether couples with high extended-family integration are characteristically different from couples that are likely to break up. In other words, does greater family integration affect the stability of the focal couple’s relationship by integrating them within a family network and reducing uncertainty about their partner choice, as theory would suggest? Or, do these ties simply reflect unobserved heterogeneity across couples, such as commitment to family more generally, or closer proximity to family members that allows for more frequent contact, temperaments that facilitate positive relationships, or parents’ desire for their children to know their grandparents?

Our fixed effects models offer the most rigorous test of causality because they rely on only within-subject change and control for time-invariant individual characteristics. Yet, these models did not account for unmeasured time-varying characteristics, so our results could still be biased by variables correlated with both intergenerational relationships and union stability that changeover the observation period. Also, these models do not account for the potential reciprocal nature of the association between union stability and family ties in long-term relationships. Given prior research (Bryant & Conger, 1999; Bryant et al. 2001), we might expect that the influence of these ties increases the longer couples are together.

Analyses of missing data do not point to serious selection bias with our sample compared to the larger sample of parents at the time of birth in the Fragile Families Study. We observed relatively minor differences between our sample and the excluded cases. At the same time, our sample is drawn from a study of urban births in which the average age of mothers is fairly young—early to late twenties, so our results cannot be generalized to older parents with children, to couples without children, nor to those who live in rural or suburban areas.

This study adds to our understanding of union stability among unmarried couples with children by examining the role of extended family relationships. Given the high instability in couple relationships after a nonmarital birth and the deleterious effects of multiple transitions on children (Fomby & Cherlin, 2007; Osborne & McLanahan, 2007), understanding what keeps couples together is important. This study demonstrates the continued influence of social integration on individuals and the broad reach of Durkheim’s theory. Although Durkheim attributed higher suicide rates to individualism and social disintegration within different religious denominations in Europe more than a century ago, we might imagine how similar social forces affect outcomes for disadvantaged families, and especially fathers, in U.S. urban areas today. The continued economic hardship of low-income fathers—partially the result of industrial shifts in urban areas over the past few decades (Wilson, 1996)—may have contributed to the social disengagement of young fathers and led to poorer extended family relationships(among other things) and, consequently, less stable relationships with the mothers. The consequences of this disintegration are vast and have far-reaching, negative implications for children.

Future research should continue to examine the role of extended family ties in couple-level outcomes with particular emphasis on fathers and their relationships with mothers’ families. Exploration of the characteristics which distinguish fathers who are—and who are not—well-integrated into mothers’ families would be a reasonable place to start.