Parents’ Relationship Quality and Children’s Behavior in Stable Married and Cohabiting Families

Abstract

Although an extensive literature has shown that family structure is linked with child well-being, less well understood is how the dynamics within families affect children, in particular the extent to which positive mother–father relationship quality is linked with children’s outcomes. In this study the authors used data from the Fragile Families and Child Wellbeing Study (N = 773) to examine how couple supportiveness in stable coresident families is related to children’s externalizing and internalizing behavioral problems over ages 3 through 9. Using latent growth curve and fixed effects models, they found that parents’ greater supportiveness has a slight association with lower levels of children’s behavioral problems. Using cross-lagged structural equation models to examine the direction of the association, they also found some evidence that parents’ relationship quality and children’s behavioral problems are reciprocally related. Overall, this study suggests that more positive couple interactions are beneficial for children residing with both of their biological parents.

An extensive sociological literature over many decades has examined how family structure and instability are linked to children’s well-being. Overall, this research suggests that living in a stable two-parent family is associated with positive development for children (Fomby & Cherlin, 2007; McLanahan, Tach, & Schneider, 2013). Less well understood is how relationships and dynamics within stable families are related to child outcomes. Family systems theory suggests that children’s development is intrinsically related to interactions among other actors within the family (Bronfenbrenner, 1979). The mother–father relationship is often perceived to be at the center of this system, influencing all aspects of family functioning (Easterbrooks & Emde, 1988). Therefore, parents’ ability to communicate effectively, generate emotional closeness, and support each other’s decisions likely has implications for their children’s well-being and development.

One dimension of children’s well-being that may be particularly affected by the quality of their parents’ relationship is their behavior. Children may learn certain patterns of behavior from observing their parents (Bandura, 1978), and/or couples’ relationship quality may affect the quality of parent–child interactions (Erel & Burman, 1995). Greater behavioral problems during childhood have been linked to a range of adverse outcomes later in life, including lower odds of completing high school and attending college (McLeod & Kaiser, 2004), greater risk of unemployment after leaving school (Fergusson & Horwood, 1998), and higher odds of mental health problems and criminal behavior (Fergusson, Horwood, & Ridder, 2005). Understanding the extent to which parental relationship quality is linked to behavioral problems may illuminate a key aspect of how stable families contribute to children’s long-term success.

BACKGROUND

Theoretical Perspectives

Family systems theory suggests that the family is a complex, dynamic, and integrated whole, in which each member influences and is influenced by all other members (Minuchin, 1988; O’Brien, 2005). The changing relationships and characteristics of particular family members are believed to have implications for the growth and development of other individuals within the family. Given the centrality of parents for children’s growth and development (Rossi & Rossi, 1990), we expect that the nature of—and changes in—the quality of parents’ couple relationship should be linked to (changes in) child well-being.

Social exchange theory highlights that relationships develop and change—and interdependence grows—as a function of the social exchanges that occur between partners (Kelley & Thibaut, 1978). Although negative exchanges (e.g., conflict and tension) have been the focus of much theoretical and empirical research about couple relationships and individual well-being (e.g., Grych & Fincham, 1990), reciprocity in giving and receiving positive emotional support is fundamental to the formation of high-quality, intimate relationships, which represents a unique form of “couple-specific capital” (Hohmann-Marriott, 2008). Supportiveness in the relationship between mothers and fathers is thus an important aspect of couple relationship quality that may have implications for children over time.

Parents’ relationship quality may affect children directly or indirectly via the parent–child relationship (Easterbrooks & Emde, 1988). In terms of direct effects, social learning theory suggests that children model the behavior and interactions of significant others, in particular their parents (Bandura, 1978). Witnessing positive and low-stress interactions between parents may produce similar behavioral styles in children. Couples’ relationship quality may also affect children’s well-being indirectly via parenting behaviors (Belsky, 1984; Engfer, 1988). Marital quality and parenting are shown to be positively related, as the qualities of the dyadic couple relationship “spill over” to affect the character of the parent–child relationship (Erel & Burman, 1995). As such, conflict or tension—or conversely, supportiveness and positive affect—in the parental union may affect parent–child relations and, ultimately, children’s outcomes.

The association between parents’ relationship quality and children’s behavior may vary over the course of children’s development, as parents’ involvement in their children’s lives—and children’s understanding of their parents’ relationship—change over time (O’Brien, 2005). During infancy, parents’ primary role is to provide for the physical, social, and emotional needs of children who are fully dependent on them (Bornstein, 2002). During toddlerhood and the preschool years, as children develop greater independence and acquire new social and cognitive skills, they begin to test parental limits, increasing the need for consistent disciplinary practices and raising parental stress (Schoppe-Sullivan, Mangelsdorf, Frosch, & McHale, 2004). During middle childhood and adolescence, children continue to develop their own identities and become increasingly independent as they engage in new activities and social relationships (Collins, Madsen, & Susman-Stillman, 2002). In light of these developments, we might expect parents’ relationship quality to have a greater effect on children’s behavior at younger ages, whereas children’s behavior may have a greater influence on parents’ relationship quality as they get older. Longitudinal research that uses multiple time points can shed light on differences in how parents’ relationship quality is linked to children’s behavior as children age.

Empirical Findings

A large body of literature, especially in the fields of psychology and child development, has examined how parents’ relationship quality is linked to children’s well-being; the vast majority of this research has focused on negative features of the couple relationship and has found that higher conflict and discord are associated with higher behavioral problems and maladjustment among children (for reviews, see, e.g., Cummings & Davies, 2002, and Reid & Crisafulli, 1990). Also, parental conflict has been cited as an explanation for the effects of divorce and family instability on children (Fomby & Osborne, 2010; Jekielek, 1998); in other words, marital dissolution is associated with diminished child well-being in part because of high levels of parental conflict that often ensue in the divorce process (Amato, 2000).

Fewer studies have examined the association between positive aspects of parents’ relationship and children’s behavior; those that have suggest that parents’ supportiveness, positive affect, and ability to communicate are positively linked with children’s well-being (Conger, Rueter, & Elder, 1999; Howes & Markman, 1989; Miller, Cowan, Cowan, Hetherington, & Clingempeel, 1993). However, many such studies use data on small, nonrepresentative samples and/or do not include data across multiple time points. Given the wealth of evidence about the benefit of stable two-parent families for children (e.g., McLanahan et al., 2013), more research is needed to understand whether and how positive parental interactions matter for children’s development (Heinrich, Cronrath, Degen, & Snyder, 2010).

Within this literature questions also remain about the causal nature of the association between couple relationship quality and child behavior; it is possible that the association is driven by common factors that affect them both; that is, social selection. To better evaluate causality, it is essential to control for variables that may be related to both relationship quality and child behavior. Previous research has suggested that such factors include demographic characteristics (e.g., marital status, age, race, education, income, and employment; Buehler et al., 1997; Pardini, Fite, & Burke, 2008), parents’ physical and mental health (Engfer, 1988), religiosity (Wolfinger & Wilcox, 2008), family of origin (Cowan, Cohn, Cowan, & Pearson, 1996), prior fertility (Bronte-Tinkew, Horowitz, & Scott, 2009), and children’s gender (Reid & Crisafulli, 1990) and temperament (Caspi, Henry, McGee, Moffitt, & Silva, 1995); therefore, we included such variables as controls in our analyses. In addition, to further test the causal nature of this association, we considered models that estimate whether changes in parents’ relationship quality are associated with changes in children’s well-being within the same families over time. By focusing on within-family changes this technique allowed us to control for some unobserved factors that could also be driving between-family differences.

It is also important to consider that the nature of the association between parents’ relationship quality and children’s behavior may be bidirectional. Although family systems theory emphasizes the dynamic and reciprocal nature of effects, prior empirical research has typically assumed that parents primarily influence children (Heinrich et al., 2010; O’Brien, 2005), which likely stems from the emphasis in social science research on parental socialization of children (Maccoby, 1992). However, children can also affect parents (Bell, 1968); for instance, recurring behavioral problems may increase parents’ stress and serve to erode their relationship over time. Because of this potential reverse causality it is important to account for children’s effects on parents when examining how parents’ relationship quality affects children (Hawkins, Amato, & King, 2007). The small number of studies that have explicitly examined such bidirectional effects have found more consistent evidence that parents’ relationship quality influences children’s behavior than vice versa (Cui, Donnellan, & Conger, 2007; Schermerhorn, Cummings, DeCarlo, & Davies, 2007), but this literature has focused on negative dimensions of the couple relationship, such as marital discord (Jenkins, Simpson, Dunn, Rasbash, & O’Connor, 2005). We extend this body of work by examining reciprocal associations between supportiveness in the couple relationship and children’s behavior across child ages 3, 5, and 9. By considering both the preschool and middle childhood years, we were able to evaluate whether different patterns emerge for child versus parent effects across key stages of childhood. We expected children’s effects on parents to increase as children age; during the toddler years, parents likely have greater influence on children, but as children become more independent the effects of their behavior on parents may be greater.

Moderating Factors

Recently, scholars have urged that greater attention be paid to the context in which family processes occur (Crosnoe & Cavanagh, 2010). Even among couples living together, one factor that may moderate the association between parental relationship quality and children’s behavior is marital status. Marriage is more “institutionalized” as a context for childrearing (Cherlin, 2005), whereas cohabiting relationships are characterized by less commitment and more instability (Bumpass & Lu, 2000). As a consequence, the degree of supportiveness in the relationship may be of greater consequence to cohabiting parents than married parents if the former are seeking signals about whether to stay with their partner (Brown, 2000). If children sense this greater uncertainty in their parents’ relationship, we might expect relationship quality to have a greater effect on children’s behavior in cohabiting families.

A second potential moderating factor is child gender, and past research has uncovered mixed results. Some studies have found no significant difference in how parents’ relationship quality is linked to well-being for boys versus girls (Buehler et al., 1997), whereas other studies have shown that parental discord has a greater effect on boys (Emery & O’Leary, 1982; Reid & Crisafulli, 1990). One explanation for the latter is that fathers may be more likely than mothers to disengage from their children if parental relationship quality is low (Krishnakumar & Buehler, 2000). If sons tend to be closer to their father than daughters, this might account for the greater sensitivity of boys to parental relationship quality.

The Present Study

In this study, we extended the literature on parental relationship quality and children’s behavior by examining how biological parents’ supportiveness in the couple relationship is related to children’s behavioral problems throughout early and middle childhood. In light of the theoretical and empirical considerations discussed above, we hypothesized that the degree of supportiveness in parents’ relationship will be negatively related to children’s behavioral problems. We anticipated finding bidirectional effects, but we expected that parents will have a stronger effect on children than children have on parents, especially when children are younger. Finally, we expected that these associations will be stronger in cohabiting than married-parent families and stronger for boys than girls.

METHOD

Data

We used data from the Fragile Families and Child Wellbeing Study (http://www.fragilefamilies.princeton.edu/), a longitudinal study of 4,897 births that occurred in 20 large U.S. cities between 1998 and 2000. Unmarried couples were oversampled and the sample, when weighted, is representative of births in cities with populations over 200,000. Mothers and fathers were interviewed in the hospital within 48 hours of the focal child’s birth, and follow-up interviews were conducted via telephone when the focal child was approximately 1, 3, 5, and 9 years old. Completion rates for these four survey waves were 89%, 86%, 85%, and 72% for mothers and 69%, 65%, 64%, and 54% for fathers. Mothers were also asked to complete an in-home child assessment at the 3-, 5-, and 9-year surveys to obtain more detailed information about children’s development and behavior. Of mothers who completed each core survey, approximately 78% completed the in-home survey at 3 years, 81% completed it at 5 years, and 89% completed it at 9 years.

To examine the association between couple’s supportiveness and children’s behavioral problems, we restricted our sample to 967 couples who were coresident from the 1- through 9-year surveys. This group represents about 36% of the nearly 2,700 couples who were living together at 1 year. From this sample, we excluded eight cases (0.8%) in which the focal child did not reside with the parents at each survey wave, 38 cases (3.9%) in which the parents divorced but continued living together, 51 cases (5.3%) that were missing information on couples’ supportiveness at two or more waves, and 97 cases (10.0%) that lacked information on children’s behavioral problems at two or more waves. Our final analytic sample consisted of 773 children with stably coresident parents. Given the typical instability of cohabiting relationships, it is important to note that the cohabiting couples in our sample were an especially select group.

Among our 773 cases, 74 (9.6%) were missing information on one or more of the covariates in the multivariate models. Cases with missing information differed from the rest of the sample in terms of some key demographic characteristics: They had lower incomes, were more likely to be a member of a racial/ethnic minority group, and were more likely to be unmarried at the focal child’s birth. We used multiple imputation (Royston, 2004) to impute missing covariates as well as to impute wave-missing responses to the supportiveness measure.

In our analyses of differences between married and cohabiting couples we distinguished parents by their marital status at the time of their child’s birth. Previous research suggests that couples who marry after their child is born are more similar to couples who are unmarried at the birth than to those who are married (McLanahan, 2004). However, coresident couples who married in the years following their child’s birth may have differed in meaningful ways from those who remained unmarried. For this reason, we also ran our moderation analyses differentiating couples based on their marital status at 9 years.

Measures

Children’s behavioral problems

We examined children’s externalizing and internalizing behavioral problems, measured from mothers’ responses to items from the Child Behavior Checklist (Achenbach, 1992). This is a widely used scale with strong psychometric properties for discriminating among children with and without emotional and behavioral disorders (Ebesutani et al., 2010). For each item, mothers indicated how true each behavior was of the focal child on a scale that ranged from 0 (not true) to 2 (very/often true).

Externalizing behavioral problems were measured as the mean of mothers’ responses to two behavior subscales: Aggressive and Delinquent. The Aggressive subscale contained items such as “[Child] argues a lot,” and the Delinquent subscale contained items such as “[Child] lies or cheats.” Similarly, internalizing behavioral problems were measured as the mean of mothers’ responses to the Anxious/Depressed and Withdrawn subscales. The Anxious/Depressed subscale contained items such as “[Child] feels [he/she] has to be perfect,” and the Withdrawn subscale contained items such as “[Child] would rather be alone than with others.” The number of items included in the externalizing-behavior measure equaled 22 items at 3 years (α = .85), 30 items at 5 years (α = .84), and 35 items at 9 years (α = .89). The number of items included in the internalizing-behavior measure equaled 25 items at 3 years (α = .80), 22 items at 5 years (α = .78), and 21 items at 9 years (α = .79).

Couples’ supportiveness

Our measure of the level of supportiveness in the couple’s relationship was constructed from mothers’ and fathers’ responses to five items at the 1-, 3-, 5-, and 9-year surveys. At each wave, parents reported how often their partner (a) “is fair and willing to compromise,” (b) “expresses love and affection for [them],” (c) “encourages or helps [them] with things that are important to [them],” (d) “listens to [them] when [they] need someone to talk to,” and (e) “really understands [their] hurts and joys.” Response options included 1 (“never”), 2 (“sometimes”), and 3 (“often”). Mothers’ and fathers’ responses were combined and averaged to create a composite measure of couples’ supportiveness (range = 1–3). Reliability scores (α) for this measure equaled .69 at 1 year, .71 at 3 years, .73 at 5 years, and .77 at 9 years.

To reflect the dyadic nature of the parental relationship, in our main analyses, we used the average of mothers’ and fathers’ reports about each other’s supportiveness. However, to examine the sensitivity of these results we also ran analyses using mothers’ and fathers’ separate reports (discussed in the RESULTS section). We also re-ran our models adding a variable indicating parental (dis)agreement about supportiveness (because the same mother–father average could reflect either mother and father agreeing, or one parent reporting high supportiveness and the other reporting low supportiveness); this variable did not alter the main effect of mean levels of supportiveness on children’s behavior, so we did not include it in our main results.

Control variables

We also included a set of control variables for characteristics of mothers, fathers, and children that were likely related to parents’ supportiveness and children’s behavioral problems. These items were measured at the baseline survey unless otherwise noted. With regard to parents’ characteristics, we controlled for their marital status at the time of the focal child’s birth, mothers’ and fathers’ ages at the birth (in years), mothers’ race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other), whether the father’s race/ethnicity differed from the mother, mothers’ education level (less than high school, high school/some college, bachelor’s degree or higher), whether the father had a higher education level than the mother, and parents’ household income-to-poverty ratio. We also included self-reports of mothers’ and fathers’ physical health on a scale that ranged from 1 (poor) to 5 (excellent) and measures of mothers’ and fathers’ risk of depression, indicated by the short form of the Composite International Diagnostic Interview—Short Form (Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998). Each parent’s attendance at religious services was represented on a scale ranging from 1 (not at all) to 5 (once a week or more). Finally we controlled for whether each parent lived with their own biological parents at age 15, the total number of children parents had together at the 1-year survey, and whether they had children with other partners (both parents, mother only, father only, or neither parent).

Regarding children’s characteristics, we controlled for children’s gender and their temperament at the 1-year survey using the composite of three items from the Emotionality, Activity, and Sociability) Temperament Questionnaire, reported by mothers (Mathieson & Tambs, 1999). Response options ranged from 1 (not at all like my child) to 5 (very much like my child) and were averaged, with higher scores indicating more difficult temperaments (α = .55). Maternal reports on these items have been shown to correlate with interviewer observations (Meadows, McLanahan, & Brooks-Gunn, 2007).

Descriptive statistics for the overall sample, as well as by parents’ marital status at birth, are shown in Table 1. Married parents were somewhat older, were less likely to be a member of a racial minority group (compared to White), and had higher levels of education and income-to-poverty ratios. Married parents were also more religious, more likely to have lived with both parents, and much less likely to have had children with another partner. Finally, married parents were more likely to have boys and to have children with less difficult temperaments than cohabiting parents. All of these differences were statistically significant at the p ≤ .05 level.

Table 1.

Sample Descriptives, by Marital Status at Time of Baby’s Birth

Characteristic	Overall (N = 773)		Married (n = 445)		Cohabiting (n = 328)
	M or %	SD	M or %	SD	M or %	SD
Marital status at baby’s birth
Married	80.6
Cohabiting	19.4
Age at baby’s birth (M, years)
Mother	29.20	5.43	29.96	4.26	26.06	8.65
Father	31.07	6.27	31.81	5.05	28.00	9.68
Mother’s race/ethnicity
White non-Hispanic	44.0		51.7		11.8
Black non-Hispanic	15.8		11.2		34.8
Hispanic	28.6		23.0		52.3
Other non-Hispanic	11.6		14.1		1.1
Parents are different race/ethnicity	10.1		10.5		8.4
Mother’s education
Less than high school	15.2		10.7		33.8
High school degree or some college	48.8		44.8		65.1
Bachelor’s degree or higher	36.0		44.5		1.1
Father has more education than mother	24.0		22.3		30.9
Income-to-poverty ratio	4.53	3.70	5.14	3.21	1.99	2.39
Physical health (M, range: 1–5)
Mother	4.13	0.84	4.20	0.67	3.80	1.39
Father	4.10	0.90	4.13	0.75	3.98	1.43
Depression
Mother	7.3		5.0		17.0
Father	2.4		2.2		2.6
Religious attendance (M, range: 1–5)
Mother	3.40	1.37	3.48	1.15	3.03	1.96
Father	3.41	1.46	3.56	1.21	2.75	2.09
Lived with both parents at age 15
Mother	65.7		71.1		43.4
Father	71.0		75.8		51.1
Parents’ number of children together	1.84	1.07	1.92	0.94	1.50	1.20
Parents have children with other partner
Both parents	5.0		2.7		14.5
Mother only	9.6		5.3		27.7
Father only	8.6		6.3		18.0
Neither parent	76.8		85.7		39.7
Child is a boy	63.6		66.1		53.0
Child “difficult” temperament (M, range: 1–5)	2.60	0.90	2.53	0.73	2.88	1.46

Note: All variables were measured at baseline or the 1-year survey and are weighted by city sampling weights. The number of cases (N) is unweighted.

Analytic Strategy

Latent growth curve models

Our first set of analyses consisted of latent growth curve models of children’s externalizing and internalizing behavioral problems, which we used to capture the dynamic associations between couples’ supportiveness and children’s behavior over child ages 3 through 9. As shown in Figure 1, these models assume that children’s behavioral problems are characterized by an underlying latent trajectory, which can be measured by an intercept and a slope. This trajectory varies as a function of characteristics that remain constant over time (in this case, the control variables). Deviations from this trajectory at a point in time are caused by events and characteristics that vary over time (in this case, couples’ supportiveness). Although in reality supportiveness changes gradually, because we observed it only at discrete points in time (i.e., each survey), its effect on children’s behavior can be viewed as a sort of “shock” that causes perturbations in behavioral problems from the underlying trajectory. We estimated these models (as well as our cross-lagged models) using Mplus Version 7.1.

FIGURE 1.

LATENT GROWTH CURVE MODEL OF CHILDREN’S BEHAVIORAL PROBLEMS OVER TIME.

Note: BEH = behavioral problems; Yrs = years; SUPP = supportiveness.

Fixed effects models

Although latent growth curve models usefully capture the dynamic association between couples’ supportiveness and children’s behavior, estimates are based on both between- and within-subject variation; thus, they could be biased by unobserved heterogeneity across individuals/families. To provide a more conservative test of causality, we also used fixed effects. Unlike standard regression or structural equation models, fixed effects models examine only change within the same families over time. In this way, the estimates control for observed and unobserved time-invariant factors that might be driving the association between supportiveness and behavior, although unmeasured time-varying variables could still bias the estimates. These models also controlled for the survey wave in order to capture normal developmental changes in behavioral problems over the study period. We also tested for differences in the association between supportiveness and children’s behavioral problems across developmental periods by interacting supportiveness with the survey wave measure as well as by estimating separate models for Years 3–5 and Years 5–9. We found no indication that couple supportiveness had a different effect on children at certain ages. In light of this, we present the results from our main-effects models only. These models were estimated in Stata Version 13.

Cross-lagged structural equation models

To examine the direction of the association between couple supportiveness and children’s behavior, we used cross-lagged structural equation models. These models allow for the evaluation of the primary direction of an effect between two variables that may be reciprocally related by estimating the effect of one variable at an earlier time point (t1) on another variable at a later time point (t2), while controlling for the effect of the second variable at t1 on the first variable at t2 (Finkel, 1995). As highlighted in Figure 2, our analyses focused on estimates for four paths (a – d) reflecting the reciprocal relationships between supportiveness and behavioral problems over Years 3–5 and 5–9. We also included correlations between concurrent measures of supportiveness and behavioral problems, and among repeated measures of each variable. Direct paths were also estimated between each control variable and behavioral problems and supportiveness at 3 years.

FIGURE 2.

CROSS-LAGGED STRUCTURAL EQUATION MODEL OF SUPPORTIVENESS AND CHILDREN’S BEHAVIORAL PROBLEMS OVER TIME.

Note. SUPP = supportiveness; BEH = behavioral problems.

Moderation analyses

Finally, we returned to latent growth curve models to evaluate differences in the association between supportiveness and behavioral problems by parents’ marital status and children’s gender. We compared models in which the associations between these variables were allowed to differ between married and cohabiting parents, and between boys and girls (unconstrained), to those in which these differences were constrained to be equal. A significant improvement in model fit in the unconstrained versus constrained models (using the χ² statistic) indicated the presence of moderating effects.

RESULTS

Descriptive Analyses

Mean scores for the measures of couples’ supportiveness and children’s behavioral problems at each survey wave are presented in Table 2. Means are displayed for the overall sample as well as by couples’ marital status and—for the measures of behavioral problems—child gender. Couples generally viewed their relationship as highly supportive, with means of about 2.70 (on the 1-to-3 scale) at each survey wave. Nevertheless, the relative stability in mean levels of supportiveness across surveys belied notable variation within couples over time. Over 85% of couples experienced at least one change of more than 0.10 (about 0.30–0.50 SD) in their reported levels of supportiveness between consecutive survey waves (results not shown). Finally, it is interesting that mean levels of supportiveness were almost identical among married and cohabiting couples.

Table 2.

Means on Couples’ Supportiveness and Children’s Behavioral Problems

Measure	1 Year		3 Years		5 Years		9 Years
	M	SD	M	SD	M	SD	M	SD
Supportiveness (range: 1–3)
Overall (N = 773)	2.74	0.24	2.74	0.24	2.72	0.27	2.68	0.30
Married couples (n = 445)	2.74	0.20	2.74	0.21	2.74	0.22	2.69	0.26
Cohabiting couples (n = 328)	2.70	0.35	2.72	0.36	2.64a	0.49	2.67	0.41
Externalizing behavioral problems (range: 0–2)
Overall (N = 767)			0.52	0.32	0.37	0.22	0.14	0.15
Married couples (n = 442)			0.52	0.27	0.36	0.18	0.14	0.12
Cohabiting couples (n = 325)			0.52	0.51	0.37	0.37	0.15	0.28
Boy children (n = 409)			0.52	0.31	0.37	0.21	0.15	0.14
Girl children (n = 358)			0.51	0.31	0.35	0.25	0.11b	0.16
Internalizing behavioral problems (range: 0–2)
Overall (N = 770)			0.35	0.23	0.23	0.20	0.16	0.15
Married couples (n = 443)			0.34	0.19	0.22	0.16	0.15	0.11
Cohabiting couples (n = 327)			0.41	0.42	0.27	0.37	0.18	0.28
Boy children (n = 410)			0.35	0.22	0.22	0.18	0.17	0.14
Girl children (n = 360)			0.34	0.25	0.24	0.23	0.15	0.17

Note: Supportiveness is based on couples’ reports. Behavioral problems are based on mothers’ reports. All figures weighted by city sampling weights. Number of cases is unweighted.

^aSignificantly different from married couples at p ≤ .05 (two-tailed test using weighted data).

^bSignificantly different from boys at p ≤ .10 (two-tailed test using weighted data).

As expected in a nonclinical population, means of children’s behavioral problems were generally quite low. Children displayed higher levels of externalizing than internalizing behavioral problems, at least until the 9-year survey, when the levels were similar. Levels of behavioral problems were almost identical across groups; the only exception was that boys displayed slightly more externalizing behaviors than girls at the 9-year survey. Finally, consistent with prior research (Bongers, Koot, Van der Ende, & Verhulst, 2003), children’s behavioral problems appeared to decrease as they got older (t tests indicated that these decreases were statistically significant at the p ≤ .05 level).

Multivariate Analyses

Results from the latent growth curve models of children’s behavioral problems for the overall sample, which we used to examine our first hypothesis—that couples’ supportiveness is negatively related to children’s behavioral problems—are displayed in Table 3. Note that coefficients from all of our multivariate analyses are reported in standard deviation units in order to facilitate comparisons across models and methods. Looking first at the results for externalizing behavioral problems, the level of parents’ supportiveness was negatively related to children’s behavioral problems (the one positive association was not statistically significant). Of note is that the only coefficients to reach statistical significance were for supportiveness measured at the same wave as children’s behavioral problems. This suggests that children’s behavior was more sensitive to the recent quality of their parents’ relationship than to earlier levels of supportiveness. The results are quite similar for internalizing behavioral problems, with higher levels of parental supportiveness generally associated with lower levels of internalizing behaviors. Again, children’s behavior was more strongly associated with recent, rather than earlier, levels of supportiveness. However, it is important to note that even for the coefficients that were statistically significant, the magnitude of the effects was small: For both externalizing and internalizing behavioral problems, a 1-SD increase in supportiveness (which was about 0.25–0.30 on the 1-to-3 scale) was associated with about a 0.10-SD decrease in children’s problematic behavior.

Table 3.

Standardized Coefficients From Latent Growth Curve Models of Children’s Behavioral Problems

Predictor	Externalizing behavioral problemsa (n = 767)				Internalizing behavioral problemsb (n = 770)
	3 Years	5 Years		9 Years	3 Years	5 Years		9 Years
	Level 1 equation
Supportiveness
1 Year	−0.07	−0.07		−0.02	0.02	0.05		0.02
3 Years	−0.09*	−0.05		0.07	−0.08†	−0.05		0.05
5 Years		−0.08†		−0.06		−0.11*		−0.07
9 Years				−0.10**				−0.12**
	Level 2 equation
	α		β		α		β
Intercept	6.14**		−5.32**		4.23**		−0.51

Note: Supportiveness was based on couples’ reports. Models include controls for parents’ marital status at focal child’s birth, mothers’ and fathers’ ages, mothers’ race/ethnicity, whether parents were of a different race, mothers’ education, whether fathers had more education than mothers, parents’ income-to-needs ratio, mothers’ and fathers’ physical health, mothers’ and fathers’ depression, mothers’ and fathers’ attendance at religious services, whether mothers and fathers lived with both of their biological parents when they were 15 years old, parents’ number of other children together, whether parents had children with other partners, children’s gender, and children’s temperament.

^aModel fit: χ²(29) = 8.37, root-mean-square error of approximation = 0.00, comparative fit index = 1.00.

^bModel fit: χ²(29) = 13.82, root-mean-square error of approximation = 0.00, comparative fit index = 1.00.

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

The results from our fixed effects and cross-lagged structural equation models are displayed in Table 4. (Note that for the fixed effects models, the sample size reflects the number of person-year observations contributed by the 773 families in our sample; standard errors were adjusted to account for multiple observations from families.) Our results indicate that a 1-SD increase in the level of supportiveness reported by couples was associated with a 0.08- (0.07-) SD decrease in children’s externalizing (internalizing) behavioral problems. These results were statistically significant at the p ≤ .01 and p ≤ .05 levels, respectively. The magnitude of these coefficients was slightly smaller than those from our latent growth curve models, which is not surprising given that these models implicitly controlled for both observed and unobserved time-invariant characteristics of parents and children that could be driving the association between couples’ supportiveness and children’s behavior.

Table 4.

Standardized Path Coefficients From Fixed-Effects and Cross-Lagged Structural Equation Models of Children’s Behavioral Problems (BEH) and Couples’ Supportiveness (SUPP)

Models	SUPP → BEH
	Fixed effects
Externalizing behavioral problemsa (n = 2,019)	−0.08**
Internalizing behavioral problemsb (n = 2,037)	−0.07*
	Structural equation
	SUPP 3 → BEH 5	BEH 3 → SUPP 5	SUPP 5 → BEH 9	BEH 5 → SUPP 9
Externalizing behavioral problemsc (n = 767)	−0.09**	−0.02	−0.02	−0.07*
Internalizing behavioral problemsd (n = 770)	−0.06†	−0.05	−0.05	−0.04

Note: Supportiveness based on couples’ reports. Fixed effects models include a control for survey wave. Structural equation models include controls for parents’ marital status at focal child’s birth, mothers’ and fathers’ ages, mothers’ race/ethnicity, whether parents were different race, mothers’ education, whether fathers had more education than mothers, parents’ income-to-needs ratio, mothers’ and fathers’ physical health, mothers’ and fathers’ depression, mothers’ and fathers’ attendance at religious services, whether mothers and fathers lived with both of their biological parents when they were 15 years old, parents’ number of other children together, whether parents had children with other partners, children’s gender, and children’s temperament.

^aModel fit: F(3, 1246.2) = 2.62, p ≤ .05, R² = .006.

^bModel fit: F(3, 1256.1) = 1.94, R² = .005.

^cModel fit: χ²(104) = 206.52, root-mean-square error of approximation = .04, comparative fit index = .91.

^dModel fit: χ²(104) = 224.55**, p ≤ .01, root-mean-square error of approximation = .04, comparative fit index = .89.

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

Because fixed effects models cannot differentiate the direction of these effects, we turned to cross-lagged structural equation models to examine our second hypothesis: that supportiveness would have a stronger effect on children’s behavior than vice versa. The results for externalizing behavior suggest that between Years 3 and 5 the level of supportiveness in the couple relationship was predictive of children’s behavioral problems 2 years later but that children’s behavioral problems were not predictive of the level of supportiveness in the couple relationship 2 years later. Between Years 5 and 9, however, this pattern was reversed: Parents’ supportiveness was not predictive of their children’s behavioral problems, but behavioral problems were predictive of parents’ supportiveness. For internalizing behaviors, between Years 3 and 5, couples’ supportiveness was marginally related to subsequent levels of children’s behavioral problems, but there was no indication that children’s behavioral problems were related to subsequent levels of supportiveness. Between Years 5 and 9 neither of these paths were statistically significant, even at the p ≤ .10 level.

Moderation Analyses

We returned to latent growth curve models to evaluate our third hypothesis: that the association between supportiveness and behavior would be stronger for cohabiting than for married couples and stronger for boys than for girls. The results from these models (available on request) revealed no evidence of differences by marital status. This was true regardless of whether we classified parents according to their marital status at the baseline or 9-year survey.

With regard to children’s gender, we found some evidence of differences in the association between supportiveness and children’s internalizing behavioral problems, but not their externalizing problems. Further examination indicated that whereas parents’ supportiveness appeared to be equally relevant for boys’ and girls’ internalizing behaviors at 3 and 9 years, at 5 years it was relevant only for girls. A 1-SD increase in parental supportiveness was associated with a 0.25-SD decrease in girls’ internalizing problems at 5 years but only a 0.02-SD decrease in boys’ internalizing problems at 5 years.

Robustness Checks

Finally, we ran a number of robustness checks to examine the sensitivity of our results to various analytic choices. First, we examined how our decision to combine mothers’ and fathers’ reports about the other parent’s supportiveness affected our results by reestimating our analyses using parents’ separate reports. In general, these results were similar to our main results regardless of whose reports were used, although fathers’ reports about mothers’ supportiveness tended to bear a weaker association to children’s behavior than mothers’ reports about fathers’ supportiveness. In other words, fathers’ support of mothers (based on mothers’ reports) was especially salient for children’s behavior; this association may be partly attributable to having (in this case) the same reporter for both the independent and dependent variables.

We also examined the sensitivity of our results to our decision to restrict our sample to parents who lived together over child ages 1 through 9. Not surprisingly, couples in the Fragile Families Study who lived together initially but broke up over the study period reported lower levels of supportiveness in their relationship (while together) than couples who remained together over the whole period. To determine whether supportiveness was also predictive of children’s behavior among couples who later broke up, we reestimated our latent growth curve models using a sample of parents who were in a relationship at the 3-year survey regardless of the subsequent stability of their relationship, as well as a sample of parents who were in a relationship at the 5-year survey regardless of subsequent relationship stability. For each sample, we then compared the associations between supportiveness and children’s behavior at 3 or 5 years (using a χ² goodness-of-fit test) for parents who remained together throughout the survey versus parents who would later split up. In both cases, we found no differences in the associations between supportiveness and children’s behavioral problems for parents who remained together versus those who would later split up, suggesting that our results were not driven by our decision to restrict our sample to couples with the most stable relationships.

DISCUSSION

This study sheds new light on how a key aspect of family functioning matters for an important domain of children’s well-being. Specifically, we examined how supportiveness in the parents’ relationship is linked to children’s externalizing and internalizing behavioral problems among stable married and cohabiting families. Although an extensive literature has shown that family structure is important for children’s outcomes, much less attention has been given to how positive family dynamics within families may be salient for child development and well-being—at least using large samples of longitudinal data with rigorous methods.

Overall, our results suggest that there is a significant association between co resident parents’ dyadic relationship quality and the level of children’s behavioral problems. We found that parents with more supportive relationships had children with fewer behavioral problems. This held true for both externalizing behaviors and internalizing behaviors, although in each case the magnitude of the effects was small (around or just under 0.10 SD). Although we might be concerned that this cross-sectional association simply reflects a spurious correlation between these two constructs, our more conservative fixed effects estimates revealed that within the same families, an increase in supportiveness was significantly linked to a decrease in children’s behavioral problems over the same time period.

Recognizing the growing emphasis on how children may affect parents, we also evaluated the directionality of this association as children grew from toddlers into middle childhood. For externalizing behaviors, we found some evidence for both parent and child effects, depending on children’s ages. Between Years 3 and 5, supportiveness was predictive of children’s subsequent behavior (but not vice versa), whereas between Years 5 and 9 children’s behavior was predictive of parents’ subsequent supportiveness (but not vice versa). For internalizing behaviors, parents’ relationship quality was marginally significantly associated with behavior over Years 3–5, but there was no indication that children affected parents’ relationship quality. The fact that we observed child effects only at older ages is consistent with our expectation that children have a greater influence on family dynamics during middle childhood compared to earlier years. It is also not surprising that child effects were observed only for externalizing behaviors, because acting out is likely to be more readily observed by parents and may be more disruptive to family dynamics than internalizing problems, which are more hidden.

Although we expected that supportiveness might matter more for children’s behavior among cohabiting parents than among married parents, we observed no significant differences by marital status. This was true regardless of whether we measured couples’ marital status at the baseline or the 9-year survey. Nevertheless, this finding must be interpreted with caution, because our sample was limited to cohabitors who stayed together over the entire period—an admittedly select group. The fact that our results were similar (i.e., no difference by marital status) when we reestimated our analyses including couples who coresided at 3 or 5 years but later broke up gives us greater confidence in our results.

We also found, contrary to our expectations, little evidence in differences in how supportiveness in the couple relationship was associated with boys’ versus girls’ behavior (with the exception of internalizing behavior at one time point); this primary lack of moderation by child gender is consistent with several other prior studies (Buehler et al., 1997). It appears that for the most part, all children, regardless of gender, benefit from parents’ ability to love, support, and communicate with each other.

Our main results relied on the average of mothers’ and fathers’ reports about the level of support provided by the other spouse or partner, thus capturing the reciprocal and dyadic nature of the couple relationship. The fact that models based on fathers’ reports alone showed a smaller association with children’s behavioral problems could suggest either that mothers’ support of fathers in the couple relationship matters less for children’s behavior (than does fathers’ support of mothers) or that the couple average (and mothers’ reports alone) are somewhat biased by the shared method variance that results when the same reporter provides information about both the independent and dependent variables (because mothers also reported on behavioral problems; Marsiglio, Amato, Day, & Lamb, 2000).

Taken together, our findings underscore the emphasis in family systems theory on the interconnectedness of family relationships and individual well-being (Minuchin, 1988; O’Brien, 2005). Although an extensive literature has highlighted the importance of stable, two-parent families for children’s well-being (e.g., McLanahan et al., 2013), our research suggests that variability within stable families also matters. In other words, even among a group of high-functioning families with very positive couple relationships, the degree to which biological parents communicate effectively and provide mutual emotional support has a small, positive association with children’s behavior. These findings are consistent with existing research on the implications of negative aspects of parents’ relationship—such as conflict and discord—for children (Buehler et al., 1997; Buehler & Gerard, 2002; Cummings, Schermerhorn, Davies, Goeke-Morey, & Cummings, 2006). Yet our results extend this literature by indicating that positive aspects of the parents’ couple relationship—such as greater trust, empathy, and effective communication—have beneficial effects for children’s externalizing and internalizing behaviors as well. Also, our findings suggest that parents not only affect children but, at least during middle childhood, the externalizing problems that children display also matter for the quality of parents’ relationship.

We believe our research provides a useful extension to the literature, but we also acknowledge several limitations. First, the Fragile Families data do not contain general measures of negative dimensions of parents’ relationship quality over time; they include only extreme measures of controlling or violent behavior, which do not reflect more typical couple conflict. Ideally, we would be able to look at both negative and positive aspects of relationship quality to better replicate prior work and consider the conjoint versus independent associations of these two aspects of relationship quality for children’s behavior. Second, although our multiple measures of both relationship quality and children’s behavior allowed us to use methods designed to account for unobserved heterogeneity, our results could still be biased by unobserved differences that varied over the survey period. For example, if mothers or fathers became unemployed between waves, the stress of such job loss could have both diminished parents’ relationship quality and increased children’s behavioral problems. Third, our results can be generalized only to families in which the parents lived together over the entire 8-year period from infancy to middle childhood (child ages 1–9). Although most marriages last at least 8 years (Bumpass & Lu, 2000), this is beyond the median duration of cohabiting relationships; hence, as noted previously, our results concern a select group of stable cohabitors with children. Fortunately, our robustness checks using couples who would later break up suggest that our selective sample was not driving our results.

This work also suggests several directions for future research. First, because there were notable differences in union stability by race/ethnicity, there may also be differences in how couple supportiveness is linked to children’s behavior. Second, as suggested earlier, it would be instructive to evaluate supportiveness along with more general measures of conflict in relationships in order to consider the multiple dimensions of relationships as linked to children’s well-being. Third, it would be useful to consider additional aspects of children’s well-being, including cognitive outcomes and other measures of socioemotional health.

This study provides new information about how supportiveness in coresident parents’ relationship is linked to children’s externalizing and internalizing behavioral problems when they are approximately 3, 5, and 9 years old. We found that supportiveness is significantly—though only slightly—associated with fewer behavioral problems and that this association persists across methods that reduce unobserved heterogeneity and evaluate directionality. These results emphasize that, beyond the structure of families, what goes on inside families (of the same stable structure) has important implications for children’s well-being.