DYADIC PARENTING AND CHILDREN’S EXTERNALIZING SYMPTOMS

Abstract

We explore dyadic parenting styles and their association with first-grade children’s externalizing behavior symptoms in a sample of 85 working-class, dual-earner families. Cluster analysis is used to create a typology of parenting types, reflecting the parental warmth, overreactivity, and laxness of both mothers and fathers in two-parent families. Three distinct groups emerged: Supportive Parents, Mixed-Support Parents and Unsupportive Parents. Results indicate that dyadic parenting styles were related to teacher-reported externalizing symptoms for boys but not for girls.

Numerous studies have established that parenting has important implications for many aspects of child development (e.g., Baumrind, 1996; Kaufmann et al., 2000, Lamborn, Mounts, & Steinberg, 1991). Much research has focused on two major dimensions of parenting namely: (a) parental warmth and responsiveness, and (b) demandingness or control (Amato & Fowler, 2002; Barber, Stolz, & Olsen, 2005). These two dimensions of parenting generally tap into the emotional climate and parental control of parent-child interactions and are the key dimensions that comprise Baumrind’s parenting typology of authoritative, authoritarian, and permissive parenting styles (Baumrind, 1991).

Despite the large literature on parenting, little research has examined the combined effects of mothers’ and fathers’ parenting on children’s development, although children reared in two-parent families clearly experience the influence of both parents (Martin, Ryan, & Brooks-Gunn, 2007). Moreover, findings suggest that parents contribute differently to children’s development (Parke et al., 2005). For example, a mother may employ more warmth as a compensatory mechanism if her partner uses more harsh disciplinary tactics. Martin et al. (2007) argued that, “it is incumbent upon parenting scholars to devise ways of depicting children’s parenting experiences in two-parent families that reflect the full complexity of the family (p. 436).” It is also important to note that parenting differs as a function of social class, such that socioeconomic disadvantage is associated with less warm and more controlling parenting (Conger, McCarty, Yang, Lahey, & Burgess, 1984). Thus, the purpose of the present study is to examine how mothers’ and fathers’ parenting, examined in combination, is related to behavior problems in their first-grade children within a unique sample of low-income, working families.

Parenting Types and Child Outcomes

One of the best known and most frequently researched parenting typologies was developed by Baumrind (1966, 1996) and is based on the dimensions of warmth/responsiveness and demandingness/control. In her typology, parents who are warm and responsive as well as effective at setting appropriate limits for their children’s behavior are termed authoritative, while parents who are high in warmth and responsiveness but low in control are referred to as permissive parents. Finally, authoritarian parents are high in control, but low in warmth. “Unengaged” parents, low on both warmth and control, and “good enough” parents, parents who fall in the average range of control and warmth, were added in later studies (i.e., Baumrind, 1996).

Decades of research examining the relationship between parenting styles and child development have found that across a wide range of ages and family types, children with an authoritative parent show better adjustment in multiple domains including academic competence, behavior problems, and psychosocial development, compared to peers with authoritarian, permissive, and neglectful parents (e.g., Baumrind, 1996; Lamborn et al., 1991). When looking specifically at children’s behavior problems, the pattern of findings is quite consistent. For example, using nationally representative data from the National Survey of Families and Households (NSFH, Sweet, Bumpass, & Call, 1988), it was found that low levels of parental support and the use of harsh punishment were associated with more behavior problems in children ages 5–11 (Amato & Fowler, 2002). In another study, children from authoritative homes had fewer behavior problems than peers with authoritarian or permissive parents, even after considering gender, grade level, ethnicity, and family income (Kaufmann et al., 2000). El-Sheik and Elmore-Staton (2004) in their sample of 9–12 year olds, found that parent-child conflict plays a more robust role in predicting externalizing outcomes in preadolescent children than internalizing outcomes. More recently, Scaramella, Neppi, Ontai, and Conger (2008) have shown that the social context of the parenting process, specifically socio-economic status, plays a significant role in predicting both parenting and adolescent outcomes. Their results indicate that across parental generations, exposure to economic disadvantages is related to harsher parenting behavior and increases in children’s externalizing problems. These findings are particularly relevant to the current study in that all parent participants were in blue-collar occupations and have low levels of education, both conditions that may place children at risk for harsher parenting.

The current research specifically focuses on children’s externalizing behavior and aims to extend the work by Scaramella et al. (2008) by examining the relationship between harsh discipline and externalizing behaviors with a sample of younger children from low-income families. The developmental literature indicates that the toddler period is noted for particularly high rates of externalizing behaviors that decline between the ages of 2 to 6 (Gilliom & Shaw, 2004). Gilliom and Shaw argued, however, that under conditions of less warm and more overreactive (i.e., harsh) parenting these normative declines may not occur. In the current study, our focus on children entering the first grade (6–7 years of age) allows us to examine an important age where differences in externalizing behaviors may occur as a function of ineffective parenting. In contrast, internalizing behaviors (e.g., depression, anxiety), are more likely to emerge in early adolescence between 13 to 15 years of age (Gilliom & Shaw, 2004).

In addition to developmental stage, research on gender differences indicates that boys have higher overall rates of externalizing problems than girls (e.g., Deater-Deckard & Plomin, 1999); however, much of this literature looks at broad age ranges from early childhood through adolescence. Thus, gender differences among 6 to 7 year old children has not been well documented. There is conflicting evidence regarding gender differences in boys’ and girls’ reactions to parenting. At least two studies have failed to find evidence of gender differences in the impact of early parental supportiveness (McCarthy, Zimmerman, Digiueseppe, & Christakis, 2005) or mothers’ overreactivity on the development of externalizing symptoms in preschoolers (O’Leary, Slep, & Reid, 1999). Other findings suggest that parental disagreements are related to different outcomes for boys and girls (O’Leary & Vidair, 2005). Overall, the research reviewed above suggests that age and gender of the child may moderate the relationship between parenting and child behavior. The current study will focus specifically on children entering the first grade, and gender will be considered in analyses; however, the empirical data point to no specific predictions regarding child gender.

The Need for a Dyadic Approach

Research on parenting has focused primarily on the unique contributions of mothers and fathers to parenting. (McBride & Rane, 1998). Two areas of research, however, that have begun to examine the combined effects of parenting are studies of parental agreement and research on coparenting (i.e., Block, Block & Morrison, 1981; Feinberg, 2003). Coparenting and parental agreement studies offer the advantage of including both parents; however, each is limited as an indicator of dyadic parenting. Coparenting is defined as the way that parents/caregivers relate to each other in the role of parent and has been consistently linked, both cross-sectionally and longitudinally, to parental adjustment, parenting style, and child adjustment (e.g., Feinberg, 2002; Van Egeren, 2003). Although coparenting has been found to be an important factor in its own right, it focuses on the quality and strength in parents’ relationship to each other rather than on parenting styles toward children.

Parental agreement, or the extent to which parents agree on child-related topics, including moral values, behavioral expectations and discipline, is another potential indicator of the degree of consistency in parenting styles between caregivers (Feinberg, 2003). Disagreements about childrearing issues have been linked to child behavior problems in preschoolers and kindergarteners (Block, Block, & Morrison, 1981; Deal, Halverson, & Wampler, 1989) and adolescents cross-sectionally and longitudinally (Vaughn, Block, & Block, 1988). However, assessments of parental agreement or disagreement about childrearing practices are limited by the lack of specificity as to the standard of parenting to which parents are agreeing.

Only a handful of researchers have examined the combined effects of both mothers’ and fathers’ parenting styles, specifically warmth and control, as they relate to child outcomes. Forehand and Nousiainen (1993) found that adolescents’ reports of more supportive parenting from both mothers and fathers were significantly associated with teachers’ reports of the adolescents’ cognitive competence and parents’ reports of their behavior problems. Chambers, Power, Loucks, and Swanson (2001) found that incarcerated adolescent males who perceived poor parenting from both parents reported the highest levels of psychological distress, while those who experienced positive parenting from both parents were least depressed. Both of these studies included only adolescents, and may not generalize to families with younger children.

More recently, using data from the Early Head Start Research and Evaluation Project, Martin et al. (2007) cross-classified parents on six different parenting behaviors representing levels of warmth and control, and linked resultant clusters to children’ cognitive outcomes. Using K-means cluster analysis they devised four primary pairings of parents (a) both supportive, (b) both unsupportive, (c) mother supportive/father unsupportive, and (d) father supportive/mother unsupportive. Children with two supportive parents scored highest on both math and language outcomes while children with two unsupportive parents scored lowest. Children with one supportive parent scored in the middle and the gender of the supportive parent was inconsequential. These authors concluded that the combined effects of two supportive parents are additive. The current study will explore mothers’ and fathers’ combined contributions to parenting, what we call a dyadic approach, using a cluster analytic strategy similar to Martin et al. (2007) and linking parental clusters to children’s behavior problems.

The Social Context of Parenting

Despite increasing recognition by many researchers of the importance of considering family context (Pinderhughes, Nix, Foster, Jones, & The Conduct Problems Prevention Research Group, 2001), the majority of what we know about parenting is based upon samples of White, middle-class families (Feinberg, 2003), with some notable exceptions (e.g., Bluestone & Tamis-LeMonda, 1999; Calzada & Eyberg, 2002; Tudge, Hogan, Snezhkova, Kulakova, & Etz, 2000). Of interest in the current project is how family processes may differ for dual-earner, working-class families. Research indicates that parenting styles within lower socioeconomic status (SES) families may be more authoritarian and permissive than in middle-class families (Querido, Warner, & Eyberg, 2002). Dual-earner families with fewer financial resources are likely to be more stressed, which may diminish effective parenting skills. However, in dual-earner families, fathers may be more involved in parenting and, as such, the importance of dyadic parenting may take on increased meaning.

The Current Study

The purpose of the present study was to examine how both mothers’ and fathers’ parenting, assessed by warmth, overreactivity, and laxness, are related to the behavior problems of six- to seven-year-old children. We examined externalizing behaviors for three main reasons: (a) the clinical literature indicates that externalizing behaviors can be assessed as early as six years of age while internalizing behaviors are more likely to emerge beyond age 12, (b) early externalizing behaviors in toddlers decline normatively from ages 2–6, however, this pattern of decline may be interrupted by unsupportive parenting, and (c) internalizing behaviors are harder for teachers to assess in group settings since depressed children may simply appear as quiet.

Our first aim was to develop a typology of parenting across couples. Since demographic and psychological characteristics are highly associated with parenting behaviors, it is likely that many couples will exhibit similar styles (Martin et al., 2007). Thus, we predicted that there would be groups in which both mothers and fathers exhibit moderate to high warmth and low laxness and overreactivity (i.e., authoritative parenting). Similarly, there is likely to be a group where parents match on low warmth and high overreactivity and laxness (i.e., authoritarian parenting). Finally, we expected to find groups in which mothers may be high on supportive parenting while fathers are low and, vice versa, one in which fathers are highly supportive and mothers are low.

A secondary aim upon developing the typologies was to examine demographic differences across the established parenting types to provide more information about the social context of the parenting groups. Based on previous research, we expected lower income and younger parents to be over represented in groups that use harsher discipline styles (Scaramella et al., 2008).

If we are successful in identifying meaningful typologies of parents, our third research question addresses differences in children’s externalizing behaviors as a function of the parenting cluster groups. There is little empirical basis for predictions about how cross-classified parenting styles across mothers and fathers may be related to child outcomes. Based on Martin et al.’s (2007) results, which looked only at cognitive child outcomes, we predicted that typologies in which both parents report supportive parenting behaviors (i.e., high warmth, low overreactivity, low laxness) will have children with fewer behavior problems than typologies where both parent are unsupportive (i.e., low warmth, high overreactivity, high laxness). In cases where there is only one supportive parent, we predicted that children would have fewer behavior problems than when there is no supportive parent but more problems than children with two supportive parents.

The final research question examined the issue of similarity in parental ratings in a way not addressed with the cluster analytic approach, which only presents groups of couples with similar profiles. Specifically, intraclass correlations were used as indicators of the similarity between dyads, and this level of agreement between partners, regardless of how high or low their parenting scores, is related to child behavior problems. We hypothesized that, even when highly correlated, parents’ intraclass correlations will be unrelated to child outcomes since simple agreement does not reflect the type of parenting that is used (e.g., supportive or unsupportive).

These research questions were examined in a sample of working-class, dual-earner couples as their oldest child enters the first grade. Cluster analysis was used to identify meaningful types of dyadic parenting styles for 6–7 year old children. Specifically, we were interested in identifying groups of parents, both mothers and fathers, who exhibit unique profiles across a variety of parenting measures (e.g., warmth, harsh discipline, laxness). As Crouter and Manke (1997) noted, it is difficult to combine more than two or three variables in an a priori fashion, using cutoffs for “high” and “low” values, which can often result in large numbers of groups, some of which are small and not substantively interesting. Moreover, it is often not accurate to identify meaningful cutoff points when mothers’ and fathers’ scores have different distributions. Thus, cluster analysis is a useful technique for simultaneously taking into consideration several qualities of mothers’ and fathers’ parenting. Parental cluster groups that emerge from the cluster analysis, reflecting both mothers’ and fathers’ parenting styles, were then used to examine the relationship between dyadic parenting and children’s externalizing behavior. To avoid the problem of shared method variance, mothers’ and fathers’ ratings of their own parenting were used in the study, along with teacher reports of children’s externalizing behaviors.

Method

The data in the current analysis were based on a subsample of 85 two-parent families from the Work and Family Transitions Project, a 5-year longitudinal study of working-class, dual-earner couples experiencing their first transition to parenthood. Data were drawn from the interviews conducted when the oldest child was in the first grade. Each family was paid $200 for a single 3-hour visit to the participants’ home. Of the total sample, 85 intact two-parent families and 79 teachers completed the follow-up portion of the study and were included in the analyses.

Participants

Participating couples were recruited into the larger project from prenatal education classes at local hospitals. Only couples who were still married or cohabiting at the time of the follow-up were included in the analyses (mean length of relationship was 10.1 years). The average age of fathers was 37.3 years, while the average age of the mothers was 35.4 years. Approximately 94% of the men and 95% of the women were White. All parents were employed in working-class occupations. On average, men worked 47.5 hours per week with an annual income of $44, 651 (median = $43,610), while women worked 31.7 hours per week with an annual income of $28,269 (median = $26,000). According to 2004 U.S. Census data, the median household income in Massachusetts was $53,657; however, as this figure includes a significant number of single parent households, it is likely a substantial underestimation of the equivalent figure for dual-earner families. The majority of parents in the study were working at the time of the interview (85% of mothers and 94% of fathers); however, given the nature of the sample, it is not surprising that the families experienced many transitions in and out of employment (due to seasonal work and variable weekly work hours for example). Fifty-nine percent of the couples worked the same shift as their partner, while 41% worked opposite shifts. About half of the mothers had earned a high school diploma or GED as their highest level of education (n = 45), 2 had no degree, and 38 women had an Associate’s degree. Nearly 75% of the men had earned a high school diploma or its equivalent (n = 64), and 21 men had an Associate’s degree. The average age of the target children in the sample was nearly 7 years old (82.4 months). Of the children in the sample with complete data from teachers, 35 were male, 43 were female, and one family had mixed gender twins. The majority of the families in the study had two children at the time of follow up (n=58), 11 families had only one child, and the remaining 16 had three or more children (average age of second child was 4 years old).

Measures and Variables

Parental Warmth (Alabama Parenting Questionnaire, APQ)

The APQ is a shortened version of a 42-item self-report measure of parenting (Shelton, Frick & Wootton, 1996). The version used in the present study was designed to reflect only positive parenting behaviors with 6 items reflecting parental warmth. Parents responded on a 5-point scale ranging from “never” to “always.” A sample item included: “You praise your child if he/she behaves well.” The alpha coefficients for warmth were .80 for mothers and .74 for fathers.

Parental Overreactivity and Laxness (The Parenting Scale,(TPS)

The TPS is a 30 item self-report measure, which indicates the extent to which parents employ dysfunctional discipline practices with their young children (Arnold, O’Leary, Wolff, & Acker, 1993). We used a revised version containing 21 items reflecting two subscales: Laxness and Overreactivity. The subscale of overreactivity contains 10 items reflecting the degree to which parents respond to their children with anger, irritability, and meanness. The 11-item laxness subscale captures parents’ tendency to give in, to allow rules to be unenforced, or to provide positive consequences for rule violations. Items are rated on a 7-point scale, with parents indicating their position between two anchor statements. For example, an item from the overreactivity subscale asked parents to select where they fall on a continuum between the following alternatives: “When my child misbehaves… I handle it without getting upset,” or “I get so frustrated or angry that my child can see I’m upset.” A sample item from the laxness subscale included: “When I say my child can’t do something… I let my child do it anyway,” or “I stick to what I said.” Alpha coefficients were .83 for mothers’ laxness and .84 for mothers’ overreactivity. For fathers, the alpha coefficients were .85 for laxness and .78 for overreactivity.

Children’s Externalizing Behaviors (Behavior Assessment System for Children, BASC)

The BASC (Reynolds & Kamphaus, 1992) is an instrument designed to evaluate the behavior and/or self-perceptions of children aged 2.5 to 18 years of age. It includes parent and teacher versions and assesses various aspects of behavior and personality, including both positive (adaptive) and negative (clinical) dimensions. The teacher version (TRS-C) consists of 148 items. The Externalizing Problems composite scale (scale reliability = .93) assesses children’s behavior problems. The Externalizing Problems scale is comprised of three subscales: Aggression, Hyperactivity, and Conduct Problems, with alphas of .93, .92, and .62, respectively.

Results

Means and standard deviations of the parenting attributes of laxness, over-reactivity, and warmth for mothers and fathers in the three groups are reported in Table 1. Bivariate correlations were used to examine the intercorrelations between parenting variables. Overall, the pattern of correlations revealed only a few modest links between mothers’and fathers’ parenting. Specifially, mother overreactivity was positively correlated with mother laxness (r = .32) and father overreactivity (r = .28) and negatively correlated with mother warmth (r = −.38) and father warmth (r = −.24).

Table 1.

Mothers’ and Fathers’ Scores for Parenting Attributes for the Total Sample as well as the Cluster Groups

	Total Sample (n = 85)		Group 1 (n = 29)		Group 2 (n = 28)		Group 3 (n = 28)
			Supportive Parenting Couples		Mixed-Support Couples		Unsupportive Parenting Couple
	M	SD	M	SD	M	SD	M	SD
Mothers’ Laxness	2.91	.83	2.53	.77	2.93	.70	3.33	.79
Boys	2.94	.83	2.49	.67	3.20	.88	3.20	.78
Girls	2.92	.82	2.57 a	.87	2.76 ab	.53	3.45 b	.80
Mothers’ Overreactivity	2.95	.84	2.20	.57	3.10	.44	3.59	.78
Boys	3.10	.76	2.34 a	.40	3.30 b	.36	3.75 b	.56
Girls	2.83	.89	2.07 a	.68	2.96 b	.44	3.45 b	.92
Mothers’ Warmth	4.38	.46	4.59	.25	4.65	.31	3.90	.38
Boys	4.48	.45	4.65 a	.21	4.79 a	.17	4.03 b	.44
Girls	4.31	.47	4.53 a	.27	4.56 a	.35	3.79 b	.29
Fathers’ Laxness	2.80	.84	2.77	.99	2.94	.70	2.74	.81
Boys	2.66	.82	2.55	.96	3.14	.63	2.37	.65
Girls	2.94	.84	2.97	1.00	2.81	.73	3.10	.81
Fathers’ Overreactivity	2.69	.72	2.18	.47	3.31	.49	2.64	.65
Boys	2.71	.75	2.23 a	.54	3.42 b	.43	2.62 a	.73
Girls	2.70	.69	2.13 a	.42	3.25 b	.53	2.67 a	.61
Fathers’ Warmth	4.27	.44	4.52	.28	4.24	.42	4.06	.49
Boys	4.34	.41	4.52	.31	4.29	.35	4.18	.48
Girls	4.22	.47	4.51 a	.25	4.21 ab	.47	3.96 b	.49

Note. Means with different subscripts between boys and girls differ at p < .05.

The first goal was to identify patterns of parenting attributes based on fathers’ and mothers’ scores on warmth, over-reactivity, and laxness via cluster analysis. As a first step, the six parenting variables (three for mothers, three for fathers) were standardized to ensure equal variances across each of the clustering variables. Cluster groups were created using the SPSS program and selecting the K-means, two-step cluster option, with the log-likelihood method chosen as the distance measure. This index detects profile differences across clustering variables to determine group membership. Other similarity measures (e.g., Euclidian distance) were not chosen because they consider mostly level (i.e., high, medium, or low; Aldenderfer & Blashfield, 1984), which can result in couples with very different profiles being grouped together. Cluster analysis is an inherently atheoretical statistical application that relies on theoretical models to make sense of the patterns in the data. Thus, the exploratory analyses investigated different cluster solutions with 2, 3, and 4 groups, and it was found that the cluster solution with 3 groups made the most conceptual sense. A two-cluster solution was overly simplistic, and a 4 cluster solution yielded cell sizes that were too small. Because cluster solutions can be sensitive to the order of cases in the data, the cluster solution was replicated on the data resorted in random order and was found to be invariant to order effects. As Figure 1 illustrates, the three cluster groups represented conceptually distinct types of profiles based on both partners’ parenting styles.

Figure 1.

Cluster Group Profiles of Mothers’ (M) and Fathers’ (F) Laxness, Over-reactivity and Warmth as Parents

Group 1: Supportive Parenting group

Mothers in group 1 (n = 29) scored significantly lower than mothers in group 3 on laxness, and significantly lower than mothers in groups 2 and 3 on over-reactivity. In addition, mothers in group 1 were significantly higher in their reports of parenting warmth than mothers in group 3. Fathers in group 1 scored significantly lower on over-reactivity than fathers in either of the other two groups, and were also relatively higher in warmth than both groups 2 and 3. Notably, the couples in group 1 were the only group in which both fathers and mothers reported more parental warmth than the other groups, and were also lower on over-reactivity. Parents in this cluster, designated the Supportive Parenting group were characterized by parental warmth and lack of over-reactivity.

Group 2: Mixed-Support Parenting group

Mothers in group 2 (n = 28) were average on laxness, and were not significantly different from groups 1 and 3. Mothers in this group were also average on over-reactivity, being significantly more over-reactive than group 1 and significantly less over-reactive than group 3. Mothers were characterized by a higher level of warmth in this group relative to mothers in group 3. Thus, mothers in group 2 were high on warmth but average in their over-reactivity and laxness. Fathers in group 2 were the highest of any group on over-reactivity, and were lower on warmth than group 1. Fathers in this group, unlike their partners, were characterized by low warmth. To characterize the mixed support received by mothers and fathers, this group was termed the Mixed-Support Parenting group.

Group 3: Unsupportive Parenting group

Mothers in group 3 (n = 28) were significantly higher in their laxness than group 1. In addition, mothers in this group were significantly more over-reactive than mothers in either group 2 or group 1. With respect to warmth, mothers in this sample were significantly lower than the other groups. Overall, mothers in group 3 showed the most negative parenting style out of the three groups. Fathers in group 3 were about average relative to other fathers on laxness and were significantly more over-reactive and significantly less warm than group 1. Thus, group 3 was unique for being the only group in which both parents were low on warmth. They were designated the Unsupportive Parenting group.

In order to test whether cluster groupings meaningfully differentiate between groups, a mixed model analysis of variance was used, a technique where dyad membership can be treated as a repeated measure, thus accounting for the non-independence of the data (Helms, Crouter, & McHale, 2003). A 3 (cluster) x 3 (parenting attribute: lax, overreactive, warm) x 2 (spouse) mixed model analysis of variance (ANOVA) was conducted on the unstandardized clustering variables with parenting attributes and spouse treated as within-groups factors. Because cell sizes were unequal, we examined Type III sums of squares. Significant univariate findings were followed up with Tukey tests. Because cluster analysis is designed to identify homogeneous groups that are maximally different, it was not surprising that the omnibus test for cluster group was significant, F(2, 82) = 15.21, p < .001. Follow-up tests confirmed that the patterning of spouses’ scores differed across the three groups (see Table 1). Fathers’ laxness was the only parenting attribute that did not differ across clusters; in addition, clusters did not differentiate couples on this variable. Cluster groups were not different from what would be expected by chance with respect to child gender (χ² = .52, df = 2, ns); boys and girls were equally represented in the three groups.

Correlates of the Typology of Parenting Styles

To place our parenting typologies in context, our second goal was to examine possible correlates of the groups. Specifically, a series of Chi-square analyses revealed that cluster groups were not different in terms of how likely they were to have their oldest child diagnosed with a medical, emotional or developmental problem (χ² = .90, df = 2, ns). Groups were also equally representative of whether parents worked opposite shifts from each other (χ² = .32, df = 2, ns). Results indicated that mothers and fathers had similar levels of education within the three cluster groups (χ² = .39, df = 2, ns). Chi-square analysis regarding whether clusters differed in the number of children in the family was invalid due to low cell counts (i.e., the majority of families had 2 children, while cell counts for clusters with 1 child or with 3 or more were low).

Analyses were performed to determine if the cluster groups differed on mothers’ and fathers’ ages, work hours, (individual) incomes, family income, and involvement (see Table 2). Results showed that mothers and fathers in group 2, the Mixed-Support Parenting group, were significantly younger and had lower family incomes than parents in groups 1 and 3. There were no significant differences between groups in terms of mothers’ work hours, fathers’ work hours, or mothers’ and fathers’ individual incomes.

Table 2.

Means and Standard Deviations for Correlates of Cluster Membership

	Group 1 (n = 29)		Group 2 (n = 28)		Group 3 (n = 28)
	M	SD	M	SD	M	SD
Mother age**	36.75 a	4.40	32.73 b	3.81	36.51 a	4.10
Mother wk hours	35.92	15.85	30.70	16.06	25.66	15.59
Mother income	$34,365	$19,404	$27,352	$18,478	$24,498	$16,916
Mother involvement	5.37	.75	5.41	.90	4.89	.81
Father age**	39.69 a	4.93	34.44 b	3.39	37.93 a	2.99
Father wk hours	49.08	10.65	46.99	8.35	46.18	8.55
Father income	$59,195	$28,208	$46,809	$13,603	$52,833	$16,290
Father involvement	4.08	.96	3.80	.91	3.79	1.08
Family income*	$93,560 a	$31,443	$74,162 b	$20,048	$77,332 a	$16,612

Note. Means with different subscripts differ at p < .05.

Between subjects effects are indicated as follows:

^*p < .05;

^**p < .01.

Cluster Group and Teacher Reports of Behavior Problems

Table 3 contains the means and standard deviations of teachers’ reports of boys’ and girls’ overall externalizing symptoms and subscale scores. For the full sample, the mean for externalizing symptoms was 46.11 (s.d. = 6.71) (boys m = 48.79, s.d. = 8.42; girls m = 44.35, s.d. = 4.56). A 3 (cluster) X 2 (child gender) ANOVA was performed with cluster and gender as between-subjects factors to determine if the cluster groups were different on teacher reports of behavior problems and if there was a gender difference. Results indicated there was a main effect of cluster group membership such that teachers reported the lowest levels of externalizing symptoms for cluster 1 and the highest for cluster 3 [F(2, 73) = 4.30, p < .05]. In addition, there was a main effect of child gender indicating that teachers rated boys as having more externalizing symptoms than girls [F(1, 73) = 11.42, p < .01], which was qualified by a child gender X cluster group interaction [F(2, 73) = 3.21, p < .05]. As shown in Figure 2, boys displayed a higher level of externalizing symptoms than girls overall, but the difference was much greater for boys in the Unsupportive Parenting group.

Table 3.

Means (and Standard Deviations) for Teacher Reports of Externalizing Symptoms

	Group 1 (n = 27 Couples)	Group 2 (n = 27 Couples)	Group 3 (n = 25 Couples)
	Supportive Parenting Couples	Mixed-Support Couples	Unsupportive Parenting Couples
	M (SD)	M (SD)	M (SD)
Externalizing
Full sample	44.07 (3.74)	46.96 (6.55)	48.04 (8.77)
Boys	45.15 (3.93)a	48.40 (8.34)ab	53.90 (10.65)b
Girls	43.07 (3.39)	46.12 (5.35)	44.13 (4.26)
Aggression
Full sample	43.81 (3.18)	48.07 (7.35)	48.00 (8.14)
Boys	44.08 (3.15)a	49.10 (8.60)ab	52.70 (9.33)b
Girls	43.57 (3.30)	47.47 (6.72)	44.87 (5.59)
Hyperactivity
Full sample	44.67 (5.21)	46.78 (8.53)	49.04 (12.49)
Boys	46.31 (5.88)a	49.20 (10.04)ab	58.20 (14.93)b
Girls	43.14 (4.15)	45.35 (7.47)	42.93 (4.91)
Conduct problems
Full sample	45.07 (3.74)	46.70 (4.29)	47.16 (5.10)
Boys	46.08 (4.05)	47.00 (5.66)	49.40 (6.02)
Girls	44.14 (3.30)	46.53 (3.43)	45.67 (3.90)

Note. Overall between-subject effects are bolded.

Means with different subscripts differ at p < .05.

Figure 2.

Boys’ and Girls Externalizing Symptoms as a Function of Cluster Membership

Next, a 3 (cluster) X 2 (child gender) multivariate analysis of variance (MANOVA) was computed with the three subscales of externalizing behavior problems (aggression, hyperactivity, and conduct problems) as the outcome measures and cluster and gender as between-subjects factors (see Table 3). Results indicated that there was a main effect of cluster for aggression [F(2, 73) = 4.94, p < .05], and hyperactivity [F(2, 73) = 3.31, p < .05], but the cluster groups were not different on conduct problems. Tukey tests indicated that children in the Supportive Parenting group had the lowest levels of teacher-reported aggression, compared to the Mixed-Support and Unsupportive Parenting groups. A significant interaction between child gender and cluster group emerged for hyperactivity [F(2, 73) = 4.36, p < .05] such that boys in group 3 had more symptoms than girls in that group on teacher reports of hyperactivity.

Intraclass Correlations and Children’s Behavior Problems

It has been argued that parental agreement in parenting styles is associated with a higher quality of parenting (i.e., Deal, Halverson, & Wampler, 1989; Vaughn, Block, & Block, 1988). To test this hypothesis, intraclass correlations (ICC) were calculated as a measure of the similarity between spouses. ICC is a statistic that represents how closely two individuals’ scores on a set of items match and ranges from −1 to +1, indicating perfect similarity and total dissimilarity, respectively. Unlike cluster analysis, ICC are not dependent on level (how high an individual scores) or rank order (like a Pearson correlation), but rather with the consistency between two people’s scores on a given measure. As a first step, ICC’s on mothers’ and fathers’ warmth, overreactivity and laxness were correlated with children’s behavior problems. Secondly, ICC scores were examined as moderaters of the relationship between parenting styles and children’s behavior problems.

ICC’s were calculated based on a procedure described by Maguire (1999) to reflect the similarity between mothers and fathers in terms of their warmth (mean = .404; SD = .296), laxness (mean = .245; SD = .233), and over-reactivity (mean = .457; SD = .223). Bivariate correlations were conducted between ICC scores and parents’ levels of warmth, laxness, and over-reactivity, but no significant correlations emerged. Next, an ANOVA with parent cluster as the grouping variable indicated that the clusters were not significantly different in their ICC’s for warmth, over-reactivity, or laxness. For example, mothers and fathers in the Supportive Parenting group were not more similar to each other in their warmth, over-reactivity, or laxness, than parents in either of the other two groups. Next, parents’ ICC scores on warmth, laxness, and over-reactivity were associated with children’s externalizing symptoms and the subscales of aggression, hyperactivity, and conduct problems. Again, no significant relationships emerged.

Finally, analyses were conducted to explore whether similarity (ICC) would moderate the relationship between dyadic parenting (cluster groupings) and children’s behavior problems. Interaction terms were created between the three cluster groups and parents’ ICC’s for warmth, overreactivity, and laxness and tested in regression analyses. Results indicated that parental similarity in warmth and over-reactivity (but not laxness) did moderate the relationship between cluster group membership and children’s behavior problems as follows. Although the overall interaction of cluster group and over-reactivity ICC and externalizing behavior [B = 21.40, s.e. = 8.97, p < .05] was significant, upon further analysis the relationship held only for the subscales of aggression and conduct problems. Children’ conduct problems in the Unsupportive Parenting group were higher when parents were more similar in over-reactive parenting, while the similarity between parental over-reactivity was unrelated to differences in conduct problems in the Supportive Parenting and Mixed-Support Parenting groups. In addition, parental similarity in warmth also moderated the relationship between cluster group and children’s behavior problems. Children in the Unsupportive Parenting group had fewer conduct problems when their parents were similar (as opposed to different) in their warmth, while children in the Mixed-Support Parenting group had fewer conduct problems when their parents were less similar in their warmth [B = − 8.86, s.e. = 4.40, p < .05].

Discussion

The present study explored the combined parenting styles of mothers and fathers as they related to the behavior problems of their first-grade children. Exploratory cluster analysis was used to categorize a sample of working-class families based on their parenting styles. The relationship between dyadic parenting and children’s teacher-reported behavior problems was also examined, as was the question of gender differences in the association between parenting styles and externalizing symptoms. Cluster analyses revealed three distinct groups of parents, replicating in part Baumrind’s categorization of parenting at the individual level. Specifically, the Supportive Parenting group contained parents with styles similar to Baumrind’s (1966) authoritative parent, while parents in the Unsupportive Parenting group were similar to the authoritarian style described by Baumrind. Mothers and fathers in the Mixed-Support Parenting group were the most distinct from each other in their parenting styles. Mothers in this group displayed high warmth and average laxness and over-reactivity, similar to Baumrind’s “good enough” parent. Fathers, on the other hand, were low in warmth and the highest of any group on over-reactivity, making them similar to the authoritarian parenting style. It is interesting to note that no comparable group emerged from this sample in which fathers were the more authoritative parent.

We then examined linkages between parenting clusters and child behavior problems. As hypothesized, Supportive Parenting was associated with fewer reports of externalizing behavior problems as compared to the less supportive parenting styles found in the Unsupportive Parenting group. In examining the unique dimensions of externalizing behavior, hyperactivity was highest among the children of the Unsupportive Parenting group, a finding consistent with prior research on adolescents (Chambers et al., 2001; Forehand & Nousiainen, 1993). Results also suggested that children’s level of aggression was lowest in the Supportive Parenting group, compared to both the Mixed-Support Parenting and the Unsupportive Parenting groups. Taken together, findings support the hypothesis that having two supportive parents is associated with the fewest behavior problems, followed by having one good-enough parent, and finally, that having two unsupportive parents is associated with the highest degree of behavior problems in children. It is unclear from the current analyses whether it matters which parent (mother or father) is the supportive parent, since the cluster analysis on the current sample did not yield a comparison group in which the father was more effective.

Turning to the findings related to child gender, results suggested that poor dyadic parenting was associated with hyperactivity and aggression for boys but not for girls. The lack of findings for girls could be related to the overall lower base rate of externalizing behavior of girls in this sample. Another possibility is that girls may manifest symptoms differently from boys via internalizing problems that may emerge later in development.

In an attempt to examine spousal similarity on assessments of parenting, parental agreement was explored using intraclass correlations. Level of agreement on all indices of parenting was unrelated to children’s outcomes. However, examining the interaction between couples’ ICC scores and cluster group membership provided a method for testing the hypothesis that consistency in supportive parenting would be related to fewer behavior problems while agreement on unsupportive parenting behaviors would be related to more behavior problems. As predicted, parents’ similarity in over-reactivity and warmth (but not laxness) did moderate the relationship between dyadic parenting styles and children’s externalizing symptoms. Thus, the results support the hypothesis that consistency in supportive parenting styles is related to fewer behavior problems in children, while consistency in unsupportive parenting (in this case overreactivity) is linked to negative outcomes for children, particularly in the context of an overall less supportive parenting environment.

In term of study limitations, the cross-sectional design makes it impossible to determine the directionality of effects. Results could be interpreted as the influence of parenting on children’s behavior, of children’s behavior on their parents, or the interaction of the two, or as the result of some unidentified third factor. It is most likely that a combination of bi-directional, interactive effects between parents and children play an important role in determining outcomes for children. Parenting processes may affect children differently at different development stages and cumulatively over time; accordingly, longitudinal studies could add greatly to our understanding of these processes. Another limitation concerns the generalizability of the findings. Cluster analysis is inherently sample-specific, and thus the clusters that emerged in the current sample are not necessarily representative of other populations. However, the advantages afforded by cluster analysis, especially the fact that groupings arise from patterns in the data, rather than forcing parents into groups based on median splits or some other artificial distinction, outweigh the disadvantages. Although limited in power by the small sample size, our findings suggest that there are key differences in the relationship between particular parenting styles and child outcomes which depend on child gender.

By design, this study focused on dual-earner, working-class couples, which may pull for greater father involvement since both parents are employed. While an important contribution was to highlight an understudied group, the within-group design limits the populations to which the findings can be extended. Another drawback was the lack of ethnic and racial diversity among the sample. Further research already underway seeks to address the lack of diversity by recruiting a sample of Black and Latino families. In addition, future work should examine coparenting that can occur in single mother families in conjunction with grandmother, sisters, boyfriend, or other kin, a virtually unexplored aspect of parenting that children experience

An important contribution of the current study was to extend research on the combined contributions of both mothers and fathers to children’s development. Our results indicate that there is a benefit to be gained from having two consistently “supportive” parents rather than one or none; however, of note was the fact that this finding only emerged for boys. Future work, examining longer term effects of parenting may find positive effects of parenting on girls’ depression or anxiety (i.e., internalizing behaviors).

The results of this study have important clinical implications. First, our findings similar to those of Martin et al. (2007) highlight the benefits of having two supportive parents, as compared to only one supportive parent. Given that low-income families are often the target of early intervention services and our findings emerged from such a sample, we would propose that parenting programs and education materials should highlight the benefits of focusing equally on mothers and fathers as important contributors to child outcomes. As Markham, Ganong, and Coleman (2007) found in their coparenting work with divorced couples, mothers especially need to be convinced of the importance of both parents for children. Further, this is likely to be the case in all relationships and these data point the additive effects of two supportive parents.

In addition, given that the prototypical child client at most outpatient clinics is a boy with externalizing behavior, the results suggest that clinicians working with boys with behavior problems should attempt to include both parents in treatment and have at least part of the focus be on improving parenting skills with respect to warmth and discipline. Future work needs to explore the long term effects of combined parenting on both sons and daughters and longitudinal designs are need to capture these important developmental effects.