An Overview of the Risks and Protectors for Children of Separation and Divorce

2. The Association Between Divorce and Children's Adjustment

Divorce involves considerable changes and stressors for children, such as a decreased standard of living, change in family type, decreased contact with one parent, and interparental conflict.  Divorce is one of the most prevalent life stressors experienced by children today (Hetherington, Bridges and Insabella, 1998; Lengua et al., 2000).  Therefore, during the past five decades, numerous retrospective and prospective studies have been dedicated to identifying the possible negative effects of divorce on the adjustment of children.  What follows is a brief overview of the findings of these studies.

2.1  Retrospective Studies

Various studies have retrospectively compared children from divorced and married families to examine the implications of parental divorce on child well-being (see Stewart, 2001, for a comprehensive review of this research).  Such studies, and qualitative reviews of the research literature, have produced inconsistent findings.  Some studies and reviews have found that children from divorced families, compared to children from intact families, experience lower levels of well-being across various domains (e.g. social and emotional functioning) (Hetherington, 1981; Krantz, 1988; Peterson and Zill, 1986).  Other studies and reviews have failed to find significant results and have suggested that most children recover from divorce with few enduring negative consequences (Edwards, 1987).  Emery (1988) concluded that although divorce is associated with a number of negative child outcomes, serious problems are not markedly over-represented among divorced families.

To clarify the confusion and contradictory nature of the research findings, Amato and Keith (1991a) conducted a meta-analysis to estimate the impact of parental divorce on the well-being of children and adolescents.  Including 92 studies conducted prior to 1991 that collectively involved more than 13,000 children, the meta-analysis examined eight of the most frequently studied and reviewed outcomes of child well-being in relation to divorce.  They included academic achievement (standardized achievement tests, grades, teachers' ratings, and intelligence), conduct (misbehaviour, aggression and delinquency), psychological adjustment (depression, anxiety and happiness), self-concept (self-esteem, perceived competence and internal locus of control), social adjustment (popularity, loneliness and cooperativeness), mother-child relations (affection, help and quality of interaction), father-child relations (affection, help and quality of interaction) and other.

Results of the meta-analysis revealed two important trends.  First, compared to children from intact homes, children from divorced families did, indeed, experience lower levels of well-being.  In fact, more than two thirds of studies included in the meta-analysis found lower levels of well-being in children of divorced parents.  Second, while the differences between children of divorced and intact families were statistically significant, the size of the differences (effect size) was small.  More specifically, across all eight outcome measures, children from divorced families scored between one fifth and one eighth of a standard deviation below children from intact families.  The strongest mean effect sizes (effect sizes are reported in terms of standard deviations and refer to the magnitude of an association between two variables) were for father-child relations (-.26) and conduct (-.23), and the weakest but still statistically significant effect sizes were for psychological adjustment (-.08) and self-concept (-.09) (Amato and Keith, 1991a).  This finding indicates that the stress of divorce may manifest itself most strongly in children's relationships with their fathers and their externalizing behaviour and may have the least amount of impact on children's psychological adjustment and self-concept.  It is also of interest to note that these adjustment differences exist not only in the short-term, but may also persist into adulthood to influence subsequent generations of children (Kiernan and Hobcraft, 1997; Kiernan and Mueller, 1998; O'Connor et al., 1999; Rodgers and Prior, 1998).  In a second meta-analysis that compared the adjustment of adults who grew up in divorced or intact families, larger but still moderate effect sizes were found (Amato and Keith, 1991b).

Why do different studies produce discrepant results with respect to the effects of divorce on child well-being?  In an attempt to answer this question, Amato and Keith (1991a) employed meta-analytic techniques to elucidate study characteristics that account for variations in effect sizes.  First, they found that methodologically strong studies (i.e. those that utilized general population samples, control variables or large sample sizes) reported smaller differences between children from divorced and non-divorced families and that methodologically unsophisticated studies may overestimate the effects of divorce on children.  Second, the meta-analysis revealed that effect sizes based on the reports of parents were weaker than effect sizes based on other sources.  This indicates that divorced parents (usually mothers) may underestimate their children's problems and that researchers and clinicians should not rely solely on parental reports of child functioning.

Although the research literature contains numerous suggestions that divorce has more negative consequences for boys than for girls (Hetherington, Cox and Cox, 1982; Wallerstein and Kelly, 1980), the meta-analysis indicated that, in general, boys and girls do not differ in their response to parental divorce.  There was one exception to this generality, however, with boys from divorced families exhibiting more difficulty adjusting socially than girls.  A fourth factor found to be significantly associated with effect sizes for psychological adjustment, social adjustment, mother-child relations and father-child relations was the age of the child.  Effect sizes were largest for children in primary school and high school, and smallest for college-aged children (Amato and Keith, 1991a).  Finally, on the basis of longitudinal studies (Hetherington et al., 1982), it was hypothesized that effect sizes would be strongest for studies taking place shortly after the divorce.  The results of the meta-analysis indicated that this hypothesis was true only in relation to child conduct problems.  Compared to children whose parents had been divorced for more than two years, children whose parents had been divorced for two years or less had more conduct problems.  This finding indicates that conduct problems become less pronounced over time (Amato and Keith,1991a).

Many studies conducted from 1990 reported that children from divorced families scored lower than children from intact families on various outcomes, including academic success, conduct, psychological adjustment, self-concept and social competence following parental separation (Amato, 2000; Juby and Farrington, 2001).  In addition, the small effect sizes reported in Amato and Keith's (1991a) meta-analysis (mean effect size of .18) appear to be consistent with those established by studies conducted in the 1990s (mean effect size of .19) (Amato, 2000).

Most outcomes for children are measured on continuous scales.  The results for outcomes measured dichotomously show a similar picture.  For instance, studies of nationally representative samples in the U.S. indicate that children from divorced families are approximately twice as likely to receive psychological help than are children from married families (Zill, Morrison and Coiro, 1993).  McLanahan and Sandefur (1994) analyzed data from five different national surveys of children and families.  Such analyses revealed that the risk of dropping out of high school and of teen pregnancy was about twice as great for children from divorced families than for children from married families.

While such data indicate that divorce is a risk factor for various psychological problems among children, the same data highlight that many children "bounce back" from the stress of parental divorce.  Therefore, although Zill et al. (1993) found that twice as many 12- to 16-year-old American children from divorced families (21 percent) as intact families (11 percent) received psychological help, 79 percent of those children coped with their parents' divorce without receiving psychological help.  Similarly, most children who experience parental separation do not drop out of high school or get pregnant as teenagers.

2.2  Prospective Studies

Longitudinal, prospective studies provide another perspective on the effects of divorce on child well-being.  While retrospective studies find correlations between parental divorce and child maladjustment, longitudinal studies remind us that correlation is not the same as causation.  Thus, the maladjustment of children of divorced families may not be solely due to parental separation.  Rather, in light of the fact that divorce is the final outcome of a process marked by dissatisfaction, unhappiness and conflict, child maladjustment may also be influenced by factors present prior to the divorce.  In fact, the longitudinal studies of Cherlin et al. (1991) found that although children from divorced families exhibited more behaviour problems than children from married families, these problems, at least for boys, were present before the divorce.  Amato and Booth (1996) also found that problems in parent-child relationships (including parental reports that their children had given them more than the average amount of problems) were present 8 to 12 years before parental separation.  Comparable results were obtained with regards to children's internalizing and externalizing behaviour, social competence, self-esteem and adolescent substance abuse (Aseltine, 1996; Doherty and Needle, 1991; Hetherington, 1999).  In light of such findings, Cherlin et al. (1991) speculated that the observed differences between children from divorced and intact families can be traced to three sources: