The Effects of Divorce on Children : A Selected Literature Review

2. LIMITATIONS OF THE RESEARCH

In order to examine the effects of marital disruption on children, three different research techniques have commonly been employed: clinical assessments; comparisons of children from divorced and intact families; and, in-depth interviews with divorced families (Amato, 1987). Clinical assessments generally involve examining children of divorce who have been referred to various counselling or clinical programs. For instance, Wallerstein and Kelly (1975) examined the effect of parental divorce by interviewing parents and children referred to divorce counselling. Although clinical assessments provide a great deal of information concerning children from maritally disrupted families, they focus on extreme cases and, therefore, the results cannot be generalized to the majority of children who experience marital disruption. In addition they present an almost invariably negative picture of children’s post-divorce adjustment and it is these studies which predominated in the early years of research on the effects of divorce. Comparative studies usually compare non-clinical samples of children from families experiencing marital disruption with children from intact families. These studies usually examine objective, quantifiable outcomes, such as academic achievement, emotional adjustment and self-esteem, through the use of tests or questionnaires. However, many of these measures do not allow an understanding of how separation and divorce are subjectively experienced and interpreted by parents and children. The third technique involves conducting in-depth interviews with parents and/or children from divorced families in order to elicit the experiences from their own perspective. Problems associated with this technique include potential bias or distortion of facts by those interviewed.

In addition to various research techniques, both cross-sectional and longitudinal research has also been conducted. Cross-sectional, the most commonly used approach, involves examining individuals at one point in time - for instance examining children of divorce shortly after the divorce in order to see whether they differ from intact families. Longitudinal studies, on the other hand, track a sample of individuals from a particular point in time (usually following marital disruption), with follow-up interviews at various times following the divorce. Cross- sectional studies rarely collect retrospective data, and little information is therefore available on the socio-economic history of the family, level of family conflict, parent-child relations, etc. prior to the divorce (Demo & Acock, 1988). As a result, no examination of causal directions or developmental effects is possible. For example, cross-sectional designs are not able to determine whether some characteristics of children seen as a consequence of divorce are present prior to marital dissolution (e.g., behavioural problems). Although longitudinal studies are better able to track the causes and effects of marital disruption on children and may include retrospective data, they are quite costly and time intensive and consequently are more rarely conducted.

While the literature examining the effects of divorce on children is extensive, many of the findings are inconclusive or inconsistent. One possible reason for these disparities is that different procedures have been used among studies. For example, as mentioned previously, Wallerstein and Kelly (1975) based their results on a clinical sample of children referred for divorce counselling to a local Community Mental Health Centre. Since these children may not be representative of all children experiencing divorce, the findings with respect to the problems experienced by children of divorce may not be generalizable to the broader population of children of divorce. Healy, Malley and Stewart (1990) also suggest that observed gender differences in adjustment to divorce may reflect the overuse of clinical samples rather than genuine gender differences. They argue that the undercontrolled behaviour of boys is more readily observed and, therefore, more likely to lead to clinical referral.

In addition to procedural variations, the definition of "family structure" may lead to differing results. Many studies examine single-parent households, which may be a due to divorce, death, a parent who has never married, etc. Since it is fairly well established that children of divorce differ from children from other single-parent households (Demo & Acock, 1988; Felner, 1977; Felner, Farber, Ginter, Boike & Cowen, 1980; Mechanic & Hansell, 1989), it is necessary to avoid grouping divorced, widowed, and never-married parents. Further, it is important to distinguish between single-parent families and those where the parent has remarried (Barber & Eccles, 1992).

Studies also vary in the extent to which they control for potentially intervening factors, such as socio-economic status of families, race/ethnicity, gender and age of children, thereby making comparisons difficult. Most studies utilize caucasian, middle-class children, from urban areas, making it difficult to generalize to other groups. Krantz (1988) suggests that caution should be used in interpreting studies which do not control for factors other than marital status. She also argues that studies which group children from divorced and intact families by socioeconomic status are problematic because divorced families tend to cluster at the lower end of these groups. Furthermore, studies that match divorced and non-divorced persons are rare, as are studies which use statistical controls of extraneous factors.

Another limitation of many studies concerns the validity of the measurements used. For instance, information provided by adults (e.g., teachers) about children may reflect stereotypes about what children of divorce should be like, rather than the actual behaviour of the child. Parental reports may be biased due to the personal involvement of the parent with the child. For example, a parent opposed to the divorce may only be aware of problem behaviours associated with the child. In an examination of the adjustment of children to divorce, Kurdek (1987) found that children, mothers and teachers do not provide similar information with regard to children’s divorce adjustment. Clinically observed behaviours in subjects can also be problematic because they are highly subjective and can be difficult to replicate. Even supposedly objective reports (such as police records) may be biased because police may be more likely to charge a child from a single-parent than intact home. Finally, assessment instruments that tap some objectively defined behaviour are often biased by the prevailing cultural norms and values. These values and norms change over time and at any point in time may be disputed.

In addition to these limitations, there is a need for sensitivity to cohort effects. It has been found that the results from some of the early studies differ from more recent studies (Amato & Keith, 1991a&b). It is possible that the older studies were conducted during a period when divorce and single-parenthood was seen as anomalous or socially unacceptable. Therefore, Barber and Eccles (1992) suggest that the results of these older studies may be time-bound experiences which are no longer prevalent today. Amato and Keith in two meta-analyses, one of children (1991b) and one of adults (1991a), conclude that the more sophisticated and recent the study, the more tenuous the connection between parental divorce and well-being of the child. This indicates that if the various interacting effects are taken into account, many of the effects vanish.

Due to these limitations, it is difficult to draw firm conclusions based on the literature. However, in recent years, an attempt has been made to control for many of these limitations. For example, most recent studies attempt to control for intervening variables such as age and gender of child, socio-economic status and conflict. Krantz (1988) argues that although there are serious limitations with some of the studies, the information need not be rejected if conclusions are made cautiously and with full recognition of their limitations. Further, she argues that biases in the available information are unlikely to distort the conclusions if the data are repeatedly replicated and biased in different directions.

Date modified: