Professional Documents
Culture Documents
Parental Divorce
Parental Divorce
PAUL R. AMATO, University of Nebraska-Lincoln LAURA SPENCER LOOMIS, WESTAT ALAN BOOTH, Pennsylvania State University
Abstract
Cross-sectional studies show that adults who grew up in conflict-ridden two-parent families or who experienced parental divorce report lower levels of psychological and marital well-being than do other adults. However, previous research has been unable to determine how parental marital conflict, divorce, and children's long-term outcomes are related. Using a 12-year longitudinal study, we find that the consequences of parental divorce depend on parental marital conflict prior to divorce. In high-conflict families, children have higher levels of well-being as young adults if their parents divorced than if they stayed together. But in low-conflict families, children have higher levels of wellbeing if their parents stayed together than if they divorced. In marriagesthat do not end in divorce, parental marital conflict is negatively associated with the well-being of
offspring.
Each year in the U.S., approximately one million children experience parental divorce (U.S. Bureau of the Census 1992). Social scientists have attempted to determine the consequences of marital dissolution for offspring. Studies of children are contradictory, and many have serious methodological limitations. Nevertheless, the majority show that children of divorce are more likely to exhibit psychological, behavioral, social, and academic problems than children raised in continuously intact two-parent families (Amato & Keith 1991a; Furstenberg & Cherlin 1991; Emery 1988). Furthermore, studies of adults suggest that the gap between children from divorced and continuously intact two-parent families persists well into adulthood (Amato & Keith 1991b).
* This research was supported in part by Grant #5 R01 AGO4146from the National Institute on Aging and the Pennsylvania State University Population Research Institute with core support from the National Institute of Child Health and Human Development Grant #1-HD28263.Direct correspondence to Paul R. Amato, Department of Sociology, University of Nebraska-Lincoln, Lincoln, NE 68588-0324.
896 / Social Forces 73:3, March 1995 In addition to divorce, chronic conflict between parents who remain married increases the risk of a variety of problems for children (Emery 1982; Grych & Fincham 1990). Prolonged exposure to interparental conflict during childhood may also create a predisposition toward psychological and marital difficulties in later life (Adam, Bouckoms & Streiner 1982; Amato & Booth 1991; Booth & Edwards 1990; Overall, Henry & Woodward 1974). In general, studies of children and adults suggest that experiencing parental divorce and growing up in a high conflict two-parent family are associated with long-term decrements in well-being. However, previous research has been unable to specify how parental divorce and parental marital conflict are related to offspring well-being. We consider four models in this article. The first is that divorce and parental marital conflict have independent effects on children. A second possibility is that the apparent "effects" of divorce are due to the conflict that precedes marital dissolution. Because marital conflict is a cause of both divorce and behavioral problems of children, the association between divorce and child behavioral problems may be spurious. A third possibility is that the effects of marital conflict are mediated by divorce; that is, marital conflict leads to divorce, and divorce, in turn, lowers children's well-being. The fourth possibility is that parental marital conflict and divorce interact, with the consequences of divorce depending on the level of conflict that precedes marital dissolution. In relation to the fourth model, when conflict between parents is overt and intense, the consequences of divorce may be positive. Many observers have suggested that children are better off in well-functioning single-parent families than in two-parent families marked by severe and persistent conflict (e.g., Barber & Eccles 1992; Emery 1988; Furstenberg & Cherlin 1991; Kurdek 1981). Furthermore, reasons given for divorce frequently include mental and physical cruelty and alcohol abuse (Kitson & Sussman 1982). Under these conditions, children may recognize divorce as an escape from an aversive environment. But when parents display little overt conflict prior to marital dissolution, children may not anticipate or welcome divorce. Under these circumstances, children may gain little but risk losing a great deal, including the absence of one parent from the household, a decline in standard of living, and a less stable and less predictable home environment. In these cases, the consequences of divorce for children are likely to be negative. This issue can be clarified by including both divorce and parental marital conflict (measured prior to divorce, in cases where divorce occurs) in statistical models to predict children's well-being. However, few data sets contain information on parental marital conflict measured prior to divorce. We use a 12-year longitudinal study of married people first interviewed in 1980, some of whom divorced over the course of the study. In 1992, we also interviewed adult offspring who had lived in the same household as the respondent (parent) at the time of the initial parental interview in 1980. Because we have detailed information on parental marital quality prior to divorce, as well as data on a range of outcomes for offspring 12 years later, we are able to estimate the effects of both parental divorce and predivorce marital conflict on offspring during the transition to adulthood.
A large number of studies have compared children in divorced and continuously intact two-parent families. A meta-analysis of 92 of these studies shows that children of divorce, compared with those living in continuously intact twoparent families, score lower on measures of academic achievement, conduct, psychological adjustment, social relations, self-concept, and the quality of mother-child and father-child relationships (Amato & Keith 1991a). However, the mean effect sizes in this literature (which reflect the differences between divorced and nondivorced groups in standard deviation units) are modest, ranging from -.08 for self-concept to -.26 for the quality of father-child relations. Similarly, a meta-analysis of 37 studies indicates that adults who experienced parental divorce as children, compared with adults raised in continuously intact two-parent homes, have lower psychological well-being, lower socioeconomic attainment, poorer quality marital relationships, and an increased propensity to divorce (Amato & Keith 1991b). As in similar studies of children, most of the effect sizes are modest, ranging from -.06 (the quality of social relations) to -.32 (psychological adjustment). Nevertheless, the fact that any association exists, given the length of time involved, is noteworthy. A number of factors can account for the small but well-replicated gap in well-being between offspring from divorced and nondivorced families. As noted above, the apparent "effects" of divorce on children may be spurious - both being caused by conflict between parents. On the other hand, divorce often generates problems for children that go beyond interparental conflict. For example, divorce frequently results in diminished contact between children and the noncustodial parent - usually the father (Furstenberg & Nord 1985; Seltzer 1991). A decrease in contact with the noncustodial parent may represent a net loss in the amount of emotional support, practical help, and supervision that children receive from adults. In addition, divorce typically results in a decline in the standard of living of noncustodial mothers and their children (Duncan & Hoffman 1985). The stress of divorce also appears to have a negative impact on the child-rearing skills of custodial parents; compared with other parents, recently divorced custodial parents are less affectionate, less communicative, more punishing, and more inconsistent in their use of discipline (Hetherington, Cox & Cox 1982). Finally, divorce often sets in motion additional events that may be stressful for children, such as moving, changing schools, and parental remarriage (Wolchik et al. 1985).
PARENTAL MARITAL CONFLICT
A variety of studies show significant negative associations between the level of conflict between married parents and children's functioning and well-being (see Emery 1982, and Grych & Fincham 1990, for reviews.) A related group of studies show that among children of divorce, well-being is inversely correlated with the level of postdivorce conflict that persists between parents (Guidubaldi et al. 1986; Johnston, Kline & Tschann 1989). The effects of family conflict may
898 / Social Forces 73:3, March 1995 be long-lasting. Clinical research shows that adults suffering from depression and other psychological disorders report more childhood family conflict than do matched controls (eg., Adam, Bouckoms & Streiner 1982; Overall, Henry & Woodward 1974). Similarly, Booth and Edwards (1990) and Amato and Booth (1991), using a representative national sample, found that people who recalled their parents' marriage as being unhappy reported more psychological and marital problems than did other adults. Several mechanisms exist through which marital conflict might have shortand long-term negative consequences for children. First, overt conflict between parents is a direct stressor. Observational studies show that children react to conflict between adults with fear, anger, aggression, or the inhibition of normal behavior (Cummings, Zahn-Waxler & Radke-Yarrow 1981; Cummings 1987). Physical violence between parents, in particular, is associated with psychological and behavioral problems among children, partly because violence between parents increases the risk of parent-to-child violence (Christopoulos 1987; Hughes 1988). Stress resulting from conflict is likely to interfere with children's ability to concentrate on school work, which may decrease academic performance. Because young children are egocentric, they may attribute blame for marital conflict to themselves (Wallerstein & Kelly 1980), resulting in feelings of guilt and low self-esteem. Children are often drawn into conflict between parents, resulting in a deterioration in parent-child relationships and a disintegration of general family cohesiveness (Amato 1986; Johnston, Kline & Tschann 1989). Furthermore, through modeling verbal or physical aggression, parents indirectly teach their children that disagreements are resolved through conflict rather than discussion. For this reason, children from high-conflict families may not learn the social skills (such as negotiation and compromise) necessary to ensure rewarding relationships in childhood and adulthood. Finally, youth from conflicted families may leave home early to escape from an aversive home environment (Aquilino 1991). This pattern may involve curtailing educational plans; it may also involve marriage at an early age to an inappropriate partner, resulting in poor marital quality and an elevated risk of divorce.
THE RELATION BETWEEN DIVORCE AND MARITAL CONFLICT
As noted above, we consider four models linking parental marital conflict, divorce, and offspring well-being: (1) conflict and divorce have independent effects on children, (2) the relation between divorce and children's well-being is due to their mutual association with conflict, (3) conflict affects children because it increases the likelihood of divorce, and (4) conflict and divorce interact, with the effects of divorce being positive when conflict is high and negative when conflict is low. A number of cross-sectional studies have included data on children in highdiscord two-parent families, low-discord two-parent families, and divorced families. The majority of these studies show that children in both divorced and high-discord two-parent families have more problems than children in lowdiscord two-parent families (e.g., Long et al. 1987; Peterson & Zill 1986). These results can be interpreted in two ways: They are consistent with the notion that
Parental Divorce and Marital Conflict / 899 both divorce and conflict have independent effects on children (model 1) and with the notion that conflict is the primary cause of children's problems (model 2). Other cross-sectional studies show that parental marital conflict, but not divorce, is negatively associated with children's well-being (e.g., Enos & Handal 1986; Long 1986). These studies suggest that conflict, rather than divorce, is the key variable (model 2). However, many cross-sectional studies reveal a particular rank order of groups, with children in high-discord two-parent families having the most problems, children in low-discord two-parent families having the fewest problems, and children of divorce falling in between (Amato & Keith 1991a). As such, these studies are consistent with the fourth model. If divorce improves the well-being of children from high-discord homes, but lowers the well-being of children from low-discord homes, then the overall mean for the divorced group is likely to fall between the two. Overall, it is difficult to distinguish between models 1, 2, and 4 on the basis of these studies. Although prospective longitudinal studies of divorce are rare, they tend to show that problems for children of divorce are apparent before marital dissolution. Block, Block, and Gjerde (1986) analyzed data from a sample of 101 families - 40 of which experienced divorce - over an 11-year period. They found that boys (but not girls) whose parents divorced exhibited an elevated number of behavioral problems several years prior to parental separation. This result is consistent with the notion that parental marital conflict is responsible for the problems often seen among children of divorce (model 2), although it is also possible that other unmeasured characteristics of the family prior to divorce (such as impaired parenting) are also relevant. Similarly, in a five-year study of adolescent adjustment and drug use, Doherty and Needle (1991) compared 48 adolescents whose parents divorced and 578 whose parents did not. They found that girls demonstrated negative reactions prior to parental separation and did not become worse afterward. Boys, on the other hand, showed problems following but not prior to parental separation. In this study, the results for girls suggest that conflict, rather than divorce, is the cause of child problems (model 2), whereas the results for boys suggest that divorce, rather than conflict, is the cause (model 3). Cherlin et al. (1991) present the only analysis in which parental marital conflict was measured prior to divorce. Using a British longitudinal data set involving over 11,000 families, they found that children who experienced parental divorce between the ages of 7 and 11 had more behavioral problems and lower academic test scores than children in intact families did. However, these differences were apparent prior to divorce, and controlling for predivorce scores substantially reduced many of the postdivorce differences, especially for boys. Adding other predivorce variables to the model, including a short measure of family conflict, reduced the coefficients to nonsignificant levels for boys, but not for girls. Cherlin et al. also carried out a similar analysis based on a five-year longitudinal subsample from the National Survey of Children and found comparable results. Overall, the results of this study for boys support the second model, whereas the results for girls are mainly consistent with the first model.
900 / Social Forces 73:3, March 1995 The weight of the evidence appears to support the notion that parental marital conflict accounts for much of the apparent ill effects of divorce (model 2). However, a number of studies are also consistent with the other perspectives outlined above. Furthermore, almost all this evidence is indirect; only one study (Cherlin et al. 1991) measured conflict prior to divorce, and that study did not search for possible interactions between conflict and divorce. No study to our knowledge has assessed all four models simultaneously. The Present Study We are able to both complement and move beyond previous longitudinal studies in this area. The Cherlin et al. (1991) study - although the best of its kind - was limited in its ability to measure characteristics of the predivorce marital relationship. For example, the British data included only a two-item measure of family conflict as rated by a "health visitor." Furthermore, in their analysis of British data, the authors included family conflict along with four other scales, such as use of children's services and mental health services; consequently, they could not assess the independent role of conflict. The present study uses a measure of marital conflict based on parents' reports at three points in time over an eight-year period. In cases of parental divorce, we are able to use the conflict score closest in time prior to the dissolution of the marriage. Previous longitudinal studies have focused on children (Block, Block & Gjerde 1986; Cherlin et al. 1991) or adolescents (Doherty & Needle 1991). We complement these studies by examining a different age group: young adults, most of whom are in their 20s. Young adulthood is a critical time, as it typically involves leaving the parental home, completing education, beginning full-time employment, marriage, and the first years of parenthood. The extent to which people experience success or failure in these early tasks sets the stage for achievement and well-being later in life. In terms of outcomes, we include measures of psychological well-being (psychological distress, overall life happiness), the quality of intimate relations (either marital or cohabiting), and ties with kin and friends. We investigate psychological well-being and relationship quality because these variables have been shown in previous studies of adults to be related to parental divorce (Amato & Keith 1991b). Ties with kin are relevant because a number of studies have shown that parental divorce decreases contact with kin, especially on the father's side (Rossi & Rossi 1990). We include ties with friends because they reflect the formation and maintenance of close attachments other than marital ones. If parental marital conflict and divorce interfere with the learning of skills for conflict management, we would expect evidence of weaker ties with friends in adulthood. Unlike previous studies, ours tests hypotheses representing each of the four models stated formally below: Hypothesis 1. When parental divorce and parental marital conflict (measured prior to divorce) are entered simultaneously in regression equations, both have negative partial associations with offspring well-being.
Parental Divorce and Marital Conflict / 901 Hypothesis 2. Parental divorce is negatively associated with offspring well-being, but the magnitude of the association decreases and is no longer statistically significant when parental marital conflict (measured prior to divorce) is controlled. Hypothesis 3. Parental marital conflict is negatively associated with offspring well-being, but the magnitude of the association decreases and is no longer statistically significant when parental divorce is controlled. Hypothesis 4. Parental divorce and parental marital conflict interact so that parental divorce is positively associated with offspring well-being when predivorce conflict is high and negatively associated with offspring well-being when predivorce conflict is low. Finally, we consider how the age of the child at the time of divorce is related to later outcomes. We are unable to examine the implications of parental marital dissolution for children who are quite young because our sample does not contain children younger than 9 at the time of divorce. Nevertheless, within the age range of our sample, we examine correlations between child's age at time of divorce and later well-being. Research Procedures
THE SAMPLE
A 12-year longitudinal study of marital instability over the life course (Booth et al. 1991) was used to test the hypotheses we have advanced. A national sample of 2,033 married persons (not couples) 55 years of age and under were interviewed by telephone in 1980. The sample was re-interviewed in 1983,1988, and 1992. Sample households were chosen through a clustered random-digitdialing procedure, and the husband or wife was selected for an interview using a second random procedure. A random sample of offspring were also interviewed as part of the fourth wave of interviews. Of the 58% of the original sample of persons who were successfully re-interviewed in 1992 (N - 1,183), 625 had offspring who were 19 years of age or older in 1992 and who had lived in the parental household in 1980. The respondent was asked to provide the name and telephone number of the child. In cases where there was more than one eligible child, a random procedure was used to select one. About 86% (N - 537) of the parents gave us names and telephone numbers of children. We were able to obtain interviews with 471 of them for an overall completion rate of 75%. Of the 66 we were unable to interview, 18 could not be contacted, 22 refused, and 26 agreed to a mail version but did not return it. The ages of offspring who were successfully interviewed ranged from 19 to 40 with a median of 23.5 years. The original sample of married persons was compared with national data and was found to be representative with respect to age, race, household size, presence of children, tenure, and region, although residents of large metropolitan areas were slightly underrepresented. Subsequent waves were slightly less representative with respect to African Americans and Hispanics, younger respondents, renters, and those without college education, but no more so than investigations using personal interviews (Booth & Johnson 1985).
902 / Social Forces 73:3, March 1995 Probit was also used to see whether 1988 parental demographic and marital quality variables affected the probability of a noninterview with offspring. Those factors unrelated to obtaining an interview are parent's age, parent's gender, parent's education, the presence of children in the parental household, parental divorce between 1980 and 1988, and parental marital conflict. The only significant variable was home ownership, with renters (parents) being less likely to yield an interview with offspring than homeowners. Overall, it appears that families that yielded offspring interviews are highly representative of the larger sample of parents at the time of the fourth interview. The 1980 parental interview provided information on the quality of the parents' marriages. Subsequent interviews provided information on whether parents had divorced and, if still married, their marital quality. The 1992 offspring interview provided data on children's psychological well-being, marital happiness, and social support.
INDEPENDENT VARIABLES
Offspring were placed in the parental divorce group if (1) they were living with both biological or adoptive parents in 1980, and if (2) they experienced a parental divorce (as reported by parents) after 1980. (For this purpose, we considered cases of permanent separation to be equivalent to divorce.) Fortytwo offspring met these criteria and were coded 1; those who were not exposed to a parental divorce were coded 0. Offspring whose parents had divorced prior to 1980 were omitted from the analysis because they were living with a stepparent at the time of the 1980 interview. Omitting these individuals brought the total N down to 419. Of the 42 offspring in the divorced group, 8 experienced divorce between the ages of 9 and 12, 24 as teenagers, and 10 in their early 20s (however, most of these were still living with parents at the time of divorce). Because of the wide age range, we carried out all analyses twice: once with the oldest group included and once with this group excluded. Our measure of marital conflict was based on parents' reports of the severity and amount of verbal and physical conflict. The four items referred to the frequency of disagreements, the conflict over the household division of labor, the number of serious quarrels in the past two months, and whether spouses have ever slapped, hit, pushed, kicked, or thrown things at one another in anger. Scores on this scale ranged from 0 to 12, with higher scores indicating greater disagreement (ax - .54). In 1980, all parental marriages were intact. For children whose parents subsequently divorced, the parental marital conflict variable was based on the conflict score closest in time prior to marital dissolution (either 1980, 1983, or 1988). For children whose parents did not divorce, marital conflict was based on the mean conflict score across the three time periods. (We also carried out analyses in which parental marital conflict was based entirely on the 1980 scores. These results of these analyses were essentially the same as those reported below.) Although the reliability coefficient for our measure of conflict at a single point in time is modest, aggregating observations over time improves the reliability of measurement.
Three additional measures of parental marital quality were available in the data set: an 11-item measure of marital happiness (a - .87), a 5-item measure of marital interaction (a - .63), and a 12-item measure of divorce proneness (a - .91). Our results were very similar regardless of which measure of parental marital quality we used. Consequently, to maintain continuity with previous research in this area, we report only the results based on the measure of marital conflict.
DEPENDT VARIABLES
Information on all dependent variables was obtained from the offspring in the 1992 interview. Initially, we thought of examining aspects of socioeconomic achievement but decided against it because the majority of offspring had either not finished their education or not worked long enough to establish a stable occupational status. Psychological well-being was assessed by two variables. The first, psychological distress, was based on eight items from the Langner (1962) scale of symptoms. Higher scores indicated greater distress, and the alpha coefficient for this scale was .67. The second indicator was a widely used measure of overall happiness: "Taking all things together, how would you say you are these days? Would you say you are very happy, pretty happy, or not too happy?" For this item, higher scores reflected greater happiness. For offspring who were married, relationships were assessed with an 11-item measure of respondent's happiness with the relationship. This measure tapped both global feeling about the marriage (overall happiness, happiness compared to three years ago, strength of love) and the person's feelings about specific aspects of the relationship (amount of understanding received, amount of affection, extent to which spouses agree about things, the sexual relationship, the spouse as someone to do things with, and the spouse's faithfulness). For offspring who were cohabiting, the items were modified slightly by replacing "wife" or "husband" with "partner." Cohabiting relationships were combined with marital relationships in the analysis because there were too few to consider separately. Also, it is apparent that cohabitation is increasingly a substitute for marriage (Bumpass & Sweet 1989). Scores on this 11-item scale ranged from 11 to 33, with higher scores reflecting greater happiness. This scale had a reliability cqefficient of .84. Relations with relatives was measured by the item: "Now I want to ask you about your relationships with relatives who don't live with you, such as brothers, sisters, grandparents, in-laws, aunts, uncles, and cousins. Are there relatives you have that you feel emotionally close to? (If yes) How many would that be?" Ties with friends was assessed by the question: "Now I want to ask about your friends. Are there people whom you consider very close friends who are not relatives? (Ifyes) How many very close friends would that be?" The two variables were moderately positively correlated (r =.21, p <.01) and were added to produce a summary measure of social resources. Because the distribution was skewed, a logarithmic transformation was used to normalize the data.
Because they precede both parental divorce and offspring outcomes in time and could influence both, we controlled for the interviewed parent's sex, age, race (white or black), and education, as well as the education and age of the spouse. We controlled for the offspring's sex and age for the same reasons. Results We used multiple regression (ordinary least squares) to assess the associations between offspring outcomes and parental divorce and marital quality. We regressed offspring outcomes on parental divorce (equation 1), parental marital conflict (equation 2), and both parental divorce and parental marital conflict (equation 3). The control variables were included in each model. Significant associations between offspring well-being and both parental marital conflict and divorce in equation 3 would support the first hypothesis. Alternatively, a substantial decrease in the magnitude of the association between offspring wellbeing and parental divorce between equations one and three would support the second hypothesis. And a substantial decrease in the magnitude of the association between offspring well-being and parental marital conflict between equations 2 and 3 would support the third hypothesis. To examine interactions between parental divorce and parental marital conflict, a multiplicative term was created by multiplying divorce by conflict. The control variables and the components of the interaction term were included in equation 4. A significant interaction would indicate that the consequences of divorce depend on the level of marital conflict prior to marital dissolution - a result that would support hypothesis 4. These results are shown in Tables 1, 2, 3, and 4. Neither divorce nor parental marital conflict showed significant associations with any offspring outcome in equations 1, 2, or 3. These results provide little support for the first three hypotheses. However, parental marital conflict and parental divorce yielded statistically significant interactions across all four offspring outcomes (equation 4). Although the full equation (equation 4) in Table 3 involving happiness in the relationship was not statistically significant, we should take the interaction seriously. The full equation did not attain significance partly because the sample size was reduced for this outcome and partly because most of the control variables were not associated with relationship happiness. Of course, the control variables were included, not to increase the variance accounted for, but because they precede the independent and dependent variables and could therefore be a cause of both. When we ran the analysis again with the control variables omitted, the full equation involving divorce, conflict, and happiness in the relationship became statistically significant (p <.05). Furthermore, the nature of the interaction is consistent with the interactions observed for the other three outcomes. The nature of the relationship between parental divorce, parental marital conflict, and offspring well-being is the same in all cases. This is shown in Figure 1 where the divorce and marital conflict relationship is plotted for each
Parental Divorce and Marital Conflict / 905 TABLE 1: Unstandardized Coefficients for the Regression of Offspring's Psychological Distress on Parent's Divorce, Marital Conflict, Divorce and Marital Conflict Interaction, and Controls Model I Parent Divorce Parent's sex (1 - female) Parent's race (1 - black) Parent's education (respondent) Parent's education (spouse) Parent's age (respondent) Parent's age (spouse) Offspring's age Offspring's sex (1 - female) Parent divorce (1 - divorced) Parent marital conflict Parent divorce x marital conflict Intercept R2 (N- 416) * p <.05 p <.01 **p < .001 16.850** 0.048* 16.356** 0.050* 16.408** 0.053* Model 2 Parent Marital Conflict -0.072 -0.103 -0.003 -0.182*** -0.009 0.001 -0.028 0.420 Model 3 Parent Divorce + Marital Conflict -0.065 -0.047 0.004 -0.180** -0.011 -0.000# -0.026 0.398 -0.441 0.081 0.095 Model 4 Divorce x Conflict 0.028 -0.124 -0.000# -0.187** 0.001 -0.008 -0.026 0.444 1.657 0.170* -0.421* 16.020** 0.065**
906 / Social Forces 73:3, March 1995 TABLE 2: Unstandardized Coefficients for the Regression of Offspring's Overall Happiness on Parent's Divorce, Marital Conflict, Divorce and Marital Conflict Interaction, and Controls Model 1 Parent Divorce Parent's sex (1 - female) Parent's race (1 - black) Parent's education (respondent) Parent's education (spouse) Parent's age (respondent) Parent's age (spouse) Offspring's age Offspring's sex (1 - female) Parent divorce (1 - divorced) Parent marital conflict Parent divorce x marital conflict Intercept R2
(N - 419)
*
Model 2 Parent Marital Conflict -0.137* -0.183 -0.024 0.051** -0.001 0.005 0.007 0.116*
Model 3 Parent Divorce + Marital Conflict -0.137* -0.181 -0.024 0.051** -0.001 0.005 0.008 0.115* -0.017
Model 4 Divorce x Conflict -0.160* -0.163 -0.022 0.053*** -0.005 0.007 0.008 0.104* -0.522* -0.032* 0.101"*
-0.014
-0.014
1.601** 0.083***
1.667** 0.085***
1.669** 0.085***
1.762** 0.102**
Parental Divorce and Marital Conflict / 907 TABLE 3: Unstandardized Coefficients for the Regression of Offspring's Marital Happiness on Parent's Divorce, Marital Conflict, Divorce and Marital Conflict Interaction, and Controls Model 1 Parent Divorce Parent's sex (1 - female) Parent's race (1 - black) Parent's education (respondent) Parent's education (spouse) Parent's age (respondent) Parent's age (spouse) Offspring's age Offspring's sex (1 - female) Parent divorce (1 - divorced) Parent marital conflict Parent divorce x marital conflict Intercept R2 (N - 161) * p < .0 5 **p < .01 28.409*** 0.063 28.803*** 0.061 28.529*** 0.063 Model 2 Parent Marital Conflict -1.031 -1.904 -0.127 0.194 0.107 -0.060 -0.076 -0.314 Model 3 Parent Divorce + Marital Conflict -1.050 -1.864 -0.127 0.193 0.107 -0.054 -0.075 -0.327 0.747 -0.011 -0.026 Model 4 Divorce x Conflict -1.111 -1.824 -0.141 0.208 0.106 -0.060 -0.072 -0.397 -4.672 -0.141 1.154* 29.193** 0.088
Model 2 Parent Marital Conflict 0.010 -0.066 0.001 0.002 -0.001 0.005 -0.010* 0.020
Model 3 Parent Divorce + Marital Conflict 0.009 -0.074 0.001 0.002 -0.001 0.005 -0.010"* 0.023 0.054
Model 4 Divorce x Conflict -0.001 -0.063 0.001 0.003 -0.002 0.006 -0.010"* 0.017 -0.206 -0.016* 0.052**
-0.005
-0.007
0.881** 0.030
0.922"** 0.029
0.917" 0.032
0.965** 0.050*
(N
418)
'p<.01 *p<.O01
*p<.05
Parental Divorce and Marital Conflict / 909 of the four dependent variables. In general, psychological well-being, the quality of intimate relations, and support from kin and friends are lowest when low marital conflict is followed by divorce and when high marital conflict is not followed by divorce. Similarly, in all cases, the adverse outcomes associated with a high-conflict parental marriage that does not dissolve are more severe than those associated a with high-conflict parental marriage that breaks up. The figure also reveals that it was only among offspring from nondivorced families that conflict was negatively related to well-being. To explore this notion further, we omitted those offspring from the sample who had experienced a parental divorce and regressed the dependent variables for the remaining offspring on parental marital conflict. This revealed that conflict was significantly associated with offspring psychological distress (P - .11, p < .05), offspring overall happiness (p - -.10, p < .05), and offspring social resources (P- -.13, p <.05). The coefficient for offspring marital happiness, although negative, was not significant ( =-.11); this may have been due to the reduced sample size for this outcome. Because of the relatively small number of cases of parental divorce in our sample, we were concerned that a few outliers might be disproportionately affecting the results reported in Tables 1-4. Consequently, we conducted an analysis of residuals first using the Cook's distance procedure. No outliers were indicated. We then used DFBETA to detect outliers with respect to each interaction term involving the conflict measures. When we removed the three most influential cases, the pattern of findings remained the same, although with the reduced degrees of freedom (due to the fact that all of the influential cases were in the divorced group), there was a reduction in the significance level of some of the effects. Overall, the general pattern of interactions did not appear to be the result of a few outliers. We also carried out a series of multiple regression analyses identical to those described above, except that we excluded the 10 cases in which parental divorce had occurred after the age of 18. The results were similar to those reported in Tables 1 through 4. No significant additive effects appeared for parental marital conflict or divorce. For psychological distress, overall happiness, and marital happiness, the interaction terms were significant, in the same direction, and slightly stronger than those reported above. For offspring social resources, however, the interaction term was in the same direction but no longer significant. As a final step in the analysis, we considered the age of offspring at the time of divorce, as well as the length of time that had passed since divorce. Given the small number of cases of divorce, we relied on zero-order correlations. Parental marital conflict was not associated significantly with either age at time of divorce (r- -.01) or time since divorce (r - .05). Age at time of divorce was significantly associated with our measure of social resources (r = -.35, p <.05, two-tailed) but not with any other offspring outcome. Examination of means confirmed that offspring who were in their 20s at the time of parental divorce reported the lowest level of social resources. This accounts for why the interaction between divorce and marital conflict for this outcome was not significant when older offspring were removed from the analysis, as noted above. Length of time since divorce was not associated significantly with any
910 / Social Forces 73:3, March 1995 FIGURE 1: Offspring Outcomes as a Function of the Interaction between Parental Marital Conflict and Parental Divorce While Controlling for Parent's Age, Sex, Race, and Education and Offspring's Sex and Age Offspring Psychological Distress
3.5
3.0
2.5
2.0
1.5
0I
10
12
10
12
Parental Conflict
6 Parenta Conflic
10
12
6 Parental Conflict
10
Parents divorced
Parental Divorce and Marital Conflict / 911 above. Length of time since divorce was not associated significantly with any offspring outcome. Discussion Several perspectives are available in examining parental marital conflict, parental divorce, and offspring well-being. These views reflect a debate within the social scientific literature and the wider society about the relative importance of family structure (divorce) and family process (conflict) for children's well-being. Some have argued that family processes, such as conflict, are considerably more important than family structure in understanding children's well-being (e.g., Demo 1992; Marotz-Baden et al. 1979). Others have argued that although family processes are important, family structure differences in children's well-being are real and nontrivial (e.g., Hetherington 1989; McLanahan & Booth 1988). However, our data strongly support a different perspective - one that posits that the long-term consequences of divorce depend on the level of parental conflict prior to separation. Our results show that if conflict between parents is relatively high, offspring are better off in early adulthood if their parents divorced than if they remained married. This result is consistent with the notion, advanced by a number of observers, that children are better off in divorced single-parent families than in two-parent families marked by high levels of discord (Barber & Eccles 1992; Emery 1988; Furstenberg & Cherlin 1991; Kurdek 1981). On the other hand, if conflict between parents is relatively low, offspring are worse off in early adulthood if their parents divorced. Although our data suggest no simple, additive effects due to divorce, they nevertheless indicate that family structure cannot be ignored in understanding children's long-term outcomes. In fact, many of our results are the opposite of what one would expect to find on the basis of a pure conflict interpretation of divorce effects (model 2 described above). If conflict between parents is primarily responsible for the problems encountered by children of divorce, then those exposed to the most conflict prior to divorce should show the lowest level of well-being following divorce. However, our data reveal the opposite pattern: those offspring of divorce who were exposed to the least conflict show the lowest level of wellbeing in early adulthood. These results are consistent with a study by Wheaton (1990), who found that the mental health consequences of life transitions depend on the level of stress prior to the transition. His study showed that life changes (such as job loss, divorce, ending a romantic relationship, and retirement) are nonproblematic - even beneficial - when they are preceded by a high level of stress. On the other hand, when stress is low, transitions out of these roles are detrimental to psychological well-being. Although Wheaton's study dealt with adults, a similar process may operate for children. Depending on the level of stress associated with living in a twoparent household, parental divorce may be beneficial or detrimental to children. When conflict between parents is severe, divorce removes children from an
912 / Social Forces 73:3, March 1995 aversive and stressful home environment. Under these circumstances, the decline in stress may well outweigh the loss of other resources, resulting in an overall positive outcome. On the other hand, when overt conflict between parents is not excessive, children are unlikely to anticipate or welcome divorce. Under these circumstances, divorce represents a loss of resources for the child with no compensating gain. These findings are consistent with the qualitative observations of Wallerstein and Kelly (1980), who found that children who had little awareness of parental marital unhappiness prior to divorce reacted more intensely and negatively than did those who had been exposed to prolonged marital conflict. A curious pattern is evident in Figure 1: offspring who experienced high levels of parental marital conflict followed by parental divorce appeared to be doing particularly well in early adulthood - as well as those from low-conflict nondivorced families. Given the small number of cases of divorce in our sample, this particular aspect of our data should probably not be overinterpreted. The fact that the regression lines cross should be regarded as the key finding - not the predicted scores for individuals at the endpoints of the regression line. On the other hand, our results are consistent with perspectives emphasizing the potentially positive implications of divorce for children (e.g., Barber & Eccles 1992). Adjusting successfully to a family crisis may give children strengths not readily available to those raised in harmonious families, and these strengths may translate into enhanced competence and well-being in adulthood. Although speculative, this notion warrants further investigation. The fact that we found no simple, additive effects of parental divorce is contrary to some previous cross-sectional studies that show modest but significant differences between offspring raised in divorced and continuously intact two-parent families, irrespective of parental marital conflict (Amato & Keith 1991a; Cherlin & Furstenberg 1991; Emery 1988). The absence of observed effects may be due to the fact that most offspring in our sample experienced divorce as teenagers, rather than at a younger age. In fact, no children in the present study experienced divorce younger than age 9. If the effects of divorce are most pronounced for very young children, then our analysis underestimates the overall effect of divorce. However, Amato and Keith (1991a, 1991b) showed that many previous studies have failed to find significant differences between offspring raised in divorced and intact families. The general absence of significant additive effects of divorce in the present study, therefore, should not be seen as an anomaly. Furthermore, our results help to explain why effect sizes in this literature tend to be weak (Amato & Keith 1991a, 1991b). Because divorce is problematic for some children but not for others (depending on the level of conflict prior to divorce), calculating a mean score based on all children leads to a modest estimate of the effect of divorce. Our study also suggests that interparental conflict has modest but detrimental long-term consequences for offspring, provided that no parental divorce occurs. This finding is consistent with several earlier studies of adults (e.g., Adam, Bouckoms & Streiner 1982; Amato & Booth 1991; Booth & Edwards 1990; Overall, Henry & Woodward 1974.) However, previous studies have not had the advantage of independent measures collected over a long period of
Parental Divorce and Marital Conflict / 913 time. Instead, researchers have required respondents to recall the quality of their parents' marriage and provide information on their own current well-being. Estimates from such studies suffer from common method variance - a problem that artificially inflates correlations - as well as errors associated with using retrospective data. As far as we are aware, our study is the first to use longitudinal data from independent sources to confirm that parental marital discord (that is not followed by divorce) is associated with lowered offspring well-being in later life. In conclusion, it is often claimed that children are better off in single-parent families than in two-parent families marked by high levels of discord. Our study provides the strongest evidence yet in support of this notion. Divorce can remove a child from a hostile, dysfunctional, and perhaps abusive, environment. At the same time, our data indicate that the consequences of divorce for children are not always benign. Not all cases of divorce are preceded by a period of sustained overt conflict between spouses. Instead, some unhappily married parents exhibit a pattern of quiet disengagement. Other parents may leave a moderately happy marriage for reasons of personal fulfillment or if a more attractive partner comes along. And some relationships may be happy for a period of years, then deteriorate quickly as new circumstances arise (such as an extramarital affair or a wife's entering the labor force). In these cases, children may react with shock and disbelief, and divorce is likely to represent an unwelcome and uncontrollable long-term decline in quality of life. Clearly, the consequences of divorce for children can be benefitial or harmful, depending on the benefits and costs involved. Our results suggest that the impact of divorce on children's lives can be better understood by giving greater attention to what is happening in the family prior to divorce.
References Adams, Kenneth S., Anthony Bouckorns, and David Streiner. 1982. "Parental Loss and Family
Stability in Attempted Suicide." Archives of General Psychiatry 39:1081-85. Amato, Paul. R. 1986. "Marital Conflict, the Parent-Child Relationship, and Child Self-esteem." Family Relations 35:403-10. Amato, Paul R., and Alan Booth. 1991. "The Consequences of Parental Divorce and Marital Unhappiness for Adult Well-being." Social Forces 69:895-914. Amato, Paul R., and Bruce Keith. 1991a. "Parental Divorce and Adult Well-being- A Metaanalysis." Journal of Marriageand the Family 53:43-58.
.1991b. "Parental Divorce and the Well-being of Children: A Meta-analysis.' Psychological Bulletin 110:26-46. Aquilino, William S. 1991. "Family Structure and Home-leaving- A Further Specification of the Relationship." Journalof Marriageand the Family 53:999-1010.
Barber, Bonnie L, and Jacquelynne S. Eccles. 1992. "Long-term Influences of Divorce and Single Parenting on Adolescent Family- and Work-related Values, Behaviors, and Aspirations.' PsychologicalBulletin 111:108-26.
I0
R I
Journal of Communication
Editor Mark Levy University of Maryland
An official journal of the International Communication Association, Journal of Communication is aquarterly publication devoted to communication theory, research, policy, and practice. JOCbrings its readers the latest, broadest, and most important research inthe field of communication and has the widest readership of any communication journal. Its regular features include: Influential articles of theoretical and topical interest The Journalseeks the broadest range of methodological and conceptual approaches and features empirical, historical, critical, feminist, comparative, interpersonal, and policy studies. The most extensive book review section available No other journal provides as much information on current books incommunication and related fields. Each issue contains an extensive book review section and often an essay exploring books on aparticular theme. Multifaceted symposia of selected studies on current issues Most issues of the Journalfeature asymposium of original articles on a current theme, from telecommunications to storytelling, from video music to televised violence. Recent and forthcoming symposia include " Communication inChina " Focus on Latin America
Please enter myone-year subscription (4 issues) to the JournalofCommunication, 0 Institutions: $78 Add $15 for subscriptions outside the US. [I Individuals: $36.50
Name Addr City/st aJzp mXCOC
Send order with payment to the Journals Department. Oxford University Press, 2001 Road. Cary, NC 27513. Evans For credit card orders and (MasterCard. VISA, American Express are accepted), include account number, expiration date, and fax: signature. Toorder by 919-677-1714. Or call 800-852-7323