parental predivorce relations and offspring postdivorce well-being

16
Journal of Marriage and Family 63 (February 2001): 197–212 197 ALAN BOOTH AND PAUL R. AMATO The Pennsylvania State University l Parental Predivorce Relations and Offspring Postdivorce Well-Being This 2-part study uses national longitudinal inter- view data from parents and their adult children to examine the way in which predivorce marital conflict influences the impact of divorce on chil- dren. In the 1st study, we find that the dissolution of low-conflict marriages appears to have nega- tive effects on offspring’s lives, whereas the dis- solution of high-conflict marriages appears to have beneficial effects. The dissolution of low-con- flict marriages is associated with the quality of children’s intimate relationships, social support from friends and relatives, and general psycho- logical well-being. The 2nd study considers how parents in low-conflict marriages that end in di- vorce differ from other parents before divorce. We find that low-conflict parents who divorce are less integrated into the community, have fewer imped- iments to divorce, have more favorable attitudes toward divorce, are more predisposed to engage in risky behavior, and are less likely to have ex- perienced a parental divorce. Evidence suggests that parents’ marital conflict and divorce have adverse effects on offspring, ef- fects that often persist into young adulthood (Amato & Booth, 1997; Cherlin, Chase-Lansdale, & McRae, 1998; McLanahan & Sandefur, 1994). Department of Sociology, The Pennsylvania State Univer- sity, 212 Oswald Tower, University Park, PA 16803 ([email protected]). Key Words: adult child well-being, parental conflict, pa- rental divorce, parental marital commitment. Recent studies by Amato, Loomis, and Booth (1995) and Jekielek (1998) have drawn attention to the importance of predivorce conflict in under- standing the impact of divorce on offspring well- being. Children appear to benefit from the divorce of parents in high-conflict marriages. On the other hand, children appear to suffer from the divorce of parents in low-conflict marriages. These two studies contain some limitations, however. Amato, Loomis, and Booth (1995) used a relatively small sample of offspring from divorced families, and Jekielek (1998) relied on a single source of infor- mation (parents) for data on independent and de- pendent variables. Furthermore, neither study ex- amined how predivorce conflict is related to postdivorce factors (such as the amount of acri- mony following marital dissolution, residential mobility, changes in schools, and the number of family transitions) that may affect the long-term well-being of children. These two prior studies raise important ques- tions. Can the results be replicated using larger samples, more stringent methods, and a more comprehensive set of outcomes? How is predi- vorce conflict related to postdivorce conflict, as well as to other postdivorce factors that may affect children? Finally, there are no studies of parents who seldom disagree or fight, yet end their mar- riages in divorce, a seemingly incongruous marital outcome, but one that appears to be fairly com- mon (Amato & Booth, 1997). What qualities dis- tinguish parents who dissolve low-conflict mar- riages from parents who dissolve high-conflict

Upload: alan-booth

Post on 23-Jul-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Parental Predivorce Relations and Offspring Postdivorce Well-Being

Journal of Marriage and Family 63 (February 2001): 197–212 197

ALAN BOOTH AND PAUL R. AMATO

The Pennsylvania State University

l

Parental Predivorce Relations and Offspring Postdivorce

Well-Being

This 2-part study uses national longitudinal inter-view data from parents and their adult childrento examine the way in which predivorce maritalconflict influences the impact of divorce on chil-dren. In the 1st study, we find that the dissolutionof low-conflict marriages appears to have nega-tive effects on offspring’s lives, whereas the dis-solution of high-conflict marriages appears tohave beneficial effects. The dissolution of low-con-flict marriages is associated with the quality ofchildren’s intimate relationships, social supportfrom friends and relatives, and general psycho-logical well-being. The 2nd study considers howparents in low-conflict marriages that end in di-vorce differ from other parents before divorce. Wefind that low-conflict parents who divorce are lessintegrated into the community, have fewer imped-iments to divorce, have more favorable attitudestoward divorce, are more predisposed to engagein risky behavior, and are less likely to have ex-perienced a parental divorce.

Evidence suggests that parents’ marital conflictand divorce have adverse effects on offspring, ef-fects that often persist into young adulthood(Amato & Booth, 1997; Cherlin, Chase-Lansdale,& McRae, 1998; McLanahan & Sandefur, 1994).

Department of Sociology, The Pennsylvania State Univer-sity, 212 Oswald Tower, University Park, PA 16803([email protected]).

Key Words: adult child well-being, parental conflict, pa-rental divorce, parental marital commitment.

Recent studies by Amato, Loomis, and Booth(1995) and Jekielek (1998) have drawn attentionto the importance of predivorce conflict in under-standing the impact of divorce on offspring well-being. Children appear to benefit from the divorceof parents in high-conflict marriages. On the otherhand, children appear to suffer from the divorceof parents in low-conflict marriages. These twostudies contain some limitations, however. Amato,Loomis, and Booth (1995) used a relatively smallsample of offspring from divorced families, andJekielek (1998) relied on a single source of infor-mation (parents) for data on independent and de-pendent variables. Furthermore, neither study ex-amined how predivorce conflict is related topostdivorce factors (such as the amount of acri-mony following marital dissolution, residentialmobility, changes in schools, and the number offamily transitions) that may affect the long-termwell-being of children.

These two prior studies raise important ques-tions. Can the results be replicated using largersamples, more stringent methods, and a morecomprehensive set of outcomes? How is predi-vorce conflict related to postdivorce conflict, aswell as to other postdivorce factors that may affectchildren? Finally, there are no studies of parentswho seldom disagree or fight, yet end their mar-riages in divorce, a seemingly incongruous maritaloutcome, but one that appears to be fairly com-mon (Amato & Booth, 1997). What qualities dis-tinguish parents who dissolve low-conflict mar-riages from parents who dissolve high-conflict

Page 2: Parental Predivorce Relations and Offspring Postdivorce Well-Being

198 Journal of Marriage and Family

marriages or from parents who remain in low-con-flict marriages? Understanding the characteristicsof parents in low-conflict marriages that end indivorce may provide insights into why these mar-riages end and why these divorces are especiallydifficult for children.

The research reported here involves two stud-ies. Study 1 replicates and extends the Amato,Loomis, and Booth (1995) and Jekielek (1998)studies. Study 2 focuses on the characteristics ofparents in low-conflict marriages that end in di-vorce and how these parents differ from parentsin high-conflict marriages that end in divorce andfrom parents in low-conflict marriages that remaintogether.

STUDY 1

The two prior studies that demonstrated the inter-active effects of parents’ marital conflict and di-vorce drew on different populations and used dif-ferent outcomes. The analytic methods in the twostudies were similar, however.

Amato, Loomis, and Booth (1995) drew on theMarital Instability Over the Life Course study(Booth, Amato, Johnson, & Edwards, 1993), aninvestigation that began in 1980 with a nationalsample of more than 2,000 married persons. Thoseindividuals were interviewed in 1983, 1988, and1992. In 1992, a random sample of offspring whowere living in the parents’ home in 1980 and whohad reached the age of 19 were interviewed. Ofthese individuals, 42 had experienced a parentaldivorce between 1980 and 1992. One advantageof the study was that data on marital conflict wereobtained from parents before divorce, and data onchildren’s well-being were obtained from off-spring after the divorce. This design avoided prob-lems associated with same-source bias and the useof retrospective data. The dependent variables in-cluded overall happiness, psychological distress,the number of close kin and friends, and (amongmarried offspring) marital happiness. Ordinaryleast squares (OLS) regression revealed significantinteractions between parents’ marital conflict andparental divorce. For all four outcomes, offspringfrom high-conflict families were better off if a di-vorce occurred, whereas offspring from low-con-flict families were better off if a divorce did notoccur.

Jekielek (1998) drew on the 1988 and 1992waves of the National Longitudinal Study ofYouth. Her study included 1,640 children aged 6to 14 who were living with both biological parents

in 1988. Of these children, 239 experienced a pa-rental divorce by 1992. Marital conflict was basedon parents’ reports in 1988. The dependent vari-ables were scales of children’s anxiety and de-pression, completed by parents in 1992. Jekielekused OLS regression and included a parental con-flict by divorce interaction term in the analysis.Elevated levels of anxiety and depression amongchildren were observed following divorce, partic-ularly divorces that occurred within the previous2 years. Significant interactions revealed, howev-er, that when marital conflict was high in 1988,children were better off in 1992 if a divorce oc-curred than if parents remained married. Childrenwith the highest levels of anxiety and depressioneither had low-conflict parents who obtained a di-vorce or high-conflict parents who did not obtaina divorce.

Conceptual Perspective

Researchers have employed a variety of theoriesand conceptual perspectives to explain how di-vorce affects children. The largest number of stud-ies were begun with the assumption that maritaldisruption is a potentially stressful life transitionto which children must adjust. Many researcherslink their work to established stress perspectives,such as family stress and coping theory (Hill,1949; McCubbin & Patterson, 1983), generalstress theory (Thoits, 1995), and the risk and re-siliency perspective (Cowan, Cowan, & Schulz,1996). Because stress frameworks dominate theliterature on divorce, we give them priority here.

A stress perspective views marital dissolutionas a process that begins while parents are still liv-ing together and ends long after the legal divorceis concluded. The disruption process typically setsinto motion numerous events that most childrenexperience as stressful. For example, divorcesometimes results in less effective parenting fromthe custodial parent, a decrease in involvementwith the noncustodial parent, exposure to continu-ing interparental discord, a decline in economicresources, and other disruptive life events, such asmoving and changing schools (Amato, 1993;McLanahan & Sandefur, 1994; Simons & Asso-ciates, 1996). These stressors, in turn, increase therisk of negative emotional, behavioral, and healthoutcomes for children. The severity and durationof these negative outcomes varies, depending onthe presence of a variety of moderating or protec-tive factors in children’s lives. Moderators includethe child’s resources (self-efficacy, coping skills),

Page 3: Parental Predivorce Relations and Offspring Postdivorce Well-Being

199Parental Predivorce Relations

as well as interpersonal relationships (social sup-port) and community factors (community servic-es, supportive government policies). Another keymoderator refers to the manner in which childrendefine divorce, with some viewing it as a personaltragedy and others viewing it as an escape froman aversive and dysfunctional home life.

Some researchers have argued that stress per-spectives tend to focus exclusively on the negativeaspects of divorce and ignore positive outcomesfor children (Barber & Eccles, 1992; Gately &Schwebel, 1991). The notion that divorce can bebeneficial is not inconsistent with a stress per-spective, however. Thoits (1995), for example, ar-gued that potentially stressful events, such as mar-ital disruption, can have positive long-termconsequences when people resolve their problemssuccessfully. Similarly, Wheaton (1990) demon-strated, using longitudinal data, that when individ-uals are moderately happy with their marriages,divorce is followed by a decline in psychologicalwell-being. But when individuals are unhappywith their marriages, divorce is followed by anincrease in psychological well-being. Divorce rep-resents an increase in stress for the former groupsand a decrease in stress for the latter group. Con-sequently, divorce is likely to be defined differ-ently by these two groups of individuals.

We believe that a similar process applies tochildren. Children in households where parentsengage in a long-term process of overt, unresolvedconflict are at risk for a variety of developmentaland emotional problems (Davies & Cummings,1994; Emery, 1999). When a divorce occurs, thesechildren are freed from a dysfunctional home en-vironment and may genuinely welcome the shiftto a calmer single-parent household. Under thesecircumstances, children’s conflict-related symp-toms are likely to improve over time. In contrast,children in households in which parents engage inrelatively little overt conflict are at low risk fordevelopmental and emotional problems. Even ifparents are not highly committed to the marriage,children still benefit from the economies of scaleprovided by a two-parent household (and usuallytwo parental incomes) and by having ready accessto both parents. They also benefit from stability intheir neighborhood and friendships, as well asfrom regular school attendance. Children are like-ly to view divorce under these circumstances asan unexpected, unwelcome, and uncontrollableevent, an event that sets into motion a series ofstressful circumstances (a decline in standard ofliving, loss of contact with one parent, and mov-

ing) with no compensating advantages. Underthese circumstances, children may exhibit a vari-ety of stress-related symptoms, including loweracademic achievement, behavioral and emotionalproblems, loss of affection for parents, and diffi-culties in forming close, long-term ties with oth-ers.

A variety of studies support this interpretation.In an Australian study, many children from di-vorced families reported deep sadness and linger-ing hopes for parental reconciliation, whereas oth-er children reported relief and a belief that thedivorce was ‘‘for the best.’’ Those children whoreported relief tended to come from the most con-flicted family environments prior to divorce(Amato, 1987). In a qualitative study of collegestudents from divorced families (Duran-Aydintug,1997), those who recalled the highest levels ofstress at the time of parental separation alsoclaimed that their parents had ‘‘kept them in thedark’’ about the deteriorating state of the mar-riage. These students also were more likely thanothers to report having relatively low levels oftrust in intimate relationships. Similarly, Jacobson(1978) found that most parents inform their chil-dren about the impending divorce close to thetime of separation—usually within a week, andoften on the same day. Jacobson reported thatchildren had the most difficult time adjusting tothe divorce when they had the least time to pre-pare for it.

Several studies indicate that children who placethe blame for the divorce on themselves tend tobe more poorly adjusted (Bussell, 1995). Healy,Steward, and Copeland (1993) studied primaryschool children 6 months after parental separationand found that one third reported some feelingsof self-blame. Self-blame, in turn, was related toa variety of child problems, including depression,externalizing problems, and lowered feelings ofself-competence. Presumably, when children dis-cern few external reasons for the divorce (whichis more likely to occur when the divorce is pre-ceded by relatively little conflict), they are morelikely to make self-attributions for its occurrence(Grych & Finchan, 1990). In addition to self-blame, Kim, Sandler, and Jenn-Yum (1997) foundthat children’s perceived lack of control overevents mediated some of the impact of divorce-related stress on adjustment. It is likely that chil-dren perceive less control over events when thedivorce is unanticipated, that is, when it is notpreceded by a long period of overt marital discord.In general, studies suggest that children have the

Page 4: Parental Predivorce Relations and Offspring Postdivorce Well-Being

200 Journal of Marriage and Family

most difficulty when they do not anticipate theirparents’ divorce, when they attribute blame tothemselves, and when they perceive events sur-rounding the divorce as uncontrollable—all out-comes that are most likely to occur when divorceis not preceded by a prolonged period of overtconflict between parents.

In summary, divorce has the potential to be astressful process for children, a process that placeschildren at risk for a variety of behavioral, emo-tional, and academic problems. When the homeenvironment is highly stressful, divorce may low-er the level of stress to which children are ex-posed, however, leading to improvements in chil-dren’s well-being. On the other hand, when thehome environment is relatively benign, divorce islikely to increase the level of stress to which chil-dren are exposed, leading to a deterioration inchildren’s well-being. Children’s lack of prepara-tion, self-attributions of blame, and a perceivedlack of control over events are likely to exacerbatethe degree of stress experienced around the timeof divorce.

Contributions of the Current Study

The study presented here serves as a replicationand extension of the Amato, Loomis, and Booth(1995) study. Compared with the previous study,it uses a larger sample, a wider range of outcomes,and a more sophisticated method of data analysisbased on structural equation modeling. The largersample of children from divorced families (85 inthe current study vs. 42 in the original study) in-creases our confidence in the findings. The morecomprehensive exploration of outcomes gives usadditional insight into the pervasiveness of the im-pact of marital conflict and divorce on children.The use of structural equation methods allows usto incorporate measurement error into our modelsand adopt a parsimonious data-reduction strategywith respect to our dependent variables.

Neither Amato and his colleagues nor Jekielekexplored other aspects of the divorce that may berelated to predivorce conflict. Conflict surround-ing marital disruption (over custody, child rearing,visitation, or child support) may be the first timeoffspring from low-conflict marriages have en-countered severe interparental discord, much of itdirectly or indirectly involving them. In contrast,for offspring from high-conflict marriages, the de-parture of one parent from the household may re-sult in an overall reduction in turmoil, which maybenefit children. Indeed, it is not clear whether

pre- and postdivorce conflict between parents arepositively or negatively correlated.

Another aspect of divorce that has been shownto be a source of stress for parents and childrenis a change in residence, especially if it involvesmoving to lower quality housing (Booth & Ama-to, 1993; South, Crowder, & Trent, 1998). Movingoften entails a change in schools, which can resultin lower offspring achievement (McLanahan &Sandefur, 1994). Again, this may be more seriousfor offspring of parents in low-conflict marriages,who see it as an additional cost, whereas thosefrom high-conflict marriages may view it as ‘‘asmall price to pay’’ to get out of a conflict-riddenhome.

The number of parent relationship transitionshas been shown to have adverse consequences forchildren (Amato & Booth, 1991b; Booth, Brink-erhoff, & White, 1984; Wu, 1996). There are sev-eral reasons the number of transitions may begreater among low-conflict parents than high-con-flict parents. Because people in the former groupare not prone to conflict, they may be more de-sirable partners and therefore more likely to re-marry. The former group, however, also may havehigher standards for continuing a marriage, andmild perturbations in the relationship may be suf-ficient for the dissolution of second- and higherorder marriages. Thus, low-conflict parents mayexperience more subsequent marital transitionsthan do high-conflict parents, which would ac-count for the adverse outcomes we observe.

The focus of previous studies was largely onoffspring’s psychological well-being. Amato,Loomis, and Booth (1995) took into account over-all happiness and psychological distress, whereasJekielek (1998) focused on anxiety and depres-sion. In addition, Amato and his colleagues in-cluded the number of people to whom offspring‘‘felt close,’’ as well as marital happiness amongmarried offspring. The current study focuses notonly on offspring’s psychological well-being, butalso on the size of offspring’s kin support net-works, friend support networks, the quality of in-timate relationships in general, offspring’s affec-tion for parents, and educational attainment. Byexamining a variety of domains, we gain a cleareridea of which aspects of children’s lives are mostaffected.

METHOD

Sample. A 17-year longitudinal study of maritalinstability over the life course (Booth, Amato,

Page 5: Parental Predivorce Relations and Offspring Postdivorce Well-Being

201Parental Predivorce Relations

TABLE 1. MEANS AND STANDARD DEVIATIONS FOR ALL

VARIABLES

Variables X S

Independent variablesDivorced (1980–1992)Martial conflict

Dependent variablesGlobal happinessLife satisfaction

.143.98

2.373.88

.341.88

.54

.53Self-esteemPsychological distressNumber of close relativesNumber of relatives talk withNumber of close friendsNumber of friends talk with

3.341.623.892.514.153.28

.40

.404.022.663.562.94

Happiness intimate relationsInteraction intimate relationsAffection for mothersAffection for fathers

.00

.002.702.51

1.001.00.40.52

Control variablesOffspring femaleOffspring ageOffspring nonwhiteParents’ education

.4824.28

.1413.86

.504.74.35

2.20

Note: Sample size is 629.

Johnson, & Edwards, 1998) was used to examinethe questions we have advanced. A national sam-ple of 2,033 married persons (not couples) age 55and under were interviewed by telephone in 1980.Husbands or wives were selected for an interviewusing a second random procedure. The samplewas interviewed again in 1983, 1988, 1992, and1997.

A random sample of offspring also were inter-viewed as part of the 1992 and 1997 waves ofinterviews. Of the 58% of the original sample ofpersons reinterviewed in 1992 (n 5 1,183), 625had offspring aged 19 or older who had lived inthe parental household in 1980. In 1997, 275 ofthe 1,193 reinterviewed respondents had offspringin the household in 1980 who had reached the ageof 19 and had not been interviewed in 1992. Over-all, 87% (n 5 787) of the parents provided namesand telephone numbers of children. We were ableto obtain interviews with 691 (88% of those forwhom we had names) for an overall completionrate of 77%. Our analyses pooled data from 471offspring interviewed in 1992 and 220 offspringinterviewed in 1997. In households where therewas more than one eligible child, a random pro-cedure was used to select the child for inclusionin the study. In households with two eligible chil-dren, the name and telephone number of the sec-ond offspring also were obtained. In householdswith three or more eligible children, the secondchild was selected at random. The second childwas interviewed when it was not possible to ob-tain an interview with the first. Five percent of theinterviews were with offspring other than the firstselected. The ages of offspring who were suc-cessfully interviewed ranged from 19 to 47, witha median age of 23.

The original sample of married persons wascompared with 1980 Census data and found to berepresentative with respect to age, race, householdsize, presence of children, tenure, and region. Sub-sequent waves were slightly less representativewith respect to African Americans, younger re-spondents, male respondents, renters, householdswhere the husband had not been to college, south-ern residents, and individuals residing in metro-politan areas. Nonetheless, within-category dis-crepancies between the sample and the populationwere slight, with most being less than 3%.

Logistic regression was used to assess demo-graphic and marital factors in 1988 (for offspringinterviewed in 1992) and in 1992 (for offspringinterviewed in 1997) that influenced whether weobtained the names and telephone numbers of off-

spring from parents and whether an offspring in-terview was obtained. Older respondents, respon-dents who had experienced divorce, and moredivorce-prone individuals (people thinking aboutor taking steps to end the marriage) were slightlyless likely to provide a name and phone number.Similarly, the offspring of such respondents wereslightly less likely to grant an interview. Giventhat we were less likely to obtain interviews withoffspring when their parents had troubled mar-riages, the restricted range of the independent var-iables may lead to slight underestimates of the ef-fects of marital discord and divorce on offspring.Consequently, the conclusions drawn from ouranalysis may be more conservative than conclu-sions drawn from an analysis based on all eligibleoffspring.

Measurement

The 1980 parental interview provided informationon the quality of the parents’ marriage. Subse-quent interviews provided information on whetherparents had divorced and, if still married, the levelof marital conflict. The 1992 and 1997, offspringinterviews provided data on children’s psycholog-ical well-being, kin support, friend support, thequality of intimate relations, parent-child rela-tions, and educational achievement. Table 1 con-tains means and standard deviations for the majorvariables.

Page 6: Parental Predivorce Relations and Offspring Postdivorce Well-Being

202 Journal of Marriage and Family

Marital conflict. Our measure of marital conflictwas based on four items from the parent inter-view. These items dealt with (a) whether couplesargue over the household division of labor, (b) thefrequency of disagreements in general, (c) thenumber of serious quarrels in the past 2 months,and (d) whether spouses were so angry that theyslapped, hit, pushed, kicked, or threw things atone another. Alpha reliability coefficients were.54, .52, .47, and .52 in 1980, 1983, 1988, and1992, respectively. Correlations between time pe-riods ranged from .53 to .65, with a mean corre-lation of .59 (all p , .001). To provide an estimateof the general level of conflict in the marriage, weaveraged conflict scores across the four inter-views. Averaging the four scores improved the re-liability of measurement, resulting in a combinedalpha coefficient of .85. In cases of divorce, how-ever, conflict was based on the mean of all conflictscores obtained before the separation. (In one al-ternative analysis not shown, we used the conflictscore closest in time to separation for marriagesthat ended in divorce. In another analysis, we re-lied entirely on conflict in 1980 for all cases.These alternative procedures yielded results thatwere substantively identical to those shown be-low.)

Divorce. Between 1980 and 1997, 85 cases of di-vorce occurred (0 5 not divorced, 1 5 divorced).Of these cases, 34 (40%) were between ages 0and 12 at the time of parental divorce, 40 (47%)were between ages 13 and 19, and 11 (13%) werebetween ages 20 and 25. We included the oldergroup because research shows that parental divor-ce is associated with problematic child outcomes,even if it occurs when offspring are in their early20s (Cooney & Kurtz, 1996). Furthermore, omit-ting the oldest offspring from the analyses did notchange the results reported later. Our analysis alsoincluded 544 offspring who never experienced aparental divorce. Offspring whose parents di-vorced before 1980, and those who experiencedthe death of a parent, were excluded from theanalysis.

To get an overview of offspring well-being inearly adulthood, we included indicators from sixdomains: psychological well-being, support fromkin, support from friends, the quality of intimaterelationships, the quality of relations with parents,and educational attainment. Data came from the1992 or 1997 interview, depending on when off-spring were first interviewed. (The questions usedin both interview years were identical.)

Psychological well-being. We used four indicatorsof offspring psychological well-being. Globalhappiness was based on the question, ‘‘Taking allthings together, how happy would you say you arethese days? Would you say you are (1) not veryhappy, (2) somewhat happy, or (3) very happy?’’We measured life satisfaction by having offspringrate their satisfaction in seven areas: neighbor-hood, job or career, house or apartment, friends,hobbies or leisure activities, family life outside thecurrent marriage, and financial situation. Responseoptions ranged from 1 (no satisfaction) to 5 (agreat deal of satisfaction), and the mean ratingserved as the scale score (alpha 5 .65). We alsoincluded six items from the Rosenberg (1965)self-esteem scale. Sample items included ‘‘I feelthat I have a number of good qualities’’ and ‘‘Icertainly feel useless at times.’’ Offspring’s re-sponses ranged from 1 (disagree strongly) to 4(agree strongly), and the mean across the sixitems served as the scale score (alpha 5 .77). Fi-nally, we used nine items from the Langner (1962)scale of psychological distress. Examples of itemsincluded ‘‘How often in the last year have youfound yourself wondering if anything is worth-while anymore?’’ and ‘‘How often would you sayyou felt isolated or alone, somewhat apart fromothers, even among friends?’’ Response optionsranged from 1 (never) to 3 (often), and the meanserved as the scale score (alpha 5 .73).

Kin support. To assess the size of offspring’s kinsupport networks, we asked, ‘‘Are there relativesyou have whom you feel emotionally close to? Ifyes, how many would that be?’’ In answering thequestions, respondents were instructed to excludetheir parents or any relatives currently living withthem. We also asked, ‘‘How many of your rela-tives do you feel you could talk with about prob-lems you may have in your family?’’ Because asmall number of individuals provided very highestimates, we recoded the maximum number ofrelatives at 11 for both questions.

Friend support. To assess the size of offspring’snetwork of friends, we asked, ‘‘Are there peoplewhom you consider very close friends who arenot relatives? If yes, how many close friendswould that be?’’ We then asked, ‘‘How manyclose friends do you have that you can talk withabout problems you may have with your family?’’Respondents were instructed not to refer to rela-tives in answering these questions. We recoded asmall number of very high estimates to 11.

Page 7: Parental Predivorce Relations and Offspring Postdivorce Well-Being

203Parental Predivorce Relations

Intimate relations. Married offspring were admin-istered an 11-item marital happiness scale. Themeasure tapped global feelings about the marriage(e.g., overall happiness, strength of love) and theperson’s feelings about specific aspects of the re-lationship (e.g., the amount of understanding re-ceived, affection, the extent to which spousesagree about things, the sexual relationship, thespouse as someone with whom to do things, andthe spouse’s faithfulness). The marital happinessscale was adapted and administered to cohabitingcouples. Another adaptation of the marital happi-ness scale was devised for individuals ‘‘datingsomeone seriously.’’ Scale items for individuals incohabiting or dating relationships were identicalto the scale items for married individuals, exceptthat ‘‘spouse’’ was replaced with ‘‘friend.’’

For those not steadily dating, two questionswere asked: ‘‘Do you seem to have difficulty dat-ing people with whom you feel you could developa serious relationship?’’ and ‘‘Would you preferto date more often, less often, or about the sameamount as you are now.’’ A 4-point scale was de-rived from these two questions, with those re-sponding ‘‘no’’ to the first question and ‘‘aboutthe same’’ to the second question receiving thehighest score. Those replying ‘‘yes’’ to the firstquestion and indicating they wanted to date ‘‘moreoften’’ or ‘‘less often’’ received the lowest score.The four scales (for those who were married, co-habiting, dating someone steadily, and not datingsomeone steadily) were standardized (z scored)and combined to form a single variable reflectingoverall happiness with intimate relationships.

To obtain a measure of frequency of interac-tion, those in marital or cohabiting relationshipswere asked a series of questions about how fre-quently they engaged in five activities with theirpartners: eating the main meal, shopping, visitingfriends, working on projects around the home, andgoing out for recreation (1 5 never, 4 5 almostalways). The items were summed to provide ageneral interaction score. For those not married orcohabiting, interaction was based on the numberof times in the previous month respondents datedor went out with their current partner (for thosedating someone steadily) or any other individual(for those not dating someone steadily). The fourvariables were standardized (z scored) and com-bined to form a single variable reflecting interac-tion in intimate relationships.

Parent-child relations. Six-item scales were de-vised to assess the current quality of relationships

with mothers and fathers. The first five items dealtwith the extent to which relations with mothers(or fathers) were characterized by understanding,trust, respect, fairness, and affection (1 5 verylittle, 2 5 some, and 3 5 a great deal). The lastquestion was ‘‘Overall, how would you describeyour relationship with your mother? Would yousay you are very close, somewhat close, or notvery close?’’ The scale had an alpha coefficientof .87 for mothers and .90 for fathers.

Socioeconomic achievement. We attempted to cre-ate a latent socioeconomic status variable basedon education, income, and occupational status. Wewere not successful, however, because of the lowcorrelations between variables. For example, ed-ucation and income correlated at r 5 .15. Thisresult was due to the young age of the sample,with many not yet having had time to achieve in-come levels appropriate to their educations. Con-sequently, we relied on years of education as asingle, observed variable. As it turned out, thisvariable did not yield any significant results, sowe excluded it from the measurement model andthe main results described below.

Measurement model. Rather than examine eachoutcome separately, we subjected the 12 measuresto a confirmatory factor analysis. We used theAnalysis of Moment Structures (AMOS) program(Arbuckle, 1997) with maximum likelihood esti-mation for this purpose. Figure 1 shows the mea-surement model based on all 12 measures. As weanticipated, a model based on five latent vari-ables—psychological well-being, kin support,friend support, intimate relationships, and parent-child relationships—fit the data reasonably well.Although the model was statistically significant,the goodness-of-fit indicators were high (GFI 5.98, AGFI 5 .97). All of the paths between latentvariables and indicators were significant (p ,.001). Some of the latent variables were moder-ately correlated. For example, psychological well-being correlated at .58 (p , .001) with the qualityof parent-child relations and at .51 (p , .001)with the quality of intimate relationships. Decreas-ing the number of latent variables, however, re-sulted in significantly poorer fitting models, so themodel with five latent variables was retained forall analyses.

Control variables. Control variables included off-spring gender (0 5 male respondent, 1 5 femalerespondent), age, race (0 5 White, 1 5 non-

Page 8: Parental Predivorce Relations and Offspring Postdivorce Well-Being

204 Journal of Marriage and Family

FIGURE 1. MEASUREMENT MODEL FOR OFFSPRING

OUTCOMES

White), and the mean of the mothers’ and fathers’years of education. ‘‘Age’’ refers to age in 1992for offspring first interviewed in 1992 and to agein 1997 for offspring interviewed in 1997.

Events following divorce. Postdivorce factors thatmay be associated with predivorce marital con-flict, as well as offspring well-being, were askedof the 85 offspring who experienced a parentaldivorce between 1980 and 1997. Postdivorce con-flict between parents was assessed with four ques-tions regarding the amount of conflict followingseparation with respect to (a) where you lived, (b)how you were raised, (c) visitation, and (d) childsupport payments (1 5 none, 2 5 some, 3 5 agreat deal). The mean of the four items served asthe scale score, and the final scale was logged toreduce skewness at the high end. The scale hadan alpha coefficient of .80. Residential mobilitywas assessed with the question, ‘‘During the yearfollowing the separation, did you move?’’ Forty-six percent reported a move (1 5 moved, 0 5 didnot move). Offspring who were in school at thetime of separation were asked, ‘‘During the yearfollowing the separation, did you changeschools?’’ Twenty-five percent of offspringchanged schools (1 5 changed schools, 0 5 didnot change schools). Family transitions were ob-tained by counting whether the biological parents

had remarried, and if so, whether they subsequent-ly divorced. The number of transitions rangedfrom 0 to 3, with 12% reporting three changes,30% two changes, 28% one change, and the re-mainder none.

RESULTS

Parental Conflict and Divorce: Additive Models

Preliminary analysis (not shown) involved a seriesof structural equation models in which all five di-mensions of offspring well-being were predictedby marital conflict, divorce, and the control vari-ables. The first model included conflict but ex-cluded divorce, the second model included divor-ce but excluded conflict, and the third modelincluded conflict as well as divorce. All threemodels included the full set of control variables.Model 1 revealed that marital conflict was asso-ciated negatively with psychological well-being(B 5 2.025, p , .01), kin support (B 5 2.197,p , .05), and parent-child relations (B 5 2.033,p , .001), but not with the other outcomes. (Un-standardized B coefficients are presented in thetext.) Model 2 revealed that divorce was associ-ated negatively with parent-child relations (B 52.216, p , .001) but not with the other outcomes.Finally, Model 3, with conflict and divorce in theequation together, revealed associations betweenconflict and psychological well-being (B 52.024, p , .05), conflict and kin support (B 52.192, p , .05), conflict and parent-child rela-tions (B 5 2.023, p , .001), and divorce andparent-child relations (B 5 2.186, p , .001).

These results suggest that conflict had moreimplications than divorce for offspring’s long-term well-being, at least in models that excludedinteraction terms. With regard to parent-child re-lations, however, parents’ marital conflict and di-vorce had independent estimated effects. Further-more, relatively little of the total estimated effectof divorce on parent-child relations was spurious(due to conflict that preceded divorce), as reflectedin the minor decline in the divorce coefficient be-tween Model 2 and Model 3 (2.216 vs. 2.186).The difference between offspring from divorcedand nondivorced families in Model 3 was rela-tively large, representing .67 of a standard devi-ation on the latent variable. Additional analyseswith mother-child relations and father-child rela-tions treated as separate observed variables re-vealed that marital conflict and divorce had sig-nificant independent associations with each

Page 9: Parental Predivorce Relations and Offspring Postdivorce Well-Being

205Parental Predivorce Relations

TABLE 2. UNSTANDARDIZED COEFFICIENTS (MAXIMUM LIKELIHOOD ESTIMATES) SHOWING ASSOCIATIONS BETWEEN

PARENTS’ MARITAL CONFLICT, DIVORCE, AND DIMENSIONS OF OFFSPRING WELL-BEING IN EARLY ADULTHOOD

Predictors

Dimensions of Offspring Well-being

PsychologicalWell-being

KinSupport

FriendSupport

IntimateRelations

Parent-childRelations

ConflictDivorceConflict 3 divorceOffspring age

2.05****2.59****

.11****

.01**

2.26***21.96*

.37*2.09***

2.17**22.43**

.47***2.11****

2.06**2.55*

.15**2.01

2.02****2.20**

.00

.00FemaleNonwhiteParents’ educationR squared

.05

.04

.02***

.09****

.67**

.382.04

.04***

2.75***2.13

.19***

.06***

.02

.00

.03

.04**

.012.01

.00

.16****

Note: Model x2 5 191.91; df 5 94; GFI 5 .97; AGFI 5 .94. Significance tests for coefficients are two-tailed. Samplesize is 629.

*p , .10. **p , .05. ***p , .01. ****p , .001.

variable, although the estimated effects tended tobe larger for fathers than for mothers.

Parental Conflict and Divorce: InteractiveModels

Our major analysis involved a model in which thefive offspring outcomes were predicted simulta-neously by marital conflict, divorce, the interac-tion between marital conflict and divorce, and thecontrol variables. These results appear in Table 2.The interaction term was significant for psycho-logical well-being, friend support, and intimate re-lations, and was marginally significant for kinsupport. For each of these four outcomes, the di-rection of the interaction was the same.

The R2 values for all dependent variables (withthe exception of parent-child relations) were low,but this is not unusual for models that includeskewed dichotomous predictors, such as divorce.The AMOS program does not provide significancetests for R2 coefficients; consequently, we used abootstrap procedure with 500 replications to esti-mate standard errors (Arbuckle, 1997). This pro-cedure revealed that the R2 values for all depen-dent variables were fully significant (p , .05),with the exception of the coefficient for intimaterelations, which approached significance (p ,.10).

To reveal the nature of the interactions moreclearly, we calculated the regression lines separate-ly for offspring from divorced and nondivorcedfamilies. The AMOS program does not calculateintercepts for latent variables. Consequently, to es-timate intercepts for each dependent variable, wemultiplied the means of each predictor variable bythe B coefficients in Table 2, added these values,

and subtracted the total from 0, which is the meanof the latent variable. We then used the full regres-sion equations to predict latent variable means forvarious steps on the conflict scale, with all controlvariables set at their means. These values were thendivided by the standard deviation of the latent var-iable so the results could be represented as z scores.These results are presented in Figure 2. Althoughconflict scores ranged from 1 to 12, only a fewfamilies (1%) had scores greater than 9, so the fig-ure is extended to 9 only.

Consider the results for psychological well-be-ing. When marital conflict was low, marital dis-solution was associated with a relatively low levelof well-being. For example, with a marital conflictscore of 3, offspring from divorced families hada predicted score that was .71 of a standard de-viation lower than that of offspring from non-di-vorced families. Over one fifth (22%) of offspringof divorce had parents with conflict scores of 3 orlower. When conflict was relatively high, however,marital dissolution was associated with a relative-ly high level of well-being. For example, with amarital conflict score of 7, offspring from di-vorced families had a mean well-being score thatwas .49 of a standard deviation higher than thatof offspring from nondivorced families. Amongoffspring from divorced families, 25% had parentswith conflict scores of 7 or higher.

The median conflict score for offspring in di-vorced families was 5.0. At this score, offspringfrom divorced families scored .11 standard devi-ations below offspring from nondivorced families.This result indicates that at the typical level ofmarital conflict that existed prior to divorce, mar-ital disruption was associated with slightly lowerpsychological well-being for offspring. The re-

Page 10: Parental Predivorce Relations and Offspring Postdivorce Well-Being

206 Journal of Marriage and Family

FIGURE 2. OFFSPRING OUTCOMES AS A FUNCTION OF THE INTERACTION BETWEEN PARENTAL MARITAL CONFLICT AND

PARENTAL DIVORCE

gression lines crossed above a conflict score of5.3, suggesting that this is the point at which di-vorce became beneficial rather than detrimental tochildren. In our study, 55% of children from di-vorced families were below that point, and 45%were above it. The results for kin support, friendsupport, and intimate relations revealed a similarpattern. In general, the regression lines crossedclose to a conflict score of 5—the median scorefor couples who later divorced.

To further understand the process by which di-vorce among low-conflict parent influences off-spring well-being, we undertook analysis to see ifone or more of the outcome variables mediatedthe impact of the others. Our analysis (not shown)revealed that although psychological well-being isa primary outcome, it appears that the various out-comes are reciprocally related and reinforce oneanother.

Events Following Divorce

To examine the role of events following divorce,the analysis was limited to the 85 offspring who

had experienced a divorce. Because of its pivotalrole, psychological well-being was selected as thedependent variable for analysis and was regressedon parents’ marital conflict, along with parents’education and offspring’s race, age, and gender.Parental conflict had a statistically significant re-lationship with well-being (b 5 .25, p 5 .05). Inother words, among offspring from divorced fam-ilies, those who experienced the most conflict be-fore marital disruption had the highest levels ofpsychological well-being.

We then entered four additional variables—postdivorce conflict between parents, moving,changing schools, and the number of subsequentmarital transitions among parents—singly andthen collectively into the regression equation.None of these variables was significantly associ-ated with psychological well-being. In addition,no variable reduced the predivorce marital conflictcoefficient by more than a trivial amount. With allof the variables entered at the same time, the betacoefficient for predivorce conflict was .26 (p 5.07). Interestingly, postdivorce conflict was cor-

Page 11: Parental Predivorce Relations and Offspring Postdivorce Well-Being

207Parental Predivorce Relations

related negatively with predivorce conflict (r 52.11), although the coefficient did not attain sig-nificance (p . .10). This indicates that predivorceconflict between parents was not a good predictorof postdivorce conflict. We also considered wheth-er children’s age at the time of divorce made adifference, but this variable was not correlatedsignificantly with psychological well-being orwith predivorce conflict. Overall, we found no ev-idence that events following the divorce were re-lated either to predivorce conflict or to children’soutcomes, and correspondingly, no evidence thatthese variables explained the association betweenpredivorce conflict and psychological well-beingamong children from divorced families.

STUDY 2

The idea that parents who seldom disagree or fightmight elect to end their marriages through divorceappears to be incongruous, yet our analyses sug-gest that this is not an uncommon event. In thissecond study, we explore a range of attributes thatmay provide insight into why individuals in rela-tively low-conflict relationships divorce.

To guide our analysis, we draw on conceptualperspectives that emphasize the importance of re-lationship commitment. Levinger (1976), workingfrom an exchange theory perspective, argued thatthe likelihood of marital dissolution increases tothe extent that people receive few rewards fromthe relationship, face few barriers to ending therelationship, and perceive good alternatives to therelationship. According to this perspective, indi-viduals may be moderately satisfied with a currentrelationship, but if they perceive weak barriers toleaving the relationship and if they believe thatsuperior alternatives to the relationship exist, thentheir commitment to the relationship will be low.From a somewhat different perspective, Johnson,Caughlin, and Huston (1999) distinguished be-tween three forms of commitment. Personal com-mitment exists when one wants the relationship tocontinue, moral commitment exists when onefeels obligated to remain in the relationship, andstructural commitment occurs when one feels con-strained to remain in the relationship. We assumethat individuals in low-conflict marriages that staytogether are not exceedingly different from indi-viduals in low-conflict marriages that end in di-vorce in terms of personal commitment (or howrewarding they find the relationship to be). In-stead, we assume that these two groups differ pri-

marily in their levels of moral and structural com-mitment.

We examine four general factors that may re-flect low moral and structural commitment andhence increase the risk of divorce among low-con-flict couples: weak community integration, non-traditional attitudes, risk-taking, and economic in-dependence. With regard to communityintegration, individuals who own their ownhomes, have lived in their communities for a longtime, have many friends, and are affiliated withreligious organizations are embedded within net-works of others who may disapprove of divorceand encourage couples to resolve their maritalproblems (Booth, Edwards, & Johnson, 1991).Couples who are not well-integrated into localcommunities and networks lack these structuralbarriers to divorce (Johnson et al., 1999). In ad-dition, owning a home creates a shared investmentthat is difficult to divide, which may reinforcestructural commitment (Booth, Johnson, White, &Edwards, 1986).

People with accepting attitudes toward divorce,by definition, do not perceive divorce as a ethicallapse and hence may have lower moral commit-ment to their marriages (Johnson et al., 1999).Furthermore, because organized religion is a pro-family organization, church attendance is likely tobe an indicator of traditional attitudes and hencemoral commitment to marriage (Johnson et al.,1999). Premarital cohabitation also may serve asan indicator of a lack of support for traditionalmarriage. In addition, individuals who divorce andenter a second or higher order marriage are likelyto have more favorable attitudes toward maritaldissolution than are individuals in first marriages(Thornton, 1985). People in higher order marriag-es also were successful in finding a new spouseafter their previous marriages ended, which maymake alternatives to the present marriage seemlike real possibilities. People who experienced pa-rental divorce as children also may be less tradi-tional because they have seen their parents modelbehavior that is antithetical to the norm of life-long marriage (Amato & Booth, 1991a). Peoplewho are prone to risk taking may be less com-mitted to their marriages. Risk takers, as opposedto those who are risk averse, may view barriers toleaving the relationship as less problematic andview alternatives to the current relationship asmore desirable. In addition, low marital durationmay be a hallmark of low-conflict marriages thatdissolve. That is, individuals with weak moral andstructural commitments to their marriages may

Page 12: Parental Predivorce Relations and Offspring Postdivorce Well-Being

208 Journal of Marriage and Family

initiate divorce at the first signs of incompatibility,whereas individuals with strong moral and struc-tural commitments may wait a long time beforeending a troubled marriage. A shorter marital du-ration also may indicate a willingness to ‘‘rushinto new marriages,’’ which may indicate a per-sonality trait that reflects a willingness to takerisks.

Finally, we assume that structural commitmentto the relationship is weaker when wives are em-ployed, because the wife’s employment makes iteasier for either spouse to leave the marriage. Fur-thermore, employed wives encounter a widerrange of alternative partners than do wives whowork as full-time homemakers.

In the study presented here, we compare low-conflict marriages that end in divorce with (a)low-conflict marriages that remain together, and(b) high-conflict marriages that end in divorce.(For the sake of logical completeness, we also in-clude high-conflict marriages that remain togeth-er.) We assume that divorces among high-conflictcouples occur for an obvious reason, that is, be-cause of the severe level of discord in the rela-tionship. Because these relationships (which in-volve fundamental disagreements, frequentquarrels, and a high likelihood of violence) arerelatively unrewarding, spouses may have weakpersonal commitment to the marriage. In contrast,we assume that divorces among low-conflict cou-ples occur primarily because levels of moral andstructural commitment are low. Consequently,low-conflict couples who divorce should scorelower than couples in group (a) as well as group(b) on indicators of moral and structural commit-ment.

Sample and Measures

For this analysis, we examined all parents in thesample, regardless of whether they had a focalchild in Study 1. We relied on the 1980 interviewfor data on all variables (including conflict) exceptwhether respondents divorced (1983–1997), thenumber of moves (available in 1983), and the risk-taking personality variable (available in 1988).The parents were divided into four groups: low-conflict marriages that did not end in divorce (n5 730), high-conflict marriages that did not endin divorce (n 5 450), high-conflict marriage thatended in divorce (n 5 115), and low-conflict mar-riage that ended in divorce (n 5 100). To formthe low- and high-conflict groups, we used the

median conflict score of marriages that ended indivorce.

The attributes were measured in the followingway. Home ownership was coded 1 if respondentsowned or were buying their home and 0 other-wise. In 1983, respondents were asked how manytimes they had moved since the fall of 1980. Themaximum number of moves was three. To avoida skewness problem, the number of moves wastruncated at two. The number of friends was basedon a question asking about the number of friends(not relatives) to whom respondents felt close.Church attendance was tapped with the questions,‘‘How often do you and your (husband/wife) at-tend church together? Weekly or more, once amonth or more but less than weekly, once a yearor more but less than monthly, or less than oncea year?’’ The item was coded so that higher scoresindicated greater participation.

A scale was created to measure positive atti-tudes toward divorce by calculating the mean offour attitude items. Sample items included ‘‘Cou-ples are able to get divorced too easily today’’ and‘‘The personal happiness of an individual is moreimportant than putting up with a bad marriage.’’Response options (1 5 disagree strongly, 2 5 dis-agree, 3 5 agree, 4 5 agree strongly) were codedin the direction of positive, accepting attitudes to-ward divorce. The alpha reliability for this scalewas .58. Premarital cohabitation was gleaned fromthe question, ‘‘Did you live with your (husband/wife) before you got married?’’ (1 5 yes, 0 5no). Prior divorce was obtained from a maritalhistory (1 5 prior divorce, 0 5 no prior divorce).With regard to parental divorce, respondents weregiven a score of 1 if their parents had divorcedand 0 if they had not.

A scale was devised to measure the extent towhich individuals were willing to make risky de-cisions in seven areas. Sample items included, ‘‘Ifyou were offered a job that you know would pro-vide a modest but adequate income for the rest ofyour life and, at the same time, a job that paidmuch more but its long-term prospects were stillunknown, which would you take?’’ and ‘‘Imagineyou were unmarried and dating two people whoindicated they were interested in marrying you.One you found exciting to be with and with whomyou always had a great time, but you were uncer-tain how you would get along after several yearsof marriage. The other was not nearly as exciting,but you were fairly certain you would get alongfor many years. Which would you choose?’’ Foreach item, the low-risk choice was scored 1 and

Page 13: Parental Predivorce Relations and Offspring Postdivorce Well-Being

209Parental Predivorce Relations

TABLE 3. MAXIMUM LIKELIHOOD ESTIMATES OF BEING IN A RELATIONSHIP OTHER THAN A LOW-CONFLICT MARRIAGE

THAT ENDED IN DIVORCE

Variable

Low-Conflict Marriage,No Divorce

B Exp B

High-Conflict Marriage,No Divorce

B Exp B

High Conflict MarriageThat Ended in Divorce

B Exp B

Home ownershipNumber of changes in residenceNumber of friendsChurch attendanceFavorable divorce attitudesPremarital cohabitation

.72***21.02***

.08***

.36***2.61**2.77**

(2.05)(0.36)(1.08)(1.43)(0.54)(0.46)

.062.02

.012.18

.33

.41

(1.06)(0.98)(1.01)(0.83)(1.39)(1.50)

.55*2.96***

.02

.20*2.332.26

(1.73)(0.38)(1.02)(1.22)(0.71)(0.77)

Prior divorceParents divorcedRisk-takerYears marriedWife in labor force

21.05***.22

21.61***.06**

2.83***

(0.34)(1.24)(0.20)(1.06)(0.43)

2.271.23***

2.562.012.60**

(0.76)(3.42)(0.57)(0.99)(0.54)

2.92***.65*

21.36***.03**

2.65***

(0.39)(1.91)(0.25)(0.97)(0.52)

Note: The figures in parentheses are the odds of being in relationships other than the excluded category of interest. Eachvariable is represented by a separate equation.

*p , .06. **p , .05. ***p , .01.

the high-risk choice was scored 2. The individu-al’s score was the mean of the items, and the scalehad an alpha coefficient of .64.

Finally, the wives’ labor force participationwas scored 1 if employed and 0 if not.

Multinomial logistic regression analysis wasused to compare the category of low-conflict mar-riages that end in divorce (the omitted category)with the other categories on each of the parents’attributes. A separate analysis was carried out foreach variable, and the results are shown in Table3. The most important comparisons involve theomitted group with the groups represented in col-umns 1 and 3. Our group of interest (low-conflictdivorce) is similar to the low-conflict marriagegroup (column 1) because of the low level of con-flict but differs because of the divorce. Corre-spondingly, our group of interest is similar to thehigh-conflict divorced group (column 3) becauseof the divorce but differs because of the low levelof conflict. The low-conflict divorced group (col-umn 2) has little in common with the low-conflictdivorced group, but it is included for the sake ofcompleteness.

RESULTS

Low-conflict parents who ended their marriagesthrough divorce were less involved in the com-munity than were the parents represented in col-umns 1 and 3. The former group were less likelyto own a home, moved more often, had fewerfriends (compared with the low-conflict marriagegroup), and attended church less often. Similarly,

the omitted group had more favorable attitudes to-ward divorce and were more likely to cohabit be-fore marriage (compared with the low-conflictmarriage group). Although the differences withrespect to attitudes and cohabitation were not sta-tistically significant in column 3 (compared tohigh-conflict marriages that ended in divorce), thecoefficients were in the expected directions. In ad-dition, the low-conflict divorce group was morelikely than the other groups to have been marriedbefore their 1980 marriage. The low-conflict di-vorced group, however, was not significantly morelikely to have experienced a parental divorce. In-deed, individuals in this group were the least like-ly of all to have grown up in a divorced family,a finding contrary to our expectations.

The low-conflict divorced group scored signif-icantly higher than other groups in terms of risk-taking. Furthermore, they had been married a rel-atively short period of time, suggesting a tendencyto end marriages quickly. Finally, wives in thelow-conflict divorced group were significantlymore likely than were wives in the other groupsto be in the paid labor force.

When all of the variables were entered into asingle equation, a few were no longer significant(results not shown). These shared variance withother variables, however, reflecting similar attri-butes. For example, home ownership was not sig-nificant with the number of moves in the equation,and premarital cohabitation and attitudes towarddivorce were not significant with church atten-dance in the equation. We included the table withthe separate equations so that the reader could

Page 14: Parental Predivorce Relations and Offspring Postdivorce Well-Being

210 Journal of Marriage and Family

gain an appreciation of the range of factors char-acterizing low-conflict divorce parents.

In summary, parents in low-conflict marriagesthat end in divorce appeared to lack strong ties tohome, community, and friends, thus reflectingweak structural barriers to divorce. They also mayhave found dissolving a marriage to be easy be-cause they did not have assets that were difficultto divide (a home) and had independent sourcesof income. In terms of attitudes, these individualsheld favorable attitudes toward divorce and weremore likely to be in a second or higher order mar-riage. Moreover, these individuals had a predis-position to taking risks and tended to leave mar-riages relatively quickly. All of these factors,taken together, suggest that individuals in low-conflict marriages who divorced tended to haveweak moral and structural commitments to mar-riage. These individuals also were relatively un-likely to have experienced parental divorce aschildren. Having grown up with continuouslymarried parents, these individuals may have beensomewhat naive about the consequences of divor-ce for children. Indeed, holding relatively favor-able attitudes toward divorce in general, these in-dividuals may have underestimated the extent towhich divorce was a stressful event in their chil-dren’s lives. Perhaps the willingness of these in-dividuals to take risks by leaving a low-conflictmarriage reflects a willingness to take risks withtheir children’s lives as well.

CONCLUSIONS

We can no longer think of the effects of divorceon children in monolithic terms. The study pre-sented here, combined with earlier work (Amato& Booth, 1997; Amato et al., 1995; Jekielek,1998) suggests that marital conflict influences theimpact of divorce in major ways. The dissolutionof low-conflict marriages appears to have a strongnegative influence on offspring, whereas divorceamong high-conflict couples appears to have a rel-atively benign or even beneficial effect. Further-more, the dissolution of low-conflict marriages isassociated with a wide range of offspring out-comes, including the quality of intimate hetero-sexual relations, the amount of social supportfrom friends and relatives, and general psycholog-ical well-being. Our results differed somewhat forrelations with parents, however. For this outcome,parental divorce appeared to lower the quality ofrelations with parents regardless of the level ofconflict preceding the divorce. Conflict was in-

dependently associated with parent-child relation-ships. In other words, marital conflict loweredchildren’s closeness to parents, and divorce low-ered it even further. These associations were sig-nificant for both parents but were stronger for fa-thers than for mothers. In contrast, we found noevidence that either parental divorce or maritalconflict was related to offspring’s educational at-tainment. Nevertheless, the weight of the evidencesuggests that most of the effects of divorce aremoderated by the level of marital conflict that pre-cedes divorce.

Presumably, escape from a high-conflict mar-riage benefits children because it removes themfrom an aversive, stressful home environment. Incontrast, a divorce that is not preceded by a pro-longed period of overt discord may represent anunexpected, unwelcome, and uncontrollableevent, an event that children are likely to experi-ence as stressful (Amato, 1987; Bussell, 1995;Duran-Aydintug, 1997; Grych & Finchan, 1990;Healy et al., 1993; Jacobson, 1978; Kim et al.,1997). Divorce under these circumstances repre-sents a major change in what may otherwise havebeen a secure and (from a child’s perspective)seemingly well-functioning family. Marital dis-ruption entails one parent leaving the home (andan uncertain future with regard to this parent), acustodial parent often overwhelmed with the de-mands of single parenthood, and a decline in stan-dard of living. Daily interaction with adults whocare for them (and care for each other) is desirablebecause it helps children negotiate significant andfrequent changes in their own lives. Cognitive andphysical developmental changes, starting school,changing schools as the child progresses from onelevel to next, and the increased relevance of peersand romantic involvements represent an intensiveand extensive set of changes over a period of 18years. Stable relations with two caring adults cando much to help children adjust to the immensechanges in their lives. The loss of these resourcesis a distinct disadvantage to children and may re-sult in stress overload. By contrast, the stress gen-erated by parents who are continuously fighting islikely to exceed the stress associated with livingin a single-parent household. In general, our re-sults suggest that divorce may be beneficial orharmful to children, depending on whether it re-duces or increases the amount of stress to whichchildren are exposed

We searched for other mechanisms throughwhich low-conflict marriages that end in divorcemight have adverse effects. We explored the idea

Page 15: Parental Predivorce Relations and Offspring Postdivorce Well-Being

211Parental Predivorce Relations

that parental conflict over custody, visitation, childsupport, and the division of property might be par-ticularly damaging to offspring who have littleprior exposure to parent conflict. Postdivorce con-flict (which was negatively but not significantlyassociated with predivorce conflict) was not as-sociated with offspring well-being, however. Wealso found no evidence that changes in residenceor schools, or the number of subsequent familytransitions, were related to offspring’s well-being.Consequently, we believe that an explanationbased on children’s perceptions of the divorce—the extent to which it is unexpected, unwelcome,and unpredictable—accounts for the differentialimpact of divorces preceded by high and low lev-els of marital discord.

We also examined the attributes of people whodissolve a low-conflict marriage to ascertain thereasons for these divorces. This line of inquirywas productive. Low-conflict couples who divorcediffer in a number of important ways from high-conflict couples who divorce and low-conflictcouples who remain married. They are not inte-grated into the major institutions in their com-munity. They are not home owners, move fre-quently, have few friends, and attend churchinfrequently. Low-conflict parents who divorcehave few impediments to marital dissolution. Bothparents are employed, which means that divorceis less costly to both parties. The couple does notown a house, an asset difficult to divide in theevent of marital dissolution. Moreover, their pre-vious divorce experience makes the prospect ofanother divorce less foreboding. These individualsalso hold positive attitudes toward divorce, atti-tudes that weaken ethical barriers to ending a mar-riage. Low-conflict parents who divorce also tendto be risk takers. This personality characteristicnot only predisposes them to seek out other ro-mantic partners (which increases the likelihood ofdissolving the marriage), but also causes them tofeel less inhibited about ending the current mar-riage. Finally, these individuals are less likely tohave experienced parental divorce as children.Therefore, they may have little understanding ofwhat effect their own divorce may have on theirchildren. In general, this constellation of charac-teristics is likely to reflect weak moral and struc-tural commitments to marriage (Johnson et al.,1999).

The study reported here has achieved twothings. First, through the application of rigorousmethods to longitudinal data, we have providedstrong evidence that children from high-conflict

families are better off, in the long run, if theirparents divorce than if their parents remain mar-ried. On the other hand, children from low-conflictfamilies that end in divorce experience inordinateadversity, both psychologically and socially. Sec-ond, we have presented compelling evidence thatthe tendency to divorce among low-conflict par-ents is related to low involvement in communityinstitutions, few moral or economic impedimentsto divorce, a willingness to engage in risky be-havior, and a lack of knowledge about the impactof divorce on children—all factors that lower peo-ple’s moral and structural commitment to mar-riage. Curiously, our results suggest that divorceswith the greatest potential to harm children occurin marriages that have the greatest potential forreconciliation or for managing a divorce such thatit is less detrimental to their children. After all, amarriage characterized by mild disengagement hasmore on which it can build than does a marriagecharacterized by chronic hostility. Parent fromlow-conflict marriages, perhaps aided by educa-tional and therapeutic interventions, should beable to buffer the effects of divorce on their chil-dren. This might include preparing children forthe dissolution, as well as establishing plans andactivities that would minimize change and insta-bility in the child’s life.

NOTE

We are deeply indebted to Christopher Knoester for hisassistance with portions of the statistical analysis. Thisresearch was supported in part by Grant 5 RO1AG04146 from the National Institute on Aging and thePennsylvania State University Population Research In-stitute, with core support from the National Institute ofChild Health and Human Development Grant 1HD282663.

REFERENCES

Amato, P. R. (1987). Children’s reactions to parentalseparation and divorce: The views of children andcustodial mothers. Australian Journal of Social Is-sues, 22, 610–623.

Amato, P. R. (1993). Children’s adjustment to divorce:Theories, hypotheses, and empirical support. Journalof Marriage and the Family, 55, 23–38.

Amato P. R., & Booth, A. (1991a). Consequences ofdivorce for attitudes toward divorce. Journal of Fam-ily Issues, 12, 306–322.

Amato, P., & Booth, A. (1991b). Consequences of pa-rental divorce and marital unhappiness for adult well-being. Social Forces, 69, 895–914.

Amato, P., & Booth, A. (1997). A generation at risk:Growing up in a era of family upheaval. CambridgeMA: Harvard University Press.

Amato, P., Loomis, L., & Booth, A. (1995). Parental

Page 16: Parental Predivorce Relations and Offspring Postdivorce Well-Being

212 Journal of Marriage and Family

divorce, marital conflict, and offspring well-beingduring early adulthood. Social Forces, 73, 895–915.

Arbuckle, J. (1997). AMOS 3.5 for Windows. Chicago:Smallwaters.

Barber, B. L., & Eccles, J. S. (1992). Long-term influ-ence of divorce and single parenting on adolescentfamily- and work-related values, behavior, and aspi-rations. Psychological Bulletin, 111, 108–126.

Booth, A., & Amato, P. (1993). Divorce, residentialchange, and stress. Journal of Divorce, 18, 205–213.

Booth, A., Amato, P., Johnson, D., & Edwards, J.(1993). Marital instability over the life course: Meth-odology report for fourth wave. Lincoln, NE: Uni-versity of Nebraska Bureau of Sociological Research.

Booth, A., Amato, P., Johnson, D., & Edwards, J.(1998). Marital instability over the life course: Meth-odology report for fifth wave. Lincoln, NE: Univer-sity of Nebraska Bureau of Sociological Research.

Booth, A., Brinkerhoff, D., & White, L. (1984). Theimpact of parental divorce on courtship. Journal ofMarriage and the Family, 46, 85–94.

Booth, A., Edwards, J. N., & Johnson, D. R. (1991).Social integration and divorce. Social Forces, 40,207–224.

Booth, A., Johnson, D., White, L., & Edwards, J.(1986). Divorce and marital instability over the lifecourse. Journal of Family Issues, 7, 421–442.

Bussell, D. A. (1995). A pilot study of African Amer-ican children’s cognitive and emotional reactions toparental separation. Journal of Divorce and Remar-riage, 25, 1995.

Cherlin, A. J., Chase-Lansdale, P. L., & McRae, C.(1998). Effects of divorce on mental health through-out the life course. American Sociological Review,63, 239–249.

Cooney, T. M., & Kurtz, J. (1996). Mental health out-comes following recent parental divorce. Journal ofFamily Issues, 17, 495–513.

Cowan, P. A., Cowan, C. P., & Schulz, M. S. (1996).Thinking about risk and resilience in families. In E.M. Hetherington & E. A. Blechman (Eds.), Stress,coping, and resiliency in children and families (pp.1–38). Mahwah, NJ: Erlbaum.

Davies, P. T., & Cummings, E. M. (1994). Marital con-flict and child adjustment: An emotional security hy-pothesis. Psychological Bulletin, 116, 387–411.

Duran-Aydintug, C. (1997). Adult children of divorcerevisited: When they speak up. Journal of Divorceand Remarriage, 27, 71–83.

Emery, R. E. (1999). Marriage, divorce, and children’sadjustment. (2nd ed.). Thousand Oaks, CA: Sage.

Gately, D. W., & Schwebel, A. I. (1991). The challengemodel of children’s adjustment to parental divorce:Exploring favorable postdivorce outcomes in chil-dren. Journal of Family Psychology, 5, 60–81.

Grych, J., & Finchan, F. (1990). Marital conflict and

children’s adjustment: A cognitive-conceptual frame-work. Psychological Bulletin, 108, 267–290.

Healy, J. M., Steward, A. J., & Copeland, A. P. (1993).The role of self-blame in children’s adjustment to pa-rental separation. Personality and Social PsychologyBulletin, 19, 279–289.

Hill, R. (1949). Families under stress. New York: Harp-er and Brothers.

Jacobson, D. S. (1978). The impact of marital separa-tion/divorce on children: Parent-child communicationand child adjustment. Journal of Divorce, 2, 175–194.

Jekielek, S. (1998). Parental conflict, marital disruptionand children’s emotional well-being. Social Forces,76, 905–936.

Johnson, M. P., Caughlin, J. P., & Huston, T. L. (1999).The tripartite nature of marital commitment: Personal,moral, and structural reasons to stay married. Journalof Marriage and the Family, 61, 160–177.

Kim, L., Sandler, I. N., & Jenn-Yum, T. (1997). Locusof control as a stress moderator and mediator in chil-dren of divorce. Journal of Abnormal Child Psychol-ogy, 25, 145–155.

Langner, T. (1962). A twenty-two item screening scoreof psychiatric symptoms indicating impairment. Jour-nal of Health and Social Behavior, 3, 269–276.

Levinger, G. (1976). A socio-psychological perspectiveon marital dissolution. Journal of Social Issues, 52,21–47.

McCubbin, H. I., & Patterson, J. M. (1983). Stress andthe Family. In H. I. McCubbin & C. R. Figley (Eds.),Family stress, coping, and social support (pp. 5–25).Springfield, IL: Charles C. Thomas.

McLanahan, S., & Sandefur, G. (1994). Growing upwith a single parent: What hurts, what helps. Cam-bridge, MA: Harvard University Press.

Rosenberg, M. (1965). Society and adolescent self-im-age. Princeton, NJ: Princeton University Press.

Simons, R. L., and Associates. (1996). Understandingdifferences between divorced and intact families.Thousand Oaks, CA: Sage.

South, S. J., Crowder, K. D., & Trent, K. (1998). Chil-dren’s residential mobility and neighborhood environ-ment following parental divorce and remarriage. So-cial Forces, 77, 667–693.

Thoits, P. A. (1995). Stress, coping, and social supportprocesses: Where are we? What next? Journal ofHealth and Social Behavior [extra issue], 53–79.

Thornton, A. (1985). Changing attitudes toward sepa-ration and divorce: Causes and consequences. Amer-ican Journal of Sociology, 90, 856–872.

Wheaton, B. (1990). Life transitions, role histories, andmental health. American Sociological Review, 55,209–223.

Wu, L. (1996). Effects of family instability, income, andincome instability on risk of premarital birth. Amer-ican Sociological Review, 61, 386–406.