Parental Divorce and Adult Well-Being: A Meta-Analysis

Author(s): Paul R. Amato and Bruce Keith

Source: Journal of Marriage and Family, Vol. 53, No. 1 (Feb., 1991), pp. 43-58
Published by: National Council on Family Relations
Stable URL: http://www.jstor.org/stable/353132
Accessed: 20-04-2016 23:41 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/353132?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Wiley, National Council on Family Relations are collaborating with JSTOR to digitize, preserve and
extend access to Journal of Marriage and Family

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 PAUL R. AMATO AND BRUCE KEITH University of Nebraska-Lincoln

Parental Divorce and Adult Well-being:

A Meta-analysis

A meta-analysis was conducted of studies dealing mains of well-being. Some studies have focused
with the long-term consequences of parental on socioeconomic outcomes, such as educational
divorce for adult well-being. Effect sizes were attainment, earnings, and welfare dependency
calculated for 15 outcome variables across 37 (e.g., Hill, Augustyniak, and Ponza, 1987; Keith
studies involving over 81,000 individuals. Mean and Finlay, 1988; McLanahan, 1985, 1988). Other
effect sizes were significant and negative for all studies have examined marital quality, marital in-
outcomes; this indicated that adults who ex- stability, and divorce (e.g., Glenn and Kramer,
perienced parental divorce exhibited lower levels 1987; McLanahan and Bumpass, 1988; Pope and
of well-being than did adults whose parents were Mueller, 1976). Still others have dealt with
continuously married. The strongest estimated ef- psychological well-being (e.g., Amato, 1988;
fects occurred in the areas of one-parent family Glenn and Kramer, 1985; Kulka and Weingarten,
status, psychological adjustment, behavior/con- 1979) or psychiatric problems (e.g., Greer, 1966;
duct, and educational attainment. Effect sizes Roy, 1985; Sethi, 1964). Although inconsistencies
were significantly stronger in clinical studies than appear in this literature, a number of studies have
in studies based on community samples. In addi- found that adults from divorced families of origin
tion, effects sizes tended to be stronger for whites experience lower levels of well-being than do
than for blacks, stronger in earlier studies than in adults whose parents were continuously married.
more recent studies, and stronger in studies that Why would we expect parental divorce to have
did not use statistical controls than in studies that long-term consequences for quality of life in
did. adulthood? Three theoretical perspectives can be
brought to bear on this question; these models
focus on economic deprivation, family structure,
A number of studies have examined the conse- and psychological stress as mediating processes
quences of parental divorce for later well-being in (see McLanahan, 1985, 1988).
adulthood. The usual analytic strategy has been to First, divorce typically leads to a dramatic
compare adults who experienced parental divorce decline in the standard of living of mother-
as children with adults whose parents were con- custody families, often pushing them below the
tinuously married. In many sociological studies, poverty line (Duncan and Hoffman, 1985; Weitz-
statistical controls are used to take into account man, 1985). It is well known that economic
other family-of-origin characteristics associated deprivation presents obstacles to children's school
with divorce, such as parental education and oc- attainment. Most single mothers are unable to af-
cupational status. In clinical studies, matching of ford private lessons, educational toys and games,
subjects is sometimes used for the same purpose. books, home computers, and other commodities
This literature has dealt with several broad do- that facilitate children's academic success.
Moreover, many single-parent families live in
Department of Sociology, University of Nebraska-Lincoln, Lin- neighborhoods where school programs are poorly
coln, NE 68588-0324. financed. And adolescents in one-parent families

Journal of Marriage and the Family 53 (February 1991): 43-58 43

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 44 Journal of Marriage and the Family

may be pressured to leave school early to seek Finlay, 1988). Variations in results are to be ex-
employment. Low educational achievement is pected, given that studies differ in the types of
likely to be associated with low occupational at- outcomes examined, sampling methodology, the
tainment, unemployment, poverty, and welfare sex and racial composition of samples, sample
dependency later in life. These problems, in turn, size, the year sample data were collected, and the
may contribute to marital and psychological dif- use of control variables in analyses.
ficulties. At this point in time, a sufficient number of
Second, a focus on family structure suggests studies exists to make a meta-analysis of this
that living with one parent rather than two results literature useful. Meta-analysis is a technique for
in certain disadvantages. For example, it is well conducting a quantitative literature review (Glass,
known that divorce is associated with a decrease McGaw, and Smith, 1981; Hedges and Olkin,
in the quantity and quality of contact between 1985; Wolf, 1986). In meta-analysis, one or more
children and noncustodial fathers (Amato, 1987a, effect sizes are calculated for each study. Express-
1987b; Furstenberg and Nord, 1987; White, ing results in a common metric makes it possible
Brinkerhoff, and Booth, 1985). Custodial moth- to combine results across studies to reach general
ers are often in the labor force and, as a result, are conclusions. Effect sizes also can be related to
constrained in the number of hours they can study characteristics, such as the nature of the
devote to their children. For these reasons, many sample or the analysis used. This makes it possible
children in one-parent families experience a to test hypotheses about why some studies show
decrease in parental attention, help, and supervi- larger effects than others.
sion, and may therefore be less likely to do well in The purpose of this analysis is twofold. First,
school. Furthermore, the loss of parental role we seek to estimate the impact of parental divorce
models may result in the inadequate learning of on various forms of well-being across all available
social skills-such as cooperation and the resolv- studies. Outcomes in the domains of psychologi-
ing of disputes through compromise-that con- cal well-being, family well-being, socioeconomic
tribute to successful performance in adult occupa- well-being, and physical health are examined. Sec-
tional and marital roles. And a decrease in paren- ond, the magnitude of effect sizes may vary sys-
tal contact during childhood may result in tematically from study to study. Accordingly, we
restricted kin networks later in life, with conse- examine associations between effect sizes and the
quences for social support and psychological well- following variables: the population sampled
being. (general community vs. clinical), sex of respon-
Finally, divorce is likely to be a stressful ex- dents, ethnicity of respondents (black vs. white),
perience for children. It is usually preceded, and sample size, the year data were collected, and
often followed, by a period of interparental con- whether or not statistical controls were used.
flict. Divorce is often associated with other dif-
ficult events for children, such as moving, chang-
ing schools, losing contact with grandparents, and METHOD

parental remarriage (Wolchik, Sandler, Braver,

Selection of Studies
and Fogas, 1985). These stressful events may
disrupt children's school attainment, social rela- Studies were located by searching manually
tionships, and personality development-all of through yearly volumes of the Social Science In-
which may have long-lasting implications. dex and Psychological Abstracts. Computerized
Although sound theoretical reasons exist for data bases were also utilized. Additional studies
presuming that parental divorce has a detrimental were found in the reference lists of published ar-
impact on children's lives, some studies fail to ticles. Several unpublished studies were obtained
support this notion. For example, Nock (1982) through personal contacts of the authors.
found that family disruption during childhood To be included in the meta-analysis, studies
was not associated with levels of psychological had to meet four criteria. First, the study had to
well-being in adulthood. Furthermore, some contain a sample of individuals from divorced (or
studies find associations for some outcomes but permanently separated) family backgrounds and a
not for others (e.g., Amato, 1988) or for some comparison sample of individuals whose biologi-
subsamples but not for others (e.g., Keith and cal parents had been continuously married. This

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 Parental Divorce and Adult Well-being 45

resulted in the exclusion of many studies that con- tact group in well-being; a positive sign indicated
tained a sample of individuals from one-parent the opposite.
family backgrounds but did not distinguish be- Significance tests require that effect sizes be in-
tween different causes of one-parent family dependent of one another. Consequently, the
status, such as parental divorce, the death of a mean effect size was calculated for all "indepen-
parent, or out-of-wedlock birth. Second, at least dent samples" in a study. If a study reported
one quantitative measure of well-being was re- analyses based on two different data sets, each
quired; this measure, however, could be a dichot- was considered an independent sample. Similarly,
omy, such as "graduated" versus "did not if a study reported data separately for males and
graduate from high school." Third, information females or for blacks and whites, these were con-
had to be presented to allow the calculation of at sidered to be independent samples. For each in-
least one effect size; usually this involved means dependent sample in a study, multiple outcomes
and standard deviations, t values, F ratios, cor- were often reported. If outcomes were in the same
relation coefficients, or percentage differences. category of well-being, the mean effect size was
Fourth, respondents had to be 18 years of age or calculated. However, if outcomes were in dif-
older; studies based entirely on college students ferent categories, they were counted as separate
were excluded. A total of 37 studies that met these observations. In other words, effects sizes were
criteria were located. These studies involved aggregated within categories of outcomes within
81,678 individuals (excluding studies based on the independent samples. (Occasionally, two studies
same samples). were based on the same data set; in these cases,
the effect size was averaged across the two studies
and counted only once.) Aggregation reduced the
Calculation of Effect Sizes number of effect sizes from 593 to 272.
Procedures described by Hedges and Olkin (1985) The effect sizes in this literature can be inter-
were used to calculate effect sizes. The most direct preted in three ways, depending on the nature of
method of calculation required subtracting the the statistical analysis conducted in the study.
mean score for the intact group from the mean Some articles reported unadjusted, zero-order dif-
score for the divorced group and dividing this dif- ferences between groups. Effect sizes based on
ference by the within-group standard deviation. these comparisons can be said to reflect the total
The effect size, therefore, is the difference be- covariation between parental divorce and adult
tween groups expressed in standard deviation outcomes. Some of this covariation may be
units. When means and standard deviations were causal, and some may be spurious. Other studies
not available, formulas given by Hedges and used multivariate methods in which residualized
Olkin (1985) were used to transform t values, F differences between groups were reported. Many
ratios, and correlation coefficients into effect of these studies controlled for predivorce family-
sizes. Percentage differences were transformed in- of-origin and respondent characteristics, such as
to effect sizes through the probit procedure parental education, parental occupational status,
described by Glass, McGaw, and Smith (1981). family size, ethnicity, sex of respondent, and age
When only probability values were reported, a of respondent (or year of birth). Because these
table was used to estimate t. When a result was variables precede parental divorce in time, they
said to be significant, and when no other details could affect both parental divorce and later adult
were provided, a p value of .05 was assumed. outcomes. Controlling for these variables resulted
When a result was said simply to be nonsignifi- in effect sizes that reflect the total estimated effect
cant, a p value of .5 was assumed. Because these of parental divorce, including both direct and in-
methods produce slight overestimates of the effect direct effects. Other multivariate analyses con-
size when the sample size is small, the correction trolled for postdivorce characteristics of the
proposed by Hedges and Olkin (1985) was used to respondent. For example, a study may have esti-
calculate "d," the unbiased estimator. mated the effect of parental divorce on current
Signs were affixed to effect sizes to reflect earnings while controlling for years of education.
comparative well-being. A negative sign indicated Effect sizes based on these analyses reflect the
that the divorced group scored lower than the in- direct effect of parental divorce; they do not

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 46 Journal of Marriage and the Family

reflect indirect effects through control variables. ethnicity (black, white, or mixed), sex (male,
Because these effect sizes have different inter- female, or mixed), size of sample, and the year
pretations, when studies reported both zero-order data were collected (if data were collected over a
and partial associations, we calculated effect sizes period of years, the middle year was coded).
for each. This procedure is consistent with Hedges The coding of most variables was straightfor-
and Becker (1986), who suggested that researchers ward. The classification of outcomes into the 15
calculate effect sizes in whatever metric is categories, however, left some room for disagree-
available (that is, unadjusted or adjusted scores) ment. Consequently, a reliability check was car-
and compare them later to see if they differ. ried out in which each author independently
An example may make this procedure clearer. classified 100 outcomes. The resulting Cohen's
Imagine a study that included three measures of kappa value was .84 (p < .001), indicating a
psychological well-being and two measures of moderately high level of agreement. A check was
marital quality. Furthermore, assume that this also made on the reliability of effect size calcula-
study reported both zero-order and multivariate tions. Each author independently calculated all
results separately for males and females. For pres- effect sizes from five articles, representing a total
ent purposes, the study would yield a total of of 27 separate effect sizes. The product-moment
eight effect sizes: four for males and four for correlation between the two sets of calculations
females, two reflecting psychological well-being was .98 (p < .001).
(the mean of the three effect sizes, calculated
from both unadjusted and adjusted differences)
and two reflecting marital quality (the mean of the RESULTS

two effect sizes, calculated from both unadjusted

Overall Effect Sizes
and adjusted differences).
As a preliminary step, effect sizes across all out-
comes were examined. Figure 1 presents stem-
Variables
and-leaf plots for effect sizes based on three types
Outcomes were coded into the following 15 of analyses: those without statistical controls or
categories: (1) psychological well-being (emo- matching of subjects (n = 130), those that con-
tional adjustment, depression, anxiety, life trolled for predivorce variables (n = 108), and
satisfaction); (2) behavior/conduct (criminal those that controlled for postdivorce variables (n
behavior, drug use, alcoholism, suicide, teenage = 34). (Stem-and-leaf plots contain the first digit
pregnancy, teenage marriage); (3) use of mental of the effect size in the first column. The second
health services; (4) self-concept (self-esteem, self- digit is represented in subsequent columns-one
efficacy, sense of power, internal locus of for each case. As such, they reveal the shape of
control); (5) social well-being (number of friends, the frequency distribution.) Several trends emerge
social participation, social support, contact with from this preliminary view.
parents and extended family); (6) marital quality For all three distributions, the majority of ef-
(marital satisfaction, marital disagreements, fect sizes were negative (84%, 900/%, and 91%,
marital instability); (7) separation or divorce; (8) respectively). This indicates that adult children of
one-parent family status; (9) quality of relations divorce generally exhibited a lower level of well-
with one's children; (10) quality of general family being when compared with other adults. How-
relations (overall ratings of family life); (11) ever, the average effect size was relatively weak.
educational attainment (high school graduation, In fact, the median values for the three distribu-
years of education); (12) occupational quality (oc- tions were remarkably similar: -.158, -.154, and
cupational prestige, job autonomy, job satisfac- -.172, respectively. In other words, across all
tion); (13) material quality of life (income, assets studies and all outcome measures, adults from
held, housing quality, welfare dependency, divorced families of origin scored about one-sixth
perceived economic strain); (14) physical health of a standard deviation below adults from con-
(chronic problems, disability); and (15) other. tinuously intact families of origin. The similarity
To compare effect sizes across studies, the of the three distributions also suggests that dif-
following study characteristics were coded: nature ferences between groups changed relatively little
of sample (clinical vs. general community), when control variables were used in analyses.

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
Parental Divorce and Adult Well-being 47

FIGURE 1. STEM-AND-LEAF PLOTS FOR THREE TYPES OF EFFECT SIZES:

WITHOUT CONTROL VARIABLES, WITH PREDIVORCE CONTROL
VARIABLES, AND WITH POSTDIVORCE CONTROL VARIABLES

Effect Sizes -1.0

without Controls - .9
- .8 59
- .7 6
- .6 778
-.5
- .4 1223788
- .3 00444566777799
- .2 00000011122222578899999
- .1 00111222223333445555556666677788899999
- .0 113344445555677777889
.0 000001344667888
.1 0012
.2 8
.3
.4
.5
.6 5

Effect Sizes -1.0

with Controls - .9 02
for Predivorce - .8
Variables - .7 8
- .6 4
- .5 118
- .4 23449
- .3 133455578
- .2 00011222334666778889
- .1 1222233345556666677788999
- .0 1112222344444555566667777999999
.0 000011469
.1 2
.2
.3
.4 2

Effect Sizes -1.0 0

with Controls - .9 7
for Postdivorce - .8 4
Variables - .7
- .6 9
- .5 6
- .4 39
- .3 0125
- .2 2379
- .1 1222346689
- .0 577889
.0
.1 47
.2 8
.3

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 48 Journal of Marriage and the Family

FIGURE 2. STEM-AND-LEAF PLOT FOR EFFECT SIZES USED IN ANALYSIS

-1.0 1
- .9 027
- .8 49
- .7 68
- .6 489
- .5 1156
- .4 223344589
- .3 0133444555677789
- .2 000001122222334556667788899999
- .1 0111122222233333344455555666666677778899999
- .0 11112222334444455555566667777778999999
.0 0000000111469
.1 122
.2
.3
.4 2
.5
.6 5

However, because the distributions represent ef- distribution is presented in Figure 2. Overall, 89%
fect sizes based on different outcomes, this con- of effect sizes were negative. The median effect
clusion is premature. The manner in which results size was -.164 and the mean was -.205.
were affected by the presence or absence of con- Significance tests were carried out on individual
trol variables is directly examined below. effect sizes (Hedges and Becker, 1986); this re-
For subsequent analyses and significance test- vealed that 45%0 were significant and negative,
ing, it was necessary to include only one type of whereas only 1% were significant and positive
effect size from each study for each type of out- (p < .05, two-tailed). These results clearly show
come. It was decided to select effect sizes in the that the estimated effects of parental divorce were
following order, depending on which were considerably more likely to be negative than posi-
available: (a) effect sizes based on analyses in tive. However, more than half (540%o) of all effect
which controls were used for predivorce family- sizes were not significantly different from zero.
of-origin characteristics, (b) effect sizes based on
zero-order comparisons, and (c) effect sizes based
Mean Effect Sizes within
on analyses in which controls were used for post-
Domains of Well-being
divorce characteristics. Effect sizes based on pre-
divorce controls were preferred because they The next step in the analysis involved calculating
represent the total estimated effect of di- the mean effect sizes for the various domains of
vorce-the most relevant figure for the present well-being. This involved weighting the effect
purpose. Effect sizes based on postdivorce control sizes by their corresponding sample sizes, using
variables were avoided when possible for two procedures outlined by Hedges and Olkin (1985).
reasons. First, because the particular control These results are presented in Table 1.
variables varied from study to study, the effect The mean effect sizes, found in column 2, were
sizes are not comparable. Second, because they consistently negative, and all but two attained
represent only the direct effects of parental significance. The strongest effects were for
divorce, these effect sizes provide an underesti- psychological adjustment and one-parent family
mate of the total effect. (As a check, later analyses status. These represented differences between
were replicated with the order of effect sizes groups of about one-third of a standard devia-
changed, and the results were virtually identical to tion. Other noteworthy effects occurred for
those reported below.) behavior/ conduct, use of mental health services,
This selection process resulted in a total of 169 separation or divorce, educational attainment,
separate effect sizes across all outcomes. This and material quality of life. These differences

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 Parental Divorce and Adult Well-being 49

TABLE 1. DATA ON MEAN EFFECT SIZES FOR COMPARISONS OF ADULT SUBJECTS FROM
DIVORCED AND INTACT FAMILY BACKGROUNDS ACROSS 15 DOMAINS OF WELL-BEING

X Effect Fail-safe
Domain of Well-being n Size N H
Psychological adjustment 23 -.32*** 1,594 388.51***
Behavior / conduct 9 -.28*** 418 37.36***
Self-concept 3 -.09 2.74
Use of mental health services 6 -.21*** 26 5.41
Social relations 8 -.06** 25 15.67*
Parent-child relations 13 -.02 21.71*
Marital quality 16 -. 13*** 141 12.92
Separation or divorce 24 -.22*** 1,995 113.99***
General family quality 5 -.12"** 29 12.36*
One-parent family status 6 -.36*** 313 84.91***
Educational attainment 18 -.28*** 1,263 38.03**
Material quality of life 10 -.18*** 112 22.74**
Occupational quality 9 -.11*** 36 21.88**
Physical health 11 -.10*** 76 13.74
Other 6 -.10*** 48 36.35***

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity.
*p < .05. **p < .01. ***p < .001.

were in the order of one-fifth to one-fourth of a category is greater than expected by chance.
standard deviation. Another way of interpreting the statistic is to view
Our literature review almost certainly failed to it as a test of study-by-effect-size interaction. In
locate some relevant studies, particularly those other words, it addresses the question, Are the
that are unpublished. Because a bias exists to estimated effects of divorce significantly stronger
publish only those studies that report significant in some studies than in others? If effect sizes are
results, it is likely that the effect sizes in Table 1 not homogeneous, it is advisable to examine study
are overestimates. To address this problem, we characteristics to see if variation in effect sizes
calculated "fail-safe N" values for each mean ef- between studies can be accounted for. Table 1
fect size (Rosenthal, 1979). This refers to the reveals a number of significant H values; conse-
number of additional studies that fail to reject the quently, the next step in the analysis involved
null hypothesis that would need to be included in comparing effect sizes across studies.
our meta-analysis to decrease the overall effect
size to a nonsignificant level (that is, p > .05).
Some of the fail-safe N values are small. For
Between-Study Comparisons:
Univariate Analysis
example, for use of mental health services, only 26
additional studies finding no significant dif- The first study characteristic we examined was
ference would need to be included to make the whether the sample was drawn from a clinical or a
result nonsignificant. It is possible that 26 addi- nonclinical population. In fact, the clinical and
tional studies with null findings exist, so our faith noniclinical studies differed in many ways. Clinical
in the robustness of this finding is weak. studies were generally conducted by psychiatrists,
However, the fail-safe N values for some used nonrandom selection procedures, were based
categories are extremely large. For example, it is on comparatively small samples, rarely distin-
unlikely that 1,995 additional studies of the in- guished between males and females or between
tergenerational transmission of divorce exist that blacks and whites in analyses, relied on
all yield null findings. For this reason, our faith in psychiatrists' judgments of well-being, and were
the robustness of this finding is very strong in- mainly published in the 1950s and 1960s. In con-
deed. trast, nonclinical studies were generally conducted
The final column contains the Hedges and by sociologists, used random sampling proce-
Olkin (1985) "H" statistic, which is a measure of dures, were based on comparatively large
effect size homogeneity. A significant H value in- samples, often distinguished between males and
dicates that the amount of heterogeneity within a females and between blacks and whites in

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 50 Journal of Marriage and the Family

TABLE 2. DIFFERENCES IN EFFECT SIZES BETWEEN CLINICAL AND NONCLINICAL SAMPLES

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 298.69***
Clinical 6 -.81*** 35.64***
Nonclinical 17 -.16*** 54.18***

Behavior / conduct 18.11**

Clinical 7 -.53*** 18.23**
Nonclinical 2 -.23*** 1.02

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity.
*p < .05. **p < .01. ***p < .001.

analyses, relied on respondent's self-reports of sons are not relevant.) It is clear that for both out-
well-being, and were mainly published in the comes, effect sizes were considerably larger for
1970s and 1980s. clinical than for nonclinical studies-five times
Table 2 contains mean effect sizes for clinical larger for psychological adjustment and over
and nonclinical studies for two outcomes: psycho- twice as large for behavior/conduct. Further-
logical adjustment and behavior/conduct. (Two more, the H-between values indicate that the dif-
other types of outcome-use of mental health ser- ferences in mean effect sizes between clinical and
vices and marital quality-occurred in clinical nonclinical samples were significant in both cases.
studies. But the H-within values of these variables (The significant H-within values, however, reveal
were not significant, so between-study compari- that a considerable degree of heterogeneity still

TABLE 3. DIFFERENCES IN EFFECT SIZES BETWEEN MALE, FEMALE, AND MIXED-SEX SAMPLES

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 336.89***
Male 6 -.14*** 3.14
Female 7 -.19*** 42.08***
Mixed 4 -.12** 6.40
Social relations 0.22
Male 4 -.05 4.55
Female 4 -.07* 10.90*
Mixed - -
Separation and divorce 15.57***
Male 10 -.18*** 9.61
Female 14 -.25*** 88.21***
Mixed 2 -.04 0.60
One-parent family status 63.01***
Male 2 -.67*** 17.99***
Female 4 -.20*** 3.91
Mixed - -

Educational attainment 9.08*

Male 7 -.20*** 8.65
Female 7 -.32*** 8.84
Mixed 4 -.30*** 11.46*
Material quality of life 0.18
Male 5 -.16*** 14.24**
Female 5 -.19*** 8.32
Mixed - -
Occupational 0.78 quality
Male 3 -.13* 1.85
Female 3 -.13** 0.98
Mixed 3 -.08 18.27***
Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity. Studies based on clinical samples are excluded.
*p < .05. **p < .01. ***p < .001.

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 Parental Divorce and Adult Well-being 51

exists within both clinical and nonclinical predictor of effect sizes for psychological adjust-
samples, except for nonclinical studies of ment, separation/divorce, one-parent family
behavior / conduct.) status, and educational attainment. For these
Tables 3 through 7 present further analyses of variables, post hoc tests were used to compare
effect sizes for outcome variables that contained a specific pairs of effect sizes (Hedges and Olkin,
significant amount of heterogeneity, as noted in 1985). This procedure indicated that the dif-
Table 1. The category of general family relations ferences between all-male and all-female samples
was excluded because of the small number of ef- were not significant for either psychological well-
fect sizes and because further between-study com- being or for separation and divorce. Instead,
parisons revealed little of interest. The category of mixed-sex samples scored lower than the other
parent-child relations was not included because its two. The interpretation of this pattern is unclear.
overall effect size was nonsignificant. The However, for educational attainment, post hoc
"other" category was excluded because of am- tests revealed that parental divorce had a stronger
biguity in interpretation. Because of the extreme estimated impact on females than males (p <
differences between clinical and nonclinical .05). On the other hand, for one-parent family
studies, clinical studies were omitted from these status, parental divorce had a stronger estimated
tables. Consequently, behavior/conduct was impact on males than females. (This is apparent
omitted because only two studies were nonclini- from the H-between value, since there were only
cal. two groups.)
Table 3 contains mean effect sizes for all-male, Table 4 contains comparisons between white,
all-female, and mixed-sex samples. The H-be- black, and mixed-race samples. Significant dif-
tween values reveal that sex was a significant ferences were apparent for all variables. Post hoc

TABLE 4. DIFFERENCES IN EFFECT SIZES BETWEEN WHITE, BLACK, AND MIXED-RACE SAMPLES

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 337.93***
Whites 11 -. 18** 43.78***
Blacks 2 -.12* 0.88
Mixed 4 -.10* 5.91
Social relations 13.04**
Whites 4 -.05* 2.38
Blacks 2 .01 0.14
Mixed 2 -.41*** 0.11

Separation and divorce 38.61**

Whites 12 -.26*** 55.02***
Blacks 10 -.16*** 4.64
Mixed 4 .03 15.72**

One-parent family status 4.11"

Whites 3 -.40*** 79.41***
Blacks 3 -.28*** 1.39
Mixed - -
Educational attainment 9.78**
Whites 8 -.31*** 8.41
Blacks 5 -.17*** 7.11
Mixed 5 -.32*** 12.73*

Material quality of life 11.14**

Whites 4 -.24*** 5.07
Blacks 4 -.12* 6.17
Mixed 2 .05 0.36

Occupational quality 14.84***

Whites 4 -.20*** 5.32
Blacks 2 -.06 0.28
Mixed 3 .05 1.44

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity. Studies based on clinical samples are excluded.
*p < .05. **p < .01. ***p < .001.

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 52 Journal of Marriage and the Family

TABLE 5. DIFFERENCES IN EFFECT SIZES BETWEEN STUDIES WITH

SMALL, MEDIUM, AND LARGE SAMPLES

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 338.89***
< 615 4 -.28*** 27.30***
616-1,554 6 -.13** 3.85
1,555 + 7 -.15*** 18.48**
Social relations 13.04**
< 615 2 -.41*** 0.11
616-1,554 2 .01 0.14
1,555 + 4 -.05* 2.38
Separation and divorce 6.48*
< 615 7 -.18*** 0.40
616-1,554 9 -.17*** 55.66***
1,555 + 10 -.23*** 51.45***
One-parent family status 5.70
< 615 3 -.32*** 0.47
616-1,554 1 -.21** 0.00
1,555 + 2 -.41*** 78.74***
Educational attainment 11.55**
< 615 5 -.19* 4.71
616-1,554 6 -.19*** 10.93
1,555 + 7 -.32*** 10.83
Material quality of life 3.04
< 615 5 -.10 12.34*
616-1,554 3 -.14* 7.24*
1,555 + 2 -.22*** 0.12
Occupational quality 4.50
< 615 3 -.03 0.09
616-1,554 3 -.22*** 6.14*
1,555 + 3 -.09** 11.15**

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. His the Hedges and Olkin (1985) measure of hetero-
geneity. Studies based on clinical samples are excluded.
*p < .05. **p < .01. ***p < .001.

tests revealed that for several outcomes, it was the samples often yielded stable findings, as reflected
mixed group that differed from the other two-a in nonsignificant H-within values.
pattern that is difficult to interpret. However, the Table 6 contains information on the year data
consequences of parental divorce were significant- were collected. Significant trends over time were
ly greater for whites than for blacks for separation apparent for psychological adjustment, separa-
and divorce (p < .01), one-parent family status, tion and divorce, material quality of life, and oc-
and educational attainment (both p<.05). These cupational quality. For the first three variables,
results suggest that the negative impact of divorce effect sizes for studies conducted in the 1980s were
is more substantial for whites than for blacks. smaller than those for studies carried out in earlier
Table 5 presents mean effect sizes for small, decades. The picture was less clear for occupa-
medium, and large sample sizes. Sample size was tional quality, but this is due to the fact that only
significantly related to effect sizes in four do- one study was available for each of the earlier
mains. However, the pattern was not consistent. decades. However, combining the two studies
For psychological adjustment and social relations, from earlier decades and comparing this mean
effect sizes were greatest in the smallest samples. with that for the 1980s revealed a significant dif-
But for separation and divorce and educational ference (-.30 vs. -.08, p < .01). Overall, these
attainment, effect sizes were greatest in the largest results suggest that the effects of divorce have
samples. It is also noteworthy that the smallest become less strong over time.
samples did not consistently reveal the largest Finally, Table 7 contains comparisons based
degree of heterogeneity-a common finding in on whether or not control variables were used in
meta-analysis. Instead, studies based on small analyses. For four variables-psychological ad-

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 ParentalDivorce andAdult Well-being 53

TABLE 6. DIFFERENCES IN EFFECT SIZES BETWEEN STUDIES CONDUCTED IN THE 1950-60s, 1970S, AND 1980s

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 341.70***
1950-60s 4 -.31*** 21.01***
1970s 6 -.16*** 19.29**
1980s 7 -.13*** 6.51"
Social relations 0.86
1950-60s -
1970s 2 -.09* 0.72
1980s 6 -.05 14.09*
Separation and divorce 25.81***
1950-60s 8 -.24*** 26.84***
1970s 9 -.30*** 15.25
1980s 9 -.15** 46.09***
One-parent family status 0.03
1950-60s - -
1970s 2 -.35*** 0.32
1980s 4 -.36*** 84.56***
Educational attainment 3.99
1950-60s - -
1970s 11 -.32*** 17.76
1980s 7 -.25*** 16.28*
Material quality of life 6.42*
1950-60s - -
1970s 4 -.36*** 2.91
1980s 6 -.15*** 13.41"
Occupational quality 10.11**
1950-60s 1 -.11 0.00
1970s 1 -.39*** 0.00
1980s 7 -.08** 11.77

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity. Studies based on clinical samples are excluded.
*p < .05. **p < .01. ***p < .001.

justment, separation and divorce, educational at- we used a stepwise procedure to enter variables in
tainment, and occupational quality-effect sizes the equations. Standard errors for b values were
were significantly greater in studies that reported adjusted by using formulas appropriate for the
only zero-order differences. These results suggest analysis of effect sizes. A .05 alpha level was used
that studies that do not use control variables tend for all significance tests. Because of the small
to overestimate the effects of parental divorce. number of effect sizes and the correlation between
predictors, we recommend that the results of this
analysis be treated with some caution.
Between-Study Comparisons:
In relation to psychological adjustment, two
Multivariate Analysis
predictors attained significance. First, the year of
As a final step, a series of weighted multiple- the study was positively associated with effect
regression analyses were carried out with pro- sizes (b = .009), indicating that more recent
cedures described by Hedges and Olkin (1985) and studies yielded less negative effects. Second,
Hedges and Becker (1986). One analysis was car- same-sex samples yielded stronger negative find-
ried out for each of the outcomes described in ings than did mixed-sex samples (b = -.10).
Tables 3 through 7. (Clinical studies were ex- The stepwise regression for social relations
cluded.) Sex, race, sample size, year of study, and revealed only one significant predictor: mixed-
presence of control variables were used as predic- race samples yielded stronger negative effect sizes
tors. Sex and race were represented as dummy than did same-race samples (b = -.37).
variables, sample size (x 100) and year of study In relation to separation and divorce, the step-
were continuous, and presence of controls was a wise procedure revealed that white samples were
dichotomy. Because we had no theoretical model, associated with significantly stronger negative ef-

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 54 Journal of Marriage and the Family

TABLE 7. DIFFERENCES IN EFFECT SIZES BETWEEN STUDIES THAT USED AND

DID NOT USE STATISTICAL CONTROLS IN ANALYSES

Domain of Well-being n X Effect Size H-between H-within

Psychological adjustment 336.54***
No controls 8 -.21*** 32.01***
Controls 9 -.14** 19.96*
Social relations 0.00
No controls - -
Controls 8 -.06** 15.67*
Separation and divorce 4.45*
No controls 6 -.33*** 18.58**
Controls 20 -.21*** 90.96***
One-parent family status 0.00
No controls - -
Controls 6 -.36*** 84.91***
Educational attainment 10.74**
No controls 5 -.35*** 5.29
Controls 13 -.24*** 22.00*
Material quality of life 0.00
No controls - -
Controls 10 -.18*** 22.74**

Occupational quality 7.11**

No controls 2 -.30** 3.00
Controls 7 -.08** 11.77

Note: n refers to the number of effect sizes based on independent samples. A negative effect size indicates that the
divorced group scores lower than the intact group in well-being. H is the Hedges and Olkin (1985) measure of hetero-
geneity. Studies based on clinical samples are excluded.
*p < .05. **p < .01. ***p < .001.

fect sizes than were black or mixed-race samples resulting in less negative effect sizes (b = .16). In
(b = -.14 and -.43, respectively). Female samples addition, same-race samples exhibited more nega-
and mixed-sex samples yielded more negative ef- tive effect sizes than did mixed-race samples (b =
fect sizes than did male samples (b = -.11 and -.19).
-.29, respectively). In addition, the total sample Finally, one-parent family status was predicted
size (x 100) and the year of the study were by sex, with male samples revealing more strongly
positively associated with effect sizes (b = .0008 negative effect sizes than female samples (b =
and .007, respectively), indicating less negative ef- -.47).
fects in larger samples and in more recent years.
Educational attainment was predicted by total
DISCUSSION
sample size (x 100) (b = -.004) in the stepwise
procedure, indicating that larger samples showed The present analysis, based on data from over
stronger negative effects than did smaller samples. 81,000 people in 37 studies, suggests that parental
Sex was also significant, with female samples and divorce (or permanent separation) has broad
mixed-sex samples exhibiting stronger negative ef- negative consequences for quality of life in
fect sizes than male samples (b = -.11 and -.14, adulthood. The data in Table 1 show that out-
respectively). comes associated with parental divorce include
In relation to material quality of life, the step- psychological well-being (depression, low life
wise procedure revealed that more recent studies satisfaction), family well-being (low marital quali-
yielded less negative effect sizes than did earlier ty, divorce), socioeconomic well-being (low
studies (b = .026). Furthermore, white samples educational attainment, income, and occupa-
yielded more negative effect sizes than did black tional prestige), and physical health. These results
or mixed-race samples (b = .17 and .27, respec- lead to a pessimistic conclusion: the argument that
tively). parental divorce presents few problems for
Occupational quality was predicted by the use children's long-term development is simply incon-
of control variables, with the use of controls sistent with the literature on this topic.

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 Parental Divorce and Adult Well-being 55

However, several qualifications temper the is, part of the association between parental
seriousness of this conclusion. First, the effect divorce and adult well-being is spurious. These
sizes in this literature are generally weak. The considerations suggest that studies that do not
largest mean effect sizes are in the order of one- control for family-of-origin characteristics are
third of a standard deviation between groups; likely to overestimate the consequences of paren-
many are closer to one-tenth of a standard devia- tal divorce.
tion. Of course, the fact that any effect at all is A fourth qualification concerns change over
observed, given the length of time involved, is time. It is sometimes pointed out that divorce has
noteworthy. One must also keep in mind that become more common, and more accepted, dur-
measurement error in the dependent variables will ing the last three decades; consequently, the ef-
lead to an underestimate of the true effect size. fects of divorce on children may be less strong to-
Nevertheless, on the basis of the research present- day than in the past. The present meta-analysis
ly conducted, it is clear that differences between provides evidence in support of this contention.
groups are small in absolute terms. In Table 6, studies of psychological adjustment,
Second, the data in Table 2 show that clinical separation/divorce, material quality of life, and
research presents a more serious picture of the ef- occupational quality suggest that the impact of
fects of divorce than does sociological research. parental divorce has diminished over time. These
Clinical studies are based on individuals who are results, with the exception of occupational quali-
suffering from disorders severe enough to bring ty, continued to be significant in the multivariate
them to the attention of mental health practi- analyses. And although the effect sizes for studies
tioners. People who experienced parental divorce carried out in the 1980s continue to be significant-
as children are clearly overrepresented in these ly greater than zero, they are small in absolute
groups. But clinical studies also show that a varie- terms. Incidentally, the finding of diminishing ef-
ty of other family-of-origin problems are fects over time is one that would be difficult to
associated with adult disorders, such as parental demonstrate without the use of meta-analysis.
death, parental mental illness, parental A final qualification refers to the fact that
alcoholism, marital violence, and problematic divorce is not a unitary experience for children.
parent-child relationships (Adam, Bouckoms, Although we have no data in the present analysis
and Streiner, 1982; Burbach and Borduin, 1986; on this point, a number of investigators have
Cadoret, Troughton, O'Gorman, and Heywood, shown that divorce is less disruptive if both
1986; Oltman, McGarry, and Friedman, 1952). parents maintain positive relationships with the
Parental divorce is one additional factor in child, if parental conflict subsides after the
multiproblem families that predispose individuals separation/divorce, and if the level of socio-
to major psychological and behavioral problems economic resources available to the child is not
later in life. But in the general community, where severely lessened (see Hetherington and Camara,
most people function well enough to escape the 1988; Krantz, 1988; Kurdek, 1981, for reviews).
attention of mental health practitioners, the link We should keep in mind that the effect sizes
between parental divorce and later negative out- reported in Table 1 are averages based on large
comes is weak. Although it is often stated, our numbers of individuals, with the consequences of
meta-analysis provides clear evidence that the divorce being strong for some people and minimal
results of clinical studies should not be generalized for others.
to the larger community, at least in relation to the Several other noteworthy findings emerge
present topic. from the meta-analysis. Studies of children tend
A third qualification has to do with the to show that boys in divorced families experience
sophistication of the analysis used. Table 7 reveals more problems than do girls (Hetherington, Cox,
that studies that use control variables show and Cox, 1982; Wallerstein and Kelly, 1980). But
weaker effect sizes than do studies that present in Table 3, only two sex differences are apparent.
zero-order differences between groups. (This was First, parental divorce is associated with one-
replicated in the multivariate analysis for occupa- parent family status more strongly among males
tional quality.) Variables such as parental educa- than females. In other words, although the ma-
tion and occupational status are associated with jority of one-parent families are female, parental
both parental divorce and later outcomes-that divorce increases the risk of being a single parent

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 56 Journal of Marriage and the Family

more for males than for females. The reason for ies. And studies conducted in the 1980s show
this, however, is unclear. Second, the univariate weaker effects than do those conducted in earlier
and multivariate analyses indicate that parental decades. In other words, the more sophisticated
divorce is linked to low educational attainment and recent the study, the more tenuous is the con-
more strongly among women than men. This re- nection between parental divorce and adult well-
sult is consistent with previous research showing being.
that noncustodial fathers are more likely to main-
tain contact with sons than with daughters and
NOTE
more likely to provide child support payments to
sons than to daughters (Hetherington, Cox, and This research was supported with a grant from the
University of Nebraska-Lincoln Research Council.
Cox, 1982). More to the point, Wallerstein and
Corbin (1986) noted that noncustodial fathers are
less willing to finance the education of daughters REFERENCES
than sons. The multivariate analysis also indicates Adam, Kenneth S., Anthony Bouckoms, and David
that among adult children of divorce, females are Streiner. 1982. "Parental loss and family stability."
more likely to have their own marriages end in Archives of General Psychiatry 39: 1081-1085.
divorce than are males. The results of the present Amato, Paul R. 1987a. Children in Austrialian families:
The growth of competence. Sydney: Prentice-Hall.
analysis, therefore, do not support the widely held
Amato, Paul R. 1987b. "Family processes in one-
contention that parental divorce has more parent, stepparent, and intact families: The child's
detrimental consequences for males than for point of view." Journal of Marriage and the Family
females. 49: 327-337.
Amato, Paul R. 1987c. Unpublished data from the
The data in Table 4 provide evidence of racial
Australian Family Formation Study. Department of
differences, with parental divorce having more Sociology, University of Nebraska-Lincoln.
serious consequences for whites than for blacks in Amato, Paul R. 1988. "Long-term implications of
the areas of separation/divorce, one-parent fami- parental divorce for adult self-concept." Journal of
ly status, educational attainment, material quality Family Issues 9: 201-213.
Amato, Paul R. In press. "Parental absence during
of life, and occupational quality. These differ-
childhood and adult depression." Sociological
ences were also observed in the multivariate
Quarterly.
analyses for separation/divorce and material Amato, Paul R., and Alan Booth. In press. "Long-term
quality of life. These results are consistent with consequences of parental divorce and marital
Hetherington, Camara, and Featherman (1983), unhappiness for adult well-being." Social Forces.
Amato, Paul R. and Bruce Keith. 1990. "Separation
who noted, in a review of the literature, that the
from a parent during childhood and adult
implications of single-parent family status for socioeconomic attainment." Unpublished
school achievement are weaker among black manuscript, Department of Sociology, University of
Nebraska-Lincoln.
children than among white children. The weaker
Bachman, G. Jerald, Patrick M. O'Malley, and Jerome
estimated effects for blacks may be due to the
Johnston. 1978. Youth in Transition (Vol. 6).
general prevalence of single-parent families in the Adolescence to Adulthood: Change and Stability in
black community. In addition, given the the Lives of Young Men. Ann Arbor: University of
numerous sources of inequality in American Michigan, Institute for Social Research.
society, parental divorce may lower the quality of Booth, Alan, and John N. Edwards. 1989. "Transmis-
sion of marital and family quality over the genera-
life only marginally for disadvantaged minorities.
tions: The effects of parental divorce and unhap-
But among whites, for whom opportunities are piness." Journal of Divorce 13: 41-58.
less restricted, parental divorce may have a more Brill, Norman Q., and Edward H. Liston. 1966.
pronounced impact. "Parental loss in adults with emotional disorders."
Archives of General Psychiatry 14: 307-313.
In conclusion, individuals who experienced
Burbach, D. J., and C. M. Borduin. 1986. "Parent-
parental divorce as children, compared with those child relations and the etiology of depression: A
parents were continuously married, have lower review of methods and findings." Clinical
quality of life as adults. However, these dif- Psychology Review 6: 133-153.
ferences appear to be small rather than large. Fur- Cadoret, Remi J., Ed Troughton, Thomas W.
O'Gorman, and Ellen Heywood. 1986. "An adop-
thermore, studies based on nonclinical samples tion study of genetic and environmental factors in
with controls in place for relevant predivorce drug abuse." Archives of General Psychiatry 43:
variables show weaker effects than do other stud- 1131-1136.

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 Parental Divorce and Adult Well-being 57

Crook, Thomas, and Allen Raskin. 1975. "Association Families: Parenting and Child Development. Hills-
of childhood parental loss with attempted suicide dale, NJ: Lawrence Erlbaum.
and depression." Journal of Consulting and Clinical Hill, Martha S., with Sue Augustyniak and Michael
Psychology 43: 277. Ponza. 1987. "Effects of parental divorce on
Duncan, Greg J., and Saul D. Hoffman. 1985. "A re- children's attainments: An empirical comparison of
consideration of the economic consequences of five hypotheses." Unpublished manuscript, Survey
marital disruption." Demography 22: 485-498. Research Center, Institute for Social Research,
Duszynski, R., John W. Shaffer, and Caroline B. University of Michigan.
Thomas. 1981. "Neoplasm and traumatic events in Keith, Verna M., and Barbara Finlay. 1988. "The
childhood." Archives of General Psychiatry 38: impact of parental divorce on children's educational
327-331. attainment, marital timing, and likelihood of
Faravelli, C. E. Sacchetti, A. Ambonetti, G. Conte, S. divorce." Journal of Marriage and the Family 50:
Pallanti, and A. Vita. 1986. "Early life events and 797-809.
affective disorder revisited." British Journal of Koller, K. M. 1970. "Family background in prison
Psychiatry 148: 288-295. groups: A comparative study of parental depriva-
Furstenberg, Frank F., and Christine W. Nord. 1987. tion." British Journal of Psychiatry 117: 371-380.
"Parenting apart: Patterns of childrearing after Koller, K. M., and J. N. Castanos. 1969. "Family
marital disruption." Journal of Marriage and the background and life situation in alcoholics." Ar-
Family 47: 893-904. chives of General Psychiatry 21: 602-610.
Glass, G., B. McGaw, and M. L. Smith. 1981. Meta- Krantz, Susan E. 1988. "The impact of divorce on
analysis in Social Research. Beverly Hills, CA: Sage. children." In S. M. Dornbusch and M. H. Strober
Glenn, Norval D., and Kathryn B. Kramer. 1985. "The (eds.), Feminism, Children, and the New Families.
psychological well-being of adult children of New York: Guilford.
divorce." Journal of Marriage and the Family 47: Kulka, Richard A., and Helen Weingarten. 1979. "The
905-912. long-term effects of parental divorce in childhood
Glenn, Norval D., and Kathryn B. Kramer. 1987. "The on adult adjustment." Journal of Social Issues 35:
marriages and divorces of the children of divorce." 50-78.
Journal of Marriage and the Family 49: 811-825. Kurdek, Lawrence. 1981. "An integrative perspective
Glick, Paul. C. 1980. "Marriage experiences of family on children's divorce adjustment." American Psy-
life specialists." Family Relations 29: 111-118. chologist 36: 856-866.
Greer, S. 1966. "Parental loss and attempted suicide: A McLanahan, Sara. 1985. "Family structure and the
further report." British Journal of Psychiatry 112: reproduction of poverty." American Journal of
465-470. Sociology 90: 873-901.
Greer, S., J. C. Gunn, and K. M. Koller. 1966. "Aetiol- McLanahan, Sara. 1988. "Family structure and de-
ogical factors in attempted suicide." British Medical pendency: Early transitions to female household
Journal: 1352-1355. headship." Demography 25: 1-16.
Hallstrom, T. 1987. "The relationships of childhood McLanahan, Sara, and Larry Bumpass. 1988. "Inter-
sociodemographic factors and early parental loss to generational consequences of family disruption."
major depresion in adult life." Acta Psychiatrica American Journal of Sociology 94: 130-152.
Scandinavica 75: 212-216. Mueller, Charles W., and Hallowell Pope. 1977. "Mari-
Hedges, Larry V., and Betsy Jane Becker. 1986. "Sta- tal instability: A study of its transmission between
tistical methods in the meta-analysis of research on generations. Journal of Marriage and the Family 39:
gender differences." Pp. 14-50 in Janet S. Hyde and 83-93.
Marcia C. Linn (eds.), The Psychology of Gender: Nock, Steven L. 1982. "Enduring effects of marital
Advances through Meta-analysis. Baltimore: Johns disruption and subsequent living arrangements."
Hopkins University Press. Journal of Family Issues 3: 25-40.
Hedges, Larry V., and Ingram Olkin. 1985. Statistical Oltman, Jane E., John J. McGarry, and Samuel Fried-
Methods for Meta-analysis. New York: Academic man. 1952. "Parental deprivation and the 'broken
Press. home' in dementia praecox and other mental dis-
Hetherington, E. Mavis, and Kathleen A. Camara. orders." American Journal of Psychiatry 108:
1988. "The effects of family dissolution and 685-693.
reconstitution on children." Pp. 420-431 in Norval Overall, John E., B. W. Henry, and Arthur Woodward.
D. Glenn and Marion T. Coleman (eds.), Family 1974. "Dependence of marital problems on parental
Relations: A Reader. Chicago: Dorsey Press. family history." Journal of Abnormal Psychology
Hetherington, E. Mavis, Kathleen A. Camara, and 83: 446-450.
David L. Featherman. 1983. "Achievement and in- Pope, Hallowell, and Charles W. Mueller. 1976. "The
tellectual functioning in children in one-parent intergenerational transmission of marital instability:
households." Pp. 205-284 in Janet T. Spence (ed.), Comparisons by race and sex." Journal of Social
Achievement and Achievement Motives. San Fran- Issues 32: 49-67.
cisco: W. H. Freeman. Rosenthal, Robert. 1979. "The 'file drawer' problem
Hetherington, E. Mavis, Martha Cox, and Roger Cox. and tolerance for null results." Psychological Bulle-
1982. "Effects of divorce on parents and children." tin 86: 638-641.
Pp. 223-288 in Michael Lamb (ed.), Nontraditional Roy, Alec. 1985. "Early parental separation and adult

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms
 58 Journal of Marriage and the Family

depression." Archives of General Psychiatry 42: Wallerstein, Judith S., and Joan B. Kelly. 1980. Sur-
987-991. viving the Breakup: How Children and Parents
Schooler, Carmi. 1972. "Childhood family structure Cope with Divorce. London: Grant McIntyre.
and adult characteristics." Sociometry 35: 255-269. Weitzman, Lenore J. 1985. The Divorce Revolution.
Sethi, Brij B. 1964. "Relationship of separation to New York: Free Press.
depression." Archives of General Psychiatry 10: White, Lynn, David Brinkeroff, and Alan Booth. 1985.
486-496. "The effect of marital disruption on child's attach-
Tennant, Christopher, Jane Hurry, and Paul Bebbing- ment to parents." Journal of Family Issues 6: 5-22.
ton. 1982. "The relation of childhood separation ex- Wolchik, S., I. N. Sandler, S. T. Braver, and B. Fogas.
periences to adult depressive and anxiety states." 1985. "Events of parental divorce: Stressfulness rat-
British Journal of Psychiatry 141: 475-482. ings by children, parents, and clinicians." American
Wadsworth, M. E. J., and M. Maclean. 1986. "Par- Journal of Community Psychology 14: 59-74.
ents' divorce and children's life chances." Children Wolf, Fredric M. 1986. Meta-analysis: Quantitative
and Youth Services Review 8: 145-159. Methods for Research Synthesis: Sage University
Wallerstein, Judith S., and Shauna B. Corbin. 1986. Paper series on Quantitative Applications in the
"Father-child relationships after divorce: Child sup- Social Sciences, Series No. 07-001. Beverly Hills,
port and educational opportunities." Family Law CA: Sage.
Quarterly 20: 109-128.

APPENDIX

STUDIES USED IN THE META-ANALYSIS

Adam, Bouckoms, and Streiner, 1982 Hallstrom, 1987

Amato, 1987c Hill, Augustyniak, and Ponza, 1987
Amato, 1988 Keith and Finlay, 1988
Amato, in press Koller, 1970
Amato and Booth, in press Koller and Castanos, 1969
Amato and Keith, 1990 Kulka and Weingarten, 1979
Bachman, O'Malley, and Johnston, 1978 McLanahan, 1985
Booth and Edwards, 1989 McLanahan, 1988
Brill and Liston, 1966 McLanahan and Bumpass, 1988
Cadoret, Troughton, O'Gorman, and Heywood, 1986 Mueller and Pope, 1977
Crook and Raskin, 1975 Nock, 1982
Duszynski, Shaffer, and Thomas, 1981 Oltman, McGarry, and Friedman, 1952
Faravelli, Sacchetti, Ambonetti, Conte, Pallanti, and Overall, Henry, and Woodward, 1974
Vita, 1986 Pope and Mueller, 1976
Glenn and Kramer, 1985 Roy, 1985
Glenn and Kramer, 1987 Schooler, 1972
Glick, 1980 Sethi, 1964
Greer, 1966 Tennant, Hurry, and Bebbington, 1982
Greer, Gunn, and Koller, 1966 Wadsworth and Maclean, 1986

This content downloaded from 161.116.100.129 on Wed, 20 Apr 2016 23:41:13 UTC
All use subject to http://about.jstor.org/terms