You are on page 1of 10

JEAN M.

TWENGE San Diego State University

W. KEITH CAMPBELL University of Georgia*

CRAIG A. FOSTER United States Air Force Academy**

Parenthood and Marital Satisfaction:


A Meta-Analytic Review

This meta-analysis finds that parents report lower people make. In contrast to previous historical
marital satisfaction compared with nonparents (d eras, individuals today have an unusual amount of
5 2.19, r 5 2.10). There is also a significant freedom to decide if and when they will have chil-
negative correlation between marital satisfaction dren and how many they will have. In many cases,
and number of children (d 5 2.13, r 5 2.06). this is a mutual decision reached between marital
The difference in marital satisfaction is most pro- partners. In making this decision, couples some-
nounced among mothers of infants (38% of moth- times consider an important question: ‘‘How will
ers of infants have high marital satisfaction, com- children affect our relationship?’’ Folk wisdom
pared with 62% of childless women). For men, suggests that babies bring couples closer together,
the effect remains similar across ages of children. and some couples name greater closeness as a rea-
The effect of parenthood on marital satisfaction is son for having a baby (Brinley, 1991). Unfortu-
more negative among high socioeconomic groups, nately, some scientific research suggests that cou-
younger birth cohorts, and in more recent years. ples actually grow less satisfied with their marital
The data suggest that marital satisfaction de- relationship after having children (see Belsky &
creases after the birth of a child due to role con- Pensky, 1988, for a review). Others may argue
flicts and restriction of freedom. that children will have no effect (that an unhappy
marriage will remain unhappy, and a happy mar-
Deciding whether to have children is one of the riage will remain happy). Which viewpoint is cor-
most important and life-changing decisions many rect? If there is an effect, how large is it? What
variables determine the effect of children on mar-
Department of Psychology, San Diego State University, 5500 ital satisfaction? Finally, why might children af-
Campanile Drive, San Diego, CA 92182-4611 (jtwenge@ fect marital satisfaction? We seek to answer these
mail.sdsu.edu). questions in this meta-analytic review of parent-
hood and marital satisfaction.
*Department of Psychology, University of Georgia, Athens,
GA 30602-3013. Understanding the transition to parenthood is
not simply a matter of scientific interest. Delin-
**DFBL, 2354 Fairchild Drive, M2, United States Air eating the relationship between parenthood and
Force Academy, CO 80840-6228.
marital satisfaction could have profound practical
Key Words: birth cohort, children, gender, marital satis- value for those who wish to make an informed
faction, marriage. decision about having children. Not only would

574 Journal of Marriage and Family 65 (August 2003): 574–583


Parenthood and Marital Satisfaction 575

potential parents be able to make better decisions, may result in a number of consequences that de-
but also—and perhaps more importantly—future crease marital satisfaction.
parents could prepare for and hopefully counteract Nevertheless, there are certainly many instanc-
the potentially negative effect of children on their es when children improve or increase satisfaction
marriages. In short, social scientists and the public or at least do not have a negative effect. Further,
alike may benefit from a clarification of the rela- the decrease in satisfaction may be experienced
tionship between parenthood and marital quality. disproportionately by some individuals (e.g., by
Here, we first review the specific arguments pre- women more than men). Examining moderator
dicting a decrease in marital satisfaction during variables may help address these questions.
the transition to parenthood. Second, we outline
the moderator variables that may have an influ- Moderator Variables
ence on the effect size. Third, we outline four gen-
eral theories that address the influence of children Past research on the effect of children on marital
on marital satisfaction. Finally, we present the re- satisfaction has been understandably hindered by
sults of the meta-analysis and test each of the rel- the difficulty of including multiple moderators in
evant theories. single studies. Most studies, for example, include
only one type of sample and thus cannot compare
across many variables. Meta-analysis offers the
WHY MIGHT THE TRANSITION TO PARENTHOOD potential to examine moderators that have not
DECREASE MARITAL SATISFACTION? been compared in the literature thus far because
they are only evident when one looks across stud-
The seminal research on this topic was conducted ies. The moderators examined in the present meta-
by LeMasters (1957), who suggested that the ad- analysis include the following: (a) gender of par-
dition (or removal) of a family member could ent, (b) age of child, (c) socioeconomic status
force a reorganization of the family system. This (SES) of parents, and (d) birth cohort (i.e., the
reorganization may be described as a crisis: a de- birth year of the parents). We also examine the
cisive change for which old patterns are inade- measure of marital satisfaction used. All of these
quate. In interviews with parents of sound mental variables could moderate the effect size. For ex-
health who had at least one child under 5 years ample, any differences in marital satisfaction be-
old, mothers reported experiencing a loss of sleep, tween parents and nonparents could be stronger
chronic tiredness, confinement to the home, guilt for women compared with men.
over not being better mothers, worry about their
appearance, and other dissatisfactions. Fathers re-
ported similar experiences and added problems Theoretical Models
such as experiencing a decline in their wife’s sex- As noted earlier, children likely exert multiple in-
ual responsiveness, economic pressure resulting fluences on marital satisfaction. Explanations of
from their wife’s withdrawal from the workplace, these influences vary; nevertheless, it is important
and general disenchantment with the parental role. to elaborate and understand the general theoretical
Further research has supported and extended underpinnings of the effect. In the present paper,
LeMasters’ (1957) findings. The transition to par- we combine the insights of past research into four
enthood may (a) increase chores, stress, and general theoretical models. We call these (a) the
strain, partially due to decreased time for discus- role conflict model, (b) the restriction of freedom
sion (Anderson, Russell, & Schumm, 1983; Lo- model, (c) the sexual dissatisfaction model, and
pata, 1971); (b) interfere with couple companion- (d) the financial cost model. Each model is asso-
ship (Glenn & Weaver, 1978; White, 1983); (c) ciated with a pattern of predictions on the mod-
interfere with the couple’s sex life (Blumstein & erator variables. In meta-analysis, each of the in-
Schwartz, 1983); (d) overload the number of so- dividual predictions can be tested empirically.
cial roles accumulated by the new parents (Rollins
& Galligan, 1978); (e) exacerbate inequity be- Role conflict model. The addition of children to a
tween partners that underbenefits wives (Feeney, family can lead to a reorganization of social roles
Peterson, & Noller, 1994); and (f) create negative along traditional lines. This is especially true for
evaluations of marriage, especially among nontra- women, who are often expected to take on a pri-
ditional women (Belsky, Lang, & Huston, 1986). marily caregiving role for the child. The father
Thus it appears that the transition to parenthood also may be expected to take on a greater bread-
576 Journal of Marriage and Family

winning role than before. A problem arises when Restriction of freedom model. Children greatly di-
the couple does not desire traditional roles. Wom- minish the freedom of individuals in a marriage.
en, for example, may give up or downgrade their Children demand a great deal of time and atten-
professional roles for the sake of their caregiving tion. These demands directly interfere with the
roles. The result may be more power for husbands pursuit of the parents’ own pleasures, and thus
(who now are the significant or sole breadwinners; children restrict the experience of freedom for par-
Feeney et al., 1994). Women also may experience ents. The restriction of freedom model generates
some psychological stress as they see themselves several specific predictions, many of which are
primarily in a caregiving role, especially if they similar to those for the role conflict model, prob-
are accustomed to a professional role (Belsky et ably because role conflict involves giving up free-
al., 1986). In addition, many women are overbur- dom to pursue professional goals for the purpose
dened and feel that they are not receiving enough of childrearing. We present both models, however,
help from their partners (Hochschild, 1989). Fi- because their explanations for the effects are dif-
nally, becoming a parent adds another social role, ferent practically and theoretically. For example,
and having more roles may lead to stress and con- a woman who desires a traditional role might not
flict (Rollins & Galligan, 1978). feel role conflict but may still experience a restric-
The role conflict model generates several spe- tion of freedom.
cific predictions for the moderator variables. First, The restriction of freedom model predicts that
women should experience greater dissatisfaction women should experience greater dissatisfaction
compared with men. Women are the primary care- than men, as they generally experience the greater
takers, so their roles change more when a child is restriction of freedom. Similarly, the effect should
born. Second, the effect should be greater when be greater when children are infants, because in-
children are infants. Older children demand less fants demand the most time and attention. High
direct caregiving and spend much of their time in SES individuals (especially women) should suffer
school. Thus they create less role conflict for par- the greatest dissatisfaction, as they are more likely
ent caregivers. Next, high SES individuals (espe- to leave or scale back desirable careers. More re-
cially women) should suffer the lowest satisfac- cent studies should also find greater dissatisfac-
tion with the presence of children. High SES tion. Recent decades have seen increasing individ-
women may have successful careers that they may ualism and autonomy (Frum, 2000; Gough, 1991;
scale back or give up with the arrival of a child. Twenge & Campbell, 2001), and individuals today
In general, high SES people enjoy the freedom are less tied to family and social responsibilities
and individualism that money and professional ca- than those in the past (Putnam, 2000). In addition,
reers bring; the loss of freedom and reduced au- couples have children later in life on average and
tonomy experienced after the birth of a child may thus become accustomed to a life of adult free-
require a greater adjustment (e.g., Jenkins, Ras- doms before they become parents. Therefore, the
bash, & O’Connor, 2003). restriction of freedom brought on by parenthood
Last, there might be a birth cohort effect. Re- might come as a shock to more recent birth co-
cent birth cohorts, where women expect and ex- horts.
perience more nontraditional roles, should report
the largest negative effects of children. Women Sexual dissatisfaction model. Children often inter-
today have more prestigious work roles and have fere with parents’ sex lives (Blumstein &
taken on the assertive personality required for Schwartz, 1983). It is more difficult to have a sex-
these roles (Twenge, 2001). The adjustment from ual relationship with children in the house. The
professional career woman to mother is much decrease in sexual intimacy may have negative
more radical than the smaller adjustments of pre- consequences for marital satisfaction.
vious decades (when women had less prestigious The sexual dissatisfaction model generates sev-
jobs or were not employed at all before having eral specific predictions. Because men report
children). In addition, having children is now seen greater sexual desire in marriage (Baumeister, Ca-
as a choice, rather than a social imperative (Yan- tanese, & Vohs, 2001), men should experience the
kelovich, 1981). Thus parents might believe that greater dissatisfaction. Next, the effect of children
it is their ‘‘fault’’ when the transition to parent- on sexual satisfaction might be largest for parents
hood is difficult, rather than seeing it as the natural with infants, because infants demand attention
next step in a marriage and something that every- throughout the evening and night. There should
one experiences. be no effect for either SES or birth cohort, as these
Parenthood and Marital Satisfaction 577

variables are unlikely to interact with sexual sat- Thus the sample of data points includes journal
isfaction. articles, masters’ theses, and dissertations.
We searched for two types of studies, which
Financial cost model. Children are expensive. we analyze separately: studies comparing the mar-
They can place significant stress on family fi- ital satisfaction of people with children to those
nances, which may lead to dissatisfaction with the without, and studies examining the correlation be-
marriage (e.g., LeMasters, 1957). The financial tween marital satisfaction and number of children.
cost model generates several specific predictions. Studies reporting separate means or correlations
LeMasters suggested that the financial burden of by relevant variables (e.g., sex) were included as
children should be greater for husbands than separate data points in the analysis; thus some
wives because husbands are the traditional bread- studies yielded more than one data point. The
winners. Therefore, this model suggests greater searches yielded studies representative of the lit-
dissatisfaction among men. Children are generally erature: four studies (3%) sampled exclusively
more expensive as they age, so the effect should White respondents; 22 (15%) of the studies sam-
be greater for older children. There should be a pled a minority group exclusively (Black, Hispan-
larger effect for low SES couples who will strug- ic, or Asian); and most samples (122) were pre-
gle to make ends meet with the addition of chil- dominantly White but included all racial and
dren. As for birth cohort, the effect should be ethnic groups (82%). The searches yielded 97 ar-
stronger with recent cohorts because children are ticles containing 148 data points that met the cri-
more expensive for recent than for earlier cohorts teria for inclusion in the meta-analysis. These
(for example, the increase in the cost of a college studies included a total of 47,692 respondents.
education has far outstripped inflation, as have
housing and day care costs). Criteria for Inclusion
We used several criteria for deciding which arti-
PRESENT META-ANALYSIS cles to include in the meta-analysis. First, an ar-
ticle needed to have a sample that included both
We conducted a meta-analysis of the effect of
parents and nonparents. These took several forms:
children on marital satisfaction. We had three pri-
Some couples were studied longitudinally, allow-
mary goals for this analysis: (a) to verify the mag-
ing their satisfaction to be examined before and
nitude and direction of any existing effect of chil-
after the birth of their first child. In other studies,
dren on marital satisfaction; (b) to investigate the
surveys compared the marital satisfaction of cou-
scope and moderators of the effect (e.g., Does the
ples with children to that of couples without chil-
effect occur for men and women? Are there birth
dren or simply reported a correlation between
cohort differences?); and (c) to test four proposed
marital satisfaction and the number of children. In
theoretical explanations for the effect of children
some cases, articles reported separate statistics on
on marital satisfaction.
marital satisfaction for couples with children of
different ages; in these cases, we compared non-
METHOD parents to parents with the youngest children (to
most accurately capture the transition to parent-
Sample of Studies hood period). Second, the article needed to report
statistics sufficient for the calculation of an effect
We searched for articles using three techniques: size (d). We coded d from r, t, F, or means and
(a) PsycLIT (1974–2000) and Sociofile (1974– standard deviations (see, e.g., Wolf, 1986). A refer-
2000) using the keywords children and marital ence list of articles and dissertations included in the
satisfaction; (b) reference sections of all included meta-analysis is available by request to J. M. T.
journal articles for additional relevant articles; and
(c) dissertation abstracts, a major source of the
Weights for d
unpublished data of masters’ theses and disserta-
tions, using the same keywords (dissertations un- We corrected for bias in d using two techniques.
available through interlibrary loan were examined For studies comparing couples with children to
at the Library of Congress). We did not search for those without, we used the formulas for w pre-
conference papers, given the difficulty of system- sented in Hedges and Becker (1986). For studies
atically searching for and obtaining such papers. reporting the correlation between marital satisfac-
578 Journal of Marriage and Family

tion and number of children, we converted each r as separate data points. We coded each study for
to z and then weighted by n 2 3 (Hedges & Olkin, the following variables: sex; age; race/ethnicity of
1985). For ease of presentation, we converted the sample; length of marriage; age of children; SES
z back to d in the results. In all cases, positive (coded as low, middle, or high); measure of mar-
effect sizes indicate that parents’ marital satisfac- ital satisfaction; measurement of children (pres-
tion was higher than that of nonparents’ or than ence or absence vs. number); year of data collec-
those with fewer children, whereas negative effect tion (coded as 2 years prior to publication unless
sizes indicate that parents’ marital satisfaction was noted otherwise in the article; longitudinal studies
lower than that of nonparents’ or than those with were coded by subtracting 2 years plus the num-
fewer children. Thus a negative effect size means ber of years in the longitudinal study); and birth
that parenthood is associated with lower marital cohort (year minus the mean age of the sample).
satisfaction, and a positive effect size means that Some of these variables (e.g., length of marriage,
parenthood is associated with higher marital sat- racial group) were not significant moderators and
isfaction. In a few cases, effect sizes were simply are thus not discussed further.
reported as nonsignificant; for such effect sizes, a
null effect (d 5 .00) was entered.
RESULTS
Analyzing the two types of studies separately
proved useful for theoretical reasons as well.
Comparing Parents to Nonparents
Comparing parents and nonparents is the most di-
rect way of addressing how parenthood affects Parents had significantly lower marital satisfaction
marital satisfaction. The correlation with number than nonparents (d 5 2.19, r 5 2.10) across 90
of children is less direct. Because samples usually studies (n of respondents 5 31,331). The 95%
include many more parents than nonparents, stud- confidence interval was 2.22 to 2.16; because
ies reporting a correlation with the number of chil- this interval does not include zero, the d is sig-
dren are concerned with the effect of a larger ver- nificant. This is a small effect size (around .20;
sus a smaller family on marital satisfaction. Cohen, 1977), but it is significantly different from
zero. As Rosenthal (1994) explains, an r of .10
means that 55% of nonparents have higher than
Analyses
average marital satisfaction, compared with 45%
In performing the analyses, we employed standard of parents. HT was 221.40, p , .001, showing that
techniques used in meta-analysis (e.g., Hedges & these samples demonstrate significant heteroge-
Becker, 1986). First, we weighted the effect sizes neity and suggesting that there are significant
as explained in the previous section. For all stud- moderators.
ies and for moderator variable subgroups (e.g., Women showed a slight but significantly larger
women with infants), we computed an effect size effect size for parenthood than men (see Table 1).
and heterogeneity statistic (called total H or HT Thus parenthood had a stronger negative effect on
for all samples and H within or HW for subgroups). women’s marital satisfaction compared with
These statistics are computed using a x2 distribu- men’s.
tion and indicate whether the samples are signif- Among those samples reporting the informa-
icantly heterogeneous from each other. This anal- tion, we also examined how the age of a couple’s
ysis determines only that there is significant child or their youngest child influenced the effect.
variation among the samples; analyses for specific The effect size in studies comparing childless cou-
moderator variables must be performed to ascer- ples to those with an infant was significantly high-
tain the source of this variation. To compare sub- er (d 5 2.38; p , .001) than in those comparing
groups of moderator variables with each other, we childless couples to those with a child older than
used H between (HB), which is analogous to an 2 years (d 5 2.20).
analysis of variance (ANOVA). Does age of child interact with gender in its
effect on marital satisfaction? We found that it
does. Men’s marital satisfaction is relatively in-
Theoretical Moderators of Effect Size Magnitude
variant across children’s age (d 5 2.25 for men
We coded each study for moderator variables that with infants, d 5 2.30 for men with older chil-
might influence the effect size. A study that re- dren). In contrast, the marital satisfaction of wom-
ported means or correlations separately for mod- en with infants, compared with that of childless
erator variable subgroups (e.g., sex) was included women, is very low (d 5 2.50), whereas the dif-
Parenthood and Marital Satisfaction 579

TABLE 1. VARIABLES MODERATING THE EFFECT SIZE OF PARENTS’ VERSUS NONPARENTS’ MARITAL SATISFACTION

Variable and Categories H Between k d 95% CI for d H Within

Gender 4.38*
Men 29 2.13 2.17 to 2.09 44.95**
Women 38 2.19 2.23 to 2.15 111.34***
Age of child 10.83***
Infant 33 2.38 2.44 to 2.32 41.25*
Older than 2 years 16 2.20 2.29 to 2.11 22.88
Gender by age of child 27.26***
Women with infants 16 2.50 2.60 to 2.40 19.20
Women with older children 10 2.14 2.24 to 2.04 9.90
Men with infants 13 2.25 2.25 to 2.03 9.23
Men with older children 5 2.30 2.52 to 2.08 6.45
Socioeconomic status 57.72***
Low 3 .05 2.19 to .29 1.59
Middle 46 2.14 2.17 to 2.11 89.70***
High 27 2.45 2.53 to 2.37 26.73
Measure 35.68***
Standardized measures 43 2.36 2.41 to 2.31 50.67***
One item and not standard 47 2.15 2.18 to 2.12 124.59***
Note: k 5 number of samples in each group; d 5 difference in terms of standard deviations; CI 5 confidence interval.
*p , .05. **p , .01. ***p , .001.

ference is smaller for women with older children groups, and more recent birth cohorts. The sexual
(d 5 2.14). Thus the largest difference in marital dissatisfaction model, however, predicted a larger
satisfaction is between childless women and wom- effect for men and no effect for SES or birth co-
en with infants. Only 38% of women with infants hort. The financial cost model predicted larger ef-
have higher than average marital satisfaction, fects for men, parents of older children, and low
compared with 62% of childless women (using the SES groups (exactly the opposite of the results we
binomial display of Rosenthal, 1994). found); however, this model did correctly predict
SES was also a significant moderator. The the larger effect size for more recent cohorts.
presence of children has a more negative effect on We also examined the measure of marital sat-
marital satisfaction among high SES groups (see isfaction used. Standardized measures (the Locke-
Table 1). Wallace, the Global Distress Scale, and the Dyadic
We also found historical and birth cohort ef- Adjustment Scale) yielded a significantly higher
fects. Year of data collection was significantly cor- effect size compared with single item and non-
related with the effect size, weighted r 5 2.21, p standardized measures (see Table 1). It seems like-
, .001. More recent samples showed a more neg- ly that standardized measures are more valid and
ative effect of children on marital satisfaction. reliable than single-item measures and thus have
Similarly, those from more recent birth cohorts more power to discriminate between parents and
demonstrated a more negative effect size (the cor- nonparents. In addition, many single-item mea-
relation between birth cohort and d, weighted r 5 sures may not assess the specific areas of conflict
2.42, p , .001). Thus, more recent generations increased by parenthood. Overall, the moderator
of parents have experienced greater dissatisfaction variable results are best explained by the role con-
with their marriages after the birth of a child. flict and restriction of freedom models.
Compared with earlier generations, more recent
cohorts experienced a 42% greater drop in marital
Correlation Between Marital Satisfaction and
satisfaction with the transition to parenthood (us-
Number of Children
ing the guidelines for r of Rosenthal, 1994).
The moderator variable results are best ex- What is the correlation between the number of
plained by the role conflict and restriction of free- children in a family and the parents’ marital sat-
dom models. Both of these models predicted larg- isfaction? The relationship was negative, with a d
er effects for women, parents of infants, high SES of 2.13 (z 5 2.06, r 5 2.06) across 58 samples
580 Journal of Marriage and Family

(n of respondents 5 16,361). Thus the more chil- Size of the Effect


dren in the family, the lower the parents’ marital
satisfaction. For this analysis, each effect size was How large is a large effect? This is a subjective
converted to Fisher’s z and then weighted by n 2 judgment. Cohen (1977) would classify the effect
3. We then converted the zs to ds in order to use size comparing parents and nonparents (d 5
the same measure of effect size used in the other 2.19) as small (under his guidelines, a d around
analysis. The 95% confidence interval was 2.15 .20 is a small effect). Another way to understand
effect sizes is the binomial effect size display (Ro-
to 2.11; because this does not include zero, the
senthal, 1994). For example, the effect size com-
effect is significant. HT was 153.12, suggesting
paring childless women and those with infants
significant heterogeneity.
was d 5 2.50, r 5 2.24. Converted to a binomial
Gender, age of youngest child, birth cohort,
display, 62% of childless women have high mar-
and measurement of marital satisfaction were not
ital satisfaction compared with only 38% of wom-
significant moderator variables in this analysis.
en with infants. Understood this way, the effect is
The effect size differed, however, with the SES of
substantial. For the overall effect (d 5 2.19, r 5
the sample, HB 5 24.11, p , .001. In contrast to
2.10), the comparison is as follows: 55% of peo-
the other analysis, the lower SES (d 5 2.26, k 5
ple without children have high marital satisfaction
7) and higher SES groups (d 5 2.28, k 5 13) compared with 45% of people with children. This
demonstrated similar, stronger effect sizes. In the is a smaller, but still noticeable, effect.
previous analyses, the high SES group had the
most negative effect size. Middle SES groups
showed a weaker effect (d 5 2.08, k 5 31). Be- Moderator Variables
cause this analysis examines the correlation with
number of children, it could be that subsequent The transition to parenthood. The effect size is
children decrease marital satisfaction more signif- stronger when parents are compared with nonpar-
icantly among those in the lower SES strata. Al- ents than when correlating marital satisfaction and
ternatively, the high or low SES groups might number of children. In addition, marital satisfac-
have a restriction of range problem in number of tion is lower among parents of infants, particularly
children, which might have changed the effect mothers. Taken together, these results suggest that
the transition to parenthood is crucial. Becoming
size in this analysis.
a parent has the largest effect on marital satisfac-
There was a significant correlation between the
tion.
effect size and year of data collection, r 5 2.31,
p , .01. Thus, children had a more negative effect
on marital satisfaction in recent years, compared Gender. In studies comparing parents with non-
with past years, consistent with the results of the parents, women reported a slightly greater in-
previous analysis. Birth cohort was not signifi- crease in marital dissatisfaction with parenthood.
cantly correlated with effect size. This effect was strengthened when the interaction
The theoretical models did not receive much with the age of the child was introduced. Mothers
support from these results. None of the models of infants are significantly more dissatisfied with
their marriages than any other group (e.g., men
predicted a stronger effect size for middle SES
with infants, men with older children, and women
groups. The role conflict, restriction of freedom,
with older children.) The large effect for mothers
and financial cost models, however, predicted that
of infants suggests that marital satisfaction de-
the effect of parenthood on marital satisfaction
creases the most among the individuals who are
would grow more negative over historical time.
most responsible for the child. In the vast majority
None of the predictions of the sexual dissatisfac- of cases, this is the mother (e.g., Hochschild,
tion model were supported by these data. 1989).

DISCUSSION Socioeconomic status. In both analyses, high SES


groups showed more marital dissatisfaction with
The results of this meta-analysis show that parents parenthood when compared with middle-class in-
report lower marital satisfaction than nonparents. dividuals. In the analysis comparing parents and
In addition, people with more children report nonparents, high SES individuals were also more
greater marital dissatisfaction than people with dissatisfied than low SES groups. Apparently par-
fewer children. enthood has a greater effect on marital satisfaction
Parenthood and Marital Satisfaction 581

when one is particularly well educated or well- single-item questions about marital satisfaction
off. As noted in the introduction, high SES indi- and instead administer multiple-item, standardized
viduals may experience more role conflict and scales.
greater reduction in freedom with the birth of a
child (e.g., Jenkins et al., 2003).
Examining the Four Theoretical Models
Birth cohort. Both analyses demonstrated an ef- The analysis of the moderator variables made it
fect for year of data collection, with more recent possible to test the theoretical perspectives out-
studies finding more marital dissatisfaction with lined in the discussion. For the data on presence
parenthood. In addition, the analysis comparing versus absence of children, the moderator variable
parents and nonparents found an effect for birth results strongly favor the role conflict model and
cohort, with more recent generations showing the restricted freedom model, with almost all of
greater dissatisfaction (the number of children their predictions supported by the data. For ex-
analysis did not show a birth cohort difference). ample, both models predicted a stronger negative
In many ways, historical effects are cultural ef- effect for women, parents of infants, high SES
fects, with birth cohort a proxy for the culture of individuals, and more recent years and birth co-
the larger social environment (Twenge, 2001, horts. The sexual dissatisfaction model and the fi-
2002). The culture of previous generations valued nancial cost model received less support. Al-
social conformity more and favored closer con- though some of the predictions of these models
nections with family and community (Fukuyama, were supported, several others were not (e.g., both
1999; Putnam, 2000; Yankelovich, 1981). Overall, predicted a stronger effect for men, and neither
the 1950s and 1960s were more pronatalist, en- predicted the SES results).
couraging a social norm of childbearing; this at-
titude was also reflected in the high birth rates of
Strengths and Limitations
these eras. Couples in previous decades also mar-
ried and had children earlier in life. In contrast, One of the most significant strengths of this meta-
the culture of recent years appears to value indi- analysis is its summary of a very large body of
vidual freedom and choice (Frum, 2000; Yanke- research. We located and analyzed data from 148
lovich), which is also reflected in the lower birth samples and 47,692 individuals. This represents
rate. Perhaps because of a focus on individual the largest analysis to date of the effect of children
goals, couples now, on average, marry later and on marital satisfaction. With so many data points
have children later. For many people today, the to analyze, we were able to address many ques-
preparenthood adult period is one of extraordinary tions that individual studies cannot. For example,
freedom and self-focus. Consequently, there may we were able to compare samples from many dif-
be a greater contrast between young adult life and ferent historical periods and with different ages of
the early years of parenthood, a contrast that may children, as well as studies using different mea-
make the transition more difficult and lead to sures of marital satisfaction.
greater marital dissatisfaction. This analysis also has limitations. Although we
found an overall effect size for parents versus
Measurement. In the analysis comparing parents nonparents and for parents with fewer versus more
and nonparents, standardized measures showed a children, there were not enough data points to
larger effect than nonstandard and single-item compare couples with one child versus those with
measures. Standardized measures are more valid two versus three or more children. Thus we were
and reliable, strengthening their utility. In addi- unable to determine the effect of additional chil-
tion, many of them include items that reflect the dren on marital satisfaction. From the correlation-
models presented previously; for example, they al analysis it appears that every additional child
may ask specific questions about traditional values has a small detrimental effect on marital satisfac-
and feelings of freedom. These results suggest that tion, but this correlation could be driven primarily
single-item ratings of marital satisfaction may be by the contrast between parents and nonparents.
particularly unreliable and weak. More than likely, We do not know to what extent second, third, and
participants answer defensively or holistically, so subsequent children affect marital satisfaction.
the question does not capture the sources of con- In addition, the studies reviewed here are pri-
flict and interaction that standardized scales mea- marily correlational, and thus we cannot prove
sure. If possible, future researchers should avoid that having children causes marital dissatisfaction.
582 Journal of Marriage and Family

Having children might, for example, create a high- ties that children will bring. Belsky (1988) dem-
er barrier for divorce, so that unhappy couples onstrated that awareness of parenthood’s pitfalls
who might have ended their marriages if they decreases, if not eliminates, the negative influence
were childless instead stay in them for the sake of of children on marital satisfaction. Accordingly,
the children (Hetherington, 1999). This might the research reviewed and quantified here should
lower the average marital satisfaction for parents not be viewed as an indictment of parenthood, but
compared with those who are not parents. There rather as a reminder that relationship stressors are
may be other plausible alternative explanations as minimized through awareness, planning, and com-
well. munication.

Implications and Conclusions REFERENCES


Our results are not intended to imply that married Anderson, S. A., Russell, C. S., & Schumm, W. R.
couples should avoid having children. The nega- (1983). Perceived marital quality and family life cy-
cle categories: A further analysis. Journal of Mar-
tive effects on marital satisfaction are relatively riage and the Family, 45, 127–139.
small and primarily affect certain groups and in- Baumeister, R. F. (1991). Meanings of life. New York:
dividuals in certain life stages (e.g., women with Guilford.
high SES and parents of infants). This knowledge Baumeister, R. F., Catanese, K. R., & Vohs, K. D.
(2001). Is there a gender difference in strength of sex
might help couples set reasonable expectations or drive? Theoretical views, conceptual distinctions, and
make structural changes that will help diminish a review of relevant evidence. Personality and Social
the burden of child care. For example, a more eq- Psychology Review, 5, 242–273.
uitable sharing of responsibilities between spouses Belsky, J., Lang, M., & Huston, T. L. (1986). Sex typing
might improve women’s satisfaction with their and division of labor as determinants of marital
change across the transition to parenthood. Journal
marriages. This would not only reduce the burden of Personality and Social Psychology, 50, 517–522.
on women as caregivers, but might also decrease Belsky, J., & Pensky, E. (1988). Marital change across
women’s resentment and dissatisfaction in their the transition to parenthood. Marriage and Family
marriages. For example, Gjerdingen and Chaloner Review, 12, 133–156.
Blumstein, P., & Schwartz, P. (1983). American couples.
(1994) found that mothers whose husbands partic- New York: BasicBooks.
ipated more in childcare enjoyed better mental Brinley, M. B. (1991). Should you have a baby? In L.
health and higher marital satisfaction. The use of Cargan (Ed.), Marriages and families: Coping with
reliable day care might also help lessen the burden change (pp. 248–253). Englewood Cliffs, NJ: Pren-
on parents and could increase marital satisfaction. tice Hall.
Cohen, J. (1977). Statistical power analyses for the be-
We are hopeful that researchers can identify these havioral sciences. New York: Academic Press.
mitigating factors. Feeney, J., Peterson, C., & Noller, P. (1994). Equity and
It is also important to stress that marital satis- marital satisfaction over the family life cycle. Per-
faction is not everything. A drop in marital sat- sonal Relationships, 1, 83–99.
isfaction may be more than offset by other gains. Frum, D. (2000). How we got here: The 70s, the decade
that brought you modern life (for better or worse).
For example, couples with children may find more New York: Basic Books.
meaning in life (cf. Baumeister, 1991). They may Fukuyama, F. (1999). The great disruption: Human na-
feel increased satisfaction and joy as they raise ture and the reconstitution of social order. New York:
their children, teach them, and learn from them in Free Press.
Gjerdingen, D. K., & Chaloner, K. (1994). Mothers’ ex-
turn. Nevertheless, we do caution individuals not perience with household roles and social support dur-
to decide to have children in order to strengthen ing the first postpartum year. Women and Health, 21,
their marriage. On average, this tactic will back- 57–74.
fire, and all parties will suffer. Glenn, N. D., & Weaver, C. N. (1978). A multivariate
Becoming a parent is likely to be the biggest multisurvey of marital happiness. Journal of Mar-
riage and the Family, 40, 269–282.
and most permanent decision in an individual’s Gough, H. (1991, August). Scales and combinations of
life. Although the practical significance of the scales: What do they tell us, what do they mean?
children–marital satisfaction effect may be small Paper presented at the 99th Annual Convention of the
on average (10%), the consequences of its dele- American Psychological Association, San Francisco,
terious effect may be severe for partners already CA.
Hedges, L. V., & Becker, B. J. (1986). Statistical meth-
in unstable relationships. Partners in both healthy ods in the meta-analysis of research in gender differ-
and unhealthy romantic relationships should be ences. In J. S. Hyde & M. C. Linn (Eds.), The psy-
prepared to realize the potential joys and difficul- chology of gender: Advances through meta-analysis
Parenthood and Marital Satisfaction 583

(pp. 14–50). Baltimore, MD: Johns Hopkins Univer- Rosenthal, R. (1994). Parametric measures of effect
sity Press. size. In H. Cooper & L. V. Hedges (Eds.), The hand-
Hedges, L. V., & Olkin, I. (1985). Statistical methods book of research synthesis (pp. 231–244). New York:
for meta-analysis. Orlando, FL: Academic Press. Russell Sage Foundation.
Hetherington, E. M. (1999). Should we stay together for Twenge, J. M. (2001). Changes in women’s assertive-
the sake of the children? In E. M. Hetherington (Ed.), ness in response to status and roles: A cross-temporal
Coping with divorce, single parenting, and remar- meta-analysis, 1931–1993. Journal of Personality
riage: A risk and resiliency perspective (pp. 93–116). and Social Psychology, 81, 133–145.
Mahwah, NJ: Lawrence Erlbaum. Twenge, J. M. (2002). Birth cohort, social change, and
Hochschild, A. R. (1989). The second shift. New York: personality: The interplay of dysphoria and individ-
Avon Books. ualism in the 20th century. In D. Cervone & W. Mis-
Jenkins, J. M., Rasbash, J., & O’Connor, T. G. (2003). chel (Eds.), Advances in personality science (pp.
The role of the shared family context in differential 196–218). New York: Guilford.
parenting. Developmental Psychology, 39, 99–113.
Twenge, J. M., & Campbell, W. K. (2001). Age and
LeMasters, E. E. (1957). Parenthood as crisis. Marriage
and Family Living, 19, 352–355. birth cohort differences in self-esteem: A cross-tem-
Lopata, H. Z. (1971). Occupation: Housewife. Westport, poral meta-analysis. Personality and Social Psychol-
CT: Greenwood. ogy Review, 5, 321–344.
Putnam, R. D. (2000). Bowling alone: The collapse and White, L. K. (1983). Determinants of spousal interac-
revival of American community. New York: Simon & tion: Marital structure or marital happiness. Journal
Schuster. of Marriage and Family, 45, 511–519.
Rollins, B. C., & Galligan, R. (1978). The developing Wolf, F. M. (1986). Meta-analysis: Quantitative meth-
child and marital satisfaction of parents. In R. M. ods for research synthesis. Beverly Hills, CA: Sage.
Lerner & G. B. Spanier (Eds.), Child influences on Yankelovich, D. (1981). New rules: Searching for self-
marital and family interaction: A life-span perspec- fulfillment in a world turned upside down. New York:
tive. New York: Academic Press. Random House.

You might also like