DEMOGRAPHIC RESEARCH
VOLUME 30, ARTICLE 18, PAGES 535546
PUBLISHED 26 FEBRUARY 2014
http://www.demographic-research.org/Volumes/Vol30/18/
DOI: 10.4054/DemRes.2014.30.18
Descriptive Finding
When one spouse has an affair, who is more
likely to leave?
Paula England
Paul D. Allison
Liana C. Sayer
© 2014 Paula England, Paul D. Allison & Liana C. Sayer.
This open-access work is published under the terms of the Creative Commons
Attribution NonCommercial License 2.0 Germany, which permits use,
reproduction & distribution in any medium for non-commercial purposes,
provided the original author(s) and source are given credit.
See http:// creativecommons.org/licenses/by-nc/2.0/de/
Table of Contents
1
Background
536
2
Data and methods
536
3
Results
538
4
Discussion
544
References
546
Demographic Research: Volume 30, Article 18
Descriptive Finding
http://www.demographic-research.org 535
When one spouse has an affair, who is more likely to leave?
1
Paula England
2
Paul D. Allison
3
Liana C. Sayer
4
Abstract
OBJECTIVE
We examine whether having an affair around the time a marriage broke up is associated
with being the person who wanted the divorce more or the person who was left. We
also examine predictors of having an affair around the end of the marriage.
METHODS
We use the National Survey of Families and Households, using each ex-spouse’s
reports of which spouse wanted the divorce more and whether either was having an
affair around the end of the marriage. We combine latent class models with logistic
regression, treating either spouse’s report as a fallible indicator of the reality of whether
each had an affair and who wanted the divorce more.
RESULTS
We find that a spouse having an affair is more likely to be the one who wanted the
divorce more. We find little gender difference in who has affairs preceding divorce.
CONCLUSIONS
Results suggest that it is more common to leave because one is having an affair, or to
have an affair because one has decided to leave, than it is to discover one’s spouse
having an affair and initiate a divorce.
1
We acknowledge funding from National Institute for Child Development of the U.S. National Institutes of
Health and the U.S. National Science Foundation.
2
Corresponding author. New York University, U.S.A. E-Mail: pengland@nyu.edu.
3
University of Pennsylvania, U.S.A.
4
University of Maryland, U.S.A.
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
536 http://www.demographic-research.org
1. Background
Despite trends toward permissiveness in premarital sex, most Americans disapprove of
extramarital sex. In a 1992 American survey, 77% declared it ―always wrong,‖ and only
25% of married men and 15% of married women claimed to have ever engaged in it
(Laumann et al. 1994). Extramarital sex has been found to be more common among
those who married young, divorced previously, had parents who divorced, attend
church less frequently, or are black or male (Treas and Giesen 2000; Amato and Rogers
1997). Extramarital sexual affairs often foreshadow divorces, and longitudinal analysis
suggests that they can be both cause and consequence of other marital problems that
predict divorce (Previti and Amato 2004). While we know that affairs precede some
divorces, past research has not examined which is the more common scenario—that the
person having the affair leaves to form a union with their new-found partner, or that one
spouse discovers the other’s affair and leaves as a result. Our analysis speaks to this
question, using data on a sample of U.S. couples who divorced to assess whether one
spouse having an affair around the end of the marriage predicts which spouse wanted
the divorce more. We also examine the distinct predictors of husbands and wives
having affairs toward the end of marriages that ended in divorce.
2. Data and methods
We use waves 1, 2, and 3 of the National Survey of Families and Households. The
NSFH started from a probability sample of U.S. households, with interviews in
1987–88, 1992–24, and 2001–02. In households containing a married couple at Wave 1,
if spouses separated or divorced, each ex-spouse was interviewed separately at the next
wave and asked about the divorce. We started with a sample of all Wave 1 married
couples in which neither spouse was over 55 and use all 747 cases of separation and/or
divorce that occurred between Waves 1 and 3; we consider marriages to be dissolved at
the point of separation.
Our key dependent variable is which spouse wanted the divorce more. At the next
wave after separation, each ex-spouse was asked which person wanted the divorce
more. We collapsed categories to 3: he wanted it more, they wanted it equally, she
wanted it more. (On item wording and the degree of disagreement between spouses, see
Sayer et al. 2011.) The person who wanted the divorce more might plausibly be the one
who left, or who initiated the divorce, but we do not know this for certain. We use each
partner’s report of who wanted the divorce more, a latent class analysis, and logistic
regression to estimate how much either spouse having an affair around the time the
marriage ended is associated with which spouse wanted the divorce more.
Demographic Research: Volume 30, Article 18
http://www.demographic-research.org 537
Whether each partner was having an affair was measured by two questions asked
in the separate interviews of each ex-spouse: ―Was your husband/wife involved with
someone else just before your marriage ended?‖ and ―Were you involved with someone
else just before your marriage ended?‖ In addition to using these measures, along with
other covariates, to predict who wanted the divorce more, two additional analyses
assess the net association of the other covariates with whether the husband and whether
the wife had an affair before the end of the marriage. (We cannot include marriages that
did not dissolve in our analysis of predictors of affairs because the questions about
affairs were asked only of those who divorced.) These additional two models are also
latent class analyses; they treat his and her report of whether the husband was having an
affair as alternative indicators of whether he actually was, which is the dependent
variable, and analogously for the model predicting whether the wife was having an
affair.
Covariates in our models include spouses’ relative age (wife greater than three
years younger, wife greater than three years older, or within three years), wife’s age at
marriage, the duration of the marriage in years at the point of separation, whether either
spouse was black, whether his/her parents divorced while s/he was growing up, whether
s/he ever cohabited, whether s/he was previously divorced, whether the first birth to the
couple was conceived before marriage, their number of children under 18 at the most
recent wave before the divorce, whether there were children present that are the wife’s
but not the husband’s (and vice versa) at Wave 1, his and her education at Wave 1 in
years, his and her annual earnings at the most recent wave before the divorce, and the
number of months that he and she had been employed (part- or full-time) out of the two
year period ending 12 months before the separation. We also include reports of each
partner’s religious attendance, whether each believes marriage is for life, whether each
reports disagreements about household tasks, and whether each thinks their finances are
unfair to the other spouse at the most recent wave.
Our first regression model predicts who left the marriage from the various reports
of whether each had an affair and other covariates. Although the survey gave each ex-
spouse the option to say that each partner wanted the divorce equally, our modeling
procedures make the simplifying and plausible assumption that one partner wanted the
divorce more than the other, even if only slightly so. Therefore the latent class software
uses the information available to assign each case to one of the two categories—she or
he wanted the divorce more. This means that, when one person said the divorce was
equally wanted, the other partner’s report is used by the model to assess which partner
wanted it more, and the values of other variables are also used, insofar as the model
shows them to be correlated to who wanted the divorce more. Thus, we treat the
outcome as a dichotomy (he left or she left) that is measured with two fallible indicators
of the true status. The true status is modeled as a logistic regression with the several
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
538 http://www.demographic-research.org
covariates. In this regression, reports of affairs by self and other are treated as separate
observed independent variables, not as fallible indicators of latent variables.
Next we seek to predict whether husbands and whether wives had affairs, among
divorcing couples. For this we use two logistic regression models predicting whether he
had and whether she had an affair, respectively, each embedded in a latent class
analysis that treats each spouses’ report of whether he (she) was having an affair as a
fallible indicator of whether he (she) was involved in an affair.
For each regression, we estimated the latent class and logistic regression models
simultaneously by maximum likelihood (with robust standard errors) using methods
described by Yamaguchi (2000) and implemented in Mplus software. Missing data on
the dependent variables were handled by full information maximum likelihood. Missing
data on predictors were multiply imputed under a multivariate normal model, under the
missing-at-random assumption, using PROC MI in SAS. The units of analysis for all
regressions are the 747 divorcing couples. (No imputation is used for the descriptive
proportions in Tables 1 or 2.) All analyses are unweighted.
3. Results
Table 1 shows how prevalent reports of affairs are among these 747 divorcing couples.
If we use each spouse’s own self-report we conclude that only 13.8% of divorces
involve wives having an affair toward the end of the marriage, and 12.9% involve
husbands having an affair, with a trivial gender difference. If we use each spouse’s
report of whether the other was having an affair, we conclude that 33.3% of the men
and 30.8% of the women had affairs, much larger numbers, also with a small gender
difference. In the aggregate, gender differences are trivial, but within couples they are
large as each spouse is more likely to say that the other had an affair than that they
themself did. We suspect that this difference reflects that 1) having an affair is
stigmatized, so individuals under-report their own affairs more than they under-report
their partners’ affairs, and 2) individuals know for certain if they weren't having an
affair, but may suspect that their spouses were having affairs even if they actually
weren't having affairs. Table 1 also shows the proportion of divorces in which each
spouse reported that both had affairs, neither did, just the other person did, or just the
spouse himself or herself did. Whether we use men’s or women’s reports,
approximately 45% of the divorces involved no one having an affair. Reports that both
had an affair are quite rare (4.1% of women and 7.1% of men say this).
Demographic Research: Volume 30, Article 18
http://www.demographic-research.org 539
Table 1: Proportion of divorces after which husbands and wives made various
claims about whether self or spouse had an affair around the end of
the marriage
Responses regarding one spouse's affair (irrespective of report about other spouse):
0.138
0.308
0.333
0.129
Responses regarding both spouse's affairs:
0.041
0.071
0.131
0.341
0.382
0.137
0.446
0.450
Table 2 shows a strong association between claims that a particular spouse had an
affair and that this spouse was the one who wanted the divorce more. For example, men
claim that they wanted the divorce more in 41.7% of divorces in which they admit to
having had an affair, but in only 28.0% of those in which they say their ex-wife had an
affair. Women claim that the husband wanted the divorce more in only 4.2% of the
cases where they admit to having had an affair, but in 69.7% of cases where they say
that their ex-husband had an affair.
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
540 http://www.demographic-research.org
Table 2: Proportion of divorces after which husband said he left, husband said
she left, wife said he left, wife said she left; for all divorces and
subsamples defined by ex-spouses' reports of affairs
Proportion of divorces in which:
Husband said
He Left
Husband
said She Left
Wife said He
Left
Wife Said
She Left
All divorces (747)
0.186
0.297
0.190
0.454
Subsample:
Husband said she had affair (230)
0.280
0.563
0.148
0.348
Husband said he had affair (96)
0.417
0.059
0.282
0.073
Wife said she had affair (103)
0.065
0.176
0.042
0.236
Wife said he had affair (249)
0.504
0.230
0.697
0.286
Note: Statistics not shown for cases where either spouse said both wanted the divorce equally, or did not answer the question. Thus,
the proportion of divorces in a given row for husband said he left plus husband said she left do not add to 1; and the analogous
statement for wives’ reports is true.
Turning to the regressions in Table 3 to see if this association between having an
affair and being the one to want the divorce more holds in multivariate models
containing controls, the basic answer is yes; all associations are in the predicted
direction, and three of the four are significant. Our models predict whether she versus
he wanted the divorce more from both measures of whether he had an affair as well as
both measures of whether she did, controlling for covariates. When either he or she
reports that he had an affair around the end of the marriage, the wife is between 70%
and 80% less likely to have been the one who wanted the divorce more (OR=.241 for
husband’s report; OR=.221 for the wife’s report). Despite much disagreement regarding
whether he had an affair (with only 12.9% of husbands admitting this but 33.3% of
wives saying that their husband did, Table 1), his and her report of his affair both
significantly predict him being the one who wanted the divorce, with similar OR sizes.
In the case of reports of the wife having an affair, only the husband’s report has a
significant positive association with her (versus him) wanting the divorce more
(OR=4.166). Women’s report that they had an affair is not significant but nonetheless
shows the predicted positive sign (OR=1.395). (This corresponds with the one
exception in Table 2 to the overall pattern of bivariate association; the proportion of
marriages in which women said they wanted the divorce more is not higher when she
says she had an affair than when she says he has an affair.) Despite this one exception,
overall, the evidence suggests that when one partner is having an affair and a divorce
occurs, the much more likely scenario is that the partner having an affair leaves, rather
than their partner. This further suggests either that those who want to leave seek out
Demographic Research: Volume 30, Article 18
http://www.demographic-research.org 541
affairs or that those who have affairs often decide to leave to be able to form a union
with the new partner. Initiating a divorce because you discover a partner having an
affair appears much less common.
Table 3: Odds ratios from latent class logistic regression model predicting that
wife (versus husband) wanted divorce more among couples who
divorced
Independent Variables
OR
Two-tailed P-value
Husband Reports He had Affair
0.241
*
0.014
Wife Reports Husband had Affair
0.221
***
0.000
Wife Reports She had Affair
1.395
0.582
Husband Reports Wife had Affair
4.166
*
0.018
Relative age (reference=within 3 years)
Wife >3 years older than husband (0, 1)
0.803
0.705
Wife >3 years younger than husband (0, 1)
1.968
*
0.038
Wife's age at marriage
0.942
*
0.048
Duration of marriage (years)
0.988
0.663
Either spouse Black (0, 1)
2.989
*
0.026
Husband's parents divorced (0, 1)
0.882
0.704
Wife's parents divorced (0, 1)
1.374
0.367
Husband ever cohabited (0, 1)
0.757
0.439
Wife ever cohabited (0, 1)
1.904
0.068
Husband previously divorced (0, 1)
0.662
0.248
Wife previously divorced (0, 1)
4.297
**
0.005
First birth conceived before marriage (0, 1)
1.373
0.446
Number of Children < 18
0.982
0.893
Child(ren) present not husband's (0, 1)
0.866
0.771
Child(ren) present not wife's (0, 1)
0.327
*
0.033
Husband's Education (years)
0.922
0.223
Wife's Education (years)
0.971
0.708
Husband's Annual Earnings ($)
1.020
0.912
Wife's Annual Earnings ($)
1.094
0.447
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
542 http://www.demographic-research.org
Table 3: (Continued)
Independent Variables
OR
Two-tailed P-value
# months husband employed last 2 years
1
1.023
0.175
# months wife employed last 2 years
1
1.000
0.979
Husband's Religious Attendance
1.057
0.536
Wife's Religious Attendance
0.886
0.166
Husband Believes Marriage is for Life
0.679
0.251
Wife Believes Marriage is for Life
1.477
0.258
Husband Reports Disagreements re: Household Tasks
0.977
0.867
Wife Reports Disagreements re: Household Tasks
1.025
0.845
Husband Thinks Finances Unfair to Wife
1.000
1.000
Wife Thinks Finances Unfair to Husband
1.723
0.227
***=p<.001, **=p<.01, *= p<.05, +=p<.10, two-tailed tests.
1
The two year period ends 12 months before the month of the separation.
The models also show other predictors of who wanted the divorce more (net of
whether either spouse was having an affair). The wife is approximately twice as likely
to be the one wanting the divorce more when she is younger than the husband
(OR=1.968); put another way, the husband is much less likely to leave a wife who is
younger than himself. In marriages where either spouse is black, the woman is
approximately three times as likely to be the one who is leaving than is true for spouses
of other races (OR=2.989). (Results not shown reveal that in most of these couples
where at least one partner is black, both partners are black, and in cases where only one
is black, it is generally the husband.) Women who were married and divorced from a
previous husband are approximately four times more likely to be the one wanting the
divorce more in the more recent marriage (OR=4.297); but whether men were divorced
before does not affect whether they were the one wanting the current divorce. This may
be because women initiate a much higher proportion of divorces than men;
consequently, the fact that a woman is previously divorced may be a stronger indicator
of general willingness to leave a marriage. If the wife was older (than other wives)
when she entered this marriage, she is less likely to be the one who wanted the divorce
more (OR=.942). She is over 65% less likely to be the one wanting the divorce
(OR=.327) when the household contains one or more of his children that are not also
hers.
We now turn to Table 4, with separate models predicting whether the husband and
whether the wife had an affair near the end of the marriage. Only one variable is a
Demographic Research: Volume 30, Article 18
http://www.demographic-research.org 543
significant net predictor of whether the wife has an affair; women who married at older
ages are less likely to have affairs. The model predicting whether men have affairs
contains more significant predictors: Men are almost nine times as likely to have an
affair if their wife is more than three years older than themselves (OR=8.706), and
about 70% less likely to do so if their wife is more than three years younger than
themselves (OR=.289). For each year the marriage has lasted, men are 6.2% more likely
to have had an affair (OR=1.062). Men are nearly four times more likely to have affairs
in couples where one or both partners are black (OR=3.999); given that husbands are
black in most of these couples, this suggests that black men are more likely than other
men to have affairs. Men are nearly 25% less likely to have an affair if the couple has
children under 18 (OR=.766). The remaining significant coefficients have no obvious
interpretation; men are less likely to have affairs if the wife reports that they have had
disagreements about household tasks and less likely if the wife cohabited with anyone
before the marriage.
Table 4: Odds ratios from latent class logistic regression models predicting
husband having affair and wife having affair, among couples who
divorced
Predicting Husbands'
Affairs
Predicting Wives'
Affairs
Independent Variables
OR
Two-tailed
P-value
OR
Two-tailed
P-value
Relative age (reference=within 3 years)
Wife >3 years older than husband (0, 1)
8.706
**
0.006
0.348
0.670
Wife >3 years younger than husband (0, 1)
0.289
**
0.006
1.621
0.179
Wife's age at marriage
0.948
0.201
0.890
*
0.025
Duration of marriage (years)
1.062
**
0.006
0.965
0.104
Either spouse Black (0, 1)
3.999
**
0.007
0.828
0.752
Husband's parents divorced (0, 1)
0.857
0.662
1.397
0.419
Wife's parents divorced (0, 1)
0.832
0.551
0.805
0.569
Husband ever cohabited (0, 1)
2.056
0.172
1.155
0.763
Wife ever cohabited (0, 1)
0.523
+
0.090
1.107
0.771
Husband previously divorced (0, 1)
1.680
0.264
0.861
0.784
Wife previously divorced (0, 1)
1.266
0.681
0.967
0.957
First birth conceived before marriage (0, 1)
1.026
0.942
1.070
0.856
Number of Children < 18
0.766
+
0.063
1.094
0.552
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
544 http://www.demographic-research.org
Table 4: (Continued)
Predicting Husbands'
Affairs
Predicting Wives'
Affairs
Independent Variables
OR
Two-tailed
P-value
OR
Two-tailed
P-value
Child(ren) present not husband's (0, 1)
0.527
0.188
1.436
0.613
Child(ren) present not wife's (0, 1)
0.610
0.431
0.926
0.912
Husband's Education (years)
0.984
0.821
1.073
0.318
Wife's Education (years)
0.944
0.512
0.934
0.545
Husband's Annual Earnings ($)
0.897
0.347
1.020
0.891
Wife's Annual Earnings ($)
1.146
0.288
1.169
0.175
# months husband employed last 2 years
1
1.025
0.157
1.007
0.697
# months wife employed last 2 years
1
1.011
0.520
0.979
0.194
Husband's Religious Attendance
0.949
0.533
1.001
0.993
Wife's Religious Attendance
1.171
0.131
0.868
0.260
Husband Believes Marriage is for Life
0.549
+
0.088
0.660
0.589
Wife Believes Marriage is for Life
0.954
0.879
1.069
0.900
Husband Reports Disagreements re:
Household Tasks
1.210
0.254
0.925
0.714
Wife Reports Disagreements re: Household
Tasks
0.738
*
0.036
1.057
0.691
Husband Thinks Finances Unfair to Wife
1.404
0.475
0.362
0.232
Wife Thinks Finances Unfair to Husband
1.481
0.320
0.948
0.917
***=p<.001, **=p<.01, *= p<.05, +=p<.10, two-tailed tests.
1
The two year period ends 12 months before the month of the separation.
4. Discussion
Divorces are sometimes preceded by affairs. By either men’s or women’s reports,
almost half (45%) of divorces involve neither person having an affair, but it is rare for
either spouse to report that both had an affair. It is difficult to estimate the proportion of
cases in which men versus women had an affair because both sexes are much more
likely to think the other was having an affair than to state that they were; 13–14% of
either men or women say that they themselves were having an affair, whereas
approximately a third of each sex said that they think their partner had an affair. Past
Demographic Research: Volume 30, Article 18
http://www.demographic-research.org 545
findings not limited to divorcing couples suggest that men have more affairs than
women. Our finding of no aggregate gender difference between men or women having
affairs around the end of a marriage suggests that women who do have affairs may be
more likely to leave than men; future research should study this directly.
Among spouses who divorced, when the woman is younger than the man, he is
less likely to have had an affair, and he is less likely to be the one who wanted the
divorce; by contrast, men have affairs more and are disproportionately the ones wanting
the divorce more when the wife is older than the husband. Men’s preference for
younger wives has been found in studies of mate selection (England and McClintock
2009); our results suggest that this preference affects decisions to have affairs and
divorce as well.
We find that the person who has the affair is most often the one who wanted the
divorce more. We have avoided causal language, as we cannot be sure whether this
means that one is more open to affairs because of having decided to leave a marriage,
that having an affair makes one decide to leave to live with the new partner, or that both
affairs and divorce are driven by the same antecedents. However, our results do
strongly suggest that the usual scenario involving affairs is not that one spouse
discovers the other having an affair and leaves because of it.
England, Allison & Sayer: When one spouse has an affair, who is more likely to leave?
546 http://www.demographic-research.org
References
Amato, P.R. and Rogers, S.J. (1997). A longitudinal study of marital problems and
subsequent divorce. Journal of Marriage and the Family 59(3): 612–624.
doi:10.2307/353949.
England, P. and McClintock, E.A. (2009). The gendered double standard of aging in
U.S. marriage markets. Population and Development Review 35(4): 797–816.
doi:10.1111/j.1728-4457.2009.00309.x.
Laumann, E.O., Gagnon, J.H., Michael, R.T., and Michaels, S. (1994). The social
organization of sexuality: Sexual practices in the United States. Chicago:
University of Chicago Press.
Previti, D. and Amato, P.R. (2004). Is infidelity a cause or a consequence of poor
marital quality? Journal of Social and Personal Relationships 21(2): 217–30.
doi:10.1177/0265407504041384.
Sayer, L., England, P., Allison, P.D., and Kangas, N. (2011). She left, he left: How
employment and satisfaction affect women’s and men’s decisions to leave
marriages. American Journal of Sociology 116(6): 1982–2018. doi:10.1086/
658173.
Treas, J. and Giesen, D. (2000). Sexual infidelity among married and cohabiting
Americans. Journal of Marriage and the Family 62(1): 48–60. doi:10.1111/
j.1741-3737.2000.00048.x.
Yamaguchi, K. (2000). Multinomial logit latent-class regression models: An analysis of
the predictors of gender-role attitudes among Japanese women. American
Journal of Sociology 105(6): 1702–1740. doi:10.1086/210470.