Economics Letters 121 (2013) 458–462

Contents lists available at ScienceDirect

Economics Letters
journal homepage: www.elsevier.com/locate/ecolet

New facts on infidelity✩

Effrosyni Adamopoulou ∗
Bank of Italy, Research Department, Structural Economic Analysis, Labor Market Division, Via Nazionale 91, 00184, Rome, Italy

highlights
• We establish new empirical facts in line with the recent theoretical literature.
• People that are unfaithful to their current partner were unfaithful also in the past.
• Infidelity displays seasonality, peaking around summertime.
• In the US young males and females are equally likely to be unfaithful.
• In this context, socioeconomic status does not seem to be a driver of infidelity.

article info abstract

Article history: We establish new empirical facts, in line with the recent theoretical literature on infidelity. Infidelity
Received 13 February 2013 displays seasonality and state dependence. In the US socioeconomic status is not a driver of infidelity and
Received in revised form females and males are equally likely to be unfaithful.
10 September 2013
© 2013 Elsevier B.V. All rights reserved.
Accepted 20 September 2013
Available online 27 September 2013

JEL classification:
A13
D10
J12

Keywords:
Infidelity
Multiple dating
Seasonality
State dependence

1. Introduction
Since the pioneering work of Becker (1974) there have been
many studies trying to model different aspects of family behav-
ior from an economic point of view. Among them is the one of Fair
(1978) that models infidelity using a model of time allocation. He
first develops a theory of extramarital affairs and then tests it em-
pirically using data from magazine surveys. These data have certain
limitations regarding non-representability of the sample as well as
the scarcity of information on economic variables (e.g., wage). This
✩ First version: October 2012. I am grateful to Nezih Guner for his valuable advice
and guidance. Many thanks to Federico Cingano, Michael Haliassos, Francesco
Manaresi, and an anonymous referee for useful suggestions. This research was
initiated and mainly conducted while at the Universidad Carlos III de Madrid.
The views expressed in the paper are those of the author and do not necessarily
correspond to those of the Bank of Italy. All the remaining errors are mine.
∗ Tel.: +39 0647922594.
E-mail address: Effrosyni.Adamopoulou@esterni.bancaditalia.it.
0165-1765/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.econlet.2013.09.025
is why there have been various follow-up studies, most of them
very recent, trying to check the validity of his results (Wells, 2003;
Li and Racine, 2004), providing new interpretations (Elmslie and
Tebaldi, 2008; Smith, 2012), or extending his model allowing for
additional variables (Treas and Giesen, 2000; Potter, 2011; Brooks
and Monaco, 2012). None of these studies however addresses the
issue of state dependence.
Infidelity is related to divorce, which in turn might affect chil-
dren’s welfare (Weiss and Willis, 1985). Moreover, infidelity might
have health implications in terms of sexually transmitted diseases.
Recent studies (Pongou, 2009a; Pongou and Serrano, 2013) pro-
pose a theoretical model of (in)fidelity networks as the mechanism
underlying the gender gap in HIV/AIDS prevalence. This is why un-
derstanding the determinants of infidelity matters.
In this paper we use the information on the whole history of
infidelity events of the respondents and we show that those who
cheat have done so also in the past. We find that infidelity displays
seasonality, peaking in the summer. In line with the theoretical
predictions in Pongou and Serrano (2013), we find that among
 E. Adamopoulou / Economics Letters 121 (2013) 458–462
Fig. 1. Seasonality in infidelity behavior, 1996–2001.
unfaithful individuals in the US, about half are men and about half
are women. We also show that socioeconomic status does not seem
to be related with the infidelity behavior of young adults in the US.
2. Data
We use data from the Longitudinal Study of Adolescent Health
(Add Health). Add Health is a survey of a nationally representative
sample of adolescents in grades 7–12 in the United States during
the 1994–95 school year.1 In Wave I the study started with an in-
school questionnaire that was administered to more than 90,000
students from 80 high schools and 52 middle schools. A subsample
of them (around 20,000) were also asked to complete in-home
interviews and were followed in subsequent waves (II, III, and
IV). Wave IV was conducted in 2008, when the sample was aged
24–34. The respondents in Wave IV had to answer questions about
their educational background, employment, tobacco and alcohol
consumption, criminal activities as well as their marital status and
details regarding their relationship with their current partner.2 In
particular, the information about infidelity comes from the answer
to the question ‘‘During the time you and your current partner
1 This research uses data from Add Health, a program project directed by Kathleen
Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen
Mullan Harris at the University of North Carolina at Chapel Hill, and funded by
grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child
Health and Human Development, with cooperative funding from 23 other federal
agencies and foundations. Special acknowledgment is due to Ronald R. Rindfuss
and Barbara Entwisle for assistance in the original design. The information on
how to obtain the Add Health data files is available on the Add Health website
(http://www.cpc.unc.edu/addhealth). No direct support was received from grant
P01-HD31921 for this analysis.
2 These detailed questions were referring to only one current partner. If there
were multiple current partners, priority was given as follows: marriage partner,
cohabitation partner, pregnancy partner, dating partner. If two or more partners
fell in the same type of relationship, the longer/longest relationship was selected. If
two or more partners fell in the same type of relationship, and they were of the same
duration, then the respondent was asked to pick the partner they cared about the
most. If there were no current partners then the most recent partner was selected.
If there were no current partner and no most recent partner, end dates for each
marriage, cohabitation, and relationship with a pregnancy were reviewed to select
the one partner with the most recent end date. If two or more partners had the same
end date, the longer/longest relationship was selected.
459
have had a sexual relationship have you ever had any other sexual
partners?’’
Our objective is to combine the information about current
infidelity behavior with past events of infidelity. To do so we use
data from Wave III. Wave III in-home interviews took place during
the years 2001–2002 when the respondents were aged 18–26. In
Wave III the respondents had to list all their current and previous
sexual relationships with detailed information on the starting and
ending date, whether they cohabited and how long, when they got
married, etc. Using this information we create a monthly panel
for the years 1996–2001 which allows us to detect infidelity. If
the respondent had more than one relationship in a given month,
we keep the one with the longest overall duration and treat the
event as infidelity. Our definition of infidelity is not restricted
to extramarital affairs but includes multiple dating and infidelity
among cohabiting couples.
Fig. 1 shows the average incidence of infidelity among mar-
ried, cohabiting, and dating couples at each month during the years
1996–2001. Infidelity in all types of relationships displays season-
ality peaking during summertime. A simple explanation for this
feature might be that most people travel during summer, which
in turn facilitates cross-space cheating. Cheating is less likely to be
detected when an individual chooses partners in different places
rather than in the same area. This interpretation is in line with the
predictions of the model of strategic cross-ethnic cheating devel-
oped in Pongou (2009b).3 , 4
Our final sample consists of respondents with non-missing re-
lationship history (from Wave III) and information about their cur-
rent partner (from Wave IV). We first consider those respondents
who started being in a relationship with their current partner in
2001 or after (the reference point is the Wave III interview date)
3 In Pongou (2009b) limited communication across ethnic groups fosters cross-
group anonymity, providing incentives for individuals to choose their sexual
partners across groups rather than within, in order to minimize the probability of
being caught.
4 Another pattern that Fig. 1 reveals is that infidelity has increased over time
among dating couples but remained relatively constant among cohabiting and
married couples. A possible explanation for this pattern may be that as respondents
age, those that choose not to get married or cohabit are those most prone to
infidelity. In other words, the increase in the incidence of infidelity among dating
couples might be due to a composition effect.
460 E. Adamopoulou / Economics Letters 121 (2013) 458–462

Table 1
Individual characteristics in Wave IVa, b .
Characteristics Cheaters Non-cheaters

% women 43.20 51.06

Mean age 28.35 28.33
(1.56) (1.62)
% African American 22.89 13.07
Mean beauty 3.47 3.45
(0.82) (0.83)
Mean religiosity 1.22 1.44
(1.37) (1.50)
Mean earnings in thousand dollars 33.61 34.72
(39.13) (42.97)
% with college education 29.01 34.07
% with children 82.97 89.48
% dissatisfied with current partner 54.85 24.75
Mean current relationship duration in years 3.43 3.34
(2.56) (2.58)
Mean # other relationships <6 months after 2001 4.82 2.07
(10.03) (4.68)
% ever cheated before 29.24 20.97
Number of observations 1146 4068
a
Individuals with non-missing relationship history and information about their current partner.
b
Corrected for survey design.

and we relate infidelity in the current relationship with the infi- Table 2
Determinants of infidelity in Wave IV.
delity history before the Wave III interview date.5
In Wave IV, 21.5% of the respondents report that they have (1) (2) (3)
cheated on their current partner at least once. The percentage of Married −0.114*** −0.116*** −0.109***
cheaters reduces to 12.9% among those that are currently mar- (0.018) (0.024) (0.023)
ried. Table 1 shows the descriptive statistics for cheaters and non- Female −0.053*** −0.066*** −0.058***
(0.018) (0.020) (0.021)
cheaters. In accordance with other surveys (see Smith, 2012), the
Age 0.287* 0.224 0.145
percentage of men who report infidelity in Wave IV (56.8%) is larger (0.148) (0.169) (0.166)
than the one of women. This feature though can be due to gen- Age squared −0.005* −0.004 −0.003
der differences in reporting infidelity.6 Indeed, when we use the (0.003) (0.003) (0.003)
Race 0.146*** 0.120*** 0.113***
relationship history in order to detect infidelity before Wave III,
(0.025) (0.029) (0.028)
cheaters are more likely to be women (55.8%). By pooling the in- Beauty 0.006 0.002 0.003
fidelity reports from Wave IV with the incidence of contempora- (0.009) (0.010) (0.010)
neous relationships before Wave III one can argue that men and Religiosity −0.015*** −0.012** −0.011*
women are equally likely to be unfaithful. This is in accordance (0.005) (0.006) (0.006)
Earnings −0.0001 −0.0001 −0.0002
with the model of Pongou and Serrano (2013). Their findings imply (0.0002) (0.0002) (0.0002)
that in societies where women have achieved more equalities with College −0.018 −0.012 −0.013
men as in the US, among unfaithful individuals, about half should (0.017) (0.018) (0.019)
be men and half should be women. Children −0.105*** −0.086***
(0.030) (0.029)
Those who cheat are more likely to be African Americans, less Dissatisfied with partner 0.190*** 0.183***
religious, less educated, and poorer with respect to non-cheaters. (0.022) (0.022)
They are less likely to have children than non-cheaters, they are Relationship duration 0.013*** 0.017***
more dissatisfied from their current partner, and the average dura- (0.004) (0.004)
Mean # other relationships 0.009***
tion of their current relationship is longer. Regarding the relation-
<6 months (0.002)
ship history the differences are striking. Cheaters are much more Ever cheated before 0.081***
likely to have cheated in the past and they report having numerous (0.021)
relationships that lasted less than 6 months.7 All the variables are R2 0.059 0.124 0.144
explained in Table A.1 in Appendix. Number of observations 4606 3520 3472
Cross sectional weights used, state dummies included.
*
p < 0.1.
**
p < 0.05.
5 As a robustness check, we repeated the analysis using as a reference point the
***
p < 0.01 (robust s.e. clustered at school level).
year 2000 and the results remained practically unchanged.
6 An alternative explanation is provided in Pongou (2009a) and Pongou and 3. Regression analysis
Serrano (2013). If the optimal number of partners is greater for men than for women
(either because men have stronger preferences for sex than women, or because
women are more severely punished for infidelity than men), this might lead to men
In this section, we estimate linear probability models in order
being more unfaithful than women. to study the determinants of infidelity. We first present the results
7 In Wave IV, the respondents answered the following question regarding short for all types of relationships (Table 2) and then separately for mar-
term relationships: Not counting the persons you have ever married, the persons ried and unmarried (Tables 3 and 4). We start with a regression
you have ever lived with for one month or more, the persons you have ever had a including only individual characteristics, we then add relation-
relationship that resulted in a pregnancy, the persons with whom you are currently
having a relationship, and the persons with whom you had a relationship that lasted
ship specific variables, and lastly variables regarding the relation-
six months or more, with how many other people have you had a romantic or sexual ship history and previous infidelity experience. In all regressions
relationship that lasted less than 6 months since 2001? we also control for state fixed effects in order to capture regional
 E. Adamopoulou / Economics Letters 121 (2013) 458–462 461

Table 3 when we include only individual characteristics as controls. Being

Determinants of infidelity among married in Wave IV. married or being female decreases the incidence of infidelity.8 Sim-
(1) (2) (3) ilarly, being more religious is related with lower infidelity. There
Female −0.038* −0.053** −0.052** are racial differences as well. Being African American significantly
(0.021) (0.021) (0.021) increases the likelihood of being unfaithful in the relationship. The
Age 0.137 0.102 0.085 beauty of the respondent, as assessed by the interviewer in Wave
(0.167) (0.165) (0.166) IV, does not seem to matter.9 In column 2 we add relationship
Age squared −0.003 −0.002 −0.002 specific variables, namely whether the respondent has a child, the
(0.003) (0.003) (0.003)
relationship duration, and the degree of dissatisfaction with the
Race 0.122*** 0.107** 0.096**
(0.045) (0.043) (0.044) current partner. Given that these variables might be endogenous,
Beauty 0.017 0.018 0.017 we do not interpret the effects as causal. We do include them,
(0.012) (0.012) (0.012) however, in order to compare our estimates with the ones in Fair
Religiosity −0.017** −0.014** −0.012* (1978). Similarly to the results in Fair (1978), we find that higher
(0.007) (0.007) (0.007)
dissatisfaction with the partner is positively related with the in-
Earnings 0.0002 0.0002 0.0002
(0.0003) (0.0003) (0.0003)
cidence of infidelity and that infidelity is more likely to occur in
College −0.011 −0.007 −0.011 relationships of long duration. Moreover, having a child is related
(0.022) (0.022) (0.022) with lower infidelity.
Children −0.066* −0.058 We now add variables regarding the relationship history and
(0.037) (0.037) past infidelity experiences of the individuals (Table 2, column 3).
Dissatisfied with partner 0.186*** 0.185***
Having a history of multiple short term relationships significantly
(0.037) (0.035)
Relationship duration 0.004 0.008*
increases the incidence of infidelity. Moreover, those who have
(0.005) (0.005) cheated in a past relationship they are more likely to cheat in
Mean # other relationships 0.008** their current relationship as well. This suggests that there is state
<6 months (0.003) dependence in the infidelity behavior. Learning might be one form
Ever cheated before 0.061** of state dependence. If an individual cheats many times, he/she
(0.029)
might learn the way to keep it secret from his/her partner.
R2 0.061 0.113 0.131
Number of observations 1694 1570 1561
We also perform the analysis for married and unmarried
separately (Tables 3 and 4) and we obtain similar results. For
Cross sectional weights used, state dummies included.
unmarried individuals having cheated in a previous relationship
*
p < 0.1.
**
p < 0.05. significantly increases the probability of cheating in the current
***
p < 0.01 (robust s.e. clustered at school level). relationship. Even for married individuals past infidelity has a
significant effect on current infidelity (although the coefficient is
Table 4 smaller in magnitude).
Determinants of infidelity among unmarried in Wave IV. We do not find evidence that economic variables matter.10
Earnings and education do not seem to have any statistically signif-
(1) (2) (3)
icant effect in almost all specifications. The coefficient of income is
Female −0.058** −0.070** −0.055* statistically significant only in the subsample of unmarried and it is
(0.024) (0.029) (0.030)
negative. The relationship between socioeconomic status and infi-
Age 0.283 0.172 0.028
(0.214) (0.262) (0.253) delity may differ across contexts. Fortson (2008) using the number
Age squared −0.005 −0.003 −0.0005 of times that male respondents were away from home as a proxy of
(0.004) (0.005) (0.004) infidelity finds that more educated men in sub-Saharan Africa are
Race 0.145*** 0.113*** 0.110*** more likely to have concurrent relationships. Hence, in developing
(0.031) (0.040) (0.041) countries socioeconomic status is positively related with infidelity,
Beauty 0.002 −0.007 −0.004
(0.013) (0.015) (0.015)
while in developed countries, like the US, there does not seem to
Religiosity −0.014** −0.012 −0.014 be any relationship between socioeconomic status and infidelity,
(0.007) (0.010) (0.010) and in the case of unmarried, it turns out to be negative.
Earnings −0.0003 −0.0004** −0.0006***
(0.0002) (0.0002) (0.0002) 4. Conclusions
College −0.024 −0.010 −0.012
(0.024) (0.029) (0.031)
Children −0.132*** −0.107*** Infidelity is an aspect of human behavior that cannot be easily
(0.040) (0.039) observed. Studies of infidelity have to deal with various data
Dissatisfied with partner 0.188*** 0.180*** limitations ranging from non-representability of the sample to the
(0.026) (0.027) lack of information on key variables. In this paper we establish
Relationship duration 0.017*** 0.021*** new empirical facts on infidelity using data on the whole history
(0.006) (0.006)
of romantic relationships of young adults in the US. We find that
Mean # other relationships 0.010***
<6 months (0.002)
infidelity displays state dependence and seasonality. We also find
Ever cheated before 0.094*** that in the US, males and females are equally likely to be unfaithful
(0.034) and that in this context, socioeconomic status does not seem to
R2 0.047 0.117 0.143 matter. Our findings are in line with the predictions of the recent
Number of observations 2912 1950 1911 theoretical literature on infidelity.
Cross sectional weights used, state dummies included.
*
p < 0.1.
**
p < 0.05.
8 The dependent variable in the regressions is self-reported infidelity; this is why
***
p < 0.01 (robust s.e. clustered at school level).
the variable female appears with a negative coefficient.
9 Using the body mass index as a proxy of physical appearance produces very
differences, e.g., in social norms about infidelity or the availability similar estimates.
of opportunities (sex ratio and number of bars etc.). Table 2, col- 10 This may also be due to the relatively young age of the individuals in the sample
umn 1 shows the results for infidelity in all types of relationships (around 28 years old).
462 E. Adamopoulou / Economics Letters 121 (2013) 458–462

Table A.1
Definition of variables.
Variable Type Values

0 if currently not married
Married Binary
1 if currently married

0 if male
Gender Binary
1 if female
Age Continuous In years

0 if not African American
African American Binary
1 if African American

 1 very unattractive
Beauty 2 unattractive


(Assessed by Ordinal 3 about average
the interviewer)
4 attractive



5 very attractive
0 never

1 a few times


Religiosity 2 several times


(Attendance in Ordinal 3 once a month
religious services) 4 two or three times a month




5 once a week


6 more than once a week
Total Earnings Continuous In thousand $

0 if no college or college not completed
College Binary
1 if completed college

0 if no child
Child Binary
1 if at least one child

Dissatisfied with 0 very happy
Binary
current partner 1 fairly happy/not too happy
Relationship duration Continuous In years
# of relationships after 2001
Continuous
lasting less than 6 months 
0 if never cheated before 2001
Ever cheated before 2001 Binary
1 if cheated before 2001

Appendix Li, Q., Racine, J., 2004. Predictor relevance and extramarital affairs. Journal of
Applied Econometrics 19, 533–535.
Pongou, R., 2009a. An economic theory of fidelity in network formation. Mimeo.
See Table A.1.
Pongou, R., 2009b. Anonymity and infidelity: ethnic identity, strategic cross-ethnic
sexual network formation, and HIV/AIDS in Africa. Mimeo.
References Pongou, R., Serrano, R., 2013. Fidelity networks and long-run trends in HIV/AIDS
gender gaps. American Economic Review: Papers & Proceedings 103, 298–302.
Becker, G., 1974. A theory of marriage: part II. Journal of Political Economy 82, Potter, J., 2011. Reexamining the economics of marital infidelity. Economics Bulletin
S11–S26. 31, 41–52.
Brooks, T., Monaco, K., 2012. Your cheatin’ heart: joint production, joint Smith, I., 2012. Reinterpreting the economics of extramarital affairs. Review of
consumption and the likelihood of extramarital sex. Applied Economics Letters Economics of the Household 10, 319–343.
20, 272–275. Treas, J., Giesen, D., 2000. Sexual infidelity among married and cohabiting
Elmslie, B., Tebaldi, E., 2008. So, what did you do last night? the economics of
americans. Journal of Marriage and the Family 62, 48–60.
infidelity. Kyklos 391–410.
Weiss, Y., Willis, R., 1985. Children as collective goods and divorce settlements.
Fair, R., 1978. A theory of extramarital affairs. Journal of Political Economy 86,
45–61. Journal of Labor Economics 5, 268–292.
Fortson, J., 2008. The gradient in sub-saharan Africa: socioeconomic status and Wells, C., 2003. Retesting Fair’s (1978) model on infidelity. Journal of Applied
HIV/AIDS. Demography 45, 303–322. Econometrics 18, 237–239.