NBER WORKING PAPER SERIES
MONEY, SEX AND HAPPINESS: AN EMPIRICAL STUDY
David G. Blanchflower
Andrew J. Oswald
Working Paper 10499
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
For helpful suggestions, we thank three referees and James Konow, Richard Layard and Lynne Segal.The views expressed herein are those of the author(s) and not necessarily those of the National Bureau of Economic Research.
?2004 by David G. Blanchflower and Andrew J. Oswald. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including ? notice, is given to the source.
Money, Sex, and Happiness: An Empirical Study
David G. Blanchflower and Andrew J. Oswald
NBER Working Paper No. 10499
JEL No. I1, J3
This paper studies the links between income, sexual behavior and reported happiness. It uses recent
data on a random sample of 16,000 adult Americans. The paper finds that sexual activity enters
strongly positively in happiness equations. Greater income does not buy more sex, nor more sexual
partners. The typical American has sexual intercourse 2-3 times a month. Married people have more
sex than those who are single, divorced, widowed or separated. Sexual activity appears to have
greater effects on the happiness of highly educated people than those with low levels of education.
The happiness-maximizing number of sexual partners in the previous year is calculated to be 1.
Highly educated females tend to have fewer sexual partners. Homosexuality has no statistically
significant effect on happiness. Our conclusions are based on pooled cross-section equations in
which it is not possible to correct for the endogeneity of sexual activity. The statistical results should
be treated cautiously.
David G. Blanchflower
Department of Economics
6106 Rockefeller Hall
Hanover, NH 03755-3514
Andrew J. Oswald
Department of Economics
Money, Sex and Happiness: An Empirical Study
An emerging branch of economics has begun to examine the empirical determinants of
happiness (for example, Easterlin 2001 and Frey and Stutzer 2002). This paper continues that
avenue of research in a different sphere. It focuses on the -- still relatively unexplored -- links
between income, sexual activity and wellbeing.
Human beings are interested in sex. There are also scientific reasons to study it. For
example, recent work by Daniel Kahneman, Alan Krueger, David Schkade, Norbert Schwarz and
Arthur Stone (Kahneman et al 2003) finds, among a sample of 1000 employed women, that sex
is rated retrospectively as the activity that produces the single largest amount of happiness.
Commuting to and from work produces the lowest levels of psychological wellbeing. These two
activities come top and bottom, respectively, of a list of 19 activities.
In this paper we estimate what may be the first econometric happiness equations in which
sexual activity is an independent variable. Like the rest of the recent wellbeing literature, we
study the numbers that people report when asked questions about how happy they feel with life.
Our data set is a randomly selected group of approximately 16,000 Americans. Although, for the
sake of persuasive identification, it would be desirable to have instrumental variables for sexual
activity, in this paper we follow the simpler route of providing single-equation estimates with no
adjustment for possible endogeneity. Our instinct is that solving the endogeneity problem --
working out whether sex causes happiness or causality runs in the reverse direction -- will be
particularly difficult here. Future work will have to return to this issue.
There are limitations to wellbeing statistics. An inquiry in this field also faces the
disadvantage that controlled experiments cannot be done. To understand the connections
between happiness and intimate behavior such as sexual activity is likely to be particularly
difficult. Nevertheless, it seems implausible that happiness can be understood without, in part,
listening to what human beings say about their own lives and levels of happiness. This paper
examines such data.
Surveys have for many years recorded individuals' responses to questions about well-
being. They have been studied by psychologists1, sociologists and political scientists2, and
more recently economists3. As yet, however, there seems to have been little attempt to link
happiness surveys to information on sexual behavior.
There are similarities between our work and the earlier research of Edward Laumann4
and coauthors (Laumann et al 1994, Michael et al 1994). Laumann, Robert Michael and
colleagues collected sexual data on 3400 Americans at the start of the 1990s. Laumann‟s seminal
research does not estimate the kinds of equations we do, nor focus on happiness data, but a
number of our findings on sexual patterns -- particularly on frequency and numbers of partners --
replicate his research team‟s conclusions5.
2. Measuring Happiness
How should we conceptualize „happiness‟? One definition is the degree to which an
individual judges the overall quality of his or her life as favorable (Veenhoven 1991, 1993).
1 Earlier work includes Andrews (1991), Argyle (1989), Campbell, Converse and Rodgers (1976), Campbell (1981), Chen and Spector (1991), Diener (1984), Diener et al (undated, 1999), Douthitt et al (1992), Fox and Kahneman (1992), Frisch et al (1992), Larsen et al (1984), Morawetz et al (1977), Mullis (1992), Shin (1980), Veenhoven (1991, 1993), Van Praag, Bernard and Kapteyn (1973), and Warr (1980, 1990).
2 For example, Inglehart (1990) and Gallie et al (1998).
3 However, see especially the modern work of Andrew Clark, Bruno Frey and Yew Kwang Ng (Clark, 1996; Clark and Oswald, 1994, 1996, 2002a; Frey and Stutzer, 1999, 2000; Ng, 1996, 1997). Blanchflower and Oswald (2004) is on some decades of British and US data. See also Easterlin and Schaeffer (1999), Frank (1985, 1997), Blanchflower (2001), Blanchflower and Oswald (1998, 2000), MacCulloch (1996), Di Tella and MacCulloch (1999), Oswald (1997, 2003), Di Tella et al (2001, 2003), and Winkelmann and Winkelmann (1998). Other recent work has been done by Graham (2001), Graham and Pettinato (2002), Gardner and Oswald (2001), Hollander (2001), Helliwell (2001), Johansson-Stenman et al (2002), McBride (2001) and Selnik (2002). Clark and Oswald (2002b) is a review written for epidemiologists.
4 We thank referees for drawing our attention to this work.
5 Other modern research by economists on sex includes Black et al (2003) and Moffat (2000).
Psychologists draw a distinction between the well-being from life as a whole and the well-being
associated with a single area of life: these they term "context-free" and "context-specific".
Our approach is to assume that people can decide for themselves how happy they feel.
There has been debate in the psychology literature on whether a well-being measure can be -- in
that literature‟s terminology -- reliable and valid. Self-reported measures are recognized to be a
reflection of at least four factors: circumstances, aspirations, comparisons with others, and a
person's baseline happiness or dispositional outlook (e.g. Warr 1980, Chen and Spector, 1991)).
There is known to be a connection between the subjective and the objective. Konow and Earley
(1999) describes evidence that a person‟s recorded happiness levels are correlated with factors
1. Objective characteristics like unemployment.
2. The person‟s recall of positive versus negative life-events.
3. Assessments of the person‟s happiness by friends and family members.
4. Assessments of the person‟s happiness by his or her spouse.
5. Duration of authentic or so-called Duchenne smiles (a Duchenne smile occurs when both the
zygomatic major and obicularus orus facial muscles fire, and human beings identify these as
6. Heart rate and blood-pressure measures responses to stress, and psychosomatic illnesses such
as digestive disorders and headaches.
7. Skin-resistance measures of response to stress
8. Electroencephelogram measures of prefrontal brain activity.
As in Blanchflower and Oswald (2004), we refer readers to the checks on self-reported
happiness statistics that are discussed in Argyle (1989) and Myers (1993), and to psychologists‟
articles on reliability and validity, such as Fordyce (1985), Larsen, Diener, and Emmons (1984),
Pavot and Diener (1993), and Watson and Clark (1991).
Although also based on the General Social Surveys, the Blanchflower and Oswald (2004)
study had no information on sexual activity. Generalizing that paper‟s framework slightly, the
idea used here is that there exists a reported well-being function
r = h(u(y, s, z, t)) + e (1)
where r is some self-reported number or level (perhaps the integer 4 on a satisfaction scale, or
“very happy” on an ordinal happiness scale); u(…) is to be thought of as the person‟s true well-
being or utility; h(.) is a function relating actual to reported well-being; y is real income; s is
sexual activity; z is a set of demographic and personal characteristics; t is the time period; and e
is an error term. As plotted in Figure 1, the function h(.) rises in steps as u increases. It is
assumed, as seems plausible, that u(…) is a function that is observable only to the individual. Its
structure cannot be conveyed unambiguously to the interviewer or any other individual. The
error term, e, then subsumes among other factors the inability of human beings to communicate
accurately their happiness level (your „two‟ may be my „three‟)6. The measurement error in
reported well-being data would be less easily handled if well-being were to be used as an
This approach is somewhat utilitarian, in the Benthamite sense, and is also reminiscent of
the experienced-utility idea advocated by Kahneman et al (1997). The structure of equation 1
makes it suitable for estimation as, for example, an ordered probit or logit. In this way, „true‟
utility is the latent variable, and the subjectivity of responses can be thought of as going into the
6 We accept the social scientist‟s traditional distrust of a person‟s subjective „utility‟. An analogy might be to a time before human beings had accurate ways of measuring people‟s height. Self-reported heights would contain information but be subject to large error. They would predominantly be useful as ordinal data, and would be more valuable when averaged across people than used as individual observations.
error term. For simplicity, this paper also reports various kinds of ordinary least squares
It is possible to view self-reported well-being questions in the psychology literature as
assessments of a person‟s lifetime or expected stock value of future utilities. Equation 1 would
then be rewritten as an integral over the u(…) terms. Nevertheless, this paper will use a
happiness question that seems more naturally interpreted as a flow rather than a stock.
Easterlin (1974, and more recently 1995, 2001) was among the first social scientists to
study data over time on the reported level of happiness in the United States. One of his aims was
to argue that individual well-being is the same across poor countries and rich countries. He
suggests that we should think of people as getting utility from a comparison of themselves with
others. Duesenberry (1949), Hirsch (1976), Scitovsky (1976), Layard (1980), Frank (1985,
1999) and Schor (1998) have argued a similar thesis; see also Cooper, Garcia-Penalosa and Funk
(2001), Ferrer-i-Carbonell (2002) and Keely (1999). A slightly different form of wellbeing data
has been used recently by Ravallion and Lokshin (2001).
This paper draws upon the General Social Surveys of the United States. In order to
obtain information on sexual behavior, income and reported happiness, we use cross-sections
from the years 1988 to 2002 (though, because of missing variables, not every year is available
for regression equations). The key question asked is:
Taken all together, how would you say things are these days -- would you say that you are very
happy, pretty happy, or not too happy? (GSS Question 157)
The same wording has been used in each year. It is known that there is a reasonable amount of
stability in the proportion of people giving different well-being scores. The bulk of survey
respondents place themselves in the middle category „pretty happy‟. Overall, approximately
12% of Americans describe their lives as not too happy, while 56% say they are pretty happy,
and 32% say they are very happy. Most of our statistical work uses the ordering -- not a literal
or exact interpretation of the words.
3. Measuring Sexual Activity
Before reporting the structure of the estimated happiness equations, it is useful to
describe the data set‟s information on sexual activity.
This is a sensitive area about which to question people, but there is a body of knowledge
on how best it can be done (see, for example, chapter 2 of Michael et al, 1994). Respondents in
the GSS are asked how many sexual partners they had in the previous year, how many times they
had sexual intercourse, and the gender of their sexual partners. The survey is confidential and
face-to-face. As with other variables, there is likely to be measurement error in these sexual
data. One bias might stem from bravado; people may wish to appear to the survey interviewer to
be enjoying more sex than they do. Another might stem from modesty or a wish to conceal
extra-marital affairs; this would tend to lead to under-reporting. Our instinct from examining the
data is that, if anything, the former bias dominates, especially among men. Nevertheless, in this
paper we take the numbers at face value and study the implied patterns in American society.
According to our data, Americans have less dramatic sex lives than might have been
imagined from television and other media. Table 1 provides cross-tabulations and describes the
main patterns. The data set here is for a slightly longer span of years than in later regressions,
because not all survey questions are asked in every year.
First, the median American adult has sex approximately 2 or 3 times a month (all the sex
described in this paper refers to sex with a partner; masturbation is not discussed). Among those
aged under 40 years old, the median individual has sex once a week. About 10% of under-40
Americans say they have sexual intercourse at least 4 times a week. Approximately the same
proportion say they are celibate and have no sex. In the whole US population, 6% of adults say
they have sex 4 or more times a week, and 22% report having no sex.
In the over-40 category, the frequency of sex is much lower. Among older women the
median amount of sex is once a month, while for males it is 2-3 times a month (not shown
separately in Table 1). We cannot tell whether this discrepancy is because males, relatively,
have exaggerated memories, or have younger sexual partners, or visit female prostitutes. Among
Americans over 40, 13% of women and 20% of men say they have sexual intercourse twice or
more times a week. A third of over-40s say they are celibate.
Second, the modal and median American had one sexual partner last year. This is true
for more than three quarters of both males and females (see Table 1). Although it might be
thought that young people would have many more sexual partners than the old, only 11% of
under-40 Americans reported themselves as having 3 or more sexual partners in the previous
year. Subdividing this group by gender, among the under-40s 84% of US women and 70% of
US men had at most one sexual partner in the previous year (not shown in the table). For this
age group, 3% of US women and 10% of US men say they had 4 or more sexual partners in the
previous year. Monogamy is dominant among the old. For those over the age of 40, 96% of
women and 89% of men say they had at most one sexual partner in the last 12 months. And 40%
of American females over the age of 40 did not have sexual intercourse in the previous year. The
figure for American males is 20%.
Third, a small proportion of people in the GSS survey report homosexual activity.
Among males, 2.6% say they had a male sexual partner in the previous year. Among females,
just under 1.5% report having had a female partner. About 0.5% of females and 0.5% of males
report themselves as bisexual.
Fourth, Table 1 gives the happiness distributions for different groups in US society.
Although it is not shown explicitly in the table, people who have no sexual activity are
noticeably less happy than average. Happiness scores of people who had no sex last year (and
are therefore classified as neither heterosexual nor homosexual) are: very happy, 23%; pretty
happy, 60%; not at all happy, 17%. This contrasts with the numbers for the whole sample: very
happy, 32%; pretty happy, 56%; not at all happy, 12%. We return to this issue, using regression
equations, in the next section.
Fifth, a few men report large numbers of sexual partners (four males in our sample of
approximately 7000 said they had more than 100 partners in the previous year, whereas no
women said that, and only four women out of nearly 9000 reported having more than 20 partners
in the year). Taking the data set as a whole, almost the only way to make the men‟s and
women‟s answers consistent is for there to be some women in the United States who have
enormous numbers of sexual partners without reporting that fact in our survey data. It is
possible that this is because of the existence of prostitutes. An alternative explanation is that
men tend to overestimate.
Sixth, Table 1 does not find particularly strong correlations between sexual activity and
education, nor between sexual activity and (perceived) high or low income. However, marriage
and sexual frequency are highly correlated; unmarried people say they have much less sex than
those who are married. Nine out of ten married Americans report a single sexual partner in the
4. Happiness Equations with Sexual-Activity Variables
Table 2 reports happiness equations for the United States using pooled cross-section data
from 1988 to 2000. For simplicity, these assume cardinality and are Ordinary Least Squares
estimates where „very happy‟ is coded as 3, „pretty happy‟ is coded as 2, and „not at all happy‟ is