Social Indicators Research

, Volume 116, Issue 1, pp 115–135

Sex and the Pursuit of Happiness: How Other People’s Sex Lives are Related to our Sense of Well-Being

Article

DOI: 10.1007/s11205-013-0267-1

Cite this article as:
Wadsworth, T. Soc Indic Res (2014) 116: 115. doi:10.1007/s11205-013-0267-1

Abstract

A growing literature suggests that income, marriage, friendship, sex, and a variety of other factors influence self-reported happiness. Why these characteristics matter has been less examined. Scholars have recently demonstrated that part of the effect of income is relative. More income makes people happier, in part, because it sets them above their peers. Until now, the role of relative comparison in the study of happiness has been limited to income. The current work extends this focus to another activity—sex. Using GSS data, I examine how respondents’ frequency of sex, as well as the average sexual frequency of their cohort, influences their happiness. The findings suggest that happiness is positively correlated with their own sexual frequency, but inversely correlated with the sexual frequency of others.

Keywords

Happiness Subjective well-being Sex Reference groups Social comparison 

Over the last several decades there has been a growing body of research on the social and psychological determinants of happiness. The role that some factors, such as marital status (Haring-Hidore et al. 1985; Veenhoven 1994), physical health (Dolan et al. 2008), sex (Blanchflower and Oswald 2004), and friendship networks (Myers 1999), play in increasing happiness has been consistent with common cultural understandings of what leads to the “good life”. Yet, the lack of influence of other factors, such as parenthood (McLanahan and Adams 1989), intelligence (Layard 2005) and physical appearance (Keltner and Harker 2001), has been less consistent with widespread cultural assumptions.

One of the correlates of happiness that has been most intriguing to economists, psychologists, and sociologists, has been income. A number of individual-level cross-sectional studies have shown a positive relationship between income and self-reported happiness (Easterlin 1974, 2001, 2003; Graham and Pettinato 2002), yet longitudinal studies have demonstrated that increases in income are not accompanied by increases in happiness (Dolan et al. 2008). Macro-level research has offered analogous findings. As nations’ GDPs increase, national “happiness levels” tend to remain stable (Layard 2005). A number of explanations have been offered for this apparent contradiction. Some have suggested that growth in income is often accompanied by increases in workload, stress, or other negative factors which offset the positive effects of increased income (Fischer 2008). Others have proposed that the causal direction of the cross-sectional relationship between income and happiness has been misspecified—income and happiness are related not because more money makes people happy, but because happy people are more productive, resulting in higher earnings. This being the case, we would not expect to see a relationship across time when analyzing longitudinal data.

These explanations notwithstanding, perhaps the most controversial suggestion, and the one that has received the most attention, is that the relationship between income and happiness is based on relative, rather than absolute, income. From this perspective, income does not lead to happiness through its ability to satisfy our material needs or desires, but because it makes us more able to satisfy our needs and desires than others. In this sense, income is valuable primarily as a status marker or benchmark of comparison. If we are economically more successful than others in our reference group we are happier. A number of studies have demonstrated the utility of relative income (Solnick and Hemenway 1998; Layard 2005), and few would likely dispute that the concept of “keeping up with the Joneses” is a common narrative in US culture. If relative, more so than absolute, economic standing influences happiness, we would expect a positive cross-sectional relationship. Yet, growth in income over time might not lead to increases in happiness because our reference group, “the Jones”, may also be experiencing income growth.

While the degree to which the mechanism by which income leads to happiness is driven by relative versus absolute standing is still being debated (Veenhoven 1991; Easterlin 2003), less attention has been given to the question of whether other determinants of happiness may also have an important “relative” dimension. In other words, is the happiness attributed to our objective situation often fueled by our relative position, or is income an exception? If, in fact, the social or psychological determinants of happiness often rest on relative comparisons, what does this tell us about the ability of personal, environmental, and situational characteristics to increase our subjective well-being, either as individuals, or as a whole society?

In a recent paper, Blanchflower and Oswald (2004) offer what they suggest is the first ever analysis of the relationship between frequency of sexual behavior and happiness. Using General Social Survey (GSS) data, the authors demonstrate that, controlling for a variety of factors, individuals of all ages, genders and educational levels who engage in more frequent sexual behavior report higher levels of happiness than those who engage in sex less frequently. While the authors note the potential bi-directionality of the relationship, the findings are consistent with popular conceptions concerning what makes people happy, or what contributes to a fulfilling life. However, is it possible that the happiness resulting from such a basic human behavior is driven not solely by the objective experience, but by relative comparison? Similar to the process by which income primarily makes people happy only if they are earning more than others, are individuals happier because they are more frequently experiencing the joys of sex or because they believe they are experiencing more joy of sex than others in their reference group? The current research uses the GSS to explore this issue. I examine the degree to which a variety of factors, including both frequency of sex, and the average level of sexual activity of the respondents’ cohort shape self-reported happiness. More important than explicating the relationship between sex and happiness, the present research can deepen our understanding of a social psychological process that may be fundamental to our understanding of the determinants of happiness. To my knowledge, this is the only empirical analysis of how other people’s sex lives, or any characteristic or activity other than income, is related to respondents’ self-reported happiness.

1 The Pursuit of Happiness and the Role of Comparison

Philosophical explorations into the question of what makes us happy certainly predate the social scientific investigation of the correlates of subjective well-being. However, the recent rise of humanistic and positive psychology, as well as “happiness economics” has popularized more systematic empirical approaches to understanding happiness. This research has generally included two streams of investigation, one focusing on macro-level forces thought to shape average levels of happiness across large aggregates (usually nations), and the other examining individual-level correlates of happiness. Despite different levels of analysis, the two approaches have often been used together to explore theoretical propositions concerning what makes people happy.

Easterlin (1974) was the first to point to an interesting relationship between income and happiness. He demonstrated that within countries, individuals with higher incomes tended to report higher levels of subjective well-being. However, in international comparisons he noted that there was no consistent relationship between a nation’s wealth and the average level of self-reported happiness (with the exception of countries with large segments of the population living in extreme deprivation). Further, he demonstrated that as country’s gross national product (GNP) grew, there was no subsequent increase in happiness. These three findings formed the basis of the “Easterlin Paradox”. Easterlin (1974) suggested that this paradox resulted from the mechanism—reference group comparison—by which income and wealth lead to happiness. Over the last decade, the Easterlin Paradox has been challenged and new analyses have suggested that the three findings upon which it is based may have been overstated (Veenhoven and Hagerty 2006; Stevenson and Wolfers 2008) and that income growth may increase happiness for both individuals and nations.

While the most controversial aspect of the Easterlin Paradox (the claim that absolute income does not influence subjective well-being) is now hotly contested, few have argued against the relevance of relative income. In fact, recent work at the individual level has offered stronger empirical findings demonstrating this process. Drawing on data from the National Survey of Families and Households (NSFH), Luttmer (2005) demonstrated that neighbors’ incomes have a negative effect on respondents’ happiness. And that, consistent with theories of assessments based on reference groups, the influence of neighbors’ income is stronger for respondents who socialize more with people in their neighborhood. Similarly, both Firebaugh and Tach (2008) and McBride (2001) demonstrate that the self-reported happiness of GSS respondents is negatively correlated with the income of their peers. The two studies conceptualize peer/reference groups differently (discussed in more detail below), but their findings are remarkably similar. While not as directly focused on happiness, evidence for the importance of relative income can also be found in research on preferences. Solnick and Hemenway (1998) found that when a sample of Harvard public health students were given the option of living in a world in which they would make $50,000 dollars a year and everyone else would make $25,000, or one in which they would make $100,000 and everyone else would make $250,000, a majority of the respondents chose the first option—a world in which they would have half as much purchasing power but would be relatively high earners. The students appeared to place more value on relative than absolute income.

The primacy of relative versus absolute standing is suggested in a number of theoretical paradigms. The concept of the “reference group” was first introduced by Hyman in 1960 and began to have a significant impact soon thereafter. In their explication of Stouffer et al.’s, The American Soldier (1949), Merton and Kitt (1950) explore the degree to which subjective evaluations of individuals’ situations are heavily influenced by the circumstances of others. They found that the degree to which objective conditions brought happiness, frustration or a sense of inequity was profoundly shaped by the complex process through which respondents compared themselves to others. Out of reference group theory, grew the concept of relative deprivation which has been used to explain a wide range of behaviors, including violent crime (Blau and Blau 1982), social movements (Rose 1982), and job satisfaction (Johnson and Johnson 1995).

Concurrent with the growth of the concept of the reference group, Festinger (1954) outlined a theory of social comparison in which he proposed a number of processes by which individuals compare their “opinions and abilities” with those of others. While Festinger was focused more on when, and with whom, individuals draw comparisons than on the emotional outcomes of the process, his work further specifies the role of reference groups identified by Merton and Kitt (1950). The evidence he offers from experimental research concerning the in-group nature of reference groups is consistent with Merton and Kitt’s observations—individuals tend to compare themselves to others “like them”. Yet exactly how individuals determine who is in their reference group and thus should be used as a point of comparison remains unclear. From an empirical standpoint, individuals with similar relevant characteristics are often assumed to comprise the reference group from which people draw their comparisons. In their research on relative income effects on happiness, Firebaugh and Tach (2008) treat other GSS respondents of the same age in the same year as the reference group. Similarly, McBride (2001) draws on Current Population Survey data to generate average income estimates for individuals within 5 years of the age of the GSS respondents. The significant negative relationship between reference group income and happiness identified in both studies offers some validation to the process of treating others of the same age at the same time as the reference group—at least when income is the concept of interest.

In addition to the question of who we compare ourselves to, and the emotional response to such comparisons, scholars have questioned whether social comparisons are more common in some realms than others. Based on his research on hedonic adaptation (to what degree our happiness is influenced by the changing conditions of our own lives) Easterlin suggests that social comparison is “less in family life and health than in the material goods domain, because these circumstances are less accessible to public scrutiny than material possessions.”(Easterlin 2003:11181). However, Merton and Kitt (1950) suggest that while some of the domains in which servicemen compared themselves to others were easily observable (e.g. promotions and deferment status), others were more private and/or subjective in nature (e.g. levels of sacrifice or hardship). The influence of comparisons in domains less exposed to public scrutiny is unknown, as outside of the research on income, little empirical evidence is available as to whether individuals’ happiness is altered by the successes and failures of others. While we know with some certainty that respondents’ self-reported happiness is negatively affected when individuals in their reference group earn more money, is this also the case when they appear to have smarter children, more loving spouses, fewer health problems, or more active sex lives?

2 Sex and Happiness

As noted above, Blanchflower and Oswald (2004) published findings demonstrating that higher levels of sexual activity were correlated with higher self-reported happiness among GSS respondents. The authors report that, “Frequency of sexual activity is shown to be positively associated with happiness. The effect of sex on happiness is statistically well-determined, monotonic and large. This is true for males and females, and for those under and over the age of 40.” (2004:411). The authors also offer evidence suggesting that the “happiness maximizing number of sexual partners in the previous year is 1” (2004:411).

In related work drawing on data from the Global Study of Sexual Attitudes and Behaviors, a survey of 27,500 men and women aged 40–80 years in 29 countries, Laumann et al. (2006) found evidence suggesting that “sexual well-being” (having an active sex life) is correlated with overall well-being. Both the authors and critics of these studies note that the findings demonstrate an association between sex and happiness rather than a causal connection. It is possible, perhaps likely, that other unobserved factors contribute to both participation in sexual activity and overall levels of well-being (one criticism of Blanchflower and Oswald’s work is that they did not include a measure of health in their models, a well established correlate of happiness that is also likely related to frequency of sexual activity). As with income, it is also possible that it is happiness that increases the frequency of sex versus the frequency of sex increasing happiness. What is not addressed is exactly how or why sexual activity influences happiness. And the possibility that the observed relationship between sex and happiness may be influenced by the sex lives of others is not even suggested. I propose that just as income may influence happiness in a number of different ways (e.g. providing material comforts, increasing material security, increasing the amount of respect or deference offered by others, increasing self esteem), the same is likely true of sexual frequency, and some of these processes may be influenced by social comparisons.

Among other benefits associated with happiness, research suggests that frequent sexual activity increases physical pleasure, decreases physical pain, and promotes sleep (Uryvaev and Petrov 1996; Brody 2006). Such physiological effects are likely to increase happiness directly, and are independent of reference group comparisons. However, researchers have also found that sex can increase self esteem and confidence (Buss 2003; Meston and Buss 2007). For instance, Meston and Buss (2007) suggest that some individuals engage in sexual activity as a way of confirming their desirability. To the degree that sexual activity influences self-esteem and is used as a way of assessing the quality of one’s life, comparisons with others, be they specific or general, would be expected. While I am unable to assess the degree to which individuals compare their sex lives to the sex lives of others, the current research examines a potential outcome of such comparisons—the possibility that other people’s sex lives influence respondent’s levels of happiness.

Unlike income, the fruits of which show clearly and are often intentionally displayed to impress others, sexual activity is largely hidden from public view. For the sex lives of reference groups to be influential in shaping happiness, individuals must have some idea of how much sex others “like them” are having. While there is no empirical research on exactly how much individuals know about the sex lives of others, there is plenty of evidence that information concerning normative sexual behavior is learned through discussions within peer groups and friendship networks (Lefkowitz et al. 2004), and that information about sex norms are frequently presented in both general and gender specific media (Andre et al. 1989; Bielay and Herold 1995). Concerning the latter, one process through which the media distributes information about sex is by presenting “survey” results. In a quick search of some of the larger circulation magazines serving various populations I found that Cosmopolitan, Glamour, Men’s Health, Men’s Journal, and The AARP (American Association of Retired Persons) Magazine, with a combined US circulation of about 30 million (many of which end up in doctor’s offices and other locations where multiple readers are likely to be exposed to them), all present findings from their own sex surveys. These surveys vary greatly in both depth and methodological rigor (especially sampling procedures), but are quite successful in reaching large audiences. Many magazines also report information from surveys conducted by others—one of the more common being the annual Durex Global Sex Survey, which collects information from age/gender/nationality groups concerning frequency of sex and various sexual practices. In addition to the print media, the findings from these surveys appear online in advice columns, blogs, and other media formats. It should be noted, that these media also generally promote the message that “an active sex life is an important part of a healthy lifestyle”. While exposure to survey results and associated messages certainly varies across individuals, the widespread availability of information on other peoples’ sex lives increases general awareness of behaviors and norms around sexual activity—including frequency of sex. Thus, while information about sexual behavior at the individual level is likely hidden from public scrutiny (with the exception in many cases of close friends and acquaintances), at the aggregate level information concerning normative sexual behavior is available in much the same format as information about income, wealth and other personal characteristics.1

3 Data and Analysis

The present work draws on GSS data to examine the influence of sexual activity, reference group sexual activity and a host of other factors on self-reported happiness. The GSS is a multi-wave survey based on a national probability sample of the English speaking, non-institutional, adult population of the United States (NORC 2009). The GSS is an almost ideal data set for the present purposes for several reasons. Most importantly, unlike many of the large national data sets, the GSS asks questions about both sexual behavior and happiness. As the GSS has been administered regularly for over three decades (annually from 1972 to 1993 and biennially since 1994) and includes over fifty thousand respondents in total, it provides a large number of respondents from a wide range of years. Questions pertaining to frequency of sexual activity were first asked in the GSS in 1989.2 For the current analysis, I include respondents who were interviewed between 1993 and 2006, and as will be discussed below, generate “sex reference groups” from information provided by respondents between 1989 and 2006. This allows me to perform cross sectional analyses using 8 years of data which include fifteen thousand three hundred and eighty-six respondents (N = 15,386).

3.1 Dependent Variable

The outcome variable of interest is self-reported happiness. Since the inception of the GSS, respondents have been asked the following question every year:

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?

Over 26 years of data collection, about 56 % of the respondents have suggested that they are “pretty happy”, 32 % say they are “very happy”, and about 12 % report being “not too happy.” The measure demonstrates a good degree of variation and its distribution is fairly consistent across time. Several attempts to validate self-reported measures of happiness have found strong support for the concept. Self-reported happiness is generally highly correlated with spousal reports (Costa and McCrae 1988), reports of friends and family members (Sandvik et al. 1993), Duchenne smiles (Eckman et al. 1990)3, heart rate and blood pressure measures (Shedler et al. 1993), skin resistance/response to stress measures, and electroencephalogram measures of prefrontal brain activity (Sutton and Davidson 1997). Much previous work has discussed these measures in detail (Argyle 1989; Larsen et al. 1984; Myers 1993). Definitions and descriptive statistics for all variables are presented in Table 1.
Table 1

Descriptive statistics for all variables

Variables

Mean

SD

Self-reported happiness

2.21

.62

Demographic characteristics

 Year

1998

4.3

 Age

43.7

16.4

 Male (reference category)

46.2 %

.50

 Female

53.8 %

.50

 White (reference category)

81 %

.39

 Black

12.1 %

.33

 Other non-white

6.9 %

.25

Socio-economic characteristics

 Family income

50.7

40.6

 Educational achievement

13.4

2.8

Employment status

 Full-time work (reference category)

54.6 %

.50

 Part-time work

11.7 %

.32

 Temporarily unemployed

2.1 %

.14

 Laid off

3 %

.17

 Retired

11.3 %

.32

 Keep house

11.3 %

.32

 Out of labor market

2 %

.14

 Student

3.9 %

.19

Marital status

 Married (reference category)

56.1 %

.50

 Divorced

11.5 %

.32

 Separated

2.7 %

16.3

 Widowed

6.1 %

.24

 Never married

23.5 %

.42

Sexual activity characteristics

 Frequency of sexual activity during the previous year

  No sex in the previous year (reference category)

18 %

.38

  Once

7.8 %

.27

  About once per month

11 %

.31

  Two to three times per month

16.8 %

.37

  About once per week

18.8 %

.39

  Two to three times per week

21 %

.41

  Four or more times per week

6.5 %

24.7

 Number of sexual partners during the last year

1.2 %

2.77

 Frequency of peer/reference group sexual activity

2.93

.95

Health and family history characteristics

 Self-rated health

1.93

.80

 Parents divorced at age 16

15.5 %

.36

Geographic region of residence

 New England (reference category)

4.6 %

.21

 Mid Atlantic

14.1 %

.35

 East North Central

17 %

.38

 West North Central

7.5 %

.26

 South Atlantic

18.9 %

.39

 East South Central

6.5 %

.25

 West South Central

9.4 %

.29

 Mountain

7.4 %

.26

 Pacific

14.7 %

.35

3.2 Independent Variables

The main independent variables concern the frequency of sex for both the respondents and their likely reference groups. The GSS asks respondents how often they have had sex during the last 12 months. The answer categories include: not at all, once or twice, about once a month, 2 or 3 times a month, about once a week, 2 or 3 times a week, more than three times a week, don’t know, or no answer. Two potential issues with this question concern respondents being unwilling to answer the question and the truthfulness of their answers. The first concern appears to be largely unfounded. During the 11 years this question was asked, the non-response rate was only 7.4 %. Respondents who chose not to answer the question were dropped from the analysis. The second issue is more difficult to assess. Blanchflower and Oswald (2004) note that respondents may underreport due to modesty or the desire to conceal an extramarital affair, or over report in the interest of bravado. In their research with the GSS the latter appears to be more pronounced (2004:397), especially among males. However, as self-report data are the only way of estimating sexual frequency, and there is no way to validate respondent’s answers, I treat the responses as accurate measures of respondents’ frequency of sexual activity, with the acknowledgement that a certain amount of error is likely. As the frequency of sex variable is ordinal in nature, there are two primary ways of including it in the equation. It can either be transformed, usually by treating the midpoint of the category as a fixed point or an indicator of each category can be included in the equation, while leaving one category out to serve as reference. I have followed Blanchflower and Oswald (2004) and chosen the latter approach for two reasons. First, it is unclear whether the relationship between frequency of sex and happiness is linear. To the degree that a positive relationship is observed, there may be key threshold effects which would be masked by modeling the relationship as linear. Related, keeping the data in ordinal form allows us to assess how much of an effect each categorical increase has on happiness in comparison to the reference category—no sexual activity in the last 12 months.

In addition to measuring sexual frequency, I include a measure of the number of sexual partners a respondent reports during the 12 months prior to the survey. The answer options include 0, 1, 2, 3, 4, 510, 1120, 21100, and more than 100. I start by treating this as a continuous variable and follow the common practice of recoding each category to the midpoint of possible values. I also explore other treatments of the variable which I discuss below.

In order to estimate the relative influence of sexual activity, I generate estimates of reference group sexual activity. These estimates are based on age-, gender-, and time-specific groupings of GSS respondents between 1989 and 2006. For each year I generated the average frequency of sexual activity for males and females ages 18–21, 22–25, 26–29, 30–33, 34–37, 38–41, 42–45, 46–49, 50–53, 54–57, 58–61, 62–65, 66–69, 70–73, 74–77, and 78 and over, during the previous 5 year period. These are treated as estimates of the average sexual frequency of individuals in the same age/gender/year cohort. For instance, the reference group for a 32 year old male respondent from the 2001 wave of the GSS would be other male respondents age 30–33 who were surveyed from 1997 to 2001.4 Current “sex reference group” estimates are based on the self-reports of between 55 and 152 respondents. Table 2 reports these estimates.
Table 2

Sexual frequency of male and female peer/reference groups

Age

Year

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

(a) Male

18–21

2.79

2.85

2.76

3.03

3.03

2.98

3.11

3.11

2.91

2.98

3.15

3.12

3.13

3.02

22–25

3.67

3.82

3.88

3.83

3.83

3.71

3.59

3.50

3.31

3.35

3.38

3.46

3.56

3.60

26–29

3.74

3.69

3.68

3.73

3.73

3.65

3.68

3.70

3.66

3.75

3.89

3.91

4.07

3.78

30–33

3.84

3.71

3.64

3.62

3.62

3.62

3.66

3.70

3.75

3.76

3.78

3.75

3.70

3.66

34–37

3.95

3.76

3.74

3.68

3.68

3.50

3.48

3.55

3.46

3.53

3.75

3.80

3.87

3.74

38–41

3.76

3.63

3.60

3.65

3.65

3.59

3.59

3.60

3.54

3.50

3.52

3.54

3.44

3.34

42–45

3.45

3.36

3.38

3.35

3.35

3.44

3.46

3.46

3.47

3.55

3.58

3.53

3.64

3.40

46–49

3.18

3.13

3.10

3.19

3.19

3.24

3.27

3.18

3.10

3.14

3.09

3.14

3.33

3.01

50–53

3.27

3.12

3.12

3.09

3.09

3.14

3.20

3.15

3.18

3.06

2.91

2.86

2.71

2.61

54–57

2.83

2.92

3.01

3.03

3.03

3.01

2.96

3.05

3.16

3.13

3.18

3.03

2.80

2.79

58–61

2.80

2.86

2.89

2.94

2.94

2.92

2.80

2.76

2.69

2.66

2.66

2.59

2.49

2.62

62–65

2.65

2.64

2.64

2.78

2.78

2.58

2.48

2.45

2.23

2.25

2.37

2.23

2.08

2.35

66–69

2.28

2.11

2.11

2.02

2.02

2.07

2.18

2.07

2.09

2.09

1.96

1.93

1.98

1.72

70–73

1.69

1.60

1.54

1.66

1.66

1.70

1.83

1.87

1.82

1.94

2.11

2.14

2.30

1.69

74–77

1.05

.99

1.05

1.45

1.45

1.54

1.82

1.43

1.02

.95

.67

.91

1.15

1.38

78 and over

.86

.81

.81

.77

.77

.65

.61

.59

.53

.57

.62

.58

.6

.59

(b) Female

18–21

3.27

3.35

3.39

3.42

3.42

3.22

3.21

3.18

3.02

3.15

3.27

3.20

3.23

3.19

22–25

3.78

3.83

3.84

3.72

3.72

3.73

3.70

3.77

3.86

3.80

3.81

3.67

3.55

3.58

26–29

3.79

3.90

3.92

3.95

3.95

3.83

3.78

3.81

3.75

3.73

3.79

3.89

3.88

3.80

30–33

3.74

3.79

3.79

3.82

3.82

3.69

3.60

3.63

3.60

3.57

3.60

3.64

3.62

3.66

34–37

3.49

3.52

3.53

3.53

3.53

3.54

3.50

3.49

3.49

3.45

3.39

3.50

3.53

3.42

38–41

3.40

3.31

3.28

3.43

3.43

3.38

3.49

3.38

3.23

3.22

3.17

3.26

3.33

3.39

42–45

3.14

3.07

3.00

3.03

3.03

2.96

2.97

2.91

2.83

2.81

2.79

2.87

2.92

2.97

46–49

3.06

2.87

2.82

2.80

2.80

2.78

2.79

2.78

2.74

2.81

2.85

2.85

2.89

2.84

50–53

2.54

2.80

2.80

2.87

2.87

2.68

2.54

2.58

2.47

2.46

2.54

2.51

2.44

2.35

54–57

2.25

2.06

2.03

1.99

1.99

2.08

2.23

2.05

2.07

2.10

1.90

1.78

1.86

1.83

58–61

1.78

1.76

1.71

1.86

1.86

1.70

1.58

1.66

1.53

1.51

1.65

1.60

1.49

1.56

62–65

1.37

1.45

1.46

1.51

1.51

1.57

1.71

1.65

1.50

1.51

1.51

1.52

1.53

1.53

66–69

1.15

1.11

1.09

1.09

1.09

.95

.89

.88

.78

.88

.97

1.20

1.40

1.23

70–73

.50

.66

.67

.66

.66

.81

.69

.64

.69

.62

.49

.45

.39

.54

74–77

.35

.36

.38

.35

.35

.36

.33

.38

.38

.47

.57

.56

.58

.45

78 and over

.11

.12

.13

.14

.14

.17

.19

.17

.19

.18

.15

.23

.30

.24

Estimates are based on ordinal response categories. The question asks how often respondents have had sex during the last 12 months

0 = not at all; 1 = once or twice; 2 = about once a month; 3 = two or three times per month; 4 = about once a week; 5 = two or three times a week; 6 = four or more times a week

This approach to the creation of reference group sexual activity raises a number of questions concerning both the validity and the meaning of such a measure. The validity issue with the reference group measure is similar to the concern over using respondent self reports of frequency of sex. Some respondents may not be telling the truth. However, reporting error may be less of an issue when self-reports are used to create reference group measures than when used as direct indicators of the respondent’s behavior. From the perspective of reference group theory, how much sex others are having is not as important as how much sex respondents think others are having. If reports of sexual activity are being exaggerated or lowered in response to surveyors’ queries they may also be likely to be exaggerated if and when the information is shared with peers, on magazine or internet polls, or in any other venues.

A second question concerns who the reference groups are who are considered in comparative assessments. As noted above, both Firebaugh and Tach (2008) and McBride (2001) found that income comparisons appeared to be made with individuals of the same age/cohort. This is consistent with Festinger’s suggestions (1954) that we compare ourselves to those “like us”. In addition to the general practice of in-group comparisons, as income is commonly thought to vary by age and thus income norms will change across the age spectrum, similarly aged others may provide the best benchmark. Frequency of sexual activity is similar in that it too follows a common pattern across the life course (there is onset in the mid to late teen years, a significant increase into the late twenties, followed by a steady decline into the mid fifties which accelerates between the mid-fifties and the end of life). Given this pattern, age is most likely an important parameter in setting the boundaries of sex reference groups. While the age/frequency pattern of sexual activity is fairly consistent across genders, males typically report greater frequency of sex. This could be attributable to overestimates in male reporting or due to variation in the distribution of sexual activity by gender.5 In addition to gender-specific distributions, given the personal nature of the information and common patterns of friendship networks, which tend to be comprised of same versus opposite gender dyads, most direct information about the sexual activity of others is likely to come through same-sex friendships. Media sources which provide more indirect information on sexual activity may be either gender specific in orientation (e.g. Cosmopolitan Magazine or the Oprah Winfrey Show vs. Maxim Magazine or Men’s Journal) or offer gender-specific information (readership surveys, etc.). Both of these explanations suggest that for the purposes of sex, gender is likely an important factor to consider in reference group composition.

In addition to examining the influence of respondent and reference group sexual frequency, I explore the role of a number of demographic, economic, and social characteristics, many of which have been found to be significant predictors of subjective well-being in previous research. Basic demographic characteristics include age, gender (male = 0, female = 1), and race (white, black, other nonwhite). Employment status and family income comprise the economic characteristics. Employment status is an indicator of the respondent’s labor market participation during the week before the interview. Response categories include: employed full time, employed part time, temporarily not working, retired, student, keeping house, out of the labor market, and laid off or unemployed. The GSS collects information on both individual and family income. Firebaugh and Schroeder (2008) found that family income has a greater effect on happiness. This makes sense, as family income is likely a better reflection of both socio-economic status and individual consumption patterns. This is especially true for non-working respondents with working spouses. For this reason, I include a measure of family income in the present analyses. The GSS has asked the same question to measure family income over the life of the survey: In which of these groups did your total family income, from all sources, fall last yearbefore taxes that is? The response categories have varied across the survey waves as the national income distribution shifted upward as a result of inflation. I follow standard research practice by taking the midpoint dollar value for each of the categories selected. For the top category, which includes no upper bound (e.g. $75,000 and over), I impute values using a formula based on the Pareto curve (Hout 2004). This formula takes into account both the next-to-highest income category’s midpoint, as well as the frequencies of both the next-to-highest and highest categories when imputing the midpoint of the highest category.6 In addition, to correct for inflation, all incomes are converted to year 2000 dollars.

Social and personal characteristics include marital status (married, divorced, separated, widowed, never married), whether the respondent’s parents were divorced when the respondent was 16 years old, and self-reported health. Marital status has been shown in previous research to be significantly related to subjective well-being (Haring-Hidore et al. 1985; Veenhoven 1994) and is also likely to be related to sexual frequency as marriage may provide a stable sexual partner. I include a measure of parental divorce as some have suggested that adolescents experiencing parental divorce may be both less happy as adults (Amato et al. 1995) and may be more likely to struggle with intimacy issues (Wallerstein 1991), which could significantly influence frequency of sex. Self-reported health has been shown to be one of the strongest predictors of happiness (Dolan et al. 2008). In the GSS, respondents are asked whether they would say their health, in general, is excellent, good, fair, or poor. This is an especially important indicator in the present work for two reasons. First, given its influence on happiness, its absence from statistical models would significantly increase the likelihood of omitted variable bias. Second, health status, along with age is likely to play an important role in shaping sexual activity.

3.3 Analytic Approach

As the outcome variable of interest, self-reported happiness, is categorical in nature, I first estimated a conventional (restricted) ordered logistic regression model, or Polytomous Universal Model (PLUM) to assess the influence of the independent variables on the dependent variable. This is an extension of the linear model to ordinal categorical data and allows one to assess the degree to which changes in the independent variables increase or decrease the probability that a respondent will report a higher or lower level of happiness. However, a Wald test (Brant 1990) demonstrated that some of the variables in the model violated the assumption of proportional odds (also referred to as parallel regression), which states that the influence of the independent variables is constant across the different levels of the dependent variable (e.g. variables would have the same effect on the probability of a respondent reporting that they were pretty happy or very happy rather than not too happy as they would on the probability of a respondent reporting that they were very happy rather than not too happy or pretty happy. In response, I estimated partial proportional odds-ordered logit models. This approach allowed me to relax the assumption of proportional odds only for the variables that violate it, but maintain the restriction for those that don’t (Williams 2006). For independent variables that do not conform to proportional odds, the partial proportional odds-ordered logit model estimates n − 1 coefficients for each independent variable, with n representing the number of categories of the dependent variable. In the present case, the first coefficient represents the influence that the independent variable has on the probability that a respondent reports being pretty happy or very happy rather than not too happy, and the second the probability that a respondent reports being very happy rather than not too happy or pretty happy. For independent variables that do not violate the assumption of proportional odds, the coefficients are restricted to be equal and thus only one is reported. For ease of interpretation, I also report the odds ratios.

4 Findings

Before proceeding to the multi-variate analysis, we can first identify some of the correlates of both frequency of sexual activity and happiness by examining the distribution of responses across a number of demographic, economic and social categories. Table 3 provides these statistics.
Table 3

The distribution of sexual frequency and happiness

 

All

Males

Females

Age

<35

Age

35–55

Age

>55

Income

<$25 k

Income

$25–$50 k

Income

>$50 k

White

Black

Other non-white

Educ

<HS

Educ

HS grad

Educ

College grad

N = 15,386

N = 7,099

N = 8,287

N = 5,220

N = 6,400

N = 3,554

N = 4,368

N = 4,617

N = 5,626

N = 12,458

N = 1,867

N = 1,062

N = 2,371

N = 8,856

N = 4,157

Frequency of sex in the previous year

 No sex

18

13.4

22

10.8

9.8

43.1

28.4

16.8

8.9

18.1

18.7

16

29.5

20.5

17.7

 1 or 2 times

7.8

8.5

7.2

7.5

6.9

9.8

9.1

7.8

6.6

7.5

8.8

9.5

8

8

7.5

 About 1 time per month

11

11.2

10.9

7.4

12.3

14

9

11.3

12.6

11.5

9.6

8.1

10.7

10.2

12.1

 2 or 3 times per month

16.8

17.9

15.9

15.1

19.8

13.9

14.4

16.3

20

17.1

15.9

15.6

13.7

16.2

17.4

 Weekly

18.8

19.6

18.2

18.8

22.8

11.9

11.8

19.6

24.4

19.2

15.4

20.4

14

17.1

21.1

 2–3 times per week

21

21.8

20.3

28.6

23.2

5.9

18.7

22.2

22.4

20.4

23.8

23

16.9

21

19.7

 4 or more times per week

6.5

7.6

5.6

11.8

5.1

1.4

8.7

6.2

5.1

6.2

7.8

7.3

7.3

7

4.4

Self reported happiness

 Very happy

32.1

31.4

32.7

28.7

31.4

38.1

24.5

31.1

39.3

33.7

23.3

29

25.2

28.9

37.2

 Pretty happy

57

58

56.1

60.4

57.6

51.3

57.4

59.5

54.9

56.9

57.4

57.7

56.3

59.2

54.9

 Not too happy

10.9

10.6

11.2

10.9

11

10.6

18.1

9.4

5.8

9.4

19.3

13.3

18.5

11.8

7.8

 

Employed full time

Student

Retired

Keeping house

Out of labor market

Laid off

Married

Never married

Divorced

Widowed

Health is good or excellent

Health is fair or poor

N = 8,410

N = 604

N = 1,739

N = 1,734

N = 311

N = 465

N = 8,636

N = 3,618

N = 1,776

N = 935

N = 9,424

N = 2,442

Frequency of sex in the previous year

 No sex

10.2

24.5

51.4

19.4

33.6

14.3

6.4

24.3

30.2

76.5

15

30.1

 1 or 2 times

7

11.4

10

7.3

9.7

7.7

6.3

11.3

9.4

4

7.4

10.2

 About 1 time per month

10.7

7.4

12.7

11.3

12.3

10.9

13

8.5

9.4

4.5

10.8

11.4

 2 or 3 times per month

18.3

15.2

11.6

16.6

10.8

14.5

20.6

14.2

11

5.4

17.2

14.7

 Weekly

21.6

13.3

8.5

18

12.3

19.3

24.2

13

13.1

4.8

20.7

12.2

 2–3 times per week

25.2

18.6

4.7

20.1

12

21.9

23.7

19.7

19.6

3.5

22.7

14.5

 4 or more times per week

7

9.5

1.2

7.2

9.2

11.4

5.6

9.1

7.3

1.3

6.2

7

Self reported happiness

 Very happy

32.4

28.1

38.8

34.1

21

15.1

40.7

21.4

19.6

24.7

35.8

17.5

 Pretty happy

59

60.2

50

53.1

49.9

55.9

53

64

63.3

55

56.4

58.3

 Not too happy

8.6

11.7

11.2

12.8

29

29

6.3

14.6

17

20.3

7.9

24.2

 

No sex

1 or 2 times

About 1 time per month

2 or 3 times per month

Weekly

2–3 times per week

4 or more times per week

N = 2,769

N = 1,201

N = 1,696

N = 2,590

N = 2,899

N = 3,230

N = 1,002

Self reported happiness

 Very happy

24.1

24.5

30.7

34.7

36.5

35.6

34.5

 Pretty happy

58.5

61.5

57.9

56.7

55.9

56.2

53.1

 Not too happy

17.4

14

11.4

8.7

7.6

8.2

12.4

All frequencies are given in percentages

Concerning the outcome of interest, levels of self-reported happiness exhibit significant variation across a number of dimensions. Without controlling for other factors, respondents who are older, who come from households with earnings over fifty thousand dollars a year, who are white, who are more educated, who are married and who are in good health are more likely to report that they are “very happy”. Respondents who are from households with earnings of less than twenty-five thousand dollars a year, who are Black, who have not graduated from high school, and who are out of the labor market, laid off, or widowed are much more likely to report being “not too happy”. These initial relationships are consistent with much of the literature on the economic and social determinants of happiness. Concerning the relationship between sexual frequency and happiness, consistent with the findings of Blanchflower and Oswald (2004), respondents who are having more sex report greater levels of happiness—although there appears to be a threshold effect. As sexual frequency increased, the proportion of respondents reporting that they were very happy increased and the proportion reporting not too happy decreased. However, once a sexual frequency of about once a week was reached, more sex did not result in more happiness. In fact, the proportion of individuals reporting that they were not too happy begins to increase again for individuals who are having sex four or more times per week. These patterns should be interpreted with caution as they are not controlling for known correlates of both happiness and sexual frequency. I did not examine the bivariate relationship between respondent happiness and reference group sexual frequency as this relationship is inherently dependent on other factors such as age and gender in order to identify the appropriate reference group.

Table 4 shows the results of a partial proportional odds-ordered logit model in which self-reported happiness scores are regressed on sexual frequency and peer sexual frequency, along with other individual characteristics. A number of factors exhibit significant relationships with happiness. Foremost, given the focus of the current work, both respondents’ frequency of sexual activity and the sexual frequency of their age/gender reference groups are significantly related to happiness. Respondents who report having sex two to three times a month or more are significantly happier than respondents who report not having sex at all during the previous 12 months. Starting with the category of about once a month, the growth of the coefficients representing each of the frequency categories suggests a monotonic increase in self-reported happiness. The parameter estimates become significant upon reaching a frequency of 2 or 3 times per month. The odds ratios demonstrate that respondents who report having sex 2 or 3 times per month are 33 % more likely to report a higher level of happiness than those who did not have sex during the previous year, those who report having sex about once a week are 44 % more likely, and those reporting sexual frequency of 2 to 3 times a week are 55 % more likely. The one exception to this monotonic pattern concerns the respondents who reported having sex four or more times per week. Holding all else constant, these respondents were no more likely to report being pretty happy or very happy rather than not too happy, but were 87 % more likely to report being very happy rather than not too happy or pretty happy. One interpretation of this finding is that the highest level of sexual frequency does not differentiate unhappiness from moderate happiness, but significantly distinguishes respondents of moderate happiness from respondents reporting the most happiness. While not statistically significant, respondents who indicated having sex once or twice over the previous 12 months reported being less happy than individuals who did not have sex at all. When the number of partners was included in the equation as a continuous variable it was not significantly related to happiness. However, when treated as three dichotomous variables (no partners, one partner (the reference category), or multiple partners) the coefficients for both no partners and multiple partners are negative and significant, suggesting that a single partner is optimal for maximizing self-reported happiness. These findings are consistent with Blanchflower and Oswald (2004).
Table 4

Partial proportional odds-ordered logit findings regressing all characteristic on happiness

Variables

Not too happy versus pretty happy or very happy

Not too happy or pretty happy versus very happy

Coef. (SE)

Odds ratio

Coef. (SE)

Odds ratio

Demographic characteristics

 Year

.008 (.004)

1.007

PR

 

 Age

−.005 (.004)

.995

.004 (.003)

1.004

 Female

.081 (.043)

1.084

PR

 

 Black

−.503 (.083)***

.604

−.156 (.070)*

.855

 Other non-white

−.244 (.120)*

.783

.032 (.089)

1.032

Socio-economic characteristics

 Family income

.007 (.001)***

1.007

.002 (.001)***

1.002

 Educational achievement

.033 (.007)***

1.034

PR

 

Employment status

 Part-time work

−.070 (.062)

.932

PR

 

 Temporarily unemployed

−.205 (.132)

.814

PR

 

 Laid off

−.869 (.134)***

.420

−.520 (.159)***

.594

 Retired

.371 (.075)***

1.449

PR

 

 Keep house

.062 (.065)

1.064

PR

 

 Out of labor market

−.476 (.162)**

.621

.168 (.171)

1.183

 Student

.194 (.118)

1.215

PR

 

Marital status

 Divorced

−.824 (.057)***

.438

PR

 

 Separated

−1.177 (.112)***

.308

PR

 

 Widowed

−.841 (.087)***

.431

PR

 

 Never married

−.658 (.059)***

.518

PR

 

Sexual activity characteristics

 Frequency of sexual activity during previous year

  Once

−.068 (.082)

.935

PR

 

  About once per month

.098 (.077)

1.102

PR

 

  Two to three times per month

.286 (.073)***

1.331

PR

 

  About once per week

.364 (.073)***

1.439

PR

 

  Two to three times per week

.440 (.073)***

1.552

PR

 

  Four or more times per week

.229 (.129)

1.258

.624 (.102)***

1.867

 Number of sexual partners during the last year

.006 (.007)

1.006

PR

 

 Frequency of peer/reference group sex

−.154 (.053)**

.857

PR

 

Health and family history characteristics

 Self-rated health

.672 (.029)***

1.957

PR

 

 Parents divorced at age 16

−.131 (.050)**

.877

PR

 

Geographic region of residence

 Mid Atlantic

−.060 (.096)

.942

PR

 

 East North Central

−.076 (.093)

.927

PR

 

 West North Central

.101 (.102)

1.106

PR

 

 South Atlantic

.077 (112)

.926

.120 (.095)

1.128

 East South Central

.336 (.110)**

1.396

PR

 

 West South Central

.075 (.101)

1.078

PR

 

 Mountain

.003 (105)

1.003

PR

 

 Pacific

−.165 (.117)

.848

.042 (.099)

1.043

Standard errors are in parentheses

“PR” indicates that the parallel regression assumption was met and therefore the coefficients are the same across the different levels of the dependent variable

p < .05, ** p < .01, *** p < .001

Turning to the effect of reference group sexual frequency, we see a relationship similar to that which has been identified in the literature on income and happiness. The coefficient representing the influence of frequency of reference group sex on respondent happiness is negative, and consistent across the different values of the dependent variable, suggesting that the more sex others report having, the less happy respondents are. The odds ratio of .857 indicates that as the sexual frequency of respondents’ reference groups increases by one category (e.g. from a sexual frequency of two to three times per month to a frequency of about once a week), respondents’ probability of reporting a higher level of happiness decreases by about 14 %.

Together, the findings demonstrate a significant (almost monotonic) positive relationship between respondent sexual frequency and happiness along with a significant negative relationship between reference group sexual frequency and respondent happiness. The relationship between sex and happiness appears to be both absolute and relative. For the most part, other variables in the model exhibit relationships in the expected directions. Individuals from households with higher incomes and those with more education reported higher levels of self-reported happiness. Compared to those employed fulltime, retired respondents reported higher, and laid off respondents (as well as those out of the labor market in the first equation) reported lower levels of happiness. Divorced, separated, never married, and widowed respondents all reported significantly lower levels of happiness than their married counterparts. Respondents who were Black (and other non-white in the first equation), those with worse self-reported health, and those whose parents were divorced when the respondent was sixteen all reported being less happy. The one geographic region that demonstrated significantly higher levels of happiness than the Northeast was the East South Central region.

Another way of modeling the influence of reference group sexual activity is to look at the difference between respondents’ sexual activity and that of their peers. To do so, I created a new indicator by subtracting reference group sexual frequency from respondent self-reported sexual frequency. A positive number indicates that a respondent reports a higher sexual frequency than their reference group, and a negative number suggests a lower sexual frequency. Not surprisingly, collinearity diagnostics suggested that including this measure in a model with the indicators of the respondent’s frequency of sex would introduce significant collinearity to the model and risk substantial collinearity bias. When estimating this model without the respondent sexual frequency categories, the coefficient representing the difference between the respondents’ sexual frequency and that of their reference group is positive and significant—respondents who are having more sex than their reference group are happier and those having less sex than their reference group are less happy. For every one category difference in sexual frequency between the respondent and their reference group, the likelihood of reporting a higher level of happiness increases by 10.4 %. These findings are consistent with the models presented above.

5 Discussion and Conclusion

The findings broadly support the existing literature on the individual-level correlates of happiness. Consistent with previous research, income, education, and race, as well as employment, marital, and health status are all significantly related to respondents’ self reported levels of happiness. The current work also examines the little explored role of sexual activity on shaping happiness. While improving on Blanchflower and Oswald’s work (2004) by adding an indicator of physical health to the statistical models, including more recent waves of data, and correcting for potential violations of the proportional odds assumption, the current research replicates their findings—frequency of sexual activity is positively associated with happiness. Its effect is statistically significant after a certain threshold (sexual frequency of more than once per month), monotonic (with the one small exception mentioned above), and robust. More importantly, the findings suggest that the overall process by which sex is associated with happiness is intricately connected to our perceptions of the sex lives of others. Controlling for a variety of characteristics, the frequency of sexual activity among respondents’ reference groups is negatively associated with respondents’ happiness.

One of the critiques that has dominated much of the correlates of happiness literature concerns causal direction. We know with some confidence that across samples of individuals, higher incomes, better health, and marriage are associated with happiness. But do these, and other characteristics actually make us happy or are happy people more likely to have higher incomes (perhaps due to higher productivity or likability), enjoy better health (research suggests a positive outlook is good for our health), and enter into more stable relationships? The same question has been raised concerning the association between sex and happiness. Perhaps people who are happier have more energy for, or interest in, sex. While still interesting, the emerging field of happiness studies in economics, psychology, and sociology, as well as the growing self-help sections in bookstores across the country, have been more focused on the causes, rather than the consequences of happiness. It is understandable that people are more interested in how to be happier than in what their lives would be like if they became happier. The finding that the sexual activity of reference groups is associated with respondents’ happiness is much less susceptible to this critique. Few would argue that one’s happiness can influence the sexual activity of their cohort members across the nation.

A limitation of the present work concerns the inability to measure the process of social comparison or to assess the degree to which respondents have information about the sexual frequency of their reference groups. While the models strongly suggest that reference groups’ sexual activity matters, and there is much evidence that information concerning normative sexual behavior (including frequency of sex) for various groups is widely available through the media, and that information concerning the behavior of more proximate peers is often disseminated privately, I am not able to model either of these processes with the current data. In discussing this project with colleagues and peers, I have found that some feel that they have a strong sense of how much sex others are having (and based on GSS data, their estimates are fairly accurate) and others feel less knowledgeable. It is important to note that the current findings do not suggest or require that all respondents are knowledgeable about, or influenced by others’ sex lives, just as the findings concerning relative income effects do not require all respondents to suffer when others make more money. Instead, they suggest a pattern of relationship among the sample. The process of social comparison seems like the most likely explanation for this pattern, but it is possible that the relationship between respondents’ happiness and the sexual activity of their reference groups is being driven by some other process. Future research, most likely more qualitative in nature, should address this possibility by more thoroughly exploring the role of social comparison in the realm of sexual behavior.

While GSS data do not directly measure the process by which one’s happiness is influenced by social comparisons, the current findings, along with the growing evidence that the income of others influences respondents’ happiness, suggests an important process that deserves more attention in the rapidly growing literature on the causes and correlates of happiness. Individual situations are not evaluated in isolation. The emotional response to what is occurring in an individual’s life arises within a social context and cannot be separated from what is occurring in the lives of others. The contextual nature of the evaluation of self is certainly not new to social psychology, but it has only begun to be incorporated into our understanding of what makes people happy.

For it to be practically useful and theoretically informative, research on happiness needs to explore both what makes people happy as well as how and why it makes people happy. In the case of income and sex, part of the how and why appears to stem from the process of relative comparison. This has a number of implications. As has been noted by economists, the finding as it relates to income suggests that income growth at the national level will not increase average happiness levels. A rising tide may lift all boats, but being higher in the water only provides a better view if other boats are lower. Similarly, initial findings demonstrating a positive relationship between sex and happiness should be interpreted cautiously. Broad cultural changes in sexual practices that influence sexual frequency (whether they be sexual revolutions or the availability of Viagra) may do less to increase happiness than might be expected. In fact, some scholars have suggested that the “sexual revolution” of the sixties and seventies may have increased some types of stress and anxiety, as sexual expectations were much higher than they had been in the past.

More important than further specifying the relationship between sex and happiness, the current findings contribute to the growing body of literature that views social comparison as fundamental to our understanding of the causes and correlates of happiness. This is the first research that I am aware of that explores the role of social comparison on happiness in any realm other than income. The finding that happiness is shaped by both our own sexual behavior as well as that of our reference groups calls to question the degree to which the influence of other well established correlates of happiness (e.g. health, marital status, labor market participation) are also influenced by the process of comparison. If our understanding of the causes and correlates of happiness is to continue to mature, it is essential that future research more deeply explore the role of reference group comparison in shaping subjective well-being.

Footnotes
1

To the degree that individuals evaluate and compare their sex lives to the sex lives of others, it is likely that both frequency as well as quality is considered. Given data restrictions, I have no way of measuring quality in the present research. It should also be noted that information about quality is generally much less available to individuals, and thus may play a smaller role in social comparisons.

 
2

The sexual frequency question was asked to about half of the respondents in 1990 and was not included in the survey in 1992. During all of the other post 1988 waves of data collection the sexual frequency question was included in most of the interviews.

 
3

“Duchenne” smiles involve the contraction of muscles that both raise the corners of the mouth and the cheeks. They are often treated as the expression of genuine emotions because few people can voluntarily contract the outer portion of the orbicularis oculi muscle, which is responsible for raising the cheeks.

 
4

While respondents from the previous 5 years were used in the estimation of reference groups, as the GSS was only administered every 2 years, the estimates are actually based on 2 or 3 waves of respondents.

 
5

Given the structure of the sexual frequency categories in most surveys, the high values are right censored. For instance in the GSS, respondents who reported engaging in sex 4 times a week would be in the same category as those reporting having sex 25 times a week. Thus, one female outlier could move several males into a higher category. Blanchflower and Oswald (2004) suggest that the gender distribution of prostitutes and their customers may partially explain this pattern.

 
6

Another common practice is to multiply the lower bound of the upper category by 1.5 and to impute this value as the midpoint (Firebaugh and Tach 2008). Substituting this value had no effect on the substantive findings reported below.

 

Acknowledgments

I thank Jerald Herting, Stefanie Mollborn and Fred Pampel for helpful comments on earlier drafts.

Copyright information

© Springer Science+Business Media Dordrecht 2013

Authors and Affiliations

  1. 1.Department of SociologyUCB327, University of Colorado at BoulderBoulderUSA

Personalised recommendations