Why Are Fewer Young Adults Having Casual Sex?

Fewer young adults are engaging in casual sexual intercourse now than in the past, but the reasons for this decline are unknown. The authors use data from the 2007 through 2017 waves of the Panel Study of Income Dynamics Transition into Adulthood Supplement to quantify some of the proximate sources of the decline in the likelihood that unpartnered young adults ages 18 to 23 have recently had sexual intercourse. Among young women, the decline in the frequency of drinking alcohol explains about one quarter of the drop in the propensity to have casual sex. Among young men, declines in drinking frequency, an increase in computer gaming, and the growing percentage who coreside with their parents all contribute significantly to the decline in casual sex. The authors find no evidence that trends in young adults’ economic circumstances, internet use, or television watching explain the recent decline in casual sexual activity.

The percentage of young adults who engage in sexual intercourse has been declining in recent years. Twenge, Sherman, and Wells (2017a) report that the percentage of adults ages 20 to 24 who did not have sex in the past year increased from 11.67 percent in 2000–2009 to 15.17 percent in 2010–2014. Using a slightly different age range and time frame, Ueda et al. (2020) found that the percentage of sexually inactive men ages 18 to 24 increased from 18.9 percent in 2000–2002 to 30.9 percent in 2016–2018, and the percentage of sexually inactive young women increased from 15.1 percent to 19.1 percent over the same period. Although similar declines in sexual activity are found among older age groups, the decline in sexual activity has been most acute among teenagers and young adults (Abma and Martinez 2017; Ethier, Kann, and McManus 2018). But although there has been considerable speculation as to the forces driving the decline in young adult sexual activity (e.g., Ingraham 2019; Julian 2018; Wilcox and Sturgeon 2018), few studies have attempted to rigorously identify the sources of this downswing in young adult sexual activity.

An important form of adolescent and young adult sexual activity is encounters that occur outside of a committed relationship or partnership, encounters we refer to here mainly as “casual sex.” In recent decades casual sexual encounters have become normative behavior among emerging adults of both sexes (Bogle 2008). Casual sexual encounters could involve people who are strangers or acquaintances, previous partners, or friends (Garcia et al. 2012). Some casual sexual encounters may serve as a sort of trial or rehearsal for determining whether a long-term romantic partnership is desirable (Manning, Giordano, and Longmore 2006).

The decline in casual sexual activity could have both positive and negative consequences for young adults and broader society. On one hand, less frequent casual sex may lead to fewer unwanted pregnancies, sexually transmitted diseases, and mental health problems (Arnett 2018; Bersamin et al. 2014; Townsend et al. 2020). On the other hand, sexual inactivity may hinder young adults’ psychosocial development and diminish their physical and emotional gratification (Julian 2018). Nationally representative data on the frequency of, or trends in, casual sex are rare (cf. Johnson 2013). However, the decline in sexual activity among unmarried young adults (Ueda et al. 2020) suggests that a decline in casual sex parallels the more general decline in sexual frequency among young adults.

The purpose of this study is to identify some of the reasons for the decline in the frequency with which adolescents and young adults (hereafter referred to simply as young adults) who are not in romantic relationships report having had recent sexual intercourse. We use data from the Panel Study of Income Dynamics Transition into Adulthood Supplement (PSID-TAS) for the period from 2007 to 2017. We first document a decline in the percentage of young adults who report having had recent sexual intercourse without having been in romantic partnerships, and we then perform a formal mediation analyses to isolate the factors that explain this downward trend. We examine eight potential drivers of the decline in casual sexual behavior among young adults, reflecting trends in employment, earnings, financial debt load, coresidence with parents, use of electronic media, television watching, computer video gaming, and alcohol consumption.

Our study builds on a recent analysis by Lei and South (2021) that examined reasons for the recent decline in sexual activity among young adults. Lei and South found that about one third of the decline in young adult sexual activity can be attributed to a decline in the percentage of young adults who are in romantic relationships, including not only formal marriage and nonmarital cohabitation but also less formal dating relationships (see also Twenge et al. 2017a). Declining earnings and alcohol consumption also explain some of the decline in young adult sexual activity.

Our study extends Lei and South’s (2021) research in three main ways. First, while Lei and South examined trends in sexual activity among both partnered and unpartnered young adults, we focus only on sexual activity that occurs outside of a romantic relationship. Second, we consider the possibility that the sources of the recent decline in casual sexual activity differ between young women and young men. Casual sexual activity may have different meanings and motivations for women and men (Auster, Faulkner, and Klingenstein 2018; Manning, Giordano, and Longmore 2006), and thus the determinants of the decline might differ between the sexes. Third, given our focus on casual sexual activity, we explore the role played by possible explanations of the decline not considered by Lei and South, including trends in student debt load.

Data and Methods Data Data for this analysis come from the 2007 through 2017 waves of the PSID-TAS. Begun in 1968, the main, or core, PSID collects data on a representative sample of families on a variety of topics. Adult children and their families are added to the study when they leave to form own households. Interviews were conducted annually through 1997 and have been conducted biennially since then. In 1997, the PSID launched the Child Development Supplement (CDS), collecting data on children ages 0 to 12 in PSID families and their primary caregivers. The original PSID-CDS was conducted every 5 years, with a new sample introduced in 2014. Recognizing the need to collect data on PSID sample members as they age out of the CDS to join the core PSID, in 2005 the PSID instituted the Transition into Adulthood Supplement, initially collecting data on CDS participants ages 18 to 21 whose families remained part of the core PSID. Every two years the PSID-TAS reinterviews the initial sample until they reach age 28 and adds new members to the sample as they reach age 18. With the proper weights applied, the PSID-TAS is representative of the cohort of young adults born between 1985 and 1999 who are no longer attending high school (PSID 2016). Sample Selection For this analysis, we select PSID-TAS respondents who were between the ages of 18 and 23 in the 2007, 2009, 2013, 2015, or 2017 interview wave. We omit observations for the 2005 wave because the initial PSID-TAS interview did not ask questions about respondents’ sexual activity. We omit observations from the 2011 wave because a processing error by the PSID staff made the questions regarding sexual activity unusable. We select respondents ages 18 to 23 so as to consistently bound the age range of the sample; respondents at older ages can be observed only in the later interview years. Finally, as described below, we select only respondents who report not being married, nonmaritally cohabiting, or in dating relationships. Our sample consists of 945 women (who contribute 1,296 person-year observations) and 1,040 men (who contribute 1,514 person-year observations). Dependent Variable To measure whether the respondents recently engaged in casual sex, we first exclude those who at each interview report being married, in nonmarital cohabiting relationships, or in (nonmarital, noncohabiting) romantic relationships. We then consider respondents to have had casual sex if they gave a nonzero response to a question asking them how many times during the four weeks preceding the interview they had engaged in sexual intercourse. We acknowledge several limitations to this measure of casual sexual activity. First, the question asked of the respondents does not define “sexual intercourse.” Some respondents may consider certain types of sexual activity (e.g., oral sex) to be intercourse, while others may not (Gute, Eshbaugh, and Wiersma 2008; Horowitz and Spicer 2013). Second, it is not possible to distinguish same-sex from opposite-sex encounters. Third, individuals who are in romantic relationships may have engaged in casual sex with someone other than their partners; these out-of-relationship incidents will not be captured in our measure. Fourth, the slight temporal discrepancy between the time frame for sexual activity and the measurement of relationship status could create measurement error in the dependent variable. Some respondents might have been in romantic relationships at the time of sexual activity but were no longer in relationships by the time of the survey, which would lead to an overestimate of the percentage of respondents who had engaged in casual sex. Conversely, some respondents might have engaged in casual sex during the four weeks prior to the interview but subsequently formed romantic relationships, either with the sexual partners or with someone else, by the time of the interview; this misclassification would lead to an underestimate of the percentage of respondents who had engaged in casual sex. We believe that these limitations to the measurement of casual sex are at least partially compensated for by the PSID-TAS’s rich measures of potential explanations for the decline in young adult sexual activity. Independent Variables The trend in the probability that the PSID-TAS respondents report having had casual sex is indicated by a continuous variable for the year the survey was taken (2007 = 7, 2009 = 9, 2013 = 13, 2015 = 15, 2017 = 17). Each unit difference in survey year thus provides an estimate of the annual change in the likelihood of having had casual sex. Several of the independent variables tap the respondents’ financial situations. A dummy variable distinguishes employed from nonemployed respondents. Earnings refers to earnings from work in the calendar year preceding each interview. It is measured in constant 2007 dollars and logged to reduce skewness. Debt load is measured by the total value of outstanding student loans, credit card loans, and other personal loans measured in constant logged 2007 dollars. Parental coresidence is measured by a dummy variable indicating whether the respondents lived in their parents’ homes in the fall and winter preceding each PSID-TAS interview. We use several variables to capture the frequency with which young adults use electronic media. Respondents were asked the frequency with which they used the Internet in the prior 12 months for (1) games, (2) e-mail, (3) school-related projects, (4) shopping, and (5) visiting social networking Web sites such as Facebook. For each type of Internet use, the possible responses range from 1 (never) to 5 (every day). Given the potentially unique importance of the rise in video gaming for explaining the decline in young men’s likelihood of having casual sex, we treat the response to this use of the Internet as a separate variable. We then average the responses to the other four Internet uses. Because not all items are available in every interview year, we first standardize each item and then calculate the average of the available items in that year. The PSID-TAS also asks at each interview how often respondents watched (non-news) television shows in the year preceding the interview; possible response categories range from 1 (less than once a month) to 6 (every day). Alcohol consumption is measured by a question asking respondents how often they drank alcohol in the past year, with response categories ranging from 1 (never) to 7 (every day). Control Variables In addition to variables that might help explain the decline in the frequency of casual sex, the models also include several other potential predictors of the likelihood that unpartnered young adults have recently engaged in sexual intercourse. Frequency of sexual activity increases with age over the young adult life course (Twenge, Sherman, and Wells 2017b); we measure respondents’ age in years. Some research finds racial and ethnic differences in young adults’ propensity to have nonromantic or nonrelationship sex (Manning, Longmore, and Giordano 2005). Respondents are categorized into four racial/ethnic groups: non-Hispanic black, Hispanic, non-Hispanic white, and other races. College students and college graduates tend to have fewer sexual partners than other young adults (Lyons et al. 2013). Our measure of educational status combines respondents’ educational attainment with their college enrollment, categorizing respondents as not enrolled in and not graduated from college, college graduates not currently enrolled in college, or current college enrollees. Religiosity is a strong predictor of young adult sexual behavior (e.g., Weitzman 2020). We tap respondents’ religiosity by a question asking respondents how often they attended religious services in the 12 months preceding each interview (1 = never, 6 = more than once a week). We treat this as a continuous variable. Healthier youth may be more likely to have sexual intercourse (Lei and South 2021). Respondent’s self-rated health is measured conventionally with responses ranging from 1 (poor) to 5 (excellent). Family structure is known to affect children’s sexual behavior (South, Haynie, and Bose 2005). Our measure of respondents’ family structure contrasts respondents whose families, when the respondents were age 18, consisted of two married parents, one biological and one stepparent, or one unmarried parent. Youth from more advantaged families tend to delay the transition to sexual activity (Baumer and South 2001). We capture family-of-origin socioeconomic status with parental educational attainment, measured by the highest year of school completed by the respondent’s more educated parent. These latter two variables are derived from questions asked of the PSID-TAS respondents’ household heads—most often their fathers or mothers—as registered in the core PSID. Analytical Strategy Our focal analyses pool respondents across interview waves and estimate logistic regression models of the odds that a PSID-TAS respondent has engaged in unpartnered sexual intercourse in the four weeks preceding each interview. The first model is a baseline model that includes as covariates only year of the survey and the control variables. Subsequent models then add measures of each of the hypothesized explanations for the decline in the probability of engaging in casual sex to determine the degree to which they mediate the effect of survey year. Because some PSID-TAS respondents will appear in more than one survey wave, we compute robust standard errors of the regression coefficients that adjust for this nonindependence of observations. Comparing coefficients across nested logistic regression models is not as straightforward as for linear models because of the rescaling of the model in the presence of mediators (Breen, Karlson, and Holm 2013). Accordingly, we formally assess the degree to which the hypothesized explanations “mediate” the time trend in young adult casual sexual activity using the procedure developed by Karlson, Holm, and Breen (2012) (KHB) and implemented through the Stata command khb (Kohler, Karlson, and Holm 2011). The KHB method resolves the rescaling problem and allows direct comparison of coefficients between nested models. The variables that contain any missing data have on average only about 2 percent of their values missing. To retain these cases, we use multiple imputation, generating 10 imputed data sets. All of the descriptive and multivariate analyses were conducted using multiply imputed data, and the results were combined using Rubin’s (1987) rules.

Results Table 1 presents weighted descriptive statistics for all variables used in the analysis, separately for women and men. Twenty-eight percent of the women and 34 percent of the men who were unpartnered at the time of the survey report having engaged in sexual intercourse in the past four weeks. On average, respondents are about 20 years old at the time of the survey. Most are non-Hispanic white. More of the women than the men are enrolled in college. The respondents’ self-rated health typically falls between the categories of “good” and “very good,” and their attendance at religious services tends to fall between the categories of “a few times a year” and “about once a month.” Most respondents come from a family containing their two biological parents. Table 1. Descriptive Statistics for Variables Used in Analysis of Casual Sex: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version Slightly more than half of the respondents are employed at the time of the survey. Men have higher earnings and less debt than women. About half of the respondents live with their parents. The mean frequency of playing computer games falls between “at least once a month” and “once a week” for women and at “once a week” for men. The average standardized frequency of using the Internet for purposes other than gaming is higher among women than men. For both women and men, the average frequency of watching television falls between “several times a week” and “almost every day.” Among women, the mean frequency of drinking alcohol during the past year falls between “about once a month” and “several times a month,” while for men the mean falls slightly above “several times a month.” Table 2 begins to shed light on the ability of the hypothesized mediators to explain the trend in young adult casual sexual activity by presenting descriptive statistics for the focal variables for the study frame’s initial year (2007) and final year (2017). The decline in casual sex is apparent for both women and men. In 2007, 31 percent of the young unpartnered women reported having sexual intercourse during the past month, but in 2017 only 22 percent did so. The percentage of men who reported engaging in casual sex dropped from 38 percent in 2007 to 24 percent in 2017. These declines in causal sexual activity are generally consistent with the decline in sexual intercourse among all young adults observed in other nationally representative surveys (e.g., Twenge et al 2017a; Ueda et al. 2020). Table 2. Descriptive Statistics for Casual Sex and Potential Explanatory Factors, 2007 and 2017: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version Some but not all of the hypothesized mediators emerge as possible candidates for explaining these declines in casual sex. Between 2007 and 2017, more young women became employed, but there was no significant change in their mean earnings, and their total debt load actually declined significantly. Young men’s employment, earnings, and debt did not change significantly between 2007 and 2017. Both women and men became significantly more likely to live with their parents. Both women and men played computer games and used the Internet more frequently in 2017 than in 2007. Among men, however, the frequency of watching television fell significantly. Both women and men drank alcohol significantly less often in 2017 than in 2007. Table 3 presents a series of logistic regression models of the odds that young women engaged in casual sex during the four weeks preceding each PSID-TAS interview. Model 1 is a baseline model that includes as predictors year of the survey and the control variables. The statistically significantly coefficient for survey year indicates that, net of the effects of the controls, the estimated odds that young women had casual sex dropped by about 6 percent per year ([1 – e−.065] × 100) between 2007 and 2017. The odds of having casual sex increase significantly with age. Compared with non-Hispanic white women, non-Hispanic black women are significantly more likely, and women of “other races” significantly less likely, to have casual sex. College graduates are less likely than those not in college (and without a college degree) to have sexual intercourse outside of a romantic relationship. Frequent attendance at religious services is inversely associated with the odds of having casual sex. Table 3. Logistic Regression Models of Casual Sex among Young Women: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version The subsequent models in Table 3 add to model 1 the hypothesized explanations for the decline in casual sex. Model 2 adds young women’s employment status and earnings. The difference between employed and nonemployed women in the likelihood of having casual sex is not significant, but the coefficient for earnings is significant and positive. However, these factors explain very little of the decline in casual sex; the coefficient for survey year changes only trivially, from –.065 to –.064, when young women’s employment and earnings are controlled. Model 3 adds financial debt load. The coefficient for debt is statistically nonsignificant and its inclusion in the model also has little effect on the coefficient for survey year. Model 4 of Table 3 adds to the baseline model the dummy variable for parental coresidence. As expected, young women who live with their parents are significantly less likely than young women who live independently to have casual sex. However, controlling for parental coresidence does not appear to explain much of the time trend in the likelihood of having casual sex. The coefficient for survey year drops by only about 6 percent (from –.065 to –.061) when parental coresidence is controlled. Model 5 adds the frequency of playing computer games to the baseline model. The frequency of playing computer games is neither strongly nor significantly associated with the odds of having casual sex. Consequently, even though young women’s frequency of computer gaming has increased over time (Table 2), this cannot explain their decline in casual sex. In fact, the coefficient for survey year increases slightly in absolute value when the frequency of computer gaming is controlled. In model 6, measures of the frequency with which young women use the Internet and watch television are added to the baseline model. Frequency of watching television is not significantly associated with the odds of having casual sex. And perhaps surprisingly, frequency of using the Internet is positively and significantly associated with the probability of having casual sex. Perhaps young women are using the Internet to find potential sexual partners via social networking and dating apps. Because young women’s use of the Internet has increased over time (Table 2), trends in Internet use tend to suppress what would have otherwise been a larger decline in the likelihood of having casual sex. The negative coefficient for survey year becomes stronger, increasing from –.065 to –.076, when frequency of Internet use is controlled. Frequency of alcohol consumption is added in model 7. Frequent drinking is significantly and positively associated with the odds of having casual sex. Moreover, controlling for frequency of drinking reduces the coefficient for survey year, which drops from –.065 in model 1 to –.051 in model 7. Model 8 includes as mediators those factors that, in prior models, explained at least some of the decline in young women’s propensity to have casual sex. We omit from this model the hypothesized mediators that unexpectedly serve to suppress the drop in casual sex—computer gaming, Internet use, and TV watching—because including them would allow the other mediators to explain a larger amount of the decline in casual sex than actually occurred. The coefficients for earnings, parental coresidence (at a borderline level), and alcohol consumption are statistically significant. The coefficient for survey year drops from –.065 to –.049 in model 8. The association between the frequency of drinking and the propensity to engage in casual sex is particularly noteworthy. The odds that young women who report drinking daily (scored 7) engage in casual sex are almost seven times the odds for young women who report never drinking (scored 1) (e[7 − 1][.320] = 6.8). The analysis presented in Table 4 hints at the factors that might explain the recent decline in casual sexual intercourse among young women. However, because the outcome variable is categorical, a more formal assessment of mediation is needed. The KHB mediation analysis is shown in Table 4. The only factor that explains a statistically significant portion of the downward trend in casual sex is the frequency of drinking, which alone accounts for about one quarter (23.9 percent) of the decline. Adding to this the other factors that explain at least some of the decline (employment and earnings, debt load, coresidence with parents, and drinking) only boosts the percentage explained from 23.9 percent to 27.9 percent. As anticipated by the results in Table 3, the trends in Internet use and television watching significantly suppress the decline in young women’s propensity to engage in casual sexual intercourse. The decline in young women’s casual sexual activity would have been even greater had young women’s Internet use and television watching not increased as they did. Table 4. Mediation Analysis of Factors Explaining the Decline in Casual Sex among Young Women: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version Explaining the Decline in Casual Sex among Young Men Tables 5 and 6 present a largely parallel analysis for young men. As shown in model 1 in Table 5, net of the effects of the control variables, the odds that young men engage in casual sex are estimated to have declined by about 7 percent per year ([1 – e–.073] × 100) between 2007 and 2017. Model 1 also shows that young men who are older, are black, attend religious services infrequently, and do not live with both biological parents at age 18 are more likely than other young men to have had casual sex in the prior month. Table 5. Logistic Regression Models of Casual Sex among Young Men: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version Table 6. Mediation Analysis of Factors Explaining the Decline in Casual Sex among Young Men: Panel Study of Income Dynamics Transition into Adulthood Supplement, 2007 to 2017. View larger version Although several of the hypothesized mediating factors are significantly associated with the odds that young men have casual sex (Table 5, models 2–7), only three of them appear to explain a nontrivial proportion of the decline in casual sexual activity. As shown in model 4, young men who live with their parents are significantly less likely than their counterparts who live independently to report having had casual sex in the past month, and controlling for parent coresidence reduces the coefficient for survey year from –.073 to –.067. As shown in model 5, the frequency with which young men play computer games is significantly and inversely associated with the likelihood of having casual sex, and controlling for computer gaming drops the time trend coefficient from –.073 to –.054. And as shown in model 7, frequency of alcohol consumption is significantly and positively associated with the odds of having casual sex, and controlling for drinking drops the coefficient for survey year from –.073 to –.051. The coefficients for all three of these mediators remain statistically significant in the “full” model (model 8), which omits the mediators that suppress the time trend even if only slightly (debt load, Internet use, and television watching). The estimated odds that young men who live with their parents engage in casual sex are only 63 percent of the corresponding odds for young men who live independently (e−.457 = .63). The odds that young men who report playing Internet computer games every day (scored 5) have casual sex are less than half the odds for young men who never game (scored 1) (e[5 − 1][−.185] = .48). And the odds that young men who report drinking daily have casual sex are about 5.5 times the odds for young men who abstain from alcohol (e[7 − 1][.286] = 5.56). Table 6 presents the formal mediation analysis for men. The most important factor driving the decline in casual sex among young men is the decline in drinking, which alone explains more than one third of the decline. Considered individually, the increase in computer gaming explains about one quarter of the decline in young men’s propensity to have nonrelationship sex, and the rise in parental coresidence explains a little more than 10 percent. No other mediator explains a statistically significant portion of the decline in young men’s probability of engaging in casual sex. Considered together, the mediators that explain at least some portion of the decline (employment and earnings, parental coresidence, computer gaming, and drinking) explain 57 percent of the drop in young men’s casual sexual activity. Additional Analyses We performed several additional analyses. First, we tested for the statistical significance of the sex difference in the associations between the hypothesized mediating variables and the log odds of having casual sex. We pooled the samples of women and men and estimated a model that included product terms capturing the interaction between respondent’s sex and each of the mediating factors. The only statistically significant interaction involved the frequency of playing computer games, with the slope being more negative among men than among women. Second, because some of the hypothesized mediating variables might be endogenous to one or more of the other mediators, we estimated the models in Tables 3 and 5 controlling for all of the other mediating variables. That is, we entered into the model a given mediator after including all of the other mediators (as well as the background controls). The results were highly consistent with the findings reported in the tables. The main exception is that among men, the increase in parental coresidence no longer explains a statistically significant proportion of the decline in casual sex when the other mediators are controlled. Even in this case, however, proportion of the secular decline in the log odds of having casual sex attributable to the rise in parental coresidence, almost 20 percent, is reasonably large.

Discussion and Conclusion Adolescents and young adults are increasingly less likely to have sexual intercourse outside of a romantic relationship, but the causes of this decline in casual sex have not been rigorously evaluated. We use data from the PSID-TAS to quantify the sources of the decrease in young adult casual sexual activity between 2007 and 2017. We find that about one quarter of the drop in young women’s propensity to have casual sex is attributable to a decline in their frequency of drinking alcohol. Of the various sources of the decline in sexual activity considered in this analysis, the decline in alcohol consumption is the only factor that explains a significant portion of the decline in young women’s probability of engaging in casual sex. A somewhat different story emerges for young men. As with young women, a decline in the frequency of drinking alcohol is an important source of young men’s diminished likelihood of having casual sex. But unlike for young women, among young men increases in the frequency of playing computer games and in the tendency to reside in the parental home also play important roles. Although both young women and young men play computer games more frequently now than in the past, gaming inhibits only young men’s casual sex behavior. The factors hypothesized to explain the decline in casual sexual intercourse explain a greater portion of the decline in young men’s than in young women’s propensity to engage in casual sex. We find no evidence that some other transformations in the lives of emerging young adults can explain the decline in their casual sexual activity. Trends in young adults’ financial insecurity, including their student debt load, do not appear to underlie their change in casual sexual activity. Nor does an increase in time spent watching television. And among young women the increase in the use of the Internet appears to actually suppress what would otherwise have been a larger drop in the propensity to engage in sex with someone who is not a romantic partner. We acknowledge that our findings raise questions as to what factors are driving changes in these proximate sources of the decline in young adult casual sexual activity. Further research is needed to identify the causes of trends in young adult alcohol consumption, computer gaming, and parental coresidence. Although changes in each behavior may have unique determinants, it is possible that some hard-to-quantify change in the young adult cultural zeitgeist is driving changes in these proximate determinants of casual sexual activity, as well as trends in casual sex. Growing individualism and reduced sociability might lead to less partying (and hence less drinking), more computer gaming, and less autonomous living, while also diminishing the desire for sexual intercourse—at least the type of casual encounters captured in this analysis. Causation could also run in the reverse direction if a diminished desire for casual sex leads youth to party, and drink, less frequently and to play more computer games, perhaps all the while living in the proverbial parents’ basement. Quantifying these potentially distal sources of change in young adults’ causal sexual activity is likely to be difficult, so qualitative studies may have much to offer here. Future research might also profit by redressing some of the limitations of this analysis. The small sample size makes it difficult to detect significant associations or subgroup differences. Our measure of casual sexual activity is rather crude and could both undercount and overcount the nonromantic sexual encounters considered to be casual sex. The measure is also insensitive to heterogeneity in the types of these encounters. Trends in, and determinants of, one-time sex with strangers might differ substantially from sexual encounters between friends or former romantic partners and from encounters that one or both participants hope will lead to a more serious romantic relationship. College “hookups” might be a unique subtype of casual sexual encounters driven by a distinct set of factors (Allison 2016; England and Ronen 2015). We note as well that our analysis leaves unexplained a substantial portion of the decline in young adults’ casual sexual behavior, particularly among young women. Trends in the hypothesized mediating factors included in this analysis explain more than half of the decline in young men’s odds of engaging in casual sex but account for only about one quarter of the decline in young women’s probability of having a nonromantic sexual encounter. Further research is needed to identify additional causes of the decline in casual sexual activity among young adults. Perhaps the intensifying concern with interpersonal sexual violence and sexual coercion as exemplified in the #MeToo movement has begun inhibiting presumably voluntary casual sexual encounters between young women and men. The impact of this and other broad cultural shifts will also likely be difficult to measure but may well require consideration in order to develop a comprehensive assessment of the decline in young adults’ casual sexual activity.

Scott J. South https://orcid.org/0000-0002-0006-5133 Lei Lei https://orcid.org/0000-0003-2161-1217