Abstract

This article describes life-cycle change in the sexual behavior of undergraduates in “hookups” (which are outside traditional dates or relationships) during years 1–4 of college, explains a decline in the use of condoms, and shows how changes in the odds of coitus and condom use depend on family background, school gender imbalance, and whether the partners attend the same college. Coitus becomes increasingly likely as students progress through college. Condom-use rates, in contrast, decline precipitously between the freshman and sophomore years, before stabilizing. This rapid normalization of unprotected sex in hookups arises as students adapt to a high-education culture. Condom-use rates in the freshman year are lowest among students who have the most highly educated mothers. After the freshman year, students with lower-educated mothers converge to the same lower rate. This decline is pronounced when the partners attend the same college. Implications for health and social policy are discussed.

Introduction

This article describes life-cycle changes in undergraduate uncommitted sexual behavior during years 1–4 of college, explains a decline in the use of condoms, and shows how changes in the odds of coitus and condom use depend on family background, school gender imbalance, and whether the partners attend the same college. My aim is to show how the sexual risk-taking behaviors of undergraduates develop, sometimes rapidly, in part due to the environment in which they are embedded.

Despite the level of interest in the sexual culture on college campuses, social scientists have a limited understanding of undergraduate uncommitted sexual behavior and no picture of behavioral change over the course of a four-year college education. Our understanding is generally limited to qualitative studies of how women face a campus culture that conflicts with traditional gender norms (e.g., Bogle 2008; Hamilton and Armstrong 2009), and small-scale surveys of sexual behavior, typically limited to single classes, which treat sexual behaviors as static and ignore relationship context (Lewis, Malow, and Ireland 1997; Sheeran, Abraham, and Orbell 1999; Parsons et al. 2000; Beckman, Harvey, and Tiersky 2010). These studies reveal that students observe considerable social pressure to conform to peer norms, and use condoms with imperfect consistency. However, these studies do not inform us as to behavioral differences between lower- and upperclassmen, and what mechanisms explain changes in students’ behaviors.

My analysis focuses on differences between lower- and upperclassmen in the odds of coitus and condom use in sexual events that occur outside a traditional date or relationship—hereafter, hookups. I discuss these differences as behavioral changes between years 1–4 of college, but later address the limits of this interpretation, as I do not follow the same individuals over time. To provide context, I first describe prior research on undergraduate sexual behavior. Whereas past research on college students has implicitly assumed that coital and contraceptive behavior is largely static, I show that students’ behaviors substantively change during college. I thereafter propose a parsimonious model that accounts for a sharp decline in condom use. Finally, I discuss potential implications for health and social policy in the conclusion.

To preview the results, (1) the odds of coitus when hooking up double between the freshman and senior years; (2) when coitus occurs during hookups, the odds of condom use decrease by half by the sophomore year; (3) the entirety of the decline in condom use is accounted for by lower-SES students adopting the behavior exhibited by higher-SES students at the onset of college; (4) change in female behavior is somewhat sensitive to a school's gender composition; and (5) the decline in condom use is especially pronounced in cases where the partner attended the respondent's college. These results suggest that the college environment substantially affects undergraduate sexual behavior. Consequentially, models that assume that behaviors are static or develop linearly with age may lead to a misunderstanding of the determinants of sexual risk behavior.

Background

Studies of Undergraduate Sexual Scripts

As early as the 1930s, sociologists addressed undergraduate sexual behavior, when Waller (1937) expressed the concern that a thrill-seeking college sexual culture would lead to the end of marriage as we knew it. Waller (1937) described an environment in which (1) dating that was not intended to lead to marriage was sought out by both genders; (2) men and women were rated in a rigid social hierarchy by their appearance, material wealth, and ability to find dates; and (3) failure to “rate” led to psychological stress and rationalizations to cope with social and sexual rejection.

Whereas Waller (1937) discussed the transition from courtship to dating, more recent scholarship discusses the transition from dating to a more casual form of sexual behavior commonly known as “hooking up” (Bogle 2008; Armstrong, England, and Fogarty 2012). Unlike the previous transition, hooking up may not replace dating. Rather, it may remain normative only during college (Bogle 2008).

In the early part of the prior century, the scarcity of women on college campuses may have empowered women (Waller 1937). The subsequent reversal, so that women outnumber men, combined with the rise of hookups, leads to the concern that the college social scene now disempowers women; gender norms suggest that hookups are more desirable for men, whereas, due to the double standard, they continue to conflict with norms for women's behavior even after the sexual revolution, such that women are more stigmatized than men for casual sex (Guttentag and Secord 1983; Bogle 2008; Hamilton and Armstrong 2009; Kreager and Staff 2009; Armstrong, Hamilton, and England 2010; Armstrong, England, and Fogarty 2012). Uecker and Regnerus (2010), analyzing a sample of undergraduate women, find that a relative scarcity of men affects women's behaviors; since this affects sexual activity within as well as outside relationships, they argue that the effect is due not only to the shortage of available male partners, but also to the balance of power within dyads.

A key qualitative similarity between college culture in the earlier part of the twentieth century and now remains the extent to which social status and class background affect students’ ability to navigate the social environment in college. Students from different socioeconomic backgrounds may find college more or less similar to the environments in which they grew up. Qualitative research has documented the sense of difficulty expressed by lower-SES students in adapting to and coping with the social environment in college (Hamilton and Armstrong 2009; Armstrong, Hamilton, and England 2010), but it remains unclear how socioeconomic status affects the development of students’ sexual risk behaviors.

Studies of Social Status and Health Behavior

Prior research shows a positive relationship between health behavior and socioeconomic status (Link and Phelan 1995; Pappas et al. 1993; Beaglehole 1990). Higher-SES respondents tend to know more about the risks associated with unhealthy behaviors such as smoking and overeating (Link and Phelan 1995; Beaglehole 1990). Those activities are equally dangerous regardless of the prior or concurrent activities of others, but the potential consequences of sexual activity relate to the prior, not directly observable, behaviors of the partner. SES's relationship with condom use may differ from its relationship with food and drug habits.

Past research shows that higher-SES women more effectively prevent unintended births (by preventing unplanned pregnancy and more consistently aborting unplanned births [England, McClintock, and Shafer 2011; Musick et al. 2009]), but it does not necessarily follow that higher-SES men and women use condoms more. In contrast to hormonal birth control, intrauterine devices, and abortion, the decision to use a condom must be made while in the course of sexual activity, during which judgment may be compromised, potentially inhibiting condom use. Further, if higher-SES women more consistently use other methods of birth control (even if these do not prevent infection), higher-SES men and women may consider condom use less important.

Studies of Undergraduate Condom Use

Prior research in the HIV/AIDS literature discusses the psychosocial correlates of condom use among college students (see Lewis, Malow, and Ireland [1997] for a review). These studies are typified by three key limitations: (1) they have a small sample size, for example from a single university or from a single section of a course; (2) they implicitly assume that lower- and upperclassmen exhibit the same behaviors; and (3) they assess condom use at last intercourse without disentangling intercourse that occurs in relationships and dates from intercourse that occurs in hookups. Regardless, this research indicates that college students do not consistently use condoms, though they are aware of the potential consequences of infection (Lewis, Malow, and Ireland 1997; Sheeran, Abraham, and Orbell 1999; Parsons et al. 2000; Beckman, Harvey, and Tiersky 2010).

The assumption that relationship type does not matter, or that it is merely one factor among many that affects what happens when sex occurs, is problematic. While pathogens may not care about the social context of sexual activity, humans do (Tavory and Swidler 2009). Prior research on sexual relationships among 1990s high school students shows that contraceptive practices differ between students who do and students who do not engage in romantic behavior (such as hand-holding) with their partners (Manlove, Ryan, and Franzetta 2004; Manning, Longmore, and Giordano 2005). These distinctions may also matter in college. Moreover, additively controlling for relationship context does not address whether its importance changes over time, varies by group, or interacts with other factors.

Within relationships, higher-SES adolescents use condoms less than lower-SES adolescents, although they are more apt to use some kind of contraception (England, McClintock, and Shafer 2011; Manning et al. 2009). Whether the same holds for intercourse within hookups, however, is an open question. Students may trust their partners in a relationship, but less so in hookups. Students may trust higher-SES partners more than lower-SES partners, and this may lead to differences in behavior when both partners are high status, where status is a function of SES and other qualities: qualitative evidence suggests that individuals underestimate risk in unprotected sex based on factors including perceived trust and the personal appearance of the partner (East et al. 2007; Hoffman and Cohen 1999).

Educational assortative mating increases the probability that spouses will have similar levels of educational attainment and that they will have attended institutions with similar characteristics (Mare 1991; Schwartz and Mare 2005; Arum, Roksa, and Budig 2008). College attendance and socioeconomic status may also affect how partners sort into hookups. Students may assess potential hookup partners quickly and subjectively based on SES, attractiveness, and network ties (Hoffman and Cohen 1999; East et al. 2007; Downing-Matibag and Geisinger 2009). An intriguing aspect of college attendance in particular is that students may easily observe whether the partner attends the same college; a relative stranger could thus be accorded the status of being (loosely) within one's social network, which may in turn affect perceptions of trust and safety.

Students may underestimate risks based on their perceptions of the community in which they are embedded (Downing-Matibag and Geisinger 2009). For students from lower-SES backgrounds in particular, this pool may be of substantially higher SES than that which they had previously had access to, because of both their backgrounds and their future prospects. Within the college environment, furthermore, they may observe fewer consequences of unprotected intercourse than they observed in high school. This may help desensitize students to the risks of unprotected intercourse. Lower-SES students may be particularly likely to reduce condom use while in college, as they experience an elevation in the socioeconomic status of their peers between high school and college, and because of this find themselves in an environment where visible, negative consequences of unprotected sex (such as pregnancy and HIV) are less common than they were in their high school. Both factors may lead them to let their guard down. Consistent with this, a recent study following 279 women through their first nine months of college at a Northeastern university found that condom use declined more among low-SES women (Walsh et al. 2013). Unfortunately, their data were insufficient to separately examine sex in hookups.

In sum, prior research on undergraduate condom use shows that students, though they may be aware of the potential consequences of unprotected intercourse, do not use condoms consistently. In part a product of the HIV/AIDS literature, which typically studies populations with high infection rates, these studies consider coitus, regardless of its context. In relatively low-risk populations, such as students attending four-year colleges and universities in the United States, the aggregate risk of infection within a relationship is, technically, nonzero, although if monogamy is practiced, it may be fairly low. But with a casual liaison, one may well be with a partner who has had other casual liaisons. So, to assume that the decision-making process is similar across relationships, dates, and hookups may be unrealistic. Finally, analyses to date have shown associations between risk behaviors, but have not analyzed whether, or explained why, lower- and upperclassmen behave differently.

The Present Study

Attending a four-year college or university substantially shapes social structure in ways that may affect sexual behavior. Students have reduced parental supervision while adjusting to a new social environment. Many students are exposed to new behaviors as the socioeconomic characteristics, and even the gender balance, of their peer groups, change (DiPrete and Buchmann 2013; DiPrete and Buchmann 2006). This article analyzes data on coitus and condom use outside a traditional date or relationship in the late adolescent society of American undergraduates. It ascertains the typical behavioral patterns of coitus and condom use when coitus occurs, how quickly these behaviors develop, and analyzes potential explanations for the development of these behaviors.

Data

I utilize data from the Online College Social Life Survey—a 15 to 20-minute survey administered between 2005 and 2011 (OCSLS, Paula England, PI). The survey, which is described further by Armstrong, England, and Fogarty (2012), recruited students from undergraduate classes at 21 four-year colleges and universities. It contains data on more than 10,000 hookups among students in a moderately selective environment. This contrasts with the samples typically used to analyze the sexual behaviors of undergraduates, which are generally at least an order of magnitude smaller, and sometimes rely on a sample of students within a course (see Lewis, Malow, and Ireland [1997] and Sheeran, Abraham, and Orbell [1999] for reviews). These data permit comparing lower- and upperclassmen, by sex and by SES, subdividing hookups by whether the partners attend the same college, and stepping back from condom use at last sex so as to first address whether coitus occurs. In contrast to the Add Health data frequently used to assess the sexual and health behaviors of 1990s high school students, however, students were not reinterviewed, and the data were collected by snowballing an increasingly large number of professors who were willing to offer the survey to the students in their classes.1 While not a probability sample, these data notably represent a near census in classes where students were recruited (Armstrong, England, and Fogarty 2012).2 Despite their limitations, these data remain the best available to answer the questions posed in this analysis. (For copies of the data, the questionnaire, and a codebook with detailed information on variable construction, see https://files.nyu.edu/jmb736/public/ocsls/. This website also includes the program code that constructs the dataset and codebook and the program that executes the analysis described below.)

I use the large sample of undergraduates to analyze a synthetic cohort. This approach contrasts freshmen, sophomores, juniors, and seniors who were part of the same single wave of data collection. This synthetic cohort approach, which takes cross-sectional differences between freshmen and upperclassmen (net of controls) to illuminate how students change across their college years, rests on the assumption that the sexual behaviors of first-year students and later undergraduates at four-year colleges and universities are comparable, except that the latter have completed 1–3 additional transitions through the college years. As discussed in more detail in the Method section, the analysis begins by testing whether compositional differences account for any differences between students by year of college. As further discussed in the Results section, the time trends are unaffected by demographic controls, making the assumption of the synthetic cohort approach more reasonable. However, it is impossible to rule out unobserved differences by year of college. A particular concern might be differential attrition by socioeconomic status. Such attrition could be a problem if, among lower-SES students, those most likely to use condoms drop out in large numbers. However, no such attrition is evident in these data; there are no fewer low-SES seniors than there are low-SES freshmen, proportionally.

The survey asks students which sexual behaviors they have ever experienced, and what sexual behaviors occurred on their most recent date, at the last sexual occurrence within a relationship lasting at least six months, and on their most recent hookup since starting college. For the latter, students are instructed to use whatever definition of hookup is understood by their peers, with the clarification that this cannot be an event that occurred within a relationship.

In contrast to the list of behaviors students have ever experienced, which cannot decrease as they age, typical sexual behaviors for hookups may change in any direction. The former are described in table 1, which may be of particular use for readers unfamiliar with the sexual behaviors of college students. These figures pertain to the proportion of students who have passed specific thresholds for sexual activity, by year of college. These statistics necessarily show a monotonically non-decreasing trend, since they report the percentage of students who have exhibited a behavior at least once. Many students continue to experience first events as they progress through college. Panel (a) reports percentages for all students; panel (b) reports percentages for non-virgins; and panel (c) reports percentages for students who have had intercourse outside a relationship. These figures are discussed further below.

Table 1.

Percent of Students Who Have Passed Sexual Activity Thresholds, by Gender and Grade

 (a)
 
(b)
 
(c)
 
All students
 
Non-virgins
 
S's who have had non-exclusive intercourse
 
Women Men Women Men Women Men 
Have you ever had vaginal intercourse? 
 Freshmen 63% 60%     
 Sophomores 64% 64%     
 Juniors 74% 69%     
 Seniors 77% 70%     
Have you ever had intercourse outside an exclusive relationship? 
 Freshmen 37% 40% 56% 59%   
 Sophomores 39% 46% 58% 64%   
 Juniors 49% 50% 63% 64%   
 Seniors 53% 53% 66% 67%   
Have you ever had two ongoing sexual partnerships involving intercourse at the same time? 
 Freshmen 8% 12% 13% 17% 21% 27% 
 Sophomores 9% 15% 14% 21% 22% 29% 
 Juniors 13% 17% 17% 22% 24% 31% 
 Seniors 16% 20% 19% 26% 28% 36% 
Have you ever had sex with one person, then had sex with another partner, then had sex with the first person? 
 Freshmen 20% 22% 31% 32% 49% 49% 
 Sophomores 23% 27% 34% 38% 55% 54% 
 Juniors 30% 31% 39% 39% 57% 58% 
 Seniors 35% 36% 43% 46% 61% 62% 
 (a)
 
(b)
 
(c)
 
All students
 
Non-virgins
 
S's who have had non-exclusive intercourse
 
Women Men Women Men Women Men 
Have you ever had vaginal intercourse? 
 Freshmen 63% 60%     
 Sophomores 64% 64%     
 Juniors 74% 69%     
 Seniors 77% 70%     
Have you ever had intercourse outside an exclusive relationship? 
 Freshmen 37% 40% 56% 59%   
 Sophomores 39% 46% 58% 64%   
 Juniors 49% 50% 63% 64%   
 Seniors 53% 53% 66% 67%   
Have you ever had two ongoing sexual partnerships involving intercourse at the same time? 
 Freshmen 8% 12% 13% 17% 21% 27% 
 Sophomores 9% 15% 14% 21% 22% 29% 
 Juniors 13% 17% 17% 22% 24% 31% 
 Seniors 16% 20% 19% 26% 28% 36% 
Have you ever had sex with one person, then had sex with another partner, then had sex with the first person? 
 Freshmen 20% 22% 31% 32% 49% 49% 
 Sophomores 23% 27% 34% 38% 55% 54% 
 Juniors 30% 31% 39% 39% 57% 58% 
 Seniors 35% 36% 43% 46% 61% 62% 

In this paper, I focus on undergraduates’ most recent hookups. Hooking up is normative, but this does not mean that nearly all college students engage in intercourse when they hook up; hookups may involve no more than kissing or making out. Though the survey does not ask students whether they have ever had sex during a hookup, it does ask whether they have ever had intercourse outside a relationship. Among freshmen, 37 percent of women and 40 percent of men, and among seniors, 53 percent of either gender, report that they have had intercourse outside a relationship. Thus, while a majority of undergraduates engage in intercourse outside a relationship, nearly half do not.

This contrast is somewhat misleading, and is better understood after accounting for the fact that many students have never been sexually active. Among freshmen, 37 percent of women, and 40 percent of men, report that they have not had vaginal intercourse. (These figures closely track the proportion of virgins in the general population, wherein 37 percent and 36 percent of never-married women and men, respectively, between the ages of 18 and 19, report that they have never had intercourse [Martinez, Copen, and Abma 2011]). By the senior year, the proportion of students who have not had intercourse decreases to 23 percent of women and 30 percent of men—still a nontrivial amount. Among students who have had intercourse, a substantial majority—57 percent of freshmen and 66 percent of seniors—report that they have had intercourse outside the context of an exclusive relationship. Among the students who have ever had intercourse, in sum, it is fairly normal to have had intercourse outside a relationship. This is evident in the freshman year, and changes comparatively little thereafter.

This distinction between the behavior of students who have and who have not ever had sex is important, because it suggests that, conditional on becoming sexually active for the first time, there is little change over the course of college in the proportion of students who have ever had intercourse outside a relationship. There appears to be a fairly static group of students who will have intercourse outside a relationship, provided they have had sex in some context. This indicates not that some students participate in a casual sexual culture and others do not, but rather that some students avoid intercourse altogether, a second and larger group has sex only within a relationship, and a third group, the largest, also engages in intercourse outside a relationship.

Students who have engaged in intercourse outside a relationship report substantially higher rates of non-monogamous behaviors than all students or all non-virgins. Nearly a third of seniors who have had non-relationship sex—28 percent of women and 36 percent of men—report having had two ongoing sexual partnerships at the same time. Nearly two-thirds of seniors who have had non-relationship sex, furthermore, report having had sex with one person, then with another person, and then with the first person.

These figures, in sum, provide some insight into the population of undergraduates, hopefully providing greater depth and accuracy than might be gleaned from representations of undergraduates in popular culture. The subject of this paper, however, is changes in what typically happens when students hook up. As students progress through college, does intercourse become more common when they hook up, and when students have coital hookups, do the odds of condom use increase, decrease, or remain the same?

Prior research shows that teenagers are much less likely to use a condom at first sex if they were already in a relationship with their partner (Manning, Longmore, and Giordano 2000). This pattern is also evident among high school students who have had multiple partners (Manlove, Ryan, and Franzetta 2004). Similarly, in the data used here, 52 percent of the women and 58 percent of the men reported using a condom the last time they had vaginal intercourse within a long-term relationship, in contrast to 67 percent of women and 74 percent of men when they last had sex within a hookup. Students use condoms more often when they hook up, but condom non-use remains nontrivial.

Most undergraduates report engaging in intercourse outside the context of a relationship by their senior year, and when they do so, it is often unprotected sex. Are these behaviors affected by the college environment? No prior evidence exists as to whether, to what extent, and why undergraduates’ sexual risk behaviors change as they progress through college.

Method

I restrict the sample to self-identifying heterosexual students in their 1st, 2nd, 3rd, and 4th year of college who are of typical age, 18, 19, 20, or 21 plus or minus one year, respectively. (Reestimating all models without these restrictions produces similar results.) Table 2 reports the means for all the variables used in the analysis. The average student attends a school that is 53 percent female, but due to the representation of women in the sampled classes, many of which were in Sociology, two-thirds of my sample is female. My analytic sample includes 7,144 female and 3,131 male respondents reporting on their most recent hookup with a member of the other sex since starting college. (Questions that pertain to a specific recent event minimize recall error and disclosure bias [Morris 1993; Younge et al. 2008]). Of these, 2,582 (36 percent) and 1,252 (40 percent), respectively, involved coitus.

Table 2.

Descriptive Statistics

 Hookups
 
Hookups with sex
 
Women Men Women Men 
Coitus occurred 0.36 0.40   
Used a condom   0.67 0.74 
Year of college     
 Freshman year 0.35 0.33 0.29 0.28 
 Sophomore year 0.24 0.26 0.23 0.27 
 Junior year 0.21 0.21 0.24 0.22 
 Senior year 0.20 0.19 0.25 0.24 
Mother's educational attainment 
 High school education or less 0.21 0.19 0.24 0.21 
 Some college education 0.24 0.23 0.25 0.24 
 Bachelor's degree 0.33 0.32 0.31 0.33 
 Graduate degree 0.22 0.25 0.20 0.22 
Population density during high school 
 Rural 0.06 0.05 0.06 0.06 
 Small town 0.24 0.21 0.26 0.22 
 Suburban 0.36 0.37 0.34 0.35 
 Medium city 0.22 0.24 0.22 0.24 
 Large city 0.12 0.13 0.12 0.14 
Respondent's race 
 White 0.73 0.73 0.74 0.73 
 Black 0.05 0.06 0.06 0.07 
 Hispanic 0.11 0.10 0.10 0.10 
 Asian 0.07 0.07 0.06 0.06 
 Other race 0.04 0.04 0.04 0.04 
 Mixed race 0.13 0.13 0.13 0.13 
Partner's race     
 White 0.70 0.72 0.69 0.73 
 Black 0.08 0.03 0.09 0.04 
 Hispanic 0.08 0.06 0.08 0.06 
 Asian 0.04 0.08 0.04 0.07 
 Mixed race 0.08 0.08 0.08 0.08 
 Other or not indicated 0.02 0.02 0.02 0.02 
Childhood religion     
 Catholic 0.41 0.41 0.40 0.42 
 Jewish 0.09 0.11 0.08 0.10 
 Other 0.16 0.13 0.17 0.13 
 Protestant 0.14 0.16 0.13 0.15 
 None 0.21 0.20 0.22 0.20 
Immigrant 0.07 0.09 0.07 0.09 
Observations 7,144 3,131 2,582 1,252 
 Hookups
 
Hookups with sex
 
Women Men Women Men 
Coitus occurred 0.36 0.40   
Used a condom   0.67 0.74 
Year of college     
 Freshman year 0.35 0.33 0.29 0.28 
 Sophomore year 0.24 0.26 0.23 0.27 
 Junior year 0.21 0.21 0.24 0.22 
 Senior year 0.20 0.19 0.25 0.24 
Mother's educational attainment 
 High school education or less 0.21 0.19 0.24 0.21 
 Some college education 0.24 0.23 0.25 0.24 
 Bachelor's degree 0.33 0.32 0.31 0.33 
 Graduate degree 0.22 0.25 0.20 0.22 
Population density during high school 
 Rural 0.06 0.05 0.06 0.06 
 Small town 0.24 0.21 0.26 0.22 
 Suburban 0.36 0.37 0.34 0.35 
 Medium city 0.22 0.24 0.22 0.24 
 Large city 0.12 0.13 0.12 0.14 
Respondent's race 
 White 0.73 0.73 0.74 0.73 
 Black 0.05 0.06 0.06 0.07 
 Hispanic 0.11 0.10 0.10 0.10 
 Asian 0.07 0.07 0.06 0.06 
 Other race 0.04 0.04 0.04 0.04 
 Mixed race 0.13 0.13 0.13 0.13 
Partner's race     
 White 0.70 0.72 0.69 0.73 
 Black 0.08 0.03 0.09 0.04 
 Hispanic 0.08 0.06 0.08 0.06 
 Asian 0.04 0.08 0.04 0.07 
 Mixed race 0.08 0.08 0.08 0.08 
 Other or not indicated 0.02 0.02 0.02 0.02 
Childhood religion     
 Catholic 0.41 0.41 0.40 0.42 
 Jewish 0.09 0.11 0.08 0.10 
 Other 0.16 0.13 0.17 0.13 
 Protestant 0.14 0.16 0.13 0.15 
 None 0.21 0.20 0.22 0.20 
Immigrant 0.07 0.09 0.07 0.09 
Observations 7,144 3,131 2,582 1,252 

More non-virgins hook up than date, but that does not make hooking up universal. More than two-thirds of all students (73 percent of women and 72 percent of men), and about four-fifths of non-virgins (79 percent of women and 82 percent of men), hook up by the senior year. A similar portion of non-virgin seniors have gone out on a date with someone with whom they were not already in an exclusive relationship (83 percent of women and 80 percent of men). Among freshmen and sophomores (of both sexes), furthermore, more non-virgins have hooked up than have gone on dates. Hooking up, in sum, is at least as normative, if not more so, than dating, but I caution against generalizing the results that follow to the fraction of students who do not hook up.

Because the decision to have coitus is not independent of the subsequent decision to use a condom, I analyze these separately. I subsequently discuss predicting the distribution of three outcomes—when intercourse does not occur (no coitus); when intercourse occurs with condom use (protected coitus); and when intercourse occurs without condom use (unprotected coitus)—as the product of a sequential process in which students first make a decision about coitus. (Students may later revisit the original coital decision—e.g., due to condom availability—but this information is not available.)

Equations

I estimate a series of logistic probability models, separately by sex. The key independent variable compares lower- and upperclassmen. To make the results of the coitus and condom-use analyses easy to compare, I estimate corresponding equations in which I change only the response variable.

The first model includes no control variables. The next model ascertains whether compositional differences affect the estimated time trend. This second model includes controls for student characteristics and school random effects. (School fixed effects produce similar results, but absorb school gender composition, which I measure in a third model.) The demographic controls in this model include the respondent's and partner's race, the educational attainment of the respondent's mother, the religion in which the respondent was raised, whether the respondent's high school was urban, suburban, or rural, and whether the respondent is an immigrant.

I use mother's education as an indicator of the socioeconomic status of the family the student grew up in. The survey did not ask father's education, or about parental occupations or earnings.

There are three race variables. First, I include a categorical based on a question in which the respondent could choose only one race. Second, I include a dichotomy, “mixed race,” which indicates whether the respondent selected multiple races from an open-ended list. A third variable indicates the partner's race, based on the respondent's report. As the respondent could select multiple races for the partner, and was not separately asked in a subsequent question to pick just one, the partner's race variable differs from the respondent's race variable in that “mixed race” is mutually exclusive with respect to the other categories.

A third model addresses whether change over time varies based on the respondent's SES, and the gender composition of the respondent's college. The environmental change induced by attending college may be greater for students from less educated backgrounds, given that parental education predicts who goes to college. The gender composition of the college may also affect sexual behavior, either directly, as students compete among balanced or imbalanced pools of potential partners, or indirectly, as students may search for partners outside the college environment, which may in turn also influence their behavior with on-campus partners. This model includes an interaction between year of college and SES, as well as an interaction between year of college and an indicator for whether the school has more women than average. This variable equals 1 if a respondent attends a school in which 53 percent or more of the student body is female. (This threshold was chosen because it balances the sample at the student and school levels; half of the students and exactly half of the schools fall on either side of this threshold. Results are insensitive to minor arbitrary adjustments of this variable.)

In this third equation, I collapse the years of college into a dichotomy to compare students during and after the freshman year. This eases interpretation and maintains cell size, given the number of coefficients for the interaction terms. Results from likelihood ratio tests comparing more complex specifications (not dichotomizing the years of college), and odds ratios from these equations, are reported in a supplement.3 These supplementary analyses support the results of the main analysis.

Sensitivity Analyses

Since students may feel safer with students from their campus, lower-SES students may take longer to integrate into the campus social scene, and students could hook up off campus more (or less) as they age, I twice repeat the above analyses: I separately analyze hookups in which the partner does or does not attend the respondent's college. Since sexual behavior may relate to familiarity or affection, I also estimate models that control for prior hookups with the same partner. Duration of exposure to sexual activity may affect the odds of sex and condom use, but it is exogenous only among students who had sex before college; I estimate additional models that ask whether this explains all, some, or none of the relationship between college transitions and the odds of coitus and condom use when coitus occurs. Finally, students may become desensitized as they accumulate partners. Introducing such a control poses a reverse-causality problem: students who are more likely to have non-romantic intercourse should be more likely to accumulate partners. This dilemma may be of lesser concern when modeling condom use; for symmetry, I report results for both outcomes.

Results

How Do the Odds of Coitus and Condom Use Change in Years 1–4 of College?

Table 3 reports odds ratios from Equations 1 and 2, before and after adjusting for respondent characteristics and school random effects, respectively. The left panel presents models estimating coitus, whereas the right panel presents models estimating condom use in events involving coitus. The similarity between the Equation 1 odds ratios and the Equation 2 odds ratios for grade indicates that the time trends discussed below are not substantively related to compositional differences between students across grades. This increases confidence in the synthetic cohort approach taken here.

Table 3.

Odds Ratios for Equations 1 and 2 (before and after including controls)

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 1
 
Equation 2
 
Equation 1
 
Equation 2
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.15* 1.42*** 1.29*** 1.53*** 0.69** 0.66* 0.68** 0.62* 
 Junior 1.62*** 1.38** 1.79*** 1.53*** 0.72** 0.63* 0.72** 0.63* 
 Senior 1.82*** 1.89*** 2.09*** 2.27*** 0.63*** 0.62* 0.63*** 0.62* 
Mother's educational attainment 
High school education or less 
 Some college education   0.85* 0.87   0.92 0.90 
 Bachelor's degree   0.80** 0.94   0.86 1.03 
 Graduate degree   0.84* 0.73*   0.89 0.80 
Population density during high school         
 Rural   0.98 1.44*   0.90 0.86 
 Small town   1.25** 1.03   1.12 0.90 
 Suburban         
 Medium city   1.13^ 1.06   1.23^ 0.88 
 Large city   1.32** 1.26^   1.29^ 1.37 
Respondent's race         
 White         
 Black   1.02 1.05   0.96 2.63* 
 Hispanic   0.77** 1.00   1.32 1.16 
 Asian   0.90 0.72^   0.99 1.19 
 Other race   0.81 1.01   1.24 0.97 
 Mixed race   1.12 1.18   1.13 1.10 
Partner's race         
 White         
 Black   1.39** 0.96   1.77** 0.73 
 Hispanic   0.95 0.86   0.94 0.87 
 Asian   0.89 0.76^   0.82 1.20 
 Mixed race   1.25* 1.03   0.91 1.11 
 Other or not indicated   0.68* 0.78   0.68 1.22 
Childhood religion         
 Catholic   0.84* 0.96   0.92 1.53* 
 Jewish   0.77* 0.90   1.30 1.43 
 Other   0.93 1.09   0.84 1.05 
 Protestant   0.79** 0.91   1.24 1.33 
 None         
Immigrant   0.98 0.97   1.04 1.43 
Observations 7144 3131 7144 3131 2582 1252 2582 1252 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 1
 
Equation 2
 
Equation 1
 
Equation 2
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.15* 1.42*** 1.29*** 1.53*** 0.69** 0.66* 0.68** 0.62* 
 Junior 1.62*** 1.38** 1.79*** 1.53*** 0.72** 0.63* 0.72** 0.63* 
 Senior 1.82*** 1.89*** 2.09*** 2.27*** 0.63*** 0.62* 0.63*** 0.62* 
Mother's educational attainment 
High school education or less 
 Some college education   0.85* 0.87   0.92 0.90 
 Bachelor's degree   0.80** 0.94   0.86 1.03 
 Graduate degree   0.84* 0.73*   0.89 0.80 
Population density during high school         
 Rural   0.98 1.44*   0.90 0.86 
 Small town   1.25** 1.03   1.12 0.90 
 Suburban         
 Medium city   1.13^ 1.06   1.23^ 0.88 
 Large city   1.32** 1.26^   1.29^ 1.37 
Respondent's race         
 White         
 Black   1.02 1.05   0.96 2.63* 
 Hispanic   0.77** 1.00   1.32 1.16 
 Asian   0.90 0.72^   0.99 1.19 
 Other race   0.81 1.01   1.24 0.97 
 Mixed race   1.12 1.18   1.13 1.10 
Partner's race         
 White         
 Black   1.39** 0.96   1.77** 0.73 
 Hispanic   0.95 0.86   0.94 0.87 
 Asian   0.89 0.76^   0.82 1.20 
 Mixed race   1.25* 1.03   0.91 1.11 
 Other or not indicated   0.68* 0.78   0.68 1.22 
Childhood religion         
 Catholic   0.84* 0.96   0.92 1.53* 
 Jewish   0.77* 0.90   1.30 1.43 
 Other   0.93 1.09   0.84 1.05 
 Protestant   0.79** 0.91   1.24 1.33 
 None         
Immigrant   0.98 0.97   1.04 1.43 
Observations 7144 3131 7144 3131 2582 1252 2582 1252 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

Results show that the odds of coitus approximately double between the freshman and senior years, while the odds of condom use in coital events decrease by about 40 percent. The odds ratio for coitus, comparing year 4 to year 1, net of controls, is 2.09 for women and 2.27 for men (both significant at p < .001). The odds ratio for condom use in coital events, comparing year 4 to year 1, net of controls, is 0.63 (p < .001) for women and 0.62 (p < .01) for men.

The time trends are strikingly similar regardless of the sex of the respondent. Since desirability bias would affect the results in opposite ways for men and women, this consistency lends additional support to the accuracy of the reporting.

The four panels in figure 1 graph the predicted probabilities of coitus and condom use in coital events by sex and year of college. The left panels report the results for women, and the right panels report the results for men. The top panels report the predicted probabilities of coitus, and the bottom panels report the predicted probabilities of condom use in coital events.

Figure 1.

Predicted probabilities of coitus and condom use

Figure 1.

Predicted probabilities of coitus and condom use

Among women, the predicted probability of coitus net of controls (model 2) increases from 27 percent in the freshman year to 33 percent, 40 percent, and 43 percent, in the sophomore, junior, and senior years (figure 1, panel A). The predicted probability of condom use in coital events, meanwhile, declines from 73 percent in the freshman year to 64 percent, 66 percent, and 63 percent in the sophomore, junior, and senior years (figure 1, panel B).

Among men, the predicted probability of coitus net of controls (model 2) increases from 33 percent in the freshman year to 42 percent, 42 percent, and 52 percent in the sophomore, junior, and senior years (figure 1, panel C). The predicted probability of condom use in coital events, meanwhile, declines from 80 percent in the freshman year to 72 percent, 72 percent, and 72 percent in the sophomore, junior, and senior years (figure 1, panel D).

Whereas the odds of coitus increase at a comparatively gradual pace between the freshman and senior years, the odds of condom use, in contrast, decrease more abruptly. A fairly rapid normalization of unprotected casual sex occurs between the freshman and sophomore years, and thereafter condom use in coital hookups decreases little.

Figures 2a and 2b show for women and men, respectively, how changes in students’ coital and condom-use behaviors contribute to unprotected casual sex when students hook up. The probability of unprotected casual sex increases from 7 percent to 16 percent among women, and from 6 percent to 15 percent among men, between the freshman and senior years. The increasing odds of coitus explain more of this than the decreasing odds of condom use. If sophomores, juniors, and seniors were no less likely to use condoms, the probability of unprotected casual sex would increase from 7 percent to 12 percent among women, and from 6 percent to 10 percent among men, between the freshman and senior years. If sophomores, juniors, and seniors were no more likely than freshmen to have sex, the probability of unprotected casual sex would increase from 7 percent to 10 percent among women, and from 6 percent to 9 percent among men, between the freshman and senior years. To prevent seniors from having more unprotected casual sex than freshmen, women would have to use condoms 83 percent of the time, and men 87 percent of the time, instead of 63 percent and 72 percent of the time, respectively, in the senior year; in proportionate terms, this means an intervention that increases senior women's condom use by 31 percent, and senior men's by 22 percent. Alternatively, women and men in their senior year would have to have, respectively, 53 percent and 56 percent proportionately less sex. A conventional analysis, which only evaluates the odds of condom use at last intercourse, would miss these points; if condom use but not coitus is a legitimate policy target, then an intervention to affect condom use would have to be particularly effective.

Figure 2.

Predicted probabilities of unprotected intercourse

Figure 2.

Predicted probabilities of unprotected intercourse

What Could Account for Changes in Students’ Behaviors?

Table 4 reports odds ratios from model 3, which assesses student and environmental variation in the time trends. Grade is collapsed into a dichotomous categorical variable indicating whether or not a student is beyond the freshman year. This increases cell size and eases interpretation, to allow for two interactions with grade: the educational attainment of the respondent's mother, and whether the school has more women than average. Additional results that do not collapse grade into a dichotomy support the results shown below, pooling the sexes to increase power when estimating condom use, and are available in the online supplement.

Table 4.

Odds Ratios for Equation 3

 Coitus
 
Condom use
 
Women Men Women Men 
During/after freshman year 
 First year of college 
 Years 2/3/4 of college 1.31* 1.37 0.51** 0.36* 
Mother's educational attainment     
 High school education or less 
 Some college education 0.83 0.72 0.90 0.46^ 
 Bachelor's Degree 0.88 0.77 0.73 0.74 
 Graduate Degree 0.82 0.58** 0.55* 0.35* 
Grade × mother's educational attainment     
 Years 2/3/4 × some college 1.03 1.31 1.02 2.35^ 
 Years 2/3/4 × bachelor's degree 0.83 1.31 1.22 1.51 
 Years 2/3/4 × graduate degree 1.00 1.40 1.92* 2.94* 
School gender composition     
 < 53% Female     
 ≥ 53% Female 1.14 1.06 1.05 1.21 
 ≥ 53% Female × years 2/3/4 1.50*** 0.97 1.13 0.86 
Population density during high school 
 Rural 0.96 1.42* 0.88 0.87 
 Small town 1.29*** 1.02 1.11 0.91 
 Suburban     
 Medium city 1.15* 1.05 1.23^ 0.89 
 Large city 1.27** 1.25^ 1.29^ 1.37 
Respondent's race     
 White     
 Black 1.03 1.04 0.98 2.61* 
 Hispanic 0.76** 1.02 1.31 1.20 
 Asian 0.91 0.73^ 0.99 1.22 
 Other 0.82 1.01 1.23 0.92 
 Mixed 1.14 1.17 1.13 1.08 
Partner's race     
 White     
 Black 1.38** 0.95 1.78** 0.74 
 Hispanic 0.98 0.84 0.94 0.84 
 Asian 0.87 0.78 0.81 1.19 
 Mixed 1.27* 1.01 0.91 1.14 
 Other or not indicated 0.73^ 0.77 0.67 1.23 
Childhood religion     
 Catholic 0.84* 0.95 0.92 1.53* 
 Jewish 0.77* 0.89 1.34 1.49 
 Other 0.91 1.06 0.84 1.08 
 Protestant 0.77** 0.90 1.24 1.32 
 None     
Immigrant 0.99 0.98 1.04 1.41 
Observations 7144 3131 2582 1252 
 Coitus
 
Condom use
 
Women Men Women Men 
During/after freshman year 
 First year of college 
 Years 2/3/4 of college 1.31* 1.37 0.51** 0.36* 
Mother's educational attainment     
 High school education or less 
 Some college education 0.83 0.72 0.90 0.46^ 
 Bachelor's Degree 0.88 0.77 0.73 0.74 
 Graduate Degree 0.82 0.58** 0.55* 0.35* 
Grade × mother's educational attainment     
 Years 2/3/4 × some college 1.03 1.31 1.02 2.35^ 
 Years 2/3/4 × bachelor's degree 0.83 1.31 1.22 1.51 
 Years 2/3/4 × graduate degree 1.00 1.40 1.92* 2.94* 
School gender composition     
 < 53% Female     
 ≥ 53% Female 1.14 1.06 1.05 1.21 
 ≥ 53% Female × years 2/3/4 1.50*** 0.97 1.13 0.86 
Population density during high school 
 Rural 0.96 1.42* 0.88 0.87 
 Small town 1.29*** 1.02 1.11 0.91 
 Suburban     
 Medium city 1.15* 1.05 1.23^ 0.89 
 Large city 1.27** 1.25^ 1.29^ 1.37 
Respondent's race     
 White     
 Black 1.03 1.04 0.98 2.61* 
 Hispanic 0.76** 1.02 1.31 1.20 
 Asian 0.91 0.73^ 0.99 1.22 
 Other 0.82 1.01 1.23 0.92 
 Mixed 1.14 1.17 1.13 1.08 
Partner's race     
 White     
 Black 1.38** 0.95 1.78** 0.74 
 Hispanic 0.98 0.84 0.94 0.84 
 Asian 0.87 0.78 0.81 1.19 
 Mixed 1.27* 1.01 0.91 1.14 
 Other or not indicated 0.73^ 0.77 0.67 1.23 
Childhood religion     
 Catholic 0.84* 0.95 0.92 1.53* 
 Jewish 0.77* 0.89 1.34 1.49 
 Other 0.91 1.06 0.84 1.08 
 Protestant 0.77** 0.90 1.24 1.32 
 None     
Immigrant 0.99 0.98 1.04 1.41 
Observations 7144 3131 2582 1252 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

In the models estimating the odds of coitus, the interaction between grade and educational background is not significant for either sex, while the interaction between grade and the gender composition of the student body is significant among women but not men. Among women, the odds ratio for the interaction between grade and an above-average ratio of females to males is 1.50 (p < .001). This indicates that the odds of coitus increase more, among women, at schools with fewer men. Significant differences between schools with and without above-average ratios of females to males are not evident during the freshman year, when the odds ratio for the baseline term, for an above-average ratio of females to males, is a nonsignificant 1.15. Among men, the odds ratio for the interaction between grade and an above-average ratio of females to males is a nonsignificant 0.97. A difference in the rate at which the odds of coitus increase conditional on school gender composition, therefore, is evident among women but not men. That the sex ratio relates to female, but not male, sexual behavior does not support the notion that an unbalanced sex ratio empowers the scarcer sex. One possibility is that a scarcity of potential male partners causes women to search more off-campus. To eliminate this as a potential confounding factor, this finding is discussed further under the Sensitivity Analyses section, with respect to regressions on hookups in which the partners attend the same college.

Condom-use rates decline among lower-SES students, so that, after the freshman year, they use condoms at the comparatively lower rate already exhibited in the freshman year by their peers from higher-SES backgrounds. The associations in these (table 4) models with interactions are more easily understood by examining the two panels in figure 3, which graph the predicted probabilities of condom use in coital events by mother's education and year of college, separately by sex. The results for women are shown in the left panel, and the results for men are shown in the right panel. Among lower-SES women, the predicted probability of condom use is 78 percent in year 1 and 65 percent thereafter. Among higher-SES women, in contrast, condom use scarcely changes, at 67 percent in year 1 and 66 percent thereafter. Among lower-SES men, the predicted probability of condom use is 87 percent in year 1 and 71 percent thereafter. Among higher-SES men, in contrast, condom-use rates scarcely change, at 70 percent in year 1 and 71 percent thereafter.

Figure 3.

Predicted probabilities of condom use by the educational attainment of the respondent's mother

Figure 3.

Predicted probabilities of condom use by the educational attainment of the respondent's mother

The relationship between condom use and the gender composition of the student body is nonsignificant. That these estimates do not approach statistical significance appears encouraging, as it would be concerning if sex-based imbalances in the pool of potential partners affected safer-sex practices. In events with partners from the same college, there is a marginally significant positive association [table 5]. If we were to interpret this, this could indicate that women feel less safe with partners from their colleges when there are fewer men available.

Table 5.

Selected Odds Ratios for Equation 2–3, Estimated for Events in which the Partner Attended the Respondent's College

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 2
 
Equation 3
 
Equation 2
 
Equation 3
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.12 1.59***   0.56*** 0.60*   
 Junior 1.57*** 1.64***   0.69* 0.59^   
 Senior 1.74*** 2.15***   0.60** 0.51*   
During/after freshman year         
 First year of college         
 Years 2/3/4 of college   1.30 1.07   0.26** 0.32^ 
Mother's educational attainment         
 High school education or less 
 Some college education 0.91 0.96 0.98 0.67 0.80 0.77 0.58 0.29^ 
 Bachelor's degree 0.87 0.97 0.99 0.71 0.63* 0.83 0.37* 0.78 
 Graduate degree 0.87 0.86 0.93 0.51** 0.62* 0.65 0.29** 0.22* 
Grade × mother's educational attainment         
 Years 2/3/4 × some college   0.89 1.71^   1.45 3.40^ 
 Years 2/3/4 × bachelor's degree   0.81 1.61^   1.90 1.06 
 Years 2/3/4 × graduate degree   0.94 2.15*   2.81* 3.85^ 
School gender composition         
 < 53% Female         
 ≥ 53% Female   0.88 1.12   0.86 2.00 
 ≥ 53% Female × years 2/3/4   1.64*** 1.07   1.68^ 0.74 
Observations 4294 2107 4294 2107 1375 784 1375 784 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 2
 
Equation 3
 
Equation 2
 
Equation 3
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.12 1.59***   0.56*** 0.60*   
 Junior 1.57*** 1.64***   0.69* 0.59^   
 Senior 1.74*** 2.15***   0.60** 0.51*   
During/after freshman year         
 First year of college         
 Years 2/3/4 of college   1.30 1.07   0.26** 0.32^ 
Mother's educational attainment         
 High school education or less 
 Some college education 0.91 0.96 0.98 0.67 0.80 0.77 0.58 0.29^ 
 Bachelor's degree 0.87 0.97 0.99 0.71 0.63* 0.83 0.37* 0.78 
 Graduate degree 0.87 0.86 0.93 0.51** 0.62* 0.65 0.29** 0.22* 
Grade × mother's educational attainment         
 Years 2/3/4 × some college   0.89 1.71^   1.45 3.40^ 
 Years 2/3/4 × bachelor's degree   0.81 1.61^   1.90 1.06 
 Years 2/3/4 × graduate degree   0.94 2.15*   2.81* 3.85^ 
School gender composition         
 < 53% Female         
 ≥ 53% Female   0.88 1.12   0.86 2.00 
 ≥ 53% Female × years 2/3/4   1.64*** 1.07   1.68^ 0.74 
Observations 4294 2107 4294 2107 1375 784 1375 784 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

Students from highly educated backgrounds use condoms at a lower rate, relative to their peers, in the first year of college. These students continue to use condoms at approximately the same comparatively low rate later on. Students from less educated families, in contrast, use condoms more consistently during year 1, but thereafter quickly adopt the behaviors typically exhibited by their peers from more highly educated backgrounds. These results are consistent with the view that college is perceived as a safer environment because of the characteristics of the pool of potential partners it exposes students to. An interpretation equally consistent with the data is that it may take longer for lower-SES students to integrate into the social activities on their campus, which, conceivably, may not encourage condom use.

Sensitivity Analyses

Table 5 reports selected odds ratios from Equations 2–3, reestimated using only those events in which the partners attend the same college. These results are similar to those discussed above, with one exception: estimating men's odds of coitus, the interaction between grade and high-SES is significant. With the caveat that this loses significance if the models use school fixed instead of random effects, this indicates that high-SES men, but not women, are less likely to have intercourse when they hook up in the freshman year with partners from their college, and that this difference dissolves later on. The predicted probabilities of coitus in same-school events during the freshman year among women and men whose mothers have graduate degrees are 25 percent and 34 percent, respectively. The corresponding figures for women and men whose mothers did not attend college are 36 percent and 39 percent, respectively. After the freshman year, the predicted probability of coitus is about 40 percent regardless of sex or SES group.

Table 6 reports selected odds ratios from Equations 2–3, reestimated using only those events in which the respondent reported that the partner did not attend the same college. The time trends for the odds of coitus are generally significant, while the time trends for the odds of condom use are not. These results provide additional support for the conclusion that undergraduate condom-use rates decline in part due to a perception that their peers in college are comparatively less risky or because lower-SES students take longer to integrate into the campus social environment. Students seem to trust their partners more, or are otherwise less careful due to the social context, after the freshman year, especially when they attend the same college.

Table 6.

Selected Odds Ratios for Equation 2–3, Estimated for Events in which the Partner Did Not Attend the Respondent's College

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 2
 
Equation 3
 
Equation 2
 
Equation 3
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.31* 1.24   0.86 0.68   
 Junior 1.68*** 1.27   0.82 0.74   
 Senior 2.00*** 2.31***   0.73^ 0.91   
During/after freshman year         
 First year of college         
 Years 2/3/4 of college   1.32 1.76^   0.80 0.43 
Mother's educational attainment         
 High School education or less 
 Some college education 0.78* 0.81 0.66^ 0.74 0.94 0.98 0.93 0.74 
 Bachelor's degree 0.74** 0.95 0.78 1.07 0.98 1.02 0.96 0.51 
 Graduate degree 0.80^ 0.62* 0.77 0.92 1.11 0.74 0.63 0.39 
Grade × mother's educational attainment         
 Years 2/3/4 × some college   1.23 1.04   1.04 1.42 
 Years 2/3/4 × bachelor's degree   0.91 0.81   1.00 2.48 
 Years 2/3/4 × graduate degree   1.04 0.57   2.09^ 2.33 
School gender composition         
 < 53% Female         
 ≥ 53% Female   1.00 0.97   1.41 0.86 
 ≥ 53% Female × years 2/3/4   1.39^ 1.01   0.74 1.08 
Observations 2778 967 2778 967 1199 448 1199 448 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
Equation 2
 
Equation 3
 
Equation 2
 
Equation 3
 
Women Men Women Men Women Men Women Men 
Year of college         
 Freshman         
 Sophomore 1.31* 1.24   0.86 0.68   
 Junior 1.68*** 1.27   0.82 0.74   
 Senior 2.00*** 2.31***   0.73^ 0.91   
During/after freshman year         
 First year of college         
 Years 2/3/4 of college   1.32 1.76^   0.80 0.43 
Mother's educational attainment         
 High School education or less 
 Some college education 0.78* 0.81 0.66^ 0.74 0.94 0.98 0.93 0.74 
 Bachelor's degree 0.74** 0.95 0.78 1.07 0.98 1.02 0.96 0.51 
 Graduate degree 0.80^ 0.62* 0.77 0.92 1.11 0.74 0.63 0.39 
Grade × mother's educational attainment         
 Years 2/3/4 × some college   1.23 1.04   1.04 1.42 
 Years 2/3/4 × bachelor's degree   0.91 0.81   1.00 2.48 
 Years 2/3/4 × graduate degree   1.04 0.57   2.09^ 2.33 
School gender composition         
 < 53% Female         
 ≥ 53% Female   1.00 0.97   1.41 0.86 
 ≥ 53% Female × years 2/3/4   1.39^ 1.01   0.74 1.08 
Observations 2778 967 2778 967 1199 448 1199 448 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

Appendix table 1 shows the time trends after modifying Equation 2 to adjust for the number of prior hookups with the partner. This does not notably change the relationship between grade and coitus or condom use. Hookups with repeat partners may not be safer than hookups with first-time partners because hooking up more than once with the same person does not imply monogamy.

Appendix table 2 shows the time trends from Equation 2, modified to adjust for age at first intercourse. This does not affect my findings.

Finally, appendix table 3 shows the time trends from Equation 2, modified to adjust for the accumulation of intercourse partners; this could lead to desensitization, which could in turn increase the odds of coitus and decrease the odds of condom use when coitus occurs. As discussed in the Method section, including this control poses a reverse-causality problem: someone who is more likely to have sex will accumulate more sex partners. This issue is less of a concern in models estimating the odds of condom use. Results for both outcomes are presented for symmetry. As expected, this explains a substantial portion of the increase in the odds of coitus after the freshman year. However, the accumulation of partners does not drive the decline in condom use.

Discussion

Limitations

One limitation of the survey is that it asked students about condom use but not other forms of contraception, such as the pill, which prevent pregnancy but not infection. As a result, I cannot test whether students are more concerned about pregnancy than disease. If this is true, then the observed life-cycle trends could reflect increased access to the pill, IUDs, and abortion; however, I suspect this is not the case, as it is unclear why freshmen would have access to fewer forms of contraception than sophomores.

Another limitation is that the data are not longitudinal. I interpret differences among freshmen, sophomores, juniors, and seniors as change as students progress through college, but these are inferences from cross-sectional comparisons. I found no evidence that compositional differences affect the trends, which is reassuring regarding the validity of the synthetic cohort approach taken; unobserved differences associated with dropping out and sexual behavior could bias the results, but to obviate my conclusions, any such differences would have to account for a doubling in the odds of coitus, and a halving in the odds of condom use.

The results show that, during the freshman year, high-SES students use condoms less than low-SES students, and after the freshman year, high-SES students continue to use condoms at this lower rate while low-SES students converge to the behavior of high-SES students. If one questioned the synthetic cohort approach, one might also question this conclusion because, if low-SES students drop out of school at higher rates, the low-SES students in the sample in higher grades may be unusually select on unmeasured variables. However, this does not appear to be a problem in these data: with respect to all hookups, in the freshman year, 21 percent of women and 19 percent of men reported that their mothers did not attend college, in contrast to 22 percent of women and 21 percent of men in the senior year; with respect to coital hookups, in the freshman year, 23 percent of women and 23 percent of men reported that their mothers did not attend college, in contrast to 24 percent of women and 22 percent of men in the senior year. These figures are inconsistent with the idea that the findings presented in this paper could be driven by lower-SES students dropping out at higher rates. Further, to affect the results, students who drop out would have to be individuals with higher odds of condom use. In addition, my conclusions are compatible with the findings of Walsh et al. (2013), who also find an inverse relationship between SES and a decline in condom use.

Conclusion

These results demonstrate an overall decline across the years of college in the use of condoms when undergraduates hook up. Though condom non-use may send signals about faithfulness within relationships, within hookups, in contrast, pleasure and unconcern remain viable explanations of condom non-use. Of these, only decreasing concern plausibly explains a decline. Bearman, Moody, and Stovel (2004) show that network ties affect partner choice among adolescents within a high school. The findings presented in this article further suggest that loose ties (attending the same college) affect what partners do after selecting each other, even when their behavior is not directly observable by their peers. Lower-SES students adopt the same condom-use behavior as their higher-SES peers by the sophomore year, a convergence driven by events in which both partners attend the same college.

This suggests an additional question: Do undergraduates perceive sexually transmitted infection to be a problem for those from lower-class or less educated backgrounds? Sexually transmitted infection remains a salient concern on American college campuses; half of all new infections occur among adolescents and young adults (Weinstock, Berman, and Cates 2004). However, intervention efforts that focus on risk irrespective of partner characteristics may prove less effective than interventions that explicitly compare infection risks between groups. Students mating with partners from their campus and students who grew up in higher-SES environments may feel safer, such that they perceive warnings about the potential consequences of sexual behavior to apply less to them. To a certain extent, for certain infectious diseases, there may be some validity to this mindset, but even if the risk of infection is less, without a condom it is certainly not 0, and the consequences of infection can be serious.

The behaviors that develop in college may not last. College may not discourage condom use directly, but rather, expose undergraduates to a pool of potential partners with whom they feel safer or otherwise place them in social situations that do not encourage condom use. Declines in condom use, for example, are generally smaller and insignificant in events in which the partner did not attend the respondent's college. This view is compatible with Bogle's (2008) finding that graduates return to the dating script after college because they do not feel safe hooking up with partners met outside college.

Past studies looked at college students as a static group, rather than analyzing differences between students during and subsequent to the freshman year. This article advances the view that sexual behavior is affected by aspects of the environment in which late adolescents are embedded. One such aspect is the aggregate socioeconomic status of the pool of potential partners. Another, which the former may affect, is the implicit feeling of trust and safety that derives from the partners being fellow students at their college, or from becoming more integrated into the college social environment.

This analysis shows substantially lower condom use among sophomores, juniors, and seniors than among freshmen when they have intercourse in the uncommitted context of a hookup, male or female. An important question is why. This does not appear to relate to sexual experience: neither age at first intercourse, the cumulative number of sexual partners, nor the number of times one has previously hooked up with the focal partner meaningfully affect the estimates for the decline in condom use. It could relate to age irrespective of cumulative sexual experience. Although the odds of intercourse increase with each year of college, the decline in condom use occurs abruptly in a single transition. The odds of condom use drop substantially between the freshman and sophomore years, after which the decline is trivial (women) to nonexistent (men). This is incompatible with a simple aging effect. An alternative explanation is that these findings relate to the college environment—that students’ perceptions of safety and trust, or their integration into the college social environment more generally, affects condom use. They may come to perceive sexually transmitted infection as someone elses worry. It is possible that students choose safer partners when they disengage in condom use. Unfortunately, the net change in risk is unclear.

Another interpretation of these findings is that becoming embedded within certain environments encourages students to take on the role of high-SES individuals, among whom condom use appears less normative. Consistent with this, changes in condom-use rates as undergraduates progress through college decline as low-SES students adopt the lower rates of condom use of their high-SES peers. One possibility is that changing perceptions of safety and trust affect condom use, but students may adopt roles irrespective of these concerns, if they ascribe these behaviors to individuals they wish to or feel pressure to emulate.

About the Author

Jonathan Marc Bearak is a doctoral candidate in sociology at New York University. His main fields of research include education, health, and earnings inequality, with an emphasis on gender, family structure, and relationship formation. He is also interested in quantitative methods and statistical software. Recent articles have been published in American Sociological Review and Demographic Research.

1
Randomly sampling students from college and university rosters could better ensure representativeness, provided the response rate is high. However, using rosters at one of the schools resulted in unacceptably low response rates—about 50 percent after five attempts.
2
The response rate is nearly perfect because the survey was administered by professors who agreed to offer course credit. Early attempts that did not require that professors offer credit resulted in low response rates. Thus, the survey was administered only by professors who were willing to offer the survey, with the option of an alternative assignment, for credit (Armstrong, England, and Fogarty 2012). This arrangement was approved by each college's Institutional Review Board.
3
When estimating the odds of condom use, I pooled the sexes to increase power. (Results are similar by sex, as discussed in the Results section.) As noted in supplemental table S3, however, controlling for each individual year of college, instead of during/after the freshman year, does not significantly improve model fit. In addition, as can be seen in the Results section when discussing Equation 2, the odds ratios for the sophomore, junior, and senior years are virtually identical.

References

Armstrong
Elizabeth
Hamilton
Laura
England
Paula
.
2010
. “
Is Hooking Up Bad for Young Women?
Contexts
 
9
:
22
-
27
.
Armstrong
Elizabeth
England
Paula
Fogarty
Alison C. K.
.
2012
. “
Accounting for Women's Orgasm and Sexual Enjoyment in College Hookups and Relationships
.”
American Sociological Review
 
77
:
435
-
62
.
Arum
Richard
Roksa
Josipa
Budig
Michelle J.
.
2008
. “
The Romance of College Attendance: Higher Education Stratification and Mate Selection
.”
Research on Stratification and Mobility
 
26
(2)
:
107
-
21
.
Beaglehole
Robert
.
1990
. “
International Trends in Coronary Heart Disease Mortality, Morbidity, and Risk Factors
.”
Epidemiologic Reviews
 
12
:
1
-
15
.
Bearman
Peter S.
Moody
James
Stovel
Katherine
.
2004
. “
Chains of Affection: The Structure of Adolescent Romantic and Sexual Networks
.”
American Journal of Sociology
 
110
:
44
-
91
.
Beckman
Linda J.
Harvey
S. Marie
Tiersky
Lana A.
.
2010
. “
Attitudes about Condoms and Condom Use among College Students
.”
Journal of American College Health
 
44
(6)
:
243
-
50
.
Bogle
Kathleen A
.
2008
.
Hooking Up: Sex, Dating, and Relationships on Campus
.
New York
:
New York University Press
.
England
Paula
McClintock
Elizabeth Aura
Shafer
Emily Fitzgibbons
.
2011
. “
Birth Control and Early, Unintended Births: Evidence of a Class Gradient
.”
In Social Class and Changing Families in an Unequal America, edited by Marcia Carlson and Paula England
 .
Stanford, CA
:
Stanford University Press
.
DiPrete
T. A.
Buchmannn
C.
.
2006
. “
Gender-Specific Trends in the Value of Education and the Emerging Gender Gap in College Completion
.”
Demography
 
43
:
1
-
24
.
DiPrete
Thomas A.
Buchman
Claudia
.
2013
.
The Rise of Women: The Growing Gender Gap in Education and What It Means for American Schools
.
New York
:
Russell Sage Foundation
.
Downing-Matibag
T. M.
Geisinger
B.
.
2009
. “
Hooking Up and Sexual Risk-Taking among College Students: A Health Belief Model Perspective
.”
Qualitative Health Research
 
19
(9)
:
1196
-
209
.
East
Leah
Jackson
Debra
O'Brien
Louise
Peters
Kathleen
.
2007
. “
Use of the Male Condom by Heterosexual Adolescents and Young People: Literature Review
.”
Journal of Advanced Nursing
 
59
:
103
-
10
.
Guttentag
Marcia
Secord
Paul F.
.
1983
.
Too Many Women? The Sex Ratio Question
.
Newbury Park, CA
:
Sage
.
Hamilton
Laura
Armstrong
Elizabeth
.
2009
. “
Gendered Sexuality in Young Adulthood: Double Binds and Flawed Options
.”
Gender & Society
 
23
:
589
-
616
.
Hoffman
V.
Cohen
D.
.
1999
. “
A Night with Venus: Partner Assessments and High-Risk Sexual Encounters
.”
AIDS Care
 
11
:
555
-
66
.
Kreager
Derek
Staff
Jeremy
.
2009
. “
The Sexual Double Standard and Adolescent Peer Acceptance
.”
Social Psychology Quarterly
 
72
:
143
-
64
.
Lewis
John E.
Malow
Robert M.
Ireland
Susan J.
.
1997
. “
HIV/AIDS Risk in Heterosexual College Students: A Review of a Decade of Literature
.”
Journal of American College Health
 
45
:
147
-
58
.
Link
Bruce G.
Phelan
Jo
.
1995
. “
Social Conditions as Fundamental Causes of Disease
.”
Journal of Health and Social Behavior
 
35
:
80
-
94
.
Manlove
Jennifer
Ryan
Suzanne
Franzetta
Kerry
.
2004
. “
Contraceptive Use and Consistency in US Teenagers’ Most Recent Sexual Relationships
.”
Perspectives on Sexual and Reproductive Health
 
36
(6)
:
265
-
75
.
Manning
Wendy D.
Flanigan
Christine M.
Giordano
Peggy C.
Longmore
Monica A.
.
2009
. “
Relationship Dynamics and Consistency of Condom Use among Adolescents
.”
Perspectives on Sexual and Reproductive Health
 
41
:
181
-
90
.
Manning
Wendy D.
Longmore
Monica A.
Giordano
Peggy C.
.
2000
. “
The Relationship Context of Contraceptive Use at First Intercourse
.”
Family Planning Perspectives
 
32
:
104
-
10
.
Manning
Wendy D.
Longmore
Monica A.
Giordano
Peggy C.
.
2005
. “
Adolescents’ Involvement in Non-Romantic Sexual Activity
.”
Social Science Research
 
34
:
384
-
407
.
Mare
Robert D
.
1991
. “
Five Decades of Educational Assortative Mating
.”
American Sociological Review
 
56
(1)
:
15
-
32
.
Martinez
G.
Copen
C. E.
Abma
J. C.
.
2011
. “
Teenagers in the United States: Sexual Activity, Contraceptive Use, and Childbearing, 2006–2010 National Survey of Family Growth
.”
National Center for Health Statistics, Vital Health Stat 23
.
Morris
Martina
.
1993
. “
Telling Tails Explain the Discrepancy in Sexual Partner Reports
.”
Nature
 
365
:
437
-
40
.
Musick
Kelly
England
Paula
Edgington
Sarah
Kongas
Nicole
.
2009
. “
Education Differences in Intended and Unintended Fertility
.”
Social Forces
 
88
(2)
:
543
-
72
.
Pappas
G.
Queen
S.
Hadden
W.
Fisher
G.
.
1993
. “
The Increasing Disparity in Mortality between Socioeconomic Groups in the United States, 1960 and 1986
.”
New England Journal of Medicine
 
329
:
103
-
09
.
Parsons
J. T.
Halkitis
P. N.
Bimbi
D.
Borkowski
T.
.
2000
. “
Perceptions of the Benefits and Costs Associated with Condom Use and Unprotected Sex among Late Adolescent College Students
.”
Journal of Adolescence
 
23
(4)
:
377
-
91
.
Schwartz
Christine R.
Mare
Robert D.
.
2005
. “
Trends in Educational Assortative Mating from 1940 to 2003
.”
Demography
 
42
(4)
:
621
-
46
.
Sheeran
P.
Abraham
C.
Orbell
S.
.
1999
. “
Psychosocial Correlates of Heterosexual Condom Use: A Meta-Analysis
.”
Psychological Bulletin
 
125
(1)
:
90
-
132
.
Tavory
Iddo
Swidler
Ann
.
2009
. “
Condom Semiotics: Meaning and Condom Use in Rural Malawi
.”
American Sociological Review
 
74
(2)
:
171
-
89
.
Uecker
Jeremy E.
Regnerus
Mark D.
.
2010
. “
Bare Market: Campus Sex Ratios, Romantic Relationships, and Sexual Behaviors
.”
Sociological Quarterly
 
51
:
408
-
35
.
Waller
Willard
.
1937
. “
The Rating and Dating Complex
.”
American Sociological Review
 
2
:
727
-
34
.
Walsh
Jennifer L.
Fielder
Robyn L.
Carey
Kate B.
Carey
Michael P.
.
2013
. “
Changes in Women's Condom Use over the First Year of College
.”
Journal of Sex Research
 
50
(2)
:
128
-
138
.
Weinstock
Hillard
Berman
Stuart
Cates
Willard
.
2004
. “
Sexually Transmitted Diseases among American Youth: Incidence and Prevalence Estimates, 2000
.”
Perspectives on Sexual and Reproductive Health
 
36
:
6
-
10
.
Younge
Sinead N.
Salazar
Laura F.
Crosby
Richard F.
DiClemente
Richard J.
Wingood
Gina M.
Rose
Eve
.
2008
. “
Condom Use at Last Sex as a Proxy for Other Measures of Condom Use: Is It Good Enough?
Adolescence
 
43
:
927
-
31
.

Appendix

Table A1.

Selected Odds Ratios from Equation 2, Modified to Adjust for the Number of Times the Student Previously Hooked Up with this Partner

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.29*** 1.33** 1.19^ 1.36* 1.31* 1.12 0.67** 0.64* 0.54*** 0.64^ 0.85 0.67 
 Junior 1.76*** 1.36** 1.54*** 1.40* 1.80*** 1.28 0.72** 0.66* 0.70* 0.67 0.80 0.63 
 Senior 1.95*** 2.03*** 1.69*** 1.86*** 1.98*** 2.32*** 0.65*** 0.63* 0.62** 0.55* 0.75^ 0.78 
Prior hookups with this partner 
 0             
 1–2 1.57*** 1.45*** 1.54*** 1.35* 1.64*** 1.46* 1.13 1.75* 0.91 1.62 1.33 2.46* 
 3–5 2.22*** 2.34*** 2.40*** 2.55*** 2.04*** 2.02** 1.02 1.43 0.88 1.19 1.05 2.07^ 
≥ 6 4.83*** 3.82*** 4.97*** 3.31*** 4.59*** 4.69*** 0.58*** 0.46*** 0.46*** 0.45*** 0.66** 0.53* 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.29*** 1.33** 1.19^ 1.36* 1.31* 1.12 0.67** 0.64* 0.54*** 0.64^ 0.85 0.67 
 Junior 1.76*** 1.36** 1.54*** 1.40* 1.80*** 1.28 0.72** 0.66* 0.70* 0.67 0.80 0.63 
 Senior 1.95*** 2.03*** 1.69*** 1.86*** 1.98*** 2.32*** 0.65*** 0.63* 0.62** 0.55* 0.75^ 0.78 
Prior hookups with this partner 
 0             
 1–2 1.57*** 1.45*** 1.54*** 1.35* 1.64*** 1.46* 1.13 1.75* 0.91 1.62 1.33 2.46* 
 3–5 2.22*** 2.34*** 2.40*** 2.55*** 2.04*** 2.02** 1.02 1.43 0.88 1.19 1.05 2.07^ 
≥ 6 4.83*** 3.82*** 4.97*** 3.31*** 4.59*** 4.69*** 0.58*** 0.46*** 0.46*** 0.45*** 0.66** 0.53* 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

Table A2.

Selected Odds Ratios from Equation 2, Modified to Adjust for the Student's Age at First Intercourse

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.23** 1.48*** 1.14 1.53** 1.32* 1.28 0.65*** 0.61** 0.53*** 0.59* 0.83 0.68 
 Junior 1.55*** 1.46** 1.54*** 1.62** 1.45** 1.15 0.68** 0.61* 0.64* 0.57* 0.80 0.72 
 Senior 1.71*** 2.11*** 1.60*** 2.07*** 1.75*** 2.26*** 0.59*** 0.58** 0.56** 0.48** 0.69* 0.84 
Age at first intercourse, years             
 ≤ 15 1.57*** 2.12*** 1.68*** 2.65*** 1.50** 1.51^ 0.61*** 0.75 0.59** 0.90 0.63* 0.52^ 
 16 1.16^ 1.71*** 1.08 1.92*** 1.29^ 1.34 0.86 0.67* 0.90 0.94 0.88 0.45* 
 17             
 18 0.93 1.08 0.94 1.25 0.89 0.81 1.04 0.92 1.06 0.95 1.07 0.88 
 ≥ 19 0.58*** 0.72* 0.53*** 0.82 0.65** 0.51* 1.10 1.18 1.15 1.51 1.05 0.91 
 Not applicable 0.02*** 0.07*** 0.02*** 0.10*** 0.02*** 0.03*** 1.00 0.69 4.07 0.96 0.44 0.36 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.23** 1.48*** 1.14 1.53** 1.32* 1.28 0.65*** 0.61** 0.53*** 0.59* 0.83 0.68 
 Junior 1.55*** 1.46** 1.54*** 1.62** 1.45** 1.15 0.68** 0.61* 0.64* 0.57* 0.80 0.72 
 Senior 1.71*** 2.11*** 1.60*** 2.07*** 1.75*** 2.26*** 0.59*** 0.58** 0.56** 0.48** 0.69* 0.84 
Age at first intercourse, years             
 ≤ 15 1.57*** 2.12*** 1.68*** 2.65*** 1.50** 1.51^ 0.61*** 0.75 0.59** 0.90 0.63* 0.52^ 
 16 1.16^ 1.71*** 1.08 1.92*** 1.29^ 1.34 0.86 0.67* 0.90 0.94 0.88 0.45* 
 17             
 18 0.93 1.08 0.94 1.25 0.89 0.81 1.04 0.92 1.06 0.95 1.07 0.88 
 ≥ 19 0.58*** 0.72* 0.53*** 0.82 0.65** 0.51* 1.10 1.18 1.15 1.51 1.05 0.91 
 Not applicable 0.02*** 0.07*** 0.02*** 0.10*** 0.02*** 0.03*** 1.00 0.69 4.07 0.96 0.44 0.36 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 

*** p < .001 ** p < .01 * p < .05 ^ p < .1

Table A3.

Selected Odds Ratios from Equation 2, Modified to Adjust for the Student's Cumulative Number of Sexual Partners

 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.03 1.22^ 0.97 1.27^ 1.08 0.95 0.69** 0.62** 0.58** 0.60^ 0.86 0.67 
 Junior 1.23* 1.08 1.24* 1.22 1.09 0.75 0.74* 0.63* 0.71^ 0.59^ 0.82 0.75 
 Senior 1.18* 1.29* 1.09 1.26 1.16 1.19 0.65*** 0.63* 0.64* 0.52* 0.73^ 0.93 
Cumulative intercourse partners 
 0–1             
 2–3 6.11*** 4.14*** 6.58*** 4.80*** 5.88*** 3.56*** 1.02 1.26 1.04 0.95 1.04 1.39 
 ≥ 4 14.37*** 10.62*** 16.14*** 11.93*** 13.64*** 10.96*** 0.87 1.07 0.72 0.93 1.02 1.03 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 
 Models estimating the odds of coitus
 
Models estimating the odds of condom use
 
All events
 
Same-school
 
Not-same-school
 
Coital events
 
Same-school
 
Not-same-school
 
Women Men Women Men Women Men Women Men Women Men Women Men 
Year of college             
 Freshman             
 Sophomore 1.03 1.22^ 0.97 1.27^ 1.08 0.95 0.69** 0.62** 0.58** 0.60^ 0.86 0.67 
 Junior 1.23* 1.08 1.24* 1.22 1.09 0.75 0.74* 0.63* 0.71^ 0.59^ 0.82 0.75 
 Senior 1.18* 1.29* 1.09 1.26 1.16 1.19 0.65*** 0.63* 0.64* 0.52* 0.73^ 0.93 
Cumulative intercourse partners 
 0–1             
 2–3 6.11*** 4.14*** 6.58*** 4.80*** 5.88*** 3.56*** 1.02 1.26 1.04 0.95 1.04 1.39 
 ≥ 4 14.37*** 10.62*** 16.14*** 11.93*** 13.64*** 10.96*** 0.87 1.07 0.72 0.93 1.02 1.03 
Observations 7144 3131 4294 2107 2778 967 2582 1252 1375 784 1199 448 

*** p < .001 ** p < .01 * p < .05 ^ p < .1