Union Status and Sexual Risk Behavior Among Men in Their 30s

Abstract

CONTEXT

Understanding the relationship between union status and men’s sexual risk behavior in their 30s is important to ensure appropriate reproductive health services for men in middle adulthood.

METHODS

Data from 1,083 men aged 34–41 who participated in the 2008–2010 wave of the National Survey of Adolescent Males were used to examine differentials in sexual risk behaviors by union status, past risk behavior and selected characteristics. Bivariate tabulations were done to assess relationships between current risk behavior and background variables, multinomial regression analysis was conducted to identify associations between union status and past risk behavior, and logistic regression analysis was used to assess associations between current behavior and both union status and past behavior.

RESULTS

Eight percent of men in their 30s had had three or more sexual partners in the last 12 months, 10% had had at least one risky partner and 8% had had concurrent partners. Men living outside coresidential unions reported higher levels of these behaviors (24%, 29% and 24%, respectively) than did married (1–2%) or cohabiting men (7–12%). In multivariate analyses that controlled for past risk behavior, married men were less likely than cohabiting men to have had at least one risky partner or concurrent partners in the last year (odds ratio, 0.2 for each), while men who were not in a coresidential union had an increased likelihood of reporting each risk behavior (2.2–5.3).

CONCLUSIONS

Men in their 30s, especially those who are not married, engage in risky sexual behaviors. Further studies are needed to assess what contributes to behavioral differences by union status and what types of services might help men in this age-group reduce their risk.

Research since the early 1990s has consistently found that sexual risk behavior declines as people get older.^1,2 For example, the most recent National Survey of Family Growth, conducted in 2006–2010, found that the prevalence of sexual risk behavior declines monotonically from 15% among males aged 20–24 to less than 10% among men older than 40.² Why do risk profiles change with age? Some research suggests that the propensity for risky behavior may decline in adulthood for biological reasons, including changes occurring during adolescent and young adult brain development.³ The focus of this article, however, is on a different set of possible explanations for sexual risk—social and behavioral ones, particularly union status.

Married people are less likely than others to engage in sexual risk behavior,⁴ and older people are more likely than younger individuals to be married. Taylor and colleagues found that the greater risk for STDs among blacks than among other races or ethnicities was reduced when they controlled for marriage status.⁵ Moreover, associations have also been found between nonsexual risk behaviors and being unmarried. For example, blacks in long-term stable unions are less likely to smoke or use illegal drugs, and black women in such unions are less likely to drink heavily, than blacks who are not in such unions.⁶ In a series of studies of long-term outcomes among young men who were at risk of criminal behavior in childhood, Laub and colleagues have argued that getting married is associated with desistance from crime,^7–10 and others have found this as well.¹¹ The transition to marriage has also been associated with desistance from binge drinking and marijuana use.¹²

Several hypotheses may help to explain why marriage is associated with relatively low levels of and desistance from risky behavior. Proponents of the selection hypothesis argue that people with a low propensity to engage in risky behavior are more likely than others to marry and vice versa.¹² This hypothesis has been tested using longitudinal data and examining change within individuals. For example, in a long-term cohort study, Green and colleagues identified distinct latent marriage trajectories and found that after premarital substance use was controlled for, being stably married was associated with low levels of legal and illegal substance use.⁶ Duncan and colleagues argued that their finding of a relationship between marriage and desistance from risk behaviors was unlikely to be due to selection since they focused on ever-married people and assessed behavior change shortly after marriage and over the longer term.¹²

Another hypothesis regarding marriage and sexual risk behavior is the monitoring hypothesis. Proponents suggest that since coresidential partners spend more time together than partners in visiting unions, the former have fewer opportunities to engage in covert behavior. The prevalence of widespread nonmarital cohabitation provides an opportunity for testing the monitoring hypothesis. If the association between marriage and sexual risk behavior is largely due to the ability of coresidential partners to observe and inhibit each other’s risk behavior, then we should see no difference in behavior between married and cohabiting individuals.

A third hypothesis focuses on role socialization,^13,14 and posits that strong norms regulate appropriate behavior for married couples. These norms proscribe any behavior that puts one at risk and prescribe behavior that is future-oriented, cautious and responsible. The underlying idea is that the transition to marriage marks both an end to parental willingness to bail adult children out of trouble and the beginning of a time when the welfare of others depends on one’s own behavior. Role socialization has most often been invoked to examine the association between marriage and substance use,^12,14 yet it is even more applicable to the study of sexual risk behavior, since getting married involves taking vows of sexual exclusivity.

The role socialization hypothesis has a long history, but recent applications have been somewhat different from those of the past. In older formulations, the hypothesis was tested by examining differences between being single and being married.^13–16 Now, researchers are more likely to assess differences between two groups of coresidential couples: those who cohabit (i.e., outside of marriage) and those who are married.¹² A number of investigators have called cohabitation “an incomplete institution.”^17–19 That is, it is an arrangement not yet characterized by widely agreed upon norms of behavior or consistent expectations about role performance—certainly it does not involve vows of sexual exclusivity, as marriage does. If the role socialization hypothesis is correct, then we should observe higher levels of sexual risk behavior among individuals who are cohabiting than among those who are married.

In the current study, we consider how coresidential unions, including marriage, are associated with risky sexual behavior. We examine the sexual behavior of a nationally representative sample of men who have been followed since late adolescence, and focus on behavior when they were in their middle to late 30s. This is a key age range, since it is a period when risk behavior has typically declined to low levels;^1,2 furthermore, because age and marriage are correlated,¹² holding the age-group constant is an analytic advantage for isolating associations between union status and sexual risk behavior. The use of longitudinal data allows us to emulate some of the best studies on marriage and substance use. These researchers employed a longitudinal design to better understand the selection hypothesis.^6–12 In addition, because these data include information on cohabitation as well as marriage, we are able to test both the role socialization and monitoring hypotheses.

METHODS

Data

We used data from the National Survey of Adolescent Males (NSAM), which began in 1988 with in-person interviews with a nationally representative sample of 1,880 never-married males aged 15–19 who were living in households in the conterminous United States. The NSAM used a multistage, stratified sample that oversampled blacks and Latinos. Nonresponse was somewhat less common in black households than in white households. Weights are available to compensate for nonresponse, and the weights were poststratified to correspond with the March 1987 Current Population Survey. The second wave of NSAM occurred in late 1990 and early 1991, when respondents were aged 17–22; in that wave, 1,676 interviews were conducted, for an 89% follow-up rate (not including 11 men who had died between 1988 and 1990). Respondents were interviewed for the third wave in 1995, at ages 21–26; this wave included 1,377 respondents, of whom 1,290 had been interviewed in all three waves.

In 2008, the institutional review board of the Johns Hopkins Bloomberg School of Public Health approved the fourth wave of data collection among respondents who were not incarcerated. Telephone interviews were completed with 634 men, and in-person interviews with 449 men, by the end of the field period in August 2010. These 1,083 individuals, aged 34–41,^* represented 62% of original respondents who were not incarcerated, incapacitated or deceased. We developed longitudinal weights to adjust for nonresponse. Attrition analysis indicated that the 2008–2010 respondents were somewhat more economically advantaged than the original sample. The response rate was significantly lower among blacks than among other races or ethnicities, and was lower among men whose mothers had no more than a high school education than among those whose mothers had at least some college education; response rates were also lower in the Northeast and West than in the South and Midwest. In this study, we used data from all 1,083 men who responded to the fourth round of data collection.

Measures

We examined three outcome measures from the fourth round of interviews. The first is the number of partners in the last 12 months, scored as 1 if the man reported having had vaginal intercourse with three or more females in the last year and as 0 otherwise (based on preliminary analysis). The second outcome is whether a man had had a risky partner in the last 12 months. We considered a respondent to have had a risky partner if he had had sex with someone only once; had had sex with a sex worker, an injection-drug user, a man, or a person with HIV or AIDS; or had performed sex work. The third outcome variable is having had concurrent partners in the last 12 months; responses were scored as 1 if the man reported having had two or more partners in at least one month and as 0 otherwise. This is how concurrency has been measured in NSAM previously, and this measure is one of the most strongly associated with having had an STD.²⁰

We assessed two independent variables. The first is the man’s current union status, categorized as married and living with a spouse, cohabiting or not in a coresidential union. The second is past sexual risk behavior, which we based on reports from earlier survey waves of the three behavior outcomes used in the present study. This measure distinguishes among three groups of men: those who were never in a high-risk group^* in the first three waves; those who were in a high-risk group in only one wave (experimenters); and those who were in a high-risk group in two or three waves (repeaters).

In addition, we considered a number of control variables: race or ethnicity (white, black, Hispanic), mother’s education (completed at least high school or a GED vs. did not complete or missing), and respondent’s current educational level (completed high school or more vs. did not complete or had a GED), current employment status, and current or most recent wage (lowest quartile of U.S. wages vs. top three quartiles).²¹ For ordinal variables, cut points reflected nonlinear associations with the outcomes in exploratory bivariate tabulations.

Analysis

We tested the selection hypothesis in three ways. First, we examined whether the association between being in a coresidential union and low levels of current risk behavior varied depending on past sexual risk behavior. If selection fully accounts for the low levels of current sexual risk taken by people living with a partner, we expect that the association between union status and current risky behavior would be attenuated when the sample is stratified by past risky behavior. Second, we directly examined the relationship between past risk behavior and current union status. If the two are associated, selection likely is playing a big role. Third, we evaluated the association between union status and current sexual risk behavior in models that controlled for past risk behavior.

To test the monitoring hypothesis, we compared the current sexual risk behavior of married men with that of men outside coresidential unions, as well as the behavior of those in cohabiting unions with that of individuals outside a union. If the monitoring hypothesis is correct, then both groups in coresidential unions should exhibit lower levels of risk behavior than males outside such unions.

To test the role socialization hypothesis, we reasoned that if there are distinct normative expectations that married people reduce risk behavior, we should observe lower levels of risk among married than among cohabiting men.

Of course, role socialization and monitoring are not mutually exclusive hypotheses; the processes they posit could both be operating. Therefore, we make three predictions. First, if only role socialization is operating, we should observe that married men are less likely to exhibit current sexual risk behavior than either cohabiting men or men outside a coresidential union, and no differences should exist between these last two groups. Second, if only monitoring is operating, we should observe that men outside unions are more likely than either married or cohabiting men to exhibit current risky sexual behavior, and the two coresidential groups should not differ. Third, if both role socialization and monitoring are operating, we should observe differences between each pair of groups: Married men should exhibit the lowest level of current risky sexual behavior (since they experience both processes), cohabiting men should exhibit an intermediate level (since they experience only monitoring) and men outside unions should exhibit the highest level.

We present cross-tabulations with weighted percentages and assessed differences with likelihood ratio chi-square tests. We used multivariate multinomial regression analysis to evaluate associations between current union status and past risk behavior, and multivariate logistic regression analysis to assess associations between current risk behavior and both union status and past behavior. The direction and magnitude of the estimates were similar in weighted and unweighted models, so we present results from the unweighted models, as these estimates are preferable.²²

RESULTS

Descriptive and Bivariate Findings

Of all surveyed men, 64% were married, 11% were cohabiting and 26% were not in a coresidential union (Table 1). Overall, 53% were classified as never having been at high risk, 30% were classified as experimenters and 17% were considered repeaters. Three-fourths of respondents were white, and the rest were either black or Hispanic. More than eight in 10 men had mothers with at least a high school diploma or GED, and nine in 10 respondents had at least a high school education themselves. Eighty-nine percent were currently employed, and the current or most recent wages of 17% fell into the lowest quartile.

TABLE 1.

Percentage distribution of men aged 34–41, and percentage reporting sexual risk behaviors in the last 12 months, all by selected characteristics, National Survey of Adolescent Males, 2008–2010

Characteristic	Weighted % (N=1,083)	≥3 partners	Any risky partner	Concurrent partners
Total	100.0	7.6	9.6	8.2
Current union status
Married	63.7	1.1*,†	2.1*,†	1.3*,†
Cohabiting	10.6	6.9†	7.9†	11.8†
Not in union	25.7	23.8	29.0	24.4
Past risk group‡
Never high-risk	53.0	3.9*,†	7.6†	3.4*,†
Experimenter	30.2	12.0	9.6	11.6
Repeater	16.8	11.2	15.9	15.2
Race/ethnicity
White	76.1	7.4	9.7	6.9
Black	13.9	8.8	11.4	13.5
Hispanic	10.0	7.0	5.8	10.6
Mother’s education
≥high school/GED	82.1	7.1	9.5	8.2
<high school/missing	17.9	9.8	10.0	8.2
Education
≥high school	89.5	7.6	9.0	8.6
<high school/GED	10.5	7.4	14.1	4.6
Currently employed
Yes	88.9	6.7	9.4	7.3
No	11.1	14.2	10.6	15.2
Current/most recent wage§
Lowest quartile	16.6	5.0	10.7	5.7
Top three quartiles	83.4	8.1	9.3	8.6

^*Differs from percentage in second category at p<.001 for the likelihood ratio chi-square test.

^†DIffers from percentage in third category at p<.001 for the likelihood ratio chi-square test.

^‡Men were classified as experimenters if they were in a high-risk group in the year preceding one survey wave, and as repeaters if they were in such a group for two or three waves (see page TK).

^§Based on U.S. wage distribution (reference 21).

In the 12 months prior to being interviewed, 8% of respondents had had three or more sex partners, 10% had had at least one risky partner and 8% had had concurrent partners. Reports of current risk behavior differed significantly by union status. Twenty-four percent of men outside coresidential unions had had three or more partners, as had 7% of cohabiting men and 1% of married men. The proportions in each union category who had had at least one risky partner were 29%, 8% and 2%, respectively. Finally, 24% of men not in a coresidential union had had concurrent partners, and this was reported by 12% of cohabiting and 1% of married men. Current risk behavior also differed by past risk behavior: Men who had never exhibited risky behavior in the past were less likely than either experimenters or repeaters to report any of the risky behaviors in the last year, but the differences were smaller than those observed by union status. No significant differences in reporting of sexual risk behavior were found by socioeconomic characteristics.

With one exception, married and cohabiting respondents reported significantly lower levels of all three sexual risk behaviors than did males who were not living in a union, regardless of past risk behavior (Table 2). For example, among men who had never been at high-risk, 13% of those outside unions had had three or more partners in the last 12 months, compared with 0.2–1% of cohabiting or married men. The patterns were similar for the risky and concurrent partner measures. Among respondents classified as experimenters, much larger proportions of men who were not in a coresidential union than of those who were reported the different behaviors in the last year. Finally, among repeaters, this pattern held only for reports of having had a risky partner.

TABLE 2.

Percentage distribution of men, and percentage reporting sexual risk behaviors in the last 12 months, by current union status, according to past sexual risk

Past risk group and current union status	Weighted %	≥3 partners	Any risky partner	Concurrent partners
Never high-risk
Married	65.8	0.2†	1.4*,†	0.2*,†
Cohabiting	5.7	1.3†	12.7†	6.7
Not in union	28.5	12.9	20.9	10.9
Experimenters
Married	60.8	1.4†	0.9*,†	1.7*,†
Cohabiting	16.5	0.9†	3.4†	4.5†
Not in union	22.7	48.4	37.4	43.4
Repeaters
Married	62.0	3.6*,†	6.5†	4.1*,†
Cohabiting	15.7	24.7	11.1†	30.4
Not in union	22.3	22.8	47.5	35.2

^*Differs from percentage in second category at p<.001 for the likelihood ratio chi-square test.

^†DIffers from percentage in third category at p <.001 for the likelihood ratio chi-square test.

Multivariate Findings

Past risk behavior was not associated with current union status in a consistent or compelling way (Table 3). Experimenters were less likely than those with no observed past risk to be married rather than cohabiting (odds ratio, 0.6), and were more likely to be cohabiting rather than living outside a coresidential union (1.5). No associations were found between repeaters and those with no past risk regarding union status. Compared with blacks, whites were more likely to be married rather than cohabiting (2.5) or not in a union (1.5), and less likely to be cohabiting than living outside a union (0.6). Finally, Hispanics were more likely than blacks to be married rather than cohabiting (1.8).

TABLE 3.

Odds ratios from multivariate multinomial regression analysis assessing associations between current union status and selected characteristics

Characteristic	Married vs. cohabiting	Married vs. not in union	Cohabiting vs. not in union
Past risk group
Never high-risk (ref)	1.00	1.00	1.00
Experimenter	0.64*	0.98	1.53*
Repeater	0.76	1.02	1.35
Race/ethnicity
White	2.54***	1.52*	0.60*
Black (ref)	1.00	1.00	1.00
Hispanic	1.77*	1.14	0.64
Mother’s education
≥high school/GED (ref)	1.00	1.00	1.00
<high school/missing	1.26	1.05	0.84
Education
≥high school	2.73***	1.59*	0.58*
<high school/GED (ref)	1.00	1.00	1.00
Currently employed
Yes	1.28	2.12***	1.65
No (ref)	1.00	1.00	1.00
Current/most recent wage
Lowest quartile	0.46***	0.37***	0.81
Top three quartiles (ref)	1.00	1.00	1.00

^*p<.05.

^***p<.001.

Note: ref=reference group.

In the past year, married men were less likely than cohabiting men to have had a risky partner or concurrent partners (odds ratio, 0.2 for each—Table 4). Men not in coresidential unions were more likely than cohabiting men to report all three risky behaviors: three or more partners (5.3), a risky partner (3.7) and concurrent partners (2.2). Estimates of the differences between married men and those outside unions found that the former were less likely to exhibit these sexual risk behaviors (not shown). Experimenters were more likely than men with no observed past risk to have had three or more partners in the last year (5.2), and repeaters were more likely than men with no past risk to report all three risk behaviors (1.9–5.0). No differences were found in current risk behavior between experimenters and repeaters (not shown).

TABLE 4.

Odds ratios from multivariate logistic regression analysis assessing associations between sexual risk behavior in the last 12 months and selected characteristics

Characteristic	≥3 partners	Any risky partner	Concurrent partners
Current union status
Married	0.44†	0.17***	0.18***
Cohabiting (ref)	1.00	1.00	1.00
Not in union	5.27***	3.67***	2.16**
Past risk group
Never high-risk (ref)	1.00	1.00	1.00
Experimenter	5.24**	1.29	1.56
Repeater	5.00**	1.86*	2.02*
Race/ethnicity
White	0.94	1.13	0.71
Black (ref)	1.00	1.00	1.00
Hispanic	0.82	0.48*	0.90

^*p<.05.

^**p<.01.

^***p<.001.

^†p<.10.

Note: ref=reference group.

DISCUSSION

Our findings indicated that the overall levels of sexual risk behavior were low among men in their 30s. We also found, however, that among the one-fourth of surveyed men currently living outside coresidential unions, levels of risky behavior were as high as those reported by these same men when they were adolescents and young adults.²⁰ We observed few differences in outcomes by respondents’ socioeconomic characteristics, which suggests that the socioeconomic inequalities that characterize younger men’s sexual risk behavior²⁰ do not persist into adulthood.

We found little support for the selection hypothesis. The association between being married and current levels of sexual risk behavior was evident among all men, regardless of whether they had exhibited risky behavior in the past. If selection were the full explanation for the marriage association, the association would have disappeared or been reduced once we stratified by past risk behavior. In fact, married men who had repeatedly exhibited risky behavior in the past had very low levels of current risk, and men outside coresidential unions who had not previously exhibited risk behavior reported relatively high levels of current risky behavior. Moreover, when we examined the selection hypothesis directly by looking at the association between past risk behavior and current union status, we found no consistent pattern. In particular, we did not find that men who exhibited high levels of past risk behavior were more likely to be living outside unions. Finally, the association between being in a coresidential union and reporting low levels of risky behavior persisted in models that controlled for past risk.

In contrast, our findings support both the monitoring and the role socialization hypotheses. Both married and cohabiting men exhibited lower levels of risky sexual behavior than men outside coresidential unions, which suggests monitoring. Men who were married reported the lowest levels of risk behavior. Most notably, the odds of their having had a risky partner or concurrent partners were less than one-fifth those of cohabiting men. These findings are consistent with the idea that a normative dimension to marriage inhibits risky sexual behavior, in ways similar to those that have been found for criminal behavior and substance use.^6–12

An important extension of the present study is to examine how men’s union histories, as opposed to current union status, are associated with risky sexual behavior. Examination of the full complement of longitudinal data available in the NSAM will allow us to study trajectories of risk behavior over time and explore how these trajectories may be related to family life. Other researchers have documented an association between fatherhood and risky behavior,¹¹ and this may be an important link to consider when examining how cohabiting men and married men differ. Future research should also explore why men with the riskiest sexual histories end up cohabiting later in life.

Limitations

As noted earlier, men of lower socioeconomic status in past waves were less likely to be followed up in the fourth wave, and this might partially explain why we observed so few differences in the levels of risky behavior by socioeconomic characteristics. Another limitation was that the measures of current risky behavior were self-reported and might be subject to response bias. Furthermore, the observational nature of our study design precluded us from drawing causal inferences from our findings. Finally, we did not control for coresidential union history or fatherhood status. If we had, we might have found that risky sexual behavior is associated with men having moved in and out of coresidential unions and having fathered children with multiple partners, rather than with being in a current union. Testing this hypothesis is possible with the NSAM data, and we will pursue it in future research.

Conclusions

Our findings indicate that at least for unmarried men, the idea that sexual risk behavior diminishes over time in a developmental process that derives from brain maturation or other biological factors is incomplete. In fact, our data suggest that some men maintain their early levels of risky behavior or even increase them as they age, because some who never reported risky behavior during the transition to adulthood reported it in their 30s.

No matter what level of past risk men exhibited, the lowest levels of current risk behavior were found among married men, the next lowest among cohabiting men and the highest among men outside a coresidential union. Our findings were consistent with the ideas that institutional norms regarding marriage may influence men to avoid risky behavior, and that behavior monitoring among married and cohabiting men may also be operating.

One implication of this study is that interventions intended to reduce men’s levels of risky sexual behavior should not focus exclusively on younger men or those in very high risk groups (e.g., men who have sex with men). In particular, clinicians should be aware that adult men living outside unions, as well as cohabiting men, may be engaging in risky sexual behaviors and may benefit from appropriate counseling and care.

Biography

Nan Marie Astone is senior fellow, Center for Labor, Human Services and Population, Urban Institute, Washington, DC. Joseph H. Pleck is professor emeritus, Department of Human and Community Development, University of Illinois at Urbana-Champaign. Jacinda M. Dariotis is assistant professor, Mark Emerson is research associate and Freya L. Sonenstein is professor—all in the Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, Baltimore. Arik V. Marcell is assistant professor, Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, and Department of Pediatrics, Johns Hopkins School of Medicine, Baltimore. Samuel Shapiro is a student, Department of Economics, Johns Hopkins Kreiger School of Arts and Sciences, Baltimore.