Unpacking the racial disparity in HIV rates: the effect of race on risky sexual behavior among Black young men who have sex with men (YMSM)
- First Online:
- Cite this article as:
- Clerkin, E.M., Newcomb, M.E. & Mustanski, B. J Behav Med (2011) 34: 237. doi:10.1007/s10865-010-9306-4
- 531 Downloads
The purpose of this study is to evaluate the large disparity in HIV prevalence rates between young Black and White Americans, including young men who have sex with men (YMSM). Research focusing on individual behaviors has proven insufficient to explain the disproportionately high rate of HIV among Black YMSM. The purpose of the present study was to gain a greater understanding of the pronounced racial disparity in HIV by evaluating whether YMSM are more likely to engage in risky sexual behaviors as a function of their partner’s race. Participants included 117 YMSM from a longitudinal study evaluating lesbian, gay, bisexual, and transgender youth (ages 16–20 at baseline), who reported characteristics and risk behaviors of up to 9 sexual partners over an 18-month period. Results indicated that participants were less likely to have unprotected sex with Black partners, and this finding was not driven by a response bias (i.e., Black YMSM did not appear to be minimizing their reports of unprotected sex). Furthermore, there was support for the hypothesis that participants’ sexual networks were partially determined by their race insofar as sexual partnerships were much more likely to be intra-racial (as opposed to interracial). It is possible that dyad- and sexual network-level factors may be needed to understand racial disparities in HIV among YMSM.
KeywordsHIV disparity YMSM Gay Race
One of the most alarming findings of the US HIV/AIDS epidemic is the large disparity in prevalence rates between Black and White Americans. This disparity persists even within groups with high prevalence, such as men who have sex with men (MSM), who were recently found to have 44 times the prevalence of HIV as other men who do not have sex with men and more than 40 times the prevalence of women (CDC 2010; Easterbrook et al. 1993; Lemp et al. 1994; Torian et al. 2002; Valleroy et al. 2000). Demographic factors have proven to be key in characterizing the HIV epidemic. Recent evidence indicates that most new infections among Black MSM occur among young Black MSM (13–29 years), while most new infections among White MSM occur at older ages (30–39 years). In fact, there were more than two times as many 13–24 year-old Black (versus White or Hispanic) MSM diagnosed with HIV or AIDS in 2006 (CDC 2009). Reasons for this pronounced racial disparity remain unclear. Research has shown that Black MSM are no more likely than White MSM to report engaging in unprotected anal intercourse, the greatest risk factor for contracting HIV among MSM (Easterbrook et al. 1993; Lemp et al. 1994; Harawa et al. 2004; Koblin et al. 2006; Ruiz and Facer 1998; Vitinghoff et al. 1999). Black MSM are also no more likely than their white counterparts to report substance use or greater numbers of sexual partners (Millett et al. 2006, 2007). In fact, in their meta-analysis of race differences among MSM, Millett et al. (2007) suggest that self-reported behavioral risk factors for contracting HIV are unable to explain the disproportionately high rate of HIV among Black MSM.
Various possibilities have been suggested to explain this paradoxical disconnect between reported risk behaviors and HIV prevalence, as well as to understand the influence of race on sexual risk-taking behaviors. One possibility is that racial differences exist in reporting biases, such that Black MSM may be more likely to minimize their reports of unprotected sex compared to other groups (see discussion of this possibility in Millett et al. 2007). According to this explanation, the racial disparity in HIV prevalence is due to behaviors that are not well detected in self-report surveys. A second possibility is that instead of individual behavior, sexual partner characteristics and sexual network factors (e.g., background levels of untreated HIV prevalence) underlie the racial disparity in HIV rates. These network factors have been found to underlie racial/ethnic group differences in sexually transmitted infections (STIs; Laumann and Youm 1999). Further, recent evidence suggests that network factors may also play a role in understanding risk for HIV infection among adult MSM (Bingham et al. 2003). For instance, Bingham et al. (2003) found that the relative odds of being infected with HIV was reduced for Black (relative to White) participants when the effects of having older and more Black partners were held constant. The authors concluded that particularly for Black MSM, characteristics of sexual partnership (i.e., having an older, Black male partner) might help to explain the racial disparity in HIV rates. However, to our knowledge there have been no studies that have directly examined the effects of partner race on risky sexual behavior among YMSM.
The purpose of the current study is to evaluate whether YMSM are more likely to engage in risky sexual behaviors as a function of their partner’s race. In particular, we were interested in exploring whether participant and sexual partner race had an impact on the frequency of unprotected sex within a given sexual partnership. This study was uniquely suited to evaluate several competing hypotheses regarding the effect of race on HIV risk behaviors. First, we consider whether there was a response bias whereby Black YMSM minimized their reports of unprotected sex. If there were a response bias, we would expect for Black YMSM to report less risky sexual behavior (e.g., unprotected anal or vaginal sex acts) with non-Black partners than non-Black participants reported with Black partners. Second, we explored the possibility that YMSM of all races were more likely to engage in risky sexual behavior with Black partners (relative to sexual partners of other races), which would tend to drive sexually transmitted infections into the Black YMSM population, while reducing its diffusion into other racial groups (Raymond and McFarland 2009). This pattern would help explain the disproportionate rates of HIV and other sexually transmitted infections within Black YMSM populations. Third, we investigated whether all YMSM would be more likely to use condoms with Black partners. This follows from evidence that some MSM utilize seroadaptive behaviors, which enhance the likelihood of condom use with serodiscordant or unknown status partners (Snowden et al. 2009). While speculative, it is possible that participants would be more likely to use condoms with Black partners given the greater objective and perceived prevalence rate of HIV among Black MSM (compared to non-Black MSM; Raymond and McFarland 2009). This pattern would not explain the racial disparity in HIV, but it is consistent with previous findings of similar or lower levels of sexual risk taking among Black YMSM (Millett et al. 2007). Finally, in a separate analysis, we evaluated whether participants’ sexual networks were partially associated with their race. An affinity towards sexual partnerships with individuals within one’s own racial group could potentially exacerbate infection rates within closed sexual networks, particularly when those sexual networks have high HIV prevalence rates.
Our study takes advantage of a unique longitudinal approach by evaluating ethnically-diverse YMSM who reported on the characteristics and risk behaviors of up to 9 sexual partners over an 18-month window. This design allowed us to test how sexual partners’ race affected risk behavior in the same person across multiple partnerships of potentially varying race, thereby helping to control for the putative confounding effects of any racial differences in reporting bias.
Description of MSM sample (N = 117)
Sexual orientation label
Heterosexual (same-sex attracted)
Asian or Pacific Islander
Procedure and design
Participants in the original sample were recruited through both venue (38%) and incentivized snowball sampling (62%). Participants were paid $40 for their participation at baseline interview and 12-month follow up (~2 h) and $25 for their participant at the 6-month follow-up interview (~1 h). At each of these three time points, data were collected on characteristics of up to 3 sexual partnerships during the 6 months preceding data collection (i.e., up to 9 partners across an 18-month window). For every partner, participants were asked to report on sexual behavior (type and frequency), use of condoms, and partner characteristics (e.g., the race and perceived HIV status of their partner). To minimize response and/or recall bias in this retrospective “follow-back” approach, the self-administered questionnaire required participants to create a timeline of the previous 6 months by indicating unique events throughout this period. This was intended to help participants remember specific events surrounding their sexual experiences. The project was approved by the Institutional Review Board.
The demographics questionnaire assessed participant characteristics, including age, race/ethnicity, and sexual orientation.
AIDS-Risk Behavior Assessment (ARBA) was designed for use with adolescents to evaluate sexual and drug behaviors related to HIV (Donenberg et al. 2001). The ARBA is self-administered using audio computer-assisted self-interview (ACASI) technology, and it employs a skip structure so that questions answered in the negative are not probed for more information. The ARBA has been utilized with a variety of youth populations, including ethnically-diverse adolescents, adolescents with psychiatric disorders, and YMSM (Donenberg et al. 2001; Garofalo et al. 2007; Nappi et al. 2009). In the current study we utilized the ARBA’s retrospective partner-by-partner questionnaire, which is intended to evaluate situational relationship and partner characteristics associated with up to 3 sexual partners during the prior 6 months.
Analyses were conducted with Hierarchical Linear Modeling (HLM) and procedures outlined by Raudenbush and Bryk (2002). This approach helped to account for clustering of the partnerships within participants and allows for estimation of both within-person and between-person effects. In our model, characteristics of sexual partnerships (Level 1) were nested within participants (Level 2). Our outcome variable, Sexual RISK, refers to the number of unprotected anal or vaginal sex acts with a given partnership. At Level 1, HLM estimates the within-participant effects of sexual partnership characteristics. In this case, we were interested in the effect of sexual partner’s race (PARTNER RACE) on the rate of unprotected sex with that partner. At Level 2, HLM allows for the evaluation of between-participant effects. Specifically, one variable (SELF RACE) was entered into Level 2 of the HLM to evaluate for the main effect of this between-subjects variable. Finally, given that there were both male and female partners, the gender of the sexual partner (PARTNER GENDER) was entered into the model to control for the effect of this variable on RISK; 12% of sexual partnerships were with women. Note that HLM estimates the value of the intercept when the predictor variable value is equal to zero. More specifically, Race was dichotomized such that participants who identified as Black were coded as 1 and participants not identifying as Black were coded as 0. Thus, even if participants identified as both Black and as another race, they were coded as 1. Maximum likelihood estimation and a Poisson distribution was used in estimating the counts of unprotected sex. The model also accounted for over-dispersion in the outcome variable resulting from the presence of outliers and an over-preponderance of cases with values of zero. Estimates are from the population-average model using robust standard errors.
Across all three waves of data collection (which covered 18 months), the remaining 117 participants reported a total of 416 sexual partnerships. Sexual RISK data—the number of unprotected anal or vaginal sex acts—was missing for three of these partnerships and PARTNER RACE was missing for six of these partnerships; these data were therefore not included in the analyses. The mean number of unprotected vaginal or anal sexual encounters with each sexual partnership was 5.74 (SD = 32.56).
Primary effects are presented using the Event Rate Ratio (ERR), which provides an estimate of the change in event-rate of the outcome variable for one unit increase in the independent variable. In the current study, the ERR refers to the change in the event-rate for unprotected sex for non-Black (coded as 0) relative to Black (coded as 1) participants.
Summary of all main effects of participant and partner race on sexual risk
Event Rate Ratio
Robust standard error
Self race (β01)
Partner gender (β10)
Partner race (β20)
Finally, sexual partnerships were homophilous (i.e., tendency to associate with similar others) with regards to race. Individuals were more likely to have sex with a partner of the same race (Odds Ratio = 38.20; 95% CI: 17.94–81.35; P < 0.001). Indeed, for non-Black participants, 160 sexual encounters were also with non-Black partners (88.4% of reported intra-racial partnerships) and only 21 sexual encounters were with Black partners (11.6% of reported interracial partnerships). Similarly, for Black participants, 191 sexual encounters were also with Black partners (83.4% of reported intra-racial partnerships) and 38 sexual encounters were with non-Black partners (6.6% of reported interracial partnerships).
To evaluate the robustness of the model, a series of follow-up analyses were conducted. The pattern of findings for our primary model remained the same when we a) controlled for HIV status of the participant; b) controlled for whether sexual partners were novel or repeated to help account for dependency in the dataset; c) controlled for perceived serostatus of one’s sexual partner; and d) when the count of unprotected sex acts was winsorized (at 3 SD) to help account for the effect of outliers (winsorized range = 0–98). There were very few cases where participants reported having sex with a partner known to be HIV positive, therefore we did not include this variable in our primary analyses. Additionally, it is important to note that the moderating effect of SELF RACE on PARTNER RACE did not reach significance. Thus, the interaction term was not included in the model to increase interpretability.
In the current study, we evaluated whether partner and participant race influenced sexual risk. In particular, we explored three competing hypotheses: (a) whether Black YMSM minimized their reports of unprotected sex; (b) whether participants would be more likely to engage in risky sexual behaviors with Black partners; and (c) whether participants would be less likely to engage in risky sexual behaviors with Black partners. We were also interested in evaluating whether participants were more likely to have sexual partnerships with individuals of their same race. Taken together, our findings were most consistent with the third hypothesis—participants were more likely to use condoms with Black partners. Moreover, our findings suggested that participants’ sexual networks were partially associated with their race.
These results have meaningful implications for understanding the disproportionately high rates of HIV among Black YMSM. Consistent with Millett et al. (2007), these findings suggest that individual-level behavioral risk factors, such as condom use, are insufficient to explain the racial disparity in HIV rates among YMSM. Importantly, the rate of sexual risk was similar in interracial partnerships where one member of the partnership was Black and the other member of the partnership was non-Black (i.e., between Black participants/non-Black partners and Non-Black participants/Black partners). This suggests that a response bias was not driving the results in the current study. In fact, the Event Rates were opposite to what would have been expected if Black YMSM were minimizing their reports of risk. Black participants actually reported somewhat more risky sexual behavior with non-Black partners than non-Black participants reported with Black partners. If Black participants were minimizing their rates of sexual risk, we would have expected them to report lower levels of risky sexual behavior with non-Black partners than non-Black participants reported with Black partners. In other words, if Black youth were minimizing their reports of risky sex, we would expect to see less risk reported when Black youth were the participants (as opposed to when non-Black youth were reporting about sex with Black partners).
The rate of sexual risk was lowest when both partners were Black and highest when both partners were non-Black. This pattern has at least two explanations. First, it is possible that YMSM know that HIV rates are disproportionately high within Black populations. As such, participants may have been trying to reduce their risk of infection by reducing their risk behaviors when having sex with Black partners. This phenomenon is referred to as “seroadaption” in the HIV literature to represent behavioral strategies undertaken to reduce risk of HIV transmission or acquisition by selecting partners of the same serostatus or by modifying risk behaviors based on perceived partner status. Second, it is possible that Black (relative to non-Black) YMSM are more interested in or willing to use condoms. Here Black youth are the agents in promoting safer sexual behavior within the partnership. Thus, condom use may be facilitated when both members of a sexual partnership are Black.
At face value, it is difficult to reconcile this finding with the fact that HIV rates are disproportionately high among Black YMSM. If condom use is greater with Black partners, one would expect that HIV rates among Black YMSM would be lowest in this group. To understand the contradiction between higher rates of HIV infection and lower rates of sexual risk among Black YMSM, it is important to consider the role of sexual networks in HIV transmission rates. As Tieu et al. (2009) note, sexual networks are crucial in understanding the spread of HIV because an individual’s acquisition of HIV results both from one’s individual risk behaviors, as well as from the risk behaviors and HIV prevalence rates of other individuals in his or her sexual network. Various studies have proposed that HIV rates are disproportionately high among Black MSM because they are more likely to choose partners of the same race or ethnicity (Berry et al. 2007; Raymond and McFarland 2009). Consistent with this idea, the present study found that Black participants were more likely to have a Black partner, and non-Black participants were more likely to have a non-Black partner.
It is outside of the scope of the present study to address why racial homophily was so pronounced, although it is likely tied to the concept of “assortative mating,” or more segregated partner choices among Black individuals (Laumann and Youm 1999; see also Anderson 1999). Past researchers have proposed that the sexual networks of Black MSM are more highly interconnected and constrained than the networks of other races in part because of the preferences and attitudes espoused by non-Black MSM. For instance, one cross-sectional survey conducted among MSM in San Francisco found that Black MSM were reported as the least preferable sexual partners (Raymond and McFarland 2009). They were also less likely to be considered friends, more likely to be considered difficult to meet, and more likely to be perceived as unwelcome in various gay-friendly venues throughout San Francisco. Ultimately, it will be important to better understand the racial homophily that characterizes sexual partnerships among YMSM, and the structural factors that drive it, in an attempt to potentially reduce the racial disparity in HIV rates.
Limitations and conclusions
The findings from the present study must be interpreted in light of several limitations. First, given the demographic makeup of our sample, we were unable to evaluate racial groupings other than Black and non-Black. While these groupings were adequate to address the hypotheses in the current study, it will be valuable for future work to consider more nuanced racial differences in sexual risk, as well as to replicate the findings from this study in a more nationally representative sample given that network characteristics may change with the demographic profile of a city. Second, we lacked the ability to evaluate why racial homophily existed in the current sample, and interracial partnerships were relatively rare so the estimates for these groupings should be interpreted with some caution. Finally, due to the sample size we had limited power to detect between-subjects effects, including the effect of participant race on sexual risk. Nevertheless, it is important to point out that while this effect did not reach significance, the effect size for participant race was actually larger than the effect size for partner race (which clearly did reach significance). These results converge to suggest that Black youth are more likely to assert condom use in a relationship, and therefore unprotected sex is least likely in a relationship where both partners are Black.
Notwithstanding these limitations, this study provides insight into the relationship between race and sexual risk among YMSM. Of the three competing hypothesis that were tested in the present study, our findings were most consistent with the hypothesis that sexual risk taking was less (as opposed to more) common in partnerships with Black YMSM. The data from the present study did not support the hypothesis that Black YMSM minimized their reports of unprotected sex. Finally, there was strong support for the notion that sexual networks were partially associated with race insofar as sexual partnerships were racially homophilous. More clearly determining the mechanisms underlying sexual network factors will provide an important next step for research focused on understanding and addressing the pronounced racial disparity in HIV rates.
The authors would like to thank the IMPACT program, the youth who participated in this research, and the staff at the Broadway Youth Center. We would also like to thank Dr. Don Hedeker for his statistical consultation. This research was funded in part by a grant from the American Foundation for Suicide Prevention and a scholar’s award from the William T. Grant Foundation to Dr. Brian Mustanski.