INTRODUCTION
Among the central and enduring issues in the study of race and American politics are questions about the relationship between marginalization and voting behavior, and recent elections have raised the salience of such concerns. The 2016 U.S. presidential race threatened several policy gains important to people of color and members of other marginalized groups, as debates about immigration, racial justice, reproductive rights, gendered violence, protections for transgender people, and marriage equality highlighted deep cleavages among Americans (Schuster et al., Reference Schuster, Reisner and Onorato2016; Stack Reference Stack2016). For instance, the Affordable Care Act, against which Donald Trump campaigned heavily, brought the rate of Black uninsured down from 21% to approximately 13% (Kaiser Family Foundation 2016).Footnote 1 The modest but important police reforms achieved in response to the demands of racial justice movements were similarly threatened by a Trump victory, who campaigned on a “law and order” platform that took explicit aim at these changes (Roberts and Cleveland, 2016; Trump Reference Trump2016). Trump’s selection of anti-gay stalwart Mike Pence as his running mate also signaled the possible reversal of recently won rights in areas such as marriage equality for the approximately 10 million voters who identify as lesbian, gay, bisexual or transgender (LGBT) (Grinberg Reference Grinberg2016; Stack Reference Stack2018).
Indeed, much was at stake in November 2016 for African Americans and LGBT people, two of the most reliable Democratic voting blocs (Dawson Reference Dawson1994; Egan Reference Egan2012; Frymer Reference Frymer1999; Strolovitch et al., Reference Strolovitch, Wong and Proctor2017; Tate Reference Tate1993). But although both Black and LGBT voters can prove pivotal in local and national elections (Egan Reference Egan2012; Lewis et al., Reference Lewis, Rogers and Sherrill2011; Schaffner and Senic, Reference Schaffner and Senic2006), we know little about the determinants of voting among Black LGBT people, whose marginalized status is constituted by the intersection of race and sexuality and whose concerns have historically been given short-shrift by both mainstream civil rights and LGBT rights organizations (Battle and Harris Reference Battle and Harris2013; Cohen Reference Cohen1999; Harris and Battle, 2013; Hunter Reference Hunter2010; Strolovitch Reference Strolovitch2007, Reference Strolovitch2012; Strolovitch et al., Reference Strolovitch, Wong and Proctor2017; VanDaalen and Santos, Reference VanDaalen and Santos2017).Footnote 2 And although decades of research on the determinants of political participation has made it almost an article of faith that members of higher status groups vote at higher rates than their lower-socioeconomic-status (SES) counterparts, recent work has suggested that there are circumstances under which members of marginalized groups might, in fact, vote at higher rates. Some of this research suggests that this might happen when voters are particularly enthusiastic about a candidate (see Cohen Reference Cohen2012), but research on “threat as a motivator” suggests that turnout might also increase when groups perceive candidates or elections as particularly threatening (Hansen Reference Hansen1985; Miller and Krosnick Reference Miller and Krosnick2004). We argue that when such threats are faced by marginalized groups, the concern to protect hard-earned rights can activate a sense of what we call “political hypervigilance,” and we examine whether the threatening political environment of 2016 heightened this vigilance among Black LGBT people, even as a lack of enthusiasm for the Democratic candidate led to a decline in turnout among African Americans more generally—from 66.6% in 2012 to 59.6% in 2016 (U.S. Census Bureau 2017).
LITERATURE REVIEW
To begin to understand these issues, we consider the dominant explanations that scholars have offered for civic participation in the United States. We start by drawing on political science scholarship to first explain why members of some groups vote and participate at higher rates than others. Next, we discuss the smaller body of research on marginalization and civic engagement, exploring the circumstances under which non-dominant groups do, at times—and somewhat counter-intuitively—vote and engage in other forms of political participation at higher rates than we might expect. We focus in particular on work showing that LGBT people are somewhat distinctive in this regard, professing higher levels of civic duty and engaging more frequently in protest activity than their straight counterparts. Finally, we explore recent scholarship that addresses civic participation among people whose marginalized status is structured by the intersection of race and sexuality, which is the substantive issue with which we are most concerned.
Explanations for Civic Participation in the United States
A large body of research has illuminated much about the broad drivers of political engagement in the United States (Brady et al., Reference Brady, Verba and Lehman Schlozman1995; Campbell Reference Campbell2003; Majic Reference Majic2014; Pierson Reference Pierson1993; Putnam Reference Putnam2001; Rosenstone and Hansen, Reference Rosenstone and Hansen2002; Uslaner and Brown, Reference Uslaner and Brown2005; Verba et al., Reference Verba, Lehman Schlozman and Henry1995). Emphasizing socioeconomic, resource-based, psychological/attitudinal, and policy-specific explanations for why individuals and groups participate at differential rates in American politics, this research has demonstrated some of the ways in which factors such as social location, financial resources, demographics and other markers of socio-economic status determine the extent of civic involvement (Brady et al., Reference Brady, Verba and Lehman Schlozman1995).
Socioeconomic factors: Among the most consistent findings of this work is that social and economic privilege is associated with greater levels of civic engagement. Henry Brady and his co-authors (1995) argue, for example, that the uneven distribution of resources such as time, money, and political skills leads to higher rates of participation among people from higher socioeconomic groups. This finding about resources echoes evidence from other research demonstrating that socioeconomic factors such as family background and education are similarly associated with civic knowledge and propensity to participate (Brady et al., Reference Brady, Verba and Lehman Schlozman1995; Uslaner and Brown, Reference Uslaner and Brown2005). Together, the findings of this body of work paint a picture in which wealthy, White, older, and college educated men are the most reliably enfranchised civic participants in the United States.
Psychological and attitudinal factors: Scholars have also shown that psychological and attitudinal factors play important roles in determining political participation in the United States. For example, higher levels of trust, optimism, and a sense of agency over one’s future are associated with increased civic participation across sociodemographic groups (Avery Reference Avery2006; Cohen Reference Cohen2012; Uslaner and Brown, Reference Uslaner and Brown2005). Likewise, the prototypical disaffected non-voter has low levels of trust in government to do the right thing, a lack of confidence that the government will be responsive to their needs, and a sense of alienation from the political system as a whole (Cohen Reference Cohen2012).
Policy feedback and threat as a motivator: Work addressing what scholars have come to call “policy feedback” (Campbell Reference Campbell2003; Pierson Reference Pierson1993) and “threat as a motivator” (Hansen 1986; Miller and Krosnick, Reference Miller and Krosnick2004) suggests that groups with a stake in a particular policy will turn out to protect these policies. Scholars have shown, for example, that among the ways in which “policies create politics” (Schattschneider Reference Schattschneider1960) is that new government programs create new stakeholders which, in turn, create new patterns of civic participation among these beneficiaries. Building on behavioral economics ideas such as prospect theory, which contends that people are loss-averse and therefore more likely to take an action to avoid financial losses than to pursue gains (Kahneman and Tversky, 1979), scholars have shown similarly that people are as motivated by a desire to “avert political threats” as they are by the possibility of desirable political opportunities (Hansen Reference Hansen1985; Miller and Krosnick, Reference Miller and Krosnick2004). Research has shown that beneficiaries of programs such as Medicare and Social Security, for example, turn out in large numbers to protect these policies when they are under threat (Campbell Reference Campbell2003), and that supporters of reproductive rights are more likely to contribute money to organizations whose solicitations emphasize threats to extant rights than to those whose requests are framed around opportunities to expand those rights (Hansen Reference Hansen1985; Miller and Krosnick, Reference Miller and Krosnick2004).
Marginalization and Political Engagement
While we are interested in political participation among members of marginalized groups, voting research has typically taken the experiences and behavior of wealthy, White, and college educated men as its normative and empirical baseline (Junn Reference Junn2007). A smaller but important and growing body of scholarship, however, has explored the civic engagement of members of marginalized groups (see, Cohen Reference Cohen1999, Reference Cohen2012; Egan et al., Reference Egan, Edelman and Sherrill2008; Majic Reference Majic2011, Reference Majic2014). Studies of voting behavior have shown, for example, that Black voter turnout has historically lagged behind White turnout, due, in no small part, to discriminatory laws and practices which have had as their goal the suppression of Black voters. Although the 15th Amendment to the Constitution and later the Voting Rights Act were enacted to protect voting rights for African Americans, factors including felony disenfranchisement, voter I.D. laws, and Supreme Court decisions about districting have eaten away at many of the most robust protections (Bentele and O’Brien, Reference Bentele and O’Brien2013; Nunnally Reference Nunnally2012). Similarly, turnout among women lagged well behind that of men in every presidential election from the ratification of the 19th Amendment in 1920 until 1976 (Center for American Women and Politics 2017).
Recognizing that these patterns in voting behavior do not capture the full range of ways that groups might be politically engaged, scholars of race, gender, and sexuality have explored a broader range of forms of political engagement. Many of the findings of this research have challenged the assumptions of traditional models that take as given that members of marginalized groups—including people of color, LGBT people, women, hourly workers, and others—participate at lower rates than members of dominant groups. Samantha Majic (Reference Majic2014), for example, has shown that sex workers—a highly stigmatized and low-resource group—are more engaged in activities ranging from voting to non-profit agencies and community-based work than traditional models would predict. Some research has even found evidence that under some conditions, members of marginalized groups vote at higher rates than their counterparts in dominant groups. Voter turnout among eligible women, for example, which had lagged behind men’s through 1976, surpassed it in 1980 and has remained higher in every presidential election since then (Center for American Women and Politics 2017). At 66.6%, voter turnout among African Americans in the 2012 election surpassed the rate for Whites for the first time since the U.S. Census began making this data available (U.S. Census Bureau 2017). Similarly, using data from the 2012 National Study of Learning, Voting, and Engagement (NSLVE), scholars at Tufts University found that Black college students participated at higher rates in the 2012 presidential elections than students of other races and ethnicities (Jonathan M. Tisch College of Civic Life 2012), perhaps because the Obama presidency heightened their sense of political efficacy (Cohen Reference Cohen2012).
The Political Distinctiveness of LGBT People
The traditional emphasis on voter turnout has been particularly unilluminating when it comes to understanding the political behavior of LGBT people, as surveys almost never included questions about LGBT identification until the 1990s (Egan Reference Egan2012). In the absence of this information, scholars have explored other avenues of LGBT political engagement, particularly the important role of social movements, political advocacy, and sustained mobilization in increasing access to HIV/AIDS-related treatment and resources (Cohen Reference Cohen1999; Gould Reference Gould2009). For example, the AIDS Drug Assistance program was established in 1987 under Republican President Ronald Reagan’s administration due to the persistent pressure of activists. This policy victory was particularly important to marginalized and stigmatized groups—low-income, LGBT, and HIV positive people who are disproportionately people of color—and exemplifies the importance of vigilance and commitment to political participation on the part of such groups and their allies.
More recent data has made it possible to explore more traditional measures of LGBT political attitudes and behavior (Egan Reference Egan2012; Egan et al., Reference Egan, Edelman and Sherrill2008; Swank and Fahs Reference Swank and Fahs2016, Reference Swank and Fahs2013). Pat Egan and his colleagues (2008), for example, found that lesbian, gay and bisexual (LGB) people are overwhelmingly likely to identify as and to vote for Democrats and also that they profess a greater sense of civic duty than their straight counterparts. LGB-identified voters also tend to show more liberal preferences on issues beyond those having to do specifically with LGBT rights, including foreign policy, environmental issues and the role of government (Egan et al., Reference Egan, Edelman and Sherrill2008; Egan and Sherrill, Reference Egan and Sherrill2006). And echoing the finding about Black students in the NSLVE study cited above, Eric Swank and Breanne Fahs (2016) found that college students who identified as LGBT were twice as likely as their straight counterparts to engage in protest. These differences were particularly pronounced among LGB respondents who rejected other forms of social hierarchy and who were members of other marginalized groups as well (Swank and Fahs, Reference Swank and Fahs2016).
Civic Participation at the Intersection of Race and Sexuality
Although the studies described in the previous two sections explain a great deal about the political behavior of African Americans and of LGBT people, samples of Black LGBT people in standard surveys are typically too small to conduct meaningful analyses. Those studies that have examined civic participation among sexual minorities of color, however, provide suggestive evidence that their intersectionally-marginalized status is related to even more politically-distinctive behavior. Work by scholars such as Angelique Harris and Juan Battle (Reference Battle and Harris2013), Dara Strolovitch, Janelle Wong, and Andrew Proctor (2017), and Rachel VanDaalen and Carlos Santos (Reference VanDaalen and Santos2017), for example, suggests that multiply marginalized individuals might be particularly socially and politically engaged, and also provides hints about the nature of that engagement. For example, Harris and Battle (2013) have shown that among “same-gender loving” Black women and men, feeling connected to LGBT communities was the most important predictor of sociopolitical involvement. While Harris and Battle find that a sense of community increases political engagement, scholars have found that some kinds of threats—such as the perception that heterosexism (Swank and Fahs, Reference Swank and Fahs2013) and racism (VanDaalen and Santos, Reference VanDaalen and Santos2017) are problems—is associated with increased political involvement, in the former case, among LGBT people of all races and in the latter case among LGBT people of color.
THEORY Political Hypervigilance
We argue that threat acts as a particular motivator for political participation among members of marginalized groups, for whom the concern to protect hard-earned rights is ongoing and can be particularly pronounced. Bringing the idea that civic engagement and political participation are often motivated by threats to rights and resources into conversation with ideas developed by psychologists to describe the ways in which trauma can instill the kinds of long-term heightened awareness of threats in one’s environment (e.g., post-traumatic stress disorder) (Dalgleish et al., Reference Dalgleish, Moradi, Taghavi, Neshat-Doost and Yule2001; Loo et al., Reference Loo, Fairbank, Scurfield, Ruch, King, Adams and Chemtob2001), we introduce the concept of political hypervigilance. Political hypervigilance is meant to crystallize the idea that the ongoing dangers faced by members of politically vulnerable groups lead them to remain on high alert for any sign that the hard-won and tenuous social and civil rights that protect them and their loved ones are under threat. On this view, the politics created by policies such as Medicare, Social Security, the Affordable Care Act and by rights such as same-sex marriage, are likely to be most salient to the beneficiaries who would be most vulnerable should they be lost, and that members of marginalized and intersectionally-marginalized groups will therefore be more—not less—likely to mobilize when they are under threat.
HYPOTHESES
Together, the studies that we have summarized suggest reasons to expect that LGBT people of color might be more politically engaged than their straight counterparts. Considered through the lens of our theory of political hypervigilance, these findings suggest further that this might be particularly true in threatening political contexts such as the 2016 presidential election—a context in which, as we discussed previously, the stakes for African Americans and LGBT people were particularly pronounced.
To explore whether Black LGBT people were more motivated to participate that year, we use original data from the 2016 National Survey on HIV in the Black Community (NSHBC), which allow us to examine the relationship among sexual orientation, policy views, and the intention to vote among African Americans in the 2016 election. Using these data, we test our theory of political hypervigilance, which holds that the particular threats faced by people whose marginalization is defined by the intersection of one or more forms of marginalization—operationalized in this case as being both African American and engaging in same-sex sexual behavior—increases their motivation to participate civically.
To these ends, we test two hypotheses. In particular, we expect that:
H1: African Americans who report same-sex behavior will also report greater intentions to participate civically.
H2: African Americans who demonstrate greater support for LGBT people and issues will also report greater intentions to participate civically.
Because our data include both straight Black respondents as well as African Americans who engage in same-sex sex, we are able to explore both the attitudes and behavior of Black LGBT people as well as some of the implications of African Americans’ attitudes toward LGBT people.Footnote 3
DATA AND METHODS
Black LGBT people are a relatively small and hard-to-reach population, and the difficulty of surveying them has meant that, with some important exceptions (such as Cohen Reference Cohen1999), they are rarely the focus of political science research (Harnois Reference Harnois2015). The NSHBC consequently offers a rare opportunity to study the political engagement of Black sexual minorities.
The 2016 National Survey on HIV in the Black Community
This study uses original data from the 2016 National Survey on HIV in the Black Community (NSHBC). The NSHBC is a probability-based web panel of African- Americans between the ages of 18 and 50. The panel is designed to be representative of adults living in households in the United States. To restrict our analyses to eligible voters, we only included respondents who were born in the United States. Panel members were recruited through random-digit dialing and address-based sampling. Address-based sampling enables the inclusion of households that are served only by cell phones or have no telephone service. Households without access to the internet are provided with access and hardware if needed. Similar probability-based web panels have been used in previous publications (Kelly et al., Reference Kelly, Squiers, Bann, Stine, Hansen and Lynch2015; Lantz et al., Reference Lantz, Douglas Evans, Mead, Alvarez and Stewart2016; Meng et al., Reference Meng, McLaughlin, Pariera and Murphy2016; Pynnonen et al., Reference Pynnonen, Handelsman, King, Singer, Davis and Lesperance2016). Panel members received an e-mail request to participate in the survey on February 12, 2016. Email reminders were sent to non-responders on day three of the field period. Additional email reminders to non-responders were sent on day 6, 10, 16, 25, and 35 of the field period. Data collection was completed on April 17, 2016. The study was approved by the Boston Children’s Hospital Institutional Review Board.
The survey included sixty-nine close-ended items including measures of voter registration, voting intentions, views toward LGBT populations, experiences with racism, HIV testing behavior, and demographic information. Data were weighted to adjust for nonresponse so that people responding to the initial screening questions matched the age, sex, and other characteristics of the total population 18 to 50, as estimated from the most recent Current Population Survey conducted by the US Census Bureau.
Key Variables
We operationalize our outcome of interest—“civic participation”—using a survey question that asked, “How likely are you to vote in the November 2016 elections for the U.S. President or Congress?” (hereafter referenced using the variable name Voting Intention). Of the possible responses, “Definitely will vote” was coded as 1 and all other categories (“Probably will vote,” “Possibly will vote,” and “Will not vote”) were set equal to 0. Please see footnote for additional details about our decision, and Appendix A for related regression results.Footnote 4,Footnote 5
Our analysis includes three key explanatory variables: respondent same-sex sexual behavior, anti-transgender attitudes, and attitudes about LGBT rights. To operationalize Any Same-sex sexual behavior, we used the survey item that asked, “Have your past or current partners been (Choose all that apply): Male, Female, Transgender (Male to Female), Transgender (Female to Male).” Men who reported having had a male partner and women who reported having had a female partner were coded 1, while respondents reporting having only partners of a different sex were coded 0. Table 1 reports the results of this variable disaggregated by sex to show Male-to-Male and Female-to-Female same- sex sexual behavior. Because only a tiny portion of respondents (N=6) reported having had sex with a transgender partner, these individuals were not included in the Any Same-sex Sexual Behavior variable.
Table 1. Characteristics of survey sample eligible to vote by November 2016 presidential election voting intention (definitely will vote vs. all other intentions).
a Will not vote, possibly will vote, probably will vote.
b Counts are unweighted. Percentages are weighted.
c Differences between definitely will vote and all other intentions significant at p<0.05. NS = not significant.
We used three items to measure attitudes toward trans people and about LGBT rights. The first item gauged support for same-sex marriage, asking respondents to choose their level of agreement with the statement: “The idea of same-sex marriages (two men or two women) seems ridiculous to me” (Stoever and Morera, Reference Stoever and Morera2007). In the logistic regression analysis, this measure was coded as a binary variable with “Strongly Disagree” and “Disagree” coded as 1, and “Neither Agree nor Disagree,” “Agree,” and “Strongly Agree” coded as 0. Anti-transgender attitudes were measured using two separate items that asked respondents to indicate their level of agreement with the following two statements: “I would feel comfortable if I learned that my best friend was transgender” and “Society should view transgender people as normal” (Walch et al., Reference Walch, Ngamake, Francisco, Stitt and Shingler2012). As in the case of the same-sex marriage variable, these variables were coded 1 for responses that denoted greater acceptance of transgender individuals. That is, the responses “Agree” or “Strongly Agree” were coded as 1, and all other selections were coded as 0.
Control Variables
In addition to these variables of interest, we also included controls for three well- established predictors of civic participation in U.S. elections: Age (1= Age 45 to 50), Gender (1= Female), Unemployment (1 = Employed), Income (1= Income >=$50,000), Education (1= Graduated high school), and religiosity (1=Attends services at least weekly) (Verba et al.,Reference Verba, Lehman Schlozman and Henry1995). These variables were all coded 1 for the value that we hypothesized to have greater associations with voting behavior.
Caveats and Data Limitations
While the NSHBC offers the opportunity to study the political engagement of Black LGBT people, we recognize that the data also present some limitations for our analyses. Most significant is that self-reported voting intention represents a somewhat limited measure of political participation, particularly if what we really care about is whether or not people do, in fact, vote. Unfortunately, however, the NSHBC did not include measures of actual voting behavior, nor did the survey ask about other activities we would have liked to explore, such as participation in rallies and protests, donating time to political campaigns, and the like. Although we are limited by the constraints of the NSHBC, and although self-reported intention to vote does not perfectly predict actual voter turnout, scholars have shown that the response that one will “definitely vote” is the best self-reported predictor of actual voting behavior (Harvard Kennedy School 2013; Rogers and Aida Reference Rogers and Aida2014). That is, voters are quite good at predicting that they will vote, even if they are less accurate in their assessments that they will not vote. Todd Rogers and Masahiko Aida (2014) found, for example, that only 87% of those who predicted that they would vote actually did so in the 2008 general election. As such, while the available measure does not perfectly predict actual voter turnout (and while it cannot rule out the possibility of biases between those who do and not vote among those who say they will), voting intention serves as a reasonable proxy through which to gauge political motivation among segments of the electorate.Footnote 6
Similarly, we recognize that same-sex sexual behavior is not the same as self- identified LGBT status. Egan (Reference Egan2012), for example, found that only approximately 68% of respondents in the 2008 and 2010 General Social Survey (GSS) who reported having had a same sex partner in the previous five years actually identified as lesbian, gay, or bisexual. Scholars have also documented significant differences between those who choose to identify as LGBT and those who engage in same-sex sexual behavior but do not identify as LGBT (Egan et al., Reference Egan, Edelman and Sherrill2008). For example, Egan and colleagues (2008) find that identification as LGBT is the key driver of political distinctiveness among sexual minorities. But while the lack of a more direct measure could confound our analysis, we show below (Appendix C) that the attitudinal and demographic characteristics of respondents who report same sex behavior are in line with the distinctly liberal and younger age distribution of LGBT-identified individuals in the general population of the United States (Egan et al., Reference Egan, Edelman and Sherrill2008; Gallup 2017; Lewis et al., Reference Lewis, Rogers and Sherrill2011).
We also acknowledge that the observational nature of the data available to us do not allow us to asses directly whether or not political hypervigilance drives the observed differences in voting intentions. Our ability to assess such claims is further compromised by the lack of a measure of political knowledge, which is likely a component of knowing that one’s rights are under threat and acting accordingly. As a partial remedy for this absence, we include a control for education, which many scholars have identified as a key predictor of political knowledge (Verba et al., Reference Verba, Lehman Schlozman and Henry1995). In addition, the analysis did not include information about the date on which respondents completed the survey, and we thus lack information about the proximity of the survey date to each state’s primary election. Given that the data were collected during the 2016 primary season, proximity to local races may affect survey results. Finally, social desirability bias may play a role in the extent to which respondents are willing to report their comfort or discomfort with transgendered people and gay marriage, for example. However, the NSHBC data were collected using a confidential web panel, a mode that has been shown to reduce the role of social desirability bias in survey responses (Kreuter et al., Reference Kreuter, Presser and Tourangeau2008).
STATISTICAL ANALYSIS AND STUDY RESULTS
Means and standard deviations for continuous variables and counts with percentages for categorical variables are presented in Tables 1 and 2. We conducted bivariate analyses (cross tabular and unadjusted logistic regression analyses) to explore statistically significant relationships among attitudinal and demographic variables and intention to vote. We then conducted a series of logistic regression analyses, controlling for Age, Gender, Unemployment, Income, Education, and Religiosity to examine factors driving voting intentions among all respondents eligible to vote. All covariates were coded as binary (0,1) variables, with the category hypothesized to be positively associated with definite voting intentions set equal to 1.
Table 2. Results from bivariate and multivariate regression analyses for November 2016 presidential election voting intentions, among those eligible to vote.
NS = not significant.
We regressed respondents’ stated intention to vote on the independent variables using logistic regression. The discrimination ability of the logistic models was measured by c- statistics with calibration assessed using Hosmer-Lemeshow chi-square statistics and their associated p-values. We employed an alpha of 0.05 in all statistical tests to determine statistical significance.
Descriptive Statistics
As the data in Table 1 make clear, slightly more than half of respondents in the sample are female (54%) and almost two-thirds of respondents were between 25 and 44 years of age (62%). Almost half of respondents had a household income of $50,000 or more (48%) and over half had completed some college (55%). Almost one-third of respondents were unemployed (28%), most were single (62%), and fewer than 40% of respondents were married or living with a partner. More than one third of respondents attended religious services at least weekly (39%), while the majority (61%) attended services monthly or less frequently. Ten percent (n=40) of the men and 12% of the women (n=58) in our sample reported having engaged in same-sex sex. These proportions are similar to the percentage of those reporting same-sex sex in the broader public but are also, as we expected, higher than the proportion of LGBT-identified individuals in the general population (Egan Reference Egan2012; Gallup 2017; Semlyen Reference Semlyen2017). A recent study by Gallup (2017), for example, found that 4.6% of non-Hispanic Blacks identified as LGBT. Similarly, the younger age distribution of NSHBC respondents reporting same sex behavior is similar to that of LGBT-identified people in broader studies (Egan et al., Reference Egan, Edelman and Sherrill2008; Gallup 2017). Although not statistically significant, the NSHBC subsample of those reporting same-sex sexual behavior reports lower average income levels and higher levels of education than those who do not report such behavior, in keeping with the pattern of socioeconomic status among the overall LGBT population (Egan et al., Reference Egan, Edelman and Sherrill2008; Gallup 2017).
We present respondents’ views about LGBT people in Table 1. These data make clear that relatively similar proportions of respondents agreed (30%) and disagreed (33%) with the statement that society should view transgender people as normal. A similar distribution is evident with respect to same-sex marriage, with 34% of respondents expressing an accepting view toward same-sex marriage—by either disagreeing or strongly disagreeing with the statement that “The idea of same sex marriages (two men or two women) seems ridiculous to me”—and 39% of respondents agreeing or strongly agreeing with this negative view. Similarly, just over a third of respondents (36%) agreed that they would feel comfortable if they learned their best friend was transgender, while just under a third (29%) said that they would feel uncomfortable.
Table 1 also presents the breakdown of professed voting intentions as they vary by demographic categories and by respondents’ attitudes towards LGBT people and issues. More than half of all NSHBC respondents (57%) reported that they “definitely will vote” in the 2016 elections. Sixty-three percent of women and 53% of men said they would definitely vote. A disproportionately high percentage (80%) of those reporting any same- sex sexual behavior responded that they definitely will vote. Similarly, although individuals reporting same-sex sex made up 10% of the overall sample, they comprised 14% of those who answered that they “definitely will vote.” Older respondents (35 to 44 and 45 to 50) were also over-represented among those reporting that they definitely intended to vote. For example, while 18% of respondents were between the ages of 45- 50, they comprised 21% of those who said that they would definitely vote.
Socioeconomic indicators such as household income, education level, and employment status were also positively and significantly associated with intentions to vote. But although previous work on Black civic engagement finds that religiosity increases civic participation among African Americans, frequency of religious services attendance was not significantly associated with the intention to vote among the respondents in our sample (Dawson et al., Reference Dawson, Brown and Allen1990).
Multivariate Results: Greater LGBT Tolerance, Greater Intended Electoral Engagement
The results of our logistic regression analysis are presented in Table 2, and reveal pronounced differences in professed voting intentions between people who reported sex with a partner of the same sex and those who did not (even after controlling for gender, age, income, education, religiosity, and unemployment status). More specifically, with an odds ratio of 3.2 (95% C.I.: 1.7, 5.1), the odds that a respondent who reported same-sex sexual behavior also reported that they would definitely vote was more than three times the odds for their heterosexual counterparts.
It was not only people who engaged in same-sex sex who were more likely to say they would vote; people who held more supportive attitudes toward same-sex marriage and about transgender people were also more likely to say that they planned to vote in the 2016 election.Footnote 7 For instance, those who disagreed with the statement that “The idea of same sex marriage seems ridiculous” had 1.8 times the odds of saying that they would definitely vote than those who were neutral about or agreed with the statement (O.R. = 1.8, 95% C.I.: 1,3, 2.4). Respondents who agreed or strongly agreed that they would feel comfortable if they learned their best friend was transgender had almost three times the odds of saying that they would definitely vote in 2016 than those who were neutral or reported less comfort (O.R. = 2.7, 95% C.I: 2.0, 3.6). Finally, broader support for transgender people was also positively associated with increased intentions to vote, as people who responded that society should view transgender people as normal were twice as likely as those who did not agree or were neutral on the issue to say that they definitely would vote in 2016.
DISCUSSION AND IMPLICATIONS
The findings of our analyses provide strong support for both of our hypotheses and, more broadly, for our theory of political hypervigilance. More specifically, the finding that African Americans who report having engaged in same-sex sex were more likely to say that they intended to vote in the 2016 presidential election lends credence to the argument that under some circumstances, multiple marginalization increases rather than decreases political participation. And because previous research has shown that those who claim an LGBT identity are more politically engaged than those who have same-sex sex but do not identify as LGBT (Egan Reference Egan2012), it is possible that our analyses understate both the extent and implications of this vigilance for voting.
Our findings also provide evidence of hypervigilance among those who express greater support for LGBT people and rights, who were also more likely to say that they would vote. Additional evidence that hypervigilance plays a role in increasing political activity among members of intersectionally marginalized groups is apparent among Black women (the group by and about whom the theory of intersectionality was developed), who were far more likely than their male counterparts to say they planned to vote (O.R. = 1.9; we also replicated this finding in a separate analysis of a sub-sample of more than 4,400 Black respondents to the 2016 Cooperative Congressional Election Study; O.R. = 1.2) (Crenshaw Reference Crenshaw1991).Footnote 8 Though not the only way to interpret our results, we argue that together they do suggest that politically vulnerable groups perceived the 2016 election as a threat to themselves and their loved ones and acted accordingly.
These findings also echo—and suggest an interesting wrinkle to—Egan’s (2012) arguments about the “political distinctiveness” of LGBT voters. In particular, they suggest that the political distinctiveness and civic duty evident among LGBT people and those who care about them may be driven by the relatedly “distinctive” sense of vulnerability and threat that members of marginalized groups feel in the face of potential losses to hard-won policy gains on which they often depend for their survival. In other words, greater political participation on the part of sexual minorities and of those who hold supportive views of LGBT people is a function not only of identity and ideology but also of material interests common to these communities. One particularly salient example of such a threat is the one posed by the 2016 election to programs for HIV/AIDS prevention and treatment—policies that affect people of color who have same-sex sex (even those who do not identify as LGBT) as the HIV risk for men who have sex with men remains high regardless of whether individuals identify as LGBT. The Affordable Care Act and state Medicaid expansions that have been targeted by the Trump administration have helped to expand and facilitate access to health care for low income people and people of color in general, and have also expanded access to lifesaving AIDS medications for people for whom HIV positive status was a pre-existing condition.
CONCLUSION
What is the relationship between marginalization and voting behavior? Whereas traditional research on political participation has typically taken as given that voting is positively associated with socioeconomic status and privilege, there also are circumstances under which members of marginalized groups turn out at higher rates than their higher-status counterparts. We argue that these high rates of participation are the result, in part, of a phenomenon that we call political hypervigilance—a particular way in which threat acts as a motivator for members of marginalized groups.
We recognize that our reliance on cross-sectional data and the absence of several theoretically-relevant independent and dependent variables limits our ability to make causal claims. Future research will try to address these concerns through survey experiments that allow us to test our claims about political hypervigilance more directly. This research will also include measures of actual voting behavior as well as other kinds of political engagement. Qualitative interviews and focus groups could also be used to explore the meaning of political hypervigilance for Black LGBT people themselves and to examine in greater depth why they participate at higher rates than their straight counterparts.
Although our results are situated in the context of one particular presidential election season, they are suggestive about the role of political hypervigilance in the political behavior of members of marginalized and intersectionally-marginalized groups. In particular, they suggest that in a context in which threats to LGBT communities and communities of color are likely to continue and to intensify, Black LGBT people will continue to remain vigilant about threats to policies in areas such as funding for HIV prevention and treatment programs. More generally, the concept of political hypervigilance also promises to help us understand the central role played by members of marginalized groups in so many contemporary social movements, from the foundational role of queer women of color in the Black Lives Matter movement—arguably the most consequential racial justice movement of the last several decades—to the Women’s March, #MeToo, and #SayHerName movements for which intersectional frameworks have been key organizing principles.
ACKNOWLEDGMENTS
We would like to thank the Women and Society Seminar at Columbia University, as well as Drs. Patrick Egan, Bob Blendon, Logan Casey, Keren Ladin, Andrea Acevedo, Madina Agénor, and Peter Levine for their helpful feedback and suggestions. We would also like to acknowledge Dr. Tony Earls and the NSHBC National Advisory Committee for their contributions and support. The research for this paper was made possible with help from the Harvard University Center for AIDS Research (CFAR), an NIH funded program (P30 AI060354), which is supported by the following NIH Co-Funding and Participating Institutes and Centers: NIAID, NCI, NICHD, NIDCR, NHLBI, NIDA, NIMH, NIA, NIDDK, NIGMS, NIMHD, FIC, and OAR. Funding was also provided by the UCLA Center for HIV Identification, Prevention and Treatment Services (CHIPTS) which is supported by NIH grant no P30MH058107. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Appendix A. Results from multivariate regression analyses for self-reported Voter Registration, among those eligible to vote.
Appendix B:Results from bivariate and multivariate regression analyses of Voter Intentions and Voter Registration among Black respondents age 18 to 50, Cooperative Congressional Election Study, 2016.
Appendix C. Characteristics of survey sample by sexual behavior
