Abstract
Gay marriage bans have received widespread support in the past decade in states across America. The authors evaluate the extent to which context mitigates or exacerbates support for gay rights. Their question is grounded in the debate over whether exposure to out-groups creates acceptance through contact or whether exposure induces hostility through threat. Using data from twelve marriage ban elections, the authors find conditional support for the threat hypothesis. While the presence of a gay population is unimportant to the majority of the population, evangelical Christians who live proximate to larger gay populations are more likely to support bans on gay marriage.
In 1978, Californians voted down Proposition 6, The Briggs Initiative, in a 58 percent no to 41 percent yes vote. The initiative would have required the firing of gay school teachers and officials and anyone who had openly pro-gay positions working in schools. The measure was the subject of an intense national campaign pitting gay rights opponent Anita Bryant and California State Senator John Briggs against local gay rights activist and politician Harvey Milk. Milk biographer Randy Shilts recounts the famous speech given by Milk after the Gay Freedom Day Parade in June 1978:
1
Gay people, we will not win our rights by staying quietly in our closets. . . . We are coming out. We are coming out to fight the lies, the myths, the distortions. We are coming out to tell the truths about gays, for I am tired of the conspiracy of silence, so I’m going to talk about it. And I want you to talk about it. You must come out. Come out to your parents, your relatives. I know that it is hard and that it will hurt them, but think of how they will hurt you in the voting booths. (Shilts 1982, 224)
Since the 1950s, leaders of the gay rights movement had argued that “one necessary precursor to equality was for gay men and lesbians to emerge from their closets” (Tadlock, Gordon, and Popp 2007, 193). The implicit argument in this call to arms is that voting behavior will be conditioned by social networks. In particular, interaction with groups seen as different will lead to feelings of understanding and ultimately to less reactionary views on policies affecting those groups. In the political science literature, this view is commonly called the intergroup contact hypothesis. In accordance with this hypothesis, contact with gays by the predominantly straight community should make them less likely to support policies viewed as restricting gay rights.
In contrast, a long research tradition in American politics involving racial and ethnic politics, dating back to Key (1949), has argued that the presence and proximity of large minority groups triggers increasing feelings of threat by dominant groups, which leads to an increased likelihood that members of majority groups will support antiminority policies precisely when they come into contact with the out-group. In accordance with this hypothesis, commonly termed the intergroup conflict or threat hypothesis, we would expect that straight voters who live in closer proximity to larger and/or apparent gay and lesbian communities will be more likely to support restrictive gay rights policies.
The American states offer a place ripe for investigation of the validity of intergroup contact and conflict theories around gay rights policies. The central reason for this is that policy issues often become matters of majority opinion in states as citizens are asked to vote on ballot initiatives, on referendums, and in ratification of constitutional amendments. In total, twenty-nine states have passed constitutional amendments banning gay marriage and another eleven states have passed statutes banning gay marriage. All of the constitutional amendments have been adopted since 1998. In contrast, only five states currently allow gay marriage—Massachusetts, Iowa, New Hampshire, Vermont, and Connecticut; in each case, these laws were enacted by courts and legislatures. 2 Simply put, the ability of nongay majorities to enact laws has contributed significantly to policies regarding gay rights in the United States. 3
Our core research question is set on resolving the tension between intergroup contact and conflict theories. Given that gay marriage bans have had overwhelming success in passing muster with voters, what is the trajectory of such policies as the gay rights movement grows? As the presence of gays in communities becomes more apparent and open, should we expect a more or less tolerant response from the majority straight community around them? And should we expect different groups to have different responses to living among openly gay neighbors? Or, as Harvey Milk famously argued, will agitation overcome “the conspiracy of silence?” To answer these questions, we analyze individual-level data on twelve marriage bans during the 2004 and 2006 elections, examining the effect of gay context on voting behavior. Our models demonstrate that the effect of context is generally unimpressive when examining the entire sample of voters, but the conditional effects are large and theoretically significant. Specifically, we find evidence of a macro-social context effect in that those living in more urban areas are considerably less likely to support gay marriage ban ballot measures. However, we also find that those who identify as evangelical or born-again Christians are more supportive of gay marriage bans in areas with larger gay populations. This finding in particular suggests that the intergroup hostility and threat experienced by white populations living in areas with numerous black residents that is so well documented by political scientists is extendable to the symbolic threat that homosexual couples pose to evangelical Christians. Thus, the findings of this article contribute to the broader study of context and voting behavior.
Context in Studies of Ballot Measures
Almost the entirety of the literature on context and public opinion and voting behavior focuses on racial and/or ethnic context and attitudes toward racial/ethnic minorities and policies linked to these groups. The psychological mechanism of contextual theories is well established: interaction and exposure to “out-groups” alters the heuristics that individuals draw on when it comes to expressing their opinions on policy issues (see Kinder and Mendelberg 1995). Where the consensus ends, however, is in how people use these interactions to develop opinions.
Intergroup contact theory argues that tolerance of out-groups by members of majority groups becomes more likely as interaction among in- and out-group members increases. The chief mechanism here is social; people become less threatened by those perceived as being “different” once they get to know others who may come from different social, racial, or ethnic backgrounds. Key to this theory then is the idea of meaningful interactions (see Bramfield 1946). In general, research suggests that for intergroup contact to produce increased inclusion, the interaction must be made between individuals of equal status who are working toward common goals (for a thorough review of the intergroup contact literature, see Dovidio, Gaertner, and Kawakami 2003).
In contrast, intergroup conflict/threat theory argues that proximity and competition fuel threat, which ultimately leads to group conflict (Kinder and Mendelberg 1995). The most well-known formulation of this theory was first articulated by Key (1949). Subsequent treatments of this question have argued that as the threat of intergroup competition becomes more intense, dominant groups tend to respond to out-groups with hostility (see Giles and Evans 1985; Giles and Hertz 1994). While many of the early studies considered race politics and the impact of racial context on school desegregation (Giles, Cataldo, and Gatlin 1975; Pettigrew 1957) and defection of white Democrats in the South (Huckfeldt and Kohfed 1989; Giles and Hertz 1994), as well as attitudes toward African Americans and policies that are associated with African American populations (Behrens, Uggen, and Manza 2003; Taylor 1998), subsequent research has also focused on white response to emergent Latino and Asian immigrant populations (Dixon 2006; Hood and Morris 1997, 1998, 2000; Stein, Post, and Rinden 2000; also see Oliver and Wong 2003 and Dixon and Rosenbaum 2004). Hood and Morris (1998), for example, investigating immigration policy attitudes, find evidence of both contact and threat responses to Latino immigrant populations—their findings indicate that contact and threat are contingent on the immigration status of the Latino immigrant population.
To be sure, a thorough review of the literature indicates that intergroup contact and threat theories can be thought of as complementary rather than adversarial. For example, in highly segregated areas, large minority populations might be observable to members of the majority population, without much or any meaningful interactions occurring between groups. Meanwhile, contact effects can be established in areas with very few minorities by many and repeated meaningful personal interactions. Some have argued, however, that increased minority populations can imply increased opportunities for contact (see Stein, Post, and Rinden 2000). Furthermore, much of the literature has treated the two theories as contradictory (Stein, Post, and Rinden 2000; Rocha et al. 2011; Hood and Morris 1997, 1998, 2000; McClaren 2003; Dixon 2006; Schlueter and Scheepers 2010; Wagner, Christ, Pettigrew, Stellmacher and Wolf 2006; Campbell 2006).
A number of studies have evaluated the relative importance of conflict and threat experiences in voting on ballot measures in the American states. This type of design is especially pertinent to examining the effects of context on political outcomes because they rely on expressed opinions about policy measures rather than responses to survey questions about hypothetical policies. Since we know that people often have not given much thought to survey questions before they answer them, we can expect greater error to be present in responses to hypothetical questions than to questions of how individuals actually voted (Zaller and Feldman 1992).
On similar questions, however, consistent results have not been the norm in this literature. Tolbert and Hero (1996), for instance, use a county-level design and find that support for Proposition 187, a 1996 California ballot measure that sought to restrict access to public services for illegal immigrants, was higher in “bifurcated” political contexts (areas with both large white, nonethnic populations and large minority populations). Branton et al. (2007) find that Democrats living in closer proximity to the border were more likely to support both Propositions 187 and 227. With individual data, however, Hood and Morris (2000) find that white voters in counties with a larger proportional Hispanic population did not become more likely to vote for Proposition 187. Very little work investigates the impact of social exposure to homosexuals, but what little research exists indicates that interpersonal contact is associated with warmer feelings toward homosexuals and less support for policies aimed at discriminating against gays and lesbians (Herek and Capitanio 1996; Herek and Glunt 1993; Brewer 2008, 57).
On a host of issues, then, there is evidence that the presence of out-groups induces understanding and tolerance and that contact can generate competition and ultimately threat. It is difficult to establish where gay rights issues fit within this rubric. On one hand, gays are not an emergent or immigrant group, and so we would not expect that the type of economic competition at the root of many threat theories would be induced here. On the other hand, opinions about gay rights might hinge on symbolic responses rather than economic competition.
Some recent research has made the argument that actual contact is often unnecessary to induce either positive or negative feelings about a group within a community. Particularly on civil rights issues, overarching “macro-social” factors can also drive responses. The factors include the relative economic and social characteristics of a neighborhood (Moore and Ovadia 2006). One of the overarching messages of opponents to gay marriage has been the “threat” legally recognized gay unions poses to the institution of marriage, religious liberty, and social values (Rom 2007, 4; Wald and Glover 2007; Wilcox, Brewer, et al. 2007). Experimental research finds that opposition to same-sex marriage is at least partially driven by the perception that it threatens heterosexual rights, social status, and identity (Schmitt, Lehmiller, and Walsh 2007).
The question of how individuals react to frames is especially important. A long tradition of research suggests that attitude acquisition is dependent on elite framing (see Iyengar 1991; Kinder and Sanders 1990) and especially on elite framing of trusted elites (Zaller 1992). However, some recent research has also suggested that there are limits to the effects of framing (Brewer 2001; Druckman 2001; Nelson, Clawson, and Oxley 1997), especially because often there are multiple frames being presented at the same time (Chong and Druckman 2007). Given that much evidence suggests some amount of deliberation (Brewer 2001) in the face of value-based frames, how should we expect individuals to react to frames? We argue that this implies two things: (1) individuals will make relevant use of their environments to contextualize the relative truth of the information they are receiving and (2) different individuals are likely to respond to different frames based on their social identity and personal values.
Of particular relevance is the work of Haider-Markel and Meier (1996), who argue that when gay rights policies are highly salient, the scope of conflict regarding these policies appears to fit the mold of “morality policy” rather than “interest group politics.” Participation and opinion formation in morality policy areas require very little information to participate in such issues, as everyone views their expertise as sufficient to have an opinion. Morality politics involves significant involvement of advocacy coalitions, particularly those from religious groups. The implication here is that when gay rights issues are less salient, victories are to be won for gay rights groups through traditional insider strategies for interest groups. Therefore, once salience increases, we should expect a high degree of policy congruence with majority opinion to restrict gay rights in most states. Lax and Phillips (2009) find that, indeed, a high degree of policy responsiveness exists on gay rights issues, although they report less policy congruence. 4 In addition, Lacey (2005) shows that salient ballot measures on “easy issues” often provoke higher levels of voter turnout. Brewer (2008) argues that while there is public feedback in the discourse on gay rights issues, elites have tended to use the malleability of public opinion on gay marriage to structure opinions. Conservative Christian elites, he argues, aggressively took the lead on framing marriage as a moral threat when the public began to be more supportive of antidiscrimination measures and gays in the military (also see Rimmerman and Wilcox 2007).
We read this literature as suggesting that when the issue of gay marriage is elevated to salient national status, as it was with the attention that Republican and conservative elites from both the political and religious worlds devoted to it, we would expect battle lines to be drawn by advocacy groups. Since salience transforms these issues into “easy issues” (Carmines and Stimson 1989), the possibility of social context to alter views on these issues might be more limited. In addition, the large amount of ambivalence present in gay marriage opinions (Brewer 2008, 96) means that elite cues should be particularly effective in structuring the opinions of group members (Zaller 1992). Indeed, leaders in the Religious Right clearly outlined for their members the threat posed by gay marriage (Lucas 2007; also see Wilcox, Merolla, and Beer 2007).
For this reason we expect evangelical Christians in particular to have a heightened sense of threat when they live among openly gay populations. The campaign against same-sex marriage, in fact, centered around the “threat” that gay marriage posed to family values, the institution of marriage, and religious freedom. As Wilcox, Merolla, and Beer (2007, 57) note, “Some Christian Right activists thought that the United States would be judged in the manner of Sodom and Gomorrah if it allowed same-sex marriage to continue; others believed that the basic fabric of society itself would unravel and heterosexual marriage would collapse as an institution.” Political leaders were rather unclear about their beliefs about same-sex marriage during the 2004 campaign. As a result, the Christian Right was at the forefront of framing the issue for attentive audiences (Wilcox, Merolla, and Beer 2007; Campbell and Monson 2008). 5 The frame that emerged from the leaders of the Christian Right uniformly focused on the threat posed by same-sex marriage (Lucas 2007; Wilcox, Merolla, and Beer 2007, 59). For example, the Family Research Council, a leading evangelical political group, placed the threat to marriage as the first reason it cites against same-sex marriage and notes in their pamphlet that “the slippery slope leading to the destruction of marriage as we know it draws ever closer with the decision of the Massachusetts Supreme Judicial Court to compel the state legislature to grant homosexual sex partners the legal status of married people” and concludes with a citation to a Harvard sociologist that no society that has condoned gay marriage has survived (Daily, n.d., 3, 10). Also, Pat Robertson, a longtime Christian Right leader, suggested several times on the Christian Broadcasting Network that the “ultimate conclusion” of legal same-sex marriage was a downward spiral to legal polygamy, bestiality, child molestation, and pedophilia. The Republican Party also made a specific attempt to court evangelical Christians by acquiring church directories to use for mailing and contacting voters with persuasive material. Focus on the Family, also “involved pastors in delivering their pro-traditional family message by distributing materials at the Southern Baptist Convention” (Magleby, Monson, and Patterson 2007, 23) and mailing pastors sample sermons to help them advise congregants on how to vote (Wilcox and Larson 2006, 71). The message that evangelical Christians heard from their leaders was clear: the sanctity of marriage as between solely a man and a woman must be preserved, or it will have harmful effects on the moral values of the nation and the communities we live in.
Therefore, generally speaking, we have a clear sense that context provides important information from which individuals might make decisions, but that information seems to be conditional on the group consuming the information. As ballot-initiative politics is likely to elevate gay marriage bans to “morality policy,” we would likely expect threat responses from members of specific groups, whose leaders are against gay marriage, particularly those with strong religious cues. In short, groups whose leaders cue a symbolic threat response should have stronger feelings in favor of gay marriage bans as their exposure to gay populations increases.
According to Sears et al. (1980), symbolic attitudes (as opposed to self-interest) structure much of opinion formation, particularly on controversial issues. As Sears (1993, 113) writes, “Political symbols often evoke and mobilize human emotions.” Given that gays already represent a static part of the population, however, we would not expect economic competition to be a driving factor in threat responses as individuals become aware of gays in their neighborhood, community, or workplace. Therefore, among groups who do not have a symbolic response based on morality politics activated by the campaign (e.g., Democrats and nonevangelicals), it is possible that exposure would instigate a contact effect (i.e., exposure to gay populations would therefore decrease support for gay marriage bans). To be sure, positive context effects on gay marriage opinions may have less to do with actual contact and more to do with local social forces contained in educated and urban areas. In addition, given that conservative Christian elites were out in front on this issue, we should expect the greatest structuring of opinions to occur among those most likely to receive the message: evangelical or born-again Christians.
The logical next step is to turn to empirics and examine the relative weight of contact versus conflict responses in conditioning voting behavior, with deference to the fact that the effects of context are likely to be conditional. Our general findings reveal that indeed the context effects attenuate one another, leaving us with null results in the baseline model. When conditional effects are examined, however, the story becomes more apparent. People in groups biased toward antigay opinions and whose leaders emphasize the threat of gay marriage become more stalwart in their opinions when they are proximate to larger gay populations, suggesting an intergroup conflict or threat effect. However, groups without an elite-cued bias toward antigay policies do not change their voting behavior on gay marriage ban ballot propositions as they live among larger or smaller gay communities. Instead, those who live in neighborhoods where macro-social forces are more favorable toward out-groups (urban, wealthy, educated, and liberal neighborhoods) show a decrease in support for gay marriage bans.
Research Design
To test for context effects in voting on gay marriage bans, we employ exit poll data from twelve states during the 2004 and 2006 elections. 6 We match each individual with census data on the percentage of households in the area in which the individual lives that are occupied by unmarried partners of the same sex aggregated to the congressional district (CD). We have, admittedly, chosen this level of “context” based on practicality, as it is the most precise geocode obtainable in the exit poll data. Thus, the contextual level we employ for our contextual analysis is quite large and noisy. However, there are some virtues to using CDs. First and foremost, while CDs vary in geographic size considerably, they contain roughly the same number of citizens. 7 In addition, if we appropriately control for any posited confounding factors, the noisiness of this measure should dampen any effects that we find. Thus, we consider this a relatively conservative test of our hypotheses and one that should be biased against finding evidence in favor of our hypothesis. 8
The exit poll data also allow us to control for all relevant factors including age, gender, race, ethnicity, income, and, more importantly, ideology, party identification, the approval rating of President Bush, and if the respondent self-identifies as a born-again Christian. 9 To maximize our control variables, we only use data from state exit polls that included all the variables listed above. These included Arkansas, Georgia, Kentucky, Michigan, Mississippi, Montana, North Dakota, Ohio, Oklahoma, and Utah in 2004 and Tennessee and Virginia in 2006. By including the political controls—Bush approval, party identification, and ideology—any finding above and beyond these controls is noteworthy given the increasing relevance of partisanship on voter choice and political preferences and evaluations (Hetherington 2001; Bartels 2000, 2002; Miller and Shanks 1996; D. Green, Palmquist, and Schickler 2002). We must also note that our religious measure certainly leaves something to be desired as it was coded as a dummy variable. We admittedly lump all nonevangelical Christians into a single category, thus lumping together Jews and Muslims as well as mainline and liberal Christian denominations. Although a more precise measure would be preferable, the noise this variable produces should lead us away from finding statistically significant results in our conditional models.
Our contextual controls are generated from the 2000 U.S. census. The percentage of same-sex partner households is measured as the percentage of all households in the CD that are occupied by unmarried partners of the same sex. 10 We also include contextual controls for the percentage of the population that is college educated, the percentage living in an urban environment, and the median household income for the CD. Urban environments as well as those with a larger college-educated population are expected to be friendlier to gay populations and more liberal overall. Thus, we include them to make sure we are not inadvertently measuring these political variables with our same-sex couples variable. We attained each of these variables from the 2000 U.S. Census Long Form. For comparison and interpretation purposes, all variables have been recoded to range from 0 to 1, where 0 is the minimum and 1 is the maximum observed value in the data set. Included in the appendix is a full description of each individual- and contextual-level variable. With the inclusion of all variables we are able to include 13,452 voters located in eighty-seven CDs in the completed data set.
To be clear, the constraints of available data unfortunately leave us with less than optimal measures on several fronts. However, in each and every case the limitations of the data should bias our results away from finding evidence of our hypothesis; thus, we consider this study to be a very conservative test of our hypotheses.
Models and Results
Our data consist of individual units nested within higher order contextual levels with an inherently interactive hypothesis between first- and second-order variables. Such an organization of data suggests a multilevel model. Since our outcome variable of vote is dichotomous, we employ a hierarchical generalized linear mixed model with a logit link function which is coded 1 for yes votes on gay marriage bans and 0 for no votes against these bans. We present two sets of models. In Table 1, we present general contextual effects, where level 2 variables are modeled on our dependent variable, a dichotomous measure of support for gay marriage bans, linearly as a shift in the constant term of the level 1 model. These models allow us to assess evidence for context effects generally. This model is presented in Table 1.
Hierarchical Generalized Linear Model of Voting Yes on Gay Marriage Bans, 2004–2006 Exit Poll Data—Base Models
Data are from the Inter-university Consortium for Political and Social Research exit poll series archive. States with marriage ban ballot propositions are Arkansas, Georgia, Kentucky, Michigan, Mississippi, Montana, North Dakota, Ohio, Oklahoma, and Utah in 2004; Tennessee and Virginia in 2006. Models are restricted maximum likelihood multilevel models with a logit-link function run in HLM6. p values are two-tailed tests. Level 1 N = 13,452; level 2 N = 87.
The differences between models 1 and 2 are the specifications of the level 2 variables. As the individual-level variables do not vary in interpretation across the 2 models, we can interpret them generally. Among the demographic, racial, and socioeconomic controls, we find that younger voters, those with higher family incomes, women, and whites are more likely to oppose gay marriage bans. 11 Meanwhile, among the political controls, conservatives, Republicans, and Bush approvers are most likely to support the bans. Also in line with our general expectations, born-again Christians are also much more likely to support gay marriage bans than are other voters.
We present three separate level 2 specifications because of the high degree of collinearity between most of our congressional-level variables. The first model includes only the census measure of percentage gay households. The second model includes controls for percentage college educated, percentage urban, and median household income. All of the variables have been recoded to vary between 0 and 1. In model 1, we observe what appears to be a contextual effect and more specifically an effect consistent with group contact theories that posit neighborhood characteristics to imply more or less frequent opportunities for contact—as the percentage of gay households increases, the probability of voting for the marriage ban decreases. However, given the correlation between percentage gay households and percentage urban (R = .46), the effect is completely mitigated in model 2. 12 Note that among the four variables, only percentage urban reaches significance. Theoretically, it is difficult to parse the differential effects of percentage gay and percentage urban. To be sure, however, we are hard-pressed to conclude that this is any strong evidence of an intergroup contact effect. However, a contextual effect does exist. Some have argued that contextual effects on civil liberties may have more to do with “macro-social” forces than actual interactions with individuals (see Moore and Ovadia 2006). Given the relatively high correlations between CDs with higher rates of college education, higher median household incomes, percentage of gay households, and urbanicity, it may be that in urban areas, such macro-social forces exist in regard to gay marriage. To be sure, it is also true that our inability to test for personal contact interactions with gays limits our ability to form solid conclusions about the intergroup contact effect given the null results. Indeed, one reading of the importance of urbanicity in promoting more tolerant attitudes toward gays is that urban areas are those that actually foster the greatest contact between gays and nongays, irrespective of the absolute size of the gay population in the district.
In Table 2, we present models to probe for interactive context effects; specifically as the percentage of gay households in a CD increases, so too does the probability of supporting gay marriage bans among evangelical or born-again Christians. Model 3 presents the cross-level interaction with only the percentage of gay households in the model. This model demonstrates that among all voters who do not identify as born-again Christians there is an apparent intergroup contact effect, while among voters who do identify as born-again Christians there is an intergroup threat or conflict effect. Model 4 adds the additional contextual controls, demonstrating again that the effect of an increase in gay households, even among the non-born-again subset of the data, is not statistically significant. Urbanicity continues to dominate among the contextual effects. However, among born-again or evangelical Christians, as the gay population increases, so too does evangelical support for gay marriage bans. Model 5 further demonstrates that this effect is not mitigated by urbanicity. Given the potential difficulty in interpreting these effects given that we have cross-level interactions in a nonlinear multilevel model, an illustrative figure of the substance of this effect is presented in Figure 1. 13
Conditional contextual effects of % gay households
Hierarchical Generalized Linear Model of Voting Yes on Gay Marriage Bans, 2004–2006 Exit Poll Data—Interactive Models
Data are from the Inter-university Consortium for Political and Social Research exit poll series archive. States with marriage ban ballot propositions are Arkansas, Georgia, Kentucky, Michigan, Mississippi, Montana, North Dakota, Ohio, Oklahoma, and Utah in 2004; Tennessee and Virginia in 2006. Models are restricted maximum likelihood multilevel models with a logit-link function run in HLM6. p values are two-tailed tests. Level 1 N = 13,452, level 2 N = 87.
Figure 1 demonstrates the effect of the percentage of households occupied by gay residents on the predicted probability of voting for gay marriage bans across twelve states. Effects are modeled by splitting out voters who identified as born-again Christians and all other voters (i.e., those who did not identify as born-again). The scale on percentage gay households has been recoded to range from 0 to 1, so the figure represents the actual range of the context variable. Among voters who do not identify as born-again Christians, there is, as the model coefficients suggest, no effect from an increase in gay population on the probability of voting for a gay marriage ban. However, among born-again Christians, the probability of supporting a gay marriage ban ballot measure increases by 10 percentage points from .77 to .87, a highly significant effect given the overwhelming support from evangelicals, conservatives, and Republicans for gay marriage bans generally. 14
Discussion
Decades of research in political science and sociology have explored the political consequences of minority social context. To date this research has focused almost exclusively on interaction between racial and ethnic groups. Over the past two decades, led mainly by the politicization of gay rights issues as social issues by elites, the gay community has emerged as a new politically relevant minority. In this article we examined the effect of exposure to gay couples on same-sex marriage ban initiative voting behavior.
The story that emerges from our analysis is complex. Given that same-sex partners are not economically threatening to majority groups, the threat they pose to majority populations is conditioned on feelings of symbolic threat that not all majority group members possess. The threat response to this minority group, therefore, is evident only among another minority population—born-again Christians. The evangelical community has been at the forefront of the movement against gay rights and same-sex marriage in particular. To some members of this group, gay marriage is “the defining battle in the culture wars—more important and immediately critical than abortion” (Wilcox, Merolla, and Beer 2007, 57) and poses a significant threat to both individual families and the country as a whole. This symbolic threat is present in almost all evangelical Christians, but, as our analysis shows, it is heightened when they live among large populations of same-sex couples.
The remainder of our sample appeared to be unaffected by living among larger or smaller same-sex populations. Unfortunately, our measure of context is large and does not measure interpersonal contact. Thus, as we have noted throughout, it may not adequately capture the contact effect. Certainly, future studies will be necessary to clarify if the effects of context on communities that do not receive elite cues against the minority experience no effect or if that contact leads to warmer feelings toward the minority.
However, although we find little support for the contact hypothesis per se, we did find evidence of macro-social factors at work. Residents of urban areas that house significantly more diversity in general and more liberal and educated populations were much less likely to support gay marriage bans. These findings together also point to a classic distinction between intergroup contact and threat theories: substantive interactions matter. Nonevangelicals who live in concentrated urban areas may be more likely to interact with gays and lesbians on a daily basis, resulting in the types of interaction one would expect with a null or contact effect. Evangelical Christians on the other hand are unlikely to experience substantial interpersonal contact with same-sex couples. Evangelical churches expect significant time from their members and facilitate church-based social networks that keep believers together in and outside of church (Iannaccone 1994; Finke and Starke 1992). Thus, while evangelicals may live in communities with large gay populations, their interaction with these groups may be very limited.
Furthermore, some evangelical political action groups have long campaigned against homosexual lifestyles and profess homosexuals to be child abusers, mentally ill, and/or heretics. The same-sex marriage movement gave a political outlet for an otherwise amorphous threat. Sharing a neighborhood with same-sex couples may therefore heighten the threat evangelicals feel from open gay and lesbian lifestyles.
Conclusion
So in light of these findings, what is the effect of the “conspiracy of silence” that Harvey Milk thought would ultimately lead to the majority harming gays at the voting booth? The effects appear to be conditional. For the majority of the population, the openness of gay populations in neighborhoods seems to have little additional affect above and beyond other contextual factors associated with more political and social tolerance, although such macro-social factors might predict more meaningful contact between gays and nongays, something we are unable to test for with the data presented herein. For communities predisposed to be against gay rights though, openness appears to instigate a threat response to marriage in the voting booth. Furthermore, this finding suggests that threats can be perceived as symbolic and that the political or economic relevance of a group in the market or in an election may not be necessary to induce threatened responses. Indeed, the role that elites play in constructing frames for their constituents is especially important.
Finally, our findings are interesting in light of recent survey research on the dynamics of opinion change on gay rights issues. By most accounts, attitudes toward same-sex couples are becoming more liberal. A recent Pew study found that nationwide opposition and support for same-sex marriage are converging. Opposition to same-sex marriage was found to be at an all-time low of 48 percent and support at an all-time high of 42 percent (Pew Research Center 2010). In some states support for same-sex marriage recently surpassed 50 percent (Gelman, Lax, and Phillips 2010). In some ways, the shift in public opinion to more pro-gay attitudes is the best signal that an overall contact effect is occurring. Certainly, gay issues and populations have been more present in society in the past decade than the previous one. However, our results are also consistent with the findings of Lupia et al. (2010), who argue that opinion–policy congruence is conditioned not only on shifting opinion but also on structural or institutional differences. In our data, behavior is conditioned further by contextual differences. To the extent that gay communities become more apparent and prominent, we would likely expect greater pressure and coherence within the evangelical community against progay policies.
Footnotes
Appendix
Acknowledgements
Authors’ names are listed alphabetically; contributions are equal. We would like to thank the editors, the anonymous reviewers, Beth Reingold, H. Whitt Kilburn, Scott McClurg, Ben Bishin, Brian Krueger, and the participants of the University at Buffalo Political Science Workshop Series for their comments and suggestions. Any errors present in the manuscript belong to the authors alone.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.