Electoral Studies 30 (2011) 607-620
ELSEVIER
Contents lists available at ScienceDirect
Electoral Studies
journal homepage: www.elsevier.com/locate/electstud
Two steps forward, one step back? Bias in the 2008 presidential election
Brett V. Benson a, Jennifer L. Merollab, John G. Geera*
a Vanderbilt University, USA
b Claremont Graduate University, USA
ARTICLE   INFO ABSTRACT
The candidates running during the 2008 presidential campaign were the most diverse in America's history. Prior to this historic election, female and minority candidates had little success in pursuing the presidency. Barack Obama's victory signals a decline in those barriers. Yet some groups, especially religious ones, continue to face barriers, including Atheists, Mormons and Muslims. The paper takes a close look at bias in presidential voting. This examination will provide an opportunity to consider new hypotheses about why barriers remain, shedding light on the nature and extent of bias within the American public. We consider social desirability, ideology, social contact, and group threat explanations. To test our ideas, we rely on list experiments using national representative samples in 2007 and in 2008. These data provide a unique opportunity to advance our understanding of the 2008 election, in particular, and the role of bias, in general. The results also offer some insight into future presidential elections.
© 2011 Elsevier Ltd. All rights reserved.
Article history:
Received 4 October 2010
Received in revised form 15 May 2011
Accepted 16 May 2011
Keywords:
Presidential elections Religion
Presidential candidates Public opinion
For most of American history, the majority of eligible citizens were eliminated from being considered for the presidency. If a qualified individual happened to be African-American, female, Catholic, or Jewish (to name just a few), they had little chance at winning the White House. The presidency seemed to be an office set aside for Protestant, white males. In 1960, John Kennedy broke down an important barrier for Catholics when he defeated Richard Nixon for the presidency. Of course, the 2008 Presidential election saw many barriers cast aside or weakened. Barack Obama's election as the first African-American president is historic. But we also saw Hillary Clinton succeed far more than any previous female contender for the Oval Office. Bill Richardson mounted a credible campaign as the first serious Latino candidate for president. And Mitt Romney was the first Mormon to ever win a presidential primary.
What happened in 2008 appears, at first glance, to be part of a more general pattern. Data from Gallup, for
example, offers an increasingly optimistic picture about the willingness of Americans to support candidates other than protestant white males (See Fig. 1). In 1937, nearly two-thirds of the public said they would not vote for a qualified woman for president. By the turn of the 21st century, only 7% of the country expressed such reservations. Fifty years ago, 53% of Americans were not willing to vote for a qualified African-American1 for president. In 2008, the proportion stood at 5% and, of course, America now has President Obama. In the 1950s, about a quarter of the electorate was unwilling to support a Catholic for president. That share dropped dramatically following Kennedy's time in office and now stands at just 4%. A similar pattern exists for possible support for a qualified Jewish candidate. In 1937, 47% of Americans would not support a qualified Jew and that share now stands at 5% as of 2007.2
In sum, there is good reason to believe that the long standing barriers facing most Americans are lowering and
* Corresponding author. E-mail   addresses:   brett.v.benson@vanderbilt.edu   (B.V. Benson), jennifer.merolla@cgu.edu   (J.L.   Merolla), john.g.geer@vanderbilt.edu (J.G. Geer).
1 As one indication of this change, the question in the late 1950s asked about qualified "negroes."
2 See Gallup Poll data, which are available through The Roper Center and can be located with the IPoll search engine.
0261-3794/$ - see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.electstud.2011.05.004
608
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
80
-^Catholic Mormon —^Latino
Fig. 1. Percentage who would not vote for a qualified candidate with the given characteristic for President, Gallup 1958-2007.
perhaps even ending. But ominous signs remain. In August 2010, one third of the public did not even think a Muslim should be allowed to run for president.3 This question does not even ask respondents about voting for a Muslim. When we do that, over half of the public seems unwilling to vote for a Muslim for president.4 Or consider the hurdles that Mitt Romney faced in his quest for the Republican nomination, particularly among Republican primary voters in the South. Concerns about his religion were so severe that he felt compelled to deliver a speech in November 2007 ("A Symphony of Faith") that sought to ease concerns about his Mormon faith. Furthermore, contrary to the trends noted above in the Gallup data, the percentage of individuals who would not vote for a qualified Mormon has remained virtually the same since Mitt Romney's father, Governor George Romney (R, MI), sought the presidency over forty years ago.5 In April 2011, nearly 50% of Americans remain "uncomfortable" with the possibility of a Mormon being president.6 This kind of evidence indicates that while some groups may be facing less bias in the pursuit of political office, others have made little or no progress.
This paper takes a closer look at the biases facing a wide range of potential contenders for the presidency. This is an important empirical question in and of itself. But by pursuing this topic, we also advance broader empirical and theoretical concerns. First, the public's willingness to
3 Survey by Time, August 16-17, 2010, available through The Roper Center.
4 Survey by Bloomberg/LA Times, June 24-27, 2006.
5 The data from 1967 come from a Gallup poll in April 19-April 24 that year, relying on 2190 personal interviews. In March 2007, Gallup asked a similar question and 20% indicated they would not vote for a qualified Mormon for president (March 2-March4,2007 and based on 1010 telephone interviews). Using a different question in July, 2008, Newsweek reports that about a quarter of Americans would not support a qualified Mormon for president (Survey by Newsweek. Methodology: Interviewing conducted by Princeton Survey Research Associates International, July 9-July 10,2008 and based on 1209 telephone interviews using a national sample of adults).
6 Survey by NBC News, Wall Street Journal. Methodology: Conducted by Hart and Mclnturff Research Companies, March 31-April 4, 2011.
support "minority" candidates for president sheds light on the overall level of discrimination present in the country. If people are not willing to vote for a qualified candidate from a particular group, then that sends strong signals about prejudice. Second, we tackle this topic with an eye towards the role that social desirability may play. It is quite possible that the increased willingness of Americans to say that they would vote for African-Americans or women, for instance, is misleading, since the answers may reflect a shift in how people respond to such questions not a change in their actual attitudes. In other words, citizens may feel pressure to say they are not sexist or racist, when in fact they hold such views. We collected our data with this concern in mind.
Third, we seek to understand factors that may or may not mitigate bias with respect to a candidate's religious characteristics. We start by considering that religion may be a proxy for ideological views and bias may reflect such considerations. We, then, consider how social contact with the relevant religious minority influences the public's thinking about candidates. Here, we evaluate the predictions generated from social contact and group threat theory, which offer interesting cross-cutting hypotheses. According to the former, increased social contact with groups should diminish prejudice, while according to the latter an increased presence of a minority group may actually intensify it. Both theories have merit, but we argue that which one is applicable depends on the political salience of religious threat in different electoral contexts for particular groups. We try to weave these competing explanations into a single, more compelling narrative.
Our data come from two internet-based experiments run through Polimetrix. The first was run in November 2007just prior to Romney's speech in Texas about his Mormon faith. The second was conducted right before the balloting in November 2008. Besides running a series of experiments, which we will describe below, a key part of our project involves an over-sample of southern born-again Christians.7 Evangelicals Christians have been a focus of research for a while (e.g. Layman, 1997, 2001; Campbell, 2006). But we wanted to focus on the South, since the political salience of religious threat should be high for this group. The Romney campaign worried from day one about southern evangelicals and the role these voters would play in the 2008 presidential nomination process.8 Our impression was that
7 One of the big issues confronting this project is the definition of "born-again" Christians. We will from time to time use the term "born-again" and evangelicals interchangeably. Our measurement strategy, as one will see, is straightforward. We simply ask people if they view themselves as born again. There is a large literature that deals with this issue (e.g. Layman, 2001). Some will quibble with this choice, calling for alternative measurement strategies. Our approach is consistent with the work of Campbell and Monson (2008). We sought a simple analytical approach to this issue and employed self-identification to get at it.
8 We want to thank Vin Weber, a close Romney advisor in 2007/08, who shared with us his many insights about the campaign, which confirmed our reasoning for focusing on southern evangelicals. For example, Romney's team viewed evangelicals as part of the dominant culture in the South, making their views much more of a hurdle for Romney than in communities where evangelicals were less prevalent.
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
609
this group was very much opposed to Romney's candidacy, even though Romney's stated views on social issues meshed well with those of southern evangelicals. Certainly our view squares with commentary in the news media about the many concerns born-again Christians had about a Mormon as president.9 We followed the same procedure when collecting our data in October 2008.
The findings tell a mixed tale about bias in presidential campaigns. Our data suggest that the public is more willing to vote for a female or an African-American, at least in the abstract. But that progress is undermined by what appears to be religious intolerance against certain faiths. And the bias is far greater among southern born-agains, as we expected. Increased social contact appears to diminish bias among the general population as expected; however, the role of social contact is more nuanced for southern evangelicals, a group for whom religious threat is particularly relevant. Social contact diminishes bias in contexts in which the threat posed by the religious minority is not politically salient, while social contact does not have such positive effects when threat is politically salient.
1. Establishing a baseline
How much bias exists against various minority candidates? The Gallup data are of interest, but they are potentially misleading. Individuals may simply not be willing to admit directly to being biased against certain groups because it is considered socially unacceptable, while they may think it is more acceptable to express bias against other groups. For instance, it is likely to be more socially acceptable to say one would not vote for a qualified Atheist than a qualified female for President. If so, bias will appear less intense against a woman than actually exists. In order to rule out the possibility that we are observing social desirability effects for some groups and not others, we relied on list experiments. Such data will also help provide an important baseline for assessing the amount of prejudice that exists among the American public.
List experiments are commonly used to tap attitudes that individuals may be less willing to express publicly. In such an experiment, subjects are presented with a series of statements and are asked to indicate the number of statements that they agree (or disagree) with. A control condition is given a list of statements and a treatment condition is given one additional statement, which pertains to the bias that the researcher is trying to uncover. Since individuals are only asked to indicate the number of statements they agree with, they do not need to state directly any biased views. If there is no bias, then the average number of statements that the subjects agree with should be the same across the control and treatment group. We get an estimate of bias by subtracting the mean number of statements between the control and treatment group. List experiments have been widely used to deal with attitudes
9 For examples of such media coverage see: http://www.Washington monthly.com/features/2005/0509.sullivanl.html; http://www.boston.com/ news/globe/editoriaLopinion/oped/articles/2007/07/01/anti_mormonism_ gets_personal/; http://www.nytimes.eom/2007/n/28/us/politics/28repubs. html?_r=l.
related to race (e.g., Gilens et al., 1998; Kuklinski et al., 1997; Kuklinski et al., 1997b; Sniderman and Carmines, 1997). More recently, scholars have applied list experiments to study propensities to vote for a Jewish candidate (Kane et al., 2004) and a female candidate for President (Streb et al., 2008).10
We embedded list experiments in two internet-based survey studies conducted by Polimetrix, one of which was in the field from November 12-29th, 2007, just before the start of the Republican presidential primary season, and the other was in the field in the last week of October 2008. Both samples include 1800 subjects, with an over-sample of 600 southern born-again Christians.11 With proper weights12, about 52% of the samples consisted of females, and with respect to race, 76% of the samples were Caucasian, 11% African-American, and 12% Latino. About 25% of the samples had a college degree or higher, while 30% were age 55 or older.
In our first list experiment, in November 2007, respondents in the control condition (n = 330) were exposed to three core statements: "I could not support someone who has smoked marijuana for President of the United States; I could not support someone over 75 for President of the United States; I could not support a Muslim for President of the United States." To look at bias against "minority" candidates, we had five experimental conditions in which we inserted an additional statement. One treatment group (n = 280) saw the following statement: "I could not support a Mormon for President." In the four other treatments we asked about a Woman (n = 298), an African-American (n = 286), an Atheist (n = 286), and an Evangelical for President (n = 320). We wanted to look at both atheists and evangelicals in an effort to consider various forms of religious bias that would complement our focus on Mormons.
In our second study, in October 2008, we ran a similar list experiment, but added an extra treatment condition to consider explicitly bias against Muslims (i.e. "I could not support a qualified Muslim for President"). We replaced the Muslim control statement in the original list experiment with the following statement: "I could not support a Communist for President of the United States." For our first
10 We recognize that this method measures explicit prejudice and may not tap into prejudice that lies outside of conscious awareness, but may nonetheless influence support for minority candidates. Scholars of social psychology have developed implicit measures to get at bias that individuals may not state explicitly (for a review see Fazio and Olson, 2003). With respect to the 2008 election, scholars have found that individuals who exhibit prejudice on the implicit association test and the affect misattribution procedure, as well as on explicit measures of prejudice, were less likely to vote for Obama (Greenwald et al., 2009; Payne et al., 2010; Pasek et al., 2009). In applying another type of implicit measure, subliminal priming, to candidate preference, Kam (2007) had a similar finding in which implicit and explicit prejudice influenced support for a Hispanic candidate in a judicial election, but only in the absence of party cues. When party cues were present, neither measure influenced candidate support. Thus, while explicit bias against females and African-Americans for political office may be minimal, implicit bias may still be pronounced. Our focus in this paper is only on explicit bias since it appears that even explicit bias has not declined dramatically for candidates with certain religious characteristics, such as Mormons and Atheists.
11 Polimetrix uses a screen in which an individual is asked if they are born again or not. The over-sample is of individuals residing in the South who indicated being born again.
12 We utilize the survey weights Polimetrix employs based on known marginals of the general population of the United States from the 2005 American Community Survey and the 2004 National Annenberg Survey.
610
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
study, we were focused more on bias against a Mormon candidate, reflecting the controversy around Romney's candidacy. Given what had unfolded in the 2008 campaign, we added Muslim as a treatment condition to expand further our examination of bias against candidates with different religious characteristics.
The distribution of subjects across the control and treated groups in our second study is as follows: Control (n = 261); Mormon (n = 270); Woman (n = 281); African-American (n = 256); Atheist (n = 229); Evangelical (n = 242); and, Muslim (n = 258). If social desirability effects are driving the differences in support for female and black candidates relative to the different religious minority groups, then we should find that bias against the female and African-American candidate is higher in our list experiment relative to the Gallup data.
In the top half of Fig. 2, we present the proportion of individuals in 2007 who would not vote for various candidates among the nationally representative sample and among born-again southerners. This proportion is arrived at by subtracting the mean number of statements that people agree with in the control group from the mean number of statements that people agree with in the relevant treatment group.13 Looking first at the 2007 national sample, we can rule out the hypothesis that the differences in bias we observed in the Gallup data concerning gender, race, and religious characteristics are due to social desirability effects. According to difference in means tests between the control group and each treatment group, we do not detect any significant bias against a female (p = 0.18), an African-American (p = 0.22), or an Evangelical presidential candidate (p = 0.12), while we do find significant bias toward an Atheist (p = 0.00) and a Mormon candidate (p = 0.00). Nearly 50% are not willing to support an Atheist, while 33% are not willing to support a Mormon. The results for Atheists are similar to those registered by Gallup, while the bias is almost doubled for the Mormon candidate with this less direct way of asking the question. If we turn to born-again southerners14, a higher percentage is significantly less willing to support an Atheist (p = 0.00),
13 The mean number of statements agreed to in the control group for the 2007 study is 1.33 for the whole sample and 1.92 for the born-again sub-sample. For 2008, the values are 1.29 and 1.23. For both samples, we did distribution tests to ensure that subjects were evenly distributed across experimental conditions across many demographic and political indicators, including being born-again, political interest, partisanship, ideology, age, gender, race, education, marital status, church attendance, and income. There was an uneven distribution by race across experimental conditions in the 2007 study, but controlling for race in a regression does not alter the results. Full regression results are available upon request.
14 The n for each experimental condition for the southern born-again sub-sample is as follows: control (n = 133); Mormon (n = 109); African-American (n = 128); Woman (n = 102); Atheist (n = 109); and Evangelical (n = 138). Theoretically, we should not observe the negative proportion on the evangelical candidate, which is created from subtracting the control mean from the evangelical treatment mean. List experiments assume that the number of control statements that people agree with does not change with the presence of the treatment statement. However, the difference between the two conditions is not statistically significant. This also happens in the whole sample for the comparison between the control group and those in the African-American treatment group, though the difference is essentially nil, -0.01, and not at all significant.
77%, and a Mormon candidate (p = 0.01), 39%. Thirty percent are also less willing to support a female candidate (p = 0.03).15
The results for the 2008 study are displayed in the bottom half of Fig. 2. For the whole sample, subjects report being significantly less willing to support an Atheist, Evangelical, Mormon, and a Muslim candidate.16 The proportion not willing to vote for an Atheist is higher, at 65%, while the proportion not willing to support a Mormon is lower, at 20%. Furthermore, 58% are not willing to support a Muslim, while 29% are not willing to support an Evangelical. If we turn to born-again southerners17, we find significant differences between the treatment and control for all groups, with the bias most pronounced for Atheists and Muslims.18 It is a bit surprising that born-again southerners show bias against an African-American and Evangelical candidate, when they did not in the 2007 study. We suspected that some respondents may be projecting to particular candidates running for office, namely Barack Obama and Sarah Palin, but the results remain the same if we run a regression and control for feelings toward these two candidates.
All in all, Fig. 2 tells a mixed story. The general public does not seem to harbor much explicit bias against female and black candidates, at least when asked about an abstract candidacy.19 However, we do observe explicit bias against certain types of candidates on the basis of religious affiliation, and for the Mormon candidate, the bias is more pronounced than what is found in the Gallup surveys.
While it does not appear that social desirability is driving individual responses for most of the candidate types, might the bias reported above reflect ideological inferences and not necessarily prejudice against religious groups? People might reasonably assume that Mormons are conservative in light of the well-known views this
15 One possibility is that the bias against a Mormon for both samples and the bias against a female among southern born-again evangelicals reflects antipathy towards Mitt Romney and Hillary Clinton. We ran a regression on the number of statements the respondent agrees with and included dummy variables for each treatment condition (the control group is the baseline) and feeling thermometers towards Mitt Romney and Hillary Clinton. In a regression context, the coefficient on the treatment variable is the difference in the mean between the control group and the given treatment. Including the feeling thermometers does not lead to any substantial change in the treatment effects. These results are available upon request.
16 These are all significant at p < 0.05. Our findings are also in line with previous comparisons between feelings toward blacks compared to feelings toward atheists in the 2004 National Election study (Kalkan et al., 2009).
17 The n for each experimental condition for the southern born-again sub-sample is as follows: control (n = 104); Mormon (n = 112); African-American (n = 99); Woman (n = 107); Atheist (n = 91); Evangelical (n = 81); and, Muslim (n = 83).
18 These are all significant at p < 0.05.
19 Again, we recognize that females and African-Americans still face many uphill battles in running for office, even if people are less inclined to explicitly state that they would not support a female or African-American for office. In footnote 10, we cited literature linking implicit measures of prejudice to less support for Obama and a Hispanic candidate. Studies have also linked explicit measures of racial prejudice to lower support for African-American candidates during real and hypothetical elections (e.g., Citrin et al., 1990; Terkildsen, 1993). With respect to women, experimental studies of hypothetical elections have found evidence of gender stereotypes in voting behavior (e.g., Ekstrand and Eckert, 1981; Huddy and Terkildsen, 1993; Sanbonmatsu, 2002).
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
611
2007
-0.2
-0.4
African-American
Atheist
Evangelical Candidate Type
Mormon
Woman
2008
0.9
0.8
5 0.7 ■o
■C 0.6
0
Q. Q.
• 0.5
s
01
c
= 0.4
Š
o
= 0.3
o
t
S. 0.2 p
0.1
-0.1
■ whole sample □ bom-again
African-American
Atheist
Evangelical
Mormon
Muslim
Woman
Candidate type
Fig. 2. Difference in the Mean Number of Statements Agreed to in the Control Group and each Treatment Group, Whole Sample and Born-Again Southerners.
612
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
group takes on social issues. Atheists, by contrast, are often perceived as being liberal because of that group's opinions on most social issues. If survey respondents are making inferences from religious affiliation to ideological beliefs, then an unwillingness to support a candidate from that group may be tied to ideological differences rather than intolerance.
To disentangle religious bias from perceived ideological differences, we collected additional data in the fall 2008 study. Specifically, we asked subjects to place themselves and each group (African-Americans, Women, Atheists, Mormons, Muslims, and Evangelicals) on five-point ideological scales. We used these measures to generate a dichotomous indicator for low or high ideological distance between the respondent and each group.21 In order to explore whether there is bias against each candidate type controlling for ideology, we regress the number of statements the respondent agrees with (in the list experiment) on a measure of whether the respondent is in a particular treatment group (the control serves as the baseline), a control for the ideological distance between the respondent and the group in the treatment, and an interaction between the treatment and the ideological distance measure. We run a separate model for each type of candidate. If ideological distance moderates the effect of the treatment, then we should find a significant p-value on the interaction term (Kam and Franzese, 2007).The coefficient on the treatment variable represents the difference in mean between the control and the given treatment group for those low in ideological distance, while we separately calculate the effect for those high in ideological distance when we find evidence of a moderating effect. We use regression rather than difference of means tests in order to see whether ideology moderates the effect of the treatment.22 The results are presented in Table 1. 23
For the whole sample, we find that ideological distance moderates the effect of the experimental treatment for the African-American and Muslim candidates since the p-value on the interaction terms are significant (both at p < 0.05).
20 For example, in a recent national survey released by the Pew Foundation, Mormons and Evangelical Protestants were the most conservative of the religious traditions studied. http://pewforum.org/docs/?DocID=429
21 We took the absolute value of the difference between the respondent's self-ideological placement and the placement of each group and then split the variable at the median distance (2.5). We dichotomize this variable so that we have enough respondents in each experimental condition that fall into different values on the ideological distance measure when we interact the two. It is important to have a big enough n per condition at different levels of ideological distance for the whole sample as well as the born-again southern sub-sample. For example, in the whole sample only 67 respondents received the Mormon treatment and are high in ideological distance, while the comparable n for the born-again southerner sub-sample is only 18.
22 An alternative strategy is to do a series of difference in means tests between the control and each treatment group for those low and high in ideological distance. The estimates of the difference in means and standard errors using this alternative strategy are identical to what we obtain using the regression analyses in Table 1. These results are available upon request.
23 We recognize that a count model may be more appropriate given the coding of the dependent variable. We ran poisson regressions and the results are consistent. These results are available upon request.
Some of the effects are sizable. Of those low in ideological distance with Muslims, 45% would not be willing to support a Muslim candidate for President (p = 0.00), while 93% of those high in ideological distance would not (p = 0.00). Ideology seems to be part of the explanation for the bias against Muslims, but even when ideological viewpoints are in alignment there is still significant bias. Interestingly, those low in ideological distance with African-Americans do not have bias against an African-American candidate, while those high in ideological distance do (28% would not support the candidate, p = 0.02). For born-again southerners, we only find that the interaction between ideology and the treatment is significant for African-Americans, and the effects are consistent to those we found for the whole sample. These data suggest that explicit bias against blacks may not be racial, but ideological, at least with respect to asking about an abstract candidacy. That result offers a different interpretation of race that warrants further consideration in future research.
Of more direct significance to this particular paper, we do not find that ideology moderates the effect of the treatment for the Mormon, Atheists, Evangelical, or Female candidate (for whom there was only bias among the southern born-again sample). Such findings are important, because one cannot dismiss the earlier findings for the religious minority candidates on ideological grounds. That is, the bias people exhibit against these groups is more than just ideology at work. We now turn our attention to examining an alternative explanation for the biases uncovered in the list experiment.
2. Social contact theory
It is not surprising that we find bias in our sample toward an Atheist and Muslim for President. For starters, both groups constitute a small segment of the U.S. population, 1.6% and 0.6% respectively, meaning that many, perhaps even most, Americans will not have direct experience with members of these groups. Further, the public perceives these two groups to hold values that are in conflict with the vast majority of Americans.24 After the events of 9/11, many studies documented a link between concerns about terrorism and negative attitudes toward Muslims, as well as a willingness to restrict their civil liberties (e.g., Davis, 2007; Davis and Silver, 2004; Huddy et al., 2005). Kalkan et al. (2009) argue that even before 9/11, Americans viewed Muslims as falling outside of the mainstream with respect to religion and culture. Meanwhile, a long line of literature has demonstrated intolerance against Atheists. In Stouffer's classic study of political tolerance in 1954, only 37% of the public thought an Atheist should be allowed to speak in public, much less run for the highest office in the U.S. More recently, Gibson (2008) found that Atheists are still one of the most disliked groups in the U.S., and radical Muslims are the third most disliked group. There is certainly a lack of Atheists and Muslims in national political office, with only two Muslims
http://religions.pewforum.org/reports.
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620 613
Table 1
OLS Regression on the Effects of the Treatment Condition and Ideological Distance on the Mean Number of Statements the Respondent Agrees with, Fall 2008.
Whole Sample
	African-American	Female	Atheist	Evangelical	Mormon	Muslim
	Coef. (S.E.)	Coef. (S.E.)	Coef. (S.E.)	Coef. (S.E.)	Coef. (S.E.)	Coef. (S.E.)
Treatment	-0.035 (0.093)	0.068 (0.085)	0.579** (0.133)	0.255** (0.092)	0.178* (0.092)	0.449** (0.118)
Ideological Distance	-0.072 (0.093)	0.097 (0.163)	0.146 (0.099)	-0.143 (0.138)	-0.087 (0.132)	0.028 (0.114)
Ideological Dist. * Treatment	0.319** (0.150)	0.012 (0.275)	0.277(0.180)	0.308 (0.225)	0.206 (0.326)	0.483** (0.229)
Constant	1.298** (0.058)	1.296** (0.052)	1.223** (0.074)	1.308** (0.043)	1.301** (0.047)	1.256** (0.064)
N	390	380	323	415	417	309
R-squared	0.018	0.005	0.220	0.060	0.024	0.172
Born-Again Southerners						
	African-American	Female	Atheist	Evangelical	Mormon	Muslim
Treatment	-0.065 (0.111)	0.201* (0.111)	0.751** (0.206)	0.124(0.099)	0.211* (0.107)	0.714** (0.173)
Ideological Distance	-0.317** (0.128)	-0.016 (0.221)	-0.082 (0.136)	0.138 (0.211)	0.123 (0.172)	0.111 (0.151)
Ideological Dist. * Treatment	0.888** (0.219)	-0.174 (0.270)	0.116(0.263)	0.310(0.287)	0.097 (0.473)	0.380 (0.302)
Constant	1.340** (0.076)	1.266(0.061)	1.294** (0.101)	1.238** (0.065)	1.232** (0.075)	1.180** (0.120)
N	160	160	133	164	177	114
R-squared	0.148	0.027	0.312	0.031	0.032	0.326
**p < 0.05 (two-tailed), *p < 0.10 (two-tailed), +p < 0.10 (one-tailed).
in the U.S. Congress, and no open Atheists. Given these factors, it is not a surprise that the public would be largely unwilling to support either type of candidate for President.
However, bias against Mormons is a bit more surprising in that members of the LDS church believe in Jesus Christ and tend to hold positions on social issues that square with groups such as Southern Baptists or many Catholics. Furthermore, there is a fair representation of Mormons in government. In the 111th Congress, about 2.4% of the representatives identified themselves as members of the LDS Church.26 That being said, Mormonism is a relatively new religion that the public does not know much about. For example, in August of 2008, Fox news asked the American public if they think Mormons are Christians or not and 25% of the country did not know and another 24% answered 'no' to the question. To the same questions asked about Catholics in 2007, only 5% said they did not know and just 10% answered 'no' to the question. Part of this lack of familiarity likely stems from the fact that there are not many Mormons in the population, about 1.7%. But the size of the group may not matter as much as geographic concentration-a significant portion of the Mormon populations lives in just a handful of states. States with the highest percentage of LDS church members include Utah, 66%, Idaho, 24.1%, Wyoming, 9.5%, Nevada, 5.9%, and Arizona, 4.9% (Campbell and Manson, 2007).
Because these particular religious groups are both small and vary in where they live, we have a unique opportunity to consider whether social contact with them lowers or
25 However, in a study by the Pew Forum on Religion and Public Life and Congressional Quarterly of the 111th Congress, Representative Peter Stark, from Northern California indicated that he did not believe in a supreme being. Lin, Joanna. 5, January, 2009. "111th Congress Reflects Greater Religious Diversity in the U.S." Los Angeles Times. Retrieved at http://www.latimes.com/news/local/la-me-beliefs5-2009jan05,0,3274449.storyon March 2nd, 2009.
26 Retrieved from the web site for the Pew Forum on Religion and Public Life. http://pewforum.org/docs/?DocID=379.
increases bias. For many groups like Catholics, contact is so widespread that we lose leverage to assess the role of contact. To ground our discussion, we employ two general theories about how contact with out-groups influences attitudes, social contact theory and group threat theory. The two theories offer different theoretical insig hts.27 We hope to reconcile some of the differences by taking advantage of the different electoral contexts offered by our two studies.
According to social contact theory, increased contact with members of out-groups results in less bias against that group. In The Nature of Prejudice (1954), Allport documented the many facets of prejudice that exist toward African-Americans, but argued that one potential silver lining for the future is that increased inter-group contact could improve relations by reducing prejudice and stereotypes between groups. The more contact individuals have with a group, the more they learn about the group  and  form  evaluations  from  direct personal
27 One factor that may account for the general bias across all of these religious minorities, even though they are each quite different, is ethnocentrism. According to Kinder and Kam (2009,8), "Ethnocentrism is a mental habit. It is a predisposition to divide the human world into in-groups and out-groups...Ethnocentrism constitutes a readiness to act in favor of in-groups and in opposition to out-groups." A key feature of ethnocentrism is that individuals who are more ethnocentric are inclined to oppose out-groups generally, even if the degree to which they oppose particular groups varies (Kinder and Kam, 2009; Kam and Kinder u.p., Sniderman and Piazza, 1993). Since all of these groups make up a small proportion of the population, they serve as out-groups for the vast majority of Americans. This is an important lens to view these general questions. Even so, we have chosen to focus on social contact because of the unique geography of Mormons. To be comprehensive, we did undertake tests concerning ethnocentrism. Those results, which are available upon request, paint a mixed picture. For the whole sample and among born-again southerners, we find that only those high in ethnocentrism are unwilling to support a Mormon candidate for president. However, if we examine Atheist and Muslim candidates, evangelicals does not work as well. Respondents, whether ethnocentric or not, are not inclined to support either type of candidate for President.
614
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
experiences, which tend to be more positive on average (Sigelman and Welch, 1993). The less contact individuals have with a group, the more likely they form impressions from stereotypes present in the media or their social networks. According to Allport (1954), inter-group contact would only lead to a reduction in prejudice under conditions in which contact is among equals, where the groups share common goals, where the situation is cooperative rather than competitive, and where it is supported by authorities, law, or custom.
A long line of scholarship has demonstrated empirical support for the positive effects of inter-group contact in diminishing prejudice against racial and ethnic minorities (e.g., Johnson and Jacobson 2005; McClain et al., 2006; Oliver and Wong, 2003; Pettigrew, 1997; Sigelman and Welch, 1993; Yancey, 1999). The positive effects of social contact also extend to other target groups, such as gays and lesbians, the physically disabled, the mentally disabled and ill, and the elderly (for a metaanalysis comparing these effects, see Pettigrew and Tropp, 2006). Some scholars have also applied the theory with respect to religious groups. For example, Islam and Hewstone (1993) find that increased contact, contact that is qualitatively better, and lower inter-group anxiety improve attitudes between Hindus and Muslims in Pakistan. With respect to another set of groups, Paolini et al. (2004) show that direct and cross friendships in Northern Ireland reduce prejudice between Protestants and Catholics.28 Finally, Kalkan et al. (2009), find that knowing someone who is Muslim and having personal experiences with a Muslim increases favorable attitudes toward the group.
While a long line of literature has explored the four conditions that Allport outlined, in a meta-analysis, Pettigrew and Tropp (2006) find that social contact with out-groups can reduce prejudice even when the four conditions are not met. The positive effects of social contact appear to be most pronounced as the quality and quantity of contact increase (Voci and Hewstone, 2003; Aberson and Haag, 2007; Pettigrew, 1998), though studies even find positive effects when measures of geographical presence (Oliver and Wong, 2003) or estimates of the presence of different groups in the community (Yancey, 1999) are used. The findings for social contact theory are not however always rosy. For example, while positive and even neutral contact reduces prejudice, scholars argue that when social interactions are negative it will not likely lead to such a reduction (e.g., Paolini et al., 2010; Stephan et al., 2002).29
28 This is but one example from an extensive research program on the topic. For a summary, see Hewstone et al. 2008.
29 These studies focus on slightly different mechanisms. For example, Stephan et al., 2002 show how negative contact has a strong effect on negative racial attitudes, which is mediated through increased inter-group anxiety. Paolini et al. (2010) show how negative contact increases the salience of one's own group membership. More generally, much of the work in social psychology focuses on various mechanisms by which inter-group contact reduces prejudice. Since we will not be able to test various mechanisms, we leave this discussion out in the interest of space.
While scholars have not used social contact theory to understand support for presidential candidates with different religious characteristics, the same logic may hold. Since most Americans have not had extensive contact with Atheists, Mormons, and Muslims, that may fuel some of the bias against them as candidates. But some Americans do interact with these groups and we can, therefore, test to see if social contact with a religious minority lessens bias against possible presidential candidates from that group.
A different set of expectations arises if we turn to group threat theory, which was developed to explain attitudes of whites toward African-Americans (Key, 1949; Blumer, 1958; Blalock 1967). According to this theory, a dominant group becomes more hostile to an out-group as the size of the out-group increases, since the group is perceived to pose a threat to the dominant group's economic resources and social privilege. Key's seminal work showed that white resistance to integration and support for racially conservative candidates was strongest in areas of the South with a higher presence of African-Americans. Other work has shown that as the presence of African-Americans increases, whites are more inclined to vote for racially conservative candidates (e.g., Blalock 1967; Giles and Buckner, 1993; Huckfeld and Kohfeld 1989; Wright, 1977), are less supportive of black political figures (Glaser, 1994), and are more likely to identify with the Republican party (Giles and Hertz, 1994).30
Few studies have applied group threat theory to religious groups. One important and notable exception is Campbell (2006). Campbell argues that religious conflict is an enduring feature of U.S. politics, and if there is any group for whom religious threat is relevant, it is evangelical Christians. According to Campbell, evangelicals have awareness as a socially cohesive and politically relevant group who have a tension with secular society. He draws from group conflict theory to argue that as the geographical presence of atheists increases, evangelicals should be more inclined to vote for Republican candidates. The 1996 and 2000 presidential elections provide an opportunity to test this hypothesis, with the salience of religion being more prominent in the latter contest. Campbell's data nicely confirm this idea for 2000, while the results are more mixed for 1996.
Is there a way to reconcile the two theories in the political arena with respect to the religious characteristics of presidential contenders? Perhaps the predictions of social contact theory work, in general, but there may be exceptions for some religious groups, for certain segments of the population, and for particular contexts. Since the salience of religious conflict waxes and wanes in the political domain, as Campbell shows, social contact may reduce bias when religious conflict is not salient while it may not diminish (and even exacerbate) bias when
30 Other studies find the opposite effects with respect to prejudice (e.g., Sigelman and Welch, 1993; Oliver and Wong, 2003) and support for black candidates (Carsey, 1995). As we mentioned earlier, even some studies using the theory of social contact but with only geographical proximity type measures to proxy social interaction still find that prejudice decreases as members of an out group increase in an area.
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
615
religious conflict is salient. That is, group threat theory may be more applicable for some groups in contexts in which religious conflict is salient in politics.
For which groups might we witness bias, even given increased contact, and among whom? We contend that while the positive effects of social contact should hold among the general population, they may not hold among southern evangelicals, since religious threat is particularly relevant for this group. More specifically, social contact should not diminish bias among southern evangelicals when the religious conflict from a given group is salient in politics.
Turning first to Atheists, evangelicals perceive a high degree of threat from secularists (Campbell, 2006), and this tension in the political sphere has given rise to the now common reference of a culture war (e.g., Hunter 1991; Campbell, 2006; Fiorina et al., 2004; Layman, 2001). Campbell only explored the voting behavior of evangelicals given the increased presence of secularists in a general election context. It is possible that the salience of that religious threat waxes and wanes during different electoral conditions. For example, it could be that the threat posed by Atheists is not particularly salient during primary elections, since the battle is within the Republican primary, but becomes salient during the general election, as the Republican candidate faces a Democrat. If this is true, then we should find that increased social contact diminishes bias in the former, but not the latter situation. Given the timing of our two studies, we can shed some light on this contextual explanation.
Recall that Fig. 2 showed that born again southerners are less willing to support a Mormon candidate for President than the general population, even though both religious groups tend to be conservative Republicans. One might expect that Mormons and Evangelicals would band together in politics given that they have similar preferences on moral values (Campbell and Monson, 2007). A recent example of such coordination between members of the two groups was in the campaign for Proposition 8 in California, which changed the constitution to no longer allow gay marriage. However, the salience of religious conflict between the two groups may arise in certain contexts, such as a primary election, especially if members from the two groups are running for office. During the 2008 presidential election, the two groups would have been in a state of political competition during the primary setting since Mitt Romney, a Mormon, was competing directly against Mike Huckabee, a former Evangelical preacher, for the Republican presidential nomination. In this case, the religious conflict between the two may have been particularly salient and social contact may not reduce bias. By the general election, the two groups were no longer in competition, but had incentives to band together against a Democrat, which should yield positive effects for social contact.
We decided to look at bias against Muslims only after we ran our 2007 study, so we do not have two different electoral contexts with which to explore bias against a Muslim candidate. We can only specify how we expect
social contact to work in the context of the general election. Recent work shows that evangelicals and religious traditionalists perceive a high degree of threat from Muslims, which can result in more negative attitudes toward the group. For example, Kalkan et al. (2009) show that born-again Christians and southerners have less favorable views of Muslims and Muslim Americans.31 The salience of the threat posed by Muslims during general elections has been high since 9/11, especially as Republican candidates have sought to keep the issue of terrorism on the table (Merolla and Zechmeister, 2009). Anecdotal evidence from the 2010 general elections speaks to the salience of this threat to evangelicals. A proposition to ban Sharia law passed with overwhelming support (70%) in Oklahoma,32 a state with one of the highest proportions of evangelicals (53%) and a very small percentage of Muslims.33 Given the salience of the threat posed by Muslims to born again individuals during general elections, we do not expect social contact to diminish bias toward a Muslim candidate among this group.
We now summarize our expectations for the effects of social contact among the general sample and the southern born-again sub-sample.
HI: Among the whole sample, increased contact with members of a religious minority should decrease bias against a presidential candidate from that group.
Our specific expectations for the reactions of born again southerners to each religious minority candidate are as follows:
H2: Social contact will diminish bias toward an Atheist candidate in the primary election, while it will not diminish bias toward an Atheist candidate in the general election.
H3: Social contact will not diminish bias toward a Mormon candidate in the primary election, while it will diminish bias toward a Mormon candidate in the general election.
H4: Social contact will not diminish bias toward a Muslim candidate in the general election.
3. Social contact findings
In both studies, we included a question of how many Atheists, Evangelicals, and Mormons an individual personally knows, and added Muslims to the fall 2008 study. The response options were as follows: do not know any, know a few, know quite a few, and almost everyone I know. Social Contact theory suggests that bias should be reduced as individuals are more familiar with a given group, though individuals may need to know more than a handful of individuals from that group for positive effects to obtain (Pettigrew, 1997; Sigelman and Welch, 1993; Yancey, 1999). We therefore combine those in the first two categories and the last two categories to create
31 The most important predictor of negative attitudes toward Muslims was negative affect toward other religious out-groups.
32 http://elections.nytimes.com/2010/results/oklahoma
33 See http://religions.pewforum.org/maps.
616
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
a dichotomous measure of knowing Atheists, knowing Mormons, and knowing Muslims.34 The distribution of the measures averaged across the two studies is presented in Fig. 3 for the whole sample and for born-again southerners. These descriptive statistics reinforce our argument earlier that social contact with Atheists, Mormons, and Muslims is relatively low. For the whole sample, just over 9% indicate having high contact with Mormons, while only about 10% have high contact with Atheists. The percentages are slightly lower for born-again southerners where about 8% have high contact with Mormons and 6% report high contact with Atheists. For both the whole sample and the southern born-again sample, high contact is even lower with Muslims, with only about 4% saying they know many Muslims.
In order to see if social contact diminishes bias in the whole sample (HI), we ran a regression on the number of statements the respondent agrees with in the list experiment separately for each candidate type (with the control group serving as the baseline). As independent variables, we included a dummy variable for whether the respondent is in the given treatment condition, has high contact with the given group, and an interaction between the two.35 Given the presence of interaction terms, we separately calculate the coefficient and standard errors of the treatment variable for those low and high in social contact and illustrate these effects in Fig. 4 (for full regression results see Appendix Table A).36 As noted earlier, the coefficients on the treatment variables give us the proportion not willing to support each candidate type since it represents the mean difference between the control and the given treatment group.37
Turning first to those who do and do not know many Mormons, we only find bias against a Mormon running for President among those who know a few or no Mormons in both of our studies. Among this group, the percentage of subjects in the treatment group who say that they would not vote for a Mormon for President is 38% in the fall of 2007 and 23% in the fall of 2008, and
34 We dichotomize this measure since we will be interacting it with our treatment conditions. This gives us a big enough n in each experimental condition to analyze the effects of the treatment for those who have low and high contact with each group. We would not be able to do this if we looked at the effect of the treatment for all four categories since the sample becomes very small, particularly among those who say "almost everyone I know."
35 We also ran models with a control for race since subjects were not evenly distributed across experimental conditions on this measure for the 2007 study and the results are the same.
36 The results in Fig. 4 are consistent if we run Poisson regression models. For the analysis of the atheist treatment relative to the control in
2007, the p-value on the interaction term of atheists and high social contact is higher in the Poisson model than the OLS model. However, when we calculate the standard error of the treatment for those high in contact with atheists, it is insignificant in the OLS and Poisson regressions, so substantively the results are the same. For the same analysis in
2008, the p-value on the effect of the treatment for those high in contact with atheists is 0.13 in the OLS model and 0.08 in the Poisson model. These are very close to one another. All of these results are available upon request.
37 The results are identical if we use difference of means tests between the control and each treatment condition among those low and high in social contact. These results are available upon request.
both effects are statistically significant (p = 0.00 and p = 0.01, respectively). Meanwhile, those who know quite a few or many Mormons do not exhibit any bias. The treatment condition is not statistically significant for this group in either study.38 If we turn to the results for the Atheist candidate, in 2007, those low in social contact exhibit bias (p = 0.00), while those high in social contact do not. However, social contact does not diminish bias against an Atheist candidate in the 2008 study. Social contact also appears to reduce bias against the Muslim candidate. Individuals high in social contact do not register any bias, while 60.6% of those low in social contact would not support a qualified Muslim (p = 0.00). In four out of five cases, we find that increased social contact with members of religious minorities diminishes bias against a presidential candidate from the given group.39 This lends strong support to our first hypothesis.
We do not expect social contact to diminish bias in all cases. We argued that social contact will not diminish bias among born-again southerners in contexts when the religious threat posed by a given group is politically salient. To test our hypotheses, we run the same analysis as we did for the whole sample, but this time among born again southerners. In Fig. 5, we present the coefficient on the treatment variables by levels of social contact during the primary and general election studies (for full results see Appendix Table B).40 Turning first to the Atheist candidate and the nomination process, we do not observe bias against an Atheist candidate among born again southerners who know quite a few or many atheists, while the treatment variable is significant among those low in social contact (p = 0.00), and 85% of those low in contact would not support an Atheist candidate. If we turn to the general election context when the religious threat posed by Atheists is salient, social contact
38 Again, there are a few cases in which the coefficient is negative when it should not theoretically be negative; however, none of these are statistically significant.
39 The effects for contact could be inflated if the person is a member of the target group. The religious denomination question did not have a response option for being Mormon, but did have an option for other Christian or other religion. If we control for these variables in the Mormon v. control analyses, the results are consistent and neither control is statistically significant. If we control for those who do not identify with a religion in the Atheist v. control analysis, the control variable is significant. The results on our key measures are consistent, though the effect of the treatment among those high in contact becomes insignificant in 2008 once we control for those who do not identify with a religion (p = 0.148). Controlling for this variable then would mean that all five cases are consistent with our hypothesis. There were not enough Muslims in the sample to include members of the group as a control in the Muslim v. control analysis. If we use a measure of geographical proximity instead of social contact, we find that areas with a higher presence of Mormons and Atheists do not exhibit any bias. There was not enough variation in the presence of Muslims to run a similar analysis. All of these results are available upon request.
40 In Poisson models, the results are consistent for all of the results in Fig. 5. The only minor deviation is in the fall 2007 data in the analysis of the atheist condition relative to the control group. The p-value on the interaction of the atheist treatment and high contact is 0.12 in the Poisson model and 0.02 in the OLS model. However, when we calculate the effect of the treatment for those high in contact it is not significant in either model, so again the results are substantively the same. These results are available upon request.
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620 617 100 t-
8. -o-1
-0.2
Low Contact	High Contact	Low Contact	High Contact
2007		2008	
	Contact and Year		
Fig. 4. Coefficient on Treatment Condition by Social Contact, whole sample 2007 and 2008.
618
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
1.6
I Atheist
I Mormon
I Muslim
Low Contact       High Contact       Low Contact       High Contact
2007
2008
Contact and Year
Fig. 5. Coefficient on Treatment Condition by Social Contact, Born Again Southerners 2007 and 2008.
does not diminish bias among those high in contact with Atheists.41 In fact, the proportion not willing to support an Atheist for president is even higher among those high in social contact. These results support our second hypothesis.
We expected the opposite pattern to hold for the Mormon candidate since the salience of religious threat from Mormons was present in the primary but not the general election (H3). In the context of the primary, we find that social contact is not only ineffective in reducing bias it actually exacerbates bias, as it did for the Atheist candidate in the general election. While 37% of those low in contact with Mormons would not support a qualified Mormon (p = 0.02), 65% of those high in contact would not (p = 0.01 ).42 If we look at the results for the general election we find, as expected, that high social contact now washes away the bias against a Mormon candidate. While 34% of born-again southerners low in social
41 If we use a measure of geographical proximity instead of social contact, similar results obtain. We could not do a similar analysis for the Mormon treatment verses the control group because of a low n for southern evangelicals living in states with a high presence of Mormons. These results are available upon request.
42 If we look at all born again individuals, not just those in the South, this result does not hold. Those high in social contact do not exhibit any significant bias, while 47% of those low in social contact would not support a Mormon candidate (p = 0.00). This makes sense in that the locus of political competition between Romney and Huckabee would be more pronounced among southerners.
contact with Mormons would not support a Mormon for President (p = 0.00), there is no bias among those high in contact.
We only had one electoral context to explore for a Muslim candidate, the general election. Our expectation was that social contact would not diminish bias against the Muslim candidate among born again southerners since the religious threat posed by Muslims was salient during the general election (H4). The results are generally supportive of this expectation. About 90% of those low in contact with Muslims would not support a Muslim candidate for president, while 62% of those high in contact would not. While social contact appears to slightly diminish bias, it is still quite high among those high in social contact.
4. Conclusions
Barack Obama's victory in 2008 symbolizes in dramatic fashion the decline of bias in the United States. But we should not overstate the amount of that decline. This paper shows that bias has a strong religious dimension, which is consistent with the growing literature about what others have called "culture wars".43 In other words, it appears that America has taken two steps forward, one step back.
James Davison Hunter's Culture Wars inspired this term.
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
These empirical findings are important in and of themselves. But this paper has also given us a new lens with which to view social contact theory and group threat theory. To understand how contact may lessen or increase bias, we need to understand the electoral context and its relevance for certain groups. In exploring bias against religious minorities, we argued that the positive effects of social contact may not hold for southern evangelicals when the religious threat posed by the group is politically salient. This may even occur among groups who otherwise share similar political goals, such as evangelicals and Mormons. In this sense, the theory provides a much richer understanding of the dynamics of inter-group contact than existing theories. While our results are in line with this theoretical argument, we recognize that we do not have a direct test of how the political salience of religious threat conditions the effects of social contact. That is, we did not directly manipulate the political salience of religious threat, relying instead on the electoral climate that was taking place at the time of our studies. As a result, we offer our findings with appropriate caution.
Our findings also have implications for current and future presidential politics. Candidates from religious minorities, such as Mitt Romney, cannot take much comfort in our results. As Romney seeks the Republican presidential nomination in 2012, he will likely be thrust into a competitive context with an evangelical candidate. Such a context will likely spike the bias many groups, such as evangelicals, hold toward Mormons, which in turn will undermine Romney's, or any Mormon's, chances of success. It does not mean that Mormon candidates cannot overcome this hurdle. It just indicates a higher jump may be needed. But if Romney (or perhaps former Utah Governor Jon Huntsman) can overcome this hurdle, as did John Kennedy, bias will likely decline, as the fears many have about Mormons (or any minority group) would fade as the public learns more about them. Muslims clearly face a much tougher and steeper climb, as do Atheists. It is hard to imagine the conditions that would make such candidacies credible any time soon. Nonetheless, if Oba-ma's election has taught us anything, it tells us that progress can be made and sometimes at surprising speed. The path is not simple or easy. But the overall trend is positive and that offers some reason for optimism, at least in the long run.
Acknowledgements
We want to thank Cindy Kam, Geoff Layman, Dave Campbell, Dave Lewis, Efren Perez, Jim Gibson, Lynn Vavreck and Jeff Lewis for helpful comments on this paper and the design of our survey. The reviewers for Electoral Studies also offered insightful suggestions for which we are grateful. Geoff Evans has offered thoughtful advice as well. We also want to thank Sam Luks of Poli-metrix for her kind assistance with the data collection. Finally, we express our appreciation to Simon Jackman and Lynn Vavreck for spearheading the collections of the 2008 data under the Cooperative Campaign Analysis Project.
619
Appendix.
Table A
OLS Regression on the Effects of the Treatment Condition and Social Contact on the Mean Number of Statements the Respondent Agrees with, Whole Sample.
2007 Study
Atheist Mormon
Candidate Candidate
Coef. (S.E.) Coef.(S.E.) Treatment 0.504** (0.117)   0.377** (0.117)
High Contact      -0.595** (0.205) 0.279 (0.244) Treatment -0.511+(0.331) -0.514+(0.369)
*High Contact Constant 1.355** (0.070)   1.250** (0.070)
N 611 609
R-squared 0.102 0.027
2008 Study
Atheist Mormon Muslim
Candidate Candidate Candidate
Coef. (S.E.) Coef. (S.E.)        Coef. (S.E.)
Treatment 0.648** (0.088) 0.234** (0.096)   0.606** (0.091)
High Contact      -0.182+(0.139) 0.008(0.117)     0.428+(0.275)
Treatment -0.102 (0.372) -0.307+ (0.207) -0.556* (0.335)
*High Contact
Constant 1.309** (0.047) 1.284** (0.047)   1.268** (0.043)
N 485 528 509
R-squared 0.160 0.023 0.120
**p < 0.05 (two-tailed), *p < 0.10 (two-tailed), +p < 0.10 (one-tailed). Table B
OLS Regression on the Effects of the Treatment Condition and Social Contact on the Mean Number of Statements the Respondent Agrees with, Born-Again Southerners.
2007 Study
Atheist Mormon
Candidate Candidate
Coef. (S.E.) Coef. (S.E.)
Treatment 0.849** (0.141) 0.366** (0.154)
High Contact       -0.234 (0.257) -0.034 (0.212) Treatment         -1.192** (0.517) 0.284(0.296)
*High Contact
Constant 1.542** (0.081) 1.537** (0.082)
N 240 240
R-squared 0.195 0.195
2008 Study
Atheist Mormon Muslim
Candidate Candidate Candidate
Coef. (S.E.) Coef. (S.E.) Coef. (S.E.)
Treatment 0.771** (0.118) 0.336** (0.108) 0.892** (0.113)
High Contact       -0.241** (0.063) -0.098 (0.143) -0.134(0.125)
Treatment 0.602+(0.427) -0.264(0.217) -0.274(0.299)
*High Contact
Constant 1.241** (0.063) 1.233** (0.063) 1.235** (0.063)
N 195 216 187
R-squared 0.261 0.061 0.323
**p < 0.05 (two-tailed), *p < 0.10 (two-tailed), +p < 0.10 (one-tailed).
References
Aberson, C.L., Haag, S.C., 2007. Contact, perspective taking, and anxiety as predictors of stereotype endorsement, explicit attitudes and implicit attitudes. Group Processes & Intergroup Relations 10,179-201.
Allport, G., 1954. The Nature of Prejudice. Addison-Wesley, Cambridge.
620
B.V. Benson et al. / Electoral Studies 30 (2011) 607-620
Blalock, H., 1967. Toward a Theory of Minority-Group Relations. Wiley, New York.
Blumer, H., 1958. Race prejudice as a sense of group position. Pacific Sociological Association 1, 3-7.
Campbell, D.E., 2006. Religious 'threat' in contemporary presidential elections. Journal of Politics 68 (1), 104-115.
Campbell, D.E., Monson, J.Q., 2007. Dry kindling: a political profile of American mormons. In: Wilson, J.M. (Ed.), From Pews to Polling Places: Faith and Politics in the American Religious Mosaic. Georgetown University Press, Washington, D.C.
Campbell, D.E., Monson, J.Q., 2008. The religion card: gay marriage and the 2004 election. Public Opinion Quarterly 72 (3), 399-419.
Carsey, T, 1995. The contextual effects of race on white voter behavior: the 1989 New York mayoral election. Journal of Politics 57, 221 -228.
Citrin, J., Sears, D.O., Green, D.P., 1990. White reactions to black candidates: when does race matter? Public Opinion Quarterly 54, 74-96.
Davis, D.W., 2007. Negative Liberty: Public Opinion and the Terrorist Attacks on America. Russell Sage Foundation, New York.
Davis, D.W., Silver, B.D., 2004. Civil liberties vs. Security: public opinion in the context of the terrorist attacks on America. American Journal of Political Science 48 (1), 28-46.
Ekstrand, L., Eckert, W., 1981. The Impact of candidate sex on voter choice. The Western Political Science Quarterly 34 (1), 78-87.
Fazio, R.H., Olson, M.A., 2003. Implicit measures in social cognition research: their meaning and use. Annual Review of Psychology 54,297-327.
Fiorina, M., Abrams, S., Pope, J., 2004. Culture War? the Myth of a Polarized America. Longman, New York.
Gibson, J.L., 2008. Intolerance and political repression in the United states: a half century after McCarthyism. American Journal of Political Science 52 (1), 96-108.
Gilens, M., Sniderman, P.M., Kulinski, J.H., 1998. Affirmative action and the politics of realignment. British Journal of Political Science 28,159-183.
Giles, M.W., Buckner, M.A., 1993. David Duke and black threat: an old hypothesis revisited. Journal of Politics 55, 702-713.
Giles, M.W., Hertz, K., 1994. Racial threat and partisan identification. American Political Science Review 88, 317-326.
Glaser, J.M., 1994. Back to the black belt: racial environment and white racial attitudes in the south. Journal of Politics 56, 21-41.
Greenwald, A.G., Tucker Smith, C, Sriram, N., Bar-Anan, Y, Nosek, B.A., 2009. Implicit race attitudes predicted vote in 2008 U.S. Presidential election. Analyses of Social Issues and Public Policy 9 (1), 241-253.
Hewstone, M., Kenworthy, J.B., Cairns, Ed, Tausch, N., Hughes, Joanne, Tarn, T, Voci, A., von Hecker, U, Pinder, C, 2008. Stepping stones to reconciliation in Northern Ireland: intergroup contact, forgiveness, and trust. In: Arie, N., Malloy, T.E., Fisher, J.D. (Eds.), The Social Psychology of Intergroup Reconciliation. Oxford University Press, New York.
Huckfeld, R., Kohfeld, C.W., 1989. Race and the Decline of Class in American Politics. University of Illinois Press, Urbana, IL.
Huddy, L., Terkildsen, N., 1993. The Consequences of gender stereotypes for women candidates at different levels and types of office. Political Research Quarterly 46 (3), 503-525.
Huddy, L., Feldman, S., Taber, C, Lahav, G., 2005. Threat, anxiety, and support of Antiterrorism Policies. American Journal of Political Science 49 (3), 593-608.
Hunter, J.D., 1991. Culture Wars: The Struggle to Define America. Basic Books, NY.
Islam, M.R., Hewstone, M., 1993. Dimensions of contact as predictors of
intergroup anxiety, perceived out-group Variability, and out-group
attitude: an Integrative model. Personality and Social Psychology
Bulletin 19, 700-710. Johnson, B.R., Jacobson, C.K., 2005. Contact in context: an examination of
social settings on whites' attitudes toward Interracial marriage. Social
Psychology Quarterly 68 (4), 387-399. Kalkan, K.O., Layman, G.C., Uslaner, E.M., 2009. Band of others'? Attitudes
towards Muslims in Contemporary American Society. Journal of
Politics 71, 847-862. Kam, CD., 2007. Implicit attitudes, explicit choices: when subliminal
priming predicts candidate preference. Journal of Politics 29,
343-367.
Kam, CD., Franzese Jr., R.J., 2007. Modeling and Interpreting Interaction Hypotheses in Regression Analysis. University of Michigan, Ann Arbor, MI.
Kam CD. and D. Kinder, u.p. That one: ethnocentrism as a Short-term Force in the 2008 American presidential election.
Kane, J.G., Craig, S.C, Wald, K.D., 2004. Religion and presidential politics in Florida: a list experiment. Social Science Quarterly 85, 281-293.
Key, V.O., 1949. Southern Politics in State and Nation. University of Tennessee Press, Knoxville.
Kinder, D., Kam, CD., 2009. Us against Them: Ethnocentric Foundations of American Opinion. University of Chicago Press, Chicago.
Kuklinski, J.H., Cobb, M.D., Gilens, M., 1997a. Racial attitudes and the New South. Journal of Politics 59, 323-349.
Kuklinski, J.H., Sniderman, P.M., Knight, K., Piazza, T, Tetlock, P.E., Lawrence, C, Mellers, B., 1997b. Racial prejudice and attitudes toward Affirmative Action. American Journal of Political Science 41,402-419.
Layman, C, 1997. Religion and political behavior in the United states. The Public Opinion Quarterly 61, 288-316.
Layman, C, 2001. The Great Divide: Religious and Cultural Conflict in American Party Politics. Columbia University Press, New York.
McClain, P., et al., 2006. Racial distancing in a southern City: Latino Immigrants'views of black Americans. Journal of Politics 68, 571-584.
Merolla, J.L., Zechmeister, E.J., 2009. Democracy at Risk: How Terrorist Threats Affect the Public. University of Chicago Press, Chicago.
Oliver, E.J., Wong, J., 2003. Intergroup prejudice in Multiethnic settings. American Journal of Political Science 47 (4), 567-582.
Paolini, S., Hewstone, M., Cairns, Ed, Voci, A., 2004. Effects of direct and Indirect cross-group friendships on Judgments of Catholics and protestants in Northern Ireland: the mediating role of an anxiety-reduction Mechanism. Personality and Social Psychology Bulletin 30, 770-786.
Paolini, S., Harwood, J., Rubin, M., 2010. Negative intergroup contact makes group Memberships salient: Explaining why intergroup conflict Endures. Personality and Social Psychology Bulletin 36, 1723-1738.
Pasek, J., Tahk, A., Lelkes, Y, Krosnick, JA., Keith Payne, B., Akhtar, O., Tompson, T, 2009. Determinants of Turnout and candidate choice in the 2008 U.S. Presidential election: Illuminating the Impact of racial prejudice and other considerations. Public Opinion Quarterly 73,943-994.
Payne, K.B., Krosnick, J.A., Pasek, J., Lelkes, Y, Akhtar, O., Tompson, Trevor, 2010. Implicit and explicit prejudice in the 2008 American presidential election. Journal of Experimental Social Psychology 46,367-374.
Pettigrew, T.F., 1997. Generalized intergroup contact effects on prejudice. Personality and Social Psychology Bulletin 23,173-185.
Pettigrew, T.F., 1998. Intergroup contact theory. American Review Psychology 49, 65-85.
Pettigrew, T.F., Tropp, L.R., 2006. A meta-Analytic test of intergroup contact theory. Journal of Personality and Social Psychology 90, 751-783.
Sanbonmatsu, K., 2002. Gender stereotypes and vote choice. American Journal of Political Science 46 (1), 20-34.
Sigelman, Lee, Welch, S., 1993. The contact hypothesis Revisited: black-white interaction and positive racial attitudes. Social Forces 71, 781-795.
Sniderman, P.M., Carmines, E.G., 1997. Reaching beyond Race. Harvard University Press, Cambridge, Massachusetts.
Sniderman, P.M., Piazza, T, 1993. The Scar of Race. Harvard University Press, Cambridge, MA.
Stephan, W.G., Boniecki, K.A., Ybarra, O., Bettencourt, A., Ervin, K.S., Jackson, L.A., McNatt, P.S., Lausanne Renfro, C, 2002. The role of threats in the racial attitudes of blacks and whites. Personality and Social Psychology Bulletin 28,1242-1254.
Streb, M.J., Burrell, B., Frederick, B., Genovese, M.A., 2008. Social desirability effects and support for a female president. Public Opinion Quarterly 72 (1), 76-89.
Terkildsen, N., 1993. When white voters evaluate black candidates: the processing implications of candidate Skin Color, prejudice and self-Monitoring. American Journal of Political Science 37 (4), 1032-1053.
Voci, A., Hewstone, M., 2003. Intergroup contact and prejudice towards Immigrants in Italy: the Mediational role of anxiety and the Mod-erational role of group salience. Group Processes and Intergroup Relations 6, 37-54.
Wright, G.C, 1977. Contextual models of electoral behavior: the southern Wallace vote. American Political Science Review 71, 497-508.
Yancey, C, 1999. An examination of the effects of Residential and church integration on racial attitudes of whites. Sociological Perspectives 42, 279-304.