Experiments Manipulating Candidates’ Religious Affiliations, Behavior, and Rhetoric

One of the oldest and largest branches of experimentation in religion and politics involves manipulating voter information about or perceptions of candidates’ religious characteristics.² These scholars have drawn on social psychological theories elucidating how individuals perceive and evaluate each other. The dependent variables involved in candidate studies have been diverse. Most often, the intention has been to understand electoral behavior. Sometimes the goal is simply to gauge prejudice against candidates from particular religious out-groups, especially Muslims, Mormons, Jews, and atheists. Other work focuses on how candidate information shapes voters’ inferences about those candidates—including their ideology and party, competence and intelligence, as well as trustworthiness, patriotism, or religious affiliation itself.

The conceptually simplest experiments manipulating candidate religion aim simply to measure voter preferences for candidates from some religious groups over others. These experiments maintain a relatively static view of religion as a label ascribed to elites. Beginning in the early 2000s, and continuing through the present, scholars have been asking “what if?”: what would happen if an atheist ran for president? a Jew? These questions have in part been driven by the increasing religious diversity of the candidate pool in the United States. For instance, in the wake of Mitt Romney’s 2008 and 2012 presidential campaigns, scholars explored the extent to which American voters might discriminate against Mormon candidates (Calfano, Friesen, & Djupe, 2013; Campbell, Green, & Monson, 2012); Joe Lieberman’s historic vice presidential nomination led Kane, Craig, and Wald (2004) to investigate discrimination against Jewish candidates.

List experimental studies have asked, would voters discriminate against “Religious Outgroup X” if they thought no one was watching (as is the case in the privacy of the polling booth)? Respondents are randomly assigned to receive one of two lists of items and asked to report the number of items with which they agree. For instance, interviewers in Kane et al.’s (2004) study of Americans’ discrimination against Jewish candidates read control group respondents the following prompt:

Now I’m going to read you FOUR things that sometimes make people angry or upset. After I read all four statements, just tell me how many of them upset you. I don’t want to know which ones, just how many:

Interviewers gave respondents in the treatment group not four but five items; the fifth item read, “Five: a Jewish candidate running for vice president” (p. 284).³

In list experiments, the difference in the number of items reported in the two groups is the key finding of the research; for instance, if respondents in the control condition report on average 2.0 items and respondents in the treatment report on average 2.4 items, researchers deduce that 40 percent of respondents in the treatment chose the experimental item. Researchers can further analyze the differences between treatment and control in varying subpopulations—for instance, testing if levels of discrimination are higher among evangelical Christians. Hence, list experiments enable researchers to estimate population and subpopulation average levels of discrimination, but they do not allow for finer-grained, multivariate analysis of the factors associated with greater or lesser levels of discrimination. Using their list experiment, Kane et al. (2004) found minimal levels of discrimination against Jewish candidates. Nearly a decade later, however, Benson, Merolla, and Geer (2011) found substantial discrimination against atheist and Muslim candidates in a list experimental study.

Most other candidate-related experiments have involved creating fake candidate profiles, holding all aspects of the candidate constant in treatment and control groups, except for key traits that are varied. To explore voter discrimination, researchers gauge how different candidate religious affiliations affect respondents’ reported support for and self-reported likelihood of voting for those candidates. For instance, using such methods, Calfano, Friesen, et al. (2013) found that U.S. voters discriminate against Mormon candidates under some circumstances, which can be mitigated by religious value priming. Outside the United States, Smith (2019) showed that Brazilian evangelical voters are more likely to vote for fake candidates described as evangelical, while Campbell and Cowley (2014) found that giving British parliamentary candidates stereotypically Muslim and Jewish names slightly reduced voter support for those candidates.

Moving beyond religious affiliation, scholars have manipulated other religious characteristics of fake candidates. Boas (2014), for example, explored voter evaluation of evangelical pastor candidates in Brazil (see also Boas, 2016). Meanwhile, many studies manipulate a candidate’s level of religiosity, often without explicitly specifying the candidate’s religious affiliation; in these cases, it seems safe to assume, as Sumaktoyo, Ottati, and Untoro (2016) argue, that voters typically impute the local majority religion to the candidates. Similarly, reminding voters about the candidate’s religious affiliation and the historical voting patterns of the religious group help to reinforce the voting coalition (Boas, 2015).

Benstead, Jamal, and Lust (2015) used photos as the treatment, manipulating the religious/secular dress of fake candidates in Tunisia. More typically, scholars have varied candidates’ supposed levels of religiosity using textual cues (Castle, Layman, Campbell, & Green, 2017; Clifford & Gaskins 2016; McLaughlin & Wise 2014; Smith, 2019; Sumaktoyo et al., 2016).⁴ McLaughlin and Wise, for instance, found that telling American respondents that the religious candidate had decided to run for office after “prayerful consideration” alienated many voters yet attracted religious voters’ support. Meanwhile, also studying the United States, Castle and coauthors manipulated both the fake candidate’s ascribed level of church engagement and the religious content of his language; more religious candidates were found to attract Republican voters, while more secular candidates attracted Democrats.

In many of the studies discussed until this point, the goal has been to understand approval or support for candidates with different religious characteristics. However, an important group of studies has gone further, exploring how cues regarding candidates’ religious traits can serve as heuristics that voters use to draw inferences about a candidate’s ideology and partisanship (Campbell, Green, & Layman, 2011; McDermott, 2007, 2009; McLaughlin & Thompson, 2016; McLaughlin & Wise, 2014), trustworthiness (Clifford & Gaskins, 2016), and patriotism (Braman & Sinno, 2009).⁵ One of the earliest studies took as inspiration the erroneous perception that Ted Kennedy was pro-life in 1980 when he was running for president (Granberg, 1985). To understand the cross-sectional results, Granberg primed respondents by asking them for the religious affiliation of the candidates, and those choosing Catholic were more likely to attribute a pro-life stance to Kennedy. In one innovative study, Berinsky and Mendelberg (2005) showed that priming “discredited stereotypes” that respondents rejected (“Jews are shady”) could cue them to rely more strongly on socially acceptable stereotypes (“Jews are liberal”) in evaluating candidate positions.

A related body of work drills into how candidates use religious language or “God talk” to shape the ways voters perceive them. This work goes beyond the black box of religion as a label others ascribe to an individual to consider strategic religious speech. Several authors have shown that using religious rhetoric can boost candidate support, particularly among religious conservatives (Calfano & Djupe, 2009; Chapp, 2012; Sumaktoyo et al., 2016) and those with high levels of “external religious motivation” (Jennings, 2016). Such language can sometimes serve as an effective “dog whistle”—signaling shared religious background to evangelicals while conveying little to religious liberals and the nonreligious (Albertson, 2014). It can also reduce voter prejudice against candidates who are members of religious or racial outgroups (Calfano, Friesen, et al., 2013; Calfano & Paolino, 2010). Finally, God talk not only affects candidate support; it also—perhaps not surprisingly—effectively signals candidate ideology (Calfano, Djupe, & Wilson, 2013; Weber & Thornton, 2012). Conversely, Djupe, Lewis, Jelen, and Dahan (2014) discover that “rights talk”—framing claims with religious implications in the language of rights, rather than religion—signals politicians’ ideological moderation and helps build support for religious liberty for evangelical dissenters (Djupe, Lewis, & Jelen, 2016).

Some of the most interesting work manipulating candidates’ religious traits and language has taken an intersectional approach. A few scholars have asked how candidates’ religious traits intersect with other attributes such as their race and gender. For instance, do U.S. voters evaluate religious African American candidates differently than they evaluate religious White candidates (McLaughlin & Thompson, 2016)? Does religiosity affect Tunisians’ evaluations of female candidates differently from the way it affects their evaluations of male candidates (Benstead et al., 2015)? Or do voters evaluate female candidates differently from male ones, or Black candidates differently from White ones, when they use God talk (Calfano & Djupe 2011; Calfano & Paolino, 2010)? In all cases, the answer is yes—voter perceptions of candidates depend on the complex intersection of candidates’ religion, race, and gender.

Finally, one striking study explored how describing a candidate in various ways could affect the inferences voters drew about his religious affiliation itself (as Muslim or Christian). Studying the U.S. 2008 election, Layman, Kalkan, and Green (2014) sought to understand the widespread but false belief that Barack Obama was Muslim (he is actually Christian). Though the researchers were ethically constrained to present only true information, they manipulated the way Obama was described: with or without his middle name “Hussein” and focusing or not on his family background and childhood exposure to Islam. They found that both respondent ideology and political awareness affected the extent to which the treatment led respondents to report falsely that Obama was Muslim.

Yet the dependent variable in candidate experiments does not always need to be related to candidate inferences or preferences. A new and rapidly growing body of literature reverses the causal arrow, examining how politics affects citizens’ religious affiliations. Thus one intriguing recent study examined whether candidate information—in particular, exposure to fake candidates who exhibited varying levels of religiosity—affected respondents’ own secular orientations (Campbell, Layman, Green, & Sumaktoyo, 2018).

Experiments Manipulating Appeals Attributed to Religious Elites and Institutions

Another approach to experimentation grew from a finding that directly challenged the notion that religion was stable, socialized, and united. Studying the United States, Smith (2008) and Djupe and Gilbert (2009) found that the political speech of Catholic and Protestant (respectively) clergy had essentially no discernable effect on congregants—clergy political opinions and speech patterns were largely uncorrelated with congregant opinions. Instead, Djupe and Gilbert found evidence of defensiveness—disagreement drove down accurate perception of clergy cues. Yet in other contexts, research discussed later indicates that the political positions of clergy do sometimes directly affect congregants’ attitudes. Such mixed findings have motivated scholars to analyze clergy–congregant communications to understand when clergy are able to circumvent congregants’ defenses.

Beyond political messages, researchers have also examined the impact of clergy speech on religious topics. While congregants might be motivated to ignore clergy political speech, few would deny their clergy the role of articulating and interpreting religious messages. Importantly, there is no shortage of religious values (regarding how the world should work and how individuals should act), religious beliefs (regarding how the world works and is constituted), or religious behaviors that may have political effects. Of course, there is no need to explicitly attribute religious cues to religious elites; religious messages can be manipulated experimentally without attributing them to any particular messenger, as if in a vacuum. But it is important to consider theory in which those cues are at least implicitly offered by religious elites in the context of religious institutions (or at least in some context). Those elites have different attributes (e.g., men vs. women in some denominations), and people have often complicated histories with both the elites and their institutions. Or, put another way, would religious messages have the same influence coming from any source in any other context? It is likely or perhaps just plausible that people think about their own clergy and congregation when primed with religious cues of some sort. Fortunately, some research has begun to disentangle these threads and highlight the leverage that can be gained by making the elite or institutional source explicit.

A few experiments have manipulated the physical appearance of the elite source, comparing clergy to other authority figures. Condra, Isaqzadeh, and Linardi (2019) randomly assigned Afghan clerics to wear religious or civilian dress, comparing the effects on charitable donations; they also manipulated the use of a scriptural appeal versus a common good message. They found that clerics were able to induce higher compliance rates but not to increase the size of donations, while including scriptural messages in the cleric condition also increased donations. They concluded that clergy can induce norm compliance, but religious messages are more powerful in inducing generosity than the particular source conveying that message. Similarly, in a lab-in-the-field experiment in Ghana, McCauley (2014b) compared subjects’ cooperation with a Pentecostal preacher versus various forms of a regional “strong man”—the same experimental confederate but dressed and described differently. Examining giving in a dictator game, he found evidence to suggest “Pentecostal exclusivity, excessive allegiance to leaders, and a shift away from ethnic-based patronage to Pentecostal patronage” (p. 761).

Most experimental work in this area manipulates not the physical manifestation of a clergyperson but his or her political or religious speech and opinions. Clergy influence on key public issues has been confirmed across a wide variety of studies, each offering some complications that make aggregation difficult. In the Brazilian context, Boas and Smith (2015; also Smith, 2019) find that campaign-related messages from (hypothetical) evangelical clergy can shift vote choice. And a number of studies examine immigration attitudes in the United States. For instance, there is some evidence that pro-immigration messages from in-group clergy (from the same denomination) are able to move Americans’ opinion on immigration, though the degree of influence depends on the religiosity of the recipient (Wallsten & Nteta, 2016). Similarly, Margolis (2018a, 2018b) finds that American evangelicals’ attitudes on immigration shifted in response to the advertised messages of the Evangelical Immigration Table, though immigration reform opponents were demobilized by them. More general messages of tolerance also seem to work. Respondents exposed to a clergy’s argument to be respectful of all people expressed warmer feelings toward immigrants and more liberal immigration attitudes as a result (Djupe, Neiheisel, & Olson, 2015).

However, other work has found effects conditional to time—evangelicals exposed to anti-Trump messages from a credible religious elite in the 2016 election expressed less warmth toward the Republican nominee in September, but similar messages did nothing a week before the election (Djupe & Calfano, 2018). In the case of evangelical support for Trump, partisanship appears to dominate religious considerations.

Likewise, other work finds that in-groups are not necessarily ready consumers of whatever arguments are proffered to them. Evangelicals have always lagged behind others in their support for environmental protection, which made the National Association of Evangelicals’ shift to support for taking action on climate change in 2003 potentially highly influential. Experimental evidence suggested that the particular justification clergy offered for the decision mattered (Djupe & Gwiasda, 2010). Using an evangelical decision-making process (through reflection and prayer) was influential for evangelicals, but only among those who were not personally invested in the issue (captured through low issue importance). Another study affirmed the mechanism—that credibility and trust go up when arguments are accompanied by a trusted decision process language (Djupe & Calfano, 2009; see also Djupe et al., 2016).

A good amount of other work has found religious elite influence bounded by other considerations. In their study of reactions to extreme proposals calling for stripping rights and liberties from an opponent, Calfano and Djupe (2015) found that participants summarily dismissed and ignored clergy advocating for gay rights. They judged anti-gay rights clergy, especially evangelical clergy, as credible, and more so when the clergy expressed intolerant opinions (stripping rights from those they disagreed with—in this case lesbian, gay, bisexual, and transgender activists). Adoption of their arguments depended on in-group status, and intolerant appeals only received less support from non-evangelicals. That is, evangelicals supported the in-group clergyperson, even when they called for a clearly unconstitutional “law to stop homosexuals from lobbying for their immoral agenda.”

Other work offers almost painful complications to theories of clergy credibility. For instance, McCauley (2014a) finds that the effectiveness of postconflict prejudice reduction strategies in the Ivory Coast depends on the religious group. Muslims react well to theological messages regardless of the religious or political source delivering them, while Christians react well to politicians delivering an anti-bias message regardless of whether it has theological content. Adkins, Layman, Campbell, and Green (2013) find that when religious and nonreligious leaders’ stances are presented in opposing pairs, those stances primarily tend to move liberals, Democrats, and secular respondents away from religious leaders’ positions, and especially those attributed to evangelical leaders—thus more out-group rejection than in-group embrace of positions. In contrast, Robinson (2010) finds evangelical activists are more likely to embrace a religious elite’s argument when the elite is plausibly within the Christian conservative movement than when the elite is outside the movement (Catholics vs. mainline Protestants in this case), especially when the activists have had greater contact with that group.

To summarize, a large body of research demonstrates that religious elites’ political and religious positions can affect citizens’ attitudes and behavior on a very wide array of public issues. However, clergy do not always fulfill their potential. The barriers to religious elite influence depend in part on whether clergy adopt behavior and communication strategies that send signals of credibility to adherents. Even more importantly, though, clergy influence also depends on the broader social and political context—including the strength of partisan and ethnic ties, the dominant political and secular norms, and the strength of messages citizens received from other, nonreligious, social and political elites.

Yet it is important to note that the contextual limits on religious influence apply to religious influence on politics more broadly—they do not only inhibit clergy influence in particular. For instance, in the Brazilian context, Smith (2019) finds that citizen adherence to secular norms conditions their responses to information about candidate religiosity. Citizens more strongly committed to separation of church and state are turned off by candidates who advertise their religiosity, while those who favor the intermixing of church and state are attracted to candidates who publicize their own religiosity. In short, context and norms likely always impose limits on the scope of religious influence on politics.

Experiments Priming Subjects’ Own Religious or Political Beliefs or Manipulating Other Religious Attributes of Subjects

Can researchers randomize whether citizens are, say, Buddhists or Christians? As discussed earlier, experiments involving religion face a natural limit: it is probably impossible to randomly assign individuals to one religious tradition or another. Even if it were possible, most people would likely consider doing so unethical. Yet creative researchers have found partial workarounds. In one shrewd approach, Clingingsmith, Khwaja, and Kremer (2009) studied the effects of an actual lottery drawing for Pakistanis to participate in the annual Hajj (pilgrimage) to Mecca. The scholars concluded that performing Hajj attaches participants to the international form of the religion and less to the localized forms.

It is also possible to create a sense of religious experiences. Throughout research in this area, there is a tight connection between observational research and experimentation. Hirsh, Walberg, and Peterson (2013) ask, if religious people tend to be conservative and “spiritual” people tend to be liberal, why? The researchers induced subjects to complete a four-minute, guided meditation exercise delivered by video, expecting that spiritualism that involves the “dissolution of self-boundaries and an enhanced sense of connection with the world” (p. 15) would lead to more liberal orientations. Indeed, the treatment reduced participants’ levels of social dominance orientation, an orientation closely linked to conservative political attitudes and identity.

Yet more commonly, scholars seeking to test the impact of subjects’ religious experiences prime participants with religious elements of their own traditions. In effect, they offer a reminder of some message citizens have likely been exposed to before. Priming is the act of raising a consideration in memory so that future decisions may draw upon that consideration. The essential logic is that people have a wide variety of considerations that compete for their attention and that, over time, religious messages get buried by secular ones. The exhortation from Christian clergy that religion is not just a Sunday practice explicitly acknowledges the need to keep religion “at the top of the mind.”

There are now several strands of priming work: some prime respondents’ beliefs and behaviors, some prime their religious identity, and others prime religious values. This section takes them up in turn.

Perhaps the simplest type of prime stimulates respondents to ponder words associated with their own religious tradition. Scholars might give respondents puzzles that induce them to reflect on words such as “God,” “Jesus,” “Mohammed,” or “karma.” For example, Shariff and Norenzayan (2007) asked randomly selected respondents to unscramble sentences that included five target words: spirit, divine, God, sacred, and prophet. Such studies provide evidence that religious reminders affect behavior, promoting trust and cooperation and inhibiting social defection, perhaps because the primes remind subjects that God is watching and will punish noncompliance (e.g., Gervais & Norenzayan, 2012; Shariff & Norenzayan, 2007). Word primes also affect broader social and political dispositions. It turns out that priming Buddhist respondents with Buddhist concepts decreases ethnic prejudice in most cases (Clobert, Saroglou, & Hwang, 2015)—the opposite finding of priming Christian respondents with basic concepts from their own religion (Johnson, Rowatt, & LaBouff, 2010).

A surprisingly uncommon approach is to prime particular religious beliefs. When Be’ery and Ben-Nun Bloom (2015) primed the message that God is in control, it served to elevate support for the welfare state, not opposition to it as cross-sectional studies have found from those with higher religiosity. Likewise, priming the self-affirming message of the “God-given potential” citizens possess (emphasis in original) in Nairobi, Kenya, led to greater political participation (McClendon & Riedl, 2015). And studying Italy, France, Ireland, and Turkey, Warner, Kilinc, Hale, Cohen, and Johnson (2015) primed subjects’ own religious beliefs by asking them to write an essay on one of six topics: “duty to God, community expectations, similarity, deservedness, and God’s Grace.” They found that priming these varying religious messages affected charitable contributions.

Another approach has been to prime respondents’ own religious beliefs, identity, and practices by randomly varying the placement of relevant survey questions on those topics. In a series of papers, Ben-Nun Bloom and Arikan (2013; Ben-Nun Bloom, Arikan, & Courtemanche, 2015) primed citizens’ religious beliefs by asking about belief in God, life after death, heaven, hell, and the existence of a soul before asking about support for democracy, though it is effectively a measure of political tolerance. In both cases, priming belief undercut democratic stances, either with a generalized measure (Ben-Nun Bloom & Arikan, 2013) or involving immigrants specifically (Ben-Nun Bloom et al., 2015). Priming belief systems, they argue, serves as an instrument of mental closure around the group, akin to dogmatism (see also Sagioglou & Forstmann 2013; see Albertson [2011] for an implicit attitudes context).

In those same studies, Ben-Nun Bloom and Arikan (2013) also used question ordering experiments to prime religious behavior (worship attendance) and found the exact opposite pattern. Compared to the control, those primed to consider the communal worship experience expressed greater support for democracy (tolerance) and decreased prejudice. In this case, consideration of the diverse set of people in the pews and the difficulty of collective action with them has a direct logical extension to democratic processes and norms.

The boundaries between the self/group and others are probed more directly in a series of studies by Djupe and Calfano (2013a, 2013b, 2013c). Their inspiration comes from the religious economies approach, which arrays religious groups on a scale from inclusive to exclusive as a way to differentiate their offerings (Stark & Finke, 2000). While this takes form in the practices of houses of worship, clergy will also communicate these commitments as values—commands of how to engage the world. Priming inclusive values, where believers are encouraged to reach out and involve/engage new people and reduce group boundaries, while priming exclusion, where believers are encouraged to keep to themselves and protect the in-group, should enhance boundaries with the world.

This is just what Djupe and Calfano (2013c) found in a range of experiments. In Springfield, Missouri, they found that priming inclusive values weakened threat perceived by a least liked group and thereby enhanced tolerance. In another study (Djupe & Calfano, 2013a) that involved religious adherents from a range of religious traditions sitting in their home pews, they found that priming inclusive values did not affect attitudes on American foreign intervention since they are effectively chronically accessible as measured through clergy’s self-reports. Instead priming exclusive values augmented support for go-it-alone military interventions and reduced support for cooperative (United Nations) interventions. And, lastly, they found that priming these values affected immigration attitudes in consistent ways—inclusion boosts support for immigration reform, while exclusive values undercuts support (Djupe & Calfano, 2013b). Hsiung and Djupe (2018) used these values, along with some other information, as part of religious worldviews that, when primed, affect social and political trust. The key revelation is that these values are expected to vary across time as the needs of the congregation change and thus lead to a range of potential policy attitude outcomes even as general patterns obtain.

Finally, a related approach primes religious identity directly. Of course, reminders about participants’ beliefs and behaviors could prime identity, so the separation between these various strands is not a clean one (e.g., Ben-Nun Bloom & Arikan, 2013; Ben-Nun Bloom et al., 2015). But McCauley (2014c) takes a different tack in studies of Ghana and Ivory Coast, priming religious versus ethnic identity by mentioning religious (as opposed to ethnic) groups in society in the context of five-minute radio broadcast treatments. He finds that priming religious identity leads respondents to prioritize society-wide policy, rather than an emphasis on “club goods” produced by priming ethnic identity.

Experiments Manipulating the Characteristics of Other Citizens

Yet another type of experiment does not manipulate the attributes/behavior of clergy or the respondent but rather varies how fellow citizens are described. The aim of such studies is generally to test how those varying descriptions affect respondents’ out-group attitudes. For instance, Karpowitz, Monson, and Patterson (2016) conducted an online study of norms toward derogatory speech regarding religious out-groups. Respondents read the following prompt:

While discussing religion and politics on a national television news program, a political commentator recently made the following statement, “People don’t know much about [Mormons/Catholics/Jews/Muslims/evangelicals/Mitt Romney]. When they find out, they are amazed at how weird they (he) really are (is). They’re (He’s) just not normal. What a strange group (guy)—they’re (he’s) disgusting, really.”

(Karpowitz, Monson, & Patterson, 2016, p. 518)

The authors sought to understand how varying the identity of the target group affected respondents’ levels of discomfort with the derogatory speech, using a battery of items such as “People should be reprimanded for making statements like this” and “I feel uncomfortable when I read comments like this.” They also assessed how the varying identity of the out-group affected respondents’ perceptions of the commentator as prejudiced (or not). Results revealed that both Democrats and Republicans exhibited the highest level of unease when the negative comments were made toward Jews. By contrast, there were very large partisan differences in perceptions of derogatory speech about Muslims. While Democrats were nearly as protective of Muslims as of Jews, Republicans exhibited the lowest levels of unease toward negative speech about Muslims of any target group.

Note that randomizing the target group was critical. If each respondent had been presented with negative comments toward all six targets, respondents with strong norms of fairness might have been inclined to self-censor, aligning responses to reduce the perception that they were being unfair or inconsistent (see also Ben-Nun Bloom et al., 2015). Similar dynamics played out in Aarøe’s (2012) study of Europeans’ tolerance toward religious displays. Her study used the following prompt: “Do you agree or disagree that a ban should be introduced prohibiting judges from wearing a [necklace with a Christian cross/necklace with a Muslim crescent/Muslim headscarf] at work?” She found that respondent opposition to the religious display was affected both by the religious group under consideration and by the prominence of the display.

Alternatively, researchers could randomize not the target out-group but the way a single out-group is described. For instance, Calfano, Djupe, Cox, and Jones (2016) examined how descriptions of Muslim Americans as patriotic or unpatriotic affected anti-Muslim attitudes. They found that among respondents who trust the Fox News Channel, positive descriptions of Muslims actually led to a backlash in which respondents expressed more negative views toward Muslims.

Others have manipulated how other countries are described in religious terms. Focused on the public’s role in foreign policymaking through their attitudes, Lacina and Lee (2013) explored whether opinion shifted in response to learning the regime type (democratic vs. nondemocratic) and the dominant religious affiliation (Islam vs. Christian), finding that religious affiliation was the dominant force shaping trust and threat perceptions. Isani and Silverman (2016) elaborate on these findings, exploring how the way that Islam is conveyed to people matters to their response: exposure to a message involving “shari’a” invokes fear, especially among conservatives, and more Islamic cues produce additive effects. On the flip side, religious Americans show “my brother’s keeper” effects, meaning a willingness to intervene in other nations’ affairs in order to protect fellow Christians (Wu & Knuppe, 2016).

Experiments Manipulating Religious Institutional Cues Received by Clergy

One last set of experiments in religion and politics seeks to understand the behavior of a narrow but important set of actors: clergy. Several recent studies explore how varying institutional cues affect the political behavior and attitudes of clergy using question order experiments. A question order experiment randomly varies the order in which questions are presented in the survey as a means of influencing what clergy are thinking about at the moment they answer the question or questions that serve as the dependent variable. The questions the clergy receive immediately before the dependent variable serve as a prime to consider certain influences.

Thus Calfano and Oldmixon (2016) surveyed Roman Catholic priests in the United States, asking them which of various entities (the bishop, Church doctrine, etc.) was the most important source of guidance in their decision making. One-third of respondents received the dependent variable (i.e., the question about sources of guidance) immediately after receiving an “institutional cue”: a four-question battery of Likert-type agree/disagree statements designed to make priests think about the institutional demands they faced. Another third received the dependent variable immediately after receiving an “interpersonal cue” designed only to make them think about pressures within their own parishes. The last third received neither prime before answering the dependent variable. The researchers discovered that priming priests to think about the Catholic Church hierarchy pushed them to rely on guidance from their bishop in discharging their responsibilities (see also Calfano, Michelson, & Oldmixon, 2017).

Other studies have followed a similar pattern. Smith (2016, 2019) included a question order experiment in her survey of clergy in Brazil, which served to prime feelings of market pressure from other churches for congregants. Studying Roman Catholic priests in Ireland and Northern Ireland, Calfano, Oldmixon, and Suiter (2014) found that institutional primes led priests to adopt more populist economic attitudes and more conservative cultural attitudes, relative to interpersonal primes or no primes. They argued that both types of attitudes were “(religious) establishment preferences” (p. 397).