Conservative Christian Persecution Discourse and Support for Political Violence: Experimental Evidence from the United States

Abstract

International Relations scholarship on religious freedom points to religious persecution as a major driver of political violence around the globe. If correct, the perceived persecution of conservative Christians in the United States (U.S.) may contribute to the radicalization of individuals who self-identify as conservative and Christian. Yet, in focusing on country-level indicators, previous empirical research on the “religious freedom peace” is generally silent on the role of individual-level perceptions in the formation and mobilization of grievances. This article represents a first attempt to fill this gap. As such, it asks if the religious freedom discourse articulated in conservative American media contributes to the radicalization of its domestic consumers through the cultivation of perceptions of persecution that are divorced from the generally high levels of religious freedom otherwise experienced in the U.S. Although the results of an original online survey experiment demonstrate that persecution discourse does indeed shape perceptions of threat to religious liberty, I find no support for the idea that it also leads to increased support for political violence, either directly or indirectly through misperceptions of persecution.

1. Introduction

Widespread perceptions of threat to religious freedom have taken hold among American Christian conservatives in recent years.1 Conservative politicians, pundits, and activists intentionally cultivate these perceptions by framing issues of major concern as violations of Christians’ religious freedom. This has been the case since at-least the late-1970s, but this tactic has surged in popularity since the Obama Administration first tried to implement the Affordable Care Act’s (ACA) contraception mandate in 2012. Subsequent decisions by the Supreme Court to expand the rights of LGBTQ—lesbian, gay, bisexual, transgender, and queer—people have further exacerbated this trend to the point that one conservative essayist could ominously title her book on the subject It’s Dangerous to Believe (Eberstadt 2016).

The growing popularity of this idea—that it is dangerous to overtly practice, express, or even believe in one’s Christian faith in the United States (U.S.)—is concerning for several reasons. First, although some academics and critical Evangelical commentators uncharitably pathologize (mis)perceptions of threat to religious freedom as part of a Christian or Evangelical “persecution complex” (Castelli 2007a; Noble 2014; Singer 2019; Wiedel 2014; also see Wilkinson [2017] 2019), members of religious majorities that perceive themselves as part of a persecuted minority experience adverse psychological effects similar to those reported by members of minority religious traditions who have really experienced discrimination on account of their beliefs (Parent et al. 2018; Vang et al. 2019; but see Rosik and Smith 2009).

Second, and most importantly for the present study, although such experiences may be triggered by different events for religious minorities and majorities, perceptions of threat also lead to increased outgroup hostility for members of both minority and majority religious groups (Pasek and Cook 2019). This is unsurprising because (mis)perceptions of religious abuse are primarily cultivated for political purposes (Castelli 2007a, 2007b; Djupe et al. 2016; Jones 2016; Knippa 2015; Lewis 2017; Lynn 2006; Michaelson 2013; Moss 2013; Whitehead et al. 2018; Williams 2018; Wolraich 2010).2 Nevertheless, this may be especially problematic because the International Relations scholarship on the “religious freedom peace” identifies religious oppression as one of the world’s chief drivers of political violence (e.g., Farr 2008; Grim and Finke 2011; Iannaccone and Berman 2006; Inboden 2012; Preston 2013; Saiya 2015, 2018, 2019; Toft et al. 2011).

Even so, the causes and consequences of (mis)perceptions of persecution have not been studied extensively (but see Basedau et al. 2017; Goidel et al. 2016; Whitehead et al. 2018). Moreover, while students of the religious freedom peace thesis (RFPT) have linked higher levels of religious freedom to reductions in various kinds of political violence at the country-level (Grim and Finke 2011; Henne et al. 2020; Muchlinski 2014; Saiya 2015, 2018, 2019; Saiya and Scime 2015), the theory is premised on untested expectations about the formation and mobilization of grievances at the individual-level.

This is an important gap, because (mis)perceptions of persecution persist even though the U.S. is, by objective accounting, one of the most religiously free countries in the world (Fox 2009; Kohl and Pina 2016; also see Fox et al. 2018). Yet, if the individual-level process of radicalization presumed by the RFPT is driven by perceived instances of religious oppression (Cox 2015), as appears to have been the case when Timothy McVeigh cited the government’s persecution of the Branch Davidians at Waco among his motives for the Oklahoma City bombing in 1995 (Smith 2015), conservative American Christians could be at increased risk of radicalizing.

The present study examines this possibility. However, it also explores the prospect that perceptions of persecution may not themselves suffice to independently motivate individuals to action. Though the RFPT is generally silent on collective action problems, disaffected groups need the opportunity to mobilize their members to act on their grievances (among others, see Snyder and Tilly 1972). An absence of religious repression may itself create structures of opportunity for violent or nonviolent political action (Tarrow 1989), but even under such conditions, it is conceivable that perceptions of persecution must be cultivated and mobilized through in-group processes before leading to the kinds of violence predicted by the RFPT. Even if the RFPT is itself mute on this issue, research on religion and nonviolent political activism suggests that religious leaders play a central role in mobilizing individual grievances in countries throughout the world (e.g., Koter 2013; Omelicheva and Ahmed 2018; Spenkuch and Tillmann 2018). Meanwhile, research on “dog-whistle” politics and elite cueing suggest the same about violent activism. Indeed, something as simple as metaphor or narrative could play an important role in catalyzing support for violence (e.g., Chyzh et al. 2019; Kalmoe 2014; also see Jones and Paris 2018).

By connecting these two bodies of literature with the RFPT, I therefore theorize that while opinion leaders may shape perceptions of threat to religious freedom, they must call for violence, even if tacitly, before individuals respond to perceived instances of persecution with increased support for political violence. To elaborate, this paper asks if the religious freedom discourse articulated by American conservative Christian politicians, media personalities, authors, and filmmakers contribute to the radicalization of conservative audiences either indirectly, through the cultivation of (mis)perceptions of persecution, or directly, through tacit appeals to violence. I employ an online survey experiment to address such questions and find that while the media can indeed shape perceptions of threat to religious liberty, there is no etiological evidence that conservative Christian persecution discourse cues support for political violence.

By focusing on the relationship between perceptions of persecution and attitudes towards political violence in the U.S., this study seeks to make several theoretical and empirical contributions. First, insofar as it examines the relationship between religious persecution and religious freedom on the one hand, and political violence and peace on the other, this study is informed by and contributes to the International Relations literature on the religious freedom peace. More specifically, the relationship between religious persecution and violence has been tested on data that primarily measure direct manifestations of religious regulation at the country-level. By experimentally examining the way individuals’ perceptions of their circumstances shape their willingness to support violence though, this research tentatively suggests that perceptions of religious freedom are unlikely to contribute to the process of radicalization presupposed by the RFPT. The implication is that scholars of religion and international relations must do more to theorize the mechanisms linking religious freedom and violence at the country-level.

Second, with the proliferation of interest in the roles played by ideas and ideologies in extremist recruitment, researchers have turned to the analysis of frames that mediate the impact of various background conditions on people’s perceptions, attitudes, and behavioral choices. This study contributes to this research by assessing the impact of media frames (i.e., persecution discourse) on individuals’ support for political violence. Third, this study adds to the burgeoning literature on the relationship between language and violence. Fourth, by investigating the causal effects of religious freedom framing in the American case, this study builds on research in American politics on the relationship between the conservative media and perceptions of threat to religious freedom. It can, therefore, also be viewed as a contribution to the “domestic turn” in International Relations research (Kaarbo 2015; also see Werner et al. 2003, and their colleagues). Finally, because this study employs an individual-level survey experiment to test the RFPT, it provides a concrete example of how the experimental method can be employed by scholars of religion and politics to expand our knowledge base about an array of complex phenomena situated at the nexus of religion and international relations.

That said, the remainder of the paper turns to unpacking the argument and presenting the findings. To that end, conservative American Christian persecution discourse is briefly introduced in the next section. The subsequent section discusses (mis)perceptions of persecution in the American case. The relationship between persecution discourse, perceptions of threat to religious liberty, and support for political violence is theorized in the following two sections. The first elaborates on the RFPT while the second explores how responses to perceptions of persecution may be shaped by contributors to the discourse. The experimental research design is presented afterward, followed by the results of the experiment. I conclude by summarizing this paper’s contributions and highlighting the implications of the findings for future research.

2. American Conservative Christian Persecution Discourse

Persecution discourse is a collection of narratives about anti-Christian marginalization rooted in conservative American identity politics.3 As illustrated by the bibliography of select books, movies, and television series presented in

Table 1. Select chronological bibliography of American conservative Christian persecution discourse by wave.

Wave I (Pre-9/11)	Wave II (Post-9/11)
Arnold, O. Carroll. 1978. Religious Freedom on Trial. Valley Forge, Pennsylvania: Judson Press. *	McQuaid, Elwood. 2003. Persecuted: The Growing Intolerance Toward Christianity. Eugene, Oregon: Harvest House Publishers.
Buzzard, Lynn R., and Samuel Ericsson. 1982. The Battle for Religious Liberty. Elgin, Illinois: David C. Cook Publishing.	Sears, Alan and Craig Osten. 2003. The Homosexual Agenda: Exposing the Principal Threat to Religious Freedom Today. Nashville, Tennessee: Broadman and Holman Publishers.
Jurjevich, Ratibor-Ray M. 1985. The War on Christ in America: Christian Fortress in America Under Siege—Christophobes of the Media and of the Supreme Court in Action. Denver, Colorado: Ichthys Books.	Parshall, Janet. 2008. Speechless: Silencing the Christians. Tupelo, Mississippi: American Family Association.†
Sekulow, Jay. 1990. From Intimidation to Victory: Regaining the Christian Right to Speak. Lake Mary, Florida: Creation House.	Gregory, David. 2010. The Last Christian. Colorado Springs, Colorado: WaterBrook Press.‡
Fournier, Keith A. 1993. Religious Cleansing in the American Republic. Washington, D.C.: Liberty, Life, and Family Publications.	Hicks, Marybeth. 2011. Don’t Let the Kids Drink the Kool-Aid: Confronting the Left’s Assault on Our Families, Faith, and Freedom. Washington, D.C.: Regnery Publishing.
Schaefer, Frank. 1994. Religious Apartheid: The Separation of Religion from American Public Life. Muskegon, Michigan: Gospel Films.†	Kern, Sally. 2011. The Stoning of Sally Kern: The Liberal Attack on Christian Conservatism—and Why We Must Take a Stand. Lake Mary, Florida: FrontLine.
Whitehead, John W. 1994. Religious Apartheid: The Separation of Religion from American Public Life. Chicago, Illinois: Moody Press.	• Schlafly, Phyllis and George Neumayr. 2012. No Higher Power: Obama’s War on Religious Freedom. Washington, D.C.: Regnery.
Kennedy, D. James and Jerry Newcombe. 1996. The Gates of Hell Shall Not Prevail: The Attack on Christianity and What You Need to Know to Combat It. Nashville, Tennessee: Thomas Nelson Publishers.	Cronk, Harold. 2014. God’s Not Dead. Scottsdale, Arizona: Pure Flix. ¶
Sekulow, Jay Alan. 1996. Knowing your Rights: Taking Back Our Religious Liberties: A Handbook. Virginia Beach, Virginia: Life, Liberty, and Family Publications.	Starnes, Todd. 2014. God Less America: Real Stories from the Front Lines of the Attack on Traditional Values. Lake Mary, Florida: FrontLine.
Sekulow, Jay and Keith Fournier. 1996. And Nothing but the Truth: Real-Life Stories of Americans Defending Their Faith and Protecting Their Families. Nashville, Tennessee: Thomas Nelson Publishers.	Eberstadt, Mary. 2016. It’s Dangerous to Believe: Religious Freedom and Its Enemies. New York, New York: Harper Collins.
McDowell, Josh and Bob Hostetler. 1998. The New Tolerance: How a Cultural Movement Threatens to Destroy You, Your Faith, and Your Children. Carol Stream, Illinois: Tyndale House Publishers.	Erickson, Erick and Bill Blankschaen. 2016. You Will Be Made to Care: The War on Faith, Family, and Your Freedom to Believe. Washington, D.C.: Regnery Publishing.
McCalvany, Don. 1999. Storm Warning: The Coming Persecution of Christians and Traditionalists in America. Oklahoma City, Oklahoma: Hearthstone Publishing.	Dreher, Rod. 2017. The Benedict Option: A Strategy for Christians in a Post-Christian Nation. New York, New York: Sentinel.
Emmanuel, Tristan. 2000. Christophobia: The Real Reason Behind Hate Crime Legislation. Ontario, Canada: Freedom Press.	Starr, Ken. 2021. Religious Liberty in Crisis: Exercising Your Faith in an Age of Uncertainty. New York, New York: Encounter Books.

* Entries are nonfiction texts, unless specified otherwise. ^† Documentary film or television series. ^‡ Christian fiction (text). ^¶ Christian fiction (film).

, it appears to have emerged in two somewhat distinct waves as a genre of conservative polemic. The first wave emerged in the 1970s and 1980s emphasizing the disestablishment of official religion in public spaces (e.g., LaHaye 1980; McGraw 1976; also see Diamond 1998). It was especially invested in resisting the removal of prayer from public schools. By contrast, the second wave largely coalesced after 9/11 and shifted emphasis to sexual freedom. As such, persecution discourse is today primarily concerned with limiting the expansion of women’s and LGBTQ people’s rights (e.g., Eberstadt 2016; Sears and Osten 2003). However, this may change if a distinct third wave emerges in reaction to restrictions caused by the Covid-19 pandemic.

While the first two waves each have a comparatively unique focus, the discourse writ large shares a common thesis: traditionalist Christians, at least those who are uncompromisingly committed to their sincerely held religious beliefs, are unfairly subjected to protests, unemployment, and other forms of “soft” persecution if they attempt to act on their beliefs in public (e.g., Eberstadt 2016).4 This persecution, so the narrative goes, is aimed at forcing believers to abandon their sincerely held religious beliefs by acquiescing to the nebulous forces of secularism, postmodernism, and cultural relativism.

To advance its core argument, persecution discourse attempts to frame every ostensible criticism of Christianity,5 and, among other things, every conservative political defeat in the culture war,6 as part of a well-orchestrated campaign of anti-Christian discrimination and hostility. As Table 1 makes clear, its messaging is spread and reinforced by multiple media, including by a Christian film industry that emphasizes its viewers’ victimhood (e.g., see the God’s Not Dead trilogy) and a veritable genre of popular persecution literature. The message is also spread by sympathetic news coverage in the conservative press. Prior to his termination, the most prominent example of the latter was quite certainly Bill O’Reilly’s perennial coverage of the putative “War on Christmas,” which began on Fox News circa 2004 (e.g., O’Reilly 2004).

Of course, all this is to say nothing of the conservative think-tanks and activist organizations whose public relations campaigns, legal actions, and lobbying efforts promote the same narratives (e.g., Gallagher 2006; Messner 2008, 2011; also see Andrews 1994; Bennett 2017; Hacker 2005; Stahl 2016; Steinfels 2006).

3. (Mis)Perceptions of Persecution among Conservative Christian Americans

Persecution discourse notwithstanding, the historical struggle for religious freedom in the U.S., like the fight for other civil rights, has been quite onerous (Joshi 2020; Muñoz 2016; Smith 2015; Wenger 2017; also see Michaelson 2013). Not only have some clergy, churches, and “cults” been targeted by the repressive apparatus of the state itself (Goldstein 2001), but an intolerant American public has, at times, displayed wildly antisemitic, anti-Catholic, Islamophobic, and other anti-religious attitudes towards those outside of the Protestant mainstream (Boyle and Sheen 1997; Beydoun 2018; Dinnerstein 1995; Jenkins 2003; Smith 2015; also see Joshi 2020).7

Nevertheless, there is little reliable evidence that American Christians, namely, Evangelical or fundamentalist Protestants, face anything today similar to the level of animosity experienced, either presently or historically, by the country’s minority religious groups or by groups in many other countries. For example, as illustrated by the Religion and State Project’s comparative data on government religious restrictions, both the religious majority and religious minorities in the U.S. are counted among the most privileged in the world, and they fare significantly better than their counterparts in the worst offending countries (e.g., Vietnam, North Korea, Iran). However, the data also clearly show that the American religious majority possess a greater degree of religious freedom than American religious minorities. Where the U.S. is tied for 14th place globally for its respect for majority religious freedom, it falls to 36th place where its religious minorities are concerned (Fox et al. 2018).8

Moreover, while anti-Christian hate crimes occur every year in the U.S., the rate at which Christians suffer from such incidents is, according to the Federal Bureau of Investigations (FBI n.d.), far exceeded by the disproportionate targeting of minority religious groups.9 Meanwhile, more mild episodes of perceived anti-Christian discrimination have been documented elsewhere, such as on secular college campuses (Hyers and Hyers 2008), and in secular professional associations (Rosik et al. 2016). However, students report comparable experiences regardless of their religious background (Boucher and Kucinskas 2016), even when attending private Christian schools (Rosik and Smith 2009), and the same is true of Christian professional associations (Rosik et al. 2016).

Additionally, while there is some evidence that non-religious progressives harbor zealous anti-fundamentalist Christian attitudes (Bolce and de Maio 1999; Yancy and Williamson 2015), and that such attitudes have been shaped by the media (Bolce and de Maio 2008; Shields 2011), research on political tolerance demonstrates that Christian fundamentalists themselves comprise the only social group tolerated by the majority of Americans (Gibson 2008). Admittedly, it is true that some leftwing activists have framed the Christian Right as a “radical,” “extreme,” or “theocratic” movement (e.g., Hedges 2006; Lynn 2006; Swenson 2004; Wolraich 2010; also see George and Wilcox 1996). Indeed, some scholars and progressive Christian thought leaders even see conservative Christianity as a threat to religious freedom in its own right (e.g., Lynn 2006; Miller 2000; also see Guth et al. 1997). Yet, in recent years, the percentage of the public at-large that supports religious freedom for even the most extreme groups has reached an all-time high (Freedom Forum Institute 2019).

Despite this reality, American Christians increasingly perceive their ability to practice their faiths freely to be in jeopardy. For example, the Public Religion Research Institute (PRRI) found that the percentage of Americans agreeing that their religious freedom was under threat increased from 39% in 2012 to 55% in 2014 (Cox and Jones 2012; PRRI 2014). Meanwhile, the percentage of Evangelicals reporting that Christians are more likely to face discrimination than Muslims rose from 45% in 2013 to 57% in 2017 (Green 2017; Jones et al. 2014). Furthermore, where just 20% of Southern Baptist preachers surveyed in 1988 agreed that their religious freedom was threatened in America, that number had exploded to 90% by 2008 (Lewis 2017, p. 82).10

Goidel et al. (2016) suggest that these shifts result from exposure to persecution discourse, especially as pedaled by conservative media outlets such as Fox News. As such, my first hypothesis aims to test the effects of persecution discourse on perceptions of threat to religious liberty:

Hypothesis 1 (H1).

Exposure to persecution discourse increases perceptions of threat to religious liberty.

Establishing this relationship is necessary because persecution discourse may indirectly affect support for political violence through such perceptions. The next sections turn to this possibility.

4. The Religious Freedom Peace Thesis (RFPT)

The widespread (mis)perception of persecution among conservative American Christians represents an interesting puzzle because the idea that religious oppression is a chief driver of political violence around the world has gained substantial currency in the literature on the role of religion in international relations (e.g., Grim and Finke 2011; Farr 2008; Iannaccone and Berman 2006; Muchlinski 2014; Saiya 2015, 2018, 2019; Toft et al. 2011).11

Insofar as it attempts to understand the conditions under which religious actors turn to violent political action, this idea—the RFPT—can be thought of as a neo-Weberian approach to religion. Neo-Weberian approaches to religion seek to understand if and how different social, political, and economic contexts condition the diverse manners in which religious actors behave. As a result, they transcend different theories of international politics (Lynch 2009, 2014). However, since the RFPT is an inherently Lockean idea about violence analogous to the Kant-inspired democratic peace thesis, it can be viewed as a branch of liberal international theory (Owen 1994; Preston 2013; also see Locke [1689] 2010).

That said, scholars advancing this Democratic Peace-like argument reason that a free marketplace of religious ideas drives the moderation of religious actors competing for adherents (Iannaccone and Berman 2006), while a closed marketplace of religious ideas instead cultivates easily exploitable grievances (Grim and Finke 2011). Taken together, this suggests that the expansion of religious freedom around the world is key to reducing terrorism and other forms of political violence while promoting both domestic and international stability (Farr 2008).

Though much of this literature primarily concerns theoretical expectations (e.g., Farr 2008; Toft et al. 2011), a burgeoning body of empirical scholarship has lent credence to such notions. For instance, the religious persecution that can occur under the reign of oppressive regimes has been linked to the radicalization of Muslim groups in multiple contexts (Hafez 2003; Omelicheva 2010). Indeed, historical evidence confirms that state repression has been a source of radicalization for once peaceful religious movements in India, the Middle East, and elsewhere, going back to the emergence of agriculture (Armstrong 2014). Meanwhile, despite evidence that suggests religious violence may instead lead to religious repression (Henne and Klocek 2019; Murphy 2013), several cross-national studies have connected higher levels of religious freedom at the country-level to general reductions in religiously motivated outbursts of hate crime, terrorism, and even war (Grim and Finke 2011; Muchlinski 2014; Saiya 2015, 2018, 2019; Saiya and Scime 2015).12

That said, the RFPT neglects the way religious freedom may constitute a structure of political opportunity for group formation and mobilization (Snyder and Tilly 1972; Tarrow 1989) even though it is apparent that religiously free countries can be a breeding ground for heterodox religious movements, including at least some with extremist tendencies. For example, although the U.S. is often counted among the most religiously free nations in the world (Boyle and Sheen 1997; Cingranelli and Richards 2010; Fox 2009; Fox et al. 2018; Gastil 1985, esp. 157–58; Kohl and Pina 2016; Marshall 2008; Weigel 1982), the same constitutional guarantees of free religious exercise that make it possible for surfing churches, yoga churches, biker churches, and a diverse host of other benign religious movements to flourish throughout the country have also made it possible for white supremacist churches and other extreme religious movements to take root on American soil (Arriga 2013; DeHart 2000; Singular 2001).

Similarly, the widespread respect for freedom of religion in Japan is the reason the country’s response to Aum Shinrikyo’s 1995 sarin gas attacks on the Tokyo subway system is praised in the religious freedom peace literature as a resounding success. That is, students of the RFPT reason that efforts to clamp down on Aum would have created a cycle of escalating violence. Yet, because the Japanese government instead responded by prosecuting only those members responsible for the attack, it diffused the situation. In other words, the remaining members of Aum opted for nonviolence because their religious freedom was protected (Grim and Finke 2011; Shah et al. 2012). However, this disregards the fact that the same protections did not prevent Aum’s turn to violence in the first place. Indeed, it altogether overlooks the conspiracy theory of victimhood Aum’s leaders cultivated prior to the attack. Despite Japan’s respect for their religious freedom, Aum members came to see themselves as an oppressed religious minority, and those that acted on their grievances by carrying out the Tokyo attack aimed to punish the people and institutions they blamed for their organization’s putative persecution (Reader 2011).

Although most of the empirical scholarship investigating the RFPT has focused on country-level characteristics, this ultimately suggests that perceptions of persecution could lead to political violence independent of a regime’s actual respect for religious freedom (Cox 2015). Regardless, the RFPT identifies grievances (about religious discrimination or persecution) as the causal mechanism driving the relationship between religious freedom and political violence, and such grievances obviously represent an individual-level phenomenon (Fox et al. 2017; Moghadam 2005). Thus, American Christians aggrieved by the sense that they cannot freely practice their faith provide an opportunity to test the RFPT’s expectations at the individual-level, which is precisely the aim of the second hypothesis:

Hypothesis 2 (H2).

Perceptions of threat to religious freedom increase support for political violence.

5. Mobilizing Responses to Perceived Persecution

While it is important to understand how individuals respond to perceived instances of anti-Christian persecution, the RFPT is largely silent on the conditions under which such instances may lead to radicalization at the individual-level. However, research on the effects of religion (Omelicheva and Ahmed 2018) and religious discrimination (Fox et al. 2017) on political activism shows that subjective experience may not become politically salient free of any in-group processes that seek to cue attitudes or mobilize grievances (Snyder and Tilly 1972). Because leaders set agendas (Fox et al. 2017), frame opinions (Omelicheva and Ahmed 2018; Zaller and Feldman 1992), and encourage political action (Koter 2013; Omelicheva and Ahmed 2018; Spenkuch and Tillmann 2018), it is reasonable to assume that they play a central role in this process. Of course, politicians and pundits are especially able to cue attitudes through their asymmetric command of media attention (Gamson and Wolfsfeld 1993; Hershey 1992; Zaller and Feldman 1992), and since the media may exert an outsized influence in shaping conservatives’ perception of threat to religious freedom (Goidel et al. 2016), it (and persecution discourse more specifically) likely also plays a pivotal role in shaping how conservative American Christians respond to instances of perceived persecution.

This means that the political or media elite may need to incite violence before conservative Christians’ act on their perceptions of religious persecution in a violent manner. This would be consistent with research on the effects of “dog-whistle” politics, at any rate. After all, several studies demonstrate a link between violent rhetoric, including fiction, metaphorical language, and explicit calls to violence, on the one hand, and both increased support for violence and actual instances of violence, on the other (Asal and Vitek 2018; Chyzh et al. 2019; Glaser et al. 2002; Jones and Paris 2018; Kalmoe 2014). It is unclear how frequently violence is appealed to in American persecution discourse, but in recent years, a few opinion shapers have indeed resorted to violent language in defense of American Christians’ religious freedom.

For example, writing just before President Barack Obama won reelection in 2012, Schlafly and Neumayr (2012, pp. 169–70) remind their readers how the founders of the American republic rejected the tyranny of King George III, using Christianity to justify their civil disobedience towards England. They called for conservatives to disobey President Obama in the same way, before adding, somewhat ominously:

The tree of liberty is fertilized by the blood of tyrants, said Thomas Jefferson. What would Jefferson say today? Next to the growing tyranny of Obama, the grievances Jefferson enumerated against the British crown in the Declaration of Independence look almost minor. King George III, for all of his injustices, never forced the colonists to submit to the false gods of secularism.

(Schlafly and Neumayr 2012, p. 170; emphasis added).

Unfortunately, an embrace of rhetoric not unlike that which Schlafly and Neumayr express has spread to other fora as well. At a 2014 conference for the Faith and Freedom Coalition, for instance, then-Louisiana Governor Bobby Jindal discussed the “rebellion brewing […] for a hostile takeover in Washington, D.C.” meant to stop the war on religious freedom in America (quoted in Rosenberg 2014). A few weeks later, one Barb Wire blogger predicted “widespread civil disobedience” and threateningly hinted of an impending Christian revolt in a piece that was republished by several conservative media outlets, including The Blaze, Townhall, and World Net Daily (Barber 2014a, 2014b, 2014c, 2014d).

Admittedly, such calls for “civil disobedience” could be interpreted as advocating for nonviolent resistance (e.g., Sears and Osten 2003; Whitehead 1994). Moreover, radical beliefs and practices need not necessarily lead to violent political action (Khalil 2013). Still, it is not inconceivable that similar sentiments about revolts or rebellions, especially when intermixed with rhetoric about “the blood of tyrants,” could be interpreted by sympathizers as tacit calls for violent political action (Kalmoe 2014; also see Chyzh et al. 2019). Yet, if even such rhetoric is given the benefit of the doubt, there are others willing to make the call more explicit. Writing for the same blog mentioned above, Jolly (2019) argues that conservative Christians must resist leftwing efforts to take away their rights by “arm[ing] themselves and rebell[ing] against the tyrants that [are] oppressing them” as happened in the Revolution. “Am I calling for a rebellion?,” he adds, “Yes, I am. […] I urge conservatives and Christians to be ready to take up arms” (Jolly 2019).

Such sentiments are unsurprising when one remembers that broad social movements can easily lead to the emergence of sympathizers willing to act individually on their grievances by turning to violence in lieu of either the pen or ballot (LaFree and Ackerman 2009), but such calls, I suspect, can also mobilize larger segments of an audience. Consequently, the third hypothesis considers this possibility:

Hypothesis 3 (H3).

Violent appeals to defend religious freedom increase support for political violence.

Where my third hypothesis considers the direct effect of violent metaphor in persecution discourse, drawing on the expectations of the RFPT, my fourth hypothesis anticipates an indirect effect:

Hypothesis 4 (H4).

Violent appeals to defend religious freedom increase support for political violence among individuals who perceive a high degree of threat to religious freedom.

Alternatively, if the conservative media follows the example of Sears and Osten (2003), who believe conservatives and Christians can defend their religious freedom by taking peaceful religious, political, and legal action against the secular left, most consumers of persecution discourse will likely respond in kind. Written at the height of the battle against gay rights, just before a series of referenda banning gay marriage in several states helped President Gorge W. Bush achieve reelection in 2004 (Jones 2016; Lewis 2017; Putnam and Campbell 2010), Sears and Osten’s (2003) work calls readers to pray, proselytize, protest, write or call political officials, support activist ministries, and among other things, vote for pro-family Republican political candidates (Sears and Osten 2003, pp. 205–11).

Consistent with theories of democratic learning (e.g., Peffley and Rohrschneider 2003), experience agitating for religious freedom throughout this process might increase religious tolerance (Lewis 2017). This would be interesting, given the idea that political violence is perpetrated by individuals who are themselves motivated by intolerance towards their victims (Chyzh et al. 2019; Sisk 2011; Stark and Corcoran 2014). Of course, exposure to preaching about religious liberty has already been found to increase political tolerance (Djupe et al. 2016), which suggests, contrary to the expectations of the RFPT, that the mere belief that religious freedom is threatened may not itself be sufficient to elicit violent reactions (Lewis 2017). Far from radicalizing its readers then, efforts by the conservative media to channel its consumers into the democratic process might instead reinforce respect for democratic norms (Omelicheva and Ahmed 2018). This logic undergirds the fifth hypothesis:

Hypothesis 5 (H5).

Nonviolent appeals to defend religious freedom decrease support for political violence.

6. Research Design

This study employs an online survey experiment to test the five hypotheses articulated above. Thus, the survey was specifically designed to determine whether exposure to persecution discourse shapes perceptions of threat to religious freedom, and to determine whether those perceptions are sufficient to independently increase support for political violence, or whether another factor (i.e., a call to violently defend religious freedom, or an appeal to nonviolent political action) is necessary to elicit this or the opposite effect. As such, participants were randomly assigned to one of three treatments where they read a brief report (≤229 words) on religious freedom similar to what might be published on a popular online conservative website, such as Fox News or The Blaze. A fourth condition included the control, an article about Martian exploration (181 words). Respondents were then asked several questions (see Appendix A) about their perceptions of threat to religious freedom and support for political violence, which were used in the analysis below.

6.1. Mechanical Turk and Participant Recruitment

Participants for the experiment were first recruited using Amazon’s Mechanical Turk (MTurk). MTurk is an online marketplace employing crowdsourcing, where businesses and individuals (i.e., “requesters”) can pay workers around the world to complete what Amazon calls Human Intelligence Tasks (HITs). A HIT can include any computerized task that cannot be performed by artificial intelligence, and that includes online political opinion surveys. Following previous scholarship (Avdan and Webb 2018, 2019), my survey was advertised as an MTurk HIT and opened to voting age American citizens who were compensated $0.40 (USD) for their participation (see Berinsky et al. 2012 for a discussion of compensation).13 Thus, over a period of about 17.5 hours between 14 and 15 April 2020, 962 participants were recruited for the experiment. An additional 38 participants were recruited during a 3.5-hour window on 17 April 2020, yielding a grand total of 1000 initial participants. After addressing data quality concerns (see Dennis et al. 2020), this produced a dataset of 816 surveys (81.6%).14

Of course, using MTurk could raise concerns for some. For example, when compared to the general public, MTurk respondents tend to be younger (Berinsky et al. 2012), less religious (Lewis et al. 2015; Mullinix et al. 2015), and more liberal (Berinsky et al. 2012; Levay et al. 2016; Clifford et al. 2015). Not only has youth been shown to increase support for political violence (Kalmoe 2014; Tessler and Robins 2007; Zhirkov et al. 2014), but because the present study is concerned with attitudes among Christian conservatives, issues with religion and ideology could be especially problematic. However, religious subgroups behave consistently on MTurk with the way their counterparts might in a nationally representative sample (Lewis et al. 2015). As such, the religious composition of MTurk only remains problematic in small samples that seek to explore the moderating effects of specific religion variables (Lewis et al. 2015; Mullinix et al. 2015). The same is true for ideology. That is, conservatives on MTurk also reflect the same values and traits as conservatives drawn from nationally representative samples, their underrepresentation notwithstanding (Clifford et al. 2015).

Moreover, MTurk samples are also more representative of age, as well as a host of other characteristics, including ethnicity, geographic distribution, socioeconomic status, and education, than are alternatively recruited convenience samples, including for example, undergraduate students (Avdan and Webb 2018, 2019; Berinsky et al. 2012; Buhrmester et al. 2011; Casler et al. 2013; Horton et al. 2011; Mason and Suri 2012; Paolacci et al. 2010). Yet, even if this were not the case, because I am interested in comparing differences between experimental conditions, these issues are further minimized by randomly assigning participants to conditions and through regression with control variables. Randomization gives participants an equal opportunity of being placed in any given condition, while regression provides a robustness check to ensure that the results are driven by the experimental conditions rather than specific participant characteristics (Avdan and Webb 2018, 2019; Levay et al. 2016; McDermott 2002). Thus, aside from lingering ethical concerns regarding the low-pay MTurk workers receive (e.g., Semuels 2018), the platform stands out as a useful tool for engaging in relatively valid and reliable experimental research.

6.2. Experimental Procedures and Treatments

After agreeing to participate, respondents were redirected to Qualtrics, a survey hosting and data management website that facilitates randomization and location validation. Once there, they were asked 24–26 questions (see Appendix A), which took an average of 6.5 minutes to answer. After being randomized into one of the three treatment conditions (used collectively to test H1), where one item (i.e., the response cue) was manipulated in three different ways, respondents read a fictitious article about an incident framed as an attack on religious freedom. The scenario conveyed therein is useful for the purposes of the experiment for at least two interrelated reasons. First, the treatment emphasizes and synthesizes the core themes that emerged in each wave of persecution discourse: religion in public space and resistance to sexual freedom. Second, it mirrors one of the more common, specific, and potentially legitimate concerns expressed in persecution discourse about the rights of religious student organizations to exclude LGBTQ students from membership (e.g., Eberstadt 2016; Limbaugh 2003; Sears and Osten 2003).15

That said, respondents in the first condition were given no cues about how they should respond to the story. Those in the second condition (used to test H5) were cued to respond to the story with nonviolent political action. Meanwhile, respondents in the third condition (used to test H3 and H4) saw Schlafly and Neumayr’s (2012) metaphorical appeal to violence added to the story. Finally, respondents in the control condition (which functions as a baseline for comparison) read a brief but unrelated semi-fictitious article (adapted from Strickland 2019) about Martian gas fluctuations. All four conditions are detailed in

Table 2. Summary of experimental conditions.

Fictional News Article about Religious Freedom *
Arcadia State College is accused of banning a Christian student organization for requiring its leaders to live a Christian lifestyle. Christian Outreach, the club at the center of the controversy, was derecognized last week for its refusal to comply with the school’s nondiscrimination policy by allowing openly gay or transgendered students to serve as the club’s officers. Now students belonging to the club are no longer welcome on campus. [Treatment Condition Inserted Here] Activists like Lynn are concerned because this episode is part of a larger trend. In 2014, InterVarsity Christian Fellowship experienced the same problems on a much larger scale when the University of California system derecognized the club at all 23 of its schools. Similar incidents also occurred at the University of Michigan in 2013. For its part, Christian Outreach will lose access to free meeting space and programs sponsored by the Arcadia student activities office, including an annual recruitment fair hosted at the start of the school year.
Condition 1: No Response	Condition 2: Nonviolent Response	Condition 3: Violent Response†
[Another university is attacking religious freedom,” said George Lynn of the Christian Religious Liberty Defense Fund. “It’s just not safe to be a practicing Christian on campus in the United States anymore.”]	[“Anyone wishing to show their support for Christian Outreach should contact their state legislators or the Governor’s office,” said George Lynn of the Christian Religious Liberty Defense Fund. “Let them know that what’s happening at these schools is not right. Tell them that Arcadia needs to reverse this decision and that they need to vote for legislation that will prevent this sort of thing from happening again.”]	[Thomas Jefferson once said that the tree of liberty is fertilized by the blood of tyrants,” said George Lynn of the Christian Religious Liberty Defense Fund. “What would he say today? Next to the growing tyranny of America’s institutions of higher learning, the grievances Jefferson enumerated against the British crown almost look minor. King George III, for all his injustices, never forced the colonists to submit to the false gods of secularism.]
Control Condition: Fictional News Article about Martian Gas Fluctuations‡
The Curiosity rover on Mars recently reported some strange fluctuations in the level of oxygen on the Red Planet. As the Martian seasons change from warm to cold, carbon dioxide gas freezes at the planet’s poles. This causes air pressure to lower. Normally, other gasses like nitrogen and argon follow the same patterns. But for some unknown reason, oxygen apparently doesn’t. Surprisingly, it rocketed 30% during the warm months and dropped to back to normal in the fall. A study on the strange phenomenon was published in the Journal of Geophysical Research: Planets, but its authors are still puzzled by the data. After Sushil Atreya called the findings “just mind boggling,” people took to the web to joke about hibernating aliens. NASA has warned against such expectations before though because geological processes can affect gas production. But Melissa Trainer, another of the paper’s authors had this to say: “We don’t totally understand it.” She added, “For me, this is an open call to all the smart people out there who are interested in this: See what you can come up with.”

* Adapted from Starnes (2013) and Stetzer (2014). ^† Adapted from Schlafly and Neumayr (2012, 170). ^‡ Adapted from Strickland (2019).

.

6.3. The Dependent Variables

The first dependent variable (used to test H1) doubles as an independent variable (used to test H2 and H4) in the second phase of the analysis. Language for the variable, Religious Liberty Threat Perceptions, was adapted from a PRRI question about threats to religious liberty (Cox and Jones 2012). The question asks respondents to rate the degree to which they agree with the statement, “In America today, the right of religious liberty is being threatened.” Language for the second dependent variable, Support for Political Violence (used to test H2–H5), was taken from Wave 3 of the World Values Survey (World Values Survey Association 2015). Respondents are asked to rate the extent to which they agree with the statement, “Using violence to pursue political goals is never justified.”16 Responses to both questions range from “Completely Agree,” and “Agree,” on one hand, to “Disagree,” and “Completely Disagree” on the other, with a neutral category, “Neither Agree Nor Disagree,” situated in the middle. Responses to both questions were also converted to a 5-point Likert scale, coded such that higher scores correspond to higher perceptions of threat (e.g., 5 for “Completely Agree”) or higher levels of support for political violence (e.g., 5 for “Completely Disagree”).

6.4. Control Variables

As discussed in Appendix B, several additional questions from various sources regarding participants’ political beliefs and demographics were also included on the survey (Avdan and Webb 2018; Dennis et al. 2020; Funk and Strauss 2018; Gallup 2021; Pennycook et al. 2015; Pew Research Center 2021). They serve as the basis for a series of control variables employed in the regressions. Religious Attendance, Media Consumption, Education, and Income represent scales similar to the dependent variables. Religious Attendance, for example, is a 6-point scale ranging from 0 for low attendance at religious services to 5 for high attendance. Media Consumption is an 11-point scale indicating how many hours of news the respondent listens to or reads each week. Similarly, Education values closer to 5 indicate higher levels of education, while higher levels of annual household income are likewise indicated by scores closer to 5. Age is a continuous variable indicating a participant’s age in years, while Duration is a continuous variable that represents the amount of time in seconds that it took the participant to finish their survey. The rest of the control variables are dichotomous indicators for Conservatives (all others = 0), Republicans (all others = 0), Atheists and Agnostics (all others = 0), Catholics (all others = 0), Protestants (all others = 0), Other Religions (all others = 0), Born-Again Christians (all others = 0), White (all others = 0), and Male (all others = 0). Summary statistics and post-randomization demographic distributions are thoroughly described in

Table 3. Summary statistics and post-randomization demographic distributions.

	Min.	Max.	Mean	Std. Dev.	Variance
Religious Liberty Threat Perceptions	1	5	2.85	1.36	1.85
Support for Political Violence	1	5	1.75	1.02	1.06
Age	20	79	42.39	13.72	188.24
Religious Attendance	0	5	1.43	1.60	2.54
Media Consumption	0	10	5.20	3.31	10.97
Education	1	5	3.67	0.97	0.93
Income	1	5	3	1.31	1.72
Duration	103	3047	393.45	269.76	72,771.63
	Control Condition	Condition 1 (No Response)	Condition 2 (Nonviolent Response)	Condition 3 (Violent Response)	Full Sample (All Conditions)
Conservatives	51/206 (24.76%)	60/207 (28.99%)	71/207 (34.30%)	66/196 (33.67%)	250/816 (30.64%)
All Others (Liberals, Moderates, Others)	155/206 (75.24%)	147/207 (71.01%)	136/207 (65.70%)	128/196 (65.31%)	566/816 (69.36%)
Republicans	54/206 (26.21%)	60/207 (28.99%)	74/207 (35.75%)	66/196 (33.67%)	254/816 (31.13%)
All Others (Democrats, Independents, Others)	152/206 (73.79%)	147/207 (71.01%)	133/207 (64.25%)	130/196 (66.33%)	562/816 (68.87%)
Atheists/Agnostics	80/206 (38.83%)	62/207 (29.95%)	72/207 (34.78%)	61/196 (31.12%)	275/816 (33.70%)
Catholics	27/206 (13.11%)	45/207 (21.74%)	43/207 (20.77%)	40/196 (20.41%)	155/816 (19.00%)
Protestants	61/206 (29.61%)	51/207 (24.64%)	47/207 (27.71%)	52/196 (26.53%)	211/816 (25.86%)
Other Religious Adherents	38/206 (18.45%)	49/207 (23.67%)	45/207 (21.74%)	43/196 (21.94%)	175/816 (21.45%)
Born-Again Christians	38/206 (18.45%)	34/207 (16.43%)	41/207 (19.81%)	32/196 (16.33%)	145/816 (17.77%)
White	157/206 (76.21%)	162/207 (78.26%)	159/207 (76.81%)	145/196 (73.98%)	623/816 (76.35%)
All Others (Non-Whites)	49/206 (23.79%)	45/207 (21.74%)	48/207 (23.19%)	51/196 (26.02%)	193/816 (23.65%)
Male	100/206 (48.54%)	102/207 (49.28%)	102/207 (49.28%)	94/196 (47.96%)	398/816 (48.77%)
All Others (Female and Other)	106/206 (51.46%)	105/207 (50.72%)	105/207 (50.72%)	102/196 (47.96%)	418/816 (51.23%)

In the lower portion of the table, observation frequencies are shown over the full sample N, with the observation percentage shown in parentheses below. The full sample includes 816 observations; none are missing.

.

6.5. Analytic Method

Because the results of an analysis of variance (ANOVA) test can highlight statistically significant differences in groups’ survey responses, it can be used to compare each group’s mean score for the dependent variables. For this reason, a series of ANOVA tests are employed to examine the differences between experimental conditions in respondent answers to the religious liberty and political violence questions discussed above. However, to confirm that the results are driven by the experimental framing conditions rather than respondents’ attitudes or demographic characteristics, OLS regression with control variables is employed for a series of robustness checks.

7. The Effects of Persecution Discourse on Perceptions of Threat to Religious Liberty

The hypotheses above consider the effects of exposure to persecution discourse in the media on two different outcomes: perception of threat to religious liberty and support for political violence. The effects of such exposure on religious liberty threat perceptions are summarized first, in Figure 1 and

Table 4. Summary of experimental effects on religious liberty threat perceptions.

	N	Df	Sum Sq.	Mean Sq.	F
Full Sample	816	3	11.46	3.82	2.07
Conservatives Only	250	3	14.92	4.97	3.63 *
Non-Conservatives Only	566	3	0.25	0.08	0.05
Republicans Only	254	3	10.91	3.64	2.71 *
Non-Republicans Only	562	3	0.85	0.28	0.18
Catholics Only	155	3	5.64	1.88	1.23
Protestants Only	211	3	10.92	3.64	2.25
Born-Again Christians Only	144	3	6.19	2.06	1.42
Atheists/Agnostics Only	275	3	2.64	0.88	0.68

Results are from one-way ANOVA. * p < 0.05.

. Figure 1 consists of four boxplots that illustrate the effects of exposure to each of the separate experimental conditions on perceptions of religious liberty threat. Each box represents the interquartile range, with the thick line illustrating the median, while the whiskers plot minimum and maximum responses. Where the median disappears, it overlaps with the top or bottom of the box. In the control condition, for instance, it overlaps with the 25th percentile. The difference in the mean response between each condition is illustrated by the placement of the circles. The mean level of religious liberty threat perception for individuals exposed to an article about Martian exploration in the control condition is 2.67. Individuals in the other three groups were all exposed to an article about the banning of a Christian student organization on a secular college campus. The mean level of threat to religious liberty perceived by individuals who were not told how to respond to this putative attack on religious freedom rises slightly to 2.85. The mean response for individuals cued to respond with nonviolence rose to 2.92, while the mean response for individuals cued with Schlafly and Neumayr’s (2012) metaphorical appeal to violence rose to 2.99. Although this appears to represent a linear increase in means across conditions, the results of the ANOVA test listed at the top of Table 4 show that none of these differences can be considered significant. Thus, at least where the full sample is concerned, this suggests that persecution discourse does not drive perceptions of threat to religious liberty.

The results of the OLS analyses presented in

Table 5. Determinants of religious freedom threat perception.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Condition 1: No	0.176	0.118	0.0757	0.119	0.0840	−0.0320
Response	(0.134)	(0.118)	(0.112)	(0.112)	(0.111)	(0.129)
Condition 2: Non-	0.248	0.125	0.143	0.117	0.125	−0.0345
violent Response	(0.134)	(0.118)	(0.113)	(0.113)	(0.111)	(0.132)
Condition 3: Violent	0.320 *	0.175	0.155	0.179	0.163	−0.0459
Response	(0.136)	(0.120)	(0.114)	(0.114)	(0.113)	(0.134)
Conservative		0.741 ***	0.642 ***	0.557 ***	0.564 ***	0.105
		(0.158)	(0.151)	(0.152)	(0.150)	(0.220)
Condition 1 ×						0.469
Conservative						(0.251)
Condition 2 ×						0.594 *
Conservative						(0.245)
Condition 3 ×						0.733 **
Conservative						(0.246)
Republican		0.634 ***	0.464 **	0.544 ***	0.458 **	0.444 **
		(0.157)	(0.151)	(0.150)	(0.149)	(0.149)
Catholic			0.768 ***		0.562 ***	0.559 ***
			(0.118)		(0.133)	(0.132)
Protestant			0.991 ***		0.634 ***	0.635 ***
			(0.110)		(0.133)	(0.132)
Other Religion			0.670 ***		0.522 ***	0.522 ***
			(0.113)		(0.117)	(0.117)
Born-Again				0.417 ***	0.356 **	0.361 **
				(0.121)	(0.126)	(0.126)
Religious				0.181 ***	0.0935 **	0.0960 **
Attendance				(0.0292)	(0.0336)	(0.0335)
White		−0.357 ***	−0.296 **	−0.258 **	−0.255 **	−0.267 **
		(0.102)	(0.0976)	(0.0982)	(0.0971)	(0.0969)
Male		0.0687	0.114	0.124	0.134	0.141
		(0.0840)	(0.0803)	(0.0803)	(0.0794)	(0.0792)
Education		−0.171 ***	−0.172 ***	−0.164 ***	−0.164 ***	−0.163 ***
		(0.0473)	(0.0449)	(0.0453)	(0.0445)	(0.0445)
Income		0.00630	−0.00489	−0.00766	−0.00666	−0.00124
		(0.0351)	(0.0333)	(0.0335)	(0.0330)	(0.0329)
Age		0.00869 *	0.00426	0.00643 *	0.00443	0.00440
		(0.00337)	(0.00324)	(0.00322)	(0.00320)	(0.00319)
Media Consumption		−0.0368 **	−0.0326 *	−0.0324 *	−0.0309 *	−0.0296 *
		(0.0136)	(0.0130)	(0.0130)	(0.0128)	(0.0128)
Survey Duration		0.000104	0.0000180	0.0000280	−0.0000128	−0.0000210
		(0.000160)	(0.000152)	(0.000152)	(0.000150)	(0.000150)
Constant	2.670 ***	2.957 ***	2.673 ***	2.723 ***	2.596 ***	2.692 ***
	(0.0946)	(0.239)	(0.230)	(0.230)	(0.228)	(0.230)
Adj. R2	0.004	0.232	0.308	0.303	0.326	0.332
N	816	816	816	816	816	816

Results are OLS; standard errors are in parentheses. * p < 0.05; ** p < 0.01; *** p < 0.001.

substantiate this conclusion. Although the results of the ANOVA test do not suggest that the mean responses of any of the categories are significantly different from one another, the first model hints that individuals cued to respond violently are more likely to perceive a threat to religious freedom than are individuals who read the unrelated article about Mars. However, when other models integrate a range of control variables to confirm that these findings are not driven by respondents’ demographic characteristics, the effects of the violent response condition evaporate. Nevertheless, the results of the regression analysis appear relatively unsurprising. Conservatives are more likely than non-conservatives (i.e., liberals and “others”) to perceive a threat to religious liberty, even when controlling for exposure to the experimental conditions. The same goes for Republicans.

Meanwhile, Catholics, Protestants, and adherents of other religions are all more likely than Atheists and Agnostics (the omitted baseline religious group) to perceive a threat to religious liberty, even when controlling for exposure to different conditions. Similarly, born-again Christians are more likely to perceive a threat to religious liberty than individuals that do not identify as born-again. Age also appears to contribute to perceptions of religious liberty threat, but not after controlling for religious attendance, which appears to be a better predictor. By contrast, education and media consumption reduce the likelihood of perceiving such a threat, perhaps suggesting that increased political awareness can inoculate individuals from voter mobilization strategies that rely on religious liberty framing. Though previous research suggests that media consumption drives perceptions of threat to religious liberty (Goidel et al. 2016), this discrepancy could result from the fact that the experiment specifically accounts for exposure to persecution discourse. Because Goidel et al. (2016) find the opposite, the only surprise to come out of the regression analysis is that whites in the experiment are less likely than non-whites to perceive threats to religious liberty. Perhaps this is because non-white respondents are more likely to identify with marginalized religious traditions, such as Islam. Since race remains significant even when controlling for such religious identities though, perhaps this finding has something to do with the nature of the MTurk sample itself.

Although the findings above are largely unsurprising, they do not account for the possibility that the effects of persecution discourse, which is a product of conservative, Christian identity politics, may vary across subgroups. Since conservatives, Republicans, and born-again Christians are the targets of such discourse, and are known to perceive a greater threat to religious liberty than their political or religious counterparts, it is possible that they are more receptive to the religious liberty messaging in each of the three treatment conditions. The regressions in Table 5 appear to confirm as much. They show that conservatives exposed to the nonviolent and violent response conditions were more likely to perceive a threat to religious liberty than conservatives in the Mars control. Similar interactions exist for Protestants and Republicans exposed to the violent response condition. Although no similar effects were detected for born-again Christianity (omitted), this could be because so few MTurk users identify as born-again (17.7% of the present sample). At any rate, these findings are confirmed by the rest of the ANOVA results presented in Table 4 and further illustrated in Figure 2 and Figure 3, which show that persecution discourse alters perceptions of religious liberty threat for certain conservative Christian affinity groups.

8. The Effects of Persecution Discourse on Support for Political Violence

Next, the effects of persecution discourse on support for political violence are presented in

Table 6. Summary of experimental effects on support for political violence.

	N	Df	Sum Sq.	Mean Sq.	F
Full Sample	816	3	0.11	0.04	0.04
Conservatives Only	250	3	0.95	0.32	0.29
Republicans Only	254	3	0.11	0.4	0.03
Protestants Only	211	3	2.07	0.69	0.66
Born-Again Christians Only	145	3	1.95	0.65	0.70
Individuals Perceiving Threat to Religious Liberty	293	3	0.54	0.18	0.15

Results are from one-way ANOVA.

and illustrated in Figure 4. In brief, although research on the RFPT and dog-whistle politics led me to hypothesize that exposure to persecution discourse could lead directly or indirectly to increased support for political violence, there appears to be no evidence in support of such a relationship. The ANOVA results in Table 6 show that individuals exposed to persecution discourse are no more or less likely than individuals exposed only to the Mars control to support political violence, regardless of whether or how they are cued to respond to incidents framed as attacks on religious freedom. Indeed, this is true even when the sample is limited to conservative respondents, Republican respondents, Protestant respondents, born-again Christian respondents, or respondents who agree or completely agree that religious liberty is threatened in the U.S.

In the latter case, the fact that religious liberty threat perceptions exert no effect on support for political violence, either directly or indirectly through exposure to persecution discourse, contradicts H2. This means there is no evidence supporting the RFPT’s contention that individual-level perceptions are the causal mechanism responsible for the country-level relationship between religious freedom and political violence.

Furthermore, the fact that the violent response condition exerts no effect on support for political violence contradicts H3. This implies that leaders’ tacit appeals to violence are not sufficient to move individuals to act violently on their grievances about religious freedom. The fact that the violent response condition exerts no effect for even those people who perceive their religious freedom as being in jeopardy contravenes H4 and further undermines this possibility.

H5 fairs no better. Although they do not support political violence in general, there is no evidence to suggest that individuals cued to respond with nonviolent political action behave any differently than their counterparts.

Finally, the regressions included as robustness checks in

Table 7. Determinants of support for political violence.

	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
Condition 1: No	−0.0327	−0.0373	−0.0392	−0.0423	−0.0414	−0.0446
Response	(0.101)	(0.0986)	(0.0991)	(0.0986)	(0.0990)	(0.0990)
Condition 2: Non-	−0.0134	−0.0242	−0.0281	−0.0246	−0.0230	−0.0277
violent Response	(0.101)	(0.0991)	(0.0994)	(0.0991)	(0.0994)	(0.0995)
Condition 3: Violent	−0.0119	−0.0135	−0.0169	−0.0215	−0.0216	−0.0277
Response	(0.103)	(0.100)	(0.100)	(0.100)	(0.100)	(0.101)
Religious Liberty						0.0375
Threat Perception						(0.0316)
Conservative		−0.161	−0.153	−0.141	−0.136	−0.157
		(0.132)	(0.133)	(0.134)	(0.134)	(0.135)
Republican		0.282 *	0.274 *	0.288 *	0.278 *	0.261
		(0.131)	(0.133)	(0.132)	(0.133)	(0.133)
Catholic			0.0912		0.103	0.0823
			(0.104)		(0.119)	(0.120)
Protestant			−0.0284		0.0442	0.0205
			(0.0971)		(0.119)	(0.120)
Other Religion			−0.127		−0.107	−0.127
			(0.0995)		(0.105)	(0.106)
Born-Again				−0.175	−0.170	−0.184
				(0.106)	(0.113)	(0.113)
Religious				0.0112	−0.000249	−0.00375
Attendance				(0.0257)	(0.0300)	(0.0301)
White		−0.0936	−0.108	−0.0948	−0.111	−0.102
		(0.0856)	(0.0860)	(0.0862)	(0.0866)	(0.0870)
Male		0.349 ***	0.335 ***	0.341 ***	0.330 ***	0.325 ***
		(0.0703)	(0.0708)	(0.0705)	(0.0708)	(0.0709)
Education		0.0228	0.0215	0.0164	0.0162	0.0223
		(0.0396)	(0.0396)	(0.0398)	(0.0397)	(0.0401)
Income		−0.0372	−0.0400	−0.0395	−0.0419	−0.0416
		(0.0293)	(0.0294)	(0.0295)	(0.0295)	(0.0295)
Age		−0.0126 ***	−0.0125 ***	−0.0124 ***	−0.0125 ***	−0.0127 ***
		(0.00282)	(0.00285)	(0.00283)	(0.00285)	(0.00286)
Media Consumption		−0.00649	−0.00687	−0.00659	−0.00713	−0.00597
		(0.0114)	(0.0114)	(0.0114)	(0.0114)	(0.0115)
Survey Duration		−0.0000359	−0.0000178	−0.0000241	−0.00000690	−0.00000642
		(0.000134)	(0.000134)	(0.000134)	(0.000134)	(0.000134)
Constant	1.767 ***	2.246 ***	2.287 ***	2.277 ***	2.313 ***	2.216 ***
	(0.0718)	(0.200)	(0.203)	(0.202)	(0.204)	(0.219)
Adj. R2	−0.004	0.058	0.060	0.059	0.060	0.061
N	816	816	816	816	816	816

Results are OLS; standard errors are in parentheses. * p < 0.05; *** p < 0.001.

confirm these non-findings. Although young MTurk users, and users identifying as Republican or men, are generally more likely to support political violence than older users, or users not identifying as Republican or men, exposure to persecution discourse has no direct effect on support for political violence. Nor does it appear to have an indirect effect on support for political violence through its ability to increase perceptions of threat to religious liberty. In fact, while the variables for the treatment conditions are insignificant, their effects do not even point in the theorized direction.

9. Conclusions

Because the RFPT suggests that religious repression leads to violence, and because conservative Christians in the U.S. are increasingly likely to perceive their rights to practice their religion freely to be threatened by their government, this study set out to understand if American Christians could be at increased risk of supporting political violence. However, research on religious liberty in the U.S. suggests that exposure to persecution discourse in the media shapes religious liberty threat perceptions, while the literature on elite cueing and dog-whistle politics shows that leaders can provoke their followers to engage in both violent and nonviolent forms of political activism. As such, this study also sought to determine if religious liberty threat perceptions could motivate individuals to action independent of a leader’s call to violence, or if exposure to dog whistles was required to move individuals to defend their religious freedom violently.

To these ends, I articulated five hypotheses. Following the American politics literature, I first hypothesized that exposure to persecution discourse would increase perceptions of threat to religious liberty. The first section of results discussed above tested this hypothesis by comparing mean levels of religious liberty threat perception across experimental conditions. Consistent with previous work on framing (Djupe et al. 2016) and work on religious liberty in American politics (Goidel et al. 2016), I found that while conservatives were generally more likely to perceive religious freedom as under threat than non-conservatives, they were especially likely to perceive threats to religious freedom if they were exposed to persecution discourse (as compared to conservatives that instead read an article about Martian exploration). While perceptions of threat to religious liberty among MTurk users generally are not impacted to the same degree, additional evidence indicates that persecution discourse exerts similar effects on the perceptions of Republicans, and certain religious groups, such as born-again Christians.

Next, since country-level research on international religious freedom suggests that religiously free countries may be less prone to spasms of political violence than religiously unfree countries, the remaining four hypotheses attempt to plumb the possibility that individual-level perceptions of religious freedom operate according to the same logic, independent of country-level characteristics. They do so by investigating the direct and indirect effects of exposure to persecution discourse on individuals’ willingness to support political violence in the U.S., one of the most religiously free countries in the world. Yet, there is no indication that persecution discourse affects individuals’ willingness to support political violence, either directly or indirectly through altered religious liberty threat perceptions. There is a similar lack of evidence suggesting that such perceptions exert an independent influence on support for political violence, just as there is no evidence that appeals to nonviolence reduce support for violence. Resultantly, there is no support for the remaining four hypotheses.

Of course, it is possible that the null results in the second part of the analysis are a product of weaknesses in the experimental design. For example, a sample drawn from a non-American or non-Christian-majority population or drawn from multiple countries might reveal a different story. Then again, perhaps the violent metaphor included in the treatment was inappropriate to cue support for violence in this case because respondents appropriately read it as hyperbole. If so, future research on language and violence may need to account for differences in the effects of contextually germane and contextually inappropriate dog-whistle appeals to violence. That is, issue area may matter. If the travails of a Christian student organization are not germane, a similar experiment invoking reproductive health, restrictions resulting from the Covid-19 pandemic, or violent anti-religious harassment, could yield different results. After all, the events at the U.S. capitol in January 2021 certainly demonstrate that American conservatives can be incited to violence under the right conditions. The nature of those conditions just needs to be properly elucidated.

With that in mind, future research may find it appropriate to adopt a new measure of religious liberty threat perception. Rather than asking an agree-disagree question and converting the results of such a question to a small Likert-scale, researchers interested in investigating the individual-level implications of the RFPT may instead want to ask about the degree to which a respondent agrees that religious freedom is threatened. A feeling thermometer ranging from 0 for completely disagree to 100 for completely agree would allow for greater variance in the data and any resulting analysis employing such a measure may find different results.

Alternatively, MTurk tends to consist disproportionately of nonreligious workers, meaning born-again Christians especially—a group likely to perceive a greater degree of threat to religious liberty—are underrepresented in the sample. Said another way, the size of the subsample of born-again Christians participating in the experiment may not yield enough power to determine how exposure to persecution discourse really affects their willingness to support political violence. This seems unlikely since this was not a problem for the first analysis. However, the mean score for religious attendance (1.43) is also quite low. Religiosity is much higher for the general public, so perhaps the disproportionate inclusion of nonreligious individuals in the sample biased the results. Future experimental research investigating this link could avoid this possibility though by using the more costly TurkPrime (now CloudResearch) to recruit a larger population of more religious people, including a greater number of born-again Christians (Lewis et al. 2015; Mullinix et al. 2015).17

Of course, it is also possible that the single exposure to persecution discourse presented by the experiment is incapable of matching the effects of the kind of iterative exposure conservative media consumers are likely to encounter in real life. Longitudinal research might therefore challenge these findings. Additional areas of concern might include the varying word length of each treatment in the faux news article presented to the participants, the experimental treatments’ relative position in the article, or even the type of elite presented in the treatment as attempting to shape respondents’ behaviors.

Nonetheless, there is nothing presented above evocative of a relationship between exposure to persecution discourse and support for political violence. Consequently, it remains conceivable that the null results in the second phase of the analysis are more indicative of weaknesses in the theory that motivated this study. Given the expectations of the RFPT, this is both striking and important for the direction of future research on religion and political violence. Since previous research on the RFPT consistently finds support for the contention that religious freedom reduces political violence at the country-level, this suggests that more work is needed to theorize the mechanisms that facilitate this relationship. At present, however, the RFPT hinges on the idea that encounters with a religiously repressive state or society at the individual-level, be they real or imagined, are what erode the religious freedom peace. The null findings in the second phase of the analysis simply do not support this contention.

Yet, this is not to say perceptions do not matter. For instance, misperceptions of persecution could be operative at the group-level as opposed to the individual-level. Regardless, the record of violence carried out by Timothy McVeigh (Smith 2015), along with such groups as Aum Shinrikyo in Japan (Reader 2011), or even Jim Jones’ Peoples Temple in Guyana (Guinn 2017; Moore 2011), still points to a place for misperceptions of threat to religious liberty in the lead up to political violence. Students of the RFPT must do more to investigate what that place may be though. Perhaps further research will show that some individuals and groups that are already predisposed to committing violence merely appropriate and instrumentalize religious liberty—as may happen with religion more generally (e.g., Pape 2006)—to legitimate extremist agendas and violent activity. If true, the spread of persecution discourse might still be a cause for concern as violent or potentially violent non-state actors seek to position themselves as defenders of Christians’ (or other religious groups’) religious freedom, both within and beyond the American border.

Already, groups such as the Oath Keepers, a rightwing, anti-government militia, and the more diffuse, anti-abortion terrorists who identify as the Army of God, have signaled a willingness to do precisely this. First, both routinely adopt persecution discourse framing in posts published to their respective websites (e.g., Army of God n.d.; Codrea 2015). Second, when Kim Davis, a former Kentucky county clerk, came to national attention for her refusal on religious grounds to comply with the Supreme Court’s Obergefell decision legalizing same-sex marriage, she was charged with contempt of court, resulting in an offer by the Oath Keepers to protect her and prevent her incarceration by force of arms (Rhodes 2015). For its part, the Army of God website recently published a Turner Diaries-esque “novel” by Eric Rudolph (2017) called Enemies, Foreign and Domestic, which uses religious freedom framing to justify the bombing of the San Francisco LGBTQ Pride Parade.18

From a normative perspective, it might be tempting to represent the findings discussed above as “good” news, at least insofar as they fail to show increased support for political violence at the individual-level. Since public support for violence creates space for actual instances of violence to occur (Tessler and Robins 2007), this study might have foreshadowed serious domestic security problems if the findings had shown that exposure to persecution discourse triggers support for political violence. Yet, by reminding us that social movement messaging can be appropriated by potentially violent sympathizers (LaFree and Ackerman 2009), the preceding discussion limns the need for further research by hinting at the existence of an alternative causal pathway for persecution discourse to influence violent actors at the group-level.

The results above notwithstanding, further research is also needed because an alternative causal pathway implies a risk of contagion beyond the borders of the U.S. After all, similar misperceptions of persecution have been documented or studied in several other religiously free, historically Christian-majority, Western nations, including the United Kingdom, Australia, and Canada (A Boost for Believers 2019; Moulin 2016; Vang et al. 2019; also see Shah et al. 2016). Although it is nowhere as thoroughly developed as it is in the U.S., these same countries have already seen the emergence of their own nascent discourses of Christian persecution (e.g., Hodge [1993] 2015; Hutchinson 2017; Meritus 2015).

Then again, if the link between religious freedom and political violence is itself spurious (see Muchlinski 2014), such concerns could be moot.