Convention, Annenberg/Pew Archive of Presidential Campaign
These passages bear certain similarities. Both passages, for example, use a personal story (Clinton’s time at Georgetown and the birth of his daughter, respectively) to connect with the audience. In addition, both passages make a reference to God in support of a particular social agenda. Nevertheless, these passages are quite distinct in terms of the religious identities Clinton is evoking.
In the first passage, Clinton uses a religious reference to illustrate his connections with a particular religious group—Catholics—in a speech at the Catholic University of Notre Dame. Indeed, later in this same speech Clinton takes the time to illustrate the commonalities between the Baptist (his own religious affiliation) and Catholic outlook. This passage is indicative of religious subgroup identity rhetoric, or appeals to distinct religious subsets within the population, such as Catholics, Jews, specific Protestant denominations, and religious interest groups such as Focus on the Family.
There are several reasons to suspect that subgroup identity cues may be important in contemporary politics. First, subgroup references are uniquely equipped to resonate with specific segments of the electorate. The United States is characterized by tremendous denominational pluralism (Putnam and Campbell 2010), and the degree to which politicians make reference to these diverse groups is significant. Insofar as individuals’ religious group affiliations are an important determinant of their voting behavior, politicians have an incentive to align themselves with religious groups through their rhetoric. In addition, even though subgroup appeals are distinct from culture wars rhetoric, they may have a similar impact. By reminding people of their denominational affiliations, the appeals may make voters inclined to more readily think about religious outgroups. We know very little about whether subgroup rhetoric is typically used in the spirit of religious favoritism, or, as in the Clinton example, in the spirit of inclusivity.
Of course, there are potential costs associated with subgroups appeals. Although it is unlikely that Clinton’s single speech at Notre Dame alienated non-Catholic voters, sustained appeals to religious subgroups could have precisely this effect. For example, because about 20 percent of Americans are non-Christian, we can imagine that regularly invoking Christianity would seriously alienate these voters, accentuate tensions between religious subgroups, and suggest that the candidate might deliver representation with an explicitly Christian outlook.1 Thus, although subgroup appeals have the potential to create a close bond between a candidate and a narrow constituency, these appeals are also potentially problematic, defining the political community in an exclusive manner.
In contrast, the second Clinton passage contains no reference to a particular religious subgroup. Rather, the passage uses religious language to ascribe a certain sacred quality to America. Clinton is evoking a shared “faith” that in America everyone has the ability to achieve to the extent that abilities are “God-given.” The passage constitutes a subtle ingroup appeal insofar as Clinton uses first-person plural pronouns (“our”) and references to America to build a sense of shared group cohesiveness in the audience. The implied group is unmistakably religious in nature because the egalitarian ideal posited by Clinton is inextricably linked with a sense of faith. As Michael Billig (2003) has argued, ambiguous references to shared identity combined with references to universal principles such as equality and freedom frequently operate to establish a sense of oneness among all message recipients. And, of course, in the American case the universal principles posited by Billig are often religious in nature. America is often portrayed as a “new Israel,” endowed by God with special responsibilities and given a favored place in the world order (Fowler et al. 2004; Domke and Coe, 2010). These tenets are central to Americans’ self-understanding, providing, as Robert Bellah has written, “a religious dimension for the whole fabric of American life, including the political sphere” (1967, 3). Civil religion is thus intertwined with identity, and this identity gains expression through political rhetoric.
Note that the second Clinton passage does not live up to all the criteria of Bellah’s civil religion—Clinton does not specifically say that America has a special place in the world order. Clinton does, however, implicitly assert a shared belief in God and identifies “America” as a place where people have faith. Moreover, Clinton takes pains to draw his audience together in a shared concern with phrases such as “our cause” and “family and friends.” In short, the second passage, in contrast with the first, unites the audience members—and all Americans—in a common cause that has something to do with religious faith, even if the substance of this faith is somewhat unspecified.
Culture wars rhetoric stands in sharp contrast to both subgroup appeals and the inclusive language of civil religion. Reagan’s 1984 campaign made a point of employing this genre in the context of the debate about allowing prayer in public schools, arguing, “If our opponents were as vigorous in supporting our voluntary prayer amendment as they are in raising taxes, maybe we could get the Lord back in the schoolrooms and drugs and violence out.” Several features of this passage stand out. First, it implicitly references two competing sides of an explicitly religious debate. In saying “our opponents,” Reagan does not call the outgroup or the ingroup by name, instead arguing that there is some other unspecified “other” in American politics, determined to remove God from the schools (and raise taxes). Moreover, by yoking the idea of school prayer together with drugs and violence, Reagan is escalating the stakes in the debate, indicating that the absence of religion in schools is responsible for a fundamental moral breakdown.
Although it is clear that religious identities are evoked in political rhetoric, we know relatively little about how often candidates use these identities, how this has changed over time, and how these identity cues serve the interests of building a cohesive and inclusive political community. And, although it might be tempting for some to draw the simple conclusion that civil religion is “good” for democracy and that subgroup and culture wars appeals are “bad,” this oversimplification would be a mistake. Religious identity rhetoric can have crosscutting and ambiguous consequences. For example, although subgroup appeals could call religious differences to mind, they could also serve an important democratic function by recognizing underserved and marginalized religious groups.2 Likewise, although civil religion is often theorized to serve important democratic functions (Hart 2005), it is certainly possible to imagine this civil religion manifesting itself as something particular and exclusive. To the extent that civil religion rhetoric privileges a particular view of the sacred, members of excluded religious and spiritual traditions may find themselves rhetorically marginalized, not only from a religious group but also from a particular conception of American identity.
Understanding Religious Identities in Political Rhetoric
To analyze how religious language is used in politics, I began by collecting a large sample of stump speeches from major party candidates, 1980-2008. There are several reasons why this data source is appropriate for capturing the overall thrust of campaign rhetoric. First, it is important to consider that I am analyzing rhetoric over multiple campaigns.
Mediums of campaign communication have changed radically in the past thirty years, with a rise in cable television and Internet use and a decline in newspaper circulation (Baum and Kernell 1999; Putnam 2000; Kaid 2003). Candidate stump speeches, in contrast, have been a relatively constant staple during this period. Scott Althaus, Peter Nardulli, and Daron Shaw (2002) find evidence of only a slight increase in candidate appearances from 1972 to 2000. Moreover, candidates have behaved in strategically similar ways during this time period (Althaus, Nardulli, and Shaw 2002, 69). This supports my using stump style speeches as a data source in that candidates can be compared on roughly even ground. Second, I avoided media interpretations of campaigns because news reports are not necessarily accurate reflections of campaign discourse; for example, media reports often accentuate negativity (Jamieson and Waldman 1997). Finally, because the later chapters in this book link candidate rhetoric with opinion changes over the months preceding a campaign, I needed to develop an indicator to capture the overall thrust of a campaign rather than episodic campaign events (Druckman 2004). For this purpose, my measuring the overall campaign climate by examining a large sample of speeches (as opposed to, say, the content of a single convention address) is preferable given what is known about preference formation. For example, James Druckman and Kjersten Nelson (2003) find evidence suggesting that the impact of a single political event is fairly short-lived.
To assess the overall thrust of a candidate’s rhetoric, I began by gathering a large number of stump speeches from a variety of sources: (1) the Annenberg/Pew Archive of Presidential Campaign Discourse (2000), (2) the Stanford University Political Communication Lab (2000), (3) the University of California-Santa Barbara American Presidency Project, and (4) a Lexis-Nexus search of news wire transcripts of presidential campaign speeches.3 In all, I collected 1,329 speeches made over the course of 16 campaigns.
From this large campaign speech database, I selected a sample of unambiguously religious passages that is the unit of analysis for all subsequent discussion in this book. The immediate objection to this approach is that that what seems like religious rhetoric to some might not be religious rhetoric to others. For example, Bethany Albertson (2005) has found evidence that the biblical phrase “wonder-working power” works as coded political rhetoric to persuade narrow subsets of Christians, even though this reference goes over the heads of most Americans. Although this finding supports the contention that religious rhetoric resonates with certain segments of the electorate, it also suggests that the task of identifying religious language is mired in subjectivity. To overcome this challenge, I focused on religious rhetoric that is explicitly religious. The choice to focus on explicit references is a design feature that clearly limits the inferences I ultimately could draw from these data; however, this sacrifice was necessary to reliably analyze the large volume of campaign speeches.
I define explicitly religious as any passage containing a set of religious indicator words—words that indicate or signal religion to the vast majority of Americans. Fortunately, the content analysis program Linguistic Inquiry and Word Count (LIWC) contains a predefined set of religious words that have already gone through a rigorous selection process (Pen- nebaker, Francis, and Booth 2001), suggesting that these indicator words are well recognized as religious.4 This list—sixty explicitly religious indicator words in all—represents all potential indicators of religious rhetoric. From this list of indicator words, I eliminated words that appear infrequently, words that are religious only in certain contexts, and words that often appear as jargon. For example, I eliminated minister because it is usually not an indicator of religious rhetoric (e.g., prime minister or foreign minister). In other words, I developed a set of indicators that reliably mark only those passages that have meaningful religious content. These cuts yielded a list of twelve religious rhetoric indicators.5 Using these indicators, I selected the entire passage in which that word appeared, using a set of linguistic rules to ensure that passages were sampled in the appropriate context.6
I performed several checks to ensure the adequacy of this sampling procedure. First, using a computer content analysis program, I compared samples drawn using different selection procedures. The samples were indistinguishable along key dimensions such as emotive language and pronoun use (Pennebaker, Mehl, and Niederhoffer 2002). Second, I qualitatively examined complete speeches to make sure this sampling procedure did not exclude any obviously religious passages. Overall, this procedure did a good job capturing the religious content of a speech. In fact, religious words that are not indicators themselves (such as minister) are still generally included in the sample because they typically appear in close proximity to one of the keyword indicators. This selection procedure resulted in a sample of 3,495 religious passages. From this religious rhetoric sample, I drew a random probability sample, clustered by candidate, to make the task of content analysis more manageable.
I randomly selected 75 passages per candidate, yielding a sample of 1,200. This sample is the basis of all subsequent analysis in this book.
I content-analyzed each of the 1,200 passages along several dimensions relating to identity to test my expectations about the contours of the political community (Neuendorf 2002).7 First, each passage was coded for ingroup and outgroup references. My use of ingroup and outgroup borrow from the social psychological concepts by the same name. Essentially, ingroups are an individual’s own social group attachments, whereas outgroups are social “others.” The coding scheme used in this book is intended to capture the rhetorical analogs of these social psychological concepts—the cues that can bring ingroup and outgroup distinctions to the “top of the head” (Zaller 1992). Ingroup references were examined further to uncover the specific social identities to which candidates were appealing—be they religious (faithful Christians), partisan (faithful Republicans), or geographical (faithful Wisconsinites). While I analyzed religious rhetoric for numerous identities, two specific identity references are of particular interest in this chapter: Civil religion identity and Subgroup identity.8
The Culture wars variable is a subset of all outgroup references. References to outgroups were coded to discern whether the reference to the outgroup signaled a deep-seated conflict in American politics or only primed religious differences in a superficial way. In addition, only references to American political conflicts were coded as culture wars rhetoric. For example, even though the Soviet Union was the target of religious outgroup rhetoric (particularly in the early 1980s), this, if anything, was a point of unity for Americans, not evidence of a culture war.
I also coded whether candidates characterized religion as something shared by all Americans or as something personal and a matter of individual choice—essentially whether American religiosity is framed in Shared or Pluralistic terms. Religious identities have the potential to be either inclusive or exclusive. Rhetorically, candidates can emphasize the unique aspects of multiple faith traditions (or no faith tradition at all) or posit a shared American faith that is broad enough to encompass all the essential aspects of American religious diversity. Accordingly, my coding scheme assessed whether each religious passage implied a common or shared religion or God in America, or discussed religion in a way that is more pluralistic or relative.
A Favored nation variable was included to test the expectation that the language of civil religion identity is closely connected to many of the belief elements identified in the civil religion literature. One of the more common assertions made is that America as a kind of new Israel, “endowed . . . with special opportunities, and assigned . . . special responsibilities to do good” (Fowler et al. 2004). Given these expectations, all passages were coded to determine whether they made no reference to country, linked religion and country without ascribing a favored-nation status to the United States, or explicitly indicated that the United States is blessed in some way (Domke and Coe 2010).
Finally, a God concept variable tests the extent to which civil religion rhetoric privileges a particular view of the sacred. The relationship between civil religion and subgroup identities has long been a source of debate in the academic literature. Some scholars see civil religion as integrative, representing a distillation of the common elements of multiple religious faiths so as to serve as a values touchstone on which American democracy rests (Wimberley and Christenson 1981). Others, including Bellah (1967), view civil religion as transcending all subgroup beliefs, positing that a divine power is directly influencing American national affairs.
What is clear is that no matter how civil religion is conceptualized, it is generally not theorized to privilege a particular religious subgroup. Nevertheless, there is reason to suspect that civil religion rhetoric might implicitly do exactly this, particularly because adherence to civil religion ideals seem to be related to denominational traditions (Wimberley and Christenson 1981). One way to test this is to take advantage of the fact that, although most large religious traditions in America have a monotheistic belief structure, they often conceptualize God in very different ways. Methodists, for example, tend to think of God as being more distant than do Catholics or Evangelicals. Evangelicals, in contrast, think of God as being more vengeful (Noffke and McFadden 2001). George Lakoff (1996) has suggested that conservatives tend to think of God in terms of a “strict parent,” whereas liberals tend to think of God with “nurturing” imagery (see also Jensen 2009). Paul Froese and Christopher Bader (2010) convincingly argue that Americans generally understand God in terms of how engaged God is the world and whether God is judgmental. These differences in God image are related to both denominational traditions and political differences.
Although God concept has been widely studied, there is relatively little guidance for reliably analyzing speeches for the image of God they project. Accordingly, I analyzed religious rhetoric for seven different con?ceptualizations of God deemed to reasonably approximate the variety of God images in the academic literature. God concept codes for “companion,” “paternal,” “judge,” and “savior,” given evidence that these images predict sociopolitical attitudes (Welch and Leege 1988). To this list I added a “maternal” God concept (Roof and Roof 1984), as well as a category for God as “provider.”9 An additional “no imagery” category was also added. To the extent that any one of these is rhetorically more prominent than the others, this will suggest that the divine elements in civil religion are more closely associated with certain religious traditions and that civil religion contains subtle elements of religious exclusivity.
Speaking of Faith in Politics
The results of the content analysis provide strong confirmation of the centrality of identity in religious political rhetoric.10 To begin, it is helpful to get a sense of how often candidates used religious language in stump speeches in general (as opposed to looking directly at the sample of religious rhetoric passages). Figure 3.1 shows the LIWC religious word
FIGURE 3.1 Frequency of religious rhetoric in presidential campaigns
Linguistic Inquiry and Word Count religious word scores for all major party presidential candidates, 1980-2008.
scores for each campaign. Several things about figure 3.1 stand out. First, although candidates are varied in their use of religious language, this variance is somewhat underwhelming, with religious words accounting for between approximately one-fifth and two-fifths of a percentage point of all total words employed in every campaign (note that the у-axis scale in figure 3.1. ranges from 0 to 1 percent).11 This is similar to other findings in political science. For example, in analyzing political rhetoric in presidential campaigns over the past fifty years, Roderick Hart (2000, 48) finds relatively little systematic variation in the frequency of religious language use among major party candidates.
There still are some noteworthy differences between the candidates in frequency of religious word use. It is particularly interesting to note that the variation illustrated in figure 3.1 violates much of the conventional wisdom about the use of religious rhetoric in presidential campaigns. For example, John Kerry in 2004 actually scores higher than George W. Bush in 2004 on religious word frequency, a finding that at first blush is out of step with the widely held perception that Bush was the more regular religious communicator. Likewise, Walter Mondale’s 1984 campaign had the highest rate of religious language use in the sample—a surprising finding, given that candidates such as Jimmy Carter and George W. Bush are much more well known for their public religiosity. This is further evidence that a nuanced and theoretically driven look at the language of religious identity is necessary to account for electoral and public opinion dynamics. It is not how often candidates talk about religion that matters—it is how this rhetoric is crafted.