The Effects of Religious Rhetoric on Candidate Evaluation
To capture a candidate’s religious rhetoric style, I examined the effects of the subgroup,12 civil religion, and culture wars variables from chapter 3, along with the LIWC anger and anxiety scores and a scaled measure of candidate enthusiasm from chapter 4.1 3 Averaging these measures over the courses of an entire campaign provided a good estimate of the thrust of the candidate’s rhetorical approach to religious identity rhetoric and emotive language (see Druckman 2004).14
To examine the impact of religious campaign rhetoric on individual preferences, I pooled seven American National Election Surveys (the 1980, 1984, 1988, 1992, 1996, 2000, and 2004 ANES) from 1980 to 2004.15 The use of ANES data is appropriate for several reasons. First, the survey is administered in the context of the presidential election campaigns in our sample, and it contains both pre- and postelection questionnaires. Thus, it is possible to examine how individual attitudes have changed over the course of the campaign. Second, the ANES contains several measures of individual religiosity. As suggested earlier, persuasion can work indirectly by priming latent considerations, but only among those who already have some form of religious attachment. It is thus important to examine the variation in the extent to which religious predispositions become predictors of attitude change, as a function of religious rhetoric.
To truly capture the effects of religious rhetoric on the political process, it is not enough to simply measure the evaluation of a candidate at one particular time. Voters enter the election season with a host of preconceived notions about candidates; what is ultimately of interest is the extent to which a candidate’s rhetorical choices shape the basis of these beliefs. Accordingly, I examine the candidate assessments of survey respondents after the election while statistically controlling for preelection feelings. This allows me to examine how well campaign rhetoric can predict changes in individuals’ feelings toward candidates over the course of electoral campaigns.16
The theory of identity priming suggests that different dimensions of religiosity should be activated or deactivated depending on the rhetorical thrust of a given campaign. I examined two specific dimensions of interest. Religious commitment was measured as church attendance and the importance of religion in an individual’s life.17 Given that religious commitment is a nondenominational measure of religious adherence, I expect that civil religion rhetoric should activate religious commitment as an evaluative dimension. Of course, religious commitment is a highly imperfect approximation of civil religion identity; however, given the need for a consistent measure of nondeminational religiosity across several elections, religious commitment is the best available indicator. Second, I expect that subgroup and culture wars rhetoric should activate Religious orthodoxy. The ANES is somewhat limited in its ability to consistently measure religious orthodoxy spanning 1980-2004.и Therefore, I replicated a measure of religious orthodoxy developed by Layman (2001, 78-87) that involves using religious tradition to rank individuals’ denominational orthodoxy and then further differentiated respondents based on church attendance.
The model contains several control variables of theoretical importance in assessing campaign effects. I controlled for Partisan identification, Race, and Gender because each of these variables can ultimately exist as other identities that can be activated or deactivated over the course of a campaign. I also controlled for the Date of the survey interview because individuals surveyed two months before an election have theoretically been exposed to less religious rhetoric than those surveyed the day before the election.19 In addition, I controlled for Media exposure because religious rhetoric could disproportionally affect those individuals most exposed to religious messages across levels of religiosity (Zaller 1992).20