Religion’s Varied Role from Election to Election
In this chapter, I use an analytical strategy called multilevel modeling to analyze how the religious rhetoric unique to a candidate influences vote choice. These models can be complex;2 1 however, the logic underlying this statistical method is a fairly straightforward way to test for identity priming and emotive rhetoric in campaigns. Specifically, we are interested in whether the strength of the relationship between religiosity and change in candidate evaluation is related to systematic differences in candidate rhetoric.
Consider figure 5.1. If we ignore whether the lines are dashed or solid, figure 5.1 looks like little more than a random array of lines. But this figure actually provides valuable information about the role of religion in campaigns. Aligned along the y axis are the survey respondents’ changes in candidate evaluation for all presidential candidates from 1980 to 2004. A positive value indicates that the respondents evaluations of the candidate improved over the course of the campaign, whereas a negative value indicates that the voters came to see the candidate less favorably. Along the x axis I have plotted individuals’ levels of religious commitment. Thus, if religiosity was never activated by campaign dynamics (i.e., if it made no difference in the evaluation of candidates), we would expect to see fourteen horizontal lines. To be sure, people’s opinions of candidates might get better or worse over the course of a campaign
FIGURE 5.1 Religious commitment and change in candidate evaluation by level of civil religion rhetoric, 1980-2004
Relationship between religious commitment and change in candidate evaluation for all Republican and Democratic presidential candidates from 1980 to 2004. In most cases, more religious voters showed a larger, more positive change in their evaluation of candidates when the candidates frequently invoked American civil religion than did less religious voters.
(as reflected in the y intercepts of these horizontal lines). But a line with a slope of 0 for any candidate would indicate that the religiously committed did not display any more (or less) change in opinion than the uncommitted and that, in all likelihood, religious factors were not activated by campaign rhetoric.
Figure 5.1 should make it clear that religiosity did play a markedly different role from one campaign to the next. For example, consistent with conventional wisdom, religious voters came to see George W. Bush much more positively as the campaign progressed—suggesting that Bush’s rhetoric changed perceptions in a way that resonated with voters. In other words, while Bush generally fared poorly over the course of the campaign, the slope of the line for Bush is steeper than for Kerry, suggesting that Bush successfully activated religious commitment as an evaluative criterion. While Bush’s lower overall numbers could be due to any number of factors (related to foreign policy, educational policy, et cetera), the damage was abated considerably among religiously committed respondents. Interestingly, the findings in figure 5.1 also rule out the possibility that the effect of religion on candidate evaluation is limited to a candidate’s partisanship. For example, all else being equal, George H. W. Bush actually fared worse with religious than non-religious voters over the course of the 1992 campaign (a negative slope), whereas Clinton garnered comparatively more favor with the religiously committed as the election progressed (a positive slope). Keeping in mind that the y-axis is change in candidate evaluation, figure 5.1 illustrates that campaigns do not universally reinforce existing partisan allegiances between Republicans and the religiously committed.
In technical terms, my central purpose in this chapter is to explain the variance in the slopes of the lines (Raudenbush and Byrk 2002). The question is: Why do religiosity committed (or orthodox) voters warm up to some candidates over the course of an election but sour on others? Or stated differently, what activates religious cleavages in some elections? Multilevel models allow us to address precisely this question by bringing order to the cacophony of lines in figure 5.1 by modeling the varied slopes as a function of candidate rhetoric. Accordingly, I have also distinguished the lines in figure 5.1 by the level of civil religion rhetoric the candidate used in that campaign. Although the figure breaks candidate rhetoric into only two very rough categories (low and high), a clear trend is nevertheless evident. When candidates used high levels of civil religion rhetoric (solid lines), the more religiously committed voters changed their evalu?ation more favorably than did less religious voters (i.e., the slope is positive). When candidates failed to do so (dashed lines), the less religiously committed voters changed their evaluation more favorably than did more religious voters (i.e., the slope is negative). (The exception is Gore in 2000, who had the highest level of civil religion rhetoric in the “low” category.) Thus, the level of candidate rhetoric can be used to meaningfully interpret the relationship between religious commitment and candidate evaluation. The data indicate that, when candidates use religious language, voters respond in predictable and politically powerful ways.