Three questions help to introduce the concept of civil religion:  What is civil religion?  Does it exist?  And, what does it do?  Indeed, from previous conceptualization and research, it appears most people hold civil religious views although they probably would not know what civil religion is if they were asked to define the term.  Nor would they likely know how their civil religiosity effects what else they believe and how they behave.  Such is the case, we surmise, for the role of one's civil religiosity and the choice he or she makes for a U.S. presidential candidate. 

    What is civil religion?  For our conceptual and operational purposes in this research, we attempt to follow the definition proposed by Robert Bellah in his influential article, "Civil Religion in America," where he reintroduced the concept into social scientific and religious scholarship in 1967 and which he reprinted and further elaborated a few years later (1970) and elsewhere (e.g., 1974) in his work.
 

    Accordingly, he states, "There are…certain common elements of religious orientation that a great majority of Americans share…that provide a religious dimension for the whole fabric of American life, including the political sphere…[and] is expressed in a set of beliefs, symbols, and rituals that I am calling the American civil religion (Bellah 1967: 3-4).  Later, he (Bellah 1970: 168)  adds, "I conceive of the central tradition of the American civil religion…as the subordination of the nation to ethical principles that transcend it and in terms of which it should be judged."   He (Bellah 1974: 255) further elaborates that through its civil religion, "…The nation is not an ultimate end in itself but stands under transcendent judgment and has value only insofar as it realizes…a 'higher law'."

    The basis for civil religion in America can be traced as far back as the Declaration of Independence.  It opens with reference to "the laws of nature and of nature's God" and goes on to say that,

We hold these truths to be self evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness. 
That to secure these rights, Governments are instituted among Men, deriving their just Powers from the consent of the governed….

    The basic civil religious principles from the Declaration of Independence are echoed in President Lincoln's Gettysburg Address where he states:

Four score and seven years ago our fathers brought forth on this continent, a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.
Now we are engaged in a great civil war, testing whether that nation, or any nation so conceived and so dedicated, can long endure.   
…That we here highly resolve that these dead shall not have died in vain—that this nation, under God, shall have a new birth of freedom—and that government of the people, by the people, for the people, shall not perish from the earth.

    From the civil religious basis of the Declaration of Independence, Bellah (1980: 11) sees the foundations of American civil religion in reference to "…a suprapolitical sovereignty, to a God who stands above the nation and whose ends are standards by which to judge the nation and indeed only in terms of which the nation's existence is justified, [that] becomes a permanent feature of American political life ever after."  In addition, Bellah surmises, "The existence of this highest level of religious symbolism in the political life of the republic justifies the assertion that there is a civil religion in America."

    The idea of civil religious transcendence characterizes societal beliefs and symbols held in common by the citizenry and aid in promoting social solidarity and political legitimacy in a society.  These beliefs, symbols, and even rituals are used to associate the values, history, and institutions of the country with divine favor or even with a national religious destiny.  Wimberley and Swatos (1998:94) note that, "Civil religion is unique in U.S. culture—and arguably in other cultures as well—in that it does not claim an identifiable social group short of the entire society itself." 

    Overall, from the conceptualization of civil religion, we see that, first, a civil religion transcends our country's sectarian or denominational religions as well as its partisan political orientations.  Second, and according to this civil religion, the moral and ethical standards of the nation's people are ultimately derived from a higher authority. Third, the nation's government is justified by the higher authority through its people.  Fourth, people's rights and other national blessings come from the higher power.  And fifth, the nation stands under the higher standards and judgment of the divine presence.

Does civil religiosity exist in the United States?  In addition to historical references to the ideas of civil religion expressed in the Declaration of Independence, presidential addresses since the time of George Washington, and other documents (see Bellah 1967; 1970; 1980), empirical research from surveys of Americans show a dimension of civil religious beliefs apart from other religious and political orientations (Wimberley 1976; 1979; Wimberley et al. 1976).  Furthermore, this empirically measured dimension of civil religious beliefs is found to cut across socioeconomic distinctions (Christenson and Wimberley1978) and denominational religious affiliations (Wimberley and Christenson 1981), but it does not infringe upon beliefs in the separation of church and state (Wimberley and Christenson 1980).  The research, for example, reveals that civil religious beliefs are indeed widely shared across denominational identities and are found most in such American-based groups as Mormons, Adventists, and Pentecostals and least among Unitarians, Jews, and people without a religious preference. 

What does civil religion do?  Although empirical research to date is limited, two previous studies examine the potential political effects of civil religiosity.  One of these (Wimberley 1980) finds that people's civil religious beliefs appeared to influence their preference for candidate Richard Nixon over George McGovern in the 1972 presidential election.  Another study (Wimberley and Christenson 1982) finds that civil religiosity typically fares as well or better than other religious, political, and socioeconomic indicators of people's public policy preferences.  Only occupational status and age tended to match or exceed civil religiosity's correlations with support for programs in health and welfare programs, transportation, education, and for controlling crime and addictions, or for opposition to environmental and arts programs. 

    Since the contemporary concept of civil religion was introduced in Bellah's 1967 Daedalus article and beginning with the first known empirical research by Thomas and Flippen (1972) who analyzed evidence of civil religion in newspaper editorials, there has been relatively little empirical research in comparison to the amount of conceptualization and commentary on civil religiosity.  Since the beginning of the 21st Century, however, the study of civil religion in America has experienced a resurgence not seen since the 1980s.
 
    The bulk of this recent attention includes further elaboration on the conceptualization of civil religiosity (e.g., Cristi 2001; Marvin 2002; Santiago 2009) or is limited to commentary and other articles using civil religion as a backdrop (e.g., Silk 1999; Adams 2003; Will 2004; Benne 2005; Meizel 2006; ; Hecht 2007; Butterworth 2008).  In the popular press, some (e.g., Meacham 2006) may not use the concept per se but provide historical context for the development of civil religion in the United States or view what is taken to be its misuse in a more contemporary national context (e.g., Phillips 2006).  Relatively few of the academic works (e.g., Flere and Lavric 2007; Zeljko 2008) that make up the newer wave of civil religious writings are actually based on empirical research.  As in earlier decades, therefore, most recent studies take civil religiosity as a given without delving into any empirical evidence of what it is or does.  Still, calls continue for more empirical work on the outcomes of civil religion in America (Swatos 2006).   

Civil religion and the presidency.  Overall, the conceptualization underlying the pattern of theory for civil religion is extensive.  And although the empirical research on civil religion is limited and by no means extensive, the evidence to date strongly supports the theoretical claims that civil religiosity exists apart from other political and religious commitments, party partisanship, denominational affiliations, and socioeconomic differences, and that civil religion matters in the way people perceive the presidency and policy orientations.

    Earlier, Wimberley (1980) suggested that voter preferences for particular presidential candidates were a theoretically appropriate place to look for the influence of civil religion in the United States. Many of the presidential icons of American civil religion have revealed multiple civil religious roles.  The saintly Washington, the prophetic Jefferson and Wilson, and the martyred Lincoln and Kennedy are all remembered as such in the collective American memory.  But first and foremost, these presidents were elected leaders of the American people.  As such, they became the high priests of American civil religion's amalgamation of Judeo-Christian and political beliefs.  As Bellah (1967: 4) describes it, the religious legitimacy of their political authority is marked by the rituals of the inauguration.

    The role of a U.S. president as the nation's civil religious leader is portrayed by several religious scholars including Marty, Hammond, and Bellah.  Marty (1974: 145-147) describes the most common civil religious role of a president as that of a priest with ceremonial duties for the country.  A less common role is that of a prophet who stands in judgment over the nation when it does not meet its ideals.  Hammond (1976: 177) explains that a president is a civil religious symbol of the country who perform in a priestly role.  Similarly, Bellah (1967: 8) sees a president as a "national Magistrate" when he serves in his official civil religious role.  Regardless of the specifics of the role, each analyst views the president as a religious leader—a civil religious leader—in addition to his role and responsibilities as the nation's political leader.  That the president may be implicitly viewed as the nation's civil religious leader helps to account for the potential importance of the civil religious factor in the minds of the electorate during presidential elections.

     Civil religion, the presidency, and Reagan.  American religious scholar David Adams (2003:23) suggests that the public expression of civil religion in America has been shaped by three presidents who he refers to as civil religion's three greatest theologians.  In the context of American civil religion, he likens Thomas Jefferson's role to that of Moses, Abraham Lincoln to Jesus, and Ronald Reagan to Paul.  Adams also points out that although each of these presidents were religious, none were active in church religion.  This might be expected from a transcendent civil religious leadership perspective that rises above sectarianism.  Furthermore, Adams notes, each took a public role as a moral leader and helped to voice and frame the direction of the nation during their terms.

    Most people may not consciously think of Reagan as the metaphorical successor to Lincoln as the sacrificial American Jesus, although many may be able to view Reagan in a civil religious missionary role analogous to Paul's spreading of the gospel.  Others (Mathisen 1989) further suggest that Reagan offered a priestly form of civil religious leadership that may have been more acceptable than the sometimes prophetic civil religion of his predecessor Jimmy Carter (Pierard and Linder 1988).  In any event, there was obviously something about Reagan that induced a large number of people to re-elect him to the presidency in 1984 with nearly 60 percent of the popular vote.
 
    The term Reagan-Democrat was used by the media in an attempt to explain the reason for the size of the popular and electoral-vote victories of Ronald Reagan's re-election.  In the Reagan years prior to that election, a large number of voters who had customarily identified with the Democratic Party began to defect to the Republicans in national elections while remaining registered as Democrats (Luebke 1990; 1998; Greenberg 1996). 

    An explanation for this may lay in Wimberley's 1980 study of the 1972 presidential election. That study of the 1972 Nixon-McGovern presidential race found that civil religious beliefs of the study's community and church-attender samples were better predictors of candidate support than were political party loyalties.  However, the effect of civil religious beliefs in that study may have been tempered by the fact that one of the candidates, Nixon, was an incumbent president.  As such, he may have been regarded by the voters—although largely unaware of the underlying concept—as having civil religious legitimation by virtue of already holding the office of president.  As in other presidential elections involving an incumbent, this may have diminished the public's civil religious perception of George McGovern and given an advantage to Richard Nixon.  Similarly, as an incumbent, Ronald Reagan may have had a civil religious advantage over Walter Mondale in 1984. 

Implications for Testing

    Now, as in the past, most discussion of American civil religion tends to be theoretical, if not philosophical or historical.  Also, the analytical work that has been done is rarely empirical.  Therefore, the analysis to be reported here offers an empirical contribution to the study of civil religiosity in the United States.  It examines the relationship between civil religion and the public's perception of the American presidency.  This study looks at several hypotheses that link civil religion to the electorate's preferences for presidential candidates.

    As reviewed previously, theoreticians and conceptual analysts such as Bellah (1967) and Marty (1974) emphasize the importance of civil religion in the public's perception of the presidency.  Wimberley's (1980) research found that in the Nixon-McGovern presidential race of 1972, civil religious beliefs were a major factor in predicting greater voting preferences for incumbent Richard Nixon.  Although that study of civil religion's influence in the Nixon-McGovern election showed a civil religious disposition toward voting for Nixon, it measures neither people's perceptions of each candidate's civil religiosity nor how the voters own civil religiosity corresponds with their perceptions of each candidate's civil religiosity (Wimberley 1980:  57-58).

    Presidential voting choices.  Therefore, several hypotheses are proposed to extend the previous line of research.  First, as in the study of the 1972 Nixon-McGovern presidential race, it is hypothesized that people who are more civil religious will tend to have a preference for one candidate rather than the other and, furthermore, will prefer an incumbent candidate due to that incumbent's civil religious legitimacy attached to that position.  Second, it is predicted that people will prefer the presidential candidate they perceive to be the more civil religious.  And, third, people will prefer the presidential candidate who is perceived to hold civil religious beliefs that are more similar to their own civil religious beliefs.

    Sample.  This study draws from the third in a line of longitudinal community studies on political and religious commitment in Raleigh, North Carolina.  The first was conducted prior to the 1972 Nixon-McGovern election (e.g., Wimberley 1976; 1980).  The second study took place in the weeks before the 1976 presidential election (e.g., Wimberley 1979).  In each wave of this series, the same community sample frame, similar data collection techniques, many of the same attitudinal, behavioral, and background items,  and similar analysis techniques were repeated.

    One of these longitudinal studies measured several civil religion variables unique to the analysis to be conducted here.  Therefore, for the present study, data was collected from voting-aged adults during October and early November just prior to the 1984 presidential, senate, house, state, and local elections.  To draw the sample, the 1980 U.S. Census tracts for Raleigh were ranked according to their percentage of high school graduates.  A randomly selected tract on this ranked list was taken as a starting point, and eleven tracks were selected systematically throughout the list.  These eleven selected tracts were the basis for random samples of city blocks.  Volunteer data collectors worked their ways around each of these blocks, starting at one corner and stopping at every third residence to drop off a questionnaire.  The data collectors returned at a later time to pick up the questionnaires.  Along with this drop-off and pick-up method of distribution, efforts were made to match the race of the volunteer data collectors with the predominant race residing in each tract (Bell 1989:7).  These two actions may help account for the excellent return rate of 98 percent of those with whom questionnaires were given.  A total of 349 respondents were obtained by the time data collection had to be terminated immediately before election day. 
 
    The questionnaire was constructed to draw on the religious and political dispositions of the respondents along with basic demographic and socioeconomic background information.  Political variables covered such things as party affiliation and preference, political inclinations, and voting behavior.  These variables were mirrored in the religious items (Wimberley 1978) that covered religious affiliation, preference, inclinations, and behaviors.  The demographic and socioeconomic data collected included gender, race, occupation, age, marital status, geographic history, income, and education.

    Presidential candidate preference.  The dependent variable at issue is the respondents' choices between the majority parties' presidential candidates, Democrat Walter Mondale and Republican Ronald Reagan.  A total of 42 respondents who preferred neither candidate, or who were undecided, and those not stating a preference were omitted from this analysis.  This narrows the presidential analysis sample to 307 members.  In other words, no missing data assignments were made to the dependent variable. To measure the presidential candidate preferences, the 169 choosing Mondale are coded as 0; the 138 respondents for Reagan are scored as 1.

    Civil religious index.  Among the items on the questionnaire were a series of Likert-type, agreement-response items that had been developed with various samples in previous studies (Wimberley et al., 1976; Wimberley 1976; 1979; Christenson and Wimberley 1978) and found to be useful indicators of civil religious beliefs.  The items used here are: 

    Principal axis factoring shows that these five items converge to suggest a single civil religious dimension.  Because only one factor was extracted from this set of statements, the factor was not rotated.  The items' factor loadings on this dimension were all fairly high and ranged from .540 to .755.  In the order these five items are presented in the previous paragraph, these loadings are .540, .666, .586, .620, and .755 respectively.  Cronbach's alpha reliability for these items is .771 and in the conventionally acceptable range for an index of so few items.    

    In each of the statements, the agreement alternatives are coded on a continuum from 0 = Strongly Disagree to 1 = Disagree, 2 = Uncertain, 3 = Agree, and 4 = Strongly Agree. Missing data for each item were assigned the item's mean. The values of these items are summed to create a civil religious index.  The summed index scores range from 0 to 20 and provide an index running from 0 for the lowest level of civil religious beliefs when a respondent strongly disagreed with all of the items, to the highest level and score of 20 representing strong agreement on all of the civil religious belief statements.
 

    Perceptions of the candidates' civil religiosity.  A second civil religion measure is based on each respondent's perception of the civil religiosity of the two major candidates.  The items that comprise this measure are unique to this wave of the community-study data and are essential to testing the last two hypotheses in this analysis.  After asking a respondent how he or she personally reacted to the five civil religious index items described previously, the respondent was then asked, "Regardless of your own opinions," which presidential candidate would be "more in agreement" with each civil religious measurement item?  The candidate a respondent picked more often is scored as the one the respondent perceived to be more civil religious. 

    If the candidate who received the most choices was Mondale, the candidate-perception measure was scored as a -1.  If Reagan received the most choices, the score assigned was +1.  In other words, the candidate who is perceived to be more civil religious on the most items is assigned a non-zero score.  The use of negative or lower scoring values to designate Mondale and positive or higher scoring values to designate Reagan helps to assure that such low to high scores will correlate or regress positively with the dependent variable's candidate preference scores that are also coded as low for Mondale and high for Reagan.  This system will be used in the scoring for several other independent variables in this analysis. This helps to make positive and negative correlation and regression coefficients more directly and consistently interpretable in the analysis to follow.

    If there was missing data on as many as four of the five perception items, the candidate who received a majority of civil religious perceptions on the item or items a respondent did answer are scored as the candidate perceived to be more civil religious.  A score of 0 indicates that the respondent did not answer either one or three of the perception items and that the candidates were tied at 2:2 or 1:1—having answered a total of four or two, respectively, of the five perception items—on the remaining items.  Because there is an odd number of five items used to measure civil religious beliefs, other numbers of tied items are not possible.  If a respondent failed to check any of the five perception items, this response pattern is classified as missing data for the entire candidate perception measure.

    Respondent-candidate congruence on civil religiosity.  To test the third hypothesis, a final measure of civil religiosity is created here to show the correspondence or congruence, between each respondent's personal civil religiosity and the perceived civil religiosity of each of the two major candidates.  In other words, the congruence measure indicates the candidate with which a respondent agrees more on civil religiosity.  This third measure of civil religiosity is based on the respondents' agreements and disagreements with the five civil religious statements described earlier plus the five items on how respondents perceived the civil religiosity of the major presidential candidates. 

    Similar to the scoring technique for the candidate-perception measure, the candidate with whom the respondent is congruent on a majority of the civil religion belief statements is the candidate who is scored most congruent.  For example, if candidate Mondale was congruent with a respondent on most of the civil religious beliefs, the congruence measure is scored as -1.  When Reagan was congruent with a respondent on most of the items, the score is +1.   

    To be congruent, a respondent's own civil religiosity on an item has to match the perceived civil religiosity of a candidate in regard to the same belief statement.  For example, a respondent would be congruent with Reagan on a particular civil religious belief item if he or she "Agrees" or "Strongly Agrees" with that item and also thinks that Reagan is more civil religious than Mondale on that item.  A second way in which a respondent could be more consistent with Reagan would be when he or she "Strongly Disagrees," "Disagrees," or is "Uncertain" on one of the civil religion items, but thinks Mondale would be more in agreement with the item.  In either case, the respondent-candidate congruence favors Reagan.

    If a respondent did not answer a specific civil religion item or did not answer a candidate perception question for that item, this was treated as missing data.  However, the congruence score can still be established from the majority of the matches on any remaining items a respondent answered.  When there was missing data on all five item-pairs, the case was coded as having missing data for the entire congruence measure. 

    Religious and political control variables.  Other questionnaire items provided information on additional political and religious positions of the respondents.  These items include both behaviors and beliefs that are to be used as controls in statistical analyses of the hypotheses in order to help ensure the civil religious effects are not due to other such factors.
 
    Corresponding political and religious dimensions of behaviors and beliefs were found in previous analyses of data from one of the earlier longitudinal surveys of the community sampled here (Wimberley 1978).  From those findings, measures of church membership, church attendance, and political party preferences were adopted for use in this study, along with items measuring liberal-conservative political beliefs and liberal-conservative religious beliefs.  The conceptual basis of these measures may be traced back to the rites, beliefs, and groups sharing them as described by Durkheim ([1912] 1965:51-53). 

    The political belief and religious belief measures in this study have parallel, three-point, 0, 1, and 2 codes for response categories that run from "Liberal" to "Moderate" and to "Conservative" respectively.  Whether one is a church member is a measured simply as "No" or "Yes" and is scored as 0 or 1. Church attendance is measured as three categories in which "never" and "a few times a year" are the low category, "at least once per month" and "every week" are the mid category, and "more than once a week is the highest category.  A three-point political party preference variable is scored into its implicit categories:  Democrats were assigned the lowest coded value, those claiming "neither party" received the mid-value, and Republicans were coded highest.

    Analysis categories.  Several of the independent variables inherently have three-point value levels given the nature of the measures upon which they are based.  As it turns out, this helps put the independent variables on a similar footing and helps make their regression coefficients more comparable in their interpretations.  As indicated in the preceding variable descriptions, civil religious perceptions of the candidates, respondent-candidate congruence on civil religiosity, political party identification, political conservatism, and religious conservatism are examples.  In addition, the civil religion index readily collapses into scores for agreement, disagreement, and mid-level stands on civil religious statements.  And, as shown among the control variables described above, church attendance readily lends itself to three ordinal rankings. 

    Consequently, all of the independent variables except church membership—which is naturally a yes-no dichotomous variable—are measured as three-point variables. 

    Basic relationships.  Statistical correlation analyses are a beginning point for assessing any connections between civil religiosity and presidential voting preferences.  The first and most basic question is whether there is a relationship between a respondent's civil religiosity and the preference for a particular presidential candidate?  The answer from the bivariate correlation analysis is yes.  The Pearson's r correlation coefficient for the data analyzed here is .277 with a statistical significance exceeding .001. 

    Therefore, the basic, bivariate correlation between one's personal civil religiosity and the preference for a particular presidential candidate is found to hold with this data in absence of any control variables.  This is consistent with the findings from earlier correlation and regression analyses of the Nixon-McGovern race (Wimberley 1980).  In the present analysis of the Reagan-Mondale race, Ronald Reagan more often attracted the potential voters who scored highest in civil religiosity.

    Given that civil religious people do prefer one candidate more than the other, why is this the case?  Although this connection between civil religiosity and candidate preference exists and favors Reagan, it is not immediately apparent whether this is because Reagan was perceived to be the more civil religious candidate.  In other words, is there an empirical relationship between the perceived civil religiosity of the presidential candidates and the preference of one candidate versus the other?  Judging from a correlation of .317 having statistical significances less than .001, the answer is again yes.  This suggests a stronger link with Reagan as the candidate most often perceived to be more civil religious.  Furthermore, the congruence between the respondent's civil religiosity and his or her perception of the candidates' civil religiosity is also notably strong at .416 and again statistically significant at less than .001.
 
    Basically, Reagan tends to have been chosen by those who are more civil religious, those who perceive Reagan to be more civil religious than Mondale, and those who feel compatible with Reagan's perceived civil religiosity.  The question remains, however, whether these three zero-order correlations will hold up when each relationship is analyzed while controlling for the effects of the remaining two civil religious measures and other control variables in regression analyses.
 
    Regression models.  To further test the hypotheses in the presence of controls for the effects of other political and religious variables that may statistically influence preferences for presidential candidates, a series of six regression models are formulated to look at the relationship between civil religiosity and candidate choice in presidential elections.  The models start with the three civil religious measures as the independent variables in Model A and successively add a religious or political control variable to the predictions of candidate choice in each succeeding model.  Since candidate preference is a dichotomous nominal dependent variable—i.e., Mondale or Reagan—these regressions are analyzed by logistic regression techniques rather than ordinary least squares regressions in order to avoid violating the assumption of homoscedasticity in the statistical relationships. 
 
    Model A.  In the first regression model tested, Model A of Table 1, only the three independent variables representing the civil religion measures are tested.  Of these three predictors of relationships with presidential candidate choice, only two are statistically significant.  Specifically, the respondents' perceptions of the candidates' civil religiosities and the congruence of the respondents' own civil religiosity with this perception are statistically significant when the effects of the other two civil religious indicators are held constant, or controlled, in the logistic regressions.  Perceptually, the presidential candidate who certain respondents see as the more civil religious is about 3.4 times more likely to be favored.  Respondents having civil religious views more congruent with a particular presidential candidate are about 3.7 times more likely to vote for that candidate. 

    Accordingly, in Model A, the first hypothesis that the more civil religious respondents will prefer one candidate more than the other—does not hold when the effects of the candidates' perceived civil religiosity and the congruence between the respondents and candidates are held statistically constant.  The other two hypotheses, however, are supported by highly significant regression coefficients.  We shall also see this pattern of support for hypotheses two and three repeated in all the other regression models to be analyzed here. 

    Model A also has a Nagelkerke R² of .390.  The Nagelkerke R² (also known as Cragg & Uhler's R²) attempts to estimate a pseudo R² for logistic regression that is analogous to the R² used as a goodness-of-fit measure in ordinary least squares regression. As such, this simulated version of an R² varies from 0.0 to 1.0 with higher values indicating better model fit.  If this was an ordinary-least-squares regression, a value of .390 would mean that approximately 39 percent of the statistical variation in the candidate choice is explained by these three independent variables.  However, these pseudo R² statistics should not be interpreted the same way as an ordinary-least-squares R² and are only included here to give a general indication of goodness-of-fit between the models.

    Model B.  The regression impacts of civil religious perceptions and congruence remain about the same as found in Model A, once religious beliefs are added to the regression equation as a control variable.  The influence of religious liberalism-conservatism is also highly significant statistically.  The predictive clout of religious ideology, as estimated by its logistic regression odds coefficient, indicates that conservative believers were about 2.2 times more likely to prefer Reagan while liberals favored Mondale at the same rate.  Although religious belief is a statistically significant predictor of presidential candidate preference in Model B, the predictive ability of the respondents' own civil religiosity index scores remains flat as first noticed in Model A, and as will be observed in the other presidential-candidate regression models.  With the inclusion of religious beliefs into this four-predictor regression model, the Nagelkerke R² increases to .425. 

    Models C and D.  Introducing the additional control variables of church membership and church attendance in Models C and D respectively show little improvement in the predictive power of the civil religious measures representing the three hypotheses.  While statistically significant at the .05 level, the odds coefficient related to church membership approaches 1.0, indicating it has a very small effect on candidate choice. Once church attendance is brought into the equation, neither of the church variables or religious beliefs have a significant effect. This is reflected in the Nagelkerke R² rising ever so slightly to .434 in Model D.
 
    Model E.  When political beliefs (liberal or conservative) are introduced into the analysis in Model E, it is indicated that political conservatives were nearly 7.0 times more likely to favor Reagan just as liberals were more likely to prefer Mondale.  The lack of statistically significant influences of liberal-conservative religious beliefs, church membership, and church attendance suggests that in absence of the political beliefs variable as a control from Models B, C, and D, the three church-religious factors were essentially carrying and masking the influence of political beliefs. 

    However, the presence of the political belief variable in Model E does not detract from the predictive power of the candidates' perceived civil religiosity or the civil religious congruence of the respondents and candidates.  The Nagelkerke R² for Model E is .559   

    Model F.  Only four of the independent variables are statistically significant in the final model:  Perceived candidate civil religiosity, respondent-candidate civil religious congruence, political beliefs, and a new control for the respondents' political party affiliation.  Party affiliation has a logistic regression coefficient of almost 6.0.  This means one is about six times more likely to favor the candidate of his or her own party.  This coefficient consequently provides a standard of comparison for other three-point predictor variables in the models.  Although it comes as no surprise that political party identity is a strong indicator of candidate preference in Model F, the predictive power of candidates' perceived civil religiosities, civil-religious congruence, and of political beliefs is only slightly lowered by the presence of partisan identity in the Model F equation.  
  
    Model F, the full model containing the three civil religious predictors, the church-religion controls and the political control variables, produces a Nagelkerke R² of .693.
 
    Overall, political partisanship is the major predictor of presidential voting preferences and political liberal-conservative beliefs also factor heavily into the equations for choosing between presidential candidates.  None of the traditional church-religious variables have a statistically significant effect in the full model. The important finding for the hypotheses tested here, where the possible effects of the civil religious variables are tested here along beside church-religious and political control variables, is that two of the three civil religious variables are found to show significant and appreciable effects on presidential voting preferences. 

    Both the civil religious perceptions of the candidates and civil religious congruence between the candidates and the respondents are statistically significant predictors at the .001 level in final Model F and in Models A through E that precede it.  These significant findings first suggest that, as measured here, people—regardless of their own civil religious commitment—may be about three times more likely to vote for a presidential candidate they perceive to be more civil religious.  Secondly, people may be about two and one-half times more likely to favor a presidential candidate who they believe is more in line with their own type of civil religious views.

    Also, these findings suggest that the traditional religious indicators of church membership and church attendance have only small and statistically insignificant effects on presidential voting choices when civil religious and political factors are taken into account.  The standardized, log-odds coefficients—shown in Table 1 in parentheses—yield the same interpretations as the logistic regression coefficients discussed for these analysis models. 

    What is the role of civil religion in presidential elections?  Judging from our correlation and logistic regression analyses, civil religion does have a role in the politics of choosing a president in the United States.  The evidence shows that civil religious variables do correlate with choices between presidential candidates.  And when customary political and religious predictors of candidate preference are added as controls on civil religiosity's impacts, certain civil religious factors retain their importance.  In other words, the impacts of civil religion are not to be explained away entirely by other types political and religious factors commonly used to predict voter choices in presidential elections.
 
    More importantly, these results show that one's perception that a presidential candidate is more civil religious than another has a major impact on who a potential voter prefers to be president.  Basically, a person strongly tends to prefer the presidential candidate whom he or she believes to be the most civil religious.  This finding is even more impressive, according to the analyses at hand, in that it does not matter so much about one's own civil religiosity.  Regardless of one's own civil religious beliefs, major church-religious orientations or political orientations, one is very likely to favor the candidate seen as the more civil religious and/or more congruent with one's own civil religious beliefs.  This research suggests that prospective voters may be about three times more likely to prefer a presidential candidate who is perceived to be more civil religious and about two and one-half times more likely to favor a presidential candidate who is perceived to be congruent with the voter's own civil religiosity.

    The statistically significant predictors of a prospective voter's choice for a presidential candidate also includes the prospective voter's liberal-conservative political beliefs and political party affiliation.  For example, a politically conservative voter is more likely to choose a politically conservative candidate, and a voter who identifies with a particular political party is more likely to choose a candidate of the same party.  But regardless of the voter's own level of civil religiosity, he or she is more likely to prefer the presidential candidate who is seen to be more civil religious and who is civil-religiously congruent with the voter.

    Furthermore, this study suggests that it is not just religion in general that voters want to see in a presidential candidate.  Rather, it is how the candidates are perceived to believe in civil religion and to have the same civil religious beliefs as the voters that is important to many voters.  Beliefs that a divine being transcends the denominational or sectarian religions in the nation, justifies the existence of the nation, is the ultimate authority and basis for the nation's laws and freedoms, may provide guidance for the nation, and has the power to judge and bless or punish the nation.

    The evidence at hand suggests that civil religion's impact on candidate choice overwhelms the influences of church-related religious identity, behavior, and beliefs.  In other words, this research finds the perception of a candidate's civil religiosity may be what primarily drives voting choices, instead of church-religious factors such as church membership, attendance, or liberal-conservative religious beliefs that are customarily measured in studies of voting choices.

    Also, when political liberal-conservatism and political party affiliation are introduced into the statistical analysis as controls on church membership, church attendance, and religious beliefs, these church religious factors lose their clout in predicting one's preference for a presidential candidate.  For example, the role of religious conservatism fades in the presence of one's political conservatism.  But the importance of civil religious perceptions of the candidates and voter-candidate congruence on civil religion are not explained away by other religious or political factors however important the latter are in the statistical analyses reported here.

     We suggest that civil religious findings of this study have implications for other presidential elections that follow.  The pattern of findings observed in the present research fit with the social scientific conceptualization of civil religion (e.g., Bellah 1967) and fit with other empirical research on civil religiosity to date (e.g., Wimberley 1980).  Historically, we know that presidents beginning with George Washington have made civil religious appeals in their public speeches (e.g., Bellah 1967).  This current study helps connect these appeals with the public's civil religious perceptions of other candidates for the office of U.S. president.
 

    Much has been made of voters' moral values in recent presidential elections, but essentially nothing is said about one of the most important American political and religious values:  civil religion.  We do not know the role of civil religion in more recent presidential races, though moral values were depicted by exit-poll data to have been a major factor in determining the race between Bush and Kerry (Cable Network News 2004). 

    To our knowledge, however, no other research has collected data on civil religiosity of the voters other than the Nixon-McGovern race in 1972, or the perceived civil religiosity of presidential candidates or the voter-candidate civil religious congruence in any elections before or after Wimberley's 1984 study.  Therefore, the empirical answer to the role of civil religion in other presidential elections has not been gauged, and it remains unknown if not unknowable. 

    Of course, this analysis reports the results of just one study of one U.S. community during one presidential election.  Party partisanship is the mainstay indicator in other studies and voting polls in presidential elections.  Any factor that compares favorably to voters' political party affiliations in predicting preferences for presidential candidates is a concept to be taken seriously.  In this case, a candidate's perceived civil religiosity and voter-candidate congruence in civil religiosity remain statistically significant with notable regression impacts after the effects of political partisanship and liberal-conservative political beliefs are controlled.   

    More importantly, the results of this study are consistent with what has been conceptualized about civil religion in general and conceptualizations of civil religion and the presidency in particular (e.g., Bellah 1967; 1970; 1974; 1975; 1980; 1998; Marty 1974), plus with what has been found in the limited body of empirical research on civil religion and political behavior that led to this study (Wimberley 1980; Wimberley and Christenson 1982).  Therefore, we posit that civil religion has played a significant and substantive role in more recent presidential elections and will likely do so in future presidential elections. 

    We hope that new research and theoretical refinements will continue to explore, predict, and explain civil religion's relative importance in comparison with other, customary types of political and religious beliefs and behaviors used to study the electorate.  Given the potential importance of presidential and other voting choices, the conceptualization and research of civil religiosity to date may be just the start.  More generally, unless there is more empirical, social-scientific research on the concept of civil religion, it risks collapsing from the weight of its own extensive conceptualization and claims.  Civil religion's empirical role in voter decisions among prospective candidates appears to be a good place to continue.

Footnotes

(1) Based on a paper presented at the meeting of the Southern Sociological Society,  Richmond, VA, April 2008.  The authors especially appreciate the assistance of Druscie Simpson of the North Carolina State Archives, who helped facilitate retrieving the original electronic data file for this project; Alisha Curry who assisted in the latter phases of the research; and Dale Wimberley and Steve McDonald who offered suggestions on the interpretations of the analysis.  The authors, of course, are responsible for the research and interpretations involved.  

(2) A Brief Explanatory Note about  "The Role of Civil Religion in Choosing a President."   In many ways, this article is a missive from the past.  A little known fact about Ronald Wimberley was that in the 1980s he became the leading quantitative expert in the study of civil religion. Even after the fashion for civil religion lessened in sociology, he maintained an interest in it that he shared with me as a doctoral student.  After my graduation, we slowly began to envision a series of articles looking at voting patterns using datasets that he and I had gathered.  These datasets were (and are still) unique in the questions asked and this article was to be the first of the series. We planned to follow it with others dealing with the influences of civil religion, political beliefs, the media, and the opinions of friends and family on national and state elections.

    Ronald Wimberley had told me that he had made his final revisions to this paper before he died in 2011, but events prevented him from ever sending it to me. In the months after, his family and I unsuccessfully attempted to find the revision, and a series of electronic failures and file losses on my part led me to give up on finishing the article.  However, this past spring when I changed offices at my university, I found a printed copy of one of the last versions. There are things that I would want to do differently now, but this paper is essentially what we planned five years ago. As I have gone through it, I am surprised how much of it I hear in his voice.

References