Abstract
The authors examine the effect of religiosity on intended and recalled voter turnout in presidential elections. They argue that the trade-off between time spent in worship and time spent in political activities, specifically voter turnout, is strongest for mainline Protestants, weaker for Catholics, and nonexistent for evangelical Protestants. Evangelical Protestants increasingly recognize the connection between their religious beliefs and politics, with the result that they have formed a habit of voting. This argument has important implications for American voting behavior literature. Going beyond partisan voting patterns, the findings demonstrate that evangelical Protestants manifest unique patterns as they relate to turnout patterns.
Many democratic theorists argue that political participation is essential for a well-functioning democracy (Pateman 1970; Dahl 1989). While there are many avenues for citizen participation in American democracy, through interest groups and political parties for example, the fundamental means by which citizens participate in government is through elections. Elections serve the function of translating citizen opinion into governmental policy, through the means of their chosen representatives. The repeated nature of elections enables citizens to hold their elected officials accountable for their actions while in office (Manin 1997), and it is clear historically that citizens are able to and do effectively exercise this ability. However, scholars demonstrate that participation in elections is not equal and that, as a result, the voices of some are privileged over others (see Verba and Nie 1972; Piven and Cloward 1989; Rosenstone and Hansen 1993). Consequently, mechanisms that promote citizen participation in elections are considered vital for the health of democracy. Accordingly, it is important to discover both what causes citizens to participate in elections and what routinizes subsequent participation.
Toward this goal, scholars increasingly find that identification and attachment with a religion increase the likelihood of political participation and engagement among citizens (Conger and Green 2002; Green 2007; D. E. Campbell and Monson 2008; Claassen and Povtak 2010). Religion promotes the development of civic skills that enable citizens to participate politically (Verba, Schlozman, and Brady 1995; Putnam 2000; Putnam and Campbell 2010). It also provides an important community or social network to which citizens belong and a social setting in which similar beliefs are promulgated among group members (Berelson, Lazarsfeld, and McPhee 1954; Verba, Schlozman, and Brady 1995).
We argue that the effect of religious commitment on voting behavior is an important relationship that is more nuanced than previous scholarship suggests. As with previous scholars, we agree that active participation in a religion fosters political participation. However, we contend that this positive effect is conditional on the particular religious tradition and historical circumstances. Specifically, the trade-off between time spent in worship and time spent in political activities is strongest for mainline Protestants, weaker for Catholics, and nonexistent for evangelical Protestants. Evangelical Protestants increasingly recognize the connection between their religious beliefs and politics, with the result that they have formed a habit of voting (Djupe and Grant 2001; Gerber, Green, and Shachar 2003).
This argument has important implications for American voting behavior literature. Scholars increasingly find that evangelical Protestants behave in a unique fashion politically, especially in relation to partisan voting patterns (Layman 2001; D. E. Campbell and Monson 2008). Going beyond partisan voting patterns, our findings demonstrate that evangelical Protestants manifest unique patterns as they relate to turnout patterns. Our article is divide into five parts: (1) we examine previous scholarship regarding the relationship between religious commitment and political participation, (2) we present our argument concerning religious commitment and turnout among the various religious traditions, (3) we describe the data and method, (4) we describe and analyze our results, and (5) we conclude the article in the final section.
Religious Groups and Voting in Elections
Turnout literature demonstrates that religion is a significant factor in voter turnout and that religious affiliation (belonging to a religion) in particular has a positive effect on voter turnout (Milbrath and Goel 1977; Hougland and Christenson 1983; Martinson and Wilkening 1987; Verba, Schlozman, and Brady 1995; Cassel 1999; Jones-Correa and Leal 2001; Greenberg 2000; Putnam 2000; Claassen and Povtak 2010). In addition, scholars find that religion interacts with demographic and ethnic factors to increase political participation for subsets of American voters (Strate et al. 1989; Harris 1995; Roof and McKinney 1987; Jones-Correa and Leal 2001).
Scholars identify social capital as the mechanism in the positive relationship between religious affiliation and political participation. In this view, members of religious traditions have greater ability to acquire skills that facilitate political participation (Wald, Owen, and Hill 1990; Verba, Schlozman, and Brady 1995; Putnam 2000; Jones-Correa and Leal 2001). Churches that are associated with the Judeo-Christian traditions of the United States generally have the greatest positive influence on political participation (Leege and Kellstedt 1993; Wald and Calhoun-Brown 2011).
Voting in Elections and the “Economy of Time”
While identifying with a dominant religious orientation plays a positive role in voting, research also shows a positive effect historically between religiosity (time spent in church) and voting in elections. The logical extension of the social capital argument is that the more time spent in religious institutions, the more skills that can be acquired, resulting in higher rates of political participation (see Hougland and Christenson 1983; Leege et al. 1993; Wald, Owen, and Hill 1990; Layman 1997). However, scholars have since discovered a curvilinear effect of religiosity on political participation (Iannaccone 1990, 1994; D. E. Campbell 2004; Driskell, Embry, and Lyon 2008). It is logical to suggest that “the more one is involved in a religious tradition, the less time there is available to be involved in other activities, (political, civic or otherwise)” (Driskell, Embry, and Lyon 2008, 304-5). Iannaccone (1990) concludes that increased religiosity leads to less political participation and ascribes this to an “economy of time” effect.
Briefly summarizing the “economy of time” argument, as individuals attend church more regularly, they are more likely to participate in politics. However, at the highest level of attendance, typically attending more than once a week, individuals experience a time trade-off between participating in church and participating in secular matters. At this point, their level of political participation begins to decline. In reality, the “economy of time effect” can apply equally to a motivational trade-off, as citizens may experience a motivational trade-off between participating in church and participating in secular matters at the highest level of attendance.
While the economy of time effect arguably should be more apparent in more demanding forms of political activity, such as participating in a campaign, voting in an election involves a distinct time element. Scholars have long noted that voting requires a significant amount of time to complete at the cost of other activities. As Downs (1957, 265) argues, “Time is the principle cost of voting.” Take a working parent, for example. She must make a number of time-related decisions to vote. If she votes in the morning, she must decide how she will get her children ready for school. If she votes at lunch time, she must sacrifice her lunch break. If she votes after work and after picking up her children from after school, she must find a babysitter for her children or take her children with her to vote. And if she votes later in the evening, she sacrifices other activities such as helping her children with homework or attending choir practice. In such a situation, it is easy to see how a significant portion of potential voters will forgo voting.
We identify two logics for the relationship between religiosity and political participation (voter turnout in particular). Under the first logic, social capital theory posits that there is a linear relationship; the more skills acquired, the more it facilitates civic participation. Under the second logic, the economy of time suggests that a trade-off exists in how individuals allocate time for or are motivated to participate in political activity. We posit that these divergent logics can be reconciled once religious tradition and changing historical circumstances are taken into account more fully.
Economy of Time and Religious Tradition, Traditionally Conceived
Presently, scholars associate the economy of time thesis most prominently with white evangelical and African American Protestants (Iannaccone 1994; Emerson and Smith 2000; Greenberg 2000; Wuthnow 2002; D. E. Campbell 2004; Driskell, Embry, and Lyon 2008). Scholars use the fact that white evangelical and African American Protestants have on average higher church attendance but lower political participation than mainline Protestants and Jewish Americans as evidence of an economy-of-time differential across religious orientations (Driskell, Embry, and Lyon 2008). 1 Scholars attribute the curvilinear effect that is evident between religiosity and political participation to the relatively lower participation of evangelical Protestants at the highest level of religiosity. By extension, evangelical Protestants have less interest in secular matters such as voting. As Steensland et al. (2000, 293-94) point out, evangelical Protestants are those denominations that have “typically sought more separation from the broader culture, emphasized missionary activity and individual conversion, and taught strict adherence to particular religious doctrines.”
Meanwhile, mainline Protestants are denominations that “typically emphasize an accommodating stance toward modernity, a proactive view on issues of social and economic justice, and pluralism in their tolerance of varied individual beliefs” (Steensland et al. 2000, 294). Mainline Protestants are believed to have greater connection to politics and civic life (Wuthnow 1999). Mainline Protestants and Jewish Americans exhibit the highest rates of voter turnout compared to the other religious traditions in America (Kellstedt et al. 1994). However, literature does not directly test for an economy of time effect as it relates to voter turnout, either within or across Protestant religious orientations. Furthermore, Catholics historically are associated with lower levels of political participation than Protestants (either mainline or evangelical). Scholars posit that this is the case because the hierarchical structure of the Catholic Church does not foster the development of civic skills (Verba, Schlozman, and Brady 1995).
Religion as a Public Activity and Economy of Time
The traditionally conceived relationship between the religious traditions and the economy of time thesis needs to be revised. We argue three things: (1) greater religious participation results in reduced political participation; (2) when religious groups recognize an explicit connection between their private religious life and public or civic life, it provides incentive for political participation thereby nullifying the economy of time effect (Djupe and Grant 2001); and (3) some religious orientations are more likely to view religion as deprivatized than are others. From these assertions, we form our hypotheses of how the economy of time thesis affects evangelical Protestants, mainline Protestants, and Catholics. Our hypotheses are presented in Figure 1.
Relationship between religious tradition and economy of time thesis.
Our model asserts that religious group attachment affects individuals in the dominant Christian orientations in specific ways. Evangelical Protestants are the most cohesive group of the three orientations, and this greater group cohesion deflects the influence of the economy of time (and motivation) on the probability of voting. Traditionally, evangelical Protestants, with their emphasis on saving souls and the afterlife, are thought of as those “too heavenly minded to be of any earthly good” and consequently are shown by scholars to exhibit an economy of time effect and to participate politically at rates lower than other religious groups (D. E. Campbell 2004; Driskell, Embry, and Lyon 2008). However, after a coalescence of events in the early 1970s, including the decision in Roe v. Wade (1973), a general feeling of societal degradation, and mobilizing as a political force through the 1976 Carter presidential campaign, evangelical Protestants developed a strong political orientation (Wald and Calhoun-Brown 2011). Accordingly, more recent literature indicates that evangelical Protestants participate at levels similar to individuals with other religious orientations when they are given additional motivation to do so (Kellstedt and Green 1993; D. E. Campbell 2004), in particular focusing on the effect of social and moral issues on conservative and partisan voting patterns (Green et al. 1996; D. E. Campbell 2004; Hillygus and Shields 2005; D. A. Smith, DeSantis, and Kassel 2006; Green 2007; D. E. Campbell and Monson 2008).
As a result of the increased importance of social and moral issues to politics, evangelical Protestants increasingly connect their private beliefs in these matters to taking public action on them. Their main motivation is the fact that many of the issues for which they advocate, abortion, stem cell research, gay rights, and increasingly gay marriage, are already legal. As a result, evangelicals increasingly view government as a threat to their fundamental beliefs and values. Evangelical Protestants increasingly equate political participation with protection of their beliefs and values. It follows that individuals who are committed to certain core beliefs should be more likely to see the necessity of political participation. Once the connection between belief and participation is routinized by the group, voting becomes a part of group behavior, a habit. For evangelical Protestants, voting is what “people like me do on election day” (Gerber, Green, and Shachar 2003, 548).
This habit of participation is the result of the cohesiveness of evangelicals. As Djupe and Gilbert (2009, 146) argue, “Individuals come to resemble the congregation when it is more cohesive. It is simply more difficult to stand apart when there is less social support for that position.” Through clergy cues, small group interactions, and the deprivatization of their religion, the political participation of evangelicals was cemented. Voting became a part of group behavior.
Figure 1 shows that our prediction for evangelical Protestants is bifurcated because belief is more likely to motivate evangelicals. This means that the manner in which evangelical membership is measured affects our prediction concerning the evangelical voting–religiosity relationship. When evangelical Protestantism is measured by a belief (the solid line), there is a nonlinear relationship between voting and time spent in worship. There is a “pull effect” toward political participation by those who are most integrated in the group. Individuals who attend church once or twice a month to once a week are pulled toward the extreme level of participation to conform to group norms.
Conversely, when evangelical membership is measured by belonging (denominational affiliation), the relationship between voting and religiosity is more linear (the dashed line). We still observe a slight “pull effect” toward the political participation level of the most integrated individuals, but the pull effect does not produce participation levels equal to that of evangelicals with the highest religiosity. In the case of both measurement approaches, we argue that evangelical Protestants have been mobilized into political participation as a result of changing historical circumstances, which in turn has caused them to exhibit a habit-based group behavior of political participation.
In contrast, and contrary to conventional wisdom, the economy of time thesis is most likely to affect mainline Protestants (regardless of how mainline Protestant membership is determined). Mainline Protestants take more moderate to liberal views on social and moral issues (Steensland et al. 2000; Layman 2001; Putnam and Campbell 2010). Mainline Protestant sects secularized Protestant principles into the fabric of American public life. For mainline Protestants, civic work, including aspects of social justice such as caring for the poor, is an expression of religiosity. For example, Mockabee, Monson, and Grant (2001, 679) find that mainline Protestants (probability = 0.18) are significantly more likely to view volunteering as the most important activity of their religion than are evangelical Protestants (probability = 0.07). 2 Liberal issues such as gay marriage and abortion do not cause a contradiction with their beliefs, nor do these issues increase mainline Protestant motivation to participate in political life above existing levels because government is not seen as threatening to their beliefs and values but as a necessary part of the secular world.
In addition, mainline Protestants prefer individual action to collective involvement in politics (Wuthnow 2002; Beyerlein and Chaves 2003). 3 Of the religious traditions, mainline Protestants (85 percent) are the most disapproving of clergy actively seeking to persuade their congregants in how to vote (Putnam and Campbell 2010). Instead, they prefer to be involved in politics as individuals pursuing their own causes. Consequently, mainline Protestants are likely to exhibit an economy of time effect. The highest level of church work reduces a person’s motivation and capacity to participate in the secular world.
Catholics present a less clear-cut story in the literature. While scholars historically argued that Catholics participate at lower rates than do Protestants, more recent literature suggests that this may not be an accurate portrayal of the Catholic vote. Jones-Correa and Leal (2001) demonstrate, looking at both affiliation and church attendance, that Catholics, both Latino and Anglo, do not participate at lower levels than their Protestant counterparts. In addition, the Catholic clergy is increasingly active in politics, motivated in large part by social issues and encourages their members to be active in politics as well (G. A. Smith 2008). Meanwhile, research finds that the Catholic Church is increasingly fragmented, with those low in religiosity leaning in a moderate to liberal direction in belief and practice and those high in religiosity leaning in a conservative direction (see Wilson 2007). 4 Catholics with high religiosity increasingly see the connection between private religious beliefs and public action on social issues. Consequently, Catholics are less subject to the economy of time effect than mainline Protestants but more so than evangelicals because of the fragmented nature of the Catholic Church.
Data and Measurements
We use two data sets to test our hypotheses regarding the effect of religious commitment on voter turnout across the religious traditions looking at the 2004 presidential election, the National Annenberg Election Survey and the Pew Forum on Religion and Public Life U.S. Religious Landscape Survey. Each data set offers specific advantages in testing our hypotheses, and the use of two distinct data sets provides a robustness check on our findings.
The National Annenberg Election Survey (NAES), conducted during the 2000–2008 presidential elections, constitutes the “largest academic public opinion studies of the American electorate ever conducted within a campaign cycle” (Romer et al. 2006, 14). We employ the rolling cross-national study that was conducted from July 16 to November 1, 2004. The data cover roughly a week before the Democratic National Convention through the Monday before the presidential election. Overall, the NAES sample that we use covers 119 polling days, with between 89 and 448 people interviewed on each of the days for a total of 32,931 respondents. 5 Because of the large sample size, the NAES yields greater leverage for the subgroup findings of this analysis. This is an improvement over studies that use National Election Study (NES) data to examine group level questions. For the NES data, investigators must pool subpopulations across election cycles to obtain sample sizes that are sufficiently large to make valid inference about smaller religious groupings (e.g., Claassen and Povtak 2010).
The second data set is the Religious Landscape Survey (Landscape). The survey was conducted in 2007 on a representative sample of the continental U.S. population with a total of 35,556 respondents and an overall response rate of 24 percent. It focuses on the religious life of Americans but also contains questions regarding the political attitudes and behaviors of Americans, including voter turnout in the 2004 election. 6 We use this data set for two reasons: (1) the large sample size allows for more accurate inference regarding the political behavior of religious subgroups, and (2) the survey allows for tests of hypotheses using more precise and accurate religious information than employed in previous research.
Dependent Variable
The use of two data sets allows us to test our model using two estimates of individual-level voter turnout: intended turnout and recalled turnout. These two measures of estimated turnout, while flawed, have been used extensively in the literature on voter turnout (Silver, Anderson, and Abramson 1986; Ansolabehere and Iyengar 1994; Hillygus 2005; Curtice and Sparrow 2010). While not a true measure of actual turnout, intended turnout captures respondent’s valuation of the importance of voting prior to the actual election. For the NAES data, which uses intended turnout, our estimated turnout rate is 87 percent. Meanwhile, recalled turnout is our dependent variable for the 2007 Landscape data. While the tendency of individuals to overestimate their actual turnout to vote in postelection surveys has been examined extensively in political behavior literature (e.g., Powell 1986; Silver, Anderson, and Abramson 1986; Curtice and Sparrow 2010), the measure remains useful because of the lack of availability of actual individual level turnout data. For the Landscape data, our estimated turnout rate is 76.1 percent.
In the present analysis, we expect there to be differences across and within the religious groups in their reported turnout over these two measures. In particular, we expect an overestimation effect for intended turnout among evangelicals because (1) they have greater group cohesion at high levels of religiosity (as detailed earlier) and (2) they receive greater campaign information through churches and Christian media compared to other religious groups (Wilcox and Sigelman 2001; Guth et al. 2007). Consequently, they not only have pressure to comply with group behavior but also are more likely to access campaign information as they are interviewed prior to the election (Zaller 1992). In turn, after the election, both the group pressure and information effects are no longer present, thereby reducing their likelihood of overestimating turnout. In contrast, mainline Protestants are the most disapproving of and likely to “selectively resist” clergy speaking about politics, taking away an important source of campaign information (Putnam and Campbell 2010; Geddes and Zaller 1989). 7 Consequently, mainline Protestants should be less predisposed to overestimate than are evangelical Protestants.
Religious Traditions
Scholars debate the proper way to conceptualize religion in America, through either the ethno-religious perspective or the theological restructuring perspective (see Smidt, Kellstedt, and Guth 2009). The traditional approach to conceptualizing (and measuring) religion is the ethno-religious perspective adopted from the work of Emile Durkheim (1915). This perspective emphasizes the importance of religious affiliation, arguing that religion is a social phenomenon, such that mere identification with a religion shapes political attitudes and responses. In contrast, the theological restructuring perspective, adopted from Max Weber (1930), emphasizes the importance of religious beliefs in shaping political attitudes and responses, arguing that religion is essentially the embodiment of specific beliefs. This perspective is emphasized in the recent culture wars debate (Hunter 1991; Fiorina, Abrams, and Pope 2005). 8 We test our hypotheses using both conceptualizations of religious tradition: (1) religious affiliation, which operates as an enduring social grouping mechanism and shapes political behavior (Smidt, Kellstedt, and Guth 2009); 9 and (2) religious belief, which has a profound impact on political behavior often overlooked in the literature (Driskell, Embry, and Lyon 2008; D. E. Campbell and Monson 2008; Wielhouwer 2009).
The use of the NAES and the Landscape Surveys allows us to test our hypotheses using measures of both theoretical perspectives. The Landscape Survey probes respondents to uncover their religious affiliation, with an emphasis on properly assigning individuals to either the mainline Protestant or evangelical Protestant category. As a result, this survey offers precise measures of the religious affiliation capturing the belonging aspect of religion. 10 In contrast, the NAES assigns respondents into more general religious categories, including Catholic, Protestant, and Jewish (among others). The NAES includes the question, “Do you consider yourself an evangelical or born-again Christian?” We categorize Protestants who answer yes to this question as evangelical and Protestants who answer no as mainline. In turn, the NAES survey allows us to capture the believing aspect of religion (but see Smidt, Kellstedt, and Guth 2009). Across both surveys, we generally follow the religious categorization scheme of Steensland et al. (2000) and include the following categories: Catholic, mainline Protestant, evangelical Protestant, Jewish, Mormon, other religion, no denomination, atheist or agnostic, and refused.
As we hypothesized, we expect the belief-based measure to result in higher estimates of turnout by evangelical Protestants with high but not the highest religiosity. This is because of the “pull effect” of evangelicals with the highest religiosity. This approach to measurement captures self-identified evangelical Protestants in traditional mainline Protestant churches. This condition occurs for at least two reasons: (1) marriage of evangelical Protestants to mainline Protestants who then choose to attend mainline churches and (2) some evangelicals seeking the status advantage associated with mainline Protestant churches (Wuthnow 2002). As a consequence, the voting behavior of evangelical Protestants with high but not the highest religiosity is doubly reinforced by mainline Protestant commitment to participation in civic life and evangelical Protestant information effect (through likely exposure to Christian media and clergy). In the belonging-based measure (religious affiliation), these evangelical Protestants in mainline Protestant churches are excluded from the evangelical category.
Church Attendance
We contend that religiosity affects an individual’s capacity or motivation to vote in presidential elections. Church attendance is our measure of religiosity and the second key independent variable of interest. Both surveys ask respondents how often they attend worship services. The NAES includes five response categories of never, few, monthly, weekly, and more than weekly (at least two times a week). The Landscape Survey includes six response categories: (1) never, (2) seldom, (3) few, (4) monthly, (5) weekly, and (6) more than weekly. Because we expect religiosity to have a nonlinear effect on voter turnout for certain religious orientations, we operationalize church attendance as a set of dummy variables as opposed to an ordinal variable.
Figure 2 shows the distribution of attendance across the two surveys. The overall NAES data (panel 2-1) is somewhat bimodal, with a spike at the few-times-a-year and weekly attendance levels. Meanwhile, the overall Landscape plot (panel 2-2) is less bimodal, with the large spike at the weekly attendance level. These two conditions occur because the few times a year category in the NAES data generally captures individuals who fall in the seldom and few times a year categories in the Landscape data. In addition, the congruence of belief-based and denominational-based measurements of evangelical and mainline Protestantism is demonstrated by the fact that in both data sets evangelical Protestants have a greater tendency to attend church service more than once a week than do mainline Protestants.
Religious orientation and church attendance.
Issue and Salience Control
Our argument is that evangelical Protestants in particular connect their private religious beliefs with the necessity of political participation to protect those beliefs. Mobilized first by specific issues, they subsequently form a habit of voting and so are likely to vote regardless of specific issues in the campaign. As a consequence, it is vital to control for the alternative explanation that specific issues mobilize evangelicals to vote (Wilcox 1989; Rozell and Wilcox 1996; Carmines and Layman 1997; D. E. Campbell 2004; Hillygus and Shields 2005; D. A. Smith, DeSantis, and Kassel 2006; D. E. Campbell and Monson 2008).
To do this, we control for gay marriage amendments on states’ ballots. In 2004, fourteen states had gay marriage ballot initiatives. 11 While the media credited these initiatives in the electoral victory of Bush (e.g., Dao 2004), scholars generally find that these initiatives increased the vote for Bush at the individual level but that these initiatives had little effect on evangelical turnout specifically (Abramowitz 2004; Hillygus and Shields 2005; D. A. Smith, DeSantis, and Kassel 2006; D. E. Campbell and Monson 2008). We include a control for same-sex marriage initiatives to control for issue-mobilization effects (Table 1). 12
Descriptive Statistics and Relation to Self-Reported Voter Turnout.
Data Source: 2004 National Annenberg Election Survey and 2007 Pew Forum on Religion and Public Life U.S. Religious Landscape Survey.
No denomination for the Pew Religious Landscape data is included in the atheist or agnostic category. Refused for the Pew Religious Landscape data is included in the atheist or agnostic category.
We also realize that the intensity of the campaign in any given state may produce higher turnout generally (Box-Steffensmeier, Darmofal, and Farrell 2009). Accordingly, we control for the presence of battleground states, those states in which neither candidate holds a substantial lead and so receive greater amounts of campaign activity. In 2004, eighteen states were classified as battleground states, and Table 1 shows campaign intensity had little effect on turnout. 13
Additional Controls
We include a number of additional control variables to ensure that our results are not spurious. These include the standard socioeconomic status measure of income and education, party identification, age, gender, and minority status. Each of these variables has been shown to have an effect on voter turnout (A. Campbell et al. 1960; Verba and Nie 1972; Wolfinger and Rosenstone 1980; Schlozman et al. 1995; Verba, Schlozman, and Brady 1995).
Statistical Model
To test our analytical models, we generate predicted probabilities of our dependent variable, voter turnout. We then plot these predicted probabilities by church attendance (our measure of religiosity). To generate these predicted probabilities, we run eight multilevel logistic mixed-effect models to estimate the probability of voter turnout. 14 We run two overall models for the NAES and Landscape data and submodels for Catholic, mainline Protestant, and evangelical Protestant for each survey. Subsetting the data by Christian orientations accounts for the fact that we expect Christian orientation to interact with religiosity (see Achen 2002).
We use a logistic multilevel mixed-effect regression model to analyze individual-level and state-level factors, covariates, and random effects that affect voter turnout and enter the independent variables and the level 1 control variables on the first level of the multilevel model. These level 1 variables include church attendance, race or ethnic group, age, gender (woman), income, education, and party identification. We enter attendance, race, and party identification as sets of dummy variables.
Level 2 of the multilevel model controls for the effects of the same sex ballot initiative and residing in a battleground state. We enter these state-level variables on level 2 of the model to control for these state-level effects. Last, we account for variation that may occur because of the two sampling processes. For the NAES data, the 119 polling days are modeled as random intercepts to account for the proximity of the election (Box-Steffensmeier, Darmofal, and Farrell 2009). For the Landscape data set (which sampled states on denominational characteristics), state sample size is modeled as random intercepts. Table 2 shows the specifications for these statistical models.
Multilevel Mixed Effect Logistic Regression Models of Intention to Vote.
Data Source: 2004 National Annenberg Election Survey and 2007 Pew Forum on Religion and Public Life U.S. Religious Landscape Survey.
Dependent variable is intention to vote in the Annenberg model and recalled vote in the Landscape model. Education is measured as 1 = 8th grade or less, 2 = some high school, 3 = high school degree, 4 = some college, 5 = 4-year degree, 6 = graduate degree. We also control for respondents’ income. In the NAES data, income is ordinal, and in the Landscape models income is a set of dummy variables.
Findings
The two overall models of turnout shown in Table 2 provide some evidence of an economy of time effect. In the overall NAES model of intent to turn out, the coefficient estimate for more-than-weekly attendance is less than the coefficient estimate for weekly attendance. Meanwhile, the coefficient estimate for more-than-weekly attendance in the overall Landscape recalled turnout model is greater than the coefficient estimate for weekly attendance but only slightly so. Indeed, the estimate for more-than-weekly is not significantly different from weekly attendance effect on recalled turnout. this lack of difference between the effect of weekly and more-than-weekly attendance on recalled turnout demonstrates the nonlinear effect of attendance on political participation.
To more definitively demonstrate this effect, we plot the predicted probability of intent-to-turnout and recalled turnout by level of attendance in plots 3-1 and 3-2 of Figure 3. Plot 3-1 of Figure 3 shows the NAES overall model and shows that the highest level of attendance does not result in the highest level of predicted probability of turnout to vote. Plot 3-2 of Figure 3 shows a similar result. Again, the highest level of turnout does not result in the highest level of predicted recalled turnout.
Predicted turnout, church attendance, and Christian religious orientation.
We contend that the economy of time affects mainline Protestants most. Both the NAES and Landscape mainline models strongly support our theoretical contentions. For both mainline models, Table 2 shows that the coefficient estimate for more-than-weekly attendance is lower than the coefficient estimate for weekly attendance. These two mainline Protestant models are the only Christian orientation models for which this is true.
Next, we plot the predicted probability of intent-to-turnout and recalled turnout by Christian orientation and level of attendance in plots 3-3 and 3-4 of Figure 3. We evaluate each religious grouping at the group mean for that attendance level. Our analysis is similar to evaluating unadjusted means for each group in an analysis of covariance model (Agresti and Finlay 1997). The figure shows graphically the effect of increased attendance on turnout for mainline Protestants, evangelical Protestants, and Catholics. Again, mainline Protestants (the circle symbol) are highly susceptible to the economy of time effect. Plot 3-3 of Figure 3 shows that the predicted probability of intended turnout for mainline Protestants at the more-than-weekly attendance level is the same as that of mainline Protestants at the none attendance level. The probability of recalled turnout (plot 3-4) also declines significantly moving from weekly attendance to more-than-weekly attendance.
Following our group cohesion argument, the confidence interval for mainline Protestants at the more-than-weekly attendance level is wider, particularly in plot 3-3. While the wider confidence interval is a function of sample size, we contend that it is also a function of the more individualistic behavior of mainline Protestants. The difference across the two mainline models is important and signifies that our findings are quite robust. Across the NAES and Landscape models, we use different measures of mainline Protestantism (one measure that is belief based and one that is belonging based). Consequently, the economy of time affects the political behavior of mainline Protestants in the same nonlinear manner regardless of the manner in which mainline Protestant membership is determined.
On the other hand, plots 3-3 and 3-4 support the idea that evangelical Protestants (the diamond symbol) are more cohesive as a group than are mainline Protestants or Catholics but are also bifurcated by the manner in which evangelical Protestantism is measured. Plot 3-3 shows that there is no difference in the probability of intended turnout for evangelical Protestants at the once-or-twice-a-month, weekly and more-than-weekly attendance levels (the relationship flatlines). Meanwhile, plot 3-4 shows that evangelical Protestants (the diamond symbol) are resistant to the economy of time in that the probability of recalled turnout increases as attendance increases (the relationship is more linear). 15 The plot also shows no significant difference between evangelical Protestants who attend church service at the seldom and few attendance levels. Alternatively, there are significant differences in the probability of recalled turnout for mainline Protestants and Catholics at these two levels of attendance. Our findings suggest a significant and abrupt divide in the political behavior of evangelical Protestants with high religiosity and evangelical Protestants with low religiosity, which is evidence of greater group cohesion.
In addition, we find greater disparity in intended and recalled turnout at each level of attendance among evangelical Protestants than among mainline Protestants and Catholics. We find this effect for three reasons: (1) the “pull effect” of greater evangelical group cohesion, (2) greater access to campaign information in the run-up to the election, and (3) the inclusion of self-identified evangelicals who attend mainline Protestant churches among the belief-based evangelicals.
Finally, our findings for Catholics are more mixed (the square symbol in plots 3-3 and 3-4). On one hand, Catholic turnout patterns are fragmented (as related to attendance). Across both the NAES and the Landscape Catholic models, Catholics with lower attendance are less likely to participate in voting, and this lower participation results in a step-like function. By step-like function, we mean there is no difference in the probability of turnout between two categories. It is important that the no difference comes at the same levels across both models: (1) between rarely and once a month attendance for the NAES Catholic model and (2) between seldom and monthly attendance for the Landscape model. In turn, we find fragmentation in the Catholic voting patterns across both models.
On the other hand, we do not find that Catholics are more subject to the economy of time effect than are evangelicals. In fact, we find some evidence that Catholics are more resistant to the economy of time in the Landscape Catholic model than are evangelicals.
Discussion and Conclusion
How and where citizens spend their time has political consequences. While social capital theorists argue that religious participation affects political participation positively (Putnam 2000), our findings demonstrate that the relationship between religious participation and political participation, in the form of voter turnout, is more nuanced than previous scholarship suggests. As with past scholars, we find that time spent in church does not automatically translate into increased political participation but exhibits an economy of time effect, where the highest level of church attendance brings about decreased levels of political participation (Iannaccone 1990, 1994; Driskell, Embry, and Lyon 2008). However, this effect is dependent on, and so varies with, specific Christian traditions.
Our hypothesis is confirmed that mainline Protestants are the most affected by the economy of time effect. They experience a trade-off between participation in religious and secular affairs. This suggests that as a result of their individual-based rather than group-based view of religion (Wuthnow 2002; Beyerlein and Chaves 2003) and their more liberal policy stances (Steensland et al. 2000), increased time spent in worship does not motivate mainline Protestants to engage in politics. In contrast, evangelical Protestants do not experience a time trade-off between participation in religious and political affairs. This suggests that evangelical Protestants are indeed habitualized into voting as a result of the deprivatization of their religion (Djupe and Grant 2001; Gerber, Green, and Shachar 2003). They increasingly see that voting is what members of their group “do on election day” (Gerber, Green, and Shachar 2003). Regardless of whether evangelicals resided in a state with a same-sex marriage amendment in the 2004 election, they remain immune to the economy of time effect. 16 As further confirmation that political participation is distinctive from other types of community volunteering, we find that evangelical Protestants are most susceptible to the economy of time as it relates to church-organized community volunteering. 17
Catholics present a more fragmented picture, as hypothesized. We do not find that they are more subject to the economy of time effect than evangelicals, suggesting that Catholics are undergoing a change not only in their political affiliations but also in their rates of political involvement (Wilson 2007; G. A. Smith 2008). This finding, along with previous research, makes Catholics an interesting and important group to examine in future research.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.