Abstract
This study investigates the direction and extent to which religious belonging and regular church attendance are related to distinct political preferences among U.S. Latinos. The key question is whether Latino churchgoers are more committed than infrequent attenders to liberal policy views and the Democratic Party, or whether Latino religious commitments are related to conservative policy views and Republican Party support. Findings indicate that Latino Protestants are more likely to hold conservative views, while Latino Catholics—the vast majority of religious Latinos—are more likely to hold liberal views, or show no political differences, if they attend church regularly.
Are Latino 1 voters up for grabs politically on the basis of their religious commitments? One response, often given by Republican Party officials, is that traditional family values and generally conservative views on social issues such as abortion and gay marriage (Nicholson and Segura 2005) make Latinos natural allies of the Republican Party (de la Garza and Cortina 2007). High rates of religious belonging and regular churchgoing (Perl, Greely, and Gray 2006), widespread reliance on spiritual guidance in daily life (Pantoja, Barreto, and Anderson 2008; Putnam and Campbell 2010), and the rising share of evangelical Latinos in the United States (McDaniel and Ellison 2008) all suggest Hispanics’ potential openness to conservative religious appeals.
A key implication of these claims is that Latinos with stronger religious commitments should espouse stronger Republican Party attachments and more conservative policy views than Latinos with more tentative religious commitments. Indeed, research shows this to be the general pattern for non-Hispanic whites in the United States (Green 2007; Guth et al. 1995; Layman 1997, 2001; Layman and Carmines 1997; Mockabee 2007; Smidt 2013; Wilcox 1986). Does this pattern also hold for Latinos in contemporary American politics? Or do Latinos more closely resemble African Americans, for whom traditional religious beliefs are associated with greater racial solidarity and Democratic Party support (Brown and Brown 2003; Calhoun-Brown 1996; Reese and Brown 1995)?
Answers to these questions have broad implications for how we understand Latino and American politics because they provide a window into the relationship between a widespread institution of civil society that touches the lives of most Americans—houses of worship—and the political allegiances of a rapidly growing demographic group (Suro et al. 2007). More than 90 percent of Latinos identify with some religious faith (Perl, Greely, and Gray 2006; Putnam and Campbell 2010), a remarkably high percentage that suggests the potential impact of Latino religious commitments. Whether Latino religious identifiers resemble white Americans or look more like African Americans speaks to practical and theoretical claims about the future place of Latinos in American electoral politics. The practical is whether Republican operatives are right to believe that traditional religious appeals will persuade Latino voters. The theoretical is how religious commitments relate to political preferences in a multiethnic American polity.
The answers turn out to be straightforward. Latino Catholics, who represent between 65 and 70 percent of the overall Latino population (Perl, Greely, and Gray 2006), as well as Latino Catholics who attend church regularly—those who are the most committed to their religious faith and potentially the most receptive to traditional religious appeals—are no more conservative in their political preferences than, respectively, Latinos who do not identify with an organized religion and Latino Catholics who are less committed to their faith. Instead, committed Latino Catholics are both more likely to support immediate amnesty for undocumented immigrants and more likely to support government welfare policy. Far from being open to religious appeals, Latino Catholics appear sympathetic to liberal Democratic views on Latino-salient issues, evidence suggestive of a process of cultural reinforcement that takes place within Catholic parishes.
I also find that Latino Protestants are consistently more conservative in their political preferences than Latinos who do not identify with an organized religion, results that confirm previous findings (Kelly and Kelly 2005; Kelly and Morgan 2008; J. Lee and Pachon 2007). I extend previous work in this area by also showing that committed Latino Protestants—those who attend church on a regular basis—are substantially more conservative than their less-committed fellow Protestants. Overall, while Protestant identification is on the rise among Latinos in the United States (Putnam and Campbell 2010; Smidt 2013), and religious commitments among this group are associated with more conservative political views, Latino Protestants still comprise a small share of the overall Latino population. Republican claims of Latino openness to conservative religious appeals are overstated, and Democrats, not Republicans, are more likely to persuade religious Latinos with their policy prescriptions.
Previous Research
Previous research shows that across the major faith traditions, distinct origins and practices are associated with distinct links between religious belonging and political behavior (Campbell 2004; Kohut et al. 2000; Layman 2001). Historically, Protestants have tended to support the Republicans, while Catholics have tended to support the Democrats (Kohut et al. 2000; Layman 2001; J. Lee and Pachon 2007). In recent years, however, among white religious identifiers, political divisions across the major faiths have declined, while distinctions within faiths have increased, leading to growing differences between traditional believers who attend church regularly and heterodox religious identifiers who worship less often (Green 2007; Guth et al. 1995; Layman 1997, 2001; Layman and Carmines 1997; Mockabee 2007; Smidt 2013).
Research looking at Latino religion and politics has disaggregated this group into the major faith traditions and found wide differences between Catholics on the one hand, and Protestants—both mainline and evangelical—on the other (DeSipio 2007; Kelly and Kelly 2005; Kelly and Morgan 2008; J. Lee and Pachon 2007). DeSipio (2007) and J. Lee and Pachon (2007), for example, both find non-Catholic Latinos more likely to support the Republican Party and its candidates than Catholic Latinos. Digging deeper, Kelly and Kelly (2005) use interaction terms to test whether religious commitment moderates relationships between Republican Party identification and religious affiliation, national origin, and nativity status among English-speaking Latinos. They find virtually no significant interactions but do find that Protestant Latinos were more likely than Catholic Latinos to identify with the Republican Party. Kelly and Morgan (2008) test whether religious traditionalism moderates relationships between religious affiliation and policy views, ideology, and party identification among English-speaking Latinos and whites. They find that traditionalism among Catholic, evangelical Protestant, and other religious Latinos is associated with more conservative ideology, greater support for restrictions on abortion, and traditional roles for women in society (see Kelly and Morgan 2008, 255, Table 5). These previous results lend support to claims by Republican leaders and strategists that, because of their upward mobility and conservatism on social issues, Latino voters may be persuaded by religious political appeals (de la Garza and Cortina 2007; Nicholson and Segura 2005).
African Americans, however, serve as a case in contrast because scholars find that black religiosity and regular churchgoing are linked to racial empowerment and Democratic Party support (Alex-Assensoh and Assensoh 2001; Brown and Brown 2003; Calhoun-Brown 1996; Dawson 1994; Harris 1994, 1999; McClerking and McDaniel 2005; McDaniel 2008; Tate 1991). African Americans, who are predominantly Protestant and religiously traditional (Fowler et al. 2004; Harris 1999; Mckenzie 2004), show strong positive relationships between greater church attendance, traditional religious beliefs and heightened racial group identification (Wilcox and Gomez 1990) and system blame (Reese and Brown 1995), attitudes that are linked to support for liberal policy positions and the Democratic Party (Dawson 1994).
The role of religion in American politics is further complicated by ethnic and racial group differences in relationships between religious beliefs and political attitudes (McDaniel and Ellison 2008; McKenzie and Rouse 2013). McDaniel and Ellison (2008) focus on biblical interpretations to show that white biblical literalists are consistently more conservative in their policy views than non-literalists; black literalists are consistently more liberal in their policy views than non-literalists; and Latino biblical literalists, compared with non-literalists, fall in-between the other two racial groups. McKenzie and Rouse (2013) emphasize racial and ethnic group differences in religious conservatism to explain core political values, finding that white religious conservatives are consistently opposed to egalitarian views, Latino religious conservatives are somewhat less consistent in their opposition to such views, and black religious conservatives are no different in their views than blacks who are religiously liberal.
Latinos, however, are different from whites and blacks on a number of dimensions, including a high proportion who are immigrants and Spanish speakers, substantial national-origin heterogeneity, and a large majority who are Catholic and not Protestant (Fowler et al. 2004; Levitt 2002; Perl, Greely, and Gray 2006). In addition, the American Catholic Church has never been a political or social sanctuary for Latinos in the way that black churches have been for African Americans. Many Irish- and Italian American Catholics resisted the incorporation of Mexican American and other Latino worshippers into their local parishes (Sagarena 2009). This relatively weak historical connection between Latinos and official Catholic institutions has resulted in the underrepresentation of Latinos in Catholic leadership positions (Levitt 2002).
Nonetheless, the Catholic Church eventually supported the creation of separate Spanish-language parishes (Lopez 2009) and pursued changes to Church policy that placed greater emphasis on individual morality and showed greater support for liberal positions on economic policy and approaches to immigration, even while maintaining its conservative stance on social issues such as abortion and homosexuality (Cavendish 2007; Levitt 2002; Sagarena 2009). These changes have made the Catholic Church more accommodating to poor and minority worshippers overall, raising the possibility that regular churchgoing among Latino Catholics reinforces liberal policy views and identification with the Democratic Party. Less clear is whether Latinos today exhibit patterns of religious commitments and political attitudes that more closely mirror the divergent paths of black or white Americans.
Theory and Hypotheses
Previous research provides an important starting point in the study of Latino religion and politics, but existing findings are tentative because of previous data limitations and conceptual ambiguity about the process linking religious belonging to political preferences. The present study contributes to literatures on religion and politics and Latino political behavior by showing that relationships between Latino religious belonging and political preferences go deeper than mere differences across the major faith traditions, the primary conclusion of existing scholarship (e.g., DeSipio 2007; Kelly and Kelly 2005; Kelly and Morgan 2008; J. Lee and Pachon 2007). I argue instead that by investigating variation in religious commitments within faiths, we can test claims about Latino openness to conservative religious appeals among those who should be most receptive to such appeals: churchgoers who are regularly exposed to teachings of their faith.
By conceptualizing church attendance as a measure of within-fath religious commitment and a proxy for exposure to religious teachings, we can evaluate whether commitment to a particular faith is associated with political preferences that are distinct from those among committed identifiers of a different faith. Because previous research has shown significant political differences between Hispanic Catholics and non-Catholics, I, too, expect relationships between Latino churchgoing and political preferences to vary systematically by faith, with committed Latino Catholics holding views on politics and policy that are distinct from those among committed Latino Protestants or committed Latino identifiers of other faiths. This perspective also suggests that any connection between faith and politics should be most evident among those who are regularly exposed to religious teachings in their place of worship. I test these and related expectations using nationally representative data on Latinos that allow me to overcome three major limitations of previous research in this area.
First, studies that have relied on surveys focused primarily on religion often lack questions on standard political attitudes such as party identification and policy views (e.g., Pantoja, Barreto, and Anderson 2008), or else lack information about language use, national origin, and migration experiences (e.g., McDaniel and Ellison 2008; McKenzie and Rouse 2013), deficiencies that limit the scope of research findings directly applicable to the diverse Latino population and its current political preferences.
The second major limitation in previous research on religion and politics derives from the practice of pooling multiple years of English-only surveys to conduct analyses on Latinos (e.g., Kelly and Kelly 2005; Kelly and Morgan 2008). This approach results in a sample overrepresented with wealthier and more acculturated Latinos who tend to hold more conservative political preferences (Barreto and Pedraza 2008; Branton 2007) and stronger party attachments (Barreto and Pedraza 2008; Hajnal and Lee 2011). Contrary to the claim of no expected differences in “relationships between religion and politics . . . among Spanish speakers compared to those who speak at least some English” (Kelly and Kelly 2005, 88, fn. 3), the previous research on Latino acculturation suggests that religious commitments should be linked to conservative political views to a greater extent among English-speaking Latinos than among bilingual or Spanish-only Latinos. Studies of Latino religion and politics that rely on English-only surveys cannot test this claim and are limited to examining a truncated range of Latino political views.
I overcome these first two limitations of previous research by analyzing survey responses of English- and Spanish-speaking Latinos who answered several religion questions, a broad set of policy items, political preferences, and detailed items on ethnic identity, language use, and migration experiences, factors that are systematically related to Latino public opinion (Branton 2007; T. Lee and Pérez 2014; Sanchez 2006). Using these more appropriate data, the first set of hypotheses suggest that if Latinos are indeed open to conservative religious appeals, then Latinos who belong to a religious faith should hold more conservative political preferences than Latinos without any religious identity.
In addition, however, because the Catholic Church has taken divergent ideological positions on social issues such as abortion and gay marriage, compared with those on economic welfare and immigration, Latino Catholic preferences should similarly diverge by policy domain.
I expect Latino Protestants to hold consistently more conservative views than non-identifying Latinos.
The first set of hypotheses will test for differences in Latino political preferences by faith using more appropriate data than in previous works.
The next question is whether Latino churchgoers, regardless of their particular faith, are more likely than infrequent attenders to support the Democratic Party, its presidential candidate, and liberal policy views much like African Americans who attend church regularly, or whether Latino churchgoing is associated with Republican Party support and conservative policy views as it is for white Americans. If relationships between churchgoing and political preferences among Latinos mirror those of white Americans, then the following hypothesis will be supported.
If, however, previous research showing significant political differences between Catholic and non-Catholic Latinos (e.g., DeSipio 2007; Kelly and Kelly 2005; Kelly and Morgan 2008; J. Lee and Pachon 2007) is consistent with the current data, then assessing average relationships between churchgoing and Latino political preferences could mask countervailing differences across faiths. For example, Latino Catholic churchgoers might be more liberal, while Latino Protestant churchgoers might be more conservative—both compared with infrequent attenders of their respective faiths—and thereby produce null results for average relationships between churchgoing and preferences.
This suggests a third shortcoming in research on religion and politics, which is the common analytic strategy of modeling political attitudes as a function of dichotomous variables for religious belonging while statistically controlling for religious traditionalism, Biblical literalism, or the frequency of church attendance (e.g., Jones-Correa and Leal 2001; McDaniel and Ellison 2008; McKenzie and Rouse 2013). This approach makes two often implicit assumptions: (1) that religious traditionalism, Biblical literalism, or the frequency of churchgoing are related to political preferences in the same direction, on average, regardless of the particular faith to which individuals belong; and (2) that individuals of a given faith have, on average, the same relationship with politics regardless of the strength of their commitment to that faith. Both of these assumptions should be tested in the study of religion and politics generally, and particularly so among Latinos who are divided politically along lines of the major faiths (DeSipio 2007; Kelly and Kelly 2005; Kelly and Morgan 2008; J. Lee and Pachon 2007).
For Latinos, I argue that religious commitment measured as regular churchgoing is not simply an indicator of behavior, but also a measure of exposure to the church environment. If religious teachings and expectations about political behavior are communicated within the church environment, and previous research suggests that they are (Brown and Brown 2003; Djupe and Gilbert 2003; McClerking and McDaniel 2005; Reese and Brown 1995), then variation in exposure to these teachings should moderate relationships between belonging to a particular faith and political preferences. Differences in political preferences between Latinos who attend church regularly and those who attend infrequently, within a single faith, represent a churchgoing gap in Latino public opinion. Because of previous research showing links between Latino Protestantism and more conservative political views (e.g., Kelly and Kelly 2005; Kelly and Morgan 2008), I expect committed Latino Protestants to hold more conservative views than Latino Protestants who are less committed, while committed Latino Catholics should hold more liberal views than less-committed Catholics.
The previous discussion suggests a third set of hypotheses about variation in Latino religious commitments and political preferences.
Conversely, because of the history of Latino exclusion and segregation from white Catholic worshippers (Lopez 2009; Sagarena 2009), as well as the Catholic Church’s outspoken support for liberal immigration reform and economic welfare policies, I expect committed Latino Catholics to hold more liberal political and policy views than less-committed Latino Catholics.
Implicit in the preceding hypotheses is a unidirectional causal explanation: that exposure to messages, norms, and values in church influences the political preferences of individual worshippers. Parishioners, however, may bring their pre-existing attitudes into the church environment, or else seek places of worship where religious teachings are consistent with their pre-existing attitudes. Although previous scholars have argued that politics in church are a rare occurrence (Putnam and Campbell 2010; Wald, Owen, and Hill 1988), making self-selection into churches for political reasons similarly rare, others have shown that clergy speeches and other church activities are endogenous to the political preferences and needs of a congregation (Djupe and Gilbert 2003). Existing survey data such as those analyzed here do not allow for a conclusive test of either claim, but the current data do allow an investigation into whether regular churchgoers of a particular faith are politically distinct from those who do not attend church regularly. Inferences from such an investigation among Latinos show the extent to which religious commitments politically divide the Latino community.
Data and Method
I test my expectations using the 2006 Latino National Survey (“LNS”; Fraga et al. 2006), which was conducted via telephone between November 2005 and August 2006 using fully bilingual (English and Spanish) callers who interviewed 8,634 Hispanics or Latinos. Most LNS respondents are Mexican origin (70.2%), foreign-born (64.0%), and Spanish-dominant (58.0%). These data contain information on Latino political preferences, whether respondents belong to a particular faith, whether they had a born-again experience, and how often they attend worship services. Religious belonging is captured with indicator variables for Catholics, Protestants, 2 other religious identifiers, 3 and non-identifiers who “don’t identify with any religious denomination.” For ease of interpretation, church attendance is dichotomized to represent regular churchgoing, defined as attending religious services at least once per month or more often. 4 I use the born-again item to control for religious identifiers with more traditional beliefs (Wilcox 1986). 5
In addition, the LNS asked detailed demographic and immigration queries that, combined with the political and religious items, allow me to describe patterns of religious identification, regular churchgoing, and political preferences among the heterogeneous Latino population. Importantly, because this survey was conducted with fully bilingual callers, estimates of the Catholic and non-Catholic Latino populations are more accurate than estimates derived from surveys conducted in English with a Spanish callback option (e.g., the Pew Forum on Religion and Public Life’s 2007 U.S. Religious Landscape Survey) and especially those that were conducted entirely in English (e.g., the American National Election Studies conducted before 2008). Surveys conducted in English or that use a Spanish callback option are likely to be biased because they under-sample recent immigrants and Spanish-dominant Latinos, compared with surveys using fully bilingual callers, and it is precisely these less-acculturated Latinos who identify as Catholic at higher rates than English-speaking Latinos (Lugo et al. 2008, 41–43; Perl, Greely, and Gray 2006). Comparing the distribution of religious identification in the LNS with those elicited by surveys that interviewed Latinos using fully bilingual callers (e.g., Suro et al. 2007) demonstrates the appropriateness of the LNS data for the current investigation. 6
I use the three religious items on the LNS to test whether and to what extent religious commitment among Latinos of various faiths is associated with dependent variables measuring Republican two-party identification, support for George W. Bush in 2004, and various policy attitudes. The policy items include two social policies—views on abortion and gay marriage—that are most likely to be related to religious commitments, one item on immigration policy, and one item on government-provided income support (economic welfare). Immigration and economic welfare represent two policy areas in which the Catholic Church has taken strong liberal positions (Levitt 2002; Sagarena 2009), and they also represent policy domains of particular salience to the Latino community (Sanchez 2006).
I code all of the dependent variables dichotomously because, for three of the four policy items, the survey provided only discrete response options that do not constitute an ordered scale and would require the use of a multinomial modeling strategy that makes presentation of results for several dependent variables difficult to interpret. 7 Dichotomizing also allows me to specify the policy items such that they represent the most extreme views, which makes any significant differences between more and less religiously committed Latinos substantively meaningful. In addition, because the focus of this investigation is on comparing relationships between churchgoing and political preferences within and across denominations, rather than across response options for any single policy item, a dichotomous approach is appropriate.
Party identification is coded as two-party identification, pitting Latino Democratic (1) against Republican (0) identification, both including leaners, to isolate relationships between religious belonging, churchgoing, and Latino preferences between the two major parties. I remove “pure” independents and non-respondents from the analysis of Latino party identification because of very high rates of non-response (29% overall) and the lack of meaningful attachments to the two parties that this may signify (cf. Hajnal and Lee 2011). I code support for Bush in 2004 in a similar dichotomous fashion, pitting Latinos who voted for or preferred President Bush (1) against Latinos who voted for or preferred Senator John F. Kerry (0). 8
Below, I first examine bivariate distributions of Latino political preferences and policy attitudes by religious affiliation and regularity of church attendance, and then analyze political preferences and policy attitudes as dependent variables in models that include measures of religious belonging, born-again experiences, regular churchgoing, the interactions of religious belonging and regular churchgoing, plus demographic controls. All of the models use a national sampling weight that accounts for non-response by age and education and allows for generalizations to the U.S. Latino population. The models also use robust standard errors clustered by sampling unit, procedures that create more demanding tests of statistical significance between variables to account for non-independence of observations (Williams 2000).
All of the predicted probabilities reported are derived from models with a focused set of demographic control variables that include age, gender, education, income, national origin, race, nativity, citizenship, immigrant age of arrival, generation, language dominance, and survey language use. These variables each capture important life experiences that previous research shows are related to Latino party identification, candidate support, and policy attitudes (Alvarez and Bedolla 2003; Beck, Corak, and Tienda 2012; Branton 2007; Cain, Kiewiet, and Uhlaner 1991; Hajnal and Lee 2011; T. Lee and Pérez 2014; Nicholson and Segura 2005; Uhlaner and Garcia 2005). Without these demographic controls, any observed relationships between religious variables and political preferences or policy views could be spurious (King, Keohane, and Verba 1994).
Results
The LNS data confirm high rates of religious belonging, particularly Catholic identification, within the Latino community. Table 1 shows that fully 93 percent of Latinos explicitly identify with a religious faith, 41 percent say they had a born-again experience, 71 percent identify as Catholic, and another 13 percent as Protestant. Jehovah’s Witnesses, Mormons, Jews, and other religious Latinos together comprise about 10 percent of the sample, while those not identifying with any major religion (“non-identifiers”) account for about 7 percent, a figure substantially smaller than the 16 percent of whites estimated not to hold any religious affiliation in 2007 (Lugo et al. 2008). For each religious subgroup, Table 1 also shows the percent that attends church regularly. Among Latinos of any religious faith, 75 percent attend church on a regular basis, while, unsurprisingly, those not identifying with any faith are the least likely to attend church regularly. Nonetheless, 19 percent of Latino religious non-identifiers still report attending church at least once a month or more often. Latino Protestants and those with born-again experiences are the most regular churchgoers, with 85 percent and 84 percent of these subgroups, respectively, reporting regular church attendance. Latino Catholics and other religious identifiers report regular church attendance at similar rates: 73 percent and 69 percent, respectively.
Latino Religious Belonging and Regular Churchgoing.
Cells show the percentage of Latinos with the particular attribute (columns) that belong to the religious subgroup (rows designated with “all” in parentheses) or that belong to the religious subgroup and attend church regularly (rows designated with “regular” in parentheses).
Table 1 also shows rates of religious belonging and regular churchgoing across major Latino subgroups defined by national origin, nativity, generation, and partisanship. In general, Mexican-origin Latinos are the most likely to identify as Catholic, and more likely to attend church regularly, compared with Puerto Rican- or Cuban-origin Latinos. By contrast, U.S.-born Latinos are less likely to identify as Catholic, and less likely to attend church regularly, than immigrant Latinos; these differences are no more pronounced among the third generation, indicating that the biggest drop-off in religiosity among Hispanics occurs between the first and second generation. More Democratic than Republican Latinos identify as Catholics, but more Republican than Democratic Latinos identify as Protestants. Latino independents look like Democrats in their religious belonging, with higher rates of Catholicism, and lower rates of Protestantism, than among Latino Republicans. Republican Protestants are more likely to attend church regularly than Democratic Protestants, but otherwise, there are few partisan differences in rates of regular church attendance.
How are Latino political preferences related to their religious commitments? Bivariate results not shown (but available in the online appendix at http://prq.sagepub.com/supplemental/) demonstrate considerable variation in Latino political preferences and policy support by religious belonging and regular churchgoing. Among all Latinos, there is a significant churchgoing gap for every dependent variable except two-party identification, with regular churchgoers more likely to say they supported Bush in 2004, more likely to oppose abortion under any circumstance, more likely to oppose gay marriage and civil unions, more likely to support immediate amnesty for undocumented immigrants, and more likely to support economic welfare than infrequent attenders. These results, however, mask differences across faiths. Among Protestant and other religious Latinos, regular churchgoers are significantly and substantively more likely than infrequent attenders to identify as Republican (twenty-nine and seventeen points more likely, respectively), to support Bush (twenty-eight and sixteen points more likely, respectively), to oppose abortion (thirty-six and twenty-one points more likely, respectively), and to oppose gay marriage and civil unions (sixteen and twelve points more likely, respectively). For both Protestant and other religious Latinos, however, there are no differences between regular and infrequent churchgoers in their views about immigrant amnesty and economic welfare.
To a lesser degree, Catholic Latinos who attend church regularly are also more likely than infrequent attenders to support Bush (five points more likely) and oppose abortion (four points more likely), but this largest religious subgroup among Latinos is no more or less likely to identify with the Republicans, to oppose gay marriage, or to support economic welfare if they attend church regularly. Conversely, churchgoing Catholics are significantly more likely than infrequent attenders to support immediate amnesty for undocumented immigrants (ten points more likely).
To account for systematic differences between Latinos of different faiths, I use logistic regression to model each of the binary dependent variables as a function of one of two sets of religious variables plus control variables described previously. The first set of religious variables includes indicators for religious belonging (non-identifiers are the excluded category), an indicator for regular church attendance, and an indicator for born-again experiences; the second set includes religious belonging, regular church attendance, born-again experiences, and the interactions of religious belonging indicators and regular church attendance. In order to correctly interpret logistic regression coefficients, especially those of interaction terms, one must calculate the difference in predicted probability of the dependent variable when changing the value of an independent variable of interest and holding other independent variables constant (Ai and Norton 2003; Brambor, Clark, and Golder 2006; King, Tomz, and Wittenberg 2000).
Figure 1 plots partial results for the dependent variables regressed on the first set of religious variables and all control variables (left-hand panel), showing differences in the probability of the dependent variables between Latino religious non-identifiers and Latinos of any faith, Catholics and Protestants. 9 The right-hand panel of Figure 1 plots partial results for the dependent variables regressed on regular churchgoing and its interaction with religious belonging indicators, plus all other religious and control variables, which show differences in the probability of the dependent variables between regular and infrequent Latino churchgoers of any faith, and among Catholics and Protestants. 10
Latino religious belonging, regular churchgoing and political preferences.
One of the main conclusions to draw when examining results for each dependent variable in Figure 1 is that, without accounting for regular church attendance, analysts will have a biased view of how Latino religious belonging is related to political preferences. First, Latino political divisions are only apparent when we disaggregate the group into its major faith traditions: comparing Latinos of any religious faith with non-identifiers produces no significant differences on any of the dependent variables examined (Hypothesis 1a is not supported). Instead, there are systematic differences in Latino preferences by faith, with Latino Catholics no more conservative than non-identifiers (Hypothesis 1b is not supported) and Latino Protestants substantially more conservative than non-identifiers (Hypothesis 1c is partially supported). Specifically, the results indicate that Latino Catholics are no different than religious non-identifiers in terms of their party identification, candidate preferences, and policy views. In contrast, Latino Protestants hold more conservative views than non-identifiers in terms of their greater likelihood of Republican Party identification (8.8 percentage points greater, p = .058), support for President Bush in 2004 (21.4 percentage points greater, p < .001), and opposition to gay marriage (15.3 percentage points greater, p < .05). This divergence between Latino Catholics and Protestants, without accounting for the moderating role of church attendance, largely confirms existing research.
Examining religious commitments expressed as regular churchgoing (Figure 1, right-hand panel), among Latinos of any faith there is a positive association between regular churchgoing and more conservative party identification, candidate preferences and social policy views, but also greater support for economic welfare. Specifically, regular churchgoing among Latinos of any faith is positively related to increased likelihood of Republican identification (4.0 percentage points greater, p < .05), support for Bush in 2004 (9.1 percentage points greater, p < .001), opposition to abortion (4.0 percentage points greater, p < .05), opposition to gay marriage (8.4 percentage points greater, p < .001), and support for economic welfare (2.8 percentage points greater, p < .01). These results are largely supportive of Hypothesis 2 because Latino churchgoing on average is associated with more conservative political preferences and social policy views, and only associated with support for economic welfare to a small degree.
However, as I argued previously, it should be among the most committed religious identifiers that we see the strongest connections between religious belonging and political preferences. To test this expectation, I examine regular churchgoing among Latino Catholics and Protestants separately as shown in the right-hand panel of Figure 1. The results are clear: Latino Catholics who attend church regularly are no more conservative than infrequent attenders in terms of their party identification, presidential candidate support, and social policy attitudes on abortion and gay marriage. Latino Protestants, on the other hand, are significantly and substantively more likely to identify with the Republican Party (23.9 points more, p < .001), support Bush in 2004 (22.8 points more, p < .01), oppose abortion under any circumstance (16.3 points more, p < .001), and oppose gay marriage and civil unions (18.8 points more, p < .01), compared with Latino Protestants who are less committed to their faith. These results suggest that Latino Protestant identity alone, without regular exposure to church teachings, has few associations with conservative political preferences.
In contrast to the null results on party identification, candidate preferences, and social policy views, Latino Catholics’ views on Latino-salient issues of immigrant amnesty and economic welfare are significantly and positively related to regular church attendance. Latino Catholics who attend church regularly are more likely to support immediate amnesty (3.6 percentage points more likely, p < .05) and economic welfare (2.9 percentage points more likely, p < .05) than Latino Catholics who infrequently attend church. Tellingly, Latino Protestants show an identical pattern of results on these Latino-salient issues, with point-estimates that are even more liberal than those obtained for Catholics. 11 Hypothesis 3 and its sub-hypotheses, therefore, are largely supported.
Robustness Checks
Because the goal of assessing whether regular churchgoing divides Latinos politically might be threatened by omitted variable bias (King, Keohane, and Verba 1994) or mismeasured concepts, I preform robustness checks of the previous results. First, potential omitted variables relevant to the previous analysis include civic engagement, interest in politics, campaign contact, political ideology, ethnic identity, and ethnic commonality. I test whether these attributes are related to churchgoing and political preferences by including their measures in models described previously and re-calculating differences in probability for religious belonging and regular churchgoing. I also test two alternative definitions of churchgoing (defined as weekly or more frequent and a categorical measure spanning the range of attendance frequency) and whether Latino born-again experiences or survey language use moderate the previous results.
In all cases, the findings are virtually unchanged and the substantive conclusions remain the same (but see the online appendix for more details). On support for immigrant amnesty, there is a small difference from the previous results when controls for Latino identity are added to the models, which results in the significance of Catholic churchgoing dropping below conventional levels (p = .06). This suggests that ethnic identity is part of the process linking Latino Catholic churchgoing to preferences on immigration policy. On opposition to abortion, there is a small difference from the reported results when regular churchgoing is defined as weekly or more frequent attendance, which results in a significant (but substantively small) positive association for Catholic churchgoers. Born-again experiences have few moderating effects on the relationships reported previously: born-again Catholics who attend church regularly are not more conservative than regularly attending Catholics without born-again experiences, and born-again Protestants who attend church regularly are about as conservative as regularly attending Protestants without born-again experiences.
On survey language use, the results indicate that English-dominant Latinos hold stronger and more consistent relationships between regular churchgoing and the dependent variables, in contrast to Spanish-dominant Latinos who mainly show null results. In fact, for Spanish-dominant Latinos, regular churchgoing is not related to any of the outcomes of interest for any faith except Protestants’ opposition to abortion, with Spanish-dominant Protestants who attend church regularly 16.7 percentage points (p < .01) more likely to oppose abortion than Spanish-dominant Protestants who attend less frequently. For English-dominant Latinos, the results mirror those obtained for Latinos overall.
Discussion
The previous findings make three specific contributions to the literature on Latino religion and politics. First, as scholars have previously shown, Latinos remain a highly religious group in aggregate. In addition, the current data show overwhelming rates of religious belonging and regular churchgoing across several group attributes including national origin, nativity, generational status, and partisanship. Latinos’ religious commitments, however, do not mean that they are up for grabs politically because the vast majority of religious Latinos—those who identify as Catholic—do not hold political and policy preferences that are more conservative than those among non-identifiers. This is in contrast to the pattern of political and policy preferences among Latino Protestants, who consistently hold more conservative views than non-identifiers. Latino Protestants, however, comprise a small share of the overall Latino population.
Second, although regular churchgoing is associated with more conservative political views among Latinos on average, the results also show that the particular faith to which Latinos belong significantly moderates relationships between regular churchgoing and Latino public opinion. If Latino religious identifiers relied on their religious commitments alone to direct their political choices, without regard to the content or frequency of exposure to their particular faith, we should see few differences across faiths in the relationships between churchgoing and politics. The results, however, allow us to soundly reject this view because the evidence consistently shows that the particular faith to which religious Latinos belong matters for the direction and extent to which churchgoing is linked to party identification, candidate preferences, and policy views.
Third, looking specifically at Latinos who are most likely to be persuaded by traditional religious appeals—those who are most committed to their faith and regularly exposed to religious teachings in their place of worship—the results are again clear. Committed Latino Catholics who attend church regularly are no more conservative in their party identification, presidential candidate support, or social policy views than less-committed Latino Catholics. Instead, committed Latino Catholics are more liberal on immigration and economic welfare policies. These findings are consistent with a story of cultural reinforcement that takes place within Latino churches, especially Catholic parishes, where regular interactions between parishioners and church leaders help connect ideas about faith to politics.
For Latino Protestants who are committed to their faith and attend church regularly, the results are strikingly different on party identification, presidential candidate support, and social policy views. On these key political outcomes, Latino Protestants who attend church regularly show large differences from Protestants who attend church infrequently, and these churchgoing gaps are all in a substantially more conservative direction. In fact, Latino Protestants who do not attend church regularly look no different than Latino Catholics on these outcomes, suggesting the importance of regular exposure to the church environment for cementing Latino Protestants’ conservative political views. However, on outcomes that are salient to the Latino community such as immigration and economic welfare policies, committed Latino Protestants look no different than their less-committed brethren, instead following the same general pattern as committed Catholics. Latino Protestants may be open to conservative religious appeals on the basis of their party attachments and social policy views, but not in terms of their views on immigration and economic policy.
Conclusion
The present study advances how we understand and operationalize relationships between religion and politics among Latinos and other groups in American politics. For students of race and ethnicity, the wide-ranging dependent variables examined here, the specific attention given to unpacking the cleavage potential of religious commitments, and the large and diverse Latino sample analyzed provide the most complete empirical portrait of Latino religious practice currently available.
In addition, building on previous research by McDaniel and Ellison (2008) and Kelly and Kelly (2005), both published in this journal, the interactions strategy pursued here demonstrates the utility of conceptualizing religious belonging as an identity that varies in strength depending on religious commitment or the regularity of church attendance. Previous works that look only at average effects of religious conservatism (as in McKenzie and Rouse 2013) or average effects of churchgoing (as in Jones-Correa and Leal 2001) implicitly assume that the direction of these effects is the same for all Latinos. However, if the major faiths diverge in terms of their expressed teachings and the role of religion in politics, examining average effects of churchgoing for all individuals will mask differences across individuals of different faiths. Instead, this article has shown that committed Latino Catholics, committed Latino Protestants, and committed Latinos of other faiths are politically distinct from their less-committed, same-faith brethren. In particular, committed Latino Catholics are more liberal on immigration and economic welfare policies, while committed Latino Protestants are more conservative on party identification, candidate support, and social policy views, than their less-committed fellow identifiers.
These findings represent an important contribution to our understanding of religion in American politics more generally. To paint an accurate picture of the links between religious practice and political preferences, analysts should specify the extent to which regular churchgoing within faiths moderates these relationships, in addition to assessing direct relationships between politics and individual attributes such as religious belonging, beliefs, or behaviors. While the current study only investigates how belonging to a particular faith moderates the political consequences of churchgoing behavior, the beliefs dimension of religious practice is an area ripe for additional research.
For Latinos in particular, religious commitment expressed as regular church attendance should strengthen religious beliefs that are emphasized in church as well as their relationships to political preferences. Moreover, as I argued previously, it is precisely among the most religiously committed who attend church regularly that we should see evidence of a stronger connection between faith and politics. Although some parishioners undoubtedly bring their pre-existing political attitudes with them into the church environment, it is also in particular church environments where messages from the pulpit, social interactions, and other group activities convey ideas about religion, society and politics to those who are regularly in attendance. Whether through regular exposure to political content in church, or through the development of stronger religious beliefs, regular church attendance within specific faiths evinces political attitudes that are distinct from those held by religious identifiers without regular exposure to their faith’s worship environment.
Footnotes
Acknowledgements
I am grateful to Natalie Masuoka, Rachel Stein, Gary Segura, David Laitin, Paul Sniderman, Simon Jackman, Shaun Bowler, Eric McDaniel, Martin Gilens, three anonymous reviewers and PRQ guest editors for their feedback on this research. I also thank Michael Hoffman and Elizabeth Nugent for their research assistance. All remaining errors are mine alone.
Author’s Note
Previous drafts of this article were presented at the Western Political Science Association annual meeting, the Politics of Race, Immigration and Ethnicity Consortium workshop at Texas A&M University, and the Rooney Center workshop at the University of Notre Dame, in all cases benefiting from the feedback of audience members.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
