This paper uses the Latino Immigrant National Election Study (LINES) to better understand the relationship between religion and immigrant political and civic engagement. Over the last half century, both American religion and the immigration landscape have changed in important ways. The LINES, which includes a number of religious questions from the American National Election Study and a rare focus on Latino newcomers, provides the opportunity to better understand the contemporary relationship between the two. We find that measures of religious belongings, beliefs, and behaviors (the Three Bs) are not generally associated with the civic and political engagements of Latino immigrants. We posit that such null results may be explained by the varying religious experiences of immigrants—some developing bridging social capital through religious institutions, but others experiencing what might be called segmented religious assimilation.
migration, immigrants, religion and politics, segmented assimilation, social capital, Latino politics
The goal of this article is to better understand how religion shapes the political engagement of the Latino first generation. The Latino Immigrant National Election Study (LINES) is a unique dataset that provides rare perspectives on Latino immigrants, an understudied population. Although a growing number of surveys now include enough Latino respondents to allow separate analysis, such as the American National Election Studies (ANES), the share of foreign-born respondents is typically low. This is especially the case for the study of religion and politics, as few surveys include the range of questions necessary to understand contemporary dynamics. The LINES combines a large sample of Latino immigrants with a number of religion questions from the ANES. We can therefore explore how the religious belongings, behaviors, and beliefs of the Latino first generation influence their engagement in civic affairs and electoral politics.
Understanding religion is increasingly important to the study of American politics—although this is not often recognized—because of the considerable, and intertwined, transformations of demography and religion. As is well known, the 1965 Immigration Act brought from across the globe large numbers of people previously barred from migration. Less noted, however, is that these individuals brought new religious traditions, theologies, practices, symbols, and interpretations to America, thereby [End Page 125] disrupting established patterns of religious life. Immigrants not only founded new churches but also joined, and diversified, existing denominations.
This would come to have implications for politics and policy. Many of these churches, while traditionally conservative, recognized that immigrants from the developing world were not only joining congregations but also potential targets of evangelization. These immigrants and their children are increasingly seen as the future of many denominations, which has caused institutional reassessments of policies such as immigration reform. The Southern Baptist Convention (2011), for example, has adopted resolutions supporting comprehensive immigration reform with a path to legal status for unauthorized immigrants.
These changes are overlapping with longer-standing transformations in American religious identity and practice. For much of American history, denominational affiliation was a key dividing line with political, social, and economic implications. This was reflected in the title of Will Herberg’s (1955) classic book Protestant, Catholic, Jew. Changes were nevertheless emerging that would complicate traditional categories, including the fundamentalist and modernist split within Protestantism and the concomitant development of different beliefs and practices within traditions and denominations. Scholars realized that old categorizations no longer described the reality of religion in American life (Kellstedt et al. 1996; Layman 1997, 2001; Green 2007; Olson and Warber 2008; Smidt, Kellstedt, and Guth 2009) and that religion was fundamentally restructuring (Wuthnow 1990).
The concepts of belief (such as biblical literalism) and behavior (such as church attendance) were therefore added to belonging (denomination). These Three Bs better captured the “new religion gaps” (Green 2007) and allowed scholars to understand how contemporary religion was shaping politics. For instance, conservatives and liberals from different denominations are increasingly working together to achieve political and policy goals. To add to the complexity, religious orientations may also have different political effects by race and ethnicity (McDaniel and Ellison 2008; Leal and Patterson 2013), and many Americans are now moving away from any religious affiliation (Campbell, Green, and Layman 2013; Corral, Leal, and Tafoya 2015).
Demographic trends are generating renewed research across the social sciences on minority populations, and the Latino and Asian American religious experiences are the subject of increased attention. For Latinos, scholars have explored such topics as religion and civic activism (Espinosa, Elizondo, and Miranda 2005), the role of faith during the migration journey (Hagan 2008), the activism of Latino faith-based organizations (Wilson 2008), the role of religion in the farmworkers movement (Prouty 2006), historical accounts (Sandoval 2006), the place of Latinos in American religion and Catholicism (Stevens-Arroyo 1980; Díaz-Stevens and Stevens-Arroyo 1998; Matovina and Riebe-Estrella 2002; Matovina 2012), Latino ministry (Dahm 2004), and Latino theology (De La Torre and Aponte 2001). While space considerations preclude a more detailed discussion of this literature, it can be found in the previous work of the first two authors of this article (Leal 2010; Leal and Patterson 2013, 2014).
Although Catholicism has long been a core component of Latino cultures, it is less so every year. In 2006, the Pew Forum on Religion & Public Life (2007) found that 68 percent of Latinos were Catholic, a figure that had declined to 55 percent by 2013 (Pew Research Center 2014). Many of these former Catholics are now attending evangelical and Pentecostal congregations, but the number of Latino “Nones” also increased (from 8 percent to 18 percent). At the same time, the large number of migrants from Mexico, Central America, the Dominican Republic, and other Spanish-speaking nations in the Western Hemisphere has augmented Catholic numbers. Even if many ultimately change religious identities in the United States or become converts before the migration experience, the American Catholic population would have shrunk over the last two decades without “the infusion of millions of Latino Catholics” (Keysar 2014, 11). The paradox is that while Latinos are decreasingly Catholic, American Catholicism is increasingly Latino.
These changes in Latino religious affiliation are particularly clear across the generations. [End Page 126] The 2006 Pew survey found that while foreign-born Latinos were 74 percent Catholic, their native-born counterparts were 58 percent. The 2013 Pew survey notes that 60 percent of the foreign born are Catholic—a 14 percentage-point drop in seven years—and that 50 percent of the second generation and 45 percent of the third generation are Catholic (Pew Forum 2007; Pew Research Center 2014). The figures vary slightly in different surveys, but the trends are clear.
Although the religion and politics literature addresses a wide and growing range of topics, this paper is interested more specifically in the role of religion in shaping Latino immigrant political engagement.
This paper summarizes what we know about immigrants and religion by examining the literature on how religion can shape Latino political engagement. In particular, the debate between Verba, Schlozman, and Brady (1995) and Jones-Correa and Leal (2001) raised the question of whether Catholicism enhances or dampens Latino political participation. We contribute to this discussion by asking how religious measures—and particularly Catholicism—are associated with variables for electoral and civic participation among the Latino first generation. We discuss the variables in the 2012 LINES and how they allow for (and in some instances, limit) new research on these topics. The paper also presents models that add to our understanding of how religion shapes the political engagement of Latino immigrants and how the results—which predominantly show little association between religion and politics—can be interpreted in multiple ways.
IMMIGRANTS AND RELIGION
We might begin by remembering the words of Koheleth that “there is nothing new under the sun” (Ecclesiastes 1:9). Indeed, research on religion and immigration probes a question that is perhaps as old as humanity itself. As the Psalmist so poignantly reflected on the central crisis of the Hebrew people’s Babylonian exile, “How can we sing the songs of Zion while in a foreign land?” (Psalms 137:4)
Research demonstrates the important role of religion in the lives of immigrants and immigrant communities (for example, Leonard et al. 2005; Foley and Hodge 2007; Chen and Jeung 2012). In surveying this research, a logical starting point would be why, in theoretical terms, religion matters to the immigrant experience. Thomas Tweed (2006) offers a potential explanation, arguing that religion is inherently spatial: “As spatial practices, religions are active verbs linked with unsubstantial nouns by bridged prepositions: from, with, in, between, through, and most important across … religions designate where we are from, identify whom we are with, and prescribe how we move across” (79). Religion enables, in Tweed’s view, “homemaking”—“It draws boundaries around us and them; it constructs collective identity and, concomitantly, imagines degrees of social distance” (111). In his terms, it enables both “dwelling” and “crossing,” that is, “finding a place and moving across space” (59). In essence, religion shares the central concerns of the immigrant experience: singing the songs of home while coming to terms with a new life in a foreign land.
As the exilic Psalm suggests, religion is a powerful and multifunctional force. In broad terms, it can reinforce ethnic identities and provide a space within which ethnic differences may be expressed, while also promoting participation in, or assimilation into, a new host country. Research supports this dual role (see, for example, Dolan 1975; Yang 1999; Yang and Ebaugh 2001; Cherry 2013) while revealing important complexities. This is why scholars have approached the question of religion and immigration from different angles.
Over the past half century, beginning at least with Herberg’s (1955) seminal work, our understanding of religion and immigration has grown steadily and respect has increased for religion’s explanatory power. Social scientists have addressed issues such as the functional effects of religion in promoting or inhibiting civic participation (Verba, Schlozman, and Brady 1995; Putnam 2001; Jones-Correa and Leal 2001), the role of religion in political reasoning (Lee and Pachon 2007), and group and identity politics (Chen and Jeung 2012). Scholars have also given greater nuance to such matters as the relationship between religion and ethnicity, and how religion facilitates [End Page 127] involvement in the social and political life of a new host country. For example, this may depend on the relative strength and relatedness of ethnic and religious identities (Greeley 1971; Abramson 1973; Hammond and Warner 1993) or majority-minority status in home and host countries (Yang and Ebaugh 2001).
In addition, many religious dynamics are transnational, so that we cannot easily divide home country and American religion. Some immigrants are pre-acculturated before the journey by transnational congregational networks, which not only helps facilitate border crossing but can also change religious practices in the home nation and thereby prepare the immigrant for reception in the United States (Levitt 2002). Religious capital is therefore an important part of the migration experience, but religions are global networks and the directional arrows of personal and institutional change point in all directions.
Religion and Latino Political Participation
Research has also provided insight into the degree with which religion and communities of faith promote or inhibit participation in the social and political life of host countries among foreign-born populations and subsequent generations. More general works on civic participation have found church attendance to promote greater involvement (Verba, Schlozman, and Brady 1995; Putnam 2001). Sidney Verba and his colleagues (1995) identify the importance of civic skills and civic resources to civic participation, but they also note the unique role of religious institutions in providing the opportunity to access such skills. Although socioeconomic status (SES) and organizational resources are conceptually distinct, they note that “those with high levels of educational attainment are likely to be slotted into the kinds of prestigious and lucrative jobs and organizational affiliations that provide further political resources” and the only organization that can “provide a counterbalance to this cumulative resource process” is religion (Verba, Schlozman, and Brady 1995, 18).
Verba and his colleagues (1995) analyze time-based nonelectoral activities, voting, political contributions, and political discussion, finding that SES factors and not racial-ethnic measures are statistically significant. However, SES is developed by a variety of organizational experiences, particularly that of religion. They argue that unfortunately for Latinos, Catholicism was less likely to develop the same level of politically relevant civic skills as many Protestant denominations did. Because Latinos were predominantly Catholic at the time of the Citizens Participation Study (though the authors acknowledge a growing Protestant share), Verba and his colleagues posit that Catholicism served to dampen Latino political participation.
Michal Jones-Correa and David Leal (2001) argue that if Catholicism helped explain disparities in political participation across groups, then it should also shape participation within groups. They compare Latino and Anglo (non-Hispanic white) electoral and nonelectoral participation, finding that Catholic affiliation never reduced Latino participation—and in fact served to increase several types of engagement. By contrast, attendance at religious services was consistently significant, thus suggesting that the associational role of churches—regardless of denomination—was also important. Jones-Correa and Leal observe that “while churches play an important part in American civic life in general, in the absence of other civic associations they play a disproportionate role in the civic and political lives of Latinos” (2001, 763). In addition, religious institutions may be particularly important to immigrant communities: “As they did for previous waves of immigrants before them, Catholic churches may serve as ethnic associations as much as they do religious institutions” (764).
For Anglos, however, religious denomination was consistently insignificant across four political participation models. Although Catholicism may have once shaped Irish, Italian, and Polish American political participation, “it is not likely to be true today. For most Anglos, the difference among churches is simply denominational, not ethnic. Political appeals in Latino Catholic parishes might have a resonance they do not have in Anglo Catholic parishes” (Jones-Correa and Leal 2001, 764).
Louis DeSipio (2007), examining the 2000 Hispanic Churches in American Public Life (HCAPL) survey, notes that Catholic voter turnout [End Page 128] was one-third higher than that of non-Catholics, which is inconsistent with the demobilization argument. He also finds that Catholics were slightly more likely than Protestants to report that their churches had become more involved in political and social questions. However, Catholics were less likely than Protestants to believe that a political candidate’s faith and morals were important considerations in their vote choice.
Joanne Ibarra and David Leal (2013) replicate the Jones-Correa and Leal (2001) models using the 2006 Latino National Survey (LNS) and the 2008 ANES (with Latino oversample). None of the models indicate that Latino Catholics incur a mobilization penalty, though only the LNS found that Catholicism was positively associated with Latino political participation (voter registration and voter turnout in the 2004 presidential election). In the ANES findings, the religious variables—denomination, born-again status, religious importance, and church attendance—were consistently insignificant. For both Anglos and Latinos, religion did not shape voting, an index of nonelectoral participation, or campaign contributions.
Jongho Lee and Harry Pachon (2007) find no evidence that denomination or church attendance shaped political participation. Using the 2004 Washington Post/Univision/Tomas Rivera Policy Institute survey of 1,600 registered Latino voters, they test whether religious variables affected presidential vote preference, intensity of vote preference, interest in the presidential campaign, and whether they were contacted to register or to vote. Although evangelicals were more likely to support George W. Bush, no religious tradition variable was statistically significant in the other models. These null results support neither the Verba, Scholzman, and Brady (1995) nor the Jones-Correa and Leal (2001) findings because they suggest that religion was not an important factor (aside from vote direction) in 2004.
Research by sociologists and other scholars tends to support the argument that religious institutions can provide important, politically relevant skills to immigrants. Cristina Mora (2013) observes that Catholic churches provide opportunities for civic engagement for Mexican immigrants. First, participating in small groups allows immigrants the opportunity to build skills and develop networks. Second, churches provide links to nonreligious civic groups. This qualitative paper provides additional evidence that is consistent with the majority of the quantitative literature discussed previously. In addition, the edited volume by Vargas-Ramos and Stevens-Arroyo (2012) includes multiple chapters—using both qualitative and quantitative methods and across a variety of settings—that counter the argument that Catholicism demobilizes Latino populations.
Researchers have also addressed such questions for additional faiths and ethnic groups. Peggy Levitt’s study of Boston-area immigrant groups, including Hindus and Muslims, finds that “even when religious institutions did not have explicit political agendas, people learned about fundraising, organizing and leadership by participation, which they applied to other settings” (2008, 778). Moreover, the greatest accumulation of civic skills occurred among those whose congregations were in close proximity to a native-born congregation, as interaction between the two offered a kind of intermediary education in U.S. political and civic life. Levitt ominously observed that “in contrast, members of stand-alone congregations, with few U.S. ties, were on their own” (780, emphasis added).
A recurring theme in this literature is the important role of the church not only as an institution in promoting the civic skills necessary for participation but also as a source of organizational support for participation. This is especially true of Catholic churches. For example, Matt Barreto and his colleagues (2009) find the Catholic Church played a central role in disseminating information about the immigrant rights marches, and Catholic identification was a strong predictor of participation in, or support for, these marches. Similarly, Kraig Beyerlein and Mark Chaves (2003) note that Catholic churches, more than other congregations, organized demonstrations and marches and lobbied elected officials. Cecilia Menjívar (2003) observes the same among Salvadoran immigrants as Catholic churches encourage immigrants to work collectively to transform their communities. In contrast, evangelical [End Page 129] Christian churches attended by Salvadorans place greater emphasis on individual salvation.
DATA AND MODELS
The LINES survey addresses each of the three traditional approaches to measuring religion in survey research: believing, behaving, and belonging. For religious belief, the LINES asked respondents two relevant questions: “Do you consider religion to be an important part of your life, or not?” and “Would you say your religion provides [some/quite a bit/a great deal] of guidance in your day-to-day living?” One might also include in this category the follow-up question for Christian respondents about whether they identify as born-again or evangelical Christians.
For the most part, the LINES follows the lead of the ANES in its religious questions. However, some deviations are notable. The first of these concerns religious belief. The LINES does not include a question on the literal interpretation of scripture, which is included in the ANES. This is a common question on several surveys, including not only the ANES but also the Cooperative Congressional Election Study (CCES), the General Social Survey (GSS), and many Pew studies.
Questions about biblical literalism reflect the role of an interpretive community in articulating a religious worldview that translates religious teachings and texts into a particular vision for social and political life (Fish 1980; McDaniel and Ellison 2008). This question is often used in research as a proxy for religious traditionalism, and the LINES does not include another such measure. However, the literalism question is primarily appropriate for Protestant traditions more so than for Catholicism.
We would also have liked a variable for charismatic or Pentecostal beliefs. A vibrant, charismatic spirituality can be found not only among Pentecostals but also increasingly among Catholics. This is especially the case in Latin America. In 1970, Pentecostals and charismatics represented no more than 4 percent of Latin America’s population. According to the World Christian Database, by 2005 their numbers had increased to more than 25 percent (Pew Forum 2006). Not only is charismatic spirituality qualitatively different from noncharismatic spirituality, but its political implications are also debated.
The LINES also includes questions about religious behavior. Here, the survey follows the ANES closely and includes questions on the frequency of prayer and church attendance: “Outside of attending religious services do you pray?” and “Do you ever attend religious services, apart from occasional weddings, baptisms, or funerals?” Response options account for variation in frequency. These are helpful measures, particularly the church attendance question, which have shown predictive power in estimating social and political behaviors.
The LINES includes questions on religious affiliation. Here we find another departure from the ANES and certainly the most problematic feature of the LINES’s religious measures. Like the ANES, the LINES asks respondents, “Do you mostly attend a place of worship that is Protestant, Catholic, Jewish, or something else?” or for those who do not attend religious services, “Do you consider yourself to be” any of these religious affiliations. Although the LINES stops here, the ANES offers respondents an opportunity to further define their religious affiliation. This is important for several reasons. First, because a growing number of Protestants do not identify as such (or even understand the term), only asking respondents whether they are Protestant, Catholic, Jewish, or something else will underrepresent Protestants. Many contemporary Protestants instead identify as “just Christian” or evangelical or sometimes with a denominational moniker.
A similar problem exists with the something else option in the LINES, which groups religious others and the unaffiliated, and unfortunately risks also including Protestants who do not identify with this term (an issue of growing concern to survey researchers). We can create an estimate of the Latino immigrant Nones, but the measure is not ideal.
Our paper examines the religious, demographic, and socioeconomic determinants of Latino immigrant political engagement. This is not an exact replication of the Jones-Correa and Leal (2001) paper, as the LINES contains additional variables that are specifically relevant [End Page 130] to immigrant populations. We are guided by the expectation that if Catholic churches boost electoral participation because they serve as community centers that can uniquely connect Latinos to political information as well as political mobilization efforts, the political and civic effects might be even stronger among immigrants.
The first measure is a dummy variable for whether the respondent voted in the 2012 presidential election. This question was only asked of naturalized citizens. We also analyze the Campaign Activity Battery (four questions) and the Civic Participation Battery (four questions). These questions were asked of all respondents, as nonelectoral participation is not limited to citizens or legal permanent residents (see Leal 2002). We also create an index variable (from 0 to 4) and dummy variable (whether the respondent participated in any activity) for each of the batteries.
The campaign measures that we use include whether a campaign attempted to mobilize the respondent, attending a campaign rally, wearing a campaign button, and working for a political campaign.1 The civic participation measures include whether the respondent attended a civic rally, attended a civic meeting, signed an Internet petition, or signed a paper petition.2
To our knowledge, these analyses have not been previously conducted specifically for Latino immigrants. A variety of studies have examined the importance of religion for immigrant political engagement, but not in the manner found in this paper.
The independent variables are standard demographic measures (education, income, gender), national-origin groups measures (Mexican, Cuban, and Central American), and variables specifically applicable to immigrants (took the survey in Spanish, was brought to the United States before age sixteen, and the percentage of life lived in the United States). See appendix A for descriptive statistics for all the independent and dependent variables. In addition, a correlation matrix of the independent variables (not included) indicates that multi-collinearity should not be a concern.
For each model, we first run regression models using the variables by themselves and then using the weight variable provided by the LINES (wgtrake). It was based on education, gender, and age.
The LINES survey contains considerable missing data, both of the normal individual nonresponse variety and also because some questions were not asked of the respondents in the supplement to the second-wave survey. We therefore conduct three types of analysis for most questions: the first two use the unweighted and weighted survey data (as described), and the third uses AMELIA II in R to impute all missing data (Honaker, King, and Blackwell 2012). In this way, we provide evidence for scholars with different views of the value and propriety of weighting datasets and imputing missing data (for encouraging perspectives on imputing independent and dependent variables, see Graham 2009; Young and Johnson 2010; Hollenbach et al. 2014). As we will show, regardless of whether the weight measure is used or the data are imputed, the key religious results do not vary, which adds to our confidence in the findings.3 [End Page 131]
The voter turnout dependent variable, a simple dummy variable, was modeled using logit regression. Our index measures for civic and campaign participation denote counts of activities reported by the respondents. In both the LINES and the ANES, rates of participation are heavily skewed to the 0 or 1 counts. The skewed nature of these dependent variables means that ordinary least squares (OLS) is an inadequate method of estimation. The go-to approach for estimating count data is Poisson regression, but our dependent variables show overly dispersed variance (higher than mean) and an excess of 0 observations, both of which violate traditional assumptions. Under these conditions, a zero-inflated negative binomial regression is ideal but unavailable when estimating subpopulation parameters, imputed data, or models with too few observations. We therefore chose negative binomial regression as the next best procedure, which addresses excessive variance. We make no assumptions about the data-generating process for 0 (no participation) observations. They serve as our base for understanding the characteristics of respondents who reportedly engaged in political and civic activities.
In addition, we collapse the two index measures into two dummy variables for whether the respondent participated in any civic or campaign activity. These models are analyzed using logit analysis.
For ease of interpretation, all coefficients are presented as odds ratios (for Logit regressions) or incidence rate ratios (for the negative binomial regressions). Values greater than 1 represent positive associations between the independent and dependent variables, and values less than one represent negative associations.
Last, because of considerable second-wave attrition in the LINES, a second survey firm was engaged to increase the sample size. We were concerned that respondents in these two different samples would potentially yield different responses to survey questions. To address this, we tested the possibility that respondents represent separate data-generating processes. First, Amelia II, in addition to treating respondent identification numbers as unique, allowed us to consider each survey firm as a unique cross-section and generated imputations of missing observations therein. The generated datasets were then analyzed using generalized negative binomial regression, which enables users to assign a cross-section variable to test whether estimates for each differ. We applied this test to the models for the civic and campaign index measures. In each case, the auxiliary test for differences in models across cross-sections did not find statistically significant differences. That is, the data collected across the two survey firms in the second wave of the LINES are not apparently significantly different in the count models we specified.
Examining both pre- and postwave respondents, we see that the basic religious affiliations of the LINES respondents are both different than and similar to those reported by most surveys. Using the sample weights, 61 percent identified as Catholic, 8.9 percent as unaffiliated, 13.1 percent as Other, and the remaining 17 percent as Protestant (table A1).
We can compare these responses with those of the Pew Research Center’s 2013 survey, which found that the Latino foreign born are 60 percent Catholic, 16 percent evangelical Protestant, 4 percent mainline Protestant, 15 percent unaffiliated, and 4 percent Other. The share of Catholics, one of our main measures of interest, is therefore almost identical in both surveys. The differences among the other categories likely reflect the variations in how the religious affiliation question was asked by the LINES and the Pew. [End Page 132]
The first set of regressions examines voter turnout in the 2012 presidential election (see table 1). These models, in combination with the subsequent models for electoral and civic participation, allow us to contribute to the Verba, Schlozman, and Brady (1995) and Jones-Correa and Leal (2001) debate by examining how a variety of religious factors are associated [End Page 133] with political engagement. As noted, Verba, Schlozman, and Brady (1995) argue that fewer civic skills were associated with Catholic Church membership, which could help explain relatively low levels of Latino electoral participation. Jones-Correa and Leal (2001), however, find no evidence to support a Catholic demobilization argument. In fact, they note that the Catholic variable was positively associated with some forms of engagement and insignificant for the rest. They argue that churches serve as accessible community centers, where both political learning and recruitment can take place, and this role might be particularly consequential for immigrants.
The first model is for reported turnout—among naturalized citizens—in the 2012 presidential elections. As with most surveys, the reported turnout is likely higher than is the case, but the votes are not validated (consistent with ANES practice). The number of observations is relatively small because we are modeling a subset of the dataset, and missing data take their standard toll. Nevertheless, the model provides some unique evidence about the association of religion with Latino immigrant electoral participation.
The first column includes the basic regression model (using odds ratios), the second column model incorporates the weight measure, and the model in the third column analyzes the data as imputed by Amelia II in R. Across all three models, the Catholic and church attendance variables are statistically insignificant. Catholicism and more regular church attendance do not demobilize Latino immigrants, but neither do they encourage it. The findings therefore fall between the argument that religion bolsters or dampens minority political engagement. The only consistently statistically significant religious factor is the reporting of an Other religious affiliation; these individuals are much less likely to report voting in 2012.
Only a relatively few other variables are statistically significant in the models. In terms of SES, age is positively associated with the vote in two models (unweighted and imputed), and income and education are only statistically significant for one instance each. We conclude that the standard SES model does not appear to apply to voting among the foreign born.
We do not include a model for specific vote choice in 2012, as the number of observations was too low—below one hundred (it was relevant only to the naturalized, and it was asked only on the postelection wave). Nevertheless, a separate model (not shown) indicates that no religious variable was statistically significant and that the dominant effect was party identification.
We next examine the campaign participation batteries. As noted previously, the items in each battery have been combined into index and dummy variables. We model the unweighted measures (first two columns), the weighted measures (second two columns), and the imputed measures (last two columns). Table 2 indicates that few religious effects were present.
No religious variables are statistically significant in the first four models (unweighted and weighted). In addition, the Catholic measure is statistically insignificant in all models, and the church attendance variable is only statistically significant in the campaign index model with all missing data imputed. In addition, the other religious variables are almost entirely insignificant across the models. The one exception is the measure for the importance of religion, which is statistically significant only in the imputed models.
We also see that education (positive) is statistically significant across almost all models, as the SES theory of political engagement might predict. Women are also less likely to vote than men are, although neither age nor education was generally associated with voter turnout. Also, the variables that take into account features of the immigrant experience (such as percentage of life in the United States, immigrating to the United States as a child, or taking the survey in Spanish) are not significant in any of the models.
We also modeled (without the survey weights or missing data imputation) the individual variables that comprise the campaign index as well as an additional measure for financial contributions. These are not shown because of space considerations, but they indicate that Catholicism and church attendance are never statistically significant. The other religious [End Page 134]
[End Page 135]
variables were largely insignificant, showing only a few scattered instances of statistical significance.4
Taken together, the campaign models provide additional evidence that religion is neither positively nor negatively associated with Latino immigrant political participation. More generally, we see relatively few variables at work in any given model, which suggests that political engagement may be differently structured for immigrants and the native born (a proposition we test using 2012 ANES data).
Table 3 includes the models for civic participation. Again, we see relatively few religious variables at work, and none that is consistently significant. Most important, the Catholic and church attendance variables are always insignificant. In addition, the other religious measures show only scattered, inconsistent effects. Most notable is that in the imputed models, the religiously unaffiliated (the Nones) and those who found religion important are both more likely to engage in civic activities.
The most consistently significant measures are those that involve the immigrant experience. The greater the percentage of life lived in the United States, the more likely the respondent is to participate in four of the six models. In addition, identifying with a political party in the respondent’s nation of origin is associated with greater civic activity in five of the six models. The SES measures, by contrast, are largely insignificant (although income was the most notable among this group).
When we model the individual variables that make up the index measure, plus two questions about donating to a religious or non-religious organization, we see no consistent religious effects. One difference in these models is that the role of education is more noticeable because it is statistically significant in four of the six models, whereas it was significant only once in the index and dummy variable models.
We therefore see that religion does not appear to shape the civic engagement of Latino immigrants. On the one hand, this suggests that religion is not providing a boost to civic activism among immigrants. On the other hand, it is not negatively associated with such engagement. As we will suggest, however, these null findings could mask disparate effects—some immigrants join churches that promote a bridging social capital that promotes greater civic and political engagement, while others become members of isolated congregations and consequently experience a form of segmented religious assimilation.
Last, we created comparison models using similar dependent and independent variables from the 2012 ANES (see table A3 for descriptive statistics). The former included voter turnout, an index of campaign activities, and an index of civic participation activities (for the specific measures, see table A2). We ran these models for the Latino native-born sample as well as the Anglo (non-Hispanic white) sample (see tables 4 and 5).
For native-born Latinos, the role of religion is more evident than in the LINES models. The Catholic variable is statistically significant and positive in the two index models, whereas the church attendance variable is statistically significant and positive in the voter turnout model. Among Anglos, the Catholic variable is not significant, although the church attendance measure is positively associated with voter turnout and campaign participation.
Taken together, these ANES findings are consistent with Jones-Correa and Leal (2001). These authors note that Catholicism was never negatively, and sometimes positively, associated with Latino political engagement. For Anglos, by contrast, church attendance mattered in a way that denomination did not. They posit that such results spoke to the different roles played by churches in the lives of Latinos and Anglos. [End Page 136]
[End Page 137]
This paper used the unique LINES as well as the 2012 ANES to better understand the implications of religion for the political engagement of Latino immigrants. To date, the lack of survey data on Latino immigrants means that we have little quantitative evidence about the political implications of the Latino immigrant presence in American religion. Given this lack of previous research, it would be premature to draw strong conclusions from our results. Nevertheless, we hope that future researchers will continue to examine the religious profile of Latino immigrants and to study the role of religion in Latino and immigrant communities (for discussion, see Leal 2002, 2010; Barvosa-Carter 2004; DeSipio 2007; Matovina 2012).
More specifically, the paper contributes to the Verba, Schlozman, and Brady (1995) and Jones-Correa and Leal (2001) debate by examining the determinants of electoral, nonelectoral, and civic engagement. The various models, [End Page 138] using different weighting and imputing approaches, found consistent results—almost no religious effects among Latino immigrants. In particular, there is no support for the theory that Catholicism or church attendance shapes the political or civic activism of the first generation. As we note, although this suggests that religious beliefs, belongings, and behaviors are not enhancing the involvement of immigrants in politics, neither are they reducing it. The results are inconsistent both with the theories of Verba, Schlozman, and Brady (1995) about the demobilization potential of Catholicism and with the findings of Jones-Correa and Leal (2001) about the unique electoral benefits of Catholicism to Latinos.
By contrast, the ANES models indicate that religious denomination and service attendance play a role in the political engagement of native-born Latinos and Anglos. For the former, Catholicism shapes civic and campaign activities, while attendance is associated with voter turnout. For Anglos, it is church attendance that matters, not denomination. These results are broadly consistent with those of Jones-Correa and Leal (2001), but taken together suggest that religion plays different political and civic roles for immigrants and native-born Latinos.
How do we understand these largely statistically insignificant LINES religious effects? Any fair assessment will encompass three potential explanations: issues with the survey data that led to null results; actual null results, which are nevertheless important to researchers; and mixed effects that appear as null results. [End Page 139]
First, no dataset is without its quirks. This may particularly be the case for understudied and more difficult to study populations, which present sampling challenges. In our examination of religion and participation among Latino immigrants, the LINES manifested its share of such problems. Although the two-wave design may be a positive feature for some research questions, the drop off in wave 2 respondents required supplemental respondents in the second wave. Unfortunately, several questions from the first wave were not asked of these fresh respondents,5 and the survey also has nontrivial missing data of the standard variety. However, the main religious results do not change when we impute the missing data, which adds to our confidence in the findings.
We also have questions about potential differences that may result from the use of two survey firms, as discussed. For example, slightly more than half of the respondents (51 percent) contacted by Latino Decisions preferred to take the survey in Spanish. In the second-wave surveys, conducted by Interviewing Services of America, about 95 percent preferred to take the survey in Spanish. This suggests nonrandom drop off, which was addressed through a new sample generated by the second firm. This could have introduced potential biases or inefficiencies into the data-set. Nevertheless, as we discussed, the results do not appear to change whether we take into account the separate survey efforts in the second wave, which is encouraging.
For the religion question, the LINES could have benefited from more appropriate questions that better capture the relevant dimensions of Latino and immigrant religion. For example, the emergence of charismatic spirituality is one of the most notable features of Christianity in Latin America and the United States, among both Catholics and Protestants. The LINES has no question to assess this important fissure. The contemporary rise of nondenominational Christianity is shared among Latino populations, yet the LINES asked respondents to choose between Catholic and Protestant labels. Other survey work finds that many Protestants do not identify with this label and may erroneously fall into an Other category without more appropriate response options. It is also difficult to assess the religiously unaffiliated due to a question-branching scheme that conflates church attendance with religious affiliation. Religion and politics scholarship understands these as different politically relevant aspects of religion. To be fair, much of the religion module was adopted from the ANES, but one might ask why the LINES retained Jewish as a response option for a study of Latino immigrants yet did not find a way to better assess charismatics, nondenominational Christians, or the unaffiliated.
Second, the null findings could reflect religious measures that are not, in fact, associated with the dependent variables. Although many scholars automatically discount null findings, academia is increasingly aware of the danger that doing so poses to scholarship. Annie Franco, Neil Malhotra, and Gabor Simonovits (2014) explain how null findings are an important part of the scientific process (see also Mervis 2014). Unfortunately, the tendency in social science is to not publish or even submit statistically insignificant results, which Franco and colleagues see as a “pernicious form of publication bias.” This serves to obscure a large swath of scientific results that would otherwise help advance the scholarly conversation. At the very least, it does nothing to discourage future researchers from replicating such past work. At worst, write Franco and colleagues, “if future researchers conduct similar studies and obtain significant results by chance, then the published literature on the topic will erroneously suggest stronger effects” (2014, 1504).
Third, we might see the results in the models as reflecting the complexity of religion in immigrant communities. Rather than positing a single effect, we might instead see the immigrant [End Page 140] interaction with religion as varying. In Levitt’s (2008) study of immigrants in Boston, although churches helped connect individuals to politics, the crucial factor was whether a church facilitated the interaction of immigrants with the native born. Churches that could provide this bridging social capital were more effective in promoting political engagement. By contrast, “members of stand-alone congregations, with few U.S. ties, were on their own” (780, emphasis added).
The results in this paper might be seen as evidence for such a complex understanding of the religious profile of Latino immigrants. Perhaps some are experiencing a form of segmented religious assimilation, which not only fails to help them adapt but contributes to an isolation associated with downward mobility. More generally, Stephen Warner sought to “remind students of assimilation (and of ethnic and racial minorities) that religion is a factor that they must take into account in their models” (1998, 103). We similarly hope that researchers of religion and assimilation will see that each literature has much to offer the other.
Although we cannot draw such conclusions without contextual parish-level data, we hope that this paper will help inspire more detailed survey and data collection efforts. They will be necessary if we are to better understand the interlocking, contemporary phenomena of immigrant growth and religious dynamism. The LINES allows us to bring some unique evidence to bear on the subject, but we also hope it will prove to be only one of many surveys that researchers will examine. Only from such collective efforts can we best understand the important and emerging topics at the intersection of Latino politics, immigration, and religion.
David L. Leal is professor of government at the University of Texas at Austin.
Jerod Patterson is a doctoral candidate in the Department of Government at the University of Texas at Austin.
Joe R. Tafoya is a doctoral candidate in the Department of Government at the University of Texas at Austin.
1. We do not include contributing money to a political campaign in the index, as the more general literature on political participation finds that it is uniquely shaped by disposable income.
2. We similarly do not include in the index the questions for giving money to a religious or a nonreligious organization.
3. Our imputation of LINES data started with our specifying the models of interest for Amelia II. The procedure assumes, as do we, that data are missing at random and then generates observations based on every parameter in the model, including the dependent variable. We tasked Amelia II with generating five imputation datasets, each with respondent identification numbers as unique cases with which to create unique values. Dependent variables were treated in a manner consistent with their nominal characteristics. Specifically, Amelia II generated integer values for whatever rate of civic or campaign participation it predicted respondents to have based on observed characteristics. The same prediction occurred for whether naturalized Latino citizens voted for president in 2012. The program stacked the five datasets into one export and was analyzed by Stata’s mi estimate protocol. The procedure was informed of the size and shape of the stacked dataset, as well as identifying variables for respondents, datasets, and variables with imputed observations. This entire process yielded a complete data frame and highly reliable parameters that range across multiple datasets.
4. The born again were more likely to attend a campaign rally but less likely to wear a campaign button, the Other religious were also less likely to wear a campaign button, and those who thought religion was important were less likely to donate to a campaign.
5. These include variables that are especially relevant to studying immigrant populations, such as the number of family members at home, identification with the country of origin, and the desire to eventually return home. Some of our models would have benefited from these questions, but we did not include them because the result would be a further loss of observations (in addition to those generated by the usual missing data).