Since the foundational work of Lipset and Rokkan (Reference Lipset, Rokkan, Lipset and Rokkan1967), the influence of religious orientations and beliefs on political preferences has been intensely debated, largely because of the fact that empirical research has systematically reached conflicting findings. On the one hand, multiple studies have posited that the strength of the link between religious inclinations and political preferences in multiple societies has diminished over time as a result of a general decline of the capacity of religion to maintain its influence in the political sphere (Reference DoganDogan 1995; Reference Dalton, LeDuc, Niemi and NorrisDalton 1996; Reference Franklin, Mackie and ValenFranklin, Mackie, and Valen 2009). On the other hand, some studies have not found sufficient evidence to support the notion of a universal decline in social cleavages or, in particular, the religious cleavage. Consequently, reports on the death of these cleavages should be considered exaggerated (Reference Brooks, Nieuwbeerta and ManzaBrooks, Nieuwbeerta, and Manza 2006; Reference ElffElff 2007).Footnote 1 Both Elff (Reference Elff2007) and Knutsen (Reference Knutsen2004) found evidence that the association between religion and political preferences has remained steady in many European countries, some of which have also seen processes of political realignment among both religious and secular communities. These processes may even lead to a rise in the association between religion and political preferences (Reference Kellstedt, Green, Guth and SmidtKellstedt et al. 1994; Reference LaymanLayman 2001; Reference Brooks and ManzaBrooks and Manza 2004).
In line with the comparative literature, the current research in Chile has not reached agreement with respect to the role of religion in the formation of political preferences. On the contrary, the link between religion and politics has been the focus of much academic debate. Some researchers have stated that religion is, nowadays, an irrelevant predictor of political preferences (Reference Tironi and AgüeroTironi and Agüero 1999; Reference Torcal and MainwaringTorcal and Mainwaring 2003; Reference Álvarez and KatzAlvarez and Katz 2009), mainly arguing that the postauthoritarian Chilean party system is mostly structured around the division between supporters and opponents of the civic-military dictatorship. Other specialists have claimed not only that the religious cleavage is a historically deep political division that dates to the nineteenth century but also that it remains fully active (Reference Valenzuela and ScullyValenzuela and Scully 1997; Reference Valenzuela, Scully and SommaValenzuela, Scully, and Somma 2007).
The goal of this article is to reveal how the religious cleavage has evolved in contemporary Chilean society while considering the degree of religious heterogeneity with which people experience religion. Once this last element is considered, processes of political dealignment (or even realignment) and continuity can simultaneously coexist in time and place. To evaluate this argument, we carry out a multinomial regression analysis based on data from opinion surveys conducted between 1998 and 2014 to determine whether the relationship between individuals’ ideological preferences and religious denominations has varied over time and whether such variation corresponds to their level of religious observance (as captured by their church attendance). By focusing on religious heterogeneity, the current article tries to reconcile the existent duality in the literature regarding the possible dissolution and permanence of the religious cleavage among contemporary democratic societies.
The Decline or Persistence of the Religious Cleavage
Comparative research on political behavior around the salience of the religious cleavages has been framed in two ways, which reach opposite conclusions (Reference KnutsenKnutsen 2004). The first argues that the influence of individuals’ religion on their political preferences has significantly eroded in many postindustrial societies over the past decades (Reference DoganDogan 1995; Reference Franklin, Mackie and ValenFranklin, Mackie, and Valen 2009; Reference DaltonDalton 2013). Most such studies explain this tendency as a relentless result of general processes of religious secularization.
Secularization, understood as “the process by which sectors of society and culture are removed from the domination of religious institutions and symbols” (Reference BergerBerger 1990, 107), encompasses a multiplicity of dimensions, including separation between church and state, the loss of influence and centrality of religious organizations, and the increasing valuation of scientific knowledge. In terms of individuals’ subjectivity, secularization involves a change in the codes with which individuals interpret the world, such that religious interpretive frameworks are abandoned and replaced with outlooks that are closer to science and secular moral. In terms of political behavior, secularization is expected to erode the influence of religion over individuals’ political preferences (Reference Esmer, Pettersson, Klingemann and DaltonEsmer and Pettersson 2007).
Varied empirical evidence points to this end. For example, Norris and Inglehart (Reference Norris and Inglehart2011) showed that support to religious parties has substantively declined in postindustrial societies over the past fifty years. In this same vein, Dogan (Reference Dogan1995) indicated that, in Europe, the influence of religion over electoral behavior has diminished as religious beliefs and practices have declined.
Theories of value change associated with Inglehart’s (Reference Inglehart1990) and Dalton’s (Reference Dalton1984, Reference Dalton2013) work on postmaterialism pose another theoretical mechanism for explaining the erosion of religion’s influence. These theories hold that in postindustrial societies, membership in social groups—religious, socioeconomic, or otherwise—has lost power to predict political preferences. In these contexts, political preferences have become progressively more individual and less conditioned by group membership. For Dalton, this weakening of group membership bonds is a consequence of the growth in educational achievement and the mass media’s widespread dissemination of political information in developed societies. This triggers a process of cognitive mobilization, whereby citizens gain higher levels of autonomy because they are increasingly able to process political information at lower cost and without the need to rely on political cues provided by political parties (Reference DaltonDalton 1984).
The second approach argues that the influence of the religious cleavage has remained steady in most cases (Reference KnutsenKnutsen 2004; Reference Brooks, Nieuwbeerta and ManzaBrooks, Nieuwbeerta, and Manza 2006). A possible reason for this persistence is that when a cleavage contributed to the development of political parties’ structure, the division tends to permeate in institutions across that society, such as religious organizations, social clubs, and trade unions (Reference ValenzuelaValenzuela 1995, Reference Valenzuela1999).
Interestingly, certain interpretations of secularization theory also support the idea that the religious cleavage has not vanished in societies where it was once prominent. For example, for Elff (Reference Elff2007), even if it is possible to observe a decline in levels of religiosity across a population, this does not necessarily imply that religious orientation has ceased to influence political preference. In the context of increasingly secular populations—in which ways of life different from those valued by traditional religion and morality grow and gain social legitimacy—the probability that religious individuals feel challenged by secular lifestyles increases. Thus, the religious cleavage may gain importance because of secularization (Reference ElffElff 2007).
Similarly, some studies have shown that the relationship between religion and political preferences can become sharper in those political contexts with strong controversies related to moral issues (e.g., abortion, sexual-minority rights, divorce). In such settings, a political realignment process may take place whereby the association between religious conservatism and right-wing preferences strengthens while antipathy toward religious orthodoxy rises among those with leftist preferences (Reference Kellstedt, Green, Guth and SmidtKellstedt et al. 1994; Reference LaymanLayman 1997; Reference Bolce and De MaioBolce and De Maio 1999; Reference Manza, Wright and DillonManza and Wright 2003).
The Religious and Social Context in Contemporary Chile
As in the rest of Latin America, Chile has been historically identified as a mainly Catholic country. This religious homogeneity, which has been sustained for more than four hundred years, has been challenged during the past four decades by the growth of Pentecostalism and, more recently, the irruption of large segments of the population that do not identify with any religious group (which we refer to as the irreligious population). Valenzuela, Bargsted, and Somma (Reference Valenzuela, Bargsted and Somma2013) referred to this trend as a “double pincer” that threatens the historical hegemony of Catholicism in Chilean society.
Survey data from the Center for Public Studies (Centro de Estudios Públicos, CEP) corroborate Valenzuela and colleagues’ (Reference Valenzuela, Bargsted and Somma2013) diagnosis. Figure 1 shows a sustained decline in the adult Catholic population, from 74 percent in 1998 to 61 percent in 2014. Simultaneously, Evangelicals stabilized around 17 percent from 2006 onward.Footnote 2 The irreligious population tripled between 1998 and 2014, from a scarce 6 percent to 19 percent. According to Valenzuela, Bargsted, and Somma (Reference Valenzuela, Bargsted and Somma2013), this unprecedented growth implies that secularization is a mass phenomenon and not strictly linked to the country’s elite, as it was historically.
Figure 1 Evolution of religious denomination in Chile. Source: CEP surveys.
The religious transition that the Chilean society has experienced over the past two decades can be seen not only in the trends of religious predominance but also in levels of religious observance, as captured by church attendance among believers. Figure 2 shows how frequency of church attendance has evolved by religious denomination. The percentage of frequently attending Evangelicals is not only much higher than that of Catholics, but the percentage of frequently attending Catholics shows a clear downward trend, as opposed to Evangelicals, for whom attendance has been relatively constant, though with some annual fluctuations.
Figure 2 Evolution of frequency of religious service attendance by denomination. Source: CEP surveys.
Does this mean that the Chilean population is becoming more secular? Although we believe that there is no unambiguous response, there is certainly some degree of increasing secularization, to the extent that large portions of the population have progressively moved away from religious institutions and the most observant segment of the largest religious group has declined significantly. On the other hand, the irreligious population should not be considered as mainly atheist or agnostic, but rather as subjects who are not members of religious organizations or groups. Valenzuela, Bargsted, and Somma (Reference Valenzuela, Bargsted and Somma2013) stated that 64 percent of people who do not profess any religion believe in God. Consequently, it is more accurate to speak of a process of increasing religious fragmentation that is marked by increased heterogeneity of religious experience.
Economic and Political Context
Most macrosocial theories of values change, such as secularization or cognitive mobilization, assume that socioeconomic modernization is a key driver of social change (Reference InglehartInglehart 1990; Reference Norris and InglehartNorris and Inglehart 2011). Coincidentally, Chilean society has experienced great economic and social change during the past few decades that could be fundamental drivers of religious decline. Three specific trends are highly relevant: a high rate of economic growth, expansion of secondary and tertiary education, and the rapid increases in internet accessibility.
Figure 3 shows how these variables evolved during the period from 1960 to 2015. As the figure shows, Chilean society has experienced incredibly high growth in secondary and tertiary school enrollment.Footnote 3 Secondary school enrollment rose from 46 percent in 1970 to greater than 90 percent in recent years, and tertiary school enrollment rose from 9 percent in 1970 to 78 percent in 2013. The country has also seen drastic economic changes. After the return to democracy in 1990, the per-capita gross national domestic product (GDP) increased more than sixfold; per-capita income was CLP$2,401 in 1990, it reached $15,741 in 2014. Similarly, internet access has dramatically increased since the mid-1990s. In only twenty-two years, internet users rose from 3 percent of the population to 78 percent. Thus, these changes indicate that economic well-being, formal education, and information access for the average Chilean have all substantially increased over the past twenty-five years.
Figure 3 Evolution of key indicators of socioeconomic development in Chile, 1960–2015. Source: World Bank Indicators Database.
As in other Latin American countries, the contemporary panorama of Chilean politics is strongly influenced by the legacy of the civic-military dictatorship (1973–1990) and the gradual consolidation of the current democratic regime (Reference NaviaNavia 2010). An important aspect of these processes is the resurgence of the new party system. Since the transitional government of Patricio Aylwin (1990–1994) until 2014, Chilean parties have grouped around two big coalitions that are organized around whether their members supported or opposed the civic-military dictatorship. Right-wing parties (Renovación Nacional and Unión Demócrata Independiente) were aligned with the military regime, forming the Alianza por Chile coalition (currently joined by some smaller, more liberally oriented right-wing parties and known as Chile Vamos), whereas the centrist party (Partido Demócrata Cristiano) and center left-wing parties (Partido Socialista, Partido por la Democracia, and Partido Radical Social Demócrata) were joined under the umbrella of the Concertación de Partidos por la Democracia. The Concertación parties governed the country during four presidential periods between 1990 and 2010. The Alianza por Chile then governed for one presidential period until 2014, when the Concertación parties, joined by the Communist Party under the label of the New Majority, retook the government.
The postauthoritarian Chilean party system has some notable characteristics. It is widely considered a highly institutionalized party system (Reference Mainwaring and ScullyMainwaring and Scully 1995; Reference Payne, Zovatto, Carrillo and AllamandPayne et al. 2003, Reference Castiglioni and RoviraCastiglioni and Rovira 2016; but see Luna and Altman [Reference Luna and Altman2011] for a more skeptical perspective) with relatively low electoral volatility (Reference Mainwaring and ZocoMainwaring and Zoco 2007). In ideological terms, the system is aligned along a left–right continuum (Reference Valenzuela, Scully and SommaValenzuela, Scully, and Somma 2007; Reference Kitschelt, Hawkins, Luna, Rosas and ZechmeisterKitschelt et al. 2010; Reference Gamboa, López and BaezaGamboa, López, and Baeza 2013), which implies, as Kitschelt and colleagues (Reference Kitschelt, Hawkins, Luna, Rosas and Zechmeister2010) empirically showed, that labels of “left” and “right” are strongly connected to bundles of issues that follow party lines. The right-wing parties favor free-market economics and take conservative positions on moral and cultural issues, whereas the left-wing parties promote a more active role of the state in redistribution and lean toward a more liberally oriented moral agenda. The Partido Demócrata Cristiano, as the centrist party, has tended to adopt more conservative policies on moral issues than its coalition partners but certainly advocates a more active role of the state in economic matters. Last, it is also worth noting that levels of political disaffection in respect to Chilean political parties are high and trending upward. For example, Huneeus (Reference Huneeus2014) showed that political parties have systematically ranked lowest in trust among all political and nonpolitical institutions since the early 1990s. Likewise, levels of identification with political parties and rates of electoral participation have declined sharply since the return to democracy (Reference SiavelisSiavelis 2009; Reference Corvalan and CoxCorvalan and Cox 2013).
During both the dictatorial and the democratic periods, the Catholic Church has been an influential actor, albeit with different political strategies and alliances. After the fall of democracy in 1973, the church exercised a strong leadership role in the defense of human rights abuses by the dictatorship, positioning itself as a key opponent of the military regime and enjoying broad social legitimacy (Reference StrassnerStrassner 2006). After the return to democracy, the church hierarchy progressively reoriented its role of authority over morals and values, particularly those related to sexuality and family planning. In exercising this role, the church has opposed the more liberally oriented agenda promoted by the Concertación governments, though with some dissident voices in the minority (Reference Guzmán, Seibert and StaabGuzmán, Seibert, and Staab 2010). This agenda included approval of the first divorce law in 2004, the distribution of emergency contraceptives in public hospitals, sex education programs in public schools, and the approval of a civil-union agreement for same-sex and opposite-sex relationships, among other things. The Catholic Church hierarchy has exercised its opposition to these policies, over public opinion, as well as through the two right-wing political parties and within the Concertación through the Partido Demócrata Cristiano (Reference StrassnerStrassner 2006; Reference Guzmán, Seibert and StaabGuzmán, Seibert, and Staab 2010). As a consequence, Valenzuela, Scully, and Somma (Reference Valenzuela, Scully and Somma2007) argued that the Catholic world during the current democratic era is relatively divided. A more conservative group is lined up behind the figure of Jaime Guzmán (founder of the Unión Demócrata Independiente party and an important adviser to Augusto Pinochet) and the right-wing parties, and another more centrist faction more strongly committed to the defense of human rights and condemnation of the military regime.
Religion and Political Preferences in Chile
Discussion of the relevance of the religious cleavage in Chile has, to some extent, reproduced the duality observable in comparative research. Some authors have maintained that traditional social cleavages still operate in Chile, and others have affirmed that a new division related to the legacy of the military dictatorship has displaced them. We briefly discuss both perspectives.
According to Valenzuela and his collaborators, the Chilean political party system has been historically shaped by two great social cleavages. Since the nineteenth century, there has been a religious cleavage that distinguishes Catholic from secular forces. By the twentieth century, and in light of growing social problems, a second socioeconomic dimension, which differentiated orientations for and against economic liberalism, overlapped with the religious dimension (Reference ValenzuelaValenzuela 1995). These two original fissures are deeply rooted in society, finding support in organized society and in institutions such as the Catholic Church, Freemasons, social clubs, business associations, and unions, among others (Reference Valenzuela and ScullyValenzuela and Scully 1997). Thus, for these authors, it is highly unlikely that the base on which the political party system stands can be entirely brought down by fissures related to more recent political processes (Reference Valenzuela, Scully and SommaValenzuela, Scully, and Somma 2007).
For Tironi and Agüero (Reference Tironi and Agüero1999; see also Reference Tironi, Agüero and ValenzuelaTironi, Agüero, and Valenzuela 2001), the interruption of the Chilean democratic system reformulated the party’s structure, pushing aside the oppositions that shaped the political system before the dictatorship; therefore, religion would no longer determine the political preferences of citizens. Since the return to democracy, the political landscape has supposedly been shaped by opposition between those who supported the military dictatorship and those who supported the return to democracy, or more simply, between authoritarianism and democracy (Reference Huneeus and MaldonadoHuneeus and Maldonado 2003). On the basis of a similar diagnosis, Torcal and Mainwaring (Reference Torcal and Mainwaring2003) argued that Chile is an example of how political processes can erode social cleavages and create new oppositions. From this perspective, the emergence of cleavages in the party system is the product of political agency rather than an expression of general social divisions.
Recent investigations have not been able to overcome this rivalry. Raymond and Feltch (Reference Raymond and Barros Feltch2014), following the theory of issue evolution, argued that Chile is in the midst of a realignment process whereby religious voters are shifting their views from left- to right-wing positions. Accordingly, a consolidation of these changes is likely to bring about an increase in the effect of religion on ideological preferences. Bargsted and Somma (Reference Bargsted and Somma2016) come to opposite conclusions. They affirm and provide empirical evidence, indicating that Chilean society is undergoing a process of political dealignment whereby the explanatory capacity of all cleavages (class, religion, and political) is in decline.
Finally, it is important to mention a political particularity of the Evangelical population. According to the (scarce) existing evidence, the Evangelical world is highly depoliticized. For example, Fontaine and Beyer (Reference Fontaine and Beyer1991) showed that Evangelicals are less likely than the general population to identify as left, right, or center in political terms. Fediakova and Parker (Reference Fediakova and Parker2009) showed that the Evangelical population has low levels of interest in politics and low willingness to participate. Moreover, the Evangelical worldview sees politics as opportunistic and inconsistent with Christian values (Reference Fediakova and ParkerFediakova and Parker 2009). Therefore, in Chile, Evangelical religious groups have not shown signs of partisan or ideological alignment, and therefore we believe that Evangelism in Chile is more in line with the descriptions of Martin (Reference Martin1991), who considered that Evangelicals an apolitical group both in institutional and in partisan terms.
Hypotheses
Both the comparative and the Chilean literature on political behavior have reached conflicting conclusions regarding the link between religion and political preferences. We propose that this contradiction can be resolved by recognizing that religious communities are heterogeneous groups and that religious experiences can differ in at least two dimensions: by denomination, which entails differences in doctrinaire beliefs and practices, and by level of observance.
Given the decline in religious identification and observance in Chile, along with the socioeconomic gains across the population, we expect that for an increasingly large number of adults, religion will no longer operate as an interpretive framework that conditions their political preferences:
Hypothesis 1: The effect of Catholic or Evangelical denomination on individuals’ propensity to prefer a right-wing ideological position or, to a lesser degree, a centrist position, has declined over time.
Even though a decline suggests a displacement of religion as a relevant framework for forming political opinions, this does not mean that the magnitude of this phenomenon is similar across all religious groups. Just as Catholic groups have traditionally been associated with right-wing positions, Evangelicals have historically been more reluctant to adopt a political identification:
Hypothesis 2: The decline in the effect of religious denomination on right-wing ideological positions and, to a lesser extent, centrist positions, has been greater for Catholics than for Evangelicals.
Third, in the context of increasing religious fragmentation, religion cannot cease to have an influence on the ideological preferences of the growing irreligious population, but it could increase the likelihood that those who identify as religious will feel challenged by modern and secular lifestyles. Indeed, as shown by Elff (Reference Elff2007) and Manza and Wright (Reference Manza, Wright and Dillon2003), when studying other countries, the reaction of the religiously observant to secular trends can reinforce their original ideological positions instead of eroding them:
Hypothesis 3: The effect of church attendance over individuals’ propensity to identify with a right-wing ideological position and, to a lesser degree, with a centrist position, has increased over time.
Last, if we combine the predictions from hypotheses 2 and 3, there are reasons to believe that the varying effect of church attendance could differ by religious denomination. Indeed, if the effect of Evangelical denomination over right-wing and centrist preferences has decreased less over time than for Catholics—since the former has been historically apolitical—then the positive effect of church attendance could also increase less over time for Evangelicals:
Hypothesis 4: The increase in the effect of church attendance on right-wing ideological positions, and to a lesser degree, centrist positions, has been stronger for Catholics than for Evangelicals.
Data and Methodological Design
Our empirical analysis is based on survey data from CEP, which has conducted periodic surveys in Chile with multistage probabilistic samples representative of the national adult population since 1987.Footnote 4 Although the survey module of religion and politics has remained relatively constant for a significant number of years, prior to 1997, the data lose comparability, because of either changes to questions included or lack of continuity in the phrasing of questions or response categories. Accordingly, the analysis presented here covers the period between 1998 and 2014, a seventeen-year window that allows us to observe changes in the variables of interest.
To maximize the time span of our analysis, some control variables, such as income or marital status, have been discarded for lack of continuity. In turn, we rectified some variables to make response categories comparable over time. The combined database has 43,094 cases, corresponding to twenty-nine surveys carried out during the period in question. In the statistical models shown that follow, the number of cases slightly decreases because of nonresponse to one or multiple questions, and the analysis includes only the three major religious groups in Chile (Catholics, Evangelicals, and irreligious). Alternatively, a residual category that gathers several minority religious groups (e.g., Jewish, Muslim, Buddhist) could have been added to the analysis, but given the low number of cases and heterogeneity of the groups that would compose this category, we decided to exclude them from the analysis.
Dependent variable
We capture respondents’ ideological preferences using a self-identification item with left, center, or right positions in response to the question: “As you know, people in our country traditionally define their political positions as nearer to the left, center, or right. In this card, we represent the different political positions. Please indicate with which position do you identify more, or which you sympathize the most.” The response categories are: “right,” “center-right,” “center,” “center-left,” “left,” “independent,” and “none.” To simplify the statistical analysis, we grouped the response options for “right” and “center-right” under the common label “right”; “left” and “center-left” as “left”; and the responses “none,” “independent,” “don’t know,” and “no response” under the common category “no preference.”Footnote 5 This results in a dependent variable with four categories (right, center, left, and no preference). The addition of the “No Preference” option led us to employ a nominal-level dependent variable, as opposed to what could be a simpler ordinal variable. However, it is crucial to incorporate this alternative: if our hypotheses 1 and 2 are correct, the decline in the effect of religious denomination on right-wing preferences could be motivated either by an increase of the proportion of Catholics and Evangelicals with no preference instead of a right-wing preference (i.e., dealignment) or by a progressive replacement of their right-wing preferences with centrist and left-wing positions (i.e., realignment). Instead, if we employ ordinal coding of the dependent variable (which implies dropping those with no preference from the analysis) we would be able to observe only a realignment trend. In other words, by incorporating the “no preference” category, we allowed for our statistical analysis to capture dealignment and realignment trends simultaneously.
We used the ideological self-identification item for several reasons. First, and following our description of the Chilean party system, the policy positions of Chilean parties have been clearly structured around a left-right continuum (Reference MoralesMorales 2008). This has led multiple scholars to note that the labels of “left” and “right” constitute a meaningful political heuristic that enables citizens to easily synthesize their positions on several issues into a single ideological dimension (Reference FontaineFontaine 1995; Reference Harbers, de Vries and SteenbergenHarbers, de Vries, and Steenbergen 2013; Reference Bargsted and SommaBargsted and Somma 2016). Second, this measure enables us to obtain data for a greater number of points in time than other measures, such as vote intention. Moreover, some studies have argued that vote intention or recall may have limitations for the study of social cleavages and political preferences because campaign effects and the specific characteristics of candidates can strongly influence electoral decisions (Reference Brooks and ManzaBrooks and Manza 2004; Reference Barone, Lucchini and SartiBarone, Lucchini, and Sarti 2007). Nonetheless, electoral and ideological preferences are very strongly related (see the online supplement).
Independent variables
To contrast our hypotheses, we employed two key independent variables. Religious denomination was measured by the following question: “Could you tell me the religion or church to which you belong or feel closest?” This variable is introduced in the multinomial logit models as two binary variables—whether the respondent identifies as an Evangelical or otherwise or as irreligious or otherwiseFootnote 6—Catholic was the reference category. An important detail to consider is that the category “Evangelical” mainly comprises Pentecostals.Footnote 7 We did not distinguish Pentecostals from other forms of Evangelism because most surveys do not make a distinction between such groups.
By the second variable, frequency of church attendance—and following the specialized literature (Reference Manza, Wright and DillonManza and Wright 2003)—we captured level of religious observance using the following question: “How often do you attend religious services?” Response categories varied across time periods, although they are completely commensurable to the response scheme employed in most surveys. The ordinal response categories used in the study were: 0 (never), 1 (once a year), 2 (a few times during the year), 3 (once a month), and 4 (more than once a month). Admittedly, this is an imperfect measure, as there are other relevant ways people practice religion, such as praying or participating in religious festivities. Accordingly, some scholars have argued in favor of using composite measures instead of single measures of religiosity or religious observance (see, e.g., Reference Valenzuela, Scully and SommaValenzuela, Scully, and Somma 2007). However, we are constrained to this single variable because it is the only measure available for this purpose.Footnote 8 That said, from a theoretical point of view, we believe the variable has an important advantage over other religious behaviors, because it captures variation in individuals’ exposure to the doctrinal messages of religious leadership, arguably a key contributor to the adoption of more conservative-minded positions on political matters.
In addition to religious variables, we included four control variables in our statistical models. First was year of the survey, which varies from 1998 to 2014. Because of the results of the exploratory analysis, this variable was introduced as a linear term in the statistical models; consequently, it is treated as continuous and not as fixed effects. This allows for us to test the hypotheses of linear trends in the coevolution of our variables of interest. We centered the year of the survey at its minimum value (1998) to facilitate interpretation of the intercept and the interactive variables included in the models. The second control variable is birth cohort, which groups the year of birth of survey respondents in groups of five years, except for the first cohort, which groups all subjects born before 1920 to avoid a numerically small cohort. This scheme was used following common practices in demography (Reference Yang and LandYang and Land 2013). The variable has sixteen categories, ranging from 0 (born between 1900 and 1920) to 15 (born between 1991 and 1996). Gender was captured as a binary variable, where women were assigned the code 1. Finally, years of education, which corresponds to respondent-reported number of years of study, varies from zero to twenty-five years. This measure was prioritized over educational attainment, as it has greater availability over time.Footnote 9 (See the online supplement for summary statistics of all variables included in the regression models).
Statistical Results
To study how religion and politics have coevolved in Chilean society, we estimated two sets of multinomial logistic regression models. The first set describes the evolution of the relationship between ideological preferences and individuals’ religious denomination, including the irreligious population. The second set adds frequency of church attendance and appropriate interactions as additional predictors. The irreligious population is not included in this analysis because the CEP surveys omitted this information for these respondents on some occasions. Moreover, from the CEP records, the irreligious attend church very rarely, so there is very low variance for this group.
Religious denomination and political preferences
Table 1 shows the regression estimates of religious denomination, gender, education, year, and birth cohort, predicting ideological preferences. Model 1 is presented as a starting point for the overall effects of religious denomination on respondents’ preferences for the entire study period. According to these estimates, respondents’ propensity for a right-wing preference rather than a left-wing one is slightly lower among Evangelicals than Catholics and is considerably lower among the irreligious. In both cases, however, the statistical difference is significant (p < 0.01). Compared to Catholics, the odds that Evangelicals identify with a right-wing position are 15 percent lower (e –0.165 = 0.85), while irreligious respondents are 57 percent lower (e –0.833 = 0.43). The same pattern occurs with centrist ideological preferences, although the magnitude of coefficients is somewhat smaller. The coefficients also indicate that Evangelicals and the irreligious are, compared with Catholics, significantly (p < 0.01) less likely to have no preference than are respondents with a leftist position.
Table 1 Multinomial logit regression models of ideological preference by religious denomination and control variables.
Note: Standard errors in parenthesis. Reference category is “left-wing.” AIC = Akaike information criterion.
***p < 0.01; **p < 0.05; *p < 0.1.
The second model in Table 1 includes an interaction term between religious denomination and year. With this multiplicative term, we test our first two hypotheses, which state, in brief terms, that the influence of religious denomination over individuals’ ideological preferences has declined over time (hypothesis 1), and this decline has been sharper for Catholics than for Evangelicals (hypothesis 2). Given that Catholics are the reference category, the trajectory of this group’s political preferences are captured in the coefficients of the year variable. While all three coefficients are significant (p < 0.01), it is negatively signed for right-wing preferences and positive for centrist preferences and the “no preference” option. This confirms that Catholics’ propensity to identify with right-wing positions has declined. Among the double interaction between year and Evangelical denomination, there is only a marginally significant (p < 0.1) and negative coefficient predicting right-wing preferences, thus indicating a slightly steeper decrease of these preferences in comparison to those of Catholics. Among the irreligious population, only the interaction with year predicting the “no preference” option is significant (p < 0.01) and positive. This implies an increase in time for this alternative among the irreligious. To provide a substantive interpretation of these results, Figure 4 represents how the predicted probabilities—based on model 2—of each ideological position vary in time according to respondents’ religious denominations.Footnote 10
Figure 4 Predicted probabilities (with 95% confidence intervals) for each ideological preference by religious denomination and year.
The behavior of Evangelicals and Catholics is very similar. Both groups present a small decline of approximately four percentage points in identifying with left-wing preferences, and a slight increase of two and four percentage points, respectively, in the probability of identifying with a centrist position. In contrast, larger differences emerge when we consider right-wing preferences and the “no preference” alternative. For Catholics, the probability of identifying with a right-wing position decreased from 0.26 in 1998 to 0.16 in 2014; among Evangelicals, this declined from 0.28 to 0.13. This decline has been offset by a strong increase of thirteen and fifteen percentage points for Catholics and Evangelicals, respectively, in the probability of not mentioning any ideological preference. Among the irreligious, we observed even more pronounced growth of “no preference,” from a probability of 0.37 in 1998 to 0.64 in 2014. As can be observed, this sharp increase of the “no preference” alternative has been at the expense of left-wing and, to a lesser extent, right-wing preferences.
In sum, a process of political dealignment has been taking place in Chilean society over the past two decades. The propensity to not identify with traditional ideological positions has grown markedly for the three religious groups, although growth of the ideologically apathetic segment has been fueled by different sources. Consistent with hypothesis 1, the increase among Catholics and Evangelicals has mainly been at the expense of right-wing ideological preferences, whereas among the irreligious, this has occurred mostly at the expense of left-wing preferences. In contrast, preferences for centrist positions have remained relatively stable for all religious groups. These results also lead us to reject hypothesis 2, given that Catholics and Evangelicals have followed a similar trend. In fact, and contrary to the conventional perception of Evangelicals as an apolitical group, the empirical models indicate that their ideological preferences are distributed very similarly to those of Catholics, and their decline in identifying with right-wing positions has occurred at a slightly higher rate than for Catholics.
Also important to note is the remarkable increase in ideological apathy, as expressed by the sharp growth of the “no preference” alternative across all groups. This trend is consistent with research showing that the Chilean political system is currently affected by high levels of political apathy, whether measured by low levels of trust and identification with political parties or by low rates of electoral participation (Reference Luna and AltmanLuna and Altman 2011; Reference SiavelisSiavelis 2009; Reference Corvalan and CoxCorvalan and Cox 2013; Reference HuneeusHuneeus 2014). This growth in ideological apathy could cast some doubts on the importance and suitability of our measure of ideological preference. We believe, however, that this is not the case for two reasons. First, the distribution of left and right preferences among those who identify with ideological positions, despite an overall decrease, still captures the degree of “baseline” ideological support on which the parties can potentially rely. Likewise, changes in ideological balance in the population could presumably have direct consequences for the policy platforms and electoral strategies that political parties pursue. Second, some additional analysis shown in the online supplement suggests that respondents’ ideological positions, among those who continue to locate themselves on the scale, have remained highly correlated to several political opinion variables, such as government approval, evaluation of political leaders, and perception of the economy. These complementary results suggest that Chileans preferences for left or right—at least among those who continue to adopt these positions—continue to be a relevant political disposition that conditions respondents’ opinions on more specific political matters.
Religious observance and political preferences
We now consider hypotheses 3 and 4, that the effect of church attendance over Catholics’ and Evangelicals’ inclination to prefer a right-wing position, and less so a centrist one, has increased over time (hypothesis 3) and that this trend should be stronger among Catholics (hypothesis 4). To test hypothesis 3, model 3 from Table 2 incorporates an interaction between frequency of church attendance and year. The results show that church attendance has a positive, significant coefficient (p < 0.01), thus predicting right-wing and centrist ideological positions, but that such effect interacts positively and significantly in time only for right-wing preferences (p < 0.05). Instead, the interactions between church attendance and year predicting either a centrist position or “no preference” are both negative but not significant. Therefore, as Catholics and Evangelicals attend to church more frequently, not only do their chances of mentioning a right-wing position rise, but also this association becomes stronger moving from 1998 to 2014—favorable evidence for hypothesis 3.
Table 2 Multinomial logit regression models of ideological preference by religious denomination and church attendance.
Note: Standard errors in parenthesis. Reference category is “left-wing.” FCA = frequency of church attendance. AIC = Akaike information criterion.
***p < 0.01; **p < 0.05; *p < 0.1.
To test hypothesis 4, we include in model 4 of Table 2 a triple interaction of religious denomination, church attendance, and year. A key result refers to the significant double interaction (p < 0.05) between year and church attendance, predicting right-wing preferences. This indicates that the effect of church attendance becomes stronger for Catholics over time. In contrast, the triple interaction of Evangelical denomination, church attendance, and year is not significant for any response category. However, the triple interaction coefficient predicting a right-wing preference is almost of the same size and the opposite sign of the respective double interaction for Catholics. This implies that the effect of church attendance among Evangelicals does not increase over time as it does for Catholics.
We again plotted the predicted values of model 4 in Figure 5. In 1998, Catholics constituted a relatively homogeneous religious group, with no significant differences in their ideological preferences associated with frequency of church attendance (the difference between the least and most frequently attending groups are never statistically different; their confidence intervals overlap). However, as time passed, the differences systematically increased for right-wing preferences and for the “no preference” option. Among Catholics who go to church more than once a month, the probability of having a right-wing position decreased from 0.27 in 1998 to 0.21 in 2014. In contrast, the respective decline among Catholics who never go to church was from 0.24 to 0.11. Consequently, migration from right-wing preferences to ideological apathy has occurred at a substantially lower rate among frequently attending Catholics compared to nonattending Catholics. At the same time, nonattending Catholics have increased their probability of having no ideological preference at higher rates than have frequently attending Catholics. The difference in the probability of not opting for any ideological label between both groups rose from four percentage points in 1998 to thirteen percentage points in 2014. This implies that Chilean Catholics have become a politically more heterogeneous group.
Figure 5 Predicted probabilities (with 95% confidence intervals) for each ideological preference by religious denomination, church attendance, and year.
Interestingly, this divergence observed among the Catholic population is not reproduced among Evangelicals. As is shown in the second row of Figure 5, the differences between Evangelical subgroups are neither increasing nor decreasing. Evangelicals have evolved consistently as a politically homogeneous group, and differences by levels of church attendance in right-wing preferences and “no preference” are never significant across time. Accordingly, we find support for hypothesis 4.
In sum, Catholic and Evangelical religious groups are experimenting a generalized process of ideological dealignment, whereby higher levels of ideological apathy are replacing traditional preferences. This process is particularly visible with respect to right-wing preferences, for which the decline is more abrupt in the two religious groups. However, while Evangelicals have often behaved as a politically homogeneous group, the ideological preferences of Catholics have followed a process of differentiation associated with level of religious observance.
Conclusion
In this article, we have studied the variable influence of religious denomination and observance on Chileans’ ideological preferences over the past two decades. Our results have shown that Chilean society has experienced a generalized process of political dealignment, whereby the proportion of the population, religious and irreligious, that ceases to identify with ideological positions has strongly increased. However, the magnitude and direction of this process varies substantially among the different groups. While Catholics and Evangelicals have moved mainly from right-wing preferences to ideological apathy, the irreligious have migrated predominantly from the left to ideological apathy. Moreover, the magnitude of this dealignment is larger for the irreligious than for Catholics and Evangelicals. This pattern is quite interesting because it suggests that the very fact of identifying with a religious denomination has, if anything, undermined the process of dealignment—that is, people who identify with a religion migrate more slowly from right, center, or left positions to “none.”
Second, frequency of church attendance among those who identify with a religious group reveals the important nuances of the dealignment process. Indeed, we found an important degree of ideological persistence among frequently attending Catholics, which has led them to migrate more slowly from their predominantly right-wing preferences to ideological apathy. For many people in this group, religion remains a guiding framework for their political preferences; in contrast, nonattending Catholics have defected from this ideological position at substantially higher rates. This result has two interesting implications. First, the fact that those with any religious inclination have declined less from their ideological preferences than those without religious inclination is attributable, at least in part, to the behavior of the most observant Catholics. Second, as Catholics have shrunk in numbers (even though they remain by far the largest religious group in Chile), they have also become an increasingly politically heterogeneous group.
How can we explain this pattern of higher ideological resistance among frequently attending Catholics? In a context of an increasingly secularized society that is marked at the political level by the advance of a more liberally oriented legislative agenda, the probability that religious individuals understand themselves in conflict with the secular world increases. In other words, signs of ideological resistance can be interpreted as a reaction to an increasingly adverse social context in which new secular lifestyles challenge conservative and traditional ones. This phenomenon is similar to results from several studies obtained for North America, where processes of political realignment have occurred among some religious groups (Reference Kellstedt, Green, Guth and SmidtKellstedt et al. 1994; Reference LaymanLayman 2001; Reference Brooks and ManzaBrooks and Manza 2004). The Chilean case, for its part, highlights how the dealignment and continuity of the religious cleavage can coexist within a single society.
Although we believe our results are suggestive, there is an important aspect that has not been fully addressed and that should be considered more carefully in future research: sources of social change. It would be interesting to identify whether the observed dealignment and ideological resistance trends are occurring as a result of processes of individual change whereby people alter their own ideological preferences, or whether they are occurring as a result of generational replacement. Even though our analyses have found that birth cohorts have little statistical influence over ideological preferences, it would be interesting to evaluate whether highly observant young cohorts continue to identify in time with more conservative ideological positions. This kind of information would suggest that the lower migration rates toward ideological apathy among frequently attending Catholics would constitute a generational phenomenon, consequently encouraging political differences between religious groups to persist as a long-term social division within Chilean society.
We have also found important similarities and differences between Catholics and Evangelicals that demand further analysis. Contrary to the general expectation of the apolitical nature of Evangelicals, we have shown that members of this group have migrated from a predominately right-wing position toward ideological apathy at similar rates to those of Catholics, but church attendance does not operate as a relevant moderator of this process among Evangelicals. The possible reasons for and mechanisms behind this difference are not obvious and seem to be a relevant topic for future inquiry. More generally, we currently know very little about Evangelical political behavior, which seems increasingly odd, as Evangelicals have become a large group in Chile and are rapidly gaining in numbers in several Latin American countries (Reference Somma, Bargsted and ValenzuelaSomma, Bargsted, and Valenzuela 2017).
Last, though not the central focus of our article, our analysis has shown that the proportion of Chileans who do not identify with ideological positions has grown considerably during the period under observation. Given the sheer magnitude of this trend and the historical alignment of the party system along the left-right continuum, this movement should be cause for deep concern and the subject of future investigation, because it could have a great impact on the aggregate stability of citizen preferences and the future of the party system itself.
Additional File
The additional file for this article can be found as follows:
• Online Supplement. Contains further details about measurement and the data analysis shown in the article. DOI: https://doi.org/10.25222/larr.309.s1
Acknowledgements
Our research was supported by a CONICYT-FONDECYT Initiation Grant (no. 11140746) provided by the Chilean Ministry of Education, as well as by the Centre for Social Conflict and Cohesion Studies (Chilean grant CONICYT/FONDAP/15130009). Previous versions of this article benefited from the comments of participants at the 2015 annual conference of the World Association for Public Opinion Research (Buenos Aires, Argentina), the Permanent Seminar of the Institute of Political Science of the Pontificia Universidad Católica de Chile, and comments from Nicolás Somma, Beltrán Undurraga, and the three anonymous LARR reviewers.
Open access
