In the last few decades, Latin America s religious landscape has changed dramatically. Research Note Mapping Religious Change in Latin America

Transcription

1 Research Note Mapping Religious Change in Latin America Nicolás M. Somma Matías A. Bargsted Eduardo Valenzuela ABSTRACT Using Latinobarometer survey data, we study the evolution of religious identities among the adult populations of 17 Latin American countries between 1996 and We find several interesting patterns. First, the current religious landscape is highly dynamic and is becoming increasingly pluralist among a majority of countries. Changes derive not only from the growth of Evangelicals, as commonly assumed, but also from the sharp rise in irreligious individuals. Second, religious change cannot be convincingly explained by important theories such as secularization, religious economies, and anomie. However, the predictions derived from anomie theory seem more useful for understanding Evangelical growth. Finally, our cohort analysis indicates that aggregate religious change largely results from individual-level change across time religious conversion and apostasy rather than from generational replacement. Still, there are interesting variations across countries in that respect. In the last few decades, Latin America s religious landscape has changed dramatically (Burdick 2010; De la Torre and Martín 2016; Hagopian 2009; Levine 2014; Stark and Smith 2012). The Catholic monopoly that prevailed for four centuries over the whole region started to crumble from the 1950s on, as Pentecostal groups started to make significant inroads among the popular masses, to the point that one pathbreaking study asked whether Latin America was turning Protestant (Stoll 1990). In Brazil and Caribbean countries, religious cults with African roots like Umbanda, Candomblé, and spiritismo also gained salience and even expanded to countries like Uruguay and Argentina, which were traditionally foreign to these religious traditions. More recently, the region also saw a dramatic increase in the number of people reporting no religious affiliation at all, suggesting incipient secularization. Nicolás M. Somma is an associate professor of sociology. nsomma@uc.cl. Matías A. Bargsted is an assistant professor of sociology. mbargsted@uc.cl. Eduardo Valenzuela is dean of the Social Sciences School and professor of sociology. evalenzc@uc.cl. All are at the Instituto de Sociología, Pontificia Universidad Católica de Chile. Copyright 2017 University of Miami DOI: /laps.12013

2 120 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 The Catholic Church has reacted in many ways to this double-edged competition, from Afro and especially Evangelical groups on the one hand and secularism on the other. During the 1970s, as Evangelicals were evidently threatening the Catholic monopoly, the church espoused a progressive Catholicism inclined to social action and the improvement of the conditions of the poor (Smith 1991; Levine 1986). In countries such as Brazil and Chile, it also denounced human rights violations by authoritarian regimes (Gill 1998). During the 1980s it had to face an internal grassroots renovation movement, the Charismatic Renewal, which combined the Pentecostal emphasis on charisma and faith healing with the adoration of the Virgin that was the hallmark of Catholicism (Cleary 2011). We are interested in Latin America not only because it currently has one of the most dynamic religious landscapes in the world (Pew Research Center 2014), but also because it combines two central concerns for sociologists of religion: religious competition a main topic among those studying religion in the United States and secularization a traditional theme among European scholars (Valenzuela et al. 2013). Latin America allows for exploring how secularization and competition combine, whether they reinforce or repel each other, or whether the forces that drive one also drive the other. It is surprising that, to our knowledge, no study has traced recent religious change in Latin America using cross-country, comparable statistical data (but see Corporación Latinobarómetro 2014). A few studies provide cross-national statistics about religious identities, but for a single time point (Parker 2009a, b; Pew Research Center 2014; Stark and Smith 2012; Valenzuela et al. 2008). These studies are so rare that an article published as late as 2012 in this journal claimed to offer the first reliable and current statistics on the percentages who are Protestant and Catholic in each of 18 Latin American nations (Stark and Smith 2012, 36). Only Gill (1998, 1999) considers change rates in religious identities, but for the 1970s and 1980s, leaving open the question about what happened in the 1990s and 2000s. Also, Gill does not differentiate among Catholics, Evangelicals, and the irreligious. Fortunately, there are many excellent qualitative, historical, and comparative works on Latin American religious change that deal with the explosion of Pentecostalism (Stoll 1990; Martin 1990), secularization (Parker 1996), Catholic renewal (Cleary 2011), and the challenges to the Catholic Church (Hagopian 2009; De la Torre and Martín 2016 for a recent review). These studies advanced enormously our knowledge of religious change in the region, and we use them for developing the working hypotheses in this article. However, they do not provide nor is it their purpose an updated and comprehensive view on religious change on the basis of comparable quantitative information. Here we start addressing this gap. We provide broad empirical generalizations across time and countries, instead of thick descriptions (Geertz 1973) of specific places and times. Specifically, we address three questions. First, how did religious identities vary in Latin America during approximately the last two decades? By religious identity we refer to the way people report their religious affiliation in survey questions. We consider three main identities: Catholic, Evangelical, and irreligious.

3 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 121 They constitute about 95 percent of the religious identities in the region. Past research suggests three general trends: an increase in the proportion of those declaring to be Evangelicals, an increase in the irreligious, and a decline in Catholics. But we cannot assume that these trends happen uniformly in all countries, and we cannot assume either that they happen at all in every single Latin American country. Indeed, we show that there is substantial variation within the region. In countries such as Mexico and Paraguay, there is little religious change: the proportion of Catholics remains high and unchanged, and Evangelicals and the irreligious barely increase. In other countries, like most in Central America, the number of Evangelicals explodes. And in others, like Uruguay, the irreligious are the fastest-growing group. Thus we provide a first step in unraveling the region s internal heterogeneity by resorting to more accurate and comprehensive data than past research did. Our second question is about the potential forces behind these variations. Why do Evangelical populations explode in some countries but those of the irreligious do so in others? Why do some countries show little evidence of religious change, with Catholicism the norm for large majorities? Besides change rates, what explains robust differences in religious composition among countries for the entire period studied? Our third question refers to the sources of change behind the aggregate trends we uncover. We seek to understand what is changing among Latin Americans: are individuals themselves changing, in the sense of experiencing high rates of religious conversion or apostasy? Or is it the case that the religious affiliations of Latin Americans remain relatively constant through their lifespan, but that younger cohorts with higher rates of Evangelicalism and apostasy replace older and more Catholic cohorts? To address these questions, we decompose the individual and cohort sources of aggregate change. Because past research provides little systematic evidence to answer our second question, we resort to simple analytic techniques. This is the necessary first step before using more refined methods, such as multilevel modeling or multivariate panel regressions. To answer our third question, we employ slightly more complex statistical techniques, but given the unprecedented nature of our empirical analysis, we underscore the provisional nature of our findings. For mapping religious change we use Latinobarometer survey data between 1996 and 2013 for all 17 Latin American countries (excluding the Caribbean). 1 Latinobarometer surveys are representative of the adult national populations and have been applied almost annually from 1996 to date (except only 1999 and 2012). National samples range from 600 to 1,200 cases, with a margin of error of 5 percent to 3 percent. Latinobarometer is the best data source for our purposes. Due to its massive geographic scope, it provides a complete picture of the region. And due to its temporal scope, it allows for studying religious change across a considerable span of time. It is therefore better for our purposes than other alternatives, like the World Values Survey, which is applied to a small and uneven set of countries, or the Latin American Public Opinion Project (LAPOP), which covers the whole region only since More generally, for estimating the distribution of religious identities, national survey

4 122 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 data are clearly a better alternative than statistics provided by religious organizations, which tend to be less reliable (Stark and Smith 2012; Hagopian 2009, 11). DESCRIBING THE LATIN AMERICAN RELIGIOUS LANDSCAPE The Latin American religious landscape is structured by two main forces: religious competition (expressed in an increase of Evangelicals) and secularization. In order to show this point, we first review the empirical trends in religious affiliation since the mid-1990s. Figure 1 shows the evolution of the proportion of Catholics, Evangelicals, and people without religion by country between 1996 and The figure shows locally weighted regression (LOESS) lines in order to smooth the trends. It confirms the three processes suggested by the literature: while Catholics tend to decline, Evangelicals and the irreligious tend to grow. The figure, however, suggests interesting cross-national variations. For instance, Catholics decline drastically in Honduras but remain stable in Mexico. Evangelicals grow steeply in Nicaragua but barely in Ecuador. The irreligious explode in Uruguay but not in Venezuela. 2 To grasp these trends better, we ran OLS regression models by country, in which the dependent variable was the proportion of each religious identity in the total population and the independent variable was the year of the survey. We then plotted the beta coefficients and their respective confidence intervals for each group and country (see figure 2). Positive coefficients indicate that the group grew across time; negative coefficients indicate the opposite; coefficients close to zero indicate little change. We did not add a quadratic term for year because figure 1 suggests essentially linear trends in most cases. Figure 2 confirms the important cross-national variations suggested by figure 1. Consider Catholics. In an average year during this period, the percentage of Catholics in Nicaragua decreases 1.9 points a highly statistically significant drop. In Costa Rica, Honduras, and Uruguay, a slightly lower but still more than one-percentage point drop per year takes place. Of the 17 countries in the analysis, the annual decline in Catholic identities is negative and significant (with 95 percent confidence level) in 14. On the other hand, the proportion of Catholics does not significantly change across Bolivia, Paraguay, and Mexico. The proportion of Evangelicals increases dramatically in Nicaragua (1.7 percentage points per year, on average), Honduras (1.4), and Guatemala (1.1), and to a lesser extent but also with statistical significance in 9 other countries. However, the proportion of Evangelicals does not change (either substantially or statistically) in Argentina, Colombia, Mexico, Paraguay, and Uruguay. And the increase is moderate, and marginal statistically, in Chile, Ecuador, and Venezuela. Latin America may still be turning Protestant (Stoll 1990), but not uniformly across countries. Furthermore, the change in the numbers of irreligious people also varies dramatically. Compare Uruguay, where they grow yearly by almost 2 percentage points (a very significant change both substantially and statistically), to Paraguay, Mexico, and

5 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 123 Figure 1. Evolution of Religious Identities in Latin America, Note: The line shows data smoothed by locally estimated (LOESS) regression with a span set equal to Religious Group

6 124 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 Figure 2. Coefficient of Survey Year on Proportion of Each Religious Group, by Country, Ecuador, where they remain essentially stable. Between these two poles we have 13 countries in which the irreligious grow, often with statistical significance but at more modest rates. It is interesting that Evangelicals grow at faster rates than the irreligious in most countries, especially in the Andean and Central American countries. We are interested not only in within-country change rates across time but also in structural differences between countries across the whole period. To this end, figure 3 shows the religious composition of Latin American countries averaging all years between 1996 and Four types of countries emerge: solidly Catholic countries like Mexico, Argentina, Peru, Venezuela, Colombia, Ecuador, and Paraguay, with 80 percent or more of the population identifying as such; countries in which Evangelicals represent a sizable minority, over 20 percent (El Salvador, Guatemala, Honduras, and Nicaragua; that is, all Central American countries except Costa Rica); countries with a relatively large percentage of the irreligious, essentially Uruguay (32 percent) and Chile (16 percent) and, if removed from the previous group, El Salvador (15 percent); and countries without any distinctive pattern, which are mostly Catholic but also have some Evangelicals and irreligious (Brazil, Costa Rica, Panama, and Bolivia). In sum, any broad generalization about the Latin American religious landscape over the last two decades would be misleading. Regarding change rates, there are notable differences between, say, the secularizing Uruguay, the increasingly Evangelical Nicaragua, and the solidly Catholic Mexico. In terms of religious composition, while Catholics prevail in all countries, there are important variations across them. Furthermore, some countries show particular patterns that merit detailed attention: Uruguay and all Central American countries except Costa Rica.

7 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 125 Figure 3. Proportion of Religious Groups by Country (average, ) MAKING SENSE OF CROSS-NATIONAL RELIGIOUS VARIATIONS How can we make sense of these variations? Given the dearth of cross-national quantitative studies in the region, we resort to the secularization, anomie, and religious economies theories. They help us derive empirical hypotheses about both the correlates of change rates and the correlates of religious composition. Secularization Theory The secularization theory s main assertion is that as societies grow modern, capitalistic, and eventually wealthier, religion will weaken. This may be reflected across several dimensions from a decline in people s demand for religion or their religious beliefs and identities to a diminishing influence of religious institutions on politics. Several mechanisms explain the links between modernization and religious decline. These include increases in education and material prosperity, scientific and technical rationalization, skepticism toward established traditions (including religious ones), the pluralization of cultural worldviews and lifestyles, and functional differentiation (Berger 2011; Berger et al. 2008; Swatos and Christiano 1999; Parker 2009b for Latin America). In terms of our research questions, secularization theory predicts that as Latin American countries modernize and as the level of education, urbanization, equality, and material well-being of their people increases, new cultural options should flourish. By allowing increasing segments to question traditional religion (Catholicism in

8 126 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 our case), modernization should create a fertile ground for the expansion of both alternative, nontraditional religious identities (e.g., Evangelical religion) and secular, irreligious identities. Empirically, and regarding religious change, socioeconomic modernization should be associated with a more drastic decline in the proportion of Catholics (hypothesis 1, or H1) as well as a more pronounced increase in the proportion of Evangelicals (H2) and the irreligious (H3). Regarding average differences in religious composition for the whole period, secularization theory suggests that socioeconomic modernization should be associated with a lower proportion of Catholics and a higher proportion of Evangelicals and the irreligious. Norris and Inglehart (2011) propose a theory of existential security and religious change that can be understood as a specification within the secularization framework. They assert that people living in societies with high levels of material and existential insecurity often face collective hazards, such as famines, interpersonal violence, state repression, epidemics, and natural disasters. To compensate for these insecurities, people turn to religion in particular, traditional religion. Yet as societies develop and provide more security, demand for traditional religion decreases. People become more open to both new religious experiences (Evangelical religion in our case) and irreligious identities. The empirical implications of this theory are consistent with H1, H2, and H3. The main difference is that while secularization theory emphasizes cognitive mechanisms (e.g., pluralization of cultural worldviews), Norris and Inglehart emphasize emotional or experiential mechanisms. Anomie Theory Anomie theory might also be subsumed within the framework of secularization theory, but we treat it here as a different theory, since it leads to opposite empirical implications for our research questions. Anomie theory focuses on the social dislocations brought by modernization. Modernization means rapid socioeconomic change, migration, and urbanization. These conditions create inequalities and deprivation, cut people off from their kin and communal ties, and generate alienation and rootlessness. Since old religious identities cannot provide the tools for thriving in this new context, people are more prone to adopt new religious identities (Evangelical religion in our case). Thus whereas, according to existential security theory, material and existential stress leads to a reinforcement of traditional religion, in anomie theory it leads to religious innovation (Willems 1967; Lalive d Epinay 1969; Gill 1999, for a summary). For our purposes, this theory suggests that the poorest, the most unequal, and the least developed Latin American societies are the worst prepared to attenuate the dislocations promoted by economic change and therefore are the most traumatized by it. Regarding religious change, it is in such societies that we should see the starkest decline of Catholics (H4) and the clearest surge of Evangelicals (H5). Regarding religious composition, socioeconomic deprivation should be associated with more Evangelicals and fewer Catholics. Anomie theory does not offer clear predictions about the evolution or levels of irreligious identities.

9 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 127 Religious Economies Theory The religious economies theory emerged as a response to the failures of the secularization theory to explain the case of the United States (Finke and Stark 2005; Stark and Finke 2000). In the United States, religiosity increased as capitalist development advanced during the nineteenth and twentieth centuries. Also, in comparative perspective, the United States is very religious, despite being very modern and economically advanced (but see Baker and Smith 2015 for the deep roots of American secularism). To explain this anomaly, religious economy theorists challenge a central assumption in secularization theory that the plurality of worldviews challenges the validity of the prevailing religion and weakens religiosity. Instead, they claim that as religious diversity increases, religious groups compete to capture adherents. This improves the quality of the symbolic goods offered in the religious market, moving people to consume more religion. Additionally, religious plurality and competition increase when the state lowers the political costs and legal regulations for the expression of minority religions. Thus, countries with official and monopolist religions (such as many European nations), as well as those that disproportionately favor the prevailing religion, will have lower levels of religiosity than those with more plural and less regulated religious markets. For our purposes, this theory suggests that in Latin American countries with more regulated religious markets, and in those with greater favoritism for the predominant religion (Catholicism), religious competition should be lower. Because this lowers the quality of religious products, the proportion of Catholics should decrease (H6), and the proportion of irreligious should increase (H7). Evangelicals, in a hostile institutional context for minority religions, should decrease (H8). Conversely, countries with lower religious regulation and favoritism should have more religious competition, with Evangelicals growing at a faster rate. Also, competition means better religious products. This delays both the increase of irreligious people and the decline of Catholics. Regarding religious composition, the theory suggests that more regulation and favoritism should be associated with more irreligious (less attractive products turn people away from religion) and fewer Evangelicals (due to higher costs of expression). The prediction for Catholics is unclear: regulation pushes up their numbers because it protects them from competition, but it also lowers religious consumption, which should push them down. We assume that these opposite drives will cancel each other out, and we expect no differences in this respect.

10 128 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 DATA AND METHODS To measure religious affiliation, Latinobarometer employs the following survey question: What is your religion? Respondents freely give an answer and the interviewers mark a precoded choice. The alternatives that receive most mentions are (in decreasing order) Catholic, Evangelical (which includes Pentecostal), and none. Religious identity is important in itself because it provides an indication of the religious group or collectivity to which people feel closest in terms of doctrine, affect, and social networks (Alwin et al. 2006). The obvious limitation of this measure is that religious identity tells us nothing about individuals level or type of religiosity. We cannot differentiate between, say, practicing and nominal Catholics, or atheists and people not identified with a religious group who nonetheless believe in a supreme being (Baker and Smith 2009; Davie 1990; Stark and Smith 2012). We cannot capture religious expressions such as New Age, esotericism, syncretism, and popular religion that have been or are increasingly relevant in Latin America (De la Torre and Martín 2016). Also, the Latinobarometer surveys do not provide commonly used measures of religiosity (such as attendance at religious services or the importance of religion in one s life). Furthermore, we cannot make distinctions among religious identities (e.g., charismatic and noncharismatic Catholics). To provide a preliminary test of these hypotheses, we use six religious variables, with countries as the units of observation. Three variables indicate rates of religious change within countries across time. They are the beta coefficients of the years in which the Latinobarometer survey was applied, regressed on the proportion of Catholics, Evangelicals, and the irreligious in each country (as plotted in figure 2). Survey years range from 1996 to By correlating these coefficients (now considered as variable values) with other variables, we expect to find out which national characteristics are associated with more or less change in the proportion of Catholics, Evangelicals, and the irreligious. The other three religious variables indicate religious composition: average percentage of Catholics, Evangelicals, and the irreligious across the entire period. These six variables are calculated from Latinobarometer survey data. We use other variables for tapping the secularization, anomie, and religious economies variables. Because the first two refer to structural socioeconomic and demographic features of countries, they can be tested by resorting to a common set of measures (keeping in mind that the expected signs differ according to the theory). These include the Human Development Index (HDI, a composite measure of education, per capita gross domestic product, and life expectancy, taken from the United Nations Development Program), the Gini coefficient of income inequality, the poverty rate (measured as the percentage of people living on US$2 or less daily, adjusted by purchasing power parity), and the urbanization rate. 3 The last three come from the World Bank s World Development Indicators. 4 Since the socioeconomic and demographic conditions of different time periods may have a differential impact on the religious landscape of , we consider them at several time points across the 1980s, 1990s and 2000s (see table 1).

11 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 129 Table 1. Descriptive Statistics of Variables Used in Correlational Analysis Standard Variable Mean Deviation Minimum Maximum Catholics (slope ) Evangelicals (slope ) Irreligious (slope ) Catholics (average prop ) Evangelicals (average prop ) Irreligious (average prop ) HDI HDI HDI Gini average Gini average Gini average Poverty average Poverty average Poverty average Urbanization average Urbanization average Urbanization average Government regulation Social regulation Government favoritism Gill s regulation index Observations 17 for all variables For testing the religious economies theory, we use four measures from different sources that tap important aspects of religious markets. Three of these variables come from the dataset developed by Grim and Finke (2006). The index of governmental regulation measures the actions of the state that deny religious freedoms through any laws, policies, or administrative actions that impinge on the practice, profession, or selection of religion (Grim and Finke 2006, 13). The index of social regulation of religion refers to the restrictions placed on the practice, profession, or selection of religion by other religious groups or associations or the culture at large (Grim and Finke 2006, 19). The index of government favoritism of religion refers to the actions of the state that provide one religion or a small group of religions [Catholicism in our case] special privileges, support, or favorable sanctions (Grim and Finke 2006, 13). In all these indexes, higher values indicate more regulation or favoritism. Our fourth measure in the religious economies framework is Anthony Gill s religious regulation index (1999). Based on 21 items taken from different sources, it measures the degree to which countries regulated religious organizations during the 1970s and accounts both for outright bans on various religious activities and

12 130 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 rights (e.g., outdoor religious processions, property ownership) and preferential favoritism shown to one particular denomination (e.g., Catholicism being taught in schools exclusive to other religions) (Gill 1999, ). Higher values mean more restrictive environments for minority religions and more favorable conditions for Catholicism. Table 1 shows descriptive statistics for these variables. Regarding the analytic methods, we provide an exploratory test of the hypotheses presented in the previous section. Because quantitative research on religious identities in Latin America is scarce, we proceed with caution. Also, the small number of countries examined (N = 17) makes it difficult to carry out multivariate statistical analyses aimed at identifying causal patterns. For both reasons, at this stage we choose bivariate Pearson correlations (all our variables are continuous). Of course, we do not endorse any claim about causation with these data. COUNTRY-LEVEL CORRELATES OF RELIGIOUS AFFILIATION AND CHANGE Table 2 presents correlation coefficients between religion variables, on the one hand, and variables related to the secularization, anomie, and religious economies theories, on the other. We focus on the sign and size of coefficients more than on their standard errors and significance levels. Because we are working with the universe of Latin American countries (rather than a sample of the universe, as standard errors assume), the meaning of standard errors is ambiguous. Also, since the number of observations is small (N = 17), even considerably large coefficients may be statistically insignificant at the.05 level. Thus, we will speak about a strong correlation when the absolute coefficient is.65 or higher; a moderate correlation when it ranges between.4 and.64; and a weak one when it is lower than.4. We will also examine whether or not signs are consistent with theoretical predictions. We focus first on Catholics. Regarding religious composition, the average percentage of Catholics in the period is moderately higher in those countries with higher HDI and lower poverty rates in the 1980s. This is consistent with anomie theory (H4) and inconsistent with the secularization and existential security theories (H1). There is also a positive and moderate correlation between Gill s regulation index and the percentage of Catholics, which may be construed as supporting the religious economy theory (if we assume that more regulation protects Catholicism) or contradicting it (if we assume that more regulation lowers religious consumption). The remaining 13 correlations (81 percent out of 16) between the percentage of Catholics and the other variables are weak, and there are no strong correlations. Moving to the change rate of Catholics across time, we find (again) no strong correlations, a majority of weak correlations (88 percent), and only two moderate correlations. According to the latter, Catholics decrease less, or not at all, in countries with higher income inequality during the 1990s. This is consistent with the secularization and existential security theories (H1): inequality creates risks that keep people attached to the traditional religion. This is also inconsistent with the

13 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 131 Table 2. Bivariate Correlations Between Religious Variables and Secularization, Anomie, and Religious Economies Variables (17 Latin American countries) Catholics Evangelicals Irreligious Average Average Average % Maintenance % Growth % Growth Secularization/anomie theories HDI ***.69** HDI ***.71** HDI **.70** Gini average * Gini average *.51* Gini average Poverty average *.25.82***.69** Poverty average *.51*.52*.59* Poverty average *.52*.36.50* Urbanization average **.60*.43.52* Urbanization average **.64**.39.49* Urbanization average **.66** Religious economies theory Government regulation Social regulation Government favoritism Gill s regulation index.56* * p <.05, ** p <.01, *** p <.001 anomie theory (H4), according to which inequality promotes dislocations and therefore religious change. In addition, Gill s regulation index is moderately and positively correlated (.47) with Catholic maintenance. This is inconsistent with the religious economy theory (H6), which predicts a negative correlation. It is with Evangelical presence and change rate that we find the strongest correlations of our analysis. Regarding the former, countries that were less developed across the period have a higher percentage of Evangelicals (all strong correlations). The same happens with countries with higher poverty rates across the whole period (moderate and strong correlations) and those that were more unequal during the 1980s (moderate correlation). This is fully consistent with H5, derived from anomie theory: poverty, inequality and underdevelopment contribute to the kind of social tensions and dislocations favorable to Evangelical prevalence. Of course, the strong empirical support for H5 means a lack of support for H2 (derived

14 132 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 from secularization and existential security theories), which predicts just the opposite. It is intriguing, and contrary to the anomie theory, that urbanization is negatively related to Evangelical presence: Evangelicals are more prevalent in more rural countries. Honduras and Guatemala illustrate this finding: with urbanization rates below 50 percent in the last three decades, almost a third of their inhabitants were Evangelical in the period more than any other Latin American country. Also, all four religious economy variables correlate weakly with the prevalence of Evangelicals. When looking at the correlates of Evangelical growth rates, the evidence again mostly supports H5 instead of H2. Between 1996 and 2013, Evangelicals increased more in countries with lower HDI (all strong correlations) and higher poverty rates (two moderate and one strong correlation). As with Evangelical prevalence, correlations with the Gini index were weak: what fosters Evangelical expansion does not seem to be inequality but poverty and underdevelopment. As with prevalence, Evangelicals, surprisingly, increase less in more urbanized countries. In addition, all the correlations between Evangelical growth and the regulation measures were weak, but the correlation with Gill s regulation index was moderate, and negative, as expected (.41). In ancillary analyses we found that Guatemala, Honduras, and Nicaragua are influential cases regarding the prevalence and growth rate of Evangelicals. However, the general patterns mentioned above remain in place when these countries are excluded from the analysis. We close the examination of table 2 with the irreligious. Regarding religious composition, the average percentage of irreligious in the period tends to be higher in countries with less inequality and poverty in the 1990s, and more urbanization in the 1980s (all are moderate correlations). This provides support for secularization theory s H3, which claims that prosperity, modernization, and equality erode religion. The remaining correlations linked to secularization theory are weak, but many of them are close to moderate, and all of them have signs consistent with H3. Ancillary analyses show that these results are strongly influenced by Uruguay a very secular, developed, urbanized, and egalitarian country in Latin America (see Armet 2014 and Da Costa 2009 for the Uruguayan exceptionalism ). 5 In addition, three out of the four religious economy variables show weak correlations with the prevalence of Catholics. The exception is Gill s regulation index, which has a moderate correlation with it. However, this is a negative correlation, indicating that countries with more regulated religious environments have a lower proportion of irreligious people. This is inconsistent with H7, which predicts the opposite. Results for the change rates of irreligious people are similar to those for their prevalence. During the period, the irreligious increased faster in more developed, egalitarian, and urbanized countries, as well as in those with lower poverty rates (mostly moderate correlations). This supports H3, derived from secularization theory. Again, Uruguay plays a major role in these results for the reasons mentioned above. 6 In addition, all the correlations between irreligious growth and

15 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 133 the religious economy variables are weak, with the exception of Gill s regulation index, which is moderate (.40) but negative therefore contrary to H7. To summarize, the secularization (and existential security) theories perform only reasonably well for predicting the prevalence and growth of the irreligious, but only if Uruguay is included otherwise correlations are weak. The religious economy theory does not do a good job, either. Correlations are almost invariably weak, and when moderate, often have a sign opposite to what was expected. All things considered, anomie theory is the theory that performs the best: results are moderately consistent with it regarding the prevalence of Catholics and strongly consistent regarding the prevalence and growth of Evangelicals. Sources of Religious Change: Individual-level Change Versus Cohort Replacement Another way of approaching religious change in Latin America is to decompose the aggregate trends into their individual- and cohort-level components (Firebaugh 1988, 1997). The aggregate trends shown in figures 1 and 2 may mask two very different processes. On the one hand, the proportion of Catholics in many countries may be declining because more and more Catholics are becoming Evangelicals or irreligious. In this case, individuals are changing there is intracohort change. On the other hand, aggregate trends may result from intercohort change (Ryder, 1965). For instance, people born during different time periods may show differences in their levels of Catholic affiliation that remain stable over time. Changes in the aggregate levels of Catholic identity would result from generational replacement: older and more Catholic cohorts pass away and are replaced by younger and less Catholic cohorts. Of course, both processes may operate simultaneously. 7 It is important to make this distinction because much of the literature about religion in Latin America assumes that the main driver of change is conversion, which, by definition, implies within-cohort change across time. That is, it is assumed that individuals go through certain experiences (e.g., a miraculous cure in a Pentecostal ceremony) that make them switch faiths. The anomie theory also seems to emphasize conversion it is when people move to new contexts that religious change ensues. Secularization theories, however, seem to privilege an explanation based on cohort replacement. For example, for Norris and Inglehart (2011), the level of existential security that people experience during their formative years shapes their religious orientation. Such experience varies across cohorts. Furthermore, recent studies have found sizable intercohort differences in religious identities and beliefs. Voas and Crockett (2005) found that in Britain, a considerable part of religious change resulted from cohort differences. To a large extent, Britain secularized across time because younger cohorts were less likely to identify with a religion than older ones were. Similarly, Hout and Fischer (2014) found that around 60 percent of the growing trend among adult Americans to resist identification with any religion could be

16 134 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 attributed to cohort differences. Also, Parker (2009a) posits the existence of cohort differences among Latin Americans; secularization and new spiritual currents are more common among younger generations, who were exposed during their formative years to globalized cultures and markets (Parker 2009a, 135). To decompose the aggregate change in religious identities we apply, to each country and religious identity, the sheaf variable procedure suggested by Hout and Fischer (2014). This procedure is somewhat intricate but tremendously flexible. First, we create a trend variable for each religious group in which we replace the year the survey was applied with the LOESS smoothed percentage of people who identify with the respective religious group that year. Second, we take the log odds of this percentage and introduce this trend variable into a binary logit model predicting whether an individual identifies with the religious group or not. By construction, the coefficient of this variable should be 1. If, by chance such as for some countries in the Latinobarometer survey samples fluctuate sizably from year to year, the LOESS percentage of respondents identifying with a religious group will differ slightly from the raw data, and the trend variable coefficient will only approximate the value of 1. Third, we add to the previous regression a series of five-year birth cohort dummy variables. The percentage decrease in the size of the trend variable of the second model compared to the first model corresponds to the degree to which the aggregate trend in religious identities can be attributed to birth cohort membership, or more technically, is our estimate of the intercohort effect. 8 The most important advantage of this procedure as opposed to the simpler approach suggested by Firebaugh (1988), in which the researcher regresses religious identity on year of survey and year of birth is that the sheaf variable approach can easily capture nonlinear trends. As shown in figures 1 and 2, the majority of the trends are relatively linear, but our analysis is flexible enough to capture any departures, such those observed for Evangelicals in Costa Rica or the irreligious in Uruguay. 9 Our empirical estimates are presented in table 3. The table shows many interesting results. Changes in religious affiliation in Latin America between 1996 and 2013 operate most commonly though individual levels of religious switching or apostasy. However, a sizable portion is attributable to intercohort variation in the trends of Catholic and irreligious identities. On average, for the entire region, more than one-third of the increase trend in the irreligious (whose numbers grew from 7 percent to 11 percent) can be attributed to cohort differences among the population. Similarly, almost one-fourth of the trend for Catholics (whose numbers decreased from 79 percent to 67 percent) can be attributed to intercohort change. By contrast, only 9 percent of the trend for Evangelicals (who increased from 8 percent to 17 percent) can be attributed to cohort differences. Therefore, whereas individual change is the main driver of aggregate religious change, it is particularly so among Evangelicals. This means that the rise of Evangelicals in Latin America has been mostly through individual-level processes of religious switching or conversion. By contrast, among Catholics and the irreligious, cohort differences amount to a sizable part of the explanation.

17 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 135 Table 3. Decomposition of Change in Religious Identities by Country, Catholics Evangelicals Irreligious TC TSU TSA PR TC TSU TSA PR TC TSU TSA PR Argentina Bolivia Brazil Colombia Costa Rica Chile Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela Latin America TC: Total Change (Percent Religious Group in 2013 Percent Religious Group in 1996) TSU: Trend Coefficient (Sheaf) Unadjusted TSA: Trend Coefficient (Sheaf) Cohort Adjusted PR: Percent Change in Trend Coefficient (sheaf)

18 136 LATIN AMERICAN POLITICS AND SOCIETY 59: 1 These differences in conversion rates are consistent with a recent study of religion in Latin America (Pew Research Center 2014). Using respondents selfreported religious switching, it finds that the country average percentage of current Evangelicals who mention having been raised in another religion or in no religion at all is 9.8 percent. This is higher than the figures for the irreligious (only 6.1 percent of whom claim to have been raised religiously) and Catholics (only 1.4 percent of whom were raised in another identity). The estimates within countries show considerable diversity for Catholics and the irreligious. Among Catholics, cohort replacement captures more than 30 percent of the trend in Argentina, Brazil, and Uruguay, while it captures less than 10 percent in El Salvador, Mexico, Nicaragua, Paraguay, and Peru. In general, there is a negative but relatively weak correlation between the total amount of change in the percentage of Catholics between 1996 and 2013 and the proportion of the trend attributable to cohort replacement in each country (r =.27). Regarding the trends among the irreligious, we find several countries in which cohort replacement captures more than 30 percent of the trend: Argentina, Colombia, Chile, Honduras, Nicaragua, and Peru. On the other hand, there is not a single country in which less than 10 percent of the trend can be attributed to cohort replacement. In this case, there is a positive and more sizable correlation between the total amount of change in the irreligious and the proportion of the trend attributable to intercohort change (r =.48). There is also a considerable correlation (r =.51) between the proportion of change attributable to intercohort change for Catholics and the irreligious, suggesting that the intercohort drop in Catholic affiliation has been matched by an intergenerational rise of the irreligious. Furthermore, the trends among Evangelicals result mostly from individual conversion, with perhaps one exception represented by Costa Rica, where about 15 percent of the trend for Evangelicals can be attributed to intercohort differences. On the opposite side, we find several countries where even less than 5 percent of the trend is attributable to cohort differences: Bolivia, Brazil, Colombia, Chile, Ecuador, Guatemala, Mexico, Nicaragua, Paraguay, Peru, and Venezuela. There is no association between the total amount of change in Evangelicals and the proportion of the trend attributable to intercohort change (r =.04). CONCLUSIONS Our analysis of religious change in 17 Latin American countries across 2 decades yields some interesting findings. First, while the proportion of Catholics is declining, on average, this is not a homogenous or regionwide phenomenon. In 4 countries, the proportion of Catholics has not decreased significantly across time, or it has done so only marginally (Ecuador, Bolivia, and especially Paraguay and Mexico). From 1995 on, at least, Catholics declined only moderately in the most populous Latin American country (Brazil), and they did not decline in the second most populous one (Mexico). At the same time, some countries have experienced dramatic increases in the proportion of Evangelicals, particularly in Central Amer-

19 SOMMA, BARGSTED, VALENZUELA: RELIGIOUS CHANGE 137 ica, while in others (e.g., Chile and Uruguay), religious apathy is clearly on the rise at least in terms of religious identities. Although the longstanding Catholic monopoly is being challenged by religious competition and apostasy, in only a few countries do both processes occur simultaneously. Indeed, there are only two countries in which the proportion of both Evangelicals and the irreligious increased more than five percentage points (Chile and Nicaragua). Instead, the more common pattern is one in which the Catholic monopoly is being challenged by one specific force. In Honduras or Guatemala, the challenge comes from growing levels of Evangelicals. In Uruguay and Colombia, it comes from rising rates of the irreligious. These two processes are distinctive and should be considered on their own terms in future research. Our correlational analysis supports some important predictions derived from the anomie theory. Our clearest finding is that Evangelicals both grow faster and constitute a higher proportion of the population in countries less developed and with higher poverty rates. Yet contrary to this theory, Evangelicals thrive in less urbanized countries. A proper answer to this puzzle goes beyond this article. In addition, Evangelicals do not boom, as the religious economy theory would expect, in less regulated religious markets. Also consistent with anomie theory, Catholics remain a larger share of the population in countries with higher levels of human development and those that had less poverty during the 1980s. Another noteworthy finding is that the irreligious grow more in more developed, urbanized, and egalitarian countries, as well as in those with lower poverty rates. This is consistent with the secularization and existential security theories, which suggests that better living conditions erode the motivations for religious attachment. Yet these results should be taken with care, since they are disproportionately influenced by Uruguay a case that merits attention in its own right. In sum, none of the three theoretical approaches considered here can explain by itself the diversity of trends observed in the region. Although these theories are useful for suggesting hypotheses for our research questions, it is important to keep in mind that they were not designed to explain variations in religious identities; instead, they focus on variations in the private or public relevance of religion. Moreover, some of these theories were not built with the Latin American experience in mind. The secularization theory attempted to understand religious decline in Europe, and the religious economies theory focused on the North American experience (but see Gill 1999 and Chesnut 2003 for impressive applications to Latin America). Along with Burdick (2010, 176), we raise the question of the extent to which we need a truly Latin American theory for explaining religious change in the region. Our cohort analysis also suggests important variations in the sources of change of the aggregate trends. While the changes in the three religious groups we consider result mostly from individual-level processes of religious apostasy or switching, the role of cohort replacement among Catholics and the irreligious is certainly not negligible. This pattern seems a bit odd to us. When it comes to different forms of social identity, one would expect that religious identity, if any, would change very little