Cross-National Trends in Religious Service Attendance

Abstract

The nature of religious change and the future of religion have been central questions of social science since its inception. But empirical research on this question has been quite American-centric, encouraged by the conventional wisdom that the United States is an outlier of religiosity in the developed world, and, more pragmatically, by the availability of survey data. The dramatic growth in the number and reach of cross-national surveys over the past two decades has offered a corrective. These data have allowed research on religious trends in the United States, Canada, and Europe, putting American trends into comparative relief. This research synthesis reviews the past quarter century of cross-national comparative survey research on religious behavior, focusing on religious service attendance as a commonly measured behavior that is arguably more equivalent across societies and cultures than other measures of religiosity. The lack of evidence for religious revival is highlighted, noting instead declining rates of attendance in the United States and Canada, and either declining rates or low “bottomed-out” stability in Western Europe, most of Eastern Europe, and Australia and New Zealand. Finally, countries in Latin America, Africa, and Asia are discussed to the extent that research allows, before a call for future research—in these places in particular—is made in order to correct for the Western and Christian focus of much of the research on cross-national religious trends.

Introduction

Religious change is one of the central focuses of social science. From Comte and Durkheim to Marx and Weber, the founding fathers of social science were concerned with the future of religion. While they hypothesized varying causes and processes, they shared a prognosis: a future without religion, or at least without religion in its current form. Of course, social scientists working today have tools unavailable to the founders of social science. Among these tools, survey methods in particular offer generalizable information against which the theories of religious change, positive or negative, or stability can be tested. Survey data have been widely used to investigate religious change, through analysis of repeated cross-sectional (e.g., Chaves 1989; Firebaugh and Harley 1991; Schwadel 2011; Voas and Chaves 2016) and panel data (e.g., Wilson and Sherkat 1994; Stolzenberg, Blair-Loy, and Waite 1995; Argue, Johnson, and White 1999). These research programs have been focused on change among individuals (accounting for differences across the life course) and within a population (caused by change across birth cohorts or by immigration and emigration). The population upon which much, if not most, of this research focuses is the United States.

This American-centric focus is encouraged by the paradigm of American religious exceptionalism: the conventional wisdom that the United States is an outlier of religiosity in the developed world (Demerath 1998). In contrast to Western Europe, some social scientists cite evidence for the continuity or revival of religion and religiosity in the United States and many other countries across the globe, from post-communist Eastern Europe to developing countries in South America, Africa, and Asia. This perspective twists the old theme of American exceptionalism, characterized by the persistence of religiosity in the United States as an exception to the secularization rule (Lipset 1991, 1996), to one characterizing Europe as the secular exception in a still-religious world (Davie 1999). With such a reorientation, these scholars are less convinced than their 19th-century forbears that the path of modernization will lead to a future without religion (Berger 1999; Stark 1999; Berger, Davie, and Fokas 2008). Arguably, neither of these exceptionalism perspectives is particularly satisfying. To paraphrase Demerath (1998), instead of asserting the exceptionalism of one or another country, why not instead simply investigate their similarities and differences and use that information to extend our understanding?

Thus, this paper reviews recent research on trends in religious behavior cross-nationally, looking at the evidence for growth, stability, or decline. A comprehensive review includes trends from rich, highly developed democracies (North America, Western Europe, Australia, and New Zealand), diverse countries of post-communist Eastern Europe, and developing countries from South America, Asia, and Africa. Such a landscape view provides insight into a central question of social science—the future of religion—and evidence to the ongoing debate over secularization by reviewing trends in countries understood to be highly secularized, highly religious, and those somewhere in between. Thus, the main questions addressed are: What is the state of religious behavior? Are trends in church attendance waxing, waning, or stable? And if there is evidence for a change, where is the change happening and in which direction? In answering these questions, trends in survey estimates from recent research are reviewed for evidence of secularization, sacralization, or stability.

CROSS-NATIONAL SURVEY DATA AND MEASUREMENT EQUIVALENCE

In addition to theoretical reasons, the American focus in the religious-trends literature is grounded in a more pragmatic concern: the availability of survey data. The ability of survey data to make comparisons across national boundaries is a relatively new development in the century-old history of survey research (Converse 1987; House et al. 2004; Igo 2007). The past two decades have seen dramatic growth in the number of cross-national surveys (Andersen, Gundelach, and Lüchau 2008). Many of these studies have their roots in the European project, including the Eurobarometer. Sponsored by the European Commission, these studies began in 1973 in Western Europe but are now conducted throughout the European Union (EU), with supplemental surveys in EU candidate countries. The standard Eurobarometer is conducted biannually in the 28 EU member states using personal interviews to measure core topics and some of current interest (Nissen 2014). More recently, the European Social Survey (ESS) began in 2001 as an academically oriented omnibus survey measuring attitudes, beliefs, and behavior in countries across the continent. ESS is conducted biennially using personal interviews (European Social Survey 2015).

As these brief summaries suggest, many of the cross-national survey programs are regional in the orientation rather than global. Other examples include additional free-standing “barometer” studies in Asia (Asiabarometer), the Middle East (Arabarometer), and Latin America (Latinobarometer). But cross-national research in Europe has played a key role in the development of more global survey programs. Originating with the European Values Survey (EVS), which began in 1981, the World Values Survey (WVS) has extended beyond Europe to over 100 societies on every occupied continent (Kistler, Thöni, and Welzel 2015). The project uses a variety of data-collection methods, including personal and phone interviews and web surveys. WVS measures a variety of attitudes and behaviors, including core scales to measure trends in modernization (Inglehart and Baker 2000; Inglehart and Welzel 2005; Norris and Inglehart 2011). The International Social Survey Program (ISSP) is a collaboration among social science surveys to jointly develop and administer questionnaire modules on topics of current interest. From its beginning in the early 1980s as a joint project of four ongoing surveys from West Germany, the UK, the United States, and Australia, ISSP has grown to include over 50 countries. Data-collection mode varies by member survey and year (International Social Survey Program n.d.). Other cross-national surveys have been conducted by nonprofit and commercial organizations. The Pew Global Attitudes surveys, ongoing since 2002, have conducted surveys in 60 countries (Pew Research Center 2015), and the Gallup World Poll, ongoing since 2005, covers over 95 percent of the world’s adult population from 160 countries (Gallup 2015).

Religious service attendance has been a primary focus of social scientists using these and other surveys to study religious change, a choice justified by two main considerations. First, in addition to religious affiliation, religious service attendance is the most commonly included measure of religiosity on these cross-national surveys. Second, attendance can be argued to be equivalent across countries and cultures in a way that religious beliefs are not. Until recently, relatively few cross-national survey programs have allowed comparisons using identical questionnaires, time frames, data-collection methods, and other survey dimensions (Andersen, Gündelach, and Lüchau 2008). Even with these equivalences, the sources of potential inequivalence may still be many. First, the application of standardized survey interviewing may differ across survey research organizations, generating “house effects” that lead to variation in survey statistics (Viterna and Maynard 2002). Second, even a technically perfect translation may acquire subtle variations from the intended meaning (Johnson 1998; Harkness 2003; Harkness, Van de Vijver and Johnson 2003). Moreover, variation in language use across nations and subgroups can yield differences in the meanings.

Cross-national comparisons of behaviors rather than attitudes or beliefs are more easily justifiable given the scalar equivalence of units of time and conceptual equivalence of religiosity between traditionally Christian societies. ¹ While services may reflect different denominational traditions—varying by degree of liturgy, emphasis on emotionality, leadership by clergy or laity, and length of service—they have a common conceptual core that permits the assumption of equivalence.

Analysis of Trends

UNITED STATES AND CANADA

Much, if not most, of the cross-national research on church attendance focuses on the debate over the United States as an “exceptional” case. However, Chaves’s recent summary of religious trends notes that Americans’ attendance ² since 1990 “unambiguously has not increased…attendance has gone down or essentially remained stable.” (2011, 47). Trend analyses of the American General Social Survey (AGSS; part of ISSP) demonstrate a modest but significant 0.25 percentage point reduction in attendance each year over the past three decades. American National Election Studies (ANES) data display an L-shaped pattern, with a period of unambiguous decline in attendance rates starting in the 1960s (Chaves 2011) before reaching stability more recently (Presser and Chaves 2007). Over the complete time frame, from the mid-1960s until 2008, the self-reported rate of attendance from ANES has declined from about 43 to about 38 percent. Brenner (2011) found a similar trend using ANES, AGSS, and WVS data. To paraphrase Chaves, whether these trends should be characterized as decline or stability can be reasonably argued, but there is no evidence for increasing rates of attendance.

Cross-national comparisons often start with the case thought to be most similar to the United States: Canada. Beyer’s (1997) analysis of Gallup data found an overall reduction in weekly attendance of over 50 percent across five decades, from 67 percent in 1946 to 30 percent in 1996. Bibby similarly found a reduction in weekly attendance from 53 percent in 1957 (Gallup) to 19 and 21 percent (CGSS and Project Canada Surveys [PCS], respectively) in 2000 (Bibby 2004, 2011). Using 1985 to 2004 Canadian General Social Survey (CGSS) data, Clark and Schellenberg (2006) found that the percentage of weekly attenders declined from 41 to 31 percent and non-attenders increased from 31 to 43 percent. Eagle (2011) also used CGSS and PCS, as well as the Canadian Survey of Giving, Volunteering, and Participating, in his comparison of attendance rates. He found reductions in weekly attendance rates from 30 percent in 1975 (PCS) to 18 percent (CGSS) in 2008, and in the percentage of respondents reporting only monthly attendance in CGSS, from about 17 percent in 1986 to 10 percent in 2008.

While each of these studies focuses primarily on Canada, others put Canadian attendance in a cross-cultural context. Reimer (1995) compared Canadian and American attendance using the 1983 WVS, 1988 AGSS, and 1990 PCS data sets. He found that weekly Canadian attendance rates are nearly half those in the United States. Wilkins-Laflamme (2014) compared monthly attendance rates in US, UK, and Canadian regions using the AGSS, CGSS, British/Scottish/Northern Ireland Social Attitudes Surveys, and Northern Ireland Life and Times Surveys from the mid-1980s through the early 2000s. She found reductions in attendance in most, from dramatic (Quebec and the Atlantic Provinces) to moderate declines (much of Canada, the United States, Scotland), and low stability in others (England and British Columbia). Similarly, Aarts et al. (2008) compared attendance across the United States, Canada, and 11 European countries, using 1981–2000 WVS data. They found reductions in attendance across most of the countries included, albeit slight in Canada, compared to relative stability in the United States. These findings support the conventional wisdom that Canadian and American societies are religiously different, as Canadians more closely resemble their European progenitors than their American siblings (Marger 2013).

In summary, attendance, while likely in decline, is still higher in the United States than in Canada. However, this discussion introduces an important comparator—Europe—as the lower bound. Arguably, only in a cross-continental context can we make sense of North American trends. It matters if Europe is a single, undifferentiated mass of secularity, or if religious heterogeneity highlights important differences between countries. Moreover, the extent and direction of religious change in Europe will help make sense of the exceptionality, or unexceptionality, of North American patterns.

WESTERN EUROPE

Americans report more frequent church attendance than do Europeans, on average, and this cross-continental difference is larger than the transcontinental difference between predominantly Protestant Northern and Catholic Southern Europe (Andersen, Gündelach, and Lüchau 2008). Using 1999–2001 WVS data, Andersen et al. found 36 and 27 percent monthly or more frequent attendance in Southern and Northern Europe, respectively, compared to 60 percent in the United States. This categorization of Western European countries (Catholic/Protestant, North/South) presents the major determinants of difference in attendance rates, as Southern and Catholic countries, like Italy, Spain, Ireland, and Portugal, tend to have relatively high(er) attendance (Pollack 2003). Yet, these categories may hide important differences.

In none of the traditionally Protestant (Canada, ³ Denmark, Iceland, Sweden) nor the traditionally mixed (Germany, the Netherlands, the United Kingdom, and the United States) countries did Aarts et al. (2008) find increased attendance. Declines in attendance ranged up to 8 percentage points, although in Denmark and Iceland, where attendance rates are so low as to have bottomed out, no further trend was detected. Examining WVS and ESS data from 10 Western European countries (Spain, Belgium, Ireland, France, the Netherlands, Britain, Denmark, Norway, Iceland, and Sweden), Kaufmann, Goujon, and Skirbekk (2012) found that attendance has declined over the past four decades. The lowest weekly attendance rates, primarily in Protestant Northern Europe, are at or approaching a floor of about 5 percent.

While Catholic countries in Europe typically demonstrate higher attendance rates than Protestant countries, no consistent evidence emerges even in these countries for upward trends in attendance. Using WVS data, Aarts et al. (2008) compared attendance trends from 11 Western European countries to those from the United States and Canada. In four of the five traditionally Catholic societies (Belgium, France, Ireland, and Spain), weekly attendance declined between 4 and 14 percentage points from 1981 to 2000. In Belgium, the decline in attendance started in the late 1960s, and todayBelgian attendance resembles the low attendance rates of neighboring countries like France, Germany, and the Netherlands.

The decline in attendance in Ireland is more recent. Where Greeley (2003) found high, stable rates of attendance, Conway (2013) and others discovered a precipitous decline in Irish religious practice beginning in the 1980s. Analyzing Eurobarometer and ISSP data, Nic Ghiolla Phádraig (2009) found a decline from near-universal weekly attendance (92 percent) in 1973 to 40 percent in 2008. Adding ESS and WVS to these trends, Brenner found a reduction from about 90 percent in the 1980s to about 45 percent in 2006 (2011). A similar decline in religious behavior emerged in Spain, as regular Mass attendance plummeted from about 60 percent in 1975 to 28 percent in 2002 (Requena 2005; see also Brenner [2011]).

But has attendance increased anywhere in Western Europe? Some evidence has emerged for a revival of religious practice in Italy. Aarts et al. (2008) found an increase in attendance only in Italy—by about 6 percentage points. Introvigne and Stark (2005) found an increase in weekly attendance at Mass from 32 percent in 1981 to 40 percent in 1999, primarily using WVS data. In their analysis of more than 30 European countries in five categories—Northern, Western, Central, Eastern, and Southern Europe—using EVS data, Voas and Doebler (2011) also discovered evidence for a small uptick in religious attendance in Italy and other Southern European countries (Malta, Portugal, and Spain) in addition to an unlikely location: the Nordic countries (Denmark, Finland, Iceland, and Sweden). Using the past two decades of EVS and ESS data, Burkimsher (2013) also discovered an uptick in Denmark, the only such increase in the 14 European countries analyzed.

Neither increase, however, appears to be a substantive change in the pattern of religious decline. Voas and Doebler (2011) argued that the small uptick they discovered is likely a combination of a sampling artifact and a period effect, noting that the effect was temporary, and disappeared in the late 2000s. This conjecture is confirmed in a series of additional studies. Analyzing EVS, ESS, ISSP, and Eurobarometer data, Brenner (2011) also found an increase in Italian attendance, but characterized it as a short-lived uptick in the middle of an S-shaped function, a finding replicated by Vezzoni and Biolcani (2015). By the turn of the millennium, attendance rates had resumed their decline. Andersen, Gündelach, and Lüchau’s (2013) analyses of panel data also confirmed that no religious or spiritual revitalization has occurred or is occurring in Denmark.

In sum, there is no consistent evidence of increasing rates of attendance in Western Europe or North America. Quite the opposite—attendance is either declining or has bottomed out in the single digits. However, Western Europe is not where growth has been hypothesized to occur by critics of secularization. Rather, some expected that the removal of impediments to the free exercise of religion with the fall of communism would spur religious revival across Eastern Europe. But did this revival actually occur?

EASTERN EUROPE

The fall of communism in Eastern Europe created the possibility of increasing rates of religious behavior. In an examination of 1991 ISSP data, Gaultier (1997) argued for a religious resurgence in Eastern Europe, given the opening of the “religious marketplace” after decades of artificial constraint. This supply-side paradigm viewed the decline of communist regimes in Eastern Europe as an opportunity for the revival of previously regulated or banned religious practice. Gaultier’s limited evidence suggested that the resurgence was underway in the early 1990s, characterizing Poland’s increasing Mass attendance as a paradigmatic example. But in his analysis of ISSP data from the 1990s, Greeley found evidence to the contrary, uncovering relative stability for most Eastern European countries (Greeley 1994) and even declining attendance in Poland (Greeley 2003).

Counter to the trend in Poland, Greeley (2003) also found evidence of increasing attendance in irreligious Russia and East Germany. Expanding on these analyses, Reitsma et al.’s (2012) investigation of 42 countries in EVS and ESS from 1981 to 2007 found some evidence for increasing attendance in Eastern Europe. However, their decision to pool countries together rather than analyzing them separately adds a strong caveat to their findings. This decision washes out many of their unique histories and religious idiosyncrasies, grouping together countries as varied as religious, populous Poland and secular, sparsely populated Estonia.

Other research, treating countries individually rather than in large categories, found low and declining attendance rates across Eastern Europe. Pollack’s (2003) findings, based on an analysis of “Political Culture in Central and Eastern Europe” survey data, suggest that Eastern European attendance has either remained low or declined. He found similar patterns in many Eastern and Western European countries, noting that the Czech Republic, East Germany, Estonia, and Russia are highly secularized similar to their counterparts in Western Europe and demonstrate very low and/or declining rates of attendance (2003).

In subsequent analyses of “Church and Religion in an Enlarged Europe Survey,” Gallup, WVS, and Eurobarometer data, Pollack found evidence for declining attendance in Eastern Europe linked to birth cohorts, as younger generations are less likely to attend (see also Lois [2011]). Pollack, however, also found evidence for a small, temporary reversal—a period effect—for some Eastern European countries, as some briefly returned to the Church after the fall of their communist regimes. This slight uptick in attendance was also detected in analyses of Hungarian attendance rates from WVS and ISSP by Froese (2001), who noted that this period effect reversed itself relatively quickly.

Burkimsher’s (2013) analysis of EVS and ESS data for 24 European countries found evidence for increasing rates of attendance in Russia, and to a lesser extent Romania and Bulgaria, but also decreasing rates of Mass attendance in Poland and Slovenia and low and stable attendance in other post-communist countries like Estonia and the Czech Republic. Thus, in summary, findings in Eastern Europe also fail to demonstrate any strong and consistent pattern of increased attendance. Most Eastern European countries resemble those in Western Europe—low and stable (e.g., Estonia, Czech Republic) or declining (e.g., Slovenia) attendance. Others with relatively high attendance, like Poland, have attendance rates similar to higher-attendance Western European countries (e.g., Italy, Ireland), and demonstrate similarly negative trends. Only in three countries, namely Romania, Russia, and Bulgaria, is there somewhat consistent evidence of increasing attendance. Notably, the latter two of these countries have very low attendance rates, on par with those in Northern Europe.

AUSTRALIA AND NEW ZEALAND

Australia and New Zealand could be readily grouped with Canada and the UK, unsurprisingly perhaps given their shared history and culture (Inglehart and Baker 2000). However, somewhat fewer cross-national comparisons include these Southern Hemisphere countries. Norris and Inglehart place Australian and New Zealander attendance near Protestant Europe and Canada, as highly secularized (or secularizing) and low-attending countries (2011).

Trends in Australia demonstrate long-term reductions in the attendance rate. Inglehart and Baker (2000) noted dramatic change in Australian attendance over the 1980s and 1990s. About 40 percent of Australians reported at least monthly attendance in the 1981 WVS, but this rate was reduced to 25 percent by the mid-1990s. Smith (2009) found monthly attendance in the same data set and years declining from 28 to 17 percent. Weekly attendance trends are less clear. Analyzing 1981 WVS data, Campbell and Curtis (1994) reported a weekly attendance rate of 17 percent, and Smith’s (2009) analysis of 1991 and 1998 ISSP data reported weekly attendance rates of 13 and 19 percent, respectively.

New Zealand shows patterns of low, stable attendance similar to those in Western Europe. In an analysis of 1991 and 1998 ISSP data, Vaccarino, Kavan, and Gendall (2011; see also Ward [2006]) found that about 20 percent of New Zealanders attend regularly, defined as once a month or more often. Using 2008 ISSP data, Ward (2013) found this rate stable for the following decade. Smith’s (2009) analysis from the 1991 and 1998 ISSP found weekly attendance in New Zealand at 17 and 13 percent, respectively. In sum, like the European countries that colonized them, neither Australia nor New Zealand demonstrate evidence of increasing religious behavior; rather, trends suggest reductions in already low rates of church attendance.

LATIN AMERICA, SUB-SAHARAN AFRICA, AND ASIA

Unfortunately, little can be said about trends in attendance rates outside this list of Western democracies and post-communist European countries. While cross-national survey projects have begun collecting data in Latin America, Africa, and Asia, little published work has analyzed trends in the religious behavior measured by these studies. This oversight is especially troublesome given recent research linking religiosity and religious behavior to existential security (see Inglehart and Welzel [2005]; Norris and Inglehart [2011]). This revision to modernization and secularization theories posits that as material needs are met in a society, existential security rises, the compensatory need for religion is diminished, and attendance is reduced.

Smith (2009) and Inglehart and Baker (2000) presented rates of attendance from countries with available data in WVS: seven (and a territory) from Latin America, one from Asia, and one from Africa. The South American countries in WVS demonstrate a set of mixed trends. Argentina has participated in four waves of WVS—1981–84, 1990–93, 1995–97, and 1999–2004—longer than other South American countries. From the earliest two surveys—31 and 32 percent, respectively—to the latter two—24 and 25 percent, respectively—Argentinian weekly attendance trends look somewhat similar to Catholic countries in Europe, with declining attendance over these three decades. Inglehart and Baker (2000) found a similar trend in monthly attendance for Argentina using the same WVS data.

Data are available for Brazil and Venezuela at two points, and for each country, these two points show relative stability. ⁴ In Brazil, weekly attendance increased negligibly from 33 percent in the early 1990s to 36 percent in the late 1990s (see also Inglehart and Baker [2000]), whereas in Venezuela, attendance remained stable at 31 percent over the two surveys in the late 1990s and early 2000s. Analyses for Chile, Columbia, and Peru are also limited to only two data points, similarly stable although at a higher level of attendance. Monthly attendance in Chile declined negligibly from 47 percent in the early 1990s to 44 percent in the late 1990s (Inglehart and Baker 2000). Weekly attendance rates in Columbia and Peru remained stable: 45 and 43 percent, respectively, in the late 1990s, and 47 percent for both countries in the early 2000s (Smith 2009).

Mexico and Puerto Rico show some evidence of increasing attendance rates. Mexican WVS data show increased attendance from 43 percent in the early 1990s to 46 percent in the late 1990s to 56 percent in the early 2000s (Smith 2009), although Inglehart and Baker (2000) found that monthly attendance declined from 74 percent in 1981 to 65 percent in the late 1990s. Weekly attendance in Puerto Rico increased from 52 to 57 percent from the late 1990s to the early 2000s (Smith 2009).

Two countries, one each from Africa and Asia, round out the discussion of trends from outside the West. These are the only countries included, as both are predominantly Christian and thus comparable to the countries previously discussed. WVS data from the late 1990s and early 2000s show stable attendance rates in South Africa, 54 and 52 percent, respectively (Smith 2009). Also using WVS, Inglehart and Baker found increased monthly attendance over a longer span, from 61 percent in 1981 to 70 percent in the late 1990s (2000). The highest current rates of attendance of all the countries reviewed here are from the Philippines. In the late 1990s and early 2000s, weekly attendance rates reached 84 and 80 percent, respectively (Smith 2009).

Explaining Patterns in Attendance

Religious tradition was previously introduced as a potential cause of variation in attendance rates and trends. Pollack (2003) found that predominantly Catholic countries, like Poland and Romania, demonstrate relatively high attendance, comparable to their Western European Catholic peers (e.g., Italy, Spain, Ireland, and Portugal). Using ESRC East-West Programme data, Need and Evans (2001) examined attendance in 10 post-communist countries: five predominantly Orthodox (Belarus, Bulgaria, Romania, Russia, and Ukraine ⁵ ) and five predominantly Roman Catholic (Czech Republic, Hungary, Lithuania, Poland, and Slovakia.) Their analyses found that Catholic countries tend to have higher attendance than the Orthodox countries, with Poland demonstrating far and away the highest rate of attendance. Bruce (2000) examined 17 post-communist countries using 1995–1996 WVS data. He found that predominantly Catholic countries, like Croatia, Slovenia, and Lithuania, have higher rates of attendance, by around 30 percent, than do traditionally Orthodox countries like Russia and Serbia and pluralistic countries like Latvia and Estonia. Comparing EVS data for post-communist countries, Titarenko (2008) also found evidence for higher, although declining, rates of Mass attendance in Catholic countries (Lithuania) than in predominantly Orthodox countries (Belarus, Russia, and Ukraine). She argues that this distinction is due to Catholics’ ability and desire to educate adherents’ belief systems and the inability or lack of desire in Orthodox churches to do the same.

Halman and Pettersson (2003) agree that European religiosity is related to religious tradition rather than an East/West dichotomy: Protestant and Orthodox countries are more secular, and Catholic countries remain more religious (Martin 1978, 2005). Halman and Pettersson performed a cluster analysis of church attendance and religiosity (importance of God) with 1990 EVS data from 23 countries (2003). They summarized eight post-communist countries in five clusters: most secular (Czech Republic, Bulgaria, East Germany), more secular (Latvia, Hungary), middle (Slovakia), more religious (Romania), and most religious (Poland).

In addition to denomination, modernization also emerges as a useful concept in explaining by-country differences in attendance. Using EVS data, Conway linked Slovenia’s pattern of declining Mass attendance first to the communist regime’s “enforced secularization” and later to a “spontaneous secularization” caused by modernization and its concomitant effects of urbanization and expansion of education that started before Slovenia’s integration in the EU (2013). Norris and Inglehart (2011) too argue that the predictors of religiosity based in modernization theory used in Western Europe work in the East too. Attendance is well explained (R ² = .45) using human development factors at the societal level as well as the individual level (e.g., educational attainment). Notably, they included religious factors, like religious tradition, as well as demographic factors in their explanatory model.

One such demographic factor, gender, is often linked to differentials in practiced religion, as women are more likely to attend religious services. Yet, this “gender gap” does appear to be shrinking as gender loses its causal, or at least predictive, value. Voas, McAndrew, and Storm (2013) compared about 40 countries using EVS, ESS, and ISSP data to determine whether the gender gap in attendance was shrinking, growing, or stable. There is some evidence that the gap in Europe is narrowing, likely a function of women’s increasing material security (see Norris and Inglehart [2011]).

MEASUREMENT ERROR, OVERREPORTING

But can we take these estimates at face value? In each country examined, the trends discussed all rely on survey self-reports. Whether the respondent is reporting his or her frequency of attendance to an interviewer on the telephone or face-to-face across a dining-room table, or by writing on paper or clicking on a radio button on a website, self-reported rates of religious behavior can include measurement bias (Brenner and DeLamater forthcoming). Moreover, even if questions are deemed equivalent, errors may not be. Survey error may be random in one country but systematic in another.

Most of the analysis of measurement error of religious service attendance has focused on the United States. This research program has compared survey estimates to those based on data collected by the religious organizations themselves (Chaves and Cavendish 1994; Hadaway, Marler, and Chaves 1993, 1998; Marler and Hadaway 1999) and investigator-initiated counts at churches on Sunday mornings (Hadaway, Marler, and Chaves 1993). In each of these investigations, survey reports of attendance were found to be about double the actual rate of behavior.

In the face of critiques of these methods of validating the survey report (Caplow 1998; Hout and Greeley 1998; Woodberry 1998), other studies have turned to time-diary data as a potential comparator. Unlike conventional survey questions, chronologically based data-collection procedures like time diaries eschew direct questions about specific behaviors of interest (Robinson 1985, 1999; Stinson 1999), avoiding prompting self-reflection on the part of the respondent, and arguably yielding less biased and higher-quality data (Bolger, Davis, and Rafaeli 2003; Niemi 1993; Zuzanek and Smale 1999). Presser and Stinson (1998) compared diary and survey estimates from the United States in the early 1990s. Their analysis supported the findings of Hadaway, Marler, and Chaves (1993), arguing that “misreporting error produces an increase in attendance claims of almost 50 percent” (1998, 140).

Brenner (2011) extended the work of Presser and Stinson, comparing four decades of data from the Multinational Time Use Study (MTUS) with survey data from 14 countries. Brenner found evidence of large and consistent overreports in the American survey data, varying between about a quarter and a half of the survey report, depending on the survey and year. He also found an overreport in Canada, although less consistent and somewhat smaller than that found in the United States. Italy ⁶ and Ireland also demonstrated statistically, but not substantively, significant rates of overreporting, as the effect sizes were negligible. Subsequent work using a more complex method for computing overreporting has supported the existence of an overreport in Italy (Rossi and Scappini 2012) and the United States (Rossi and Scappini 2014). Finally, little research exists testing the validity of self-reported religious behavior outside Europe and North America. However, findings from Brenner (2014) suggest that overreporting of religious behavior does occur in non-Christian societies as well.

SUMMARY OF ATTENDANCE RATES AND TRENDS

In summary, acknowledgment of measurement errors makes US attendance look more like that in Europe, on par with Italy and Spain. When placed into the spectrum of European attendance rates, with Ireland and Poland near the top and Estonia and Scandinavia near the bottom, the United States looks unexceptional (see Voas and Chaves 2016). Moreover, a review of trends in attendance rates shows little support for widespread sacralization. Most countries reviewed here fall into one of two types: low and stable attendance, including many Northern European countries as well as some from Eastern Europe; and declining attendance, including most of Europe as well as the United States, Canada, Australia, and New Zealand. A smaller set of countries’ trends, primarily from South America, could be described as stable but somewhat higher than the European average. Only three countries, Romania, Bulgaria, and Russia (plus Puerto Rico), can be described as having increasing attendance, and the latter two have low attendance rates on par with countries in Northern Europe (see table 1).

Table 1.

Summary of Religious Service Attendance Trends

Low and stable	Declining	Mixed/ Fluctuating	High and stable	Increasing
Czech Republic	Argentina	Belarus	Brazil	Bulgaria
Denmark	Australia	Mexico	Chile	Puerto Rico
Estonia	Austria	South Africa	Colombia	Romania
Finland	Belgium		Peru	Russia
Iceland	Canada		Philippines
Latvia	France		Venezuela
Norway	Germany
Slovakia	Hungary
Sweden	Ireland
United Kingdom	Italy
	Lithuania
	Netherlands
	New Zealand
	Poland
	Portugal
	Slovenia
	Spain
	Switzerland
	Ukraine
	United States

Conclusion and Future Directions

To further elucidate the future of religious service attendance, the continued extension of this research program into the developing world is desperately needed. The lowest-hanging fruit are the traditionally Christian countries of West Asia (i.e., Armenia and Georgia). While included in relatively few cross-national religious comparisons (see Bruce [2000] for an exception), these countries could be readily compared to other post-communist countries of Eastern Europe. Other convenient comparisons can be found in Central and South America. Many of these countries have recently joined one or more of the cross-national survey programs, and even more are members of a regional survey program, the Latinobarometer. The region includes highly developed countries (Chile, Argentina), rapidly developing countries (Brazil, Panama, and others), and one of the world’s poorest, least developed countries (Haiti). Other predominantly or majority Christian countries in Sub-Saharan Africa, some of which participate in the cross-national survey programs outlined here, should also be included. These include South Africa, which was included in a study discussed here, but also Ethiopia, Ghana, Zambia, Rwanda, and Zimbabwe, which were not.

A somewhat more difficult to include, but just as important, extension to this research program would include non-Christian countries. Starting with the other Abrahamic religions, Israel would be relatively easily included, as it participates in a number of cross-national survey programs, and the nature of the practice of Judaism may lend itself well to a comparative analysis with predominantly Christian countries. Equally or more important would be an extension to the Islamic world. Many, but not all, predominantly Muslim countries participate in WVS or ISSP. Those with at least a year of survey data in one of these survey programs include countries in West (i.e., Mali) and North Africa (i.e., Algeria, Egypt, Libya, Morocco, and Tunisia), and in West (Azerbaijan, Bahrain, Jordan, Kuwait, Iran, Iraq, Palestine, Qatar, Saudi Arabia, Turkey, and Yemen), Central (Kazakhstan, Kyrgyzstan, Uzbekistan), South (Bangladesh and Pakistan), and Southeast Asia (Indonesia and Malaysia). Research could readily compare religious behavior between these countries (see Fish [2011]; Hassan [2002]) and, with care, those on which this synthesis has focused. Additionally, religiously pluralist countries like Burkina Faso, Nigeria (included in Inglehart and Baker [2000]), and Tanzania could also be included in analyses, and perhaps quite productively so, being certain to account for religious affiliation both within and between countries.

Other Asian countries in the Hindu (India) and Buddhist (Thailand) traditions, as well as other religious traditions that Inglehart refers to as Confucian (Japan, South Korea, China, Taiwan), as well as Vietnam, may be quite difficult to compare with operationalizations of Western religious behavior (see note 1). However, some enterprising social scientists with deep knowledge of one or more of these cultures may find or create equivalent measures up to such a task. Certainly, the continuing development and extension of these cross-national survey programs is far too great an opportunity to allow to pass.

Religious service attendance is not inherently more important than other ways of operationalizing religiosity, like affiliation, other behaviors (e.g., scripture reading, praying, and meditating), or beliefs (Brenner 2014; Sasaki and Suzuki 1987). It is only used here given its relative ease and ubiquity of measurement and potential for validation. Other research could use alternative operationalizations to pursue a similar question of the comparison of cross-national trends. Affiliation, given the attention paid in the research literature to the growth in the unaffiliated in the United States (Hout and Fischer 2014) and elsewhere, may be a particularly fruitful research focus.

Cross-national comparisons, even those limited to traditionally Christian countries, may still encounter inequivalences. First, countries included here differ greatly in their wealth, with North American and European countries, Australia, and New Zealand being more wealthy and many African, Asian, and Latin and South American countries being less wealthy. Measures of country wealth, including gross domestic product (GDP), are one factor considered in recent research by Inglehart and his colleagues linking changes in traditional religiosity to human development factors (Inglehart and Baker 2000; Inglehart and Welzel 2005). However, additional research could be undertaken to understand the causes of differences in religiosity that are associated with variations and change in GDP and other measures of country wealth.

Second, religion has played different roles in the political environments of the countries included here, including differences in the regulation of the religious marketplace (Aarts et al. 2010) and relationships between religious and political groups (e.g., the religious right in the United States; the Catholic Church’s opposition to communism in Poland). Relatedly, some differences between countries may be linked to evangelical movements—where they have arisen and where they have not—as well as the differences between these evangelical movements and countries’ traditional denominations. These traditional denominations and evangelical movements may have different relationships with political systems and parties, and these affiliations add additional sources of inequivalence between countries (e.g., evangelicalism’s connections with rightist movements and Catholicism’s links to liberation theology and Marxism in some Latin American countries [Stoll 1990]). Perhaps because of these country-level idiosyncrasies, it is difficult to consider political-religious associations as part of a cross-national analysis like this one. Future research, whether single- or multi-country, should to the extent possible understand that these histories have purchase in understanding religious trends.

Future research may also examine an important topic not included in the current review: the effect of immigration on religiosity in both sending and receiving countries. Research by Van Tubergen (2006) and colleagues (Van Tubergen and Sindradóttir 2011) and others suggests that religious affiliation is linked to immigrants’ origin and destination; however, future research could examine whether and how this has changed over time. Moreover, many sending and receiving country pairs (e.g., Turkey and Germany; France and its former African colonies) have long unique histories and should be considered in future work.

These limitations and caveats of the present research notwithstanding, cross-national comparisons like these are arguably useful. The landscape view presented here can be an effective call for future research to dig behind these trends and better understand the nature of religious behavior in each of these countries. Comparisons of the trends between countries that take their differences and similarities into account allow us to better understand their causes. For instance, by understanding trends in three societies with a shared history, language, and culture—the United States, Canada, and the UK—as well as the similarities and differences in these countries, we can begin to understand why the United States is the most religious, the UK the least, and Canada somewhere in between. While they share many of the factors often considered when differences in religiosity are explained, they have important differences, including some of the factors mentioned in the previous paragraphs. By looking to both the similarities and the differences, we can better understand not only the religiosity in these countries taken together, but also religiosity in each country individually.

In conclusion, this landscape view further reduces American religious exceptionalism, as well as European “irreligious exceptionalism,” as descriptive tropes. To again paraphrase Demerath, the United States and European countries are no more or less exceptional than any other country discussed here, all in terms of kind rather than degree (1998). Moreover, to extend Chaves’s comment about American trends, whether the overall trend, as well as those from some individual countries, should be characterized as decline or stability can be reasonably argued, but there is no widespread or consistent trend of increasing attendance in the countries and studies reviewed here.