ABSTRACT
This paper investigates (trends and determinants of) individual combinations of religious believing and belonging in Europe from a cross-national and longitudinal perspective. Individual level data for the period 1981–2007 in 42 countries derived from the European Values Survey and the European Social Survey are harmonised and enriched with contextual characteristics, constituting a more comprehensive data base for Europe than any previous research. Complementary hypotheses from modernisation and market theories are derived systematically and tested rigorously. We find differential trends. In western societies, the popularity of secularity and consistent religiosity has persisted. In former communist societies, the popularity of consistent religiosity has increased whereas secularity has decreased. In both western and former communist societies, solitary religiosity has been a clear yet small phenomenon, somewhat increasing in western societies and somewhat decreasing in former communist societies. A crucial hypothesis derived from modernisation theories, stating that financial and social security would decrease consistent religiosity, is empirically supported.
Introduction
The role of religion has considerably changed during the past century. Especially in Europe, church membership and attendance declined substantially (Norris and Inglehart 2004). However, Davie (1990a,b, 1994, 2002) argued that religious beliefs remained relatively popular in Europe. Several investigators argued that religious beliefs declined substantially in Britain and several other countries during the last decades (De Graaf and Need 2000; Bruce 2002), although non-traditional beliefs remained relatively stable or increased (Stark and Iannaccone 1996; Gill et al.1998).
Most research focused on one, e.g., believing (Kelley and De Graaf 1997) or the other, e.g., belonging (Ruiter and Van Tubergen 2009) and presented both phenomena (only) separately. Voas and Crockett (2005) and Pollack and Pickel (2007), focussed on both phenomena, however, only in Britain and Germany, respectively. Aarts et al. (2008) studied correlations between believing and belonging in several western countries. However, correlations indicate the degree of consistency between believing and belonging, while the core argument is about lack of consistency: believing without belonging.
We set out to explore these presumed longitudinal trends in many countries of persisting religious beliefs with declining levels of religious practice which is characterised as ‘the strong version’ of believing without belonging as opposed to so-called ‘weak interpretations’ (Voas and Crocket 2005). We like to emphasise that this ‘strong version’ is not necessarily consistent with, nor do justice to the original subtleties as intended by Davie (1994; Voas and Crocket 2005).Therefore, we will extend on this ‘state of the art’, by investigating trends in combinations of believing and belonging.
To assess such longitudinal trends in many countries, we need valid measurements on believing and belonging tapped simultaneously from as many people as possible, at as many points in time as possible, in as many countries as possible. For this purpose we rely, first, on religious self-identification, which is not a direct measurement of beliefs, however, a general indicator that is likely to be highly correlated with religious beliefs and therefore often used and considered valid from Davie (1990) up to Voas (2009). Second, we rely on religious attendance, a straightforward and valid indicator of religious belonging (from Davie (1990) up to Ruiter and van Tubergen (2009)).
We will distinguish four combinations of religious self-identification and attendance. Two of these combinations can be labelled as consistent patterns: secularity – defining oneself to be non-religious and refraining from attending religious services – versus consistent religiosity – defining oneself as religious and attending religious services. The two other combinations seem to be somewhat inconsistent. The strong version of believing without belonging means that people do define themselves to be religious, however, they do not attend religious services: therefore, we will label this phenomenon as solitary religiosity. A usually neglected category is constituted by people who do not define themselves to be religious but, nonetheless, do attend religious services: this phenomenon is characterised as habitual attendance, suggesting belonging without believing.
We aim to contribute to the debate on believing and belonging by studying the trends and determinants of these four combinations of believing and belonging from a cross-national and longitudinal perspective, based on a European database more comprehensive than any previous research, containing 42 European countries, also former communist countries, over the time span from 1981 to 2007. Focussing on trends in the prevalence of these categories over time, our first (descriptive) research question reads: to what extent has the popularity of various combinations of believing and belonging changed over time in European countries?
Next, we focus on the explanations for variations in religiosity, building on two competing paradigms (Norris and Inglehart 2004): one paradigm focuses on modernisation processes that reduce the demand for religion, the other paradigm focuses on the influence of religious market characteristics on religious consumption. We will test systematically derived hypotheses from both perspectives with regard to the determinants of combinations of believing and belonging. Therefore, our second research question reads: to what extent are characteristics of religious markets and aspects of modernisation – on the individual and the societal level – related to combinations of believing and belonging?
Theories and hypotheses
Religious markets
Religiosity can be investigated from an economic perspective of supply and demand. Religious institutions provide religious goods and consumers may or may not consume them. In this perspective, the amount of consumption of religious goods is a function of the competitiveness of the religious market, while demand for religion is implicitly assumed to be stable over time (see e.g., Stark and Bainbridge 1996). Under this assumption of stability in demand for religion, we would expect the proportion of people considering themselves to be religious to be rather stable over time.
When religious suppliers compete to gain market share, they will adjust their products to the preferences of consumers. This competition will result in a diversified supply of religious goods. Hence, people looking for religious goods have more of a choice and are possibly more likely to find a good that suits their needs (Stark and Bainbridge 1987; Finke and Stark 1988; Iannacone 1990, 1991, 1995; Stark and Finke 2000).
Key to the competition is the freedom of religion which may hold to some extent in European countries. Yet, traditional catholic western European countries have had neither too much diversity nor competition; which was at a slightly higher level in traditional protestant societies; and even a bit higher in traditionally mixed societies (Aarts et al. 2008). However, the situation has been rather different in former communist countries. In times of restricted freedom of religion because of state regulation, when there is only one recognised religious institution or religion is prohibited altogether, it is almost impossible for new religious firms to enter the market in order to diversify the supply of religious goods. Vice versa, freedom of religion will have induced competition which will have, in turn, increased religious attendance (Finke 1990; Chaves and Cann 1992; Stark and Iannaccone 1994; Iannacone 1995; Stark and Finke 2000).
This history of religious freedom may be important for trends in religiosity, particularly in the European context. In former communist countries, where atheism was the official state doctrine, the development towards more religious freedom may have resulted in more diversified supply, inducing a higher consumption of religious goods (Greeley 1994, 2003; Pickel 2010). The combination of assumed stability of religious demand and increased religious freedom in former communist countries leads to the hypotheses that, over the years, the likelihood increases to belong to category of consistent religious people as compared to belong to the category of secular people, in former communist countries. This line of reasoning may even hold for the category of habitual attendants who may attend religious services for non-religious, i.e., social motives. Given that religious freedom induces more diversified supply, adjusted to the religious demands of people, we expect that, over the years, the likelihood to belong to the category of solitary religiosity decreases as compared to the category of secular people, in these former communist countries.
Modernisation
The modernisation perspective on religion argues that there are several developments that reduced the actual demand for religion. Ruiter and Van Tubergen (2009) distinguished three different developments as part of the modernisation perspective on religion: over time increases in scientific worldviews, financial security and individualisation processes.
Scientific worldviews are characterised by methodological naturalism: it is restricted to phenomena that are somehow measurable and repeatedly observable. Transcendent religious phenomena have therefore no place in scientific worldviews. Since scientific worldviews undermine the plausibility of religious teachings, they may reduce levels of both religious belief and attendance (Lenski et al.1991; Weber [1922] 1993; Need and De Graaf 1996; Stark and Iannacone 1996; Bruce 2002).
Scientific worldviews improved technological developments, such as medical treatments that reduced infant mortality. Hence, we hypothesise that in societies with more technological developments, the likelihood (to belong to all combinations) of religiosity decreases whereas secularity increases. Since scientific worldviews are part of and taught in the curricula of educational institutions, the educational level obtained by individuals provides us with an indicator at the individual level. Therefore, we hypothesise that with increases in the educational level of individuals, the likelihood (to belong to any of the combinations) of religiosity decreases as compared to belong to the category of secular people.
Norris and Inglehart (2004) argued that financial security is an important aspect of modernisation that may account for secularisation. Traditionally, religious institutions played an important role in the care for the poor. This made the poor highly dependent on religious institutions. In contemporary societies, financial security for the poor is primarily provided by the state (De Swaan 1988), although there are differences between welfare state regimes (Esping-Andersen 1990, 1999). This societal level redistribution of income makes the poor less dependent on religious charity organisations. The size of government expenditure on social security at the societal level is an indicator of financial security for individuals (Te Grotenhuis et al.1997). Therefore, we hypothesise that in societies with high spending on social security, the likelihood (to belong to any of the combinations) of religiosity decreases as compared to belong to the category of secular people. Moreover, country level financial security is indicated by purchasing power of the gross national product per capita. Therefore we expect that in societies with a higher purchasing power, the likelihood (of all combinations) of religiosity decreases. At the individual level, income is an indicator of financial security. Therefore, we hypothesise that with increases in the individual level income, the likelihood (to belong to any of the combinations) of religiosity decreases as compared to belong to the category of secular people.
A third aspect of modernisation is individualisation. Individual religiosity as a social phenomenon is reinforced by religious networks: one's parents or partner may expect and stimulate religious participation and their social pressure may be difficult to resist (Te Grotenhuis and Scheepers 2001). Religious communities may provide sanctions on deviance from religious norms (Berger 1967; Durkheim [1912] 1995). During the process of individualisation (see Beck 1986; Pollack 1996), more people started and continued to live on their own: youngsters left their parents’ home to live on their own and divorce rates increased. The proportion of individuals living alone, without partner or parents, can be regarded as an indicator of individualisation at the level of a society. We hypothesise that the level of individualisation decreases the likelihood (to belong to any of the combinations) of religiosity as compared to belonging to the category of secular people. To the extent that social networks are important for religiosity, we would expect that living with parents and living with a partner increases the likelihood (of all combinations) of religiosity as compared to belong to the category of secular people. Urbanisation has also increased over time, which reduced social control on individual behaviour. Therefore, we expect that living in an urbanised area decreases the likelihood (of all combinations) of religiosity as compared to secularity.
Methods
Data
In order to answer our research questions, cross-national survey data from the European Values Studies (EVS) (Inglehart et al.2000), four waves collected in 1981–1984, 1989–1993, 1995–1998 (as part of WVS) and 1999–2002, respectively, and the European Social Survey (ESS 2008a,b), three rounds collected in 2002–2003, 2004–2006 and 2006–2007, respectively, have been harmonised with regard to relevant variables. Only countries that had participated in more than one wave have been analysed, since we aim at the investigation of longitudinal trends. Contextual level data have been added from World Bank (2008b), OECD (2008) and Norris (2008). These context data have been matched to individual data on the basis of country and the year in which the majority of interviews for a wave in a country were held. To account for possibly disturbing influences of the differences between EVS and ESS data collections, we included a dummy for the data collection (ESS versus EVS) in our analyses.
Measurements: operationalisations of religiosity
Focussing on so many countries at so many different points in time comes at a cost: researchers are forced to restrict themselves to those valid measurements that are actually available at all data points in so many countries.
Religious belief has been operationalised as religious self-definition, a valid measure available in so many data sources whereas other measurements lacked in one or the other source. The observations are recoded into a religious versus non-religious dichotomy. The EVS asks ‘Independently of whether you attend religious services or not, would you say you are: a religious person/not a religious person/an atheist’. In our analyses, non-religious and atheists have been grouped together. The ESS asks a similar question ‘Regardless of whether you belong to a particular religion, how religious would you say you are?’, but with different response options ranging from 0 (‘not at all religious’) to 10 (‘very religious’). Therefore, the ESS measure on religious self-definition is recoded in a way that ensures that it resembles the EVS measure as closely as possible to constitute a seamless longitudinal trend in the data.1 By using a dummy variable for ESS in our models, we corrected for differences in data collection. Furthermore, we tested whether a model without the dummy for ESS would lead to different conclusions with regard to our hypotheses, which was not the case.
Religious belonging has been operationalised as attendance to religious services. EVS asks ‘Apart from weddings, funerals and christenings, about how often do you attend religious services these days?’ and ESS asks ‘Apart from special occasions such as weddings and funerals, about how often do you attend religious services nowadays?’. Both surveys provide answer categories ranging from ‘never’ to ‘more than once a week’ (EVS) or ‘every day’ (ESS). The values have been dichotomised into attending (from ‘every day’ up to ‘attending only on specific holy days’ versus not attending (those who ‘never’ attend plus those who say they attend less often than on specific holy days).2
Operationalisations of independent variables: contextual level
Countries with a communist regime during part of the twentieth century have been coded as former communist countries, whereas the other countries function as reference category.3
Technological developments have been approximated with infant mortality as counter indicator. The effective use of technological developments can greatly decrease infant mortality (World Bank 2008a). In addition, infant mortality is one of the very few societal level indicators of technological development that is widely available for many countries and periods. Infant mortality is operationalised as the number of infants per thousand live births who die within 1 year after live birth. We have used data retrieved from the OECD (2008) as well as Norris (2008). We took the OECD data as a starting point. Missing data in the OECD database are estimated by regression on Norris (2008) data on infant mortality. In some cases, linear interpolation has been applied. The external validity of infant mortality as a counter indicator for technological development is demonstrated by a strong negative correlation (−0.78, n=141; p<0.001) of recent country level infant mortality data from the year 2005 with the recent Knowledge Index for the same year of the World Bank (2009).
Government expenditure on social security has been operationalised as the percentage of GDP devoted to total social expenditure. This information has been retrieved from the OECD (2008).4
Gross domestic product has been, more precisely, operationalised as gross domestic product purchasing power parity per capita constant 2005 international dollar. Data have been retrieved from the World Bank (2008b). Since data for 2007 were not yet available at the date of retrieval, GDP data for 2006 have been used instead.
Individualisation has been approximated by the percentage of people living alone, which is derived from individual information on living with a partner or parent. Those individuals who indicated that they neither lived with a parent nor with a partner have been coded as living alone. Subsequently, the percentage of people living alone is computed for each combination of country and wave.
Operationalisations of independent variables: individual level
Education has been coded as the age at which education was completed with 12 years as minimum and 21 years as maximum values, as it is asked in EVS. The measure of education in ESS is in terms of years of education, which has been recoded to an estimated age by adding 6 and recoding into a minimum value of 12 years and a maximum value of 21 years.
The operationalisation of income in the data collection in different countries as well as different waves varies widely: not only currencies differ, but also the categories of income. In order to get a comparable income measure, we decided to use a z-score of income within the country and wave.
Respondents were asked for their marital status in both EVS and ESS. Both being married and living as married have been coded as living with partner.
In EVS, respondents answered a question as to whether they live with their parents. The ESS asks for the relationship to household members, including (step-) parents (-in law) as an option. These measures have been recoded into a dummy for living with a parent or not. In several countries, the EVS seems to have asked this question only to people who never have been married or living as married. Therefore we have coded people who are or have been married or living as married as not living with a parent.
The categories in which urbanisation has been measured differ between countries and waves. We decided to use a z-score of urbanisation within country and wave in order to get a comparable urbanisation measure. Respondents have been asked for their gender (coded as female versus male), and birth year.
Analyses
Data have been analysed by hierarchical multinomial logistic regression analyses in order to test the different hypotheses simultaneously. In these analyses, individuals (N=254,499) are nested within combinations of country and year (N=178) and combinations of country and year are nested within countries (N=42).5 Missing data for individual or contextual determinants have been substituted by the mean of valid observations. In order to control for differences between cases with valid and mean substituted data, we added dummies for missing values in our analyses for each variable that had some missing data. Since these dummies have solely a methodological purpose, their results are not presented. The robustness of findings has further been investigated by additional analyses: one analysis excluding the dummy for ESS and another analysis excluding ESS data altogether.6
Appendix A contains descriptive results for all independent variables included in the analyses, as well as their VIF (Variance Inflation Factor) values, in order to check for problems of multi-collinearity. VIF values are satisfactory low for the variables used to test our hypotheses. Only for the highly correlated control variables age and birth year we find disturbingly high VIF values. Considering previous convincing evidence on cohort effects (Te Grotenhuis and Scheepers, 2001; Crocket and Voas 2006) we only included birth year in the analyses.
Results
Longitudinal trends
We set out to answer our first research question on the extent to which the popularity of various combinations of believing and belonging has changed over time in European countries.7
At first sight, considering the general trend model in the upper half of Table 1, results show no significant changes over the years regarding the likelihood of any kind of religiosity versus secularity. Moreover, we found no significant difference between our data sources, EVS and ESS.
| Consistent religiosity | Solitary religiosity | Habitual attendance | ||||
|---|---|---|---|---|---|---|
| B | p | B | p | B | p | |
| General trend model | ||||||
| Intercepta | 0.465 | 0.013* | −0.435 | 0.000*** | −1.139 | 0.000*** |
| Year (1990 = 0)a | 0.011 | 0.143 | 0.001 | 0.875 | −0.000 | 0.971 |
| ESS | −0.148 | 0.056 | 0.006 | 0.939 | 0.038 | 0.667 |
| Differential trends model | ||||||
| Intercepta | 0.594 | 0.022* | −0.306 | 0.013* | −1.004 | 0.000*** |
| Former communist | −0.397 | 0.267 | −0.301 | 0.108 | −0.448 | 0.144 |
| Year (1990 = 0)a | −0.001 | 0.873 | 0.004 | 0.567 | −0.022 | 0.008** |
| Year×former communist | 0.043 | 0.001** | 0.003 | 0.801 | 0.065 | 0.000*** |
| ESS | −0.174 | 0.021* | −0.026 | 0.764 | 0.021 | 0.812 |
| Consistent religiosity | Solitary religiosity | Habitual attendance | ||||
|---|---|---|---|---|---|---|
| B | p | B | p | B | p | |
| General trend model | ||||||
| Intercepta | 0.465 | 0.013* | −0.435 | 0.000*** | −1.139 | 0.000*** |
| Year (1990 = 0)a | 0.011 | 0.143 | 0.001 | 0.875 | −0.000 | 0.971 |
| ESS | −0.148 | 0.056 | 0.006 | 0.939 | 0.038 | 0.667 |
| Differential trends model | ||||||
| Intercepta | 0.594 | 0.022* | −0.306 | 0.013* | −1.004 | 0.000*** |
| Former communist | −0.397 | 0.267 | −0.301 | 0.108 | −0.448 | 0.144 |
| Year (1990 = 0)a | −0.001 | 0.873 | 0.004 | 0.567 | −0.022 | 0.008** |
| Year×former communist | 0.043 | 0.001** | 0.003 | 0.801 | 0.065 | 0.000*** |
| ESS | −0.174 | 0.021* | −0.026 | 0.764 | 0.021 | 0.812 |
Reference category: Secularity.
***p<0.001; **p<0.01; *p<0.05; (two-tailed).
aRandom effect.
Then, we set out to explore over time differences between western and former communist countries, shown in the bottom half of Table 1 under the heading of differential trends model, actually indicating differential trends for former communist and western countries. For the ease of interpretation, we present Figures 1 and 2 that contain the estimated probabilities, based on these parameter estimates of Table 1, of the various combinations of religious self-definition and attendance, varying over time and with a correction for ESS data to the EVS standard, for western and former communist countries.8
Figure 1, on longitudinal trends in western countries, shows that the trend (of predicted probabilities) of consistent religiosity remains more or less at the same level (of approximately 46–47%) which is also the case for the trend of secularity (of approximately 25–26%). The trend of solitary religiosity has slightly increased over time (from 18 to 21%). Figure 1 also shows that the category of habitual attendants is the only category that has a significantly declining trend (indicated by the parameter in Table 2) over time (from approximately 11% for 1980 to 7% for 2007).
Predicted probabilities of combinations of religiosity over time in western countries
Predicted probabilities of combinations of religiosity over time in western countries
| Consistent religiosity | Solitary religiosity | Habitual attendance | ||||
|---|---|---|---|---|---|---|
| B | p | B | p | B | p | |
| Intercepta | 8.121 | 0.000*** | 2.884 | 0.077 | 3.751 | 0.023* |
| Contextual | ||||||
| Infant mortality | −0.020 | 0.242 | 0.010 | 0.544 | −0.028 | 0.085 |
| Social expenditure | −0.028 | 0.017* | −0.014 | 0.284 | −0.014 | 0.249 |
| Purch. power (2 log) | −0.381 | 0.001** | −0.113 | 0.265 | −0.309 | 0.003** |
| Individualization | 0.355 | 0.489 | 0.471 | 0.407 | −0.909 | 0.102 |
| Former communist | −0.821 | 0.038* | −0.620 | 0.012* | −0.655 | 0.044* |
| Yeara | 0.033 | 0.001** | 0.024 | 0.011* | −0.004 | 0.683 |
| Former comm.×year | 0.028 | 0.013* | 0.004 | 0.738 | 0.046 | 0.000*** |
| ESS | −0.114 | 0.178 | 0.038 | 0.706 | −0.023 | 0.809 |
| Individual | ||||||
| Education | −0.025 | 0.000*** | −0.043 | 0.000*** | 0.015 | 0.000*** |
| Income (z) | −0.066 | 0.000*** | −0.063 | 0.000*** | −0.002 | 0.832 |
| Living with partner | 0.260 | 0.000*** | 0.085 | 0.000*** | 0.164 | 0.000*** |
| Living with parent | 0.355 | 0.000*** | 0.084 | 0.001** | 0.247 | 0.000*** |
| Town size (z) | −0.229 | 0.000*** | −0.072 | 0.000*** | −0.130 | 0.000*** |
| Female | 0.867 | 0.000*** | 0.526 | 0.000*** | 0.357 | 0.000*** |
| Birth year (1881 = 0)a | −0.025 | 0.000*** | −0.014 | 0.000*** | −0.005 | 0.004** |
| Consistent religiosity | Solitary religiosity | Habitual attendance | ||||
|---|---|---|---|---|---|---|
| B | p | B | p | B | p | |
| Intercepta | 8.121 | 0.000*** | 2.884 | 0.077 | 3.751 | 0.023* |
| Contextual | ||||||
| Infant mortality | −0.020 | 0.242 | 0.010 | 0.544 | −0.028 | 0.085 |
| Social expenditure | −0.028 | 0.017* | −0.014 | 0.284 | −0.014 | 0.249 |
| Purch. power (2 log) | −0.381 | 0.001** | −0.113 | 0.265 | −0.309 | 0.003** |
| Individualization | 0.355 | 0.489 | 0.471 | 0.407 | −0.909 | 0.102 |
| Former communist | −0.821 | 0.038* | −0.620 | 0.012* | −0.655 | 0.044* |
| Yeara | 0.033 | 0.001** | 0.024 | 0.011* | −0.004 | 0.683 |
| Former comm.×year | 0.028 | 0.013* | 0.004 | 0.738 | 0.046 | 0.000*** |
| ESS | −0.114 | 0.178 | 0.038 | 0.706 | −0.023 | 0.809 |
| Individual | ||||||
| Education | −0.025 | 0.000*** | −0.043 | 0.000*** | 0.015 | 0.000*** |
| Income (z) | −0.066 | 0.000*** | −0.063 | 0.000*** | −0.002 | 0.832 |
| Living with partner | 0.260 | 0.000*** | 0.085 | 0.000*** | 0.164 | 0.000*** |
| Living with parent | 0.355 | 0.000*** | 0.084 | 0.001** | 0.247 | 0.000*** |
| Town size (z) | −0.229 | 0.000*** | −0.072 | 0.000*** | −0.130 | 0.000*** |
| Female | 0.867 | 0.000*** | 0.526 | 0.000*** | 0.357 | 0.000*** |
| Birth year (1881 = 0)a | −0.025 | 0.000*** | −0.014 | 0.000*** | −0.005 | 0.004** |
Reference category: Secularity.
***p<0.001; **p<0.01; *p<0.05; (two-tailed).
aRandom effect.
More important, however, is our test of the hypotheses that a communist history is related to an over time increase in the likelihood of belonging to the categories of consistent religious people and habitual attendants versus belonging to the category of secular people. We find that these hypotheses are supported, as indicated by the significantly positive interaction parameters (see Table 1: year×former communist country). Figure 2 shows that, in former communist countries, consistent religiosity has gained popularity (with increasing predicted probabilities over time from approximately 41% in 1990 to 54% in 2007).9 At the same time, the popularity of secularity has decreased in these countries (from approximately 33 to 22%).10 The predicted probability of habitual attendance as compared to secularity has inclined slightly (from approximately 6 to 11%), yet significantly in this period which is due to the decline of secularity. These findings imply that these categories of consistent religious people and habitual attendants, as compared to the category of secularity, have over time become more popular in former communist countries as compared to the category of secular people which has decreased over time.11 Contrary to our expectation we did not find a significantly different trend in former communist countries in the over time development of the popularity of solitary religiosity versus secularity, where solitary religiosity declined somewhat (from approximately 20 to 13%).12
Predicted probabilities of combinations of religiosity over time in former communist countries
Predicted probabilities of combinations of religiosity over time in former communist countries
Overall, these findings imply that categories of solitary religious people as compared to secular people were smaller in both western and former communist countries; in former communist countries, its popularity decreased somewhat (yet, non-significantly); in western countries its popularity increased somewhat (yet, non-significantly). Rather new is the finding that the category of consistent religious people as compared to the category of secular people has over time become more common in former communist countries.
Contextual and individual determinants
Next, we set out to answer our second question on the contextual- and individual-level characteristics related to these combinations of believing and belonging. In these analyses, year of survey is substituted with substantial period characteristics derived from modernisation theories, e.g., social expenditure in a particular year in a particular country, and their individual level counterparts are also included. A dummy variable for former communism was included as country characteristics in order to test market perspectives on religion. In this model we, again, find that consistent religiosity and habitual attendance have increased in former communist countries over the years, as compared to secularity. With this model, we estimated to what extent the contextual and individual level determinants are related to individual combinations of religious self-definition and attendance.
The hypothesis that in societies with more technological developments, indicated by a lower infant mortality, all combinations of religiosity are less likely to prevail (as compared to the category of secularity) is rejected: this determinant has no significant effects. The hypothesis that an individual's scientific worldview, as indicated by a longer exposure to the educational system, decreases the likelihood (of belonging to all combinations) of religiosity as compared to belonging to the category of secular people is supported regarding consistent religiosity and solitary religiosity. These findings imply that the longer people have been educated, the less likely they are to belong to categories of consistent religious people or solitary religious people. However, the likelihood of habitual attendance is slightly higher among higher educated people.
Our hypothesis that social expenditure by the government decreases the likelihood (of all combinations) of religiosity (as compared to secularity) is only supported for the category of consistent religiosity. The expectation that financial security, measured by purchasing power per capita, decreases the likelihood (of belonging to all kinds) of religiosity versus secularity gains support with regard to consistent religiosity and of habitual attendance, but not for solitary religiosity. Individual financial security, as indicated by income, shows the expected negative relationship: the higher people's income, the less likely they are to belong to the categories of consistent religious people and solitary religious people as compared to the category of secular people.
We hypothesised that individualisation would decrease the likelihood (of all combinations) of religiosity versus secularity and found no evidence: none of the parameters reached significance. As expected, we found that living together with a partner as well as living with parents increase the likelihood to belong to the categories of consistent religiosity, solitary religiosity and habitual attendants as compared to belong to seculars. Individualisation of the social environment, as indicated by urbanisation, decreases the likelihood (to belong to any combination) of religiosity as compared to secularity.
Gender and birth year were taken into account as individual level control variables in our model. Females are more likely than males to belong to (any of the three different combinations of) religiosity as compared to seculars. Individuals born later, i.e., belonging to more recent cohorts, are significantly less likely to belong to any of the three different combinations of religiosity as compared to the category of seculars.
Summary of findings: In order to provide more insight into the large number of results, we present a summary of our findings (see Scheme 1).
| Consistent religiosity | Solitary religiosity | Habitual attendance | |
|---|---|---|---|
| Contextual level | |||
| Over time changes in former communist countries | + | 0 | + |
| (Lack of) Technological development | 0 | 0 | 0 |
| Social expenditure | − | 0 | 0 |
| GDP per capita purchasing power | − | 0 | − |
| Individualization | 0 | 0 | 0 |
| Individual level | |||
| Education completed | − | − | + |
| Income (z within country×survey) | − | − | 0 |
| Living with partner | + | + | + |
| Living with parent | + | + | + |
| Town size (z within country×survey) | − | − | − |
| Female | + | + | + |
| Birth year | − | − | − |
| Consistent religiosity | Solitary religiosity | Habitual attendance | |
|---|---|---|---|
| Contextual level | |||
| Over time changes in former communist countries | + | 0 | + |
| (Lack of) Technological development | 0 | 0 | 0 |
| Social expenditure | − | 0 | 0 |
| GDP per capita purchasing power | − | 0 | − |
| Individualization | 0 | 0 | 0 |
| Individual level | |||
| Education completed | − | − | + |
| Income (z within country×survey) | − | − | 0 |
| Living with partner | + | + | + |
| Living with parent | + | + | + |
| Town size (z within country×survey) | − | − | − |
| Female | + | + | + |
| Birth year | − | − | − |
Reference category: Secularity.
−, negative relationship;+, positive relationship; 0, non-significant.
Summary and discussion
This contribution set out to investigate trends and determinants of individual combinations regarding believing and belonging from a cross-national European and longitudinal perspective, taking advantage of cross-national secondary data collected over a long time span 1981–2007, constituting a more comprehensive data base for Europe than any previous research we know. Moreover, this data base provided us with possibilities to test, more systematically than previous research, a set of complementary hypotheses from both market and modernisation theories on changes in the European religious landscape, including former communist countries.
When looking at the longitudinal results, we found that the category of solitary religious people, i.e., those ‘believing without belonging’, has over time actually been a clearly discernible category, slightly increasing in western countries, slightly decreasing in former communist countries as compared to seculars. The category of ‘belonging without believing’, i.e., habitual attendants, has been smaller, slightly decreasing in western countries, slightly increasing in former communist countries. Moreover, at any point in time under study, both in western and former communist countries, the category of secular people turned out to be outnumbered by the category of consistent religious people, i.e., those belonging and believing. Yet, we have to acknowledge that these findings are limited due to valid single item measurements available over such a long time span in so many countries.
Next, we evaluated hypotheses derived from two perspectives on religion: modernisation and market theory. In general we found that consistent religiosity and habitual attendance, as compared to secularity, have over time become more common in former communist countries. This confirms statements of Greeley (1994, 2003) that consumption of religious goods has over time increased in former communist countries after the weakening of state control over religion. We have to acknowledge that we cannot ascertain possible effects of increasing religious freedom on religiosity since religious freedom indicators are not available over different periods. Yet, this result provides more evidence to … a slight reverse tendency in some eastern European countries … indicated by Pollack (2008) and is well interpretable from the market perspective on religion, at least in these former communist countries (Finke 1990; Chaves and Cann 1992; Stark and Iannaccone 1994; Iannacone 1995; Stark and Finke 2000).
The modernisation perspective also gains moderate support. Societies with high levels of financial security harbour less consistent religiosity and habitual attendance as compared to secularity. We also found that social expenditure decreases the likelihood of consistent religiosity as compared to secularity. We found, however, no support for hypotheses on technological developments: although Ruiter and Van Tubergen (2009) used a different indicator for technological developments, they also found no consistent, significant effects. We did find, however, that scientific worldviews play their role mainly at the individual level: education decreases the likelihood to belong to the categories of consistent religious and solitary religious people as compared to the category of secular people. Financial security at the individual level, indicated by income, decreases the likelihood of both consistent religiosity as well solitary religiosity. We found no significant effects of individualisation at the societal level. In this regard, other valid indicators might test these theoretical insights more appropriately. Finally, we found that living in urban settings decreases (all kinds of) religiosity in favour of secularity at the individual level. These findings support arguments that these aspects of modernisation induce a decline of religiosity (see e.g., Berger 1967; Te Grotenhuis and Scheepers 2001; Bruce 2002; Norris and Inglehart 2004).
Concluding, we propose that considering and analyzing the strong version (Voas and Crocket 2005) of combinations of believing and belonging at the individual level turns out to be a valuable path of investigation. Believing without belonging turns out to have been a relatively small, yet slightly increasing phenomenon in western societies whereas the phenomenon has been slightly decreasing in former communist societies. Applying propositions derived from the perspective of religious markets to, particularly former communist societies, has delivered new evidence regarding the increased popularity of consistent religiosity in these countries. Results show support for modernisation theories (Ruiter and van Tubergen, 2009) more particularly that financial security and, to a lesser extent, social security reduces religiosity.
Footnotes
An analysis of the two operationalisations in the fourth wave of the EVS (1999–2002) and the first round of the ESS (2002–2003) in the same countries shows that the treatment of the middle category as religious is best in terms of comparability of average levels of religiosity (recoding 0–4 as non-religious versus 5–10 as religious): the difference between ESS, round 1 data is on average only 1.77% lower regarding religious people than the previous EVS wave 4 data. Moreover, we also tested whether the findings were robust in case we used a slightly different mode of coding religious self-definition: the results turned out to be substantially similar.
We also considered and executed an even stricter analysis, in which we only included those who never attend. We found very much similar trends and results.
We considered to include the Religious Freedom Index derived from Inglehart and Norris which is, however, not available at so many points in time. Therefore, we followed the advice of anonymous reviewers to refrain from including it.
Unfortunately, the OECD measure for social expenditure is not available for several countries and periods. We tested whether leaving out social expenditure from the final model would substantially change the estimates for the other predictors in the model. This turned out not to be the case: a model without social expenditure would lead to the same conclusions with regard to the other predictors.
Time and country are actually crossed factors. However, the software in which we were able to run our analyses, HLM, does not provide the possibility of cross-classified multinomial logistic models. Therefore, we decided to treat years in which data are collected as if they are nested within countries.
We also tried to estimate a model without the EVS data. Unfortunately, probably due to the limited number of higher level cases and a high number of higher level independent variables, this model did not converge. None the less, we are confident that a model without ESS data is sufficient as a robustness test, because a model without EVS data should logically result in more or less similar deviations – but in the opposite direction – from the model on the complete data set.
We briefly describe the presence of the four types of believing and belonging in Europe, ignoring longitudinal changes, just for this descriptive purpose. Let us start with the consistent categories: the category of secularity appeared to contain 31.6% of all respondents, whereas the category of consistent religiosity contained 40.8% of the respondents. Next, we focus at the inconsistent categories: the category of solitary religious people contained 19.6%, and the fourth, often neglected, category of habitual attendants contained 8.0%. Yet, we find substantial variation in the presence of these types across countries.
Predicted probabilities are calculated from the logits of the various combinations of religiosity versus secularity. These logits are derived from the diverging trends model in Table 1. The predicted probability for secularity (the reference category) is: 1/(1 + e^logitConsistentreligiosity+e^logitSolitaryreligiosity+e^logitHabitualattendance). The predicted probability of non-reference categories is, for example for consistent religiosity: e^logitConsistentre‘ligiosity/(1 + e^logitConsistentreligiosity+e^logitSolitaryreligiosity+e^logitHabitualattendance). See e.g., Borooah (2001) for a detailed discussion.
We decomposed this trend descriptively for the different former communist countries. We found this trend to be rather strong in following countries like: Bulgaria, Belarus, Czech Republic, Hungary, Latvia, Poland, the Russian Federation, Slovakia, Slovenia, Ukraine, and East Germany; however, somewhat less strong in countries like: Estonia and Romania.
We also decomposed this trend for different former communist countries. This trend was rather strong in countries like: Bulgaria, Belarus, Czech Republic, Estonia, Hungary, Latvia, the Russian Federation, Slovakia, Slovenia, Ukraine and East Germany.
Let us explain briefly. We find that the main effects of year of survey (–0.001 for consistent religiosity versus secularity; –0.022 for habitual attendance versus secularity) apply to western European countries. However, for countries with a communist past, the estimated trend is the main effect of year of survey with the interaction added (–0.001 + 0.043 = 0.042 for consistent religiosity versus secularity; –0.022 + 0.065 = 0.043 for habitual attendance versus secularity).
Recent descriptive analyses regarding church attendance and religious self-assessment show partially similar trends in many former communist countries (Mueller, 2011): moderate over-time increases in church attendance and rather massive increases in religious self-assessment. In these analyses different data sources were employed, including data in 1990 and 2008, meanwhile ignoring available data sources from the years in between 1990 and 2008.
Acknowledgement
This research project has been supported by a major grant from NORFACE, entitled: Religious Sources of Solidarity, explicitly focusing on religious changes in Europe.
Appendix
| N | Min | Max | Mean | SD | VIF | |
|---|---|---|---|---|---|---|
| Contextual level | ||||||
| Former communist | 254,499 | 0 | 1 | 0.37 | 0.48 | 2.69 |
| Infant mortality | 254,499 | 2.40 | 28.64 | 7.65 | 4.95 | 4.11 |
| Purchasing power (2 log) | 251,365 | 1481 | 67,620 | 23,202.27 | 11,688.71 | 5.71 |
| Social expenditure | 113,402 | 13.60 | 32.00 | 22.03 | 4.02 | 1.07 |
| Individualization | 254,499 | 0.06 | 0.44 | 0.24 | 0.07 | 1.39 |
| Year | 254,499 | 1981 | 2007 | 1998.90 | 6.85 | |
| Individual level | ||||||
| Education (age completed) | 249,831 | 12 | 21 | 17.34 | 2.93 | 1.27 |
| z of income (within country * survey) | 198,748 | −3.91 | 8.38 | 0.00 | 1.00 | 1.19 |
| Living with partner | 253,916 | 0 | 1 | 0.63 | 0.48 | 1.55 |
| Living with parent | 253,922 | 0 | 1 | 0.13 | 0.34 | 1.89 |
| z of town size (within country * survey) | 249,534 | −3.27 | 2.44 | 0.00 | 1.00 | 1.04 |
| Female | 254,265 | 0 | 1 | 0.53 | 0.50 | 1.03 |
| Birth year | 253,517 | 1881 | 1992 | 1953.36 | 18.40 | 25.50 |
| Age | 253,517 | 13 | 110 | 45.53 | 17.70 | 24.66 |
| N | Min | Max | Mean | SD | VIF | |
|---|---|---|---|---|---|---|
| Contextual level | ||||||
| Former communist | 254,499 | 0 | 1 | 0.37 | 0.48 | 2.69 |
| Infant mortality | 254,499 | 2.40 | 28.64 | 7.65 | 4.95 | 4.11 |
| Purchasing power (2 log) | 251,365 | 1481 | 67,620 | 23,202.27 | 11,688.71 | 5.71 |
| Social expenditure | 113,402 | 13.60 | 32.00 | 22.03 | 4.02 | 1.07 |
| Individualization | 254,499 | 0.06 | 0.44 | 0.24 | 0.07 | 1.39 |
| Year | 254,499 | 1981 | 2007 | 1998.90 | 6.85 | |
| Individual level | ||||||
| Education (age completed) | 249,831 | 12 | 21 | 17.34 | 2.93 | 1.27 |
| z of income (within country * survey) | 198,748 | −3.91 | 8.38 | 0.00 | 1.00 | 1.19 |
| Living with partner | 253,916 | 0 | 1 | 0.63 | 0.48 | 1.55 |
| Living with parent | 253,922 | 0 | 1 | 0.13 | 0.34 | 1.89 |
| z of town size (within country * survey) | 249,534 | −3.27 | 2.44 | 0.00 | 1.00 | 1.04 |
| Female | 254,265 | 0 | 1 | 0.53 | 0.50 | 1.03 |
| Birth year | 253,517 | 1881 | 1992 | 1953.36 | 18.40 | 25.50 |
| Age | 253,517 | 13 | 110 | 45.53 | 17.70 | 24.66 |
References
Jan Reitsma was a post-doctoral researcher at the Faculty of Religious Studies of Radboud University in Nijmegen, working on a grant from NORFACE on a project labelled ‘Religious sources of solidarity’.
Ben Pelzer is an assistant professor of comparative methodology in the Faculty of Social Sciences of the same university
Peer Scheepers is a full professor of comparative methodology in the Faculty of Social Sciences of the same university and principal investigator of the NORFACE project.
Hans Schilderman is a full professor of religion and care in the Faculty of Religious Studies of the same university.