Trends in Religious Affiliation and Practice
Figure 1a–c presents the distribution of religious affiliation by birth cohorts in Britain, France and the Netherlands from 1930 to 1979 (for descriptive statistics and sample sizes, see “Appendix,” Table 5a–f). As in previous studies, the proportion of people who identify as belonging to a particular denomination decreases in each subsequent cohort in all three countries. Both in Britain and in the Netherlands, religiously affiliated people are the majority until the cohorts born in the 1970s, in which the group of “no religion” becomes the majority. Nearly 60 % of the youngest cohort in Britain state they have no religion, while in the Netherlands this proportion is slightly lower (53 %). In France, on the other hand, the proportion of non-affiliated people is relatively low and the vast majority of people define themselves as Roman Catholic, although their share is decreasing from one cohort to another. This may indicate a strong attachment to Catholic cultural identity, which is also associated with national French identity (Byrnes 2005; Hervieu-Léger 1990, 2000; Pace 2007), although this proportion may also be exaggerated as a result of the particular phrasing of the question on religious affiliation.
The increase in the “no religion” group in Britain from one cohort to the next appears to be mainly at the expense of those who identify as Protestants, as their proportion among recent cohorts is closer to that of Catholics. The latter group, on the other hand, remains stable at around 8–10 %. In the Netherlands, the share of those affiliated with the Dutch Reformed Church as well as the proportion of Roman Catholics shows a decrease across cohorts. On the other hand, the proportion of conservative Protestants (Evangelicals and Orthodox Calvinists) remains stable at around 10 %. In all countries, the share of those affiliated with the minority religions is larger among the younger cohorts, although it remains relatively low.
Figure 2a–c shows the proportion of regular attendants at religious services (at least once a month) within each birth cohort and by religious denomination. The solid line indicates the total proportion within the population who are regular attenders.
As in the case of religious affiliation, the share of religiously active people is much smaller among the younger cohorts. While over a fifth of the oldest cohorts in Britain and France are regular attenders, their proportion among the youngest cohort declines to 11 % in Britain and only 8 % in France.
In the Netherlands, attendance rates appear to be slightly higher, with almost 40 % among the oldest cohort, which decreases to 13 % among the youngest cohort. Nonetheless, the rate of decline in each country appears to slow down, as attendance rates for the 1970–1979 cohort are not very different from those who were born a decade earlier. These results are in line with recent studies, showing that the decline in religious service attendance in Western Europe is leveling off among younger cohorts (Burkimsher 2014; Kaufmann et al. 2012).
Furthermore, the decline in religious practice among Catholics appears to be steeper compared with that in other religions. This is similar to the trends found in the USA, which were attributed in part to the growing dissent to the church teachings on sexuality (Hout and Greeley 1987). As in the general population, the group of other religions also show a decline in service attendance among the younger birth cohorts, except in the case of France, where this group shows no consistent trend.
Fertility Trends by Religiosity
This section includes findings on trends in completed fertility across birth cohorts of women by religious affiliation and practice. The results from Britain (Fig. 3a) show a sharp decline in fertility among practicing Catholic women from 3.5 children on average among those born in 1930–1939 to around replacement level in the 1950–1959 cohort and then an increase to 2.5 among the youngest cohort. Nominal Catholic women experienced a similar trend with a decline from 2.5 children to 1.9 and then a rise to 2.4 children. Practicing Protestant women have generally lower fertility levels compared with their Catholics counterparts in Britain, although their fertility rates remain rather stable at around replacement level. In contrast to that, nominal Protestant women experienced a decline from 2.3 children to 1.8 children, converging with the fertility level of the non-affiliated women. Thus, fertility differences among religious groups in Britain appear to have narrowed, although Catholic women, whether practicing or nominal, still show distinctly higher fertility compared to other women.
In France, on the other hand, there is a clearer pattern of fertility divergence between practicing Catholic women and non-practicing ones, whether they are religiously affiliated or not. Following an initial decline from an average of 2.7 children, fertility of younger cohorts of practicing Catholic women rose back again to a level of 2.6, while fertility of nominal Catholic and non-affiliated women continued to decline to a level of 1.9 and 1.7, respectively (Fig. 3b). A similar decline (from an average of 2.9 children) followed by a rise in fertility has also occurred among practicing Catholic women in the Netherlands (Fig. 3c), although fertility has not returned to its initial level and the average family size for practicing Catholics in the youngest cohort is 2.4. By contrast, fertility among nominally Catholic women declined from 2.6 to 1.8 children on average.
Interestingly, practicing Dutch Reformed women experienced a substantial increase in fertility from 2.4 to three children on average, while fertility of nominal Dutch Reformed, which was initially at a similar level, declined to 2.2 children. A divergence has also occurred among practicing and nominal conservative Protestants, as the former maintained relatively high fertility levels (although with a decline from 3.5 to 2.6), while the latter experienced a decline from three to less than two children on average.
Although each country shows somewhat different trends, a general pattern emerges where practicing religious women either maintain high fertility levels or even experience an increase in family size, while non-practicing and non-affiliated women experienced a decline in fertility from one cohort to another (with the exception of nominal Catholic women in Britain). Moreover, women who state having no religion consistently show the lowest fertility compared with all other groups. These findings give some support for the research hypotheses on religious differences in fertility and the increasing fertility gap by religious practice.
The following section describes the results of a multivariate regression model for completed fertility, controlling for religious group (combined affiliation and practice), birth cohort, level of education and nativity.
Multivariate Regression Analysis
Table 1 presents the results of an OLS multivariate regression model for the number of children ever born to women aged 40 and above. The first model in each country is the basic model for the relationship between religious group and fertility, including other covariates, while the second model introduces interaction terms between religious group and birth cohort.Footnote 7 The predicted means for the average number of children by religious group and cohort are shown in “Appendix,” Table 6a–c.
The results of the first model show that women who are either practicing or nominally affiliated with a particular religion have significantly higher fertility compared with women who profess to no religion. The exception to that is in France, where no significant difference was found between non-affiliated and nominally Catholic women. This may be due to the particularly large proportion of the latter in France which may be highly heterogeneous in terms of loyalty to Catholic norms about reproduction.
The findings from the second model of the regression, which includes the interaction term between religious group and birth cohort, show some contrasting trends in each country. In Britain, although the main effect for practicing Catholic women remains strongly positive, the interaction between practicing Catholics and birth cohort is significantly negative in reference to the oldest cohort group, reflecting the sharp decline in fertility to levels that are closer to the non-affiliated. Significant negative interaction is also found for nominal Catholics from the 1950–1959 cohort and for nominal Protestants. Thus, contrary to the third hypothesis about increasing fertility differences by religiosity, the results from Britain point to a convergence of family size between nominally and practicing religious women and their non-affiliated counterparts. Nevertheless, as can be seen by the predicted means of children ever born (Table 6a–c), the gap between non-affiliated and Catholic women (practicing or nominal) increases again in the most recent cohort (1960–1965). Therefore, the apparent fertility decline among Catholic women may be temporary. On the other hand, in France and the Netherlands there is evidence of increasing divergence between non-affiliated and practicing religious women: In France, this is shown by the positive interaction effect among Practicing Catholics in the two recent cohorts. In the Netherlands, a positive interaction term is found for practicing Dutch Reformed in all cohorts compared to the earliest one. Among practicing conservative Protestants, the interaction is only significant for the 1950–1959 cohort, while no interaction effect is found for practicing Catholics.
The increasing fertility among practicing Dutch Reformed in relation to their non-affiliated counterparts could be attributed to the growing proportion of the fundamental wing in this group, as a result of their stronger attachment to the church compared with liberal members who are more likely to reduce their attendance or leave their faith altogether (Knippenberg 1998, 2005).
The results of the OLS regression and the predicted means for children ever born give further support to the first two hypotheses: across cohorts, practicing religious women have the highest fertility levels compared with other groups, while non-affiliated women are consistently at the lower end of the scale. Fertility levels of nominally religious women are generally lower than that of practicing ones, although in most cases they have significantly higher fertility than non-affiliated women. The third hypothesis, which implies an increasing fertility gap between practicing and non-practicing women among the younger cohorts, is only partially supported by the results. However, these findings are limited to women who have already completed their births. The following analysis follows religious differences in the transition to first birth by reconstructing women’s birth histories from the age of fifteen.
Event History Analysis of the Transition to First Birth
Table 2a–c summarizes the main findings from the reconstructed birth histories of women born before and after 1960 by religious group. It is shown that while the median age at first birth has increased for all women who were born after 1960, the proportion of childless women among the actively religious has decreased in all three countries. By contrast, non-affiliated women within the post-1960 cohorts have the highest proportion of childlessness and nominal religious women are usually found in between. In addition, the proportion of women who have had their first birth outside marriage has increased substantially for all women, although the largest increase was among the non-affiliated. In both cohort groups, non-affiliated women have the highest prevalence of births outside marriage, except among the pre-1960 cohorts in Britain, where nominal Catholic women show a higher percentage of first births out of wedlock (17 % compared to 11 % among the non-affiliated). This may be the consequence of socioeconomic differences between Catholics and other population groups in Britain (O’Grada and Walsh 1995).
As expected, these findings point to more traditional family behaviors among women with stronger attachment to religion. In addition, religious differences in childlessness and births out of wedlock have become more marked for the post-1960 cohorts. These findings therefore support the fourth hypothesis about the higher propensity of the more religious women to enter motherhood and the increasing gap in this propensity among religious groups from recent cohorts. The significance of these differences, controlling for other socio-demographic factors, is further explored in the logistic regression analysis, which is presented in Table 3.
The results of the logistic regression model appear to vary from one country to another. In Britain for example, no significant difference is found in the likelihood of first birth for practicing Catholic or Protestant women in comparison with non-affiliated women. However, when controlling for marital status, nominal Catholic women in the pre-1960 cohorts show significantly higher likelihood of entering motherhood compared with the reference group (odds ratio of 1.423, p < 0.01). A similar trend is found in the Netherlands, where nominal and practicing Catholic women born before 1960 show significantly higher likelihood of experiencing the transition to first birth when marital status is controlled. This could be attributed to the tendency of late entry to marriage (as well as later transition to first birth) among Catholics (Lehrer 2000, 2004b). According to Lehrer, the cost of marital dissolution for Catholics is especially high, due to the anti-divorce doctrine of the Catholic Church. Therefore, the search for a spouse will be longer for Catholics than for other religious groups (ibid). This pattern, however, is only evident among earlier cohorts. For Catholic women born after 1960 in the Netherlands, as well as among Dutch Reformed, the higher transition rates to first birth are insignificant once marital status is included in the model. Similarly, among women born after 1960 in France, the coefficient for practicing Catholic changes from significantly positive to insignificant when controlling for marital status and a negative relationship emerges for nominal Catholic women. This may be the result of differential marriage trends among religious groups in these countries; while the proportion of ever married people in Europe has declined over time (Perelli-Harris et al. 2009), the more religious segments of society are more likely to maintain high marriage rates (Régnier-Loilier and Prioux 2008). Lehrer (2004b) has argued that the stronger marriage preferences among religious women are closely linked to fertility aspirations. Thus, women who expect to have a large family would be more inclined to marry than to live in non-marital cohabitation, which is considered to be a less stable union. Therefore, the higher transition rates to first birth among more religious women in recent cohorts can in some cases be attributed to differences in marriage patterns.
The group that shows consistently higher likelihood of experiencing first birth compared with non-affiliated women among both earlier and recent cohorts is practicing conservative Protestant women in the Netherlands. On the other hand, for nominal women from this group, the higher likelihood of first birth compared with women with no religion is only evident among earlier cohorts.
Thus, while the findings from Table 2a–c support the fourth hypothesis on the increasing gap in the transition to first birth by level of religiosity, the results of the multivariate logistic regression are not consistent across countries. This is partly due to differences in the timing of first birth, as in some religious denominations women tend to start childbearing at a later age and are more likely to defer the first birth until after they are formally married. In addition, as past studies show, more religious women have higher marriage rates compared with others, which are also related to higher likelihood of entering motherhood. Indeed, the results of the logistic regression show that in all three countries, current marital status is strongly linked to the transition to first birth, as the odds for entering motherhood are considerably lower for single women and those in non-marital cohabitation compared with that of married women.
Another noteworthy finding is that in all countries, the highest level of education achieved as well as the time-varying covariate of enrollment in education is associated with lower propensity to experience the transition to first birth. In addition, a significant negative effect on first birth is found in Britain for those who were born abroad, while in France and in the Netherlands the effect is positive. This is likely to be the result of differences in the sending countries from which people immigrate to each of these destinations. For example, in recent decades a large proportion of immigrants to Britain have come from Eastern Europe, where fertility levels are often lower in relation to Britain (Coleman and Dubuc 2010). On the other hand, the majority of immigrants in France and the Netherlands arrived from countries that are characterized with relatively high fertility.Footnote 8 It should be noted, though, that if the analysis included women from non-Christian religions, the results for the effect of nativity might have been more strongly positive, since immigrants to Europe from Muslim countries tend to have higher fertility compared with the native population (Kaufmann et al. 2012).
Finally, a robustness check was made using different cutoff points (1955 and 1965) between the cohorts. In general, the change in the cutoff point did not yield very different results, except in the case of Britain, where the odds for the transition to first birth among nominal and practicing Catholic women who were born after 1965 were significantly higher compared to non-affiliated women, ceteris paribus (not shown). This trend is similar to the findings on completed family size (see Fig. 3a), where the fall in Catholic fertility is followed by a rise among the 1960–1965 cohort. Therefore, this may represent a trend of resurgence to higher fertility levels within that group.
Religiosity as a Determinant of Progression to Higher Parity
So far, the presented findings were based on retrospective data analysis, in which religiosity is measured after the time of childbearing. This section includes findings from a longitudinal analysis of the relationship between religiosity and subsequent childbearing. Table 4 presents the results of the logistic regression model for the likelihood of experiencing an additional childbirth within 3 years from the first observation. The findings from Britain show no relationship between religious indicators and the likelihood of having an additional birth within the observed time period, which is in accordance with earlier findings on fertility convergence between religious groups there. By contrast, practicing Catholic women in France show higher likelihood of progressing to higher parity compared with non-affiliated women (odds ratio of 2.421, p < 0.1).
In the Netherlands, a positive effect on parity progression was found among nominal Catholics (odds ratio of 1.794, p < 0.05) and conservative Protestant women who attend services regularly (odds ratio of 3.017, p < 0.05) compared with their non-affiliated peers. No significant effect was found among the other religious groups, although this may be due to small sample sizes; when Dutch Reformed are grouped together with conservative Protestant women, both nominal and practicing women have a significantly higher likelihood of progressing to higher parity compared with non-affiliated women (not shown). Therefore, in the Netherlands fertility patterns not only differ by religious practice, but also nominally religious women are more likely to progress to higher parity compared with non-affiliated ones. These patterns may be related to the pronounced religious diversity in the Netherlands and the fact that religion was considered an important aspect of social identity, at least during the pillarization period (Knippenberg 1998; van Poppel 1985; van Rooden 2003).