1Education has long been recognized as a key factor in explaining fertility behaviour among women, as higher education is associated with lower fertility levels via several mechanisms. For example, extended time in education leads to the postponement of women’s age at first birth (Ní Bhrolcháin and Beaujouan, 2012), which may result in higher rates of childlessness or lower completed fertility. Furthermore, education can introduce changes in traditional family values and preferences (McDonald, 2000; Esping-Andersen, 2009). Becker’s (1991) economic theory of the family predicts that a rise in women’s earning potential reduces the demand for children by increasing the opportunity costs of leaving the labour market to care for a child. Moreover, the quality–quantity trade-off model proposed by Becker and Lewis (1973) anticipates that parents with higher income invest more in their children’s quality of life, which raises the cost of each additional child and leads to lower fertility. However, some authors suggest that the relationship between education and fertility may vary according to cultural contexts, such as religious affiliation and the extent of religious involvement, which affect the perceived costs and benefits attributed to childbearing and rearing (Heaton, 1998; Lehrer, 2004; Goldscheider, 2006; Newman and Hugo, 2006).

2Recent studies have documented the continuing relevance of religion to childbearing behaviour in Western countries. Across Europe and the United States, religiously affiliated individuals have more children compared to those who report having no religion (Philipov and Berghammer, 2007; Frejka and Westoff, 2008). Moreover, those who are more devout in each denomination (measured by religious service attendance and self-assessed religiosity) register higher fertility than their less religious peers (Philipov and Berghammer, 2007; Frejka and Westoff, 2008; Zhang, 2008; Berghammer, 2012; Peri-Rotem, 2016). These patterns are attributed to the most prominent religious traditions attaching greater importance to family life and childbearing (Norris and Inglehart, 2004; Chatters and Taylor, 2005; Adsera, 2006). Nevertheless, the relevance of religion to fertility behaviour is not limited to specific religious doctrines on methods of family planning or family-size norms. According to Goldscheider (1971, 2006), the mechanism by which religion shapes fertility behaviour is better understood within the wider context of social organization, such as the social norms and gender-role perceptions associated with different religious communities. Furthermore, women’s investment in education and their level of participation in the labour force are also linked with religious affiliation and practice (Lehrer, 2004), which may affect fertility outcomes.

3The interaction between education and religion and its implications for fertility outcomes are of particular interest due to education’s association with the rejection of religious authority and traditions (Lesthaeghe and Surkyn, 1988; Surkyn and Lesthaeghe, 2004). Nonetheless, little is known about the interrelationships of education, religion, and fertility outcomes in Europe, where the religious landscape has been changing rapidly over recent decades. Focusing on completed fertility and childlessness, this study adds to the literature by analysing fertility differences by education and the way this relationship varies by religious affiliation and religiosity among women born between the late 1920s and late 1960s in Britain and France (for simplicity, this paper refers to those born in these countries as British and French, even though some may not be citizens). These countries provide useful case studies for this research because they share some sociodemographic characteristics and are geographically close. However, they differ in their religious composition and cultural heritage: the dominant religion in France is Roman Catholicism, whereas Britain is largely Protestant, with the Church of England as the established religious institution. Thus, most British Christians identify as Protestant, and only a minority (about 10% of the population) identify as Roman Catholic (Lee, 2012). Among non-Christian populations in Britain and France, Muslims comprise the largest faith community. Although the overall share of non-Christian populations is increasing, it represents less than a tenth of the populations in Britain and France (Pew Research Center, 2015). [1]

4During the latter half of the 20th century, both Britain and France experienced a marked decline in the proportion of religiously affiliated persons and religious service attendance (Lee, 2012; Baudin, 2015). In both Britain and France, the decline in religiosity is largely generational: over one-fifth of those born in the 1930s attended religious services on a monthly basis, whereas this proportion drops to around 10% among those born in the 1960s (Peri-Rotem, 2016). Nonetheless, previous studies highlight the continuing importance of religion in shaping attitudes and behaviour among actively religious individuals (Régnier-Loilier and Prioux, 2008; Peri-Rotem, 2016).

5British Roman Catholics attend church more often than their Protestant counterparts, with 45% of Roman Catholics going at least once a month compared to 18% of Anglicans (Lee, 2012). Less than a fifth of French Roman Catholics attend church on a monthly basis, which is lower than other European Catholic countries (Pew Research Center, 2013). The low rate of church attendance in France is likely due to the historical conflict between the Catholic Church and the secular state, which dates to the French Revolution and ultimately led to the 1905 constitutional separation of church and state (Hubert, 2015). In Britain, on the other hand, religious practice among the Catholic community is partly shaped by immigration flows from Ireland, Italy, and Poland, countries where religious participation remains relatively high (Greeley, 2003; Weller, 2007).

6Although only a minority of people in Britain and in France regularly attend religious services, more than half of these populations identified as Christians in 2010 (Pew Research Center, 2015). Thus, both countries are comprised of people who state they have no religious affiliation; are affiliated with a particular denomination but do not regularly attend religious services (‘nominally religious’); and are practising religious individuals.

7The fertility patterns of both countries trend towards a decline across cohorts born from the 1930s to the 1960s. Average completed cohort fertility dropped from 2.3 to 1.9 children in Britain and from 2.5 to 2.0 children in France (Human Fertility Database, 2017). However, Britain and France differ in their distributions of fertility across different age groups. Rendall et al. (2005, 2009) found that the age distribution at first birth among women born in the 1950s and 1960s is much more homogeneous in France than in Britain, as the latter is more polarized by social strata. This pattern results from highly educated women postponing birth in Britain compared to their less educated peers. These differences are attributed to each country’s distinct family-policy regimes. France is characterized by a universalistic family policy with a relatively generous child-care provision, whereas family benefits in Britain are means-tested. Consequently, British women with higher earnings prospects are more incentivized to delay motherhood until later ages, as the opportunity costs of children are higher (Rendall et al., 2009). Furthermore, each country differs in their childbearing norms. For example, Merz and Liefbroer (2012, p. 594) used comparative data from the 2006 European Social Survey to show that around 30% of French respondents expressed disapproval of voluntary childlessness compared to less than 10% of British respondents.

I – Religion, education, and fertility outcomes

8The relationship between religion and education is a subject of considerable debate. The standard theory of secularization proposes that extended mass education and advances in science promote a rational worldview and therefore undermine beliefs in God and the metaphysical (Berger, 1967; Bruce, 2002). Nevertheless, the empirical evidence on the relationship between religiosity and education is diverse. For example, Greeley (2003) found that unaffiliated individuals in Europe are more likely to be highly educated than religiously affiliated people. However, this negative correlation is restricted to those born between the 1920s and 1950s, whereas the relationship is reversed among more recent birth cohorts. Similarly, Hubert’s (2015) study on 1930s to 1960s French and other European cohorts concludes that religiously affiliated people have benefitted more than the unaffiliated from educational expansion. Moreover, she found that, among the religiously affiliated, those who frequently attend religious services tend to be better educated than rare attendees. Thus, religious adherence is not necessarily incompatible with the pursuit of higher education; rather, it is the religious and cultural context that influences this relationship.

9The interaction between religion and education may influence fertility outcomes. Previous studies have pointed to variations in education’s effect on fertility by level of religiosity, such as Newman’s and Hugo’s (2006) study of Australia, which found that fertility is more likely to drop with education among unaffiliated women when compared to their religiously affiliated peers. These relationships could be explained by the high value that religious traditions ascribe to family and children while also emphasising that traditional family roles be fulfilled (McQuillan, 2004; Norris and Inglehart, 2004; Chatters and Taylor, 2005). Thus, while increased levels of education are linked with greater emphasis on personal endeavours that may compete with familial roles, this effect may be weaker among women who are more religious.

10Alternatively, more religiously active women may perceive the opportunity costs of children as lower when compared to less religious women. For example, the costs and benefits of a large family may be influenced by the religious context, as some religious communities provide psychological and social rewards in the form of approval and higher status to those who conform to prescribed family norms (Lehrer, 2004). Newman and Hugo (2006) argued that Australian women raised in Christian religious families were often encouraged to have large families and, at the same time, pursue an education and a career. Thus, they perceive less conflict in combining family and work than do women with no religion. Other studies show that, compared to non-religious women, religiously observant women are more likely to adopt strategies for reconciling their family and work responsibilities, such as taking jobs with flexible and reduced hours (Edgell, 2006; Glass and Nath, 2006). Similarly, a study on religious differences in the labour supply of British and French women born in 1955–1975 found that practising religious women were more likely than unaffiliated women to reduce their working hours during the most common childbearing ages (25–35). However, they found no differences in overall employment rates by religiosity (Peri-Rotem, 2015).

11Other scholars highlight that religious belief and practice contribute to the development of coping strategies for dealing with new and stressful situations as well as the difficulties of daily life, such as those associated with work and family responsibilities (Pargament et al., 2000; Chatters and Taylor, 2005). In particular, members of religious social networks mutually exchange emotional and practical support (Putnam, 2000; Krause et al., 2001; Waite and Lehrer, 2003). For women with strong religious affiliations, this support may further reduce the direct and indirect costs of starting a family or having another child.

12From this perspective, although higher education is generally associated with higher rates of childlessness, highly religious women are demonstrably more inclined to start a family even after having obtained higher qualifications. Therefore, our first hypothesis postulates that the negative educational gradient in entry into motherhood is more pronounced for unaffiliated women than for nominally or practising religious women.

13Because affiliation with a religious community and attendance at religious services increase social capital, women who are more religious and seeking to expand their family may receive better support, be it emotional, material, or both. Among religious communities, the greater value and social incentives involved in having a large family may also influence the perceived costs of childbearing. More highly religious women would therefore be less likely to reduce their family size even if they have achieved higher education. Thus, our second hypothesis is that the completed fertility gradient by education is more strongly negative among unaffiliated women compared to their nominally and practising religious counterparts.

14Figure 1 illustrates the expected interrelationships between education, religiosity, and fertility. Education leads to lower fertility, whereas religiosity increases fertility levels. However, we assume that education’s effect on fertility depends on religiosity, i.e. the negative relationship between education and fertility is expected to be weaker for women who are more religious. In addition, the relationship between religiosity and education is assumed to be negative in general.

Figure 1

Expected interrelationships between education, religiosity, and fertility

Expected interrelationships between education, religiosity, and fertility

II – Data and methods

III – Findings

26Table 1 presents the sample characteristics of women aged 40–80 at the time of the British and French surveys: educational attainment, religious denomination and degree of involvement, proportion never-married, and proportion of foreign-born. Each category includes the number of children ever born and the proportion of women who have ever had children.

27Women in Britain show generally higher levels of education than those in France, although these differences have diminished among women of more recent cohorts (not shown). In both countries, education is negatively correlated with the total number of children. Similarly, the likelihood of entering motherhood decreases with education in both countries, but the gap by level of education is slightly greater in Britain.

28The proportion of religiously unaffiliated women is higher in Britain than in France (perhaps partly due to the different wording of the survey question), while the overall proportion of practising religious women is similar in both countries. As expected, women reporting no religious affiliation have the lowest completed fertility (1.8 children on average in Britain and 1.9 in France), whereas practising Catholic women have the highest completed fertility at 2.5 and 2.4 children on average in, respectively, Britain and France. Nominally affiliated women in both countries also have higher fertility compared to unaffiliated women. Consistent with the findings on completed fertility, unaffiliated women are most likely to remain childless. Just over 80% of unaffiliated women in Britain and France have ever had children, while the proportion of mothers among religiously affiliated women is close to 90%.

Table 1

Sample characteristics of women aged 40–80 (using country-specific population weights)

A. Britain % (n) Number of children per woman % ever had children Birth cohort 1928–1935 13 (402) 2.2 81 1936–1945 21 (657) 2.1 87 1946–1955 27 (833) 2.0 86 1956–1968 39 (1,185) 1.9 84 Education Lower secondary 30 (926) 2.4 90 Upper secondary 39 (1,190) 1.9 85 Tertiary 31 (960) 1.8 80 Religious denomination No religion 33 (1,023) 1.8 81 Nominal Protestant 43 (1,330) 2.0 86 Practising Protestant 12 (360) 2.1 87 Nominal Catholic 7 (210) 2.1 86 Practising Catholic 5 (155) 2.5 92 Never married 9 (276) 0.7 40 Foreign-born 4 (123) 2.0 83 Total 100 (3,077) 2.0 85 Source: BHPS (2008). B. France % (n) Number of children per woman % ever had children Birth cohort 1925–1935 17 (524) 2.5 85 1936–1945 21 (655) 2.2 90 1946–1955 30 (906) 2.0 89 1956–1965 32 (966) 2.0 88 Education Lower secondary 47 (1,423) 2.4 91 Upper secondary 34 (1,035) 2.0 88 Tertiary 19 (593) 1.9 84 Religious denomination No religion 6 (190) 1.9 82 Nominal Catholic 77 (2,347) 2.1 89 Practising Catholic 17 (514) 2.4 87 Never married 16 (474) 1.3 60 Foreign-born 9 (280) 2.4 93 Total 100 (3,051) 2.1 88

Sample characteristics of women aged 40–80 (using country-specific population weights)

29The proportion of never-married women is higher in France than in Britain (16% vs. 9%). Moreover, the proportion of never-married women with children is considerably higher in France (61% vs. 41%). This can be attributed to the higher prevalence of non-marital cohabitation and childbearing outside marriage in France, compared to the more conservative family-formation patterns in Britain (Perelli-Harris et al., 2009).

30Finally, the French sample is characterized by a higher proportion of foreign-born individuals (9% vs. 4% in Britain). The fertility levels of foreign-born women in France are also higher than the average family size in this country (2.4 children vs. 2.1); those of foreign-born women in Britain are similar to the average (2.0). This can be attributed to differences in immigration trends and to the composition of migrant populations in each country (see Kaufmann et al., 2012).

31Figure 2 reveals the distribution of educational attainment by religious group in Britain and France. The findings show no consistent pattern in the relationship between religion and education. The greatest proportion of highly educated women in Britain is found among practising religious women (46% and 39% among practising Protestant and practising Catholic women, respectively). In France, on the other hand, the greatest proportion of highly educated women is found among those with no religious affiliation (26% vs. 19% and 20% among, respectively, nominally and practising Catholic women).

Figure 2

Educational attainment by religious group (in %)

Educational attainment by religious group (in %)

1 – The odds of having children by education and religious group

32Table 2 presents the results of the logistic regression model for the odds of becoming a mother. Model 1 estimates the odds of entering motherhood as a function of combined religious affiliation and practice, woman’s education, birth cohort, and nativity. In Model 2, an interaction between religion and education is introduced. As the interrelationships between education, religion, and fertility may be partly mediated by marital status, the control variable for never-married women is added to Models 3 and 4. The findings show that in both countries, women with upper secondary or tertiary education are less likely to become mothers compared to women with a lower level of education. However, this relationship is more robust in Britain than in France. In terms of religion, unaffiliated women are less likely to become mothers compared to nominally and practising religious women in both countries (although this odds ratio is not significantly different for nominally Catholic—relative to unaffiliated—women in Britain).

Table 2

Logistic regression for motherhood among women aged 40–80 at the time of the surveys (odds ratios)

A. Britain Model 1 Model 2 Model 3 Model 4 Birth cohort 1928–1935 0.516*** 0.524*** 0.357*** 0.359*** 1936–1945 0.861 0.862 0.642** 0.635** 1946–1955 0.972 0.952 0.714* 0.690** 1956–1968 Ref. Ref. Ref. Ref. Foreign-born 0.899 0.933 0.927 0.968 Never married 0.077*** 0.075*** Religious denomination No religion Ref. Ref. Ref. Ref. Nominal Protestant 1.553*** 0.900 1.383** 0.665 Practising Protestant 2.039*** 1.841 1.754** 1.461 Nominal Catholic 1.504 0.847 1.352 0.686 Practising Catholic 3.273*** 0.754 2.350** 0.499 Education Lower secondary Ref. Ref. Ref. Ref. Upper secondary 0.571*** 0.350*** 0.595*** 0.314*** Tertiary 0.393*** 0.254*** 0.415*** 0.227*** Religious denomination × Education Nominal Protestant × Upper secondary 2.152** 2.655** Nominal Protestant × Tertiary 1.830 2.485** Practising Protestant × Upper secondary 1.093 1.402 Practising Protestant × Tertiary 1.188 1.213 Nominal Catholic × Upper secondary 1.869 2.080 Nominal Catholic × Tertiary 2.263 2.709 Practising Catholic × Upper secondary 12.236*** 12.365** Practising Catholic × Tertiary 6.245** 7.539*** N 3,077 3,077 3,077 3,077 Log pseudo-likelihood −1734.7 −1724.7 −1515.6 −1503.7

Logistic regression for motherhood among women aged 40–80 at the time of the surveys (odds ratios)

B. France Model 1 Model 2 Model 3 Model 4 Birth cohort 1928–1935 0.517*** 0.524*** 0.376*** 0.376*** 1936–1945 0.904 0.909 0.549*** 0.547*** 1946–1955 0.966 0.972 0.623*** 0.624*** 1956–1965 Ref. Ref. Ref. Ref. Foreign-born 1.745** 1.754** 1.569* 1.567 Never married 0.097*** 0.097*** Religious denomination No religion Ref. Ref. Ref. Ref. Nominal Catholic 1.803*** 1.784 1.627** 1.965* Practising Catholic 1.642** 1.576 1.591* 2.013 Education Lower secondary Ref. Ref. Ref. Ref. Upper secondary 0.654*** 0.722 0.762* 1.052 Tertiary 0.445*** 0.392* 0.553*** 0.719 Religious denomination × Education Nominal Catholic × Upper secondary 0.922 0.734 Nominal Catholic × Tertiary 1.093 0.738 Practising Catholic × Upper secondary 0.804 0.590 Practising Catholic × Tertiary 1.487 0.816 N 3,051 3,051 3,051 3,051 Log pseudo-likelihood −954.5 −953.9 −812.8 −812.3

* p < .10, ** p < .05, *** p < .01.

33Never-married women are significantly less likely to become mothers compared to ever-married women in both countries (Table 2, Models 3 and 4). Furthermore, when adding the covariate for never-married women, the magnitude of the coefficients for the religious groups in Britain and France becomes slightly less large, which may indicate that some of the relationship between religion and the transition to parenthood is explained by differential marriage patterns among affiliated and unaffiliated women.

34An interaction effect between religion and education on the odds of having children is found only in Britain. Compared to unaffiliated women with lower secondary education, there is a significant positive interaction for nominally Protestant women with upper secondary and tertiary education and for practising Catholic women in the same educational groups (Table 2A, Model 4). No significant interaction is found for practising Protestant and nominally Catholic women in Britain. However, when Protestant and Catholic denominations are merged, the interaction between religion and education is significant for all nominal and practising women (see Appendix Table A.2).

35To better understand how religion and education interact with entry into motherhood, Figure 3 illustrates the predicted probabilities for motherhood by religious group and education in Britain and France. Among unaffiliated women in Britain, those with upper secondary (medium) and tertiary (high) education are significantly less likely to have children compared to less- educated women within that group. By contrast, the differences in the predicted probabilities of motherhood by educational attainment are less significant among nominally and practising religious women. In France, on the other hand, no significant differences by education are found among unaffiliated women. Therefore, our first hypothesis is not supported in France, but it does hold in Britain: a stronger negative relationship exists between education and childlessness among religiously unaffiliated women compared to other religious groups.

Figure 3

Average predicted probabilities for motherhood by religious group and level of education (women aged 40–80 at the time of the surveys)

Average predicted probabilities for motherhood by religious group and level of education (women aged 40–80 at the time of the surveys)

2 – Completed fertility gradient by educational attainment among religious groups

36In this section, a Poisson regression model is used to predict the number of children ever born to women aged 40–80 at the time of the surveys in each country. The results of the model are presented in Table 3. The findings from Britain show that both nominally and practising religious women have a significantly higher number of children compared to unaffiliated women (Table 3A, Models 1 and 3). In France, on the other hand, only practising Catholic women show significantly higher completed fertility compared to unaffiliated women (Table 3B, Models 1 and 3). In accordance with the descriptive findings, education is negatively associated with the number of children in both countries, while never-married women have significantly lower completed fertility than ever-married women, although this relationship is somewhat more pronounced in Britain than in France.

37When including the interaction term for education and religion, a significant positive interaction is found in Britain for nominally Protestant women with upper secondary and tertiary education, as well as for practising Protestant women with tertiary education when compared to the reference category of unaffiliated less-educated women (Table 3A, Models 2 and 4). When modelling all nominally and practising religious women in Britain together, we find a significant interaction for nominally religious women with upper secondary (medium) and tertiary (high) levels of education and for practising women with higher education (see Appendix Table A.3).

38In France, a significant positive interaction is found for both nominally and practising Catholic women with upper secondary and tertiary education in relation to the reference group (Table 3B, Models 2 and 4). Therefore, consistent with our second hypothesis, the findings from Britain and France point to a significant variation in the relationship between education and completed fertility across religious groups. Moreover, the interaction between religion and education in Britain and France remains significant when foreign-born women are excluded from the analysis (not shown).

Table 3

Poisson regression for number of children ever born to women aged 40–80 at the time of the surveys (incidence rate ratios)

A. Britain Model 1 Model 2 Model 3 Model 4 Birth cohort 1928–1935 1.041 1.049 1.000 1.006 1936–1945 1.011 1.014 0.974 0.976 1946–1955 1.002 0.997 0.961 0.957 1956–1968 Ref. Ref. Ref. Ref. Foreign-born 0.972 0.982 0.979 0.990 Never married 0.346*** 0.347*** Religious denomination No religion Ref. Ref. Ref. Ref. Nominal Protestant 1.106*** 0.993 1.068** 0.957 Practising Protestant 1.159*** 0.989 1.107** 0.954 Nominal Catholic 1.173** 1.123 1.140** 1.100 Practising Catholic 1.376*** 1.245 1.279*** 1.182 Education Lower secondary Ref. Ref. Ref. Ref. Upper secondary 0.796*** 0.740*** 0.810*** 0.753*** Tertiary 0.774*** 0.658*** 0.798*** 0.680*** Religious denomination × Education Nominal Protestant × Upper secondary 1.160* 1.159** Nominal Protestant × Tertiary 1.209** 1.222** Practising Protestant × Upper secondary 1.076 1.090 Practising Protestant × Tertiary 1.414*** 1.371** Nominal Catholic × Upper secondary 0.955 0.945 Nominal Catholic × Tertiary 1.240 1.222 Practising Catholic × Upper secondary 1.054 1.017 Practising Catholic × Tertiary 1.280 1.243 Constant 2.124*** 2.286*** 2.338*** 2.511*** N 3,077 3,077 3,077 3,077 Log pseudo-likelihood −6,889.6 −6,876.1 −6,690.8 −6,678.9

Poisson regression for number of children ever born to women aged 40–80 at the time of the surveys (incidence rate ratios)

39Figure 4 illustrates the adjusted predicted means of children ever born by religious group and level of education. In Britain (Figure 4A), the group of unaffiliated women show a linear decline in completed fertility by education: from 2.3 children among the least educated women to 1.8 and 1.6 among, respectively, upper secondary and tertiary educated women. A negative fertility gradient is also found for nominally Protestant women, although the decline is less steep: from 2.2 children on average among lower secondary educated to 1.9 and 1.8 children among women with upper secondary and tertiary education. By contrast, a U-shaped fertility curve exists for practising Protestants and nominally and practising Catholics, as the least educated have the highest completed fertility, followed by a decline among those with upper secondary education and then a rise among those with tertiary education. Furthermore, nominally and practising religious women with tertiary education have significantly higher completed fertility than do unaffiliated women with the same level of education.

B. France Model 1 Model 2 Model 3 Model 4 Birth cohort 1928–1935 1.127*** 1.141*** 1.097** 1.110** 1936–1945 1.059 1.069* 1.001 1.010 1946–1955 0.975 0.980 0.927** 0.932** 1956–1965 Ref. Ref. Ref. Ref. Foreign-born 1.118** 1.132** 1.100* 1.112** Never married 0.557*** 0.562*** Religious denomination No religion Ref. Ref. Ref. Ref. Nominal Catholic 1.086 0.919 1.058 0.919 Practising Catholic 1.182** 0.937 1.162** 0.954 Education Lower secondary Ref. Ref. Ref. Ref. Upper secondary 0.853*** 0.656*** 0.875*** 0.690*** Tertiary 0.842*** 0.534*** 0.879*** 0.603*** Religious denomination × Education Nominal Catholic × Upper secondary 1.318** 1.283* Nominal Catholic × Tertiary 1.524*** 1.410*** Practising Catholic × Upper secondary 1.337* 1.297* Practising Catholic × Tertiary 2.078*** 1.852*** Constant 2.040*** 2.398*** 2.280*** 2.608*** N 3,051 3,051 3,051 3,051 Log pseudo-likelihood −4,623.9 −4,611.7 −4,529.7 −4,520.7

* p < .10, ** p < .05, *** p < .01.

40In France, there is also evidence of a U-shaped fertility gradient by education among practising Catholic women (Figure 4B), with highly educated women in this group showing the largest family size: 2.6 children on average compared to 2.3 children for the least educated, and 2.1 children for those with upper secondary (medium) education. On the other hand, fertility declines with education among nominally Catholic and unaffiliated women, although the educational gradient of completed fertility is much steeper for the unaffiliated: from 2.5 children at the lowest level of education to 1.7 at the upper secondary level, and 1.5 at tertiary education. The corresponding predicted means for nominally Catholic women are: 2.3 among the less educated, then it declines from 2.0 to 1.9 children among moderately and highly educated women. Thus, as in Britain, the differences in completed fertility by religion are particularly pronounced at a high level of education.

41Overall, these findings support our second hypothesis, according to which the educational gradient of completed fertility is more strongly negative for unaffiliated women than for those who are more religious. The steeper decline in fertility with education among unaffiliated women may reflect disparities in the perceived cost and benefits of children for women in each religious group.

Figure 4

Predicted means of children ever born by religious group and level of education (women aged 40–80 at the time of the surveys)

Predicted means of children ever born by religious group and level of education (women aged 40–80 at the time of the surveys)

Notes: Means adjusted for birth cohort, nativity, and ever-married. Bars represent 90% confidence intervals.

Conclusion

42This study examines the interrelationships between religion, education, and fertility outcomes in Britain and France, focusing on women who have completed (or nearly completed) their reproductive years. First, we explored the distribution of educational attainment by religion, using a combined measure of religious affiliation and practice. Then, different types of regression analyses were employed to estimate the odds of having children and the number of children ever born as a function of education and the combined measure for religion. The descriptive findings show no consistent relationship between religion and educational attainment in Britain and France for women born from the 1920s to the 1960s. However, we do show that, at the individual level, greater religious involvement can go hand in hand with having obtained higher education, as the proportion of highly educated women among the practising religious group is either similar (in France) or higher (in Britain) than the average proportion. Furthermore, by analysing the interaction between education and religion in relation to childbearing patterns, educational attainment is found to result in differential fertility outcomes among women with varying levels of religious adherence.

43The first hypothesis, which predicted a stronger association between education and childlessness among unaffiliated women compared to nominally and practising religious women, is supported by the findings in Britain but not in France. This may be due to the latter having smaller differences in childlessness rates by educational attainment. For example, bearing in mind both the relatively generous child-care provision in France (Rendall et al., 2009) and the greater disapproval of voluntary childlessness (Merz and Liefbroer, 2012), one may reasonably conclude that these contribute to the smaller variation in the proportion of women entering parenthood by level of education. In addition, women identified as religiously unaffiliated in France may differ from those in Britain due to differences in the survey questions on religious affiliation.

44Nevertheless, when examining the interrelationships between education, religion, and completed fertility, we find a significant interaction between religion and education in both countries. Thus, consistent with the second hypothesis, nominally and practising religious women are less likely than unaffiliated women to reduce their family size when obtaining higher qualifications. The differences in the educational gradient in completed fertility are most pronounced for practising religious women, of whom the lowest and highest educated are found to have the largest family sizes. These patterns can be explained by variations in the perceived costs and value of children among different religious groups. As described above, religious involvement may affect the costs and benefits of a large family, through social rewards or psychological and practical support (Lehrer, 2004). In particular, women who regularly attend religious services are more likely to benefit from different types of assistance (Krause et al., 2001; Waite and Lehrer 2003; Chatters and Taylor, 2005). Because, on the one hand, family life and children are highly valued by religious communities and, on the other, women who are more religious receive more support when seeking to expand their families, they perceive both the direct and indirect costs of children as lower than do other social groups. These combined high-fertility aspirations and improved social and financial resources could account for the U-shaped relationship between education and completed fertility among practising religious women. Thus, not only is there no contradiction between religious adherence and higher education, increased human capital may enable higher fertility among the more religious groups.

45Despite the unique characteristics of the British and French societies, higher religiosity is linked with higher fertility in both; and it appears to attenuate the negative association between education and completed family size. These findings shed some light on the importance of religion in understanding variations in the relationship between education and fertility. As this study shows, the relationship between education and fertility outcomes is not uniform across religious groups, and in some cases higher education may lead to higher fertility, as we find among practising Catholic women in France. For this group, completed fertility is highest among the most educated women, which may be the result of placing a high value on children while improving the resources for having them and reducing their costs. Future studies should consider these complexities when analysing the effect of religion on demographic and socioeconomic patterns.