When people shed religious identity in Ireland and Austria: Evidence from censuses.

BACKGROUND: Disaffiliation from religion is an important factor behind the rapid rise in persons claiming no religious affiliation in many advanced industrial countries. Scholars typically think of disaffiliation as a life course process that is confined to young adults, with little change occurring among older adults, yet few studies have examined this assumption outside the United States and Great Britain.
OBJECTIVE: We evaluate whether the young-adult model of disaffiliation from religion applies in Ireland and Austria, two historically Catholic-majority countries with different levels of non-affiliation growth.
METHODS: We use census data on religious affiliation in Ireland (1971-2011) and Austria (1971-2001) to track aggregate changes in the percentage reporting no religious affiliation over the life course for successive birth cohorts.
RESULTS: We find support for the young-adult model in Ireland. However, recent cohorts in Austria exhibit a distinct pattern of disaffiliation that continues into middle adulthood. Our analysis suggests that mid-life disaffiliation in Austria is connected to a religious tax, which we argue spurs nominally affiliated adults to disaffiliate themselves, as their income rises and the costs of religious affiliation increase.
CONCLUSION: Our findings offer insight into some of the social factors behind recent religious change across Europe and highlight the need for more cross-national research on the age and cohort dimensions of this change.

1. Introduction

The rising number of people claiming no religious affiliation across Europe and the United States represents one of the most significant social changes over the past century, with implications upon the role of religious institutions in civil society, as well as on family and fertility patterns (Chaves 1994; Lesthaeghe 2010). While unaffiliated populations could grow through a combination of high fertility, low mortality, and migration (Hout, Greeley, and Wilde 2001; Kaufmann, Goujon, and Skirbekk 2012; Skirbekk, Kaufmann, and Goujon 2010), disaffiliation – making the change to no religious affiliation – seems to be an important driver of growth in non-affiliation in many advanced industrial countries. Most scholars see disaffiliation, and religious allegiance-changes generally, as a life-course process confined to young adulthood (Roof and McKinney 1987; Schwadel 2010; Wuthnow 1976). This life stage is a tumultuous time during which young adults are moving out of the parental household, attending school, forming families, and developing political and religious identities (Arnett 2004; Gooren 2007). Religious participation declines during this stage, but often returns following marriage and childbearing; in contrast, changes in affiliation are less common but thought to be more permanent (Stolzenberg, Blair-Loy, and Waite 1995; Uecker, Regnerus, and Vaaler 2007).

Is the life course model of religious disaffiliation, with change concentrated in young adulthood and stability in older ages, a widely shared social phenomenon in countries experiencing non-affiliation growth? Evidence from the United States, Great Britain, and the Netherlands generally supports the life course model, with successive cohorts of non-affiliated young adults leading to overall changes in each country’s religious composition (Hout and Fischer 2002; Putnam and Campbell 2010; Need and DeGraff 1996; Te Grotenhuis and Scheepers 2001; Crockett and Voas 2006; Voas 2003). Cross-national studies also find substantial variation in levels of religious affiliation change (Barro, McCleary, and Hwang 2010; Brañas-Garza, García-Muñoz, and Neuman 2009), but few have examined the age-related dimensions of these changes (for exception see Goujon et al. 2007).

We examine life course patterns of disaffiliation, in the aggregate, across multiple cohorts in Austria and Ireland from the 1970s to the 2000s. These countries provide interesting comparative cases because, in addition to being understudied, both are historically Catholic-majority countries that have experienced markedly different levels of religious change. Both countries have also asked about religious affiliation in their censuses, which allows us to investigate the timing of disaffiliation for multiple cohorts. Our approach is unique among studies of religious change, which have relied primarily on panel data (e.g., Crockett and Voas 2006) or retrospective survey reports (e.g., Hout and Fischer 2002). We close with a brief discussion of the implications of our findings for future patterns of religious change.

2. Data and methods

We used IPUMS census micro-data from Ireland in 1971, 1981, 1991, 2002, and 2011 and Austria in 1971, 1981, 1991, and 2001 (Minnesota Population Center). Austria switched to a register-based census in 2011, which prohibited us from extending the period of observation to 2011. Both countries asked respondents to report their religious affiliation, including an option for no affiliation. Importantly, information on religious affiliation was requested in a similar manner across census years. Since the censuses do not collect data on religious belief, practice, or salience, our analysis is limited to reports of religious affiliation and non-affiliation.

We disaggregated the religious composition of each country into ten-year birth cohorts and tracked changes in the percentage reporting no affiliation in each cohort across census years. We also tracked changes in the percent reporting any Christian affiliation. In both countries, the bulk of the shift in religious composition was due to declines in the proportion reporting Christian affiliation and increases in no religion, which we attribute to disaffiliation. Of course, individuals may disaffiliate or re-affiliate between years. Our method shows aggregate or cumulative levels of disaffiliation in each cohort over time. This allows us to observe the age-transitions in which disaffiliation was most common.

To account for changes in the religious composition due to immigration rather than disaffiliation, we restricted the samples to native-born respondents in Ireland and citizens in Austria (Austria did not ask nativity in all census years). Our method does not account for the potential effects of out-migration. If out-migrants are selected on religion, then the growth in non-affiliation that we attribute to disaffiliation could, instead, reflect the abatement of Christians from the population. However, in order to produce the levels of disaffiliation we observe, out-migrants would need to be disproportionately Christian – a circumstance that we believe is unlikely (Connor and Tucker 2010). We do not follow cohorts past age 70 because of the potential for differential mortality by religious affiliation or its correlates (e.g. socioeconomic status) (Ellison and Hummer 2010). Finally, we do not show trends separately for men and women. While ancillary analyses by gender revealed that men disaffiliated themselves at higher rates (Hayes 1996; Collett and Lizardo 2009), the age patterns were similar for men and for women.

We start by presenting overall change in the religious composition of both countries during the period. We then disaggregate by birth cohort to identify the age patterns of disaffiliation across cohorts. We follow with a brief discussion of the unusual age pattern of disaffiliation in Austria, and provide a brief empirical test of one potential explanation.

3. Results

Table 1 displays the trends in the percentage of the population claiming no religious affiliation, Christian affiliation (all Christian denominations are combined, although the majority are Catholic), and other religious affiliations in Ireland from 1971–2011, and Austria from 1971–2001. Both countries maintained Christian majorities during this period, but, nevertheless, experienced growth in the non-affiliated population over time. Ireland had a larger percentage of Christians in 1971 and showed relatively small increases in non-affiliation, growing from 0.18% to 4.2% in 2011. The strength of Christianity displayed in these numbers reflects the historic presence of the Catholic Church as well as the strong connection between religious and political identities (Bruce 2002). In contrast, Austria shows a more robust pattern of disaffiliation, albeit to a lesser extent than other European countries such as Great Britain or those in Scandinavia (Crockett and Voas 2006; Bruce 2000). Compared to Ireland, Austria had a larger non-affiliated population in 1971, 4.3%, which grew steadily over the period to 12.3% in 2001. As another point of comparison, 14% of the United States population claimed no religious affiliation in 2000 (Hout and Fischer 2002).

Table 1

Percentage of population reporting Christian affiliation, no religious affiliation, and other affiliation by country and census year, 1971–2011. Restricted to native-born only in Ireland; citizens only in Austria

	Christian	No religion	Other
Ireland
1971	99.8	0.2	0.05
1981	99.0	0.9	0.05
1991	97.3	1.5	1.2
2002	96.3	2.3	1.4
2011	93.4	4.2	2.4
Austria
1971	95.2	4.3	0.5
1981	93.1	6.0	0.9
1991	89.8	8.9	1.3
2001	86.1	12.3	1.6

Figures 1 and 2 disaggregate changes in the non-affiliated population by birth cohort. Each line represents a ten-year birth cohort, each point on the line a census year (see legend). The x-axis indicates the age group of each cohort in a given census year. This method allows us to follow successive birth cohorts (e.g., born in the 1940s, 1950s, and 1960s) as they age. For example, looking at Table 1 for Ireland, we can follow the 1942–51 birth cohort from ages 20–29 in 1971 to ages 60–69 in 2011, and see how and when religious non-affiliation changes over the life course. We can then compare the experience of the 1942–51 birth cohort to that of the 1952–61 cohort (which we observe from ages 10–19 in 1971 to 50–59 in 2011) to see the ages at which religious change is most common across cohorts. While the absolute levels of non-affiliation may differ across cohorts (with younger cohorts displaying more non-affiliation), the life course model expects disaffiliation (or increases in non-affiliation) to crest in young adulthood for all birth cohorts.

Figure 1

Percent reporting no religious affiliation by birth cohort and age in Ireland, 1971–2011. Restricted to native-born only

Source: IPUMS census micro-data

Note: This figure depicts the growth of non-affiliation, and the ages at which disaffiliation occurred, for multiple birth cohorts (represented by each line). For example, it shows that disaffiliation among the 1952–61 and 1962–71 cohorts was concentrated during the transition from ages 10–19 and 20–29, in support of the young-adult model of disaffiliation. The markers denote the census year of observation for each cohort.

Figure 2

Percentage reporting no religious affiliation by birth cohort and age in Austria, 1971–2001. Restricted to citizens only

Source: IPUMS census micro-data

Note: This figure depicts the growth of non-affiliation, and the ages at which disaffiliation occurred, for multiple birth cohorts (represented by each line). For example, it shows that disaffiliation among the 1940–49 and 1950–59 cohorts occurred not only in young adulthood, but also at later ages in middle adulthood. The markers denote the census year of observation for each cohort.

Figure 1 shows cohort trends for Ireland. We see strong support for the life course model of religious disaffiliation. Across successive cohorts, religious disaffiliation is most common during the transition to adulthood. There is little aggregate change in the percent reporting no religious affiliation beyond ages 30–39. Even for the 1942–1961 cohorts (the oldest cohorts to display moderate levels of religious change) disaffiliation is concentrated in the adolescent and young-adult years. This life course pattern appears to be maintained among more recent cohorts born after 1961. It is also the driving force behind a generational decline in Christian affiliation in Ireland. Starting with the 1942–51 cohort, subsequent cohorts have disaffiliated themselves at higher rates during young adulthood, leading to an overall increase in religious non-affiliation in the population during this period. The 2011 census shows evidence of a period effect, with all cohorts born after 1950 showing moderate increases in non-affiliation. While growth remains more pronounced among young adults, this may signal a shift in the age-pattern of religious disaffiliation in the future.

Figure 2 displays cohort trends in disaffiliation for Austria. The age pattern of disaffiliation in Austria is strikingly different from that in Ireland. Unlike cohorts in Ireland, older Austrian birth cohorts already had sizeable percentages reporting no religious affiliation in 1971. The 1910–19 cohort, for example, had the largest proportion of non-affiliation in 1971 at nearly 9%. However, by 2001, younger cohorts were disaffiliating at much higher rates. Yet the ages at which disaffiliation took place among these younger cohorts represent a significant departure from the life-course model. While young adulthood is still characterized by substantial levels of disaffiliation, this trend extends beyond the young-adult years, well into mid and late life. The 1940–49 cohort stands out in this respect: by ages 52–61 in 2001, this cohort had nearly 20% reporting no affiliation, an increase from just 3% at ages 22–31 in 1971. In contrast, the previous cohort, born 1930–39, reported only 12.5% non-affiliated when it was in its 50s. The cohorts following the 1940–49 cohort appear to be on the same path of high rates of disaffiliation in young adulthood followed by continued disaffiliation well into mid-life. In fact, in 2001, the 1940–49 cohort (now in their 50s and 60s) still displayed the highest levels of non-affiliation.

One possible explanation for mid-life disaffiliation in Austria is a period effect towards secularization felt across a number of European countries. As “no religion” became a more socially accepted identity, nominal affiliates of all ages may have been less inclined to hold on to a Christian identity. A series of scandals involving the Catholic Church in Austria during the 1990s may also have contributed to mid-life disaffiliation during later periods. However, another possibility that is more closely connected to the life course involves the church tax, called Kirchenbeitrag (ecclesiastical contribution) (Goujon et al. 2007). Instituted under Nazi rule to fund Catholic and Protestant churches, Kirchenbeitrag is a mandatory tax required of all Austrian Christians that amounts to 1% of an individual’s income. We propose that because of this tax, Austrians with a nominal attachment to their Christian identity have greater economic incentive to disaffiliate with religion as they enter their prime working years.

We tested this explanation using cross-sectional data from the 2008 Generations and Gender Survey (GGS), a household survey of Austrian adults ages 18–45. We restrict our analysis to ages 25 and older. This segment of the population has largely completed schooling and entered the labor market. Ideally, a larger age range (beyond age 45) would be preferable, but the GGS is the only household survey of Austria (of which we are aware) that asks about religious affiliation in a way that is comparable to the census (for more detail see Hackett 2014). We hypothesized that income would be positively associated with reporting no religious affiliation, across all age groups. Table 2 displays the percent reporting no religious affiliation by monthly earned-income (in Euros). The first three columns show the relationship between income and religious non-affiliation for the entire age range (25–45), younger respondents (25–29), and older respondents (40–45). As expected, we see a monotonic increase in the likelihood of reporting no religious affiliation as we move from low to high income. The final three columns display predicted probabilities from multivariate logistic regression models that control for gender, marital status, children living in the household, educational attainment, and urban residence. All covariates are set at mean levels. The predicted probabilities confirm a statistically significant positive association between income and religious non-affiliation. We also tested whether the effect of income varied across age groups in models that interacted income and age (not shown). These models indicated that the association between income and religious non-affiliation was the same across all age groups. Since older age groups have higher average incomes, this supports our contention that the changing income distribution, rather than aging per se, is behind higher levels of non-affiliation at older ages. Together, this provides suggestive evidence for the role of Austria’s church tax in mid-life disaffiliation observed among recent cohorts.

Table 2

Percent reporting no religious affiliation by monthly income and age in Austria, before and after controls. Restricted to citizens only

	Observed, no controlsUnaffiliated (%)			Predicted probabilities,with controlsUnaffiliated (%)
Monthly earned income (Euros)	25–45	25–29	40–45	25–45	25–29	40–45
<1000	11	15	10	11	9	11
1000–1999	18	16	17	15	13	16
2000+	22	21	23	18	15	19
Total	17	16	18	14	12	16

Source: Generations and Gender Survey, 2008.

Note: Predicted probabilities from multivariate regression models control for gender, marital status, children living in the household, educational attainment, and urban residence. All covariates are set at mean levels.

4. Conclusion

We analyzed age patterns of disaffiliation across multiple birth cohorts in Ireland and Austria using census data. Our objective was to see whether the life course model of disaffiliation held in these contexts. Prior research on the United States, Great Britain, and the Netherlands emphasizes young adulthood as a critical period of religious change and anticipates little change later in adulthood. We found more support for the life course model in Ireland. Although overall levels of non-affiliation were low, disaffiliation was concentrated in the young-adult years and was followed by aggregate stabilization of religious affiliation later in adulthood. There is evidence that this could be changing, as some Irish cohorts continue to disaffiliate at moderate rates later in life in the most recent period. We observed a different pattern in Austria that deviated much more substantially from the life course model. While young adulthood was an important period, recent cohorts in Austria have continued to disaffiliate – and remain unaffiliated – well into middle adulthood.

We tested one potential explanation for this distinctive pattern of mid-life disaffiliation involving the church tax. We expected that the church tax would provide incentive for disaffiliation among older, higher-earning adults with nominal religious identities. Consistent with this explanation, we found that reporting no religious affiliation was positively associated with earned income in Austria. Alternatively, income could be associated with non-affiliation because higher income individuals have competing demands on their time (e.g., longer work hours) or more opportunities for leisure activities that interfere with church attendance and that erode religious identity. Income may also increase existential security, leading to the adoption of post-materialist values and secular identity (Inglehart 1990). This age-related income dynamic is likely to complement other period influences – such as a general trend towards secularization and church scandals in the 1990s – to shape Austria’s distinctive pattern of mid-life disaffiliation.

In conclusion, individual change in religious affiliation is not necessarily concentrated in the young adult years, as usually posited in life course analysis. In Austria, disaffiliation continues throughout the adult life cycle, possibly due to a tax on religious affiliation. Furthermore, the Ireland and Austria examples demonstrate additional mechanisms through which non-affiliation can increase at the country level. In each country, non-affiliation has risen through cohort replacement, as new cohorts, aging into adulthood, identified themselves as non-affiliated at higher levels than previous cohorts at the same age. In addition, there are hints of a period effect in which various cohorts have experienced at least modest increases in aggregate non-affiliation during the same period.

In order to understand changing religious landscapes, future research should further explore the demography of cross-national variation in disaffiliation patterns and changing levels of religious non-affiliation.