Integration and religiosity among the Turkish second generation in Europe: a comparative analysis across four capital cities

Abstract

Drawing on recent cross-national surveys of the Turkish second generation, we test hypotheses of secularization and of religious vitality for Muslim minorities in Europe. Secularization predicts an inverse relationship between structural integration and religiosity, such that the Turkish second generation would be less religious with higher levels of educational attainment and intermarriage. The religious vitality hypothesis predicts the maintenance of religion in the second generation, highlighting the role of religious socialization within immigrant families and communities. Taking a comparative approach, these hypotheses are tested in the context of different national approaches to the institutionalization of Islam as a minority religion in four European capital cities: Amsterdam, Berlin, Brussels and Stockholm. Across contexts, religious socialization strongly predicts second-generation religiosity, in line with religious vitality. The secularization hypothesis finds support only among the second generation in Berlin, however, where Islam is least accommodated.

Keywords:

• Integration
• religion
• Islam
• second generation
• institutions
• comparative approach

Notes

1. Results not shown, available upon request. We compared rural or urban origin, year of migration, age at migration, migration motives and educational qualifications, as well as parents’ employment status during participants’ youth. Most parents originate from rural areas in Turkey and migrated to Western Europe in the late 1960s and early 1970s, with mean age at migration between 18 and 23 years. Most fathers migrated for economic reasons, while mothers generally migrated for family reunification or formation. Most parents had little or no formal education, with somewhat higher levels in Brussels and Stockholm. Most fathers were working in unskilled or skilled manual jobs when participants were 15 years old, while mothers were mostly homemakers; in Brussels and Stockholm, however, mothers’ levels of labour market participation were higher.

2. Because mosque visiting is gendered with higher attendance among men (cf. Diehl and Koenig 2009), a direct effect of gender on mosque attendance is included in all models.

3. The levels of education distinguished correspond to the International Standard Classification of Education (ISCED) (UNESCO 1997), with low education representing ISCED-2 and below, vocational referring to ISCED-3B, upper secondary to ISCED-3A and tertiary to ISCED-5 and higher.

4. Self-categorized Muslims are participants who indicated that they currently had a religion and that this religion was Islam. Religious affiliation was used as a routing question in computer-assisted personal interviews such that participants who indicated that they had no religion did not answer questions on religious identification and practice. The wording of the routing question deviated in Berlin where participants were asked ‘Are you currently religious?’ This wording may have resulted in a greater selectivity of participants who completed the religiosity measures in Berlin. We analysed self-categorization as Muslims using all independent variables as predictors in logistic regression. These results (not shown, available upon request) replicate similar context-dependent associations of religious membership with structural integration, religious socialization and perceived discrimination.

5. Moreover, model fit statistics suggest that the distinction between the two religious practice scales fits better to the data, compared to a model where all four indicators load on one common practice scale. Akaike's Information Criterion (AIC) is used to compare these competing models, with smaller values indicating better fit (Hu and Bentler 1999; Kline 2005). AIC is 186.316 for a model with only one religious practice scale, while a model with two correlated practice scales has an AIC of 155.780, ceteris paribus.

6. The effects of Koran lessons and parental religiosity are not significant in Stockholm, yet constraining them to zero results in significantly worse model fit (▵ χ2 (2) = 8.144, p = 0.017).

7. Table 2 also shows the loadings of the religiosity dimensions on the second-order factor after including all explanatory and control variables. Despite some changes in magnitude, the loadings still differ significantly between the four cities; constraining them to be equal results in significantly worse model fit (▵ χ2 (9) = 38.733, p<.001).