Measurement of the Generalized Religiosity in the World Values Survey: Issues of Cross-Country Comparability

Religiosity is widely used by comparative researchers as dependent or explanatory variable. However, the invariance of its measurement has been rarely assessed. In the current research, the invariance of the measurement of generalized religiosity in the 6th round of the World Values Survey was examined. It was evaluated using confirmatory and multi-group confirmatory factor analysis, alignment, and multi-level structural equation modeling methods. Results demonstrated a different degree of invariance of religious measures and a range of systematic dissimilarities in the expressions of religiosity across countries. The 'confidence in religious institutions', 'belief in god', 'frequency of praying', and 'self-assessed level of religiosity' items were found to be perceived differently by populations of Abrahamic and non-Abrahamic religions. The 'frequency of praying' indicator was a non-reliable indicator for the comparison of Muslim and European Christian countries. The 'belonging to religious denomination' variable functioned noninvariantly across countries differentiated by the level of religious diversity. The cross-national differences in the 'membership in religious organizations' item functioning were accounted for by variations in the level of government restrictions of religion. The 'frequency of attendance of religious services' and 'membership in religious organizations' indicators were noninvariant in post-Soviet countries. Cultural dissimilarities were found for almost all religious measures. We concluded that indicators should be used on homogeneous samples relative to the corresponding contextual factors that we detected. The 'belonging to religious denomination', 'belief in god', 'membership in religious organizations', and 'importance of god' items should be used for cross-national analysis only of Western countries. Other measures demonstrated higher potential to be applied for comparisons of countries differentiated by their religious, social, and cultural characteristics. Finally, we proposed four measurement models for valid cross-national analysis of generalized religiosity on big and heterogeneous samples, with estimates produced with alignment optimization.