Religious Belief, Religious Denomination, and Preferences for Redistribution: A Comparison across 13 Countries

Abstract

Despite clear evidence of a link between religion and welfare state development, research on public support for redistributive social policies has been slow to incorporate religion into models of public preferences. Does religious belief influence support for redistribution? If so, does religion necessarily increase hostility to redistribution as suggested by existing research? Or, do the Catholic and Protestant traditions generate different patterns of preferences consistent with Weber’s notion of a Protestant work ethic? This research explores these questions by comparing public support for redistribution across 13 European states. The data reveal a powerful and complex effect of religious belief on support for redistribution that works on both the individual and cultural levels. The evidence challenges existing research by demonstrating that the impact of religion is not uniform, but varies across different religious traditions, with Catholicism producing significantly greater support for redistribution than Protestantism.

Notes

1. The ESS coding does not provide a clear means for distinguishing between different Protestant traditions. As such, the data does not allow for a distinction between Lutherans and Reform Protestants as suggested by the work of Manow (2004) and Kahl (2005). Given that both traditions are expected to generate negative attitudes toward redistribution relative to both Catholics and non-believers this does not pose a significant limitation to the analysis.

2. The typical individual is defined as an individual at the median of all control variables. In this analysis, the -typical- individual is a 48-year-old, middle-class woman who is employed, has completed upper secondary education, and defines herself as ‘coping on present income’.

3. Full results for each model will be provided in a web appendix.

4. Religious tradition is measured using an ordinal variable that differentiates between Catholic (2), Divided (1) and Protestant (0) societies. For a check of robustness to specification, a number of alternative specifications were also attempted including treating Divided and Catholic societies as separate dummy variables, treating Divided and Protestant societies as a single category, and finally, Catholics as a percentage of the population. All results presented were statistically and substantively robust to specification.

5. Full results for each model will be provided in a web appendix.