Institutional Arrangements and Power Threat: Diversity, Democracy, and Punitive Attitudes

Abstract

This research synthesizes studies on crime and punishment, work in political sociology, and race and ethnicity scholarship in order to theorize and empirically examine the democratic foundations of group threat theory. We argue that ethnic diversity is particularly threatening when coupled with robust democratic institutions that empower individuals to pose challenges to the extant political and social order. Making use of recent measurement advances in the study of democracy, this article uses multi-level modeling techniques across 39,926 survey respondents in 27 countries from the fifth wave of the European Social Survey to test the extent to which punitive attitudes toward criminals were associated with interaction effects of an index of ethnic diversity and democratic quality. Results strongly confirm our theoretical predictions that robust democratic institutions condition the effect of ethnic fractionalization on punitive attitudes in Europe.

Keywords:

• Group threat
• political criminology
• punishment

Disclosure Statement

The authors have no conflicts of interest to disclose.

Notes

1 Data for France are missing from the HIEF dataset. Thus, we replace the value for France with the value for ethnic diversity found in Fearon and Laitin (2003) from the EPR (V. 3) dataset. Our results are robust to using the Fearon and Laitin (2003) measure in isolation, as well as with dropping France from our analysis. Our choice to proceed with the HIEF measure is due to it being the only measure we know of that contains time-variant data on ethnic diversity that contains information for the specific year 2010.

2 We should also note that we have replicated our findings using the additional indices from the V-Dem dataset (“Liberal Democracy,” “Participatory Democracy,” “Egalitarian Democracy, and “Deliberative Democracy”. Results from these robustness checks are consistent with our main findings presented here, with an interaction term significant at the p < 0.05 level. These results will be made available by request.

3 In robustness checks, we also estimate multi-level ordered logistic regression models, a binary logistic regression model based on the dichotomized punitive sentiment variable, as well as models taking the square root of the dependent variable for an interval transformation of our ordinal variable. All results are substantively identical. Thus, we proceed with the multi-level linear model.

4 This scaling procedure does not affect the direction or statistical significance for this relationship.

5 2008 for France, 2009 for Ukraine – all other are 2010