Does Democratic Backsliding Depend on the Policy Issue? Evidence from a Conjoint Experiment

Abstract

Recent studies have shown that citizens in established democracies are willing to make trade-offs between valued political goals and democratic principles. Yet we know rather little about the particular types of policy issues that are more likely to make citizens prioritize them over democratic norms. These questions are particularly important in European multiparty systems that have experienced significant party system realignments in recent years. Drawing on the literature on democratic backsliding and issue dimensions in multiparty democracies we hypothesize that identity-related policy issues are more likely to generate trade-offs between issue congruence and democratic principles. Our theoretical expectations are tested with evidence from a pre-registered conjoint experiment embedded in a nationally representative survey (n=2406) conducted in Finland, an established multiparty democracy where identity-related issues have become increasingly politically salient. We find strong evidence that citizens are willing to make substantial trade-offs between their favoured issue positions and democratic principles. However, contrary to our expectations, these effects are similar across all issue areas, and are not more pronounced over identity-related issues. Our results therefore do not support to the expectation that voters would be willing to overlook democratic principles only for particularly polarizing political issues. Instead, our results are in line with earlier literature that stresses instrumental benefit-seeking in political behaviour.

The politicization of European integration and support for restrictive migration policies

“…As an additional, exploratory test, I estimate individual-level marginal component effects (IMCEs) using Bayesian Additive Regression Trees as proposed by Robinson and Duch (2023). Unlike AMCEs, IMCEs represent the probability of a specific respondent “choosing a profile given an attribute level” (Robinson and Duch, 2023: 8).…”

“…As an additional, exploratory test, I estimate individual-level marginal component effects (IMCEs) using Bayesian Additive Regression Trees as proposed by Robinson and Duch (2023). Unlike AMCEs, IMCEs represent the probability of a specific respondent “choosing a profile given an attribute level” (Robinson and Duch, 2023: 8). In other words, the IMCE recovers the respondent-level effect of each attribute on the outcome, given respondents’ choices in each of the multiple conjoint tasks they completed.…”