Does Providing Corruption Information Reduce Vote Share: A Meta-analysis

• In Event: Corruption and Electoral Behavior

Abstract

Do voters in democratic countries hold politicians accountable for corruption? Field experiments that provide voters with information about the corrupt acts of politicians, then monitor vote choice have become standard in political science and economics. Similarly, vote choice survey experiments commonly provide respondents with information about the corrupt acts of hypothetical candidates. What have we learned from these experiments? A meta-analysis reveals that the aggregate treatment effect of providing information about corruption on vote share in field experiments is null. Compared to field experiments, survey experiments vastly overestimate the negative effects of corruption information on electoral outcomes. Holding other candidate features fixed by design, corrupt candidates are punished by respondents by approximately 34-37 percentage points across survey experiments, depending on estimation methods. This suggests that while vote-choice survey experiments may provide information on the directionality of informational treatments, the point estimates they provide may not be representative of real-world voting behavior. I also explore the mechanisms behind these differential treatment effects, such as social desirability bias, publication bias, and when and how survey experiments fail to capture real-world voter behavior.

Author

• Trevor Incerti, Yale University