Electoral Crime Under Democracy: Information Effects from Judicial Decisions in Brazil

TL;DR

This study investigates how the public disclosure of electoral criminal records in Brazil influences voter behavior, showing that convictions significantly decrease candidates' chances of election, especially for severe crimes, highlighting the importance of credible information sources.

Contribution

It provides new evidence on the electoral impact of judicial decisions in large democracies and uses machine learning to classify crime severity, revealing differential voter responses.

Findings

Convictions reduce candidates' election probability by 10.3 percentage points.

Vote share decreases by 12.9 percentage points after convictions.

Larger penalties (6.5 p.p.) occur for severe crimes.

Abstract

This paper examines voters' responses to the disclosure of electoral crime information in large democracies. I focus on Brazil, where the electoral court makes candidates' criminal records public before every election. Using a sample of local candidates running for office between 2004 and 2016, I find that a conviction for an electoral crime reduces candidates' probability of election and vote share by 10.3 and 12.9 percentage points (p.p.), respectively. These results are not explained by (potential) changes in judge, voter, or candidate behavior over the electoral process. I additionally perform machine classification of court documents to estimate heterogeneous punishment for severe and trivial crimes. I document a larger electoral penalty (6.5 p.p.) if candidates are convicted for severe crimes. These results supplement the information shortcut literature by examining how judicial information influences voters' decisions and showing that voters react more strongly to more credible sources of information.