Misperceived Social Norms and Political Accountability: Evidence and Theory

Elections can deter corruption only if voters punish tainted incumbents. We study whether punishment depends on second-order beliefs---beliefs about how other voters will react. Before Japan’s October 2024 general election amid a funding scandal, we ran a pre-registered online survey experiment. To study this channel, we provided no new factual information about the scandal itself and instead reported a baseline statistic about perceived public intolerance of the underlying corruption: treated respondents learned that, in our baseline survey, the average respondent estimated that 67% of other respondents viewed the conduct as unacceptable. The message increased turnout by 6 percentage points and support for opposition challengers by 7 percentage points. Effects were sharply heterogeneous. Swing voters, especially those who initially overestimated how widely others would punish, became more likely to vote and back challengers. By contrast, ruling-party supporters, especially those who initially underestimated how widely others would punish, shifted toward the incumbent when they learned that intolerance of the corruption was higher than expected. More broadly, anti-corruption messages may affect voting not only by changing beliefs about wrongdoing, but also by changing beliefs about others’ reactions, helping explain why such campaigns often have mixed effects.