A Bayesian explanation for the effect of incumbency

Incumbents appear to perform better in elections because they are incumbents, yet the most commonly proposed explanations for this phenomenon are unsatisfying. I introduce a new explanation that is simple, parsimonious, and largely consistent with empirical evidence. If voters lack perfect information about electoral candidates, incumbency is an informative signal of quality, and voters will update their beliefs accordingly. I formalize these claims with a decision-theoretic model where voters receive noisy signals of candidate quality, and I discuss several empirical phenomena consistent with this explanation. For example, when voters learn that their incumbent barely won office, they are less likely to support reelection and the effect of incumbency largely disappears.