What Happens when a Candidate Doesn't Bark? "Cursed" Voters and Their Impact on Campaign Discourse

Formal models of candidate and voter election behavior often assume that the voters are Bayesian learners who are able to draw inferences not only from the choices candidates make, but also from choices that the candidates did not make. Here, we introduce a formal model and experiment designed to test the extent of voter updating and how it affects campaign strategies. We find that voter behavior tends toward quite incomplete updating, which in turn leads candidates to increase their use of negative campaigning.