Conditional Retrospective Voting in Large Elections

We introduce a solution concept in the context of large elections with private information by embedding a model of boundedly rational voters into an otherwise standard equilibrium setting. A retrospective voting equilibrium (RVE) formalizes the idea that voters evaluate alternatives based on past performance. Since counterfactual outcomes are not observed, the sample from which voters learn is potentially biased, leading to systematically biased beliefs in equilibrium. We provide an explicit learning foundation for RVE and contrast it to standard solution concepts in the literature.