Elections with incomplete information

Abstract

This dissertation consists of three chapters exploring the role of incomplete information and learning in elections. The first chapter examines the dynamics of voter learning about candidate ability in repeated elections. The dynamic process of belief revision gives rise to incentives that vary strongly over a politician's career. In particular, candidates become entrenched over time, so, even though they exert little effort, the voter cannot commit to throw incumbents out of office. I embed the basic model in a common agency framework to study seniority norms in legislative organization. The model organizes many of the stylized facts about the U.S. Congress, including the incumbency advantage, the dynamics of effort allocation over a career, the importance of constituency service, and seniority norms in committee assignments. In chapter 2, I study a simple model of campaign finance with possibly asymmetric candidates. Each candidate has the option of promising favors to interest groups in exchange for the funds they need to reveal information to the voters. When the incumbent has a sufficiently large ex-ante advantage, the challenger will be unable to raise funds at all.

(cont.) In this case, incumbent spending is unambiguously too high from the perspective of voter welfare. In fact, if the value of a good candidate is high relative to the value of favors a winner can promise, it will be socially optimal to simultaneously restrict spending by the incumbent and encourage spending by the challenger. In chapter 3, (joint with Aaron Hantman) we propose a simple model of rational learning in elections. A linear approximation to the model is used to justify a version of the Gelman-King (1990) estimator of the incumbency advantage. Restricting the model to elections for which the linear approximation should be valid produce different estimates for the incumbency advantage than those found by either Gelman and King or more traditional studies bases on the "slurge".