Essays on Strategic Voting.

Abstract

This dissertation consists of three related chapters.

The first chapter, which is written jointly with Lones Smith presents a dynamic model of deliberation by two privately informed individuals. Even by assuming the coarsest possible language to communicate information among members, it is shown that the decision is `almost instantaneous' when individuals have identical objectives. Despite the coarse syntax, the model also predicts that information aggregation can be quite effective.

The second chapter asks the question under what circumstances can a static voting mechanism aggregate dispersed information of committee members. I argue that whenever the voters are able to cast multiple votes, the quality of the joint decision increases. However, voting mechanisms are intrinsically additive ways of aggregating private information. This, naturally, is not a binding constraint if the private information is conditionally independent. However, if the `meaning' of the private information depends on other members' signals, i.e. the signals are conditionally correlated, then the joint decision by voting may be unsatisfactory. I relate this question to a representation problem in utility theory to derive abstract conditions on the joint signal distribution that are necessary and sufficient for efficient voting.

The final chapter proposes a game-theory model to study the relationship of margin and turnout in elections. Common sense suggest that any individual voter is more likely to participate in a closer election. In an equilibrium model, a closer election is a consequence of a shift in the preferences of the electorate. A change in preferences may result in a higher number of voters with strong opinions about the candidates, thus it may directly influence the number of participating voters. I show that a shift in the preferences of the electorate decreases the equilibrium margin and increases the equilibrium turnout, provided that the shift does not decrease the polarization of the preference distribution.