Laboratory Experiments in Voter Behavior

William Christie

Advisor: Daniel E. Houser

Committee Members: Thomas Stratmann, David Porter, Interdisciplinary Center for Economic Science

Truland Building, 400-J
December 02, 2010, 09:00 AM to 11:00 AM


In November 2006 a ballot initiative was presented and rejected in the state of Arizona which would have created a lottery for participating voters. Supporters argued that lotteries would increase voter turnout. From statements of the leaders of the initiative it can be expected that more such proposals will be put forward in the coming years both in Arizona and in other states. Both theory and experimental and field data suggest that uninformed swing voters might rationally abstain under certain conditions. Experimental evidence has shown that higher-quality candidates are more likely to be elected in the presence of truthful advertising when abstention is allowed. In such an election regime, a lottery drawing voters to the polls could increase the chances of a low-quality candidate being elected, thus lowering the expected social benefit of a particular campaign cycle.

This dissertation examines the issue of voter lotteries in a laboratory economics experiment. The discussion is prefaced with a review of the literature on voter turnout and abstention. We begin with decision-theoretic models and move on to game- and information-theoretic contributions. Particular attention is paid to the calculation of voter benefits and to the way in which a lottery might affect the voter’s payoff structure. A rational voter should weight the positive expected lottery payout with potential for being the pivotal voter in electing a low-quality candidate. In our experiments, the number of voters is small, so the probability of being the pivotal voter is high relative to large elections. The expected value of the lottery is smaller than the penalty for electing a low-quality, as opposed to a high-quality, candidate. Specifically, the lottery is a high-payout, low-probability event, matching the Arizona ballot initiative. Given these two conditions we formulated a research hypothesis that voter turnout should not increase significantly in the presence of such a lottery when compared to non-lottery elections.

In fact, we find that abstention rates drop almost to zero in the presence of a lottery, a difference that is highly significant compared to abstention rates in non-lottery conditions. We discuss possible explanations for this outcome, and we suggest future experiments to probe the issue further.

The dissertation concludes with a discussion of experimental economics software requirements and design practices. The type of software required for laboratory economics experiments has no close analog in the software taxonomy. Consequently there exist no teaching references beyond the study and copy of existing source code, which more often than not fails follow good design practice. Flawed software and long development times are anecdotally the cause of the biggest delays in an experimental project. We make a first attempt at remedying this by providing a basic outline of good-practice design and the specific requirements of experimental software.