Dynamics of Voting Propensity : Experimental Tests of Adaptive Learning Models

Abstract

This paper aims to deliver experimental evidence on the dispute between two behavioral models of electoral turnout. Both models share the idea that the subjects’ voting propensities are updated from their past propensities, aspirations, and realized payoffs. However, they differ in the exact specification of the feedback mechanism. The first model has a strong feedback mechanism toward 50 percent, while the other has only moderate feedback. This difference leads to two distinct distributions of voter types: the first model generates more casual voters who vote and abstain from time to time. The latter generates more habitual voting behavior. Thus far, the latter model seemed to be better supported empirically because survey data reveal more habitual voters and abstainers than casual voters. Given that the two models differ in their propensity updating mechanism in dynamic processes, a more direct test of their assumptions as well as implications with survey data is still pending. We designed a laboratory experiment in which subjects repeatedly make turnout and voting decisions. The result from experimental data is mixed, but more supportive of the second model with habitual voters and abstainers.