A Computational Model of the Citizen as Motivated Reasoner: Modeling the Dynamics of the 2000 Presidential Election

Abstract

We develop a computational model of political attitudes and beliefs that incorporates contemporary theories of social and cognitive psychology with well-documented findings from electoral behavior. We compare this model, John Q. Public (JQP), to a Bayesian learning model via computer simulations of empirically observed changes in candidate evaluations over the course of the 2000 presidential election. In these simulations, JQP clearly outperforms the Bayesian learning model. In particular, JQP reproduces responsiveness, persistence, and polarization of political attitudes, while the Bayesian model has difficulty accounting for persistence and polarization. We demonstrate that motivated reasoning - the discounting of information that challenges prior attitudes coupled with the uncritical acceptance of attitude-consistent information - is the reason our model can better account for persistence in candidate evaluations over the course of the campaign. Two implications follow from the comparison of models: (1) motivated reasoning explains the responsiveness, persistence, and polarization of political attitudes, and (2) any learning model that does not incorporate motivated reasoning will have difficulty accounting for the persistence and polarization of political attitudes.

Keywords: political cognition, information processing, motivated reasoning, political attitudes, computational modeling