The Performance of Rational Voter Models in Recent Presidential Elections

Abstract

In An Economic Theory of Democracy, Downs hypothesized that rational, utility-maximizing citizens would calculate the benefits from voting (as opposed to abstaining altogether) and then vote if this expected benefit exceeded voting cost. This hypothesis is logically derived from economic principles of rational behavior, but is counterintuitive: even if the utility increment is plausibly large, the probability of casting a pivotal, tie-breaking vote in any sizable polity is minute. Hence the expected benefit will be less than the time cost of voting, and no rational person will vote. Nevertheless, a number of statistical studies have found that voter turnout does respond to differences in the probability of casting a decisive vote, even though that probability may be small.This article uses data gathered at the state level from the 1968, 1972, 1976, and 1980 general elections to reestimate the relationship between voter turnout and the probable closeness of the election. Pooled data Least Squares Dummy Variable (LSDV) regressions are used whenever preliminary tests indicate that such pooling is permissible. The empirical results suggest that the relationship between closeness and turnout is weak, unstable, or nonexistent in all of the models tested.