U.S. Presidential Elections 1968-80: A Spatial Analysis

Abstract

A methodology is presented for empirical estimation of spatial models of voting in mass elections. The basic choice model posits utility functions that depend on spatial distance and a random error term. The parameters of the utility functions are estimated by a polytomous logit model, which is applied to spatial maps of voters and candidates for all presidential elections since 1968. These maps indicate that presidential candidates take positions at the periphery of the distribution of voters. Results based on these maps support the hypothesis of sincere, spatial voting in two-candidate elections. In three-candidate elections, the coefficients of the voter utility function adjust to proxy for the lesser viability of the minor party candidate. In contrast to choice, turnout shows only weak spatial effects. The data are inconsistent with the proposition that turnout is low because voters are not offered distinct alternatives. While the utility function used by voters appears stable through time and candidate and voter positions appear stable in the last few months prior to an election, voter and candidate positions exhibit substantial change over longer time periods.