An Individual-Level Multiequation Model of Expenditure Effects in Contested House Elections

Abstract

We shall address questions concerning the impact of candidate spending in congressional elections in a new way. We develop a multiequation model of congressional vote choice that takes the endogeneity of expenditures into account. We then estimate both this model and the more traditional single-equation model using individual level survey data. The substantial differences we find between the two models indicates that the simultaneity bias present in the single-equation model is not trivial. The challenger and incumbent expenditure terms are each significant (whereas only challenger expenditures are in the single-equation setup) and of much greater magnitude in the multiequation case. In addition, we evaluate hypotheses grounded in persuasive communications theory concerning the type of individuals most affected by the messages emanating from the campaigns. We find that individuals with higher education, those with greater interest in campaigns, and those with strongly held convictions are unaffected by candidate spending, whereas individuals lacking each of these attributes are greatly influenced by campaign expenditures.