Ideology, Electability, and Valence

• In Event: Theoretical Implications of Empirical Models

Abstract

Political scientists regularly interpret the results of conjoint experiments in the context of electoral contests. However, researchers' typical estimand, the Average Marginal Component Effect, does not generally admit statements about elections. In this paper we present a design related to forced-choice conjoint experiments that allows researchers to recover meaningful quantities of interest for the entire distribution of voters. We conduct this experiment on a representative sample of Florida Democrats and characterize how voters trade off candidate valence features (e.g. race, gender, and experience) with other attributes like ideology and electability. Besides recovering the direction of preferences for the entire distribution of voters, our design allows us to describe the marginal rates of substitution between any pair of attributes in the survey.

Authors

• Scott Abramson, University of Rochester

• Korhan Kocak, New York University Abu Dhabi

• Asya Magazinnik, MIT