Do Voting Criteria Differ across Levels of the U.S. Federal System? A Conjoint Experiment to Understand Voter Decision-making

Download data and study materials from OSF

Principal investigator:

Daniel Hopkins

Georgetown University



Sample size: 532

Field period: 02/12/2013-03/17/2014

Research on Americans' vote choices has focused primarily on federal elections. Yet America's political institutions depend on an electorate that is also able to cast reasoned votes in sub-national elections. This proposal describes how conjoint analysis, a method developed in marketing, can illuminate the role of partisanship and other criteria in voter decision-making across the federal system. While survey experiments typically employ one or two randomized treatments, conjoint analysis enables a comparison of the effects of several candidate attributes on the same scale. It thus allows for tests of our core hypothesis: that voters use the same criteria across the federal system. To illustrate the method's possibilities and the plausibility of the hypothesis, this proposal presents a pilot test conducted on Amazon's Mechanical Turk in which respondents chose between presidential or gubernatorial candidates. Conjoint analysis has the potential to extend our understanding of voter decision-making.

American voters commonly choose among candidates for President, Governor, Mayor, and other executive positions. Across the federal system, these positions have differing responsibilities. This research tests the hypothesis that given today's nationalized information environment, voters are likely to use the same criteria when voting for candidates for executive office, irrespective of the specific office for which they are standing.

Experimental Manipulations
In this experiment, GfK respondents from the four largest U.S. states were randomly assigned to be asked about candidates for President, Governor, or Mayor. The respondents then evaluated 7 pairings of candidates (always for the same office), with the candidates defined by 7 attributes: gender, income, race, religion, partisanship/ideology, and two issue positions. As in other conjoint analyses, each of these attributes is randomly chosen, with its position in the conjoint table fixed across the 7 pairings for each respondent.

The core outcome is respondents' choice between two candidate profiles, each comprised of seven randomly selected attributes. We also evaluate ratings of each individual candidate on a 1-7 scale.

Summary of Results
In preliminary analyses, we find that the use of candidate attributes is generally similar irrespective of whether the respondents evaluated candidates for mayor, governor, or president. We analyze Republican and Democratic respondents separately. Both groups place a substantial premium on candidates' partisanship/ideology, preferring those candidates who share their own partisanship. While Democrats weight partisanship/ideology somewhat less when choosing mayors relative to governors or presidents, Republicans rely on partisanship/ideology to similar degrees across the three offices in question. Positions on social issues also prove highly influential in respondents' choices between candidates, but again with little distinction across the levels of government. Candidate demographics such as race, income, and religion play a more modest role in voters' choices. Analyses of the forced choice and rating outcomes produce highly similar patterns.