Ballot Positions, Contest Differences, and Ballot Roll-Off

• In Event: How Ballot Structure Contributes to Voter Errors and Omissions

Abstract

Voter participation in American local elections has long been a concern of academics and reformers. Local contests are often presented to voters toward the end of long ballots, and voters experience fatigue and might skip the local races to get to the end of their ballots as a result. Moreover, Local contests are often non-partisan, and voters lose important decision cues when they are asked to decide between local candidates in low-information contests. Lack of exogenous variation has prevented researchers from testing theories of ballot roll-off, and the effects of ballot position versus substantive contest differences have not been convincingly identified and estimated.

In this paper, we take advantage of California's SB-25, commonly known as California's Alternative Ballot Order, to test theories of ballot roll-off. The law requires Los Angeles County to put local and school district elections on the top of the ballot and state office elections down the ballot, while maintaining the standard ballot order in other counties. Using difference-in-differences designs at the contest and precinct levels, we estimate the causal effects of ballot position versus substantive contest differences on ballot roll-off. Our preliminary results indicate that ballot position itself accounts for half of the gap between the turnout in local and state contests. Linking precinct-level turnout in each contest to individual-level voter registration and voting history files, we also examine how ballot roll-off due to ballot position versus substantive contest differences correlate with demographic and political composition.

In addition to testing theories of ballot roll-off, our results also have important policy implications. In particular, our study helps quantify the effects of moving local elections higher in the ballot. While the salience of local and school elections is increasing nationwide, our results suggest that unless those elections are moved higher in the ballot, the turnout effects will be muted.

Authors

• R. Michael Alvarez, California Institute of Technology

• Yimeng Li, Florida State University