When do voters reveal candidate gender preferences? Evidence from individual-level ballot data

Abstract

Does candidate gender matter for vote choice? Whereas experimental research suggests an average preference for female candidates, observational studies tend to find null effects. In this note, we address the recent debate on how to measure voter preferences on the aggregate and the individual level. We argue that candidate gender preferences exist, but that whether and when they are revealed varies between and within voters. Drawing on an observational design and using data from over 500,000 individual ballots in Lithuanian elections, we employ multilevel regression and exponential random graph models to show how voters' candidate gender preferences are distributed across the electorate and how they vary in size and direction. We find that about half of all voters prefer either male or female candidates. Whereas preference for male candidates tends to be revealed in the first and second preference votes, preference for female candidates is first revealed in lower preference votes. Our results help explain contradictory findings in the literature and illustrate how observational data and methods can be used to assess voter preferences within electorates.