Aversive racism at the ballot box: a field experiment on the effects of race and negative information in local elections

ABSTRACT

Research on the response of white voters to African American candidates has produced decidedly mixed results. I argue that methodological limitations are a major source of this perplexing array of findings and that a field experiment may help shed more light on the topic than is possible in laboratory experiments, surveys, or in post-election data analysis. Here, a field experiment embedded in two low salience school board elections in Columbus, Ohio produced a striking pattern consistent with the thrust of aversive racism research. To wit, simply being sent a mailing showing the race of a white or African American candidate had little effect on white voters. However, after being sent a mailing showing a candidate’s race and negative information about the candidate, white voters punished African American candidates more severely than white candidates with the same record. In short, people seemed to cast race-based votes only in circumstances in which they could deny race was a factor in their decision.

KEYWORDS:

• Race
• voting
• aversive racism
• negative ads

Notes

¹ When entered into an algorithmic database (https://www.namsor.com/) that attempts to predict race based on matching names with U.S. Census data, the names of both white candidates used in this experiment were classified as likely white. One African American candidate was classified as likely white and the other as likely Asian.

² Beyond categorical objections, several researchers caution against the use of deception in field work (Desposato, 2018; Kelman, 1967). Voters here were exposed to factual information that was widely accessible in the public record. Thus this research offers analysis rooted in the exchange of public information with voters. McDermott (2002) notes that it is the researcher’s responsibility ‘to take every reasonable precaution to reduce the potential risks to their human subjects.’ Here, subjects were not personally harmed by their participation since the entirety of the experience came to them only in the form of an ordinary presentation of campaign information already in the public conversation. Voters have been in no way personally identified in the process of the study nor in its results.

³ Precinct 60G is an excellent example of this practice. Located in northern Columbus, about ten miles from downtown, the neighborhood sends its children to three different school districts and contains two different townships. Casting votes for the school board candidates corresponding to the correct school district and trustee candidates from the proper township requires sorting precinct 60G into smaller sub-precincts. Thus the more than 1,000 voters in 60G are split into several sub-precincts, the second smallest of which has two voters, the smallest having one.

⁴ Studies have found that in low information settings, candidate attractiveness can exert a meaningful effect on voter preferences (Hart et al., 2011; Stockemer & Praino, 2015). The candidate photos used here were not chosen to put the candidates in the least flattering light, however candidate appearance remains a complicating factor for field experimental work.