According to theories of motivated reasoning, people process information in a way that allows them to maintain or improve their positive views of in-group members relative to out-group members. In American politics, this phenomenon can help explain why voters better respond to information when it supports or improves their partisan identities. Motivated reasoning, however, has not previously been used to account for the phenomenon of ethnic voting. Why do individuals repeatedly vote for candidates who share their own ethnicity? Much existing scholarship treats ethnic voting as a product of low-information environments, where ethnicity may be an important heuristic for judging candidate quality. Absent more detailed performance information, a candidate’s ethnicity may also be used to judge which ethnic groups will receive more goods and services if the candidate is elected. The authors tested these theories against the theory of motivated reasoning, which states that voters only respond to positive information when it is about a member of the same identity group, and negative information when it is about someone from a different identity group.
To understand how access to information influences ethnic voting, the authors implemented a field experiment during the 2015 National Assembly elections in Benin. Voters have poor information on these legislative races, and also vary in their shared ethnicity with the incumbent in their constituency. These factors allowed the authors to provide information to voters to examine how they updated their beliefs about the candidate, conditional on their shared ethnicity with the incumbent. The authors conducted the field experiment in 30 administrative communes. Within each commune, the authors randomly selected five villages or quarters from three blocks: urban; rural and electorally competitive; and rural and non-competitive. Villages were randomly assigned to receive or not receive information about the incumbent in the election. Information is delivered to villages as a video, which ensures consistency in treatment delivery. Information contains data on legislative activity of the politicians, which includes information such as rate of attendance in various committees and sessions. On average, 47 people per treatment village received the information video and presumably more learned about the video through those who received treatment. To measure treatment effects at the village level, the authors matched 2015 administrative data on election outcomes at the polling station level to treatment and control villages. To measure treatment effects at the individual level, the authors also collected panel survey data. Baseline data was collected 2 weeks to 1 month prior to the election, and endline data was collected immediately after the election and before the results were announced. 3,419 individuals participated in both the baseline and endline survey.
The field experiment reveals that positive information on candidate’s prior performance increased support for the incumbent, but only among co-ethnics. Similarly, negative performance information decreased support for the incumbent among non-co-ethnics but did nothing for support among co-ethnics. These results remain the same when the authors control for an individual’s prior beliefs about the incumbent’s legislative performance. These results hold when authors use both administrative election outcomes data and the survey data. The authors additionally find that the results do not support the two prevailing theories about co-ethnic voting. If ethnicity provided information to voters on candidate quality, citizens would be less likely to vote on ethnicity once they learn about incumbent performance. Similarly, if ethnicity provided a signal about which group will benefit in office, voters would be indifferent to information showing the candidate performed well or poorly in office. Neither pattern shows up in the data. Voters update their beliefs about the incumbent conditional on their shared ethnicity. These findings are consistent with ethnically motivated reasoning. Co-ethnics are less likely to learn and recall information about the incumbent when the information is negative, suggesting that information is effective only when it reinforces positive beliefs about the shared identity group. Additionally, these findings contribute to social science research on motivated reasoning by showing that the effects appear as more than just a survey measure of individual attitudes, but as real-world electoral behavior in an actual election.
This study finds that voters respond to information about incumbents in a way that amplifies ethnicity-based voting, suggesting that a better-informed electorate will not reduce its reliance on ethnic cues. Related, the study finds that the provision of politician performance information changes voter behavior only when the information aligns with voters’ ethnic identity attachments. The study provides a basis for further work to understand whether and how it is feasible to reduce ethnically motivated voting in response to new information. The authors suggest that such ethnic polarization may be overcome by priming a larger, shared identity such as nationality, or by priming a cross-cutting identity (such as cousinage in Mali). Multiple studies on discrimination also find that making people more aware of their discriminatory biases reduces such bias.