Isolating the Ethnic Effect on Political Behavior: A Survey Experiment Using Mixed Ethnicity Leaders in Benin

Abstract

Do African voters favor co-ethnic candidates? African elections have been characterized as little more than ethnic head counts (Horowitz 1985), and electoral patterns often confirm this impression. But individual-level surveys of the African voter paint a more complex portrait, highlighting the role that performance evaluations play along- side ethnicity (Bratton et al. 2011). In fact, efforts to identify the independent ethnic effect on voting preferences meet a particularly formidable challenge. Insofar as ethnicity is an ascriptive characteristic of a person, not a variable we can randomly assign, and insofar as ethnicity is typically correlated with other characteristics such as class or partisanship, identifying its independent causal effect has proven difficult. This paper offers a new identification strategy that sheds light on whether African voters favor co-ethnic candidates because of ethnicity per se. Relying on mixed-ethnicity leaders and their plausible membership in more than one ethnic group, this paper proposes and shows that survey experiments can successfully measure the independent effect of ethnicity: holding the identity of an actual African head of State constant, the manipulation randomly assigns one ethnic cue versus the other. The method is applied and tested in Benin, a small West African democracy led by a mixed-identity leader. The results confirm that, in this context, ethnicity plays an independent, causal role on the African voter’s decision, and confirm the effectiveness of the identification strategy.