Coethnics covote in Africa: studying electoral cleavages with a covoting regression model

Ethnicity is an important cleavage in Africa, yet its influence on voting is contested. Selection biases from restricted choice sets complicate micro-level analyses, while bias from ecological inferences and unobserved confounders hamper meso and macro-level approaches. Our new Covoting Regression (CVR) tackles several of these challenges. It estimates the effect of coethnicity on the probability that pairs of voters covote for the same party/candidate while conditioning on other characteristics shared between them. Thereby, CVR mirrors the micro-foundations of widely-used aggregate indicators, such as the Herfindahl-Hirschman indices of party concentration and ethnic homogeneity. Our data consists of dyadic comparisons between respondents from Afrobarometer surveys. Pooling across 28 countries, coethnicity increases covoting intentions by 17 percentage points. The effect of coethnicity is partially driven by politically relevant groups and covoting for ethnic parties. It is consistent across institutionally diverse countries and at least five times larger than that of other cleavages. Beyond ethnicity, we address key issues in studying electoral consequences of socio-economic cleavages and bridge gaps between levels of analysis.