LAUSR

I document evidence of ethnic favoritism in a panel of 163 ethnolinguistic groups partitioned across 35 African countries. In contrast to previous studies, I construct a computerized lexicostatistical measure of linguistic similarity between each ethnic group and the national leader as a novel measure of ethnic proximity. I exploit the arbitrary placement of African political borders as a source of exogenous within-group variation, where the similarity of the same partitioned group varies over time according to the ethnolinguistic identity of the national leader on each side of the border. To quantify patronage at the group level, I isolate time variation in night light luminosity resulting from changes in the ethnolinguistic identity of a leader. Using a triple-difference estimator I find that a one standard deviation increase in linguistic similarity yields a 7.0 percent increase in luminosity, which corresponds to a 2.1 percent increase in group-level GDP per capita. I then use the continuity of linguistic similarity to show that favoritism exists among groups that are not coethnic to the leader, where the mean effect of non-coethnic similarity is one quarter the size of the coethnic effect. I corroborate this evidence using individual-level data and establish that it's where an individual lives and the attached ethnolinguistic identity that predicts favoritism, not the identity of the individual respondent. I relate these results to the literature on coalition building, and provide evidence that ethnicity is one of the guiding principles behind high-level government appointments.