What causes electoral violence in post-conflict countries? The theoretical literature emphasizes
the potential role of (1) ethnic grievances, (2) political competition and (3) specialists
in violence. Our study is the first to test these three hypotheses simultaneously. Using a
unique dataset on electoral violence in Burundi, we study variations in the intensity of electoral
violence between neighboring municipalities, relying on the fact that these are more
likely to have similar unobservable characteristics. Interestingly, we find that electoral violence
did not result from ethnic grievances, which goes against the commonly held view
that this factor necessarily plays a key role in violence in the region. Rather, we show that
electoral violence is higher in municipalities characterized by acute polarization between
demobilized rebel groups, fierce political competition and a high proportion of Hutu. The
effect of political competition is stronger in the presence of numerous demobilized rebels.