Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
I develop a theory of pre‐electoral violence, in which parties resort to violent tactics if political competition is tight and violent supporters are numerous, and if polarization between violent supporters is high. The importance of these conditions varies with the type of violence (clashes, intimidation or murders). I test the model using data on the 2010 elections in Burundi. In line with the model, electoral violence before the election was more likely in municipalities characterized by both close political competition and high density of demobilized combatants. Violence was also more likely where polarization between demobilized combatants was high.