Ethnic violence in Gujarat in 2002 killed at least a thousand Muslims. Compiling data from the Times of India, we investigate variation across 216 towns and rural areas. Analysis reveals the political logic of violence. Killing was less likely where the Hindu nationalist Bharatiya Janata Party (BJP) was weakest, but was even less likely where the BJP was strong; it was most likely where the party faced the greatest electoral competition. Underemployment and Muslim in-migration also increased violence. The political logic is confirmed by analysis of the subsequent election: the BJP’s vote increased most in districts with the worst violence. Police chiefs in districts where violence was severe were more likely to be promoted.