Who takes the blame? The strategic effects of collateral damage

Can civilians caught in civil wars reward and punish armed actors for their behavior? If so, do armed actors reap strategic benefits from treating civilians well and pay for treating them poorly? Using precise geo-coded data on violence in Iraq from 2004 through 2009, we show that both sides are punished for the collateral damage they inflict. Coalition killings of civilians predict higher levels of insurgent violence and insurgent killings predict less violence in subsequent periods. This symmetric reaction is tempered by preexisting political preferences; the anti-insurgent reaction is not present in Sunni areas, where the insurgency was most popular, and the anti-Coalition reaction is not present in mixed areas. Our findings have strong policy implications, provide support for the argument that information civilians share with government forces and their allies is a key constraint on insurgent violence, and suggest theories of intrastate violence must account for civilian agency.