Using new micro-level data on violence in Eastern Ukraine, this paper evaluates the relative merits of 'identity-based' and 'economic' explanations of civil conflict. The first view expects rebellion to be most likely in areas home to the geographic concentration of ethnolinguistic minorities. The second expects more rebel activity where the opportunity costs of insurrection are low. Evidence from the armed conflict in Ukraine supports the second view more than the first. A municipality's prewar employment mix is a more robust predictor of rebel activity than local ethnolinguistic composition. Municipalities more exposed to trade shocks with Russia experienced a higher intensity of rebel violence throughout the conflict. Such localities also fell under rebel control earlier - and took longer for the government to liberate - than municipalities where the labor force was less dependent on exports to Russia.