Previous research has focused primarily on how ethnicity may trigger civil war, and its effect on conflict duration remains disputed. Rather than treating conflict as a direct consequence of ethnic cleavages, the authors argue that ethnicity per se does not affect civil war duration. Instead, its effect depends on its relationship to political institutions. They employ a dyadic approach that emphasizes the political context in which both government leaders and nonstate challengers can capitalize on the ascriptive nature of ethnicity. They show that although states can initially benefit from politicizing ethnic relations, once violent conflict breaks out, such policies may backfire on the government and make it difficult for incumbent governments to accept settlements that could terminate conflicts. Past policies of ethnic exclusion also benefit rebel organizations fighting the government, since the resulting grievances increase collective group solidarity and render individual fighters more cost tolerant. Using a new data set that codes the nexus between rebel organizations and ethnic groups, as well as information on ethnopolitical exclusion, the authors find considerable support for their propositions.