In this paper, we present a rent-seeking model of conflict, which highlights the role of ideology in determining whether the government or the rebels take the initiative. We use the model to interpret the impact of a large-scale Community-Driven Development project on civil conflict in the Philippines. The country is characterized by the presence of two rebel groups, the New Peoples Army (NPA) and the Moro Islamic Liberation Front (MILF), with two distinct ideologies. We use a unique geo-referenced panel dataset on the occurrence of conflicts in 2003 and 2006 gathered from local newspapers that we match with nationally representative household survey and budget data on all municipalities in the country. Consistent with our models predictions, using a variety of estimation strategies, we find robust evidence that the project leads to a decline in MILF-related events and to an increase in NPA-related events.