We examine the impact of counterinsurgency aid on conflict in Afghanistan from 2005 to 2009. To enable this analysis we combine unique aid project data from NATO, household data from the Afghan government, and conflict data from US government sources. Our panel data analysis accounts for district and time period fixed effects across 398 districts and 57 months. Projects in the health sector successfully promote stability, whereas those in the education sector actually provoke conflict. Our findings are robust to reverse causation, confounding aid programs, and other sources of endogeneity. The results shed new perspective on the ‘hearts and minds’ theory commonly discussed in this vein of inquiry.