How effective are social programs during conflicts? And how is the effectiveness of social programs affected by conflict intensity? In this paper, we consider the impact on child anthropometrics and household expenditures of the only major social program in Angola during the civil war. Our identification strategy is based on the political geography of program deployment. Our hypothesis is that one of the purposes of program implementation was to consolidate government political support in areas contested with UNITA, the main rebel group, as well to maintain the support of the population in areas to the rear of frontline communities. Based on a simple model of spatial competition, we show that the likelihood of treatment of a community should be increasing in the distance separating it from the government’s forward base, and decreasing in the distance to UNITA’s mainbase at the time. Using these exclusion restrictions to generate plausibly exogenous variation in treatment status, our linear instrumental variables estimates show that treatment at some point during the 1994-2000 period was associated with a 48.5% increase in household expenditures per adult equivalent, and a 0.359 increase inchild height-for-age z-scores, in 2000. Results based on the local instrumental variables estimator show that each 1,000 additional deaths attributable to the civil war within a 20 km radius of the community shifts the marginal treatment effect associated with household income per adult equivalent up by 18.2%. This particular social fund was therefore associated with substantial benefits for treated communities, with these benefits being significantly increasing in the intensity of violence faced by the inhabitants.