We estimate the causal effect of a large development program on conflict in the Philippines through a regression discontinuity design that exploits an arbitrary poverty threshold used to assign eligibility for the program. We find that barely eligible municipalities experienced a large increase in conflict casualties compared to barely ineligible ones. This increase is mostly due to insurgent-initiated incidents in the early stages of program preparation. Our results are consistent with the hypothesis that insurgents try to sabotage the program because its success would weaken their support in the population. (JEL D74, F35, I32, I38, O15, O17, O18, O19)

Most interpretations of prevalent counterinsurgency theory imply that increasing government services reduces rebel violence. Empirically, however, development programs and economic activity sometimes increase violence. Using new panel data on development spending in Iraq, we show that violence-reducing effects of development assistance are greater when: (i) projects are small; (ii) troop strength is high; and (iii) professional development expertise is available. These findings are consistent with an information-centric (“hearts and minds”) model, which implies that violence-reduction is greatest when projects are secure, valued by community members, and services derived are conditional on government control of the territory.