Mostly harmless? A subnational analysis of the aid-conflict nexus

Although most aid projects are aimed at local development, most research on the aid-conflict nexus is based on the country-year as unit of analysis. In contrast, this study examines the link between aid commitments and conflict intensity at the local level for three African countries between 1999-2008, using data from a unique dataset containing information on local aid allocations. The data shows that in general the spatial interdependence between aid and conflict is low, as aid is allocated relatively close to the capital and conflicts tend to occur in the peripheral areas. Fitting a Bayesian linear regression model the empirical analysis finds that there is no strong correlation between changes in lagged aid commitments and changes in conflict intensity. Looking at the extensive margin the results do show that fungible aid is correlated with increased conflict risk, in line with rent-seeking behaviour, but the estimated magnitude of the coefficient is very small. The results are stronger at the district level compared to the province level, suggesting that the possible link between aid and conflict is highly localised.