Social Incentives, Delivery Agents, and the Effectiveness of Development Interventions

In village economies, dense social networks support cooperation and exchange between citizens. The global shift towards hiring agents from within communities to deliver programs implies that the networks these agents are embedded in may affect delivery. We examine this using a randomized evaluation of an agricultural extension program in Uganda where we randomly pick one of two potential local delivery agents and map ties between delivery agents and farmers, between the two delivery agents and between farmers. Consistent with a model of favor exchange in social networks we find that (i) farmers tied to the chosen delivery agent are more likely to be treated than those tied to the counterfactual agent, (ii) this preferential treatment disappears when the two potential agents are tied by friendship, family or politics and (iii) when this is not the case the delivery agent actively prevents program benefits from diffusing to the ties of of the counterfactual agent. These results reveal the deep influence that social networks have on program delivery and help us to understand the highly unequal pattern of effects of the program both within and across villages.