Treatment Effect Accounting for Network Changes : Evidence from a Randomized Intervention

Networks may rewire in response to interventions. We propose a measure of the total treatment effect for those situations where an intervention may affect the network structure. To build this measure, we develop a treatment response model allowing for dynamic peer effects. We illustrate our results using unique panel data on the network of informal financial transactions before and after a field experiment that randomized access to savings accounts in Nepal. Using data on expenditures in medicines and traditional remedies, we show that standard treatment effect measures neglecting network changes could be underestimated by over one third in this context