Interference/spillover is a common concern in place-based policy evaluation. This paper develops a difference-in-differences framework that accounts for interference using panel data with multiple posttreatment periods and potentially staggered treatment adoption. I define causal parameters under interference, establish identification under a modified parallel trends assumption, and propose doubly robust estimators with valid inference. The method is illustrated through an empirical application.
Ruonan Xu (Fri,) studied this question.