Treatment effects in interactive fixed effects models with a small number of time periods

This paper considers identifying and estimating the Average Treatment Effect on Treated (ATT) when untreated potential outcomes are generated by an interactive fixed effects model. That is, in addition to time-period individual effects, we consider case where there is unobserved time invariant variable whose effect may change over which can therefore cause (in absence of participating treatment) follow different paths for treated group relative group. The models that this generalize many commonly used treatment literature including difference differences individual-specific linear trend models. Unlike majority models, do not require number periods go infinity consistently estimate ATT. Our main identification result relies having some covariate (e.g., race or sex) vary time. Using our approach, show ATT be identified with as few three panel repeated cross sections data.