Events

Professor Joshua Shea
Assistant Professor of Economics
University of Illinois Urbana Champaign

We develop a framework to identify dynamic treatment effects in a panel setting with endogenous treatment timing and repeated binary instrumental variable (IV) shocks. Agents respond dynamically to the instrument when selecting timing of treatment, generating multiple latent complier and non-complier types. Using the sequential IV shocks, we nonparametrically identify the distribution of types and dynamic treatment effects. Heterogeneous effects can be separately identified across complier types and treatment cohorts. Neither conventional monotonicity nor parallel trend assumptions are required. We propose nonparametric estimators for the distribution of latent types and treatment effects and derive their asymptotic properties. We additionally propose a model misspecification test to aid researchers in modeling selection into treatment. Simulations indicate good finite-sample performance, even in the presence of many latent types.