Causal Models for Panel Data

Speaker(s) Type Length Chair
Guido Imbens Keynote Address
Oliver Bruce Linton

Professor Imbens discusses identification and estimation of causal effects of a binary treatment in settings with panel data. The talk highlights that there are two paths to identification in the presence of unobserved confounders. First, the conventional, model-based, path based on making assumptions on the relation between the potential outcomes and the unobserved confounders. Second, a design-based path where assumptions are made about the relation between the treatment assignment and the confounders that is more commonly used in randomized experiments. In the talk, Professor Imbens introduces different sets of assumptions that follow the two paths, and develops double robust approaches to identification (rather than estimation) where both approaches are exploited.