Many Facets of Difference-in-Differences: Beyond the Canonical Framework

• In Event: Beyond Traditional DiD: Exploring New Approaches in Panel Data Analysis

Abstract

Since Card and Krueger's (1994) seminal work, the difference-in-differences (DID) approach has gained widespread use in social science research. However, current applications frequently diverge from the original DID model, particularly in lacking a never-treated group. This paper distinguishes between DID as a research design and as an estimator, highlighting their application in contexts that deviate from the canonical DID in terms of estimand and identifying assumptions. We introduce the comparative interruption design as a notable variant, where all units receive treatment simultaneously, and researchers focus on the causal moderation effect of a pre-treatment variable. We present identification results and estimation strategies for these alternative designs, urging caution against conflating diverse applications under the DID umbrella.

Author

• Yiqing Xu, Stanford University