A1451
Title: A universal difference-in-differences approach for causal inference
Authors: Chan Park - University of Pennsylvania (United States) [presenting]
Eric Tchetgen Tchetgen - The Wharton School, University of Pennsylvania (United States)
Abstract: Difference-in-differences (DiD) is a popular method for evaluating real-world policy interventions' treatment effects. Several approaches have previously developed under alternative identifying assumptions in settings where pre- and post-treatment outcomes are available. However, these approaches suffer from several limitations: either (i) they only apply to continuous outcomes and the average treatment effect on the treated, (ii) they depend on the scale of the outcome, (iii) they assume the absence of unmeasured confounding given pre-treatment covariate and outcomes, or (iv) they lack semiparametric efficiency theory. A new framework is developed for causal identification and inference in DiD settings that satisfy (i)-(iv), making it universally applicable, unlike existing DiD methods. Key to the framework is an odds ratio equi-confounding (OREC) assumption, which states that the generalized odds ratio relating treatment and treatment-free potential outcome is stable over time. Under OREC, nonparametric identification is established for any potential treatment effect on the treated in view, which would be identifiable under no unmeasured confounding. Moreover, a consistent, asymptotically linear, and semiparametric efficient estimator of treatment effects on the treated is developed by leveraging recent learning theory. The framework is illustrated with simulation studies and two real-world applications in labour economics and traffic safety evaluation.
