🤖 AI Summary
This study addresses the lack of systematic criteria for covariate selection in difference-in-differences (DID) designs, which often leads to arbitrary specifications of the conditional parallel trends assumption. It introduces causal graphical models to formally identify the set of covariates required to satisfy this assumption, highlighting the critical role of time-invariant covariates and distinguishing between treatment types and implementation strategies to clarify treatment–confounding feedback mechanisms. By integrating a multi-period DID framework with conditional independence tests and consistency analysis of adjustment sets, the paper proposes principled criteria for covariate inclusion, resolving the prevalent mismatch between adjustment sets and estimation methods. This approach substantially enhances the identification validity and robustness of causal effect estimates in DID analyses.
📝 Abstract
Difference-in-differences (DiD) identification relies mainly on a parallel trends assumption about untreated potential outcomes. Researchers often relax this assumption by assuming conditional parallel trends within units with the same covariate values. However, the process of selecting which covariates to include in this assumption is often \emph{ad hoc}. We propose a formal approach to select the variables that support conditional parallel trends based on graphical criteria. We show that the parallel trends assumption is rarely justified without conditioning on covariates, and that unconditional and conditional parallel trends can conflict with one another. We also demonstrate that a time-invariant covariate with a time-invariant effect on the outcome, which might not ordinarily be considered a confounder in DiD, may be a useful conditioning variable. We clarify that adjustment for a post-treatment covariate depends on what causes that covariate to change. Extending our framework to multiple time periods, we distinguish between treatment type and rollout strategy and examine the problem of treatment-confounder feedback. On the estimation side, we argue that the difficulty of incorporating covariates in DiD, often framed as an estimator problem, is more accurately understood as a misalignment between the adjustment set used by the estimator and the adjustment set required for identification. This misalignment affects several popular estimation procedures, and resolving it requires not a change of estimator, but a change in how covariates enter the estimation procedure. We show how to achieve this alignment for all estimators we evaluate.