The Credibility Revolution

Doubly Robust Difference-in-Differences Estimators

Properties1
tagsliterature

2 min read

Doubly Robust Difference-in-Differences Estimators

Causal Question / Estimand

The average treatment effect on the treated (ATT, Causal-Estimand) in a DiD design with covariates, when one cannot be sure whether the propensity-score model or the outcome-regression model is correctly specified.

Identification Strategy

Combines an inverse-propensity-score reweighting estimator (à la Abadie) with an outcome-regression estimator into a single doubly robust (DR) estimator: it is consistent for the ATT if either working model is correct, not necessarily both. Derives the semiparametric efficiency bound for the DiD ATT and shows the DR estimator attains it when both models are correct, for both panel and repeated cross-section data.

Key Assumptions

Conditional-Parallel-Trends (parallel trends given covariates ), Overlap (propensity score bounded away from 1), and SUTVA. Correct specification of at least one of the two working models (the double-robustness insurance).

Threats to Validity

Both working models wrong simultaneously; weak overlap inflating variance. The DR construction buys robustness to one misspecification, not to a failure of conditional parallel trends itself.

Setting / Data

n/a — econometric methodology (basis of the DRDID R package). Illustrated with a job-training (LaLonde/NSW–style) evaluation.

Key Claims

  • A DiD ATT estimator that is consistent under single-model correctness and efficient under joint correctness.
  • Explicit semiparametric efficiency bounds for the DiD ATT (panel and repeated cross-section).
  • Provides the workhorse building block for modern multi-period estimators.

Connections

Citation

Sant’Anna, P. H. C., & Zhao, J. (2020). Doubly Robust Difference-in-Differences Estimators. Journal of Econometrics, 219(1), 101–122. https://doi.org/10.1016/j.jeconom.2020.06.003

Graph View

Backlinks