Doubly robust difference-in-differences estimators

By Pedro H. C. Sant’Anna, Jun Zhao. 2018

835 citations recorded by the source corpus.

Abstract This article proposes doubly robust estimators for the average treatment effect on the treated (ATT) in difference-in-differences (DID) research designs. In contrast to alternative DID estimators, the proposed estimators are consistent if either (but not necessarily both) a propensity score or outcome regression working models are correctly specified. We also derive the semiparametric efficiency bound for the ATT in DID designs when either panel or repeated cross-section data are available, and show that our proposed estimators attain the semiparametric efficiency bound when the working models are correctly specified. Furthermore, we quantify the potential efficiency gains of having access to panel data instead of repeated cross-section data. Finally, by paying particular attention to the estimation method used to estimate the nuisance parameters, we show that one can sometimes construct doubly robust DID estimators for the ATT that are also doubly robust for inference. Simulation studies and an empirical application illustrate the desirable finite-sample performance of the proposed estimators. Open-source software for implementing the proposed policy evaluation tools is available.

Subfields: Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Deep Neural Networks for Estimation and Inference: Application to Causal Effects and Other Semiparametric Estimands (2018, methodology): “Our results can thus be used in IV models (Belloni, Chen, Chernozhukov, and Hansen, 2012), partial linear models (Belloni, Chernozhukov, and Hansen, 2014), games (Chernozhukov, Escanciano, Ichimura, Newey, and Robins, 2018), differencesin-differences (Sant’Anna and Zhao, 2018), and other areas (Bell”
Difference-in-Differences with multiple time periods (2018, methodology): “1(a) of Sant’Anna and Zhao (2020), whereas √ n(ÂTT dr,nev t≥(g−δ) −ATTt≥(g−δ)) d −→ N(0,E[Ψ t≥(g−δ)(W )Ψ dr,nev t≥(g−δ)(W ) ′])”
When Is Parallel Trends Sensitive to Functional Form? (2020, result): “Our results thus have implications for the conditional parallel trends assumptions considered in, for example, Abadie (2005) and Sant’Anna and Zhao (2020).”
How Much Should We Trust Staggered Difference-In-Differences Estimates? (2021, methodology): “…effect average, which is likely a result of using OLS varianceweighting rather than weighting explicitly by the sample share as in Callaway and Sant’Anna (2020) and Sun and Abraham (2020).18 The results of this stylized simulation show how in settings with panel data and heterogeneous…”