Locally efficient DR DiD estimators

The following functions implement the locally efficient doubly robust difference-in-differences estimators propose by Sant’Anna and Zhao (2020).

The resulting estimator remains consistent for the ATT even if either the propensity score or the outcome regression models are misspecified. If all working models are correctly specified, then the estimator achieves the semiparametric efficiency bound.

drdid()

Locally efficient doubly robust DiD estimators for the ATT

drdid_imp_panel()

Improved locally efficient doubly robust DiD estimator for the ATT, with panel data

drdid_panel()

Locally efficient doubly robust DiD estimator for the ATT, with panel data

drdid_rc()

Locally efficient doubly robust DiD estimator for the ATT, with repeated cross-section data

drdid_imp_rc()

Improved locally efficient doubly robust DiD estimator for the ATT, with repeated cross-section data

DR DiD estimators that are not locally efficient

When only repeated cross-section data are available, not all DR DiD estimators are locally efficient, see Sant’Anna and Zhao (2020). The following functions implement these DR DiD estimators that are not locally efficient, though, in practice, we recommend users to favor the estimators in the category above in detriment of these.

drdid_rc1()

Doubly robust DiD estimator for the ATT, with repeated cross-section data

drdid_imp_rc1()

Improved doubly robust DiD estimator for the ATT, with repeated cross-section data

IPW DiD estimators

The following functions implement the inverse probability weighted (IPW) difference-in-differences estimators propose by Abadie (2005), with either normalized/stabilized weights (Hajek-type estimators) or with unnormalized weigts (Horvitz-Thompson-type estimators).

The resulting IPW DiD estimator is consistent for the ATT only if the propensity score is correctly specified.

ipwdid()

Inverse probability weighted DiD estimators for the ATT

ipw_did_panel()

Inverse probability weighted DiD estimator, with panel data

std_ipw_did_panel()

Standardized inverse probability weighted DiD estimator, with panel data

ipw_did_rc()

Inverse probability weighted DiD estimator, with repeated cross-section data

std_ipw_did_rc()

Standardized inverse probability weighted DiD estimator, with repeated cross-section data

Outcome regression DiD estimators

The following functions implement the outcome regression (OR) based difference-in-differences estimators for the ATT, see e.g. Heckman, Ichimura, and Todd (1997).

The resulting OR DiD estimator is consistent for the ATT only if the outcome regression model for the evolution of the outcomes for the comparison group is correctly specified.

ordid()

Outcome regression DiD estimators for the ATT

reg_did_panel()

Outcome regression DiD estimator for the ATT, with panel data

reg_did_rc()

Outcome regression DiD estimator for the ATT, with repeated cross-section data

TWFE DiD estimators

The following functions implement the two-way fixed-effects (TWFE) difference-in-differences estimators for the ATT. As illustrated by Sant’Anna and Zhao (2020) in their simulation exercise, this class of estimators in general do not recover the ATT in DiD setups with covariates. As so, we encourage users to adopt alternative specifications.

twfe_did_panel()

Two-way fixed effects DiD estimator, with panel data

twfe_did_rc()

Two-way fixed effects DiD estimator, with repeated cross-section data

Data

Available datasets in the package.

nsw

National Supported Work Demonstration dataset

nsw_long

National Supported Work Demonstration dataset, in long format

sim_rc

Simulated repeated cross-section data

Package Documentation

DRDID-package

Doubly robust difference-in-differences estimators