What question did this study set out to answer?

This study examines how different matching methods impact the accuracy of the CS-DiD estimator in causal analysis.

July 10, 20260 citationsOpen Access

The Callaway-Sant'Anna (CS-DiD) Difference-in-Differences Estimator: A Practical Guide for Young Researchers

Key Points

This study examines how different matching methods impact the accuracy of the CS-DiD estimator in causal analysis.
Utilized a simulation with a hypothetical-synthetic panel to assess pre-treatment balance.
Compared Coarsened Exact Matching (CEM) and Tolerance-Based Exact Matching for achieving treatment effect estimates.
Evaluated results based on standardized mean difference and average treatment effect on the treated.
The choice of matching method significantly influences the accuracy of CS-DiD outputs.
No single matching method outperforms others across all scenarios.

Abstract

This study addresses a fundamental design decision encountered when applying the group-time difference-in-differences (CS-DiD) estimator developed by Callaway and Sant'Anna (2021)—namely, the choice of matching method used to achieve pre-treatment balance in a practical format aimed at young researchers. The standard Coarsened Exact Matching (CEM) method used in the literature (Iacus, King, and Porro, 2012) is compared directly against the Tolerance-Based Exact Matching approach proposed in the author's earlier work, using a parallel simulation on a hypothetical-synthetic panel. The comparison is conducted in terms of standardized mean difference (SMD) balance and final treatment effect (ATT) estimates. The findings show that the choice of matching method directly affects the accuracy of the CS-DiD output, and that no single method is superior under all conditions. The data used in this study are entirely synthetic. JEL Codes: C14, C21, C23, C54Keywords: difference-in-differences, CS-DiD, matching, CEM, tolerance-based exact matching, impact evaluation

问 AI

Bookmark

View Full Paper