A1244
Title: Distribution-free finite-sample inference for pre-trend testing in difference-in-differences estimation with panel data
Authors: Sieun Kim - Inha University (Korea, South) [presenting]
Seonghun Cho - Inha University (Korea, South)
Abstract: Difference-in-differences (DID) estimation in panel data critically relies on the validity of the parallel trends assumption. Conventional pre-trend tests, however, often exhibit low statistical power and may suffer from size distortions in small samples. To address these limitations, a martingale test for exchangeability based on conformal p-values is applied to the pre-trend testing problem, treating the control group as the reference set and the treatment group as the test set. Through simulation studies and an empirical application to minimum wage and employment data, it is demonstrated that the proposed approach achieves substantially higher power in detecting trend violations while maintaining accurate control of type-I error across a wide range of settings.
