Spatial causal inference with difference-in-differences: the impact of the confirmation of GTX-A Dongtan station on apartment prices in Hwaseong

This study applies a Difference-in-Differences (DID) model that accounts for both spatial dependence and spillover effects to address violations of the Stable Unit Treatment Value Assumption (SUTVA) in spatial causal inference.While conventional DID models estimate direct effects under the parallel trends assumption, spatial dependence and spillovers in spatial data challenge this assumption.Delgado and Florax (2015) proposed a DID model incorporating spillovers but not spatial dependence, and Bardaka et al. (2018) and Chagas et al.(2016) combined DID with Spatial Error and Spatial Lag models, respectively, to address spatial dependence.However, both studies had limitations in statistical inference of spatial causal effects.This study reviews these methodologies and aims to improve the inference for both models.Based on this framework, we analyze the impact of the confirmation of the GTX-A Dongtan station on apartment prices in Hwaseong, separating the effects into direct and indirect components.