A0496
Title: Spatial-temporal synthetic error model of causal analysis with application to policy causal effect evaluation
Authors: Yan Zhang - University of Southampton (United Kingdom) [presenting]
Zudi Lu - University of Southampton (United Kingdom)
Abstract: Causal analysis of spatial-temporal data is challenging owing to spatial-temporal interactions. The synthetic control method (SCM) is popular in estimating the causal effect of a given intervention on a single or a small number of units in a non-spatial panel data setting by weighted averaging of the control units to balance the outcomes and covariates of the treated unit. Inspired by the ideas of synthetic control method and spatial-temporal models, a spatial-temporal synthetic error model (STSEM) is proposed as a new framework of linear spatial-temporal causal inference model to infer the causal effect of some given intervention on the metric that is of interest for spatial-temporal data, with its synthetic weights determined by LASSO regression. Asymptotic properties of the proposed model are established, followed by which the significance of the causal effect can be tested. In addition, its performance is also compared in causal effect inference with the traditional SCM, the augmented SCM (ASCM), and a simplified STAR-PLR model in a simulation study and an empirical study, in which the causal effect of the Kansas tax cut on its GDP is demonstrated for inference.
