2024/4/12 CRETA Seminar - Synthetic Historical Control for Policy Evaluation without Cross-sectional Controls

Title：

2024/4/12 CRETA Seminar - Synthetic Historical Control for Policy Evaluation without Cross-sectional Controls

Introduction：

We propose a synthetic historical control method for policy evaluation without using cross-sectional controls. This method is established in the spirit of the conventional synthetic control method but is distinguished by replacing the cross-sectional controls with a set of historical controls. This replacement is crucial for policy evaluation in the absence of valid cross-sectional controls. Our method is built on a semi-parametric time-series regression that accounts for not only observed features but also a latent component of the outcome variable of the treated unit in addition to the time-varying intervention effect to be estimated. We demonstrate that the proposed method is approximately unbiased for estimating the intervention effect under suitable conditions and performs well in simulation and empirical applications.