Synthetic Control Methods: Never Use All Pre-Intervention Outcomes Together With Covariates

Abstract

It is becoming increasingly popular in applications of synthetic control methods to include the entire pre-treatment path of the outcome variable as economic predictors. We demonstrate both theoretically and empirically that using all outcome lags as separate predictors renders all other covariates irrelevant. This finding holds irrespective of how important these covariates are for accurately predicting post-treatment values of the outcome, potentially threatening the estimator's unbiasedness. We show that estimation results and corresponding policy conclusions can change considerably when the usage of outcome lags as predictors is restricted, resulting in other covariates obtaining positive weights.