Studies of large policy interventions typically do not involve randomization. Adjustments, such as matching, can remove the bias due to observed covariates, but residual confounding remains a concern. In this paper we introduce two analytical strategies to bolster inferences of the effectiveness...