Synthetic Controls as Balancing Scores

Harsh Parikh

Keywords: Causal Inference, Causality, Synthetic Control, Matching

TL;DR: Synthetic Controls are Balancing Scores and guarantee conditional ignorability for causal inference.

Abstract: We outline the factors under which conditioning on Synthetic Control (SC) weights emulates a randomized control trial where the treatment status is independent of potential outcomes. Specifically, we demonstrate that if there exist SC weights such that the treatment effects are exactly identified, and these weights are uniformly and cumulatively bounded, then SC weights are balancing scores.