Causal Inference under Selection on Observables in Operations Management Research : Matching Methods and Synthetic Controls

Majority of recent empirical papers in the Operations Management (OM) use observational data to investigate the effects of a adopting a program or policy. However, observational data does not offer random assignment of these program or policies; therefore, estimating the causal effects is challenging. In the special case where one can reasonably assume that all confounding factors are observed, i.e., selection on observables, the family of matching methods and synthetic controls can help researchers replicate a randomized experiment, the most desirable setting to draw causal inference. In this paper, we first provide an overview of matching methods and their use in the OM literature. Then, we lay out the framework and offer practical guidance for propensity score matching and coarsened exact matching that gained substantial popularity in OM recently. Finally, we discuss synthetic controls, a method that offer unique advantages over matching methods in certain scenarios and is likely to become more popular in OM in near future. We hope that this paper will stimulate a more rigorous application of matching and synthetic control methods