Double machine learning and automated confounder selection: A cautionary tale

Double machine learning (DML) has become an increasingly popular tool for automated variable selection in high-dimensional settings. Even though the ability to deal with a large number of potential covariates can render selection-on-observables assumptions more plausible, there is at same time growing risk that endogenous variables are included, which would lead violation conditional independence. This paper demonstrates DML very sensitive inclusion only few "bad controls" covariate space. The resulting bias varies nature theoretical causal model, raises concerns about feasibility selecting control data-driven way.