Motivation In empirical research, one of the challenges to causal inference is the potential presence of unobserved confounding. Even when we adjust for a wide range of observed covariates, there’s often a lingering concern: what if there are important variables we’ve failed to measure or include?