You Don't Have a Selection Problem: We Live in a Selectiony World

Empirical legal researchers often seek to make causal inferences about a class of phenomena that we call selectiony treatments: treatments that are defined by some reasons, rules, or mechanisms that bring them about. Reviewing the literature, we suggest that these types of treatments suffuse the law, because the law is all about limiting the kinds of reasons that can serve as the grounds for a particular legal status. Selectiony treatments pose a distinctive conceptual problem for causal inference: They make it difficult to define counterfactual contrasts that correspond to real-world interventions of interest. We show how this problem surfaces in the study of canonical legal phenomena like bail, incarceration, and police stops, and we clarify how it differs from familiar concerns like confounding and selection bias. We provide a clear way for researchers to think through these issues and diagnose limitations to their causal inferences by emphasizing two prerequisites for coherent causal claims: ( a ) a clear definition of the variables inside a causal model and ( b ) a clear understanding of the causal estimand that is its target.