Identification problems and decisions under ambiguity: Empirical analysis of treatment response and normative analysis of treatment choice

Identification problems that arise in the empirical analysis of treatment response generate ambiguity about the identity of optimal treatment rules. This paper develops general themes about decisions under ambiguity, next specializes to problems of treatment choice under ambiguity, and then shows how identification problems induce ambiguity in treatment choice. The main ideas are given specific form through consideration of the treatment choice problem of a planner who observes the treatments and outcomes realized in a classical randomized experiment, but who does not observe the covariates of the experimental subjects.