Elements of External Validity: Framework, Design, and Analysis

The external validity of causal findings is a focus long-standing debates in the social sciences. Although issue has been extensively studied at conceptual level, practice few empirical studies include an explicit analysis that directed toward externally valid inferences. In this article, we make three contributions to improve approaches for validity. First, propose formal framework encompasses four dimensions validity: $ X -, T Y and C-validity (populations, treatments, outcomes, contexts). proposed synthesizes diverse concerns. We then distinguish two goals generalization. To conduct effect-generalization —generalizing magnitude effects—we introduce estimators target population effects. For sign-generalization direction novel multiple-testing procedure under weaker assumptions. illustrate our methods through field, survey, lab experiments as well observational studies.