Weighting for External Validity
Working Paper 23826
DOI 10.3386/w23826
Issue Date
Revision Date
External validity is a challenge in treatment effect estimation. Even in randomized trials, the experimental sample often differs from the population of interest. If participation decisions are explained by observed variables such differences can be overcome by reweighting. However, participation may depend on unobserved variables. Even in such cases, under a common support assumption there exist weights which, if known, would allow reweighting the sample to match the population. While these weights cannot in general be estimated, we develop approximations which relate them to the role of private information in participation decisions. These approximations suggest benchmarks for assessing external validity.