Unconfoundedness as an Information Constraint
-
Series
-
Speaker(s)Christoph Breunig (University of Bonn, Germany)
-
FieldEconometrics, Data Science and Econometrics
-
LocationTinbergen Institute, Roeterseiland Campus, E5.22
Amsterdam -
Date and time
November 27, 2026
12:00 - 13:00
Abstract
This paper provides an information-theoretic characterization of unconfoundedness within a general class of models allowing for endogenous selection. We introduce a conditional mutual information restriction requiring that, given covariates, a potential outcome contain no more information about treatment than its latent heterogeneity does unconditionally. We show that this restriction is equivalent to a multiplicative factorization of treatment probabilities into separate functions of latent heterogeneity and covariates. Although the CMI restriction allows for endogenous selection, sufficiently rich variation in treatment probabilities across covariate values makes it equivalent to unconfoundedness for the corresponding potential outcome. State specific restrictions on residual selection yield exogeneity for the corresponding potential outcomes under boundary support conditions. When those conditions fail, we provide bounds on mean potential outcomes.