Identification and Inference for Welfare Gains without Unconfoundedness

TL;DR

This paper develops methods to bound and infer welfare gains from policy changes using observational or imperfect experimental data, without requiring unconfoundedness, and illustrates the approach with job training program data.

Contribution

It characterizes the sharp identified region of welfare gains and proposes estimation techniques using orthogonalized moments under various assumptions.

Findings

Bounds on welfare gains are derived under different assumptions.

Estimation methods perform well in finite samples as shown by simulations.

Application to job training data demonstrates practical relevance.

Abstract

This paper studies identification and inference of the welfare gain that results from switching from one policy (such as the status quo policy) to another policy. The welfare gain is not point identified in general when data are obtained from an observational study or a randomized experiment with imperfect compliance. I characterize the sharp identified region of the welfare gain and obtain bounds under various assumptions on the unobservables with and without instrumental variables. Estimation and inference of the lower and upper bounds are conducted using orthogonalized moment conditions to deal with the presence of infinite-dimensional nuisance parameters. I illustrate the analysis by considering hypothetical policies of assigning individuals to job training programs using experimental data from the National Job Training Partnership Act Study. Monte Carlo simulations are conducted to assess the finite sample performance of the estimators.