Finite-sample optimal estimation and inference on average treatment effects under unconfoundedness

DOI10.3982/ECTA16907MaRDI QIDQ5860133zbMATH OpenFDO

Authors Timothy B. Armstrong, Michal Kolesár

Publication date 18 November 2021

Published in Econometrica (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1712.04594

zbMATH Keywords

overlaphonest inferencetreatment effectsunconfoundednessfinite-sample inference

Mathematics Subject Classification ID

Nonparametric estimation (62G05) Applications of statistics to economics (62P20) Nonparametric tolerance and confidence regions (62G15)

Abstract: We consider estimation and inference on average treatment effects under unconfoundedness conditional on the realizations of the treatment variable and covariates. Given nonparametric smoothness and/or shape restrictions on the conditional mean of the outcome variable, we derive estimators and confidence intervals (CIs) that are optimal in finite samples when the regression errors are normal with known variance. In contrast to conventional CIs, our CIs use a larger critical value that explicitly takes into account the potential bias of the estimator. When the error distribution is unknown, feasible versions of our CIs are valid asymptotically, even when $sqrtn$-inference is not possible due to lack of overlap, or low smoothness of the conditional mean. We also derive the minimum smoothness conditions on the conditional mean that are necessary for $sqrtn$-inference. When the conditional mean is restricted to be Lipschitz with a large enough bound on the Lipschitz constant, the optimal estimator reduces to a matching estimator with the number of matches set to one. We illustrate our methods in an application to the National Supported Work Demonstration.

Recommendations

• Robust confidence intervals for average treatment effects under limited overlap
• Robust inference on average treatment effects with possibly more covariates than observations
• Inference on a New Class of Sample Average Treatment Effects
• Dealing with limited overlap in estimation of average treatment effects
• Statistical inference for the mean outcome under a possibly non-unique optimal treatment strategy

Cited in

(13)

• Estimating the treatment effect in the two-sample problem with auxiliary information
• Inference on difference-in-differences average treatment effects: a fixed-\(b\) approach
• Higher order inference on a treatment effect under low regularity conditions
• Robust confidence intervals for average treatment effects under limited overlap
• Nonlinear and nonseparable structural functions in regression discontinuity designs with a continuous treatment
• A note on variance estimation for the Oaxaca estimator of average treatment effects
• Matching estimators with few treated and many control observations
• Confidence intervals for the treatment effect on the treated
• Identifying the average treatment effect in ordered treatment models without unconfoundedness
• Augmented minimax linear estimation
• Inference on a New Class of Sample Average Treatment Effects
• Bounding average treatment effects: a linear programming approach
• Dealing with limited overlap in estimation of average treatment effects

This page was built for publication: Finite-sample optimal estimation and inference on average treatment effects under unconfoundedness

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5860133)