Correcting Strategic Misreporting Behavior On Outcomes in Estimating Treatment Effect

47 Pages Posted: 3 Nov 2023 Last revised: 10 Jan 2024

See all articles by Wendao Xue

Wendao Xue

University of Texas at Austin - Department of Information, Risk and Operations Management; University of Washington - Department of Economics

Date Written: September 25, 2023

Abstract

Self-reported outcomes are commonly used to identify average treatment effects. However, if reported outcomes are linked to misaligned incentives, individuals may strategically misreport their outcomes, thereby potentially biasing the estimation. We study the identification of the average treatment effect on the untreated (ATU) under two common scenarios -- incentives linked to the value (Scenario 1) and the rank (Scenario 2) of the reported outcomes. An optimal transport map is leveraged to facilitate identification in Scenario 2. We introduce plug-in estimators for ATU and derive consistency and asymptotic normality of the estimators in both scenarios. As an extension to the plug-in estimators, we derive the Neyman orthogonal moments and introduce double machine learning (DML) estimators in both scenarios. We illustrate the performance of plug-in estimators through Monte Carlo simulations. Utilizing a self-reported criminal activity dataset with validation subsample, we've show the efficacy of the proposed estimators.

Keywords: Causal inference, optimal transport, strategic behavior, measurement error

Suggested Citation

Xue, Wendao, Correcting Strategic Misreporting Behavior On Outcomes in Estimating Treatment Effect (September 25, 2023). Available at SSRN: https://ssrn.com/abstract=4583438 or http://dx.doi.org/10.2139/ssrn.4583438

Wendao Xue (Contact Author)

University of Texas at Austin - Department of Information, Risk and Operations Management ( email )

CBA 5.202
Austin, TX 78712
United States

University of Washington - Department of Economics ( email )

Box 353330
Seattle, WA 98195-3330
United States

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
138
Abstract Views
371
Rank
392,271
PlumX Metrics