|
Activity Number:
|
212
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
| Abstract - #305420 |
|
Title:
|
Efficient Nonparametric IV Estimation of Local Average Treatment Effects Using the Estimated Propensity Score and a Test for Unconfoundedness
|
|
Author(s):
|
Robert P. Lieli*+ and Stephen Donald and Hsu Yu-Chin
|
|
Companies:
|
The University of Texas at Austin and The University of Texas at Austin and The University of Texas at Austin
|
|
Address:
|
1 University Station C3100, Austin, TX, 78712,
|
|
Keywords:
|
treatment effects ; propensity score ; unconfoundedness ; instrumental variables ; nonparametric estimation
|
|
Abstract:
|
We study the problem of nonparametrically identifying and estimating local average treatment effects in the presence of covariates. Given a variable satisfying standard instrumental variable assumptions conditional on observables, we propose an efficient estimator that incorporates covariates by reweighting treatment and control outcomes by the estimated propensity score. The estimator is analogous to the estimator of average treatment effects developed by Hirano, Imbens and Ridder (2003) under the unconfoundedness (selection on observables) assumption. We show that under what is called the one-sided non-compliance assumption by Froelich and Melly (2008), local average treatment effect for the treated coincides with average treatment effect for the treated. This result allows us to construct a Hausman-type test for the unconfoundedness assumption by comparing the HIR estimator with ours.
|
