Further results on instrumental variables estimation of average treatment effects in the correlated random coefficient model

Abstract

I show that standard instrumental variables estimators, applied to an equation containing levels and interactions, consistently estimate the average treatment effect under weaker conditions than those needed for the two-stage plug-in estimators proposed by Heckman and Vytlacil [Journal of Human Resources 33 (1998) 974]. © 2002 Elsevier Science B.V. All rights reserved.