Estimating censored regression models in the presence of nonparametric multiplicative heteroskedasticity

Powell's (1984, Journal of Econometrics 25, 303-325) censored least absolut e deviations (CLAD) estimator for the censored linear regression model has been regarded as a desirable alternative to maximum likelihood estimation m ethods due to its robustness to conditional heteroskedasticity and distribu tional misspecification of the error term. However, the CLAD estimation pro cedure has failed in certain empirical applications due to the restrictive nature of the 'full rank' condition it requires. This condition can be espe cially problematic when the data are heavily censored. In this paper we int roduce estimation procedures for heteroskedastic censored linear regression models with a much weaker identification restriction than that required fo r the LCAD, and which are flexible enough to allow for various degrees of c ensoring. The new estimators are shown to have desirable asymptotic propert ies and perform well in small-scale simulation studies, and can thus be con sidered as viable alternatives for estimating censored regression models, e specially for applications in which the CLAD fails. (C) 2000 Elsevier Scien ce S.A. All rights reserved. JEL classification: C14; C23; C24.