Estimating censored regression models in the presence of nonparametric multiplicative heteroskedasticity.

Powell's (1984, Journal of Econometrics 25, 303}325) censored least absolute deviations(CLAD) estimator for the censored linear regression model has been regarded asa desirable alternative to maximum likelihood estimation methods due to its robustnessto conditional heteroskedasticity and distributional misspeci"cation of the error term.However, the CLAD estimation procedure has failed in certain empirical applicationsdue to the restrictive nature of the &full rank' condition it requires. This condition can beespecially problematic when the data are heavily censored. In this paper we introduceestimation procedures for heteroskedastic censored linear regression models with a muchweaker identi"cation restriction than that required for the LCAD, and which are #exibleenough to allow for various degrees of censoring. The new estimators are shown to havedesirable asymptotic properties and perform well in small-scale simulation studies, andcan thus be considered as viable alternatives for estimating censored regression models,especially for applications in which the CLAD fails