Censored Regression with Heteroskedasticity

• Activity Number: 617
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Hieu Nguyen*+ and James McDonald
• Companies: Brigham Young University and Brigham Young University
• Keywords: censored regression ; Tobit ; CLAD ; SCLS ; partially adaptive estimators

Abstract

Data censoring causes ordinary least squares estimates of linear models to be biased and inconsistent. The Tobit model circumvents this problem when the errors are normally distributed. However, the presence of either heteroskedasticity or non-normal errors results in the Tobit estimators being biased and inconsistent. Partially adaptive estimators have been shown to have the potential to significantly improve estimator performance for over a wide range of distributional characteristics. This paper explores modifications of previously considered partially adaptive estimators to also accommodate simple forms of heteroskedasticity. A simulation study is used to investigate the estimators' relative efficiency in these settings. Monte Carlo simulations suggest that the heteroskedastic partially adaptive censored regression estimators have little efficiency loss for censored normal errors with homoskedasticity and yield significant improvements with heteroskedasticity for normal and nonnormal errors. An empirical example provides an example, which supports these results.