NOTES ON ESTIMATION OF THE TOBIT MODELS BY POWELL'S LEAST ABSOLUTE DEVIATIONS ESTIMATOR

抄録

In recent economic studies, Tobit models, in which the dependent variables cannot be negative, have been widely used. Although these models are usually estimated by the Tobit Maximum Likelihood Estimator (Tobit MLE), the Tobit MLE is not robust with respect to heteroskedasticity and non-normality of the error terms.
Powell (1984) proposed a modified least absolute deviations estimator which is consistent under heteroskedasticity and non-normality. One of the major problems with Powell's estimator is its computational difficulty. Nawata (1992) proposed a new algorithm which makes possible to calculate Powell's estimator. However, Nawata's method is incomplete and sometimes gives wrong results. In this paper, I modify the algorithm and evaluate Powell's estimator by the Monte Carlo experiments styled after Paarsch (1984).