摘要：Type I (censored regression) and Type II Tobit (sample selection) models are widely used in the various fields of economics. The Type I Tobit model is a special case of the Type II Tobit model. However, the dimension of the error terms decreases and the distribution of the error terms degenerates in the Type I Tobit Model. Therefore, we cannot use the standard asymptotic theorems for the Type II Tobit Maximum Likelihood Estimator (MLE) when the sample is obtained from the Type I Tobit model. Results of Monte Carlo experiments show strange behavior that has never been reported before for the Type II MLE.
