to Tobit-type selection rules

A semlparametruz two-stage estimation method 1s proposed for the estimation of sample selectlon models which are subject to Toblt-type selectlon rules. With randomlzatlon restrIctIons on the disturbances of the model, all the regresslon coefficients m the model are, in general, ldentlfiable wIthout exclusion restrIctIons The proposed estimator is shown to be X/n-consistent and asymp-totlcally normal. Some Monte Carlo results, to demonstrate Its limte sample performance, are provided. Key wwrd.s Sample selection; Truncation; Censormg; SemIparametrIc estlmatlon. Kernel estlmatlon; Randomlzatlon, RegressIon mode