This paper presents a simple two step estimator for models with censor
ed endogenous regressors and sample selection bias. The approach unifi
es the literature on censored endogenous regressors and sample selecti
on bias and provides extensions. The procedure employs generalized res
iduals to adjust for the inconsistency caused by the endogeneity of th
e censored regressors. The paper also provides two new tests of weak e
xogeneity. An empirical example, examining the trade off between fring
e benefits and wages, indicates that the estimation procedure works we
ll. A monte carlo experiment suggests that the proposed tests have goo
d power properties.