Estimation of a censored regression panel data model using conditional moment restrictions efficiently

A new semiparametric estimator for the censored regression panel data model with fixed effects is introduced. It is based upon an estimator proposed by Honore for the case of two time periods combined with ideas of Newey to improve the efficiency. The estimator is more efficient than Honore's and is generalized to the case of a balanced or unbalanced panel of more than two waves. Estimation is performed in two steps. Using Honore's estimator in a first step, efficient GMM using conditional moment restrictions is applied. The performance of this estimator is compared to that of Honore's and other existing estimators in an empirical example concerning labour income of married females, using panel data from the Dutch Socio-Economic Panel, 1984-1988. Attention is paid to specification testing and the sensitivity of the results for the choice of smoothness parameters.