A NOTE ON THE EFFICIENT SEMIPARAMETRIC ESTIMATION OF SOME EXPONENTIALPANEL MODELS

This paper investigates the semiparametric efficiency of the condition al maximum likelihood estimation in some panel models. The nonparametr ic component of the model is the unknown distribution of the fixed eff ect. For the exponential panel model, there exists a complete sufficie nt statistic for the fixed effect. When the complete sufficient statis tic does not depend on the parameter of interest, the conditional maxi mum likelihood estimator (CMLE) achieves the semiparametric efficiency bound. In particular, the CMLE is semiparametrically efficient for th e panel Poisson regression model and the panel negative binomial model .