In this paper, I consider the estimation of non-linear panel data models with Þxed effects. I use the ModiÞed Maximum Likelihood Estimator (MMLE) that reduces the order of the bias from O(T −1 ) in the Maximum Likelihood Estimator to O(T −2 ). I evaluate its performance in Þnite samples where T is not large, using Monte Carlo simulations. In Probit and Logit models with lags of the endogenous variable and exogenous variables, the estimator is found to have small bias in a panel with eight

more »

... s. Other advantage of the MMLE is its generality. Some issues about the measure of interest that arise in this kind of models are also addressed. In contrast with linear models, the effect of interest is different for each individual and depends on the Þxed effects. Furthermore, the mean effect across all individuals may not be the relevant measure but the whole distribution may be needed. Compared with simple MLE, simulation results show that MMLE improves signiÞcantly the estimation of the distribution of the effect of interest.