Title: Second-order analytic bias reduction for nonlinear panel data models with fixed effects
Authors: Martin Schumann - TU Dortmund (Germany) [presenting]
Abstract: One of the most useful features of panel data is that it allows researchers to control for time-invariant individual heterogeneity that is not observed. However, in nonlinear panel data models with fixed effects, the maximum likelihood estimator can be severely biased due to the incidental parameters problem. While in the recent literature methods have been proposed that yield a first-order bias reduction relative to maximum likelihood,simulation results based on short panels suggest the need for higher-order bias reduction in order to further improve the small sample performances of these methods. Explicit expressions for the second-order biases of the profile likelihood and its score are provided. It is further shown that estimation of the first-order bias based on plug-in estimators creates an additional bias that contributes to the second-order bias. Finally, an estimator is constructed that corrects both the first and the second-order bias of the maximum likelihood estimator.