Combining Bias Correction Methods:
Does Basing Indirect Inference on Han and Phillips
Yield Better Small Sample Results?
I propose a new bias correction method for dynamic panel data models with fixed effects:
that of Indirect Inference (II) with the Han and Phillips bias-corrector (HP) as the base
estimator. I compare the performance of this new approach with that of several known
bias correction methods. The results are strong. The new estimator of the auto-regressive
coefficient outperforms both the HP bias corrector and Maximum Likelihood based II in
terms of bias reduction, while also exhibiting a significantly lower root mean square error.
The relative benefit of the new estimator increases for small N, T, and also increases as the
auto-regressive coefficient approaches 1.
Keywords: Bias reduction, Indirect Inference, Dynamic Panel