The Accuracy of Inference in Small Samples of Dynamic Panel Data Models

Through Monte Carlo experiments the small sample behavior is examinedof various inference techniques for dynamic panel data models whenboth the time-series and cross-section dimensions of the data set aresmall. The LSDV technique and corrected versions of it are comparedwith IV and GMM regarding: coefficient bias, accuracy of varianceestimators - both of the disturbances and of the coefficientestimators - and the actual size of coefficient tests. A reasonablysimple and consistent bias adjusted LSDV estimator, for which we findan analytical and a bootstrap consistent estimator of its variance,performs relatively well. Further higher-order refinements of thebias correction do not improve the accuracy considerably. Mosttechniques show substantial size distortions for asymptotic t tests.Finally, it is illustrated how these findings help to interpretempirical results on the relationship between so-called dynamicexternalities and local economic activity in Moroccan urban areas.