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.