Generalised Method of Moments (GMM) estimators for dynamic panel data models with unobserved individual effects are widely used in applied econometric research. One problem with these estimators is that the asymptotic standard errors for the efficient version of the GMM estimators are downward biased, even for moderately large sample sizes. Statistical inference and hypothesis testing based on these asymptotic variances can therefore be seriously misleading. This project will investigate whether the use of bootstrap methods will allow more reliable inferences to be obtained for the efficient GMM estimators. We will also consider whether bootstrap techniques can be used to reduce the small bias of these estimators. Bootstrap methods use repeated re-sampling from the data to form an estimate of the small sample distribution of a random variable. We propose to conduct a comprehensive Monte Carlo analysis to evaluate these gains for a wide range of data generating processes and alternative GMM estimators. If we confirm that bootstrap methods allow more reliable small sample inference and/or bias reduction in our study, we will make these methods widely available to applied researchers by incorporating bootstrap procedures in the popular DPD programme for the estimation of dynamic panel data models