A commonly used estimation procedure for dynamic panel data models is the Generalised Method of Moments (GMM). This estimation method results in consistent parameter estimates in a wide variety of settings and properties of the data generating processes. A major problem with this method is that inference using estimated asymptotic standard errors can be very unreliable in small samples for the efficient version of the GMM estimator, because that the estimated standard errors are downward biased. Research has investigated the finite sample properties of alternative testing procedures, including bootstrap methods. Inference using bootstrap methods for the efficient GMM estimator is shown to behave poorly in a wide variety of setting. A finite sample correction to the GMM variance estimate has been derived. Inference based on this corrected variance has been shown to improve the performance of the Wald test dramatically. The results of the research show that the recommended tests to use are the corrected Wald test, the bootstrap one-step Wald test, the LM test and a simple criterion-based test.