Further Results on the Weak Instruments Problem of the System GMM Estimator in Dynamic Panel Data Models

Abstract

AbstractIn this paper, we investigate the weak instruments problem of the generalized method of moments (GMM) estimator for dynamic panel data models. Specifically, we complement Bun and Windmeijer (2010) by considering the alternative first‐difference and level models transformed by the forward GLS transformation. We demonstrate that this transformation yields a higher concentration parameter compared with the original models. This indicates that the proposed transformation yields stronger instruments even though the instruments used are identical. The Monte Carlo simulation results show that the system GMM estimator for the transformed model, called the forward system GMM estimator, performs better than the conventional system GMM estimator.