A Comment on GMM Estimation in IS Research

다국어 초록 (Multilingual Abstract)

As a quick econometric solution to handle potential endogeneity issues in panel data models, the Generalized Method of Moments (GMM) estimator is gaining popularity in IS research. Despite the sensitivity of this estimator to model specifications and estimation strategies, a noticeable number of IS studies employing this method fail to report the detailed model specifications, robustness check results with different model specifications and estimation strategies, or test statistics, which render their empirical results less credible. We demonstrate, based on the dataset used by Arellano and Bond (1991), that passing the commonly required tests such as the m2 test and the Sargan-Hansen test does not guarantee validity of the estimate, because the size and the statistical significance of the estimate can largely depend on the choice of estimation procedure and possible moment restrictions that pass such required tests. We urge researchers not only to report the results of significant focal variables, but also to be explicit about the model specifications and estimation strategies, and to provide robustness checks with different model specifications, along with their complete test results.