Dynamic panel data models are increasingly and extensively used in operational research and performance analysis as researchers seek to better understand the dynamic behaviors of firms. However, estimation of the lagged dependent variable in conjunction with the time-invariant individual effect leads to a number of econometric issues. While several methodologies exist to overcome such complexities, there is little consensus on the appropriate method of estimation. In this paper, we evaluate the performance of different dynamic panel estimators across a range of common settings experienced by researchers. Instead of focusing on one single criterion of assessment, we employ multiple evaluative metrics across multiple experiments to provide a more extensive analysis of dynamic panel estimators. Taking all simulations into account, we find the quasi-maximum likelihood estimator to be the most robust and reliable estimator across empirical settings. We illustrate our findings with two empirical applications and show that the choice of estimator significantly affects the interpretation of firms’ productivity and efficiency persistence.
Read full abstract