Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This paper studies the GMM estimation of short panel data models with interactive fixed effects. We demonstrate that the nonlinear moment conditions proposed by Ahn et al. (2001, 2013) do not always satisfy the global identification assumption, which is necessary for consistency of the GMM estimation. Some numerical examples are provided to confirm this claim. We also demonstrate that the same problem occurs for moment conditions proposed by Hayakawa (2012) and Robertson and Sarafidis (2015), since their moment conditions become identical to those of Ahn et al. (2001, 2013) in some cases. Finally, we conduct Monte Carlo simulations and show that the starting value used in the computation of non-linear GMM estimators has a significant effect on performance.