Estimation of dynamic panel data models with a lot of heterogeneity

The commonly used 1-step and 2-step System GMM estimators for the panel AR(1) model are inconsistent under mean stationarity when ratio of variance individual effects to idiosyncratic errors is unbounded N→∞. reason their inconsistency that weight matrices select moment conditions do not identify autoregressive parameter. This paper proposes a new estimator still consistent in this case provided T>3. Unlike estimator, uses an optimal matrix remains case. We also show Arellano-Bond Random Effects Quasi MLE remain same conditions. To illustrate usefulness our we revisit growth study Levine et al. (2000 Levine, R., Loayza, N., Beck, T. (2000). Financial intermediation growth: causality causes. Journal Monetary Economics 46(1):31–77. doi:https://doi.org/10.1016/S0304-3932(00)00017-9[Crossref], [Web Science ®] , [Google Scholar]).