A0221
Title: Asymptotic analysis and efficiency of large dynamic panel models
Authors: Jose Diogo Barbosa - University of Michigan (United States) [presenting]
Marcelo J Moreira - FGV (Brazil)
Abstract: The efficient estimation of a dynamic panel model with fixed effects, time-series heteroskedasticity, and covariates is considered. This model allows natural rotational invariance conditions, such as re-ordering the individuals in the sample. An invariance principle determines a maximal invariant statistic. Its distribution yields an estimator for the structural parameters that is consistent as the number of cross-section and time-series observations increase with the sample size. Most estimators for this model are simply GMM estimators. Their respective moment conditions are found to be partial linear combinations of the first moment of the maximal invariant statistic. Consequently, this general framework allows us to obtain efficiency gains for these estimators in this simple dynamic panel data model.
