GMM Estimation of short dynamic panel data models with error cross-sectional dependence

This paper considers estimation of short dynamic panel data models with error cross-sectional dependence. It is shown that under spatially correlated errors, an additional, generally non-redundant, set of moment conditions becomes available for each i - specifically, instruments with respect to the individual(s) which unit i is spatially correlated with. We demonstrate that these moment conditions remain valid when the error term contains a common factor component, in which situation the standard moment conditions with respect to individual i itself are invalidated, and thereby the standard dynamic panel GMM estimators are inconsistent. The resulting estimators are computationally attractive and do not require estimating the number of unobserved factors. Simulated experiments show that the resulting method of moments estimators perform well in terms of both median bias and root median square error.