Improved GMM estimation of random effects panel data models with spatially correlated error components

We modify a previously suggested GMM estimator in a spatial panel regression model, which has recently received considerable interest in empirical applications, by taking into account the difference between disturbances and regression residuals. Consistency and asymptotic normality of the estimator are derived. Analytic results, simulation evidence and an empirical application to Indonesian rice data illustrate the improvement in finite samples.