Bias Correction for Fixed Effects Spatial Panel Data Models

This paper examines the finite sample properties of the quasi maximum likelihood (QML) esti-mators of the fixed effects spatial panel data (FE-SPD) models of Lee and Yu (2010). Following the general bias correction methods recently developed by Yang (2015), we derive up to third-order bias corrections for the QML estimators of the FE-SPD model, and propose a simple bootstrap method for their practical implementation. Monte Carlo results reveal that the QML estimators of the spa-tial parameters can be quite biased and that a second-order bias correction effectively removes the bias. The validity of the bootstrap method is established. Variance corrections are also considered, which together with bias corrections lead to improved inferences