Estimation of Multiequation Cross-Section Models in the Presence of Spatial Autocorrelation

We describe econometric techniques to treat spatial autocorrelation in multiequationcross-section models. The cross-section approaches discussed here are heavily based onthe spatial GMM procedure, proposed by Conley (1999). An extension for fullinformationinstrumental variable models is presented. Monte Carlo simulations areemployed in order to verify some asymptotic properties of the Spatial GMM approach.The simulations suggest that, even in the presence of spatial nonstationarity, the spatialGMM still delivers valid standard errors. Besides, usual t-statistics appear to have astandard normal distribution. An application for estimating labor and wage equationsto study regional growth and development of the Brazilian municipalities, between1991 and 2000, is presented.