Spatial data are characterized by spatial interdependence and heterogeneity which give a variety of measurement problems. Heterogeneity can be covered by some econometrical techniques for model with non-spherical error term. Though spatial interdependence is considered as a “statistical noise” in the field of econometrics to date, how to model spatial interaction is the main subject in regional modeling. However, most of empirical works fails to take into account the interdependence effects in estimation or tests.
In this paper, we consider a family of simultaneous spatial models which explicitly contain spatial interdependence. The most popular OLS estimator is neither unbiased nor consistent. We propose a new consistent and efficient estimator, that is named Two Stage Generalized Least Squares (2SGLS) Estimator. OLS bias, efficiency of 2SGLS and 2SLS estimator are numerically assessed. We conclude by discussing the applicability of 2SLS and 2SGLS estimator.
