Ökonometrische Modelle mit raumstruktureller Autokorrelation – Eine kurze Einführung / Econometric Models with Spatial Autocorrelation – An Introductory Survey

Cross section data originating from neighboring spatial units - frequently used in Regional Sciences - is not only affected by heteroscedasticity but is often also mutually dependent. This dependency can be taken into account by using models explicitly designed with spatially autocorrelated error terms or dependent variables. Due to lacking familiarity this field of Spatial Econometrics is often neglected. In order to improve this situation, the paper is not restricted to a survey about the most common spatial autocorrelation models: After a short introduction about the general inclusion of space into econometrics and how the concept of spatial autocorrelation works, possible consequences of not taking the interdependence structure into account are shown as a novel feature. After that, useful test statistics are mentioned and their general properties are discussed. Furthermore it is shown that OLS is not suitable when estimating such autocorrelation processes and that ML may cause computional burden. A favorable GMM methodology is therefore presented. The explanation of these topics is complemented by some Monte Carlo studies which examine the small sample properties of the tools mentioned.