In the last 20 years spatial econometric models, methods and techniques have been applied to a great variety of empirical problems. The essence of a spatial econometric model is the incorporation of a spatial autoregressive lag, which is scaled by the so called spatial autocorrelation parameter. From a mathematical perspective introducing a spatial autoregressive term into the linear regression model yields a system of equations, which may or may not be solvable for the dependent variable. Furthermore even if the system of equations is solvable, the dependent variable may be diverging if the number of observations approaches infinity. One can show that the solvability and boundedness of the dependent variable in spatial autoregessive models are crucially dependent on the (pre-) specified parameter space of the spatial autocorrelation parameter. Since almost all theoretical work in spatial econometrics assumes both model properties, the validity of spatial econometric methods and techniques is also crucially dependent on the (pre-) specified parameter space. (author's abstract)