Spatial autoregressive models come with a variety of estimators and it is interesting and useful to compare the estimators by location and covariance properties. In this paper, we first study the local sensitivity behavior of the main least squares estimator by using matrix derivatives. We then c… alculate the Taylor approximation of the least squares estimator in the SAR model up to the second order. Also, we compare the estimators of the spatial autoregression (SAR) model in terms of the covariance structure of the least squares estimators and we make efficiency comparisons using Kantorovich inequalities. Finally, we demonstrate our approach by an example for GDP and employment in 239 European NUTS2 regions. We find a quite good approximation behavior of the SAR estimator in the neighborhood of ρ = 0, i.e. a small spatial correlation.