Économétrie spatiale : l'autocorrélation spatiale dans les modèles de régression linéaire

The aim of this article is to describe the tools necessary to factor in automatic spatial correlation determined by the absence of independence between geographical observations, within the framework of linear regression models. While it is often accepted that spatial data in cross-sections are independent, this assumption is rarely justified. On the contrary, it needs to be tested systematically. After explaining the ways in which to model automatic spatial correlation, we therefore describe estimation and inference procedures geared to econometric models that factor in this effect explicitly. Lastly, we propose a general approach to test and factor in automatic spatial correlation in empirical work.