Spatial Dependence in linear Regression Models with an Introduction to Spatial Econometrics

Abstract

Econometric theory and practice have been dominated a focus on the time dimension. In stark contrast to the voluminous literature on serial dependence over time (e.g., the extensive review in King 1987), there is scant attention paid to its counterpart in cross-sectional data, spatial autocorrelation. For example, there is no referenee to the coneept nor to its relevanee in estimation or specification testing in any of the eommonly eited eeonometrics texts, sueh as Judge et al. (1982), Greene (1993), or Poirier (1995), or even in more advanced ones, sueh as Fomby et al. (1984), Amemiya (1985), Judge et al. (1995), and Davidson and MacKinnon (1993) (a rare exception is Johnston 1984). In contrast, spatial autocorrelation and spatial statistics in general are widely accepted as highly relevant in the analysis of cross-sectional data in the physical sciences, such as in statistieal mechanics, ecology, forestry, geology, soil science, medical imaging, and epidemiology (for a recent review, see National Researeh Counci\1991).