This paper provides a survey and critique of how spatial links are taken into account inempirical analysis by applied economists/regional scientists. Spatial spillovers and spatialinterrelationships between economic variables (e.g. unemployment, GDP, etc) are likely to beimportant, especially because of the role of local knowledge diffusion and how trade (interregionalexports and imports) can potentially act to diffuse technology. Since most empiricaleconomic studies ignore spatial autocorrelation they are thus potentially mis-specified. Thishas led to various approaches to taking account of spatial spillovers, including econometricmodels that dependent on specifying (correctly) the spatial weights matrix, W. The paperdiscusses the standard approaches (e.g., contiguity and distance measures) in constructing W,and the implications of using such approaches in terms of the potential mis-specification ofW. We then look at more recent attempts to measure W in the literature, including: Bayesian(searching for 'best fit'); non-parametric techniques; the use of spatial correlation to estimateW; and other iteration techniques. The paper then considers alternative approaches forincluding spatial spillovers in econometric models such as: constructing (weighted) spillovervariables which directly enter the model; allowing non-contiguous spatial variables to enterthe model; and the use of spatial VAR models. Lastly, we discuss the likely form of spatialspillovers and therefore whether the standard approach to measuring W is likely to besufficient.
