Spatial dependence and the representation of space in empirical models

A well-formed spatial model should most likely not produce spatial autocorrelation at all. From this perspective spatial autocorrelation is not (pure) statistical nuisance but a sign of that a model lacks a representation of an important economic phenomenon. In a Knowledge Production Function (KPF) context, this paper shows that a representation of space reflecting the potential of physical interaction between localities by means of accessibility variables on the "right-hand-side"aEuro"a simple alternative to spatial lag and spatial error which can be estimated by OLS-captures substantive spatial dependence. Results are verified with Monte Carlo simulations based on Anselin's (Int Reg Sci Rev 26(2):153-166, 2003) taxonomy of modelled and unmodelled effects. The analysis demonstrates that an accessibility representation of explanatory variables depict the network nature of spatial interaction, such that spatial dependence is actually modelled.