46th Congress of the European Regional Science Association: "Enlargement, Southern Europe and the Mediterranean", August 30th - September 3rd, 2006, Volos, Greece
This paper investigates growth differences in the urban system of the EU12 between the means of 1978/80 and 1992/94 for a data set relating to Functional Urban Regions rather than the more normal NUTS regions comparing the results of 'artisanal' methods of model selection with those generated using general to specific model selection (using PcGets). The artisanal approach tests hypotheses relating to the role of human capital, EU integration and fragmentation of urban government. The paper also explores issues of spatial dependence and mechanisms of spatial interaction. Using PcGets as suggested by Hendry and Krolzig (2004) to optimise model selection yields a model virtually identical to the artisanal approach if mechanisms of spatial interaction are ignored. Testing, however, reveals problems of spatial dependence. We interpret this as indicating that significant variables reflecting mechanisms of spatial economic adjustment have been omitted. Including such variables in the set available to PcGets leads to the inclusion of two measures of spatial adjustment. Further testing shows that problems of spatial dependence are now eliminated. We interpret this result as evidence that while PcGets provides a powerful tool for model selection when applied to cross sectional data, caution is necessary to ensure that variables relating to spatial adjustment processes are included and spatial dependence is avoided. Not only do the final results provide consistent estimates of parameters but they also support relevant theoretical insights. Moreover careful testing for spatial dependence reveals that national borders are still significant barriers to adjustment within the EU.