Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/85689 
Year of Publication: 
1999
Series/Report no.: 
Tinbergen Institute Discussion Paper No. 98-131/3
Publisher: 
Tinbergen Institute, Amsterdam and Rotterdam
Abstract: 
Innovation research has become an important topic in regional science analysis. Yet the modelling base of much innovation research is still feeble. This paper aims to map out the research potential of recent approaches in quantitative complexity analysis, in particular Neural Networks (NNs) analysis, from the perspective of their operational applicability in the space-economy. The urban context of European innovation processes is used as an empirical background. The paper addresses also the issue of space-time transferability of the tools employed.The first part of the paper is devoted to a concise conceptual overview and illustration of the innovation process, which is conceived of as a self-organising system. The second part presents empirical results on innovation processes in Europe. In this framework a comparative analysis is conducted between NN models and a conventional tool often used in spatial economics studies, viz. (non)linear regression analysis. The sensitivity of the various results, - by using 'transferability' experiments - is also examined. Theempirical experiments underline the advantages and limitations of these approaches from a methodological as well as an empirical viewpoint. They appear to offer a plausible range of values of empirical outcomes, which may highlight an acceptable degree of variation in spatial innovation processes.
Document Type: 
Working Paper

Files in This Item:
File
Size
94.91 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.