Abstract
This article explores the diversity of 66 French competitiveness clusters, which were all accredited in 2005 according to the same specifications, by characterizing the initial context in which they emerged and taking a close look at the link between this initial context and their current performance. Since French competitiveness cluster policy is based on state co-funding of R&D projects, we establish a typology based on a multiple component analysis and a hierarchical ascending classification of the R&D potential of the cluster's territory, the respective R&D efforts of companies and academic laboratories, the kinds of actors setting up the cluster and their pre-existing relationships. We then measure the differences among the five classes relating to their clusters' capacity to obtain state funding for their projects. Our results show that initial context can partially explain competitiveness clusters' performance. Competitiveness clusters in territories possessing significant R&D resources, and involving large companies capable of organizing projects, are the most efficient in obtaining state funding. In contrast, competitiveness clusters without prior cooperation experience perform poorly.
Notes
1. We use the term cluster to designate the general concept of territorialized networks. We use the term competitiveness clusters when referring to a French example.
2. Five more clusters were selected in 2007, six saw their accreditation removed in 2010 after a national assessment of the cluster policy, while six more (dedicated to green technologies) were added in 2010.
3. The scree test and the Kaiser's rule enable us to choose the first four axes to make our analysis.
4. The distances between the clusters are Euclidian. The method of classification used is the ward criterion.
5. It could have been interesting to enumerate the specific cluster initiatives that fall into each of the five classes. However, because of a non-disclosure agreement, we are not able to cite any cluster names in this paper.
6. These clusters will probably have research partners located outside their territory.
7. The results of this test are compared with those of Bonferroni and Scheffe tests. They are similar, which confirms their validity.