An overarching concern in regional science is the characterization of interactions—such as commuter flows, transport, migration, or knowledge flows—within and between subnational spatial units. In this work, we use techniques from social network analysis to address the quality, rather than the quantity, of such interactions. Given the great current interest in European R&D networks, in which organizations from the science and the industry sector distributed across European regions perform joint R&D, we focus on interactions constituting knowledge flows in the European R&D network, as inferred from Framework Programme (FP) data. To assess a specific quality of these region-to-region interactions, we make use of the concept of edge betweenness centrality, which assesses the power of a relation based on the load placed on the corresponding network edge. Betweenness centrality is calculated using the geodesic paths between all distinct pairs of network vertices. Those vertices and edges required by relatively many of the paths thus often lie between other vertices; the fraction of the shortest paths on which an edge occurs is defined as the edge betweenness centrality. Edges with high betweenness centrality have the greatest load, are strategically positioned, and potentially can act as bottlenecks for the flows. We use this idea to evaluate knowledge flows between organizations in the European R&D network, considering several ways to relate the betweenness centrality at the level of FP project participants to knowledge flows at the NUTS2 regional level. We do so by aggregating betweenness centrality values calculated using bipartite graphs linking organizations to the FP projects in which they participate, considering annual FP data between the years 1999 and 2006. We determine the most central inter-regional knowledge flows, describe how this changes over time, and consider the implications for knowledge flows in European R&D networks. We model the centrality of the flows by means of spatial interaction models, estimating how geographical, technological, and social factors influence which region pairs become bottlenecks in the flow of knowledge. The results have meaningful implications to European R&D policy, in particular concerning which region pairs constitute the core in European R&D networks and which mechanisms drive the formation of this regional core.
