Seite - (000182) - in Knowledge and Networks

176 In Table 9.1 some whole network measures are provided for each yearly collabo- ration network examined.4 The table shows two main waves of funded projects: an initial wave in 2006–2007 when six and seven projects were funded, respectively; and a second wave in 2013 when seven projects started. The number of organiza- tions involved in the collaboration network notably increased over time, going from 19 organizations in 2006 to 82 in 2013. The proportion of associated members to external partners also changed dramatically. From 2006 to 2011, most of the orga- nizations were local research institutions and firms formally engaged with the TD. In 2012 and 2013, the number of external partners exceeded that of associated members. However, the number of the external partners involved in the European R&D project since 2008 is much larger (50 organizations) and has been shrunk to only one representative node in order to simplify the analysis.5 Figure 9.1 shows the evolution of the collaboration network. Structural hole measures were used to highlight group clustering.6 These measures provide infor- mation on (i) dyadic constraint, which indicates the extent to which a single organi- zation bridges two cohesive groups; and (ii) associated constraint, which measures the degree of organizational embeddedness within a cohesive group. In the graphs, TD members with high aggregate constraint are drawn closely together, whereas low dyadic constraint is shown as longer links to highlight structural holes. We note that the one-mode projection of the data has generated perfect cliques in correspondence with joint participation in projects. However, clustering cannot be entirely attributed to this data distortion. Since the very beginning of the TD, a sub- set of organizations have collaborated on more than one project, suggesting the emergence of a substantially cohesive core. Furthermore, key brokers have charac- terized the network in each time period. In the start-up phase, only two actors within the TD bridged otherwise disconnected groups. From 2009 onward, however, the number of brokers notably increased. 4 We calculated the reported measures by disregarding isolates (the TD’s members not involved in projects in each year). 5 Given that only six members were involved in the EU project, we decided to subsume all the international partners in one representative node. In this way the pattern of collaborations internal to the TD remained stable, and at the same time the analyses and visualizations were simplified. If we were to count all the partners involved in the European project, the number of external partners collaborating with the TD since 2008 would increase by 50 units. 6 Pajek software for social network analysis (de Nooy, Mrvar, & Batagelj, 2011) was used for all analyses and visualizations in the study. L. Prota et al.