Page - (000179) - in Knowledge and Networks

173 community, the knowledge network was highly centralized, producing a selective and uneven environment. At the core of the knowledge network sat the firms with the strongest knowledge bases. Furthermore, Balland, Suire, and Vicente (2010) distinguished the effect of geo- graphical proximity and knowledge bases by project phases such as in a knowledge value chain approach. In this very thought-provoking study, two-mode data on the Global Navigation Satellite System in Europe (GNASS) were used to define two network projections. The first network projection, describing project-to-project relations, was used to divide projects into three phases of the knowledge value chain: exploration, integration, exploitation. The second network, showing an organization-to- organization network, was used to distinguish firms by their knowl- edge bases: synthetic, analytic, and symbolic. Finally, a blockmodel was used to find cohesive local clusters and pipelines linking these clusters to global partners. Results from social network analysis showed that the project network indeed took a core–periphery configuration, the core consisting of exploitative projects close to the marketing phase and with a high concentration of pipelines. By contrast, local embeddedness dominated the periphery where exploratory projects required trust to exchange sensitive information and reach closure. The relation between the core and the periphery of an innovation network was further discussed by Glückler (2014) with an analysis of BASF’s cross- departmental knowledge flow linking the center to a peripheral unit in Argentina. The study dem- onstrates that the periphery can become a particularly suitable location to develop controversial innovations due to its organizational features. The unit examined was able not only to capitalize on its local market connections to develop a new business model but also to exploit the global organizational viscosity of BASF by establish- ing contacts with distant units. From this literature it emerges that geographical and cognitive proximities as well as ownership can set in motion underlying processes that ultimately produce core–periphery structures. The extent to which an observed network approximates a core–periphery model can be captured through blockmodeling. This method is well established in social network analysis and has been used since the 1970s to reduce a network to its key topological features (Doreian et al., 2005). In this chap- ter, we use a variant of blockmodeling that permits prespecifying a theoretical core– periphery model and then using it as a benchmark against which to measure the configuration of the observed project networks over time. The temporal dimension is particularly important in this study because it allows one to understand the extent to which the district development trajectory was path dependent (Martin & Sunley, 2006). The presence of a core–periphery structure in a specific moment of a cluster life does not necessarily mean that the structure will persist in the future. In this study we consider the emergence of a core–periphery structure as an indication that an underlying process of preferential attachment is taking place. This process implies that new ties are more likely to form around already central actors, reinforcing the cohesiveness of the core against the periphery (Glückler, 2007). However, this correspondence between topologies and processes persists only if the processes are protracted. 9 Topology and Evolution of Collaboration Networks