Scientific collaboration and European Framework Programmes: a novel way to constructing scientific collaboration networks

R&D networks are considered as conduits of knowledge between organizations and therefore as crucial for the creation and recombination of new knowledge within innovation systems. Empirical analyses of the structure of R&D networks, especially those related with public collaboration programmes, mostly rely on project-level data. These analyses usually rest on the (heroic) assumption that all organizations within a project mutually collaborate at the same strength, thus assume fully connected project graphs. Qualitative empirical research, however, suggests that the intensity of scientific collaboration among the partners in a project can vary considerably. Therefore, one would expect that existing networks in publicly induced R&D should not be interpreted as analyses of strong scientific collaboration but merely as networks of weak interaction. This study explores novel construction techniques for strong scientific collaboration networks using information beyond project-level data. Two alternative regression approaches are employed for determining refined construction rules. The first set of rules is derived from a logit estimation of determinants of strong scientific collaboration obtained at the level of collaborating organization pairs. The second set of rules is derived at the level of corresponding region pairs. A model of the gravity type is used to study spatial separation effects of cross-region strong collaboration determinants. Hereby, spatial autocorrelation among flows of collaboration is taken into account by using Eigenfunction spatial filtering methods. Using these two sets of rules, corresponding R&D networks are constructed and compared with standard social network indicators. Data used refer to the European Framework Programme, and are taken from the EUPRO database, a survey of programme participants (both courtesy AIT), as well as regional databases.