Abstract
Universities stand at the crossroads of Europe's innovation landscape. They are not merely sources of knowledge, but dynamic orchestrators of collaboration — institutions that shape how ideas travel, evolve, and connect across technological and organisational divides. This dissertation explores their structural role within European research and innovation networks, asking whether universities act primarily as engines of cohesion or as brokers of diversity.
Drawing on project-level data from the European Commission's CORDIS database, covering more than 35,000 collaborative R&D projects funded under Horizon 2020 and Horizon Europe, the study models inter-organisational relationships using an Exponential Random Graph Model (ERGM). This network-based approach detects how academic participation alters the architecture of collaboration: whether it fosters tightly knit, trust-based communities, or instead opens structural bridges linking distant sectors, regions, and technological domains.
The evidence reveals a nuanced pattern: universities do not simply stabilise or connect, they do both, strategically. This "dual embeddedness" challenges the traditional dichotomy between closure and brokerage, positioning universities as adaptive nodes that maintain cohesion while keeping innovation networks permeable to new ideas.
Research Question
Within European R&D collaboration networks, do universities act primarily as builders of cohesive, trust-based clusters, or as brokers connecting otherwise disconnected actors and sectors — or both simultaneously?
Key Contributions
Methodology
Project-level data from Horizon 2020 and Horizon Europe (35,000+ projects), projected into a one-mode organisation–organisation network of 59,247 nodes and 482,113 collaborative ties.
Models the probability of a tie as a function of endogenous structural configurations (edges, triangles, GWESP) and exogenous attributes (organisation type, country, funding intensity), estimated via MCMC-MLE and MPLE.
Given the computational infeasibility of full ERGM estimation on networks of this scale, the analysis focuses on the 1,000 highest-degree organisations (136,191 edges), with structural closure tested via GWESP and MPLE approximation.
Keywords
Interested in the full working paper, data, or methodology? Send a message and I'll get back to you directly.