A new study explores how research collaboration networks shape scientific productivity
How does collaboration influence research productivity, and to what extent does scientific performance help universities attract new partners? A new open-access article by Antonio Zinilli, Lorenzo Giammei and Emanuela Varinetti, published in Scientometrics, investigates these questions using Bayesian networks and data from the RISIS research infrastructure.
The study, “Collaborative structures and research productivity: an empirical analysis with Bayesian networks” (2026), examines European universities over the period 2016–2021, focusing on two scientific domains: Life Sciences, and Physics and Engineering. It combines data on scientific publications and EU-funded research projects to explore how different forms of collaboration relate to research output.
The findings highlight the importance of collaboration networks. In Physics and Engineering, the number of direct collaborative ties between universities consistently emerges as a key factor associated with research productivity. In Life Sciences, the relationship is more variable: the breadth of collaboration networks becomes increasingly relevant over time, while universities’ positions within knowledge networks also matter in the earlier years examined.
The analysis also reveals differences between publication-based and project-based collaborations. In Physics and Engineering, higher research productivity can strengthen universities’ ability to attract business partners in EU-funded projects. In Life Sciences, productive universities tend instead to reinforce academic collaborations. These patterns underline the importance of considering disciplinary contexts and different types of partnerships when designing strategies to support scientific performance.
Methodologically, the study uses Bayesian networks to investigate complex relationships among institutional characteristics, collaboration structures and research productivity, offering a more nuanced perspective than approaches that assume a simple, one-way relationship between collaboration and scientific output.
The contribution of RISIS data is central to this analysis. The research draws on three RISIS infrastructure resources: EUPRO, for European research projects; CWTS bibliometric data, for scientific publications; and ETER, the European Tertiary Education Register. Their integration enables the researchers to connect institutional characteristics with university collaboration networks and scientific production.
The findings offer relevant insights for research policy and institutional strategy, highlighting the value of fostering collaborative networks while recognising that effective approaches may differ across scientific fields and partnership types.
Read the open-access article:
Zinilli, A., Giammei, L., & Varinetti, E. (2026). Collaborative structures and research productivity: an empirical analysis with Bayesian networks. Scientometrics, 131, 1209–1236.
