Determinants of Collaboration in European R&D Networks: Empirical Evidence from a Discrete Choice Model

Abstract

This paper focuses on inter-organizational R&D collaborations as captured by joint research projects funded within the European Framework Programmes for Research and Technological Development (EU-FPs). We identify determinants of collaboration, including actor characteristics, relational and network effects as well as geographical effects by means of a discrete choice model. Using data on EU-FP projects from the EUPRO database and from a representative survey of participants, we produce statistically significant evidence that collaboration choices in EU-FPs are primarily facilitated by prior acquaintance, thematic proximity and geographical proximity. Also, network effects are significantly related to collaboration choice. Moreover, the study shows that the impact of geographical effects increases for more intensive collaboration. The results are promising since a deeper understanding of these collaboration processes is needed for future governance of research policies in the EU.

Keywords:

• R&D collaboration
• network formation
• EU Framework Programmes
• discrete choice model
• ordered logistic regression

Acknowledgements

The work presented here was partly funded by the EU-FP6-NEST project NEMO (“Network Models, Governance and R&D Collaboration Networks”), contract no. 028875, and by the Austrian Science Fund (FWF), grant no. P21450. Furthermore, the authors wish to thank two anonymous referees for their valuable comments and suggestions.

Notes

¹ Since the start of FP1 (1984–88), more than 60,000 collaborative R&D projects have been funded in the subsequent FPs. Currently, FP7 (2007–13) is under way. In spite of slight changes in scope, the fundamental rationale of the EU-FPs has remained unchanged over time. All FPs share a few common structural key elements: first, only projects of limited duration that mobilize private and public funds at the national level are funded. Second, the focus of funding is on multinational and multi-actor collaborations that add value by operating at the European level. Thus, the projects require partners from at least a certain number of different eligible countries. Third, project proposals are submitted by self-organized consortia and the selection for funding is based on specific scientific excellence and socio-economic relevance criteria. See Roediger-Schluga and Barber (2008) for a detailed discussion on the history and different scopes of the EU-FPs since 1984.

² The EUPRO database is constructed and maintained by AIT Austrian Institute of Technology GmbH by substantially standardizing raw data on EU-FP research collaborations obtained from the CORDIS database (Roediger-Schluga and Barber, 2008).

³ This survey was conducted in 2007 by the Austrian Research Centers GmbH, Vienna, Austria and operated by b-wise GmbH, Karlsruhe, Germany.

⁴ FP5 (1998–2002) was chosen for the survey, in order to cover prior as well as subsequent bilateral collaborations, and effects of the collaboration both with respect to scientific and commercial outcomes. Thus, the survey is able to complement the EUPRO database.

⁵ We use the reported bilateral collaborations of the respondents (1,173 organizations collaborating in 1,089 projects). We draw on a subset of these organizations in order to optimally utilize the observed collaborations in the estimation model, while keeping the size of the model small. Technically speaking, we select organizations in the 2-core (de Nooy et al., 2004) of the largest component of the network of reported collaborations (203 organizations representing 17 per cent of all vertices). Finally, another 12 organizations are excluded due to non-availability of geographical distance data, so that we end up with a sample of n = 191 organizations that are used to construct the binary dependent variable as described below.

⁶ We kept Y symmetric so that Y _ij = Y _ji . By this we assume reciprocity of the links as soon as one of the partners reports intensive collaboration with another partner in the survey.

⁷ NUTS is an acronym of the French for the “nomenclature of territorial units for statistics”, which is a hierarchical system of regions used by the statistical office of the European Community for the production of regional statistics. At the top of the hierarchy are NUTS-0 regions (countries) below which are NUTS-1 regions and then NUTS-2 regions.

⁸ EESD, GROWTH, HUMAN POTENTIAL, INCO 2, INNOVATION-SME, IST and LIFE QUALITY.

⁹ The exact wording of the question was “how many FP5 projects of your organization are you aware of?” For multiple responses from an organization the numbers of known projects are summarized. In cases of missing data this number is set to zero.

¹⁰ The exact wording of the question was “which of your [project acronym] partners (i.e. persons from which organization) did you know before the project began?”

¹¹ It is important to note that this embeddedness is rather a co-occurrent than a preceding phenomenon, so that we cannot infer causality in the strict sense in this approach. However, as the evolution of large-scale R&D networks can be considered as a relatively long-term process as compared with individual collaboration decisions, we can expect results on close relatedness.

¹² Nagelkerke's R-squared is an attempt to imitate the interpretation of multiple R-squared measures from linear regressions based on the log likelihood of the final model vs. log likelihood of the null model. It is defined as where L ₀ is the log likelihood of the null model, L ₁ is the log likelihood of the model to be evaluated and n is the number of observations.