Skip to main content

Centrality Analysis in Research Networks

Final Report Summary - CAIRN (Centrality analysis in research networks)

Executive summary

The European Framework Programmes (FPs) have been pivotal for transforming informal nation-based networks of research collaborations within epistemic communities of academics and industrial researchers into formal collaboration arrangements between organisations at European level. The networks formed by the organisations have become almost as important outcome of FPs as the scientific and technological results of research projects conducted by them.

The analysis of the characteristics and structural properties of the networks, built by / through the six Framework Programmes, implemented until 2006, provides a plausible indication whether this new fabric of European RTD has become more cohesive and integrative during the past 20 years. It is valuable for understanding the contribution of European policies for transforming the fabric of research within the ERA, as well for identifying the emergence of a possible backbone of key research organisation in Europe. The study aims to explore this kind of issues by exploiting the richness of FP collaboration data using advanced methods of social network analysis.

The analysis of structural features of FP5 and FP6 networks suggests several implications for ERA. First of all, comparing the evolution of the FPs over time, we observe extensive instrumental and structural change. For the same type of instruments and for the same themes, the networks emerging are more integrated and more tightly knit. This could be interpreted as a signal of self-reinforcing pan-European thematic communities built on trust and a common operational framework that has evolved to its present state alongside the FP. Secondly, the overall success of the FP, in involving research teams from New Member States and integrating smaller peripheral communities into wider European networks, is compatible with the view that it is contributing to the construction of the 'backbone' of the ERA.

The identification of three kinds of networks as resulting from different types of sub-programmes - small world networks, distributed clusters networks and networked communities - has further repercussions for the implementation of ERA.

Small world networks tend to favour knowledge diffusion and building up of expertise across time but might be less effective to foster wider integration because of the difficulties that new players have in joining them. According to FP data for FP5 and FP6, small world networks (with high clustering) emerge in sub-programmes that are strongly oriented towards applied research and development. Such kind of networks are known for their resilience and their resistance to change due to the filtering apparatus of using highly connected nodes (or 'hubs'), and their high effectiveness in relaying information while keeping the number of links required to connect a network to a minimum.

Distributed cluster networks are found in programmes with a strong exploitative component and knowledge transfer functions. Such networks are less clustered than small world networks and represent a balance of expertise accumulation and integration, with lower barriers to joining in. Favouring the advancement of knowledge and efficient transfer within relatively closed cliques, they represent an interesting tool for ERA.

Finally, there are very evenly distributed network structures, the so called 'networked communities' with a lower clustering coefficient, which are associated with basic research. Such networks are better suited for cutting-edge research and allow a tighter integration since links are easily formed. However, they may be less suited for an efficient diffusion and exploitation of knowledge.

Generally speaking, different kinds of networks represent different answers to ERA priorities. Positioned in between the two main purposes of knowledge creation and of knowledge diffusion, there are irreducible trade-offs in opting for different kinds of orientations of sub-programmes in future FPs.

We identify the following main dimensions along the lines of which different network types are relevant: building strengths and the cohesion of the European Research Area.

These characteristic network properties are also crucial with respect to two policy objectives closely associated to the ERA concept, namely the building of European strengths in R&D, and the enhancement of cohesion of ERA. The actors that play a key role for achieving in both dimensions are universities. In many thematic areas, they are at the core of the networks built by the FPs through time, and have increased their centrality and share of participations. Because of the stability in the top positions and the wide representation of some of universities in different thematic networks, they play a double role of furthering both excellence and of contributing to cohesion. Together with the RTOs, universities form the building blocks of the ERA, acting as harbours of stability. Stability over time also suggests that policy interventions will need to take into account the specificities of these top actors and the networks in which they participate. It is therefore important that their central role is recognised in any discussion on the future evolution of the ERA.

The analysis of Integrated Projects (IP) and Networks of Excellence (NoE) in specific topics is of particular interest from an ERA perspective, as they were tasked with strengthening the ERA by enhancing collaboration at programmatic level. Both aimed at the facilitation of common research agendas, at the integration of smaller research communities and New Member States and at the promotion of virtual centres of excellence that are visible at a global scale. In accordance with the expectations attached to them, we found that they favoured large projects with many participants, but it remains to be seen whether these large-scale networks will have a structural effect on ERA after the end of funding.

Organisation rankings by theme indicate wide variation across themes but, within a given theme, relative homogeneity across instruments. Within each theme, we can distinguish between a core of stable presences in the top ranks and others that are rather volatile. Core organisations have played the role of integrator and coordinator in the building a European-level research agenda for a given topic.

Consistent with the ERA vision that sees coordination and cooperation as contributing to existing strengths and integrating the knowledge periphery, the 'top of the top' universities participating in FP6 in those instruments are spread across different countries, large and small, generic and specialised universities are all involved in the FPs.

The role of the FP in structuring the ERA could be enhanced by the suitable design of instruments that are tailored to the needs of thematic communities. Our analysis points to significant differences in the resulting networks across thematic priorities. We also observe that the exact shape of the knowledge triangle is thematically conditioned: the composition of resulting networks varies in terms of leading organisation type, with aerospace at one extreme (where industry is dominant) and life sciences at the other (where universities dominate). The even mix of organisation types represented in the top ranks of ICT is indicative of a priority that is conducive to knowledge sharing between different organisation types. Energy and environment has allowed a better integration of new organisations in the FP networks. This can be seen as a consequence of the public-good nature of much of the knowledge produced and diffused in this programme; a characteristic that requires more inclusive networks to be built. As such, this priority might represent an example of how the FP could contribute to the tackling of 'grand challenges'.

Summary description of project context and objectives

The move towards the European Research Area (ERA) is at the core of the political process since the Lisbon Agenda (European Commission 2000a). Several initiatives have been taken since then to foster its development. In 2007, the Green Paper for the ERA (European Commission 2007) identified the six axes along which ERA should focus to create the necessary conditions for a European internal market for research. The need for excellent national and European research organisations and the creation of the framework conditions and incentives to knowledge sharing are two of these axes of action. On 2 December 2008, the Council of the European Union adopted a common 2020 vision for the European Research Area (European Council, 2008), which alongside with the need for better competition stressed the need to reinforce cooperation and coordination. In July 2009 in the Lund Declaration Member States adopted 'grand challenges' as approach to coordinate policy initiatives to achieve the ERA Vision 2020. It defines as essential the promotion of cross-border cooperation, the strengthening of networks of excellent and of less-developed research organisations to enhance the overall competitiveness of European research.

Monitoring the move towards the European Research Area is therefore pivotal in this political process. Novel methodological tools applied to data on the European collaboration contribute to the challenges posed by a monitoring system that is not only based on the traditional input and output measurement. Beyond the analysis of co-publication and co-patenting, usually used as proxies for research collaboration, there are other sources of data that can be mobilised, like the data on public funding awarded to European R&D activity. At European level there are five major sources of public funding for collaborative endeavours: The European Framework Programme (the major European scheme for funding transnational research), the inter-governmental framework COST, the schemes promoted or managed by the European Science Foundation, and EUREKA. The European Research Council is also an important and interesting source of funding at European level, but it distinguishes from the previous ones because it does not require collaboration across European countries. Its aim is to promote competitiveness based on excellence at the European level.

The focus of this study is on the analysis of networks promoted by the past six European Framework Programmes (1984-2006). The main objective is to advance our understanding on transnational networks of collaborative research, identify the relevant networks, as well as the role played by the most central organisations in those networks. The study of the networks promoted by the other above mentioned European research funding sources would complement this analysis. A feasibility study has been done, but will not be reported here.

The European Framework Programme is the main instrument of European research policy. It has been conceived as an instrument of transnational collaborative research aimed at improving the international competitiveness of European industry, while at the same time strengthening EU cohesion. Since FP6 it serves as the key instrument to foster the European Research Area.

Although our intention is not to do the historical account of the European Framework Programmes, it is important here to recall its origins, and its role in promoting research collaboration across research organisations of the European Member States, as well as the rupture introduced in the range of both geographical and modes of collaboration of research organisations. The FPs are one of the answers of Europe to the challenges posed by the knowledge production of generic technologies, like the information and communication technologies or biotechnology, developed through the combination of different disciplines and skills through collaboration of heterogeneous actors (Callon, Larédo et al., 1995). The development of these technologies imply a cooperative process between knowledge producers and consequently the implementation of novel processes for sharing knowledge and resources in order to cope with the need of reducing lead times and the fast pace of technology development and diffusion (Onida and Malerba, 1989; Freeman, 1991).

European Framework programmes were modelled based on the success of ESPRIT I, the information technologies (IT) programme for collaborative research at the European level, created in 1982 by the European Commission. ESPRIT was promoted by the Commissioner for Industry, Étienne Davignon, with the support and advice from the European roundtable of the twelve biggest European companies in the IT sector. The First Framework Programme for Research and Technology Development (RTD) was created two years later, which included the ESPRIT programme and other sub-programmes in a variety of topics, to address the development of generic technologies within a multi-annual framework. Since then other FPs have been implemented regularly with an enlarged scope and a diversified set of funding instruments. The rationale behind was that universities, research institutes and firms (even competitors) from Member States should work in cooperation to reduce the technology gap of Europe in relation to the United States and Japan and increase its competitiveness. Therefore the projects funded by the FP focus either on the development of new technologies and products or on the development of technological standards. The projects have to be carried out by a consortium of research organisations, from at least two different countries, preferably with the involvement of knowledge producers, exploiters and users.

FPs were pivotal in changing the traditional nation-based informal research collaboration within epistemic communities into formal arrangements between research organisations at the European level. The durable networks of research collaboration formed by the organisations participating in FPs are almost as important as the scientific and technological outcomes of research projects supported by them.

The collaborative links established by European projects can be equated to paths through which the knowledge circulates between the organisations, and eventually joint knowledge is produced. The analysis of the characteristics and structural properties of these networks can plausible give an indication on the nature and characteristics of the new fabric of European RTD, and on the degree of its cohesiveness and integration. In addition, the analysis sheds light on the contribution of the European research policies to the transformation of research within the ERA and aims at identifying a possible backbone.

The main objective of the study was to exploit the richness of FP data through social network analysis (structure of research networks and actors centrality) to contribute to the process of monitoring the move towards the ERA. The research questions addressed in the study were the following:

1) Does the density of collaborative organisational links increase over time?
2) Is it possible to identify optimal network structures by areas of research and funding instruments?
3) Is it possible to identify a backbone of core research organisations in the European Research Area?
4) Who are the key players in the FPs, and where are they located within the FP networks?

Description of the main S&T results / foregrounds

A network analysis of the FPs is an important analytical tool for the overall evaluation of results and impact of R&D policies in the EU. The above analysis of structural features of FP5 and FP6 networks suggests several implications for ERA. The distinction between three kinds of networks - small world networks, distributed clusters networks and networked communities- as the outcome of different sub programmes has repercussions for the implementation of ERA.

In the context of ERA, small world networks might favour knowledge diffusion and building up expertise across time but might be less effective to foster wider integration because of the difficulties that new players have in joining in. In general, different kinds of networks represent different answers to ERA priorities, between the two main aims of building up expertise and of knowledge diffusion, there are irreducible trade-offs in opting for sub-programmes in future FPs.

Comparing the evolution of the FPs over time, we observe extensive instrumental and structural change. Over time, for the same type of instruments and for the same themes, the networks emerging are more integrated and more tightly knit. This could be interpreted as a signal of self-reinforcing pan-European thematic communities built on trust and a common operational framework that has evolved in its present state alongside the FP.

According to FP data for FP5 and FP6, small world networks (with high clustering) in sub-programmes emerge for sub programmes strongly oriented on direct research. Such kind of networks are known for their resilience over time and their resistance to change due to the filtering apparatus of using highly connected nodes (or 'hubs'), and its better effectiveness in relaying information while keeping the number of links required to connect a network to a minimum. In other words, in the context of ERA, such networks might favour knowledge and building up expertise across time but might be less effective to foster wider integration because of the difficulties that new players have in joining in.

Distributed cluster networks are found in programmes with a strong exploitative component and knowledge transfer functions. Such networks are less clustered than small world networks and represent a balance of expertise accumulation and integration, with less high obstacle in joining in. Favouring the advancement of knowledge and efficient transfer within relatively closed cliques, they represent an interesting tool for ERA.

Finally, there are more evenly distributed network structures, the so called 'networked communities' that with a lower clustering are associated with basic research. Such networks are better suited for cutting-edge research and allow a wider integration since links are easily formed (due to the small nature of the projects involved). However, they might be less suited for an efficient diffusion and exploitation of knowledge.

In general, different kinds of networks represent different answers to ERA priorities, between the two main aims of building up expertise and of knowledge diffusion, there are irreducible trade-offs in opting for sub-programmes in future FPs. We identify the following main dimensions along which different network types are relevant:

- Building strengths. The identified distinction between three kinds of networks as the outcome of different sub programmes has repercussions for the implementation of ERA. In the context of ERA, small world networks might favour knowledge and building up expertise across time but might be less effective to foster wider integration because of the difficulties that new players have in joining in. In general, different kinds of networks represent different answers to ERA priorities, between the two main aims of building up expertise and of knowledge diffusion, there are irreducible trade-offs in opting for sub-programmes in future FPs. Comparing the evolution of the FPs over time, we observe extensive instrumental and structural change. Over time, for the same type of instruments and for the same themes, the networks emerging are more integrated and more tightly knitted. This could be interpreted as a signal of self-reinforcing pan-European thematic communities built on trust and a common operational framework that has evolved in its present state alongside the FP.

- Cohesion of the European Research Area. Distributed cluster networks are found in programmes with a strong exploitative component and knowledge transfer functions. Such networks are less clustered than small world networks and represent a balance of expertise accumulation and integration, with less high obstacle in joining in. Favouring the advancement of knowledge and efficient transfer within relatively closed cliques, they represent an interesting cohesion tool for ERA. The overall success of the FP in involving research teams from new member states and integrating smaller peripheral communities into wider European networks shows that it is contributing to the construction of the 'backbone' of the ERA. Core organisations have played the role of integrator and coordinator in the building a European-level research agenda for a given topic. However, the rankings of top organisations provide some indications of high entry costs. Aerospace, in particular, is dominated by industry, exhibiting relative stability in the ranks of universities and research organisations, with some mobility in the ranks of industrial actors over time. High entry costs are also reflected in more inclusive thematic priorities (such as ICT), with the top ranks dominated by organisations from older member states. Discussions on the future evolution of the ERA should take into account the high entry costs for new participants and take the necessary steps to facilitate entry.

The actors that can achieve in both dimensions are universities that are at the core of the networks built by the FPs through time, increasing their centrality and share of participation. Because of stability in the top positions and, as observed previously, the wide representation of some of universities in different thematic networks, they play a double role of capacity building and cohesion. Stability over time also suggests that policy interventions will need to take into account the specificities of these top actors and the networks in which they participate. Of all organisation types, universities are the ones that form the building blocks of the ERA, acting as harbours of stability. It is therefore important that their central role is recognised in any discussion on the future evolution of the ERA.

The analysis by instrument is of particular interest from an ERA perspective, given that two of the instruments examined (IP, NoE) were tasked with strengthening the ERA: IP and NoE aimed at the facilitation of common research agendas, at the integration of smaller research communities and new Member States (NMS), and at the promotion of virtual centres of excellence that are visible at the global level.

In accordance with the expectations attached to IP and NoE, we found that they favoured large projects with many participants. The analysis of FP6 shows that the mean number of partners per organisation (mean degree) as well as the clustering coefficient in these networks.

The top 20 positions of universities are spread across different countries, in contrast to the typical concentration found in academic ranking tables. Large and small, generic and specialised universities are all involved in the FPs. This image is consistent with the ERA vision that sees coordination and cooperation (as promoted by the FP, but also other instruments such as COST and EUREKA) as contributing on existing strengths and integrating the knowledge periphery.

Analysis suggests that energy and environment has allowed a better integration of new organisations. This can be seen as a reflection of a topic dealing with the production and diffusion of public good-type knowledge, which requires more inclusive networks. As such, this priority might represent an exemplar of how the FP could contribute to the tackling of 'grand challenges'.

In real-world networks, this likelihood tends to be greater than the average probability of a tie randomly established between two nodes (Holland and Leinhardt, 1971; Watts and Strogatz, 1998) is higher than in networks based on Cost Shared Contracts (CSC and STREP). The above can be taken as an indication that the two ERA instruments shaped the structure of research collaboration networks across Europe.

The overall success of the FP in involving research teams from new member states and integrating smaller peripheral communities into wider European networks is compatible with the view that it is contributing to the construction of the 'backbone' of the ERA. Core organisations have played the role of integrator and coordinator in the building a European-level research agenda for a given topic.

Our analysis points to significant differences in the resulting networks across thematic priorities. We also observe that the exact shape (distribution) of the knowledge triangle is thematically conditioned: the composition of resulting networks varies in terms of leading organisation type with aerospace in one extreme (where industry is dominant) and life sciences in the other (where universities dominate). The above observations suggest that the role of the FP in structuring the ERA could be enhanced by the suitable design of instruments that are tailored to the needs of thematic communities.

The even mix of organisation types represented in the top ranks of ICT is indicative of a priority that is conducive to knowledge sharing between different organisation types. The above observations suggest that the role of the FP in structuring the ERA could be enhanced by the suitable design of instruments that are tailored to the needs of thematic communities.

Organisation rankings indicate wide variation across themes but, within a given theme, relative homogeneity across instruments. In other words, representation in the top ranks is primarily thematically conditioned. Within each theme, we can distinguish between a core of stable presences in the top ranks and others that are rather volatile. We observe a different mix of organisations across the various thematic priorities.