CAIRN.INFO : Chercher, repérer, avancer.

1In the last decades, the geography of innovation has analysed many forms of knowledge networks as sources of the innovative performances of territories. Such descriptions normally include scientific networks, but they generally focus on the links with firm’s R&D, knowledge spillovers, “technology transfer”, etc. The researchers interested in the concept of proximity (see for instance Boschma (2005) or the contributions of the French group “Dynamiques de Proximité” starting with their manifesto: Bellet, Colletis, Lung, 1993) contributed a lot to the understanding of local innovative processes, bringing to light the cooperation relationships between innovative actors (Torre, Rallet, 2005), underlining the role of local institutions (Talbot, Kirat, 2005) or distinguishing geographical proximity from organizational proximity (Zimmermann, 2008). Nevertheless few studies focused on the specific role of scientific networks, although many of the theoretical debates are similar and related: To what extent are scientific cooperation schemes local or global? What is the more relevant concept of proximity: organizational or geographical? Are there preferred partners in scientific production (in relationship with language, culture, history…)?

2Chalaye and Massard (2008) were probably the first in France to analyse scientific cooperation at a very precise (infra-regional) level. This study intended to help policymakers in the design and evaluation of innovative clusters. More recently Hazir and Autant-Bernard (2014) realised an analysis of R&D networks between European regions. Whatever the level of governance, it is important to understand the nature and functioning of research networks for evaluating the innovative performance of territorial systems.

3The same point can be identified in the literature related to the regional innovation systems. The diversity of European regions in terms of R&D and of absorptive capacities (Oughton et al., 2002; Pinto, 2009) has been extensively investigated but without taking into account all the dimensions of the regional innovation systems (Cooke, 2005). The concept of regional innovation system (RIS) assumes a systemic coherence of the actors, such as academic institutions, national agencies, industrial firms, knowledge-based services, local governments, etc (Leydesdorff, 2012). Those are leading to potential localized comparative advantages (Cooke, 2001; Asheim et al., 2005; Asheim, 2006). The RIS approach underwent a considerable expansion in the 90s, resulting from several simultaneous phenomena, mainly (i) the rising interest of researchers for the understanding of localized dynamics of research interactions and the impact of the science system on local innovation (Cooke, 2001; Bruijn, Lagendijk, 2005); and (ii) the efforts of policymakers to define a relevant conceptual framework for local intervention, following the paradigm of the “knowledge-based economy” (Oughton et al., 2002). Nevertheless, the concepts associated to RIS are not fully clear and justified (Héraud, 2003, Doloreux, Bitard, 2005; Tödtling, Trippl, 2005; Uyarra, 2010). Among the criticisms, the lack of consideration devoted to the variety of connections that the actors develop within the territory and with the external partners should be stressed (Barthelt et al., 2004; Malecki, Oinas, 1999).

4The local versus global connectivity in the RIS is an issue for public policies at the local level (Benneworth, 2010). It represents today a challenge for the implementation of the Horizon 2020 European program that aims at boosting innovation as well as promoting excellence in science (European Commission, 2010). Following the Teaming and Twinning concepts of the Horizon 2020 (European Commission, 2015), research organizations have to develop a set of stable scientific collaborations in order to obtain European R&D supports. The EU is indeed encouraging the creation of new knowledge by the development of new forms of local and European interactions among regions. Regarding the existing literature, the relevance of this double approach calls for a stronger grounding, especially regarding the characteristics of global and local scientific connectivity of European regions and their evolution over time.

5The optimal design of scientific connectivity is therefore at the core of the issue. The article investigates the different types of regional scientific connectivity and the relevant measures to characterize local and global scientific connectivity for European regions. The article is divided into three sections. Section 2 presents the issue of regional scientific connectivity. Section 3 proposes a typology of European regions based on co-publication statistics – at various levels – which captures the local-global connectivity issue. Section 4 discusses the relevance of the measures of connectivity in order to characterize European regions’ scientific activities.

The stakes of the regional scientific connectivity

6Contributions on regional connectivity are mostly focussed on firms (Markusen, 1994; Uyarra, 2010) or on science-industry relationships (Kratke, Brandt, 2009). Connectivity in academic research is much less studied as a basic component of scientific sub-systems, in particular at regional level. However, collaborative research is increasingly present in most academic fields and it seems to be a factor of success in contemporary science. More precisely, it can be considered as a necessity for performing research or as an advantage in a situation of hard scientific competition. The increasing complexity of science is often viewed as one of the main explanations of the intensification of collaborative behaviour in science (Adams et al., 2005; Tijssen, Van Leeuwen, 2007). In the meantime, researchers want to increase the visibility, the quality of their researches or the access to relevant infrastructures or funding opportunities, by joining efforts and names. In a sense, territories – like individual researchers and research teams – are in competition for scientific assets, looking for the good mix of cooperation and competition as a key for their respective success.

7In this context it becomes important to identify the various rationales behind the concept of scientific connectivity in order to propose a corresponding measurement. Scientific connectivity reflects the collaborative attitude of researchers in the creation of new knowledge, but it also reflects the tension of regional authorities in the local versus global development of an important component of their RIS. In the coming sub-sections, we propose to review how these issues were tackled in the literature, and ultimately we underline the poor inclusion of such issues in its practical measurement’s dimensions.

High or Low Scientific Connectivity as a Key Dimension of the Regional Scientific Sub-System

8While scientific collaborations happen to take place inside the same institution, scholars emphasise that excessively close actors may have little to exchange after a certain number of interactions (Boschma, Frenken, 2010). Indeed co-located agents combine and recombine local knowledge that eventually becomes redundant and less valuable. Regions with low scientific connectivity may face a risk of localism.

9Apart from bringing novel knowledge, collaborations outside the same institutions are essential to facilitate the recombination of various complementarity knowledge, ideas, expertise or infrastructure (Wagner, 2005). The more researchers develop external scientific collaborations with researchers localized in the same region and outside the region, the more they contribute to develop high connectivity of the whole regional scientific sub-system (Benneworth et al., 2011). Therefore, it is important to combine various types of collaborations, outside the region, but even inside with other research organizations in order to ensure a satisfactory amount of adoption and creation of knowledge (Fratesi, Senn, 2009). Depending on the regional context as well as on the disciplines concerned, a situation of ‘high connectivity’ may encompass different meanings. Thus, the well-balanced mix of regional, national and international connectivity becomes even more complex to identify. We will go into deeper details on that point in the next sub-section.

Global and Local Connectivity: Various Schemes of Scientific Collaborations in the Regional Scientific Sub-System

10The connectivity is based on a complementarity of knowledge sources: local, national, and international (Benneworth, Dassen, 2011). Hoekman et al. (2010) underline the importance of such various networks for regions. Universities and public research organisations are nowadays recognized as key actors for the stimulation of local-global connectivity (Tödling, Trippl, 2005; Benneworth et al., 2009) partially because those actors help to resist lock-in phenomena (Benneworth, Hospers, 2007).

11Analysing the local and global connectivity has a double advantage. On the one hand, it is a way to approach the complex reality of the scientific regional sub-system. On the other hand, it helps dealing with the issue of local governance in a more pragmatic way by looking at all the various forms they can have (Leydesdorff, Etzkowitz, 1998; Leydesdorff, 2003; Moreno, Miguélez, 2012). Two major stakes emerge from the literature: the internal (local) and the external coherence.

12The first issue, the internal coherence of the scientific sub-system (i.e. to what extent does the region produce its own system), is analysed by Anova and Leydesdorff (2001). The systemic coherence depends on many factors such as local history, culture and institutions. For sure, it is not granted that any “region” has a full-fledged scientific system, but many studies prove at least that “space does matter”. Scientific production is a global activity but proximities still facilitate the creation and functioning of scientific networks (Zucker et al., 1998; Zitt et al., 2000; Okubo, Zitt, 2004; Frenken et al., 2009). Face-to-face relations play an important role, at least at some stages of the collaboration. For instance, geographical proximity is a key factor for launching a research team. Science parks try to play on this spatial phenomenon for initiating networks (Beaudry, Breschi, 2000).

13The second issue considers the actors’ linkages with the external world and their impact on local development strategies (Benneworth, Dassen, 2011). International scientific collaborations have increased during the last three decades (Luukkonen et al., 1992 and 1993; Georghiou, 1998; Wagner, Leydesdorff, 2005), and this trend could indicate that “proximity does not matter so much”. Looking more carefully at the phenomenon, it is obvious that motivations for these collaborations are multifaceted, as shown by Georghiou (1998), Beaver (2001), Wagner (2005), and Wagner et al. (2001): historical links like colonial ones; financial opportunities; necessity to share costly infrastructures; knowledge complementarities. Another motivation is the reputation factor. Some authors suggest that the main reason to collaborate internationally refers to scientific excellence (Tjissen et al., 2007). It is true that international publications are more quoted than the others (Frenken et al., 2009). While motivations for international collaboration vary among disciplines (Wagner, 2005), the internationalization is a general rule.

14Interesting studies on the collaborative behaviour refine this general picture. For example, regional and national collaborations are explained by cultural as well as organizational/institutional factors (Torre, 2008; Hoekman et al., 2010). Another illustration is given by Glanzel (2001): the intensity of international collaboration seems to diminish with the “size” of the actor (measured by the number and quality of publications). Wagner et al. (2001) found a decreasing necessity for large scientific actors to go abroad for finding complementary competences. Conversly, for regions with low and intermediate levels of scientific excellence, internationalization of research is a main way to develop the production of new scientific knowledge (Persson et al., 2004).

15Local and global connections are possible sources of advantages (Benneworth et al., 2010). The variety of connections can reduce lock-in effects (Benneworth, Hospers, 2007) and can create new forms of diversity resulting not only from internal sources of knowledge but also from the capability of local actors to mobilize knowledge from the outside. The Cohen and Levinthal’s (1990) theory of absorptive capacities should be generalized to the regional scientific sub-systems: the translation/reinterpretation of external knowledge is a potential source of new scientific knowledge. In the specific case of European regions, the challenge is to reflect those complex sets of scientific connectivity within a comprehensive mapping that considers both internal and external coherences. Indeed, European regions have been the objects of many typology exercises that integrate a variety of indicators (see Navarro and Gibaja, 2010 for a detailed review on RIS’ typologies). But within this corpus we observe that the scientific sub-systems (with their interactions and dynamics) are poorly, if not, considered. The coming section proposes to fill this gap.

Scientific collaborations of European regions

16In the present section, we firstly present our original data set and the statistical approach chosen. We then describe the typology of European regions we obtained. The dynamics of our typologies are also discussed to show both the stability of our clusters and the relevance of integrating more strongly the scientific connectivity into RIS’ analysis.

Data and Methodology

17The data used were constructed by the Observatoire des Sciences et des Techniques (OST) and based on an original data set of Web of Science (WoS). Scientific co-authorship data were regionalized at the NUTS-2 level (263 European regions [1]) according to the institutional address of the authors listed in about 271 783 publications in 2012. Similar data have been collected for the year 1999, 2003, 2006 and 2009.

18We decided to work at the NUTS-2 level and not at lower geographical degree because we are not considering cluster analysis like in Chalaye and Masard (2008), but addressing regional policy. NUTS-2 is not always ideal as a perimeter in this respect, but NUTS-3 level would definitively be worst.

19Our typology is based on eight different indicators that cover all available disciplines in ‘hard’ science and considers the different geographical levels of collaboration that fall within the following institutional contexts:

  • Mono-address co-publications: scientific publications produced by one or several scientists from the same institution.
  • Intra-regional co-publications: produced by at least two researchers from different institutions located in the same region.
  • Co-publications among regions from the same country (except the region itself).
  • European co-publications: co-authorship between one institution and another outside the national boundaries but located in Europe.
  • International collaborations through two indicators: the co-publications between a European region and the USA, and the co-publications with China, Japan or other international (i.e. non-European and non-US) partners.

20Each indicator was regionalized and stabilized (smoothed over three years) to prevent unusual rebounds in the data set and to make the trend more obvious (see OST, 2009). These indicators are percentage, so that we analyse the share of collaborations at different geographical levels. Scientific size is in a sense neutralized by our method, but we also control for this potential critical mass effect through complementary data (described below). These additional regional data were projected backward on the computed axes (i.e. those will not impact our factorial analysis but rather provide information to later characterise our clusters).

Table 1

Descriptive statistics

Table 1
Co-publications (all disciplines) 1999 2003 2006 2009 2012 Mono address 145 386 154 756 165 133 184 662 174 510 Intra-regional 57 802 67 985 81 358 97 538 113 429 National 34 908 42 091 49 766 59 647 67 706 European 22 873 27 897 33 579 39 764 45 469 International USA 14 012 18 019 22 856 28 605 33 636 Other int. 9 002 10 683 12 071 13 596 15 080

Descriptive statistics

Note: Fractional counting (rounded).

21The first category of additional data is related to the size and the economic activity of the regions. We collected the total of publications and of co-publications in Humanities & Social sciences and in Hard Sciences, both in fractional and presence counting, for the overall NUTS-2 studied. We also integrate a measure of the evolution of scientific publications on the period 1999 to 2012. We considered the R&D expenditures of the region in all sectors and in Higher Education in absolute term and per capita. Finally we measure the amount of researchers in the Public sector and in Higher Education. Based on this first set of data, we are able to represent the scientific forces of the region, their potentialities both financial and in terms of human resources and their scientific dynamics over the last decade.

22The second category of illustrative variables aims at capturing the scientific position of a region based on its visibility and specialisation: the two-year impact index “all discipline” of the region in 2012 (Zitt et al., 2000) and its scientific specialisation index on eight disciplines in hard science plus the SSH in 2012. This index gives a measure of the average citation rate (received in 2 years) of the region’s publications.

23Considering the diversity of European regions, and the somewhat arbitrary division of the NUTS-2 nomenclature we also control for the number of regions within the same country. In other words we capture the necessary over-representation of intra-regional and national collaborations more likely to happen in Germany or in the UK (respectively considered in our analysis by 39 and 37 NUTS-2) than Denmark or Bulgaria (5 and 6 NUTS-2). In the same frame, we also measured the number of regions within the same country of the region having a number of publications and of co-publications above 60% of the European sample of regions (i.e. above the 6th decile).

24Ultimately we introduced a fourth group of illustrative data related to European programs funding. We collected the Interreg IVC budget dedicated to the theme: Innovation and the knowledge economy (sub-theme: Innovation, research and technology development). In addition we also got the European Regional Development Fund Projects selected 2007-2013 AR Community Amount (R&TD activities) in research centres; infrastructures (including physical plant, instrumentation and high-speed computer networks linking research centres) and centres of competence in a specific technology; and in technology transfer and improvement of cooperation networks between small businesses (SMEs), between these and other businesses and universities, postsecondary education establishments of all kinds, regional authorities, research centres and scientific and technological poles (scientific and technological parks, technopoles, etc.).

Factor Analysis and Clustering

25Our goal is to compare collaborative patterns and not absolute co-publication production. The strength of large scientific regions is indeed acknowledged in the literature; but in a context of regional development strategies, the relative patterns of collaborations become crucial. This motivates the choice for the quantitative methods used (factor analysis and hierarchical clustering). While having a representative and complete sample of the scientific co-publications at different levels we also control for additional factors impacting the scientific collaborations. We included a series of additional data, projected backward on the computed axes (see below). These illustrative variables have been dichotomized into four categories (as previously presented).

26This bibliometric data set was explored through a correspondence factor analysis (CFA). This methodology is common in typologies’ exercises (see Navarro, Gibaja, 2009 or Pinto, 2009). As emphasised by Teil (1975) the “aim of the method is to find associations and oppositions between subjects and variables, as in other multivariate methods. Its advantages (…) are related to its more sophisticated mathematics and its simultaneous representation of subjects and variables on the same factorial axes”. The aim of the CFA is to reduce the dimensions of our data set that contains all of the co- publications in 2009 at the European Nuts-2 level [see Doré et al. (1996) for detailed methodology of the CFA in a similar context]. The indicators of co-publications are overall additive. Basically the additivity property of our data set allows us to interpret the rows as profiles of collaboration. This method is equivalent to a conventional principal component analysis (PCA). But as Doré et al. (1996) noted the PCA is not applicable to categorical variables. “When dealing with counts (publications) rather than measurements, it is helpful to convert the data into Khi-2 frequency distributioni.e. CFA is preferred to a ‘traditional’ PCA method.

27We performed a hierarchical classification on the principal components of the CFA. The statistical result is a set of clusters of European regions based on their similarities and dissimilarities of scientific collaborations’ patterns. The first two axes represent about 74% of the standard deviation (Fig. 1). Hierarchical classification has been conducted in order to determine four main profiles of regions in terms of their scientific connectivity. Considering more types did not proved relevant in statistical nor analytical terms. Factor loadings and Eigen values are respectively presented in Table 2 and Table 3.

Figure 1

Interpretation of the first two axes

Figure 1

Interpretation of the first two axes

Table 2

Factor loadings of individual indicators

Table 2
Factor 1 Factor 2 Mono address -0.048 0.250 Intra-regional -0.334 -0.269 National 0.444 -0.163 European -0.041 -0.082 US 0.087 -0.042 Other International 0.010 -0.054

Factor loadings of individual indicators

Source : Authors computation, SPAD 11
Table 3

Eigen values on the first three axes

Table 3
Eigenvalue % of variance Cumulative % 1 0.063 43.0 43.0 2 0.046 31.5 74.5 3 0.024 16.3 90.8

Eigen values on the first three axes

Note: We retain only two axes in our analysis; the values given for the third axis are only highlighting the drop to zero starting after the second factor.

Results

The Typology of the Regional Scientific Connectivity

28Regions with similar structures of co-publications (i.e. connectivity profile) are allocated to the following four clusters (Fig. 2) presenting a maximum of similarities “within” and of differences “between” regions. The figure shows the position of the regions on the main factor analysis plan (Axis 1, Axis 2) which represents almost 75 % of total of total variance.

Figure 2

The four clusters on the main factor analysis plan

Figure 2

The four clusters on the main factor analysis plan

Source : Authors elaboration
Table 4

Matrix of distances between clusters

Table 4
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 1 0.000 0.508 0.823 0.681 Cluster 2 0.508 0.000 0.385 0.392 Cluster 3 0.823 0.385 0.000 0.385 Cluster 4 0.681 0.392 0.385 0.000

Matrix of distances between clusters

Source : Authors elaboration
Figure 3

Mapping of clusters, based on 2012 data

Figure 3

Mapping of clusters, based on 2012 data

Source : Authors elaboration, QGIS 2.8.3

29The cluster 1 “low connectivity, international orientation” is made of 98 regions. It presents regions with relatively strong international connectivity (co-publications with the US and other international are higher than the European average) coupled with a relative low degree of total connectivity (because of a large proportion of intra-institutional co-publications). The cluster encompasses a very significant part of the UK and German regions. Large scientific producers in terms of publications are included in this cluster like Berlin, Oxford, Karlsruhe, East Anglia or Köln regions.

3096 regions belong in the cluster 2 “high connectivity, intraregional orientation”. This cluster is typically representative of regional systems of science, where science is often co-produced with regional partners. Composed of large and small regions, this disparate group is characterised by a lower tendency to co-publish inside the same institution and at the same time by the tendency to co-produce with regional partners and sometimes with other European regions (but this group of regions does not score above the average in terms of international collaborations).

31The cluster 3 “low connectivity, intraregional orientation” is made of 43 regions. This cluster is characterised by strong regional collaborations’ patterns and a noticeable low connectivity (intra-institutional scheme of collaboration). The connectivity outside region’s borders (national, European or international) is particularly low.

32Ultimately, the cluster 4 “high connectivity, national orientation” made of 26 regions, is mainly characterised by intensive national collaboration and a high connectivity (poor intra-institutional scheme of collaboration). Some of the cluster 4 regions are cross-borders but in average they do not take advantage of their locations to co-produce in science with research organizations outside their home country.

33The results of our CFA and classification method lead to three interesting findings. We observe a clear opposition between European regions with high level of scientific connectivity (clusters 2 and 4) and two other groups of regions which have limited collaborations in general. Our results show that the degree of scientific connectivity is a key statistical factor differentiating European regions. The overall picture highlights regions that have the capacity to reduce the risk of lock-in effect and to develop a virtuous cycle of new knowledge production. It is interesting to notice that the regions with low scientific connectivity develop a relatively poor scientific visibility. Some European regional science systems are still shaped by regional or national tendency to collaborate (clusters 2, 3 and 4), while only one group of regions (cluster 1) is more likely to co-publish with international partners.

34The second main learning from our typology is the low level (in average) of co-publications between European partners, despite the recent efforts of the EU to reinforce the European Research Area. Indeed, the European co-publications represent only 10% of the total amount in 2012 [2]. The European connectivity does not appear as an important factor to differentiate amongst European regions.

35Ultimately, we can observe the non-complete uniformity of European countries regarding their scientific connectivity. The existing literature emphasizes that national system of science still shapes the collaborations patterns: the country level is characterized by the existence of institutions and scientific policies that have a major influence on the governance of research and specialization (classical concept of national innovation system). Our typology underlines that regions within the same countries appear to follow similar patterns, but some of them divert from their respective national norms.

The Dynamics of Connectivity

36In order to capture the dynamics of European regions, and at the same time to test the relevance of our typology we collected the same dataset on co-publications for 2006 and 1999 covering the same list of NUTS-2. Fig. 4 and Fig. 5 are mapping the projection of the European regions on the axes obtained for 2012 in order to keep the typology constant over time.

Figure 4

Mapping of clusters, based on 2006 data

Figure 4

Mapping of clusters, based on 2006 data

Source: Authors elaboration, QGIS 2.8.3
Figure 5

Mapping of clusters, based on 1999 data

Figure 5

Mapping of clusters, based on 1999 data

Source : Authors elaboration, QGIS 2.8.3

37By projecting the 1999 and 2006 variables/subjects on the 2012-computed axis, we observe i) the relative stability of the clusters definitions (when looking at the variables characterizing the clusters); ii) an increasing connectivity coupled with the reinforcement of the regional-oriented type of networking (when looking at the aggregation of regions into clusters); and iii) some interesting modifications of regional dynamics within countries (when looking at the dominant and less dominant types of regions in each country).

Table 5

Clusters in 2012, and projection of the regions in these clusters with 1999 and 2006 data

Table 5
1999 2006 2012 Cluster 1 “low connectivity, international orientation 110 124 98 Cluster 2 “high connectivity, intraregional orientation” 25 53 96 Cluster 3 “Low connectivity, intraregional orientation” 107 64 43 Cluster 4 “high connectivity, national orientation 21 20 26

Clusters in 2012, and projection of the regions in these clusters with 1999 and 2006 data

Note: The method consists here in projecting a region with its 1999 or 2006 data, on the axes obtained for 2012.
Source : Authors elaboration

38The different characteristics of clusters obtained in 2012 are stable enough to capture satisfyingly regions in 1999 and in 2006. In other words, thinking the regional scientific connectivity in a set of four groups based on their level of connectivity and the geography of their collaborations is relevant regardless of the period considered. Even if this typology is a good frame for the analysis, it is not fixed in proportions. The unchanged typology allows for shifts of regions from one type to another. The clusters 1 and 4 are relatively stable regarding their weight over the period considered, but the clusters 3 and 4 present opposite trajectories. Indeed, showing a strong tendency to collaboration within the region and low levels of connectivity for the rest (cluster 3) is a case that is less and less frequent. On the other side, strong intra-regional collaborations and significant (a little more than the average) European collaborations (cluster 2) is a growing trend since 1999. As a reminder, cluster 2 is relatively heterogeneous: there are both large and modest scientific producers, and therefore the interpretations should integrate the precise nature of each region under consideration. The evolution observed over the period could also reflect an increasing gap between regions in terms of scientific productions (i.e. small regions have not a large choice for intra-regional collaborations, while larger ones have a substantial portfolio of possible partners within their territory). Finally, we can observe a relative stability of the clusters regarding the different countries considered: Spain (except some regions like Madrid) have mainly regions in cluster 3; German regions remain mainly in cluster 1; the South of the United Kingdom stays in cluster 2 and the North in cluster 3; Italy remains a patchwork of various types of scientific connectivity over the period. The evolution of French regions is quite interesting since they went from a mixed situation (with many in clusters 3 and 1) to a strong majority of cluster 2 regions. It seems that regional science systems have been reinforced through the general development of scientific density and the increase of connectivity. The case of France shows that even in such a traditionally centralized country (Héraud, Lachmann 2015) the overall development of scientific activities leads to a stimulating densification of scientific networks in many territories. Our results suggest that the reforms and the decentralization of scientific policy in 2000 facilitate the development of new trajectory for scientific connectivity in French regions.

Discussion and conclusion

39Connectivity as a source of knowledge creation is recognised as a key issue for regional development. It represents one of the priorities of the European program Horizon 2020. We consider that, like firms, public research organisations should be integrated as important actors spurring both local and global connectivity. For the founding fathers of the European Smart Specialization Strategy (see e.g. Foray et al., 2009) what should be implemented in each region is an “entrepreneurial process of discovery” where universities as well as firms and territorial authorities commonly decide for the present and future specific innovation networks. All the actors of the territorial specialization process are not supposed to be internal actors. Less developed regions can be concerned by the exploitation of the new scientific and technological paradigms on specific aspects, while leading regions can envisage to do the exploration activity. Therefore, local networks of actors need global connections and must develop collaborations with certain external actors in order to reinforce their own specialization strategy.

40Our empirical results offer interesting insights regarding the understanding of scientific connectivity. Firstly we found a marked contrast between regions with high level of scientific connectivity and other regions with limited collaborations in general. Secondly, the regional capacity to combine local and global connectivity does not seem to be a key driver of the collaboration for European regions. Our clustering shows that the regional and national borders are still shaping scientific collaborations to a large extent. In addition, the clustering does not emphasize the European collaborations: they are still remaining marginal despite decades of EU efforts to build a European Research Area. We also observe that regions positioned in cross-border areas of Europe do not seize so much this opportunity to develop European links.

41As a set of sub-systems, regional scientific activities possess their internal and external coherence driven by a various set of factors: local history, governance or culture as well as development strategy, past collaborations or excellence motivations. Our results suggest that national patterns (history, culture, public policy and institutions) influence the pattern of regional scientific connectivity. National collaborative links do not necessarily increase, but the way to deal with scientific connectivity remains a national characteristic.

42The good news is the increasing conversion of regions towards the model of connectivity: many regions – characterized in the past by self-sufficiency and low scientific results a decade ago – are now developing broader research links.

43Ultimately, analysing the scientific productions of 263 European regions from 1999 to 2012, the patterns of collaboration (expressed in our typology) appear stable over time. However regions have their own trajectories between the general patterns. These findings reinforce previous studies pointing the regional anchorage and the poor Europeanization of scientific activities in Europe. But it overall complements it, by offering a larger and more dynamic picture of scientific activities in the context of regional connectivity. Let us take some examples.

44French regions have widely changed their patterns of collaborations over the last decade. It may nourish discussions regarding the relative success of regional strategies toward building more connected territories. For instance, the region Ile-de-France encourages a stronger local connectivity through interactions and networking (amongst a variety of large institutions and renowned scientific competences located on the territory). Aquitaine took a more complementary approach by favouring local connectivity (cooperation’s platforms of local institutions around specific infrastructures) and global connectivity (attracting scientific competences through international partnerships). In the meantime, the local authorities of Franche-Comté integrate only marginally the local-global connectivity as an issue of public research policies. In Alsace, the regional authorities strongly favoured the development of cross-border cooperation in several domains including academic activities, but the interregional networking is not yet so visible in the statistics (the University of Strasbourg is mainly a scientific global player).

45Spanish, Polish or Greek regions keep low profiles for collaborations, raising questions regarding their local strategies (as well as for European policies applied to those countries). The dynamics of German Länder – with a clear opposition between western nationally-oriented and eastern regionally-connected activities – also calls for caution while implementing homogenous national or European science policies that are mostly pushing for more global collaborations.

46Overall, the present contribution sheds lights on the reality of scientific collaborations and their evolution from a regional perspective. This should participate to the regional policy makers’ awareness regarding the local productions’ logics of science on their territories and the various modalities to enforce the production of new knowledge. Illustrated through the prism of different regions, the local-global connectivity appears in Europe as an increasingly crucial challenge variously identified, measured and tackled by regional authorities.

47Regional innovation systems are a complex set of interactions of local sub-systems. They have their own patterns of connectivity, which should be integrated in a more comprehensive development strategy both as an independent system (that sometimes overtakes its borders and governance perimeter – possibly being even a non-territorialised system) and as a key component of the RIS. Overarching our study, we must consider the governance mechanisms. It remains a complex parameter to both build a typology that integrates the various instruments and initiatives of local and national authorities tackling the issue of the local-global connectivity.

48We limited our analysis of scientific connectivity to co-publications measures. Future works will investigate and refine these results by including other forms of scientific collaborations (e.g. sharing of infrastructures or European project partnerships) but also by considering co-publications per discipline as well as interdisciplinary schemes of development. While perfectible, the new maps of scientific connectivity we have elaborated with European regions already challenge the European 2020 strategy. As underlined above, the smart specialization strategy of regions strongly involves the local research system (with its external links). A smart balance must be found between the necessary reinforcement of internal coherence, mainly around the project of collectively “discovering” the relevant specialization, and the challenge of positioning local research in the global world of science. In fact there is no contradiction if the territorial strategy is well defined. Local science systems, like innovative clusters, industrial fabrics, human and social capital, etc., are assets and opportunities for an adequate competitive positioning in the globalized economy.

Notes

  • [1]
    The NUTS nomenclature defines 276 NUTS-2 in Europe (Eurostat, RAMON). We retained a total of 263 NUTS-2 for our study because of data availability (e.g. no data available for Croatia for the period considered). In addition our study only considers “regions” but some European NUTS-2 regions are in fact countries like Estonia, Latvia, Luxembourg, Malta. As a result we excluded these lasts from our sample.
  • [2]
    10.84% for Cluster 1; 9.05% for Cluster 2; 8.46% for Cluster 3; and 11.15% for Cluster 4.
English

The production of science is often a collaborative activity. We call scientific connectivity the measure of such collaborative behavior, as reflected in the co-publication statistics. This article focus on the various forms of connectivity characterizing the scientific activities in Europe, viewed from a regional perspective. Are those networks local or global? Are they embedded in national systems? Is there an increasing tendency to collaborate between regions at European level? Such questions (and others) are addressed by constructing a typology of scientific connectivity in European regions (NUTS2) and observing its evolution over time (1999-2012). As a significant component of regional innovation systems (RIS), the types of scientific connectivity must be considered as a key issue for regional policies. Our results challenge political issues like specialization strategies, lock-in phenomenon, national and European policies for territorial development and cohesion, concentration of excellence, etc.
JEL Codes: R21

Keywords

  • local-global connectivity
  • science
  • regions
  • typology

References

  • En ligneADAMS, J. D., GRANT, G. C., CLEMMONS, J. R., STEPHAN, P. E. (2005), Scientific Teams and Institutional Collaborations: Evidence from US Universities, 1981-1999, Research Policy, 34(3), 259-285.
  • AMIN, A., COHENDET, P. (2004), Architectures of Knowledge: Firms, Capabilities, and Communities, Oxford, Oxford University Press.
  • ANOVA, R. W., LEYDESDORFF, L. (2001), Why Catalonia cannot be Considered as a Regional Innovation System, Scientometrics, 58(2), 215-240.
  • ASHEIM, B. T. (ed) (2006), Constructing Regioal Advantage: Principles, Perspectives, Policies, Brussels, European Commission Report.
  • En ligneASHEIM, B. T., COENEN, L. (2005), Knowledge Bases and Regional Innovation Systems: Comparing Nordic Clusters, Research Policy, 34, 1173-1190.
  • BARTHELT, H., MAINBERG, A., Maskell, P. (2004), Clusters and Knowledge: Local Buzz, Global Pipelines and the Process of Knowledge Creation, Human Geography, 28(1), 31-56.
  • BEAUDRY, C., BRESCHI, S. (2000), Does Clustering Really Help Firms’ Innovative Activities, CESPRI Working Paper n°111, University of Bocconi.
  • En ligneBEAVER, D. (2001), Reflections on Scientific Collaboration, Scientometrics, 52, 365-377.
  • BELLET M., COLLETIS G., LUNG Y. (eds) (1993), Économie de proximités, Revue d’Économie Régionale et Urbaine, 3, 357-602.
  • BENNEWORTH, P. (2010), Globalisation and Regional Studies for the 21st Century: Beyond Global Pipelines, Local Buzz, Conference of the Nordic Regional Studies Association.
  • BENNEWORTH, P., DASSEN, A. (2011), Strengthening Global-Local Connectivity in Regional Innovation Strategies, OECD Regional Development Working Paper 2011/01, Paris, OECD.
  • BENNEWORTH, P., HOSPERS, G. J. (2007), The New Economic Geography of Old Industrial Regions: Universities as Global/Local Pipelines, Environment and Planning: Government and Policy, 25(6), 779-802.
  • BENNEWORTH, P., COENEN, L., MOODYSSON, J., ASHEIM, B. T. (2009), Exploring the Multiple Roles of Lund University in Strengthening Scania’s Regional Innovation System: Towards Institutional Learning?, European Planning Studies 17(11), 1645-1664.
  • En ligneBRUIJN, P. J., LAGENDIJK, A. (2005), Regional Innovation Systems in the Lisbon Strategy, European Planning Studies, 13(8), 1153-1172.
  • En ligneBOSCHMA, R. (2005), Role of Proximity in Interaction and Performance: Conceptual and Empirical Challenges, Regional Studies 39(1), 41-45.
  • BOSCHMA, R., FRENKEN, K. (2010), The Spatial Evolution of Innovation Networks: A Proximity Perspective, in R. Boschma, R. Martin (eds) The Handbook of Evolutionary Economic Geography, Cheltenham, Edward Elgar, 120-135.
  • CHALAYE, S., MASSARD, N. (2008), La mesure de la production et des coopérations scientifiques dans l’agglomération stéphanoise: une approche par la veille scientifique territoriale, in T. Madiès, J.-C. Prager, Innovation et compétitivité des régions, Paris, La Documentation Française, 301-315.
  • En ligneCOOKE, P. (2001), Regional Innovation Systems, Clusters, and the Knowledge Economy, Industrial and Corporate Change, 10(4), 945-974.
  • En ligneCOOKE, P. (2005), Regionally Asymmetric Knowledge Capabilities and Open Innovation: Exploring ‘Globalisation 2’: A New Model of Industry Organization, Research Policy, 34, 1128-1149.
  • En ligneDOLOREUX, D., BITARD, P. (2005), Les systèmes régionaux d’innovation : discussion critique, Géographie, Economie et Société, 7, 21-36.
  • En ligneDORE, J. C., OJASOO, T., OKUBO, Y., DURAND, T., DUDOGNON, G., MIQUEL, J.-F. (1996), Correspondence Factor Analysis of the Publication Patterns of 48 Countries over the Period 1981-1992, Journal of the American Society for Information Science and Technology 47(8), 588-602.
  • EUROPEAN COMMISSION (2015), Horizon 2020. Work Programme 2014-2015: Spreading Excellence and Widening Participation (Revised), European Commission Decision C (2015) 2453 of 17 April.
  • EUROPEAN COMMISSION (2010), La contribution de la politique régionale à une croissante intelligente dans le cadre de la stratégie “Europe 2020”, Communication de la commission européenne COM (2010) 553 final.
  • FORAY, D., DAVID, P., HALL, B. (2009), Smart Specialization: The Concept, Knowledge Economists Policy Brief 9.
  • FRATESI, U., SENN, L. (eds) (2009), Growth and Innovation of Competitive Regions: The Role of Internal and External Connections, Berlin, Heidelberg, New York, Springer.
  • En ligneFRENKEN, K., HARDEMAN, S., HOEKMAN, J. (2009), Spatial Scientometrics: Towards a Cumulative Research Program, Journal of Infometrics, 3, 222-232.
  • En ligneGEORGIOU, L. (1998), Global Cooperation in Research, Research Policy, 27, 611-626.
  • En ligneGLANZEL, W. (2001), National Characteristics in International Scientific Co-Authorship Relations, Scientometrics, 51(1), 69-115.
  • En ligneHAZIR, S., AUTAN-BERNARD, C. (2014), Determinants of Cross-Regional R&D Cooperation: Some Empirical Evidence from Europe in Biotechnology, Annals of Regional Science, 53(2), 315-323.
  • En ligneHERAUD, J.-A. (2003), Regional Innovation Systems and European Research Policy: Convergence or Misunderstanding?, European Planning Studies, 11(1), 41-56.
  • En ligneHÉRAUD, J.-A., LACHMANN, J. (2015), L’évolution du système de recherche et d’innovation: ce que révèle la problématique du financement dans le cas français, Innovations, 46, 9-32.
  • En ligneHOEKMAN, J., FRENKEN, K., TJISSEN, R. J. (2010), Research Collaboration at a Distance: Changing Spatial Patterns of Scientific Collaboration within Europe, Research Policy, 39, 662-673.
  • En ligneKRATKE, S., BRANDT, A. (2009), Knowledge Networks as a Regional Development Resource: A Network Analysis of the Interlinks between Scientific Institutions and Regional Firms in the Metropolitan Region of Hannover, Germany, European Planning Studies, 17(1), 43-63.
  • En ligneLEYDESDORFF, L. (2003), The Mutual Information of University-Industry-Government Relations: An Indicator of the Triple Helix Dynamics, Scientometrics, 58(2), 445-467.
  • LEYDESDORFF, L. (2012), The Triple Helix, Quadruple Helix…. and an N-Tuple of Helices: Explanatory Models for Analysing the Knowledge-Based Economy?, Journal of the Knowledge Economy, 3(1), 25-35.
  • LUUKKONEN, T., PEARSSON, O., SIVERTTSEN, G. (1992), Understanding Patterns of International Scientific Collaboration, Science, Technology & Human values, 17(1), 101-126.
  • En ligneLUUKKONEN, T., TIJSSEN, R. J., PEARSSON, O., SIVERTTSEN, G. (1993), The Measurement of International Scientific Collaboration, Scientometrics, 28, 15-36.
  • MALECKI, E., OINAS, P. (1999), Making Connections: Technological Learning and Regional Economic Change, Aldershot, Ashgate.
  • En ligneMARKUSEN, A. (1994), Studying Regions by Studying Firms, The Professional Geographer, 46(4), 477-490.
  • En ligneMORENO, R., MIGUÉLEZ, E. (2012), A Relational Approach to the Geography of Innovation: A Typology of Regions, Journal of Economic Surveys, 26(3), 492-516.
  • NAVARRO, M., GIBAJA, J. J. (2009), Typologies of Innovation based on Statistical Analysis for European and Spanish Regions, Innova Working paper 2009.
  • OST (2009), Rapport des indicateurs de sciences et techniques, Paris, La Documentation française.
  • En ligneOKUBO, Y., ZITT, M. (2004), Searching for Research Integration across Europe: A Closer Look at International and Inter-Regional Collaboration, Science and Public Policy, 31, 213-226.
  • En ligneOUGHTON, C., LANDABASCO, M., MORGAN, K. (2002), The Regional Innovation Paradox: Innovation Policy and Industrial Policy, Journal of Technology Transfer, 27, 97-110.
  • En lignePERSSON, O., GLANZEL, W., DANELL, R. (2004), Inflationary Bibliometric Values: The Role of Scientific Collaboration and the Need for Relative Indicators in Evaluate Studies, Scientometrics, 30(3), 421-432.
  • En lignePINTO, H. (2009), The Diversity of Innovation in the European Union: Mapping Latent Dimensions and Regional Profiles, European planning studies, 17(2), 303-326.
  • TALBOT, D., KIRAT, T. (eds) (2005), Proximité et institutions: nouveaux éclairages, Economie et Institutions, 6&7.
  • En ligneTEIL, H. (1975), Correspondence Factor Analysis: An Outline of its Method, Journal of the International Association for Mathematical Geology, 7(1), 3-12.
  • TIJSSEN, R. W., VAN LEEUWEN, T. N. (2007), Research Cooperation within Europe: Bibliometric Views of Geographical Trends and Integration Processes, in Torres-Salinas, D., Moed, H. (eds), Proceeding Trends and Integration, proceedings of the ISSI 2007, CINDOC CSIC Madrid, -744.
  • En ligneTÖDLING, F., TRIPPL, M. (2005), One Size Fits All? Towards A Differentiated Regional Innovation Policy Approach, Research Policy, 34, 1203-1219.
  • En ligneTORRE, A. (2008), On the Role Played by Temporary Geographical Proximity in Knowledge Transmission, Regional Studies, 42(6), 869-889.
  • En ligneTORRE, A., RALLET, A. (2005), Proximity and Localization, Regional Studies, 39(1), 47-60.
  • UYARRA, E. (2010), What is Evolutionary about Regional Systems of Innovations? Implications for Regional Policy, Journal of Evolutionary Economy, 20, 115-137.
  • En ligneWAGNER, C. S. (2005), Six Case Studies of International Collaboration in Science, Scientometrics, 62(1), 3-26.
  • En ligneWAGNER, C. S, LEYDESDORFF, L. (2005), Network Structure, Self-Organization, and the Growth of International Collaboration in Science, Research Policy, 34, 1608-1618.
  • WAGNER, C. S, BRAHMAKULAM, I., JACKSON, B., WONG, A., YODA, T. (2001), Science and Technology Collaboration: Building Capacity in Developing Countries, Santa Monica, Rand Europe, Science and Technology.
  • En ligneZIMMERMANN, J.-B. (2008), Le territoire dans l’analyse économique. Proximité géographique et proximité organisée, Revue Française de Gestion, 184(4), 105-118.
  • En ligneZITT, M., BASSECOULARD, E., OKUBO, Y. (2000), Shadows of the Past in International Cooperation: Collaboration Profiles of the Top Five Producers of Science, Scientometrics, 47(3), 627-657.
  • En ligneZUCKER, L. G., DARBY, M. R., AMSTRONG, J. S. (1998), Geographically Localized Knowledge: Spillovers or Markets?, Economic Inquiry, 36, 65-86.
Mickael Benaim
Manchester Institute on Innovation Research
University of Manchester, UK
Jean-Alain Heraud
BETA– Université de Strasbourg, France
Valérie Mérindol
PSB Paris School of Business, France
Cette publication est la plus récente de l'auteur sur Cairn.info.
Mis en ligne sur Cairn.info le 16/09/2016
https://doi.org/10.3917/jie.021.0155
Pour citer cet article
Distribution électronique Cairn.info pour De Boeck Supérieur © De Boeck Supérieur. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
keyboard_arrow_up
Chargement
Chargement en cours.
Veuillez patienter...