Research integration

Transcription

1 Science and Public Policy, volume 31, number 3, June 2004, pages , Beech Tree Publishing, 10 Watford Close, Guildford, Surrey GU1 2EP, England Research integration Searching for research integration across Europe: a closer look at international and inter-regional collaboration in France Yoshiko Okubo and Michel Zitt Intra-European S&T coauthorships doubled in the last decade, but the collaborative network was surprisingly unaffected. The strongly voluntarist process in the European Union seems far from making collaboration unavoidable or homogeneous. We address the role of frontier regions in the collaborative network, taking France as an example. The study confirms the openness of border regions to their close neighbors abroad, but their level of preference for other regions within the same country is higher. We argue that internationalization/europeanization in science develops much more slowly than generally thought, and R&D programs aiming to develop multinational collaboration may be effective in spurring collaborative efforts, without greatly changing the patterns of durable scientific relationships established through producing coauthored articles. Yoshiko Okubo (corresponding author) is at the Laboratoire Strategie et Technologie, Ecole Centrale Paris; her home address is 362 bis rue de Vaugirard, Paris, France; Tel : ; okuboy@attglobal.net. Michel Zitt is at the LERCO, INRA, Nantes, Rue de la Geraudiere BP 71627, F NANTES Cedex 03 France; Tel: /74; Fax: ; zitt@nantes.inra.fr. Part of this article was presented at the MEXT/OECD Global Science Forum Workshop on Best Practices in International Scientific Cooperation, Tokyo, February The authors are grateful for the support of the Japanese Ministry of Education, Culture, Sport, Science and Technology (MEXT), and to Elise Bassecoulard for her technical assistance. THE NECESSITY OF SCIENTIFIC communication beyond national borders was recognized early on. When national scientific systems have been closed, it has generally been felt as an abnormality with counter-productive effects. de Beaver and Rosen (1978) showed that the rise of teamwork in its modern form is inherent to the professionalization of science as it occurred in Napoleonic France and later in England and Germany. Combining the tradition of transboundary communication with the rise of teamwork, international cooperation has become a strong feature of contemporary science. Modern means of communication have made scientific exchange even easier, reinforcing the natural trend of scientists to interact through complex patterns of cooperation and competition. International programs provide economic incentives to cooperate. However, collaboration among scientists of diverse nationality is not straightforward. Specific obstacles related to linguistic, cultural or geographical distance must be overcome in addition to the general constraints of working together. Collaboration is only one instance of internationalization and only a part of it takes the explicit form of copublished scientific articles, the contents and relations of which are amenable to bibliometric studies. The proportion of publications authored by institutions from different countries have practically doubled in the last decade, from less than 10% of all publication in 1990 to almost 20% ten years later (Observatorie des Science et des Tchnologies (OST) indicators, based on Science Citation Index (SCI) primary data). Science and Public Policy June /04/ US$08.00 Beech Tree Publishing

2 Yoshiko Okubo is a researcher at the Laboratoire Stratégie et Technologie, Ecole Centrale Paris. She is also Associate Researcher at the Center for Technology and Society, National Institute of Advanced Industrial Science and Technology (Tokyo). She holds a PhD in science, technology and society from the Conservatoire des Arts et Métiers (Paris). She worked as a consultant at the OECD and at UNESCO for seven years and has ten years experience in S&T studies in various countries including France, Japan, Sweden and the USA. Her main research interests are development of S&T indicators, internationalization in science, policy and mechanisms of S&T international cooperation and science policy of Europe and Japan. She has published in journals such as Scientometrics, Science and Public Policy, La Recherche and Research Policy. She is a member of the Editorial Advisory Board of Scientometrics. Michel Zitt holds a PhD in management sciences and a Doctorat d Etat in economics. He leads the IBIS scientometrics team at the Lereco/INRA Laboratory in Nantes and is a consultant at the Observatoire des Sciences et des Techniques in Paris. He is involved in the development of science indicators, especially those of France. His main interests are internationalization mechanisms and S&T dynamics, international collaboration, spatial dimensions of S&T activities and science technology relationships. On the methodological side, he worked on citation processes and co-citation mapping, internationalization measures, calibration of bibliometric indicators, publication patents informetric relations. He has published mainly in Scientometrics, JASIS and Research Policy. He is a member of the Editorial Board of Scientometrics and of the Scientific Committees of International Conferences Cycles in the field. In large European countries, such as France and Germany, the proportion is twice as high. The scope of collaboration for most countries has also increased and for the USA or any leading European country now encompasses almost all significant players in the world. The vigorous increase in international coauthorships over the last few decades expresses the strength of the globalization process. The contrast between the impressive growth of international coauthorships and the considerable inertia of the partnership structure of scientifically large European nations has been demonstrated in our earlier work (Zitt et al, 2000). In the span of a decade, coauthorship among European countries followed the world trend, but the international partnership profiles of France, Germany and the UK remained relatively stable. Preferential scientific relationships do exist for any nation, and they partially account for the weak mutual interactions among the three European leaders. The strongly voluntarist process in the European Union (EU) seems far from bringing about uniform collaboration. Leclerc et al (1992) stressed the difficulty of restructuring a network in the EU political context. Although research nets perform a significant function in Europe, they are still not yet visible as an integrating system at Community level (Leydesdorff, 2000). The Europeanization of science and its limits received a lot of attention in the 1990s (Narin and Whitlow, 1990; Moed et al, 1991; Leydesdorff, 1992; Grande and Peschke, 1999). In this article, we address the role of border regions in the European network of scientific collaboration, taking French regions adjacent to other countries as the basis of study. Constraints of geography and cultural differences are lower for regions on both sides of a border, so that a relatively high level of cross-border cooperation is expected. Border regions are expected to contribute to the European cooperative enterprise. At the same time, they may be a stabilizing element for collaborative partnerships in the EU zone, because of the strong scientific links created over a long period of time. First we provide some background on interpreting coauthorship. This if followed by methodology and then a depiction of the European research landscape. The final section is devoted to the interaction of border regions. Background: a new context for cooperation? Internationalization of science takes many forms (Okubo, 1996), including multinational infrastructures (Irvine et al, 1997), international funding programs, scientific interdependence, internationalization of scientific media, convergence of output level and specialization (Doré et al, 1996; Zitt and Bassecoulard, 1999; Zitt et al, 1999), and international collaboration (Okubo et al, 1992; Luukkonen et al, 1993; Miquel and Okubo, 1994; Katz and Martin, 1997; Havemann et al, 2000). Pavitt and Patel (1991) have cautioned against overestimating internationalization especially for corporate R&D, and OECD (1992) has advised the same for various aspects of scientific fields. The study of the motives underlying teamwork has produced results that can also be applied to international teamwork. Some motives have increased in strength over the last few decades. Competition has become keener, and cooperation appears to have become more complementary, than adverse, to competition, since pooling efforts can facilitate access to scarce funding, sophisticated equipment, databases, or large-scale projects at both the international and national level. The desire to combine competencies has always existed but is likely boosted by increasing specialization of scientists, which is traditionally considered a strong incentive for collaboration. In a competitive world, the incentive to copublish for better visibility (Herbertz, 1995), especially to copublish internationally, is also strong. International cooperation can even be a matter of survival for some scientific research outside the mainstream. A large literature is devoted to center periphery cooperation (Arvanitis and Gaillard, 1992; for catch-up nations within Europe, see, for example, Braun and Glaenzel, 1996). The mixture of cooperative and competitive behavior has prompted game theory models of cooperative rivalry by the new economics of science (Dasgupta and David, 1994). 214 Science and Public Policy June 2004

3 Bottom up versus top-down initiative Various intergovernmental agreements have developed and incited multinational participation in research activities. New possibilities are sought to improve organization and governance of science and technology (S&T) collaboration in diverse sectors involving funding institutions, universities, private research organizations, think tanks, industry and nonprofit organizations. The European Commission s European Research Area (ERA) initiative is Europe s most recent example of this trend at the policy level (EC, 2000). The purpose of the ERA is to create conditions that will lead to better resource use, a more coherent set of European policies, and a European R&D more attractive to researchers worldwide. The role of strong top-down processes has also come under discussion, both in principle and as to its efficiency. Literature describing incoherence between political and scientific aspects of international collaboration promoted by top-down initiatives is abundant. To fulfill the requirements of a program, technical, scientific or innovation aims may easily give way to pragmatic and political ends (Sharp, 1991). In another view, Ziman (1994) argues in his Prometheus Bound that international action should be limited to the provision of indivisible research facilities, or the improvement of the communication infrastructure. Any external interference with the strategic and technical autonomy of each research entity should be considered a restraint on trade, leading to suboptimal outcomes in terms of output efficiency. Scientists often favor bottom-up processes, in which research entities agree to work together with or without official arrangements; this is an example of the self-organizing mechanism studied by sociologists of science. Alongside international programs and transnational organizations, the development of international cooperation is anchored in the behavior of scientists who are increasingly cosmopolitan, aware of a competitive atmosphere. Organization of meetings and conferences, the worldwide market that exists for skilled scientific labor and personal migration (Halary, 1994), and ties between professors and former students are among Alongside international programs and transnational organizations, the development of international cooperation is anchored in the behavior of scientists who are increasingly cosmopolitan, aware of a competitive atmosphere the important sources of cooperation through selfregulation by scientific communities (Okubo and Sjöberg, 2000). An indirect indication of progress in global-scale multinational programs, as well as their limits, can be found in the comparison of bilateral with multilateral coauthorships (Miquel, 1996; Glaenzel and de Lange, 1997). If multilateral (>2 countries) authoring has grown by a factor of three in the decade, from 1% to 3% of published papers (source: Institute for Scientific Information (ISI)/OST), it represents only 10 15% of all international papers. This is also true for strongly multilateral papers (>5 countries), which represent a small fraction of the entire international publication. The situation varies widely according to field; physics, for example, is known for heavily coauthored papers. Nevertheless, this analysis shows that bilateral ties still prevail in international scientific collaboration, whether springing from small-scale research projects or from close relationships among individual scientists within large-scale projects. Large-scale projects do not necessarily result in multilateral coauthored research articles. This is especially true when the scope of the project is broad, thus encouraging juxtaposed research work under a common heading, rather than integrated activity (a target of the 6th EU Framework Program). Multilateral collaborative research that yields coauthored articles is, in fact, relatively rare even in fairly small international programs. The Human Frontier Science Program, which is exclusively focused on basic research, requires applicants from three different continents. It has been shown that more than 90% of articles in mainstream journals acknowledging support from this program did not have more than one grantee as an author, even though all teams had substantially benefited from the cooperation (Georghiou, 1998). Juxtaposed research might be an efficient way to stimulate researcher mobility, information exchange, crossfertilization, and access by researchers to new potential relationships, but a network specifically organized for a consortial-type effort does not necessarily outlive the project. In the last decade, new types of distributed research have emerged, such as international and intercontinental cooperation in the form of massive coordination programs, such as the Human Genome Project (Demiranda et al, 2004). Again, the need for a division of labor may be the motivating factor for collaboration, and such programs lead to various forms of joint products different from conventional article coauthorship, for example, the coauthoring of database segments. The landscape of collaboration is a subtle mix of top-down and bottom-up processes. It cannot be denied that generously funded programs investigating a new field or targeting strategic areas (information technology, molecular biology, biotechnology, nanotechnology) temporarily redesign the patterns of Science and Public Policy June

4 international cooperation, forcing it into new channels by making international cooperation a condition of funding (Marschan-Piekkari et al, 2001). As argued by Hayashi (2001), large institutional programs (such as the previous European Framework Programs) are typically top-down for general policy and priority and agenda setting, but leave ample room for scientists bottom-up practice. This may force researchers to consolidate their efforts in a new working environment. Behavior flexibility, as described by proponents of actor-network theory (Callon et al, 1986), depicts scientists reacting to changing environments with great efficiency and low cost, including a neuronal reinforcement of existing relationship webs and cost-free realignments of previous investigations along newly announced priorities. The market of funding sources constituted by a new program is an imperfect market; as actors become aware of its imperfections they may be expected to react in a way that ensures them a privileged position in the opportunity structure (Rip, 1986). To summarize, cooperation still appears mainly rooted in personal connections among increasingly international players, committed to a variety of programs, objectives and financial support sources, whatever their level institutional, regional, national or supranational. Collaboration requires trust and understanding among participants. The dynamics of interpersonal relationships are a critical element that will determine the success of any research project. 1 International cooperation involves, in addition, these dynamics playing across differences of culture, systems and values. Networking requires time, and scientific relationships, once formed, are likely to last. Selecting and committing to a partner is an important strategic choice for each individual researcher. It can be influenced by top-down decisions, but the basic structure of international collaborative networks constructed by researchers and their self-organizing mechanisms is unlikely to be profoundly altered by top-down actions. Global versus local initiative Movement towards internationalization is strong and growing steadily. The other side of this coin, however, is that, according to our reckoning, the majority of articles are still the result of a national effort, produced within each country s border. Even in strongly cooperative countries, such as France and Germany, 60% of publications were produced nationally in 1999; the figure was 65% for the UK (ISI/OST). In most mainstream countries, more than half of scientific publication is still a local product, essentially created in a sphere in which scientists share the same language and culture, the same institutional tradition, and a common national research system. Intranational collaboration, including within multinationally authored articles, is also growing. These observations suggest that the internationalization process in science can be overestimated. Despite some indications of convergence, the institutional culture of individual countries, or a group of countries with similar traditions, continues to shape innovation systems (Amable et al, 1997). Short-distance phenomena have attracted attention in recent decades, giving rise to discussions of the economics of proximity and to a cross-fertilization of economics, sociology, psychology, and even bibliometrics. For example, economists have stressed that spillovers and especially the transfer of tacit knowledge (as opposed to easily transferable codified knowledge ), are sensitive to spatial proximity (Jaffe, 1989). Scientific communication in the form of articles is to a degree codified knowledge, but the path leading to this outcome involves many non-codified exchanges enhanced by face-to-face interaction and mutual confidence (Storper and Venables, 2004). More than other forms of exchange, physical presence is mediated by proximity in a wider sense, such as shared nationality, or geographical, linguistic, and cultural proximity. Katz (1993) has shown the role of spatial proximity in fairly culturally homogenous spaces (within UK coauthorships). Zitt et al (2000) have provided examples in which cultural proximity overrides geographical proximity. Social proximity is a more complex factor; similar status in the social stratification of science is a significant factor in successful relationships at the local level (Kretschmer, 1996), but at the international level, an asymmetric relationship is often characteristic of cooperation as in professor pupil or center periphery relationships. The role of proximity or contiguity, especially for individual connections, may be formalized with various tools of graph techniques and social networks methodology, such as small world modeling (Watts and Strogatz, 1998) or the Erdös project, on mathematicians coauthorship, at the Oakland University. The importance of proximity in scientific projects has also been studied from a managerial point of view. For example, Cobbenhagen and Severijns (1999), addressing Good practice in interregional innovation policy, reported evidence that most of the elements having a negative influence on success can be attributed to geographical distance as well as cultural and language differences between the partners involved. A similar conclusion arises from another EU-related study that argues that collaboration would yield a positive result if partners are logically compatible, having similar powers and responsibilities and that there are no overriding cultural, philosophical, ideological or language constraints (RECITE Office, 1995). The articulation between local and global versions of science activity is, as we have seen, complex. Incentives to worldwide cooperation run up against the influence of physical proximity, while the increasing specialization of science also has an effect on patterns of collaboration. Border regions constitute a 216 Science and Public Policy June 2004

5 space where global incentives to international collaboration and local tendencies based on geographical and cultural affinities can both be found at work against a purely national logic. Yet, in other instances, national culture favors the globalization process over local initiatives. Historical and cultural ties with areas overseas often outweigh connections generated by geographical contiguity. Sources and methods Unless otherwise stated, figures are derived from the OST indicators based on the Science Citation Index data of the Institute for Scientific Information. 2 Regional assignments For European countries (EU-15), OST has carried out an almost complete assignment of SCI institutional addresses of authors to EU regions in the standard NUTS (Nomenclature d Unités Territoriales Standard) breakdown, 3 starting from NUTS-3 level. The NUTS-2 level is used here. Counting methods Two widely used methods are the fractional count and the full-distinct count. In the fractional count, every article has the same unitary weight, distributed among all contributing authors (affiliations). Therefore cooperation links are also counted in a way that sums to one. A bilateral cooperation within a massively multiauthored article will by this method be assigned a low rating. In the full-distinct count, each link between distinct countries is rated 1, so that the France Spain collaboration is simply the number of articles in which both countries appear, regardless of the number of French or Spanish institutions in an article. The implicit weight of an article is n(n 1)/2 for n distinct affiliations. Each of these two counting methods has its pros and cons. Fractional count is additive and consistent in scale changes. Full-distinct count carries implicit weighting and is inconsistent in scale changes, but is simpler to interpret. In practice, the two approaches are fairly convergent at the macro-level as far as patterns of collaboration are concerned, such as the ranking of partners for a given country. This convergence is less at a micro-level, since measures depend on a small number of articles. For convenience reasons the fractional count has been used in the coauthorship matrices below. Coauthorship indexes In this study, we used two indexes. The first is the gross measure of bilateral collaboration. This gross coauthorship flow, meaningful in terms of international balances, is to a large extent determined by the size of partners; for example the USA is expected to be the first partner for many countries in the world. It is useful, in addition, to provide a measure corrected for the size of partners. The mutual preference or probabilistic affinity index (PAI) is the ratio of observed coauthorship flow between two partners, c(ij), to the value that would be expected from their respective size under statistical independence hypothesis: c(i.)c(.j)/c(..). PAI = c(ij)c(..)/c(i.)c(.j) Where c(i.) and c(.j) are marginal sums and c(..) the sum of all cells. Keeping or simply discarding heavy PAI(i,i) in the diagonal, because of self-preference, perturbs the other cells and leads to undesirable effects. A correction has been introduced by forcing the diagonal to a neutral value using an iterative process. The index has also been standardized on the interval 1 and +1, with zero as the neutral point ( standardized probabilistic affinity index (), Zitt et al, 2000). A positive value means that the link between two countries is stronger than one would expect from each one s scientific size. If distances were abolished and no national preference prevailed, all values would be close to zero. Here the indexes are calculated on a complete matrix of all countries for international coauthorships, and on the EU-15 region s matrix for interregional flows. European scientific landscape Europeanization of science and technology At a macro-level, one aspect of the Europeanization of S&T is the intensity and orientation of collaborative linkages. This can be addressed in several ways. First, we can measure the proportion of European exchanges in each European country s profile, as a proportion of all international exchanges. 4 In Table 1, the most and the least Europeanized countries are listed. EU Member States represent 77.8% of Luxembourg s international cooperation, but only 35.4% of Germany s. The small cohesion countries of the EU (Spain, Portugal, Ireland and Greece) are particularly keen to work within the EU zone. In contrast, the biggest scientific countries, in particular, the trio Germany, UK and France, are the least Europeanized. This may be because of their particular linkages created with other major world players in science (such as the USA), or because of the geopolitical network each has constructed outside Science and Public Policy June

6 Table 1. EU countries collaboration with the 15 Member States as percentage of total external collaboration of EU Most Europeanized countries Least Europeanized countries % share EU index 100= %share EU index 100= Luxembourg Germany Portugal UK Ireland France Greece Sweden EU Source: OST, primary data ISI the European sphere through historical and cultural influences. Sixty percent of their international cooperation is with non-eu Member States. Over the period of a decade, Europeanization has made little progress, with a slight down-trend on the part of the most Europeanized (small) countries except Greece, and a slight up-trend for the least Europeanized countries. Let us now examine the mutual preferences as partners of science between pairs of EU nations, recalling that mutual preference is an observed/ expected ratio constructed as independent from the volume of coauthorship. 5 Pairs of nations with the strongest preference ( over 0.5) are selected in Table 2. It clearly shows that, within the EU zone, strong preferences exist, and the strongest ties re flect a mix of spatial, cultural and historical proximity, with the only exception being the pair Greece UK. We do not find in this list pairs of big European countries. In fact, preference levels among France, Germany and the UK are rather mediocre. Considering their geographical proximity, cooperative ties could have been expected to be much stronger. We have stated that historical and cultural linkages superseded spatial proximity in many cases. Alignment of disciplinary specialization of two partners may also have an impact on the overall figures. When spatial and cultural proximity overlap, for example, for France and (partly) French-speaking neighbors such as Belgium or Switzerland (outside the EU), a particularly strong bond is recorded. The The small countries of the EU (Spain, Portugal, Ireland, Greece) are particularly keen to work within the EU zone: in contrast, the biggest scientific countries, in particular, Germany, UK and France, are the least Europeanized same is true within the Scandinavian family or the German-speaking area. At the other end of the spectrum, it is not surprising to find among the lowest values (not shown), pairs of small distant partners with little common cultural background: Austria with Portugal; Greece with Sweden or Finland; Ireland with Portugal, Sweden, and so on. Table 2 shows a reduction of deviations: the strongest links tend to loosen slightly, while the weakest ones (not detailed here) tend to tighten slightly. These observations suggest that a movement towards more homogeneity is at work (for example, Frenken (2002) observes networks less constrained by geographical considerations), and that exclusive ties show a tendency to weaken. Nevertheless, the process is slow. The relative inertia of the structure of partnerships may be related to the personal nature of networks, mentioned above and Table 2. Pairs of EU Member States with the strongest mutual preference index (): evolution, Pair of countries delta * 0.9 Finland Sweden o Denmark Sweden - Spain Portugal o 0.8 Ireland UK - Netherlands Belgium Germany Austria Portugal UK France Belgium Greece UK + Denmark Finland Spain France Italy Spain Notes: Figures are based on three-year averages, for instance, 1998 indicates , and fractional counts are used Luxembourg, strongly linked to Belgium and France, is excluded because of the fluctuating figures caused by the small size of the country Indexes are calculated on a complete international matrix; for example, 0.6 corresponds to PAI 2 (twice the expectation), 0.8 to PAI 4 * Code: + (.0.03 points of index), o (weak change), - ( 0.03), ( 0.1) Source: OST, primary data: ISI 218 Science and Public Policy June 2004

7 documented in case studies by Laredo (1995) and Marschan-Piekkari et al (2001). Although a detailed analysis of the consequences of European programs on the development of collaboration is beyond the scope of this article, we can observe a relationship between EU funding policies and changes in the collaboration patterns of small countries. To analyze the effects of EU policy on competitiveness and cohesion, Sharp (1998) investigated the role of the Framework Programs (FP2 and FP3) in overall European S&T activities and found that the small EU countries of northern and southern Europe had been greatly favored on a per capita basis in comparison with large countries. As far as scientific cooperation is concerned, the main result of EU policy over a decade for these countries appears to be not so much an increase in the share of international cooperation that they carry out with the Community but rather a diversification of their partnerships within Europe, and also worldwide. Let us now describe the general situation of France, before approaching the question of border regions. France and the other EU Member States The Europeanization of French science can be assessed first by looking at Europe s share in France s overall international activities, then by paying heed to the strength of collaborative ties established between France and the 14 Member States. In the first case, we find that, although France s collaboration with EU countries doubled in volume, its propensity to adhere to the European competencies remained moderate. The portion of all French external S&T activity represented by European partnerships increased only slightly in the 1990s, from 40.4% to 41.8%, (leaving the great majority of French international activity conducted with nations outside the EU). Secondly, the network connecting France and the 14 other Member States is surprisingly stable. When the effect of country size is controlled by the indicator, France does not show a systematic preference for its geographical neighbors (Table 3). However, if the neighboring countries are (partly) French-speaking (Belgium, Luxembourg or Switzerland (not shown)), or of Latin culture (Spain, Portugal, Italy and Greece), a particularly high-level attraction is observed. In contrast, France is weakly attracted to Scandinavian and German- or Englishspeaking countries. France s relationships with scientifically big countries outside the EU (the USA or Japan, not shown here) are not strong either. By and large, traditional cultural affinities still dominate in French partnerships, although there is a slight weakening of affinities with French-speaking countries and some sign of strengthening scientific ties with non-latin countries inside and outside the EU (see also Wagner and Leydesdorff, 2003). To summarize, the Europeanization of French S&T activity can be measured in absolute terms by the rough doubling over a decade of the number of coauthored articles produced with EU countries. However, this trend should also be observed in an overall international context, where it is comparable to the general tendency worldwide. In relative terms, France s scientific relations with EU Member States show only a slight movement towards a greater homogeneity of collaborative preferences across the Union. A regional view Features of the French border regions France is located at one of the geographical crossroads of Europe. It is divided into 21 administrative regions, of which ten have an international border 6 that directly connects France to six EU countries and one non-eu country (Switzerland), not counting the city states of Monaco and Andorra. In terms of S&T, most of the prominent French regions are near borders that they share with rich and developed neighbors (Belgium, western border of Germany, Switzerland, northwest Italy, northeast Spain). The Table 3. Mutual preference index () between France and other EU countries: evolution between and Positive preference Negative preference Positive preference Negative preference 0.9 Luxembourg <0 Germany 0.9 <0 Germany, UK 0.8 < 0.1 UK 0.8 < 0.1 Netherlands, Ireland 0.7 Belgium, Spain < 0.2 Sweden, Ireland 0.7 Luxembourg, Belgium < Portugal < 0.3 Austria 0.6 < 0.3 Sweden, Denmark, Austria 0.5 < 0.4 Denmark 0.5 Portugal, Spain < 0.4 Finland 0.4 Italy, Greece < Italy, Greece < < 0.6 Finland 0.3 < < < < < Netherlands < < 0.9 Note: Figures based on 3-years averages, e.g signifies , and based on fractional counts. Indexes are calculated on a complete international matrix. Source: OST, primary data ISI Science and Public Policy June

8 northern and eastern French regions are on the rim of the European blue banana. The effects of European proximity are perceptible, particularly in foreign direct investment in border regions (German, Swiss, UK, Belgian). The enactment of the French Decentralization Law, the creation of the European Economic Community, the Maastricht treaty establishing the EU, and the increasing attempts of the Community to promote cross-border interregional cooperation have played an important role in developing border exchanges (for example, INTERREG I ; INTERREG II ; INTERREG III ; Estrosi, 1997). Most of the French border regions are cultural bridges, since political borders do not exactly follow linguistic or cultural borders. A part of northern France has Flemish roots; Alsace and part of Lorraine have a Germanic heritage; a majority of the population of the Catalan and Basque border cultural regions (in the extreme eastern and western portions of the Pyrénées, respectively) live within Spain but spread across the border into Languedoc-Roussillon and Aquitaine; the French Riviera bears a strong Italian imprint. If we transpose to the regional level national findings, we can expect that border regions combining spatial and cultural proximity will be the locus of particularly intense cooperation flows. Geographical factors should also be taken into consideration; physical, economic and social characteristics of border regions often resemble those of their neighbor, offering in principle clear opportunities for cooperation. However, this spatial border kinship is not without exception, for example, the Basque region is industrial only on the Spanish side. The Paris region (Ile-de-France) has considerable scientific weight in France and represents a particularly centralized aspect of French S&T structure. This region does not have any international border, but is not far from France s northern border. Ile-de- France produces 38% of French scientific publication while the ten border regions account for about 45%, and the inland regions only about 17% (fractional count, OST, 2002). In the current study we examine collaborative patterns between six French border regions and their Table 4. Structure of European partnership of six French border regions, French region Major partners of the region in rank order* Nord-Pas-de-Calais Belgium (27%), Germany, UK, Italy Alsace Germany (36%), UK, Italy Provence-ACA Italy (24%), UK, Germany, Spain Aquitaine UK, Spain ( 22%), Germany Midi-Pyrénées Germany, UK, Spain (16%), Belgium Languedoc- UK, Spain (19%), Roussillon Germany, Italy adjacent countries: Aquitaine, Midi-Pyrénées, and Languedoc-Roussillon with Spain; the region of Provence-Alpes-Côte-d Azur with Italy; Alsace with Germany; and Nord-Pas-de-Calais with Belgium. We shall first characterize the international profile of each region, before taking a closer look at interregional networks. Border regions and adjacent countries European partnership as % of international partnership Notes: *Only partner countries who show >10% of interactions are listed Border country in bold, with % of EU coauthorships of the region Table 4 shows the European and international partnerships of six French border regions and the corresponding EU countries. Belgium, the first partner of Nord-Pas-de-Calais (NPC), represents 27% of the total European collaborative activity of the region, while 40.5% of NPC s entire international activity is undertaken with EU countries. The table indicates that active collaboration takes place between the border regions and neighboring nations: 16 36% of the entire collaboration that a border region conducts with EU Member States takes place with its adjacent country, regardless of that country s scientific size. Taking a different angle, Table 5 shows the share of each border region in French collaboration with that region s neighboring country, the share of each Table 5. Role of six French border regions in the European partnership of France, French region Neighboring country A Region s share in total France/neighboringcountry partnership (%) B Region s share in total France/EU partnership (%) C Region s share in total French publications Nord-Pas de Calais Belgium Alsace Germany Provence-ACA Italy Aquitaine Spain Midi-Pyrénées Spain Languedoc-Roussillon Spain Science and Public Policy June 2004

9 border region in French/EU collaboration, and each region s share of total French publication. For example, NPC accounts for 7.5% of total France Belgium collaboration (column A), 3.1% of total France-EU collaboration (column B), and produces 3.3% of French total publication (column C). For all French border regions, their partnerships with their neighboring countries (column A) represent between 1.4 to 2.3 times more production than would be expected either by their shares of EU partnership in general (column B), or by the region s scientific size (column C). Alsace, for example, represents nearly 10% of the France Germany relationship. A strong mutual affinity between Alsace and Germany is perceived, whereas the France Germany relationship is, in general, weak. Alsace is thus an important actor in linking French and Germany in science. Table 5 illustrates how international collaboration between two regions can be relatively high when cultural and geographical proximities overlap. While a border region s preference for a neighboring nation can be impressive, the role of French border regions in the total EU collaboration is not overwhelming. Globally, the ten border regions represent 48% of France/EU collaboration, the same percentage of all French international collaboration, and 45% of national production, 7 while the ten non-border regions (Ile-de-France is excluded) account for 12% of French/EU collaboration, 13% of international collaboration, and 16% of national production. Ile-de-France taken alone shows a partnership profile closer to that of a border region. For regions shown in Table 5, EU collaboration figures (column B) are only slightly higher than expected from their scientific capacities (column C). There is little doubt that, enhanced by the effect of proximity, border regions are active local players in collaborating with adjacent countries. They therefore play a significant role as a bridge, but they do not necessarily play a significant role in the share of overall French international production. For example, the high level of centralization in France, with a strong concentration of science in Ile-de-France, mechanically limits the role that border regions can play in the nation s gross scientific transactions. The case of Ile-de-France, which maintains the same level of international collaboration as border regions, reminds us that other factors besides proximity, such as scientific excellence, contribute to international scientific activity. Enhanced by the effect of proximity, border regions are active local players in collaborating with adjacent countries: they therefore play a significant role as a bridge, but not necessarily in the share of overall French international production Let us now take a closer look at interregional networks. Interregional perspective The preferential S&T partners of six border regions are presented in Tables Nord-Pas-de-Calais (NPC) This region has a strong industrial tradition with an average level of scientific activity. The preference, as measured by, of NPC for Belgian regions is high (Table 6) and has not changed in ten years. In , the regional level of the French Belgian network was dense: linkages between NPC and Brabant Wallon or Oost-Vlanderen showed an above 0.9, which is quite high, and NPC and Vlaams Brabant figured above 0.6. Other Belgian partners, such as Hainaut, Liège and Namur also showed a strong link with NPC, even if these numbers are based on low gross volumes. In terms of mutual preference, NPC is more strongly linked with non-border regions of Belgium than with its adjacent regions. Alsace As previously mentioned, the overall mutual preference between France and Germany is practically neutral, an amazing result when considering the geographical proximity and the opportunity of cooperation offered by the European programs. Within this context, Alsace shows a strong collaborative tendency with nearby German regions (Table 7). Outside the EU, Alsace maintains close ties with neighboring Switzerland. Ten years ago, Alsace was more concentrated on national exchange, and Freiburg and Catalunya were its only foreign partners, both with a of at least 0.3 and a Table 6. Mutual preferences in Nord-Pas-de-Calais regional partnerships, Neighboring regions in France Other French regions Belgian neighboring region* Other Belgian regions Other EU regions 0.9 Picardie 10 regions Brabant, Liège, Bruxelles regions Hainaut (ns) Oost-Vlaanderen Castilla y Leon 0.3 Note: * NPC s neighboring region, West Vlaanderen was not among the preferred partners in Science and Public Policy June

10 Table 7. Mutual preferences in Alsace s regional partnerships, Neighboring regions in France Other French regions German neighboring region* Other German regions Other EU regions 0.9 Lorraine, Franche-Comté Ile-de-France, Languedoc-R 0.6 Other French Freiburg 0.3 Karlsruhe, Rheinhessen- Pfalz Brabant-Wallon (B), Liège (B) Table 8. Mutual preferences in Provence-Alpes-Côte-d Azur s regional partnerships, Neighboring regions in France Other French regions Italian neighboring region* Other Italian regions Other EU regions 0.9 Languedoc-R, Rhône- 5 regions** Alpes 0.6 Other French 0.3 Basse-Normandie Val d Aoste (ns) K Makedonia(GR), Aragon (E), Bruxelles (B), Liège (B), Brabant (B) Notes: * Liguria slightly below the threshold 0.3, Piemonte not among the preferred partners ** plus Corsica, overseas departments significant volume of copublication. Alsace therefore became more active in European cooperation, especially with neighboring foreign regions, though the preference values with the latter do not exceed 0.6. Provence-Alpes-Côte d Azur (PACA) This is the third largest scientific region in France, located just to the south of the Rhône Alpes region. PACA s collaborative ties are so strong with French regions that its external relations seem to be shadowed by this strong internal preference (Table 8). A decade ago, substantial exchanges took place with Tuscany, Liguria, and Catalonia, as seen in an value slightly below 0.3. In , however, PACA s only significant Italian partner was Val d Aoste with an under the threshold of 0.3. Regional proximity apparently does not, in the case of PACA, favor border collaboration. Southwest regions Three French regions (Tables 9 11) with a high level of academic activity border on Spain: Aquitaine (capital city Bordeaux), Midi- Pyrénées (Toulouse) and Languedoc-Roussillon (Montpelier). Spain has emerged as a major scientific player in Europe in the last two decades, while its S&T activity is concentrated particularly in the Madrid and Barcelona areas. In , partner ranking was slightly different. Aquitaine with Basque country (>0.9) and Aquitaine with Madrid both reflected a high level of preference, while several other Spanish regions with Aquitaine had a significant total exchange. For Midi-Pyrénées, Catalonia was the primary EU partner (>0.6), along with several Spanish regions that also showed a positive mutual preference, although exchange in absolute terms was low. Languedoc- Roussillon with Catalonia and with Madrid also showed a high mutual preference. These observations suggest that all three southwest border regions of France withdrew somewhat from regional border collaboration and extended their network more widely to other European regions. These observations highlight some basic facts about interregional cooperation. First, national partnership is of prime importance. Even for the border regions, which are likely to be open to international exchange, and sometimes share some cultural background with a region across the border, preferred partners are national ones. While an greater than 0.9 occurs frequently among French partners, no border region Table 9. Mutual preferences in Aquitaine s regional partnerships, Neighboring regions in France Other French regions Spanish neighboring region* Other Spanish regions Other EU regions 0.9 Poitou-C, Limousin, Midi-P Pays-de-Loire, Centre, Bretagne 0.6 Other French Basque Ctry, Valencia 0.3 Franche-C Navarra (ns) Catalunya Lisboa (P) Note: * Aragon not among the preferred partners 222 Science and Public Policy June 2004

11 Table 10. Mutual preferences in Midi-Pyrénées regional partnerships, French neighbors Other French Spanish neighbors* Other Spanish Other EU 0.9 Limousin, Languedoc, Aquitaine, Auvergne 8 regions regions K Makedonia (GR), 0.3 Champagne-Ardenne, Basse-Nor (ns) Liege (B) Catalunya Basque Ctry, Murcia Mittelfranken (D) Brabant (B) Note: * Aragon not among the preferred partners Table 11. Mutual preferences in Languedoc-Roussillon s regional partnerships, French neighbors Other French Spanish neighbors Other Spanish Other EU 0.9 Provence, Auverge, Midi-Pyrénées Franche-C., Bourgogne, Pays de Loire, Ile-de- France, Alsace 0.6 Other French Catalunya K Makedonia (GR), Liège (B) 0.3 Basse-Normandie Castilla y L, Andalucia reaches such a high value with a foreign region, except for the Belgian regions and NPC. Similarly, while most relationships between a French border region and non-adjacent French regions yield an of over 0.6, the figures for French border regions with foreign partners are mostly lower. Second, as might be expected, geographical proximity, often reinforced by cultural affinities, plays a decisive role in shaping a region s partnership network, both inside the country and across the border. In all the examples, the between two neighboring French regions is higher than 0.9, and readings of 0.98 or greater are not infrequent. Concerning foreign partnerships, neighboring country regions are, in general, the major partners of French border regions. In the neighboring country, next-door regions are, in most cases, the preferred partners. This hierarchy of preference is particularly clear in Alsace and Languedoc. Figure 1 shows the effect of geographical proximity on the mutual preference between Alsace and all European regions within a 1000km radius. 9 It is interesting to compare the between Alsace and French regions (Figure 2) with that between Alsace and German regions (Figure 3). 10 Alsace s preferred partners in S&T are clearly French, whatever the distance separating two regions. It is linked with German regions to a lesser degree, and drops sharply as distance from Alsace increases. A national preference is clearly demonstrated in these figures. The relationship between scientific attraction and geographical distance is significant (Figures 2 and 3), but it only explains a small portion of variance (R-square about 0.2). For several partner countries illustrated in Figure 1, such as the UK, the relationship is less clear. From these observations it is tempting to conclude that geographical proximity is merely one variable among others in a general model of cooperation (at the international level see the attempt by Nagpaul, 2003) ,8 0,6 0,4 0,2 0-0,2-0,4-0,6 SWE NLD LUX ITA UK FRA ESP DNK DEU BEL AUT 0,8 0,6 0,4 0,2 y = -0,0002x + 0,9189 R 2 = 0,2036 trend -0, DISTANCE Figure 1. affinity of Alsace to all regions within a 1000km radius DISTANCE Figure 2. Detail of Alsace with French regions (excluding Corsica and Monaco) Science and Public Policy June