International Scientific Collaboration: Patterns, Purposes, and Trends in EU Countries, Lecture notes of Technology

Download International Scientific Collaboration: Patterns, Purposes, and Trends in EU Countries and more Lecture notes Technology in PDF only on Docsity! Fraunhofer ISI, Idea Consult, SPRU The Impact of Collaboration on Europe's Scientific and Technological Performance Final Report Karlsruhe, Brussels, Brighton March 2009 Contents III 4.2.1 Introduction ........................................................................................ 84 4.2.1.1 Aim .................................................................................................... 84 4.2.1.2 Data Issues ....................................................................................... 84 4.2.1.3 Indicators Constructed ....................................................................... 85 4.2.2 Main Results: EPO data .................................................................... 86 4.2.2.1 Aggregate level: EU-15...................................................................... 86 4.2.2.2 Differences amongst the EU-15 countries.......................................... 87 4.2.2.3 Differences amongst Technical Fields ............................................... 89 4.2.2.4 Analysis according to EU-27 at the Aggregate level .......................... 91 4.2.2.5 Differences amongst EU-27 countries ............................................... 91 4.2.2.6 Collaborations with non-EU countries ................................................ 94 4.2.3 Main Results: USPTO data ................................................................ 95 4.2.3.1 Aggregate level: EU-27...................................................................... 95 4.2.3.2 Differences amongst the EU-27 ......................................................... 96 4.2.3.3 Differences amongst Technical Fields ............................................... 98 4.2.4 Conclusions ..................................................................................... 100 4.3 Network Analysis of Publications and Patents ................................. 100 4.3.1 Introduction ...................................................................................... 100 4.3.2 Methods to identify changes in cooperation among countries .......... 102 4.3.3 Science and research cooperations within the European Research Area ................................................................................ 106 4.3.3.1 Analysis of overall network density measures in publications .......... 106 4.3.3.2 Visualualisation of network structures for publications ..................... 107 4.3.3.3 Analysis of co-operation portfolios in publications ............................ 116 4.3.3.4 Analysis of breadth of co-operation portfolios in publications ........... 119 4.3.3.5 Analysis of the changing role of dominant players in publications ..................................................................................... 121 4.3.4 Innovation and Cooperation in the European Research Area .......... 123 4.3.4.1 Analysis of overall network density measures in patents ................. 124 4.3.4.2 Visualualisation of network structures for patents ............................ 125 4.3.4.3 Analysis of breadth of co-operation portfolios in publications ........... 129 4.3.4.4 Analysis of the changing role of dominant players in patents ........... 131 4.3.5 Conclusions ..................................................................................... 132 IV Contents 5 The Impact of Collaboration on Europe's Scientific and Technological Performance ............................................................................ 134 5.1 Introduction ...................................................................................... 134 5.1.1 Objectives of the ERA ...................................................................... 134 5.1.2 Objectives and implementation of the survey ................................... 136 5.2 Profile analysis of the respondents ................................................... 138 5.2.1 General Characteristics .................................................................... 138 5.2.2 Co-authors and co-inventors ............................................................ 141 5.3 Results ............................................................................................. 142 5.3.1 Collaboration behaviour and intensity .............................................. 142 5.3.1.1 Collaboration partners ...................................................................... 142 5.3.1.2 Collaboration intensity ...................................................................... 143 5.3.1.3 Types of collaboration ...................................................................... 144 5.3.2 Motives and barriers to collaboration ................................................ 145 5.3.2.1 Motives to collaboration ................................................................... 145 5.3.2.2 Barriers to collaboration ................................................................... 148 5.3.3 Value added of collaboration ............................................................ 150 5.3.4 Differences in motives, barriers and value added for intra-EU versus extra-EU collaboration .......................................................... 152 5.3.5 Role of ERA in collaboration ............................................................ 153 5.3.6 Preliminary conclusions ................................................................... 155 6 Policy relevant reflections ............................................................................... 158 6.1 The European Research Area .......................................................... 158 6.2 Collaboration from the researcher’s perspective .............................. 159 6.3 Higher levels of collaboration? ......................................................... 160 6.3.1 The ‘science’ sphere ........................................................................ 160 6.3.2 The ‘technology’ sphere ................................................................... 161 6.3.3 Network analysis .............................................................................. 162 6.4 Indications of increasing ‘integration’? .............................................. 163 6.5 Future monitoring of progress .......................................................... 163 6.6 Future research ................................................................................ 165 Contents V 7 References ....................................................................................................... 166 8 Appendices ...................................................................................................... 175 VIII Contents Figure 4-5: Shares of EU-single, Intra- and Extra-EU publications in EU-27 countries by scientific fields, 2004-2006 ................................................. 79 Figure 4-6: Shares of EU-single, Intra- and Extra-EU publications in EU-27 countries by scientific fields (alternative definition), 2004-2006 .............. 80 Figure 4-7: Number of publications and co-publications in EU-15 countries ............. 82 Figure 4-8: Shares of single-country, Intra-EU and Extra-EU publications in EU-15 countries, 2004-2006................................................................... 83 Figure 4-9: The extent of international co-patenting EU-15 as a whole (EPO) .......... 87 Figure 4-10: International Co-Patenting amongst EU-15 (EPO Data), 2000- 2004 ....................................................................................................... 88 Figure 4-11: Trends in International Co-Patenting amongst EU-15, 1990 to 2004 (EPO Data) .................................................................................... 89 Figure 4-12: International Co-Patenting of EU-15 countries in 6 Technical Fields, 2000-2004 (EPO Data) ............................................................... 90 Figure 4-13: Trends International Co-Patenting of EU-15 countries in 6 Technical Fields, 1990 to 2004 (EPO Data) ........................................... 90 Figure 4-14: Trends in international co-patenting: EU-27 as a whole (EPO Data) ...................................................................................................... 91 Figure 4-15: Differences amongst EU-27: All International Collaborations, 2000-2004 (EPO Data) ........................................................................... 92 Figure 4-16: Differences amongst EU-27: Intra-EU Collaborations, 2000-2004 (EPO) ..................................................................................................... 93 Figure 4-17: Differences amongst EU-27: Extra-EU Collaborations, 2000- 2004 (EPO) ............................................................................................ 93 Figure 4-18: Extra-EU Collaborations (1), 1990-94 to 2000-2004 (EPO) .................... 94 Figure 4-19: Extra-EU Collaborations (2), 1990-94 to 2000-2004 (EPO) .................... 95 Figure 4-20: The extent of international co-patenting EU-27 as a whole (USPTO) ................................................................................................ 96 Figure 4-21: International Co-Patenting amongst EU 27, 2000-2004 (USPTO Data) ...................................................................................................... 97 Figure 4-22: Trends in International Co-Patenting amongst EU 27, 1990-1994 to 2000-2004 (USPTO Data) .................................................................. 98 Figure 4-23: International Co-Patenting of EU-27 countries in 6 Technical Fields, 2000-2004 (USPTO Data)........................................................... 99 Contents IX Figure 4-24: Trends International Co-Patenting of EU-27 countries in 6 Technical Fields, 1990 to 2004 (USPTO Data) ...................................... 99 Figure 4-25: Network of EU-27 plus US & CH based on all publications in the period of 1994-1996............................................................................. 109 Figure 4-26: Cooperation network based on all publications in the period of 2004-2006 ........................................................................................... 110 Figure 4-27: Cooperation network based on all publications in the period of 1994 to 1996 using geographical locations .......................................... 111 Figure 4-28: Cooperation network based on all publications in the period of 2004 to 2006 using geographical locations .......................................... 113 Figure 4-29: Cooperation network based on publications in the field of engineering for the periods of 1994-1996 and 2004-2006 using geographical locations ......................................................................... 114 Figure 4-30: Cooperation network based on publications in the field of Life Sciences for the periods of 1994-1996 and 2004-2006 using geographical locations ......................................................................... 114 Figure 4-31: Cooperation network based on publications in the field of Medicine for the periods of 1994-1996 and 2004-2006 using geographical locations ......................................................................... 115 Figure 4-32: Cooperation network based on publications in the field of Natural Sciences for the periods of 1994-1996 and 2004-2006 using geographical locations ......................................................................... 115 Figure 4-33: Heatmap of country cooperation portfolios for all publications in the period of 1994-1996 ....................................................................... 117 Figure 4-34: Heatmap of country cooperation portfolios for all publications in the period of 2004 to 2006 ................................................................... 118 Figure 4-35: Boxplots of the Entropy distribution for selected fields for the periods of 1994-1996 and 2004-2006 .................................................. 121 Figure 4-36: Cooperation network based on patents in the field of Electrical Engineering for the periods of 1994-1996 and 2002-2004 using geographical locations ......................................................................... 125 Figure 4-37: Cooperation network based on patents in the field of ICT for the periods of 1994-1996 and 2002-2004 using geographical locations .............................................................................................. 126 Figure 4-38: Cooperation network based on patents in the field of Instruments for the periods of 1994-1996 and 2002-2004 using geographical locations .............................................................................................. 126 X Contents Figure 4-39: Cooperation network based on patents in the field of Chemistry for the periods of 1994-1996 and 2002-2004 using geographical locations ............................................................................................... 127 Figure 4-40: Cooperation network based on patents in the field of Mechanical Engineering for the periods of 1994-1996 and 2002-2004 using geographical locations .......................................................................... 127 Figure 4-41: Boxplots of the Entropy distribution for selected fields for the periods of 1994-1996 and 2004-2006 ................................................... 130 Figure 5-1: Distribution of the respondents over the different fields of expertise .............................................................................................. 140 Figure 5-2: Overview of the co-authors and co-inventors in the responses of the survey ............................................................................................ 141 Figure 5-3: Frequency of collaboration per geographic type of partners – number of respondents ........................................................................ 144 Figure 5-4: Motives for international collaboration – percentage of respondents that agrees, disagrees, or that is neutral .......................... 146 Figure 5-5: Barriers to international collaboration – percentage of respondents that agrees, disagrees or is neutral .................................. 149 Figure 5-6: Value added of collaboration – percentage of respondents .................. 151 Figure A.1: Heatmap of country cooperation portfolios in Electrical Engineering for the period of 1994-1996 and 2002-2004 ...................... 192 Figure A.2: Heatmap of country cooperation portfolios in ICT for the period of 1994-1996 and 2002-2004 ................................................................... 192 Figure A.3: Heatmap of country cooperation portfolios in Instruments for the period of 1994-1996 and 2002-2004 .................................................... 193 Figure A.4: Heatmap of country cooperation portfolios in Chemistry for the period of 1994-1996 and 2002-2004 .................................................... 193 Figure A.5: Heatmap of country cooperation portfolios in Mechanical Engineering for the period of 1994-1996 and 2002-2004 ...................... 194 Introduction 3 1 Introduction Modern high-technology and frontier research are complex, knowledge and resource intensive, and often boundary-spanning. Public research is nationally and internation- ally linked and parts of huge knowledge networks. Research in multinational compa- nies is often decentralised with project members at different locations within the same country but also very often at locations in different countries. The locations of knowl- edge, competences and resources steer the knowledge flows. In consequence, the international collaboration of public research and companies plays an increasing role, both for the national competitiveness as well as for new knowledge creation in general. Over the past about 20 years globalisation and internationalisation have accelerated, and while economic integration is perceived as its dominant feature, other dimensions including R&D but also the social, cultural, political and institutional realms are highly relevant too (OECD 2005). Knowledge production and R&D are seen as key compo- nents of this development (European Commission 2007). Thus, the understanding of the process of internationalization of R&D is indispensable for policy making and taking strategic decisions. This is also true in the context of the „European Research Area" (ERA), which was launched in the year 2000, aiming at further integration of the Euro- pean research system and achieving a higher degree of coordination and cooperation among the various players at all policy levels aiming at improved efficiency and effec- tiveness of still fragmented research efforts (European Commission 2007) in order to strengthen Europe’s international competitiveness. And even though milestones have been reached towards the ERA, progress is mixed and a lot still remains to be done. Decisions concerning required actions and measures need to be taken based on reli- able and valid information. Thus, it is crucial to have adequate tools for analysing the internationalisation process and its impact. Internationalisation of S&T can take various forms such as the mobility of researchers, collaboration between partners from different countries, research activities from institu- tions abroad, informal knowledge exchange, and systematic exploitation and applica- tion of foreign knowledge e.g. by being present in other countries for know-how acquisi- tion and networking (Edler et al. 2007). Thus, a variety of approaches and methodolo- gies are required to capture and analyse internationalisation in order to arrive at a comprehensive description of the processes and trends. So far the internationalisation of industrial R&D is a major issue being discussed in the scientific literature. Indicators applied to measure globalization are R&D expenditures or R&D personnel of foreign firms. Other studies focussed on the analysis of researcher’s mobility (OECD 2002). Very few studies attempted to draw a rather comprehensive picture of R&D interna- tionalisation by combining complementary methods of analysis. 4 Introduction This report concentrates on S&T collaboration and its measurement by indicators. The intention is to analyse the feasibility to regularly monitor developments concerning the evolving degree of integration of ERA. S&T collaboration in this project will be specifi- cally reflected by co-authorships (co-publishing) and co-patenting – knowing that these indicators cannot cover all aspects and all types of S&T collaboration. The motivations behind the internationally collaborative projects are manifold and range from personal networks to resource accesses. The output of these collaborations can also be manifold ranging from informal exchanges of ideas and knowledge to codi- fied output for example in the form of co-patents or co-publications. This latter codifica- tion can be measured and quantified. Next to a quantification and a structural analysis, a detailed examination of the driving forces is necessary to allow an overall assess- ment of the developments, of the trends as well as to derive adequate policy measures to foster international knowledge flows and create tailor-made environments and framework conditions for international research collaborations, especially against the background of the development of the European Research Area (ERA). Both aspects – the quantitative and the qualitative – of international research co-operations are taken into account in the course of this report. This documentation starts in chapter two – after this introduction – with a literature re- view on collaborations and the ERA in general as well as the use of co-publications and co-patents as indications of international collaboration. The third chapter presents the methodology and the obstacles to measure collaboration by co-patents and co- publications. Furthermore, the data sources are introduced, which are analysed in the fourth chapter. The results of an analysis of co-publications and co-patents are dis- cussed against the background of EU-15 and EU-27 cross-border research collabora- tions. Structures within the EU and with non-EU partners are examined. A network analysis complements this part of the report. As the motivations and driving forces be- hind the collaborations – measured and quantified by the data base analyses – cannot directly be derived from the quantitative sources, the results of a qualitative survey are introduced in chapter five, before the report concludes with some summarising re- marks. Literature Review 5 2 Literature Review 2.1 Importance and characteristics of RESEARCH collabo- ration 2.1.1 Collaboration from a policy perspective 2.1.1.1 Increasing interest from policy-makers The 'history' of the interest of European policy-makers for (international) collaboration in Science and Technology goes back to the period shortly after World War II. Indeed, the first initiatives in this regard were taken in the early 1950s with the creation of European intergovernmental research organisations such as CERN in 1954 (later on followed by ILL in 1967, EMBL in 1974, ESRF in 1996 etc). The rationale underlying the creation of large-scale, co-funded research organisations was that in some scien- tific or technological fields requiring large investments and complex infrastructures (typically 'Big Science fields'), operating research activities at world-class level would be too costly and too complex to be hosted by one single country. Creating inter- governmental organisations with co-funding from member countries was thus neces- sary to keep developing scientific and technological research in specific fields of strate- gic importance and according to the highest quality standards. These inter- governmental organisations were in a way the very first attempts to joint and to inte- grate European research activities, with the main aim to reach the higher critical mass required. They were also made possible by the more general trend towards European integration that finds its origin in the immediate aftermath of World War II. These first initiatives were reinforced by the development of a true EU-wide research policy from the early 1980s on. The Founding Treaties of the European Community did not initially provide the Community with an extensive responsibility in the field of Re- search. Until the late 1970s, European research policy mainly consisted of sectoral initiatives in areas such as nuclear energy, coal and steel and agriculture. A true Com- munity research policy, shifting from an ad hoc approach without an explicit legal base, towards an integrated vision for research only started in the 1980s, with the first EC Research Framework Programme (1984). On the basis of the positive experiences with this first pilot FP, a separate chapter on research and technology development was included in the Single European Act in 1986. Articles 163 to 173 of the Treaty establishing the European Union describe the objec- tives of EU RTD and define the Framework Programme as the basic mechanism for implementing this policy. The text of the Treaty is of course the basis, but not the sole 8 Literature Review • the development of a European research policy which not only addresses the fund- ing of research activities, but also takes account of all relevant aspects of other EU and national policies (EC 2002). In the last few years, and particularly with the measures implemented in the 6th Frame- work Programme (2002-2006), ERA has been transformed from a theoretical concept to a practical policy approach embodying many different dimensions. EU-wide, cross- border collaboration among scientists, research organisations (including universities) and enterprises is at the heart of the strategies and instruments implemented in the ERA context. In the following paragraphs we review the main instruments put in place under the ERA initiatives and with (expected) impact on collaboration activities). 2.1.1.3 ERA Policy instrument in support of collaboration ERA concerns both the Community and the Member States (including their regions) and the response has been significant at both levels. At EU level a number of actions have been launched since 2000 in support of ERA, notably through the 6th Framework Programme. Progress on some of these actions has been good though somewhat re- strained at times, while for others it has been more limited, pointing to the limits of what can be achieved at Community level alone. One of the notable developments has been the ERA-NET instrument which has made a start at addressing the inefficiency and fragmentation inherent in a system comprising numerous research funding schemes, spread across policy levels. The ERA-Net scheme was launched in 2002 as part of FP6. It aims at stimulating the cooperation and coordination between national (regional) research programmes, including their mutual opening and the development of joint calls. It typically targets research pro- grammes owners or managers (ministries, government agencies or research councils) and invites them to submit proposals in self-nominated topic areas (bottom-up princi- ple). The ERA-Net scheme is one of the flagship instruments of the European Com- mission for the further development of an integrated 'European Research Area' (ERA). Though the interest it provoked suggests that it responded to existing needs, the vol- ume of funding involved in the resulting joint activities is still marginal. Moreover, na- tional and regional 'programme-owners' are reluctant to restructure their programmes in a way which would enable the development of genuine joint programmes. Another area where good progress has been made is research infrastructures. A first major milestone was reached with the adoption of the European Strategy Forum for Research Infrastructures (ESFRI) Roadmap. However, the Roadmap will only be a success if the proposed projects are realised. For this to happen there is still a long Literature Review 9 way to go: New approaches are required - new legal, institutional and financial tools need to be developed. In the area of international cooperation, ITER2 has been a very visible success, and has demonstrated that Europe has the will and the capacity for leadership to address global challenges with partners around the world. However, while Europe is increas- ingly engaged in global science, research and infrastructure initiatives, these initiatives are far from systematic and often poorly coordinated with those of the Member States. Despite the success of important measures aimed at better exploiting human resources (such as the Marie Curie scheme, the European Charter for Researchers and the sci- entific visa package), Europe still lacks an open, competitive and attractive labour mar- ket for researchers. Some bright researchers and S&T graduates are still leaving Europe, others do not enter a research career in Europe or exit early, others miss op- portunities to move into positions where their capacities could be better used and de- veloped. At national level too, Member States have been involved in implementing actions which can help achieve ERA, for example: Some convergence in national policy making is materialising, driven in part by discus- sion and interaction between Member States and the Community level, such as through the Open Method of Coordination (OMC - launched in the context of the 3% Action Plan and overseen by CREST since 2003) or as a follow-up to Commission Communications. Trans-national and international cooperation are elements of most Member State re- search policies but, with some exceptions, still remain marginal in regard to the overall policy mix. In general, there is little evidence that national policy makers have taken ownership of the ERA concept, or have advanced far in their practical reflections on how national policy can contribute to constructing ERA, by building policy coherence across borders and across policy levels. Thus, progress at national level has also been mixed. As a result of these mixed outcomes, the EC has launched in the Spring 2007 a broad- based public consultation on the future of the ERA. The basic piece of evidence under- lying this consultation in the (2007) Green paper on the ERA, which has emphasised once again the importance of improved collaboration and co-operation with / between 2 International Thermonuclear Experimental Reactor. 10 Literature Review the key actors of Europe's research systems. According to the Green Paper, "the ERA that scientists, companies and citizens need should have the following 6 key features": 1. An adequate flow of competent researchers with high levels of mobility be- tween institutions, disciplines, sectors and countries; 2. World-class research infrastructures, integrated, networked and accessible to research teams from across Europe and the world, notably thanks to new gen- erations of electronic communication infrastructures; 3. Excellent research institutions engaged in effective public-private cooperation and partnerships, forming the core of research and innovation 'clusters' including 'virtual research communities', mostly specialised in interdisciplinary areas and attracting a critical mass of human and financial resources; 4. Effective knowledge-sharing notably between public research and industry, as well as with the public at large; 5. Well-coordinated research programmes and priorities, including a significant volume of jointly-programmed public research investment at European level in- volving common priorities, coordinated implementation and joint evaluation; and 6. A wide opening of the European Research Area to the world with special emphasis on neighbouring countries and a strong commitment to addressing global challenges with Europe's partners3. Improved coordination and cooperation with / between the key actors of the ERA are not only instrumental to increase the flows of people (pt 1) or knowledge (pt 4), or to achieve economies of scale (pt 2), they are also a mean to step up the quality of re- search towards world-class excellence through "effective participation in innovation clusters including virtual research communities" (pt 3). As we will see below, this increasing interest in collaboration from the policy side has influenced the patterns and intensities of national and international collaboration. 2.1.2 What does collaboration entail? 2.1.2.1 What is (research) collaboration? Collaboration in research and/or development is assumed to be 'a good thing' and thus it should be encouraged (Katz/Martin 1997). However, the interpretation of 'collabora- tion' is not an easy task. In an attempt to define 'research collaboration' on the level of 3 European Commission, The European Research Area: New Perspectives, Green Paper, Presented by the Commission (SEC(2007) 412), COM(2007) 161 final, Brussels, 4 April 2007, p. 2-3. Literature Review 13 Figure 2-1: Taxonomy of research partnerships by organizational structure Organizational structure of research partnerships Formal agreements Informal agreements Research corporations that are equity based Undefined arrangementResearch joint ventures Research contracts Source: Hagedoorn et al. (2000) According to Hagedoorn et al. little is known about informal partnerships, except the fact that a lot of companies are involved with one another in short term research en- deavours4. Firms team-up with other firms and/or universities in many various ways. Informal arrangements in this respect are mostly undefined and thus difficult to meas- ure. Two types of formal agreements are distinguished: equity joint ventures that focus on R&D (research corporations) and research joint ventures (RJV) which are mainly con- tractual arrangements (see also Hagedoorn 1990). Research joint ventures, such as joint R&D pacts or consortia to cover non-equity agreements, are created so that firms can undertake joint R&D activities. Although the success of such arrangements de- pends on the commitment of the partners, the collaboration can be terminated with only a relatively small loss compared to equity based arrangements. A specific subgroup of RJV are research contracts that concern R&D cooperation in which one firm contracts another firm to perform a particular research project. These types of collaboration may well lead to joint publications and/or patents, but their interpretation should be different. 4 As a result, the effect of informal research partnerships on the innovation performance of firms and in the end the financial performance of firms has not been studied in detail. This is largely caused by the difficulties of collecting data on informal partnerships. 14 Literature Review 2.1.2.3 Drivers of collaboration 2.1.2.3.1 The changing nature of research The current complexity in knowledge production and diffusion has proven the relatively older models of innovation, like the 'linear or chain-linked' model to be rather simplistic (see e.g. Kline/Rosenberg 1986). A reciprocal model (or network model) of knowledge production and diffusion is more appropriate in that sense (Gibbons et al. 1994). In the latter, the very nature of knowledge is evolving to a more network-oriented structure, with greater emphasis on strategic alliances, knowledge demand and supply chains and a growing transdisciplinarity and heterogeneity. Knowledge is not discrete and co- herent, and the production of it is not defined by clear rules and governed by settled routines. Instead, it is based on a mix of theory and practice, of abstraction and aggre- gation, coupling ideas and data from different origins and sources. The combination of these different origins and sources lead to further cross-fertilization and creation of new opportunities and becomes visible through intensified science – industry interactions. Two elements herein are important. Firstly, one of the drivers for increasing collabora- tion is the dynamic evolution of knowledge and knowledge creation itself (cfr. supra) that makes knowledge generation and thus research, a social process in which many different actors (academia, industry, policy, etc.) play a role. This has lead to the in- creasing awareness of research entities of their network embeddedness and to a boost in the search for partners (Mansfield 1991; Howells 2000). Secondly, over the last decade we have observed an increasing interwoveness be- tween science and technology. In many technological domains, and also sectors of industry we find an increasing proximity and interwoveness between science (i.e. crea- tion, discovery, examination, classification, reorganization and dissemination of knowl- edge on physical, biological or social subjects) and technology (i.e. the creation and use of artefacts, crafts and items of knowledge as well as various forms of social or- ganization). Modern technological areas have become highly scientific (Toynbee 1963; De Solla Price 1965; Narin and his colleagues at CHI; Schmoch 1997) thereby stimu- lating scientists to collaborate with one another. As a result, the classical distinction between industry and academia has faded, and increasing collaboration between the two has been observed. 2.1.2.3.2 Motives for collaboration As the above elaboration and explanation is of a more exogenous nature, there are also more endogenous factors that stimulate collaboration. Scientists are likely to col- Literature Review 15 laborate for reasons that go beyond scientific compatibility and complementarity. Among these factors are the following (Wagner et al. 2001; Katz/Martin 1997): • Geographic proximity: neighbouring countries often have similar research or com- plementary interests and common publication profiles. • History: Ties that form human, linguistic or other ties, as a result of historical interac- tions (including colonial relationships) support present day collaborations. • Common language: A shared language facilitates collaboration. • Specific problems and issues: Common problems, such as disease control or natu- ral disaster mitigation. • Economic factors: Factors include investment in a particular field because of re- search priorities set by scientists and policymakers, individual scientists collaborat- ing with particular universities, and the need to share facilities and equipment. Moreover, the costs of collaboration (travel, communication) have decreased strongly. • Expertise: Collaborations can be driven by the need for the best, or most appropri- ate, expertise to pursue the objectives of the scientific query. Many developing countries have institutions and individuals with world-class expertise. • Research equipment, databases, and laboratories: The presence of particular re- search equipment, databases, and laboratories in a country can give rise to interna- tional collaboration. • Political factors: Globalisation and internationalisation, the ambitions concerning ERA, support to third countries in dealing with global challenges etc. (see section 2.1.1). Hagedoorn et al. (2000) have further elaborated on these motives from a company perspective, thereby distinguishing among five approaches towards collaboration (see also Miotti/Sachwald 2003): 1) competitive force approach (strengthening positions), 2) strategic network approach (influencing agenda setting), 3) resource based view of the firm (scarce and unique capabilities), 4) dynamic capabilities (access and development of skills and capabilities), 5) strategic options to new technologies (forward looking). 2.1.3 Measuring 'collaboration' 2.1.3.1 Introduction As illustrated in the previous sections, collaboration is of particular interest for policy makers, in view of the various policy initiatives targeting the promotion and enhance- ment of scientific collaboration among researchers and research institutions. Also, in- ter-sectoral or interdisciplinary research collaboration is promoted by bringing together 18 Literature Review Katz and Martin (1997) argue that co-authorship should only be seen as a "partial indi- cator" to measure research collaboration because only those activities, which eventu- ally lead to jointly authored papers, are reflected. Not all collaborations, however, result in publications and conversely, a joint paper does not always mean that the results presented are based on research collaboration. Glänzel and Schubert (2004) on the other hand outline that this is in particular the case as far as intramural6 collaboration is concerned while in the case of international collaboration the parties involved are, as a rule, well acknowledged. They conclude that even taken into consideration those prob- lems as well as the phenomena of multi-institutional authors,7 co-authorship "seems to reflect research collaboration between institutions, regions, and countries in an ade- quate manner" (Glänzel/Schubert 2004: 259) and thus may be used as an analytical tool. Since the end of the 1970s a growing number of papers dealt with the issue of interna- tional research collaboration. This increasing interest in the topic is to be seen in the context of the significant increase of the extent of international research collaboration during the past decades (see chapter: Trends in scientific trans-national co-publishing). Among the first to have dealt with the topic of international collaboration in research are de Beaver and Rosen (1978, 1979). According to their findings until World War II inter- national collaboration grew rather slowly while afterwards a more rapid increase was observed. Apart from dealing with the degree or the intensity of international research collaboration its consequences for productivity is an interesting question that is being dealt with. Again already de Beaver and Rosen (1978, 1979) analysed this relationship (see section: Trends in scientific trans-national co-publishing). Based on their findings they concluded that collaboration results in higher publication activity. Another interesting and relevant issue is the question for what reason international col- laboration is engaged in. What are the motive and drivers that might explain this phe- nomenon? This issue will be dealt with in a specific chapter at the end (see section: Motives and drivers). Intra-scientific factors are to a large extent motivating interna- tional collaboration. In particular the desire to enhance the scientific knowledge, ex- changing skills and data and to enhance professionality (de Beaver/Rosen 1979; Luukkonen et al. 1993) is relevant but other factors are of relevance too. 6 Intramural collaboration = collaboration within a research group, a department or an institu- te. 7 Multi-institutional authors are authors involved with two or even more institutions. Literature Review 19 The analysis of patterns of collaboration in the sense of identifying networks between actors (here: countries) and the rule certain actors play within those networks is of in- creasing interest in the context of analysing international collaboration. Meanwhile also here a growing number of publications can be found, often focussing on specific fields of science or on specific countries or regions. 2.2.1 Methodologies to analyse trans-national co-publishing The analysis of international collaboration is primarily done by using co-authorship data gathered from the Science Citation Index (SCI) so far. The SCI is an international mul- tidisciplinary data base produced by Thomson Reuter. The data base can be accessed via the internet. Alternatively commercial hosts such as STN also offer access to the data base, which is furthermore also available on CD Rom. The SCI for a long time had, compared to other bibliographic data bases, the advantage that affiliation informa- tion for all authors of a scientific publication was covered. Only this information enabled the analysis of co-authorships at various levels of aggregation (e.g. inter-institutional or international co-authorships). From a technical point of view co-publication data can be dealt with in different ways. Publications can be assigned to a country using (1) fractional counting or (2) whole counting8. Fractional counting assumes that all authors named, contributed equally to the publication and thus, each author or institution is assigned the same fraction. Thus, a paper is accounted to a country proportionally to the number of addresses given from each country. However, the underlying assumption of equal contributions is not proven and thus, the integer counting method, which assigns a paper fully to each participating country can be used as an alternative too. This method leads to the fact that, if publica- tion shares of all countries are added up, the sum exceeds 100 per cent. Thus, it fa- vours those countries with a high propensity to collaborate internationally. However, both methods are applied in recent studies analysing international research collabora- tion. Results and interpretations gained from the analyses may differ depending on the counting method used. Based on data from the NSF's Science and Engineering Indica- tors Report 2002 (National Science Board 2002), where a 10 per cent decline of the US publication output was found for the period 1992-1999, it was stated that the fractional counting method is "biased against growth, and highlighted the possible effect of dis- placement of papers from 'established' countries, particularly the USA, by those from developing ones. In addition, it was observed that the absolute – wholly counted – number – of US papers did show growth, and seemed at least to suggest that this pat- 8 For a more detailed discussion of the use of fractional versus whole counting see for in- stance Persson and Danell (2004). 20 Literature Review tern may reflect more properly the trend in the US science system's performance than the fractional counting method" (Moed 2005: 274). Consequently, in order to allow for an informed assessment of the data presented it is important to always state precisely which method was used. Luukkonen et al. (1993) outline problems occurring when calculating indicators reflect- ing patterns and degree of international collaboration. Often analyses are based on simple shares of joint papers. As a relevant factor influencing a country's collaboration propensity and intensity they identified its size. Already Frame and Carpenter (1979) stated that the size of a country influences its propensity to collaborate internationally. As a consequence, in order to derive a picture adequately reflecting collaboration be- tween countries, absolute as well as relative measures should be taken into account and indicators reflecting international collaboration need to be normalized taking into account the size of the countries under investigation. Often used to normalize collabo- ration data are the Salton's or Jaccard Index. However, somewhat contradictory to the above mentioned relationship are findings by the same authors. Luukkonen et al. (1992: 123) found, based on a macro-level analysis of international collaboration, that the relationship between the size of scientific output and the rate of international collaboration is relatively weak. Their analysis, carried out for 30 countries, is based on SCI data for 1981 to 1986. In the paper it is attempted to explain country-to-country differences in the rates of international collaboration. In or- der to do so geopolitical, historical factors and language as well as the relevance of social, intellectual, cognitive and economic factors were taken into consideration. Simi- larly also Narin et al. (1991) stated the relatively weak relationship between the size of a country and the degree of international co-authorship. However also in the Narin et al. paper conflicting statements may be found. While on page 317 they argue that "first, international coauthorship is increasing steadily, and second, it is higher for scientifi- cally smaller countries", they also attempt to explain this phenomenon. "The second point is, of course, a direct consequence of scientific size. Scientists in countries such as Italy have far more scientists outside their country to cooperate with, and far fewer inside, than scientists of much larger countries such as the United States or the United Kingdom." (Narin et al. 1991: 317). This result is based on the analysis of only five ma- jor EU countries. Using a larger country set, including non-EU countries, they state that "total coauthorship seems to be determined by factors other than size" (Narin et al. 1991: 319). Schubert and Braun (1990) again conclude, based on an analysis of SCI data for the years 1981-1985, on the contrary that the "general tendency is that scien- tifically small countries have more foreign co-authorships than scientifically large coun- tries. Their explanation is similar to that given by Narin et al. (1991) cited above. Literature Review 23 ble exhibits the general trends on co-authorship data for the period 1988 to 2005. While in 1988 the rate of internationally co-authored papers was slightly above 8% it mean- while grew to more than 20%. Table 2-2: Share of worldwide S&E articles co-authored domestically and interna- tionally: 1988-2005 (National Science Board 2008: 5.42) All coauthorship Domestic coauthorship only International coauthorship 1988 40,0 31,7 8,3 1989 41,1 32,2 8,9 1990 42,2 32,7 9,5 1991 44,1 33,4 10,6 1992 45,1 33,7 11,4 1993 46,4 34,0 12,4 1994 47,5 34,4 13,1 1995 49,1 35,2 13,9 1996 50,4 35,7 14,7 1997 51,9 36,3 15,6 1998 52,9 36,6 16,3 1999 54,2 37,1 17,1 2000 55,1 37,4 17,7 2001 56,8 38,1 18,6 2002 57,8 38,6 19,2 2003 59,1 39,3 19,8 2004 60,2 40,1 20,1 2005 61,2 40,7 20,4 NOTES: Article counts from set of journals covered by Science Citation Index (SCI) and Social Sciences Citation Index (SSCI). Articles are classified by year they entered database, rather than year of publication, and assigned to region/country/economy on basis of institutional ad- dress(es) listed on article. Articles on whole-count basis, i.e., each collaborating institution or country credited one count. Internationally coauthored articles may also have multiple domestic coauthors. SOURCES: Thomson Scientific, SCI and SSCI, http://scientific.thomson.com/products/categories/citation/; ipIQ, Inc.; and National Science Foundation, Division of Science Resources Statistics, special tabulations. Hinze et al. (2007) specifically look at the developments in Germany and for the EU-27 countries. According to their findings the share of internationally co-authored papers for Germany grew from about 19% in 1990 to about 44% in 2006. Within this period the share of the EU-27 countries increased from about 11% to about 23%. Differences concerning the share of internationally co-authored papers can be found for individual countries, which can also be seen from the following table which is drawn from Moed (2005). 24 Literature Review Table 2-3: International collaboration for 15 major countries (Moed 2005: 298) Country International coauthorship (in %) USA 14,6 UK 22,6 Japan 13,6 Germany 28,8 France 30,2 Canada 26,2 Italy 30,8 Australia 23,6 India 12,2 China 23,5 Netherlands 31,9 Russia 26,4 Spain 28,4 Sweden 33,8 Switzerland 42,1 According to his data international co-authorship is highest in Switzerland and lowest in India and Japan, but also for the USA the share is comparably low. Hinze et al. (2007) found similar results even though at higher absolute levels, which can be partly ex- plained by the fact that data for a different time period was used. In addition Hinze et al. analysed a slightly different set of countries which includes the Scandinavian countries. The latter expressing similarly high levels of internationally co-authored papers as it was found for Switzerland. According to Hinze et al. the share of internationally co- authored papers for Switzerland recently amounts to about 58%, following are Den- mark (about 56%), Austria (about 55%), Norway (about 52%) and Sweden (about 50%). Similar levels as identified for Germany can be found for France (43%), the UK and Canada (42%). Also here the USA are at the lower end with about 25% and thus, at a level comparable to that of Korea and slightly above Japan. For the latter two countries their geographic locations as well as language problems were assumed to impede stronger inclusion into international networks. Frietsch et al. (2008) identified a kind of cultural effect as the Asian countries – they used data for Japan, Korea, China and India in comparison to other OECD countries – collaborate internationally on a much lower level. And this level seems to be persistent and similar over time, even given the strong increase of absolute numbers of scientific publications emerging out of these countries. Also Narin et al. (1991) highlight the steady increase of internationally co-authored pa- pers. They furthermore found out that this holds true for either inside as well as outside the EU. They found some evidence that the intra-EU increase was slightly higher in areas specifically targeted by the Commission (Narin et al. 1991: 323). Literature Review 25 As already shown by the data presented above, variations can be found among coun- tries but also among fields of research (Moed 2005: 285; Luukkonen et al. 1992; Hinze et al. 2007). Drawing on data gathered within the context of the recently performed exercise within the EU-27 countries the comparison between research fields shows that the share is highest in Multidisciplinary Research (51%) and Physics and Geo- sciences (49%) and lowest in Chemical Engineering (22%) and Basic Chemistry (23%). Similar results were found by Moed (2005). Hinze et al. (2007) specifically analysed patterns of collaboration for Germany. While they found increasing international collaboration with all countries analysed9, the in- crease differed between countries. For instance, it was highest for Korea, which might at least partly be explained by the still comparably low degree of collaboration between both countries in total (p. 19). Differences exist also if fields of science are compared. Comparing data for 1996 and 2006 for four segments of science – engineering, natural sciences, life sciences and medicine – increasing shares of international co-publication activities were found for all those segments. The share of international co-authorship is, however, highest in the natural sciences. It grew from about 36% in 1996 to about 53% in 2006. Second are the life sciences. Here the respective values are 34% in 1996 and 48% in 2006. In engineering in 1996 about 28% of all publications were internationally co-authored, while in 2006 this share came up to 42%. Lowest are the respective rates for Medicine with about 21% in 1996 and 36% in 2006. Still, if normalised for the gen- eral growth of the individual segments the annual increase was found to be highest in Medicine (5.5%) and lowest for the life sciences (3.3%). For the social sciences and / or humanities no data was presented, as they are much more nationally oriented. Fur- thermore, country comparisons are hardly possible based in this data set as non- English speaking countries are underrepresented for the reason that arts and humani- ties are also much more frequently published in national languages. In addition, for Germany it was also found that international collaboration is growing more rapidly with other EU countries that for instance with the US (Hinze et al. 2007: 19/20). A finding confirming what was stated by Narin et al. (1991) more generally for intra-EU collaboration before. These are first findings which may indeed point towards the development or emergence of a European Research Area. Other studies focussing on international collaboration for particular countries exist such as for Korea (Kim 2005), China (Zhou/Leydesdorf 2006), and Turkey (Uzun 2006). 9 In their study they analysed co-authorships between German authors and authors from 15 mainly European but also other countries e.g. USA, Japan, Korea. 28 Literature Review 2.2.3 Relationships between collaboration and impact of research As already mentioned at the beginning the relationship between collaboration and pro- ductivity is one interesting question that was dealt with by a number of studies basically starting with the pioneering work by de Beaver and Rosen (1978, 1979) and their con- clusion that collaboration enhances productivity. A couple of studies analysed the relationship between international collaboration and the impact of the respective research. In particular it was investigated whether higher citation rates may be found for internationally co-authored papers. Narin et al. (1991) as well as Gomez et al. (1995) reported that internationally co-authored papers on av- erage tend to have higher citation rates. Similar results were found by Glänzel and Schubert (2001) and Glänzel and Schubert (2004). Also Katz and Martin (1997: 15) state that on average a paper written by multiple authors is likely to be more frequently cited and thus, has a higher impact. Glänzel (2001) confirms the above finding at the national level. However, he also found differences between fields of science if pairs of countries were analysed. While in bio- medical research the observed citation rates of almost all pairs of countries were above the domestic values this seems to be different in chemistry and mathematics where for some pairs of countries very low citation rates were found. In particular this was the case if developing or Eastern European countries were concerned. Thus he concludes that international co-authorship does not pay for all partners. However, the explana- tions for these findings remain unclear. According to the findings by Narin et al. (1991) internationally co-authored papers were cited two times higher than single institutional, single country papers. At the same time it was found that there was no difference concerning the citation impact between intra- EU papers and internationally co-authored papers from Non-EU authors. With other words, for internationally co-authored European papers the impact was as high as for any others in the world. More recently Moed (2005) analysed the relationship between international collabora- tion and citation impact. His particular focus was on bilateral international collaboration. According to his findings the picture is rather mixed. Whether or not international col- laboration leads to higher citation rates depends on who is collaborating with whom. In the case that scientifically advanced countries collaborate with each other in a specific area there seems to be a positive effect, meaning that the citation impact of those pa- pers is most often higher than it is the case for purely domestic papers. However, in the case that scientifically advanced countries engage in collaborative papers with scien- tifically less advanced countries the outcome may negatively affect the citation rates of the advanced country (p. 290). Literature Review 29 2.3 Co-Patenting 2.3.1 General Overview The literature on co-patenting is comparatively small.10 To inform the literature review a brief bibliometric analysis was carried out. In the Web of Knowledge not more than 50 publications can be identified that are related to co-invention or co-patenting.11 A sub- stantial share of these papers are not immediately relevant to this study as they refer to particular instances of co-inventions and describe accounts of individual inventors rather than offer studies and analyses of co-patenting. The most relevant papers and reports are summarised in Appendix 1. Apart from very few but notable exceptions (e.g. Guellec/van Pottelsberghe de la Pot- terie 2001; Guellec/Pluvia Zuniga 2007; Edler 2004; Edler et al. 2003), co-patenting data has been used in general overviews of patent statistics, such as the OECD (2007) compendium of patent statistics, or as supplementary information in reports on interna- tionalisation of R&D (see e.g. the 2005 report by Arthur D Little on the UK). The focus of these studies and reports is here indeed on transnational knowledge flows or trans- fers at the country level. In addition, co-patenting data tends also to be offered as com- plementary information on international collaboration in domain studies of new tech- nologies (e.g. OECD 2005; Glänzel et al. 2003a, 2003b). Increasingly co-patenting is being explored in other contexts, for instance, academia- industry collaboration (e.g. Lissoni et al. 2008) or econometric studies to explore re- search and development collaboration within the context of regional innovation systems (e.g. Maggioni et al. 2007). While the focus of this review is on international flows and exchange processes, we will briefly point to examples of co-patenting studies covering also these aspects. Apart from co-patent data, other tools and techniques have been employed to explore international knowledge flows. Also here we will briefly refer to exemplary studies. 10 A recent literature review (Fontana/Geuna 2008) confirms the impression that there are very few studies that focus exclusively on co-patenting. 11 The Web of Knowledge search covered the Social Sciences, Arts and Humanities, and Science Citation Indexes as well as the proceedings databases by Thomson-ISI. A search in the Scopus database identifies around fifty publications; after closer inspection only a handful of publications seemed pertinent to this study. 30 Literature Review 2.3.2 Different Types of International Co-patenting Generally speaking, one can distinguish different types of international co-patenting (e.g. Grupp 1997; Grupp/Schmoch 1992; Hullmann 2001): • Domestic inventor, foreign applicant (or assignee) • Domestic applicant, foreign inventor • Domestic and foreign inventors • Domestic and foreign applicants Figure 2-2 illustrates the various possible combinations. These categories can be used to track co-patenting. However, it is much more difficult to make judgements about knowledge flow or exchange processes on the basis of patenting data. The direction of the knowledge flow is often challenging to trace. For instance, it is difficult to say much about the extent and direction of knowledge flows if inventor teams are from different countries. One would assume that knowledge has been transferred in both directions but it is impossible to make judgments as to which partner benefited the most. Also, having both domestic applicants and inventors does not necessarily mean the invention is utilised in that country. Licensing arrangements (which cannot be tracked through patent analysis) might well assign the exclusive right of use of the invention to a foreign company.12 When interpreting co-patenting data it is also important to recognise that there is a dif- ference between the concept of domestic inventors and a perspective that is based on the nationality of inventors. Patent data can offer only information on the former, not on the latter aspect. This means that co-invention data is based on residential addresses of inventors as they are listed in patent documents. A foreign national would be counted as domestic as long as he or she lives within the country analysed.13 At the organisational level, co-patenting activity is at times seen as an "imperfect proxy for collaboration among firms since it only picks up collaboration which result[s] in pat- enting, and since it also may involve inventors from the same company located across its various subsidiaries, the data reflects both inter – and intra-firm international col- laboration" (ADL 2005: 93). 12 Also note that the term 'co-patenting' can be defined in different ways. While most analyst will define co-inventions and co-assignations as co-patenting, the situation might be less clear when one looks at combinations of foreign ownership of domestic inventions, domes- tic ownership of foreign inventions. Some analysts (e.g. Hullmann 2001) follow this very broad perspective while other may view 'foreign owned national inventions' and 'nationally owned foreign inventions' as a group of indicators which is essentially different from co- patenting. 13 These sorts of aspects will be emphasised even more if one focuses on regions. Literature Review 33 all most economies have become more strongly involved in cross-border inventive activity over the course of this decade. The share of foreign inventions owned by domestic companies has more than doubled in Brazil, Finland, India and Sweden as compared to the early 1990's.17 Comparing EU and US, it is interesting to note that more than 50% of inventions with cross-border ownership in 2001-03 were made with inventors located in European countries, which represented twice the number of inventions made by US inventors (Figure 2-5). The breakdown at the country-level points to the importance of geographical and cultural proximity in the choice of loca- tion. European countries own inventions from other EU countries more frequently than from other locations. When intra-EU locations are netted out, the United States is the leading location.18 • the share for a given country of patents with a domestic inventor and a foreign applicant in the country's total domestic inventions ('SHIA'). It reflects the extent to which foreign firms control (own) domestic inventions (Figure 2-5). Recent OECD data suggests that, on average, nearly 17% of all inventions filed at the EPO were owned or co-owned by a foreign resident in 2001-03, which is a substantial increase from less than 12% in 1991-93. Having said this, one must note that there is consid- erable variation from country to country.19 than 30% in the cases of Belgium, Ireland, the Netherlands, Singapore, Sweden and Swit- zerland. In some of the aforementioned countries, foreign countries have established re- search labs to a larger extent than elsewhere (e.g., in Belgium in the pharmaceuticals and biotechnology areas and Switzerland in the electronics sector; Singapore pursues a policy of attracting companies and researchers to build up science-based industry, with Biopolis being the most prominent amongst a number of initiatives). In other countries, such as Sweden, some of the national players (esp. in the research-intensive pharmaceuticals sec- tor have merged with or been acquired by other MNE's. Italy, Japan, Korea and Spain report the weakest share of inventions made abroad (less than 10%). A number of factors could explain this situation. It is argued that especially Asian firms tend to do most of the R&D in their respective home countries (e.g. Pa- tel/Frietsch 2007). In other cases, countries specialisation in sectors that are less technolo- gy-intensive might be a reason to explain the comparatively low share. 17 A significant rise is also reported for France, where the share increased from 11% to 21% in 2001-3. 18 For other countries the OECD (2007) compendium reports: "Canada, India, Israel, Korea, Japan and Singapore own more patents with US inventors than with EU inventors. China shows a more even distribution of domestic ownership across regions while the Russian Federation collaborates mostly with other countries." 19 In countries, such as the Russian Federation, Luxembourg and Hungary, over 50% of do- mestic inventions are foreign-owned, having increased over the past decade. In contrast to this markedly decreased foreign ownership of patents is reported for Finland, India, Korea, Poland and Singapore, foreign ownership. The US and Germany have declining shares of foreign ownership (between 14 and 15%). Korea and Japan report the lowest shares in 2001-03 (with less than 5%). 34 Literature Review SHIA and SHAI are not entirely unproblematic indicators in terms of the extent to which they truly trace knowledge flows and cross-country ownership relations. While patent documents always need to include a complete set of inventors and their addresses, they do not always contain as complete information on applicants, or assignees. When analysing indicators on foreign-owned inventions or nationally owned inventions, one must also take into account that especially multinational corporations, which account for most cross-border inventions, have differing patenting practices. There are consid- erable differences in where they file patent applications for their inventions (e.g. Patel/Frietsch 2007) and whether they are owned centrally at the MNE's headquarters or in affiliates in other countries where most of the research and development activity leading to invention was carried out (e.g. Guellec/Pluvia Zuniga 2007). One needs to bear in mind these limitations when interpreting these types of patent indicators. A third indicator Guellec and van Pottelsberghe de la Potterie (2001) used is the share for a given country of patents with a foreign resident as co-inventor in the popu- lation of patents with a domestic inventor ('SHII'). One could argue that this indica- tor is captures more the essence of co-patenting than the two aforementioned meas- ures. This measure is also methodologically on safer ground as all patent documents must contain complete information on inventors and their addresses. As mentioned earlier, this is not necessarily the case for applicants and assignee. The authors report a steady increase in cross-border co-inventive activity from around 2% in 1985 over almost 5% in 1995 to 7% in 2001-03 (see Figure 2-11 and Figure 2-12 for a country comparison based on recent data). Having said this, one needs to note that international co-inventive activity varies widely between large and small countries with small and less developed economies more strongly engaged in international col- laboration.20 Co-inventions are discussed further in the section on field specific studies below. 20 The OECD report states that "co-invention is particularly high in Luxembourg (52%), fol- lowed by the Russian Federation, Singapore, the Czech Republic and Poland. This reflects these countries' need to overcome limitations due to the size of internal markets and/or the lack of the necessary infrastructure to develop technology. Large countries, such as Fran- ce, Germany, the United Kingdom and the United States report international co-operation of between 12 and 23% in 2001-03, the greatest expansion in the extent of international collaboration from the early 1990s. In France, for instance, it increased from 8 to 16% in 2001-03. The breakdown of collaboration by main partner country (Figure 2-8) reveals pat- terns similar to those reported for cross-border ownership. European countries collaborate essentially with other EU countries; whereas Canada, China, India, Israel, Korea and Ja- pan collaborates the most with the United States. More than 20% of inventions made in In- dia and Canada are co-invented with a US inventor, Brazil and South Africa collaborating mainly with EU inventors" (OECD 2007: 37f). Literature Review 35 Most of the above cited work has been carried out at the OECD.21 Other reports utilise this data. A recent consultancy report by Arthur D Little (ADL 2005) on the internation- alisation of R&D may serve as an example. The United Kingdom is discussed in terms of co-patenting as an exception among large countries, with around 40% of domestic inventions owned by foreign residents, compared to 30% in the early 1990s. One of the reasons is the relatively large number of foreign research labs, for instance, those of US and Japanese corporations. The report draws on basic co-patenting statistics as one of many indicators. Other indicators included are human resource statistics, R&D investment by foreign and domestic firms, foreign direct investment, and co- publications. Figure 2-3: Crossborder ownership of inventions – Global Source: Guellec and Pluvia Zuniga (2007), Fig. 1 Figure 2-4: Domestic Ownership of inventions made abroad, 2001-2003 Source: OECD (2007), Tab. 6.1.3 21 The Third European Report on Science and Technology Indicators (2003) did not include overall data on co-patenting. The report called "A more research-intensive and integrated European Research Area" (European Commission 2008a) also focus on other measures. 38 Literature Review • countries tend to cooperate relatively more in the technology fields in which they are less specialised, • Chemicals are the technology area with the highest propensity of international col- laboration whereas construction and consumer goods related collaborative patent activity is commonly the least international. Figure 2-10 offers a comparison of selected countries' co-patenting (on the basis of inventor addresses). Figure 2-10: International co-patenting of selected countries by technology fields (2004) Electrical Eng. Chemicals Machine Building Instruments Process Tech. Constr./Cons. goods Source: Frietsch/Schmoch (2006) 2.3.3.3 Studies of Technological Domains Our review of the literature could also identify a number of studies with a focus on emergent technologies or already established science-based technologies. These stud- ies cover technology areas, such as nano or biotechnology (e.g. Hullmann 2001; Glän- zel et al. 2003a, 2003b). Some studies limit themselves to presenting co-patenting data in co-invention and co-assignee format whereas others try to associate co-patenting indicators with international knowledge transfers. Both approaches are described be- low. 2.3.3.3.1 Co-invention and co-assignee analyses The studies by Glänzel, Meyer and colleagues (2003a,b) can be used as an example as to how studies of technological domains draw on co-patenting data. Again, co- Literature Review 39 patenting is one indicator of many. The focus is on collaborative activity at the country level. The studies distinguish two approaches: 1. Co-inventions: This measure tracks the composition of inventor teams either at the individual, organisational or country level. A co-invention link points to indi- viduals who generated technology in a common endeavour. It is similar to co- authorships in publications. Some studies even refer to co-inventions as 'patent co-authorships' (e.g. Tsuda et al. 2006). This might suggest that co-authorship and co-patenting are very similar processes. While there are similarities there are also considerable differences. Whereas both patents and papers are generated by teams rather than individuals, patents result to a larger extent from efforts of individuals and small teams rather than larger groups. Co-inventions appear to occur less frequently as cross-institutional collaborations than scientific publica- tions. While one can trace co-authorship networks, co-invention networks occur at best in a rudimentary form (e.g. Meyer/Bhattacharya 2004).22 2. Co-assignation: This link connects actors that share the ownership of a patent. Co-assignations of patents point to a shared interest in utilising a patented inven- tion rather than co-operation in the creation of a technology. Co-assignations oc- cur usually at the organisational level and not the individual level.23 Studies often use Salton's measure to analyse collaborative activity. In the context of co-patent analyses, Salton's measure is defined as the number of inventions (assigna- tions) shared by two countries which is divided by the geometric mean of the total number of inventions (assignations) attributed to the two respective countries, or: ji ij pp p r . = , with pij = the number of links between the countries i and j and pi(pj) the total of inven- tions (assignations) for the country i (j). Co-inventions across countries are quite frequent. For instance, Glänzel et al. (2003b) found that around 27.9% of all biotechnology patent applications with the EPO were 22 See also Appendix 2 for a comparison of co-inventions with co-authorships. 23 Co-assignations indicate joint ownership of invention and may point to joint exploitation intent of the partner organizations. They occur between business firms but links between public sector research institutes and firms may play also an important role. These collabo- rations may happen more frequently within a national than international context. Licensing (on which no data is available) is still the more common form of joint exploitation of patents. This may also be one reason as to why the share of co-assigned patents is relatively low. Also note that information on assignee and applicant organizations is less complete in pa- tent documents than for inventors. All caveats discussed earlier in this context apply also here. 40 Literature Review international co-inventions in the EPO system (12,412 out of a total of around 44,483 for the period 1992-2001). However, the share of co-assigned patents is relatively small in comparison to co- inventions, even though observers find there is an increasing trend to joint ownership and exploitation of inventions in areas, such as biotechnology (Pyka/Saviotti 2002). Glänzel et al. (2003b) identified 3,926 European patent applications (in relation to around 45,000 patent applications in this area during 1992-2001).24 The co-patenting matrices in Figure 2-11 and Figure 2-12 provide examples of cross- border collaboration in terms of co-inventions and co-assignations in biotechnology. The Appendix includes data for nanotechnology. 24 They traced 1,764 international co-assignations out of around 45,000 biotech patents granted in the US (1992-2001), respectively. Literature Review 43 2.3.3.4.1 Econometric studies of innovative activities in the context of regional innovation systems Econometric analyses increasingly draw on co-patent data to analyse research and development collaboration in a regional context. Two studies are used here as an illus- tration for this type of work. Cantner and Graf (2004) draw on co-patenting data to explore cooperation and spe- cialization in German technology regions. The authors use co-patenting as one of the proxies for research collaborations and find that technologically moderately specialized regions show the highest number of research co-operations, and the higher a regions specialization, the more co-operations take place with partners inside that region. Maggioni et al. (2007) use co-patent data as one aspect in their exploration of the im- portance of traditional 'geographical' spill-overs vis-à-vis 'relational' spillovers. Combin- ing participation in the same research networks (within the EU Fifth Framework Pro- gramme) and EPO co-patent applications, the authors examine the factors that under- lie patenting activity. They make a distinction between structural features, geographical and relational spill-overs to test whether hierarchical relationships based on a-spatial networks between geographically distant excellence centres prevail over diffusive pat- terns based on spatial contiguity. 2.3.3.4.2 Studies exploring university patenting with industrial partners Increasingly links between industry and public sector research are explored. For in- stance, Van Looy et al. (2003) studied the co-patenting activity of knowledge generat- ing institutes. Co-patenting between universities and public research institutes on the one hand and industrial companies on the other is becoming a more common topic as public sector research organisations are increasingly aware of the need to manage their IP actively. In some countries this is more pronounced than in others and also institutional practices vary greatly (see e.g. Lissoni et al. 2008; Meyer et al. 2008). Le- gal frameworks for university patenting can also have an impact on co-patenting. If co-patenting is understood as a combination of different organisations owning and through its members having contributed to the inventions, then one must include stud- ies of university/invented but not owned patents. Considerable work on this has been carried out on this topic (e.g. Lissoni et al. 2008; Meyer 2003) suggesting that cross- sectoral (university-industry) knowledge flows are far more frequent than assumed pre- viously. 44 Literature Review 2.3.3.5 Other Studies Tracking Transnational Knowledge Flows Comparing a range of approaches to trace knowledge exchange networks, Klitkou et al. (2007) raise concerns about using a single approach, such as co-patenting, as the only tool to understand knowledge flows at the science-technology interface. They ar- gue that only applying a range of indicators allows the analyst to form a proper view of exchange processes. Transnational knowledge flows have been explored through other ways than co-patenting. Already in the 1970s and 1980s studies were carried out on the foreign dependence of countries' technology bases (e.g. Carpenter/Narin 1983) which explored the extent to which patents of a given country cite patents originating in other countries. While one might remain sceptical about the nature of the 'knowledge flow' that is captured by a patent citation, a larger number of links can be traced. This literature and the related works on knowledge spill-overs (see e.g. Branstetter 2001; Hanel 1994; Hu/Jaffe 2001; Singh 2004) is considerably more established than work on co-patenting. Another possibility to explore transnational knowledge flows is now beginning to be explored on the basis of cross-national science-technology links. Glänzel et al. (2008) presented a bibliometric and patent study on the emergence of China that also looked at patent references and citations of scientific papers at the country level. Table 2-4 presents an overview and points to the substantially increased role of China as an ap- parent absorber of scientific knowledge. As with co-patent data, also here one needs to be aware that size effects and time lags.25 Table 2-4: The fifteen leading countries according to science-technology links based on patent citations based on the SCIE and DII databases (qMOCR denotes the ratio of the mean citation of papers cited by patents to that of all pa- pers) Rank Patent references Patent citations 1991 2001 1991 2001 Ctry Share Ctry Share Ctry Share qMOCR Ctry Share qMOCR 1 USA 30.6% USA 26.3% USA 53.3% 3.17 USA 46.0% 3.63 2 DEU 9.3% DEU 9.2% JPN 10.2% 2.72 JPN 12.0% 4.30 3 JPN 7.6% CHN 7.9% GBR 8.3% 3.66 DEU 10.0% 4.05 4 FRA 7.1% FRA 6.8% DEU 6.8% 3.35 GBR 9.1% 4.08 5 GBR 6.5% JPN 6.8% FRA 5.2% 3.50 FRA 5.8% 4.92 25 Time lags can be even more considerable as the measures are patent citation-based (time lag due to patent process causes publication delay plus subsequent uptake). Literature Review 45 Rank Patent references Patent citations 1991 2001 1991 2001 Ctry Share Ctry Share Ctry Share qMOCR Ctry Share qMOCR 6 CAN 3.6% GBR 6.3% CAN 4.6% 3.25 CAN 4.3% 3.98 7 ITA 3.2% RUS 6.1% ITA 2.4% 3.26 ITA 3.4% 3.17 8 IND 2.2% IND 3.8% NLD 2.3% 2.92 CHE 2.8% 3.18 9 CHE 1.8% ITA 3.8% CHE 2.1% 2.95 NLD 2.7% 3.18 10 NLD 1.7% KOR 3.4% SWE 2.0% 2.84 SWE 2.7% 3.33 11 POL 1.6% CAN 3.4% AUS 1.8% 3.23 AUS 2.4% 5.09 12 ESP 1.4% ESP 3.2% BEL 1.2% 3.09 ESP 2.0% 7.21 13 RUS 1.4% NLD 2.4% ISR 1.1% 3.62 KOR 2.0% 2.58 14 BLG 1.1% CHE 2.1% ESP 0.9% 3.09 CHN 2.0% 8.02 15 CHN 1.0% POL 1.9% DNK 0.9% 2.85 BEL 1.6% 4.10 Source: Glänzel et al. (2008) 2.3.4 Concluding Remarks The analysis of co-patent data can help improve our understanding of transnational knowledge flows, especially when used in combination with other data that can support the interpretation of co-patent statistics. Co-patent data allows us to distinguish relatively 'open' economies with foreign R&D labs in the country and larger shares of 'foreign-owned' inventions from others with more 'closed' approach. However, one must bear in mind the overall specialization of countries on certain sectors and technology areas when interpreting co-patent data. Patenting intensity varies from sector to sector and technology to technology. Interna- tional co-patenting is driven by multinational corporations that vary considerably in their own internal practices. It is important to recognize that co-patenting can encompass collaboration between a domestic and a foreign company but also does include techno- logical developments within internationally active corporations that are, for instance, driven by an R&D team with members located across a number of countries. It is also difficult to make judgments about the directions of knowledge flows on the basis of co-patents. Research labs of foreign companies are a case in point. It would be overly simplistic to say they merely absorb knowledge from their host countries. It might be more realistic to assume they engage in a process that involves some ex- change between domestic and international researchers and engineers. As patent documents allow the analysts only to make judgments about the residence, not about the nationality of inventors it would be misleading to draw on co-patent data to make strong claims about 'brain-drain' from one country or region to another. To explore 48 Feasibility Study This feasibility study focuses on co-publications and co-patents. It tries to offer a fea- sible framework for a regular monitoring based on these indicators and it discusses limits and caveats of different perspective. 3.2 Patent offices and the availability, topicality and com- parability of patent data A patent application has to satisfy at least three criteria: novelty, inventive step and industrial applicability. The criterion of novelty implies not only novelty for a national system or for the applicant, but novelty on a world-wide scale. Furthermore, any publi- cation – for example in a scientific paper or contribution to a conference – or any im- plementation of the invention in any product or process is considered prior art and in- hibits patent protection. The second criterion – the inventive step26 – means that an inventive act had to take place, which is defined by the fact that the new idea is not obvious to a person skilled in the art.27 The third requirement of industrial applicability is generally fulfilled because of the considerable costs of patent applications which are only spent with a realistic market perspective. Starting from a simple legal perspective, patents give, for a limited period, an exclusive right of usage to the applicant for securing a quasi monopolistic revenue. From the per- spective of analysing innovation systems, patents can be interpreted as an indicator of the codified knowledge of enterprises, and in a wider perspective of countries. The fo- cus of the statistical patent analysis is directed towards technological innovations, es- pecially visible in the manufacturing sector.28 In consequence, patents only give an indication of these patentable and patented research results. They are not capable of the totality of possible innovation outputs, for example as they are defined by the OECD (2005). However, it can be plausibly assumed that any patent application is pre- ceded by mostly large investment in the research and development process (Grupp 1998: 145-147; Kash/Kingston 2001). From this point of view, patents can be seen as a success or output indicator of research and development (R&D) processes (Freeman 1982: 8). On the other hand, most – but not all – technological inventions will flow into a product or process that will then be offered on national or international markets. Thus, 26 In US patent law, the corresponding requirement is called "non-obviousness". 27 See Art. 56 of the European Patent Convention (EPC):http://www.european-patent-office. org/legal/epc/e/ar56.html#A56. 28 As to the appropriateness of patents as a technology indicator, see Schmoch and Hinze (2004) and the references cited there. Feasibility Study 49 patents can also be interpreted as an input indicator (or throughput indicators) with regard to future market activities of enterprises, sectors or countries and therefore act as an early sign for future competitiveness. Patents belong to the most important innovation indicators and are a reliable source to measure R&D performance especially in the industry sector. Though patents are only capable of technological innovations – and even here they cover only a fraction of all innovative activities – they can be interpreted, on the one hand, as an output indicator of R&D processes. On the other hand, patents also point to the future by the promise of implementing the technologies and opening new markets or gaining new market shares with new products. Especially in high-technology areas, patents can help to measure present and future competitiveness of companies, sectors, or economies (Frietsch/ Schmoch 2006; Schmoch 2004b). However, the frequent use and the availability of patent data may give the impression that it is a simple and straightforward to use indicator. The opposite is the truth. As an innovation indicator, patents are rather complex as they do not only demand deep knowledge of the data sources, their reliability and validity, or their interpretability. But a mandatory prerequisite is also a deep knowledge of the central legal framework condi- tions, the application processes, different patent systems at different patent offices, incentives and disincentives as well as strategic aspects of patent filings and finally some idea about the decision processes in companies or research institutions, which apply for patents or decide not to do so. Furthermore, some knowledge on technolo- gies and their representation in patent documents is a profitable asset for any differen- tiated patent analysis. The most frequent and most misleading assumption by unfamiliar users is that there is one (and only one) patent application per invention, implicitly assuming that any inven- tion is only filed once and any patent is the same as the other and any patent can be compared or summed up with any other patent. This is by far not the case. Patent of- fices administer patent applications, they examine the claims and they grant a tempo- rary monopoly for the exclusive use of patents. But any patent office can only do this in the territory of its responsibility. If a patent protection is reached in Germany and France, for example, the technology can still be used freely in the UK, in Spain, in Italy etc. Therefore, more than one patent office is approached by an applicant if a broad coverage is intended. As a consequence, the first question in any patent analysis should be: which patents are to be analysed? And the answer to this question is highly dependent on the scope or the range of the intended analysis. 50 Feasibility Study But for the interpretation of the result of the statistical analysis, the analyst should be aware of a caveat that is directly related to the selection of a certain patent office, namely the possible home advantage or home bias. The probability that a national ap- plicant files a patent at his/her home office is usually higher than for any applicant from any other country. This means, for example, US applicants have a home advantage at the USPTO (United States Patent and Trademark Office), Japanese applicants have a strong home advantage at the JPO (Japanese Patent Office) and German applicants show a strong home bias towards the DPMA (German Patent and Trademark Office). Applicants from smaller countries with no large home market often directly file in a lar- ger neighbour country or at international patent authorities. For example, traditionally, Swiss applicants show a strong focus on the German market – and thereby on the German Patent Office – Belgian applicants direct their activities towards France and also Germany, or Canadian applicants file more patents in the US than in Canada. However, in their individual home countries or home offices, respectively, they still have a strong home advantage. Using German patent filings to generally compare the Ger- man strengths and weaknesses of the German technological competitiveness with their counterparts from other countries is not advisable. If the interest is exclusively lying on the German market, this might be a good approach. A measurement of German and international applicants on the same scale or the same standard is not possible with this approach. As a matter of fact, applicants file most of their patents at the national patent office of their resident country. Multinationals (MNEs) usually file more frequently in the country where their headquarter is located or where the research laboratory is resident, from which the invention is originated. With these first filings a priority is claimed, meaning that this is the first codified documentation of the invention. This is important for any subsequent patent application at any other office and – especially under the first to invent system – it is important to document by whom and when the invention was made. Any patent application has to fulfil the criteria of worldwide novelty, which means that the object of the patent must not be used, filed or published anywhere else in the world at any time before the first filing. Unfortunately, this also holds for patent filings by the same applicant with the same objective that are to be filed at any other office sub- sequently. Here the Paris Convention – signed in 1883 – puts remedy as it allows the application of a patent at any additional office within a one year period after the priority date – the date of worldwide first filing – and with reference to this first filing, which is called priority. This also means that after this first year an application of a certain inven- tion at any patent office worldwide is not possible any more – never again! Claiming priority at a national office first has several advantages for the applicant and is therefore still very frequently used. First of all, there is a language advantage as the Feasibility Study 53 procedures all recommend massive investments in fees, translation costs and costs of lawyers. Figure 3-1 depicts the usual and standardised time line of applying for a PCT patent. The timeliness of an "ideal" procedure is as follows. At the beginning there is a national first filing – a national priority. Within one year – the priority year – a subsequent filing of this patent is made via the PCT procedure. At this stage it is – in principle – possible to name all member countries of the PCT as designated countries. The number of countries named does not have any impact on the pricing or the application fees. It is important to note that the PCT application will be published 18 months after priority date – which means the same time like the national priority itself – and is then also ac- cessible and usable for patent statistics. By asking for a preliminary search report, which gives a first indication of success or failure of the application process, entering the national or regional phase can be postponed up to 30 month after priority date. At this point the applicant has to decide at which offices the application shall really enter the national/regional phase and thereby decides on the number of subsequent exami- nation and granting procedures, which are – from then on – individual applications and may individually fail or succeed and each has to be paid individually. And at the very end of each process there will be a number of national granted patents. In the context of this study it is important to note that a PCT filing might enter the national/regional phase both at the EPO and the USPTO (United States Patent and Trademark Office), but with an additional delay of 12 months after publication date. The cohorts of total patent applications per priority year are considerably affected by this fact. 54 Feasibility Study Figure 3-1: Timeliness of PCT filings Source: Felix (2007: 2). As can be seen in Figure 3-2, there are three paths to apply for EPO patents: 1) a na- tional priority subsequently filed at the EPO under the Paris Convention, 2) a direct first filing to the EPO, which is possible since 1994, and 3) a PCT filing – of course originat- ing in a national priority – that enters the regional phase at the EPO. And this is a rather frequently used path as about 60% of all EPO applications are filed this way. And the reason why PCTs with designation to more than one European country are filed in this way is the common examination and granting procedure at the EPO com- pared to using the PCT route to enter the designated European countries directly (Felix 2006). The procedures at the USPTO are rather similar to the procedures at the EPO in this respect. Three possible paths for an application exist: 1) a priority at another office that is subsequently filed at the USPTO under the Paris Convention, 2) a first filing to the USPTO (a priority itself), and 3) a PCT filing that subsequently enters the national phase at the USPTO. Feasibility Study 55 Figure 3-2: Process of an EPO filing PCT application (regional phase) Granting Procedure (Search Report) Grant FR DE GB BE EPO application National application (Priority) Source: Frietsch and Schmoch (forthcoming) The United States Patent and Trademark Office (USPTO) still covers the most impor- tant national market for technologies in the world, namely the US market. However, it is still a national market. Some countries, especially the upcoming and emerging coun- tries like South Korea or India, have a special focus on the US market and hardly file every patent on a worldwide scale. In consequence, the home bias of US appli- cants/inventors is considerable and the imbalance of European, North-American and emerging countries cannot be neglected. Different to the EPO – for example – the USPTO only published granted patents in- stead of applications until the publication year 2001. Since then, they publish both, ap- plications after 18 months and grants immediately after the granting procedure is fin- ished (which might take up to 7 years and more after priority). Though, pure national filings are still exempted from the pre-grant publication demand so that some applica- tions are still unpublished until the granting of the invention. In this transition phase from grant- to pre-grant-publication it might not be meaningful to analyse longer time series at the USPTO, though it seems that the transition to the new system as such has been successfully finished already in the mid of the first decade of the new century (Schmoch 2008; Schmoch forthcoming). As a matter of fact the analysis of longer time series has to be restricted to granted patents only. Figure 3-3 shows the absolute num- ber of total granted patents and total number of applications at the USPTO by priority year. It can be seen that the cohorts of the years after 2000 are still incomplete. Fur- thermore, the application cohorts are incomplete for the years 2004 and 2005 – for similar reasons as in the case of the EPO. 58 Feasibility Study which covers patent information from many sources and offices, an extraction of the examiners database is made twice a year and is made available for the scientific pub- lic. This extraction is the PATSTAT database. Next to information on applicants and inventors – their names and addresses – addi- tional information is available on the technical features (claims), the title, the classifica- tion and several dates like priority, application or publication dates are accessible. What does all this mean for the analysis of co-patents in the context of ERA? Patents do not reflect all inventions and also not all innovations that occur in research or innovation systems. Patents are restricted to technological innovations only. Patent documents are only valid in the scope of the authority/office where it is filed. In conse- quence, the same invention/innovation might be filed at several offices. Therefore, summing up patent documents across offices might result in multiple counts of the same invention/innovation. Furthermore, companies located in the country of the office will have a home advantage compared to non-national applicants/inventors. We use EPO patent filings for this study as the EPO is a transnational authority so that the home advantage is somehow balanced. Furthermore, we are interested in the Euro- pean market that is – to some considerable extent – covered by the European Patent Office. However, European applicants/inventors still have a higher probability to file here than outside Europe. On the other side, when USPTO patent filings are taken into account, the home bias for US-residents and even for Canadian applicants/inventors is much larger than for Europeans. Next to the totals, this bias also affects co-patents of European and North-American applicants/inventors, so that the structures might be different when USPTO and EPO co-patents are compared. The timeliness of the application processes and the timeliness of the publication of the patents have a direct impact on the topicality of the data to be analysed. Mid 2008 (date of data extraction for this report), EPO filings are completely published only for the priority years up to 2006 (publication phase of 18 months). However, due to the fact that more than 60% enter the EPO via the PCT route and the fact that entering the re- gional phase at the EPO might be postponed up to 30 months, the last two priority years (2005 and 2006) are still incomplete. This is why the analysis is restricted to the priority years before 2005. USPTO data is not analysed due to the missing topicality of the grants and the strong home bias, which has an unfavourable impact on the Euro- pean applicants/inventors. The pre-grant publications of the application are not yet ready for long-term analyses and the system is still shifting. For future analyses it should be considered to analyse alternative data sets. The triadic patent approach could be one, but the fact that more and more international patent Feasibility Study 59 applications are filed via the PCT-route also reduces the topicality of the available in- formation. Frietsch and Schmoch (2007; forthcoming) recently suggested an approach of Transnational Patents that integrates EPO and PCT applications and that provides an analytical framework for structural analyses of the most important innovation ori- ented nations with a high topicality of 18 month after priority date. This latter approach already proved its feasibility and reliability in several studies (Frietsch et al. 2008; Frietsch, Schmoch 2007; Patel, Frietsch 2007) and could also be an approach for measuring co-patents in the context of the ERA. Patents are one – among others – output of (mainly private) R&D. They are reliable and valid as they have to fulfill high standards, which are examined by patent offices. However, as an innovation indicator, patents are rather complex. One of the dimen- sions of this complexity is the fact that there might be applications of the same tech- nology/invention at different offices, leading to different patents (documents). There- fore, for the analysis is has to be decided which office’s documents are to be ana- lysed. Due to international filings procedures and as a matter of fact, the restriction to EPO applications – as it was requested in this project by DG-RTD – is not appropri- ate for a permanent and topical monitoring system of collaborations. Instead, the use of patent families are recommended, where the so called Transnational Patents (EPO or PCT applications) have been suggested as the most appropriate one. USPTO data is not recommended for several reasons: 1) it covers a pure national office, where US-inventors and –applicants have a home advantage; 2) time series are not long enough as until 2001 only grants and not applications have been pub- lished. 3.3 Publication data – SCI versus SCOPUS For the analysis of scientific publications in this report the Science Citation Index (SCI), an internationally, multidisciplinary database, was used. The SCI is widely acknowl- edged and accepted for evaluating research institutions and analyzing research per- formance of regions or countries. The SCI covers about 6,000 of the most significant and relevant scientific journals from a wide range of scientific and technological fields. However, the SCI has certain limits and restrictions, which are worth to be kept in mind for the analysis. First of all, even coverage of about 6,000 journals does not reflect the totality of papers published worldwide. In addition, the main focus is on reviewed jour- nals and (selected) conference proceedings – with some additional, but minor docu- ment types like letters, notes and reviews also included. Furthermore, neither so called "grey literature", nor monographs or edited books are contained. Secondly, the SCI 60 Feasibility Study covers sciences and engineering, whereas social sciences or arts and humanities are covered by the Social Science Citation Index (SSCI), which is not taken into account here. The reason is that these latter fields are very nationally oriented so that interna- tional comparisons are almost meaningless. Concerning the SCI it still has to be ac- knowledged that sciences (especially life sciences) reach a higher representation rate than engineering journals (Schmoch 2004a) – and engineering is the scientific and technological strength of many European research and innovation systems. Thirdly, the SCI is biased in favour of English-speaking countries, especially with respect to the USA. Though, other countries are catching-up, especially China, so that the US- American share in all SCI-publications is decreasing steadily over time. Scientific publications for the period 1990-2006 were retrieved. A special focus is on the period 2000-2006. Only the so called citable items namely articles, general reviews, notes and letters were included in the analysis. The data used for this study was re- trieved online via the host STN. Similar to other online-retrieval accesses – for example like the "Web of Knowledge" – this kind of data access has certain advantages and disadvantages. The advantages are flexibility, topicality and – in the case of STN – professional and uncomplicated handling. However, the limits are to be seen in the analytical possibilities as well as in the high costs of certain analyses. For example, fractional counting of authors or classes is impossible and citation analyses – next to the fact of high costs for the analysis of large datasets – are not possible at all, for ex- ample like expected citation rates (see for example Moed et al. 2004; van Raan 1988). Though, these kinds of analyses were not in the scope of this study, future studies might want to analyse not only quantity but also quality – measured by citations – of scientific publications. In this case an in-house or offline database is mandatory. For a very long time the SCI held a monopoly for this kind of analyses of a large num- ber of scientific areas. Since a few years Elsevier publisher provides an alternative da- tabase called SCOPUS, which has not yet reached the same dissemination for bibli- ometric analyses like the SCI, but which is able to overcome some of the disadvan- tages of the SCI – though by the cost of some additional disadvantages. SCOPUS claims to cover about 16,000 reviewed journals. It has a broader coverage of engineer- ing publications and it also has a broader coverage of non-US authors and journals, especially from Europe. However, disadvantages are shorter time series (about 50% of the publications only date back to 1996), a less differentiated classification scheme and less clear indication of the quality of the underlying journals and the reason for their inclusion (or exclusion). However, future studies might want to compare results from the two data sets or maybe exclusively use the SCOPUS database. Feasibility Study 63 The IPC has – in contrast to the vast majority of other classification schemes – a huge advantage, namely the fact that the assignment of the patents to the classes is done by patent examiners, who are experts in their fields. In consequence, the quality of the assignment in conjunction with the very deep differentiation scheme of the IPC pro- vides a sound and reliable foundation for any analysis. Though, each individual patent is usually assigned more than one patent class as it might have links to and features of different technologies. Until the 7th version of the IPC, which was in use before 2006, one main class and several secondary classes have been assigned. Since the 8th ver- sion of the IPC this is not done any more. So until the 7th edition it was possible to come to unique assignments to one group of technologies by using the main class only. This is not possible any more so that the whole set of IPC-classes has to be taken into account.33 In consequence, in the analyses conducted here double/multiple counts of patents are possible according to the number of different symbols. It has to be kept in mind that a patent is a vested right of a technology – and not of a product. The number of 70,000 symbols – or even 20,000 symbols – are by far too much to be examined in structural analyses like the one undertaken here. And the big- gest challenge indeed is to aggregate the IPC classes to technology fields that can be used for this kind of structural analyses. Next to the fact that it is mandatory to have a deep knowledge and understanding of the patent system and the motivations and way of thinking of the applicants, it is a mandatory prerequisite to have a sound knowledge on technologies to aggregate these classes. Instead of setting up a new aggregation scheme, we decided to make use of an existing and established one (Schmoch/Gauch 2004). In our case, this differentiates between 19 technological fields, which we aggre- gate to 6 technological areas that are analysed. The classification scheme can be found in Annex 5. The category codes of the SCI and their aggregation to scientific fields The Science Citation Index classifies all journals in the database in almost 200 so called Category Codes. Two things are important to be known and kept in mind for the analysis. On the one hand, the journals are classified and not the individual articles. This means that – for example – a paper on pharmaceuticals which is published in a chemistry journal might be classified as chemistry and not as pharmaceuticals. On the other hand – and this balances the first effect to some extent – the journals are as- signed more than one Category Code. In consequence, also here double counts of 33 It would be possible to use the first instead on behalf of the main class, but in the PATSTAT database – which is a relational database – the information on the position of each individual IPC-class is not available for all documents/offices. 64 Feasibility Study journals/articles may occur. Next to the fact that no unique assignment is possible, this also means that the sum of the individual classes is higher than the total number of publications. For structural analyses this is usually no problem, but for the calculation of indicators like shares or specialisation indices, the reference to the sum instead of the total needs to be made. An alternative to the use of the existing Category Codes would be the use of keywords to define scientific fields. By the way, an approach that cannot exclude double/multiple counts, too. However, certain caveats exist and clear disadvantages of this strategy can be identified. The first question to be addressed is, if the keywords are searched in the titles – title of the article and/or title of the journal –, in the abstracts34, in the journal keywords, or in any combination of them. A restriction to one of them might drop out a lot of relevant articles as the decisive keywords might not appear in the title, for exam- ple. A search in all available items might result in the inclusion of irrelevant arti- cles/journals. Here a first problem of false positives and false negatives emerges. However, the biggest challenge of a keyword strategy is the clear, distinct and correct definition of the scientific fields. Next to a deep knowledge of each individual field, the keywords for each class must be checked and verified for false positives and false negatives. Using too general terms does not allow a strict distinction between fields and too special terms might drop out a lot of relevant documents. To sum up, a key- word strategy is very complex and hard work. It only makes sense for individual – maybe new and upcoming – scientific fields, but not for a structural analysis like it is undertaken in this study. The classification scheme used for the analyses in this report can be found in Annex 5. As discussed, it makes use of the Category Codes provided by the database provider and it differentiates between 26 disciplines, while a further aggregation is possible, but not used here. Double/multiple assignments are possible. 34 Abstracts are not available in the SCI databases before the publication year 1991. Feasibility Study 65 Patents are classified according to a very detailed and sophisticated classification scheme (IPC). The reliability of and validity of this classification is very high as the assignment is usually done by patent examiners, who are experts in their fields. However, this classification is not very handy for statistical analyses. Therefore, for practical reasons and the reason of comparability, it is recommended to rely any regular monitoring on existing and established aggregations of the IPC. The SCI uses Category Codes to classify journals and not individual articles within the journals. Each journal is assigned multiple Category Codes so that no unique assignment and classification is the consequence. A keyword definition of scientific fields is not recommended and would be a study in its own to properly define these classes. 3.6 Full counts versus fractional counts and intensive ver- sus extensive definition of intra-EU publications 3.6.1 Full counts versus fractional counts In the previous sections it was already mentioned that double or multiple counts can occur. This – first of all – happens in the context of the classification schemes as unique assignments to classes are not possible, both for patents as well as for scien- tific papers. At the same time, double counts might also occur in the light of multiple authors/inventors from different countries. This is of special interest against the back- ground of the research question addressed in this project, namely the measurement of international collaborations by co-patents and co-publications. In the case of single authored articles or single invented patents no problem occurs. The same holds if only one nationality of the authors/inventors is given. The patent or article is counted for one country only. For example, three researchers from France, Belgium and the USA collaborated in a research project and have published a joint paper. How to count this document? The first option would be to count it for the first author only. Next to technicalities of the databases, which might not allow a realisation of this approach, it first of all ignores the contribution of the other countries. In addition, international collaborations are in the scope of this project, so ignoring international co- patents or co-publications would pervert this question. The second option is to apply a fractional counting of authors, so that ⅓ of this publication is assigned to France, Bel- gium and the USA each. This approach assumes equal contributions of each author. The third option is to assign the article to each of the countries that occur and fully count it for each of the countries. In the example above it would mean that the article 68 Feasibility Study Figure 3-5: Number of publications and co-publications in EU-27 countries (alterna- tive definition) Source: STN – SCISEARCH; Fraunhofer ISI computations The original definition of intra-EU, which focuses on exclusive intra-EU co-patents and co-publications, has been requested by DG-RTD and is therefore kept for the analyses. We have presented the differences of the two perspectives in this section. In addition, we will present more detailed analyses in the data analysis chapter, which compares the two perspectives at least for publications. A similar comparison also on the basis of patent data would have gone even more beyond the resources and the scope of this project. 3.6.3 The Simplified Data Extraction Routine The simplified data extraction strategy works as follows: first, the total number of publi- cations/patents per country was identified. Second, the total number of pure national publications/patents was calculated and third, the total number of intra-EU collabora- tions was counted. The remaining number of extra-EU collaborations was simply the differences between the sums of the pure national and the extra-EU publica- tions/patents and the total. Feasibility Study 69 As no unique assignments of classes or unique identifications of the country of origin, both of patents and publications, might occur, the way of counting patents and publi- cations (and co-patents and co-publications) has to be decided before the analysis. There is no right or wrong answer to these questions, but it is a matter of analytical scope and research interest. In addition, practical reasons were considered. In the case of classes/categories as well as of authors/inventors, we decided to use multiple assignments for this study. This approach results in a difference of the abso- lute numbers between the total and the sum of the individual publications/patents. DG-RTD requested a definition of intra-EU collaborations that was very strict, defin- ing intra-EU very exclusive as any co-patent or co-publications where only au- thors/inventors from EU Member Countries participated. Trilateral collaborations of two member country authors/inventors and one non-member country author/ inventor were counted as extra-EU collaborations. For this feasibility study we have also checked the alternative definition where also this latter example would have been counted as intra-EU collaboration. For future studies and a regular monitoring system this alternative definition is recommended. 3.7 Summarising conclusions: Do co-patents and co- publications reflect collaboration? The literature review revealed that especially to measure science collaborations, inter- national co-publications are rather frequently used, though restrictions of interpretability were discussed (Katz/Martin 1997; Laudel 2002). Co-patents are less often used so far, especially as they also cover intra-company (in MNEs) collaborations in interna- tional teams. If the ERA is seen as a vision about coordinating national research activities and poli- cies and creating an internal market for research with the free circulation of research- ers, ideas and technology, then indicators to measure this circulation can be used. The ERA was initiated to overcome three weaknesses: insufficient research funding; inade- quate framework conditions to stimulate research and its exploitation; and finally the fragmentation of activities and resources. Improving the co-operation and co-ordination among key players within the European Union is a key factor to overcome these short- comings. In consequence, international – or better trans-national – co-patents and co- publications are an adequate mean to measure this at first sight. However, the litera- ture review revealed some limits of these indicators, which need to be kept in mind when the data is interpreted. One of these limits is of course that co-patents focus on technological inventions – and even here only on a fraction of the totality of inventions 70 Feasibility Study – and co-publications reflect one among possible other outcomes of the research sys- tem. Especially informal knowledge flows and exchanges are hardly to be covered by these indicators. In addition, the ERA is also made of additional instruments, means and infrastructures, which foster the exchange of knowledge and ideas, but which can- not be covered by patents and publications. Furthermore, the elements listed in the Green Paper on the ERA cover more than just the knowledge flows between compa- nies and/or public research institutions. The motives to collaborate in general, but es- pecially across country borders were identified as: geographical or cultural (language, history) proximities, economic or political factors as well as access to knowledge or resources (equipment etc.). In general, applying indicators on co-authorship and co-patenting data provides a quantitative approach to the question of international collaborations. However, it has to be stressed that neither patent nor publication indicators as such allow an interpretation of the motivation behind, the direction of knowledge flows or the initiation of the joint work. Furthermore, intensities of networks and exchange as well as the quality of the collaboration or its output cannot be measured by the approaches presented in the literature review and also not by the indicators used in this report. The data presented in this study first of all gives a general overview of the structure and the quantity of international collaboration in patenting and publishing in Europe. The emergence of the European Research Area cannot directly be measured by this approach and first of all has to be benchmarked against the networking activities of extra-EU partners. Furthermore, the change of these patterns and structures over time give an indication of positive or negative trends of trans-national co-operation in Europe. In this respect co-patents and co-publications are – first and foremost – in- struments to monitor the past and present knowledge flows and network activities. However, the motivations and driving factors have to be examined and analysed with other instruments, for example like the survey of collaborating researchers and inven- tors, where the indicators approach is complementary to (and vice versa). The quanti- tative indicators can provide framing information as well as insertions for further re- search. The results have to be interpreted with caution and the limits and idiosyncrasies of the data, the data sources and the indicators as such have to be taken into account. A very detailed analysis of individual member countries as well as of technology areas and scientific fields cannot be recommended. A stubborn and stolid exercise on all possible or desired links does not provide new and far-reaching insights. It is recommended to focus on EU-27 as the partner countries as well as groups like Asia, North-America, and EFTA. On the side of the individual countries to be analysed, it is recommended to Data Analysis 73 4 Data Analysis 4.1 Scientific Publications – Trends and Structures of EU- 27 countries 4.1.1 Introduction The motivations and driving forces behind international collaborations – of which we focus on collaborations of EU-27 countries with a special attention to intra-EU collabo- rations – are manifold and range from seeking access to data or laboratory equipment to access to complementary knowledge (see literature review). Further impacts stem from geographical proximities as well as cultural similarities – among which language is the most important one. Three different indicators were analysed: 1) Number of EU-single publications, cover- ing all publications where authors from only one EU-country were on the list. In other words: pure national publications; 2) Extra-EU co-publications contain all collaborative publications, where at least one author from outside the EU-27 has jointly published a paper with at least one author from within the EU (also covering all publications where authors from two or more EU-countries have collaborated with researchers outside the EU); 3) The number of intra-EU co-publications reflect the activities that emerge out of the collaboration within EU-27 countries where no author from outside the EU was in- volved35. 4.1.2 Trends and Structures Figure 4-1 depicts the absolute number of publications and international co- publications of EU-27 countries in the years 1990-2006. The number of publications is steadily increasing. However, the number of single-country publications does not grow with the same pace, meaning that the international co-publications were growing much faster and have driven the overall development in the period under observation here. The second main lesson to be learnt from this graph is the fact that the pure intra-EU co-publications did not increase to the same extent like the extra-EU co-publications. The literature review has provided several reasons why researchers collaborate inter- 35 In the course of the analysis this decision proved to be too limited. Therefore, we have decided to additionally also collect the number of intra-EU-27 co-publications where also at least one author from a third country outside Europe could have been on the list. As a di- rect consequence, in relation to this we also calculated the number of extra-EU publica- tions of EU-27 countries in collaboration with a third country as excluding other EU-27 countries – this matches with indicator No. 2. 74 Data Analysis nationally. Main motivations are the access to data and equipment, but especially the access to complementary knowledge. Increasing complexity and convergence of fields and even disciplines make the joining of knowledge even mandatory. However, as a matter of fact this additional knowledge cannot be found only within the EU in the same way or to the same extent as in collaboration with external partners. The reasons might be manifold – and it has to be objected at this point that we have made a distinction between extra-EU and pure intra-EU publications, whereas a large number of publica- tions is also made in trilateral collaborations between EU-authors with external partners (see methodological discussion in chapter 3 of this report) – and can be summarised with two main arguments: 1) The other EU-27 Member countries are not always the providers of complementary knowledge to the collaboration-seeking country and 2) the structures and networks, also supported for example by the opening of the Framework Program to third countries, are not restricted to the intra-EU collaborations. A further reason is the location of international research centres – like CERN, for example – out- side the EU. However, the main explanation is the fact that the USA are still the most important actor in science and technology and therefore also offer a wide spread of opportunities and possibilities for collaboration. Figure 4-1: Number of publications and co-publications in EU-27 countries 0 50 100 150 200 250 300 350 400 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 N o . o f P u b lic at io n s (i n t h o u sa n d ) 0 15 30 45 60 75 90 105 120 N o . o f P u b licatio n s (in th o u san d ) EU-27 totals EU-27 single country Intra EU-27 Extra EU-27 Totals, Single-country Intra, Extra Source: STN – SCISEARCH; Fraunhofer ISI computations Data Analysis 75 Next to the overall trend of increasing international collaboration in science and tech- nology in general, another explanation can be found in the differing internationalisation and international orientation of countries within the EU. Furthermore, researchers within the member countries are engaged in different fields and disciplines, which re- sults in different structures and different total shares. Figure 4-2 provides the shares of intra-, extra- and single-EU publications of the 27 individual member countries. As a rule of thumb it can be said that larger countries have lower shares of international co- publications. However, this is only a rough rule of thumb as countries like Greece, Po- land, Slovenia, Lithuania and even Finland show high shares of single-country publica- tions. In consequence, there must be other mechanisms active that explain the national orientation. In an internationally comparative study like this it is not possible to dig deep into the structures of each individual country and extract detailed information. Here it is more interesting to derive overall patterns that can be found in many countries. On the other hand, countries at the high end of the scale in terms of international collabora- tions like Bulgaria, Latvia, Malta, Cyprus, or Luxembourg have low absolute numbers so that their profile is simply the result of a size effect. However, if they publish at all, they collaborate internationally much more frequently. Another finding of Figure 4-2 is the fact that several countries – most of them are the large ones again – have high shares of extra-EU collaborations, which reflects their networking especially with Switzerland, North-America, but also other countries and areas like China, Japan or South-America. Eastern European countries seem to have higher shares of intra-EU co-publications than most of the other nations. Though still lower shares than Malta, Cyprus and Luxemburg. On average the share of extra-EU co-publications is more than 26% and the share of single-country publications in all EU-27 countries is about 57%. In consequence, the share of pure intra-EU co-publications is 17% – but it has to be stressed again that this figure only covers international co-publications where no researcher from a non-EU country was involved at all. Therefore, Figure 4-3 depicts the shares of intra- and extra-EU co-publications based on the alternative definition, where intra-publications do not only cover exclusive col- laborations within the EU, but also publications between EU-members and third country partners. In consequence, extra-EU is defined as exclusive co-publications of one member country and at least one third country. On average, the share of intra-EU co-publications raises to 27.5% whereas in the countermove the shares of extra-EU co-publications is only 15.6% under this definition. The relation between intra- and extra-EU collaborations in comparison of the member 78 Data Analysis Figure 4-4: Shares of Intra and Extra-EU publications in EU-27 countries by scientific fields, 2004-2006 0 10 20 30 40 50 60 70 80 Nucl ea r t ec hnolo gy Med ica l e ngin ee rin g Geo sc ien ce s Phys ic s Mat hem at ics M ed ici ne Food, n utri tio n Com pute rs Bio te ch Bio lo gy Mea su rin g, c ontro l Eco lo gy, c lim at e Tota l f ie ld s Phar m ac y Optic s Org an ic ch em ist ry Oth er Bas ic ch em ist ry Envi ro n e ngin ee rin g Chem ica l e ngin ee rin g Mat er ial s re se ar ch Polym er s Mec han ic al en gin ee rin g Elec tri ca l e ngin ee rin g Ther m al pro ce ss es M ulti dis cip lin ar y Civ il e ngin ee rin g EU-27 total EU-27 single country Source: STN – SCISEARCH; Fraunhofer ISI computations One of the arguments introduced in the literature review and also mentioned above is the seeking for complementary knowledge. Differences in this cannot only be found between countries, but especially between scientific fields. Figure 4-4 depicts the shares of EU-27 countries of the worldwide publications as well as the shares of EU- single country publications by 26 scientific fields. The two trend lines in the graph re- flect the fact that the shares of pure national publications increase while the EU-shares of the worldwide total slightly decrease when the fields are displayed in this order. To put it in other words, where the EU is strongly engaged, the shares of international col- laboration are lower. However, one has to admit that this effect is not very strong and the variation between the fields is sometimes considerable. Tough, the correlation coef- ficient reaches a value of -0.29, indicating a medium co-variation of the two data series. The shares of EU-single, intra- and extra-EU publications by scientific fields are dis- played in Figure 4-5. Chemical Engineering as well as Food and Nutrition are at the top of the list in terms of EU-single country publications, while Nuclear Technology, Geol- ogy and Physics reach the lowest shares of pure national publications. It seems that more applied fields are less international while basic research areas seem to reach Data Analysis 79 higher shares of internationalisation. The necessity to internationalise in the different fields has – of course – also an impact on the fields' performance in this respect. CERN plays an important role in Physics and Nuclear Technology, for example. Geoscientists analyse and compare sediments, samples, or tectonics, which are not necessarily in their home country. High shares of extra-EU collaborations can be found in Civil as well as Mechanical Engineering, where Switzerland plays a considerable role. On the other hand, pure intra-EU co-authorships are exceptionally frequent – compared to the other fields with similar international activity – in Organic Chemistry, Medical Engineering, Computers or Electrical Engineering. However, shares below the average of intra-EU collaborations are especially reached by Chemical, Civil, and Mechanical Engineering. Figure 4-5: Shares of EU-single, Intra- and Extra-EU publications in EU-27 countries by scientific fields, 2004-2006 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Chem ica l e ngin ee rin g Food, n utri tio n Ther m al pro ce ss es Org an ic ch em ist ry Civi l e ngin ee rin g Med ici ne Med ica l e ngin ee rin g Com pute rs Mec han ica l e ngin ee rin g Phar m ac y Envir on en gin ee rin g Oth er Polym er s Bas ic ch em ist ry Tota l f iel ds Elec tri ca l e ngin ee rin g Mat er ial s r es ea rc h Mat hem at ics Mea su rin g, c ontro l Bio lo gy Eco lo gy, cli m at e Optic s Bio te ch Nucle ar te ch nolo gy Geo sc ien ce s Phys ics Multi disc ip lin ar y EU-27 single country Extra EU-27 Intra EU-27 Source: STN – SCISEARCH; Fraunhofer ISI computations While Figure 4-5 is based on the definition of exclusive intra-EU publications, Figure 4-6 uses the alternative perspective and allows intra-EU publications also to emerge out of collaborations with third country authors in addition. In general, the shares of intra-EU publications are much higher, of course, but it is still the engineering fields that reach collaboration rates below the average. Extra-EU co-authorship shares are lowest in Food and Nutrition – a finding that fits with patent analyses, which show that this kind of consumer market oriented applied technologies are mainly targeting national or re- 80 Data Analysis gional markets. Nuclear Technology and Physics are reaching highest shares when the alternative definition is taken into account. Together with the above finding of excep- tionally high shares of extra-EU collaborations in Figure 4-5, it can be stated that in these two fields trilateral co-authorships are very common. Figure 4-6: Shares of EU-single, Intra- and Extra-EU publications in EU-27 countries by scientific fields (alternative definition), 2004-2006 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Chem ica l e ngin ee rin g Food, n utri tio n Org an ic ch em ist ry Ther m al pro ce ss es Med ica l e ngin ee rin g Civi l e ngin ee rin g Med ici ne Com pute rs Mec han ica l e ngin ee rin g Phar m ac y Envir on en gin ee rin g Polym er s Elec tri ca l e ngin ee rin g Oth er Bas ic ch em ist ry Tota l f iel ds Mat hem at ics Mat er ial s r es ea rc h Mea su rin g, c ontro l Bio lo gy Optic s Bio te ch Eco lo gy, cli m at e Nucle ar te ch nolo gy Geo sc ien ce s Phys ics Multi disc ip lin ar y EU-27 single country Extra EU-27 Intra EU-27 Source: STN – SCISEARCH; Fraunhofer ISI computations 4.1.3 EU-15 co-publications Until the year 2004 the European Union only consisted of 15 member countries. As we analyse the time series 2004-2006 and as the integration especially in research net- works might take some time, we discuss the structures of co-publications of the former EU-15 countries in this section additionally. In Figure 4-7 the number of international co-publications in the EU-15 countries is dis- played. Next to the totals also the single country, the intra- and the extra-EU collabora- tions are depicted. In this case the original definition of intra-EU-collaborations was applied, covering only those international co-publications where at least authors from two different EU-15 countries were involved and at the same time no author from a non-EU country made a contribution. The definition is also strict in terms of EU-15, Data Analysis 83 Figure 4-8: Shares of single-country, Intra-EU and Extra-EU publications in EU-15 countries, 2004-2006 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Gre ec e Spain Ita ly UK Fra nce Ger m an y Fin lan d Net her lan ds Swed en Ire lan d Portu gal Aust ria Den m ar k Belg iu m Luxe m bourg single country Intra EU-15 Extra EU-15 Source: STN – SCISEARCH; Fraunhofer ISI computations 4.1.4 Conclusions This chapter analyses international co-authorship patterns of EU-27 countries in total, by each country as well as by 26 scientific fields with a technological perspective. It is found that the shares of international collaborations have been increasing considerably over time, while the main driving force behind this trend was the collaboration with ex- tra-EU partners. The absolute number of pure national publications has hardly been changing since the mid 1990s, but is slightly increasing nowadays. The main results derived from the analysis of the international activity rates of the indi- vidual member countries are some indications that larger countries have lower shares of international collaborative publications. Some of these countries show considerable exchange with third countries outside the EU, among which Switzerland, North- America, but also other countries and areas like China, Japan or South-America play an important role. The analysis by 26 scientific fields especially lead to the result that where the EU is strongly engaged, the shares of international collaboration are lower. Another finding is that more applied fields are less internationally oriented, while basic research areas 84 Data Analysis seem to reach higher shares of international co-authorship. Chemical Engineering or Food and Nutrition are rather nationally oriented fields, while Nuclear Technology, Ge- ology and Physics are very internationally oriented. Though, trilateral collaborations – between at least two EU member countries and one non-EU country – are rather fre- quent in Nuclear Technology research and in Physics. An additional analysis of EU-15 collaborations revealed similar trends like in the case of all EU-27 countries, though on a slightly lower level, of course. Some indications were found that support the relevance of geographical proximity. The integration of the recent acceded countries might happen on this path. However, the statistical analysis provided here covers a period of the accession that is too short to prove the integrating effects of geography. Future studies might find more evidence for this. 4.2 Patent Analysis 4.2.1 Introduction 4.2.1.1 Aim This section is focussed on identifying the main trends in international co-patenting involving the EU countries, to get an indication of the extent of international collabora- tion in technological activities. We analyse data according to: (a) EU as a whole (both EU-15 and (EU-27); (b) for each individual EU country and (c) according to 6 aggregate technical fields. In each case the aim is to show the extent of international co-patenting according to whether it involves other EU countries or those outside the EU. 4.2.1.2 Data Issues The source for data on EPO and USPTO patenting is the PATSTAT database (more specifically the version released in October 2007) as supplied by the patent offices. PATSTAT is based on an internal database of the EPO which is used by patent exam- iners to search for prior art and especially to establish worldwide novelty. It is not pri- marily compiled to for statistical purposes, but to serve procedural needs stemming from the application procedure and the patent law. PATSTAT stores information on bibliographic details of the applicants, the inventors and of patents, including the date of first filing (priority) and legal status. There are a number of issues regarding the data that need to be borne in mind when interpreting the results reported here. Data Analysis 85 • Patents have been aggregated according to priority years (beginning in 1990). This choice effectively means that the data are only complete for 2004 (as there is a con- siderable time lag of before publication). Additionally when PATSTAT contains no in- formation on a specific priority application we have used the filing year as an indica- tion of the priority year. • The 'country' designation of a patent is the country address of the inventor. Thus international co-patents are defined as any patents where at least two different in- ventors from two different countries have been involved. Where more than one in- ventor from the same country appears on a patent this is counted as a single patent for that country. This means, for example, that if two French and one British inventor have cooperated, this filing is counted both as one French co-patent and as one British co-patent application. • The 'technical' designation of a patent is based on the IPC class to which the patent belongs. Here we have used all the IPC classes that appear for a particular patent in the PATSTAT database and allocated them to one of the following aggregate classes: − Electrical engineering − ICT − Instruments − Chemistry − Mechanical engineering − Other. An additional point to note is that the analysis for EPO patents is based on direct ap- plications to the EPO plus the PCT applications that have entered the regional phase at the EPO. At the same time the USPTO data refer to patents granted, as this was one of the requirements mentioned in the ToR. As all patents used here are dated by the year of priority this means in effect that the data for the USPTO are not complete for the latest years (as these patents have not been examined as yet). 4.2.1.3 Indicators Constructed The analysis below is based on the following data extracted from the PATSTAT data- base for the priority years 1990 to 2004: • Total number of single country patents, where all inventors are within the same country. • Total number of intra-EU co-patents, where all inventors are within the EU-15. • Total number of co-patents with countries outside the EU, where at least one inven- tor is from outside the EU.